0
We're unable to sign you in at this time. Please try again in a few minutes.
Retry
We were able to sign you in, but your subscription(s) could not be found. Please try again in a few minutes.
Retry
There may be a problem with your account. Please contact the AMA Service Center to resolve this issue.
Contact the AMA Service Center:
Telephone: 1 (800) 262-2350 or 1 (312) 670-7827  *   Email: subscriptions@jamanetwork.com
Error Message ......
Original Investigation | ONLINE FIRST

Similar Outcomes With Hemodialysis and Peritoneal Dialysis in Patients With End-Stage Renal Disease FREE

Rajnish Mehrotra, MD; Yi-Wen Chiu, MD; Kamyar Kalantar-Zadeh, MD; Joanne Bargman, MD; Edward Vonesh, PhD
[+] Author Affiliations

Author Affiliations: Division of Nephrology and Hypertension, Los Angeles Biomedical Research Institute at Harbor–UCLA Medical Center, Torrance, California (Drs Mehrotra, Chiu, and Kalantar-Zadeh); Department of Medicine, David Geffen School of Medicine, University of California, Los Angeles (Drs Mehrotra and Kalantar-Zadeh); Department of Medicine, Kaohsiung Medical University, Kaohsiung, Taiwan (Dr Chiu); Division of Nephrology, University Health Network, Toronto, Ontario, Canada (Dr Bargman); and Department of Preventive Medicine, Northwestern University Feinberg School of Medicine, Chicago, Illinois (Dr Vonesh).


Arch Intern Med. 2011;171(2):110-118. doi:10.1001/archinternmed.2010.352.
Text Size: A A A
Published online

At the end of 2008, there were over 500 000 Americans living with end-stage renal disease (ESRD).1 Even though patients with ESRD constitute less than 1% of the Medicare population, they account for 5.8% of the expenses for the program (excluding the costs for the prescription benefit).1For eligible subjects, renal transplant is associated with a higher life expectancy, better quality of life, and substantially lower long-term expenses compared with continued treatment with dialysis.1,2 However, limited availability of organ donors means that most patients with ESRD are treated with dialysis for prolonged periods; in-center hemodialysis (HD) and home peritoneal dialysis (PD) are the 2 most common forms of dialysis therapy. In 2007, the annual, per-person costs for PD patients were almost $20 000 lower than that of HD patients; the substantial cost advantage for PD is robust even when data are adjusted for better health and higher rates of transfer to HD for PD patients.1,3 Recognizing the potential for cost savings, the Centers for Medicare & Medicaid Services has continued to provide financial incentives to promote greater use of PD. However, the relative use of PD for the treatment of ESRD has continued to decline, and in 2008, only 7% of dialysis patients were treated with this dialysis modality.1

Even though the determinants of the use of dialysis modality are largely nonmedical, there has been considerable interest in understanding if there is any difference in the mortality outcomes of HD and PD patients.4,5 A randomized controlled trial is probably the best way to determine if the differences in outcomes between HD and PD patients are attributable to the dialysis therapy. However, given the disparate effects of the 2 dialysis modalities on patients' lifestyles, attempts to date at randomizing patients to the different dialysis modalities have been unsuccessful.6 Numerous observational studies have been undertaken, including those from various national registries from different parts of the world.4 These comparisons have highlighted the difficulties in comparing dialysis modalities when the allocation to therapy is not random. To exemplify the complexity of these comparisons, at least 4 clinically important statistical interactions have been identified that affect the comparison of risk between HD and PD—change in relative risk over time and different relative risks based on age, diabetic status, and coexisting diseases.4 Thus, patients treated with PD have a lower risk for death in the first few years of dialysis therapy, and this is greatest in young, nondiabetic individuals with no coexisting illnesses.

We have recently demonstrated that over the last 8 years, there has been a consistent and substantial reduction in mortality rates for new patients staring PD in the United States.7 In contrast, no such improvements were observed for HD patients.7 This differential change in outcomes mandates a re-examination of outcomes of HD and PD using contemporary cohorts. We undertook this study to test the hypothesis that initial dialysis modality has no effect on the life expectancy of patients with ESRD in the United States using marginal structural models.

DATA SOURCE

The study protocol was reviewed and approved as exempt by the institutional review board at the Los Angeles Biomedical Research Center, Los Angeles, California. The data for all patients with incident ESRD in the United States over a 9-year period (1996-2004) were obtained from the patient and MEDVID files of the US Renal Data System (USRDS), the national registry for all patients with ESRD. The MEDVID file contains data from the Medical Evidence Form 2728, a form required to be filed for every new patient with ESRD in the United States and contains demographic and clinical information at the start time of renal replacement therapy. The data were linked to the RXHIST60 file, which uses information from claims data and other sources to provide start and end dates for each treatment modality and to assign dialysis modality, and to the Facility File for dialysis unit characteristics.

DEFINITIONS

For individuals eligible but not receiving Medicare benefits prior to the first dialysis treatment, coverage for in-center hemodialysis begins on day 90; thus, complete information is available from that point onward. Hence, as per convention, the dialysis modality 90 days after the first service date and continuous treatment for at least 60 days (“60-d rule”), was considered to be the initial modality. Presence or absence of various coexisting illnesses and the initial laboratory results were obtained from the Medical Evidence Form 2728. The unit affiliation was defined as the dialysis facility where the patient was being treated on day 90 of ESRD. Data for each patient were linked with the facility data from the same year as the one in which the patient first started renal replacement therapy. For this analysis, only patients whose initial modality on day 90 was either in-center HD, continuous ambulatory PD, or automated PD were included; home HD or “other” PD patients were excluded.

COVARIATES

For all statistical analyses, the models were adjusted for demographics (age, sex, race, and current employment status), facility characteristics (period-prevalent patient census, and for-profit or not-for-profit status), cause of ESRD, 10 different comorbid conditions (cardiac arrest or dysrhythmia, cerebrovascular disease, congestive heart failure, ischemic heart disease or myocardial infarction, peripheral vascular disease, limited activities of daily living, chronic obstructive pulmonary disease, current smokers, diabetes primary or contributing, and malignant neoplasm), baseline estimated glomerular filtration rate, body mass index, and selected laboratory variables (serum albumin, blood urea nitrogen, and hemoglobin).

STATISTICAL ANALYSES

Baseline characteristics of HD and PD patients were compared using Pearson χ2 tests for categorical variables and the unpaired t test for continuous variables. Patients were followed up until death, transplant, or transfer to home HD or “other PD” treatment, whichever happened first, or last follow-up examination (September 30, 2007). Patients who switched modality after 90 days were treated in survival analysis according to their modality on day 90. A detailed description of the statistical methods is presented in the eAppendix. Supplemental data are presented in eTables 1 and 2 and eFigures 1 through 7.

Nonproportional hazards models using a piecewise exponential survival model were used to compare case mix–adjusted survival of HD and PD patients.8,9 Outcomes were compared using a marginal structural model with inverse probability of treatment and censoring weighting (IPTCW).10,11 As a first step, propensity scores (PSs) were calculated as a single summary measure of confounding. The PS is the estimated probability of being treated initially with PD, given a patient's set of baseline demographic, clinical, and laboratory characteristics; adjustment for PS is sufficient to remove bias due to these measured confounding variables.12 Proportional and nonproportional hazards marginal structural models use weighted regression with weights equal to 1/PS (ie, inverse probability of treatment weighting) to provide hazard ratio (HR) estimates that are unbiased estimates of a causal treatment effect under the unverifiable assumption that there are no unmeasured confounding variables.1012 While this assumption is likely not to hold in observational studies like ours, the use of proportional and nonproportional hazards marginal structural models will tend to minimize the bias associated with each and every measured confounding variable in our study. By further weighting on the inverse probability of remaining uncensored, marginal structural models using IPTCW provide further protection against any differential selection bias resulting from transplant and other censoring mechanisms.10,11 The transplant rates, adjusted for all measured baseline characteristics, in the United States are 30% to 60% higher for PD patients than for HD patients with incident ESRD (eFigure 5). Robust standard errors based on an empirical sandwich estimator of the variance-covariance matrix of regression estimates were used to calculate confidence intervals. Additional details about the methods including model assessment and sensitivity analyses are provided in the eAppendix.

Results were summarized through up to 5 years of follow-up and are presented in terms of adjusted HRs and adjusted population survival curves. Adjusted population survival curves were obtained by exponentiation of the adjusted cumulative hazard functions for PD and HD. Based on these survival curve estimates, median life expectancies were computed using life table methodology.13 Hazard ratios were computed under a proportional hazards marginal structural model as a means for summarizing the overall relative risk of death between PD and HD over the course of a 5-year follow-up. Absolute risks are presented in terms of the 1-, 2-, 3-, 4-, and 5-year adjusted survival estimates for PD and HD. All P values for comparing survival between PD and HD were adjusted to reflect simultaneous inference over a 5-year follow-up using the Sidak method.14

Finally, consistent with previous findings, we performed a set of a priori subgroup analyses in which the impact of treatment modality on outcome was examined within 8 subgroups of patients defined by age, diabetes, and presence or absence of comorbid conditions.15 All analyses were performed using the SAS statistical software package version 9.2 (SAS Institute Inc, Cary, North Carolina).

PATIENT CHARACTERISTICS

Over the 9-year study period, 684 426 patients started dialysis therapy in the United States, survived at least 90 days, and had a Medical Evidence Form 2728 submitted (HD patients, 620 020; PD patients, 64 406). Given the large sample size, the differences in all patient characteristics between HD and PD patients (except sex distribution) were statistically significant (Table 1). The PD patients with incident ESRD were younger, more likely to be white, and less likely to have other comorbidities. Even though the differences in the baseline laboratory variables were statistically significant, they were not clinically meaningful (Table 1).

Table Graphic Jump LocationTable 1. Selected Characteristics of HD and PD Patients With Incident ESRD in the United States (1996-2004) at the Start of Renal Replacement Therapy

Preliminary analyses revealed that the concordance index for the logistic regression model used to compute PSs was 0.766, indicating a reasonably good prediction of modality selection. Moreover, there was excellent balance with respect to all measured covariates between PD and HD patients having similar PSs (eFigures 1-3), suggesting that PS adjustment via a marginal structural model using IPTCW was sufficient in removing the majority of confounding due to all measured baseline covariates.

OVERALL COMPARISON OF OUTCOMES OF HD AND PD PATIENTS WITH INCIDENT ESRD BY COHORT PERIOD

Adjusted survival of HD and PD patients at 1 year-intervals for up to 5 years of follow-up for each of the 3 cohort periods are summarized in Table 2. Adjusted HRs comparing the relative risk of death for PD and HD patients with incident ESRD over 5 years of follow-up for each of the 3 cohort periods (1996-1998, 1999-2001, and 2002-2004) are summarized in Table 3. For the 2002-2004 patients with incident ESRD, there was no significant difference in the risk for death between those treated with HD or PD through 5 years (HR, 1.03; 95% confidence interval [CI], 0.99-1.06 [P = .10]). The adjusted median life expectancy increased for both HD and PD patients with incident ESRD and was very similar for the 2002-2004 cohort (HD and PD, 38.4 vs 36.6 months, respectively). Improvement in outcomes over time—with no difference in outcomes among the 2002-2004 cohort of patients treated with the 2 dialysis modalities—was observed in the adjusted population survival curves (Figure 1).

Place holder to copy figure label and caption
Figure 1.

Adjusted population survival curves comparing the outcome of peritoneal dialysis (PD) and hemodialysis (HD) patients with incident end-stage renal disease in the United States stratified by cohort period. A, 1996-1998 Cohort: adjusted median life expectancy, 37.2 months for HD patients and 31.7 months for PD patients; B, 1999-2001 cohort: adjusted median life expectancy, 37.3 months for HD patients and 33.0 months for PD patients; C, 2002-2004 cohort: adjusted median life expectancy, 38.4 months for HD patients and 36.6 months for PD patients; and D, overall: adjusted median life expectancy, 37.6 months for HD patients and 33.7 months for PD patients.

Graphic Jump Location
Table Graphic Jump LocationTable 2. Adjusted Survival of Patients Treated With HD and PD for Up to 5 Years in Each of 3 Cohort Periods
Table Graphic Jump LocationTable 3. Adjusted Hazard Ratios (HRs) for Mortality, Stratified by Cohort Periods, Using Marginal Structural Models With Inverse Probability of Treatment and Censoring Weightinga
OUTCOME OF HD AND PD PATIENTS WITH INCIDENT ESRD STRATIFIED BY AGE, DIABETIC STATUS, AND COMORBIDITY FOR 3 COHORT PERIODS

Results of the comparisons of patients treated with HD and PD in 8 subgroups for each of the 3 cohort periods are summarized in Table 4, and the adjusted population survival curves are presented in Figures 2, 3, 4, and 5. Among younger nondiabetic patients with no additional comorbidity, the relative risk for death for patients treated with PD progressively decreased; from 1999 onwards, patients treated with PD had a significantly lower risk for death compared with those treated with HD. Furthermore, the higher risk for death among older patients with additional comorbidity who were treated with PD dissipated over time, such that for the 2002-2004 incident cohort, there was no significant difference in risk for death in patients treated with HD or PD.

Place holder to copy figure label and caption
Figure 2.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were younger than 65 years with no additional comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location
Place holder to copy figure label and caption
Figure 3.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were younger than 65 years and had 1 or more comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location
Place holder to copy figure label and caption
Figure 4.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were 65 years or older with no additional comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location
Place holder to copy figure label and caption
Figure 5.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were 65 years or older and had 1 or more comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location
Table Graphic Jump LocationTable 4. Hazard Ratios (HRs) for Mortality in 3 Different Cohort Periods in 8 Subgroups, Based on Age, Diabetic Status, and Level of Baseline Comorbidity Adjusted Using IPTCW

Diabetic patients who started dialysis between 1996 and 2001 and were treated with PD had a significantly higher risk for death irrespective of age or additional comorbidity. The relative risk for death of diabetic patients treated with PD, who started dialysis during 2002-2004, was lower than that seen in previous years for each of the 4 subgroups. During this period, there was no significant difference in the risk for death for younger diabetic patients with no additional comorbidity who were treated with HD or PD. The higher risk of death for PD patients was seen in the other 3 subgroups of diabetic patients; the HRs were lower than that seen in the previous years.

In the largest study to our knowledge to date, we were unable to demonstrate any significant overall difference in outcomes of patients with ESRD who began treatment with either HD or PD in 2002-2004—the most contemporary cohort for which data are available. Furthermore, progressive improvements in outcomes of PD patients (relative to HD patients) were seen in virtually all of the 8 subgroups examined.

Studies from different parts of the world have compared the outcomes of patients with ESRD treated with HD vs PD.1527 Even though there is heterogeneity in the results of these comparisons, there are some consistent themes. Patients treated with PD have been shown to have a lower risk for death early during the course of ESRD—the magnitude and length of time over which this lower risk for death is evident depends on age, diabetes status, and the presence of associated comorbidity. After the first few years of treatment, many studies have also shown a higher risk of death for patients treated with PD. However, virtually all the studies to date have enrolled patients with incident ESRD from the 1990s. Recent data have shown a differential improvement in outcomes of HD and PD patients, and thus it is probably inappropriate to use the results of studies of older cohorts when making therapeutic decisions in the present era.7,25 It is this differential change in outcomes of HD and PD patients that formed the rationale for us to study the secular trends in relative outcomes of patients treated with the 2 dialysis modalities. To our knowledge, only 4 published studies that have compared the outcomes of HD and PD patients with incident ESRD have included subjects who started dialysis treatment after 20002325,28; however, secular trends between PD and HD were not examined in 3 of the 4 studies (from the Netherlands, Taiwan, and the United States).23,24,28 Data from the Australia and New Zealand registry are consistent with our findings, and in that study the adjusted relative risk for death of PD patients also decreased over time such that no significant difference in outcomes of HD and PD patients were seen among those who commenced dialysis therapy in 2005.25 Consistent with the findings from Australia and New Zealand, no significant difference in the risk of death is demonstrable among patients with incident ESRD in 2002-2004 treated with either HD or PD in the United States for up to 5 years. Improvements in relative outcomes were observed in each of the 8 subgroups analyzed. Thus, a higher risk of death seen particularly in diabetic patients with additional comorbidity, though still present in the most recent cohorts, was progressively attenuated over time. In contrast, the lower risk of death among younger nondiabetic patients with no additional comorbidity persisted with progressively lower HRs in the most recent cohorts.

In addition to examining secular trends, our study has significant additional strengths. First, to our knowledge, it is the largest study done in this field to date. Our findings are relevant because we also examined secular trends of outcomes in 8 subgroups of patients. Furthermore, because our study includes data from all centers in the United States, our findings have considerable external validity.

Second, marginal structural models such as those used in this study provide robust estimates in observational studies, at least to the extent to which the measured baseline characteristics capture information relevant to all known and unknown confounders. This approach has not been used in either registry or cohort studies that have compared outcomes of HD and PD in the United States.

Third, to our knowledge, this is the first study from the United States that has adjusted for the probability of censoring. In survival studies of dialysis patients, renal transplant is one of the most common reasons for censoring participants. However, this has the risk of introducing substantial bias when considering the outcomes of maintenance dialysis patients. The unadjusted transplant rate for PD patients with incident ESRD is more than 2-fold higher than for HD patients.7 In this study, the adjusted transplant rate for incident PD patients was still 30% to 60% higher. This differential transplant rate between PD and HD suggests that some degree of selection bias due to censoring for transplant does exist and that perhaps healthier patients are removed at a faster rate over time from the PD cohorts when compared with HD cohorts. It is conceivable that a higher transplant rate may have accounted for some of the apparent increase in risk of death in patients treated with PD with increasing dialysis vintage. Furthermore, the transplant waiting times have become longer in many parts of the country. The prolongation in transplant waiting times would lead to inclusion of a larger number of healthier PD patients for longer periods in the later cohorts than HD patients (since PD patients have a higher transplant rate). This may have explained the differential improvement in outcomes of HD and PD patients.7 In the marginal structural models used in this study, we weighted the analysis by the inverse probability of censoring, and this allowed us to adjust for selection bias from unmeasured confounders associated with censoring. This is a considerable strength of the present study and further reduces bias.

Fourth, rather than focusing on time-dependent HRs, we focused on the accumulated effects of the initial choice of dialysis modality by performing an “intent-to-treat” survival analysis based on the dialysis modality on day 90. The rate of transfer of PD patients to HD is higher than that of HD patients to PD.7 An analysis that does not consider these transfers, as performed herein, allows us to consider the effect of initial stable treatment modality on day 90 and encompasses the higher risk associated with the transfer of patients from one modality to the other. Moreover, it effectively preserves the conditional randomization property that PSs impart on PD vs HD comparisons. Specifically, the set of all measured baseline characteristics will be conditionally independent of modality selection given the PS. This means that PD and HD patients with similar PSs will be similar with respect to all measured baseline characteristics as well as any unmeasured confounders that are strongly associated with those characteristics.

Our study does not allow us to determine the causes of the differential improvement in outcomes of HD and PD patients. A greater reduction in risk for infectious complications, greater improvements in prescription management of PD patients, and a more selective assignment of patients to the therapy are possible explanations for our findings. The important question to consider is if the overall equivalency of outcomes will persist if a larger proportion of patients with incident ESRD begin treatment with PD. The use of PSs and the marginal structural model with IPTCW reduces the possibility of confounding; however, in observational studies the possibility of residual confounding remains. However, a larger proportion of patients with incident ESRD are treated with PD in Canada (18%) and Australia and New Zealand (42%), and similar outcomes have been reported with the 2 dialysis therapies in the most recent data.25 Nevertheless, the effect of expanding PD use on outcomes of patients remains speculative.

Our study is not without limitations. The assignment of patients to the 2 therapies was not random, and thus one has to be cautious before inferring causality. However, randomization of patients to 2 therapies with disparate effects on lifestyle is challenging, and previous attempts at conducting a randomized controlled comparison of HD and PD have been unsuccessful.6 It is unlikely that a randomized, controlled comparison of the 2 dialysis therapies will be undertaken in the industrialized world. Furthermore, information on additional diseases was obtained from the Medical Evidence Form 2728, a source that has been shown to lead to underestimation of the comorbidity.29 In addition, data on only baseline comorbidity were available. As a result of these reasons, we cannot exclude residual confounding. The 2002-2004 cohort has limited 5-year follow-up in that only those patients with incident ESRD from January-September of 2002 would have a maximum follow-up of 5 years. Finally, we did not consider transfers from one dialysis modality to the other. Thus, the equivalency of outcomes pertains only to the dialysis modality on day 90, and no conclusions can be made regarding the direct effects of the modality.

In conclusion, in this study we demonstrate a reduction in the adjusted relative risk of death of PD patients in the United States compared with those beginning treatment with HD. In the most recent cohorts, the life expectancy of patients treated with either HD or PD on day 90 of ESRD was remarkably similar. The lower costs of peritoneal dialysis and equivalent outcomes with the 2 therapies provide support for a larger use of PD for the treatment of ESRD in the United States, particularly in subgroups in which the patients treated with PD have similar or lower risk for death when compared with HD (nondiabetic and younger diabetic patients with no additional comorbidity; almost two-thirds of patients with incident ESRD). However, the improvement in PD outcomes may have been a result of more selective assignment of patients to the therapy over the last decade. Thus, should such an expansion of PD use be undertaken, close monitoring of outcomes of patients treated with different dialysis modalities should continue.

Correspondence: Rajnish Mehrotra, MD, Division of Nephrology and Hypertension, Harbor-UCLA Medical Center, 1124 W Carson St, Torrance, CA 90502 (rmehrotra@labiomed.org).

Accepted for Publication: April 27, 2010.

Published Online: September 27, 2010. doi:10.1001/archinternmed.2010.352

Author Contributions:Study concept and design: Mehrotra, Bargman, and Vonesh. Analysis and interpretation of data: Mehrotra, Chiu, Kalantar-Zadeh, Bargman, and Vonesh. Drafting of the manuscript: Mehrotra and Vonesh. Critical revision of the manuscript for important intellectual content: Mehrotra, Chiu, Kalantar-Zadeh, Bargman, and Vonesh. Statistical analysis: Kalantar-Zadeh and Vonesh. Obtained funding: Mehrotra. Administrative, technical, and material support: Chiu.

Financial Disclosure: Dr Mehrotra has received grant support, honoraria, and served as an ad hoc consultant for Baxter Health Care. Dr Bargman has received honoraria from Baxter Health Care. Dr Vonesh is a consultant for Baxter Healthcare.

Funding/Support: This work was supported by research grants from Baxter Health Care and National Institutes of Health (grant R21 DK077341). Dr Mehrotra is also supported by grant RR18298 from the National Institutes of Health and by the American Heart Association. Drs Mehrotra and Kalantar-Sadeh are supported by a grant from DaVita Inc.

Disclaimer: These data were supplied by the USRDS and the findings do not represent the opinion of the US government or the USRDS.

Additional Contributions: Eileen Eriksen managed the data obtained from the USRDS.

US Renal Data System,Annual Data Report. Bethesda, MD US Dept of Public Health and Human Services, Public Health Service, National Institutes of Health2009;
Wolfe  RAAshby  VBMilford  EL  et al.  Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med 1999;341 (23) 1725- 1730
PubMed Link to Article
Shih  YCGuo  AJust  PMMujais  S Impact of initial dialysis modality and modality switches on Medicare expenditures of end-stage renal disease patients. Kidney Int 2005;68 (1) 319- 329
PubMed Link to Article
Khawar  OKalantar-Zadeh  KLo  WKJohnson  DMehrotra  R Is the declining use of long-term peritoneal dialysis justified by outcome data? Clin J Am Soc Nephrol 2007;2 (6) 1317- 1328
PubMed Link to Article
Mehrotra  R The John F. Maher Award Recipient Lecture 2006. The continuum of chronic kidney disease and end-stage renal disease: challenges and opportunities for chronic peritoneal dialysis in the United States. Perit Dial Int 2007;27 (2) 125- 130
PubMed
Korevaar  JCFeith  GWDekker  FW  et al. NECOSAD Study Group, Effect of starting with hemodialysis compared with peritoneal dialysis in patients new on dialysis treatment: a randomized controlled trial. Kidney Int 2003;64 (6) 2222- 2228
PubMed Link to Article
Mehrotra  RKermah  DFried  L  et al.  Chronic peritoneal dialysis in the United States: declining utilization despite improving outcomes. J Am Soc Nephrol 2007;18 (10) 2781- 2788
PubMed Link to Article
Holford  TR The analysis of rates and of survivorship using log-linear models. Biometrics 1980;36 (2) 299- 305
PubMed Link to Article
Vonesh  ESchaubel  DEHao  WCollins  AJ Statistical methods for comparing mortality among ESRD patients: examples of regional/international variations. Kidney Int 2000;57 ((suppl 74)) S19- S27
Link to Article
Robins  JMHernán  MABrumback  B Marginal structural models and causal inference in epidemiology. Epidemiology 2000;11 (5) 550- 560
PubMed Link to Article
Hernán  MABrumback  BRobins  JM Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology 2000;11 (5) 561- 570
PubMed Link to Article
Rosenbaum  PRRubin  DB The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70 (1) 41- 55
Link to Article
Lee  ETStatistical Methods for Survival Data Analysis. Belmont, CA Lifetime Learning Publications1980;
Sidak  Z Rectangular confidence regions for the means of multivariate normal distributions. J Am Stat Assoc 1967;62 (318) 626- 633
Vonesh  EFSnyder  JJFoley  RNCollins  AJ The differential impact of risk factors on mortality in hemodialysis and peritoneal dialysis. Kidney Int 2004;66 (6) 2389- 2401
PubMed Link to Article
Nelson  CBPort  FKWolfe  RAGuire  KE Comparison of continuous ambulatory peritoneal dialysis and hemodialysis patient survival with evaluation of trends during the 1980s. J Am Soc Nephrol 1992;3 (5) 1147- 1155
PubMed
Held  PJPort  FKTurenne  MNGaylin  DSHamburger  RJWolfe  RA Continuous ambulatory peritoneal dialysis and hemodialysis: comparison of patient mortality with adjustment for comorbid conditions. Kidney Int 1994;45 (4) 1163- 1169
PubMed Link to Article
Fenton  SSSchaubel  DEDesmeules  M  et al.  Hemodialysis versus peritoneal dialysis: a comparison of adjusted mortality rates. Am J Kidney Dis 1997;30 (3) 334- 342
PubMed Link to Article
Vonesh  EFMoran  J Mortality in end-stage renal disease: a reassessment of differences between patients treated with hemodialysis and peritoneal dialysis. J Am Soc Nephrol 1999;10 (2) 354- 365
PubMed
Collins  AJHao  WXia  H  et al.  Mortality risks of peritoneal dialysis and hemodialysis. Am J Kidney Dis 1999;34 (6) 1065- 1074
PubMed Link to Article
Xue  JLChen  SCEbben  JP  et al.  Peritoneal and hemodialysis: 1, differences in patient characteristics at initiation. Kidney Int 2002;61 (2) 734- 740
PubMed Link to Article
Heaf  JGLøkkegaard  HMadsen  M Initial survival advantage of peritoneal dialysis relative to haemodialysis. Nephrol Dial Transplant 2002;17 (1) 112- 117
PubMed Link to Article
Liem  YSWong  JBHunink  MGde Charro  FTWinkelmayer  WC Comparison of hemodialysis and peritoneal dialysis survival in the Netherlands. Kidney Int 2007;71 (2) 153- 158
PubMed Link to Article
Huang  CCCheng  KFWu  HD Survival analysis: comparing peritoneal dialysis and hemodialysis in Taiwan. Perit Dial Int 2008;28 ((suppl 3)) S15- S20
PubMed
McDonald  SPMarshall  MRJohnson  DWPolkinghorne  KR Relationship between dialysis modality and mortality. J Am Soc Nephrol 2009;20 (1) 155- 163
PubMed Link to Article
Jaar  BGCoresh  JPlantinga  LC  et al.  Comparing the risk for death with peritoneal dialysis and hemodialysis in a national cohort of patients with chronic kidney disease. Ann Intern Med 2005;143 (3) 174- 183
PubMed Link to Article
Termorshuizen  FKorevaar  JCDekker  FWVan Manen  JGBoeschoten  EWKrediet  RTNetherlands Cooperative Study on the Adequacy of Dialysis Study Group, Hemodialysis and peritoneal dialysis: comparison of adjusted mortality rates according to the duration of dialysis: analysis of the Netherlands Cooperative Study on the Adequacy of Dialysis 2. J Am Soc Nephrol 2003;14 (11) 2851- 2860
PubMed Link to Article
Weinhandl  EDFoley  RNGilbertson  DTArneson  TJSnyder  JJCollins  AJ Propensity-matched mortality comparison of incident hemodialysis and peritoneal dialysis patients. J Am Soc Nephrol 2010;21 (3) 499- 506
PubMed Link to Article
Longenecker  JCCoresh  JKlag  MJ  et al.  Validation of comorbid conditions on the end-stage renal disease medical evidence report: the CHOICE study: choices for Healthy Outcomes in Caring for ESRD. J Am Soc Nephrol 2000;11 (3) 520- 529
PubMed

Figures

Place holder to copy figure label and caption
Figure 1.

Adjusted population survival curves comparing the outcome of peritoneal dialysis (PD) and hemodialysis (HD) patients with incident end-stage renal disease in the United States stratified by cohort period. A, 1996-1998 Cohort: adjusted median life expectancy, 37.2 months for HD patients and 31.7 months for PD patients; B, 1999-2001 cohort: adjusted median life expectancy, 37.3 months for HD patients and 33.0 months for PD patients; C, 2002-2004 cohort: adjusted median life expectancy, 38.4 months for HD patients and 36.6 months for PD patients; and D, overall: adjusted median life expectancy, 37.6 months for HD patients and 33.7 months for PD patients.

Graphic Jump Location
Place holder to copy figure label and caption
Figure 2.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were younger than 65 years with no additional comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location
Place holder to copy figure label and caption
Figure 3.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were younger than 65 years and had 1 or more comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location
Place holder to copy figure label and caption
Figure 4.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were 65 years or older with no additional comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location
Place holder to copy figure label and caption
Figure 5.

Adjusted population survival curves comparing the outcomes of PD and HD patients with incident ESRD who were 65 years or older and had 1 or more comorbidity, stratified by DM status and cohort period. A-C, ESRD + non-DM cohort. D-F, ESRD + DM cohort. DM indicates diabetes mellitus; ESRD, end-stage renal disease; HD, hemodialysis; and PD, peritoneal dialysis.

Graphic Jump Location

Tables

Table Graphic Jump LocationTable 1. Selected Characteristics of HD and PD Patients With Incident ESRD in the United States (1996-2004) at the Start of Renal Replacement Therapy
Table Graphic Jump LocationTable 2. Adjusted Survival of Patients Treated With HD and PD for Up to 5 Years in Each of 3 Cohort Periods
Table Graphic Jump LocationTable 3. Adjusted Hazard Ratios (HRs) for Mortality, Stratified by Cohort Periods, Using Marginal Structural Models With Inverse Probability of Treatment and Censoring Weightinga
Table Graphic Jump LocationTable 4. Hazard Ratios (HRs) for Mortality in 3 Different Cohort Periods in 8 Subgroups, Based on Age, Diabetic Status, and Level of Baseline Comorbidity Adjusted Using IPTCW

References

US Renal Data System,Annual Data Report. Bethesda, MD US Dept of Public Health and Human Services, Public Health Service, National Institutes of Health2009;
Wolfe  RAAshby  VBMilford  EL  et al.  Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med 1999;341 (23) 1725- 1730
PubMed Link to Article
Shih  YCGuo  AJust  PMMujais  S Impact of initial dialysis modality and modality switches on Medicare expenditures of end-stage renal disease patients. Kidney Int 2005;68 (1) 319- 329
PubMed Link to Article
Khawar  OKalantar-Zadeh  KLo  WKJohnson  DMehrotra  R Is the declining use of long-term peritoneal dialysis justified by outcome data? Clin J Am Soc Nephrol 2007;2 (6) 1317- 1328
PubMed Link to Article
Mehrotra  R The John F. Maher Award Recipient Lecture 2006. The continuum of chronic kidney disease and end-stage renal disease: challenges and opportunities for chronic peritoneal dialysis in the United States. Perit Dial Int 2007;27 (2) 125- 130
PubMed
Korevaar  JCFeith  GWDekker  FW  et al. NECOSAD Study Group, Effect of starting with hemodialysis compared with peritoneal dialysis in patients new on dialysis treatment: a randomized controlled trial. Kidney Int 2003;64 (6) 2222- 2228
PubMed Link to Article
Mehrotra  RKermah  DFried  L  et al.  Chronic peritoneal dialysis in the United States: declining utilization despite improving outcomes. J Am Soc Nephrol 2007;18 (10) 2781- 2788
PubMed Link to Article
Holford  TR The analysis of rates and of survivorship using log-linear models. Biometrics 1980;36 (2) 299- 305
PubMed Link to Article
Vonesh  ESchaubel  DEHao  WCollins  AJ Statistical methods for comparing mortality among ESRD patients: examples of regional/international variations. Kidney Int 2000;57 ((suppl 74)) S19- S27
Link to Article
Robins  JMHernán  MABrumback  B Marginal structural models and causal inference in epidemiology. Epidemiology 2000;11 (5) 550- 560
PubMed Link to Article
Hernán  MABrumback  BRobins  JM Marginal structural models to estimate the causal effect of zidovudine on the survival of HIV-positive men. Epidemiology 2000;11 (5) 561- 570
PubMed Link to Article
Rosenbaum  PRRubin  DB The central role of the propensity score in observational studies for causal effects. Biometrika 1983;70 (1) 41- 55
Link to Article
Lee  ETStatistical Methods for Survival Data Analysis. Belmont, CA Lifetime Learning Publications1980;
Sidak  Z Rectangular confidence regions for the means of multivariate normal distributions. J Am Stat Assoc 1967;62 (318) 626- 633
Vonesh  EFSnyder  JJFoley  RNCollins  AJ The differential impact of risk factors on mortality in hemodialysis and peritoneal dialysis. Kidney Int 2004;66 (6) 2389- 2401
PubMed Link to Article
Nelson  CBPort  FKWolfe  RAGuire  KE Comparison of continuous ambulatory peritoneal dialysis and hemodialysis patient survival with evaluation of trends during the 1980s. J Am Soc Nephrol 1992;3 (5) 1147- 1155
PubMed
Held  PJPort  FKTurenne  MNGaylin  DSHamburger  RJWolfe  RA Continuous ambulatory peritoneal dialysis and hemodialysis: comparison of patient mortality with adjustment for comorbid conditions. Kidney Int 1994;45 (4) 1163- 1169
PubMed Link to Article
Fenton  SSSchaubel  DEDesmeules  M  et al.  Hemodialysis versus peritoneal dialysis: a comparison of adjusted mortality rates. Am J Kidney Dis 1997;30 (3) 334- 342
PubMed Link to Article
Vonesh  EFMoran  J Mortality in end-stage renal disease: a reassessment of differences between patients treated with hemodialysis and peritoneal dialysis. J Am Soc Nephrol 1999;10 (2) 354- 365
PubMed
Collins  AJHao  WXia  H  et al.  Mortality risks of peritoneal dialysis and hemodialysis. Am J Kidney Dis 1999;34 (6) 1065- 1074
PubMed Link to Article
Xue  JLChen  SCEbben  JP  et al.  Peritoneal and hemodialysis: 1, differences in patient characteristics at initiation. Kidney Int 2002;61 (2) 734- 740
PubMed Link to Article
Heaf  JGLøkkegaard  HMadsen  M Initial survival advantage of peritoneal dialysis relative to haemodialysis. Nephrol Dial Transplant 2002;17 (1) 112- 117
PubMed Link to Article
Liem  YSWong  JBHunink  MGde Charro  FTWinkelmayer  WC Comparison of hemodialysis and peritoneal dialysis survival in the Netherlands. Kidney Int 2007;71 (2) 153- 158
PubMed Link to Article
Huang  CCCheng  KFWu  HD Survival analysis: comparing peritoneal dialysis and hemodialysis in Taiwan. Perit Dial Int 2008;28 ((suppl 3)) S15- S20
PubMed
McDonald  SPMarshall  MRJohnson  DWPolkinghorne  KR Relationship between dialysis modality and mortality. J Am Soc Nephrol 2009;20 (1) 155- 163
PubMed Link to Article
Jaar  BGCoresh  JPlantinga  LC  et al.  Comparing the risk for death with peritoneal dialysis and hemodialysis in a national cohort of patients with chronic kidney disease. Ann Intern Med 2005;143 (3) 174- 183
PubMed Link to Article
Termorshuizen  FKorevaar  JCDekker  FWVan Manen  JGBoeschoten  EWKrediet  RTNetherlands Cooperative Study on the Adequacy of Dialysis Study Group, Hemodialysis and peritoneal dialysis: comparison of adjusted mortality rates according to the duration of dialysis: analysis of the Netherlands Cooperative Study on the Adequacy of Dialysis 2. J Am Soc Nephrol 2003;14 (11) 2851- 2860
PubMed Link to Article
Weinhandl  EDFoley  RNGilbertson  DTArneson  TJSnyder  JJCollins  AJ Propensity-matched mortality comparison of incident hemodialysis and peritoneal dialysis patients. J Am Soc Nephrol 2010;21 (3) 499- 506
PubMed Link to Article
Longenecker  JCCoresh  JKlag  MJ  et al.  Validation of comorbid conditions on the end-stage renal disease medical evidence report: the CHOICE study: choices for Healthy Outcomes in Caring for ESRD. J Am Soc Nephrol 2000;11 (3) 520- 529
PubMed

Correspondence

CME


You need to register in order to view this quiz.
Submit a Comment

Multimedia

Supplemental Content
Supplemental Content

Some tools below are only available to our subscribers or users with an online account.

Web of Science® Times Cited: 95

Related Content

Customize your page view by dragging & repositioning the boxes below.

See Also...
Articles Related By Topic
Related Collections
PubMed Articles
JAMAevidence.com

Care at the Close of Life EDUCATION GUIDES
Practical Considerations in Dialysis Withdrawal


Hemodialysis