Peritoneal dialysis (PD) use varies greatly across regions, which can be understood by reviewing the following steps in the process of care: identification of all potential candidates, assessment for PD eligibility, determination of eligibility, offer of PD, choice of PD, and receipt of PD. This framework can be used to categorize interventions that are designed to increased PD use. In general, PD and HD are felt to be equivalent with respect to clinical outcomes and quality of life, and PD is cost saving compared with HD. Of all the steps, patient choice, if provided, is a major modifiable factor explaining variance in use. However, with an increasing focus on patient-centered care, it is worth noting that patients usually have strong preferences for one therapy over another, and this fact must be reconciled with home-first policies.
Keywordscost effectiveness, eligibility, hospitalization, mortality, peritoneal dialysis, patient choice, shared decision making, patient-centered care
Peritoneal Dialysis Use, 470
Framework for Understanding Peritoneal Dialysis Use, 470
Identification of Potential Peritoneal Dialysis Patients, 470
Assessment of Peritoneal Dialysis Eligibility, 471
Offer and Choice, 472
Insertion of the Peritoneal Dialysis Catheter, 473
Receipt of Peritoneal Dialysis, 473
Outcomes of Peritoneal Dialysis Therapy, 473
Attempts at Randomized Comparisons of Peritoneal Dialysis and Hemodialysis, 473
Data Validity, 474
Comparing the Outcomes of Patients Treated With Peritoneal Dialysis and Hemodialysis: Survival, 474
Comparing the Outcomes of Patients Treated With Peritoneal Dialysis and Hemodialysis: Hospitalization, 477
Comparing the Outcomes of Patients Treated With Peritoneal Dialysis and Hemodialysis: Cost, 477
Comparing the Outcomes of Patients Treated With Peritoneal Dialysis and Hemodialysis: Quality of Life, 478
Peritoneal dialysis (PD) is associated with equivalent outcomes to hemodialysis but can be provided in most countries at significantly lower costs. Individuals receiving PD generally report good quality of life and high satisfaction with their care and can live a more independent lifestyle. Given the current pressures on healthcare budgets, PD is actively promoted in many countries. The stated justification for these policies is “to increase patient access to care, control costs through lower infrastructure and capital investments, empower patients, and optimize treatment provision”.
However, the use of PD varies widely among regions, suggesting that optimization of PD use is a complex and variable process. It is important to understand the variance in use among regions, recent changes in PD use, and what factors may explain the observed variance. It is also relevant to understand how the outcomes of patients on PD compare with those on hemodialysis, given that PD is promoted in many countries.
Peritoneal Dialysis Use
Jain et al. conducted a comprehensive review of PD use between 1997 and 2008 by contacting a representative from the national dialysis registries for each country. PD use was reported as number of patients on PD, PD incidence, PD prevalence per million population, and PD prevalence as a percentage of dialysis patients. Focusing on the latter statistic, countries can be classified into high use (50% of greater), moderate use (10% to 50% use) and low use (less than 10%) ( Table 30.1 ).
|Country||Prevalent Use of PD (%)|
|Use between 10% and 50%|
This study, although comprehensive in its methods, does not reflect the most recent PD use statistics, and some countries have implemented policies to increase PD use since 2008. For example, in the United States, the Center for Medicare and Medicaid Services (CMS) implemented an expanded prospective payment system that bundled most of the services provided to dialysis patients starting January 1, 2011 to promote PD use. Prevalent PD use, according to the US Renal Data System (USRDS) renal report, has increased from 7.2% to 9.8% from 2010 to 2014.
In Thailand a PD-first policy was implemented in 2008 that dramatically increased prevalent PD use from 9.5% in 2008 to 30.0% in 2014, according to the Thailand Renal Replacement 2014 report. In Ontario, Canada, the Ministry of Health and Long-Term Care began actively promoting PD in 2005. The policy was followed by the formation of a renal agency called the Ontario Renal Network in 2011, which has a strategic goal to increase home dialysis. Prevalent PD use has grown modestly from 18% in 2005 to 19.4% in 2015.
Framework for Understanding Peritoneal Dialysis Use
The use of PD can be explained by a variety of frameworks. For example, Liu et al. identifies the importance of government policies, economic factors, and provider and patient education. Other explanations focus on system, provider, and patient factors. Our preferred framework is to focus on the process of care that requires patients to pass through six steps to successfully start PD therapy. At each step, the impact of policies, education, and provider and patient factors can be considered.
Identification of Potential Peritoneal Dialysis Patients
PD use will be increased in programs that have a robust intake and accounting process to identify all potential PD candidates. The majority of patients are followed in predialysis clinics by nephrologists or interdisciplinary teams. The receipt of predialysis care is strongly associated with PD use. For example, a predialysis education program implemented by Fresenius Medical Care in the United States was associated with an adjusted odds ratio (OR) of 5.13 (95% confidence interval [CI], 3.58 to 7.35) for selecting PD.
In our quality improvement system (DMAR system) we have found that interdisciplinary meetings conducted every 2 weeks to identify and review all new patients starting dialysis can greatly improve the identification of potential PD patients. These meetings are particularly important for patients who start dialysis acutely in the hospital. Hughes et al. found that 46% of the patients starting dialysis in Canada began as an inpatient, despite 62% having received at least 4 months of predialysis care.
Patients starting dialysis urgently in the hospital are normally treated with hemodialysis using a central venous catheter. The advent of urgent start PD programs, usually defined as PD catheter use within 2 weeks of insertion, allows these patients to consider PD as their initial dialysis modality rather than receive hemodialysis (HD) as a default. Urgent-start PD studies indicate that this strategy can be successful, with acceptable rates of complications, although catheter dysfunction was increased in some studies because patients were acutely ill and could not undergo the same level of preparation before PD catheter insertion as those with elective PD starts. The impact on PD use is less clear, but urgent-start PD should increase PD incidence use by offering acute patients the option to start on PD rather than hemodialysis. Studies are generally small, with sample sizes of fewer than 100 patients, and some patients used their catheter for PD within 48 hours of insertion.
Identification of PD candidates among patients starting dialysis acutely may also be facilitated by use of nurse navigators. Rioux et al., in a study from Toronto, Canada, found that 21.5% of patients starting dialysis urgently without prior modality education chose PD when targeted by an advanced nephrology practice nurse. Similarly, use of a nurse navigator for urgent starts was found to increase the likelihood of transitioning from in-center hemodialysis to independent forms of dialysis (OR 3.75; 95% CI, 1.08 to 13.05) in a single-center study from Vancouver, Canada.
Assessment of Peritoneal Dialysis Eligibility
All end-stage renal disease (ESRD) patients who are likely to initiate or have initiated chronic dialysis therapy should be assessed for PD eligibility. The average rate of eligibility among multiple dialysis programs in Ontario was 65%, although experienced programs have considered up to 80% of their patients eligible for PD. Eligibility is generally determined by either a nephrologist or interdisciplinary team based on the presence of contraindications to PD, barriers to self-care PD, and the availability of support.
Contraindications to PD are not completely objective, and providers in programs may have different levels of comfort treating patients with different medical conditions ( Fig. 30.1 ). Some programs may also use extended PD catheters or presternal PD catheters, which allow patients with a large pannus or a colostomy/ileostomy to receive peritoneal dialysis. Some programs may also have access to surgeons who can perform PD catheter insertion using advanced techniques such as adhesiolysis and hernia repair, which expands eligibility for PD. With these considerations in mind, our prospective study found that 22% of patients were considered by their nephrology team to have a contraindication to PD.
Studies also indicate physical, cognitive, and social barriers to self-care PD are common. Our prospective study found that 53% of patients assessed for PD (excluding those with contraindications) had at least one physical barrier to PD, such as decreased strength, dexterity, vision or hearing. At least one cognitive barrier to PD was present in up to 15% of patients and included language barriers, dementia, a history of nonadherence, and active psychiatric conditions. When PD patients are screened for cognitive dysfunction using a battery of validated tests, up to 45% of PD patients have evidence of cognitive decline. Social barriers may be present when patients are primary caregivers for others (children or elderly parents), have employment issues, or live in residences such as small apartments or retirement homes that make PD more difficult. Although these physical, cognitive, and social barriers are common in PD patients, it is not known if they explain variance in PD use among programs within or between countries.
In contrast, the availability of support likely influences PD use at the program or regional level. We previously reported that the availability of home care support for PD increased eligibility from 65% to 80% in an elderly dialysis population in a single-center study. In France, assisted PD was chosen by 1232 out of 1613 (76%) of PD patients older than 75 years. We also reported that the availability of family support increased PD eligibility from 18% to 46% depending on the presence of barriers to self-care PD. The availability of home care assistance varies from country to country, but this model of PD delivery has been reported in studies from Canada, United Kingdom, France, Denmark, Switzerland, Spain, Brazil, Taiwan, and China.
Offer and Choice
Once a patient is deemed potentially eligible for PD, he or she is provided modality education. Modality education takes many forms and can be provided by physicians, nurses, or specialized educators. The format of educational programs can include one-to-one sessions, group education, printed materials, videos, or websites. Patients may also now access material directly from the Internet or benefit from peer-to-peer education. Given these many options, it is not surprising that education varies widely among dialysis programs. Regardless of the format of modality education, it is necessary to increase PD use. For example, Mehrotra et al. reported that only 34% of patients in the ESRD Network 18 reported being offered any form of PD when they started renal replacement therapy. This may partly explain the low penetration of PD in the United States because receipt of education is consistently associated with choosing PD. A recent review of educational studies we conducted found that modality education increased PD choice by a factor of 4.6 in a randomized study and a factor of 2.2 in prospective observational studies. Decision aids for patients, such as the Yorkshire Dialysis Decision Aid, “My Kidneys, My Choice Decision Aid”, and “My Life, My Dialysis Choice” may also allow patients to make better choices. A prospective study by Winterbottom et al. found their Yorkshire dialysis decision aid was associated with higher information scores, higher decision clarity and feelings of control, and higher levels of shared decision making among patients’ families. These are all positive results, but they did not evaluate the impact on actual home dialysis use.
Patient choice, or whether patients are given a choice at all, is a relatively easy factor to modify and likely explains much of variance in PD among programs. In prospective studies, approximately 50% of patients chose PD when they were offered it in a systematic fashion. However, the choice rate will vary widely from program to program depending on their philosophy. Some programs employ a neutral approach where patients are provided fact-based education with little promotion, similar to consent for a clinical trial. For example, France has a legal requirement to offer patients modality choice. However, most programs likely encourage patients to consider home-based therapies and actively promote them by strongly emphasizing the benefits of home dialysis. Other countries, like Hong Kong, Thailand, and Mexico, employ PD-first policies where choice is explicitly restricted. Other countries such as New Zealand may not explicitly restrict choice but make it challenging for some patients to receive in-center hemodialysis, resulting in high home dialysis rates.
PD-first policies may conflict with the patient-centered care and the principle of shared decision making. Shared decision making is defined as “an approach where clinicians and patients share the best available evidence when faced with the task of making decisions, and where patients are supported to consider options, to achieve informed preferences”. Other research suggests shared decision models, when employed thoughtfully, can increase home dialysis without restricting patient choice.
It is important to note that expressed choice before the initiation of dialysis does not always translate into actual receipt of PD therapy. For example, Leibmen et al. found of 124 patients who chose PD at the time of education, 59 started dialysis therapy with PD, and 65 started with HD. On day 91 of dialysis therapy, a total of 60 patients were on PD therapy and 55 were on HD therapy. Proponents of buried PD catheters have argued this technique can “lock in” patient choice and is one of the advantages of placing buried PD catheters at higher levels of GFR. Buried catheters are inserted so the external limb of the catheter is buried under the skin, permitting healing and tissue ingrowth of the cuffs in a sterile environment. Buried catheters cannot be flushed and must be externalized before use.
The availability of support can also increase PD choice. Family or home care support was associated with increased likelihood of choosing PD, and the impact of support was more pronounced in patients with barriers to self-care.
Insertion of the Peritoneal Dialysis Catheter
Patients who chose PD must undergo PD catheter insertion to receive PD. PD catheter insertion among eligible patients is an objective measure of choice and can be used to compare choice rates among programs in epidemiological studies. Methods of insertion include percutaneous techniques, generally performed by nephrologists and radiologists, and surgical techniques, including laparoscopic and open insertion, performed by surgeons. Randomized trials have not consistently found significant differences in complication rates between the different methods, and the most recent guidelines of the International Society for Peritoneal Dialysis (ISPD) recommend using local expertise and availability to guide the method of insertion. Some advocates of laparoscopic placement with advanced techniques report outstanding results with very low complications rates, but it is not known whether these results are generalizable to all programs.
A study we conducted in Ontario, Canada found patients who had PD catheters inserted by nephrologists were more likely to receive at least 4 weeks of PD use. However, it was not known if this result was due to be better timing of insertion, patient selection, or reduce complication rates. Higher-risk patients with previous abdominal surgeries and scarring may be selected for surgical placement of PD catheters. In this study, 17% of PD catheters did not provide at least 4 weeks of PD therapy during follow-up, but the reasons could not be determined from the administrative data. Asif et al. reported an increase in the prevalent use of PD at three centers from 16% to 18% to 22% to 32% when nephrologist-based insertion of PD catheters was introduced.
In summary, methods of insertion vary widely and it is not clear the impact the method of insertion has on PD use. Laparoscopic insertion with advanced techniques could provide superior technical results, but its impact on use is unknown. In contrast, nephrologist-based insertion may also improve use, but the mechanism is not known. Regardless of technique, high rates of insertion-related complications, particularly ones causing early technique failure, will negatively affect PD use.
Receipt of Peritoneal Dialysis
Once the catheter is in place, PD can be initiated. ISPD guidelines recommend at least 2 weeks of healing before using PD catheter. However, urgent-start PD programs have used PD catheters 24 to 48 hours after insertion. In contrast, buried catheter studies are often exteriorized months or years after initial placement ( Table 30.2 ).
|Steps in Process of Care||Factors Potentially Increasing PD Use|
|Identify PD candidates||Predialysis care|
|Urgent-start PD programs|
|Nurse (modality) navigators|
|Interdisciplinary meetings to review new patients|
|Assess for PD eligibility||Predialysis care|
|Nurse (modality) navigators|
|Interdisciplinary meetings to review new patients|
|Determine patient is eligible||Availability of family support|
|Availability of home care support|
|Offer PD||Modality education|
|Peer support programs|
|Patient chooses PD||PD-first policies|
|Shared decision-making models|
|Availability of family support|
|Availability of home care support|
|Patient receives PD||Method of PD insertion|
|Operator inserting PD catheter|
|Timing of PD catheter insertion|
Outcomes of Peritoneal Dialysis Therapy
Kidney transplantation is felt to provide the best patient outcomes and quality of life and is the most cost-effective form of renal replacement therapy available in settings that have the necessary health system infrastructure to support transplant programs. However, donor organs are a scarce resource and not all patients are eligible for a transplant. As a consequence, the vast majority will receive dialysis as the first form of treatment for end-stage kidney failure.
In an ideal world, information about patient outcomes from several, well-designed randomized controlled trials comparing different dialysis treatments would be available to guide shared decision making and help patients and providers choose between the different options. Unfortunately, attempts at randomized comparisons have been unsuccessful. Although there are numerous potential comparators (e.g., transplantation, home hemodialysis, or conservative management), we focus our discussion on studies that have compared the outcomes of patients treated with PD and HD.
Attempts at Randomized Comparisons of Peritoneal Dialysis and Hemodialysis
Investigators in the Netherlands attempted to conduct a randomized trial of incident dialysis patients at 38 participating sites. Although 63% of screened patients were deemed eligible for the trial, only 38 of 773 eligible patients consented to participate and it was stopped because of poor accrual. This study was nonetheless informative because it demonstrated the fact that patients appeared to form strong preferences for one modality or another when educated about their treatment options and were unwilling to subject themselves to randomization. This is an important consideration, particularly in environments where PD-first policies are being contemplated.
A subsequent feasibility study was conducted in China to determine whether patients would be willing to be randomly assigned to either PD or HD ( clinicaltrials.gov NCT00510549 ). The investigators had planned to enroll 50 patients to determine whether a larger trial looking at the impact of treatment modality on mortality, morbidity, quality of life, and cost effectiveness was feasible. According to clinicaltrials.gov , recruitment was completed in August 2008 and a protocol for the larger trial was registered in 2011 ( clinicaltrials.gov NCT01413074 ). Recruitment was terminated in 2013 after 416 patients were enrolled, and final data collection for the primary outcome measure was to be completed in December of that year. The results of the pilot and larger trial have not been published to our knowledge.
In the absence of data from randomized controlled trials, we must rely on outcomes of observational studies. Although there is a general feeling among experts that the outcomes of patients treated with PD and HD are equivalent, much of the existing literature is based on registry studies with unclear data quality, which have included patients who might not be eligible for both therapies and patients who are acutely ill, have a poor prognosis, and are treated almost exclusively with HD. These and other issues may introduce bias into nonrandomized comparisons. Given that many jurisdictions are actively promoting PD and some have implemented PD-first policies, it is important to be aware of the methodological challenges inherent in such comparisons and the level of confidence we should have in prior comparative studies.
It is important that data regarding baseline patient characteristics, changes in treatment modality, and patient-level outcomes are valid to ensure a fair and unbiased comparison of PD and HD. If differential reporting of patient characteristics or outcomes occurs, this can lead to systematic bias. Ideally, data collected as part of a well-designed prospective cohort study or data from a validated source would be used for PD versus HD comparisons. However, the majority of studies comparing outcomes in individuals treated with PD and HD have been based on data from regional or national dialysis registries, particularly in recent years.
Registries can be an important source of information and are often used for comparative studies of PD and HD. It is generally assumed that data from dialysis registries are valid, but very little work has been done in this area to confirm the validity of the data sets and the available information raises some important concerns. Longenecker et al. compared comorbidity information collected using Form 2728 (Medical Evidence Report for end-stage renal disease in the US Renal Data System Registry) to data from the prospective Choice for Healthy Outcomes in Caring for ESRD (CHOICE) study. The authors found that comorbidities were “significantly underreported on Form 2728” and found that performance of the registry data were systematically different among those treated with PD compared with HD. In fact, they cautioned, “bias may be introduced if Form 2728 data are used to study associations between comorbid conditions, treatment, and mortality or other outcomes.” Subsequent studies examining the validity of Form 2728 for identifying glomerular diseases have shown poor agreement with kidney biopsy results.
Validation studies of other national registries, such as the Canadian Organ Replacement Register and the Australia and New Zealand Dialysis and Transplant Registry (ANZDATA), have been conducted. Moist et al. found that there was underreporting of comorbidities in CORR, similar to the USRDS. However, when hazard ratios were calculated for individual comorbidities using CORR data and compared with those obtained using information taken from chart review, there was good agreement between the two sources. Although this is somewhat reassuring, the authors didn’t specifically look to see if there were systematic differences between the coding of comorbidities in PD and HD patients or determine the validity of other important variables required to conduct an unbiased comparison of these two modalities (e.g., changes in treatment status or the occurrence of important patient outcomes). Similar findings were reported in a retrospective audit of 175 HD patients registered in the ANZDATA registry. Gray et al. found that primary renal disease was correct in 86% of cases; date of first dialysis was correct in 94% of cases, but there was significant underreporting of most comorbidities (low sensitivity and relatively high specificity). No information was available for PD patients or for important patient outcomes.
Valid data are necessary foundation for robust, unbiased comparisons of PD to HD. However, several other methodological considerations can influence the observed results. It follows that it is important to ensure that studies are well designed, in an attempt to minimize bias and avoid misleading results and conclusions. These issues will be explored further in the context of the relevant outcomes discussed next.
Comparing the Outcomes of Patients Treated With Peritoneal Dialysis and Hemodialysis: Survival
Studies examining the impact of dialysis treatment modality on patient survival have been inconclusive, but most experts believe that PD and HD are associated with similar survival. Some studies have found a benefit with PD, whereas others have reported a survival advantage of HD or little or no difference between modalities.
Eligibility for Peritoneal Dialysis and Hemodialysis
It is important to note that not all individuals with kidney failure are candidates for PD or other home dialysis therapies. Some patients have medical and social conditions that are considered contraindications to treatment. These include morbid obesity, massive polycystic kidneys, diverticulitis, other disease states that make the peritoneal cavity unsuitable for PD, and living in a nursing or retirement home that does not permit PD. Many patients without an absolute contraindication to therapy still have barriers to home dialysis that must be overcome to perform it safely. Examples of barriers include decreased physical strength and manual dexterity, poor vision, cognitive impairment, and a fear of social isolation. After taking barriers and contraindications into consideration, two-thirds of patients are candidates for both PD and HD and must choose between them with the help of their healthcare providers.
In the setting of a randomized controlled trial, all participants would be screened to ensure that they are eligible for the two strategies being compared. The choice of dialysis treatment modality is only relevant to those who are eligible for both therapies because this is the only group that is faced with the decision in clinical practice. Ideally, observational studies comparing PD with HD would include only those eligible for both therapies and therefore faced with a choice between the two in clinical practice. However, eligibility for PD and HD is not routinely recorded in national dialysis registries and has not generally been accounted for, even in prospective studies. This is relevant because approximately one-third of patients are deemed ineligible for PD, either because of medical or social contraindications or their constellation of barriers to doing home therapies. Including ineligible patients in comparisons of PD and HD means that patients who are generally older and sicker are included in the PD group, and this may bias analyses against PD. Several studies have conducted comparisons of outcomes in patients who are judged to be eligible for both therapies according to the local multidisciplinary team and based on a structured assessment. However, the majority of studies have not addressed this issue. Some authors have attempted to overcome the lack of information about PD eligibility by using propensity scores. In this approach, statistical methods (e.g., matching on propensity score or adjustment for propensity score) are used to make groups more similar with respect to the likelihood of receiving PD. This approach attempts to level the playing field but is not the same as comparing eligible patients.
In addition to restricting analyses to patients who are eligible for both PD and HD, the definition of a “chronic” dialysis patient varies. Many registries follow patients for 90 days from the time of dialysis initiation before calling them a chronic dialysis patient. Although the impact of cohort selection on the results of PD versus HD comparisons has not been studied extensively, a Canadian study by Quinn et al. explored the impact of several different cohort definitions and reported that results differed significantly according to the definition employed. Restricting the analysis to elective, outpatient dialysis starts that had received at least 4 months of predialysis care identified no difference in mortality in those treated with PD compared with HD. When the results were compared with those obtained using traditional definitions of chronic dialysis, they differed in important ways. The interaction between diabetes status and treatment modality was no longer present and the change in the relative hazard of death over time that has been described in prior studies was no longer present, suggesting it is related to the inclusion of urgent dialysis starts, rather than an effect of dialysis modality.
The Impact of Urgent Starts
There are several potential trajectories for patients who develop end-stage kidney failure: slowly progressive kidney disease, acute kidney injury superimposed on chronic kidney disease, and acute kidney injury in an individual with no known history of kidney disease. Patients who have an acute component to their kidney failure may not have been adequately prepared to start dialysis, may have a poorer prognosis, and are preferentially treated with HD in most centers. It is important to note that the increased mortality observed in this population persists for up to 6 months, so analyses based on traditional definitions of chronic dialysis may not overcome this inherent selection bias by restricting the cohort to “chronic, stable” patients.
Incident Versus Prevalent Patients
It is generally well accepted that comparisons of HD and PD should be conducted in incident patients, rather than prevalent patients. Outcomes in the incident patient population are more relevant to clinical decision making because decisions about modality choice are generally made before the start of dialysis or soon thereafter in patients who start dialysis in an unplanned manner (Quinn J Nephrol 2007). Comparisons among prevalent patients are difficult to interpret given the heterogeneous patient population, variable time on therapy, and influence of other interventions and complications that have occurred since the initiation of renal replacement therapy.
Changes in Treatment Modality
The method used to account for switches between treatment modalities must also be considered. The two main approaches have been described as intention to treat (ITT), in which patients are classified according to their initial treatment modality, or as treated (AT), in which modality is treated as a time-varying exposure and outcomes are attributed to the current dialysis modality. In ITT approaches, patients can be censored when they switch to a different modality or modality switches can be ignored. In AT approaches, it is possible to introduce a lag period where outcomes that occur during a prespecified time after a change in modality are attributed to the prior treatment. This is done because patients may suffer a complication that leads to a change in modality (e.g., fungal peritonitis) but that also adversely affects prognosis or the likelihood of other complications. Although it is generally recommended that both approaches should be performed, ITT analyses are likely the most clinically relevant to inform decision making.
It is important to consider the setting in which a study is conducted. Differences in the way that care is structured and the proportion of patients treated with PD and HD may influence the results of observational comparisons. As a result comparisons of HD to PD in one country may not be generalizable to another because of different outcomes by country, center experience with a given modality, cost considerations that are unique to that jurisdiction, or patient factors that influence quality-of-life measures. Finally, the outcomes of patients treated with PD appear to have improved over time in the United States and many other parts of the world, whereas the outcomes of patients treated with HD have not.
Case-mix severity refers to an individual’s risk for experiencing adverse outcomes based on measured clinical characteristics and is important to account for when comparing outcomes in two or more groups. Groups may differ on the likelihood of experiencing a particular event because important risk factors are unevenly distributed between them. As a result, they confound the observed relationship between group assignment and the risk of the outcome. Failing to account for differences in case mix between groups may lead to biased or inaccurate results. The drivers of case-mix severity vary depending on the population being studied, as well as the exposure and outcome being examined. In comparison of PD and HD, demographic information, comorbidity data, laboratory variables, measures of functional status, genetic makeup, and treatment-related factors, such as those related to dialysis therapy, are all potentially relevant.
In addition to accounting for case-mix severity, it is important to screen for important effect modifiers and account for them when present. In mortality comparisons, age, race, ischemic heart disease, congestive heart failure, diabetes, and the presence or absence of comorbidities have all been found to be effect modifiers. This means, for example, that the impact of treatment modality on survival may be different for patients with diabetes mellitus than it is for those without diabetes. In situations where effect modification is present, results should be presented for all subgroups separately.
The important considerations when modeling mortality data in this patient population are reviewed by Vonesh et al. They concluded that Cox and Poisson survival models appeared to produce consistent results and the choice of model should depend on the nature of the available data and the preference of the investigator. Since the publication the report by Vonesh et al., other approaches, including marginal structural models, the use of propensity score techniques, and methods to account for competing risks (e.g., transplantation), have been employed. Although an in-depth review of the relative advantages and disadvantages of each approach is beyond the scope of this chapter, choice of statistical approach does not appear to have a significant impact on the observed results.
The relative benefit of PD and HD has been found to change over time in a number of studies. Most reported that the benefit of PD was greatest in the first 2 years of therapy and then declined. It has been suggested that PD is associated with a better preservation of residual renal function compared with HD and that this may explain the change in the relative hazard of death over time. Residual renal function tends to decrease with time on therapy and is a potent predictor of survival on both PD and HD. The loss of residual renal function is slower in PD patients, and some feel that this is the reason that there is an initial survival advantage that gradually disappears. Alternatively, it may be explained by a tendency to start PD patients while they have a higher residual renal function than their HD counterparts.
Another potential explanation for the initial survival advantage noted with PD is that patients who start dialysis urgently on HD require a temporary form of dialysis access. This typically means that the patients have a central venous catheter placed to access the bloodstream for dialysis. Catheters are associated with a risk for bacteremia and sepsis that is present soon after the catheter is placed, particularly in the acutely ill patient. In this situation, it is possible that treatment with HD is associated with higher short-term mortality as a result of the prevalence of central venous catheters in this population, rather than simply being a surrogate for a higher risk patient. Perl et al. found that the initial survival advantage of PD disappeared when PD patients were compared with HD patients who initiated dialysis with a functioning arteriovenous fistula or graft for vascular access. HD patients who started dialysis with a central venous catheter for vascular access had an 80% higher mortality in the first year of therapy (adjusted hazard ratio [HR] 1.8; 95% CI, 1.6 to 1.9).
However, it is likely that the observed initial survival advantage with PD is related to selection bias rather than an effect of treatment modality or choice of vascular access. The mortality rate on HD has been found to initially be quite high, fall to a nadir at 12 to 18 months, and then rise in a relatively linear fashion thereafter. In contrast, the mortality rate on PD has been found to rise in a linear fashion from baseline and meet the HD curve between 30 and 36 months. After that time, the mortality rates were fairly consistent. If HD is the default option in a higher proportion of urgent starts, it might explain why patients treated with HD have elevated mortality rates shortly after the start of therapy. In studies that include urgent starts, the vast majority would be included in the HD group rather than the PD group, and given that they have a worse prognosis, this would tend to artificially inflate the mortality rate on HD at the start of therapy. A study by Quinn et al. supports this hypothesis. When traditional definitions of chronic dialysis were used, the relative risk of death on PD compared with HD increased over time, consistent with prior studies. However, the association disappeared when the authors excluded patients who started dialysis therapy urgently, suggesting that the apparent increase in the relative risk of death over time on PD compared with HD is due to bias rather than a specific effect of dialysis modality or vascular access.
Some have suggested that it may be worth testing a strategy of switching patients from PD to HD after a specified period to optimize survival in ESRD patients, based on the fact that the relative hazard of death on PD increases with time. In the absence of a controlled trial suggesting that elective switch at a particular point in a PD patient’s treatment history is beneficial, it is our opinion that this approach should not be recommended based on the existing evidence.