Confidential: For Review Only - BMJ...Confidential: For Review Only 3 49 Conclusions: Visit-to-visit...

Confidential: For Review O

nly

Blood pressure variability and cardiovascular disease: A

systematic review and meta-analysis

Journal: BMJ

Manuscript ID BMJ.2015.029896

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the Author: 14-Oct-2015

Complete List of Authors: Stevens, Sarah; University of Oxford, Primary Care Health Sciences

Wood, Sally; University of Oxford, Primary Care Health Sciences

Koshiaris, Constantinos; University of Oxford, Primary Care Health

Sciences

Law, Kathryn; University of Oxford, Primary Care Health Sciences

Glasziou, Paul; Bond University, CREBP

Stevens, Richard; University of Oxford, Nuffield Dept Primary Care Health

Sciences

McManus, Richard; University of Oxford, Dept of Primary Care Health

Sciences

Keywords: blood pressure, variability, cardiovascular, review, ambulatory, mortality

https://mc.manuscriptcentral.com/bmj

BMJ


nly

1

Blood pressure variability and cardiovascular disease: A systematic review and 1

meta-analysis 2

3

Sarah L Stevens, Sally Wood, Constantinos Koshiaris, Kathryn Law, Paul Glasziou, Richard J Stevens, 4

Richard J McManus 5

Sarah L Stevens, Statistician, Nuffield Department of Primary Care Health Sciences, University of 6

Oxford, Radcliffe Observatory Quarter, Woodstock Road, Oxford, OX2 6GG, United Kingdom. Sally 7

Wood, General Practitioner, Nuffield Department of Primary Care Health Sciences, University of 8

Oxford, Radcliffe Observatory Quarter, Woodstock Road, Oxford, OX2 6GG, United Kingdom. 9

Constantinos Koshiaris, Statistician, Nuffield Department of Primary Care Health Sciences, University 10

of Oxford, Radcliffe Observatory Quarter, Woodstock Road, Oxford, OX2 6GG, United Kingdom. 11

Kathryn Law, General Practitioner, Nuffield Department of Primary Care Health Sciences, University 12

of Oxford, Radcliffe Observatory Quarter, Woodstock Road, Oxford, OX2 6GG, United Kingdom. Paul 13

Glasziou, Professor of Evidence-Based Medicine, Faculty of Health Sciences and Medicine, Bond 14

University, QLD 4229, Australia. Richard J Stevens, Associate Professor of Medical Statistics, Nuffield 15

Department of Primary Care Health Sciences, University of Oxford, Radcliffe Observatory Quarter, 16

Woodstock Road, Oxford, OX2 6GG, United Kingdom. Richard J McManus, Professor of Primary Care, 17

Nuffield Department of Primary Care Health Sciences, University of Oxford, Radcliffe Observatory 18

Quarter, Woodstock Road, Oxford, OX2 6GG, United Kingdom 19

Correspondence to: Richard J Stevens. Email: [email protected]. Phone: 01865 289355. 20

21

Word count: 3755 22

23

Page 1 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

2

Abstract 24

Objective: Variability in blood pressure (BP) has been increasingly recognized as an independent risk 25

factor for cardiovascular disease (CVD) over and above mean BP. Results from previous studies have 26

been conflicting, reflecting heterogeneity of approach and failure to consider important 27

confounders. We aimed to review studies assessing the prognostic utility of long, short and very 28

short term BP variability (measured through clinic, home and ambulatory monitoring), independent 29

of mean BP, for CVD events and mortality. 30

Data Sources: Medline, Embase, Cinahl and Web of Science were searched to 27th February 2014 for 31

full text articles in English. 32

Eligibility criteria for selecting studies: Prospective cohort studies or clinical trials in adults with at 33

least 2500 person-years follow-up were included. Studies in haemodialysis patients, where the 34

inherent nature of the condition may have a direct impact on BP variability were excluded. 35

Standardized hazard ratios were extracted and, if there was no evidence of confounding, combined 36

using random effects meta-analysis in primary analyses. Outcomes included all-cause and CVD 37

mortality and CVD events. Measures of variability included standard deviation, coefficient of 38

variation, variation independent of mean and average real variability but not night dipping or day-39

night variation. 40

Results: Thirty-one papers representing seventeen observational cohort studies and twelve clinical 41

trial cohorts were identified. Ambulatory, home and clinic monitoring were studied in fifteen, four 42

and twenty-one papers respectively (eight studied both clinic and ambulatory monitoring). Increased 43

visit-to-visit variability in clinic systolic BP was significantly associated with risk of all-cause mortality 44

(HR=1.15, 95% CI [1.07 to 1.23]), coronary heart disease (HR=1.10, 95% CI [1.00 to 1.21]) and stroke 45

(HR=1.15, 95% CI [1.02 to 1.30]). Increased day-by-day variability in home systolic BP was 46

significantly associated with all-cause mortality (HR=1.15, 95% CI [1.06 to 1.26]), as was increased 47

variability in daytime ambulatory systolic BP (HR=1.10 95% CI [1.06 to 1.15]). 48

Page 2 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

3

Conclusions: Visit-to-visit clinic BP variability is associated with cardiovascular and mortality 49

outcomes, over and above the effect of mean BP. The observed effect is similar in size to that for 50

cholesterol. The extent to which systems providing information to clinicians about BP variability 51

could alter or enhance the management of cardiovascular risk needs further investigation. 52

Systematic review registration: PROSPERO ID: CRD42014015695. 53

Page 3 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

4

Introduction 54

Blood pressure is a leading risk factor for cardiovascular disease.1,2 Most studies have used mean 55

blood pressure (BP) as the indicator of risk, measured in clinic or “out of office” settings.3–5 56

However, BP is physiologically dynamic and shows marked oscillations over the short and long term.6 57

Historically, such variability has been viewed as inhibiting accurate measurement of mean BP and as 58

a phenomenon to be overcome by improved monitoring.7 BP variability has also been recognised as 59

a potential risk factor in its own right for at least two decades.8,9 In 2010 an analysis of three cohort 60

studies and two randomized trials found that BP variability was a predictor of stroke and coronary 61

events in high risk patients.10 A companion paper proposed that the disparity between the effects of 62

beta blockers and calcium channel blockers on stroke risk could be explained by BP variability.11 63

However, understanding BP variability has been hampered by statistical and clinical methodological 64

issues. Some studies of variability have not adequately adjusted for mean BP and therefore fail to 65

rule out the possibility that high variability is a surrogate for hypertension.12,13 Others, in using 24-66

hour mean to adjust for daytime variability, may have turned high daytime variability into a 67

surrogate marker for nocturnal or 24-hour BP.14 Further studies have defined variability on the basis 68

of measurements taken during follow-up, but analysed it as a baseline risk factor15–17 which may 69

introduce problems of informative censoring or immortal time bias.18 Informative censoring occurs 70

when individuals are lost to follow-up for reasons related to the study. This could occur if individuals 71

with extreme or erratic BP readings were withdrawn from studies due to concerns over safety, in 72

particular in secondary analysis of trial data. Immortal time bias can occur if individuals are required 73

to have a certain number of repeat BP measurements in order to be included in analysis for 74

mortality outcomes. The time up until their qualifying measurement becomes 'immortal time' since 75

by definition these individuals could not have died prior to having the prerequisite number of BP 76

readings. 77

Page 4 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

5

Other studies fail to use consistent BP monitoring equipment over time, to define a consistent 78

protocol for BP measurement or to account for medication change, leaving doubt as to the source of 79

any observed variability.15,19,20 Measurement at different times of the day21 or year,22 in different 80

arms,23 or using inconsistent cuff sizes24 can affect accurate measurement, thereby inducing 81

variability. We reviewed prospective studies that quantified the longitudinal associations of (i) long 82

term variability measured through clinic BP monitoring, (ii) short term variability measured through 83

home BP monitoring and (iii) very short term variability measured through ambulatory BP 84

monitoring with cardiovascular events and mortality, independent of mean BP, in adults. We 85

focused our primary analysis on studies meeting pre-specified methodological criteria, so that any 86

apparent effect of BP variability was likely to be a true independent effect. 87

Methods 88

The protocol for this review has been published (http://www.crd.york.ac.uk/PROSPERO/, PROSPERO 89

ID: CRD42014015695). 90

91

Study selection 92

Medline, Embase, Cinahl and Web of Science were searched to 27th February 2014 for full text 93

articles in English, describing trials and prospective cohort studies, which assessed the association of 94

(i) long term variability measured through clinic BP monitoring, (ii) short term variability measured 95

through home BP monitoring and (iii) very short term variability measured through ambulatory BP 96

monitoring with cardiovascular outcomes in adults (see Supplement, Table e1 for full search 97

strategy). Studies included in a recent systematic review,25 were also screened. Titles and abstracts 98

were scrutinised by two reviewers (SW and KL/SS) with adjudication by a third (RM). 99

100

Page 5 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

6

Inclusion and exclusion criteria 101

Studies were included if they considered the following outcomes: (i) all-cause mortality, (ii) 102

cardiovascular events (including stroke, myocardial infarction, coronary heart disease, and heart 103

failure), (iii) cardiovascular mortality (including sudden death) or (iv) a combination thereof. Studies 104

that only assessed intermediate outcomes (e.g. “arterial intima-media” thickness) or studies 105

concerning “nocturnal dipping” or “day-night variation” were excluded, since these metrics of 106

variability have been considered previously.26 107

Studies in disease specific populations (e.g. hypertensive or diabetes patients) were included, except 108

studies in haemodialysis patients. This is due to the nature of haemodialysis treatment; during which 109

changes in blood pressure (intradialysis hypo- and hypertension27,28) are common and have been 110

shown to be associated with hospitalization and mortality.29,30 111

We included studies with at least 2500 person-years of follow-up where BP variability was assessed 112

in the long, short or very short term in clinics, at home or through ambulatory BP monitoring (ABPM) 113

respectively. Studies of clinic monitoring had to measure visit-to-visit variability over at least five 114

clinic visits.10 For home monitoring, at least twelve measurements over at least three days was 115

required.31 For ABPM, at least fourteen daytime readings were required.31 116

117

Data extraction 118

Study and patient characteristics data were extracted independently by two reviewers (SS/SW and 119

KL/RM), as were statistical results (SS and KC/RS) using pre-specified forms (Table e2). For each 120

paper, hazard ratios (HRs) for every variability measure and outcome were extracted. For studies 121

reporting multiple analyses, the HR from the analysis with the greatest adjustment for confounders 122

but containing only a single variability measure was extracted. Where required data were not 123

available, this was requested by contacting the study authors by email. 124

Page 6 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

7

125

Data analysis and statistical methods 126

To allow pooling across the variety of variability measures, extracted HRs were converted to 127

standardized hazard ratios, using a general method for regression models (Table e3).32 Briefly, a 128

standardized log-HR was calculated as the log-HR per unit of standardized exposure (the exposure 129

divided by its sample standard deviation). These standardized hazard ratios were pooled using a 130

random-effects meta-analysis, stratified by outcome. Separate analyses were performed for: (i) long 131

term variability measured through clinic BP monitoring, (ii) short term variability measured through 132

home BP monitoring and (iii) very short term variability measured through ambulatory BP 133

monitoring. Heterogeneity was assessed using the Chi-squared test and the I-squared statistic. 134

Where studies reported results for multiple different variability measures, HRs were selected for 135

inclusion according to the following hierarchy (preferred to least preferred): standard deviation (SD), 136

coefficient of variation (CV), variation independent of mean (VIM), average real variability (ARV), 137

standardised residual (SR), root successive variance (RSV) and other. Where HRs were calculated 138

using data from the same primary study but reported in different papers, the most recently 139

published HR was included. Hazard ratios for study subgroups were combined before inclusion. 140

Risk of bias was assessed using the QUIPS tool33 by two independent reviewers (SS and RS), with 141

adjudication by a third (RM). Information about other potential confounders, specific to studies of 142

variability were also extracted (Table e2, indicated by *). Consistency of BP measurement with 143

respect to device, cuff size, personnel and measurement is important to prevent inducing variability. 144

For example, inter-arm differences in BP will increase apparent BP variability and are also associated 145

with CVD outcomes.34 The impact of other potential confounders may be adjusted for during 146

analyses. Adjustment for mean BP is required since patients with higher absolute BP (an established 147

CVD risk factor) also have higher absolute BP variability. Similarly, adjustment for antihypertensive 148

Page 7 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

8

medication use is important since less adherent patients will have poorer outcomes and increased 149

BP variability.35 150

We decided (a-priori) to include in the main analyses only those hazard ratios which were correctly 151

adjusted for the appropriate mean BP level (e.g. adjusted for mean daytime BP if variability assessed 152

over daytime ABPM readings), where outcome ascertainment took place after the BP measurement 153

period and, for studies involving antihypertensive treatment, if at least 80% of patients were 154

adherent to medication, did not change treatment during the measurement period or if patients 155

were censored at the point of treatment change. When these requirements were not met or it was 156

unclear, hazard ratios were only included as part of sensitivity analyses. 157

Results 158

Searches identified 4800 references. Removal of duplicates and screening by two reviewers yielded 159

twenty-nine full-text articles for inclusion (Figure 1). A further two full-text articles were included 160

from a relevant systematic review .25 These thirty-one papers represented seventeen observational 161

cohort studies and twelve clinical trial cohorts. Of these twenty-nine distinct studies, we were able 162

to extract sufficient data for meta-analysis from twenty-seven studies; the authors of two studies 163

failed to respond to requests for additional data. 164

Six papers studied ABPM,13,14,36–39 four papers studied home monitoring,40–43 thirteen studied clinic 165

monitoring15–17,19,20,44–51 and eight papers studied both clinic and ABPM.10,12,52–57 The number of 166

participants per study ranged from 45752 to 58,22851 and follow-up ranged from 2514 person-years52 167

to 101,011 person-years.36 Full details of included papers are given in Table 1. 168

Risk of bias 169

Using QUIPS, a majority of studies were rated at moderate risk of bias for study participation, often 170

due to inclusion criteria based on BP readings and a potential for regression to the mean effects 171

(Table e4). Many studies did not use consistent measurement devices for BP or used mercury 172

Page 8 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

9

sphygmomanometers and varying personnel. Information about cuff size, reading arm and 173

personnel was missing in the majority of studies. Whilst ABPM studies generally considered daytime 174

and night-time variability separately, most long term studies failed to consider seasonal variation. 175

Eleven studies were at high risk of bias because the measurement period for BP variability was 176

confounded with the follow-up period or because they failed to report non-significant results. These 177

were excluded from our main analysis as were results from studies in cohorts with changing 178

antihypertensive regimens or inappropriate adjustment for mean BP. Overall, results from sixteen 179

analyses were eligible for inclusion in our pre-specified main analysis (Table 2). 180

181

i) Long term variability measured through clinic BP monitoring 182

Twenty-one papers reported results from twenty-four studies that measured BP in clinic. Five 183

studies were primarily concerned with variability in ambulatory BP and only measured clinic BP at 184

baseline12,54–56 or had fewer than five clinic measurements,57 and two studies,46,53 one of which 185

examined both clinic and between ABPM variability,53 did not present results in sufficient detail for 186

extraction. 187

Of the seventeen studies with extractable data, eleven were excluded from the main analysis (Table 188

2). A majority of studies reported results for stroke events, for which systolic BP variability was a 189

significant predictor (Figure 2; standardized HR = 1.15, 95% CI [1.02 to 1.30]; I-squared = 85.6%, 190

p<0.001). Note that the standardized hazard ratio reflects the hazard ratio per one standard 191

deviation increase in BP variability. The HRs from the UK-TIA,58 ASCOT-BPLA,59 and Dutch TIA60 trials 192

(which were determined from analyses in subgroups of patients with previous transient ischaemic 193

attack or stroke) were significantly larger than the HRs reported by other studies. Removal of these 194

studies from the meta-analysis in a post-hoc sensitivity analyses reduced heterogeneity and resulted 195

Page 9 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

10

in smaller but non-significantly different pooled HRs (HR=1.09, 95% CI [1.04 to 1.14]; I-196

squared=0.0%, p=0.573). 197

Three studies examined long term systolic BP variability and all-cause mortality (Figure e1). 198

Increased variability in systolic BP was significantly associated with increased mortality (standardized 199

HR = 1.15, 95% CI [1.07 to 1.23]). Heterogeneity between studies (I-squared = 76.3%, p=0.015), was 200

reduced after removal of a study in patients with previous stroke or vascular disease45 and did not 201

significantly alter results (HR = 1.10, 95% CI [ 1.07 to 1.14]; I-squared=0.0%, p=0.805). 202

Only one study examined systolic BP variability and coronary heart disease events, showing a 203

marginally significant relationship (HR = 1.10, 95% CI [1.00 to 1.21]) which was unchanged in when 204

adding further studies in sensitivity analyses (Figure e2). 205

Results for CVD mortality, CVD events and myocardial infarction showed similar results (HR = 1.18, 206

95% CI [1.09 to 1.28], HR = 1.18, 95% CI [1.07 to 1.30] and HR = 1.22, 95% CI [1.08 to 1.37] 207

respectively, Figures e3 to e5). Results for diastolic BP variability are also given in the online 208

supplement (Figures e6 to e9) and showed similar sized hazard ratios. 209

210

ii) Short term variability measured through home BP monitoring 211

Four papers reported results from two studies that measured BP at home. Results from all four 212

papers could be included in the main analyses, but a lack of data from distinct study populations and 213

heterogeneity in terms of the outcome studied meant it was only possible to perform formal meta-214

analysis for the all-cause mortality outcome. Systolic BP variability was a significant predictor of 215

death regardless of whether BP was measured in the morning, evening, or in both the morning and 216

evening (Figure e10). 217

Page 10 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

11

Results for other outcomes are given in the online supplement (Table e5) along with the diastolic BP 218

results (Figure e11, Table e6). Morning systolic BP variability was predictive of all outcomes except 219

CHD death. Evening systolic BP variability was predictive of CVD death, stroke death and cerebral 220

infarction death. Variability in combined morning and evening measurements was predictive of 221

stroke mortality, non-CVD mortality and cerebral infarction mortality. 222

223

iii) Very short term variability measured through ambulatory BP monitoring 224

Fourteen papers examined variability in ambulatory BP in eleven distinct studies. Of these, one study 225

considered visit-to-visit variability in ambulatory BP, and has been considered above as part of the 226

long-term variability results.53 A further five studies were excluded from the main analysis due to 227

potential confounding (Table 2). 228

A single study examined daytime systolic BP variability and all-cause mortality, showing a significant 229

association between increased daytime variability and risk of death (Figure e12; standardized HR per 230

= 1.10, 95% CI [1.04 to 1.17]). When results from a further study were included in sensitivity 231

analyses, results did not change. Meta-analysis of two studies found no association between 232

daytime ABPM variability and coronary heart disease events in the main analysis (Figure e13; HR = 233

1.04, 95% CI [0.93 to 1.17]) although this became significant when adding data from a further study 234

in sensitivity analysis (HR= 1.05, 95% CI [1.01 to 1.10]). Daytime BP variability was significantly 235

associated with increased risk of stroke (Figure e14; HR = 1.11, 95% CI [1.01 to 1.21]). 236

The effects of daytime systolic variability on CVD mortality and CVD events were similar (HR = 1.11, 237

95% CI [1.01 to 1.22] and HR = 1.04, 95% CI [0.98 to 1.11] respectively, Figures e15 and e16). Results 238

for night-time and 24-hour systolic ABPM and diastolic ABPM are also detailed in the online 239

supplement (Figures e17 to e31). 240

Page 11 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

12

Discussion 241

This review, which has systematically assessed the literature for the association of visit-to-visit, day-242

by-day or hour-by-hour blood pressure variability with cardiovascular outcomes and mortality, found 243

that visit-to-visit variability in clinic BP measurements is significantly associated with all-cause and 244

CVD mortality, CVD events, stroke and myocardial infarction, independent of mean BP. Data for BP 245

variability across days and hours were more limited but broadly supported an association of shorter 246

term BP variability with cardiovascular outcomes. Across all analyses (clinic, home and ABPM), the 247

hazard ratios for coronary heart disease events were smaller than those for stroke, suggesting that 248

the effect observed for CVD events – as with mean blood pressure - may be driven primarily by 249

cerebrovascular events. 250

This review is, to the authors’ knowledge, the first review to evaluate the effect of BP variability on 251

outcomes across all three modalities of measurement, hence allowing comparison between each 252

modality. Despite heterogeneity of approach in defining variability, this is the first review where 253

reported hazard ratios have been standardized wherever possible to allow pooling across variability 254

measures, whereas previous studies used single measures of variability only,25,61, namely standard 255

deviation or coefficient of variation. This has allowed us to pool results from greater numbers of 256

studies; for example, our meta-analysis for long term variability and stroke events includes data 257

from eleven studies in which variability was measured by standard deviation, standardised residual, 258

coefficient of variation or root mean squared error, compared to previous work including data from 259

six25 and two61 studies measuring variability by standard deviation alone. This review is also the first 260

with sufficient data to enable meta-analysis of the effect of very short term BP variability on 261

outcomes, which was a limitation of a previous review of ambulatory BP variability,26 which 262

considered the effect of night-dipping and day-night variation on outcomes. 263

The results for variability in clinic blood pressure are in line with those reported in similar 264

reviews,25,61 which may seem unsurprising given some overlap in included studies. However, 265

Page 12 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

13

previous reviews included studies in haemodialysis patients, where treatment can directly impact BP 266

variability,62 studies with few repeat measurements63 and those with short follow-up. Furthermore, 267

previous reviews did not distinguish results from studies that may be confounded e.g. by measuring 268

BP variability during follow-up.17,52 Results from our main analyses, including only studies that avoid 269

these potential sources of confounding, are therefore robust to such confounding. Hence this 270

review is the first to confirm that the apparent prognostic value of BP variability is a true prospective 271

association and can be demonstrated even in studies that avoid potential sources of confounding. 272

Studies in patients with history of cerebrovascular events reported the largest HRs, suggesting that 273

BP variability may have additional benefit for predicting recurrent events in this population.10 274

However, since results did not change substantially after removal of these studies from analysis, the 275

findings remain applicable to broader populations free from cerebrovascular disease. Some studies 276

also explicitly recruited hypertensive patients,15,39,44 where blood pressure variability could be 277

confounded by entry criteria (leading to regression-to-the-mean effects) and treatment.64 However, 278

such effects would diminish rather than exaggerate hazard ratios for variability, so this issue does 279

not affect our overall conclusions. 280

A limitation of this review is that we were unable to perform formal meta-analyses for many 281

outcomes with respect to short term, day-by-day BP variability due to a lack of data from unique 282

cohorts. A previous review was also similarly limited by paucity of data,65 despite including studies 283

which assessed target organ damage (in addition to hard outcomes) and studies which would not 284

meet our strict criteria. Meta-analyses were possible for ambulatory (hour-by-hour) BP variability 285

results, but were dominated by one very large study which included eleven cohorts.36 This study 286

reported results for average real variability in BP, which may arguably be a better measure than 287

others, e.g. SD, for accurately reflecting gradual variation in daytime BP. However results for other 288

measures of variability from other studies were similar and our pooled results were similar to those 289

obtained from the single large study. It was also not possible to determine if the effect of variability 290

Page 13 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

14

varied with the timing or number of readings.10 Despite these caveats, results supported an effect of 291

shorter term variability on cardiovascular outcome, and pooled hazard ratios were similar to those 292

observed for long-term variability (overlapping point estimates and confidence intervals). 293

In several analyses, there was significant heterogeneity between studies. In some cases this may be 294

attributable to outlying studies in specific populations e.g. patients with previous vascular disease. 295

Hazard ratios in some papers had to be converted from a categorical (e.g. from deciles10 or tertiles49) 296

to a continuous scale and this is reliant on normal distribution assumptions which may hold only 297

approximately and hence inflate heterogeneity. Conversely, it is well established that categorising 298

continuous variables can result in biased estimates of effects and increase the chance of a spurious 299

false positive result.66–68 However, not all converted hazard ratios were outliers,14,55 and we verified 300

our conversion method in simulated data (not shown). In cases of significant heterogeneity, removal 301

of outlier studies did not significantly alter the results. 302

In some cases, few studies contributed to main analyses and the validity of these meta-analyses is 303

open to debate. The addition of results from potentially confounded studies in sensitivity analyses 304

greatly increased the amount of available data but did not materially change the findings. Some 305

papers failed to report results in enough detail to allow full extraction of data, but this was mitigated 306

by direct contact with authors requesting the relevant additional data. Only two46,53 otherwise 307

eligible studies failed to contribute any quantitative data. 308

In general, there was poor reporting of study factors that may confound the relationship between BP 309

variability and outcomes. Although studies were excluded from the main analysis based on three of 310

the most important factors (pre-specified) it was not possible to do this for a wider range of factors, 311

in particular those regarding BP measurement, without greatly reducing the available data. Further 312

adjustment for confounders might be possible using individual patient data and would also facilitate 313

comparison of the prognostic utility of different variability measures, but was beyond the scope of 314

this review. 315

Page 14 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

15

The mechanism linking blood pressure variability to cardiovascular events is not well understood. 316

Short term blood pressure variability is affected by a number of factors. Behavioural, emotional and 317

postural influences on cardiovascular physiology are important along with cardiac rhythm.69,70 318

Arterial stiffness contributes to both short71,72 and long term variability.70,73,74 Meanwhile poor 319

control of blood pressure resulting in changes to, and titration of, antihypertensive medications also 320

affects variability.69 Use of certain classes of antihypertensive medication has also been linked with 321

increased visit-to-visit variability11 and may not be entirely explained by variation in adherence.35 322

The estimated standardized hazard ratio for the effect of long term BP variability on CVD mortality 323

was 1.18. For comparison, the effect of mean BP on CVD mortality reported in a previous meta-324

analysis3 corresponds to a standardised hazard ratio of approximately 1.7 (assuming between-325

person standard deviation of 15 mmHg). Note that the latter standardized hazard ratio for mean BP 326

is not adjusted for BP variability, whereas the former (for BP variability) is adjusted for mean, and 327

hence the hazard ratios for BP mean and variability cannot be directly compared. However, BP 328

variability does add additional prognostic information over and above the mean. 329

How does BP variability compare with other risk factors for CVD? A recent review75 found that the 330

standardized hazard ratio for increases in total cholesterol on CVD death was 1.19, and so variability 331

in BP has similar prognostic value to cholesterol measures. However, BP variability, unlike other risk 332

factors, is less easily assessed currently. This may be overcome though: visit-to-visit variability could 333

be automatically calculated by electronic health records, assisting physicians to take into account 334

both BP mean and variability concurrently when assessing cardiovascular risk. This may be 335

particularly important for patients with highly variable, but comparatively low mean BP or for whom 336

the traditional cardiovascular risk estimate lies close to treatment thresholds. Further work is 337

needed to determine the feasibility of different methods and the clinical impact of such additional 338

information, on subsequent risk management. 339

Page 15 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

16

In summary visit-to-visit clinic BP variability in adults is associated with cardiovascular and mortality 340

outcomes, over and above the effect of mean BP. The current evidence of an association between 341

day-by day (home) or hour-by-hour (ABPM) BP variability and outcomes is unclear and requires 342

further investigation in novel cohorts. 343

Page 16 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

17

Contributors 344

SS, SW, RS, RM and PG helped to design the study. SS, SW, KL and RM carried out article screening. 345

SS, SW, KL, KC, RS and RM extracted data. SS, KC and RS carried out statistical analyses. SS drafted 346

the original manuscript and all authors revised the paper. SS is the guarantor. 347

Competing interest statement 348

All authors have completed the Unified Competing Interest form 349

atwww.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and 350

declare: i) SS is funded by the National Institute for Health Research School for Primary Care 351

Research (NIHR SPCR), ii) RM has received grants and personal fees from Omron and grants from 352

Lloyds Pharmacy, outside the submitted work; PG reports grants from National Heart Foundation, 353

Australia, outside the submitted work, iii) no other relationships or activities that could appear to 354

have influenced the submitted work. The views expressed are those of the author(s) and not 355

necessarily those of the NHS, the NIHR or the Department of Health. 356

Copyright 357

The Corresponding Author has the right to grant on behalf of all authors and does grant on behalf of 358

all authors, a worldwide licence to the Publishers and its licensees in perpetuity, in all forms, formats 359

and media (whether known now or created in the future), to i) publish, reproduce, distribute, display 360

and store the Contribution, ii) translate the Contribution into other languages, create adaptations, 361

reprints, include within collections and create summaries, extracts and/or, abstracts of the 362

Contribution, iii) create any other derivative work(s) based on the Contribution, iv) to exploit all 363

subsidiary rights in the Contribution, v) the inclusion of electronic links from the Contribution to 364

third party material where-ever it may be located; and, vi) licence any third party to do any or all of 365

the above. 366

Page 17 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

18

References 367

1. Lim SS, Vos T, Flaxman AD, et al. A comparative risk assessment of burden of disease and 368

injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990-2010: a 369

systematic analysis for the Global Burden of Disease Study 2010. Lancet 370

2012;380(9859):2224-60. doi:10.1016/S0140-6736(12)61766-8. 371

2. Law MR, Morris JK, Wald NJ. Use of blood pressure lowering drugs in the prevention of 372

cardiovascular disease: Meta-analysis of 147 randomised trials in the context of expectations 373

from prospective epidemiological studies. BMJ 2009;338(7705):1245. 374

3. Lewington S, Clarke R, Qizilbash N, Peto R, Collins R. Age-specific relevance of usual blood 375

pressure to vascular mortality: A meta-analysis of individual data for one million adults in 61 376

prospective studies. Lancet 2002;360(9349):1903-1913. 377

4. Conen D, Bamberg F. Noninvasive 24-h ambulatory blood pressure and cardiovascular 378

disease: A systematic review and meta-analysis. J. Hypertens. 2008;26(7):1290-1299. 379

5. Ward AM, Takahashi O, Stevens R, Heneghan C. Home measurement of blood pressure and 380

cardiovascular disease: Systematic review and meta-analysis of prospective studies. J. 381

Hypertens. 2012;30(3):449-456. 382

6. Grassi G, Bombelli M, Brambilla G, Trevano FQ, Dell’oro R, Mancia G. Total Cardiovascular 383

Risk, Blood Pressure Variability and Adrenergic Overdrive in Hypertension: Evidence, 384

Mechanisms and Clinical Implications - Springer. Curr. Hypertens. Rep. 2012;14(4):333-8. 385

doi:10.1007/s11906-012-0273-8. 386

7. Takahashi O, Glasziou PP, Perera R, Shimbo T, Fukui T. Blood pressure re-screening for 387

healthy adults: What is the best measure and interval. J. Hum. Hypertens. 2012;26(9):540-388

546. 389

8. Frattola A, Parati G, Cuspidi C, Albini F, Mancia G. Prognostic value of 24-hour blood pressure 390

variability. J. Hypertens. 1993;11(10):1133-7. 391

Page 18 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

19

9. Grove JS, Reed DM, Yano K, Hwang L-J. Variability in systolic blood pressure - A risk factor for 392

coronary heart disease? Am. J. Epidemiol. 1997;145(9):771-776. 393

10. Rothwell PM, Howard SC, Dolan E, et al. Prognostic significance of visit-to-visit variability, 394

maximum systolic blood pressure, and episodic hypertension. Lancet 2010;375(9718):895-395

905. doi:10.1016/S0140-6736(10)60308-X. 396

11. Webb AJ, Fischer U, Mehta Z, Rothwell PM. Effects of antihypertensive-drug class on 397

interindividual variation in blood pressure and risk of stroke: a systematic review and meta-398

analysis. Lancet 2010;375(9718):906-915. 399

12. Bjorkland K, Lind L, Berglund L, Lithell H. Prognostic significance of 24-h ambulatory blood 400

pressure characteristics for cardiovascular morbidity in a population of elderly men. J. 401

Hypertens. 2004;22:1691-1697. 402

13. Gavish B, Ben-Dov IZ, Kark JD, Mekler J, Bursztyn M. The association of a simple blood 403

pressure-independent parameter derived from ambulatory blood pressure variability with 404

short-term mortality. Hypertens. Res. 2009;32(6):488-95. doi:10.1038/hr.2009.39. 405

14. Kikuya M, Hozawa a., Ohokubo T, et al. Prognostic Significance of Blood Pressure and Heart 406

Rate Variabilities : The Ohasama Study. Hypertension 2000;36(5):901-906. 407

doi:10.1161/01.HYP.36.5.901. 408

15. Kawai T, Ohishi M, Ito N, et al. Alteration of vascular function is an important factor in the 409

correlation between visit-to-visit blood pressure variability and cardiovascular disease. J. 410

Hypertens. 2013;31(7):1387-95; discussion 1395. doi:10.1097/HJH.0b013e328360f796. 411

16. Gao S, Hendrie HC, Wang C, et al. Redefined blood pressure variability measure and its 412

association with mortality in elderly primary care patients. Hypertension 2014;64(1):45-52. 413

doi:10.1161/HYPERTENSIONAHA.114.03576. 414

17. Lau K-K, Wong Y-K, Teo K-C, et al. Long-term prognostic implications of visit-to-visit blood 415

pressure variability in patients with ischemic stroke. Am. J. Hypertens. 2014;27(12):1486-94. 416

doi:10.1093/ajh/hpu070. 417

Page 19 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

20

18. Lévesque LE, Hanley JA, Kezouh A, Suissa S. Problem of immortal time bias in cohort studies: 418

example using statins for preventing progression of diabetes. BMJ 2010;340(april):b5087. 419

doi:10.1136/bmj.b5087. 420

19. Suchy-Dicey AM, Wallace ER, Mitchell SVE, et al. Blood pressure variability and the risk of all-421

cause mortality, incident myocardial infarction, and incident stroke in the cardiovascular 422

health study. Am. J. Hypertens. 2013;26(10):1210-7. doi:10.1093/ajh/hpt092. 423

20. Wei Y, Jin Z, Shen G, et al. Effects of intensive antihypertensive treatment on Chinese 424

hypertensive patients older than 70 years. J. Clin. Hypertens. (Greenwich). 2013;15(6):420-7. 425

doi:10.1111/jch.12094. 426

21. Pickering TG, Harshfield G a, Kleinert HD, Blank S, Laragh JH. Blood pressure during normal 427

daily activities, sleep, and exercise. Comparison of values in normal and hypertensive 428

subjects. JAMA 1982;247(7):992-996. doi:10.1001/jama.247.7.992. 429

22. Modesti PA, Morabito M, Bertolozzi I, et al. Weather-related changes in 24-hour blood 430

pressure profile: effects of age and implications for hypertension management. Hypertension 431

2006;47(2):155-61. doi:10.1161/01.HYP.0000199192.17126.d4. 432

23. Lane D, Beevers M, Barnes N, et al. Inter-arm differences in blood pressure: when are they 433

clinically significant? J. Hypertens. 2002;20(6):1089-95. 434

24. Linfors EW. Spurious Hypertension in the Obese Patient. Arch. Intern. Med. 1984;144(7):1482. 435

doi:10.1001/archinte.1984.00350190188032. 436

25. Diaz KM, Tanner RM, Falzon L, et al. Visit-to-Visit Variability of Blood Pressure and 437

Cardiovascular Disease and All-Cause Mortality: A Systematic Review and Meta-Analysis. 438

Hypertension 2014;64(5):965-82. doi:10.1161/HYPERTENSIONAHA.114.03903. 439

26. Taylor KS, Heneghan CJ, Stevens RJ, Adams EC, Nunan D, Ward A. Heterogeneity of Prognostic 440

Studies of 24-hour Blood Pressure Variability: Systematic Review and Meta-Analysis. PLoS 441

One 2015;10(5):e0126375. doi:10.1371/journal.pone.0126375. 442

Page 20 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

21

27. Daugirdas JT. Pathophysiology of dialysis hypotension: An update. Am. J. Kidney Dis. 443

2001;38(4):S11-S17. doi:10.1053/ajkd.2001.28090. 444

28. Inrig JK. Intradialytic hypertension: a less-recognized cardiovascular complication of 445

hemodialysis. Am. J. Kidney Dis. 2010;55(3):580-9. doi:10.1053/j.ajkd.2009.08.013. 446

29. Inrig JK, Oddone EZ, Hasselblad V, et al. Association of intradialytic blood pressure changes 447

with hospitalization and mortality rates in prevalent ESRD patients. Kidney Int. 448

2007;71(5):454-61. doi:10.1038/sj.ki.5002077. 449

30. Inrig JK, Patel UD, Toto RD, Szczech LA. Association of blood pressure increases during 450

hemodialysis with 2-year mortality in incident hemodialysis patients: a secondary analysis of 451

the Dialysis Morbidity and Mortality Wave 2 Study. Am. J. Kidney Dis. 2009;54(5):881-90. 452

doi:10.1053/j.ajkd.2009.05.012. 453

31. National Institute for Health and Care Excellence. Hypertension: Clinical Management of 454

Primary Hypertension in Adults (NICE CG127). National Institute for Health and Care 455

Excellence; 2011. 456

32. Bring J. How to Standardize Regression Coefficients. Am. Stat. 1994;48(3):209-213. 457

33. Hayden JA, van der Windt DA, Cartwright JL, Côté P, Bombardier C. Assessing bias in studies 458

of prognostic factors. Ann. Intern. Med. 2013;158(4):280-6. doi:10.7326/0003-4819-158-4-459

201302190-00009. 460

34. Cao K, Xu J, Shangguan Q, et al. Association of an inter-arm systolic blood pressure difference 461

with all-cause and cardiovascular mortality: An updated meta-analysis of cohort studies. Int. 462

J. Cardiol. 2015;189:211-9. doi:10.1016/j.ijcard.2015.04.079. 463

35. Muntner P, Levitan EB, Joyce C, et al. Association Between Antihypertensive Medication 464

Adherence and Visit-to-Visit Variability of Blood Pressure. J. Clin. Hypertens. 2013;15(2):112-465

117. 466

Page 21 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

22

36. Hansen TW, Thijs L, Li Y, et al. Prognostic value of reading-to-reading blood pressure 467

variability over 24 hours in 8938 subjects from 11 populations. Hypertension 468

2010;55(4):1049-57. doi:10.1161/HYPERTENSIONAHA.109.140798. 469

37. Mancia G, Bombelli M, Facchetti R, et al. Long-term prognostic value of blood pressure 470

variability in the general population: results of the Pressioni Arteriose Monitorate e Loro 471

Associazioni Study. Hypertension 2007;49(6):1265-70. 472


38. Mena LJ, Maestre GE, Hansen TW, et al. How many measurements are needed to estimate 474

blood pressure variability without loss of prognostic information? Am. J. Hypertens. 475

2014;27(1):46-55. doi:10.1093/ajh/hpt142. 476

39. Verdecchia P, Angeli F, Gattobigio R, Rapicetta C, Reboldi G. Impact of blood pressure 477

variability on cardiac and cerebrovascular complications in hypertension. Am. J. Hypertens. 478

2007;20(2):154-61. doi:10.1016/j.amjhyper.2006.07.017. 479

40. Asayama K, Kikuya M, Schutte R, et al. Home blood pressure variability as cardiovascular risk 480

factor in the population of ohasama. Hypertension 2013;61(1):61-69. 481


41. Hashimoto T, Kikuya M, Ohkubo T, et al. Home blood pressure level, blood pressure 483

variability, smoking, and stroke risk in Japanese men: the Ohasama study. Am. J. Hypertens. 484

2012;25(8):883-891. 485

42. Johansson JK, Niiranen TJ, Puukka PJ, Jula AM. Prognostic value of the variability in home-486

measured blood pressure and heart rate: the Finn-Home Study. Hypertension 487

2012;59(2):212-8. doi:10.1161/HYPERTENSIONAHA.111.178657. 488

43. Kikuya M, Ohkubo T, Metoki H, et al. Day-by-day variability of blood pressure and heart rate 489

at home as a novel predictor of prognosis: the Ohasama study. Hypertension 490

2008;52(6):1045-1050. 491

Page 22 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

23

44. Carr MJ, Bao Y, Pan J, Cruickshank K, McNamee R. The predictive ability of blood pressure in 492

elderly trial patients. J. Hypertens. 2012;30(9):1725-1733. 493

45. Hata J, Arima H, Rothwell PM, et al. Effects of visit-to-visit variability in systolic blood pressure 494

on macrovascular and microvascular complications in patients with type 2 diabetes mellitus: 495

The advance trial. Circulation 2013;128(12):1325-1334. 496

doi:10.1161/CIRCULATIONAHA.113.002717. 497

46. Hsieh Y-T, Tu S-T, Cho T-J, Chang S-J, Chen J-F, Hsieh M-C. Visit-to-visit variability in blood 498

pressure strongly predicts all-cause mortality in patients with type 2 diabetes: A 5·5-year 499

prospective analysis. Eur. J. Clin. Invest. 2012;42(3):245-253. 500

47. Kostis JB, Sedjro JE, Cabrera J, et al. Visit-to-visit blood pressure variability and cardiovascular 501

death in the Systolic Hypertension in the Elderly Program. J. Clin. Hypertens. (Greenwich). 502

2014;16(1):34-40. doi:10.1111/jch.12230. 503

48. Mallamaci F, Minutolo R, Leonardis D, et al. Long-term visit-to-visit office blood pressure 504

variability increases the risk of adverse cardiovascular outcomes in patients with chronic 505

kidney disease. Kidney Int. 2013;84(2):381-9. doi:10.1038/ki.2013.132. 506

49. McMullan CJ, Bakris GL, Phillips R a, Forman JP. Association of BP variability with mortality 507

among African Americans with CKD. Clin. J. Am. Soc. Nephrol. 2013;8(5):731-8. 508

doi:10.2215/CJN.10131012. 509

50. Poortvliet RKE, Ford I, Lloyd SM, et al. Blood Pressure Variability and Cardiovascular Risk in 510

the PROspective Study of Pravastatin in the Elderly at Risk (PROSPER). PLoS One 2012;7(12). 511

51. Shimbo D, Newman JD, Aragaki AK, et al. Association between annual visit-to-visit blood 512

pressure variability and stroke in postmenopausal women: Data from the Women’s health 513

initiative. Hypertension 2012;60(3):625-630. 514

52. Eguchi K, Hoshide S, Schwartz JE, Shimada K, Kario K. Visit-to-visit and ambulatory blood 515

pressure variability as predictors of incident cardiovascular events in patients with 516

hypertension. Am. J. Hypertens. 2012;25(9):962-968. 517

Page 23 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

24

53. Mancia G, Facchetti R, Parati G, Zanchetti A. Visit-to-visit blood pressure variability in the 518

European Lacidipine Study on Atherosclerosis: Methodological aspects and effects of 519

antihypertensive treatment. J. Hypertens. 2012;30(6):1241-1251. 520

54. Pierdomenico SD, Lapenna D, Di Tommaso R, et al. Blood pressure variability and 521

cardiovascular risk in treated hypertensive patients. Am. J. Hypertens. 2006;19(10):991-997. 522

55. Pierdomenico SD, Di Nicola M, Esposito AL, et al. Prognostic value of different indices of 523

blood pressure variability in hypertensive patients. Am. J. Hypertens. 2009;22(8):842-7. 524

doi:10.1038/ajh.2009.103. 525

56. Pierdomenico SD, Lapenna D, Bucci A, et al. Blood pressure variability and prognosis in 526

uncomplicated mild hypertension. Am. Heart J. 2005;149(5):934-8. 527

doi:10.1016/j.ahj.2004.06.029. 528

57. Pringle E, Phillips C, Thijs L, et al. Systolic blood pressure variability as a risk factor for stroke 529

and cardiovascular mortality in the elderly hypertensive population. J. Hypertens. 530

2003;21(12):2251-2257. 531

58. Farrell B, Godwin J, Richards S, Warlow C. The United Kingdom transient ischaemic attack 532

(UK-TIA) aspirin trial: final results. J. Neurol. Neurosurg. Psychiatry 1991;54(12):1044-54. 533

59. Poulter NR, Wedel H, Dahlöf B, et al. Role of blood pressure and other variables in the 534

differential cardiovascular event rates noted in the Anglo-Scandinavian Cardiac Outcomes 535

Trial-Blood Pressure Lowering Arm (ASCOT-BPLA). Lancet 2005;366(9489):907-13. 536

doi:10.1016/S0140-6736(05)67186-3. 537

60. UK-TIA Study Group. The United Kingdom transient ischaemic attack ( UK-TIA ) aspirin trial : 538

final results. J. Neurol. Neurosurg. Psychiatry 1991;54:1044-1054. 539

61. Tai C, Sun Y, Dai N, et al. Prognostic significance of visit-to-visit systolic blood pressure 540

variability: a meta-analysis of 77,299 patients. J. Clin. Hypertens. (Greenwich). 541

2015;17(2):107-15. doi:10.1111/jch.12484. 542

Page 24 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

25

62. Selvarajah V, Pasea L, Ojha S, Wilkinson IB, Tomlinson LA. Pre-dialysis systolic blood pressure-543

variability is independently associated with all-cause mortality in incident haemodialysis 544

patients. PLoS One 2014;9(1):e86514. doi:10.1371/journal.pone.0086514. 545

63. Chowdhury EK, Owen A, Krum H, Wing LMH, Nelson MR, Reid CM. Systolic blood pressure 546

variability is an important predictor of cardiovascular outcomes in elderly hypertensive 547

patients. J. Hypertens. 2014;32(3):525-33. doi:10.1097/HJH.0000000000000028. 548

64. Yudkin PL, Stratton IM. How to deal with regression to the mean in intervention studies. 549

Lancet 1996;347(8996):241-243. doi:10.1016/S0140-6736(96)90410-9. 550

65. Stergiou GS, Ntineri A, Kollias A, Ohkubo T, Imai Y, Parati G. Blood pressure variability 551

assessed by home measurements: a systematic review. Hypertens. Res. 2014;37(6):565-572. 552

doi:10.1038/hr.2014.2. 553

66. Royston P, Altman DG, Sauerbrei W. Dichotomizing continuous predictors in multiple 554

regression: a bad idea. Stat. Med. 2006;25(1):127-41. doi:10.1002/sim.2331. 555

67. Altman DG, Lausen B, Sauerbrei W, Schumacher M. Dangers of using “optimal” cutpoints in 556

the evaluation of prognostic factors. J. Natl. Cancer Inst. 1994;86(11):829-35. 557

68. Austin PC, Brunner LJ. Inflation of the type I error rate when a continuous confounding 558

variable is categorized in logistic regression analyses. Stat. Med. 2004;23(7):1159-78. 559

doi:10.1002/sim.1687. 560

69. Parati G, Ochoa JE, Lombardi C, Bilo G. Assessment and management of blood-pressure 561

variability. Nat. Rev. Cardiol. 2013;10(3):143-55. doi:10.1038/nrcardio.2013.1. 562

70. Mancia G. Short- and long-term blood pressure variability: Present and future. Hypertension 563

2012;60(2):512-517. 564

71. García-García Á, García-Ortiz L, Recio-Rodríguez JI, et al. Relationship of 24-h blood pressure 565

variability with vascular structure and function in hypertensive patients. Blood Press. Monit. 566

2013;18(2):101-6. doi:10.1097/MBP.0b013e32835ebc58. 567

Page 25 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

26

72. Schillaci G, Bilo G, Pucci G, et al. Relationship between short-term blood pressure variability 568

and large-artery stiffness in human hypertension: findings from 2 large databases. 569

Hypertension 2012;60(2):369-77. doi:10.1161/HYPERTENSIONAHA.112.197491. 570

73. Nagai M, Hoshide S, Ishikawa J, Shimada K, Kario K. Visit-to-visit blood pressure variations: 571

new independent determinants for carotid artery measures in the elderly at high risk of 572

cardiovascular disease. J. Am. Soc. Hypertens. 2011;5(3):184-92. 573

doi:10.1016/j.jash.2011.03.001. 574

74. Okada H, Fukui M, Tanaka M, et al. Visit-to-visit variability in systolic blood pressure is 575

correlated with diabetic nephropathy and atherosclerosis in patients with type 2 diabetes. 576

Atherosclerosis 2012;220(1):155-9. doi:10.1016/j.atherosclerosis.2011.10.033. 577

75. Perera R, Clarke P, Farmer A, et al. Optimal strategies for monitoring lipid levels in patients at 578

risk or with cardiovascular disease: best marker for monitoring and cost-effectiveness of 579

different monitoring frequencies. Health Technol. Assess. (Rockv). 2015;In press. 580

76. THE ESPS Group. THE EUROPEAN STROKE PREVENTION STUDY (ESPS). Lancet 581

1987;330(8572):1351-1354. doi:10.1016/S0140-6736(87)91254-2. 582

77. Dutch TIA Trial Study Group. A comparison of two doses of aspirin (30 mg vs. 283 mg a day) in 583

patients after a transient ischemic attack or minor ischemic stroke. The Dutch TIA Trial Study 584

Group. N. Engl. J. Med. 1991;325(18):1261-6. doi:10.1056/NEJM199110313251801. 585

586

Page 26 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

27

Figure 1: Study screening flowchart 587

588

589

Page 27 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only28

Table 1: Included study characteristics

Paper,

Year

Study Modality Population N Measure of Variability Follow-up Antihypertensive

medication

Outcome Measure

Asayama,

201240

Ohasama

(observational)

Home Men and women >35 years old, at

home during working hours, not

hospitalised, not incapacitated

2421 VIM, ARV, MMD (min-max

difference)

12 years (median) 27.1% Total mortality, CVD

mortality, stroke

events

Bjorklund,

200412

Uppsala

Longitudinal Study

of Adult Men

(observational)

ABPM and

clinic

(baseline)

70 year old men 872 SD 9.5 years (max,

mean 6.6 +/- 2.1

years)

30.0% CVD events

Carr,

201244

MRC Elderly trial Clinic Hypertensive patients with a mean

systolic BP at 160-209mmHg and

diastolic BP < 115mmHg at entry

4396 SR and RSV (Root successive

variance = SR divided by BP at

baseline)

5.8 years (mean) Yes (trial of

antihypertensive

medication)

Stroke events and

CHD events

Eguchi,

201252

Observational ABPM and

clinic

Asymptomatic patients aged 33-88

attending general internal medicine

clinics at three institutes in Japan, for

the evaluation and management of

hypertension

457 SD 66 months

(mean)

55.6% Hard CVD and all

CVD events

Gao,

201416

Observational Clinic US Primary care patients aged 60+ years

approached for a depression screening

study 1991-1993

2906 RMSE (root mean squared

error)

12.9 years

(median)

89.7% All-cause mortality,

CHD events or

stroke events

Gavish,

200913

Observational ABPM Non pregnant, greater than 16 years

old, good quality ABPM

3433 SD, CV and day/ night SD ratio 7.6 years (mean,

max 16 years)

59% All-cause mortality

Hansen,

201036

IDACO

(observational)

ABPM Multiple different populations in ABPM

database. Mean age 53 years.

8939 SD, ARV, mean of day and

night SD

11.3 years

(median)


CVD mortality, CVD

events, cardiac,

coronary events and

stroke events

Page 28 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


medication

Outcome Measure

Hashimoto,

201241

Ohasama

(observational)

Home Japanese men. Mean age 58.6 years 902 SD, CV 13.1 years

(median)


CVD mortality,

stroke mortality, MI

mortality, stroke

events

Hata,

201345

ADVANCE trial Clinic Patients aged 55+ years with type 2

diabetes and history of major macro- or

microvascular disease

8811 SD, CV 2.4 years

(median)

69% (at baseline but

trial of

antihypertensive

medication)

All-cause mortality,

CVD mortality

stroke, CVD events,

stroke events, MI

events

Hsieh,

201246

Observational Clinic Patients with type 2 diabetes visiting the

diabetic clinic in the Metabolism

Division at Changhua Christian hospital

Sept 2003-Apr 2005

2161 SD, CV 66.7 months

(mean)

80% All-cause and CVD

mortality

Johansson,

201242

Health 2000 study

(observational)

Home Finnish adults aged 45-74 years 1866 SD, CV, ARV and SD of day-

night difference and first-

second difference

7.8 years (mean) 30.6% All-cause mortality

and CVD events

Kawai,

201315

NOAH study

(observational)

Clinic Non-Invasive Atherosclerotic Evaluation

in Hypertension (NOAH) study.

Outpatients diagnosed with essential

hypertension sequentially recruited

between January 1998 and June 2004 at

Osaka University Medical Hospital

485 SD 7.59 years (mean) 47.3% CVD events

Kikuya,

200014

Ohasama

(observational)

ABPM Japanese general population > 40 years

(mean age 61.7 years, 40% men)

1542 SD 8.5 years (mean) 30.9% CVD mortality

Page 29 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


medication

Outcome Measure

Kikuya,

200843

Ohasama

(observational)

Home Japanese population 35-96 years (mean

age = 59.3 +/- 12.3 years). 60.5%

women.

2455 SD, CV 11.9 years

(median)


CVD mortality, Non-

CVD mortality,

stoke mortality,

CHD mortality, MI

mortality

Kostis,

201447

SHEP (trial) Clinic Systolic hypertension in the elderly

program. Average age 72.

57% women and 15% black. USA.

4736 VIM, rSSR (sum of squared

deviations between average

and trend predicted BP),

VABS2 (variance of absolute

difference between

successive daily BP (VABS2)

17 years (max) 100% CVD mortality

Lau,

201417

Observational Clinic Ischaemic stroke patients without atrial

fibrillation, Hong Kong. Average age 71

years.

632 CV 76+/- 18 months

(mean)

80% All-cause mortality,

CVD mortality, non-

fatal stroke and

nonfatal acute

coronary syndrome

Mallamaci,

201348

Observational Clinic Italians aged 18-75 with CKD stages 3

and 4, recruited in renal clinics from Oct

2005 to Nov 2007

1618 SD, CV 37 months

(median)

94% CVD events

Mancia,

200737

PAMELA

(observational)

ABPM 2012 randomly selected individuals in

Milan aged 25-74 years

2012 SD 148 months (max) Not stated All-cause mortality

and CVD mortality

Mancia,

201253

ELSA trial ABPM and

clinic

European Lacidipine Study on

Atherosclerosis trial which randomized

antihypertensive treatment for 4 years

to mildly or moderately hypertensive

patients at relatively low cardiovascular

risk

1521 (clinic

analysis)

1264 (ABPM

analysis)

SD, CV 4 years (max) Yes (trial of

antihypertensive

medication)

CVD events

Page 30 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


medication

Outcome Measure

McMullan,

201349

AASK trial Clinic African American Study of Kidney

Disease trial. African Americans with

hypertensive nephropathy. Mean age 55

years, 62% men.

908 SD 75 months (max,

median 52

months).

Yes (trial of

antihypertensive

medication)


CVD mortality, CVD

events

Mena,

201438

IDACO

(observational)

ABPM Discovery data: Copenhagen cohort

subjects equally distributed among sex

and age groups with complete ABPM

(41, 51, 61, and 71 years). Test data:

IDACO subjects 18+ years with at least

10 daytime

readings, 5 night-time readings, and 48

readings over 24 hours who were not

included in the discovery data.

1254 (discovery

data)

5353 (test data)

ARV 10.2 years

(median in test

data)

21.3% (test data) All-cause mortality,

CVD mortality, CHD

mortality, CVD

events, CHD events,

stroke events

Pierdomenico,

200556

Abruzzo, Italy

(observational)

ABPM and

clinic

(baseline)

Uncomplicated mild clinic hypertensive 1088 SD 4.74 years (mean) 87% at follow-up CVD events

Pierdomenico,

200654

Abruzzo, Italy

(observational)

ABPM and

clinic

(baseline)

Hypertensive patients undergoing ABPM

in Italy

1472 SD 4.88 years (mean) 100% CVD events

Pierdomenico,

200955

Abruzzo, Italy

(observational)

ABPM and

clinic

(baseline)

Hypertensive patients age 40+ years

who were referred for an outpatient

evaluation for hypertension in Italy

1280 SD, ARV 4.75 +/- 1.8 years

(mean, range 0.2-

7.5 years)

57.0% CVD events

Poortvliet,

201250

PROSPER trial Clinic Men and women aged 70–82 years from

Scotland, Ireland or Netherlands with

either pre-existing vascular disease

(coronary, cerebral, or peripheral) or at

high risk due to smoking, hypertension

or diabetes.

4819 (short-term

follow-up)

1808 (long-term

follow-up)

SD and CV 3 years (max,

mean 2.3 years)

or 9.3 years (max,

mean 7.1 years)

in Scottish sub-

group:

62.6% (short term

follow-up) 59.6% (long

term follow-up)


CVD mortality,

stroke events, CHD

events

Page 31 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


medication

Outcome Measure

Pringle,

200357

Syst-Eur trial ABPM and

clinic

(placebo

run-in)

Elderly patients (60+ years) with

isolated systolic hypertension

744 SD 4.4 years

(median)

100% (384 on active

treatment in trial)

CVD mortality,

stroke events, CHD

events.

Rothwell,

201010

UK-TIA trial Clinic Patients with history of TIA, mean age

60.3 years.58

2006 SD, CV, VIM. 3.3 years

(median, max

6.67 years).

27%58 Stroke events

Rothwell,

201010

European Stroke

Prevention Study

(ESPS-1) (trial)

Clinic Patients with recent cerebrovascular

event76

2500 SD, CV, VIM. 2 years (max). Not stated Stroke events

Rothwell,

201010

Dutch TIA trial. Clinic Patients with recent TIA or stroke77 3150 SD, CV, VIM. 2.6 years (mean) 42%77 Stroke events

Rothwell,

201010

ASCOT BPLA trial

(subset of

stroke/TIA patients)

Clinic Patients with previous TIA or stroke 2011 SD, CV, VIM. 5 years (median) 100%59 Stroke, CVD and

CHD events

Rothwell,

201010

ASCOT BPLA trial

ABPM sub-study

(subset of stroke/

TIA patients)

ABPM Patients with previous TIA or stroke 1905 SD, CV, VIM. 5 years (median) 100%59 Stroke, CVD and

CHD events

Shimbo,

201251

Women’s Health

Initiative

(observational)

Clinic Post-menopausal women 58228 SD 5.4 years

(median)

Not stated Stroke events

Suchy-Dicey,

201319

Cardiovascular

Health Study

(observational)

Clinic Subjects not using antihypertensive

medication during a 5 year baseline

period or those using the same

antihypertensive regimen during that

period. Mean age 71 years. 95% white.

2548

(1570 no

antihypertensive

medication,

978 with

medication)

SR (standardised residual) 9.9 years (mean) 38.4% All-cause mortality,

MI events, stroke

events.

Page 32 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


medication

Outcome Measure

Verdecchia,

200739

PIUMA study

(observational)

ABPM Initially untreated subjects with

essential hypertension. Mean age = 51

years. 47% women.

2649 SD 6 years (mean,

max 16 years)

Untreated initially -

subsequent

antihypertensive use

recorded

CHD events

Wei,

201320

PROBE trial Clinic Hypertensive Chinese patients aged 70+

years

724 SD 4 years (mean) 100% CVD events

Page 33 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Table 2: Study and analysis characteristics that may confound the relationship between blood pressure variability and outcomes

Paper,

Year

Study design characteristics Potential confounders Main analysis

(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Asayama,

2012

Not

mentioned

Not

mentioned

Omron HEM

401C

Yes (home) Adjustment

for mean

BP

Not

mentioned but

inclusion not

based on BP

readings

Morning and

evening assessed

separately

Analysis

adjusted for

medication

use and

morning

measurements

taken before

medication

No

(observational,

followed

through

registries and

questionnaires)

Measurement

only over 4

weeks

First

reading in

morning/

evening

over 28

days

Yes

Bjorklund,

2004

Not

mentioned

Right arm

for office.

Non-

dominant

arm for

ABPM

Mercury for

office.

Accutracker 2

equipment for

ABPM

Yes (ABPM) No

adjustment

Not

mentioned but

inclusion not

based on BP

readings

Day and night

variability

assessed

separately

No but ABPM No

(observational,

followed

through

registries)

Yes (ABPM) Single

ABPM

reading

every 20

mins.

No

adjustment

for mean BP

Page 34 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Carr,

2012

Not

mentioned

Not

mentioned

Hawksley

Random Zero

Sphyg.

No -

physician

and nurse

Adjustment

for mean

BP

Trends for BP

modelled to

avoid

regression

dilution bias

No diurnal

variation given

as clinic BP only.

Seasonal pattern

present in the

raw data for

mean BP.

No. Nested

case-control

study but not

matched on

medication

adherence.

Adherence

low.

No (case-

control study

nested within

trial but

adherence and

cross-over not

considered)

Measure of

variability

based on

measures

preceding

event

Mean of

two

readings

fortnightly,

then

monthly,

then 3

monthly.

Treatment

adherence

low

Eguchi,

2012

Not

mentioned

Not

mentioned

Mercury sphyg

for office.

Yes for

ABPM.

Unclear for

clinic.

Adjusted

for clinic BP

only

Not

mentioned.

Inclusion

based on

referral due to

actual or

suspected

hypertension.

Not mentioned No No

(observational,

followed

through

registries and

interviews)

Unclear for

clinic, but yes

for ABPM

Clinic:

mean of

2nd and 3rd

reading

every

month.

ABPM:

every 30

mins.

Follow-up

and

measurement

confounded.

Adjusted

ABPM

analysis for

clinic mean

Page 35 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Gao,

2014

Not

mentioned

Not

mentioned

Not mentioned

(routine data

source)

Unclear

(routine

data)

Yes (trend

adjusted)

BP not an

entry criteria

Not mentioned No (routine

data)

No

(observational.

Information

about

medications

collected but

not included in

time-

dependent

modelling)

No (routine

data)

Measure

in record

from

routine

visits

Follow-up

and

measurement

confounded.

Page 36 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Gavish,

2009

Selected

according to

arm

circumference

.

Non

dominant

Pre 1999:

Accutracker 2.

Post 1999:

SpaceLabs 90207

Yes (ABPM) Analysis of

CV and

adjusted for

mean

arterial

pressure

Not

mentioned.

Patients

referred as

part of usual

care.

Only looked at

SD over 24 hours

No but ABPM No

(observational,

followed

through

registries)

Yes (ABPM) Single

ABPM

reading

every 20

mins.

Incorrect

adjustment

for mean BP

Page 37 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Hansen,

2010

Not

mentioned in

paper -

different

cohorts

Not

mentioned

in paper -

different

cohorts

Variable

depending on

cohort:

Accutracker 2 in

Uppsala, Space

Labs

90202/90207,

Nippon Colin,

ABPM-630 in

other cohorts

Yes (ABPM) Adjusted

for mean

BP

Not recruited

on basis of BP

Taken account of

via SDdn

measure

No but ABPM No

(observational,

followed

through

registries)

Yes (ABPM) Single

ABPM

reading

every 30

mins.

Yes

Page 38 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Hashimoto,

2012

Not

mentioned

Not

mentioned

Oscillometric

device: HEM

401C

Yes (home) Adjusted

for mean

BP

Not

mentioned but

inclusion not

based on BP.

No - all readings

within one hour

of waking.

No No

(observational,

followed

through

registries and

questionnaires)

Measurements

only over 4

weeks

Single

reading in

morning/

evening

over 28

days

Yes

Hata,

2013

Not

mentioned

Not

mentioned

Automated

sphyg (Omron

HEM 705CP)

Unclear Adjusted

for mean

BP

Not recruited

based on BP.

Measurements

at

randomization

not included in

assessment of

variability

Not mentioned No (50%

changed

regimen but

results similar

for those that

did/ did not

change)

No (trial, but

treatment

during follow-

up could be

modified)

Yes Mean of 2

readings at

3, 4, 6, 12,

18 and 24

months.

Yes

Page 39 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Hsieh,

2012

Measurement

taken "after

cuff size was

corrected"

Not

specified

Standardized

automated

sphyg, Omron

HEM-1000

Not stated Adjusted

for mean

BP

Inclusion not

based on

baseline BP

measures

Not mentioned No No

(observational,

followed

through

registries)

Unclear Mean of

two

readings

every 2-6

months

No

extractable

data

Johansson,

2012

Not

mentioned

Not

mentioned

Omrom HEM

722C

Yes (home)

Adjusted

for mean

BP

Not recruited

based on BP.

SD of

morning/evening

difference

calculated

No No

(observational,

followed

through

registries)

Measurement

only over 7

days

2 readings

in morning

and

evening

for 7 days

Yes

Page 40 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Kawai,

2013

Not

mentioned

Not

mentioned

Electronic sphyg

(HEM 705IT/

HEM-711)

Unclear Yes Not

mentioned but

recruited

based on

outpatient

diagnosis of BP

not values

Not mentioned Yes - 18

people

excluded if

treatment

altered in

measurement

period.

No

(observational,

followed

through

questionnaires)

Unclear Mean of 2

readings

every 1-2

months

Follow-up

and

measurement

confounded.

Kikuya,

2000

Not specified Not

specified

ABPM630

Nippon Colin

Yes (ABPM) Adjusted

for 24-hour

mean BP

Not

mentioned but

not recruited

based on BP.

Daytime and

night time

variability

analysed

separately

No but ABPM No

(observational,

followed

through

registries)

Yes (ABPM) Single

ABPM

every 30

minutes

Incorrect

adjustment

for mean BP

Page 41 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Kikuya,

2008

Not

mentioned

Not

mentioned

HEM 401C,

Omron

Healthcare

Yes (home) Analysis by

CV and

adjusted for

mean BP

Not

mentioned

explicitly

No - single home

reading taken in

the morning

No but

analysis

adjusted for

medication

use

No

(observational,

followed

through

registries)

Measurement

only over 4

weeks

One

reading

morning/

evening

for 28 days

Yes

Kostis,

2014

Not

mentioned

Not

mentioned

Random zero

sphyg

Unclear

(trained

personnel)

Adjusted

for mean

BP

Not

mentioned

Not mentioned No but

adherence

high and

analysis

results

restricted to

those patients

who did not

cross over

similar

No (trial, but

extended

observational

follow-up after

end of trial)

Yes 2 readings

at months

0, 1,2,3

and then 3

monthly

Yes

Page 42 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Lau,

2014

Not

mentioned

Not

mentioned

Validated

automated BP

machine (GE,

DINAMAP,

PRO100,Fairfield,

CT)

Unclear Adjusted

for mean

BP

Not

mentioned but

not recruited

based on BP

Not mentioned No but

medication

adjusted for in

analysis

No

(observational,

followed

through

registries)

Unclear Mean of

2nd and 3rd

readings

every 3-4

months

Follow-up

and

measurement

confounded.

Mallamaci,

2013

Not

mentioned but

according to

ESH guidelines

Not

mentioned

but

according

to ESH

guidelines

Mercury sphyg No

nephrologist

OR nurse

Adjusted

for mean

BP

Not

mentioned but

recruited

based on eGFR

not BP

Not mentioned No No

(observational)

Unclear Mean of 3

readings

twice a

year for 3

years

Follow-up

and

measurement

confounded.

Page 43 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Mancia,

2007

Not

mentioned

Not

mentioned

Spacelabs 90207 Yes (ABPM) Adjusted

for mean

24-hour BP

Not

mentioned but

not recruited

based on BP

SD for day and

night analysed

separately

No but ABPM No

(observational,

followed

through

registries)

Yes (ABPM) Single

ABPM

reading

every 20

mins

Adjusted for

24-hour

mean in day

and night

analysis.

Mancia,

2012

Not

mentioned

Not

mentioned

Mercury sphyg Yes (ABPM)

but unclear

for clinic

(trained

physicians)

Mean BP

included in

models

Not

mentioned

Not mentioned Yes – only

clinic

measurements

after titration

period

included in

analysis

Yes (only

measurements

after titration

included,

measurement

and follow-up

period

coincides)

No Clinic:

mean of 3

readings

every 6

months.

ABPM:

mean of

24-hour

ABPM

yearly

No

extractable

data

Page 44 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

McMullan,

2013

Not

mentioned

Not

mentioned

Random zero

sphyg

Not stated Adjusted

for mean

BP

Not

mentioned but

only post-

baseline

measures used

in the analysis

No - clinic BP Measurement

period after 3

month

titration

period but

adherence low

(71%)

No (only

compliance

during first 12

months

considered)

Yes Mean of

2nd and 3rd

readings at

4, 6, 8, 10

and 12

months

Adherence

low

Page 45 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Mena,

2014

"Appropriate

cuff size"

Not

mentioned

Accutracker II in

Uppsala or

SpaceLabs 90202

and 90207,

Nippon Colin,

and ABPM 630

in the other

cohorts.

Yes (ABPM) Adjusted

for 24-hour

BP mean.

Not

mentioned but

not recruited

on BP

Only 24-hour

variability

considered

No but ABPM No

(observational,

followed

through

registries)

Yes (ABPM) Single

ABPM

reading

every 30

mins.

Yes

Page 46 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Pierdomenico

,

2005

Not

mentioned

Not

mentioned

Clinic: mercury

sphyg, ABPM:

Spacelabs 90208

Yes (ABPM) Adjusted

for mean

BP

Not

mentioned but

recruited

based on

referral and

ABPM

occurred after

baseline

Not mentioned No but ABPM No

(observational,

follow-up

through family

doctors)

Yes (ABPM) Single

ABPM

reading

every 15

mins

No

extractable

data

Page 47 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Pierdomenico

,

2006

"Appropriate

cuff size"

Not

mentioned

Clinic: mercury

sphyg, ABPM:

Spacelabs 90207

Yes (ABPM) Adjusted

for mean

BP

Not

mentioned but

recruited

based on

referral and

ABPM

occurred after

baseline

Analysis split by

day/ night.

No but ABPM No

(observational,

follow-up

through family

doctors)

Yes (ABPM) Single

ABPM

reading

every 15

mins

Yes

Page 48 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Pierdomenico

,

2009

"Appropriate

cuff size"

Not

mentioned

Clinic: mercury

sphyg, ABPM:

Spacelabs 90207

Yes (ABPM) Adjusted

for mean

BP

Not

mentioned but

recruited

based on

referral and

ABPM

occurred after

baseline

Analysis split by

daytime and

night-time

ABPM.

No but ABPM No

(observational,

follow-up

through family

doctors)

Yes (ABPM) Single

ABPM

reading

every 15

mins

Yes

Poortvliet,

2012

Not

mentioned

Not

mentioned

Omron M4 Unclear

(trained

nurses)

Adjusted

for mean

BP

Not

mentioned but

not recruited

on BP

Not mentioned No but trial of

statins

No (long-term

follow-up

observational)

Yes, patients

with events

during

measurement

period

excluded

Once

every 3

months

Yes

Page 49 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Pringle,

2003

"If arm

circumference

exceeded

31cm, larger

cuffs with

35x15 cm

bladder were

used"

Non-

dominant,

or arm

giving

highest

reading if

difference

of more

than

10mmHg

systolic

between

arms

Clinic: mercury

sphyg. ABPM:

Validated

monitors incl.

Spacelabs 90202

and 90207

Yes (ABPM) Adjusted

for mean

BP

Not

mentioned,

recruitment

based on clinic

BP not ABPM

Day-time and

night time

considered

separately and

combined

No but ABPM

and analysis

split by

treatment

group.

No (addition of

treatments

possible)

Yes (ABPM)

Single

ABPM

reading

every 30

minutes

Yes

Page 50 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Rothwell,

2010: UK TIA

Not

mentioned

Not

mentioned

Mercury sphyg. Not

mentioned

Adjusted

for mean

BP

Not

mentioned but

inclusion

based on

previous

stroke history

Not mentioned RCT of aspirin

only.

No (RCT of

aspirin only

and follow-up

part of trial but

about a third

hypertensive

so could have

other

treatment)

Patients with

events in

measurement

period

excluded

Single

reading

every 4

months

Yes

Page 51 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Rothwell,

2010: ESPS1

Not

mentioned

Both arms.

Mean of

left and

right

recorded.

Mercury sphyg. Not

mentioned

Adjusted

for mean

BP

Not

mentioned but

inclusion

based on stoke

history

Not mentioned Analysis of

placebo group

only due to

vasoactive

medication.

Yes, analysis of

placebo group

only and

follow-up part

of trial. Less

than 5% had

hypertension

and would be

eligible for

treatment)

Unclear Mean of 1

reading in

each arm

every 3

months

Follow-up

and

measurement

confounded.

Rothwell,

2010: Dutch

TIA

Not

mentioned

Not

mentioned

Mercury sphyg. Not

mentioned

Adjustment

for mean

BP

Not

mentioned but

inclusion

based on stoke

history

Not mentioned No. RCT of

aspirin with

atenolol

subgroup.

Analysis

excluding

subgroup

similar.

No, follow-up

part of trial

and subgroup

including

added atenolol

Unclear Single

reading

every 4

months

Follow-up

and

measurement

confounded.

Page 52 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Rothwell,

2010: ASCOT

BPLA

Not

mentioned

Not

mentioned

Clinic: Omron

HEM-705CP

Unclear. Adjusted

for mean

BP

Not

mentioned

explicitly

Not mentioned No. RCT of

hypertensive

medication

with titration.

No, RCT with

titration.

Follow-up as

part of trial

Unclear Clinic:

mean of

2nd and 3rd

readings

every 6

months.

ABPM:

Every 30

mins.

Follow-up

and

measurement

confounded

for clinic

analysis.

Rothwell,

2010: ASCOT

BPLA ABPM

sub-study

Not

mentioned

Not

mentioned

ABPM:

SpaceLabs

90207.

Yes (ABPM), Unclear Not

mentioned

explicitly

Day and night

considered

separately

Yes (ABPM) No, RCT with

titration.

Follow-up as

part of trial

Yes (ABPM) Clinic:

mean of

2nd and 3rd

readings

every 6

months.

ABPM:

Every 30

mins.

Adjustment

for mean

unclear

Page 53 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Shimbo,

2012

Appropriate

cuff size used

at each visit by

arm

measurement

Right Mercury sphyg Unclear

(certified

staff)

Adjusted

for mean

BP

Not recruited

based on BP

No - clinic BP No but

medication

included as

time

dependent

covariate in

models and

adjusted for

regression line

trend

Yes. Anti-

hypertensive

medication

was considered

as a

time

dependent

covariate.

Covariates

were updated

through follow

up (and post-

measurement)

whilst patient

remained at

risk.

Yes Mean of 2

readings

annually.

Yes

Page 54 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Suchy-Dicey,

2013

Not

mentioned

Right arm Initial

measurement:

random zero

sphyg.

Remaining:

mercury sphyg

Not stated Adjusted

for mean

BP

Not

mentioned but

not recruited

on BP

Not mentioned Yes - users of

changing

medications

not included in

analysis

No (only

restricted

analysis to

users of

changing

medications

during

measurement

not follow-up)

Yes Mean of

2nd and 3rd

readings, 5

times

annually.

Yes

Verdecchia,

2007

Not

mentioned

Not

mentioned

ABPM:

SpaceLabs 5200,

90202, 90207

Yes (ABPM) Adjusted

for mean

BP

Not

mentioned.

Recruitment

based on

office BP not

ABPM.

Daytime and

night-time

variability

considered

separately

No but ABPM No

(observational

follow-up

through family

doctors)

Yes (ABPM) Single

ABPM

every 15

minutes

Yes

Page 55 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


Paper,

Year


(yes or

reason for

exclusion)

Cuff Size Reading

Arm

Device used Same

person

taking

readings

Adjustment

for

appropriate

mean

Regression to

mean

considered

Diurnal or

Seasonal

variation

considered

Medication

change during

measurement

period limited

Medication

change during

follow-up

limited

Measurement

before follow-

up

Definition

of single

reading

Wei,

2013

Not

mentioned

Right arm Manual sphyg Unclear Adjusted

for mean

BP

Not

mentioned

explicitly, but

entry to trial

based on two

readings on

two days

Not mentioned No – after one

year many had

increase in

number of

medications

No (trial but

medication

change

possible)

Unclear Mean of 2

readings

every 6

months

Follow-up

and

measurement

confounded.

Page 56 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

57

Figure 2

Page 57 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

1

Online supplement

Table e1: Search strategy

CINAHL (EBSCOHost)

S24 S22 OR S23 Limiters - Clinical Queries: Prognosis - Specificity

S23 S20 OR S21 Limiters - Clinical Queries: Prognosis - High Sensitivity

S22 S20 OR S21

S21 TI ( ((blood pressure or bp or sbp or dbp) N5 (variabilit* or variation*)) ) OR AB ( ((blood pressure or bp or sbp or dbp) N5

(variabilit* or variation*)) )

S20 S3 AND S6 AND S19

S19 S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 OR S15 OR S16 OR S17 OR S18

S18 TI within subject* OR AB within subject*

S17 TI ( (dipping or dipper* or nondipping or nondipper* or non-dipping or non-dipper*) ) OR AB ( (dipping or dipper* or nondipping or nondipper* or non-dipping or non-dipper*) )

S16 TI ( (((daytime or day-time or diurnal) N5 (blood pressure or bp or sbp or sbp)) and ((night-time or nocturnal) N5 (blood pressure or bp or sbp or sbp))) ) OR AB ( (((daytime or day-time or diurnal) N5 (blood pressure or bp or sbp or sbp)) and

((night-time or nocturnal) N5 (blood pressure or bp or sbp or sbp))) )

S15 TI ( ((daytime or day-time or diurnal) N5 (night-time or nocturnal)) ) OR AB ( ((daytime or day-time or diurnal) N5 (night-

time or nocturnal)) )

S14 TI repeat* measure* OR AB repeat* measure*

S13 TI "measure* to measure*" OR AB "measure* to measure*"

S12 TI ( ("day to day" or "day by day") ) OR AB ( ("day to day" or "day by day") )

S11 TI ( "between day" OR "within day" ) OR AB ( "between day" OR "within day" )

S10 TI "visit to visit" OR AB "visit to visit"

S9 TI ( ((between or within) N3 visit*) ) OR AB ( ((between or within) N3 visit*) )

S8 TI variation*

S7 TI ( variability or variabilities ) OR AB ( variability or variabilities )

S6 S4 OR S5

S5 TI ( blood pressure or bp or sbp or dbp ) OR AB ( blood pressure or bp or sbp or dbp )

S4 (MH "Blood Pressure") OR (MH "Blood Pressure Determination")

S3 S1 OR S2

S2 TI ( hypertensive* or hypertension* or antihypertens* or anti-hypertens* ) OR AB ( hypertensive* or hypertension* or antihypertens* or anti-hypertens* )

S1 (MH "Hypertension+")

Embase (OvidSP)

1 *hypertension/

2 (hypertensive* or hypertension* or antihypertens* or anti-hypertens*).ti,ab.

3 1 or 2

4 *Blood Pressure/

5 exp *blood pressure measurement/

6 (blood pressure or bp or sbp or dbp).ti,ab.

7 4 or 5 or 6

8 (variability or variabilities).ti,ab.

9 variation?.ti.

10 ((between or within) adj3 visit?).ti,ab.

11 "visit to visit".ti,ab.

12 ((between or within) adj day?).ti,ab.

13 ("day to day" or "day by day").ti,ab.

14 "measure* to measure*".ti,ab.

15 "reading? to reading?".ti,ab.

16 repeat* measure*.ti,ab.

17 ((daytime or day-time or diurnal) adj5 (night-time or nocturnal)).ti,ab.

18 (((daytime or day-time or diurnal) adj5 (blood pressure or bp or sbp or sbp)) and ((night-time or nocturnal) adj5 (blood pressure or bp

Page 58 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

2

or sbp or sbp))).ti,ab.

19 (dipping or dipper? or nondipping or nondipper? or non-dipping or non-dipper?).ti,ab.

20 within subject?.ti,ab.

21 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20

22 3 and 7 and 21

23 blood pressure variability/

24 3 and 23

25 ((blood pressure or bp or sbp or dbp) adj5 (variabilit* or variation?)).ti,ab.

26 22 or 24 or 25

27 ((exp animal/ or exp vertebrate/ or exp invertebrate/) not human/) or animal experiment/ or animal model/ or animal tissue/ or animal

cell/ or nonhuman/

28 26 not 27

29 follow up.mp. or ep.fs. or prognos*.tw.

30 28 and 29

31 (prognos* or survival).tw.

32 28 and 31

33 30 or 32

Medline (OvidSP)

1 exp Hypertension/

2 (hypertensive* or hypertension* or antihypertens* or anti-hypertens*).ti,ab.

3 1 or 2

4 *Blood Pressure/

5 *Blood Pressure Determination/

6 (blood pressure or bp or sbp or dbp).ti,ab.

7 4 or 5 or 6

8 (variability or variabilities).ti,ab.

9 variation?.ti.

10 ((between or within) adj3 visit?).ti,ab.

11 "visit to visit".ti,ab.

12 ((between or within) adj day?).ti,ab.

13 ("day to day" or "day by day").ti,ab.

14 "measure* to measure*".ti,ab.

15 "reading? to reading?".ti,ab.

16 repeat* measure*.ti,ab.

17 ((daytime or day-time or diurnal) adj5 (night-time or nocturnal)).ti,ab.

18 (((daytime or day-time or diurnal) adj5 (blood pressure or bp or sbp or sbp)) and ((night-time or nocturnal) adj5 (blood pressure or bp

or sbp or sbp))).ti,ab.

19 (dipping or dipper? or nondipping or nondipper? or non-dipping or non-dipper?).ti,ab.

20 within subject?.ti,ab.

21 8 or 9 or 10 or 11 or 12 or 13 or 14 or 15 or 16 or 17 or 18 or 19 or 20

22 3 and 7 and 21

23 ((blood pressure or bp or sbp or dbp) adj5 (variabilit* or variation?)).ti,ab.

24 22 or 23

25 exp animal/ not human/

26 24 not 25

27 incidence.sh. or exp mortality/ or follow-up studies.sh. or prognos*.tw. or predict*.tw. or course*.tw.

28 26 and 27

29 (prognos* or first episode or cohort).tw.

30 26 and 29

31 28 or 30

Page 59 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

3

Web of Science:

# 17 #16 AND #15

# 16 Topic=(prognos* OR cohort* OR incidence OR mortality OR "follow-up" OR predict OR course)

# 15 #14 AND #13

# 14 Topic=(hypertens* OR antihypertens* OR anti-hypertens*)

# 13 #12 OR #2 OR #1

# 12 #11 AND #3

# 11 #10 OR #9 OR #8 OR #5 OR #4

# 10 Topic=("within subject*")

# 9 Topic=(dipping or dipper* or nondipping or nondipper* or non-dipping or non-dipper*)

# 8 #7 AND #6

# 7 Topic=(nightime NEAR/5 ("blood pressure" or bp or sbp or dbp)) OR Topic=(nocturnal NEAR/5 ("blood pressure" or bp or sbp

or dbp))

# 6 Topic=(daytime NEAR/5 ("blood pressure" or bp or sbp or dbp)) OR Topic=(day-time NEAR/5 ("blood pressure" or bp or sbp or

dbp)) OR Topic=(diurnal NEAR/5 ("blood pressure" or bp or sbp or dbp))

# 5 Topic=((daytime NEAR/5 (night-time or nocturnal))) OR Topic=((day-time NEAR/5 (night-time or nocturnal))) OR

Topic=((diurnal NEAR/5 (night-time or nocturnal)))

# 4 Topic=(between NEAR/3 visit*) OR Topic=(within NEAR/3 visit*) OR Topic=("between day*" OR "within day*") OR

Topic=("day to day" OR "day by day") OR Topic=("measure to measure") OR Topic=("measurement to measurement") OR

Topic=("repeat measur*" OR "repeated measur*") OR Topic=("visit to visit") OR Topic=("reading to reading" OR "readings to readings")

# 3 Topic=("blood pressure" OR bp OR sbp OR dbp)

# 2 Title=(variation* OR variability OR variabilities) AND Title=("blood pressure" OR bp OR sbp OR dbp)

# 1 Topic=("blood pressure" NEAR/5 (variability OR variabilities OR variation*)) OR Topic=(bp NEAR/5 (variability OR

variabilities OR variation*)) OR Topic=(dbp NEAR/5 (variability OR variabilities OR variation*)) OR Topic=(sbp NEAR/5

(variability OR variabilities OR variation*))

Page 60 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

4

Table e2: Extracted information

Patient characteristics

Number of participants Gender

Age Source population

Proportion on anti-hypertensive medication

Study characteristics

Type of monitoring Type of study (trial/ observational)

Measurement device used* Measurement arm*

Person taking readings* Cuff size used*

Length of monitoring period Outcomes studied

Length of follow-up Variability measures studied

Authors overall conclusion

Statistical analysis

Analysis strategy Definition of a single measurement*

Variability measure Outcome

Systolic or diastolic BP Units of reported hazard ratio

Reported hazard ratio/ 95% confidence interval Standard deviation of variability measure

Adjustment for equivalent mean BP* Diurnal/ Seasonal variation considered*

Regression to the mean considered*

Medication change during measurement period limited/ adjusted for*

Medication change during follow-up period limited/ adjusted for*

*considered as potential confounders of the effect of BP variability on outcomes.

Page 61 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


nly

5

Table e3: Standardized hazard ratios explanation

A standardized beta coefficient is calculated as the beta coefficient per unit of the standardized exposure (the

exposure divided by its sample SD). For proportional hazards models, the beta-coefficient is the logarithm of the

hazard ratio. For example, for Eguchi et al. (2012) reported that the standard deviation of SBP is 20 mm Hg and

the HR per 10 mm Hg of mean SBP is 1.18; hence the beta coefficient is 0.166 per 10 mm Hg, the standardized

beta coefficient is 0.331 per SD and the standardized hazard ratio is 1.39 per SD. This has an interpretation as

“hazard ratio per one SD of SBP”. Because our exposure measurements include SD of SBP, and the phrase

“hazard ratio per one SD of SD” may not promote clarity, we use the term “standardized hazard ratio” rather

than “hazard ratio per one SD”.

Page 62 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only

6

Table e4: Risk of bias assessment

Paper, year Study Study Participation Study Attrition Prognostic Factor Measurement Outcome Measurement Study Confounding Statistical Analysis and Reporting

Asayama, 2012

Ohasama (observational)

moderate low low low low low

Bjorklund,

2004

Uppsala Longitudinal

Study of Adult Men (observational)

low moderate low low moderate low

Carr,

2012

MRC Elderly trial moderate moderate moderate moderate low low

Eguchi,

2012

Observational moderate low low high moderate low

Gao, 2014

Observational low low moderate high high low

Gavish,

2009

Observational moderate low moderate low moderate low

Hansen,

2010

IDACO

(observational)


Hashimoto, 2012

Ohasama (observational)


Hata,

2013

ADVANCE trial low low low low low low

Hsieh,

2012

Observational low low low high low low

Johansson,

2012

Health 2000 study

(observational)

low low low low low low

Kawai, 2013

NOAH study (observational)

low moderate moderate moderate low low

Kikuya,

2000

Ohasama

(observational)

moderate low low low moderate low

Kikuya,

2008

Ohasama

(observational)

moderate moderate low low low low

Kostis, 2014

SHEP (trial)

moderate low moderate low low low

Lau,

2014

Observational low high low high low low

Mallamaci,

2013

Observational moderate low moderate high low low

Mancia,

2007

PAMELA

(observational)

moderate low low low moderate low

Mancia,

2012

ELSA trial moderate low low high low moderate

McMullan,

2013

AASK trial moderate low moderate moderate low low

Mena, 2014

IDACO (observational)


Pierdomenico, Abruzzo, Italy moderate moderate low moderate low high

Page 63 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


7

Paper, year Study Study Participation Study Attrition Prognostic Factor Measurement Outcome Measurement Study Confounding Statistical Analysis and Reporting

2005 (observational)

Pierdomenico, 2006

Abruzzo, Italy (observational)

moderate moderate low moderate low low

Pierdomenico,

2009

Abruzzo, Italy

(observational)

moderate moderate low moderate low low

Poortvliet,

2012

PROSPER trial low low low low low low

Pringle, 2003

Syst-Eur trial moderate low moderate low low low

Rothwell,

2010

UK-TIA trial low low moderate low low low

Rothwell,

2010

European Stroke

Prevention Study

(ESPS-1) (trial)

low low moderate high low low

Rothwell,

2010

Dutch TIA trial. low low moderate high low low

Rothwell,

2010

ASCOT BPLA trial moderate low low high low low

Rothwell, 2010

ASCOT BPLA ABPM substudy

low low low low moderate low

Shimbo,

2012

Women’s Health

Initiative (observational)

low moderate moderate low low low

Suchy-Dicey,

2013

Cardiovascular Health

Study (observational)

moderate low moderate low low low

Verdecchia,

2007

PIUMA study

(observational)

moderate low low moderate low moderate

Wei,

2013

PROBE trial low moderate moderate high moderate low

Page 64 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


8

Figure e1

Page 65 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


9

Figure e2

Page 66 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


10

Figure e3

Page 67 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


11

Figure e4

Page 68 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


12

Figure e5

Page 69 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


13

Figure e6

Page 70 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


14

Figure e7

Page 71 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


15

Figure e8

Page 72 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


16

Figure e9

Page 73 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


17

Figure e10

Page 74 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


18

Table e5: Hazard ratios (HRs) for cardiovascular and mortality outcomes per standard deviation increase in home systolic blood pressure (BP) variability

Morning measurements Evening measurements Morning and evening measurements

Outcome Variability

measure

Paper [Study],

year

HR (95% CI) Variability

measure

Paper [Study],

year


measure

Paper [Study],

year

HR (95% CI)

CVD mortality VIM Asayama [Ohasama],

2012

1.26

(1.07, 1.49)

VIM Asayama [Ohasama],

2012

1.23

(1.05, 1.45)

SD Kikuya [Ohasama],

2008

1.16

(0.99, 1.36)

CHD mortality SD Hashimoto

[Ohasama], 2012

0.84

(0.59, 1.19)


2008

0.99

(0.79, 1.25)


2008

1.02

(0.81, 1.29)

Stroke mortality SD Hashimoto

[Ohasama], 2012

1.47

(1.11, 1.95)


2008

1.38

(1.12, 1.70)


2008

1.31

(1.05, 1.64)

Non-CVD

mortality


2008

1.18

(1.04, 1.34)


2008

1.07

(0.94, 1.22)


2008

1.15

(1.01, 1.31)

Cerebral

infarction

mortality

SD Hashimoto

[Ohasama], 2012

1.88

(1.31, 2.69)


2008

1.42

(1.08, 1.86)


2008

1.47

(1.11, 1.95)

CVD events SD Johansson [Health

2000], 2012

1.17

(1.02, 1.34)

SD Johansson [Health

2000], 2012

1.08

(0.93, 1.26)


2000], 2012

1.06

(0.93, 1.22)

Stroke events VIM Asayama [Ohasama],

2012

1.14

(1.00, 1.30)

VIM Asayama [Ohasama],

2012

1.06

(0.93, 1.21)

- - -

Page 75 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


19

Figure e11

Page 76 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


20

Table e6: Hazard ratios (HRs) for cardiovascular and mortality outcomes per standard deviation increase in home diastolic blood pressure (BP) variability

Morning measurements Evening measurements Morning and evening measurements

Outcome Variability

measure

Paper [Study],

year


measure

Paper [Study],

year


measure

Paper [Study],

year

HR (95% CI)

CVD mortality SD Kikuya [Ohasama],

2008

1.14

(0.98, 1.32)


2008

1.13

(0.96, 1.33)

- - -

CHD mortality SD Kikuya [Ohasama],

2008

1.02

(0.81, 1.28))


2008

0.97

(0.76, 1.23)

- - -

Stroke mortality SD Kikuya [Ohasama],

2008

1.26

(1.03, 1.55)


2008

1.27

(1.04, 1.56)

- - -

MI mortality SD Kikuya [Ohasama],

2008

0.76

(0.52, 1.12)


2008

0.63

(0.41, 0.97)

- - -

Non-CVD mortality SD Kikuya [Ohasama],

2008

1.09

(0.97, 1.23)


2008

1.09

(0.96, 1.24)

- - -

Cerebral infarction

mortality


2008

1.30

(1.00, 1.68)


2008

1.33

(1.03, 1.72)

- - -

CVD events SD Johansson [Health

2000], 2012

1.23

(1.09, 1.39)


2000], 2012

1.12

(0.97, 1.28)


2000], 2012

1.16

(1.01, 1.34)

Page 77 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


21

Figure e12

Page 78 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


22

Figure e13

Page 79 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


23

Figure e14

Page 80 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


24

Figure e15

Page 81 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


25

Figure e16

Page 82 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


26

Figure e17

Page 83 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


27

Figure e18

Page 84 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


28

Figure e19

Page 85 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


29

Figure e20

Page 86 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


30

Figure e21

Page 87 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


31

Figure e22

Page 88 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


32

Figure e23

Page 89 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


33

Figure e24

Page 90 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


34

Figure e25

Page 91 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


35

Figure e26

Page 92 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


36

Figure e28

Page 93 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


37

Figure e29

Page 94 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


38

Figure e30

Page 95 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960


39

Figure e31

Page 96 of 96


BMJ

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960

Confidential: For Review Only - BMJ...Confidential: For Review Only 3 49 Conclusions: Visit-to-visit...

Documents

Transcript of Confidential: For Review Only - BMJ...Confidential: For Review Only 3 49 Conclusions: Visit-to-visit...