tap_660 199..204

Hemodialysis Versus Peritoneal Dialysis: A Case ControlStudy of Survival in Patients With Chronic Kidney

Disease Stage 5

Alexandra Maier,1 Franziska Stocks,1 Wolfgang Pommer,2 Walter Zidek,1 Martin Tepel,1

and Alexandra Scholze1

1Department of Nephrology, Charité University Medical School, and 2Department of Nephrology, VivantesHumboldt Clinic, Berlin, Germany

Abstract: It is still controversial whether the mode ofdialysis or preexisting comorbidities may influence theprognosis of patients with chronic kidney disease stage 5.Therefore, we performed a prospective case control studyto evaluate whether the mode of dialysis may influenceoutcome. We found 25 cases on peritoneal dialysis (PD)treatment and 75 age and sex-matched controls on hemo-dialysis (HD) treatment for more than 3 months. Analysiswas by intention-to-treat. During the follow up of58 months, 6 out of 25 patients (24%) died in the PD group,whereas in the HD group 26 out of 75 patients (35%) died

(relative risk 0.69 [95% CI 0.32 to 1.49]; P = 0.46). Survivalwas not significantly different between the groups as indi-cated by Mantel-Cox log-rank test (hazard ratio 0.52 [95%CI 0.25 to 1.10]; P = 0.11). Multiple variable regressionshowed that age and diabetes mellitus, but not mode ofdialysis, predicted death in patients with chronic kidneydisease. It is concluded that age and comorbidities but notmode of dialysis are important to predict survival inpatients with chronic kidney disease stage 5. Key Words:Mode of dialysis, Outcome, Chronic kidney disease.

The choice of mode of dialysis depends on severalfactors including availability of treatment options,reimbursement, support from staff and families, therequirement for training programs, and several localfactors (1–3). An investigation on health-relatedquality of life showed that patients on peritonealdialysis (PD) rated patient satisfaction better thanhemodialysis (HD) patients (4). However, there hadbeen a controversy whether the mode of dialysis, i.e.peritoneal dialysis or hemodialysis, may influence theprognosis of patients with chronic kidney diseasestage 5 (5–8). Serkes et al. did not observe any signifi-cant difference in survival between patients on PD orHD (9). On the other hand, Suzuki et al. reportedthat outcome in patients with chronic kidney diseasedepended to some extend on the mode of dialysis(10). Registry data from Canada and Denmark

showed that patients on PD have a lower risk ofdeath than those on HD using an as-treated analysis(11,12). In contrast, some studies from the UnitedStates showed that patients on HD had a more favor-able outcome compared to patients on PD (6,8). Assuggested by Vonesh et al., discrepant results can beattributed to methodologic differences, the degree ofcase-mix adjustment, and at least in part to the het-erogeneity of patient populations, who receive eitherPD or HD (7). Insufficient adjustments for age andcomorbidities are probably the most importantcauses for observed differences. Therefore we con-ducted a case control study to evaluate the effects ofage, mode of dialysis, and comorbidity on survival ofpatients with chronic kidney disease stage 5.

PATIENTS AND METHODS

In this prospective study all cases and controlshad chronic kidney disease stage 5. Cases and con-trols were treated in two nephrology centers inBerlin, Germany. The study protocol was approved

Received May 2008; revised August 2008.Address correspondence and reprint requests to Dr Martin

Tepel, Department of Nephrology, Charité—Universitätsmedizin,Hindenburgdamm 30, Berlin 12200, Germany. Email: martin.tepel@charite.de

Therapeutic Apheresis and Dialysis 13(3):199–204doi: 10.1111/j.1744-9987.2009.00660.x© 2009 The AuthorsJournal compilation © 2009 International Society for Apheresis

199

by the local ethics committees. All patients gavewritten, informed consent. The study started onOctober 1, 2002. All participants were followedfrom study entry until the first event, i.e. renal trans-plantation, changing the renal replacement therapy,the end of the study period, or death, whicheveroccurred first.

Twenty-five cases were identified as patients withchronic kidney disease stage 5 receiving PD for morethan 3 months. Patients were more than 18 years old.All cases received continuous ambulatory peritonealdialysis according to recent guidelines.

Three controls were obtained for each case. Con-trols were patients with chronic kidney disease stage5 receiving HD for more than 3 months. Controlswere more than 18 years old. All of the patientsreceiving HD were routinely dialyzed for 4 to 5 hthree times weekly using biocompatible membraneswith no dialyzer reuse according to recent guide-lines. The dialysates used were bicarbonate-based.Controls were followed up at the same time as thecase and we matched them for sex and age (yearof birth within 2 year limit). Patients were consid-ered to have switched technique when they changedfrom one type of dialysis to another and continuedto use the new technique for more than 1 monthas recommended by the literature (11). However,only three patients receiving PD switched to HDand none of the patients receiving HD switched toPD.

Baseline data collectionPatient history was raised using a standardized

questionnaire and comprised personal histories,smoking habits, cause of kidney disease judged byclinical appraisal, months of dialysis treatment, pre-existing cardiovascular disease (i.e. history of myo-cardial infarction, need for coronary angioplasty orcoronary bypass surgery, ischemic stroke, peripheralvascular disease with the need for amputation orangioplasty), presence of diabetes mellitus, andcurrent medication including angiotensin convertingenzyme inhibitors, beta-blockers, calcium channelblockers, lipid-lowering agents, and erythropoietin.Body mass index was calculated from the followingequation: body mass index = body weight (kg)/height2 (m2). Blood pressure was measured with asphygmomanometer after 10 min of recumbency.Hypertension was defined as presently existing arte-rial hypertension or a history of arterial hyperten-sion, i.e. resting blood pressure more than 140/90 mm Hg or antihypertensive medication. Smokingwas defined as former or current smoker.

OutcomeThe primary endpoint was time from study entry to

mortality from any cause. Causes of death duringfollow-up were classified as cardiovascular includingsudden death, infection, or cancer. Death occurringoutside hospital for which no other cause wasassigned was regarded as sudden death and wasincluded in the definition of cardiovascular-diseasedeath. Deaths were classified by the treating physi-cian and reviewed by one of the authors indepen-dently of the endpoint analysis. Data on mortalitywere obtained for all patients. Patients that under-went renal transplantation or change of renalreplacement therapy were censored on the day ofthat event.

StatisticsContinuous data including age, months of dialysis

treatment, and blood pressure were reported asmedian (25% percentile—75% percentile). Non-parametric Mann-Whitney test was used to detectdifferences in continuous variables between the treat-ment groups. Frequency counts were calculated forcategorical data such as gender, specific medications,and diagnostic classifications.Differences in these cat-egorical variables between the treatment groups wereanalyzed by Fisher’s exact test. All time-to-eventanalyses were performed using the Mantel-Coxlogrank test. Hazard ratio and its 95% confidenceinterval are given.This test generates a P value testingthe null hypothesis that the survival curves are iden-tical in the overall population. In other words, the nullhypothesis is that the mode of dialysis did not changesurvival.We calculated two-tailed P values because wecould not predict which group would have the longermedian survival before collecting any data. Analysiswas by intention-to-treat. The association of baselinecharacteristics including age, gender, smoking, pres-ence of diabetes mellitus, medications (angiotensin-converting-enzyme inhibitors, b-blockers, calciumchannel blockers, lipid-lowering agents, erythropoi-etin), pre-existing cardiovascular disease, systolic anddiastolic blood pressure, and mode of dialysis (PD vs.HD) to the primary endpoint was tested using Coxproportional hazard model. In a stepwise forwardCox-regression analysis, variables with a P value of0.05 or less were retained. Analyses were performedwith GraphPad prism software (version 5.0, Graph-Pad Software,San Diego,CA,USA) or SPSS software(release 12.0.0, SPSS Inc., Chicago, IL, USA). Allstatistical tests were two-sided. Two-sided P-valuesless than 0.05 were considered to indicate statisticalsignificance.

A Maier et al.200

© 2009 The AuthorsJournal compilation © 2009 International Society for ApheresisTher Apher Dial, Vol. 13, No. 3, 2009

RESULTS

The clinical and biochemical characteristics ofpatients included in this case control study aresummarized in Tables 1 and 2. Patients with chronickidney disease stage 5 received either PD (cases) orHD (controls). In the PD group 7 out of 25 patients(28%) had diabetes mellitus, and in the HD group21 out of 75 patients (28%) had diabetes mellitus(P = 1.000). Months on dialysis, body mass index,smoking, hypertension, history of cardiovasculardisease, peripheral vascular disease, ischemic stroke,use of angiotensin converting enzyme inhibitors,beta-blockers, calcium channel blockers, lipid lower-ing agents, erythropoietin, systolic blood pressure,and diastolic blood pressure were not significantlydifferent between the two groups. Hemoglobin con-centrations and platelet counts were lower whereas

serum potassium and C-reactive protein levels werehigher in patients receiving HD compared to patientsreceiving PD. Other laboratory values were not sig-nificantly different between the two groups. Duringfollow up 7 out of 25 patients (28%) in the PD groupand 16 out of 75 patients (21%) in the HD groupobtained renal transplantation (P = 0.584). Duringfollow up 3 patients (12%) receiving PD switchedto HD. None of the patients receiving HD switchedto PD.

During the follow-up the primary endpoint, i.e.mortality of all causes, was reached in 32 out of 100patients (32%). Causes of death were classified ascardiovascular including sudden death (20 patients;62% of all causes of death), infection (6 patients;19%) and cancer (6 patients; 19%). During the followup of 58 months 6 out of 25 patients (24%) died in thePD group, whereas in the HD group 26 out of 75

TABLE 1. Clinical characteristics of patients with chronic kidney disease stage 5 who receive either continuous ambulatoryperitoneal dialysis (CAPD) or hemodialysis (HD)

Characteristic CAPD HD P-value

Age (years) 57 (51–62) 58 (52–61) 0.471Gender (male/female) 13/12 39/36 1.000Months on dialysis 24 (12–48) 12 (3–41) 0.077Body mass index (kg/m2) 25.0 (22.5–29.0) 23.9 (21.8–27.6) 0.491Smoking (yes/no) 10/15 29/46 1.000Hypertension (yes/no) 17/8 55/20 0.615Diabetes mellitus (yes/no) 7/18 21/54 1.000Cardiovascular disease (yes/no) 5/20 30/45 0.091Peripheral vascular disease (yes/no) 2/23 10/65 0.725Ischemic stroke (yes/no) 1/24 8/67 0.444Angiotensin converting enzyme inhibitors (yes/no) 10/15 16/59 0.112Beta-blockers (yes/no) 15/10 44/31 1.000Calcium channel blockers (yes/no) 13/12 22/53 0.053Lipid-lowering agents (yes/no) 6/19 23/52 0.617Erythropoietin (yes/no) 8/17 41/34 0.065Systolic blood pressure (mm Hg) 151 (135–167) 143 (130–160) 0.190Diastolic blood pressure (mm Hg) 85 (77–93) 80 (72–88) 0.132

Continuous data are reported as median (25% percentile—75% percentile) and compared using Mann-Whitney test. Categorical data areanalyzed by Fisher’s exact test.

TABLE 2. Laboratory characteristics of patients with chronic kidney disease stage 5 who receive either peritoneal dialysis(PD) or hemodialysis (HD)

Characteristic PD HD P-value

Blood urea nitrogen (mmol/L) 15.8 (12.3–24.6) 18.5 (15.4–26.0) 0.173Serum sodium (mmol/L) 138 (135–140) 137 (134–139) 0.500Serum potassium (mmol/L) 3.9 (3.5–5.1) 4.6 (4.2–5.1) 0.034Serum calcium (mmol/L) 2.45 (2.33–2.50) 2.27 (2.06–251) 0.096Serum phosphate (mmol/L) 1.47 (1.35–1.93) 1.59 (1.25–1.91) 0.842Total protein (g/L) 64 (55–67) 64 (58–69) 0.684C-reactive protein (mg/L) 21 (3–27) 36 (10–44) 0.004Parathyroid hormone 120 (7–204) 230 (3–315) 0.081Serum iron (mmol/L) 9 (4–12) 12 (4–16) 0.101Leukocytes (/nL) 7.5 (5.7–8.6) 8.0 (6.5–10.6) 0.149Hemoglobin (g/dL) 11.9 (10.5–13.3) 10.1 (8.9–12.0) 0.005Platelets (/nL) 294 (219–368) 244 (192–299) 0.035

Continuous data are reported as median (25% percentile–75% percentile) and compared using Mann-Whitney test.

Dialysis Mode and Mortality 201

© 2009 The AuthorsJournal compilation © 2009 International Society for Apheresis Ther Apher Dial, Vol. 13, No. 3, 2009

patients (35%) died (relative risk 0.69 [95% CI 0.32to 1.49]; P = 0.46). Figure 1 shows Kaplan-Meier esti-mates comparing survival in the PD group and in theHD group. Survival was not significantly differentbetween the groups as indicated by Mantel-Cox log-rank test (hazard ratio 0.52 [95% CI 0.25 to 1.10];P = 0.11).

The association of baseline characteristics includ-ing age, gender, smoking, presence of diabetes melli-tus, medications (angiotensin-converting-enzymeinhibitors, b-blockers, calcium channel blockers, lipid-lowering agents, erythropoietin), pre-existing cardio-vascular disease, systolic and diastolic blood pressure,and mode of dialysis was tested using a multivariateanalysis.Age (hazard ratio 1.09 [95% CI 1.03 to 1.15];P = 0.005) and diabetes mellitus (hazard ratio 2.87[95% CI 1.10 to 7.45]; P = 0.031) were associated tothe primary endpoint. The other baseline character-istics and mode of dialysis did not significantlypredict survival.

DISCUSSION

In the present case control study we showed thatsurvival was not significantly different betweenpatients with chronic kidney disease stage 5 receivingPD or HD. Furthermore, we showed that age andpresence of diabetes mellitus are more importantthan the mode of dialysis to predict survival in thesepatients.

We conducted a case control study to exclude theproblem of differences of ages or gender betweenthe groups. The strength of the present study is theadequate matching of both groups. We confirmedthat age, gender, presence of diabetes mellitus,

months on dialysis, body mass index, smoking, hyper-tension, history of cardiovascular disease, peripheralvascular disease, ischemic stroke, use of angiotensinconverting enzyme inhibitors, beta-blockers, calciumchannel blockers, lipid lowering agents, erythropoi-etin, systolic blood pressure, and diastolic blood pres-sure were not significantly different between the twogroups. We only noticed that hemoglobin concentra-tions were lower in patients on HD compared topatients on PD, a finding that has already beenobserved by others, although the underlying causeremains largely unknown (6–8,12). As indicated inthe literature an increased inflammatory responsemay be associated with lower hemoglobin levels andincreased oxidative stress has been observed inpatients receiving HD (13,14). In accordance withthat observation another marker of inflammation, i.e.C-reactive protein, was higher in patients on HDcompared to patients on PD, whereas serum iron andparathyroid hormone were not significantly differentbetween to two groups.

Differences between the HD group and PD groupmight have affected the previously reported differ-ences in mortality between the two modes of dialysis.In the cohort study reported by Ganesh et al. patientson PD were of younger age and had significantlyhigher body mass index compared to patients on HD.Furthermore, more males and more patients with dia-betes mellitus were allocated to PD compared to HD(6). In the cohort study reported by Vonesh et al.patients on PD were generally younger and weremore likely to have diabetes mellitus (7). In thecohort study reported by Jarr et al. patients receivingPD were younger and had a significantly better case-mix profile than did patients receiving HD (8).Furthermore, the importance of body fat mass foroutcome of patients with chronic kidney disease hasrecently been described. Snyder et al. showed in aretrospective cohort study that overweight and obesepatients on PD have longer survival than those withlower body mass index (15). A similar findingcould be obtained in a prospective cohort studywith patients on HD which were followed up for83 months. That study showed that low serum leptinconcentrations indicating reduced body fat mass wasan independent predictor of mortality in patients onHD (16). Taken together, these findings underscorethe need to adequately match patients when compar-ing survival between two modes of dialysis.

Patients that underwent renal transplantation orchange of renal replacement therapy were censoredon the day of that event and analysis was based onintention-to-treat models. This approach appropri-ately addresses the question of whether the mode of

0 12 24 36 48 600

25

50

75

100

PD 6 eventsHD 26 events

HR 0.52 [95% CI 0.25-1.10]p=0.11

Months

Sur

viva

l (%

)

FIG. 1. Kaplan-Meier curve of time to primary endpoint inpatients with chronic kidney disease stage 5 receiving peritonealdialysis (PD) or hemodialysis (HD). The primary endpoint wasmortality from any cause.

A Maier et al.202

© 2009 The AuthorsJournal compilation © 2009 International Society for ApheresisTher Apher Dial, Vol. 13, No. 3, 2009

dialysis may influence the prognosis of patients withchronic kidney disease stage 5 because the changeof treatment modality is accompanied by severalchanges including life style and medication, e.g.administration of immunosuppressive drugs. Further-more, within the limitations of an observational study,this analysis was an attempt to replicate the intention-to-treat analysis in a clinical trial.As already suggestedby Jaar et al. an analysis based on intention-to-treat isthe most important matter from the clinical point ofview, because the real therapeutic choice for the clini-cian and the patient occurs primarily at the time ofinitiation of dialysis, whereas future switches may bemotivated by treatment failures over which clinicianshave little control (8). Effects of transfer from perito-neal dialysis to hemodialysis could not be detectedin the present study, because only three patientsswitched treatment modality. However, technicalfailure may affect patients’ survival anyway.

Our results which were obtained in older patientsare in partial agreement with previously reportedresults in this subgroup of patients with chronickidney disease stage 5. In an intention to treat analysisof registry data from Canada (11) and Denmark (12)did not show a significant difference in survivalbetween PD and HD in older patients. On the otherhand, Vonesh et al. showed that among patients withdiabetes mellitus with no comorbidity HD was asso-ciated with a higher risk of death among youngerpatients and with a lower risk of death among olderpatients (7).

In both groups of patients cardiovascular diseasedeath was the major cause of death, accounting forabout 62% of total mortality. The increased burdenof cardiovascular mortality in patients with chronickidney disease stage 5 has been attributed to accel-erated atherosclerosis, hypertension, dyslipidemia,endothelial dysfunction, inflammation, and mode ofdialysis (17). As detailed by Ganesh et al., PD mayoffer patients better blood pressure control, lesshemodynamic shear stress, and avoidance of thepeaks and troughs in uremic toxins that are commonin HD. On the other hand HD might provide patientswith overall greater solute clearance and greaterremoval of potentially atherogenic uremic toxinscompared to PD (6).

A limitation of the present study is the number ofpatients receiving PD. It should be noted that accord-ing to available registry data from Berlin, Germany,only about 5% of patients with chronic kidney diseasestage 5 use PD therapy. To address that problem weused a case control study design because it is knownthat case control studies are useful for studying rareconditions. To counterbalance that limitation we

extended follow up to 58 months.Assuming a propor-tional hazard approach exposure to a specific treat-ment for a longer time should amplify putativedifferences between the two modes of dialysis topredict survival in these patients. Looking at KaplanMeier survival curves may give the impression thatthere may be a difference between HD and PD.However, the difference was not significant probablydue to limited sample size.

CONCLUSION

In summary, the present case control study showedthat age and comorbidities but not mode of dialysisare important to predict survival in patients withchronic kidney disease stage 5.

Acknowledgments: This work was supported by Ernstand Berta Grimmke-Stiftung, Düsseldorf, Germany.

REFERENCES

1. Hooi LS, Lim TO, Goh A et al. Economic evaluation of centrehaemodialysis and continuous ambulatory peritoneal dialysisin Ministry of Health hospitals, Malaysia. Nephrology(Carlton) 2005;10:25–32.

2. Canaud B, Chenine L, Leray-Moragués H, Wiesen H, Tetta C.Residual renal function and dialysis modality: is it really ben-eficial to preserve residual renal function in dialysis patients?Nephrology (Carlton) 2006;11:292–6.

3. Bargman JM. Rethinking the peritoneal dialysis prescription:results of recent studies. Nephrology Carlton 2006;11:85–9.

4. Molsted S, Prescott L, Heaf J, Eidemak I. Assessment andclinical aspects of health-related quality of life in dialysispatients and patients with chronic kidney disease. NephronClin Pract 2007;106:c24–33.

5. Lupo A, Cancarini G, Catizone L et al. Comparison of survivalin CAPD and hemodialysis: a multicenter study. Adv Perit Dial1992;8:136–40.

6. Ganesh SK, Hulbert-Shearon T, Port FK, Eagle K, Stack AG.Mortality differences by dialysis modality among incidentESRD patients with and without coronary artery disease. J AmSoc Nephrol 2003;14:415–24.

7. Vonesh EF, Snyder JJ, Foley RN, Collins AJ. The differentialimpact of risk factors on mortality in hemodialysis and perito-neal dialysis. Kidney Int 2004;66:2389–401.

8. Jaar BG, Coresh J, Plantinga LC et al. Comparing the risk fordeath with peritoneal dialysis and hemodialysis in a nationalcohort of patients with chronic kidney disease. Ann Intern Med2005;143:174–83.

9. Serkes KD, Blagg CR, Nolph KD, Vonesh EF, Shapiro F.Comparison of patient and technique survival in continuousambulatory peritoneal dialysis (CAPD) and hemodialysis: amulticenter study. Perit Dial Int 1990;10:15–19.

10. Suzuki T, Kanno Y, Nakamoto H, Okada H, Sugahara S,Suzuki H. Peritoneal dialysis versus hemodialysis: a five-yearcomparison of survival and effects on the cardiovascularsystem, erythropoiesis, and calcium metabolism. Adv Perit Dial2003;19:148–54.

11. Schaubel DE, Morrison HI, Fenton SS. Comparing mortalityrates on CAPD/CCPD and hemodialysis. The Canadianexperience: fact or fiction. Perit Dial Int 1998;18:478–84.

12. Heaf JG, Lokkegaard H, Madsen M. Initial survival advantageof peritoneal dialysis relative to haemodialysis. Nephrol DialTransplant 2002;17:112–17.

Dialysis Mode and Mortality 203

© 2009 The AuthorsJournal compilation © 2009 International Society for Apheresis Ther Apher Dial, Vol. 13, No. 3, 2009

13. Horl WH, Vanrenterghem Y, Canaud B et al. Optimal treat-ment of renal anaemia (OPTA): improving the efficacy andefficiency of renal anaemia therapy in haemodialysis patientsreceiving intravenous epoetin. Nephrol Dial Transplant2005;20 (Suppl. 3):iii25–32.

14. Tepel M, van der Giet M, Statz M, Jankowski J, Zidek W. Theantioxidant acetylcysteine reduces cardiovascular events inpatients with end-stage renal failure: a randomized, controlledtrial. Circulation 2003;107:992–5.

15. Snyder JJ, Foley RN, Gilbertson DT, Vonesh EF, Collins AJ.Body size and outcomes on peritoneal dialysis in the UnitedStates. Kidney Int 2003;64:1838–44.

16. Scholze A, Rattensperger D, Zidek W, Tepel M. Low serumleptin predicts mortality in patients with chronic kidneydisease stage 5. Obesity (Silver Spring) 2007;15:1617–22.

17. Locatelli F, Marcelli D, Conte F et al. Cardiovascular diseasein chronic renal failure: the challenge continues. Nephrol DialTransplant 2000;15(Suppl. 5):69–80.

A Maier et al.204

© 2009 The AuthorsJournal compilation © 2009 International Society for ApheresisTher Apher Dial, Vol. 13, No. 3, 2009