Mahidol University · Thesis entitled THE ASSOCIATION BETWEEN PERIODONTITIS AND CHRONIC KIDNEY...
Transcript of Mahidol University · Thesis entitled THE ASSOCIATION BETWEEN PERIODONTITIS AND CHRONIC KIDNEY...
THE ASSOCIATION BETWEEN PERIODONTITIS AND CHRONIC KIDNEY DISEASE
ATTAWOOD LERTPIMONCHAI
A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR
THE DEGREE OF DOCTOR OF PHILOSOPHY (CLINICAL EPIDEMIOLOGY)
FACULTY OF GRADUATE STUDIES MAHIDOL UNIVERSITY
2017
COPYRIGHT OF MAHIDOL UNIVERSITY
Thesis entitled
THE ASSOCIATION BETWEEN PERIODONTITIS AND CHRONIC KIDNEY DISEASE
............................................................. Mr. Attawood Lertpimonchai Candidate ............................................................. Asst. Prof. Sasivimol Rattanasiri, Ph.D. (Statistics) Major advisor ... ......................................................... ............................................................. Prof. Piyamitr Sritara, Assoc. Prof. Ammarin Thakkinstian, M.D., FRCPT, FACP, FRCP (T) Ph.D. (Clinical Epidemiology & Co-advisor Community Medicine) Co-advisor ............................................................ ............................................................ Asst. Prof. Chantrakorn Champaiboon, Assoc. Prof. Atiporn Ingsathit, D.D.S., Ph.D. (Oral Biology) M.D., Ph.D. (Clinical Epidemiology) Co-advisor Co-advisor
............................................................ ........................................................... Assoc. Prof. Varaporn Akkarapatumwong, Assoc. Prof. Ammarin Thakkinstian, Ph.D. (Science) Ph.D. (Clinical Epidemiology & Acting Dean Community Medicine) Faculty of Graduate Studies Program Director Mahidol University Doctor of Philosophy Program in Clinical Epidemiology Faculty of Medicine, Ramathibodi Hospital, Mahidol University
Thesis entitled
THE ASSOCIATION BETWEEN PERIODONTITIS AND CHRONIC KIDNEY DISEASE
was submitted to the Faculty of Graduate Studies, Mahidol University
for the degree of Doctor of Philosophy (Clinical Epidemiology) on
November 1, 2017
............................................................. Mr. Attawood Lertpimonchai Candidate .. ........................................................... Asst. Prof. Chusak Okascharoen M.D., Ph.D. (Clinical Epidemiology) Chair ............................................................ .... ......................................................... Prof. Piyamitr Sritara, Asst. Prof. Sasivimol Rattanasiri, M.D., FRCPT, FACP, FRCP (T) Ph.D. (Statistics) Member Member ............................................................ ............................................................. Prof. Rangsini Mahanonda, Assoc. Prof. Atiporn Ingsathit, D.D.S., Ph.D. (Oral Biology) M.D., Ph.D. (Clinical Epidemiology) Member Member ............................................................ ............................................................. Asst. Prof. Chantrakorn Champaiboon, Assoc. Prof. Ammarin Thakkinstian, D.D.S., Ph.D. (Oral Biology) Ph.D. (Clinical Epidemiology & Member Community Medicine) Member
……………………………………... ……………………………………... Assoc. Prof. Varaporn Akkarapatumwong, Prof. Piyamitr Sritara, Ph.D. (Science) M.D., FRCPT, FACP, FRCP (T) Acting Dean Dean Faculty of Graduate Studies Faculty of Medicine Ramathibodi Mahidol University Hospital Mahidol University
iii
ACKNOWLEDGEMENTS
The success of this thesis can be succeeded by the attentive support from
my major advisor, Asst. Prof. Sasivimol Rattanasiri, the Program director, Prof Dr.
Ammarin Thakkinstian for the precious instruction and advice in all processes of my
PhD. I am heartily thankful to my co-advisors, Asso. Prof. Dr. Atiporn Ingsathit, and
Asst. Prof. Chantrakorn Champaiboon for excellence recommendations and intuitive
ideas in specific fields.
I wish to express my appreciation to Prof. Piyamitr Sritara, Prof. Rangsini
Mahanonda and Assoc. Prof. Suphot Tamsailom for introducing the EGAT project to
me and allowed to utilize the EGAT data. I also appreciate the dedication and continuing
support on statistical analysis from Dr. Win Khaing in this project.
Finally, I would like to express endless thankfulness to all my families for
their support throughout my study.
This thesis was supported by Chulalongkorn University (Government
Budget Grant 2015)
Attawood Lertpimonchai
Attawood Lertpimonchai Background and rationale / iv
THE ASSOCIATION BETWEEN PERIODONTITIS AND CHRONIC KIDNEY DISEASE
ATTAWOOD LERTPIMONCHAI 5638086 RACE/D
Ph.D. (CLINICAL EPIDEMIOLOGY)
THESIS ADVISORY COMMITTEE: SASIVIMOL RATTANASIRI, Ph.D., AMMARIN
THAKKINSTIAN, Ph.D., ATIPORN INGSATHIT, M.D., Ph.D., CHANTRAKORN
CHAMPAIBOON, D.D.S., Ph.D., PIYAMITR SRITARA, M.D., FRCPT.
ABSTRACT
Periodontitis and diabetes mellitus (DM) are suspected to be risk factors of chronic
kidney disease (CKD), but whether they were directly associated or mediated across each other is
still unknown. Therefore, this study aimed to determine the direct and mediation effects of
periodontitis through DM on CKD or vice versa. This was 10-year-retrospective non-CKD cohort
study that examined 2,635 employees of the Electric Generation Authority of Thailand (EGAT). The
interested outcome was the CKD incidence, defined as estimated glomerular filtration rate (eGFR)
< 60 ml/min per 1.73 m2 at follow-up. The periodontitis extent using the percentage of proximal sites
with severe periodontitis (Clinical attachment loss ≥ 5 mm) was used as the representativeness of
periodontal status. Mediation analysis with 1000-replication bootstrap was applied to construct 2
possible causal pathways, pathway A (periodontitis → DM → CKD), and pathway B (DM →
periodontitis → CKD). The results showed that the cumulative incidence of CKD was 1.03 cases per
100 persons per year (95% CI: 0.91, 1.16). In pathway A, each increasing percentage of proximal
sites with severe periodontitis increased the adjusted odds ratio of CKD 1.010 (95% CI: 1.005, 1.015)
and 1.007 (95% CI: 1.003, 1.013), by direct effect and indirect effect through DM, respectively.
While, pathway B indicated that the participants with DM had odds of CKD incidence around 2
times higher than the non-DM, and this effect was attributed to 6.5% mediation effect via
periodontitis. It is therefore concluded that periodontitis had significant direct and indirect effect via
DM on CKD, and it is further suggested that awareness of oral and systemic morbidities from
periodontitis should be emphasized by public health practitioners.
KEY WORDS: PERIODONTITIS / CHRONIC KIDNEY DISEASE / DIABETES MILLITUS /
MEDIATION ANALYSIS
182 pages
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / v
ความสมพนธของโรคปรทนตอกเสบกบโรคไตเร9อรง
THE ASSOCIATION BETWEEN PERIODONTITIS AND CHRONIC KIDNEY DISEASE
อรรถวฒ เลศพมลชย 5638086 RACE/D
ปร.ด. (วทยาการระบาดคลนก)
คณะกรรมการทEปรกษาวทยานพนธ: ศศวมล รตนสร, Ph.D., อมรนทร ทกขญเสถยร, Ph.D., อตพร องคสาธต, M.D., Ph.D., จนทรกร แจมไพบลย D.D.S., Ph.D., ปยะมตร ศรธรา, M.D., FRCPT.
บทคดยอ
โรคปรทนตอกเสบและโรคเบาหวานถกเชEอวาเปนปจจยเสEยงของโรคไตเร9 อรง ท9งน9 กลไกและข9นตอนของความสมพนธของโรคปรทนตอกเสบและเบาหวานตออบตการณโรคไตเร9อรงยงไมเปนทEรแนชด การศกษาน9 จงมวตถประสงคทEจะระบความสมพนธของโรคปรทนตอกเสบและโรคเบาหวานในการเพEมความเสEยงของการเกดโรคไตเร9 อรงดวยการวเคราะหโมเดลตวแปรคEนกลาง (Mediation analysis)
โดยการศกษาน9 เปนการศกษาระยะยาว 10 ป ในกลมพนกงานการไฟฟาฝายผลตแหงประเทศไทย ทEมการทางานของไตเปนปกตจานวน 2,635 ราย ตวแปรผลลพธทEสนใจคออบตการณการเกดโรคไตเร9 อรง โดยนยามจากคาอตราการกรองไตทEนอยกวา 60 มล./นาท ตอ 1.73 ตารางเมตร ในขณะทEภาวะโรคปรทนตอกเสบใชคาการกระจายตวของโรค หรอคารอยละของตาแหนงดานประชดทEมการทาลายอวยวะยดเกาะปรทนตระดบรนแรง (CAL ≥ 5 มม.) เปนตวแปรแทนสภาวะปรทนต การวเคราะหทางสถตใชการวเคราะหโมเดลตวแปรคEนกลาง ภายใตสมมตฐานการเกดโรค 2 รปแบบ คอ รปแบบ ก (โรคปรทนตอกเสบ →
โรคเบาหวาน → โรคไตเร9อรง) และ รปแบบ ข (โรคเบาหวาน →โรคปรทนตอกเสบ → โรคไตเร9อรง) ผลการศกษาพบวาอบตการณการเกดโรคไตเร9 อรงเทากบ 1.03 ตอประชากร 100 คน ตอป จากความสมพนธรปแบบ ก พบวาทกรอยละทEเพEมข9นของตาแหนงดานประชดทEมเปนโรคปรทนตอกเสบระดบรนแรงจะเพEมความเสEยงของการเกดโรคไตเร9อรงทางตรงและทางออมผานโรคเบาหวาน 1.010 (95%
CI: 1.005, 1.015) และ 1.007 (95% CI: 1.003, 1.013) เทา ตามลาดบ ในขณะทEความสมพนธรปแบบ ข
แสดงใหเหนวาตวอยางทEเปนโรคเบาหวาน มความเสEยงในการเกดโรคไตเร9 อรงมากกวาคนทEไมไดเปนประมาณ 2 เทา โดย 6.5% เปนความเสEยงทEเกดข9นผานโรคปรทนตอกเสบ โดยสรปโรคปรทนตอกเสบและโรคเบาหวานมผลเพEมความเสEยงของการเกดโรคไตเร9 อรง ท9งทางตรงและทางออม ดงน9นโรคปรทนตอกเสบจงควรไดรบการกระตนใหเหนความสาคญ
182 หนา
vi
CONTENTS
Page ACKNOWLEDGEMENTS iii
ABSTRACT (ENGLISH) iv
ABSTRACT (THAI) v
LIST OF TABLES ix
LIST OF FIGURES xi
LIST OF ABBREVIATIONS xiii
CHAPTER I BACKGROUND AND RATIONALE 1
1.1 Background and rationale 1
1.2 Research question 3
1.3 Research objectives 3
CHAPTER II LITERATURE REVIEW 4
2.1 Epidemiology of CKD 4
2.2 Identification and classification of CKD 5
2.3 Risk factors of CKD 7
2.4 Epidemiology of periodontitis 8
2.5 Periodontal medicine 8
2.6 Periodontitis classification 9
2.7 The association between OH and periodontitis: A systematic 10
review and meta-analysis
2.8 Effect of periodontitis on renal function/CKD: Biological 19
plausibility
2.9 Association between periodontitis and CKD: 19
Epidemiological studies
2.10 Association between periodontitis and DM 22
2.11 DM as the risk factor of CKD 23
CHAPTER III METHODOLOGY 67
3.1 Study design and setting 67
vii
CONTENTS (cont.)
Page
3.2 Study subjects 69
3.3 Data collection 69
3.4 Study factor and measurements 70
3.5 Primary outcome and measurements 72
3.6 Other co-variables and measurements 72
3.7 Sample size estimation 74
3.8 Data management 74
3.9 Imputation 77
3.10 Statistical analysis 79
3.11 Ethics considerations 84
CHAPTER IV RESULTS 101
4.1 Characteristic of subjects 101
4.2 Missing data and imputation results 102
4.3 Pathway A: Periodontitis → DM → CKD 102
4.4 Pathway B: DM → Periodontitis → CKD 105
4.5 Assumption checking 107
CHAPTER V DISCUSSION 142
5.1 Main findings 142
5.2 Comparison results with previous studies 142
5.3 Data management for EGAT cohort 144
5.4 Periodontal measurement and classification 144
5.5 Multiple imputation 145
5.6 Selection of covariables for mediation analysis 146
5.7 Strengths of this study 147
5.8 Limitations 147
5.9 Clinical application 148
5.10 Suggestion for further studies 149
5.11 Conclusion 149
viii
CONTENTS (cont.)
Page
REFERENCES 150
APPENDICES 172
Appendix A Modified Newcastle-Ottawa Quality Assessment Scale 173
Appendix B The GRADE APPROACH 180
Appendix C Ethical approval 181
BIOGRAPHY 182
ix
LIST OF TABLES
Table Page
2.1 2009 CKD-EPI creatinine equations 26
2.2 Search terms and search strategy: Periodontitis and OH 27
2.3 Characteristics of included studies 28
2.4 Risk of bias assessment 34
2.5 Categorization of OH level 36
2.6 Pooling effects of fair and poor versus good OH on periodontitis 37
2.7 Subgroup and sensitivity analysis according to sources of 38
heterogeneity of fair and poor versus good OH
2.8 Pooled mean difference of OH score between periodontitis 39
and non-periodontitis
2.9 Pooled effect size of oral care habit on periodontitis 40
2.10 Sources of heterogeneity of tooth brushing meta-analysis 41
2.11 Publication bias assessment by Egger test 42
2.12 Overview of the meta-analysis 43
2.13 Search terms and search strategy: Periodontitis and CKD 44
2.14 Case-control studies: Periodontitis and CKD 45
2.15 Cross-sectional studies: Periodontitis and CKD 49
2.16 Cohort studies: Periodontitis and CKD 54
3.1 Calibration of periodontal examination (weight kappa ± 1mm) 85
3.2 Imputation models: predictors and equations 86
4.1 Pattern of participation 108
4.2 Baseline characteristics of excluded cases 109
4.3 Baseline characteristics 112
4.4 Numbers of missing data 115
4.5 Within and whole wave missing data 116
4.6 RVI, FMI and Relative efficiency in GSEM final models 117
x
LIST OF TABLES (cont.)
Table Page
4.7 Comparison of characteristic between actual and imputed dataset 118
4.8 The F-test and coefficients of various periodontitis and 127
obesity definitions
4.9 Univariate GSEM of DM model: Mediation model 128
4.10 Univariate GSEM of CKD model: Outcome model 129
4.11 Multivariate GSEM of mediation and outcome models of Pathway A: 130
Periodontitis F
4.12 Multivariate GSEM of mediation and outcome models of Pathway A: 131
CDC/AAP
4.13 Casual effects of Periodontitis on CKD through DM (Pathway A) 132
4.14 Comparisons of forward stepwise method versus disjunctive clause 133
criteria for DM Model
4.15 Comparisons of forward stepwise method versus disjunctive clause 134
criteria for CKD Model
4.16 Univariate GSEM of Periodontitis model: Mediation model 135
4.17 Multivariate GSEM of mediation and outcome models of Pathway B 136
4.18 Casual effects of DM on CKD through Periodontitis (Pathway B) 137
4.19 Comparisons of forward stepwise method versus disjunctive clause 138
criteria for Periodontitis Model
xi
LIST OF FIGURES
Figure Page
2.1 Flow chart of identifying and selecting studies: Periodontitis and OH 57
2.2 Pooling effects of fair and poor versus good OH on periodontitis 58
2.3 Pooling ORs of Plaque index and Plaque score on periodontitis 59
2.4 Pooling effect of oral care habits on periodontitis 60
2.5 Funnel plots of publication bias assessment 61
2.6 Contour enhanced funnel plots 62
2.7 Summary of pooled effect of OH and oral care habits on 63
periodontitis
2.8 Flow chart of identifying and selecting studies: Periodontitis and CKD 64
2.9 Bi-directional relationship between diabetes and periodontitis 65
2.10 Mechanisms of renal injury due to diabetes 66
3.1 Structural causal pathway A: Periodontitis → DM → CKD 87
3.2 Structural causal pathway B: DM → Periodontitis → CKD 88
3.3 Periodontal measurement 89
3.4 Workflow of cleaning processes for gender 90
3.5 Workflow of cleaning processes for date visit (survey date) 91
3.6 Workflow of cleaning processes for date of birth 92
3.7 Workflow of cleaning processes for marital status 93
3.8 Workflow of cleaning processes for education 94
3.9 Workflow of cleaning processes for risk behaviors including smoking 95
and alocohol drinking
3.10 Workflow of cleaning processes for height 96
xii
LIST OF FIGURES (cont.)
Figure Page
3.11 Workflow of cleaning processes for weight, waist and hip 97
circumstance
3.12 Workflow of cleaning processes for blood pressure 98
3.13 Workflow of cleaning processes for laboratory results 99
3.14 Outline of statistical analysis 100
4.1 Flowchart of included subjects 139
4.2 Causal mediation pathway diagram of Pathway A using 140
generalized structural equation modelling
4.3 Causal mediation pathway diagram of Pathway B using 141
generalized structural equation modelling
xiii
LIST OF ABBREVIATIONS
AAP American Academy of Periodontology
ACME Average causal mediation effect
ACR Albumin-to-creatinine ratio
AGE Advanced glycosylation end products
AKI Acute kidney injury
ALT Alanine aminotransferase
AP Alkaline phosphatase
ARIC Atherosclerosis Risk in Communities
AST Aspartate aminotransferase
ASVD Atherosclerotic vascular diseases
BMI Body mass index
BOP Bleeding on probing
BP Blood pressure
CAL Clinical attachment level
CBC Complete blood count
CDC Centers for Disease Control and Prevention
CI Confidence interval
CKD Chronic kidney disease
CKD-EPI Chronic Kidney Disease Epidemiology Collaboration
COPD Chronic obstructive pulmonary disease
CPI Community periodontal index
CRF Case record (report) form
CRP C-reactive protein
CVD Cardiovascular disease
DBP Diastolic blood pressure
DE Direct effect
dl Deciliter
xiv
LIST OF ABBREVIATIONS (cont.)
DLP Dyslipidemia
DM Diabetes mellitus
DOB Date of birth
EFP European Federation of Periodontology
EGAT Electricity Generating Authority of Thailand
EKG Electrocardiography
ESRD End-stage renal disease
exp Exponential
FBS Fasting blood sugar
FMI Fraction of missing information
eGFR (estimated) Glomerular filtration rate
GI Gingival index
GSEM Generalized structural equation modelling
GGT Gamma-glutamyl transpeptidase
HD Hemodialysis
HDL High density lipoprotein HIV Human immunodeficiency virus
HR Hazard ratio
HT Hypertension
IL Interleukin
KDOQI Kidney Disease Outcomes Quality Initiative
KDIGO Kidney Disease: Improving Global Outcomes
KNHANES Korea National Health and Nutrition Examination Survey
LDL Low density lipoprotein
LOA Loss of attachment index
MAR Missing at random
MDRD Modification of Diet in Renal Disease
ME Mediation effect
MeSH Medical subject headings
xv
LIST OF ABBREVIATIONS (cont.)
mg Milligram
MICE Multiple imputation with chained equations
mmHg Millimeter mercury
MrOS Osteoporotic Fractures in Men
NHANES National Health and Nutrition Examination Survey
NSAIDs Non-steroidal anti-inflammatory drugs
OH Oral hygiene
OHI Oral hygiene index
OHI-S (Simplified) Oral hygiene index
OR Odds ratio
PD Peritoneal dialysis
PDI Periodontal disease index
PGE2 Prostaglandin E2
PI Plaque index
PISA Periodontal inflamed surface area
PKC Protein kinase C
PPD Periodontal pocket depth
PRISMA Preferred Reporting Items for Systematic Reviews and Meta-
analysis
PROSPERO International prospective register of systematic reviews
PSc Plaque score
RA Rheumatoid arthritis
RE (Gingival) Recession
RR Risk ratio
RRT Renal replacement therapy
RVI Relative variance increases
SBP Systolic blood pressure
ScC Serum cystatin C
SD Standard deviation
xvi
LIST OF ABBREVIATIONS (cont.)
SE Standard error
SMD Standardized mean difference
TNF Tumor necrosis factor TC Total cholesterol
TE Total effect
TG Triglyceride
US United State
var Variance
WHR Waist-to-hip ratio
WHO World health organization
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 1
CHAPTER I
BACKGROUND AND RATIONALE
1.1 Background and Rationale Chronic kidney disease (CKD) is defined as reduced glomerular filtration
rate (GFR), increased urinary albumin excretion, or both1. Renal impairments lead to
complex disorders that are characterized by solute retention, abnormality of hormones
regulation, and compensatory responses in other organ systems. Once, its progression
reaches to the kidney failure, patients are required renal replacement therapies (RRT),
e.g., peritoneal dialysis (PD), hemodialysis (HD), or renal transplantation. These
treatment modalities are high costs and impact on the quality of life. Moreover, CKD is
also an important cause of other subsequent health burden including infection,
cardiovascular disease (CVD) and death2. From here, CKD is recognized as a global
health problem. Despite of intensive improvements in medicine and research, the
number of patients with CKD is expected to grow continuously, worldwide3. Promising
to reduce the burden and cost of care from CKD, periodontitis, a novel and potentially
modifiable risk factor has been extensively mentioned and explored4.
Periodontitis, the most common oral diseases, is generally described as a
specific infectious disease. It causes the gum infection and destruction of tooth-
supporting periodontium. Its morbidities are not limited only to oral cavity, but also
influence overall health through systemic inflammation5. In periodontal pockets,
gingival epithelium is infected, and injured with the numerous small ulcerations
distributing throughout the pocket wall. In a patient with moderate to severe
periodontitis, the total surface area of these ulcerations is approximately the same size
of a human palm. This ulcerated epithelium always exposed to thousands of
microorganisms in dental biofilm. And thus, intense inflammatory response occurs, then
it possibly wide spreads. Bacteria and inflammatory mediators can reach to other
internal organs through bloodstream. The evidence showed that periodontitis patients
had significantly elevated level of serum C-reactive protein (CRP), a marker of systemic
Attawood Lertpimonchai Background and rationale / 2
inflammation6. As a result, it could imply that periodontitis had potentially effects on
overall systemic health, such as, increased risk of CVD events, which was confirmed by
founded periodontal pathogens in the atheromatous plaque in vessel walls of CVD
autopsy7.
Here, we hypothesize that periodontal infection might be one of the factors
accelerating deterioration in renal function. Possible mechanisms could be direct and
indirect pathways. The direct pathway may cause by bacteria and their products which
directly damaged the nephron. Additionally, the indirect pathway referred to renal
impaired by inflammatory cytokines5.
Beyond this, previous evidences showed that periodontitis related with
glycemic control impairment8, 9. Inflammatory mediators, such as interleukin (IL),
tumor necrosis factor (TNF), and bacterial endotoxins had been reported about
influencing on insulin resistance. Hence, periodontitis could increase risk of diabetes
mellitus (DM) incidence. Simultaneously, DM has been recognized as the conventional
risk factors of CKD. Metabolic pathways, including advanced glycosylation end
products (AGEs), activation of protein kinase C (PKC), oxidative stress, and
acceleration of the polyol pathway were suggested to be mediators of diabetic
nephropathy10.
On the other hand, initiated the causation with DM, periodontitis could be
the adverse consequences of DM8, 11. Impaired immunity function, imbalance of
collagen homeostasis and altered saliva compositions aggravated periodontal
inflammation, and deteriorated supporting periodontium leading to periodontitis. The
cycle of spreading periodontal pathogens and cytokines is initiated, and then, risk of
CKD is amplified.
Toward this end, with the uncertainty of complex causal pathway, the
mediation effects of periodontitis on CKD through DM, and vice versa, have been
mentioned. Periodontitis could influence directly on CKD, and it also could affect CKD
indirectly through DM. The other side of the coin, the causative pathway could be
flipped between periodontitis and DM. DM could increase risk of CKD directly, and it
could be mediated across periodontal infection.
Current publications do not answer these questions, and thus, the gaps of
knowledge are still present. Evidences from cross-sectional studies12-25 showed the
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 3
significant association between periodontitis and CKD. For example, Kshirsagar et al 21
indicated that severe periodontitis subjects had about 1.19 to 3.85 odds of having CKD
after adjustment for traditional CKD risk factors. However, a causal relationship cannot
be claimed from the cross-sectional design. Although some previous cohorts26-31
proposed independent effects of periodontitis and DM on CKD, they did not assess
effects of periodontitis on CKD that transmitted through DM. To answer these
questions, a well-planned cohort studies in general population are needed. Therefore,
this study was conducted utilizing data from the cohort of employees of the Electricity
Generating Authority of Thailand (EGAT) to assess the causative association between
periodontitis, DM and CKD using a mediation analysis. The direct effect and indirect
(mediation) effect of paths would be identified.
1.2 Research Question Does periodontitis directly affect on renal function or its effect is transmitted
or mediated through DM, or vice versa?
1.3 Research Objectives The objectives of the study were:
1.3.1 To determine direct and indirect effects of periodontitis through DM
on developing renal function in adults with normal kidney function.
1.3.2 To determine direct and indirect effect of DM on renal function
through periodontitis in adults with normal kidney function at baseline.
Attawood Lertpimonchai Literature review / 4
CHAPTER II
LITERATURE REVIEW
2.1 Epidemiology of CKD CKD is defined by the Kidney Disease: Improving Global Outcomes
(KDIGO) guideline as abnormalities of kidney structure or function, present for more
than 3 months, with implications for health32. It is associated with impaired quality of
life and substantially reduced life expectancy at all ages. Impairment and damage of
kidneys lead to complex disorders that are characterized by solute retention, hormone
deficiencies or resistance, and compensatory responses in other organ systems33.
Prevalence and incidence of CKD were diverse among publications due to
heterogeneous of populations, calculation of estimated glomerular filtration rate
(eGFR), and proteinuria measurements. Despite these limitations, the prevalence of
CKD is consistently reported to be around 10–15% in high to moderate income
countries34. The low-socioeconomic people are at higher risk of developing CKD and
of disease progression. In Thailand, the prevalence of CKD is about 17.5% (95% CI:
14.6, 20.4)35.
The burden of CKD is substantial in term of mortality and morbidity. Once,
CKD progression reaches to the kidney failure, its function is no longer able to sustain
life over the long term. The RRTs, e.g., PD, HD, or renal transplantation are required.
These treatment modalities are high costs and impact on the quality of life. Globally,
only half of those patients can afford and receive the RRTs34. CKD complications may
occur at any stage of disease, and could occur from the adverse effects of CKD
treatments. Anemia, infections, CKD mineral bone disease, CVD, cancer, drug toxicity
and cognitive impairment are common which can lead to death without progression to
kidney failure2, 34.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 5
2.2 Identification and classification of CKD According to KDIGO guideline32, eGFR and albuminuria are recommended
for CKD identification and classification. eGFR is used to determine kidney function,
while, albuminuria is the standard maker for kidney damage.
2.2.1 Estimated glomerular filtration rate
The ideal kidney function measurement or actual GFR can be measured by
renal clearance of exogenous filtration markers, such as inulin. However, it is
cumbersome and impractical. The eGFR has been alternatively proposed. Serum
creatinine and cystatin C are common biomarkers which used to assess kidney function.
Creatinine, by-product of muscle metabolism, usually produces at a constant rate and
freely filters by the glomerulus. Increasing serum creatinine represents GFR decrease.
However, it has the positive correlation with muscle mass. Hence, age, sex and ethnicity
are always used for compensation in eGFR calculation. Other factors, for example,
protein intake, physical activity, tubular secretion, extra-renal excretion and creatinine
degradation also influence serum creatinine concentration.
The serum cystatin C, the low-molecular-weight protein produces in all
nucleated cells, also has been recommended as another maker of kidney function. It is
claimed the advantage about less affected by muscle mass and diet. However, its
concentration is also influenced by age, sex, inflammation, corticosteroid use, smoking
and hyperthyroidism36.
The 2009 CKD-EPI creatinine equations (Table 2.1), the 2012 CKD-EPI
cystatin C equations, or the 2012 CKD-EPI creatinine–cystatin C equations are
recommended as the eGFR estimating equations37. If data available, a combination of
creatinine and cystitis C in eGFR estimation might improve accuracy, especially for the
lower ranges of GFR32.
2.2.2 Proteinuria
Proteinuria is clinically used for identifying the kidney damage with
increased glomerular permeability allowing the filtration of macromolecules that should
remain within the circulation. The urinary dipsticks or measurement of the albumin or
total protein concentration in a spot urine sample is the conventional methods for
Attawood Lertpimonchai Literature review / 6
proteinuria. These reagent strip devices primarily detect albumin by a colorimetric
reaction with the dipstick-impregnated reagent based on the albumin concentration
within the sample. However, it has some limitations. First, it is low validity at detecting
low concentration but clinically significance of urinary albumin. Second, false positive
is occasionally founded in concentrated or highly alkaline urine, after use of iodinated
contrast agents, or in case of gross hematuria. To compensate for variations in urine
concentration in spot-check samples, it is helpful to compare the amount of albumin in
the sample against its concentration of creatinine. This is termed the albumin-to-
creatinine ratio (ACR)1, 2.
2.2.3 CKD classification
In 2002, the Kidney Disease Outcomes Quality Initiative (KDOQI)
organization, proposed a diagnosis and classification guideline for CKD. Using eGFR,
mainly, severity of CKD was classified into five stages. The presence of proteinuria was
only mandatory for stages 1 and 2 or patients with eGFR > 60 ml/min per 1.73 m2.
Patients with stage 3 or higher were defined as CKD without proteinuria results.
However, recent studies showed that the presence of proteinuria increased risk of
mortality and morbidity. Moreover, risk of death and other complications were
discrepancy among stage 3 patients2. Therefore, in 2012, KDIGO revised the diagnosis
and classification guideline and recommended that individuals should be classified
according to eGFR and proteinuria. In addition, the diagnostic criteria for stage 3 should
be divided into 2 sub-stages, i.e., stage 3a (GFR: 45 – 59 ml/min per 1.73 m2), and stage
3b (GFR: 30 – 44 ml/min per 1.73 m2).
CKD classification by Cause, GFR and Albuminuria (CGA staging) was
proposed32. Briefly, 6 groups of eGFR and 3 levels of proteinuria are cross-tabulated,
resulting to 18 stages of CKD, i.e., G1A1 to G5A3. For example, G3aA2 represents the
stage of mildly to moderate decreased in kidney function (eGFR: 45-90 ml/min per 1.73 m2) and moderately increased albuminuria (ACR 30 - 300 mg/g). Here, it can be
used to inform the need for specialist referral, general medical management, and
indications for investigation and therapeutic interventions.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 7
2.3 Risk factors of CKD
CKD is the multi-factorial disease, both inherited and many acquired factors
were mentioned to be the causative. Much epidemiological evidences showed a link
between several risk factors and the initiation and the progression of CKD38-43. In
general, DM and hypertension (HT) are the well-known important risk factors of CKD3,
33, 34. DM and HT are the main causes of CKD in all high- and low-income countries.
The consistency results from previous literatures confirmed the increasing risk of CKD
incident and more rapidly progressive CKD from both. Metabolic pathways in glycemic
control, including AGEs, activation of PKC, oxidative stress, and acceleration of the
polyol pathway were suggested to be mediators of diabetic nephropathy. Increased
systemic and intraglomerular pressure and activation of various vasoactive hormone
also seems to be alternative common pathway of glomerulosclerosis and
tubulointerstitial fibrosis10, 44.
CKD commonly clusters within families, which implies to genetic or
familial predisposition. Genetic studies have suggested links between CKD and various
alterations or polymorphisms of candidate genes encoding putative mediators, including
the renin-angiotensin system. For instance, the APOL1 gene has been intensively
identified as a major susceptibility gene for kidney disease45, 46. Male, elderly people
and some ethnicities, i.e. African and Native Americans, might also be more susceptible
to CKD, which would explain the high proportions of these population groups in CKD
and end-stage renal disease (ESRD)33.
Nephrotoxic effects from toxic dosages of herbs, non-steroidal anti-
inflammatory drugs (NSAIDs), iodinated contrast media or interactions with
conventional medicines, as well as, environmental pollution of water by heavy metals
and of soil by organic compounds (including pesticides) are other risk factors of acute
kidney injury (AKI) and CKD. Infections are other major causes of AKI and CKD,
especially in resource-limited regions. Human immunodeficiency virus (HIV), Hepatitis
B, Hepatitis C, malaria, tuberculosis and various tropical diseases, have been shown to
be associated with increased risk of CKD. Malnutrition might be associated with a
reduction in the number of nephrons, and renal disease in later life. In addition, many
cohort studies have identified smoking, alcohol consumption, obesity, hyperlipidemia,
Attawood Lertpimonchai Literature review / 8
low/high birthweight, prematurity and kidney stones as risk factors or markers in the
general population for the development of CKD44.
2.4 Epidemiology of periodontitis Periodontal disease is generally described as specific disease that involve a
tooth supporting structure. It is infectious and recognized as one of the most common
chronic inflammatory diseases in human. From the 2009 and 2010 National Health and
Nutrition Examination Survey (NHANES), over 47% of the U.S. adults who aged 30
years and older, representing 64.7 million adults, had periodontitis. Moreover, in elders
aged 65 years and older, 64% had either moderate or severe periodontitis47.
Periodontium consists of gingiva, alveolar bone, cementum and periodontal
ligament. It always exposes to oral bacteria in dental plaque. Fortunately, with the
effective immune system, the balance between bacteria and host response is maintained
without developing periodontal disease. However, when the disequilibrium occurs,
either host hyper-responsiveness or high virulence from bacteria, periodontium will be
destructed by inflammation process. In the initial stage, patients usually do not have any
symptom leading to neglect their periodontal health care. Then, the disease progresses,
patients may suffer from easily gingival bleeding, gum swelling, dull pain, gingival
abscess, and tooth mobility. Finally, patients will lose their teeth, decreasing in occlusal
function and digestive ability. Then, it will influence the overall health, leading to
decrease their quality of life48.
2.5 Periodontal medicine Current evidences showed the potential effects of periodontal disease on
overall systemic health based on concept of inflammation diffusion, including
periodontal pathogen and cytokines5. The association between atherosclerotic vascular
diseases (ASVD) and periodontitis has been established49, and then, has been extend
investigated since 1990. Systemic inflammatory makers, such as CRP, which was the
surrogate outcome of ASVD, increased with periodontitis infection. Periodontitis also
associated with the decreased endothelial dysfunction and increased carotid intima-
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 9
media thickness. In addition, periodontal pathogens could be found in the autopsy of
atheromatous plaque along vessel walls of patients with CVD7. Moreover, periodontitis
was proved as the risk factors of future ASVD events with the cohort studies50. From
this, the joint of European Federation of Periodontology and American Academy of
Periodontology (EFP/AAP) Workshop on Periodontitis and Systemic Diseases stated
the consensus that there is consistent and strong epidemiologic evidence that
periodontitis increased risk for future CVD51.
In pregnant mothers with periodontitis, oral bacteria were found in an
amniotic fluid, along with increasing of inflammatory cytokines. It could possible lead
to adverse pregnancy outcomes, such as premature labor or preterm low-birth-weight
infants52. Furthermore, periodontitis mediators, such as IL-1ß, TNF-α, associated with
increasing insulin resistance, which developed the risk of develop DM and its
complications8. Relationships between periodontitis and other systemic disease, i.e.,
CKD, chronic obstructive pulmonary disease (COPD), rheumatoid arthritis (RA),
Alzheimer's disease and erectile dysfunction, also have been reported53.
2.6 Periodontitis classification Periodontal diseases are simply categorized based on disease severity into
two distinct forms, gingivitis and periodontitis. Gingivitis is inflammation of the
gingiva, meanwhile, periodontitis is inflammation of periodontium that extends beyond
gingiva and produces destruction of periodontal ligament and alveolar bone. From the
AAP in 1999, the severity of periodontitis depends on amount of clinical attachment
loss (CAL)54. Chronic periodontitis is characterized by periodontal pocket depth (PPD)
more than 3 mm with CAL. The extent of disease has been further classified as localized
or generalized depending on the proportion of infected sites (≤ 30% or > 30%). Severity
is measured on the amount of CAL, and it is designated as slight (1-2 mm), moderate
(3-4 mm) or severe (≥ 5 mm).
Although, the AAP 1999 classification has been world-wide accepted for
the clinical circumstance, it has not been widely adopted in periodontal research.
Various case definitions of periodontitis have been proposed55. It has great impact on
the prevalence and extent of periodontal disease56. Moreover, these discrepancies
Attawood Lertpimonchai Literature review / 10
diverged the results of investigating the link between periodontitis and systemic
diseases57. Due to lack of universally accepted, the Centers for Disease Control and
Prevention (CDC), in partnership with the AAP proposed and updated the standard case
definitions58, 59. These definitions have been proposed base on their primary objective
which was to examine and identify valid nonclinical measures for population-based
surveillance of periodontitis. Some literatures in periodontal medicine adopted the
CDC/AAP definition for categorizing periodontal status. However, as aforementioned,
it has not proposed to serve for identifying the linkage between periodontitis and
systemic health. Therefore, we conducted a systematic review, which aimed to review the
periodontitis definition in periodontal researches. Association between oral hygiene
(OH) & personal oral health care and periodontitis was selected as the topic of review,
because it was the basic and common knowledge in periodontal researches. It has been
extensive explored in the previous literatures. From this, we assume that, the various of
periodontitis definitions would been more covered with this systematic review and
meta-analysis.
2.7 The association between OH and periodontitis: A systematic review
and meta-analysis
2.7.1 Methods
A systematic review and meta-analysis was conducted in order to assess
effects of association between OH and periodontitis. The review proposal was
developed following the PRISMA guidelines60, and it was registered at PROSPERO
(registration number: CRD42015019036).
2.7.1.1 Search strategy
The MEDLINE and SCOPUS databases were used to identify
relevant studies with standardized methodological filters up to May 2016. Search
strategies were mainly constructed based on the primary objective with 3 domains (i.e.,
periodontitis, OH, and general aspects for observational studies) as follows:
(“periodontitis” OR “periodontal”) AND (“poor oral hygiene” OR “plaque index” OR
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 11
“oral hygiene index” OR “plaque score”) AND (“relation” OR “association” OR “risk
factor”). The search terms and strategies are described in Table 2.2.
2.7.1.2 Inclusion criteria
Studies were screened based on titles and abstracts; if a decision
could not be made based on this information, full papers were reviewed. Any type of
observational study (e.g., cohort, case-control, or cross-sectional studies) published in
English was included if it met the following criteria: (i) assessed associations between
OH and periodontitis in either general or specific types of adult populations; (ii) had at
least 2 outcome groups, periodontitis versus non-periodontitis, or mild, moderate and
severe periodontitis versus normal periodontium; (iii) assessed OH by standard tools,
such as the Oral Hygiene Index (OHI) or Simplified Oral Hygiene Index (OHI-S)61,
Plaque Index (PI)62, plaque control record/Plaque Score (PSc)63, or a questionnaire
including the frequency of brushing, interdental cleaning and dental visits; (iv)
reported/possibly calculated the mean and standard deviation (SD) of OH scores among
periodontitis groups or a contingency table between non-periodontitis/periodontitis and
OH groups. Studies were excluded if they had insufficient data for pooling after
contacting the authors for additional data.
Two of three reviewers (AL, SR, and SA) independently
evaluated the studies for eligibility, extracted the data, and assessed the risk of bias. Any
discrepancies between reviewers were discussed and resolved by consensus.
2.7.1.3 Study factors
The primary study factor was OH, objectively measured using
OHI, PI or PSc. Secondary study factors were oral care habits, which were subjectively
assessed using questionnaires assessing the frequency of tooth brushing, interdental
cleaning, and dental visits.
2.7.1.4 Outcome
The outcome of interest was periodontitis, which was defined
according to the original studies. The definition of periodontitis was based either on
PPD, CAL or radiographs without a restricted periodontitis definition.
2.7.1.5 Data extraction
Study characteristics, including study design (cohort, case-
control or cross-sectional study), population type (general population or specific
Attawood Lertpimonchai Literature review / 12
disease), and study base (community or hospital) were extracted. Subject characteristics
(i.e., percentage of males, smoking habits and the presence of DM) and clinical data
(i.e., periodontitis definition and details of OH assessments) were also extracted.
2.7.1.6 Risk of bias assessment
The quality of the studies was assessed using the modified
Newcastle-Ottawa Quality Assessment Scale64 (Appendix A), which considers 3
domains: the representativeness of the studied subjects, the comparability between
groups, and the ascertainment of outcome and exposure. Each domain was graded by
giving stars if there was a low risk of bias. Individual studies were categorized according
to these stars as having a low, moderate and high risk of bias if the percentage of stars
was ≥ 75%, 50-74% and < 50%, respectively.
2.7.1.7 Statistical analysis
Data were pooled if there were at least 2 studies reporting the
same outcomes and study factors. Data analysis was performed separately by the type
of OH data (i.e., categorical and continuous data) as described below.
For categorical data, the odds ratio (OR) of having periodontitis
for fair versus good OH (OR1) and poor versus good OH (OR2) along with their 95%
confidence interval (CI) were estimated for each study. Given included studies with 2
or more OH groups, a multivariate random-effects meta-analysis was applied for
pooling ORs. This method considers within-study variation using Riley’s method65, 66.
For studies in which OH was divided into more than 2 groups and ORs were reported
without frequency data, the variance-covariance was assumed to be zero.
For continuous data, the mean difference in OH scores between
periodontitis and non-periodontitis groups was estimated and pooled using a
standardized mean difference (SMD). If logistic model correlation coefficients were
reported instead of the mean and SD, the beta coefficients were then pooled using the
pooling mean method.
Heterogeneity was assessed by Cochrane’s Q test and the I2
statistic. If heterogeneity was present (Q test < 0.1 or I2 ≥ 25%), a random-effects model
(Dersimonian & Laird) was used. Otherwise, a fixed-effects model was applied with the
inverse variance method.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 13
Sources of heterogeneity were explored using a Galbraith plot
to identify outlier studies. Co-variables (i.e., population type, age, gender, smoking,
DM, index use, periodontitis definition) were then fitted one-by-one into a meta-
regression model. If there was a suggested association, a sensitivity analysis excluding
the outlier studies and/or a subgroup analysis was performed.
Finally, potential publication bias was explored using the Egger
test and a funnel plot. If either of these indicated asymmetry, a contour-enhanced funnel
plot was constructed to identify the cause of asymmetry. All analyses were performed
using STATA software version 14.2. Two-sided P < 0.05 was considered statistically
significant except for the heterogeneity test, in which P < 0.10 was used.
2.7.1.8 Grade of evidence
The system from the Grades of Recommendation, Assessment,
Development and Evaluation Working Group (GRADE Working Group)67, 68 was used
for grading the quality of evidence mainly based on the study design, risk of bias,
indirectness of evidence, publication bias, heterogeneity and imprecision of results.
2.7.2 Results
2.7.2.1 Identifying studies
A total of 2,763 studies were identified from MEDLINE and
SCOPUS, and 1934 studies remained after removing duplicates. Of these, 1,878 studies
were ineligible for reasons described in Figure 2.1, leaving 56 studies69-124 that were
eligible for review. Six studies69, 102, 103, 106, 107, 112 were excluded due to insufficient data
after contacting the authors. Of the remaining 50 studies, 45 studies70-73, 75-86, 88, 90-94, 97-
101, 104, 105, 108-111, 113-124 objectively assessed OH by oral examination. Fifteen70, 72, 77, 81-84,
86, 90-94, 101, 120 analyzed OH as categorical data, 3171, 73, 75-80, 85, 88, 97-100, 104, 105, 108-111, 113-
119, 121-124 as continuous data, and one77 as both. Eleven studies provided the association
between periodontitis and oral care habits measured by the frequency of brushing84, 87-
89, 91, 92, 95, 96, 99, 111, interdental cleaning84, 96, 99, 111 and dental visits74, 88, 89, 91, 95, 111.
2.7.2.2 Subject characteristics
The characteristics of the 50 included studies are described in
Table 2.3. Most study designs were cross-sectional, most studies investigated a general
population, and 34 studies were based in hospitals. The mean subject age ranged from
Attawood Lertpimonchai Literature review / 14
15 to 65 years. The percentages of males, smokers and diabetics are also shown in Table
2.3. While the definition of periodontitis varied across the studies, most (92%) used PPD
and/or CAL.
2.7.2.3 Risk of bias assessment
The results of the risk of bias assessments are described in Table
2.4. Most studies (72%) provided inadequate details for sample selection; hence,
representativeness was unclear. For example, some authors did not mention their
sampling methods or clearly describe their process for selecting cases and controls.
Twenty-seven studies (46%) were potentially biased due to improper statistical
adjustments for confounding factors. Almost all studies measured periodontitis via oral
examination, which was objective and valid. However, 16 studies (32%) used partial-
mouth examination protocols, 16 studies (32%) diagnosed periodontitis without data
regarding CAL, and 25 studies (50%) did not provide details about intra/inter-examiner
agreement. The numbers of studies with low, moderate and high risks of bias were 23,
19 and 8, respectively.
2.7.2.4 OH
For 15 studies with categorical OH data, 6 studies70, 84, 90, 93, 101,
120 categorized OH as good or poor, whereas 9 studies72, 77, 81-83, 86, 91, 92, 94 categorized
OH as good, fair, or poor. The criteria for classifying OH are presented in Table 2.5.
Pooled ln(ORs) determined using a multivariate meta-analysis (Figure 2.2) were 0.71
(95% CI: 0.50, 0.93) and 1.61 (95% CI: 1.22, 2.00), which yielded pooled ORs of 2.04
(95% CI: 1.65, 2.53) and 5.01 (95% CI: 3.40, 7.39), respectively, for fair and poor OH.
These results indicate that fair and poor OH increase the risk of periodontitis by
approximately 2- and 5-fold compared with good OH with an I2 of 40% and 78%,
respectively. The details of each individual study are shown in Table 2.6.
Population type appeared to be a large source of heterogeneity.
Subgroup analyses in community-based studies yielded lower heterogeneity levels, i.e.,
the I2 values were 4% and 0% for fair and poor versus good OH, respectively, with
corresponding pooled ORs of 2.23 (95% CI: 1.85, 2.69) and 4.78 (95% CI: 4.10, 5.58).
In addition, a sensitivity analysis focusing on 11 studies70, 72, 77, 81-84, 90-92, 94 of general
populations decreased the degree of heterogeneity to 22% and 49% for fair and poor
versus good OH, with pooled ORs of 2.10 (95% CI: 1.76, 2.49) and 4.21
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 15
(95% CI: 3.21, 5.51), respectively. Moreover, the periodontitis definitions and index
types used, as well as smoking behavior, also contributed to heterogeneity (Table 2.7).
Among 31 studies that measured OH on a continuous scale, 25
studies73, 79, 80, 88, 97-100, 104, 105, 108-111, 113-119, 121-124 compared OH using the mean scores
between periodontitis and non-periodontitis groups. The SMDs were highly
heterogeneous (I2 = 95.6%), with a pooled SMD of 2.04 (95% CI: 1.59, 2.50)
(Table 2.8). From these findings, it could be interpreted that periodontitis subjects had
a significantly higher OH score of 2.04 standardized units than non-periodontitis
subjects.
Six75, 77, 80, 85, 88, 97 and 371, 76, 78 studies reported the effects of PI
and PSc on periodontitis as coefficients (i.e., ln(OR)) of logistic regression models.
Pooling these corresponding effects yielded pooled ORs of 2.25 (95% CI: 1.43, 3.54)
and 1.02 (95% CI: 1.01, 1.03), and high heterogeneity was found for both (Figure 2.3).
These findings could be interpreted to indicate that each one-unit increase in the
measures of PI and PSc would increase the odds of having periodontitis by 2.25 and
1.02, respectively.
2.7.2.5 Oral healthcare habits
Ten84, 87-89, 91, 92, 95, 96, 99, 111, 484, 96, 99, 111, and 6 studies74, 88, 89, 91,
95, 111 assessed the effects of brushing, dental floss, and dental visits on periodontitis
(Table 2.9). The pooled ORs (Figure 2.4) suggested that tooth brushing and dental visits
were significantly associated with periodontitis, although the I2 values showed high
heterogeneity, at 94.5% and 60.4%, respectively. Subjects who brushed their teeth
regularly had approximately 34% significantly lower odds of having periodontitis
(pooled OR = 0.66; 95% CI: 0.47, 0.94). Smoking, the definition of regular brushing
and periodontitis were potential sources of heterogeneity (Table 2.10).
For dental visits, the sensitivity analysis was performed by
considering 4 of 6 studies that had clearly defined a regular dental visit as at least one
visit per year74, 88, 91, 111. This yielded a significant effect size of 0.56
(95% CI: 0.37, 0.83) with an I2 of 0%, indicating that subjects who regularly visited
dentists at least once a year had a 44% lower risk of periodontitis than those who did
not. The effects of interdental cleaning with dental floss on periodontitis showed little
Attawood Lertpimonchai Literature review / 16
heterogeneity (I2 = 5.1%), but the pooled OR was borderline significant (OR = 0.87;
95% CI: 0.75, 1.00).
2.7.2.6 Publication bias
Publication bias was assessed for all pooled estimates using
funnel plots (Figure 2.5) and Egger tests (Table 2.11). The results suggested symmetry
except for the mean differences in OH score, PSc, and dental visits. Contour-enhanced
funnel plots were further constructed (Figure 2.6), and these indicated that the
asymmetry of the funnels might be due to both heterogeneity and publication bias.
2.7.2.7 Quality of evidence
The scoring using the GRADE framework is shown in Table
2.12 and Appendix B. Based on observational studies, all pooled estimates were graded
as low quality68. For the effects of fair and poor OH on periodontitis, this was upgraded
to moderate quality because of large effect sizes and strong dose-response relationships.
The effects of brushing and dental visits were downgraded to very low quality due to
heterogeneity and publication bias, respectively.
2.7.3 Discussion
We conducted a systematic review and meta-analysis of the effects of OH
on periodontitis. The results suggest a dose-response relationship between OH and
periodontitis, with fair and poor OH significantly increasing the risk of having
periodontitis by 2- and 5-fold, respectively, compared with good OH. In contrast,
regular tooth brushing and dentist visits could reduce periodontitis by 34% and 32%,
respectively. These pooled OH effects and oral care habits are summarized in Table 2.12
and Figure 2.7.
The effect of OH on periodontitis was stronger than those of other risk
factors, such as DM125 (OR = 2.6; 95% CI: 1.0, 6.6), smoking126 (OR = 2.82; 95% CI:
2.36, 3.39) or obesity127 (ORs = 2.13; 95% CI: 1.40, 3.26). Our results also showed
protective effects of regular brushing, which were consistent with the findings of a
previous meta-analysis128, which reported a significant risk for severe periodontitis due
to infrequent brushing (OR = 1.44; 95% CI: 1.21, 1.71). However, our study could only
identify a small effect of interdental cleaning with dental floss, i.e., a non-significant
reduction of 13% in the risk of periodontitis. This result was also consistent with a
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 17
previous meta-analysis129, which found little benefit from self-performed flossing on
plaque or periodontal parameters.
Although the use of OH assessments varied between the included studies,
approximately half of them commonly used the PI with similar cutoff points. OH was
defined as poor for a PI > 2 or if the patient had a moderate accumulation of soft deposits
visible by the naked eye; OH was defined as fair for PI values ranging from 1 to 2 or if
the patient had a film of plaque adhering to the tooth as detected by disclosing solution
or probe.
To address concerns about the varying quality of the individual studies, a
sensitivity analysis was also performed, including only studies with a low risk of bias77,
81-84, 90-92, 101, 120. The results showed little difference compared with those of the main
analysis, but with much less heterogeneity.
Good OH and oral care habits should be encouraged and promoted in public
health campaigns. Dentists and dental hygienists should regularly educate, motivate,
and assess patients’ perceptions for improving oral health behaviors at the “chair side”.
Additionally, dental nurses or assistants should additionally encourage and provide
general, useful information. Repeated and individually tailored OH instructions are key
elements in achieving gingival health. The use of goal setting, self-monitoring and
planning are effective interventions for improving OH-related behaviors in periodontitis
patients. Recognizing the benefits of behavior changes, their own susceptibility, and the
harms of periodontitis are important messages in periodontitis prevention130.
Patients should be able to regularly access dental care for professional
cleaning alongside tailoring and monitoring their OH130. They should also be taught how
to efficiently perform plaque removal. Generally, mechanical plaque controlled by
twice-daily tooth brushing with a fluoride-based dentifrice is an accepted
recommendation. The proper duration of tooth brushing is also mentioned as an
important determinant of plaque removal; therefore, it should be stressed during tooth-
brushing instruction sessions131. The current scientific data showed that dental floss is a
less effective tool and requires the user to be instructed on specific skills in order to be
effective. Interdental brushes have been shown to be the most effective method for
interdental plaque removal132; however, the selection of interdental aids must be at the
Attawood Lertpimonchai Literature review / 18
clinician’s discretion based on a patient’s needs and dexterity and the characteristics of
a patient’s interdental spaces.
This study has some strengths. It includes studies of the effects of OH using
both objective and subjective assessments. The magnitudes of the effects were pooled
and reported. The results of subgroup analyses (i.e., population type, study base,
periodontitis definition and smoking) were also explored. We used rigorous pooling
methods (multivariate random-effects meta-analysis), which considered the variance-
covariance between the studies.
However, this study also has some limitations. Our pooled ORs were based
on summary data of observational studies. Some data were reported without adjusting
for potential confounders; thus, the pooled results might be prone to bias. Moreover, the
definition of periodontitis varied among studies, which resulted in high heterogeneity,
although the subgroup analyses did reduce this effect. Furthermore, the assessments of
publication bias using funnel plots and Egger tests with the low numbers of included
studies in some meta-analyses may not be valid. Failure to detect asymmetry cannot rule
out a reporting bias, or vice versa.
In conclusion, poor OH increases the risk of periodontitis by approximately
2- to 5-fold compared with good OH. Oral care habits, including regular brushing and
dental visits, can decrease the risk of periodontitis and should thus be promoted more in
public health concerns.
2.7.4 Conclusion
From our systematic review, lack of uniformity of periodontitis definitions
were founded among included studies (Table 2.3). Although, most of them relied on
clinical examination, differences in selected periodontal parameters were large. Only
PPD, only CAL, combination of PPD & CAL, PPD & BOP were used unsteadily.
Moreover, the cut-off points for PPD or CAL also were inconsistent between
definitions. Beyond the severity of disease, the extent of periodontitis also was
erratically used. For example, Varghese et al defined periodontitis when subjects had
destructed sites ≥ 30% of total sites, meanwhile, Pranckeviciene et al descripted
periodontitis with only 1 site with clinical attachment loss.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 19
2.8 Effect of periodontitis on renal function/CKD: Biological
plausibility Periodontal infection could be one of the factors accelerating deterioration
in renal function. Possible mechanisms include direct and indirect pathways. Bacteria
and their products could invade into systemic circulation and may cause cellular directly
damage to the nephron or its vasculature. For example, Porphyromonas gingivalis
(P.g.), periodontal pathogens, could invade capillary endothelium or cell matrix by their
polysaccharide capsules and fimbria. In addition, Gingipains, the potential proteases
secreted by P.g., may cause degradation of matrix metalloproteins, collagen, and
fibronectin of the nephrons structure133. Moreover, bacterial antigen, GroEl60 heat shock
protein, could cross-reactivity with human heat shock protein 60 which results in
endothelium damaging cascade134, 135. For indirect pathway, increasing inflammatory
mediators including IL-6, TNF-α, prostaglandin E2 (PGE2) and Thromboxane B2
accelerates atherogenesis, thrombus formation and platelet aggregation leading to
ischemia, glomerulosclerosis and severe renal insufficiency.
2.9 Association between periodontitis and CKD: Epidemiological
studies
2.9.1 Methods
A systematic review was performed by identifying studies assessing the
association between periodontitis and CKD. Studies were identified from the
MEDLINE and SCOPUS database (up to April 2017). The search terms and strategies
are showed in Table 2.13. Any type of observational study (e.g. cohort, case–control or
cross-sectional) published in English was included if it met the following criteria: (i)
assessed associations between CKD and periodontitis in either general or specific types
of adult populations; (ii) primary outcomes (kidney function or damage) were assessed
by standard measurement which recommended by KDIGO guideline, such as serum
creatinine, cystatin-C, or proteinuria; (iii) outcome was categorized at least 2 groups
with one group of normal kidney function or control, while, other(s) could be any stage
Attawood Lertpimonchai Literature review / 20
of CKD; for example, ESRD versus non-ESRD, moderate and severe CKD versus
normal kidney function. (iv) assessed periodontitis using parameters from clinical
examination, such as PPD or CAL. (v) reported the mean and SD of periodontal
parameters among CKD groups or a contingency table between non-CKD/CKD and
periodontal status. Experimental studies which assessed the effects of periodontitis
treatment on kidney function were excluded. Studies which enrolled too specific groups
of participants, i.e., renal transplant, dental implant or gingival hyperplasia/overgrowth
also were excluded.
2.9.2 Results
A total of 2,544 studies were identified from MEDLINE and SCOPUS, and
2,298 studies remained after removing duplicates. Of these, 2,257 studies were
ineligible for reasons described in Figure 2.8, leaving 41 studies12-31, 136-156 that were
eligible for review.
From literatures searching, the association between periodontitis and renal
disease has been extensively explored. Among included studies, 21 studies136-156 were
case-control with the hospital-based setting. All of them aimed to compare periodontal
status between subjects with and without renal disease. Details of each study were
provided in Table 2.14. Mostly, they focused on periodontal status in ESRD patients,
and found that, periodontal status in ESRD was worse than normal kidney function.
Results were similar with other studies which determined in CKD subjects. For
example, Borawski et al137 showed significant differences in periodontal parameters
such as PI, gingival index (GI), mean PPD, mean CAL between ESRD, Pre-dialysis
CKD and general population matched with age and gender. However, some studies
could not show these significant differences. Bot et al138, and Castillo et al140 found the
non-statistical differences of PPD or CAL among groups. Similarly, Teratani et al152
reported the non-statistical significance of extent of periodontitis. The heterogeneity
may cause by the differences in population, severity & treatment modalities of renal
disease, and methodology in assess periodontal status among these studies.
Within 14 eligible studies12-25 with cross-sectional design, 2 studies19, 22
were further excluded because study subjects were duplicated from others, as well as,
methodologies also were similarly with previous studies. One study17 was excluded
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 21
because it explored the reverse relationship, i.e., set periodontitis as the outcome.
Leaving 11 studies12-16, 18, 20, 21, 23-25 (Table 2.15), they suggested the certain association
between periodontitis and CKD. From the Atherosclerosis Risk in Communities (ARIC)
study, Kshisagar et al21 showed that the odds of having CKD is significantly greater
around 2 times in individuals with periodontitis than in those without, after adjustment
for traditional CKD risk factors. Despite of differences in periodontitis definition, group
of studies12-16, 20, 23 retrieved data from the NHANES reported the consistency
association between periodontitis and CKD. Fisher et al12-14 found that adults who had
periodontitis or edentulous were approximately twice as likely to have CKD as adults
without periodontitis in NHANES III (1988-1994) population. Similarly, Ioannidou and
Swede20 also found this association but the degree of association varied among ethnic
groups, significantly associated in non-Hispanic Blacks and Mexican-Americans, but
not so for non-Hispanic Whites. Grubbs et al16 also confirmed these with NHANES in
2001-2004. They proposed the significant adjusted odds ratios of having CKD in
periodontitis as 1.55 (95%CI: 1.16, 2.07) using normal/mild periodontitis as the referent
group.
With the advance statistical analysis, Fisher et al15 explored the mediation
analysis and structural equation modeling by cross-sectional data. The results showed
that periodontitis had a significant direct effect on CKD, and had indirectly effect
through DM duration and HT. In other words, periodontitis had a significant direct effect
on DM duration; DM duration had a significant direct effect on HT, and then, HT had a
significant direct effect on CKD. While the direct effect of DM on CKD was not
significant.
As far as we were concerned, there were six cohort studies26-31 (Table 2.16)
assessing effects of periodontitis on renal disease. Shultis et al31 showed the significant
effect of periodontitis on overt nephropathy and ESRD during 22 years in subjects with
DM type 2 in the Gila River Indian Community after adjusted confounders with hazard
ratios around 2.0 to 4.9. Iwasaki et al30 proposed the 2-year-retrospective cohort study
in elderly communities. Results showed that the most severe groups of periodontitis had
the risk of decrease kidney function than others 2.24 (95% CI: 1.05, 4.79) times.
Recently, Grubbs et al identified the consistent effect of periodontitis on CKD incidents
in 2 cohorts, i.e. the Jackson heart study28 with effect size of 2.48 (95% CI: 1.04, 5.88),
Attawood Lertpimonchai Literature review / 22
and Osteoporotic Fractures in Men (MrOS) cohorts29 with adjusted ORs of 2.04 (95%
CI: 1.21, 3.44).
2.9.3 Conclusion
Relationship between periodontitis and CKD has been extensive explored.
Implied from case-control studies, periodontal status seems to be worse in CKD and
ESRD than subjects without kidney disease. Their significant association was also
claimed from cross-sectional studies with various population and types of statistical
analysis. Moreover, the longitudinal studies suggest the pattern of causation from
periodontitis on kidney function. However, there were only half of cohort studies which
interested in the incidence of CKD by exclude CKD cases at baseline. Among these,
their included subjects were quite specific groups, i.e., DM patients in Pima Indian
community31, African-American28 and only male aged 65 years or older29. Studies
representing general population have still been required. Furthermore, the complexity
between periodontitis, DM and CKD is valuable to explore. The structure models such
as, mediation analysis, which could explain the causative pathway or the sequence of
association, also should be identified.
2.10 Association between periodontitis and DM A bi-directional relationship between periodontitis and DM was
comprehensively investigated157-159. Periodontal health in subjects with DM was worse
than normal, particularly in poor glycemic control160. Simultaneously, periodontitis
could increase risk of DM initiation, as well as, impaired glycemic control in DM
patients9. Biological pathways in bi-directional relationship are not completely
understood, but it hypothesizes that involved of inflammation and cytokine, function of
immunity, collagen homeostasis, and insulin resistance161 (Figure 2.9). Although,
periodontal pathogens in periodontitis subjects with DM were little different comparing
with subjects without DM, local inflammatory cytokines were significantly increased.
Altered levels of IL-1ß, IL-6, PGE2 and RANKL/OPG ratios in saliva and gingival
tissue were broadly reported162-164. Hyper-inflammatory phenotype of monocytes165,
defective neutrophils166, and disturbance T-cells resulting in elevation of bone
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 23
resorption cytokines, i.e., RANKL, IL-17, were described as alternative pathways of
periodontal destruction in DM patients167. In addition, oxidative stress due to
hyperglycemic condition produced many reactive oxygen species, and these substances
influenced bone metabolism166. AGEs impaired collagen stability, and decreased the
potential of tissue repair following destruction168.
Epidemiological studies reported the consistent significant association
between periodontitis and DM160, 169. Originally, it was studied in the Gila River
Community, a population of Pima Indian with a high prevalence of DM. Taylor et al
established the risk association of severe periodontitis on poor glycemic control in DM
type II9. After that, Demmer et al170 found that subjects with the most severe
periodontitis at baseline had a five times greater increase in their HbA1c values over
five years compared to those who did not have periodontitis (change in HbA1c: 0.106
± 0.03% versus 0.023 ± 0.02%). Effect of periodontal treatment on glycemic control
also intense investigated. Recent systematic reviews and meta-analysis showed that
periodontal treatment was associated with reductions in HbA1c, however, this positive
effect was explored only in short term171.
Conversely, prevalence and severity of periodontitis increased in DM
patients compared to general patients. Nelson et al125 showed the increase risk of further
interproximal bone loss in DM after adjusting effects of age and sex in Pima Indian
subjects. Jimenez et al172 also reported the significant effect of DM on self-reported
periodontitis incidence and tooth loss by cohort of men in 20 years of follow-up.
However, the risk of periodontitis was based on the level of glycemic control, directly.
In well controlled, periodontium and response to periodontal treatment appeared to be
little different comparing with normal subjects. While, the differences were obvious in
poor glycemic controlled160.
2.11 DM as the risk factor of CKD
DM usually causes the various macro- and micro-vascular complications. They can lead
to CVD, nephropathy, and retinopathy. In aspect of nephrology, DM has been
recognized as the important risk factor of CKD development and its progression.
Diabetic kidney disease (DKD), i.e., kidney damage from DM, is a common
Attawood Lertpimonchai Literature review / 24
complication of both types of DM. Around 20-30% of DM patients suffered from kidney
disease173. Recent systematic review of observation studies reported the risk of
developing kidney disease among those with DM174. In type I DM, the annual incidence
of albuminuria ranged from 1.3% to 3.8%, whereas it ranged from 3.8% to 12.7% for
type 2 DM. Moreover, within studies reporting the incidence of eGFR < 60 ml/min per
1.73 m2, the annual incidence ranged from 1.9% to 4.3%, and incidence of ESRD ranged
from 0.04% to 1.8%. From this, DM has been always included as the common predictor
in the final prediction models. Echouffo-Tcheugui and Kengne40 reviewed the
prediction models of the occurrence/presence of CKD, and found that DM was
presented as the predictor in 23 from 30 models. In addition, age, sex, body mass index,
BP, serum creatinine, proteinuria, and serum albumin or total protein were commonly
included in the final prediction models. The effect size of DM on CKD incidence was
estimated from the meta-analysis of the cohort and case-control studies. Among 10
included studies175 with data from more than 5 million subjects, the pooled adjusted risk
ratio of CKD associated with DM was 3.34 (95% CI: 2.27, 4.93) in women, and 2.84
(95% CI: 1.73, 4.68) in men. Interestingly, the risk of CKD also significantly increased
in prediabetes176, with the relative risk of 1.12 (95% CI: 1.02, 1.21).
Metabolic pathways, hemodynamic pathways, and inflammation were
mentioned as the pathogenic mechanisms for renal injury (Figure 2.10). Hyperglycemia
state initiated the multiple cellular processes in the metabolic pathway including
alteration in cellular energy production, stimulation aldose reductase and PKC,
generation of AGEs and reactive oxygen species, increased flux of polyols and
hexosamines, atypical expression of cyclin kinases and their inhibitors, and disturb
stabilization of factors controlling the extracellular matrix homeostasis. These processes
altered blood flow and capillary permeability, increased production of extracellular
matrix proteins, and stimulated oxidant and osmotic stress. And thus, they contribute
glomerular sclerosis and tubulointerstitial fibrosis177. Hemodynamic dysfunction
potential occurred from intraglomerular hypertension and hyperfiltration, which directly
traumatized the vascular components in nephrons178. The most important mediator
participating in the hemodynamic pathway was the renin–angiotensin–aldosterone
system (RAAS). Hyperglycemia induced the activation of the local renal RAAS and
increased the production of angiotensin II, which exerted a preferential constrictor effect
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 25
on the efferent glomerular arteriole, resulting in a higher intraglomerular capillary
pressure10. The relationships between this inflammatory state and the initiation and
evolution of DKD involved very complex pathways. Diverse inflammatory molecules
play significant roles, including toll-like receptors, chemokines, adhesion molecules,
and inflammatory cytokines (e.g., IL-1ß, IL-6, and TNF-α)179.
Attawood Lertpimonchai Literature review / 26
Table 2.1 2009 CKD-EPI creatinine equations
Gender Serum creatinine Equation for estimating GFR
Female ≤ 0.7 mg/dl (≤ 62 µmol/l) 144 x (SCr/0.7)-0.329 x 0.993Age [ x 1.159 if black]
Female > 0.7 mg/dl (> 62 µmol/l) 144 x (SCr/0.7)-1.209 x 0.993Age [ x 1.159 if black]
Male ≤ 0.9 mg/dl (≤ 80 µmol/l) 141 x (SCr/0.9)-0.411 x 0.993Age [ x 1.159 if black]
Male > 0.9 mg/dl (> 80 µmol/l) 141 x (SCr/0.9)-1.209 x 0.993Age [x 1.159 if black]
GFR, glomerular filtration rate; SCr, serum creatinine
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 27
Table 2.2 Search terms and search strategy: Periodontitis and oral hygiene
Item Domains Terms
1
Periodontitis
Periodontitis
2 Periodontal
3 Periodontitis [MeSH]*
4 1 OR 2 OR 3
5
Oral hygiene
Poor oral hygiene
6 Plaque index
7 Dental plaque index [MeSH]*
8 Oral hygiene index
9 Oral hygiene index [MeSH]*
10 Plaque score
11 5 OR 6 OR 7 OR 8 OR 9 OR 10
12
General
Risk factor
13 Association
14 Relation
15 Correlation
16 12 OR 13 OR 14 OR 15
17 4 AND 11 AND 16 *Options for MEDLINE
Attaw
ood Lertpimonchai
Literature review
/ 28 Table 2.3 Characteristics of included studies
Authors Study type
Study base Population OH
measurement Age
Male (%)
Smoking (%)
DM (%)
Periodontitis definition
Imaki70 Cross-
sectional Community General PI 38.1 100 56.1 N/A CPITN: 3-4
Norderyd71 Cross-
sectional Community General PSc 48 48.7 20 4
Radiography: bone loss > 1/3 of root length
Wakai72 Cross-
sectional Hospital General PI 51.1 82.1 34.4 N/A CPITN: 3-4
Papapanou73 Case-
control Hospital General PSc 50.9 47.3 32.2 N/A
≥ 1 site with PPD ≥ 5 mm AND CAL ≥ 3 mm
Hashim74 Cohort Community General OHI,
Dental visit 15 54.2 33.3 N/A ≥ 1 site with ≥ 4 mm increase in CAL
Tezal75 Cross-
sectional Community General PI 48.7 48.2 61.8 N/A Mean CAL ≥ 2 mm
Hugoson76 Cross-
sectional Community General PSc 65.0 52.7 42.9 N/A
Radiography: bone loss more than 1/3 of root length
Do77 Cross-
sectional Community General PI 40 40.3 28.9 N/A
≥ 2 sites with CAL ≥ 5 mm AND ≥ 1 site with PPD ≥ 4 mm
Meisel78 Cross-
sectional Community General PSc 51.0 46.6 49.5 6
4th - 5th quintiles of % sites with CAL > 4 mm
Alpagot79 Cohort Hospital HIV
patients PI 34.1 57.9 N/A 0
≥ 1 site with PPD ≥ 4 mm OR CAL ≥ 2 mm
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 29 Table 2.3 Characteristics of included studies (cont.)
Authors Study type
Study base Population OH
measurement Age Male (%)
Smoking (%)
DM (%) Periodontitis definition
Solis80 Cross-sectional Hospital General PI 37.4 35.3 23.5 0
≥ 2 sites with CAL ≥ 6 mm AND ≥ 1 site with PPD ≥ 5 mm
Wickholm81 Cross-sectional Community General PI 36.7 49.2 44.7 N/A ≥ 3 teeth with PPD ≥ 5 mm
Natto82 Cross-sectional Community General PI 36.4 64.9 70 N/A ≥ 10 sites with PPD ≥ 5 mm
Torrungruang83 Cross-sectional Community General PSc 60 74.4 14.3 15.8 Mean CAL > 2.5 mm
de Macedo84 Cross-sectional Community General PSc, Flossing,
Brushing N/A 33.8 31.4 N/A ≥ 4 teeth with PPD ≥ 4 mm AND CAL ≥ 3 mm at the same site
Khader85 Cross-sectional Hospital General PI 39.4 44.8 N/A N/A Khader’s risk score
Vandana86 Cross-sectional Hospital Dental
fluorosis OHI 25.36 68.6 N/A 0 CPITN: 3-4
Wang87 Cross-sectional Community General Brushing N/A 45.7 27.7 0 Mean CAL ≥ 3 mm
Akhter88 Case-control Hospital General PI, Brushing,
Dental visit 38.5 50 45.7 N/A ≥ 2 sites with CAL ≥ 6 AND ≥ 1 site with PPD ≥ 5 mm
Kumar89 Cross-sectional Community General Brushing,
Dental visit 33.9 100 N/A N/A CPITN: 3-4
Attaw
ood Lertpimonchai
Literature review
/ 30 Table 2.3 Characteristics of included studies (cont.)
Authors Study type
Study base Population OH
measurement Age Male (%)
Smoking (%)
DM (%) Periodontitis definition
Benguigui90 Cross-sectional Community General PI 58 54.9 19.2 6.7 CDC/AAP
Bawadi92 Cross-sectional Hospital General PI, Brushing 36.4 49.4 20.3 17.9 ≥ 4 teeth with PPD ≥ 4 mm AND
CAL ≥ 3 mm at the same site
Saxlin91 Cohort Community General PI, Brushing, Dental visit 41.86 27 0 0 New teeth with PPD ≥ 4 mm
Carrilho Neto93 Cross-sectional Hospital Inpatients OHI 45.7 59.7 42.7 N/A ≥ 1 site with PPD > 4 mm
Mathur94 Cross-sectional Hospital General OHI N/A 57.3 N/A N/A N/A
Teng95 Cross-sectional Hospital Psychiatric
inpatients Brushing,
Dental visit 41 62.5 42.5 N/A CPITN: 3-4
Crocombe96 Cross-sectional Community General Brushing,
Flossing N/A 50 15 4.3 ≥ 1 site with CAL ≥ 4 mm
Mannem97 Cross-sectional Hospital General PI 52.5 44.1 34.2 N/A ≥ 4 teeth with PPD ≥ 4 mm AND
CAL ≥ 3 mm at the same site
Raja98 Cross-sectional Hospital General PI 36.5 53.3 96.7 0 ≥ 4 sites with CAL ≥ 4 mm
Vogt99 Cross-sectional Hospital Pregnancy PSc, Flossing,
Brushing 27.2 0 15.87 0 ≥ 4 teeth with PPD ≥ 4 mm AND CAL ≥ 4 mm at the same site
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 31 Table 2.3 Characteristics of included studies (cont.)
Authors Study type
Study base Population OH
measurement Age Male (%)
Smoking (%)
DM (%) Periodontitis definition
Develioglu105 Case-control Hospital General PI 46.7 N/A 0 33.3 ≥ 30% sites with PPD ≥ 5 mm AND
CAL ≥ 3 mm
Fiyaz100 Case-control Hospital General OHI N/A N/A 0 0 ≥ 1 site with PPD > 4 mm
OR CAL > 1.5 mm
Palle101 Cross-sectional Hospital CVD OHI 57.2 84.1 32.3 52.2 ≥ 5 sites with CAL ≥ 5 mm
Cakmak104 Case-control Hospital General PI 38.3 49.1 0 0 ≥ 1 site with PPD ≥ 5 mm AND
CAL ≥ 4 mm
Jacob108 Case-control Hospital General PI 37.3 75.6 33.3 0 CDC/AAP
Kaur109 Case-control Hospital General PI N/A 66.7 25 0 N/A
Koseoglu110 Case-control Hospital General PI 34.0 50 0 0 ≥ 4 teeth with PPD ≥ 5 mm AND
CAL ≥ 4 mm in each jaw
Kovačević111 Cross-sectional Hospital General
OHI, Brushing, Flossing, Dental
visit 38.9 77.2 31.7 N/A CPITN: 3-4
Lavu113 Case-control Hospital General OHI 33.6 50.4 0 0 CAL > 1 mm at least 30% sites
Lutfioglu114 Case-control Hospital General PI 33.1 53.3 51.1 0 ≥ 1 site with PPD ≥ 5 mm with
radiographic evidence of bone loss
Attaw
ood Lertpimonchai
Literature review
/ 32 Table 2.3 Characteristics of included studies (cont.)
Authors Study type
Study base Population OH
measurement Age Male (%)
Smoking (%)
DM (%) Periodontitis definition
Petrović119 Case-control Hospital General PI 36.1 38.8 22.4 0 ≥ 3 quadrants with ≥ 3 sites with
PPD ≥ 3 mm AND CAL ≥ 2 mm
Pranckeviciene120 Cross-sectional Hospital Type I &
II DM PI 43.86 N/A 25.9 100 ≥ 1 site with CAL > 5 mm
Meenawat115 Case-control Hospital General PI 43.2 100 41.4 0 ≥ 4 teeth with PPD > 4 mm AND
CAL > 2 mm
Mesa116 Case-control Hospital General PSc 46.3 40.3 46.8 N/A ≥ 4 teeth with PPD ≥ 4 mm AND
CAL ≥ 3 mm at the same sites
Perayil117 Case-control Hospital General OHI 43.1 43.3 0 0 ≥ 5 teeth with PPD ≥ 5 mm AND
CAL ≥ 3 mm
Pereira118 Case-control Hospital General PSc 38.4 33.7 0 0 ≥ 4 teeth with PPD ≥ 4 mm AND
CAL ≥ 3 mm
Puri121 Case-control Hospital General OHI 39.78 N/A 0 0 AAP 1999
Singh122 Case-control Hospital General PI 43.5 52.5 0 0 ≥ 1 site with PPD ≥ 5 mm AND
CAL ≥ 2 mm
Toyman123 Case-control Hospital General PI 34.6 51.2 0 0 ≥ 6 teeth with PPD ≥ 5 mm with
radiographic evidence of bone loss
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 33 Table 2.3 Characteristics of included studies (cont.)
CAL, clinical attachment loss; CDC/AAP, periodontitis definition of the Centers for Disease Control and Prevention in collaboration with the American Academy of Periodontology, CPITN, the Community Periodontal Index of Treatment Needs; CVD, cardiovascular disease; DM, diabetes mellitus; N/A, not available; OH, oral hygiene; OHI, oral hygiene index; PI, plaque index; PPD, periodontal pocket depth; PSc, plaque score
Authors Study type
Study base Population OH
measurement Age Male (%)
Smoking (%)
DM (%) Periodontitis definition
Varghese124 Case-control Hospital General PI N/A 65.3 0 0 ≥ 30% sites with PPD ≥ 6 mm AND
CAL ≥ 5 mm
Attawood Lertpimonchai Literature review / 34
Table 2.4 Risk of bias assessment
Authors Selection Comparability Outcome/Exposure* Risk of
bias S1 S2 S3 S4 S5 C1 C2 O1 O2 O3 O4 O5 O6 Cohort
Hashim74 1 1 1 0 - 1 1 1 0 1 0 1 1 Low
Saxlin91 1 1 1 1 - 0 1 1 1 0 1 1 0.5 Low
Case-control
Akhter88 1 1 1 0 0.5 1 1 1 1 0 - - - Low
Fiyaz100 1 1 0 0 0 0 0 1 1 0 - - - High
Papapanou73 1 1 1 0 0.5 1 1 1 1 0 - - - Low
Cakmak104 1 1 0 0 0.5 1 1 1 1 0 - - - Mod
Develioglu105 1 0 0 0 0 1 0 1 1 0 - - - High
Jacob108 1 1 1 0 0.5 1 0 1 1 0 - - - Mod
Kaur109 0 0 0 0 0 1 0 1 1 0 - - - High
Koseoglu110 1 1 0 0 0.5 1 0 1 1 0 - - - Mod
Lavu113 1 1 1 0 0.5 1 1 1 1 0 - - - Low
Lutfioglu114 0 1 0 0 0.5 1 0 1 1 0 - - - High
Meenawat115 1 1 0 0 0 0 0 1 1 0 - - - High
Mesa116 1 1 1 1 1 1 1 1 1 0 - - - Low
Perayil117 1 1 1 0 1 1 0 1 1 0 - - - Mod
Pereira118 1 1 0 0 1 1 0 1 1 0 - - - Mod
Petrović119 1 0 1 0 0 1 0 1 1 0 - - - Mod
Puri121 1 1 0 0 0 1 0 1 1 0 - - - Mod
Singh122 1 1 0 0 0 0 1 1 1 0 - - - Mod
Toyman123 0 1 0 0 0.5 1 0 1 1 0 - - - High
Varghese124 1 1 0 0 0 1 0 1 1 0 - - - Mod
Cross-sectional
Alpagot79 0 1 - - - 0 0 1 0 1 1 - - Mod
Bawadi92 1 1 - - - 1 1 1 1 1 1 - - Low
Benguigui90 1 1 - - - 0 0 1 1 1 1 - - Low
Carrilho Neto93 0 1 - - - 1 1 1 1 0 0 - - Mod
Crocombe96 1 1 - - - 1 1 1 1 1 0 - - Low
Do77 1 1 - - - 1 1 1 0 1 1 - - Low
Hugoson76 1 1 - - - 1 1 0 0 0 1 - - Mod
Imaki70 0 1 - - - 1 1 1 0 0 0 - - Mod
Khader85 0 1 - - - 1 1 1 1 1 1 - - Low
Kovačević111 1 1 - - - 0 0 1 1 0 0 Mod
Kumar89 1 1 - - - 0 1 1 0 0 1 - - Mod
de Macedo84 0 1 - - - 1 1 1 1 1 1 - - Low
Mannem97 0 1 - - - 1 1 1 1 1 1 - - Low
Mathur94 0 1 - - - 0 0 1 1 0 0 - - High
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 35
Table 2.4 Risk of bias assessment (cont.)
Authors Selection Comparability Outcome/Exposure* Risk
of bias S1 S2 S3 S4 S5 C1 C2 O1 O2 O3 O4 O5 O6
Meisel78 1 1 - - - 1 1 1 0 1 0 - - Low
Natto82 0 1 - - - 1 1 1 1 0 1 - - Low
Norderyd71 1 1 - - - 1 1 0 0 0 0 - - Mod
Palle101 0 1 - - - 1 1 1 1 1 0 - - Low
Pranckeviciene120 0 1 - - - 0 1 1 1 1 1 - - Low
Raja98 0 1 - - - 1 1 1 1 1 1 - - Low
Solis80 0 1 - - - 1 1 1 1 1 1 - - Low
Teng95 0 1 - - - 1 1 1 0 0 0 - - Mod
Tezal75 1 1 - - - 1 1 1 1 1 1 - - Low
Torrungruang83 0 1 - - - 1 1 1 0 1 1 - - Low
Vandana86 0 1 - - - 0 0 1 0 0 0 - - High
Vogt99 0 1 - - - 0 0 1 1 1 0 - - Mod
Wakai72 0 1 - - - 1 1 1 0 0 0 - - Mod
Wang87 1 1 - - - 1 1 1 0 1 1 - - Low
Wickholm81 1 1 - - - 1 1 1 1 0 1 - - Low
*Outcome: Cohort and Cross-sectional study || Exposure: Case-control study
Attawood Lertpimonchai Literature review / 36
Table 2.5 Categorization of OH level
Authors Index Good OH Fair OH Poor OH
Mathur94 OHI < 1 1-1.99 ≥ 2
Vandana86 OHI < 1 1-1.99 ≥ 2
Carilho Neto93 OHI < 3 - ≥ 3
Palle101 OHI < 2 - ≥ 2
Bawadi92 PI < 1 1-1.99 ≥ 2
Do77 PI < 1 1-1.99 ≥ 2
Saxlin91 PI < 1 1-1.99 ≥ 2
Wakai72 PI < 1 1-1.99 ≥ 2
Wickholm81 PI < 1 1-1.99 ≥ 2
Benguigui90 PI Mean - ≥ Mean
Imaki70 PI ≤ 1 - > 1
Pranckeviciene120 PI < 1 - ≥ 1
Natto82 PI 0-0.69 0.7 -1.30 ≥ 1.31
de Macedo84 PSc 0-64 - ≥ 65
Torrungruang83 PSc 0-39 40-79 ≥ 80
OH, oral hygiene; OHI, oral hygiene index; PI, plaque index; PSc, plaque score
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 37 Table 2.6 Pooling effects of fair and poor versus good OH on periodontitis
Authors Index Periodontitis Non-Periodontitis OR1
* 95% CI I2 OR2** 95% CI I2
Good OH
Fair OH
Poor OH Good
OH Fair OH
Poor OH
Vandana86 OHI 21 156 100 124 605 23 1.52 0.93, 2.50 25.67 13.44, 49.06
Carilho Neto93 OHI 2.10 0.37, 11.76
Mathur94 OHI 2 60 178 18 42 0 12.86 2.83, 58.38 2641.80# 122.16, 57129.99
Palle101 OHI 6.85 2.87, 16.35
Pooled 6.16 0.85, 44.90 81% 7.64 2.14, 27.36 71%
Imaki70 PI 186 553 542 330 4.88 3.94, 6.06
Wakai72 PI 144 214 68 88 87 24 1.50 1.04, 2.16 1.73 1.01, 2.96
Do77 PI 4 36 103 24 164 244 1.32 0.43, 4.03 2.53 0.86, 7.48
Wickholm81 PI 2.72 1.89, 3.92 4.81 1.99, 10.80
Natto82 PI 1.90 1.09, 3.30 3.60 1.20, 11.00
Benguigui90 PI 87 114 42 12 4.59 2.28, 9.23
Saxlin91 PI 55 44 4 33 14 0 1.89 0.90, 3.95 5.43# 0.28, 104.11
Bawadi92 PI 15 69 21 83 119 19 3.21 1.72, 5.99 6.12 2.67, 14.01
Pranckeviciene120 PI 98 122 40 6 8.30 3.38, 20.38
Pooled 2.26 1.75, 2.92 36% 4.15 3.00, 5.72 50%
Torrungruang83 PSc 218 778 398 198 342 71 2.07 1.64, 2.60 5.09 3.71, 6.99
de Macedo84 PSc 9 33 64 66 3.56 1.58, 8.02
OVERALL POOLED 2.04 1.65, 2.53 40% 5.01 3.40, 7.39 78% * Fair versus Good OH, ** Poor versus Good OH, # Calculated with continuity correction CI, confidence interval; OH, oral hygiene; OHI, oral hygiene index; OR, odds ratio; PI, plaque index; PSc, plaque score
Attawood Lertpimonchai Literature review / 38
Table 2.7 Subgroup and sensitivity analysis according to sources of heterogeneity of fair and poor versus good OH
Factors
Fair Vs Good OH Poor vs Good OH
Number
of studies
OR1 (95% CI) I2 Number
of studies
OR2 (95% CI) I2
Overall 9 2.04 (1.65, 2.53) 40% 15 5.01 (3.40, 7.39) 78%
Subgroup analysis
Type of index
OHI
PI
PSc*
2
6
1
6.16 (0.85, 44.90)
2.26 (1.75, 2.92)
-
81%
36%
-
4
9
2
7.64 (2.14, 27.36)
4.15 (3.00, 5.72)
-
71%
50%
-
Study-based
Community
Hospital
6
3
2.23 (1.85, 2.69)
2.43 (0.72, 8.17)
4%
90%
9
6
4.78 (4.10, 5.58)
6.06 (2.08, 17.68)
0%
84%
Periodontitis definition**
Using PPD only
Using CAL or x-ray only*
Using PPD with CAL
5
1
2
1.86 (1.40, 2.46)
-
2.45 (1.58, 3.79)
43%
-
0%
7
3
4
4.86 (2.12, 11.18)
-
4.42 (2.98, 6.55)
89%
-
0%
Smoking**
< 25%
≥ 25%
3
4
2.15 (1.59, 2.91)
2.19 (1.64, 2.94)
1%
40%
4
9
5.05 (3.84, 6.62)
4.05 (2.89, 5.67)
0%
53%
Sensitivity analysis
Only studies focused on
general population
8
2.10 (1.76, 2.49)
22%
11
4.21 (3.21, 5.51)
49%
* Insufficient number of studies for pooling OR via mvmeta method ** Missing data: periodontitis definition (1 study), smoking (2 studies) CAL, clinical attachment level; CI, confidence interval; OH, oral hygiene; OHI, oral hygiene index; OR, odds ratio; PI, plaque index; PPD, periodontal pocket depth; PSc, plaque score
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 39
Table 2.8 Pooled mean difference of OH score between periodontitis and non-
periodontitis
Authors Periodontitis Non-Periodontitis
MD 95% CI N Mean SD N Mean SD
Oral hygiene index
Fiyaz100 30 2.38 0.81 30 0.53 0.41 2.88 2.15, 3.61
Perayil117 30 3.89 0.68 30 0.53 0.17 6.78 5.45, 8.11
Kovacevic111 73 1.18 0.64 28 1.18 0.64 0.00 -0.44, 0.44
Puri121 20 3.70 0.86 20 2.75 0.37 1.44 0.74, 2.13
Lavu113 177 2.77 0.83 176 0.40 0.25 3.87 3.51, 4.22
Pooled I2 = 98.9 1.71 0.65, 2.78
Plaque index
Alpagot79 111 1.79 0.74 41 0.88 0.47 1.34 0.95, 1.73
Solis80 44 1.32 0.79 105 0.87 0.57 0.70 0.34, 1.06
Akhter88 140 1.80 0.70 140 1.20 0.50 0.99 0.74, 1.23
Mannem97 77 1.19 0.44 34 0.65 0.29 1.35 0.91, 1.79
Raja98 30 1.60 0.33 30 1.31 0.26 0.98 0.44, 1.51
Cakmak104 80 1.78 0.59 40 1.05 0.57 1.25 0.84, 1.66
Develioglu105 32 2.23 0.52 16 0.19 0.10 4.74 3.60, 5.88
Jacob108 30 1.35 0.39 15 0.48 0.24 2.50 1.68, 3.31
Kaur109 20 1.69 0.35 20 0.75 0.36 2.65 1.79, 3.51
Petrovic119 36 1.20 0.78 31 0.74 0.77 0.59 0.10, 1.08
Singh122 20 2.48 0.27 20 0.57 0.10 9.38 7.18, 11.58
Koseoglu110 20 1.52 0.30 40 0.93 0.47 1.41 0.82, 2.01
Lutfioglu114 32 2.29 0.97 60 1.26 1.27 0.88 0.43, 1.32
Meenawat115 24 2.49 0.23 5 0.33 0.08 10.07 7.22, 12.92
Toyman123 21 0.89 0.90 20 0.02 0.06 1.35 0.67, 2.03
Varghese124 25 1.29 0.41 25 0.29 0.27 2.88 2.08, 3.68
Pooled I2 = 98.3 0.97 0.61, 1.32
Plaque score
Papapanou73 131 51.20 24.50 74 35.00 25.60 0.65 0.36, 0.94
Vogt99 157 73.80 14.20 177 59.60 17.90 0.87 0.65, 1.10
Pereira118 31 79.00 17.90 58 49.20 26.80 1.24 0.76, 1.71
Mesa116 41 74.47 34.79 36 46.32 35.36 0.80 0.34, 1.27
Pooled I2 = 74.2 20.44 12.85, 28.04
Overall Pooled* I2 = 95.6 2.04 1.59, 2.50 * Standardized mean difference pooling
CI, confidence interval; MD, mean difference; OH, oral hygiene; OHI, oral hygiene index; PI, plaque index; PSc, plaque score
Attawood Lertpimonchai Literature review / 40
Table 2.9 Pooled effect size of oral care habit on periodontitis
Authors Periodontitis Non-Periodontitis
OR 95% CI
I2 Good Habits
Bad habits Good
Habits Bad
habits
Brushing
de Macedo84 34 8 118 12 0.43 0.16, 1.14
Wang87 0.86 0.67, 1.11
Akhter88 14 126 38 102 0.30 0.15, 0.58
Kumar89 0.55 0.53, 0.57
Saxlin91 67 33 36 10 0.56 0.25, 1.27
Bawadi92 69 36 154 81 1.01 0.62, 1.64
Teng95 0.43 0.24, 0.75
Crocombe96 1356 788 1179 847 1.24 1.09, 1.40
Vogt99 56 101 68 109 0.89 0.57, 1.39
Kovacevic111 46 27 21 7 0.57 0.21, 1.51
Pooled 0.6 0.47, 0.94 94.5%
Floss
de Macedo84 8 34 28 102 0.86 0.36, 2.06
Crocombe96 0.90 0.77, 1.05
Vogt99 49 108 76 101 0.60 0.38, 0.95
Kovacevic111 21 52 7 21 1.21 0.45, 3.27
Pooled 0.87 0.75, 1.00 5.1%
Dental visit
Hashim74 0.66 0.35, 1.27
Akhter88 6 134 20 120 0.27 0.10, 0.69
Kumar89 0.99 0.98, 1.00
Saxlin91 68 33 35 11 0.65 0.29, 1.43
Teng95 0.65 0.35, 1.22
Kovacevic111 38 34 18 10 0.62 0.25, 1.53
Pooled 0.68 0.47, 0.98 60.4%
CI, confidence interval; OR, odds ratio
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 41
Table 2.10 Sources of heterogeneity of tooth brushing meta-analysis
* Missing data of smoking: 1 study
CAL, clinical attachment level; CI, confidence interval; OR, odds ratio; PPD, periodontal pocket depth
Factors Number of studies
OR (95% CI) I2
Overall 10 0.66 (0.47, 0.94) 94.5%
Smoking*
< 25%
≥ 25%
4
5
1.03 (0.78, 1.35)
0.51 (0.31, 0.83)
47%
69%
Definition of regular brushing
Once a day
Twice a day
3
7
0.57 (0.35, 0.94)
0.75 (0.54, 1.05)
78%
77%
Periodontitis definition
Using PPD only
Using CAL or x-ray only
Using PPD combined with CAL
4
3
3
0.55 (0.53, 0.57)
0.94 (0.64, 1.39)
0.67 (0.35, 1.29)
0%
80%
79%
Attawood Lertpimonchai Literature review / 42
Table 2.11 Publication bias assessment by Egger test
Pooling Number of studies Coefficient SE P-value
OH: Continuous data
Mean difference 25 6.00 1.75 0.002
OR: Plaque index 6 4.08 1.52 0.055
OR: Plaque score 3 -201.87 0.57 0.002
OH: Categorical data
OR: Poor Vs Good OH 15 0.17 0.78 0.827
OR: Fair Vs Good OH 9 0.68 1.01 0.523
Oral health care habit
Brushing 10 1.30 1.71 0.468
Floss 4 -0.32 0.96 0.771
Dental visit 6 -1.48 0.32 0.010
CI, confidence interval; OH, oral hygiene; OR, odds ratio; SE, standard error
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 43
Table 2.12 Overview of the meta-analysis
Number of studies
Pooled OR (95% CI) I2 Quality of
evidence*
OH
(a) Categorical data
- Fair versus Good OH
- Poor versus Good OH
(b) Continuous data
- PI: 1-unit increase
- PSc: 1-unit increase
- OH score
9
15
6
3
25
2.04 (1.65, 2.53)
5.01 (3.40, 7.39)
2.25 (1.43, 3.54)
1.02 (1.01, 1.03)
2.04 (1.59, 2.50)**
40%
78%
81.1%
74.2%
95.6%
Moderate
Oral healthcare habits
- Tooth brushing
- Interdental cleaning
- Dental visits
10
4
6
0.66 (0.47, 0.94)
0.87 (0.75, 1.00)
0.68 (0.47, 0.98)
94.5%
5.1%
60.4%
Very low
Low
Very low
* Quality of evidence: The Grades of Recommendation, Assessment, Development and Evaluation Working Group (GRADE Working Group) ** Pooled standard mean difference (SMD) CI, confidence interval; OH, oral hygiene; OHI, oral hygiene index; OR, odds ratio; PI, plaque index; PSc, plaque score
Attawood Lertpimonchai Literature review / 44
Table 2.13 Search terms and search strategy: Periodontitis and CKD
Item Domains Terms 1
Periodontitis
periodontitis 2 periodontal 3 alveolar bone loss 4 attachment loss 5 pocket depth 6 Periodontitis [MeSH]* 7 Chronic Periodontitis [MeSH]*
8 [1-7 | OR]
9
CKD
chronic kidney disease 10 CKD 11 renal disease 12 kidney failure 13 ESRD 14 ESKD 15 renal Insufficiency 16 glomerular Filtration Rate 17 *GFR 18 creatinine 19 proteinuria 20 albuminuria 21 microalbuminuria 22 macroalbuminuria 23 Kidney disease [MeSH]* 24 Kidney Failure, Chronic [MeSH]*
25 [9-24 | OR]
26
General
risk factor 27 risk study 28 association 29 relation 30 observational study 31 cross-sectional study 32 retrospective study 33 prospective study 34 cohort study 35 longitudinal 36 epidemiolog*
37 [26-36 | OR]
38 8 AND 25 AND 37
* Options for MEDLINE
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 45 Table 2.14 Case-control studies: Periodontitis and CKD
Authors (Year) Cases Controls Periodontitis Results
Oshrain (US, 1979) HD (n = 20)
Controls (n=20) Matched: N/A
Examination: 6 index teeth Index: PI, GI, PDI
All periodontal parameters were poorer in HD than Controls
Tollefsen (Norway, 1985) HD (n = 12)
Controls (n=12) Matched: age, number of teeth, Social status
Examination: - Full-mouth Index: PI, GI, PPD, Bone level
PI: HD > Controls GI, PPD, Bone level: NS
Rahman (Tuskey, 1992) HD (n = 52)
Controls (n=52) Matched: age, number of teeth, Social status
Examination: N/A Index: PI, GI, SBI, PDI, PPD
PI: HD > Controls GI, SBI, PDI, PPD: NS
Galvada (UK, 1999) HD (n = 105)
Controls (n=53) Matched: age, gender
Examination: - Full-mouth - 6 sites per tooth Index: PI, CAL
NS
Marakoglu (Turkey, 2003) HD (n = 36)
Controls (n=36) Matched: age, gender, PI
Examination: - Full-mouth - 6 sites per tooth Index: GI, PPD
NS
Attaw
ood Lertpimonchai
Literature review
/ 46 Table 2.14 Case-control studies: Periodontitis and CKD (cont.)
Authors (Year) Cases Controls Periodontitis Results
Bots (Netherlands, 2006) ESRD (n=84)
Controls (n=42) Matched: age, gender
Examination: - Half-mouth - 6 sites per tooth
Index: PI, PPD, BOP
NS
Borawski (Poland; 2007)
1. CKD (n=38) 2. HD (n = 35) 3. CAPD (n=33)
Controls (n=30) Matched: age, gender
Examination: - Full-mouth - 6 sites per tooth
Index: GI, PI, PPD, CAL
All parameters: HD > Controls
Castillo (Spain, 2007) HD (n = 52)
Controls (n=52) Matched: age, gender
Examination: - Full-mouth - 6 sites per tooth
Index: PI, CAL
NS
Marinho (Portugal, 2007)
1. CKD (n=22) 2. HD (n = 28)
Controls (n=64) Matched: age, gender, weight, education
Examination: - 6 index teeth
Index: PPD, CAL
- PPD: NS - CAL: CKD & HD > Controls
Chamani (Iran, 2009) HD (n = 55)
Controls (n=30) Matched: age, gender, PI
Examination: - Half-mouth
Index: GI, PPD, CAL, BOP
- PPD: NS - GI, BOP, CAL: HD > Controls
Throman (Sweden, 2009)
CKD & HD (n=93)
Controls (n=93) Matched: age, gender
Examination: - 2 index teeth
Index: CAL
CAL: Kidney disease > Controls
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 47 Table 2.14 Case-control studies: Periodontitis and CKD (cont.)
Authors (Year) Cases Controls Periodontitis Results
Dag (Turkey, 2010) HD (n = 43)
Controls (n=43) Matched: age, gender
Examination: - Full-mouth - 6 sites per tooth
Index: PI, GI, PPD
All parameters: HD > Controls
Bhatsange (India; 2012) HD (n = 75)
Controls (n=25) Matched: age, gender
Examination: - 6 index teeth
Index: OHI-S, PDI
All parameters: HD > Controls
Brito (2012, Brazil)
1. CKD (n=51) 2. HD (n = 40) 3. CAPD (n=40)
Controls (n=67) Matched: age, gender
Examination: - Full-mouth - 6 sites per tooth
Index: PPD, CAL
% sites with CAL > 6 mm: Kidney disease > Controls
Parkar (India, 2012) HD (n = 152)
Controls (n=152) Matched: age, gender
Examination: - 6 index teeth
Index: OHI-S, CPI, LOA
All parameters: HD > Controls
Chhokra (India, 2013) ESRD (n = 40)
Controls (n=40) Matched: age, gender
Examination: - Full-mouth - 6 sites per tooth
Index: OHI-S, GI, PPD, CAL
All parameters: ESRD > Controls
Attaw
ood Lertpimonchai
Literature review
/ 48 Table 2.14 Case-control studies: Periodontitis and CKD (cont.)
Authors (Year) Cases Controls Periodontitis Results
Teratani (Japan, 2013) HD (n = 89)
Controls (n=106) Matched: N/A
Examination: - Full-mouth - 2 sites per tooth Index: % sites with CAL ≥ 4 mm
NS
Tiwari (India, 2013) CKD (n = 30)
Controls (n=30) Matched: N/A
Examination: - Index teeth
Index: CPI
CPI: CKD > Controls
Tadakamadla (India, 2014)
CKD (n = 64) GFR: - Creatinine - Cockcroft-Gault
Controls (n=150) Matched: age, gender
Examination: - Full-mouth - 3 sites per tooth Index: OHI GI, CPI
All parameters: CKD > Controls
Zhao (China, 2014) HD (n = 102)
Controls (n=204) Matched: age, gender
- Full-mouth - 6 sites per tooth CPI, CAL
All parameters: HD > Controls
Limeres (Portugal, 2016) HD (n = 44)
Controls (n=44) Matched: age, gender
- 6 index teeth PI, GI, PPD, CAL
All parameters: HD > Controls
C, controls; CAL, clinical attachment level; CAPD, Continuous Ambulatory Peritoneal Dialysis; CKD, chronic kidney disease; CPI, Community Periodontal Index; ESRD, end-stage renal disease; GFR, glomerular filtration rate; GI, gingival index; HD, hemodialysis; LOA, loss of attachment index; N/A, not available; NS, not significant; OHI, oral hygiene index; PDI, Periodontal Disease Index; PI, plaque index; PPD, periodontal pocket depth
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 49 Table 2.15 Cross-sectional studies: Periodontitis and CKD
Authors (Year)
Participants and Setting Primary outcome Periodontitis Analysis Results
Kshirsagar (2005)
ARIC: - USA (1996-1998) - Community-based Participants: - General population - Age: 45-64 - n = 5,537
GFR: - Creatinine - MDRD equations Categorization 1. CKD: GFR < 60 2. non-CKD: GFR ≥ 60
Methods: - Full-mouth exam - 6 sites per tooth Categorization*: 1. Severe 2. Initial 3. Healthy
Binary logistic regression Adjusted: age/gender/center/race smoking/education DM/HT/CVD CRP
Adjusted OR (95% CI) Initial: 2.00 (1.23, 3.24) Severe: 2.14 (1.19, 3.85)
Fisher (2008a)
NHANES III: - USA (1988-1994) - Community-based Participants: - General population - age: ≥ 18 - n = 12,947
GFR: - Creatinine - MDRD equations Categorization 1. CKD: GFR < 60 2. non-CKD: GFR ≥ 60
Methods: - Half-mouth exam - 2 sites per tooth Categorization 1. No periodontitis 2. Periodontitis** 3. Fully edentulous
Binary logistic regression Adjusted: age/gender/race smoking/BMI/CRP DM/HT/Lipid profile Income/education
Crude OR (95% CI) Perio: 3.93 (2.95, 5.24) Eden: 10.38 (7.87, 13.70) Adjusted OR (95% CI) Perio: 1.60 (1.16, 2.21) Eden: 1.85 (1.34, 2.56)
Fisher (2008b)
NHANES III: - USA (1988-1994) - Community-based Participants: - General population - age: ≥ 40 - n = 4,053
GFR: - Creatinine - MDRD equations Categorization 1. CKD: GFR < 60 2. non-CKD: GFR ≥ 60
Methods: - Half-mouth exam - 2 sites per tooth Categorization 1. No periodontitis 2. Periodontitis** 3. Fully edentulous
Binary logistic regression Adjusted: age/gender/race smoking/BMI/CRP DM/HT/Lipid profile Income/education
Crude OR (95% CI) Perio: 1.26 (0.78 to 2.03) Eden: 3.54 (2.54 to 4.93) Adjusted OR (95% CI) Perio: 1.20 (0.76 to 1.90) Eden: 1.61 (1.09 to 2.37)
Attaw
ood Lertpimonchai
Literature review
/ 50 Table 2.15 Cross-sectional studies: Periodontitis and CKD (cont.)
Authors (Year)
Participants and Setting
Primary outcome Periodontitis Analysis Results
Fisher (2009)
NHANES III: - USA (1988-1994) - Community-based Participants: - General population - age: ≥ 18 - n = 5,978/11,955 (randomly selected)
GFR: - Creatinine - MDRD equations Categorization 1. CKD: GFR < 60 2. non-CKD: GFR ≥ 60
Methods: - Half-mouth exam - 2 sites per tooth Categorization 1. No periodontitis 2. Periodontitis** 3. Fully edentulous
Binary logistic regression Adjusted: age/gender/race smoking/BMI/CRP DM duration/HT/Lipid profile Income/education/Hospitalized albuminuria
Crude ORs (95% CI) Perio: 4.50 (3.02 to 6.71) Eden: 10.87 (6.86 to 17.20) Adjusted ORs (95% CI) Perio: 1.60 (1.07 to 2.39) Eden: 2.03 (1.31 to 3.14)
Fisher (2011)
NHANES III: - USA (1988-1994) - Community-based Participants: - General population - age: ≥ 18 - n = 11,211
GFR: - Creatinine - MDRD equations Categorization 1. CKD: GFR < 60 2. non-CKD: GFR ≥ 60
Methods: - Half-mouth exam - 2 sites per tooth Categorization 1. No periodontitis 2. Periodontitis** 3. Fully edentulous
1. Binary logistic regression 2. Mediation analysis - Periodontitis → DM → (HT) → CKD - DM → Periodontitis → CKD Adjusted: age/gender/race smoking/BMI/CRP DM duration/HT/Lipid profile Income/education/Hospitalized albuminuria
Binary logistic regression Adjusted ORs (95% CI) Perio: 1.62 (1.17–2.26) Eden: 1.83 (1.31–2.55) Mediation analysis - DM did not affect CKD directly but mediated through Periodontitis and HT - Periodontitis had a significant direct effect on CKD, and had indirect effect through diabetes duration, and then, hypertension.
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 51 Table 2.15 Cross-sectional studies: Periodontitis and CKD (cont.)
Authors (Year)
Participants and Setting
Primary outcome Periodontitis Analysis Results
Ioannidou (2011)
NHANES III: - USA (1988-1994) - Community-based Participants: - General population - age: ≥ 21 - n = 12,081 - Dentate only
GFR: - Creatinine - MDRD equations Categorization 1. CKD: GFR < 60 2. non-CKD: GFR ≥ 60
Methods: - Half-mouth exam - 2 sites per tooth Categorization: CDC/AAP 1. No/ Mild 2. Moderate/ Severe
Binary logistic regression Adjusted: age/gender smoking/BMI DM/HT/CVD Income/education
Crude ORs (95% CI) - Non-Hispanic Black: 1.85 (1.48, 2.30) - Mexican-American: 2.77 (2.15, 3.55)
Adjusted ORs (95% CI) - Non-Hispanic Black: 1.24 (0.95, 1.62) - Mexican-American: 1.59 (1.14, 2.13)
Grubbs (2011)
NHANES: - USA (2001-2004) - Community-based Participants: - General population - age: 21-75 - n = 6,199
GFR: - Creatinine - MDRD equations Categorization 1. CKD: GFR < 60 2. non-CKD: GFR ≥ 60
Methods: - Half-mouth exam - 3 sites per tooth Categorization: CDC/AAP 1. No/ Mild*** 2. Moderate/ Severe***
Binary logistic regression Adjusted: age/gender/race smoking/ DM/HT Income/education
Crude ORs (95% CI) - 2.50 (1.96, 3.19) Adjusted ORs (95% CI) - 1.55 (1.16, 2.07)
Yoshihara (2012)
Setting: - community-dwelling Participants: - Postmenopausal woman - age: 55-74 - n = 405
Cystatin C Categorization 1. Low (ScC > 0.91) 2. Normal (ScC ≤0.91)
Numbers of remaining teeth Mean CAL: - 10 index teeth - 6 sites per tooth
t-test Compare N (remaining teeth) & Mean CAL between normal and low renal function
Significant differences between renal function groups
Attaw
ood Lertpimonchai
Literature review
/ 52 Table 2.15 Cross-sectional studies: Periodontitis and CKD (cont.)
Authors (Year)
Participants and Setting
Primary outcome Periodontitis Analysis Results
Han (2013) KNHANES: - Korea (2008–2009) - Community-based Participants: - General population - age: 43.4 ± 0.2 - n = 15,729
1. Kidney function: - Creatinine / MDRD Categorization 1.1 CKD: GFR < 60 1.2 non-CKD: GFR ≥ 60 2. Proteinuria (dipstick tests; ≥ 1+) 3. Hematuria (dipstick tests; ≥ 1+)
Methods: - 6 index teeth - CPI Categorization: CPI 1. No periodontitis (CPI 0-2)
2. Periodontitis (CPI 3,4)
Binary logistic regression Adjusted: age, sex, region, education, obesity, smoking, exercise, HT, DM, Lipid, CVD
Adjusted ORs (95% CI) 1. Kidney function: 1.39 (1.03–1.89) 2. Proteinuria 1.29 (0.87–1.91) 3. Hematuria 1.29 (1.15–1.46)
Salimi (2014) NHANES III: - USA (1988-1994) - Community-based Participants: - General population - age: ≥ 20 - n = 13,270
GFR: - Continuous data - - Creatinine - MDRD equations ACR: - Continuous data - - urine albumin - urine creatinine
Methods: - Half-mouth exam - 2 sites per tooth Categorization: CDC/AAP 1. No/ Mild 2. Moderate 3. Severe
Linear regression Adjusted: Age/gender/race Income/education Smoking CVD/HT/HbA1C Total cholesterol Serum lead
1. GFR - Moderate (sig.) R2 = 0.33, β = -0.013 - Severe (sig.) R2 = 0.33, β = -0.001
2. ACR - Moderate (sig.) R2 = 0.16, β = 0.022 - Severe (sig.) R2 = 0.16, β = 0.003
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 53 Table 2.15 Cross-sectional studies: Periodontitis and CKD (cont.)
Authors (Year)
Participants and Setting
Primary outcome Periodontitis Analysis Results
Yoshihara (2016)
Setting: - Japan - Community-based Participants: - postmenopausal never smoking women - age: 55-74 - n = 332
Cystatin C Categorization 1. Low function (ScC > 0.91) 2. Normal function (ScC ≤0.91)
Methods: - Full-mouth - 2 sites / tooth (buccal / lingual) Categorization: PISA Highest quartile Vs Others
Binary logistic regression Adjusted: Osteocalcin, CRP, job, regular check-up, use of inter dental brush or floss, the number of remaining teeth and age
Adjusted ORs (95% CI) 2.44 (1.23, 4.85)
* Periodontitis definition (Kshirsagar, 2005) (1) Severe periodontitis: two or more interproximal sites with CAL ≥ 6 mm that are not on the same tooth AND one or more interproximal sites with PPD ≥ 5 mm (2) Initial periodontitis: two or more interproximal sites with CAL ≥ 4 mm (not on the same tooth) (3) Healthy/gingivitis: individuals not meeting the other 2 criteria. ** Periodontitis definition (Fisher, 2008) Periodontitis: 1 or more sites with CAL ≥ 4 mm and bleeding on the same tooth (bleeding as an indicator of active inflammation) *** Periodontitis definition (CDC/AAP) (1) Severe periodontitis: two or more interproximal sites with CAL ≥ 6 mm that are not on the same tooth AND one or more interproximal sites with PPD ≥ 5 mm (2) Moderate periodontitis: two or more interproximal sites with CAL ≥ 4 mm, or two or more interproximal sites with PPD ≥ 5 mm, not on the same tooth. (3) Normal /Mild periodontitis ACR urine albumin to creatinine ratio; BMI, body mass index; CAL, clinical attachment level; CI, confidence interval; CKD, chronic kidney disease; CPI, Community Periodontal Index; CRP, C-reactive protein; CVD, cardiovascular disease; DM, diabetes mellitus; eGFR, estimated glomerular filtration rate; HbA1C, hemoglobin A1C; HT, hypertension; OR, odds ratio; PISA, Periodontal inflamed surface area; PPD, periodontal pocket depth; SBP, systolic blood pressure; ScC, serum cystatin C; TG, triglyceride; WBC, white blood cell count
Attaw
ood Lertpimonchai
Literature review
/ 54 Table 2.16 Cohort studies: Periodontitis and CKD
Authors (Year)
Participants and Setting
Primary outcome Periodontitis Duration
(year) Analysis Results
Shultis (2007)
Gila River Indian: - Arizona, US - community-based Participants: - Type II DM only - Preserved kidney function, eGFR ≥ 60 AND ACR < 300 - age: ≥ 25 - n: 529
1. MA: ACR ≥ 300 2. ESRD
Number of missing teeth & Panoramic radiographs Categorization*: 1. none/mild 2. moderate 3. severe 4. fully edentulous
9.4
(0.03-21.6) Cox proportional hazard model Adjusted: Age/gender DM duration/BMI Smoking/ DM control
Adjusted HRs (95%CI)
1) MA - Moderate 2.0 (1.2, 3.5) - Severe 2.1 (1.2, 3.8) - Edentulous 2.6 (1.4, 4.6)
2) ESRD - Moderate 2.3 (0.6, 8.1) - Severe 3.5 (0.96, 12.4) - Edentulous 4.9 (1.4, 17.4)
Iwasaki (2012)
Niigata: - community-based Participants: - elderly Japanese - 75-year-old - n = 317
Decreased kidney function: - eGFR - creatinine - Japanese equation Categorization: 1) Decreased 2) No decreased
Full-mouth examination Categorization PISA: The highest quartile Vs Others
2 Logistic regression Adjusted: Baseline kidney function/ gender Income/education Smoking/Alcohol Proteinuria/obesity HT/Hyperglycemia
Adjusted ORs (95% CI)
- 2.24 (1.05, 4.79)
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 55 Table 2.16 Cohort studies: Periodontitis and CKD (cont.)
Authors (Year)
Participants and Setting
Primary outcome Periodontitis Duration
(year) Analysis Results
Chen (2015)
Taipei City: - National Health Examination - community-based Participants: - elderly Chinese - age > 65 - n = 100,263
Decline ≥ 30% of eGFR (creatinine: CKD-EPI)
CPI index (highest sextant) - 0-2 (No periodontitis) - 3-4 (Periodontitis)
3.8 ± 1.7
Logistic Regression Adjusted: age, sex, BMI, smoking, alcohol use, HT, DM, CVD WBC, Hb, level, HDL, triglycerides, uric acid, urea nitrogen, and albumin, baseline eGFR, urinary protein
Adjusted ORs (95% CI) - 1.59 (1.37, 1.86)
Grubbs (2015)
The Jackson Heart Study: - Mississippi, US - community-based - D-ARIC study Participants: - African Americans - preserved kidney function (eGFR ≥ 60) - age: 65.4 ± 5.2 - n = 699
eGFR < 60 OR Decline > 5% annualized loss of eGFR
Full-mouth examination Categorization: 1) CDC/AAP: Severe Vs Non-severe 2) PISA: The highest quartile Vs Others
4.8 ± 0.6
Multivariable Poisson regression Adjusted: age, gender, diabetes, hypertension, smoking, income
Adjusted ORs (95% CI) - CDC/AAP: 4.18 (1.68, 10.39) - PISA: 2.48 (1.04, 5.88)
Attaw
ood Lertpimonchai
Literature review
/ 56 Table 2.16 Cohort studies: Periodontitis and CKD (cont.)
* Periodontitis definition (Shultis, 2007) 1) None/mild periodontitis, defined as ≥ 24 teeth, of which < 6 had 25–49% bone loss and none had ≥50% bone loss 2) Moderate periodontitis, defined as ≥15 teeth, of which < 7 had 50–74% bone loss and < 4 had ≥ 75% bone loss 3) Severe periodontitis, defined as 1–14 teeth or greater bone loss than in previous categories 4) Edentulous
ACR urine albumin to creatinine ratio; BMI, body mass index; CAL, clinical attachment level; CI, confidence interval; CKD, chronic kidney disease; CPI, Community Periodontal Index; CRP, C-reactive protein; CVD, cardiovascular disease; DM, diabetes mellitus; eGFR, estimated glomerular filtration rate; HbA1C, hemoglobin A1C; HT, hypertension; MA, macroalbuminuria; OR, odds ratio; PISA, Periodontal inflamed surface area; PPD, periodontal pocket depth; SBP, systolic blood pressure; ScC, serum cystatin C; TG, triglyceride; WBC, white blood cell count
Authors (Year)
Participants and Setting
Primary outcome Periodontitis Duration
(year) Analysis Results
Grubbs (2016)
MrOS study: - US - Hosipital-based Participants: - Male only - age: 73.4 ± 4.8 - n = 761 - preserved kidney function (eGFR ≥ 60)
eGFR < 60 OR Decline > 5% annualized loss of eGFR
Half-mouth examination Categorization: 1) CDC/AAP: Severe Vs Non-severe 2) European Workshop: Severe Vs Non-severe
4.9 ± 0.3
Multivariable Poisson regression Adjusted: age, diabetes, hypertension, smoking, race, education
Adjusted ORs (95% CI) - CDC/AAP: 1.10 (0.63, 1.91) - European Workshop: 2.04 (1.21, 3.44)
Chang (2017)
MDSCAN: - Taiwan - Hosipital-based Participants: - Referred patients to a multidisciplinary subspecialty - age: 53.1 ± 8.4 - n = 1,486
Progression of color intensity in GFR and albuminuria grid from KDIGO guideline 2012
Not mentioned details of examination Categorization: Mean PPD: - < 3.8 mm - 3.8 – 4.5 mm - ≥ 4.5 mm
4.9
(2.4-7.3)
Multiple Cox regression model Adjusted: age, DM, ACR, sex, HT, creatinine, smoking, and betel nut chewing.
Adjusted HRs (95% CI) Mod periodontitis: 2.06 (1.31, 3.20) Severe Periodontitis: 3.10 (1.99, 4.58)
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 57
Figure 2.1 Flow chart of identifying and selecting studies: Periodontitis and OH
Attaw
ood Lertpimonchai
Literature review
/ 58
Figure 2.2 Pooling effects of fair and poor versus good OH on periodontitis
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 59
Figure 2.3 Pooling ORs of Plaque index and Plaque score on periodontitis
Attawood Lertpimonchai Literature review / 60
Figure 2.4 Pooling effect of oral care habits on periodontitis
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 61
Figure 2.5 Funnel plots of publication bias assessment
Attaw
ood Lertpim
onchai
Literature review / 62
Figure 2.6 Contour enhanced funnel plots
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 63
Figure 2.7 Summary of pooled effect of OH and oral care habits on periodontitis
Attawood Lertpimonchai Literature review / 64
Figure 2.8 Flow chart of identifying and selecting studies: Periodontitis and CKD
41 articles eligible for inclusion
MEDLINE
6 2 2
SCOPUS
1 9 2 0
Screening Titles and Abstracts
2 2 9 8
Excluded:
- 871 Reviews/Case reports/Letters
- 612 No kidney fn/No control group of CKD
- 410 No periodontal examination
- 159 Non-human
- 74 Children
- 48 Other gingival diseases
- 37 Renal transplants
- 34 Studies of periodontal treatment
- 11 Insufficient data
- 1 Non-English
Deleted Duplicates (244)
Case-control study(21)
Cross-sectional study(14)
Cohort study(6)
Cross-sectional study(11)
2 Duplicated study subjects & methodology
1 Reverse relationship (CKD → Periodontitis)
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 65
Figure 2.9 Bi-directional relationship between diabetes and periodontitis
Diabetes Periodontitis
Hyperglycemia
Hyperlipidemia
Insulin resistance
Defective host response
Disturb collagen homeostasis
AGE/RAGE in PMNs, Monocyte
Increase inflammatory cytokines (e.g. IL-1, TNF-α)
Decrease growth factors
Attawood Lertpimonchai Literature review / 66
Figure 2.10 Mechanisms of renal injury due to diabetes
Metabolic pathways Haemodynamic pathways Inflammation
Hyperglycemia Renin-Angiotensin-Aldosterone System
Diabetes
Mechanisms for renal injury
Cellular and Extracellular matrix-related effects
DifferentiationProliferationHypertrophy
Apoptosis
↑ Collagen & fibronectin↑ Connective tissue↑ Inhibition metalloproteinases↓ Matrix degradation
Renal functional and structural changes
Diabetic Kidney Disease
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 67
CHAPTER III
METHODOLOGY
3.1 Study design and setting A sub-cohort of a prospective EGAT cohort was conducted by retrieving
data for 10-year follow up period. Details of the EGAT cohorts were described
elsewhere180, briefly, the EGAT project included 3 parallel cohorts; EGAT1, EGAT2
and EGAT3. In total, 9,082 EGAT employees were randomly selected, and enrolled in
1985, 1998 and 2009 for EGAT1, EGAT2, EGAT3, respectively. The EGAT1
(n=3,499) and EGAT3 (n=2,584) were the headquarters which located in Nonthaburi
province, while, the EGAT2 (n=2,999) included urban employees (n=1,998) and
employees who worked at hydro-electric plants in remote areas (n=1,001). All
participants underwent medical examinations and completed self-administered
questionnaire every 5 years, except only 12-year-gap between the 1st (in 1985) and the
2nd (in 1997) surveys of EGAT1. In 2002, the 3rd survey of EGAT1 (called EGAT1/3),
periodontists were invited to collaborate with the EGAT cohort by setting up a half-
mouth examination. Then, the examination has been expanded to the full-mouth
examination since 2003 (EGAT2/2; 2nd survey of EGAT2).
This study was conducted using the EGAT2 cohort. The EGAT2/2 was
used as the baseline, whereas the EGAT2/3 (in 2008) and EGAT2/4 (in 2013) were
defined as the follow-up visit at 5 and 10 years. Two casual pathways were constructed
with the mediation analysis. Rationale and setting details of each pathway were shown
as follows:
3.1.1 Pathway A: Periodontitis → DM → CKD
3.1.1.1 Rationale
Initially, we aimed to determine the effects of periodontitis on
CKD, therefore periodontitis was set as the independent factor and CKD was the
interested outcome. As for biological background, a systemic inflammation from
Attawood Lertpimonchai Methodology / 68
periodontitis could induce the resistance of insulin receptors, then, causes the DM157. In
the same time, the hyperglycemic condition also increase risk of CKD181. Hence, we
hypothesize that periodontitis could directly affect on CKD, and it could affect through
DM (Figure 3.1).
3.1.1.2 Setting
To determine the direct and indirect effects of periodontitis in
the causal pathway A, ideally, subjects with CKD or DM in 2003 should be excluded,
remaining the non-CKD and non-DM cohorts at baseline. New case of DM and CKD
would be occurred during follow up in 2008 and 2013. The mediation effect through
DM would be identified if there was occurrence of DM, and CKD subsequently. To
apply these to ours setting, we allowed the first 5-year-interval (EGAT 2/2 to EGAT
2/3) for DM onset that was affected by periodontitis. Then, the second 5-year-interval
(EGAT 2/3 to EGAT 2/4) was a duration of CKD development after DM diagnosed.
However, typically, the onset of overt nephropathy detected by deterioration of GFR or
proteinuria is about 10 and 15 years after the onset of the DM10. As a result, the 5-year-
interval in our setting was too short follow up time. The effect of DM on CKD incidents
might be underestimated with improper follow-up duration. Therefore, we had to
compromised by including DM cases at baseline and assuming that all cases of DM
were influenced by periodontitis. From here, the duration for estimation the risk effect
of DM on overt nephropathy would be allowed to 10 years. In addition, in case that DM
and CKD were diagnosed simultaneously, we assumed that DM developed before CKD.
3.1.2 Pathway B: DM → Periodontitis → CKD
3.1.2.1 Rationale
As for a bi-directional association of periodontitis and DM, not
only periodontitis influences the glycemic control as mentioned in pathway A, but DM
might also increase risk of periodontitis11. Therefore, this causal pathway hypothesized
that periodontitis may be the mediator in the causation of DM on CKD (Figure 3.2).
3.1.2.2 Setting
In the pathway B, DM, periodontitis and CKD were set as the
independent variable, mediator, and interested outcome, respectively. Similarity with
the pathway A, preferably, CKD and periodontitis at baseline in 2003 should be
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 69
excluded. However, the prevalence of periodontitis was approximate 60%. To exclude
all periodontitis cases may cause insufficient the power of analysis. Therefore, subjects
with periodontitis at baseline were compromise enrolled.
3.2 Study subjects All studied subjects in EGAT 2 were included into this study if they
participated at least one health survey in 2003, 2008, and 2013. Subjects were excluded
if they were diagnosed as CKD at baseline in 2003 (eGFR < 60 ml/min per 1.73m2),
absence of ALL periodontal examinations due to (1) refusal, (2) systemic conditions
which required antibiotic prophylaxis before dental procedure including congenital
heart disease or valvular heart disease, previous history of bacterial endocarditis or
rheumatic fever, total joint replacement and end-stage renal disease, and (3) fully
edentulous subjects
3.3 Data collection In each survey, self-administered questionnaires were used to collect
general demographic data (age, gender, educational level, income, marital status),
behavioral data (smoking status, alcohol consumption, exercise/physical activity),
family history of illness, underlying diseases (DM, HT, stroke, CKD, dyslipidemia
(DLP), tuberculosis, allergy, autoimmune diseases, osteoarthritis, recent fracture or
Parkinson’s disease), and use of medication. Physical examinations, i.e., blood pressure
(BP), heart rate, weight, height and waist & hip circumference, were performed by
clinicians and trained personnel from Ramathibodi Hospital. Laboratory tests under
fasting state were carried out included glucose, total cholesterol, low-density lipoprotein
(LDL), high-density lipoprotein (HDL) and triglycerides, creatinine, total protein,
albumin, total bilirubin, direct bilirubin, aspartate aminotransferase (AST), alanine
aminotransferase (ALT), alkaline phosphatase (AP), gamma-glutamyl transpeptidase
(GGT) and a complete blood count (CBC). Electrocardiograms and chest X-rays were
routinely recorded in every survey.
Attawood Lertpimonchai Methodology / 70
3.4 Study factor and measurements
3.4.1 Periodontal examination
Dental examinations were carried out by experienced periodontists from the
Department of Periodontology, Faculty of Dentistry, Chulalongkorn University in
mobile dental units. Examinations were comprised of the number of missing teeth and
retained roots, periodontal examinations, and evaluation of treatment needs. Periodontal
examinations included PPD, and gingival recession (RE) which were carried out on all
fully erupted teeth, except third molars and retained roots. PPD and RE were measured
using a PCP-UNC15 probe on six sites per tooth, i.e., mesio-buccal, mid-buccal, disto-
buccal, mesio- lingual, mid-lingual, and disto-lingual. These measurements were made
in millimetres and were rounded to the nearest whole millimetre. CAL was calculated
from PPD and RE (Figure 3.3), and represented the distance from cemento-enamel
junction to the tip of a periodontal probe83.
Calibration and standardization for periodontal measurements were carried
out among six to eight examiners before the survey. The weighted kappa coefficients
(±1 mm) was used to determine the agreement of inter-examiner and intra-examiner
(Table 3.1). Between each pair of examiners, the kappa ranged from 0.72 to 1.00 for
PPD and 0.67 to 1.00 for CAL/RE. The weighted kappa coefficients (±1 mm) within
each examiner ranged from 0.85 to 1.00 for PPD and from 0.80 to 1.00 for CAL.
3.4.2 Periodontitis classification
Due to lacking uniformity of periodontitis definition in periodontal
medicine researches, we proposed classifications of periodontitis in various criteria as
follows:
3.4.2.1 CDC/AAP definitions
Using PPD and CAL58, periodontitis was categorized into 3
groups as normal/mild, moderate, and severe periodontitis. Moreover, severe
periodontitis versus non-severe periodontitis (normal/mild/moderate) was also an
alternative categorization.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 71
- Severe periodontitis: two or more interproximal sites with
CAL ≥ 6 mm that are not on the same tooth AND one or more interproximal sites with
PPD ≥ 5 mm
- Moderate periodontitis: two or more interproximal sites with
CAL ≥ 4 mm, or two or more interproximal sites with PPD ≥ 5 mm, not on the same
tooth.
- Normal /Mild periodontitis
3.4.2.2 Using the disease extent in continuous scale
Although, the CDC/AAP definition is the standard criteria, the
amounts of inflammation seem to be discrepancies within the same category. For
instance, subjects who had only 2 teeth of severe periodontitis would be grouped as the
same as subjects who had whole-mouth with severe form. Therefore, the disease extent
defined as the percentage of disease site/periodontitis sites, ranged from 0 to 100%, was
also adopted in this study. Definitions for periodontitis site are shown as follow:
- Periodontitis A: PPD ≥ 4 mm
- Periodontitis B: PPD ≥ 6 mm
- Periodontitis C: PPD ≥ 4 mm AND CAL ≥ 3 mm
- Periodontitis D: PPD ≥ 4 mm AND CAL ≥ 5 mm
- Periodontitis E: CAL ≥ 3 mm (proximal sites only)
- Periodontitis F: CAL ≥ 5 mm (proximal sites only)
Example 1: Subjects, who had 20 teeth, they would have total
120 measured sites (6 sites per tooth). If 24 sites were found with PPD ≥ 4 mm, the
extent of periodontitis would equal 20%, when defined disease site as Periodontitis A.
Example 2: Subjects, who had 20 teeth, they would have total
80 measured proximal sites (4 sites per tooth). If 24 proximal sites were found with CAL
≥ 5 mm, the extent of periodontitis would equal 30%, when defined disease site as
Periodontitis F.
Attawood Lertpimonchai Methodology / 72
3.5 Primary outcome and measurements The outcome of interest was the incidence of CKD stage III1, which was
defined as subjects who had preserved kidney function in 2003, and subsequently their
eGFR < 60 ml/min per 1.73m2 in 2008 or 2013. Serum creatinine was measured using
IDMS-standardized enzymatic assay on the Vitros 350 analyzer (Ortho-Clinical
Diagnostics, USA). The eGFR was then calculated using the Chronic Kidney Disease
Epidemiology Collaboration (CKD-EPI: 2009) equations37 (Table 3.2).
3.6 Other co-variables and measurements
3.6.1 Self-administered data
3.6.1.1 Demographic data: age, gender, marital status,
education, income
3.6.1.2 Smoking: Smoking habit was categorized as (1) never
smokers (2) quit smokers, and (3) current smokers, based on multiple questions in
questionnaires including past/current smoking habits, quantity and duration of smoking,
age at start or quit smoking.
3.6.1.3 Alcohol drinking: Alcohol classification was similar
with smoking habits. History of alcohol consumption, along with, frequency, duration,
and type of alcohol were considered.
3.6.1.4 Exercise (Physical activity): Frequency of exercise was
used to categorized physical activity as (1) none exercise (2) 1-2 times/week, and
(3) ≥ 3 times/week
3.6.1.5 NSAIDs use: Users were identified from current
medication, and then, classified as “Yes” or “No”.
3.6.1.6 Underlying diseases and current medication: DM, HT
and DLP were identified from physical examinations and laboratory, along with the
prescribed treatments.
- DM was diagnosed if an individual had fasting blood sugar
(FBS) ≥ 126 mg/dl or had been taking anti-diabetic drugs182.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 73
- HT was diagnosed if the participant had systolic blood pressure
(SBP) ≥ 140 mmHg or diastolic blood pressure (DBP) ≥ 90 mmHg, or had been taking
prescribed BP lowering drugs183.
- DLP was classified using the guidelines for the management
of dyslipidemias184. Subjects were classified as having DLP if they had HDL < 50 mg/dl
in male or HDL < 40 mg/dl in female OR LDL ≥ 160 mg/dl OR triglyceride ≥150 mg/dl
OR used any lipid-lowering medications.
3.6.1.7 Family history of DM, HT, DLP
3.6.2 Physical examination data
3.6.2.1 Body measurement: Height was measured in
centimetres and weight was measured in kilograms with dressed in normal clothing
without shoes. Hip and waist circumferences were measured in centimetres using
measuring tape by trained staffs.
3.6.2.2 Body mass index (BMI): BMI was calculated from the
recorded weight in kilograms divided by squared height in meters. BMI was
categorized185 as (1) underweight (<18.5 kg/m2), (2) normal (18.5-24.9 kg/m2), and (3)
overweight (≥ 25 kg/m2).
3.6.2.3 Waist-hip ratio (WHR): WHR was calculated from the
recorded waist in centimetres divided by hip in centimetres. WHR was categorized as
normal and obesity with cut-off value of ≥ 0.9 in male or ≥ 0.85 in female186.
3.6.2.4 Blood pressure: SBP and DBP were measured in sitting
position after 5-minute-rest using a calibrated automatic blood pressure monitors.
3.6.3 Laboratory data
Blood samples were collected after 12-hour overnight fasting. Blood
glucose was measured by plasma samples in mg/dl (Peridochrome, Boehringer
Mannheim, Mannheim, Germany). Serum total cholesterol, triglyceride, LDL, HDL
were measured in mg/dl using enzymatic-calorimetric assays (Boehringer Mannheim,
Mannheim, Germany).
Attawood Lertpimonchai Methodology / 74
3.7 Sample size estimation Sample size was estimated based on testing two independent proportions.
Type I and type II errors were set at 5% and 20%, respectively. The ORs of presence
versus absence periodontitis that could be detected was set at 1.5. The ratio of subjects
with normal periodontium versus subjects with periodontitis was set at 1:2, as for
approximating from EGAT2/2. As for previous studies39, 41, 187, 188 which was similar to
this study in terms of CKD definition and time horizon (10-12 years), the incidence of
CKD varied from 9.1 to 16.9%. Therefore, a total of 502 subjects with normal
periodontium and 1,004 subjects with periodontitis were required to detect the risk ratio
of 1.5.
The availability was explored from existed data. Among 2,651 participants
in 2003, 1,821 participants had completed data in 2008 and 2013. Among them, 1759
participants were free from CKD at baseline. As a result, a total of non-CKD participants
with completed data would be sufficient to detect the risk ratio of 1.5. Moreover, the
plan with the imputation should lead to increase more power of test.
3.8 Data management
3.8.1 Data acquirement
3.8.1.1 Demographic and medical records
Demographic and medical data were retrieved from the EGAT
databases. These were merged with the Excel worksheets of the civil registrations for
additional data.
3.8.1.2 Periodontal databases
Periodontal databases were constructed, all periodontal
parameters for EGAT2/2, 2/3 and 2/4 were computerized as follows:
- Build the periodontal databases: Databases were constructed
using Epidata version 3.1, separately by EGAT2/2, 2/3 and 2/4, because there were some
variables were differently measured for each survey. Sequencing of data entry were
designed with “tooth by tooth” system. It means that users had to entry all parameters
including PPD, and RE, tooth by tooth. If a tooth was missing, the system would not
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 75
allow users to entry any data for that tooth. In addition, databases were encoded the
specific value/range for each variable to prevent data entry error.
- Data entry (Periodontal parameters): Manual checking on case
record forms (CRF) was done by a data manager before entering data. This process
included revise of the readable of handwriting, minor missing data and consistence of
every parameters. If handwriting was not clear, the query was done directly to the
recorder. Then, data were independently entered twice by two persons. These two data
sets were then validated, any inconsistence was checked and corrected. Finally, all
records were manually checking and edited based on the original CRF, again.
3.8.2 Data cleaning
All interested variables along with the periodontal data were retrieved from
the main databases. They were renamed systematically across three EGATs in order to
combine them all together. Then, data cleaning was performed by the data cleaning
TEAM, which consisted of Asso.Prof. Ammarin Thakkinstian, Asst.Prof. Sasivimol
Rattanasiri, Dr. Attawood Lertpimonchai and Miss Sukanya Siriyotha. Regular meeting
at least twice a month was organized to solve any incorrectness or unclear data. Data
were summarized and cross-checked using STATA software. Any inconsistency or
outliners were verified and checked with the CRFs to make sure that data were valid.
All variables were assigned as the time-varying variables, if possible. Only gender and
height were considered as the fixed variables. The cleaning process of each variable was
described in Figure 3.4 to Figure 3.13
3.8.2.1 Gender
Gender was expected to be consistent among all databases.
Heterogeneity was solved by rechecking the original CRF (Figure 3.4).
3.8.2.2 Examination date
The date of examination was used to calculated “age”. It had to
be within that survey periods. If not or missing data presented, it was recoded as the
middle time of that survey (Figure 3.5).
3.8.2.3 DOB
Although, the DOB form the civil registration was available, its
validity was questioned because some errors were detected, such as inconsistency date
Attawood Lertpimonchai Methodology / 76
format, unreliable year of birth. Hence, we planned to merge the civil registration data
with the databases for improving the reliable (Figure 3.6). Similar with the gender, DOB
was expected to be consistent among all EGAT and the civil registration databases. If
the discrepancies were presented, it would be solved with the majority.
3.8.2.4 Marital status
Marital status was checked after combined data from 3 periods
together. Subjects, who were married or divorced, and subsequently became single
would be detected. Then, it would be discussed within the TEAM (Figure 3.7).
3.8.2.5 Education
The level of education cannot be decreased. Thus, the non-sense
declinations were detected and made decision with the TEAM (Figure 3.8).
3.8.2.6 Risk behaviors (smoking and alcohol drinking)
First, smoking and alcohol drinking habits were classified
within each period with multiple questions in questionnaire. If inconsistency was
present, the TEAM would consider the surrounding details of that behavior, as much as,
we could (Figure 3.9). Once, data was cleaned in each period, it would be merged and
judged with the same logic as the marital status was. For example, current smokers could
not become never smoker later.
3.8.2.7 Body measurement
Height (Figure 3.10), weight, waist and hip were summarized
and checked for outliers (exceed mean ± 4SD). If outliers presented, the original CRF
was proven. Moreover, the change within subject overtime would be checked after
merge data across periods. The astonishing change of weight, waist and hip would be
list, and then, its possibility would be validated by comparing with other relevant
variables (Figure 3.11).
3.8.2.8 Blood pressure
Concurrently present of SBP and DBP, within the proper range
of BP, SBP > DBP, and appropriate pulse pressure were used as the guideline to certify
BP (Figure 3.12).
3.8.2.9 Laboratory results
All laboratory results, which were reported in continuous data,
were checked for outliers (i.e., exceed mean ± 4SD). If outliers existed, the TEAM
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 77
would discuss the likelihood of that values (Figure 3.13). If it was not reasonable, it was
replaced with missing data.
3.8.3 Completed data with carried forward/backward methods
To complete database as much as possible, the forward/backward carry
over methods were used to replace missing data for some variables, i.e., marital status,
education, smoking and alcohol drinking. Some missing data in some circumstances
were replaced with previous or subsequent data. For example, never smokers in EGAT
2/4 whose smoking status in EGAT 2/2 and 2/3 were missing. These missing data was
replaced with the “never smokers”. The processes of carried forward/backward methods
is shown in Figure 3.7 to Figure 3.9.
3.9 Imputation
3.9.1 Imputation methods
Missing data were assumed as missing at random (MAR); the multiple
imputation using chain equation (MICE) for longitudinal data was performed for both
within-wave and whole-wave missing data189-191. Frequency and patterns or
distributions of missing variables were explored to check for the MAR assumption.
Detail of predicted variables for each chain equation was summarized in Table 3.2. Data
for all 3 periods (i.e., EGAT 2/2, 2/3 and 2/4) were combined as one dataset using wide
format, i.e., one record per one subject. This format allowed us to use the same factor
as the predictors of itself in different time. For example, serum creatinine at EGAT 2/2
could be assigned as the predictor of missing serum creatinine at EGAT 2/3.
3.9.2 Variable types in the imputation model
All analyzed variables were included in the imputation model including,
age, gender, marital status, education, income, smoking, alcohol drinking, exercise,
NSAIDs use, height, weight, waist and hip circumferences, blood pressure, family
history of DM, HT and DLP, serum glucose, cholesterol, HDL, LDL, triglyceride, uric
acid, creatinine, PPD and CAL in every site. According to the concept of MICE, some
Attawood Lertpimonchai Methodology / 78
variables in the imputed model must have completed data which called “regular
variables”. The MICE initially used them as the anchor of beginning in the chain
equations to generate set of imputed data for incomplete variables or “imputed
variables”. Here, age, gender and family history of DM, HT and DLP were assumed as
the regular variables, meanwhile, the remaining variables were registered as imputed
variables192.
3.9.3 Imputation modeling
Each model was constructed for each imputed variable based on its own data
type/distribution of that data and thus link function. In addition, predictors of each
imputed model were selected individually based on biological plausibility. Type of
models and its predictors were selected as described in the Table 3.2. In brief,
imputation modeling was selected as follows: the ordinal logit model was used for
imputation of ordinal variable (i.e., income), the multi-nominal logistic regression
model was used for imputation of nominal variable (i.e., exercise), the logistic
regression model was used for imputation of dichotomous variable (i.e., medication of
lipid drug), and the linear regression model was used for imputation of continuous
variables (i.e., height, weight, waist circumference, hip circumference, SBP, DBP).
Although, the conventional imputation could be performed and all missing
data were filled, some imputed values were still unreliable by producing outliner values
such as 8 mg/dl of serum glucose, or declination of education level. In order to solve
this problem, an interval linear regression model was applied for all laboratory results
by setting up the possible upper and lower boundary for each variable in each period.
Moreover, the interval linear regression was also applied for conditional variables, i.e.,
marital status, education, smoking and alcohol drinking by pretending them as the
continuous data, and the upper and lower boundary were set based on logical reasoning
similar with the cleaning processes. Furthermore, the truncated regression, which
regularly uses for imputing continuous variable with a restricted range, was modeled for
the PPD and RE.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 79
3.9.4 Numbers of imputations
Among eligible subjects, the missing data of studied variables range from
0.38% to 18.30%, thus, twenty imputations were initially constructed191. The sufficient
of imputations were assessed in the estimation models based on performances of
imputation measured by the largest fraction of missing information (FMI) of coefficient
estimates due to missing data.
3.9.5 Data management after multiple imputation
Composited variables or “passive variables”, i.e., BMI, WHR, DM, HT,
CAL, and periodontal status, were calculated and/or classified from the imputed
variables. The categorical variables which were imputed by interval linear regression
model were nearest rounded to be the integer. Then, all imputed variables were cleaned
and checked with the cleaning processes similar with the original data, again.
3.10 Statistical analysis Characteristics of studies subjects were described using mean and SD or
median (range) for continuous data, where appropriates; frequency and percentage for
categorical data. The outline of all analysis was summarized in Figure 3.14
3.10.1 Mediation analysis: Periodontitis → DM → CKD
A mediation analysis was conducted using rationale and statistical
procedures outlined by Baron and Kenny193. Analyses were done and reported into 2
parts separately for each casual pathway. According to pathway A (Figure 3.1), effect
of periodontitis on CKD through DM as the mediator, mediation analysis for
categorical data194, 195 were applied. With conceptual framework of mediation analysis,
DM was assigned as the mediator of periodontitis and CKD if a) periodontitis has a
statistically significant effect on DM (mediation model) b) DM is significantly
associated with the CKD incident after controlling for the periodontitis effect (outcome
model)
Attawood Lertpimonchai Methodology / 80
3.10.1.1 Equations and calculations
To explore these, two equations from causal pathway were
constructed as below.
path a
logit !"#$!" = a0 + ax1 + ∑ ek zk
path b, c’
logit %&!#$%&! = c0 + c’x1 + bm1 + ∑ ek zk
where
x1 periodontitis
m1 0, 1 for non-DM and DM
zk confounders
For mediation model, DM was regressed on periodontitis, called
path a. For the outcome model, CKD was regressed on DM and periodontitis, called
path b and c’, respectively. The generalized structural equation model (GSEM) was
applied to constructed these two equations using logit link functions for both taking into
account for within and between variations of imputed data sets, and also longitudinal
data. Then, the average causal mediation effect (ACME) was estimated using product-
of-coefficient method. Then, the total and direct effects were estimated. Simultaneously,
the ORs of these effects also calculated. Formulas are shown as follow196, 197:
Mediation effect:
ACME = ab
Total effects (TE):
TE = ab + c’
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 81
Percent of total effect mediated through DM (% MEdm):
%)*+, = ./
./ + 1′
Percent of direct effect (% DE):
%3* = 1′./ + 1′
The ORs of mediation effect (ORsACME):
ORs = expab
3.10.1.2 Selection of variables
The GSEM198, 199 was used to constructed the outcome (CKD)
and mediator (DM) model as mentioned above. In the CKD model, periodontitis and
DM were considered as exposure and mediator, respectively. Other relevant risk factors
of CKD were included as confounders including education, income, marital status,
smoking, alcohol drinking, exercise, NSAIDs use, obesity, HT, DLP, and uric acid. Age
and gender were not included in the CKD model because they had been already
accounted in the eGFR calculation. In DM model, periodontitis was recognized as the
exposure with other confounders, i.e., age, gender, education, income, marital status,
smoking, alcohol drinking, exercise, obesity, HT, DLP, and family history of DM.
All proposed periodontitis definitions, i.e., CDC/AAP
definition, and the periodontitis extent in continuous scale (Periodontitis A to F), were
applied in the univariate analysis logit model one-by-one. The definition which was
given the maximum F-test was selected for periodontitis representativeness. Similarly,
the maximum F-test criteria was applied for selecting BMI or WHR to be a represent
for obesity.
The methods of variables selection for final model is another
issue that have been recently widely discussed200-204. The conventional stepwise method
is standard and generally approach for epidemiological researches. However, when
applied the stepwise to claim causation, its rationale is questioned in reflection the
biology of the process202. Moreover, it may induce a bias and inefficiency as a result of
overfitting in the outcome regression model200, 204. The disjunctive cause criterion203 was
introduced and claimed the advantages for cause–effect relationships, in particularly,
Attawood Lertpimonchai Methodology / 82
the treatment effect and the mediation analysis. Confounders were selected if they
associated with either mediator or outcome. Therefore, in this study, the selection of
confounders for the multivariate GSEM model or final model was further performed the
confounder selection with 3 methods as follows.
- Conventional forward stepwise
The final model of outcome and mediator were constructed,
separately. Factors with p-value < 0.10 in the univariate analysis for each model were
selected and forwardly included in the final model one by one. Only significant variables
were kept. As a result, the two models might contain different confounders. - Partial disjunctive clause criteria
The significant factors in the multivariate of either CKD or DM
model from conventional stepwise were combined as the set of confounders. Then, all
of them were considered to include into both models. Although, some variables might
be significant in the DM model, but not for the CKD model, or vice versa, they might
be left to modified others and thus should be included.
- Modified fully disjunctive clause criteria
With this method, a set of confounders that were significant in
univariate analysis from either DM or CKD model was considered and included in both
models. It would be the loosest criteria resulting in being the highest number of co-
variables in final model.
3.10.2 Mediation analysis: DM → Periodontitis → CKD
A causal pathway B was constructed as the inverses association of
periodontitis and DM. Here, DM was considered as the independent variable, and
periodontitis was considered as the mediators. Steps of analyses were performed
similar to the previous approaches for pathway A; except that periodontitis was the
mediator instead of DM (switched between independent variable and mediator). Thus,
the CKD and periodontitis model were constructed. In the periodontitis model, if the
disease extent (continuous data,) was selected to representing periodontitis, the model
might be built by the linear regression model. In addition, age, gender, education,
income, smoking, alcohol drinking, exercise, HT and DLP were recognized as
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 83
confounders for periodontitis model. Then, the mediation effect estimation and
processes of variable selection were performed similar with previous one.
3.10.3 Bootstrapping
A bootstrap analysis with 1,000 replications was applied to test the
mediation effect without requiring the assumption of normality205. For each bootstrap
sample, the mediation effect, percent of total effect mediated through mediator, ORs of
mediation effect were estimated. Then, the estimations and their 95% CI were
determined using bias-corrected bootstrap technique.
3.10.4 Checking assumptions
To test the mediation effect, the basic assumptions according to the essential
steps of the Baron and Kenny should been checked. The significant effect of X regressed
on M, and then the significant effect of M regressed on Y should be presented.
Moreover, the sequential ignorability assumption was the additional assumption for
mediation analysis proposed by Imai, Keele, and Yamamoto206. First, given the
observed pretreatment/exposed confounders, assignment of treatment/exposure was
assumed to be ignorable/independently, i.e., statistically independent of exposure (here
was periodontitis) and confounders for the mediation model. The second assumption
states that the mediator was ignorable in the outcome model given the observed
exposure and pretreatment confounders. In other words, we assumed that the following
two statements of conditional independence hold:
[Yi (t', m), Mi(t)] ||�Ti | Xi = x
Yi (t', m) || Mi(t) | Ti = t, Xi = x
However, our interested independent variable was not randomly allocated
like a randomized controlled trial (RCT), it was just observed. Therefore, adjusting
confounders should lead to meet these two assumptions and the ignorability assumption
should be assumed195. Here, after estimated the effect size in the final multivariate model
with stepwise method, the residuals from the mediators and outcome models were post-
Attawood Lertpimonchai Methodology / 84
estimated. Then, Pearson's correlation was used to test the independence, assuming the
low correlation represented the higher chance of ignorability.
In addition, the misspecification assumption from causal order, causal
direction, unmeasured variables and imperfect measurement were other assumptions for
the mediation analysis. However, testing them with statistical approaches seem to be
difficult or untestable195. The reliable of conceptual framework and background
knowledge were used to support the assumption.
All analyses were performed using Stata version 14.2. All analyses were
performed based imputed data under the “mi estimate” commands, which considered
within and between imputed data variations. P-value of less than 0.05 was considered
as a threshold for statistical significance.
3.11 Ethics considerations This retrospective study used the demographic, medical and dental data
from EGAT projects which are belonging to Faculty of Medicine Ramathibodi hospital,
Mahidol University and Faculty of Dentistry, Chulalongkorn University. The
permission to access the database was asked directly to the Head of EGAT projects
(Prof. Piyamitr Sritara), as well as, the Head of Periodontal Department, Faculty of
Dentistry, Chulalongkorn University (Prof. Rangsini Mahanonda). Before making
decision about the permission, they were clearly informed about the objectives, benefits
and methodology of this study. This study was approved by Institutional Review Board
of Ramathibodi’s Ethical Committee on June 27, 2008 with MURA2008/809/S41.
(Appendix C).
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 85
Table 3.1 Calibration of periodontal examination (weight kappa ± 1mm)
CAL, clinical attachment level; PPD, periodontal pocket depth; RE, recession
PPD RE / CAL
Inter-examiner Intra-examiner Inter-examiner Intra-examiner
EGAT 2/2 0.72 - 0.90 0.85 - 0.96 0.69 - 0.79 0.80 - 0.97
EGAT 2/3 0.77 - 0.89 0.87 - 0.91 0.67 - 0.94 0.90 - 0.96
EGAT 2/4 0.74 - 1.00 0.87 - 1.00 0.78 - 1.00 0.87 - 1.00
Attaw
ood Lertpimonchai
Literature review
/ 86
Table 3.2 Imputation models: predictors and equations
Predictors
Model* Se
x
Age
Educ
atio
n
Inco
me
Mar
ital s
tatu
s
Smok
ing
Alc
ohol
Exer
cise
Wei
ght
Hei
ght
Wai
st
Hip
SBP
NSA
IDs
Lipi
d D
rug
Glu
cose
Cho
lest
erol
HD
L
LDL
Trig
lyce
ride
Uric
Aci
d
Cre
atin
ine
FM-H
T
FM-D
M
FM-D
LP
Education Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Income Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö ologit Marital status Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Smoking Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Alcohol Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Exercise Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö mlogit Height Ö Ö Ö Ö Ö Ö regress Weight Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö regress Waist Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö regress Hip Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö regress SBP Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö regress DBP Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö regress Lipid Drug Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö logit Glucose Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Cholesterol Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg HDL Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg LDL Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Triglyceride Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Uric Acid Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg Creatinine Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö intreg PPD Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö truncreg RE Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö Ö truncreg
* Type of imputation equation for each variable: intreg, interval regression; ologit, ordered logistic regression; mlogit, multinomial logistic regression; regress, linear regression; logit, logistic regression; truncreg, truncated regress DBP, diastolic blood pressure; FM-DLP, family history of dyslipidemia; FM-DM, family history of diabetes; FM-HT, family history of hypertension; HDL, high-density lipoprotein; LDL, low-density lipoprotein; PPD, periodontal pocket depth; RE gingival recession; SBP, systolic blood pressure
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 87
Figure 3.1 Structural causal pathway A: Periodontitis → DM → CKD
( A )
c
c'
ba
Periodontitis (X) CKD (Y)
Periodontitis (X)
DM (M)
CKD (Y)
( B )
Attawood Lertpimonchai Methodology / 88
Figure 3.2 Structural causal pathway B: DM → Periodontitis → CKD
( A )
c
c'
ba
DM (X) CKD (Y)
DM (X)
Periodontitis (M)
CKD (Y)
( B )
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 89
Figure 3.3 Periodontal measurement CEJ, cemento-enamel junction
Attawood Lertpimonchai Methodology / 90
Figure 3.4 Workflow of cleaning processes for gender CRF, case record form
Yes No
Check data entry error with original CRF
Consistency?
No
Recheck with the Civil Registration
Yes; data entry error
Edited data
CLEANED DATA
CLEANED DATA
MERGE DATASETS
GENDER
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 91
Figure 3.5 Workflow of cleaning processes for date visit (examination date)
Yes No
Replaced with the median time of that survey
Within actual survey date?
CLEANED DATA
DATE VISIT (Examination date)
Yes No
Missing data?
CLEANED DATA
Replaced with the median time of that survey
CLEANED DATA
Attawood Lertpimonchai Methodology / 92
Figure 3.6 Workflow of cleaning processes for date of birth CRF, case record form
Match with given age?
Yes No
Replaced with missing
Consistency?
Can we assume from the majority?
Data entry error ? (checked with original CRF)
Yes (Data entry error)
Edited data
No
W I T H I N each dataset
MERGE DATASETS
Yes No
CLEANED DATA
Data entry error ? (checked with original CRF)
Yes (Data entry error) No
Edited data
DATE OF BIRTH
(EGAT Database)
DATE OF BIRTH
(Civil Registration Database)
YesNo
CLEANED DATA
Assume data from the Civil Registration
CLEANED DATA
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 93
Figure 3.7 Workflow of cleaning processes for marital status CRF, case record form; TEAM, data cleaning team
MERGE DATASETS
MARITAL STATUS
Reasonable overtime sequence?
No. For example, married subjects became singles Yes
Discussed with TEAM
CLEANED DATA
Yes (Data entry error)
Edited data
No
Data entry error ? (checked with original CRF)
CLEANED DATA
MARITAL STATUS was single in EGAT 2/4 ?
No Yes
Replaced 2/2 & 2/3 as “single”
CLEANED DATA & CARRIED BACKWARD
F i l l i n g m i s s i n g d a t a
Attawood Lertpimonchai Methodology / 94
Figure 3.8 Workflow of cleaning processes for education CRF, case record form; TEAM, data cleaning team
No Yes
Check data entry error with original CRF
Education level decreased overtime?
No Yes; data entry error
Edited data
CLEANED DATA
CLEANED DATA
MERGE DATASETS
EDUCATION
Discussed with TEAM
EDUCATION 24 was ≤ secondary school (the lowest level) ?
NoYes
EDUCATION 22 was ≥ Bachelor's degree (the highest level) ?
NoYes
EDUCATION 23 was missing ANDEDUCATION 22 equals EDUCATION 24 ?
Replaced EDUCATION2/2 & 2/3 with
“ ≤ secondary school ”
Replaced EDUCATION 2/3 & 2/4 with
“ ≥ Bachelor's degree ”
NoYes
Replaced EDUCATION 2/3 with EDUCATION 2/2
CLEANED DATA & CARRIED FORWARD/BACKWARD
F i l l i n g m i s s i n g d a t a
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 95
Figure 3.9 Workflow of cleaning processes for risk behaviors including smoking and alcohol drinking CRF, case record form; TEAM, data cleaning team
Inconsistency among past/current habits and details of smoking or alcohol drinking ?
No YesData entry error ?
(checked with original CRF)
Yes (Data entry error)
Edited data
No
W I T H I N each dataset
MERGE DATASETS
SMOKING | ALCOHOL DRINKING
Discussed with TEAM(considered every details of
habits, such as, amount, frequency, start/quit age)
Reasonable overtime sequence?
No. For example, current smokers became never smokers Yes
Discussed with TEAM(considered every details of
habits, such as, amount, frequency, start/quit age)
CLEANED DATA
Yes (Data entry error)
Edited data
No
Data entry error ? (checked with original CRF)
CLEANED DATA
SMOKING 2/4 / ALCOHOL 2/4 was never smokers / never drinkers ?
No Yes
Replaced 2/2 & 2/3 as “never”
CLEANED DATA & CARRIED BACKWARD
F i l l i n g m i s s i n g d a t a
Attawood Lertpimonchai Methodology / 96
Figure 3.10 Workflow of cleaning processes for height cm, centimeter; CRF, case record form; obs., observes; SD, standard deviation; TEAM, data cleaning team
Within mean ± 4 S.D. ?
Yes No
Replaced with missing
From 1 dataset
How many ‘height’ present ?
From 2 datasets From 3 datasets
Average height
Difference between Max and Min > 10 cm ?
Yes No
Average height
Drop the outlier, then average height from 2 obs.
Data entry error ? (checked with original CRF)
Yes (Data entry error)
Edited data
No
W I T H I N each dataset
CLEANED DATA
CLEANED DATA
CLEANED DATA
CLEANED DATA
MERGE DATASETS
Data entry error ? (checked with original CRF)
Yes (Data entry error) No
Edited data
HEIGHT
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 97
Figure 3.11 Workflow of cleaning processes for weight, waist and hip circumstance cm, centimeter; CRF, case record form; kg, kilograms; SD, standard deviation; TEAM, data cleaning team
Within mean ± 4 SD ?
Yes NoData entry error ?
(checked with original CRF)
Yes (Data entry error)
Edited data
No
W I T H I N each dataset
MERGE DATASETS
WEIGHT | WAIST | HIP
Discussed with TEAM(comparing with height, weight, waist and hip)
Identifying impossible “Change” in body measurement over-time
Was there ∆weight ≥ 10 kg or ∆waist ≥ 15 cm or ∆hip ≥ 10 cm?
Yes No
Discussed with TEAM(comparing with height, weight, waist and hip)
Data was POSSIBLE?
YesNo
Replace the inconsistency with missing value
CLEANED DATA
Yes (Data entry error)
Edited data
No
Data entry error ? (checked with original CRF)
CLEANED DATA
CLEANED DATA
Attawood Lertpimonchai Methodology / 98
Figure 3.12 Workflow of cleaning processes for blood pressure BP, blood pressure; CRF, case record form; DBP, diastolic blood pressure; obs., observes; SBP, systolic blood pressure; SD, standard deviation; TEAM, data cleaning team
Ø SBP & DBP, simultaneously presented ?
Ø Within mean ± 4 SD ?
Ø Systolic > Diastolic ?
Ø Pulse pressure > 10 mmHg
Yes No
1 measurement
How many times of measurement ?
2 measurements 3 measurements
Average BP
Difference among SBP > 10 mmHg ?
Yes No
Average BP
Drop the SBP & DBP which SBP was the outlier,
then average SBP & DBP from 2 obs.
Data entry error ? (checked with original CRF)
Yes (Data entry error)
Edited data
No
CLEANED DATA
CLEANED DATA
CLEANED DATA
CLEANED DATA
Data entry error ? (checked with original CRF)
Yes (Data entry error) No
Edited data
BLOOD PRESSURE
Discussed with TEAM
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 99
Figure 3.13 Workflow of cleaning processes for laboratory results HDL, high-density lipoprotein; LDL, low-density lipoprotein; SD, standard deviation; TEAM, data cleaning team
Within mean ± 4 SD ?
Yes No
No
Replaced with missing value
Yes (Possible value)
L A B O R A T O R Y R E S U L T S
CLEANED DATA
CLEANED DATA
Discussed with TEAM
Data was POSSIBLE?
CLEANED DATA
Serum glucose, cholesterol, HDL LDL,
Triglyceride, uric acid, creatinine
Attawood Lertpimonchai Methodology / 100
Figure 3.14 Outline of statistical analysis
CKD, chronic kidney disease; DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension; NSAIDs, nonsteroidal anti-inflammatory drugs
DM Model:
- Periodontitis
- Age
- Gender
- Marital status
- Income
- Education
- Smoking
- Alcohol drinking
- Exercise
- HT
- DLP
- Family history o DM
Statistical analysis
Pathway A:
Periodontitis(Exposure)
CKD(Outcome)
DM(Mediator)
Pathway B:
DM(Exposure)
CKD(Outcome)
Periodontitis(Mediator)
CKD Model:
- Periodontitis
- DM
- Marital status
- Income
- Education
- Smoking
- Alcohol drinking
- NSAIDs use
- Exercise
- HT
- DLP
- Uric acid
Forward Stepwise
Final model: Mediation analysis
Partial disjunctive clause
Modified fully disjunctive clause
Periodontitis Model:
- DM
- Age
- Gender
- Marital status
- Income
- Education
- Smoking
- Alcohol drinking
- Exercise
- HT
- DLP
CKD Model:
- Periodontitis
- DM
- Marital status
- Income
- Education
- Smoking
- Alcohol drinking
- NSAIDs use
- Exercise
- HT
- DLP
- Uric acid
Forward Stepwise
Final model: Mediation analysis
Partial disjunctive clause
Modified fully disjunctive clause
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 101
CHAPTER IV
RESULTS
4.1 Characteristic of subjects Numbers of registered subjects for EGAT 2/2, 2/3 and 2/4 were of 2,686,
2,288 and 2,037, respectively. A total number of subjects who participated at least one
survey during 2003 to 2013 was 2,795. The patterns that subjects participated with the
EGAT survey were shown in Table 4.1. Among them, 1,821 (65%) subjects were
completely attended, with 14% and 21% of subjects absented from 1 and 2 times,
respectively.
From total, 126 subjects were excluded because they had CKD at baseline
(EGAT 2/2). Among excluded cases, 122 subjects were diagnosed based on actual data,
while 4 subjects, whose kidney functions at baseline were missing, were excluded from
the average imputed data. Among the non-CKD cohort, 6 subjects were further excluded
from systemic conditions that could not obtain the periodontal examination. 15 subjects
were fully edentulous and 13 subjects refused to attend all periodontal examinations. As
results, 2,635 participants were finally included for analysis (Figure 4.1).
Characteristics of excluded cases are shown in Table 4.2.
The baseline characteristics are shown in Table 4.3. With mean age of
47.7 ± 4.9 years, approximately three quarters were males and a half of them had
experiences of smoking and alcohol drinking. Prevalence of DM, HT and DLP at
baseline were 7.69%, 27.25 % and 68.54%, respectively. With the CDC/AAP
periodontitis definition, the prevalence of moderate and severe periodontitis
approximated 50% and 30%, respectively. In addition, each subject, had about 2% of
total sites with PPD ≥ 6 mm, and 10-14% of proximal sites with CAL ≥ 5 mm, by
average.
A total number of new CKD was respectively 167 and 105 for EGAT 2/3
and 2/4 survey, with a total CKD subjects of 272. As a result, a cumulative incidence of
CKD was 1.03 cases per 100 persons per year (95% CI: 0.91, 1.16). Moreover, CKD
Attawood Lertpimonchai Results / 102
incidence increased with the severity of periodontitis, in which the cumulative incidence
of CKD among normal/mild periodontitis, moderate periodontitis and severe
periodontitis were 0.72, 0.96 and 1.39 cases per 100 persons per year, respectively.
4.2 Missing data and imputation results Among twenty variables, age, gender and family history were completed
and considered/assigned as “regular” variables. After intensively cleaned and filled in
missing data with the forward/backward carried over methods, missing data was still
presented, which ranged from 0.38% to 18.30% with a median of 16.25% (Table 4.4).
The most frequent missing variables was periodontal status (18.30%) whereas the least
frequent missing variable was height (0.38%). When explored the sources of missing
data (Table 4.5), most of them were absent because of loss to follow-up which called
the whole-wave missing data. On the other hand, the within-wave missing data (i.e., the
missing data of participated subjects) was typically less than 5%. Distribution of these
missing values was explored and the results showed arbitrary patterns, thus, MAR was
assumed. Twenty imputations with MICE were completely generated to fill in all
missing observes, which were set based on the percentage of missing data with the
maximum of 18.30%. Then, the FMI from final GSEM models (Table 4.6) were
considered to confirm that our twenty imputations were adequate efficiency and power.
The highest FMI was about 23% for income in the CKD model, although we set at 20
imputations, it should be sufficient with the relative efficiency of about 99%. The
characteristics of each variable between actual and imputed data were compared which
were not much different (Table 4.7).
4.3 Pathway A: Periodontitis → DM → CKD Potential causal relationships between periodontitis, DM and CKD were
assessed following a causal diagram described in Figure 3.1. Mediation equation
(periodontitis → DM) and outcome equation (periodontitis + DM → CKD) were
constructed as follows:
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 103
4.3.1 Dealing with periodontitis and obesity variables
The F-test of simple logistic regression for the outcome model was
performed to select which classification/definition of periodontitis should be the most
suitable in term of explanation the outcome. The F-tests and coefficients of periodontitis
defined by various definitions were shown in Table 4.8. The direction of this causative
association was consistent among all classifications, in which the disease extent of
proximal sites with CAL ≥ 5 mm (Periodontitis F) was achieved the highest F-test when
compared with other definitions. Thus, it was selected for the representative of
periodontal status in further analysis. To compare with other studies, the standard
periodontitis definition defined by CDC/AAP was also alternatively selected.
Same procedures had been performed for BMI and WHR to represent
obesity. Obese group defined with WHR was finally chosen because it yielded the
largest F-test (Table 4.8).
4.3.2 Selection variables based on conventional method
4.3.2.1 Mediation model
This mediation model was performed to estimate the effect of
periodontitis on DM. The GSEM with logit link was used within the assumption that
periodontitis was the cause of DM. Periodontitis and other DM risk factors (i.e., age,
gender, income, education, marital status, exercise, smoking, alcohol, obesity, HT, DLP
and family history of DM) were considered in the univariate analysis of GSEM (see
Table 4.9) indicating p-value for all factors were < 0.1, thus should be included all in
the next step. A forward selection was performed and finally kept only periodontitis and
6 co-variables including age, education level, obesity, HT, DLP and family history of
DM.
4.3.2.2 Outcome model
A univariate GSEM with logit link was used to assess whether
periodontitis and DM along with other co-variables (i.e., marital status, income,
education, smoking, alcohol, NSAIDs use, exercise, obesity, HT, DLP, uric acid) were
associated with CKD, see Table 4.10. Almost all, including periodontitis (exposure),
DM (mediator), marital status, income, smoking, alcohol, NSAIDs use, obesity, HT,
Attawood Lertpimonchai Results / 104
DLP and serum uric acid were significantly associated from univariate analysis. A
stepwise forward selection was performed to explore the final model which strict
contained periodontitis and DM. The final model (Table 4.11) suggested periodontitis,
DM, HT, income and serum uric acid were associated with incident CKD,
independently.
The multivariate GSEMs were simultaneously constructed for
both mediation and outcome models, see Table 4.11. Results indicated the significant
independent effect of periodontitis on DM with adjusted ORs of 1.011
(95% CI: 1.006, 1.015). In addition, the coefficients of the effects of periodontitis and
DM on CKD after controlling for other co-variables were 0.010 (95% CI: 0.005, 0.015)
and 0.689 (95% CI: 0.385, 0.994), respectively.
For CDC/AAP periodontitis definitions, the univariate analysis
from GSEM showed that only severe periodontitis had significant risk effect on DM
(Table 4.9). However, it was not remained significant in the multivariate analysis
(Table 4.12). In the CKD model, the risk effect of periodontitis on CKD incidence was
also not statistically significant in the multivariate GSEM with forward stepwise
(Table 4.12).
4.3.2.3 Estimations of mediation effects
Mediation effects were estimated by products of coefficients of
mediation and outcome models, see Figure 4.2. A 1000-replication bootstrap suggested
the significant ORs from mediation (indirect) effect (Periodontitis → DM → CKD) and
direct effect (Periodontitis → CKD) of 1.007 (95% CI: 1.003, 1.013), and 1.010
(95% CI: 1.005, 1.015), respectively. The percentage of periodontitis effect contributed
through DM mediator was 42.38% (Table 4.13). From these, it could be interpreted that
every one percent increasing of periodontitis extent would increase risk of CKD through
DM around 0.7%, and it would directly increase risk of CKD about 1.0%. Suppose
subjects who had 50% of proximal sites with severe periodontitis, these subjects were
about 50% higher risk to directly develop CKD, and about 35% higher risk to develop
CKD through DM when compared with normal periodontium.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 105
4.3.3 Selection of co-variables based on disjunctive clause criteria
The partial disjunctive and modified fully disjunctive clause criteria were
performed to select co-variables in DM and CKD final models using GSEM.
Coefficients were shown and compared with the conventional forward stepwise method
in Table 4.14 and Table 4.15. With the partial disjunctive clause criteria, the income and
uric acid, which were selected from the stepwise of CKD, were added to the DM model.
In addition, education, obesity, DLP and family history of DM were added to the CKD
model. While, the modified fully disjunctive clause criteria included periodontitis and
all co-variables in both models. Comparing among 3 methods, the effect sizes of
periodontitis on CKD were very much similar, i.e., 0.010 (95% CI: 0.005, 0.015), 0.011
(95% CI: 0.006, 0.016) and 0.012 (95% CI: 0.006, 0.018) for conventional, partial and
modified fully disjunctive clause criteria, respectively. Only the uric acid was additional
significant in the DM model with coefficients of -0.190 (95% CI: -0.266, -0.155) and -0.229 (95% CI: -0.313, -0.144) from the partial and modified fully disjunctive clause
criteria, respectively. While, the direction and magnitude of other co-variables were
very not much different.
4.4 Pathway B: DM → Periodontitis → CKD
The inverse association between periodontitis and DM was also assessed.
Here, DM was considered as the exposure, and periodontitis was considered as the
mediator. The periodontitis equation was additionally constructed with adjustment for
co-variables. Whereas, the outcome model was the same as the pathway A.
4.4.1 Selection variables based on conventional method
4.4.1.1 Mediation model
The mediator model was performed to estimate the effect of DM
on periodontitis. Instead of the logit link, The GSEM was constructed with the Gaussian
family and identity link, because the disease extent, which was the continuous data, was
used representing periodontitis
Attawood Lertpimonchai Results / 106
Given periodontitis was caused by DM, it was also taking to
account simultaneously with other risk factors, including age, gender, income,
education, marital status, exercise, smoking, alcohol, obesity, HT, and DLP. Results
from the univariate analysis is shown in Table 4.16, indicating all factors significantly
related with periodontitis. Furthermore, the multivariate model from forward stepwise
selection implied that DM, age, gender, education exercise and smoking were
significant (Table 4.17). Here, it was interpreted that with identical other co-variables,
subjects with DM had the percentage of proximal sites with CAL ≥ 5 mm about 4.8%
significantly higher than non-DM.
4.4.1.2 Estimations of mediation effects
The average causal mediation effect (ACME) were estimated
using product-of-coefficient method (Figure 4.3). A 1000-replication bootstrap yielded
significant mediation (indirect) effect (DM → Periodontitis → CKD) and direct effect
(DM → CKD) of 0.048 (95% CI: 0.021, 0.096), 0.689 (95% CI: 0.366, 0.982),
respectively. The ORs of having CKD in subjects with DM was 2.09 (95% CI: 1.52, 2.83) comparing with non-DM. Within this ORs, the percentage of DM
effect contributed through periodontitis mediator was 6.53% (Table 4.18).
4.4.2 Selection of co-variables based on disjunctive clause criteria
The partial disjunctive and modified fully disjunctive clause criteria were
performed also in the periodontitis model. The effect size and pattern of existed co-
variables in stepwise were approximate in both disjunctive clause criteria (Table 4.19).
Low income, middle income and uric acid were also considered as significant factors in
the partial disjunctive clause with the effect size of 2.36 (95% CI: 0.42, 4.31), 1.51 (95% CI: 0.05, 2.26) and -0.65 (95% CI: -1.16, -0.14), respectively. Moreover, the
modified fully disjunctive clause criteria included NSAIDs as the protective factors of
periodontitis with coefficients of -1.73 (95% CI: -3.31, -0.16).
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 107
4.5 Assumption checking
The sequential ignorability assumption was tested in both pathways.
Because the post-estimation command of GSEM for residuals estimation was not
compatible with the imputed data, only actual data was used to test this assumption.
Pearson's coefficients of correlation between residuals of DM and CKD model, as well
as, periodontitis and CKD model were -0.0146 and -0.0192, respectively, which
indicated very low correlation between the two residuals. This could be implied that the
two residuals were independent and thus ignorability assumption should be held.
Attawood Lertpimonchai Results / 108
Table 4.1 Pattern of participation
Pattern EGAT 2/2
(2,686)
EGAT 2/3
(2,288)
EGAT 2/4
(2,037)
Total
(2795) %
A 1 0 0 352 12.59
B 0 1 0 25 0.90
C 0 0 1 23 0.82
A-C (1 time of participation) 400 14.31
D 1 1 0 381 13.63
E 1 0 1 132 4.72
F 0 1 1 61 2.19
D-F (2 times of participation) 574 20.54
G (Completed participation) 1,821 65.15
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 109
Table 4.2 Baseline characteristics of excluded cases
Characteristics
EGAT 2/2
n = 2,532
All excluded
cases
n = 160
CKD at
baseline
n = 126
Periodontal
reasons
n = 34
Age (year) 47.65 ± 4.85 51.10 ± 5.21 51.07 ± 5.22 51.25 ± 5.27
Gender
Male
Female
1848 (73.0)
684 (27.0)
121 (75.62)
39 (24.38)
95 (75.40)
31 (24.60)
26 (76.47)
8 (23.53)
Marital status
Single
Married
Divorce/Widows
238 (9.49)
2081 (82.97)
189 (7.54)
11 (7.19)
128 (83.66)
14 (9.15)
8 (6.50)
105 (85.37)
10 (8.13)
3 (10.00)
23 (76.67)
4 (13.33)
Income (Baht/month)
< 20,000
20,000 – 49,999
≥ 50,000
312 (12.47)
1318 (52.68)
872 (34.85)
18 (11.76)
85 (55.56)
50 (32.68)
13 (10.57)
70 (56.91)
40 (32.52)
5 (16.67)
15 (50.00)
10 (33.33)
Education
≤ Secondary school
Vocational/Diploma
≥ Bachelor’s degree
688 (27.39)
825 (32.84)
999 (39.77)
64 (41.03)
45 (28.84)
47 (30.13)
51 (40.80)
36 (28.80)
38 (30.40)
13 (41.94)
9 (29.03)
9 (29.03)
Smoking
Never smokers
Quit smokers
Current smokers
1340 (53.22)
622 (24.70)
556 (22.08)
83 (52.87)
43 (27.39)
31 (19.74)
70 (56.45)
32 (25.81)
22 (17.74)
13 (39.40)
11 (33.33)
9 (27.27)
Alcohol
Never drinkers
Quit drinkers
Current drinkers
1182 (47.04)
211 (8.40)
1120 (44.57)
73 (46.79)
20 (12.82)
63 (40.39)
60 (48.39)
17 (13.71)
47 (37.90)
13 (40.62)
3 (9.38)
16 (50.00)
Attawood Lertpimonchai Results / 110
Table 4.2 Baseline characteristics of excluded cases (cont.)
Characteristics
EGAT 2/2
n = 2,532
All excluded
cases
n = 160
CKD at
baseline
n = 126
Periodontal
reasons
n = 34
Exercise (times/week)
None
1 - 2
≥ 3
728 (29.00)
670 (26.69)
1112 (44.30)
42 (27.27)
34 (22.08)
78 (50.65)
30 (24.39)
29 (23.58)
64 (52.03)
12 (38.71)
5 (16.13)
14 (45.16)
NSAIDs use
Yes
No
246 (9.80)
2264 (90.20)
20 (12.99)
134 (87.01)
17 (13.82)
106 (86.18)
3 (9.68)
28 (90.32)
Height (cm) 163.61 ± 7.67 163.58 ± 7.73 163.39 ± 7.90 164.28 ± 7.12
Weight (kg) 65.88 ± 11.48 68.28 ± 11.85 68.94 ± 12.21 65.64 ± 9.96
Waist circumference (cm) 85.97 ± 9.87 89.23 ± 9.79 89.67 ± 9.77 87.5 ± 9.85
Hip circumference (cm) 96.49 ± 6.57 97.46 ± 6.75 97.86 ± 6.83 95.87 ± 6.25
BMI (kg/m2) 24.56 ± 3.61 25.45 ± 3.74 25.78 ± 3.84 24.13 ± 3.03
Waist to hip ratio 0.89 ± 0.07 0.91 ± 0.06 0.91 ± 0.06 0.91 ± 0.06
Diabetes Mellitus
Yes
No
194 (7.69)
2329 (92.31)
24 (15.58)
130 (84.42)
16 (13.01)
107 (86.99)
8 (25.81)
23 (74.19)
Hypertension
Yes
No
689 (27.52)
1815 (72.48)
66 (43.14)
87 (56.86)
55 (44.72)
68 (55.28)
11 (36.67)
19 (63.33)
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 111
Table 4.2 Baseline characteristics of excluded cases (cont.)
Characteristics
EGAT 2/2
n = 2,532
All excluded
cases
n = 160
CKD at
baseline
n = 126
Periodontal
reasons
n = 34
Family history of DM
Yes
No
929 (36.69)
1603 (63.31)
54 (33.75)
106 (66.25)
44 (34.92)
82 (65.08)
10 (29.41)
24 (70.59)
Total cholesterol (mg/dl) 233.79 ± 41.99 249.04 ± 53.50 249.93 ± 52.05 245.52 ± 59.69
HDL (mg/dl) 53.15 ±14.42 51.04 ± 14.43 51.57 ± 14.98 48.94 ± 12.01
LDL (mg/dl) 151.72 ± 38.74 164.14 ± 50.34 165.19 ± 47.71 159.52 ± 61.45
Triglyceride (mg/dl)** 126 (27, 1362) 140 (35, 1147) 140 (35, 892) 151 (71, 1147)
Creatinine (mg/dl) 1.01 ± 0.17 1.44 ± 0.91 1.55 ± 0.99 1.02 ± 0.14
Uric acid (mg/dl) 5.60 ± 1.47 6.26 ± 1.71 6.40 ± 1.69 5.68 ± 1.69
eGFR (ml/min per 1.73m2) 83.37 ± 12.73 58.39 ± 14.94 52.82 ± 9.32 80.34 ± 12.54
Periodontitis (CDC/AAP)
No / Mild Periodontitis
Moderate Periodontitis
Severe Periodontitis
429 (17.24)
1268 (50.94)
792 (31.82)
17 (15.18)
66 (58.93)
29 (25.89)
N/A
% sites with PPD ≥ 4 mm** 4.17 (0, 93.75) 4.42 (0, 56.25)
% sites with PPD ≥ 6 mm** 0 (0, 63.64) 0 (0, 22.92)
% sites with PPD ≥ 4 mm & CAL ≥ 5 mm** 0.88 (0, 91.67) 0.76 (0, 55.21)
% proximal sites with CAL ≥ 3 mm** 41.67 (0, 100) 43.75 (0, 100)
% proximal sites with CAL ≥ 5 mm**
2.68 (0, 100) 2.66 (0, 80.56)
Values are mean ± SD for continuous data, and frequency (%) for categorical data, except where specified.
Total numbers of subjects of each variable may be different depended on missing data.
** Median (range)
BMI, body mass index; CAL, clinical attachment level; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; FM-DM, family history of diabetes; FM-HT, family history of hypertension; HDL, high-density lipoprotein; LDL, low-density lipoprotein; N/A, not available; PPD, periodontal pocket depth
Attawood Lertpimonchai Results / 112
Table 4.3 Baseline characteristics
Characteristics EGAT 2/2 n = 2,532
EGAT 2/3 n = 2,183
EGAT 2/4 n = 1,948
Age (year) 47.65 ± 4.85 52.26 ± 4.59 56.88 ± 4.53
Gender
Male
Female
1848 (73.0)
684 (27.0)
1567 (71.8)
616 (28.2)
1368 (70.2)
580 (29.8)
Marital status
Single
Married
Divorce/Widows
238 (9.49)
2081 (82.97)
189 (7.54)
178 (8.22)
1801 (83.15)
187 (8.63)
144 (7.46)
1586 (82.18)
200 (10.36)
Income (Baht/month)
< 20,000
20,000 – 49,999
≥ 50,000
312 (12.47)
1318 (52.68)
872 (34.85)
135 (6.38)
634 (29.96)
1347 (63.66)
189 (9.82)
248 (12.88)
1488 (77.30)
Education
≤ Secondary school
Vocational/Diploma
≥ Bachelor’s degree
688 (27.39)
825 (32.84)
999 (39.77)
488 (22.49)
736 (33.92)
946 (43.59)
357 (18.50)
633 (32.80)
940 (48.70)
Smoking
Never smokers
Quit smokers
Current smokers
1340 (53.22)
622 (24.70)
556 (22.08)
1138 (52.47)
586 (27.02)
445 (20.51)
1047 (54.22)
622 (32.21)
262 (13.57)
Alcohol
Never drinkers
Quit drinkers
Current drinkers
1182 (47.04)
211 (8.40)
1120 (44.57)
782 (36.09)
312 (14.40)
1073 (49.51)
530 (27.45)
433 (22.42)
968 (50.13)
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 113
Table 4.3 Baseline characteristics (cont.)
Characteristics EGAT 2/2 n = 2,532
EGAT 2/3 n = 2,183
EGAT 2/4 n = 1,948
Exercise (times/week)
None
1 - 2
≥ 3
728 (29.00)
670 (26.69)
1112 (44.30)
1055 (48.80)
351 (16.23)
756 (34.97)
622 (32.24)
272 (14.10)
1035 (53.66)
NSAIDs use
Yes
No
246 (9.80)
2264 (90.20)
312 (14.40)
1855 (85.60)
170 (8.81)
1760 (91.19)
Height (cm) 163.61 ± 7.67 163.55 ± 7.71 163.49 ± 7.82
Weight (kg) 65.88 ± 11.48 66.24 ± 11.23 66.93 ± 11.88
Waist circumference (cm) 85.97 ± 9.87 87.14 ± 9.48 88.23 ± 10.13
Hip circumference (cm) 96.49 ± 6.57 95.72 ± 6.50 98.00 ± 6.94
BMI (kg/m2) 24.56 ± 3.61 24.72 ± 3.59 25.00 ± 3.82
Waist to hip ratio 0.89 ± 0.07 0.91 ± 0.07 0.90 ± 0.007
Central obesity
Yes
No
1283 (51.40)
1213 (48.60)
1434 (66.39)
726 (33.61)
1156 (60.05)
769 (39.95)
Diabetes Mellitus
Yes
No
194 (7.69)
2329 (92.31)
254 (11.71)
1915 (88.29)
305 (15.67)
1641 (84.33)
Hypertension
Yes
No
689 (27.52)
1815 (72.48)
827 (38.20)
1338 (61.80)
1049 (54.35)
881 (45.65)
Dyslipidemia
Yes
No
1723 (68.54)
791 (31.46)
1640 (75.75)
525 (24.25)
1468 (75.63)
473 (24.37)
Attawood Lertpimonchai Results / 114
Table 4.3 Baseline characteristics (cont.)
Characteristics EGAT 2/2 n = 2,532
EGAT 2/3 n = 2,183
EGAT 2/4 n = 1,948
Family history of DM
Yes
No
929 (36.69)
1603 (63.31)
853 (39.07)
1330 (60.93)
780 (40.04)
1168 (59.96)
Total cholesterol (mg/dl) 233.79 ± 41.99 231.33 ± 42.09 217.73 ± 43.89
HDL (mg/dl) 53.15 ±14.42 51.03 ± 12.41 57.53 ± 15.55
LDL (mg/dl) 151.72 ± 38.74 149.94 ± 30.11 144.86 ± 40.02
Triglyceride (mg/dl)** 126 (27, 1362) 128 (31, 1133) 121.5 (37, 1280)
Creatinine (mg/dl) 1.01 ± 0.17 1.02 ± 0.19 0.98 ± 0.22
Uric acid (mg/dl) 5.60 ± 1.47 5.83 ± 1.45 6.06 ± 1.45
eGFR (ml/min per 1.73m2) 83.37 ± 12.73 80.13 ± 13.60 80.45 ± 13.53
Periodontitis (CDC/AAP)
Non / Mild Periodontitis
Moderate Periodontitis
Severe Periodontitis
429 (17.24)
1268 (50.94)
792 (31.82)
227 (10.97)
1092 (52.78)
750 (36.25)
272 (14.31)
1002 (52.74)
626 (32.95)
% sites with PPD ≥ 4 mm** 4.17 (0, 93.75) 4.17 (0, 88.67) 3.97 (0, 95.83)
% sites with PPD ≥ 6 mm** 0 (0, 63.64) 0 (0, 62.67) 0 (0, 75.00)
% sites with PPD ≥ 4 mm
& CAL ≥ 5 mm** 0.88 (0, 91.67) 1.45 (0, 88.67) 1.19 (0, 95.83)
% proximal sites with CAL
≥ 3 mm** 41.67 (0, 100) 56.00 (0, 100) 55.10 (0, 100)
% proximal sites with CAL
≥ 5 mm** 2.68 (0, 100) 5.36 (0, 100) 4.35 (0, 100)
Values are mean ± SD for continuous data, and frequency (%) for categorical data, except where specified.
Total numbers of subjects of each variable may be different depended on missing data.
** Median (range)
BMI, body mass index; CAL, clinical attachment level; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; FM-DM, family history of diabetes; FM-HT, family history of hypertension; HDL, high-density lipoprotein; LDL, low-density lipoprotein; N/A, not available; PPD, periodontal pocket depth
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 115
Table 4.4 Numbers of missing data
Variables Number of Observed
% missing data
Marital status 6,619 16.27
Education 6,784 14.18
Income 6,543 17.23
Smoking 6,731 14.85
Alcohol 6,678 15.52
Exercise 6,601 16.50
NSAIDs use 6,607 16.42
Weight 6,598 16.53
Height 7,875 0.38
Waist 6,583 16.72
Hip 6,582 16.74
Blood Pressure 6,594 16.58
Glucose 6,637 16.04
Cholesterol 6,639 16.02
HDL 6,639 16.02
LDL 6,547 17.18
Triglyceride 6,639 16.02
Uric acid 6,639 16.02
Creatinine 6,623 16.22
Periodontal status 6,458 18.30
HDL, high-density lipoprotein; LDL, low-density lipoprotein; NSAIDs, nonsteroidal anti-inflammatory drugs
Attaw
ood Lertpim
onchai
Results / 116
Table 4.5 Within and whole wave missing data
HDL, high-density lipoprotein; LDL, low-density lipoprotein; NSAIDs, nonsteroidal anti-inflammatory drugs
Table 4.5 Within and whole wave missing data
Variables
PERIO-CKD cohort (N = 2,635)
EGAT 2/2 (2,532) EGAT 2/3 (2,183) EGAT 2/4 (1,948)
Within wave Whole wave (Total) Within wave Whole wave (Total) Within wave Whole wave (Total)
Observe Missing Observe Missing % Missing Observe Missing Observe Missing % Missing Observe Missing Observe Missing % Missing
Marital status 2,508 24 2,516 119 4.52 2,166 17 2173 462 17.53 1,930 18 1,930 705 26.76
Education 2,512 20 2,540 95 3.61 2,170 13 2,293 342 12.98 1,930 18 1,951 684 25.96
Income 2,502 30 2,502 133 5.05 2,116 67 2,116 519 19.70 1,925 23 1,925 710 26.94
Smoking 2,518 14 2,568 67 2.54 2,169 14 2,232 403 15.29 1,931 17 1,931 704 26.72
Alcohol 2,513 19 2,545 90 3.42 2,167 16 2,202 433 16.43 1,931 17 1,931 704 26.72
Exercise 2,510 22 2,510 125 4.74 2,162 21 2,162 473 17.95 1,929 19 1,929 706 26.79
NSAIDs use 2,510 22 2,510 125 4.74 2,167 16 2,167 468 17.76 1,930 18 1,930 705 26.76
Weight 2,505 27 2,505 130 4.93 2,165 18 2,165 470 17.84 1,928 20 1,928 707 26.83
Height 2,527 5 2,625 10 0.38 2,181 2 2,625 10 0.38 1,945 3 2,625 10 0.38
Waist 2,496 36 2,496 139 5.28 2,161 22 2,161 474 17.99 1,926 22 1,926 709 26.91
Hip 2,496 36 2,496 139 5.28 2,161 22 2,161 474 17.99 1,925 23 1,925 710 26.94
Blood Pressure 2,504 28 2,504 131 4.97 2,163 20 2,163 472 17.91 1,927 21 1,927 708 26.87
Glucose 2,523 9 2,523 112 4.25 2,168 15 2,168 467 17.72 1,946 2 1,946 689 26.15
Cholesterol 2,524 8 2,524 111 4.21 2,169 14 2,169 466 17.69 1,946 2 1,946 689 26.15
HDL 2,524 8 2,524 111 4.21 2,169 14 2,169 466 17.69 1,946 2 1,946 689 26.15
LDL 2,432 100 2,432 203 7.70 2,169 14 2,169 466 17.69 1,946 2 1,946 689 26.15
Triglyceride 2,524 8 2,524 111 4.21 2,169 14 2,169 466 17.69 1,946 2 1,946 689 26.15
Uric acid 2,524 8 2,524 111 4.21 2,169 14 2,169 466 17.69 1,946 2 1,946 689 26.15
Creatinine 2,511 21 2,511 124 4.71 2,168 15 2,168 467 17.72 1,944 4 1,944 691 26.22
Periodontal status 2,489 43 2,489 146 5.54 2,069 114 2,069 566 21.48 1,900 48 1,944 691 26.22
LDL, Low-density lipoprotein; HDL, High-density lipoprotein; NSAIDs, Nonsteroidal anti-inflammatory drugs
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 117
Table 4.6 RVI, FMI and Relative efficiency in GSEM final models
RVI FMI Relative efficiency
DM Model
Periodontitis 0.020091 0.019735 0.999014 Age 0.019497 0.019162 0.999043 HT 0.168423 0.146011 0.992752 DLP 0.052441 0.050076 0.997502 Obesity 0.150097 0.132062 0.99344 Family history of DM 0.011175 0.011064 0.999447 Education
≤ High school 0.027203 0.026555 0.998674 Vocation/Diploma
0.037777
0.036536
0.998177
Periodontitis Model
DM 0.009181 0.009106 0.999545 Age 0.020628 0.020254 0.998988 Gender: male 0.008426 0.008363 0.999582 Smoking
Quit smokers 0.057689 0.054838 0.997266 Current smokers 0.020825 0.020443 0.998979
Education ≤ High school 0.050939 0.048705 0.997571 Vocation/Diploma 0.038244 0.036973 0.998155
Exercise (times/week) 1 - 2 0.074013 0.069378 0.996543 ≥ 3 0.09528 0.087718 0.995633
CKD Model
Periodontitis 0.031003 0.030163 0.998494 DM 0.018715 0.018406 0.999081 Uric acid 0.085909 0.079719 0.99603 HT 0.148691 0.130975 0.993494 Income (Baht/month)
< 20,000 0.320169 0.247168 0.987792 20,000 - 49,999 0.306239 0.238835 0.988199
CKD, chronic kidney disease; DLP dyslipidemia; DM, diabetes mellitus; FMI, fraction of missing information; GSEM, generalized structural equation model; HT, hypertension; RVI, relative variance increase
Attawood Lertpimonchai Results / 118
Table 4.7 Comparison of characteristic between actual and imputed dataset
Characteristics EGAT 2/2 n = 2,532
Imputed data
Proportion / Mean SE
Marital status
Single
Married
Divorce/Widows
238 (9.49)
2081 (82.97)
189 (7.54)
9.50
82.80
7.70
0.006
0.007
0.005
Income (Baht/month)
< 20,000
20,000 – 49,999
≥ 50,000
312 (12.47)
1318 (52.68)
872 (34.85)
12.54
52.88
34.58
0.007
0.010
0.009
Education
≤ Secondary school
Vocational/Diploma
≥ Bachelor’s degree
688 (27.39)
825 (32.84)
999 (39.77)
27.43
33.14
39.43
0.009
0.009
0.010
Smoking
Never smokers
Quit smokers
Current smokers
1340 (53.22)
622 (24.70)
556 (22.08)
52.93
25.05
22.02
0.010
0.008
0.008
Alcohol
Never drinkers
Quit drinkers
Current drinkers
1182 (47.04)
211 (8.40)
1120 (44.57)
46.42
9.95
43.63
0.010
0.006
0.010
Exercise (times/week)
None
1 - 2
≥ 3
728 (29.00)
670 (26.69)
1112 (44.30)
28.97
26.68
44.35
0.009
0.009
0.010
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 119
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/2 n = 2,532
Imputed data
Proportion / Mean SE
NSAIDs use
Yes
No
246 (9.80)
2264 (90.20)
9.85
90.15
0.006
0.006
Height (cm) 163.61 ± 7.67 163.59 0.149
Weight (kg) 65.88 ± 11.48 65.94 0.224
Waist circumference (cm) 85.97 ± 9.87 86.04 0.194
Hip circumference (cm) 96.49 ± 6.57 96.49 0.129
BMI (kg/m2) 24.56 ± 3.61 24.58 0.070
Waist to hip ratio 0.89 ± 0.07 0.89 0.001
Central obesity
Yes
No
1283 (51.40)
1213 (48.60)
51.78
48.22
0.010
0.010
Diabetes Mellitus
Yes
No
194 (7.69)
2329 (92.31)
7.75
92.25
0.005
0.005
Hypertension
Yes
No
689 (27.52)
1815 (72.48)
27.57
72.43
0.008
0.008
Dyslipidemia
Yes
No
1723 (68.54)
791 (31.46)
68.82
31.18
0.009
0.009
Attawood Lertpimonchai Results / 120
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/2 n = 2,532
Imputed data
Proportion / Mean SE
Total cholesterol (mg/dl) 233.79 ± 41.99 233.85 0.828
HDL (mg/dl) 53.15 ±14.42 53.03 0.281
LDL (mg/dl) 151.72 ± 38.74 151.61 0.797
Triglyceride (mg/dl) 154.47 ± 116.60 155.79 2.297
Creatinine (mg/dl) 1.01 ± 0.17 1.01 0.004
Uric acid (mg/dl) 5.60 ± 1.47 5.61 0.029
eGFR (ml/min per 1.73m2) 83.37 ± 12.73 83.38 0.278
Periodontitis (CDC/AAP)
Non / Mild Periodontitis
Moderate Periodontitis
Severe Periodontitis
429 (17.24)
1268 (50.94)
792 (31.82)
16.53
50.58
32.89
0.007
0.010
0.009
% sites with PPD ≥ 4 mm 10.50 ± 14.89 10.82 0.290
% sites with PPD ≥ 6 mm 1.98 ± 5.31 1.95 0.102
% sites with PPD ≥ 4 mm & CAL ≥ 5 mm
6.13 ± 12.21 6.37 0.239
% proximal sites with CAL ≥ 3 mm
45.88 ± 29.22 46.72 0.564
% proximal sites with CAL ≥ 5 mm 10.17 ± 17.75 10.46 0.348
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 121
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/3 n = 2,183
Imputed data
Proportion / Mean SE
Marital status
Single
Married
Divorce/Widows
178 (8.22)
1801 (83.15)
187 (8.63)
7.92
82.74
9.34
0.005
0.008
0.006
Income (Baht/month)
< 20,000
20,000 – 49,999
≥ 50,000
135 (6.38)
634 (29.96)
1347 (63.66)
7.16
31.17
61.67
0.006
0.010
0.010
Education
≤ Secondary school
Vocational/Diploma
≥ Bachelor’s degree
488 (22.49)
736 (33.92)
946 (43.59)
24.71
33.68
41.61
0.009
0.009
0.010
Smoking
Never smokers
Quit smokers
Current smokers
1138 (52.47)
586 (27.02)
445 (20.51)
49.86
28.56
21.58
0.010
0.009
0.008
Alcohol
Never drinkers
Quit drinkers
Current drinkers
782 (36.09)
312 (14.40)
1073 (49.51)
33.08
19.34
47.58
0.009
0.009
0.010
Exercise (times/week)
None
1 - 2
≥ 3
1055 (48.8)
351 (16.23)
756 (34.97)
48.06
16.26
35.69
0.011
0.008
0.010
Attawood Lertpimonchai Results / 122
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/3 n = 2,183
Imputed data
Proportion / Mean SE
NSAIDs use
Yes
No
312 (14.40)
1855 (85.6)
15.07
84.93
0.008
0.008
Height (cm) 163.55 ± 7.71 163.59 0.149
Weight (kg) 66.24 ± 11.23 66.41 0.227
Waist circumference (cm) 87.14 ± 9.48 87.42 0.193
Hip circumference (cm) 95.72 ± 6.50 95.71 0.132
BMI (kg/m2) 24.72 ± 3.59 24.77 0.072
Waist to hip ratio 0.91 ± 0.07 0.91 0.001
Central obesity
Yes
No
1434 (66.39)
726 (33.61)
67.80
32.20
0.010
0.010
Diabetes Mellitus
Yes
No
254 (11.71)
1915 (88.29)
12.45
87.55
0.006
0.006
Hypertension
Yes
No
827 (38.20)
1338 (61.80)
38.82
61.18
0.010
0.010
Dyslipidemia
Yes
No
1640 (75.75)
525 (24.25)
77.11
22.89
0.009
0.009
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 123
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/3 n = 2,183
Imputed data
Proportion / Mean SE
Total cholesterol (mg/dl) 231.33 ± 42.09 231.29 0.854
HDL (mg/dl) 51.03 ± 12.41 50.68 0.254
LDL (mg/dl) 149.94 ± 30.11 149.68 0.826
Triglyceride (mg/dl) 152.39 ± 100.40 157.58 2.103
Creatinine (mg/dl) 1.02 ± 0.19 1.02 0.005
Uric acid (mg/dl) 5.83 ± 1.45 5.88 0.030
eGFR (ml/min per 1.73m2) 80.13 ± 13.60 79.80 0.346
Periodontitis (CDC/AAP)
Non / Mild Periodontitis
Moderate Periodontitis
Severe Periodontitis
227 (10.97)
1092 (52.78)
750 (36.25)
8.71
45.80
45.49
0.005
0.010
0.010
% sites with PPD ≥ 4 mm 9.95 ± 14.38 12.06 0.289
% sites with PPD ≥ 6 mm 2.07 ± 5.72 2.04 0.104
% sites with PPD ≥ 4 mm
& CAL ≥ 5 mm 6.70 ± 12.38 8.36 0.253
% proximal sites with CAL
≥ 3 mm 56.35 ± 28.47 59.15 0.521
% proximal sites with CAL
≥ 5 mm 14.04 ± 20.46 16.09 0.404
Attawood Lertpimonchai Results / 124
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/4 n = 1,948
Imputed data
Proportion / Mean SE
Marital status
Single
Married
Divorce/Widows
144 (7.46)
1586 (82.18)
200 (10.36)
6.55
80.48
12.97
0.005
0.008
0.007
Income (Baht/month)
< 20,000
20,000 – 49,999
≥ 50,000
189 (9.82)
248 (12.88)
1488 (77.30)
13.51
14.69
71.80
0.009
0.008
0.011
Education
≤ Secondary school
Vocational/Diploma
≥ Bachelor’s degree
357 (18.50)
633 (32.80)
940 (48.70)
22.72
33.93
43.35
0.008
0.010
0.010
Smoking
Never smokers
Quit smokers
Current smokers
1047 (54.22)
622 (32.21)
262 (13.57)
47.43
35.18
17.39
0.010
0.010
0.008
Alcohol
Never drinkers
Quit drinkers
Current drinkers
530 (27.45)
433 (22.42)
968 (50.13)
22.58
30.20
47.22
0.009
0.011
0.011
Exercise (times/week)
None
1 - 2
≥ 3
622 (32.24)
272 (14.10)
1035 (53.66)
32.32
13.09
54.59
0.011
0.008
0.012
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 125
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/4 n = 1,948
Imputed data
Proportion / Mean SE
NSAIDs use
Yes
No
170 (8.81)
1760 (91.19)
9.80
90.20
0.007
0.007
Height (cm) 163.49 ± 7.82 163.59 0.149
Weight (kg) 66.93 ± 11.88 66.96 0.239
Waist circumference (cm) 88.23 ± 10.13 88.56 0.208
Hip circumference (cm) 98.00 ± 6.94 97.79 0.144
BMI (kg/m2) 25.00 ± 3.82 24.97 0.078
Waist to hip ratio 0.90 ± 0.007 0.90 0001
Central obesity
Yes
No
1156 (60.05)
769 (39.95)
62.46
37.54
0.010
0.010
Diabetes Mellitus
Yes
No
384 (18.96)
1641 (81.04)
16.93
83.07
0.007
0.007
Hypertension
Yes
No
1206 (57.79)
881 (42.21)
55.42
44.58
0.010
0.010
Dyslipidemia
Yes
No
1428 (75.12)
473 (24.88)
77.94
22.06
0.009
0.009
Attawood Lertpimonchai Results / 126
Table 4.7 Comparison of characteristic between actual and imputed dataset (cont.)
Characteristics EGAT 2/4 n = 1,948
Imputed data
Proportion / Mean SE
Total cholesterol (mg/dl) 217.73 ± 43.89 216.19 0.986
HDL (mg/dl) 57.53 ± 15.55 56.60 0.315
LDL (mg/dl) 144.86 ± 40.02 143.28 0.959
Triglyceride (mg/dl) 144.08 ± 94.03 150.89 2.027
Creatinine (mg/dl) 0.98 ± 0.22 0.99 0.006
Uric acid (mg/dl) 6.06 ± 1.45 6.11 0.031
eGFR (ml/min per 1.73m2) 80.45 ± 13.53 79.93 0.405
Periodontitis (CDC/AAP)
Non / Mild Periodontitis
Moderate Periodontitis
Severe Periodontitis
272 (14.31)
1002 (52.74)
626 (32.95)
10.36
42.04
47.60
0.006
0.010
0.010
% sites with PPD ≥ 4 mm 10.58 ± 16.23 14.45 0.336
% sites with PPD ≥ 6 mm 2.17 ± 6.34 2.29 0.116
% sites with PPD ≥ 4 mm
& CAL ≥ 5 mm 6.93 ± 14.12 10.04 0.297
% proximal sites with CAL
≥ 3 mm 56.35 ± 26.47 60.93 0.493
% proximal sites with CAL
≥ 5 mm 13.46 ± 21.11 17.62 0.438
Values are mean ± SD for continuous data, and frequency (%) for categorical data, except where specified. Total numbers of subjects of each variable may be different depended on missing data. BMI, body mass index; CAL, clinical attachment level; CKD, chronic kidney disease; eGFR, estimated glomerular filtration rate; FM-DM, family history of diabetes; FM-HT, family history of hypertension; HDL, high-density lipoprotein; LDL, low-density lipoprotein; N/A, not available; PPD, periodontal pocket depth
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 127
Table 4.8 The F-test and coefficients of various periodontitis and obesity definitions
Factors Overall F-test
b t p 95% CI
LL UL
PERIODONTITIS
CDC/AAP 3gr 8.55
Normal/Mild periodontitis - Reference
Moderate periodontitis - .4429876 1.62 0.105 -.0922116 .9781869
Severe periodontitis - .9738453 3.46 0.001 .4226753 1.525015
CDC/AAP 2gr 14.43
Non-severe periodontitis - Reference
Severe periodontitis - .5998399 3.80 < 0.001 .2902356 .9094441
Periodontitis Extent
Periodontitis A 16.49 .0184575 4.06 < 0.001 .0095483 .0273667
Periodontitis B 6.35 .0282988 2.52 0.012 .0062878 .0503099
Periodontitis C 16.53 .0184865 4.07 < 0.001 .0095756 .0273974
Periodontitis D 26.89 .0257024 5.19 < 0.001 .0159878 .0354171
Periodontitis E 37.97 .0187561 6.16 < 0.001 .0127903 .024722
Periodontitis F 40.62 .0229052 6.37 < 0.001 .015861 .0299494
OBESITY
BMI 3 gr 8.55
Underweight (< 18.5) - .307886 0.57 0.569 -.7535123 1.369284
Normal (18.5 – 24.9) - Reference
Overweight/Obese (≥ 25.0) - .6407238 3.79 < 0.001 .3087904 .9726572
BMI 2 gr 14.15
Non-obese (< 25) - Reference
Overweight/Obese (≥ 25.0) - .6277721 3.76 < 0.001 .3003383 .9552058
WHR 15.73
Normal - Reference
Obese (M > 0.90 | F > 0.85) - .7214715 3.97 < 0.001 .3640045 1.078939
b, coefficient; BMI, body mass index; CI, confidence interval; F, female; M, male; p, p-value; t, t-test; WHR, waist-hip-ratio
Attawood Lertpimonchai Results / 128
Table 4.9 Univariate GSEM of DM model: Mediation model
Factors b SE t p 95% CI
LL UL
Periodontitis F 0.051 0.005 9.66 < 0.001 0.041 0.062
Periodontitis (CDC/AAP)
Moderate Periodontitis
0.200
0.262
0.76
0.445
-0.313
0.713
Severe Periodontitis 1.246 0.278 4.49 < 0.001 0.702 1.791
Age (year) 0.368 0.024 15.46 < 0.001 0.321 0.414
Sex: male 1.173 0.258 4.55 < 0.001 0.668 1.677
Income (Baht/month)
< 20,000 -0.209 0.253 -0.83 0.409 -0.707 0.288
20,000 - 49,999 -0.884 0.171 -5.16 < 0.001 -1.220 -0.548
Education
≤ High school 1.227 0.288 4.26 < 0.001 0.662 1.793
Vocation/Diploma 1.251 0.259 4.84 < 0.001 0.744 1.758
Marital status
Married 0.621 0.409 1.52 0.129 -0.180 1.422
Divorce / widows 1.206 0.473 2.55 0.011 0.279 2.134
Exercise (times/week)
1 - 2 -0.580 0.228 -2.54 0.011 -1.028 -0.132
≥ 3 0.308 0.178 1.73 0.085 -0.042 0.657
Smoking
Quit smokers 1.301 0.227 5.73 < 0.001 0.856 1.746
Current smokers 0.813 0.259 3.13 0.002 0.304 1.322
Alcohol
Quit drinkers 1.915 0.257 7.46 < 0.001 1.412 2.419
Current drinkers 1.425 0.237 6.02 < 0.001 0.961 1.890
Obesity 1.681 0.216 7.77 < 0.001 1.256 2.106
HT 2.158 0.179 12.06 < 0.001 1.807 2.509
DLP 0.953 0.225 4.24 < 0.001 0.512 1.394
Family history of DM 1.550 0.212 7.32 < 0.001 1.135 1.965 b, coefficient; CI, confidence interval; DLP dyslipidemia; DM, diabetes mellitus; HT, hypertension; p, p-value; SE, standard error; t, t-test
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 129
Table 4.10 Univariate GSEM of CKD model: Outcome model
Factors b SE t p 95% CI
LL UL
Periodontitis F 0.023 0.004 6.37 < 0.001 0.016 0.030
Periodontitis (CDC/AAP)
Moderate Periodontitis 0.443 0.273 1.62 0.105 -0.092 0.978
Severe Periodontitis 0.974 0.281 3.46 0.001 0.423 1.525
DM 1.174 0.198 5.93 < 0.001 0.786 1.562
Sex: male 0.611 0.204 2.99 0.003 0.211 1.012
Income (Baht/month)
< 20,000 0.591 0.219 2.7 0.007 0.161 1.022
20,000 - 49,999 -0.667 0.188 -3.55 < 0.001 -1.036 -0.298
Education
≤ High school 0.238 0.215 1.11 0.268 -0.183 0.660
Vocation/Diploma 0.343 0.206 1.67 0.095 -0.060 0.746
Marital status
Married 0.811 0.367 2.21 0.027 0.091 1.530
Divorce / widows 1.101 0.432 2.55 0.011 0.254 1.947
Exercise (times/week)
1 - 2 -0.461 0.245 -1.88 0.06 -0.942 0.020
≥ 3 0.219 0.178 1.24 0.217 -0.129 0.568
Alcohol
Quit drinkers 1.116 0.231 4.83 < 0.001 0.663 1.570
Current drinkers 0.661 0.201 3.3 0.001 0.268 1.055
Smoking
Quit smokers 0.811 0.199 4.08 < 0.001 0.421 1.200
Current smokers 0.324 0.230 1.41 0.16 -0.128 0.775
NSAIDs 0.470 0.226 2.07 0.039 0.023 0.916
Obesity 0.721 0.182 3.97 < 0.001 0.364 1.079
HT 1.402 0.167 8.4 < 0.001 1.075 1.730
DLP 0.653 0.193 3.38 0.001 0.274 1.032
Uric acid (mg/dl) 0.688 0.062 11.06 < 0.001 0.566 0.810 b, coefficient; CI, confidence interval; CKD, chronic kidney disease; DLP dyslipidemia; DM, diabetes mellitus; HT, hypertension; NSAIDs, nonsteroidal anti-inflammatory drugs; p, p-value; SE, standard error; t, t-test
Attawood Lertpimonchai Results / 130
Table 4.11 Multivariate GSEM of mediation and outcome models of Pathway A:
Periodontitis F
Factors b SE t p
95% CI
LL UL
DM
Mod
el
Periodontitis 0.011 0.002 4.81 < 0.001 0.006 0.015
Age (year) 0.050 0.008 6.13 < 0.001 0.034 0.066
Education
≤ High school 0.550 0.150 3.66 < 0.001 0.255 0.844
Vocation/Diploma 0.494 0.135 3.66 < 0.001 0.230 0.759
Obesity 1.076 0.124 8.69 < 0.001 0.833 1.319
Family history of DM 0.938 0.116 8.09 < 0.001 0.711 1.165
HT 0.833 0.106 7.82 < 0.001 0.624 1.042
DLP 0.595 0.128 4.64 < 0.001 0.343 0.846
CK
D M
odel
Periodontitis 0.010 0.003 3.90 < 0.001 0.005 0.015
DM 0.689 0.155 4.44 < 0.001 0.385 0.994
Income (Baht/month)
< 20,000 0.278 0.175 1.59 0.112 -0.065 0.622
20,000 - 49,999 -0.476 0.155 -3.08 0.002 -0.780 -0.172
HT 0.748 0.141 5.31 < 0.001 0.472 1.024
Uric acid (mg/dl) 0.467 0.044 10.51 < 0.001 0.380 0.554
b, coefficient; CI, confidence interval; CKD, chronic kidney disease; DLP dyslipidemia; DM, diabetes mellitus; HT, hypertension; p, p-value; SE, standard error; t, t-test
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 131
Table 4.12 Multivariate GSEM of mediation and outcome models of Pathway A:
CDC/AAP
Factors b SE t p 95% CI
LL UL
DM
Mod
el
Periodontitis (CDC/AAP)
Moderate Periodontitis -0.207 0.337 -0.61 0.54 -0.867 0.454
Severe Periodontitis 0.222 0.372 0.6 0.55 -0.507 0.952
Age (year) 0.267 0.026 10.37 < 0.001 0.216 0.318
Education
≤ High school 1.513 0.368 4.12 < 0.001 0.792 2.234
Vocation/Diploma 1.453 0.323 4.49 < 0.001 0.819 2.087
Obesity 1.506 0.280 5.39 < 0.001 0.957 2.056
Family history of DM 2.732 0.311 8.78 < 0.001 2.122 3.342
HT 1.393 0.228 6.12 < 0.001 0.946 1.840
DLP 0.638 0.282 2.26 0.024 0.084 1.192
CK
D M
odel
Periodontitis (CDC/AAP)
Moderate Periodontitis 0.221 0.287 0.77 0.441 -0.341 0.783
Severe Periodontitis 0.486 0.296 1.64 0.101 -0.094 1.066
DM 0.903 0.211 4.28 < 0.001 0.490 1.316
Income (Baht/month)
< 20,000 0.336 0.230 1.46 0.145 -0.116 0.787
20,000 - 49,999 -0.661 0.195 -3.39 0.001 -1.045 -0.278
HT 0.934 0.178 5.24 < 0.001 0.584 1.285
Uric acid (mg/dl) 0.625 0.062 10.13 < 0.001 0.504 0.746
b, coefficient; CI, confidence interval; CKD, chronic kidney disease; DLP dyslipidemia; DM, Diabetes mellitus; HT, hypertension; p, p-value; SE, standard error; t, t-test
Attawood Lertpimonchai Results / 132
Table 4.13 Casual effects of Periodontitis on CKD through DM (Pathway A)
Effects Pathway b SE Bias 95% CI*
LL UL
Direct Periodontitis ⟶ CKD 0.01003 0.00258 0.00005 0.00463 0.01494
Indirect Periodontitis ⟶ DM ⟶ CKD 0.00738 0.00228 -0.00006 0.00381 0.01314
Percent of direct effect 57.62
Percent of total effects mediated 42.38
* bias-corrected bootstrapped
b, coefficient; CI, confidence interval; CKD, chronic kidney disease; DM, Diabetes mellitus; SE, standard error;
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 133 Table 4.14 Comparisons of forward stepwise method versus disjunctive clause criteria for DM Model
Forward stepwise method Partial disjunctive clause Modified fully disjunctive clause
Coefficient 95% LCI 95% UCI Coefficient. 95% LCI 95% UCI Coefficient. 95% LCI 95% UCI
DM-Model Periodontitis 0.011 0.006 0.015 0.011 0.007 0.015 0.011 0.006 0.015 Age 0.050 0.034 0.066 0.052 0.036 0.069 0.044 0.026 0.062 Gender (male) 0.314 -0.037 0.664 Income (Baht/month)
< 20,000 0.039 -0.257 0.334 0.012 -0.290 0.314 20,000 - 49,999 0.109 -0.086 0.305 0.083 -0.115 0.281
Education ≤ High school 0.550 0.255 0.844 0.489 0.166 0.812 0.508 0.184 0.832 Vocation/Diploma 0.494 0.230 0.759 0.477 0.205 0.748 0.482 0.210 0.753
Marital status Married -0.189 -0.615 0.237 Divorce / widows -0.121 -0.642 0.401
Exercise (times/week) 1 - 2 -0.069 -0.331 0.194 ≥ 3 0.088 -0.118 0.293
Alcohol drinking Quit drinkers 0.315 0.003 0.626 Current drinkers 0.119 -0.157 0.395
Smoking Quit smokers 0.121 -0.168 0.410 Current smokers -0.121 -0.456 0.215
Obesity 1.076 0.833 1.319 1.183 0.935 1.431 1.115 0.866 1.364 HT 0.833 0.624 1.042 0.909 0.700 1.118 0.859 0.648 1.070 DLP 0.595 0.343 0.846 0.646 0.394 0.898 0.620 0.368 0.873 Family history of DM 0.938 0.711 1.165 0.941 0.713 1.170 0.926 0.697 1.156 NSAIDs use 0.675 0.452 0.897 Uric acid -0.190 -0.266 -0.115 -0.229 -0.313 -0.144
DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension; NSAIDs, nonsteroidal anti-inflammatory drugs
Attaw
ood Lertpimonchai
R
esults / 134 Table 4.15 Comparisons of forward stepwise method versus disjunctive clause criteria for CKD Model
DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension; NSAIDs, nonsteroidal anti-inflammatory drugs
Forward stepwise method Partial disjunctive clause Modified fully disjunctive clause
Coefficient 95% LCI 95% UCI Coefficient. 95% LCI 95% UCI Coefficient. 95% LCI 95% UCI
CKD-Model Periodontitis 0.010 0.005 0.015 0.011 0.006 0.016 0.012 0.006 0.018 DM 0.689 0.385 0.994 0.751 0.438 1.065 0.688 0.368 1.008 Income (Baht/month)
< 20,000 0.278 -0.065 0.622 0.366 -0.017 0.749 0.323 -0.065 0.710 20,000 - 49,999 -0.476 -0.780 -0.172 -0.408 -0.728 -0.088 -0.394 -0.719 -0.069
Education ≤ High school -0.311 -0.689 0.066 -0.338 -0.717 0.041 Vocation/Diploma -0.066 -0.380 0.249 -0.070 -0.388 0.248
Marital status Married 0.381 -0.162 0.924 Divorce / widows 0.582 -0.060 1.223
Exercise (times/week) 1 - 2 -0.191 -0.579 0.197 ≥ 3 0.104 -0.179 0.387
Alcohol drinking Quit drinkers 0.124 -0.254 0.503 Current drinkers -0.216 -0.567 0.134
Smoking Quit smokers 0.037 -0.298 0.372 Current smokers -0.219 -0.627 0.189
Obesity -0.024 -0.341 0.293 0.002 -0.318 0.321 HT 0.748 0.472 1.024 0.757 0.472 1.041 0.722 0.439 1.006 DLP 0.073 -0.248 0.394 0.055 -0.268 0.377 Family history of DM -0.268 -0.548 0.013 -0.259 -0.542 0.024 NSAIDs use 0.200 -0.181 0.582 Uric acid 0.467 0.380 0.554 0.466 0.376 0.556 0.477 0.382 0.571
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 135
Table 4.16 Univariate GSEM of Periodontitis model: Mediation model
Factors F b SE t p 95% LCI 95% UCI
DM 143.700 7.248 0.605 11.990 < 0.001 6.063 8.433
Age (year) 989.030 0.740 0.024 31.450 < 0.001 0.693 0.786
Sex: male 158.660 10.287 0.817 12.600 < 0.001 8.686 11.888
Income (Baht/month) 37.420
< 20,000 -0.537 0.527 -1.020 0.309 -1.573 0.499
20,000 - 49,999 -2.720 0.320 -8.510 < 0.001 -3.348 -2.093
Education 56.660
≤ High school 8.722 0.870 10.030 < 0.001 7.013 10.431
Vocation/Diploma 6.003 0.724 8.300 < 0.001 4.583 7.424
Marital status 28.590
Married 4.867 0.984 4.950 < 0.001 2.939 6.795
Divorce / widows 8.708 1.173 7.420 < 0.001 6.407 11.008
Exercise (times/week) 18.880
1 - 2 -2.422 0.411 -5.900 < 0.001 -3.228 -1.616
≥ 3 -0.189 0.393 -0.480 0.631 -0.963 0.585
Smoking 126.450
Quit smokers 9.923 0.632 15.710 < 0.001 8.684 11.162
Current smokers 10.988 0.731 15.030 < 0.001 9.554 12.423
Alcohol 116.600
Quit drinkers 7.279 0.480 15.160 < 0.001 6.337 8.221
Current drinkers 5.044 0.452 11.170 < 0.001 4.158 5.929
Obesity 46.890 2.585 0.378 6.850 < 0.001 1.844 3.326
HT 103.250 3.510 0.345 10.160 < 0.001 2.833 4.188
DLP 11.730 1.355 0.396 3.420 0.001 0.579 2.131
b, coefficient; CI, confidence interval; DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension; p, p-value; SE, standard error; t, t-test
Attawood Lertpimonchai Results / 136
Table 4.17 Multivariate GSEM of mediation and outcome models of Pathway B
Factors b SE t p 95% LCI 95% UCI
Peri
odon
titis
M
odel
DM 4.801 1.131 4.24 < 0.001 2.584 7.018
Age 0.636 0.044 14.51 < 0.001 0.550 0.722
Gender: male 4.204 0.637 6.59 < 0.001 2.954 5.453
Education
≤ High school 9.911 0.878 11.29 < 0.001 8.190 11.632
Vocation/Diploma 4.961 0.671 7.39 < 0.001 3.645 6.277
Exercise (times/week)
1 - 2 -1.607 0.616 -2.61 0.009 -2.814 -0.400
≥ 3 -1.273 0.576 -2.21 0.027 -2.403 -0.143
Smoking
Quit smokers 3.748 0.764 4.90 < 0.001 2.250 5.246
Current smokers 14.063 1.058 13.29 < 0.001 11.988 16.137
CK
D
Mod
el
Periodontitis 0.010 0.003 3.90 < 0.001 0.005 0.015
DM 0.689 0.155 4.44 < 0.001 0.385 0.994
Income (Baht/month)
< 20,000 0.278 0.175 1.59 0.112 -0.065 0.622
20,000 - 49,999 -0.476 0.155 -3.08 0.002 -0.780 -0.172
HT 0.748 0.141 5.31 < 0.001 0.472 1.024
Uric acid 0.467 0.044 10.51 < 0.001 0.380 0.554
b, coefficient; CI, confidence interval; CKD, chronic kidney disease; DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension; p, p-value; SE, standard error; t, t-test
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 137
Table 4.18 Casual effects of DM on CKD through Periodontitis (Pathway B)
Effects Pathway b SE Bias 95% CI*
LL UL
Direct DM ⟶ CKD 0.68944 0.15423 -0.00440 0.36568 0.98245
Indirect DM ⟶ Periodontitis ⟶ CKD 0.04816 0.01833 0.00049 0.02130 0.09555
Percent of direct effect 93.47
Percent of total effects mediated 6.53
* bias-corrected bootstrapped
b, coefficient; CI, confidence interval; CKD, chronic kidney disease; DM, Diabetes mellitus; SE, standard error
Attaw
ood Lertpimonchai
R
esults / 138 Table 4.19 Comparisons of forward stepwise method versus disjunctive clause criteria for Periodontitis Model
Forward stepwise method Partial disjunctive clause Modified fully disjunctive clause
Coefficient 95% LCI 95% UCI Coefficient. 95% LCI 95% UCI Coefficient. 95% LCI 95% UCI
Periodontitis-Model DM 4.801 2.584 7.018 4.652 2.436 6.869 4.766 2.546 6.986 Age 0.636 0.550 0.722 0.662 0.567 0.758 0.654 0.554 0.754 Gender (male) 4.204 2.954 5.453 5.298 3.743 6.853 5.578 3.899 7.258 Income (Baht/month)
< 20,000 2.364 0.417 4.311 2.314 0.359 4.269 20,000 - 49,999 1.151 0.045 2.257 1.079 -0.035 2.194
Education ≤ High school 9.911 8.190 11.632 9.028 7.199 10.857 9.141 7.298 10.984 Vocation/Diploma 4.961 3.645 6.277 4.681 3.352 6.010 4.781 3.441 6.121
Marital status Married -1.060 -3.210 1.091 Divorce / widows -0.901 -3.633 1.831
Exercise (times/week) 1 - 2 -1.607 -2.814 -0.400 -1.732 -2.941 -0.522 -1.654 -2.862 -0.445 ≥ 3 -1.273 -2.403 -0.143 -1.391 -2.520 -0.262 -1.346 -2.477 -0.215
Alcohol drinking Quit drinkers 0.498 -1.184 2.181 Current drinkers -1.180 -2.672 0.311
Smoking Quit smokers 3.748 2.250 5.246 3.784 2.292 5.277 3.994 2.462 5.526 Current smokers 14.063 11.988 16.137 13.927 11.854 15.999 14.289 12.171 16.408
Obesity 0.265 -0.854 1.383 HT -0.172 -1.371 1.027 -0.040 -1.236 1.157 DLP -0.299 -1.510 0.912 NSAIDs use -1.734 -3.305 -0.164 Uric acid -0.648 -1.159 -0.138 -0.618 -1.138 -0.097
DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension; NSAIDs, Nonsteroidal anti-inflammatory drugs
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 139
Figure 4.1 Flowchart of included subjects CKD, chronic kidney disease; EGAT, Electricity Generating Authority of Thailand; Perio, periodontitis
EGAT 2/2 EGAT 2/3 EGAT 2/4
EGAT 2/2 – 2/4
Non-CKD c o h o r t
• 126 CKD at baseline
2,669
Perio-CKD c o h o r t
2,635
Periodontal exclusion
• 6: Medical condition
• 15: Fully edentulous
• 13: Refused
n = 2,795
n = 2,686 n = 2,288 n = 2,037
Attaw
ood Lertpim
onchai
Results / 140
Figure 4.2 Causal mediation pathway diagram of Pathway A using generalized structural equation modelling
Values represent unstandardized coefficients CKD, chronic kidney disease; DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension
0.689DM
Bernoulli
logit
CKD
Bernoulli
logit
PeriodontitisAge Family history of DM
Education: ≤ High school
Education: Vocation/Diploma
Income: < 20,000
Income: 20,000-49,999
HT Uric acidObesity
DLP
Fac. of Grad. Studies, M
ahidol Univ.
Ph.D
. (Clinical Epidem
iology) / 141
Figure 4.3 Causal mediation pathway diagram of Pathway B using generalized structural equation modelling
Values represent unstandardized coefficients CKD, chronic kidney disease; DLP, dyslipidemia; DM, diabetes mellitus; HT, hypertension
0.010Periodontitis
Gaussian
identity
CKD
Bernoulli
logit
DMAge Male
Education: ≤ High school
Education: Vocation/Diploma
Income: < 20,000
Income: 20,000-49,999
HT Uric acid
Exercise: 1-2 times/week
Exercise: ≥ 3 times/week
Smoking: Quit smokers
Smoking: Current smokers
Attawood Lertpimonchai Discussion / 142
CHAPTER V
DISCUSSION
5.1 Main findings The conceptual framework of causative association between periodontitis,
DM and CKD were constructed using a mediation analysis indicating periodontitis and
DM were risk factors of CKD incidence. Both had the significant direct effect, as well
as, indirect (mediation) effect through each other. From the pathway A, periodontitis →
DM → CKD, the finding suggested that every increasing one percent of proximal sites
with severe periodontitis, the odds of developing CKD would be directly increased and
indirectly transmitted thought DM 1.010 and 1.007, respectively. The pathway B, DM
→ periodontitis → CKD, suggested that subjects with DM would had odds of CKD
incidence higher than non-DM around 2 times, and this effect was attributed to the
mediation effect via periodontitis 6.5%.
5.2 Comparison results with previous studies Similar with previous cross-control studies12-25, our results could show the
significant association of CKD and its risk factors including periodontitis and DM. But
with our cohort design which excluded CKD cases at baseline, the temporal relationship
cloud be further firmly established. All cases of interested outcome were subsequence
of periodontitis and DM.
Our results were consistent with previous six cohort studies26-31, which
found the causative association between periodontitis and CKD. However, half of
them26, 27, 30 mainly focused on kidney function declination not the CKD incidence. Both
CKD and non-CKD cases also were included into these studies. Chen et al 27 defined
the primary outcome as decline ≥ 30% of eGFR. Chang et al 26 used the progression of
color intensity in GFR and albuminuria grid (CGA staging) as the outcome. While,
Iwasaki et al 30 interested in worsening of eGFR category, which defined as ≥ 60,
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 143
30-59, and < 29 ml/min per 1.73 m2, here, the interpretation should be done with caution,
because decrease of significantly different size would be classified similarly, for
example, eGFR declination from 90 to 32 ml/min per 1.73 m2, and 31 to 28 ml/min per
1.73 m2.
Besides, there were 3 cohort studies, which interested outcomes were new
cases of altered kidney function28, 29, 31. All of them revealed the significant effects of
periodontitis, which were similar with ours. Shultis et al31 showed that the incidences
of macroalbuminuria were 2.0, 2.1, and 2.6 times as high in individuals with moderate
or severe periodontitis or those who were edentulous, respectively, compared with those
with none/mild periodontitis. However, they defined periodontal status from numbers
of remaining teeth and radiography without clinical periodontal parameters, which
might be affect the internal validity. Likewise, the MrOS cohorts29 applied the random
half-mouth protocol for periodontal examination which may be underestimated the
prevalence of periodontitis207, and then, the association between periodontitis and CKD
might be bias. Moreover, eligible subjects from these 3 cohorts were quite specific
groups, i.e., African-American28, elderly male (hospital-based)29, and Pima Indian with
type II of DM31. Generalization and implication to other populations should be done
with caution. Furthermore, the important difference between conventional and
mediation analysis is the perspective on the third variable. The conventional analysis,
as previous cohort studies, claiming the casual association regarded DM as the
confounder of periodontitis effect. While, the mediation analysis treated
DM/periodontitis as the intermediate variable (mediator). It has been claimed the
advantage in refining and understanding a possible pathway. In other words, the
mediation analysis could determine a process of how one variable effect the outcome208.
Comparing with the previous bi-directional mediation analysis using cross-
sectional survey data, Fisher et al15 found significant direct and indirect effect of
periodontitis on CKD through DM duration and HT. In addition, they also found reverse
association between diabetic duration and periodontitis, both direct and indirect
associations through HT and CKD. Our data and this previous study were different.
First, we considered only CKD incidence whereas Fisher et al considered old and new
CKD cases. Second, interestingly, the direct effect of DM on CKD was not significant
in the previous study. DM status (non-DM / well-control DM / poor-control DM) and
Attawood Lertpimonchai Discussion / 144
DM duration (assuming equal zero if subjects were non-DM) were used as
representation of DM, and both were not significant in the full-model of CKD. Hence,
they concluded that DM duration influenced CKD indirectly through HT and
periodontitis, but unexpectedly does not directly impact CKD. While, most previous
studies and ours study reported the independent and significant effect of DM on CKD.
5.3 EGAT data management The EGAT study, the first cohort study of chronic disease in Thailand, was
set up by a group of cardiologists at Ramathibodi Hospital, Bangkok, since 1985. Until
now, the cohort is actively surveyed and followed up every 5 years. Thus, the EGAT
database has continuously growth. From then to now, the data management, including
data entry process, format of database, data cleaning process, has been changed over
time depended on data managers. In this study, we retrieved and explored 3 periods of
the EGAT2 cohort, and found that validity of data should be systematically manage. In
EGAT2/2 and 2/3, the processes of data entry seem to be problematic. Data entry error
was occasionally detected. Process of data cleaning by recheck with the original CRFs
was strongly suggested. In addition, the consistency of variables is another issue that
should be aware. Within each survey, some replies from the questionnaire were
contradict to each other. Revise the questionnaire with the trained staff at the survey
site, before the end of survey, may reduce these problems. The official team of data
cleaning should be authorized to methodically solve and judge data queries. Moreover,
to validate the suspected cases, for example, single FBS ≥ 126 mg/dl or rapidly decrease
in kidney function, should be followed and documents with the proper planning.
5.4 Periodontal measurement and classification As for our results, we found that periodontitis defined form disease extent
could show the significant effect on DM and CKD incidence, meanwhile, CDC/AAP
definition was not. Various case definitions of periodontitis have been proposed for
periodontal research55. It has great impact on the prevalence and extent of periodontal
disease56. Moreover, these discrepancies diverged the results of investigating the link
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 145
between periodontitis and systemic diseases57. Due to lack of universally accepted, the
CDC and AAP proposed and updated the standard case definitions58, 59. These
definitions have been proposed for surveillance, mainly, to determine total prevalence
of periodontitis. Some previous studies16, 20, 28, 29 applied them for investigate the
relationship between periodontitis and kidney function. However, we postulate that the
plausible link between periodontitis and the systemic disease based on inflammation.
With CDC/AAP definition, in some cases, the amounts of inflammation seem to be
discrepancies within the same category. For example, subjects who had only 2 teeth
with severe periodontitis would be grouped as the same as subjects who had whole-
mouth with severe form.
Therefore, periodontitis was also alternatively dealt as continuous data by
using the disease extent. The percentage of disease sites were calculated from full-
mouth periodontal data. Six definitions of disease sites that we extracted and
summarized from previously systematic review were proposed. Results showed that the
percentage of proximal sites with severe periodontitis or CAL ≥ 5mm was the most
suitable to our primary outcome. The cut-off point of CAL at 5 mm or above was quoted
from amount of attachment loss at severe level in the AAP-1999 classification of
periodontal disease54. Only the proximal sites were considered to rule out the attachment
loss due to non-periodontitis cause, i.e., traumatic brushing at buccal and lingual sites.
5.5 Multiple imputation Missing data is undesirable, inevitable, and unignorably. It is common
problem in both observational or clinical trials study particularly with long term follow
up. There are several ways to dealing with, i.e., list-wise deletion (complete case
analysis), simple imputation with average or mode, simple regression substitution, and
MICE209, 210. With current literatures, the MICE is recommended, because results from
MICE are more valid comparing with other methods189, 210. However, currently there is
no standard guideline which method we should apply for multiple imputation in
longitudinal data. For our study, we tried to impute missing variables by using
themselves at different periods as the predictors. Moreover, to impute some conditional
variables, i.e., marital status, education level, smoking, and alcohol drinking, by
Attawood Lertpimonchai Discussion / 146
conventional MICE might be achieved some inappropriate values, for example, reversed
education level. Although, the Stata does not provide options that supported to those
conditions, alternative models for imputation, including interval linear regression model
and the truncated regression model, could be applied. The Stata is quite flexible for
imputation. Users can construct the individual imputation model for each imputed
variable with own model and predictors. It should be noted that the computations were
somewhat time-consuming, for example, with m=20 in our dataset, generating the
imputed dataset lasted about 14 hours on a personal computer (macOS Sierra, 3.3 GHz
Intel Core i7, 8 GB RAM, with 4-core Stata 14.2 MP). The high specification of
computer with the Stata multi-core version are recommended.
Comparing with the complete case analysis (i.e., actual data without
imputation), ours results from MICE were quite approximate in term of significant
factors and their coefficients, but MICE decreased uncertainty (lower standard errors),
in other words, increased precision of results.
The number of imputation, firstly, were set based on the percentage of
missing data which ranged from 0.38% to 18.30%. Then, FMI from final models were
considered to confirm that our twenty imputations were adequate efficiency and power.
A rule of thumb suggested that the number of imputation should be larger than
FMI × 100. Therefore, twenty imputations would be insufficient for this dataset.
5.6 Selection of co-variables for mediation analysis To claim the causative association, the RCT is the ideal study design. With
the randomization technique, the balances of measured and unmeasured co-variables
among groups are claimed. However, the risk factors or exposure to periodontitis could
not be randomly allocated. Therefore, confounding bias may play a role in the causal
pathway. Selection of co-variables to include and adjust in the mediation analysis is
therefore important process. For our study, the disjunctive clause criteria have been
proposed by adopted the conceptual framework of the treatment effect analysis. This
criterion can be applied to the mediation analysis by considering the mediator as the
treatment. It consists of controlling for all co-variables that satisfy (i) the co-variates are
associated with exposure, or (ii) the co-variates are associated with the outcome, or
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 147
both203. In other words, with the disjunctive clause criteria, we try to use the relevant
co-variables as much as we can to estimate the effect size. It believed that the more co-
variables that we consider could be better in explanation the unmeasured variables.
In our results, the pattern of associations and coefficients of study factor and
mediator were approximated between conventional stepwise and the disjunctive clause
criteria in all models, i.e., CKD model, DM model and periodontitis model.
Interestingly, some variables were added with the significant effect size in the
disjunctive clause criteria. It might be hidden valuable, which was not presented in the
conventional stepwise analysis, for instance, the protective effect of NSAIDs in the
periodontitis model.
5.7 Strengths of this study We constructed a bi-directional causal pathway of periodontitis, DM, and
CKD using mediation analysis. Data from EGAT cohort had been used, only new cases
of CKD were considered, thus a temporal relationship was claimed. The EGAT cohort
survey is still active with a good data management. Although some missing data
occurred, the complex MICE for longitudinal study were constructed to deal with
missing data, and it increased the precision of results. Next, the periodontal status was
examined with the gold standard protocol, full-mouth examination with six sites per
tooth, therefore misclassification of periodontal disease severity should be minimized.
Finally, many periodontitis definitions were used to identified the association with the
outcome, and the best match was proposed.
5.8 Limitations Some limitations in our study should be pointed out. CKD was classified
based on eGFR without information of micro- or macro-proteinuria. Moreover,
proteinuria by itself, was also worsening factor for kidney function, hence the CKD
incident and effect size of other co-variables might be bias. Second, the cohort had only
3 surveys with 5-year interval. With the large gaps between surveys, uncertainty of
outcome and variables among visits were present. Thus, the time varying co-variables
Attawood Lertpimonchai Discussion / 148
analysis was used to compensate for this bias. Third, our studied population might not
represent the general Thai population. EGAT employees represented older adults with
higher education and income than average Thai people. The generalization of our
findings should be made with caution. Finally, our mediation effect might be only a
small part under the complexity of association between these diseases and other chronic
conditions. Other relevant diseases or behaviors might be the candidates of other
mediators or moderators. Future investigation will be required to clarify.
5.9 Clinical application Systemic chronic inflammation from periodontitis is a risk factor not only
for DM and CKD, but also, CVD and all-cause mortality. In general population, almost
50% of adults had periodontitis, with about 10% having severe periodontitis47. It means
that large populations were at risk to develop the subsequent burdens from periodontitis.
Worse than that, awareness among patients and health-care providers was low211. Our
results pointed out the important of this oral disease. These messages may encourage
the communication between physicians and dentists. Regularly refer patients with
chronic disease such as DM to the periodontal clinic is recommended to minimize the
oral inflammation. For general dentists, not only the destructive periodontium, but also
the systemic burden form periodontitis should be emphasized. Advice for medical
check-up among periodontitis patients is also suggested to detect kidney deterioration,
as well as, other adverse events from periodontitis. To public, the basic knowledge about
periodontitis should be promoted to increase awareness. Prevention and treatment of
periodontal disease are effective and inexpensive modalities. Motivation the personal
oral health care, routine dental check-up, and professional cleaning are efficient in
control the oral inflammation, and minimized spreading systemic inflammation from
periodontitis.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 149
5.10 Suggestion for further studies To overcome our limitations, the prospective cohort study which directly
designed based on our primary research questions is recommended. Community-based
with properly sampling technique is more suitable and more representativeness of study
subjects. The study duration more than 10 to 15 years is suggested to permit the CKD
development after DM onset. In addition, the annual follow-up periods should be
proposed to serve time-varying co-variables pattern, and allow the advance “time to
event” statistical analysis. Most importantly, microproteinuria should be additionally
used for outcome verification, as well as, be considered as another risk factor of CKD.
Moreover, periodontal parameters indicated the gingival inflammation, such as, GI,
BOP and periodontal inflamed surface area (PISA), should be used to characterize the
inflammation that could be spreading to overall system.
We currently proposed the effect of periodontitis on CKD initiation.
Similarly, its progression may be also influenced by oral inflammation. The cohort of
CKD patients could be used to emphasize the important of oral health. Furthermore,
investigation the protective effect of periodontal treatment on CKD initiation and
progression is also interesting and valuable.
5.11 Conclusion In conclusion, periodontitis and DM had the significant direct and indirect
effect via each other on increasing CKD incidence. Oral and systemic morbidities from
periodontitis should be emphasized among nephrologists, general practitioners, and
patients. Its treatment and prevention also should be promoted in public.
Attawood Lertpimonchai References / 150
REFERENCES
1.Levin A, Stevens PE. Summary of KDIGO 2012 CKD Guideline: behind the scenes,
need for guidance, and a framework for moving forward. Kidney
international. 2014;85(1):49-61.
2.Levey AS, Coresh J. Chronic kidney disease. Lancet. 2012;379(9811):165-80.
3.Gansevoort RT, Correa-Rotter R, Hemmelgarn BR, Jafar TH, Heerspink HJ, Mann JF,
et al. Chronic kidney disease and cardiovascular risk: epidemiology,
mechanisms, and prevention. Lancet. 2013;382(9889):339-52.
4.Chambrone L, Foz AM, Guglielmetti MR, Pannuti CM, Artese HP, Feres M, et al.
Periodontitis and chronic kidney disease: a systematic review of the
association of diseases and the effect of periodontal treatment on estimated
glomerular filtration rate. Journal of clinical periodontology.
2013;40(5):443-56.
5.Linden GJ, Lyons A, Scannapieco FA. Periodontal systemic associations: review of
the evidence. Journal of periodontology. 2013;84(4 Suppl):S8-s19.
6.Pinho MM, Faria-Almeida R, Azevedo E, Manso MC, Martins L. Periodontitis and
atherosclerosis: an observational study. Journal of periodontal research.
2013;48(4):452-7.
7.Tonetti MS. Periodontitis and risk for atherosclerosis: an update on intervention trials.
Journal of clinical periodontology. 2009;36 Suppl 10:15-9.
8.Casanova L, Hughes FJ, Preshaw PM. Diabetes and periodontal disease: a two-way
relationship. British dental journal. 2014;217(8):433-7.
9.Taylor GW, Burt BA, Becker MP, Genco RJ, Shlossman M, Knowler WC, et al.
Severe periodontitis and risk for poor glycemic control in patients with non-
insulin-dependent diabetes mellitus. Journal of periodontology. 1996;67(10
Suppl):1085-93.
10.Satirapoj B. Review on pathophysiology and treatment of diabetic kidney disease.
Journal of the Medical Association of Thailand. 2010;93 Suppl 6:S228-41.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 151
11.Taylor GW, Burt BA, Becker MP, Genco RJ, Shlossman M, Knowler WC, et al. Non-
insulin dependent diabetes mellitus and alveolar bone loss progression over
2 years. Journal of periodontology. 1998;69(1):76-83.
12.Fisher MA, Taylor GW. A prediction model for chronic kidney disease includes
periodontal disease. Journal of periodontology. 2009;80(1):16-23.
13.Fisher MA, Taylor GW, Papapanou PN, Rahman M, Debanne SM. Clinical and
serologic markers of periodontal infection and chronic kidney disease.
Journal of periodontology. 2008;79(9):1670-8.
14.Fisher MA, Taylor GW, Shelton BJ, Jamerson KA, Rahman M, Ojo AO, et al.
Periodontal disease and other nontraditional risk factors for CKD. American
journal of kidney diseases. 2008;51(1):45-52.
15.Fisher MA, Taylor GW, West BT, McCarthy ET. Bidirectional relationship between
chronic kidney and periodontal disease: a study using structural equation
modeling. Kidney international. 2011;79(3):347-55.
16.Grubbs V, Plantinga LC, Crews DC, Bibbins-Domingo K, Saran R, Heung M, et al.
Vulnerable populations and the association between periodontal and chronic
kidney disease. Clinical journal of the American Society of Nephrology.
2011;6(4):711-7.
17.Han K, Nam GE, Kim DH, Park JB, Ko Y, Roh YK, et al. Association of Periodontitis
With Urinary Albumin Excretion in Korean Adults With Diabetes: The 2012
Korea National Health and Nutrition Examination Survey. Medicine.
2015;94(42):e1839.
18.Han SS, Shin N, Lee SM, Lee H, Kim DK, Kim YS. Correlation between
periodontitis and chronic kidney disease in Korean adults. Kidney research
and clinical practice. 2013;32(4):164-70.
19.Ioannidou E, Hall Y, Swede H, Himmelfarb J. Periodontitis associated with chronic
kidney disease among Mexican Americans. Journal of public health
dentistry. 2013;73(2):112-9.
20.Ioannidou E, Swede H. Disparities in periodontitis prevalence among chronic kidney
disease patients. Journal of dental research. 2011;90(6):730-4.
Attawood Lertpimonchai References / 152
21.Kshirsagar AV, Moss KL, Elter JR, Beck JD, Offenbacher S, Falk RJ. Periodontal
disease is associated with renal insufficiency in the Atherosclerosis Risk In
Communities (ARIC) study. American journal of kidney diseases.
2005;45(4):650-7.
22.Ricardo AC, Athavale A, Chen J, Hampole H, Garside D, Marucha P, et al.
Periodontal disease, chronic kidney disease and mortality: results from the
third National Health and Nutrition Examination Survey. BMC nephrology.
2015;16:97.
23.Salimi S, Ng N, Seliger SL, Parsa A. Periodontal disease, renal dysfunction and
heightened leukocytosis. Nephron Clinical practice. 2014;128(1-2):107-14.
24.Yoshihara A, Iwasaki M, Miyazaki H, Nakamura K. Bidirectional relationship
between renal function and periodontal disease in older Japanese women.
Journal of clinical periodontology. 2016;43(9):720-6.
25.Yoshihara A, Nakamura K, Miyazaki H. The association between renal function and
tooth loss in Japanese community-dwelling postmenopausal women.
Gerodontology. 2012;29(2):e363-7.
26.Chang JF, Yeh JC, Chiu YL, Liou JC, Hsiung JR, Tung TH. Periodontal Pocket
Depth, Hyperglycemia, and Progression of Chronic Kidney Disease: A
Population-Based Longitudinal Study. The American journal of medicine.
2017;130(1):61-9.e1.
27.Chen YT, Shih CJ, Ou SM, Hung SC, Lin CH, Tarng DC. Periodontal Disease and
Risks of Kidney Function Decline and Mortality in Older People: A
Community-Based Cohort Study. American journal of kidney diseases.
2015;66(2):223-30.
28.Grubbs V, Vittinghoff E, Beck JD, Kshirsagar AV, Wang W, Griswold ME, et al.
Association Between Periodontal Disease and Kidney Function Decline in
African Americans: The Jackson Heart Study. Journal of periodontology.
2015;86(10):1126-32.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 153
29.Grubbs V, Vittinghoff E, Taylor G, Kritz-Silverstein D, Powe N, Bibbins-Domingo
K, et al. The association of periodontal disease with kidney function decline:
a longitudinal retrospective analysis of the MrOS dental study. Nephrology,
dialysis, transplantation. 2016;31(3):466-72.
30.Iwasaki M, Taylor GW, Nesse W, Vissink A, Yoshihara A, Miyazaki H. Periodontal
disease and decreased kidney function in Japanese elderly. American journal
of kidney diseases. 2012;59(2):202-9.
31.Shultis WA, Weil EJ, Looker HC, Curtis JM, Shlossman M, Genco RJ, et al. Effect
of periodontitis on overt nephropathy and end-stage renal disease in type 2
diabetes. Diabetes care. 2007;30(2):306-11.
32.Group KDIGOKCW. KDIGO 2012 clinical practice guideline for the evaluation and
management of chronic kidney disease. Kidney international Supplement.
2013;3:1-150.
33.Jha V, Garcia-Garcia G, Iseki K, Li Z, Naicker S, Plattner B, et al. Chronic kidney
disease: global dimension and perspectives. Lancet. 2013;382(9888):260-
72.
34.Webster AC, Nagler EV, Morton RL, Masson P. Chronic Kidney Disease. Lancet.
2017;389(10075):1238-52.
35.Ingsathit A, Thakkinstian A, Chaiprasert A, Sangthawan P, Gojaseni P,
Kiattisunthorn K, et al. Prevalence and risk factors of chronic kidney disease
in the Thai adult population: Thai SEEK study. Nephrology, dialysis,
transplantation. 2010;25(5):1567-75.
36.Levey AS, Inker LA, Coresh J. GFR estimation: from physiology to public health.
American journal of kidney diseases. 2014;63(5):820-34.
37.Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, 3rd, Feldman HI, et al.
A new equation to estimate glomerular filtration rate. Annals of internal
medicine. 2009;150(9):604-12.
38.Chien KL, Lin HJ, Lee BC, Hsu HC, Lee YT, Chen MF. A prediction model for the
risk of incident chronic kidney disease. The American journal of medicine.
2010;123(9):836-46.e2.
Attawood Lertpimonchai References / 154
39.Domrongkitchaiporn S, Sritara P, Kitiyakara C, Stitchantrakul W, Krittaphol V,
Lolekha P, et al. Risk factors for development of decreased kidney function
in a southeast Asian population: a 12-year cohort study. Journal of the
American Society of Nephrology. 2005;16(3):791-9.
40.Echouffo-Tcheugui JB, Kengne AP. Risk models to predict chronic kidney disease
and its progression: a systematic review. PLoS medicine.
2012;9(11):e1001344.
41.Kshirsagar AV, Bang H, Bomback AS, Vupputuri S, Shoham DA, Kern LM, et al. A
simple algorithm to predict incident kidney disease. Archives of internal
medicine. 2008;168(22):2466-73.
42.Thakkinstian A, Ingsathit A, Chaiprasert A, Rattanasiri S, Sangthawan P, Gojaseni
P, et al. A simplified clinical prediction score of chronic kidney disease: A
cross-sectional-survey study. BMC nephrology. 2011;12(1).
43.Goto M, Kawamura T, Wakai K, Ando M, Endoh M, Tomino Y. Risk stratification
for progression of IgA nephropathy using a decision tree induction
algorithm. Nephrology, dialysis, transplantation. 2009;24(4):1242-7.
44.Meguid El Nahas A, Bello AK. Chronic kidney disease: the global challenge. Lancet.
2005;365(9456):331-40.
45.Kanji Z, Powe CE, Wenger JB, Huang C, Ankers E, Sullivan DA, et al. Genetic
variation in APOL1 associates with younger age at hemodialysis initiation.
Journal of the American Society of Nephrology. 2011;22(11):2091-7.
46.Wasser WG, Tzur S, Wolday D, Adu D, Baumstein D, Rosset S, et al. Population
genetics of chronic kidney disease: the evolving story of APOL1. Journal of
nephrology. 2012;25(5):603-18.
47.Eke PI, Dye BA, Wei L, Slade GD, Thornton-Evans GO, Borgnakke WS, et al.
Update on Prevalence of Periodontitis in Adults in the United States:
NHANES 2009 to 2012. Journal of periodontology. 2015;86(5):611-22.
48.Al-Harthi LS, Cullinan MP, Leichter JW, Thomson WM. The impact of periodontitis
on oral health-related quality of life: a review of the evidence from
observational studies. Australian dental journal. 2013;58(3):274-7; quiz 384.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 155
49.Mattila KJ, Nieminen MS, Valtonen VV, Rasi VP, Kesaniemi YA, Syrjala SL, et al.
Association between dental health and acute myocardial infarction. British
medical journal (Clinical research ed.). 1989;298(6676):779-81.
50.Dietrich T, Jimenez M, Krall Kaye EA, Vokonas PS, Garcia RI. Age-dependent
associations between chronic periodontitis/edentulism and risk of coronary
heart disease. Circulation. 2008;117(13):1668-74.
51.Tonetti MS, Van Dyke TE. Periodontitis and atherosclerotic cardiovascular disease:
consensus report of the Joint EFP/AAP Workshop on Periodontitis and
Systemic Diseases. Journal of clinical periodontology. 2013;40 Suppl
14:S24-9.
52.Sanz M, Kornman K. Periodontitis and adverse pregnancy outcomes: consensus
report of the Joint EFP/AAP Workshop on Periodontitis and Systemic
Diseases. Journal of periodontology. 2013;84(4 Suppl):S164-9.
53.Monsarrat P, Blaizot A, Kemoun P, Ravaud P, Nabet C, Sixou M, et al. Clinical
research activity in periodontal medicine: a systematic mapping of trial
registers. Journal of clinical periodontology. 2016;43(5):390-400.
54.Armitage GC. Development of a classification system for periodontal diseases and
conditions. Annals of periodontology / the American Academy of
Periodontology. 1999;4(1):1-6.
55.Beltran-Aguilar ED, Eke PI, Thornton-Evans G, Petersen PE. Recording and
surveillance systems for periodontal diseases. Periodontology 2000.
2012;60(1):40-53.
56.Costa FO, Guimaraes AN, Cota LO, Pataro AL, Segundo TK, Cortelli SC, et al.
Impact of different periodontitis case definitions on periodontal research.
Journal of oral science. 2009;51(2):199-206.
57.Ioannidou E, Shaqman M, Burleson J, Dongari-Bagtzoglou A. Periodontitis case
definition affects the association with renal function in kidney transplant
recipients. Oral diseases. 2010;16(7):636-42.
58.Page RC, Eke PI. Case definitions for use in population-based surveillance of
periodontitis. Journal of periodontology. 2007;78(7 Suppl):1387-99.
Attawood Lertpimonchai References / 156
59.Eke PI, Page RC, Wei L, Thornton-Evans G, Genco RJ. Update of the case definitions
for population-based surveillance of periodontitis. Journal of
periodontology. 2012;83(12):1449-54.
60.Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic
reviews and meta-analyses: the PRISMA statement. PLoS medicine.
2009;6(7):e1000097.
61.Greene JC, Vermillion JR. The simplified oral hygiene index. Journal of the
American Dental Association. 1964;68:7-13.
62.Silness J, Loe H. Periodontal disease in pregnancy II. Correlation between oral
hygiene and periodontal condition. Acta odontologica Scandinavica.
1964;22:121-35.
63.O'Leary TJ, Drake RB, Naylor JE. The plaque control record. Journal of
periodontology. 1972;43(1):38.
64.Wells GA, Shea B, O'Connell D, Peterson J, Welch V, Losos M, et al. The Newcastle-
Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in
meta-analyses 2015 (http://www.ohri.ca/programs/clinical_epidemiology
/oxford.asp). (Accessed 22/8/2016).
65.White IR. Multivariate random-effects meta-regression: Updates to mvmeta. The
Stata journal. 2011;11(2):255-70.
66.Riley RD, Thompson JR, Abrams KR. An alternative model for bivariate random-
effects meta-analysis when the within-study correlations are unknown.
Biostatistics. 2008;9(1):172-86.
67.Atkins D, Best D, Briss PA, Eccles M, Falck-Ytter Y, Flottorp S, et al. Grading
quality of evidence and strength of recommendations. British medical
journal (Clinical research ed.). 2004;328(7454):1490.
68.Guyatt G, Oxman AD, Akl EA, Kunz R, Vist G, Brozek J, et al. GRADE guidelines:
1. Introduction-GRADE evidence profiles and summary of findings tables.
Journal of clinical epidemiology. 2011;64(4):383-94.
69.Bagramian RA, Farghaly MM, Lopatin D, Sowers M, Syed SA, Pomerville JL. A
comparison of periodontal disease among rural Amish and non-Amish
adults. Journal of clinical periodontology. 1994;21(6):386-90.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 157
70.Imaki M, Yoshida Y, Tanada S. Relation between smoking and periodontal disease
by oral hygiene status in Japanese factory workers. Applied human science.
1997;16(2):77-81.
71.Norderyd O, Hugoson A. Risk of severe periodontal disease in a Swedish adult
population. A cross-sectional study. Journal of clinical periodontology.
1998;25(12):1022-8.
72.Wakai K, Kawamura T, Umemura O, Hara Y, Machida JI, Anno T, et al. Associations
of medical status and physical fitness with periodontal disease. Journal of
clinical periodontology. 1999;26(10):664-72.
73.Papapanou PN, Neiderud AM, Papadimitriou A, Sandros J, Dahlén G.
"Checkerboard" Assessments of Periodontal Microbiota and Serum
Antibody Responses: A Case-Control Study. Journal of periodontology.
2000;71(6):885-97.
74.Hashim R, Thomson WM, Pack ARC. Smoking in adolescence as a predictor of early
loss of periodontal attachment. Community dentistry and oral epidemiology.
2001;29(2):130-5.
75.Tezal M, Grossi SG, Ho AW, Genco RJ. The effect of alcohol consumption on
periodontal disease. Journal of periodontology. 2001;72(2):183-9.
76.Hugoson A, Ljungquist B, Breivik T. The relationship of some negative events and
psychological factors to periodontal disease in an adult Swedish population
50 to 80 years of age. Journal of clinical periodontology. 2002;29(3):247-
53.
77.Do GL, Spencer AJ, Roberts-Thomson K, Ha HD. Smoking as a risk indicator for
periodontal disease in the middle-aged Vietnamese population. Community
dentistry and oral epidemiology. 2003;31(6):437-46.
78.Meisel P, Siegemund A, Grimm R, Herrmann FH, John U, Schwahn C, et al. The
interleukin-1 polymorphism, smoking, and the risk of periodontal disease in
the population-based SHIP study. Journal of dental research.
2003;82(3):189-93.
79.Alpagot T, Duzgunes N, Wolff LF, Lee A. Risk factors for periodontitis in HIV+
patients. Journal of periodontal research. 2004;39(3):149-57.
Attawood Lertpimonchai References / 158
80.Solis ACO, Lotufo RFM, Pannuti CM, Brunheiro EC, Marques AH, Lotufo-Neto F.
Association of periodontal disease to anxiety and depression symptoms, and
psychosocial stress factors. Journal of clinical periodontology.
2004;31(8):633-8.
81.Wickholm S, Söder PÖ, Galanti MR, Söder B, Klinge B. Periodontal disease in a
group of Swedish adult snuff and cigarette users. Acta odontologica
Scandinavica. 2004;62(6):333-8.
82.Natto S, Baljoon M, Bergstrom J. Tobacco smoking and periodontal health in a Saudi
Arabian population. Journal of periodontology. 2005;76(11):1919-26.
83.Torrungruang K, Tamsailom S, Rojanasomsith K, Sutdhibhisal S, Nisapakultorn K,
Vanichjakvong O, et al. Risk indicators of periodontal disease in older Thai
adults. Journal of periodontology. 2005;76(4):558-65.
84.de Macêdo TCN, Costa MdCN, Gomes-Filho IS, Vianna MIP, Santos CT. Factors
related to periodontal disease in a rural population. Brazilian oral research.
2006;20(3):257-62.
85.Khader YS. Factors associated with periodontal diseases in Jordan: principal
component and factor analysis approach. Journal of oral science.
2006;48(2):77-84.
86.Vandana KL, Sesha Reddy M. Assessment of periodontal status in dental fluorosis
subjects using community periodontal index of treatment needs. Indian
journal of dental research. 2007;18(2):67-71.
87.Wang QT, Wu ZF, Wu YF, Shu R, Pan YP, Xia JL. Epidemiology and preventive
direction of periodontology in China. Journal of clinical periodontology.
2007;34(11):946-51.
88.Akhter R, Hassan NMM, Aida J, Takinami S, Morita M. Relationship between betel
quid additives and established periodontitis among Bangladeshi subjects.
Journal of clinical periodontology. 2008;35(1):9-15.
89.Kumar TS, Dagli RJ, Mathur A, Jain M, Balasubramanyam G, Prabu D, et al. Oral
health status and practices of dentate Bhil adult tribes of southern Rajasthan,
India. International dental journal. 2009;59(3):133-40.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 159
90.Benguigui C, Bongard V, Ruidavets JB, Chamontin B, Sixou M, Ferrières J, et al.
Metabolic syndrome, insulin resistance, and periodontitis: A cross-sectional
study in a middle-aged French population. Journal of clinical
periodontology. 2010;37(7):601-8.
91.Saxlin T, Ylostalo P, Suominen-Taipale L, Aromaa A, Knuuttila M. Overweight and
obesity weakly predict the development of periodontal infection. Journal of
clinical periodontology. 2010;37(12):1059-67.
92.Bawadi HA, Khader YS, Haroun TF, Al-Omari M, Tayyem RF. The association
between periodontal disease, physical activity and healthy diet among adults
in Jordan. Journal of periodontal research. 2011;46(1):74-81.
93.Carrilho Neto A, De Paula Ramos S, Sant'ana AC, Passanezi E. Oral health status
among hospitalized patients. International journal of dental hygiene.
2011;9(1):21-9.
94.Mathur LK, Manohar B, Shankarapillai R, Pandya D. Obesity and periodontitis: A
clinical study. Journal of Indian Society of Periodontology. 2011;15(3):240-
4.
95.Teng PR, Su JM, Chang WH, Lai TJ. Oral health of psychiatric inpatients: A survey
of central Taiwan hospitals. General Hospital Psychiatry. 2011;33(3):253-9.
96.Crocombe LA, Brennan DS, Slade GD, Loc DO. Is self interdental cleaning
associated with dental plaque levels, dental calculus, gingivitis and
periodontal disease? Journal of periodontal research. 2012;47(2):188-97.
97.Mannem S, Chava VK. The effect of stress on periodontitis: A clinicobiochemical
study. Journal of Indian Society of Periodontology. 2012;16(3):365-9.
98.Raja S, Joshi V, Shirahatti R, Byakod G, Kanteshwari IK, Babu N. "Is the
periodontium an index of your mind?" A crossectional study to evaluate the
relationship between life events and periodontitis. International journal of
clinical dentistry. 2012;5(1):39-48.
99.Vogt M, Sallum AW, Cecatti JG, Morais SS. Factors associated with the prevalence
of periodontal disease in low-risk pregnant women. Reproductive health.
2012;9:3.
Attawood Lertpimonchai References / 160
100.Fiyaz M, Ramesh A, Ramalingam K, Thomas B, Shetty S, Prakash P. Association
of salivary calcium, phosphate, pH and flow rate on oral health: A study on
90 subjects. Journal of Indian Society of Periodontology. 2013;17(4):454-
60.
101.Palle AR, Reddy CMSK, Shankar BS, Gelli V, Sudhakar J, Reddy KKM.
Association between obesity and chronic periodontitis: A cross-sectional
study. The journal of contemporary dental practice. 2013;14(2):168-73.
102.Alhabashneh R, Khader Y, herra Z, Assad F. The association between periodontal
disease and metabolic syndrome among outpatients with diabetes in Jordan.
Journal of diabetes and metabolic disorders. 2015;14(1).
103.Bhat M, Roberts-Thomson K, Do LG. Clustering of risk indicators for periodontal
disease: A population-based study. Community dental health.
2015;32(3):158-62.
104.Cakmak O, Alkan BA, Ozsoy S, Sen A, Abdulrezzaki U. Association of gingival
crevicular fluid cortisol/dehydroepiandrosterone levels with periodontal
status. Journal of periodontology. 2014;85(8):e287-e94.
105.Develioglu H, Ozdemir H, Bostanci V. Comparative analysis of the blood flow
values of patients with type 2 diabetes mellitus presenting with chronic
periodontitis, patients with chronic periodontitis only and healthy
individuals. The West Indian medical journal. 2014;63(4):359-63.
106.Hsiao CN, Ko EC, Shieh TY, Chen HS. Relationship between areca nut chewing
and periodontal status of people in a typical aboriginal community in
Southern Taiwan. Journal of dental sciences. 2015;10(3):300-8.
107.Ismail FB, Ismail G, Dumitriu AS, Baston C, Berbecar V, Jurubita R, et al.
Identification of subgingival periodontal pathogens and association with the
severity of periodontitis in patients with chronic kidney diseases: A cross-
sectional study. Biomed research international. 2015;2015.
108.Jacob PS, Nath S, Patel RP. Evaluation of interleukin-1β and 8 in gutka chewers
with periodontitis among a rural Indian population. Journal of periodontal
and implant science. 2014;44(3):126-33.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 161
109.Kaur S, Narayanswamy S, Ramesh AV. Comparative evaluation of salivary soluble
CD44 levels in periodontal health and disease. Journal of Indian Society of
Periodontology. 2014;18(6):734-8.
110.Köseoʇlu S, Saʇlam M, Pekbaʇriyanik T, Savran L, Sütçü R. Level of interleukin-35
in gingival crevicular fluid, saliva, and plasma in periodontal disease and
health. Journal of periodontology. 2015;86(8):964-71.
111.Kovačević V, Milosavljević M, Rančić N, Daković D. Assessment of the
periodontal health and community periodontal index in the Army of Serbia.
Vojnosanitetski Pregled. 2015;72(11):953-60.
112.Kvarnvik C, Soljegard E, Charalampakis G, Suksu-Art N, Dahlen G. Periodontal
disease in a remote Asian population: association between clinical and
microbiological parameters. Journal of investigative and clinical dentistry.
2015.
113.Lavu V, Venkatesan V, Bhaskar LV, Priyanka V, Kumarasamy P, Franklin Durairaj
Paul S, et al. Polymorphic Regions in Fcgr and Tnf Alpha Genes and
Susceptibility to Chronic Periodontitis in a Cohort From South India. Journal
of periodontology. 2016:1-15.
114.Lütfioğlu M, Aydoğdu A, Sakallioğlu EE, Alaçam H, Pamuk F. Gingival crevicular
fluid interleukin-8 and lipoxin A4 levels of smokers and nonsmokers with
different periodontal status: A cross-sectional study. Journal of periodontal
research. 2015.
115.Meenawat A, Govila V, Goel S, Verma S, Punn K, Srivastava V, et al. Evaluation
of the effect of nicotine and metabolites on the periodontal status and the
mRNA expression of interleukin-1β in smokers with chronic periodontitis.
Journal of Indian Society of Periodontology. 2015;19(4):381-7.
116.Mesa F, Magán-Fernández A, Muñoz R, Papay-Ramírez L, Poyatos R, Sánchez-
Fernández E, et al. Catecholamine metabolites in Urine, as chronic stress
biomarkers, are associated with higher risk of chronic periodontitis in adults.
Journal of periodontology. 2014;85(12):1755-62.
Attawood Lertpimonchai References / 162
117.Perayil J, Suresh N, Fenol A, Vyloppillil R, Bhaskar A, Menon S. Comparison of
glycated hemoglobin levels in individuals without diabetes and with and
without periodontitis before and after non-surgical periodontal therapy.
Journal of periodontology. 2014;85(12):1658-66.
118.Pereira AL, Franco GC, Cortelli SC, Aquino DR, Costa FO, Raslan SA, et al.
Influence of periodontal status and periodontopathogens on levels of oral
human β-defensin-2 in saliva. Journal of periodontology. 2013;84(10):1445-
53.
119.Petrović SM, Zelić K, Milašin J, Popović B, Pucar A, Zelić O. Detection of herpes
simplex virus type 1 in gingival crevicular fluid of gingival
sulcus/periodontal pocket using polymerase chain reaction. Srpski Arhiv za
Celokupno Lekarstvo. 2014;142(5-6):296-300.
120.Pranckeviciene A, Siudikiene J, Ostrauskas R, Machiulskiene V. Severity of
periodontal disease in adult patients with diabetes mellitus in relation to the
type of diabetes. Biomedical papers of the Medical Faculty of the University
Palacky, Olomouc, Czechoslovakia. 2014;158(1):117-23.
121.Puri K, Chhokra M, Dodwad V, Puri N. Association of interleukin-1 α (-889) gene
polymorphism in patients with generalized aggressive and chronic
periodontitis. Dental research journal. 2015;12(1):76-82.
122.Singh P, Gupta ND, Bey A, Khan S. Salivary TNF-alpha: A potential marker of
periodontal destruction. Journal of Indian Society of Periodontology.
2014;18(3):306-10.
123.Toyman U, Tüter G, Kurtiş B, Kivrak E, Bozkurt S, Yücel AA, et al. Evaluation of
gingival crevicular fluid levels of tissue plasminogen activator, plasminogen
activator inhibitor 2, matrix metalloproteinase-3 and interleukin 1-β in
patients with different periodontal diseases. Journal of periodontal research.
2015;50(1):44-51.
124.Varghese SS, Thomas H, Jayakumar ND, Sankari M, Lakshmanan R. Estimation of
salivary tumor necrosis factor-alpha in chronic and aggressive periodontitis
patients. Contemporary clinical dentistry. 2015;6:S152-S6.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 163
125.Nelson RG, Shlossman M, Budding LM, Pettitt DJ, Saad MF, Genco RJ, et al.
Periodontal disease and NIDDM in Pima Indians. Diabetes care.
1990;13(8):836-40.
126.Papapanou PN. Periodontal diseases: epidemiology. Annals of periodontology / the
American Academy of Periodontology. 1996;1(1):1-36.
127.Suvan J, D'Aiuto F, Moles DR, Petrie A, Donos N. Association between
overweight/obesity and periodontitis in adults. A systematic review. Obesity
reviews. 2011;12(5):e381-404.
128.Zimmermann H, Zimmermann N, Hagenfeld D, Veile A, Kim TS, Becher H. Is
frequency of tooth brushing a risk factor for periodontitis? A systematic
review and meta-analysis. Community dentistry and oral epidemiology.
2015;43(2):116-27.
129.Berchier CE, Slot DE, Haps S, Van der Weijden GA. The efficacy of dental floss in
addition to a toothbrush on plaque and parameters of gingival inflammation:
a systematic review. International journal of dental hygiene. 2008;6(4):265-
79.
130.Tonetti MS, Eickholz P, Loos BG, Papapanou P, van der Velden U, Armitage G, et
al. Principles in prevention of periodontal diseases: Consensus report of
group 1 of the 11th European Workshop on Periodontology on effective
prevention of periodontal and peri-implant diseases. Journal of clinical
periodontology. 2015;42 Suppl 16:S5-11.
131.Slot DE, Van der Weijden FA. Group A. Initiator paper. Plaque control: home
remedies practiced in developing countries. Journal of the International
Academy of Periodontology. 2015;17(1 Suppl):4-15.
132.Salzer S, Slot DE, Van der Weijden FA, Dorfer CE. Efficacy of inter-dental
mechanical plaque control in managing gingivitis--a meta-review. Journal
of clinical periodontology. 2015;42 Suppl 16:S92-105.
133.Imamura T. The role of gingipains in the pathogenesis of periodontal disease.
Journal of periodontology. 2003;74(1):111-8.
Attawood Lertpimonchai References / 164
134.Ford PJ, Gemmell E, Hamlet SM, Hasan A, Walker PJ, West MJ, et al. Cross-
reactivity of GroEL antibodies with human heat shock protein 60 and
quantification of pathogens in atherosclerosis. Oral microbiology and
immunology. 2005;20(5):296-302.
135.Grundtman C, Kreutmayer SB, Almanzar G, Wick MC, Wick G. Heat shock protein
60 and immune inflammatory responses in atherosclerosis. Arteriosclerosis,
thrombosis, and vascular biology. 2011;31(5):960-8.
136.Bhatsange A, Patil SR. Assessment of periodontal health status in patients
undergoing renal dialysis: A descriptive, cross-sectional study. Journal of
Indian Society of Periodontology. 2012;16(1):37-42.
137.Borawski J, Wilczynska-Borawska M, Stokowska W, Mysliwiec M. The
periodontal status of pre-dialysis chronic kidney disease and maintenance
dialysis patients. Nephrology, dialysis, transplantation. 2007;22(2):457-64.
138.Bots CP, Poorterman JH, Brand HS, Kalsbeek H, van Amerongen BM, Veerman
EC, et al. The oral health status of dentate patients with chronic renal failure
undergoing dialysis therapy. Oral diseases. 2006;12(2):176-80.
139.Brito F, Almeida S, Figueredo CM, Bregman R, Suassuna JH, Fischer RG. Extent
and severity of chronic periodontitis in chronic kidney disease patients.
Journal of periodontal research. 2012;47(4):426-30.
140.Castillo A, Mesa F, Liebana J, Garcia-Martinez O, Ruiz S, Garcia-Valdecasas J, et
al. Periodontal and oral microbiological status of an adult population
undergoing haemodialysis: a cross-sectional study. Oral diseases.
2007;13(2):198-205.
141.Chamani G, Zarei MR, Radvar M, Rashidfarrokhi F, Razazpour F. Oral health status
of dialysis patients based on their renal dialysis history in Kerman, Iran. Oral
health and preventive dentistry. 2009;7(3):269-75.
142.Chhokra M, Manocha S, Dodwad V, Gupta U, Vaish S. Establishing an Association
between Renal Failure and Periodontal Health: A Cross Sectional Study.
Journal of clinical and diagnostic research. 2013;7(10):2348-50.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 165
143.Dag A, Firat ET, Kadiroglu AK, Kale E, Yilmaz ME. Significance of elevated
gingival crevicular fluid tumor necrosis factor-alpha and interleukin-8 levels
in chronic hemodialysis patients with periodontal disease. Journal of
periodontal research. 2010;45(4):445-50.
144.Gavaldá C, Bagán JV, Scully C, Silvestre FJ, Milián MA, Jiménez Y. Renal
hemodialysis patients: Oral, salivary, dental and periodontal findings in 105
adult cases. Oral diseases. 1999;5(4):299-302.
145.Limeres J, Garcez JF, Marinho JS, Loureiro A, Diniz M, Diz P. Early tooth loss in
end-stage renal disease patients on haemodialysis. Oral diseases.
2016;22(6):530-5.
146.Marakoglu I, Gursoy UK, Demirer S, Sezer H. Periodontal status of chronic renal
failure patients receiving hemodialysis. Yonsei medical journal.
2003;44(4):648-52.
147.Oshrain HI, Mender S, Mandel ID. Periodontal status of patients with reduced
immunocapacity. Journal of periodontology. 1979;50(4):185-8.
148.Parkar SM, Ajithkrishnan CG. Periodontal status in patients undergoing
hemodialysis. Indian journal of nephrology. 2012;22(4):246-50.
149.Rahman MM, Caglayan F, Rahman B. Periodontal health parameters in patients
with chronic renal failure and renal transplants receiving
immunosuppressive therapy. The Journal of Nihon University School of
Dentistry. 1992;34(4):265-72.
150.Sobrado Marinho JS, Tomas Carmona I, Loureiro A, Limeres Posse J, Garcia
Caballero L, Diz Dios P. Oral health status in patients with moderate-severe
and terminal renal failure. Medicina oral, patologia oral y cirugia bucal.
2007;12(4):E305-10.
151.Tadakamadla J, Kumar S, Mamatha GP. Comparative evaluation of oral health
status of chronic kidney disease (CKD) patients in various stages and healthy
controls. Special care in dentistry. 2014;34(3):122-6.
152.Teratani G, Awano S, Soh I, Yoshida A, Kinoshita N, Hamasaki T, et al. Oral health
in patients on haemodialysis for diabetic nephropathy and chronic
glomerulonephritis. Clinical oral investigations. 2013;17(2):483-9.
Attawood Lertpimonchai References / 166
153.Thorman R, Neovius M, Hylander B. Clinical findings in oral health during
progression of chronic kidney disease to end-stage renal disease in a
Swedish population. Scandinavian journal of urology and nephrology.
2009;43(2):154-9.
154.Tiwari V, Saxena V, Bhambhal A, Tiwari U, Singh A, Goud S. The oral health
status of patients with renal disease in central India: a preliminary study.
Journal of renal care. 2013;39(4):208-13.
155.Tollefsen T, Johansen JR. The periodontal status of prospective and renal transplant
patients. Comparison with systemically healthy subjects. Journal of
periodontal research. 1985;20(2):220-6.
156.Zhao D, Zhang S, Chen X, Liu W, Sun N, Guo Y, et al. Evaluation of periodontitis
and bone loss in patients undergoing hemodialysis. Journal of
periodontology. 2014;85(11):1515-20.
157.Taylor GW. Bidirectional interrelationships between diabetes and periodontal
diseases: an epidemiologic perspective. Annals of periodontology / the
American Academy of Periodontology. 2001;6(1):99-112.
158.Bascones-Martinez A, Gonzalez-Febles J, Sanz-Esporrin J. Diabetes and
periodontal disease. Review of the literature. American journal of dentistry.
2014;27(2):63-7.
159.Mealey BL, Oates TW. Diabetes mellitus and periodontal diseases. Journal of
periodontology. 2006;77(8):1289-303.
160.Tsai C, Hayes C, Taylor GW. Glycemic control of type 2 diabetes and severe
periodontal disease in the US adult population. Community dentistry and
oral epidemiology. 2002;30(3):182-92.
161.Taylor JJ, Preshaw PM, Lalla E. A review of the evidence for pathogenic
mechanisms that may link periodontitis and diabetes. Journal of clinical
periodontology. 2013;40 Suppl 14:S113-34.
162.Engebretson SP, Hey-Hadavi J, Ehrhardt FJ, Hsu D, Celenti RS, Grbic JT, et al.
Gingival crevicular fluid levels of interleukin-1beta and glycemic control in
patients with chronic periodontitis and type 2 diabetes. Journal of
periodontology. 2004;75(9):1203-8.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 167
163.Ribeiro FV, de Mendonca AC, Santos VR, Bastos MF, Figueiredo LC, Duarte PM.
Cytokines and bone-related factors in systemically healthy patients with
chronic periodontitis and patients with type 2 diabetes and chronic
periodontitis. Journal of periodontology. 2011;82(8):1187-96.
164.Ross JH, Hardy DC, Schuyler CA, Slate EH, Mize TW, Huang Y. Expression of
periodontal interleukin-6 protein is increased across patients with neither
periodontal disease nor diabetes, patients with periodontal disease alone and
patients with both diseases. Journal of periodontal research. 2010;45(5):688-
94.
165.Salvi GE, Collins JG, Yalda B, Arnold RR, Lang NP, Offenbacher S. Monocytic
TNF alpha secretion patterns in IDDM patients with periodontal diseases.
Journal of clinical periodontology. 1997;24(1):8-16.
166.Chapple IL, Matthews JB. The role of reactive oxygen and antioxidant species in
periodontal tissue destruction. Periodontology 2000. 2007;43:160-232.
167.Feuerer M, Herrero L, Cipolletta D, Naaz A, Wong J, Nayer A, et al. Lean, but not
obese, fat is enriched for a unique population of regulatory T cells that affect
metabolic parameters. Nature medicine. 2009;15(8):930-9.
168.Santos VR, Lima JA, Goncalves TE, Bastos MF, Figueiredo LC, Shibli JA, et al.
Receptor activator of nuclear factor-kappa B ligand/osteoprotegerin ratio in
sites of chronic periodontitis of subjects with poorly and well-controlled type
2 diabetes. Journal of periodontology. 2010;81(10):1455-65.
169.Emrich LJ, Shlossman M, Genco RJ. Periodontal disease in non-insulin-dependent
diabetes mellitus. Journal of periodontology. 1991;62(2):123-31.
170.Demmer RT, Desvarieux M, Holtfreter B, Jacobs DR, Jr., Wallaschofski H, Nauck
M, et al. Periodontal status and A1C change: longitudinal results from the
study of health in Pomerania (SHIP). Diabetes care. 2010;33(5):1037-43.
171.Teshome A, Yitayeh A. The effect of periodontal therapy on glycemic control and
fasting plasma glucose level in type 2 diabetic patients: systematic review
and meta-analysis. BMC Oral Health. 2016;17(1):31.
172.Jimenez M, Hu FB, Marino M, Li Y, Joshipura KJ. Type 2 diabetes mellitus and 20
year incidence of periodontitis and tooth loss. Diabetes research and clinical
practice. 2012;98(3):494-500.
Attawood Lertpimonchai References / 168
173.Molitch ME, DeFronzo RA, Franz MJ, Keane WF, Mogensen CE, Parving HH, et
al. Nephropathy in diabetes. Diabetes care. 2004;27 Suppl 1:S79-83.
174.Koye DN, Shaw JE, Reid CM, Atkins RC, Reutens AT, Magliano DJ. Incidence of
chronic kidney disease among people with diabetes: a systematic review of
observational studies. Diabetic medicine. 2017;34(7):887-901.
175.Shen Y, Cai R, Sun J, Dong X, Huang R, Tian S, et al. Diabetes mellitus as a risk
factor for incident chronic kidney disease and end-stage renal disease in
women compared with men: a systematic review and meta-analysis.
Endocrine. 2017;55(1):66-76.
176.Echouffo-Tcheugui JB, Narayan KM, Weisman D, Golden SH, Jaar BG.
Association between prediabetes and risk of chronic kidney disease: a
systematic review and meta-analysis. Diabetic medicine. 2016;33(12):1615-
24.
177.Mora-Fernandez C, Dominguez-Pimentel V, de Fuentes MM, Gorriz JL, Martinez-
Castelao A, Navarro-Gonzalez JF. Diabetic kidney disease: from physiology
to therapeutics. The Journal of physiology. 2014;592(18):3997-4012.
178.Hostetter TH, Rennke HG, Brenner BM. The case for intrarenal hypertension in the
initiation and progression of diabetic and other glomerulopathies. The
American journal of medicine. 1982;72(3):375-80.
179.Navarro-Gonzalez JF, Mora-Fernandez C. The role of inflammatory cytokines in
diabetic nephropathy. Journal of the American Society of Nephrology.
2008;19(3):433-42.
180.Vathesatogkit P, Woodward M, Tanomsup S, Ratanachaiwong W, Vanavanan S,
Yamwong S, et al. Cohort profile: the electricity generating authority of
Thailand study. International journal of epidemiology. 2012;41(2):359-65.
181.McFarlane P, Gilbert RE, MacCallum L, Senior P. Chronic kidney disease in
diabetes. Canadian journal of diabetes. 2013;37 Suppl 1:S129-36.
182.Report of the expert committee on the diagnosis and classification of diabetes
mellitus. Diabetes care. 2003;26 Suppl 1:S5-20.
183.1999 World Health Organization-International Society of Hypertension Guidelines
for the Management of Hypertension. Guidelines Subcommittee. Journal of
hypertension. 1999;17(2):151-83.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 169
184.Catapano AL, Graham I, De Backer G, Wiklund O, Chapman MJ, Drexel H, et al.
2016 ESC/EAS Guidelines for the Management of Dyslipidaemias.
European heart journal. 2016;37(39):2999-3058.
185.Appropriate body-mass index for Asian populations and its implications for policy
and intervention strategies. Lancet. 2004;363(9403):157-63.
186.Nishida C, Ko GT, Kumanyika S. Body fat distribution and noncommunicable
diseases in populations: overview of the 2008 WHO Expert Consultation on
Waist Circumference and Waist-Hip Ratio. European journal of clinical
nutrition. 2010;64(1):2-5.
187.Fox CS, Gona P, Larson MG, Selhub J, Tofler G, Hwang SJ, et al. A multi-marker
approach to predict incident CKD and microalbuminuria. Journal of the
American Society of Nephrology : JASN. 2010;21(12):2143-9.
188.O'Seaghdha CM, Lyass A, Massaro JM, Meigs JB, Coresh J, D'Agostino RB, Sr.,
et al. A risk score for chronic kidney disease in the general population. The
American journal of medicine. 2012;125(3):270-7.
189.Biering K, Hjollund NH, Frydenberg M. Using multiple imputation to deal with
missing data and attrition in longitudinal studies with repeated measures of
patient-reported outcomes. Clinical epidemiology. 2015;7:91-106.
190.Rubin DB, Schenker N. Multiple imputation in health-care databases: an overview
and some applications. Statistics in medicine. 1991;10(4):585-98.
191.White IR, Royston P, Wood AM. Multiple imputation using chained equations:
Issues and guidance for practice. Statistics in medicine. 2011;30(4):377-99.
192.StataCorp. Base Reference Manual: Stata Mulitple-Imputation Reference Manual.
Stata Statistical Software: Release 14. College Station, TX: Stata Press;
2015.
193.Baron RM, Kenny DA. The moderator-mediator variable distinction in social
psychological research: conceptual, strategic, and statistical considerations.
Journal of personality and social psychology. 1986;51(6):1173-82.
194.Iacobucci D. Mediation analysis and categorical variables: The final frontier.
Journal of Consumer Psychology. 2012;22(4):582-94.
195.MacKinnon DP, Fairchild AJ, Fritz MS. Mediation analysis. Annual review of
psychology. 2007;58:593-614.
Attawood Lertpimonchai References / 170
196.MacKinnon DP, Lockwood CM, Hoffman JM, West SG, Sheets V. A comparison
of methods to test mediation and other intervening variable effects.
Psychological methods. 2002;7(1):83-104.
197.Mackinnon DP, Warsi G, Dwyer JH. A Simulation Study of Mediated Effect
Measures. Multivariate behavioral research. 1995;30(1):41.
198.Gunzler D, Chen T, Wu P, Zhang H. Introduction to mediation analysis with
structural equation modeling. Shanghai archives of psychiatry.
2013;25(6):390-4.
199.VanderWeele TJ. Invited commentary: structural equation models and
epidemiologic analysis. American journal of epidemiology.
2012;176(7):608-12.
200.Ghosh D, Zhu Y, Coffman DL. Penalized regression procedures for variable
selection in the potential outcomes framework. Statistics in medicine.
2015;34(10):1645-58.
201.Koch B, Vock DM, Wolfson J. Covariate selection with group lasso and doubly
robust estimation of causal effects. Biometrics. 2017.
202.Kozak M, Azevedo RA. Does using stepwise variable selection to build sequential
path analysis models make sense? Physiologia plantarum. 2011;141(3):197-
200.
203.VanderWeele TJ, Shpitser I. A new criterion for confounder selection. Biometrics.
2011;67(4):1406-13.
204.Vansteelandt S, Bekaert M, Claeskens G. On model selection and model
misspecification in causal inference. Statistical methods in medical research.
2012;21(1):7-30.
205.Preacher KJ, Hayes AF. Asymptotic and resampling strategies for assessing and
comparing indirect effects in multiple mediator models. Behavior research
methods. 2008;40(3):879-91.
206.Imai K, Keele L, Tingley D. A general approach to causal mediation analysis.
Psychological methods. 2010;15(4):309-34.
207.Eke PI, Thornton-Evans GO, Wei L, Borgnakke WS, Dye BA. Accuracy of
NHANES periodontal examination protocols. Journal of dental research.
2010;89(11):1208-13.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 171
208.Wu AD, Zumbo BD. Understanding and Using Mediators and Moderators. Social
Indicators Research. 2007;87(3):367.
209.Enders CK. Analyzing longitudinal data with missing values. Rehabilitation
psychology. 2011;56(4):267-88.
210.Spratt M, Carpenter J, Sterne JA, Carlin JB, Heron J, Henderson J, et al. Strategies
for multiple imputation in longitudinal studies. American journal of
epidemiology. 2010;172(4):478-87.
211.Luo H, Wu B. Self-awareness of "Gum Disease" Among US Adults. Journal of
public health management and practice. 2017;23(2):e1-e7.
Attawood Lertpimonchai Appendices / 172
APPENDICES
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 173
APPENDIX A
MODIFIED NEWCASTLE-OTTAWA QUALITY ASSESSMENT
SCALE (COHORT STUDIES)
Selection (Maximum:����)
1) Representativeness of the Exposed Cohort (depends on the study design)
a) Truly representative of the general adults in the community/hospital �
b) Somewhat representative of the average general adults in the
community/hospital �
c) Selected group of users e.g. cohorts of DM / CKD patients
d) No description of the derivation of the cohort
2) Selection of the Non-Exposed Cohort
a) Drawn from the same community/hospital as the exposed cohort �
b) Drawn from a different source
c) No description of the derivation of the non-exposed cohort
3) Ascertainment of Exposure (Oral hygiene)
a) Full-mouth examination (Plaque and/or calculus) �
b) Partial-mouth or index teeth (Plaque and/or calculus) �
c) Record only the highest score of plaque and/or calculus (½�)
d) Self-report oral hygiene with ≥ 3 questions of OH assessment (¼�)
e) Subjective judgment or no description
4) Demonstration that Outcome of Interest Was Not Present at Start of Study
a) Yes �
b) No
Attawood Lertpimonchai Appendices / 174
Comparability (Maximum:��)
Confounders must be adjusted for in the analysis.
a) Study controls and/or adjusted for SMOKING in the analysis. �
b) Study controls for any additional one or more factor(s) (age, gender,
socio-status, DM, etcetera) �
Statements of no differences between groups or that differences were not
statistically significant, are not sufficient for establishing comparability.
Outcome (Maximum:������)
1) Assessment of outcome: Periodontitis assessment
a) Clinical measurement �
b) Radiography only or No description
2) Proper representative of periodontal status
a) Full-mouth examination AND ≥ 4 sites per tooth �
b) Partial-mouth examination OR < 4sites per tooth
3) Proper “Periodontitis” definition
a) Define “Periodontitis” with details of clinical attachment level (CAL)
with or without periodontal probing depth (PD) �
b) Define “Periodontitis” without clinical attachment level (CAL)
4) Validation of periodontal measurement
a) Mention about intra/inter-examiners calibration �
b) No statement
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 175
5) Was follow-up long enough for outcomes to occur: (1 year)
a) Yes �
b) No
6) Adequacy of follow up of cohorts
a) Complete follow up�
b) ≥ 80% follow up, and provided description of those lost or some
characteristics of BOTH groups. �
c) ≥ 80% follow up, without description of those lost (½�)
d) Follow up rate < 80% or No statement
Attawood Lertpimonchai Appendices / 176
Modified Newcastle-Ottawa Quality Assessment Scale
(Cross-sectional studies)
Selection (Maximum:��)
1) Representativeness of the samples
a) Clearly define the sampling method with random technique �
b) Consecutive samples from the study setting �
c) Selected group of samples e.g. volunteers
d) No description of the derivation of the samples
2) Ascertainment of Exposure (Oral hygiene)
a) Full-mouth examination (Plaque and/or calculus) �
b) Partial-mouth or index teeth (Plaque and/or calculus) �
c) Record only the highest score of plaque and/or calculus (½�)
d) Self-report oral hygiene with ≥ 3 questions of OH assessment (¼�)
e) Subjective judgment or no description
Comparability (Maximum:��)
Confounders must be adjusted for in the analysis.
a) Study controls and/or adjusted for SMOKING in the analysis. �
b) Study controls for any additional one or more factor(s) (age, gender,
socio-status, DM, etcetera) �
Statements of no differences between groups or that differences were not
statistically significant, are not sufficient for establishing comparability.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 177
Outcome (Maximum:� ���)
1) Assessment of outcome: Periodontitis assessment
a) Clinical measurement �
b) Radiography only or No description
2) Proper representative of periodontal status
a) Full-mouth examination AND ≥ 4 sites per tooth �
b) Partial-mouth examination OR < 4sites per tooth
3) Proper “Periodontitis” definition
a) Define “Periodontitis” with details of clinical attachment level (CAL) with
or without periodontal probing depth (PD) �
b) Define “Periodontitis” without clinical attachment level (CAL)
4) Validation of periodontal measurement
a) Mention about intra/inter-examiners calibration �
b) No statement
Attawood Lertpimonchai Appendices / 178
Modified Newcastle-Ottawa Quality Assessment Scale
(Case-control studies)
Selection (Maximum:�����)
1) Is the case (periodontitis) and control (non-periodontitis) definition adequate?
a) Define “Periodontitis” with details of clinical attachment level (CAL)
with or without periodontal probing depth (PD) �
b) Define “Periodontitis” without clinical attachment level (CAL)
2) Proper representative of periodontal status
a) Full-mouth examination AND ≥ 4 sites per tooth �
b) Partial-mouth examination OR < 4sites per tooth
3) Is the periodontal measurement valid?
a) Mention about intra/inter-examiners calibration �
b) No statement
4) Representativeness of the cases
a) Consecutive or obviously representative series of cases �
b) Potential for selection biases or not stated
5) Selection of Controls: Non-periodontitis
a) Consecutive selection or using any random sampling from the same
source and the same time periods as the cases �
b) Drawn from the same source and the time periods as the cases without
specified sampling method (½�)
c) No description
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 179
Comparability (Maximum:��)
1) Comparability of cases and controls based on the design or analysis. They must be
matched in the design and/or adjusted confounders in the analysis.
a) Study controls and/or adjusted for SMOKING in the analysis. �
b) Study controls for any additional one or more factor(s) (age, gender,
socio-status, DM, etcetera) �
Statements of no differences between groups or that differences were not
statistically significant, are not sufficient for establishing comparability.
Exposure (Maximum:���)
1) Ascertainment of exposure
a) Full-mouth examination (Plaque and/or calculus) �
b) Partial-mouth or index teeth (Plaque and/or calculus) �
c) Record only the highest score of plaque and/or calculus (½�)
d) Self-report oral hygiene
e) Subjective judgment or no description
2) Same method of ascertainment for cases and controls
a) Yes � b) No
3) Non-Response rate
a) Same rate for both groups �
b) Non-respondents described
c) Rate different and no designation
Attaw
ood Lertpimonchai
A
ppendices / 180
APPENDIX B
THE GRADE APPROACH
Pooling Study design
(No. of studies) Risk of
bias1 Inconsistency2 Indirectness Imprecision Other considerations (+/-)
Quality of Evidence
Fair and poor versus good OH
Observational (15): - Cohort (1) - Cross-sectional (14)
Not serious Not serious Not serious Not serious (+) Large effect3
(+) Dose-response gradient4
(+)(+)(+)( ) MODERATE
Brushing
Observational (10): - Cohort (1) - Case control (1) - Cross-sectional (8)
Not serious Serious Not serious Not serious (+)( )( )( ) VERY LOW
Interdental cleaning Observational (4): - Cross-sectional (4) Not serious Not serious Not serious Not serious (+)(+)( )( )
LOW
Dental visits
Observational (6): - Cohort (2) - Case control (1) - Cross-sectional (3)
Not serious Not serious Not serious Notserious (-) Publication bias5 (+)( )( )( ) VERY LOW
NOTE: 1 Risk of bias: The results of modified Newcastle-Ottawa Quality Assessment Scale were used for considered (see Table S2). Proportions of low, moderate and high risk of bias studies were 10:3:2, 6:4:0, 2:2:0 and 3:3:0 for fair and poor versus good OH, brushing, interdental cleaning and dental visit, respectively. From these, the numbers of low risk of bias studies were ≥ 50% for each pooling, hence, we graded them as “Not serious”. 2 Inconsistency: Heterogeneity of all poolings were moderate to high level and might be problematic except interdental cleaning. Fortunately, sources of heterogeneity could be identified properly. Results from pooled effect of OH in “community-based studies” and pooled effect of dental visit among studies that clearly defined a regular dental visit as least once a year were quite consistent. While, the heterogeneity of brushing still had presented as moderate to high level after exploring possible sources (Table S8). Therefore, only brushing was considered as “Serious” for inconsistency, then the quality of evidence was downgraded to very low level. 3, 4 Large effect & Dose-response gradient: The effect of OH on periodontitis was significant with the large pooled OR about 2 to 5 times. Moreover, it showed the dose-response relation, in other words, risk of periodontitis increased with worsening OH level (poor > fair > good OH). Therefore, the quality of evidence was upgraded from low to moderate level. 5 Publication bias: The publication bias was suspected in the pooled effect of dental visit from the Egger test (Table S9) and the contour enhanced-funnel plot (Figure S2-c). The quality of evidence was downgraded to very low level.
Fac. of Grad. Studies, Mahidol Univ. Ph.D. (Clinical Epidemiology) / 181
APPENDIX C
ETHICAL APPROVAL
Attawood Lertpimonchai Appendices / 182
BIOGRAPHY
NAME Mr. Attawood Lertpimonchai
DATE OF BIRTH 27 April 1983
PLACE OF BIRTH Bangkok, Thailand
INSTITUTIONS ATTENDED Doctor of Dental Surgery (D.D.S.), Faculty of
Dentistry, Chulalongkorn University (2001-2006),
Master of Science (M.Sc.) in
Periodontics, Department of
Periodontology, Faculty of Dentistry,
Chulalongkorn University (2009-2011)
SCHOLARSHIP RECEIVED Faculty of Dentistry, Chulalongkorn University
RESEARCH GRANTS Chulalongkorn University (Government Budget
Grant 2015) HOME ADDRESS 20 Prachanivej 2, Samakkee 58/18, Thasai,
Amphur Muang, Nonthaburi, 11000
EMPLOYMENT ADDRESS Faculty of Dentistry, Chulalongkorn University,
Henri Dunant, Pathum Wan, Bangkok, 10330
PHONE (66) 81-726-6969
E-MAIL [email protected]