1595.full

8/13/2019 1595.full

1/7

Glycemic Control and Extended Hemodialysis Survival inPatients with Diabetes Mellitus: Comparative Results ofTraditional and Time-Dependent Cox Model Analyses

Mark E. Williams,* Eduardo Lacson Jr., Weiling Wang, J. Michael Lazarus,

and Raymond Hakim *Renal Unit, Joslin Diabetes Center, Boston, Massachusetts; and Fresenius Medical CareNorth America, Lexington, Massachusetts

Background and objectives: The benefits and risks of aggressive glycemic control in diabetes mellitus complicated byend-stage kidney failure remain uncertain but have importance because of the large patient population with inferior overallprognosis. Recent large observational studies with differing methodologies reached somewhat contrasting conclusionsregarding the association of hemoglobin A1c with survival in diabetic chronic hemodialysis patients.

Design, setting, participants, & measurements: This study supplements the authors previous analysis (which found nocorrelation) by extending the follow-up period to 3 years and using time-dependent survival models with repeated measures.Among 24,875 nationally distributed study patients, 94.5% had type 2 diabetes, allowing additional analysis in the subset with

type 1 diabetes. Data were collected at baseline and every quarter to a maximum of 3 years follow-up.Results: Adjusted standard and time-dependent Cox models indicated that only extremes of glycemia were associated with

inferior survival. There was no effect modification by serum albumin levels, a marker of protein nutrition status, and no trendassociated with random glucose measurements in a post hoc analysis. In type 1 diabetic patients, upper extreme hemoglobinA1c values indicated lower survival risk.

Conclusions: Sustained extremes of glycemia were only variably and weakly associated with decreased survival in thispopulation. In the absence of randomized, controlled trials, these results suggest that aggressive glycemic control cannot beroutinely recommended for all diabetic hemodialysis patients on the basis of reducing mortality risk. Physicians areencouraged to individualize glycemic targets based on potential risks and benefits in diabetic ESRD patients.

Clin J Am Soc Nephrol 5: 15951601, 2010. doi: 10.2215/CJN.09301209

Suboptimal glycemic control is a major determinant of mortality worldwide (1). Specifically, elevated hemoglobin A1c (HgbA1c) is an independent risk factor for

coronary heart disease in persons with diabetes (2,3). Recently,three large randomized trials have indicated that intensiveglucose lowering in patients with type 2 diabetes mellitus (whocomprise 95% of diabetic ESRD patients in the United States)did not reduce the risks of cardiovascular disease, the mostcommon source of ESRD mortality (47). Although diabeticESRD patients continue to comprise approximately half of allprecedent patients in the United States and substantial portionselsewhere (8), very few diabetic chronic kidney disease (CKD)patients have been evaluated in trials for which the resultsprovided evidence for the benefit of aggressive glucose control(9,10). Furthermore, the objectives of glycemic management inthis population remain uncertain (11). Although the KidneyDisease Outcomes Quality Initiative (K/DOQI) guidelines for

diabetes and CKD support standard HgbA1c targets in patientswith advanced kidney disease (12), the glycemic control targetswere devised for patients without advanced CKD. Support foruniversal targets for glycemic control in those with renal insuf-ficiency was lacking (because of insufficient evidence), even inrecent guidelines, such as those developed by the AmericanCollege of Physicians (13) and the Department of VeteransAffairs (14).

We previously reported findings from a large nationalESRD database that survival curves in diabetic patientsgrouped by HgbA1c levels did not differ statistically, andthat there was no correlation between HgbA1c levels and12-month mortality risk singly or when adjusted for case-mixand laboratory variables (15). However, those results werenot independently confirmed by a similar-sized retrospectivedatabase analysis that indicated that higher HgbA1c wasassociated with increased death risk in diabetic ESRD pa-tients (16). In addition to contrasting results, the two studieshad significant methodological differences, with the latterstudy using a longer follow-up period, time-dependent sur-vival models ( i.e., with repeated measures), and adjustmentsfor certain surrogates of malnutrition and inflammation.

In light of these unresolved differences, this report attemptsto build on our previous analysis by providing standardized

Received December 23, 2009. Accepted May 6, 2010.

Published online ahead of print. Publication date available at www.cjasn.org.

Correspondence: Dr. Mark E. Williams, Renal Unit, Joslin Diabetes Center, One JoslinPlace, Boston, MA 02215. Phone: 617-732-2477; Fax: 617-732-2467; E-mail:[email protected]

Copyright 2010 by the American Society of Nephrology ISSN: 1555-9041/509 1595

8/13/2019 1595.full

2/7

time-independent (Cox) and time-dependent models. It extendspatient follow-up to a maximum of 3 years and, for the firsttime, provides a separate analysis of the subgroup of patientswith type 1 diabetes mellitus. Finally, it includes an analysisstratified by serum albumin levels, the most common marker of protein nutrition status.

Materials and MethodsThe study population consisted of 24,751 patients with analyzable

information out of the 24,875 (99.5%) chronic hemodialysis patientswith a documented diagnosis of diabetes mellitus (DM) (by review of source information that were likely utilized to complete the ESRD form2728 and of dialysis unit admission records) that were actively treatedin Fresenius Medical CareNorth America (FMCNA) facilities as of January 1, 2003, with at least one HgbA1c result from the last quarter of 2002. We previously reported 1-year survival in the same cohort (15)and now extend follow-up until December 31, 2005 for a maximumduration of up to 3 years. The mean (SD) duration of follow-up was671 397 days. The main outcome was mortality, a combined endpoint of all deaths and withdrawals from dialysis. Surviving patientscontributed exposure time until they were lost to follow-up ( e.g., be-cause of kidney transplant or transfer out of FMCNA) or the end of thestudy.

Demographic variables were collated as of January 1, 2003 to includeage, gender, race, dialysis vintage, body surface area (BSA, calculatedfrom height and postdialysis weight), and vascular access type (fistula,graft, catheter, and unknown). All baseline laboratory results werepresented as the 3-month average of all available values from Novem- ber 1 to December 31, 2002. The means of available HgbA1c, equili- brated Kt/V (eKt/V), albumin, hemoglobin, phosphorous, creatinine,white blood cell count, and calcium results were also recorded everyquarter thereafter until the patient reached an end point, was lost tofollow-up, or at the study conclusion. Individual physicians deter-

mined the frequency of monitoring HgbA1c levels, whereas the otherlaboratory results were routinely measured at least once per month.When subsequent quarterly HgbA1c values were missing, the previousvalue was carried forward in time-dependent analysis. A single labo-ratory (Spectra Laboratory, Rockleigh, NJ) performed all blood tests.The Roche Cobas Integra 800 whole blood immunoturbidimetric assaystandardized according to the National Hemoglobin StandardizationProgram was utilized to determine HgbA1c, with a normal range from4.5% to 5.7%. Albumin was measured using the bromcresol greenmethod, consistent with previously referenced studies (15,16). Dialysisdose was derived from two-sample variable volume urea kinetic mod-eling and reported as eKt/V.

The primary statistical analyses were performed on all study pa-

tients, whereas secondary analyses were conducted on subgroups of type I DM patients and type II DM patients. Differences between typeI and type II DM patients were tested using the t test or 2 test whenappropriate. KaplanMeier curves were drawn to compare the survivaldistribution among varying levels of HgbA1c categorized into fivelevels, 5.5% to 8.5% at 1.0% increments, to be consistent with ourprior work (15). Standard and time-dependent Cox models were con-structed as ( 1) unadjusted, ( 2) case-mixed adjusted (including age,gender, race, dialysis vintage, BSA), and ( 3) case-mixed labs adjusted(including eKt/V, albumin, hemoglobin, phosphorous, creatinine, andcalcium as well as vascular access) models. Risk profiles were con-structed based on HgbA1c categories from 5.0% to 11.0% at 0.5%increments, also consistent with our prior work (15). A different group-ing strategy was used for the subset of type I DM with wider HgbA1ccategories from 5.0% to 10.0% at 1.0% increments because of sig-

nificantly fewer patients. For the time-dependent models, laboratoryvalues were updated every quarter and 10% of patients had missingvalues each quarter, with the last quarters value imputed for themissing data. Three sensitivity analyses were conducted. First, overallmean values were substituted for missing data with no substantialchange in the findings. Second, the primary standard Cox models formortality were repeated with adjustment for case-mix albumin alone,case-mix labs (except for albumin), and the latter series were alsomodeled with stratification by albumin ( 3.5, 3.5 to 3.9, and 4.0 g/dl)to evaluate for potential effect modification. Third, post hoc we includedadditional covariates to the case-mixed labs adjusted models such as baseline primary insurance status, number of comorbidity diagnoses,and body mass index, as well as quarterly means for erythropoietindose per treatment, normalized protein catabolic rate, ferritin, and bicarbonatedesigned to approximate as closely as possible the time-dependent models performed by Kalantar-Zadeh et al. (16). All analyseswere performed using SAS version 9.2 (Cary, NC).

ResultsCharacteristics of the study cohort ( n 24,751) along with a

breakdown between type 1 (5.5%) and type 2 (94.5%) DM areshown in Table 1. The differences between patients with types1 and 2 DM were as described previously (15). Patients withtype 1 DM tended to be younger, with proportionately moremale and white patients, and potentially better intake andmuscle mass ( i.e., higher HgbA1c, creatinine, phosphorus, andBSA). They also had proportionately more fistulas and slightlylonger vintage.

The KaplanMeier 3-year survival curves categorized by baseline HgbA1c are shown in Figure 1. Patients in the lowestHgbA1c category exhibited the worst survival, whereas thosein the highest category had the best survival ( P 0.0001). With

longer follow-up, the survival curves layered from the highestto the lowest HgbA1c categoriesfindings that were not ap-parent in our initial report of 1-year follow-up (15). Survivalmodels indicated no discernable pattern of risk betweenHgbA1c and mortality hazard ratios (HRs), shown in Figure 2.Standard Cox models (Figure 2A) showed a significant unadjusted HR of 1.20 (P 0.001) for HgbA1c 5.0% and lostsignificance when adjusted for case-mix lab (including vas-cular access type). In the other extreme, for HgbA1c 11%, HRsfrom unadjusted models were NS, but the HR for case-mixadjusted models was 1.28 ( P 0.001) and for case-mix lab-adjusted models was 1.21 ( P 0.05). Utilizing time-depen-dent Cox models (Figure 2B) exaggerated the HRs for severalHgbA1c categories 6.5% and 11% but was similarly absentof any particular trend. A sensitivity analysis utilizing standardCox models with and without adjustment for albumin as wellas stratified by albumin ( 3.5, 3.5 to 3.9, and 4.0 g/dl) indi-cated no significant effect modification (data not shown).

The subset of patients with type 2 DM comprised 94.5% of allstudy patients, hence reflecting essentially the same HR profileas the entire cohort. However, the HR profile for the smallersubset of patients with type 1 DM revealed increased death risk beyond HgbA1c levels of 9% in the adjusted standard Coxmodels (Figure 3A). For HgbA1c of 9.0% to 9.9%, the case-mixadjusted HR was 1.41 and case-mix lab-adjusted HR was 1.52(both P 0.05), whereas for HgbA1c 10%, HRs were 1.73 ( P

1596 Clinical Journal of the American Society of Nephrology Clin J Am Soc Nephrol 5: 15951601, 2010

8/13/2019 1595.full

3/7

0.05) and 2.06 (P 0.001), respectively. There was a similartrend, albeit attenuated, from the time-dependent models (Fig-ure 3B), with only HgbA1c 10% indicating increased deathrisk at case-mix adjusted HR of 1.52 and case-mix lab-adjusted HR of 1.49 (both P 0.05).

Additionally, we modified our analysis further to moreclosely approximate the previous methodology in the Kalan-tar-Zadeh report (Figure 4). In this time-dependent analysisadopting the same HgbA1c reference range (5.0% to 5.9%)and similarly constructed HgbA1c categories, our results

Table 1. Demographic characteristics of the study cohort and important subgroups

CharacteristicsDM Patients with HgbA1c

All Patients Type 1 DM Type 2 DM

Number of patients 24,751 1,367 (5.5%) 23,384 (94.5%)Age (years) 63.7 (12.1) 49.3 (14.4) 64.6a (11.4)

Percent femalec

51.5% 48.4% 52.4%Race c

white 53.1% 60.6% 52.7% black 36.4% 32.6% 36.6%other 10.6% 6.8% 10.8%

Vintage (days) 1048 (948) 1165 (1,071) 1041 a (940)BSA (m2 ) 1.86 (0.25) 1.82 (0.25) 1.86a (0.25)Access type c

fistula 29.9% 34.3% 29.7%graft 45.7% 40.8% 46.0%catheter 23.6% 24.4% 23.6%unknown 0.8% 0.5% 0.8%

Hemodialysis dose (eKt/V) 1.41 (0.29) 1.41 (0.32) 1.41 (0.28)Albumin (g/dl) 3.82 (0.38) 3.83 (0.42) 3.81 (0.38)Hemoglobin (g/dl) 11.71 (1.10) 11.71 (1.17) 11.71 (1.09)Calcium (mg/dl) 9.29 (0.70) 9.27 (0.73) 9.29 (0.70)Phosphorus (mg/dl) 5.60 (1.45) 5.89 (1.55) 5.59 a (1.44)Creatinine (mg/dl) 8.22 (2.87) 8.60 (2.61) 8.20 a (2.67)White blood cells ( 103 / l) 7.77 (2.87) 7.97 (2.39) 7.76 b (2.90)HgbA1c (%) 6.77 (1.71) 7.50 (2.00) 6.73 (1.68)

Values are shown as percentages or means (SD).a P 0.0001; b P 0.002; c P 0.0001 by 2 test: all comparing type 1 versus type 2 patients.

Figure 1. KaplanMeier survival curve. Shown are survival rates for baseline HgbA1c groups.

Clin J Am Soc Nephrol 5: 15951601, 2010 Glycemic Control and Hemodialysis in Diabetes 1597

8/13/2019 1595.full

4/7

were not changed. Because potential explanations for theattenuated relationships between HgbA1c and survival inour analyses included somewhat different laboratory adjust-ments and case-mix characteristics (comorbid status, dialysisvintage categories) of the study cohorts, we performed a posthoc analysis to include additional adjustors adapted from theKalantar-Zadeh report such as baseline primary insurancestatus, number of comorbidity diagnoses, and body massindex, as well as quarterly means for erythropoietin dose pertreatment, normalized protein catabolic rate, ferritin, and bicarbonate for the time-dependent models, (data on totaliron binding capacity, lymphocyte percentage, and smokingwere inadequate for inclusion). The results did not affect thefindings of the study presented here. Finally, we also exam-ined in a post hoc analysis Cox models for 17,182 patients whosurvived for a full year using the mean of all availablepredialysis random blood glucose measurements (three of four patients drawn at least every month) throughout theyear as the predictor variable of interest, with no associatedtrends appreciated, as with HgbA1c (data not shown).

DiscussionThe study presented here, which consists of the survival data

on the same cohort as our 2006 publication (15), supplementsthat analysis in several ways and provides a more direct com-

parison with the report by Kalantar-Zadeh (16). With extendingthe follow-up to 3 years, the study presented here allows com-parison of standard and time-dependent Cox models and pro-vides separate analysis of the type 1 diabetic hemodialysispopulation. We highlight two key findings. First, unadjusteddata using KaplanMeier survival curves, with longer follow-up, indicated that patients in the lowest overall HgbA1c cate-gory suffered inferior survival. Adjusted standard Cox analyses(worse survival at high and low HgbA1c extremes) and time-

B

A

Figure 2. (A) Standard Cox models for mortality based onHgbA1c at baseline. (B) Time-dependent Cox models for mor-tality (laboratory results adjusted quarterly).

B

A

Figure 3. (A) Standard Cox models for patients with type 1 DM(n 1367) only. (B) Time-dependent Cox model for patientswith type 1 DM as above.

Figure 4. Time-dependent Cox analysis with HgbA1c categoriesto match previous publication by Kalantar-Zadeh et al. (16).


8/13/2019 1595.full

5/7

dependent Cox analyses (survival particularly impaired at thelow HgbA1c extreme) further established that extremes of gly-cemia portend worse survival in diabetic hemodialysis pa-tients. Second, the smaller subgroup population of type 1 DMpatients indicated a lower survival risk at the upper extremerange of HgbA1c values.

Hyperglycemia is proven to play a fundamental role in dia-

betic complications in nondialysis patients (9,10). In the generaldiabetic population, glycemic control represented by HgbA1clevels 7% reduces complications of diabetes and reduces theneed for hospitalization (17). The burden of mortality attributedto higher blood glucose levels is approximately 3 times higherthan that attributed to diabetes alone (1). However, excessmortality was recently associated with intensive glycemic con-trol in the ACCORD trial, in which the intensive arm wasstopped early because of a 22% increase in all-cause mortalitycompared with the standard treatment arm (4). Such resultshave provoked a reassessment of the benefit of aggressiveglycemic control, particularly in patients whose complications

are already advanced, and have highlighted the potential risks.Although glycemic control targets in CKD currently recom-mended by K/DOQI guidelines do not meet the definition of tight in the ACCORD study, this elderly population withmultiple comorbidities is at particularly high risk of hypogly-cemic complications even with the standard target of 7%(1822).

The role of glycemic control in diabetic ESRD is of growingimportance because of the magnitude of the population world-wide and associated worse overall prognosis compared withthe general dialysis population. Evidence of mortality risk re-duction has been sought to help set glycemic guidelines on the

optimal target for glucose control.There have been no randomized clinical trials to evaluate theeffects of glycemic control in patients with advanced compli-cations such as ESRD (23), and it remains unclear to what extenttight control might justify the risks involved. However, a recentcohort study of type 2 diabetes indicated diminished cardio-vascular benefit from intensive blood glucose control in pa-tients with high levels of comorbidity (24). Until the past fewyears, there had been few studies that examined the association between HgbA1c and clinical outcomes in the diabetic dialysispopulation (2527), most of them small observational studiesperformed outside of the United States. Limited data fromolder studies appeared to suggest that poor glycemic control inpatients with diabetes on dialysis was associated with in-creased morbidity (28,29). The question whether glycemic con-trol affects survival outcomes in dialysis patients has benefitedfrom two recent large observational reports of comparable size,contrasting methodologies, and somewhat different conclu-sions.

We were previously unable to demonstrate the expectedparallel association between increasing HgbA1c and greatermortality risk over 1 year in the first large national U.S. database analysis (15). HgbA1c was associated with only a weakcorrelation of 1-year death risk at low and high extreme valuesof HgbA1c. In 24,875 hemodialysis patients who had diabetesin a large national dialysis organization database, there was no

overall relationship between glycemic control as measured by baseline HgbA1c levels, and 12-month survival, for case-mix orcase-mix laboratory-adjusted data. After our study, Kalantar-Zadeh et al. reported an analysis from a second large nationalESRD database of 23,618 patients with diabetes using time-dependent survival models with repeated measures (16). Incontrast to our earlier findings, they concluded that higher

HgbA1c ( 10%) was associated with a 41% greater death riskfor all-cause and cardiovascular death. The authors proposedthat the general populations conventional association betweenHgbA1c and survival would hold in dialysis patients if theconfounding effects of malnutrition, anemia, and other factorswere controlled for in the analysis and concluded that certainsubgroups (nonanemic, younger) would still benefit from im-proved glycemic control. In our current analyses, adjustedtime-dependent Cox models indicated that extremes of glyce-mia could be associated with inferior survival.

Although HgbA1c remains the most widely used index of glycemic control in the diabetic population, it is not without its

limitations (30). Recent studies have raised concerns about itsvalidity in ESRD. HgbA1c may not optimally represent thegeneral glycemic state in ESRD patients because of uniquephysiologic and pathologic changes in this population (31,32).In our own national database analysis, HgbA1c showed only aweak correlation with mean random glucose values ( R2 0.37).Glycated hemoglobin indicates the percentage of circulatinghemoglobin that has chemically reacted with glucose (32). Theshortened half-life of erythrocytes in uremia and the use of erythropoietin, by increasing the proportion of younger eryth-rocytes, appear to reduce HbA1c relative to plasma glucoselevels. As a result, HgbA1c tends to be lower in diabetic pa-tients with kidney impairment or on renal replacement therapy(33). Nonetheless, glycated hemoglobin remains the best avail-able tool for monitoring glycemic control (34).

In summary, sustained extremes of glycemia are associatedwith increased mortality risk in diabetic ESRD patients. Wehave previously reported similar associations with risk of hos-pitalization (31). However, the overall relationships betweenglycemic control and survival are relatively weak and maydepend in part on statistical methodology. Although random-ized controlled trials will be required to determine whethermeeting conventional HgbA1c targets can improve survivaloutcomes in this population, our results suggest that improvedsurvival should not be used to justify aggressive glycemiccontrol strategies. Finally, emphasis should be placed on therisks associated with hypoglycemia (35,36). Our preliminaryfindings indicate that risk of hypoglycemia is especially signif-icant in diabetic ESRD patients with glycemic variability (37).Therefore, pending the performance of randomized clinicaltrials, it may be prudent to consider the risks associated withaggressive glycemic control relative to expected benefits, toindividualize glycemic targets, and to tailor specific therapy onthe basis of clinical status in diabetic patients with ESRD.

AcknowledgmentsWe are grateful to the anonymous reviewers for providing analysis

that resulted in improvements in the manuscript. We thank the diligent

Clin J Am Soc Nephrol 5: 15951601, 2010 Glycemic Control and Hemodialysis in Diabetes 1599

8/13/2019 1595.full

6/7

clinical staff from dialysis units affiliated with FMCNA without whomcollection of clinical data would not have been possible.

DisclosuresE.L., W.W., J.M.L., and R.H. are employed by FMCNA.

References1. Avendano M, Mackenbach JP: Blood glucose levels: Facing

a global crisis. Lancet 368: 16311632, 20062. Selvin E, Coresh J, Golden SH, Brancati FL, Folsom AR,

Steffes MW: Glycemic control and coronary heart diseaserisk in persons with and without diabetes. The Atheroscle-rosis Risk in Communities Study. Arch Intern Med 165:19101916, 2005

3. Home P: Contributions of basal and post-prandial hyper-glycaemia to micro- and macrovascular complications inpeople with type 2 diabetes. Curr Med Res Opin 21: 889898, 2005

4. Gerstein HC, Miller ME, Byington RP: Action to ControlCardiovascular Risk in Diabetes Study Group: Effects of intensive glucose lowering in type 2 diabetes. N Engl J Med358: 25452559, 2008

5. ADVANCE Collaborative Group: Intensive blood glucosecontrol and vascular outcomes in patients with type 2diabetes. N Engl J Med 358: 25602572, 2008

6. Duckworth W, Abraira C, Moritz T, Reda D, Emanuele N,Reaven PD, Zieve F, Marks J, Davis SN, Hayward R, WarrenSR, Goldman S, McCarren M, Vitek ME, Henderson WG,Huang GD, for the VADT Investigators: Glucose controland vascular complications in veterans with type 2 diabe-tes. N Engl J Med 360: 129139, 2009

7. Skyler JS, Bergenstal R, Bonow RO: Intensive glycemiccontrol and the prevention of cardiovascular events: Im-

plications of the ACCORD, ADVANCE, and VA DiabetesTrials: A position statement of the American Diabetes As-sociation and a Scientific Statement of the American Col-lege of Cardiology Foundation and the American HeartAssociation. J Am Coll Cardiol 53: 298304, 2009

8. U.S. Renal Data System: USRDS 2009 Annual Data Report: Atlas of Chronic Kidney Disease and End-Stage Renal Disease inthe United States. Bethesda, MD, National Institutes of Health, National Institute of Diabetes and Digestive andKidney Diseases, 2009

9. Diabetes Control and Complications Trial Research Group:The effect of intensive treatment of diabetes on the devel-opment and progression of long-term complications in

insulin-dependent diabetes mellitus: The Diabetes Controland Complications Trial. N Engl J Med 329: 977986, 199310. U.K. Prospective Diabetes Study Group: Intensive blood-

glucose control with sulphonyureas or insulin comparedwith conventional treatment and risk of complications inpatients with type 2 diabetes (UKPDS 33). Lancet 352: 837853, 1998

11. Feldt-Rasmussen B: Is there a need to optimize glycemiccontrol in hemodialyzed diabetic patients? Kidney Int 70:13921394, 2006

12. K/DOQI: K/DOQI clinical practice guidelines and clinicalpractice recommendations for diabetes and chronic kidneydisease. Am J Kidney Dis 49[Suppl 2]: S12S154, 2007

13. Qaseem A, Vijan S, Snow V, Cross JT, Weiss KB, OwensDK: Glycemic control and type 2 diabetes mellitus: The

optimal hemoglobin A1c targets. A guidance statementfrom the American College of Physicians. Ann Intern Med147: 417422, 2007

14. Pogach LM, Brietzke SA, Cowan CL, Conlin P, Walder DJ,Sawin CT: Development of evidence-based clinical practiceguidelines for diabetes. Diabetes Care 27[Suppl 2]: B82B89,2004

15. Williams ME, Lacson E Jr, Teng M, Ofsthun N, Lazarus JM:Hemodialyzed type I and II diabetic patients in the US:Characteristics, glycemic control, and survival. Kidney Int70: 15031509, 2006

16. Kalantar-Zadeh K, Kopple JD, Regidor DL, Jing J, ShinabergerCS, Aronovitz J, McAllister CJ, Whellan D, Sharma K: A1cand survival in maintenance hemodialysis patients. Diabe-tes Care 30: 10491055, 2007

17. Cook CB, Hentz JG, Tsui C: Poor glycemic control in-creases the risk of hospitalization in urban African Amer-icans with diabetes. Ethn Dis 16: 880885, 2006

18. Amiel SA, Dixon T, Mann R, Jameson K: Hypoglycaemia intype 2 diabetes. Diabetic Med 25: 245254, 2008

19. Abaterusso C, Lupo A, Ortalda V, DeBiase V, Pani A,Muggeo M, Gambaro G: Treating elderly people with di-abetes and states 3 and 4 chronic kidney disease. Clin J AmSoc Nephrol 3: 11851194, 2008

20. Huang ES, Zhang QI, Gandra N, Chin MH, Meltzer DO:The effect of comorbid illness and functional status on theexpected benefits of intensive glucose control in older pa-tients with type 2 diabetes: A decision analysis. Ann Intern Med 149: 1119, 2008

21. Moen MF, Zhan M, Hsu VD, Walker LD, Einhorn LM,Seliger SL, Fink JC: Frequency of hypoglycemia and itssignificance in chronic kidney disease. Clin J Am Soc Neph-rol 4: 11211127, 2009

22. Whitmer RA, Karter AJ, Yaffe K, Quesenberry CP Jr, Selby JV: Hypoglycemic episodes and risk of dementia in olderpatients with type 2 diabetes mellitus. JAMA 301: 15651572, 2009

23. American Diabetes Association: Standards of medical carein diabetes2009. Diabetes Care 32: S87S94, 2009

24. Greenfield S, Billimek J, Pellegrini F, Franciosi M, deBerardis G,Nicolucci A, Kaplan S: Comorbidity affects the relationship between glycemic control and cardiovascular outcomes indiabetes: A cohort study. Ann Intern Med 151: 854860,2009

25. Morioka T, Emoto M, Tabata T, Shoji T, Tahara H: Glyce-mic control is a predictor of survival for diabetic patientson hemodialysis. Diabetes Care 24: 909913, 2001

26. Wu MS, Yu CC, Yang CW, Wu CH, Haung JY, Hong JJ, FanChiang CY, Huang CC, Leu ML: Poor pre-dialysis glycae-mic control is a predictor of mortality in type II diabeticpatients on maintenance haemodialysis. Nephrol DialTransplant 12: 21052110, 1997

27. Oomichi T, Emoto M, Tabata T, Morioka T, Tsujimoto Y,Tahara H, Shoji T, Nishizawa Y: Impact of glycemic controlon survival of diabetic patients on chronic regular hemo-dialysis: A 7-year observational study. Diabetes Care 29:14961500, 2006

28. Manske CL: Hyperglycemia and intensive glycemic con-trol in diabetic patients with chronic renal disease. Am J Kidney Dis 32: S157S171, 1998

29. Tzamaloukas AH, Murata GH, Zager PG: The relationship


8/13/2019 1595.full

7/7

1595.full

Documents

Transcript of 1595.full