sctcweoec

An Update on the Comparisons ofMortality Outcomes of Hemodialysis

and Peritoneal Dialysis PatientsYi-Wen Chiu, MD,*,† Sirin Jiwakanon, MD,*,‡ Lilia Lukowsky, MPH,*Uyen Duong, MD, MPH,* Kamyar Kalantar-Zadeh, MD, PhD,*,§ and

Rajnish Mehrotra, MD*,§

Summary: The number of dialysis patients continues to grow. In many parts of the world,peritoneal dialysis (PD) is a less expensive form of treatment. However, it has been questionedwhether patients treated with PD can have as good a long-term outcome as that achieved withhemodialysis (HD). This skepticism has fueled ongoing comparisons of outcomes of patientstreated with in-center HD and PD using data from national registries or prospective cohort studies.There are major challenges in comparing outcomes with two therapies when the treatmentassignment is nonrandom. Furthermore, many of the intermodality comparisons include patientswho started dialysis therapy in the 1990s. In many parts of the world, improvements in PDoutcome have outpaced those seen with in-center HD. It is not surprising, then, that virtually allthe recent observational studies from different parts of the world consistently show that long-termsurvival of HD and PD patients is remarkably similar. These studies support the case for a greateruse of PD for the treatment of end-stage renal disease. This, in turn, could allow more patients tobe treated for any given budgetary allocation to long-term dialysis.Semin Nephrol 31:152-158 © 2011 Elsevier Inc. All rights reserved.Keywords: End-stage renal disease, peritoneal dialysis, hemodialysis, mortality, epidemiology

hIatattNOaslrispTecyrt

Ever since the initial successful experi-ence with continuous ambulatory peri-toneal dialysis (PD), a large number of

tudies have tried to determine if the out-omes of PD patients are comparable withhose achieved with hemodialysis (HD). Aontrolled clinical trial in which patientsere assigned randomly to treatment with

ither PD or HD would be the best way tobtain unbiased estimates of the independentffects of dialysis modality on patient out-omes. However, the two dialysis modalities

*Los Angeles Biomedical Research Institute, Torrance, CA.

†Kaohsiung Medical University Hospital, Kaohsiung, Taiwan.

‡Hatyai Hospital, Hatyai, Songkhla, Thailand.

§David Geffen School of Medicine at University of California Los Angeles,Los Angeles, CA.

Supported in part by grant DK077341 from the National Institutes of Health(R.M. and K.K.-Z.), Baxter Health Care (R.M.), and DaVita Inc. (R.M. andK.K.-Z.). Rajnish Mehrotra has received research grants, served as ad hocconsultant, and received honoraria from Baxter Health Care.

Address reprint requests to Rajnish Mehrotra, MD, 1124 W. Carson St, Tor-rance, CA 90502. E-mail: rmehrotra@labiomed.org

t0270-9295/ - see front matter© 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.semnephrol.2011.01.004

Seminars152

ave disparate effects on patients’ daily lives.t is not surprising then that when patientsre educated about their modality options,hey often want to have a say in the choicend refuse to be randomized. The last attempto conduct a randomized, controlled, clinicalrial was made under the auspices of Theetherlands Cooperative Study of Dialysis.nly 38 (5%) of the 773 eligible subjects

greed to be randomized and, hence, thetudy was substantially underpowered to al-ow any meaningful conclusions.1 The mostecent attempt currently is being undertakenn China: a pilot study has been completeduccessfully and the clinical trial is antici-ated to begin enrollment in 2011 (Clinical-rials.gov identifier: NCT00510549). How-ver, the success and the results of thislinical trial will not be known for severalears. Until then, one has to depend on theesults of observational studies that have at-empted to compare the outcomes of patients

reated with PD and HD.

in Nephrology, Vol 31, No 2, March 2011, pp 152-158

coo

ictcosstIsdd

rwo

oab

Mortality among HD and PD patients 153

INTERMODALITYCOMPARISONS USING OBSERVATIONALSTUDIES: CHALLENGES AND PITFALLS

The observational studies that have sought tocompare the outcomes of PD and HD patients canbe grouped into two broad categories: prospec-tive cohort studies and those that use data fromnational registries of dialysis patients. The majoradvantage of prospective cohort studies is thatthey contain detailed information about patientcharacteristics and hence allow for a more com-prehensive adjustment for these differences.However, because such studies are expensive toundertake, the statistical power is limited becausethey include a relatively small number of subjects.On the other hand, national dialysis registrieshave substantially greater statistical power be-cause they often include several thousand pa-tients. However, such studies are limited by thepaucity of medical information about each indi-vidual patient. There are no recent prospectivecohort studies that have compared the outcomesof PD and HD patients, both the Choices forHealth Outcomes In Caring for ESRD (CHOICE)and The Netherlands Cooperative Study of Dialy-sis studies in the United States and The Nether-lands, respectively, enrolled patients who begandialysis therapy in the 1990s.2,3 Thus, the onlyontemporary intermodality comparisons are thenes that have used data from national registriesf dialysis patients.4-9

The inherent limitations of observational stud-es arise mainly from substantial demographic,linical, and psychosocial differences between pa-ients treated with PD and HD. Experience inomparing the outcomes of PD and HD patientsver the past 2 decades has led to clearer under-tanding of the challenges and pitfalls of suchtudies; some of these challenges and the poten-ial solutions for study design are listed in Table 1.ndeed, studies that did not adequately addressome of these challenges and pitfalls have pro-uced results that are inconsistent and not repro-ucible.10-12

Over the years, it has been appreciated that theelative risk of death for PD patients (as comparedith those treated with HD) varies over their timen dialysis.13,14 Hence, it is best to study only

incident patients and when reporting results to

make a distinction between early and late out- i

comes. If only incident patients are used, thequestion that follows is, how should the initialdialysis modality be defined? Categorizing pa-tients based on the dialysis modality on the firstday of renal replacement therapy has severalproblems. First, HD is the initial dialysis modalityfor many new PD patients. This is often, althoughnot always, a result of delayed referral to neph-rologists. Second, in the United States, previouslyuninsured but Medicare-eligible in-center HD pa-tients get insurance coverage from day 90 of end-stage renal disease. Consequently, the administra-tive datasets are not as reliable for the first 90 daysas they are thereafter. Finally, it is possible thatmany of the early deaths are secondary to theunderlying co-existing illnesses, and becausemany of these patients are unable to performhome dialysis, they are treated with in-center HD.For all these reasons, dialysis modality on day 90,with continuous treatment using that modality for60 days (60-day rule), is considered the initialdialysis modality.

The most vexing problem, however, is thenonrandom assignment of patients to the twodialysis modalities. Including demographic, clini-cal, and laboratory-related variables as covariatesin statistical models are probably not sufficient toaccount for the differences in patients who aretreated with one dialysis modality compared withthe other. Hence, propensity scores have beenused increasingly to compare therapies when theassignment is nonrandom.15 This score is theprobability of each subject being treated with agiven dialysis modality and is calculated usingdemographic, clinical, laboratory, and socioeco-nomic data.16 Any two subjects, one treated withPD and the other with HD, with the same propen-sity score are deemed to be randomly assigned totheir dialysis modality. Propensity scores can beused to match subjects, perform stratified analysisbased on quartiles or quintiles of such scores, orbe used as covariates in survival analyses. Manyrecent intermodality comparisons have used pro-pensity scores, although these reduce bias theydo not completely eliminate it.2,8,9

In addition to interaction with time, relativeutcomes of patients treated with different di-lysis modalities vary depending on age, dia-etic status, and presence/absence of co-exist-

ng illnesses.17 Hence, any subgroup analysis

154 Y.-W. Chiu et al

Table 1. Pitfalls and Challenges and Proposed Solutions When Performing Studies Comparingthe Outcomes of End-Stage Renal Disease Patients Treated With HD or PD

Pitfalls/Challenges Potential Solution

Change in relative risk over time (timeinteraction)

Include incident patients or those of recent vintage; breakdown results for different follow-up periods.

Definition of initial dialysis modality,make allowances for modalityswitches in the first 90-120 days ofend-stage renal disease

Modality on day 90 and continuous treatment with themodality for at least 60 days; the use of day 90 isdictated, in part, by administrative reasons in theUnited States, and hence is arbitrary

On the other hand, defining modality on the first day ofrenal replacement therapy has the potential tointroduce bias in favor of PD, it will exclude patientswho are referred late and start PD after a variableperiod of treatment with HD

Selection of dialysis modalities isnonrandom

Calculate propensity scores that numerically describe theprobability for a patient to be treated with a givendialysis modality based on known patientcharacteristics; outcomes can be analyzed in subgroupsbased on propensity scores or included as a covariate(directly or its inverse) in statistical models

Intent-to-treat or as-treated analysis Perform both and interpret clearlyFor as-treated analysis, how should

deaths after change in dialysismodality be handled?

Allow a grace period (generally 60 days) such that deathsthat occur within that period are assigned to theprevious dialysis modality

Subgroup analyses and statisticalinteractions

Several studies have shown statistically significantinteractions with age, diabetic status, and presence/absence of comorbidity Present results after stratifyingthe study population into subgroups based on thesevariables

Problem of overadjustment, includingvariables that may mediate thepathways of higher (or lower) riskwith different dialysis modalities

Exercise care in including variables in models that maymediate the higher (or lower) risk with each dialysismodality; for example, higher C-reactive protein levelsin HD patients may be a result of the use of a venouscatheter, which, in turn, may mediate the higher riskseen early with HD

Alternatively, lower serum albumin level is a consequenceof PD itself (secondary to peritoneal albumin loss) andinclusion of values obtained after the start of dialysismay bias results in favor of PD

Residual confounding: inability toadjust for unknown factors that maydifferentially affect the comparisonof outcomes

None

Transplantation rates of patientstreated with different dialysismodalities

Only the healthiest patients undergo renal transplantationand if transplantation rates vary by dialysis modality itmay bias results against the modality with highertransplantation rates (by removing the healthiestpatients from the cohort at a faster rate)

Use weights that account for the probability of beingcensored for transplantation

drpgcvl

wsmm

ataracco

CI

Sstta1tpitlaH

bapov

ttp1ttPfmhas

Mortality among HD and PD patients 155

necessarily needs to account for these interac-tions. Finally, the reasons why patients reachthe end of follow-up evaluation (censoring)may vary by dialysis modality. This, in turn, hasthe potential to influence survival comparisons.For example, the adjusted transplantation rate inthe United States is about 50% higher for patientstreated with PD than for patients treated withHD.9 Because only the healthiest patients un-ergo transplantation, this differential transplantates has the potential to bias the results of com-arison of outcomes of PD and HD patients. Areater efflux of healthier patients from the PDohort may explain the loss of initial survival ad-antage seen in PD patients over longer-term fol-ow-up evaluation.13,14 Conversely, longer trans-plant waiting times is likely to manifest as animprovement in survival, this may be one of manypotential explanations for a greater improvementin outcomes of PD patients than has been ob-served for those treated with HD.18 Use of

eights based on the inverse probability of cen-oring because of transplantation can be used toinimize bias and have been incorporated inore recent analyses.9,19

Notwithstanding these advances in statisticalnalyses, even the most sophisticated statisticalools cannot definitively answer a key question,re any of the differences in survival seen in suchegistry-based comparisons attributable to the di-lysis modality or to differences in patients whohose a given dialysis modality? It is with thisaveat that one should examine the current bodyf literature in this field.

ONTEMPORARYNTERMODALITY COMPARISONS

tarting from the early 1980s, a large number oftudies have compared the survival of patientsreated with PD and HD. Over the past 3 decades,he study design has evolved from single-centernd multicenter studies in the 1980s and early990s, to either prospective cohort studies orhose using data from national registries of dialysisatients thereafter. The early studies produced

nconsistent results, it was largely through at-empts to explain these inconsistencies that haveed to our current understanding of challengesnd pitfalls when comparing survival of PD and

D patients (as reviewed in the preceding sec- 1

tion). Indeed, registry-based studies from Canada,the United States, and Denmark that includedpatients who began renal replacement therapy inthe 1990s took care to avoid the common pitfalls,and came to remarkably similar conclusions (re-viewed previously20). These studies consistentlyshowed a lower risk for death for PD patientsearly during the course of end-stage renal disease(1-3 years, depending on the country).13,14,17,21

Over time, this survival advantage dissipated and,particularly in the United States, the long-termmortality risk for PD patients was higher than thatfor those treated with HD. Furthermore, this mo-dality risk relationship was modified by three im-portant variables: age, diabetic status, and pres-ence/absence of co-existing illnesses.17 Thus, the

etter the overall health of a patient, the greaternd longer was the survival advantage for PDatients. Conversely, the poorer the overall healthf a patient, the lesser and shorter was the sur-ival advantage with PD.17 Thus, most studies

indicated a robust short- and long-term survivaladvantage with PD for nondiabetic patients withno other comorbidity; older diabetic patients hada demonstrably worse long-term outcome, partic-ularly in studies from the United States. Notwith-standing these consistent observations, it hasbeen difficult to conclude whether these differ-ences are attributable to the dialysis modality.

However, the practice of both PD and HD haschanged considerably from what it was in theearly 1990s. Starting from the mid-1990s, the sur-vival of patients treated with both dialysis modal-ities has improved, albeit at different rates.22

Much greater gains have been made in the out-comes of PD patients than have been made inthose of HD patients.18 This is particularly true forhe survival of patients in the first 12 months ofreatment. Although the first-year mortality of PDatients has decreased significantly since the mid-990s, there have been no significant changes inhe first-year mortality of HD patients.18 Similarrends with greater improvements in outcomes ofD patients have been reported by investigatorsrom Canada, Denmark, and Spain at internationaleetings, however, findings from these countriesave not been published to date. These consider-tions make a strong argument against using theurvival data of patients who began dialysis in the

990s to make clinical decisions today. Studies

EA

156 Y.-W. Chiu et al

that are based, at least in part, on patients whostarted dialysis after 2000 are summarized in Ta-ble 2.4-9 As is apparent, they span countries fromurope (The Netherlands), Asia (Taiwan), Southmerica (Colombia), Oceania (Australia and New

Table 2. Studies Comparing the Mortality OIncluded Patients Who Started Treatment on

Study(year)

Cohort Period/Country

SecularTrends

Inclusion Criteria/Sample Size

Liem et al4

(2007)1987-2002

The NetherlandsNo 16,643

(HD 10,841; PD 5,802)

Huang etal5

(2008)

1995-2002Taiwan

No 48,629(HD 45,820; PD 2,809)

Sanabriaet al6

(2008)

2001-2003Colombia

No 923(HD 437; PD 486)

McDonaldet al7

(2009)

1991-2005Australia andNew Zealand

Yes 25,287(HD 14,733; PD10,554)

Weinhandlet al8

(2010)

2003United States

No 6,337 pairs(HD 6,337; PD 6,337)

Mehrotraet al9

(2011)

1996-2004United States

Yes 684,426(HD 620,020; PD

64,406)

Abbreviations: BMI, body mass index; ESRD, end-stage renal disease

Zealand), and North America (the United States).The conclusions of these studies from differentparts of the world is again remarkably consistent,if the analyses are restricted to patients whostarted dialysis in the 2000s, there is no difference

mes of Incident HD and PD Patients Thatter 2000djusted for the

FollowingFollow-Up

Period Key Results

x, renal diagnosis,of first RRT, andsis center

Up to 16years

In younger diabetic andnondiabetic patients, lowerrisk for PD patients for thefirst 15 months; no differencethereafter

In older nondiabetic patients,lower risk for PD patients inthe first 6 months, but higherrisk after the first 15 months

In older diabetic patients, nodifference in early death buthigher risk for PD patientsafter the first 15 months

x, selectedrbidity andtic status

Up to 6 years Overall similar 5-year (HD, 54%;PD, 56%) and 10-year survival(HD, 34%; PD, 35%);subgroup analysis showedhigher risk for death with PDamong all diabetic patients,and older nondiabetic patients(�55 y)

x, socioeconomics, education,ical insurance,score, selectedrbidity andatory variables,of ESRD

Up to Dec2005

No difference in overall adjustedmortality rates between HDand PD; lower death risk foryoung nondiabetic patientstreated with PD but similaroutcomes in all other groups

x, BMI, race andted comorbidity,propensity scores

Up to 12/31/05

Overall 11% lower risk for deathfor PD patients in the firstyear, but 33% higher riskafter the first 12 months; earlysurvival advantage with PDseen only in young patientswithout comorbidities

In the most recent cohort(2004), no difference in long-term mortality of HD and PDpatients

x, race andted comorbidity,propensity scores

Up to 4 years Overall mortality risk was 8%lower for PD patients; similaradjusted 4-year survival (HD,48%; PD, 47%)

x, race andted comorbidity,propensity scores

Up to 5 years No significant difference in the5-year adjusted survival of HDand PD patients (35% and33%, respectively)

Lower risk for death for youngernondiabetic PD patients;higher death risk for olderdiabetic patients (particularlythose with additionalcomorbidity) treated with PD

nal replacement therapy; SGA, subjective global assessment.

utcoor AfData A

Age, seyeardialy

Age, secomodiabe

Age, sestatumedSGAcomolaborcause

Age, seselecand

Age, seselecand

Age, seselecand

; RRT, re

sttgpwlottctippaiaa

dIPlawiTPaal

Mortality among HD and PD patients 157

in the survival (for up to 4-5 years) of patientswho begin treatment with either PD or HD. In theUnited States, the adjusted 5-year survival of PDand HD patients who started dialysis in 2002 to2004 was 33% and 35%, respectively, and themedian life expectancy was 37 and 38 months,respectively.9 Consistent with these findings, an-other study showed that the adjusted 4-year sur-vival of PD and HD patients who started dialysis in2003 in the United States was 47% and 48%,respectively.8 The trends in subgroups are theame as had been reported for patients who werereated in the 1990s. However, the survival advan-age of younger, nondiabetic patients has becomereater and the higher risk among older diabeticatients treated with PD has lessened comparedith those treated with HD.9 At this time, there is

imited evidence to determine why the outcomesf PD patients have improved more than that ofhose treated with HD. Some possible explana-ions for the differential improvement in out-omes are summarized in Table 3. Greater atten-ion to PD prescription and reduction innfectious complications are the two leading ex-lanations for the improvement in survival of PDatients. The former includes a greater use ofutomated PD. Although two large registry stud-es have shown that both continuous ambulatorynd automated PD provide similar outcomes over-ll,23,24 rapid transporters treated with automated

PD have a better survival than with continuousambulatory PD.25 A selective movement of rapidtransporters to automated PD could have contrib-uted to the improvement in outcomes of the

Table 3. Possible Explanations for the DifferenWith HD and PD in the United States

Related to Dialysis Practices

PD-relatedBetter and individualized PD prescriptionmanagement over the yearsReduced risk of infectious complicationsMore widespread use of quality-improvement pMore attention to maintenance of normal volum

HD-relatedGreater and longer use of tunneled venous ca

remaining patients treated with continuous ambu-latory PD and this, in turn, could have contributedto the overall improvement in outcomes of PDpatients.

CONCLUSIONS

Contemporary studies suggest that the effect ofdialysis modality on patient survival is rathersmall, if any. It follows then that survival studiesshould not be used in making decisions about theappropriate dialysis modality for an individual pa-tient. This should, instead, be based on a discus-sion about patients’ expectations about their life-style, which in turn can be made only after earlyand iterative dialysis modality education.26 Studieshave suggested that most patients starting dialysisin the United States often are unaware of alterna-tives to in-center HD, and a randomized con-trolled trial indicated that comprehensive modal-ity education increases selection of self-caredialysis.27,28 Hence, comprehensive dialysis mo-ality education is likely to expand the use of PD.t is important to note that the improvement inD outcomes and the consequent similarity in

ong-term outcomes of patients treated with PDnd HD has been reported from countries with aide range of PD uptake, from as low as 6% to 7%

n the United States to almost 50% in Colombia.hese studies support the case for greater use ofD in the treatment of end-stage renal disease,nd this, in turn, could allow more patients withny given budgetary allocation to be treated withong-term dialysis.

provement in Outcomes of Patients Treated

Unmeasured Confounding

mstus

rs

Residual confounding because patientsstarting PD are younger and healthierthan in previous years

Longer waiting times for transplantation:healthier patients remain dialysis-dependent for longer periods of time

Because PD patients have highertransplantation rates, outcomes of thiscohort are more likely to be affected

tial Im

rograe sta

thete

1

1

1

1

1

1

2

2

2

2

2

2

2

2

2

158 Y.-W. Chiu et al

REFERENCES1. Korevaar JC, Feith GW, Dekker FW, van Manen JG,

Boeschoten EW, Bossuyt PM, et al. Effect of startingwith hemodialysis compared with peritoneal dialysisin patients new on dialysis treatment: a randomizedcontrolled trial. Kidney Int. 2003;64:2222-8.

2. Jaar BG, Coresh J, Plantinga LC, Fink NE, Klag MJ,Levey AS, et al. Comparing the risk for death withperitoneal dialysis and hemodialysis in a national co-hort of patients with chronic kidney disease. AnnIntern Med. 2005;143:174-83.

3. Termorshuizen F, Korevaar JC, Dekker FW, VanManen JG, Boeschoten EW, Krediet RT, The Nether-lands Cooperative Study on the Adequacy of DialysisStudy Group. Hemodialysis and peritoneal dialysis:comparison of adjusted mortality rates according tothe duration of dialysis: analysis of The NetherlandsCooperative Study on the Adequacy of Dialysis 2.J Am Soc Nephrol. 2003;14:2851-60.

4. Liem YS, Wong JB, Hunink MG, de Charro FT, Win-kelmayer WC. Comparison of hemodialysis and peri-toneal dialysis survival in The Netherlands. KidneyInt. 2007;71:153-8.

5. Huang CC, Cheng KF, Wu HD. Survival analysis: com-paring peritoneal dialysis and hemodialysis in Taiwan.Perit Dial Int. 2008;28 Suppl 3:S15-20.

6. Sanabria M, Munoz J, Trillos C, Hernandez G, LatorreC, Diaz CS, et al. Dialysis outcomes in Colombia(DOC) study: a comparison of patient survival onperitoneal dialysis vs hemodialysis in Colombia. Kid-ney Int. 2008;73:Suppl 108:S165-72.

7. McDonald SP, Marshall MR, Johnson DW, Polking-horne KR. Relationship between dialysis modality andmortality. J Am Soc Nephrol. 2009;20:155-63.

8. Weinhandl ED, Foley RN, Gilbertson DT, Arneson TJ,Snyder JJ, Collins AJ. Propensity-matched mortalitycomparison of incident hemodialysis and peritonealdialysis patients. J Am Soc Nephrol. 2010;21:499-506.

9. Mehrotra R, Chiu YW, Kalantar-Zadeh K, Bargman J,Vonesh E. Hemodialysis and peritoneal dialysis are asso-ciated with similar outcomes for end-stage renal diseasetreatment in United States. Arch Intern Med. 2011;171:110-8.

10. Bloembergen WE, Port FK, Mauger EA, Wolfe RA. Acomparison of mortality between patients treatedwith hemodialysis and peritoneal dialysis. J Am SocNephrol. 1995;6:177-83.

11. Ganesh SK, Hulbert-Shearon T, Port FK, Eagle K, StackAG. Mortality differences by dialysis modality amongincident ESRD patients with and without coronary ar-tery disease. J Am Soc Nephrol. 2003;14:415-24.

12. Stack AG, Molony DA, Rahman NS, Dosekun A, Mur-thy B. Impact of dialysis modality on survival of newESRD patients with congestive heart failure in theUnited States. Kidney Int. 2003;64:1071-9.

13. Fenton SS, Schaubel DE, Desmeules M, Morrison HI,Mao Y, Copleston P, et al. Hemodialysis versus peri-toneal dialysis: a comparison of adjusted mortality

rates. Am J Kidney Dis. 1997;30:334-42.

4. Collins AJ, Hao W, Xia H, Ebben JP, Everson SE, Con-stantini EG, et al. Mortality risks of peritoneal dialysisand hemodialysis. Am J Kidney Dis. 1999;34:1065-74.

5. Rubin DB. Estimating causal effects from large datasets using propensity scores. Ann Intern Med. 1997;127:757-63.

6. D’Agostino RB Jr. Propensity score methods for biasreduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17:2265-81.

7. Vonesh EF, Snyder JJ, Foley RN, Collins AJ. The differ-ential impact of risk factors on mortality in hemodialysisand peritoneal dialysis. Kidney Int. 2004;66:2389-401.

8. Mehrotra R, Kermah D, Fried L, Kalantar-Zadeh K, Kha-war O, Norris K, et al. Chronic peritoneal dialysis in theUnited States: declining utilization despite improvingoutcomes. J Am Soc Nephrol. 2007;18:2781-8.

9. van der Wal WM, Noordzij M, Dekker FW, Boescho-ten EW, Krediet RT, Korevaar JC, et al. Comparingmortality in renal patients on hemodialysis versusperitoneal dialysis using a marginal structural model.Int J Biostat. 2010;6:1-19.

0. Khawar O, Kalantar-Zadeh K, Lo WK, Johnson D,Mehrotra R. Is the declining use of long-term perito-neal dialysis justified by outcome data? Clin J Am SocNephrol. 2007;2:1317-28.

1. Heaf JG, Lokkegaard H, Madsen M. Initial survivaladvantage of peritoneal dialysis relative to haemodi-alysis. Nephrol Dial Transplant. 2002;17:112-7.

2. United States Renal Data System. Annual Data Report.Bethesda: US Department of Public Health and Hu-man Services, Public Health Service, National Insti-tutes of Health; 2009.

3. Badve SV, Hawley CM, McDonald SP, Mudge DW,Rosman JB, Brown FG, et al. Automated and contin-uous ambulatory peritoneal dialysis have similar out-comes. Kidney Int. 2008;73:480-8.

4. Mehrotra R, Chiu YW, Kalantar-Zadeh K, Vonesh E.The outcomes of continuous ambulatory and auto-mated peritoneal dialysis are similar. Kidney Int.2009;76:97-107.

5. Johnson DW, Hawley CM, McDonald SP, Brown FG,Rosman JB, Wiggins KJ, et al. Superior survival of hightransporters treated with automated versus continuousambulatory peritoneal dialysis. Nephrol Dial Transplant.2010;25:1973-9.

6. Morton RL, Tong A, Howard K, Snelling P, WebsterAC. The views of patients and carers in treatmentdecision making for chronic kidney disease: system-atic review and thematic synthesis of qualitative stud-ies. BMJ. 2010;340:c112.

7. Mehrotra R, Marsh D, Vonesh E, Peters V, NissensonA. Patient education and access of ESRD patients torenal replacement therapies beyond in-center hemo-dialysis. Kidney Int. 2005;68:378-90.

8. Manns BJ, Taub K, VanderStraeten C, Jones H, Mills C,Visser M, et al. The impact of education on chronickidney disease patients’ plans to initiate dialysis withself-care dialysis: a randomized trial. Kidney Int. 2005;

68:1777-83.