COMPLICATIONS OF DIALYSIS - PART 10 pot

790 Mendelssohn 11. Mendelssohn DC, Skorecki KL, Cardella CJ, Thimm A. Dialysis in Toronto: a crisis in waiting. The Globe and Mail 1993; Oct. 8:A27. 12. Mendelssohn D, Roscoe J, McCready W. Public de- serves straight talk on growing dialysis crisis. The To- ronto Star 1994; March 2:A21. 13. Report of the Working Group on Renal Services, To- ronto, Ontario, January 1995. 14. Final Report of the Steering Committee of the Central East Region Dialysis Planning Study, Ontario, January 1995. 15. Ontario Legislative Assembly, Standing Committee on Social Development, Dialysis Treatment in Ontario, 1994. 16. Report of the Ontario Renal Services Advisory Com- mittee, January 1996. 17. Annual Report 1997, Vol. 1: Dialysis and Renal Trans- plantation, Canadian Organ Replacement Register, Ca- nadian Institute for Health Information, Ottawa, On- tario, March 1997. 18. U.S. Renal Data System, USRDS 1997 Annual Data Report, The National Institutes of Health, National In- stitute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, April 1997. 19. Canadian Organ Replacement Register, 1992 Ontario Report, Canadian Institute for Health Information, Don Mills, Ontario, March 1994. 20. Kjellstrand CM, Moody H. Hemodialysis in Canada: a first class medical crisis. CMAJ 1994; 150:1067–1071. 21. Mackenzie JK. Decisions for dialysis: a survey of Ca- nadian dialysis directors. Clin Inv Med 1993; 16 (suppl):B107. 22. Mendelssohn DC, Kua BT, Singer PA. Referral for dialysis in Ontario. Arch Int Med 1995; 155:2473–2478. 23. Sinclair Decision support. Quantitative Study of De- mand for Renal Dialysis in Ontario, June 1995. 24. Eadington DW. Delayed referral for dialysis: higher morbidity and higher costs. Semin Dial 1995; 8:258– 260. 25. Ratcliffe PJ, Phillips RE, Oliver DO. Late referral for maintenance dialysis. Br Med J 1984; 288:441–443. 26. Innes A, Rowe PA, Burden RP, Morgan AG. Early deaths on renal replacement therapy: the need for early nephrological referral. Nephrol Dial Transplant 1992; 7:467–471. 27. Campbell JD, Ewigman B, Hosokawa M, Van Stone JC. The timing of referral of patients with end-stage renal disease. Dial Transplant 1989; 18(12):660–668. 28. Jungers P, Zingraff J, Page B, Albouze G, Chauveau P, Page B, Hannedouche T, Man N-K. Late referral to maintenance dialysis: detrimental consequences. Ne- phrol Dial Transplant 1993; 8:1089–1093. 29. Ifudu O, Dawood M, Homel P, Friedman EA. Excess morbidity in patients starting uremia without prior care by a nephrologist. Am J Kidney Dis 1996; 28:841–845. 30. Eadington DW, Craig KJ, Winney RJ. Comorbidity and biochemical indices modulate the impact of late referral on survival on RRT (abstr). Nephrol Dial Transplant 1994; 9:960. 31. Khan IH, Catto N, Edward N, MacLeod AM. Chronic renal failure: factors influencing nephrology referral. Q J Med 1994; 87:559–564. 32. Churchill DN, Taylor DW, Keshaviah PR, for the CANADA-USA Peritoneal Dialysis Study Group. Ad- equacy of dialysis and nutrition in continuous peritoneal dialysis: association with clinical outcomes. J Am Soc Nephrol 1996; 7:198–207. 33. Churchill DN, Thorpe KE, Vonesh EF, et al. Lower probability of patient survival with continuouos peritoneal dialysis in the United States compared with Can- ada. J Am Soc Nephrol 1997; 8:965–971. 34. Toronto Region Dialysis Registry. Indicators of Ade- quacy of Treatment, 1996, Toronto, 1997. 35. Levin A, Lewis M, Mortiboy P, Faber S, Hare I, Porter E, Mendelssohn DC. Multidisciplinary predialysis programs: quantification and limitations of their impact on patient outcomes in two Canadian settings. Am J Kid- ney Dis 1997; 29:533–540. 36. Goeree R, Manalich J, Grootendorst P, et al. Cost analysis of dialysis treatments for end stage renal disease. Clin Inv Med 1995; 18:455–464. 37. Prichard SS. The costs of dialysis in Canada. Nephrol Dial Transplant 1997; 12(Suppl 1):22–24. 38. Mendelssohn DC, for the CSN Professional and Public Policy Committee. Principles of end stage renal disease care. Ann RCPSC 1997; 30:271–273. 39. Mendelssohn DC, Jung B, Blake P, Mehta R. Attitudes of North American nephrologists towards dialysis modality selection: 1) Canadian results. JASN 1997; 8: 222A. 40. NIH Consensus Statement—Morbidity and Mortality of Dialysis. Ann Int Med 1994; 121:62–70. 41. Prichard SS. Treatment modality selection in 150 con- secutive patients starting ESRD therapy. Perit Dial Int 1996; 16:69–72. 42. Nissenson AR, Pritchard SS, Cheng IKP, et al. Non- medical factors that impact on ESRD modality selection. Kidney Int 1993; 43(S40):S120–S127. 43. Inglehart JK. The American health care system: the end stage renal disease program. N Engl J Med 1993; 328: 366–371. 44. Klahr S. Anemia, dialysis, and dollars. NEJM 1996; 334:461–462. 791 45 Outcomes and Intermodality Transfers in Patients on Renal Replacement Therapy in Canada and Europe Peter G. Blake* University of Western Ontario, Ontario, Canada Wim Van Biesen** University Hospital of Gent, Gent, Belgium Norbert Lameire** University Hospital of Gent and University of Gent, Gent, Belgium Rosario Maiorca** University and Civil Hospital, Brescia, Italy PART I: IN CANADA I. INTRODUCTION The Canadian system for delivering renal replacement therapy (RRT) to patients with end-stage renal disease (ESRD) is potentially of more than local interest for a number of reasons. First, the quality and completeness of data collection achieved by the Canadian Organ Re- placement Register (CORR) is very good by interna- tional standards and so the resulting database allows questions concerning the demographics and clinical outcomes of ESRD patients to be addressed and answered in a way that has not often been possible in other countries (1). Second, the geographic and cultural proximity to the United States makes Canada a natural control for trends identified in the much larger U.S. ESRD population and so helps the observer to distin- guish between trends that are specific to one country only and those that may have broader relevance. Third, the Canadian health-care system is an interesting mix of public and private elements, which might be ex- *Part I in Canada. **Part II in Europe. pected to influence acceptance rates for RRT in con- trasting ways. Thus, there are fee-for-service payments to nephrologists for the care of ESRD patients, and these might be expected to act as an incentive to initiate more patients on RRT. However, almost all RRT is provided through publicly funded hospitals, and, given public spending constraints, this might be expected to limit growth in acceptance rates. II. DELIVERY OF RRT IN CANADA A brief description of RRT services in Canada is appropriate before looking at comparative outcomes between different RRT modalities. In Canada, health care is administered on a provincial basis within the constraints of the guiding principles of the Canada Health Act, which mandates universal coverage and access as well as predominantly public funding and administra- tion. Almost all RRT is delivered via a relatively small number of large, publicly funded hospitals, the majority of which are affiliated with universities. These RRT programs are funded out of hospital global budgets although in some provinces supplemental direct dedi- cated funding is made available for chronic dialysis because of the high cost and life-sustaining nature of 792 Blake Fig. 1 Changes with time in percentage of Canadian prevalent chronic dialysis patients being maintained on PD. (Adapted from Ref. 1.) the treatment and because of the apparent relentless growth in demand for it. In general, it costs Canadian hospitals significantly less to maintain ESRD patients on peritoneal dialysis (PD) as compared to hemodialysis (HD), and because HD expansion also tends to require more initial capital input, both government and hospitals have been in- clined to support substantial use of PD (2,3). Thus, in Canada over the past decade, about 40% of new patients have been on PD by day 90 after initiation of RRT, and overall about 35% of prevalent dialysis patients are maintained on this modality (1). Most HD is provided via hospital-based ‘‘in-center’’ units, but a significant minority of more stable patients are treated in self-care units with lower staff-to-patient ratios. There has also been a recent increase in the numbers treated in small satellite units, but home HD remains little used. There has been controversy over the high proportion of patients doing PD in the larger Canadian cities where government-funded HD expansion has had particular difficulty keeping up with population growth (3). Accordingly, there has been a perception that, in some centers, patients have been directed to PD without being given any real choice in terms of modality selection. Recently, in response to these criticisms, HD capacity has been significantly expanded and there has been greater decentralization of HD into small com- munities. This has contributed to a fall in the proportion of dialysis patients doing PD to about 30% in 1997 (1,4) (Fig. 1). As already mentioned, the majority of dialysis is delivered via teaching hospitals, and therefore most Ca- nadian nephrologists have an academic appointment and have teaching and research commitments in addi- tion to clinical responsibilities. Nephrological care of RRT patients is remunerated on a fee-for-service basis by single-payer, provincial health plans and, in all provinces, these have tended to pay substantially more for the medical care of in-center HD patients compared to that of PD and self-care HD patients (2,5). Thus, there is an incentive to physicians to maintain more patients on in-center HD, while there is an incentive to hospitals and governments to maintain more on PD. The relatively high proportion of patients on PD would suggest that the former incentive has, in practice, generally been more powerful than the latter. Recently, the largest province, Ontario, has switched to a modality- independent capitation fee for physician care of all chronic dialysis patients, but it is too early to judge if this will affect modality distribution. Acceptance rates for RRT are relatively low in Can- ada at about 120 per million population per annum, compared to over 260 per million in the United States. This discrepancy has been hotly debated and is partly accounted for by the absence of a large black population with a high incidence of ESRD in Canada (6,7). However, even the Caucasian incidence rate in Canada is 50% lower than that in the United States and is more like that seen in western Europe. It has been argued that this reflects the almost inevitable rationing of treatment that occurs in a single-payer, publicly funded and publicly delivered health-care system, but there are probably other factors, including geographical distance constraints for many potential patients outside larger cities. An alternative, more palatable explanation is that there may be a lower incidence of ESRD related to better primary health care, but this is unproven. The ongoing growth in incidence rates suggests that there is still substantial ongoing unmet need, however (7). Renal transplantation rates in Canada run at about 30 per million per year and, unlike ESRD incidence rates, have not changed appreciably in the past decade Outcomes in RRT in Canada 793 Table 1 Trends in Acceptance Rates, Mean Age, and Prevalence of DM in Incident Canadian RRT Patients, 1988–1996 Year Acceptance rate per million population Average age of incident patients % of incident patients with DM 1988 72.7 54 29.6 1989 77.6 55 29.2 1990 82.7 55 29.1 1991 93.9 56 31.4 1992 96.3 56 31.4 1993 100.3 57 33.0 1994 105.7 58 34.7 1995 108.8 58 35.0 1996 110.9 59 37.9 Source: Adapted from Ref. 1. Table 2 Adjusted Mortality Rate Ratios in Successive Cohorts of Canadian HD and PD Patients, 1981–1996 a Cohort HD PD 1981–1984 1.00 1.00 1985–1988 0.94 0.88 1989–1992 0.90 0.71 1993–1996 0.77 0.64 a Using risk ratio of 1.0 for mortality on each modality in the period 1981–1984. Source: Adapted from Ref. 9. so that waiting lists have become increasingly long (1). Again, this transplant rate is similar to that in northern European countries and is about a third less than that in the United States. While cadaveric transplant rates have stayed the same or even fallen, the use of living donors, both related and unrelated, has increased significantly in recent years. There is no doubt that in Canada, as elsewhere, transplantation is the least costly treatment for ESRD, but its use has been limited by lack of donors and also by the lack of suitability of the majority of contemporary dialysis patients for surgery and immunosuppression because of age and comorbidity (8). Against this background there are a number of key questions that need to be addressed. One is how outcomes for patients in general are changing with time and how they compare to those in the United States and elsewhere. Another is how the high proportion of patients on PD do relative to those on HD and how this compares with the situation in other countries. Im- plicit in this question is the issue of whether a policy of high PD usage as is practiced in many countries with predominantly publicly delivered health-care systems is consistent with good patient outcomes as compared to countries where HD is more dominant. A third issue is whether transplantation leads to better survival than staying on dialysis. III. OUTCOMES OF DIALYSIS A. Mortality Uncorrected one-year mortality rates for new patients initiating RRT in Canada measured from the time of first treatment fell steadily over the course of a decade, from 19.5% in the cohort of 1986 to 15.5% in the cohort of 1996 (1,9). This improvement occurred despite a rise in acceptance rates for RRT from 70 to 110 per million over this period and a consequent steady increase in the mean age of incident patients over the same period from 52 to 58 years of age (Table 1) (1). Similarly, during that time the proportion of patients with diabetes mellitus rose from 30 to 38% and the proportion whose ESRD is actually attributed to diabetes rose from 23 to 28% (Table 1) (1). The same trend towards improved survival is even more apparent when mortality rates are adjusted for age, gender, race, and province using a Poisson regression model. For HD and PD there were decreases of 23 and 36%, respectively, in the adjusted relative risk of mortality between the cohorts that started dialysis in the period 1981–1984 and the cohort that started in the period 1993–1996 (Table 2). In PD, in particular, the decline in adjusted mortality was marked in patients both above and below age 65 (Table 3). These one-year mortality rates are intermediate between those reported in the United States and those in Europe and Japan (10). However, if the first 90 days on RRT are omitted, as is the practice with U.S. registry data, the mortality rate falls to 10.7%, which is quite similar to values in Europe (10). Of course, when these comparisons are being made, it is only fair to point out that the high U.S. mortality rates may partly reflect the much higher acceptance rates for RRT in that country compared to Canada and Europe and the much higher transplant rates compared to Japan. Both of these differences likely lead to a significantly older and frailer ESRD population in the United States. This is likely a major factor underlying the recently reported excess mortality in U.S. PD patients compared to their Canadian counterparts (11). Mortality was over 90% 794 Blake Table 3 Adjusted Mortality Rates per 100 Patient-Years in Successive Cohorts of Canadian PD Patients, Classified by Age, 1981–1996 Cohort Age 45–64 Age 65ϩ 1981–1984 20.5 39.4 1985–1988 17.8 34.9 1989–1992 15.4 28.5 1993–1996 13.1 27.0 Source: Adapted from Ref. 9. higher in the United States and this difference persisted even after correction for demographics, comorbidity, and indices of adequacy and nutrition, may relate to a number of factors, but there has to be a suspicion that the degree of unmeasured comorbidity may be much greater in a country with such high acceptance and transplant rates as are found in the United States. B. HD Versus PD The relatively high proportion of ESRD patients doing PD in Canada as well as the completeness of data collection by CORR makes the country particularly suitable for studying comparative outcomes between HD and PD. As already mentioned, this important issue has been to some extent politicized in Canada because of widely expressed concerns that PD, in at least some regions of the country, has been excessively used in unsuitable patients because of a shortage of government-funded HD spots due to fiscal constraints on health-care spending. When PD and HD were first compared using CORR data in 1987, there was no significant difference in patient survival using an intention-to-treat analysis with censoring at the time of a permanent modality switch (12). These data were consistent with those in a variety of studies from the United States and Europe in the late 1980s and early 1990s which showed generally equiv- alent patient survival on the two modalities (13–16). Technique survival was, of course, notably less with PD, but as this did not appear to be affecting patient survival and as PD was less expensive, the policy of high PD usage appeared appropriate (12). In 1995, however, the U.S. registry, using a prevalent analysis, found a significantly higher mortality rate in U.S. PD patients in general, compared to their HD counterparts (17). This excess risk in PD was estimated at 19% overall and was particularly marked in older patients, diabetics, and females. This study appeared during a period when concern was being expressed in Canada about undue fiscal pressures to direct patients to PD and also at the same time as the publication of the Canada-USA (CANUSA) study, which gave rise to questions about the adequacy of delivered clearances on PD (6,18). Against this background, there was ques- tioning of the high use of PD in Canada, and the publication of CORR data in 1997 showing superior outcomes on PD, as compared to HD, caused some surprise and perhaps some confusion (19). Before considering the CORR data in more detail and why it apparently differs so much from the U.S. data, consideration has to be given to some of the pitfalls of studies comparing outcomes on the two modalities (20,21). First, registry-based comparisons are not randomized controlled trials, and so, for comparisons to be legitimate, corrections have to be made for demographic, comorbid, and other differences between patients on the two modalities. Factors such as age, sex, race, presence or absence of diabetes, and center size are relatively easy to deal with, but comorbidity and associated functional status are much more problematic and sufficiently detailed information to measure them reliably will not typically be found in a registry. Thus, the U.S. registry data on which the 1995 study by Bloembergen et al. was based had no correction for comorbidity, and even the 1997 CORR data had only a count of the number of comorbid conditions with no information as to their severity or as to how they impact on functional status, factors known to have an important impact on survival (17,19,22). A second issue that is clearly critical is the type of analysis being done. Typically this is based either on intention to treat or on the treatment actually received. A pure intention-to-treat analysis assigns patients to the modality they are using at the initiation of dialysis or, more often, at 90 days after initiation, as it may take some time to make a definitive modality selection. All subsequent events are assigned to that initial modality, even if the patient subsequently switches or is transplanted. As switches from PD to HD are much more common than those in the opposite direction, both cohorts will eventually comprise predominantly HD and transplant patients, and so the analysis will increasingly tend to obscure any survival benefit for either modality. This can be dealt with by modifying the analysis to censor patients who are transplanted or who switch modalities. This modification of intention to treat is criti- cized, however, because it excludes deaths that occur in patients shortly after a modality switch, which perhaps should be attributed to the original modality. Thus, as PD-to-HD switches are more common, this Outcomes in RRT in Canada 795 Table 4 Mortality Rate Ratios for PD Relative to HD (where HD Is 1.00) in Canadian Patients Initiating RRT, 1990–1995, Using Treatment-Received and Intention-to- Treat Methodologies Treatment-received method, relative risk (95% CI) Intention-to-treat method, hazard risk (95% CI) All 0.73 (0.69, 0.77) a 0.93 (0.87, 0.99) a Non-DM <65 yrs 0.53 (0.46, 0.60) a 0.84 (0.87, 0.99) a Non-DM >65 yrs 0.75 (0.65, 0.86) a 0.95 (0.86, 1.05) DM <65 yrs 0.76 (0.65, 0.83) a 0.90 (0.82, 1.10) DM >65 yrs 0.88 (0.75, 1.04) 1.04 (0.87, 1.24) a 95% Confidence Intervals do not cross 1.0 and so observation is significant at p < 0.05. Source: Adapted from Ref. 26. censoring may underestimate adverse outcomes on PD. This can be dealt with by having a ‘‘grace period’’ of 30, 60, 90, or 120 days after a switch during which deaths are not censored but rather attributed to the initial modality. While this type of analysis is widely used, it also presents problems. If the grace period is too short, there will be a bias in favor of the initial modality if the second modality is truly superior. How- ever, if the grace period is too long there may be a bias in favor of the second modality if the initial modality is truly superior. A typical compromise is to use 60 days, but it may be better still to test a variety of time periods and see how much the final result is altered. When a treatment-received analysis is done, switches are an even bigger issue, as deaths at some stage subsequent to the switch will have to be attributed to the new modality rather than just censored. Thus, the potential for obscuring a true treatment difference is even greater. With either type of analysis, the time at which patients enter the analysis is also important. Patients who present late for RRT tend to be sicker and also tend to be treated with HD. These patients may have a high early mortality rate and so the inclusion of patient data from the early weeks of dialysis may bias the analysis against HD (19). Typically, data from the first 90 days are excluded, in part to get around this problem but also because a significant number of patients who plan to do PD spend some initial time on HD. If data from a longer initial period is omitted, as was the policy with the U.S. registry analysis published in 1995, there is the potential to bias the analysis against PD because it is during this early period, when residual renal function is maximal, that PD may be most beneficial (17,19). For the same reason, a prevalent patient-based analysis will tend to make HD look relatively better, and an incident-based one will have the opposite effect. A third issue that should be kept in mind is that many of the statistical methods used to do these analyses, such as the Cox proportional hazards model, pre- sume that possible risk factors for mortality such as diabetic status, comorbidity, and modality confer proportional risks that remain constant with time (23). This presumption is not always justified, however, and clearly in the case of modality is not appropriate because PD, relative to HD, performs better in the first 3 years, compared to years 4 and 5 (19,24,25). The CORR study published in 1997 made a comprehensive attempt to deal with these pitfalls by correcting for the number of comorbid conditions, by re- doing the HD-versus-PD comparison using both treatment-received and intention-to-treat methodologies, and by varying both the time of initiation of analysis and the duration of postswitch grace periods (19). It thus could reasonably be said to be the most meth- odologically sophisticated analysis to address this issue in any country to date. The study comprised almost all of the over 10,000 incident patients in Canada in the period 1990–1994. In the treatment-received analysis, corrected mortality rates were found to be significantly lower on PD than on HD with a relative risk of 0.73 for all patients on PD compared to 1.00 for those on HD. Subgroup analysis showed that the advantage for PD was greatest in nondiabetics and in younger diabetics but less in older diabetics, where the trend in favor of PD did not achieve statistical significance (Ta- ble 4). In the intention-to-treat analysis, the advantage for PD did not reach significance in the original publication, but when 1995 data were added in a more recent publication, increasing the number of patients to over 14,000, statistical significance was achieved in the group as a whole and was greatest in nondiabetics less than 65 years of age (26) (Table 4). On closer analysis, it becomes apparent that the advantage of PD over HD in both analyses is concen- trated in the first 2–3 years after the initiation of treatment (Fig. 2). This raises the issue of whether PD, in the early years of ESRD, is an inherently superior therapy or whether the difference is due to unmeasured or undetected baseline comorbidity differences between patients on the two modalities. Just as in the United States, there is a suggestion in recent studies of a tendency for comorbidity to be greater in HD patients in Canada (27,28). The CORR study attempted to deal with this by correcting for diabetic status and for the number of major comorbid condition as well as for age, sex, and center size but did not take into account the 796 Blake Fig. 2 Mortality rates for PD relative to HD by follow-up interval, adjusted for age, primary renal diagnosis, and comorbid conditions and estimated using Poisson regression. (From Ref. 26.) severity, as distinct from the number, of comorbid conditions (19). It is arguable, however, whether comorbidity alone can explain all the apparent superiority of PD in Canada. Is it possible that PD is inherently superior in the early years of ESRD? One plausible explanation might be the apparently better preservation on PD of residual renal function during this time (29). Recent studies have emphasized the strong association of residual function with patient survival in PD patients (18,30). It is particularly interesting that the results of these analyses are not significantly affected by varying the way modality switches are dealt with. Thus, censoring at transplantation or at 30, 90, or 120 days, instead of at 60 days, postswitch does not alter the results appreciably (19). However, varying the time after initiation of dialysis at which patients enter the analysis has a strong impact. Thus, omitting the first 90 days from the treatment-received analysis causes the advantage for PD to lose statistical significance, and including the first 90 days in the intention-to-treat analysis increases the significance of the advantage for PD (19,26). The powerful influence that these early weeks after initiation have on the outcome of the analysis would tend to suggest that there is an early unmeasured excess of comorbidity on HD, consistent with the tendency of late or fulminant ESRD presentations to be, by their nature, directed to that modality. In support of this in- terpretation is the fact that the 1995 U.S. analysis that showed such poor results with PD used a methodology that excluded patients from the analysis for an average of 12 months postinitiation of dialysis, thus omitting a key proportion of the period when PD performs best relative to HD (18). When this methodology was changed in subsequent cohorts to include incident patients from 90 days after initiation of dialysis, the apparent benefit for HD lost its statistical significance (31). This suggests that some of the apparently strong differences in relative outcomes on the two modalities between Canada and the United States are methodological rather than real (11). The issue of how the early months on RRT are analyzed is obviously important, but it is unclear if it alone explains all the apparently better results with PD in Canada (11,31). Two recent studies clarify these issues further. Col- lins et al. studied over 116,000 U.S. incident patients from 1994 to 1996 using an intention-to-treat analysis with censoring after a period of grace of 60 days post– modality switch (32). They found that in nondiabetic patients there was a significantly higher risk of mortality on HD, while in older female diabetics there was a significantly higher risk on PD. For younger and for male diabetics, there was no significant difference. These more contemporary results are thus quite similar to the recent CORR results, again suggesting that most of the difference between the United States and Canada in relative outcomes on PD was methodological rather than real (17,19,31). Also, Murphy et al. have recently reported a cohort of 822 Canadian patients at 11 centers followed pro- spectively for up to 3 years from their time of initiation of dialysis in 1993 and 1994 (28). These patients, half of whom were on PD at 90 days after initiation, had very detailed data on comorbidity collected. Unad- justed risk of mortality was lower on PD, but there was no difference between survival on the two modalities, either overall or in any subgroup, once comorbidity was adjusted for. Outcomes in RRT in Canada 797 Table 5 Changes in Numbers, TF Rates, and Causes of TF for Canadian PD Patients, 1988–1995 Year Prevalent patients at start of year Incident patients during year Total PD patients during year Number of TF % TF % of TF due to peritonitis %ofTF due to inadequate PD 1988 1665 775 2440 284 11.6 30.6 5.3 1989 1763 835 2598 294 11.3 30.9 7.5 1990 2014 894 2908 322 11.1 32.0 7.8 1991 2188 1081 3269 334 10.2 24.3 10.2 1992 2521 1148 3669 373 10.2 24.9 11.8 1993 2826 1191 4017 441 11.0 26.3 18.6 1994 3091 1232 4323 608 14.1 24.3 24.2 1995 3353 1127 4480 627 14.0 15.6 25.7 1996 3394 UK a UK 658 UK 15.7 25.7 a UK = unknown (data not available). Source: Adapted from Ref. 1. The relative merits of HD and PD in terms of patient survival will not be definitively answered without a randomized controlled trial, but the feasibility of such a trial is questionable. In the absence of a definitive study, some conclusions can be drawn, however. The evidence suggests strongly that patient survival on PD in Canada is at least as good as that on HD and that the difference between relative outcomes in Canada and the United States may have been overstated. Given the lower cost of PD in Canada, it seems reasonable to conclude that the policy of high PD usage followed over the past two decades has been cost-effective, at least as far as survival is concerned. Furthermore, these analyses mainly deal with the period prior to the recent dramatic changes in PD prescription practices, and these might be expected to reduce PD mortality rates significantly (4,9). C. Technique Failure on PD The above findings on patient survival on PD mainly pertain to the first 2–4 years of dialysis, partly because most of the studies are relatively recent but also because only a minority of PD patients continue on the modality after that time. Much of this patient loss re- lates to death and transplantation, which are at least as likely to occur on HD, but a significant proportion is, of course, due to technique failure (TF), which contin- ues to be much more common on PD (1). It could be argued that while TF in PD is undesirable, it need not be seen as a critical problem for the modality if its rate is relatively stable and if, as has been suggested, overall patient survival is as good as, or better than, on HD. However, there is little doubt that many of the causes of TF are potentially life-threatening and expensive as well as unpleasant for the patient. Thus, if the causes of TF could be more successfully managed, PD could be an even more cost-effective treatment for ESRD. This raises the issue of whether TF rates are changing in PD. A number of forces that might be altering TF rates need to be considered. Peritonitis is generally the most common single cause of TF, but peritonitis rates have been decreasing over the past decade, mainly due to better PD technology, and this might be expected to reduce TF (34–36). More recently, there has been heightened awareness of the problem of inadequate clearances on PD, and since about 1994 prescrip- tions have changed markedly with greater use of larger- volume dwells and of automated PD (4,37,38). This too might be expected to lessen TF rates, although the trend may be too recent for much effect to be apparent yet. In contrast, the greater availability of HD facilities in many Canadian centers in recent years has made modality switching for sociomedical reasons, such as increasing age and frailty and ‘‘burnout,’’ more feasi- ble, and this may show up as an increase in TF rates, emphasizing that TF can be a difficult endpoint to eval- uate. Has there been a change in the pattern of TF in recent years, and is there any evidence that the high PD usage in Canada has been associated with excessive or increasing TF? Special cohort studies to look at TF on PD are being done using the CORR database but have not yet been published. However, certain conclusions can be drawn from annual CORR reports (1) (Table 5). First, if TF is defined as any switch from PD to HD and if the 798 Blake denominator is taken as the sum of prevalent patients at the start of the year plus incident patients during the year, then the rate of TF stayed between 10–12% per annum between 1998 and 1993 and then rose a little to 14% in 1994 and 1995, the two most recent years for which information is available (1). This recent rise in TF may not be significant, but it perhaps reflects the expansion in HD capacity that occurred in the period concerned. The fact that TF has been relatively stable at a time when comorbidity and age are increasing and when mortality is falling is probably a relative success anyway. However, it could also be argued that the absence of a decreasing rate of TF during a period of technological advance is disappointing. Again, it should be pointed out that the available data are likely still too early to detect any improvement in TF or survival consequent upon the very recent changes that have been occurring in PD prescription practices (4). If we look more closely at individual causes of TF, some interesting trends are apparent. TF due to peritonitis has fallen from over 30% of all causes, in the period 1988–1990, to about 25% in the period 1991– 1994 and to 15–16% in 1995 and 1996 (Table 5). In contrast, inadequate dialysis as a stated cause of TF has risen from only 5% in 1988 to 10% in 1991 and 25% in 1995 and 1996 (1). This dramatic increase likely represents an increased awareness of the issue of inadequate dialysis rather than a true growth in the prevalence of the problem. The increase in overall TF rate in 1995 and 1996 can be totally attributed to an increase in this category and suggests that physicians are more aware of this complication and more willing and able to transfer patients to HD when they diagnose it. All this suggests not that PD is being less well practiced but that the expectations have risen as a result of all the recent attention given to adequacy of clearances (18,37,38). III. DIALYSIS VERSUS TRANSPLANTATION It is widely accepted that renal transplantation improves quality of life and is less costly than treatment with chronic dialysis (8). The issue of whether renal transplantation actually improves survival, relative to dialysis, has been somewhat controversial, however. A recent U.S. study appeared to show that survival was improved by renal transplantation (39). This study compared patients who had received a transplant with those who had not, despite being on the waiting list. The rationale was that patients on the waiting list who did not receive a transplant were a more appropriate control group for those who were transplanted than were dialysis patients in general. Similar findings have also recently been reported from Germany (40). This issue has now been addressed in the Canadian context by Rabbat et al. (41). The database used was all 5241 patients who initiated RRT between 1990 and 1995 in the province of Ontario. Of these patients, just over 1100 were wait-listed for renal transplant and just over 700 actually received a graft in the period concerned. A time-dependent Cox nonproportional hazards model was applied, using as a reference point the time interval from the date of wait listing to the time of transplantation, with adjustment for age, gender, race, and cause of ESRD. Compared to the mortality rate during this period, the relative risk of dying in the first 30 days after transplantation was 2.91 (95% confidence interval 1.34–6.32), and this was highly significant (41). In patients surviving more than one year after transplantation, however, the relative risk decreased to 0.25 (95% confidence interval 0.14–0.42). This long-term benefit was particularly notable in patients whose primary disease was glomerulonephritis or diabetes. The overall conclusion was that patients who receive a renal transplant have a higher early mortality risk but ultimately have a net survival advantage compared to dialysis patients who are wait-listed but not transplanted (41). As with the studies comparing dialysis modalities or outcomes in different countries, there is a concern in these studies that patients who are wait-listed but do not receive a transplant may represent a more high-risk subgroup than those who do receive a transplant. The rationale here is that all patients on a transplant list are not necessarily treated equally and that physicians may exhibit subtle biases towards giving scarce organs to healthier recipients. Furthermore, the sicker patients on the list may not be called for transplant because of intercurrent medical problems or may be more often on hold and so less likely to receive a graft. This excess risk might not be detected in an analysis, such as this one, that did not have sufficient information to correct in detail for functional status and number and severity of comorbid conditions. It should also be pointed out that the comparison is complicated by the fact that the relative risks of dialysis and transplantation are not pro- portionate, being excessive for transplant in the early postoperative period and for dialysis subsequently. Nevertheless, the relative risks associated with transplantation for greater than one year are strikingly low relative to those on dialysis, both in the Canadian analysis and in those from elsewhere (39–41). While there will always be controversy about this issue, the balance Outcomes in RRT in Canada 799 of evidence would seem to suggest that transplantation improves survival. IV. CONCLUSION The Canadian model for delivery of RRT in conjunc- tion with the CORR allows some critical questions concerning cost-effective management of ESRD to be addressed, with potential implications for other juris- dictions also. While registry analyses can never be as definitive as well-constructed randomized controlled trials, some general conclusions can be drawn. A policy of high PD utilization, as practiced in Canada, appears to be both cost-effective and consistent with excellent and still improving survival rates. In particular, there is no evidence that PD is intrinsically inferior to HD in terms of patient survival, and, as used in Canada, there is at least some suggestion that it is superior in first few years of RRT. Notwithstanding this, PD is still associated with significant TF, and this does not appear to be falling despite recent technological advance. It remains unclear whether major improvements in technique survival in PD can be achieved or whether the main role of the modality will be to act as a transitional therapy for the first 2–4 years of RRT. Either way, the modality would appear to have a significant role to play in terms of delivery of cost-effective RRT. Transplantation remains the most desirable form of RRT, in terms of quality of life, cost-effectiveness, and, it appears, patient survival. A strategy to optimize the role of transplant is thus ideal, but, unfortunately, scar- city of donors and the relatively modest improvements in long-term graft survival rates have limited this option. Finally, the importance of good data collection in monitoring the effectiveness of any system of delivering RRT cannot be overstated. Not only does the data collection need to be comprehensive and accurate, but the questions that it is used to answer must be incisive and relevant and there must be an awareness of the multitude of potential confounders. REFERENCES 1. Annual Report 1998, Vol. 1. Dialysis and Renal Trans- plantation, Canadian Organ Replacement Register. Ot- tawa, ON: Canadian Institute for Health Information, March 1998. 2. Prichard SS. The cost of dialysis. Adv in Perit Dial 1988; 4:66–69. 3. Mendelssohn DC, Chery A. Dialysis utilization in the Toronto region from 1981–1992. Can Med Assoc J 1994; 150:1099–1105. 4. Blake PG, Bloembergen WE, Fenton SSA. Changes in the demographics and prescription of peritoneal dialysis during the past decade. Am J Kidney Dis 1998; 32(suppl 4):344–351. 5. Nissenson AR, Prichard SS, Cheng IKP, et al. Non- medical factors that impact on ESRD modality selection. Kidney Int 1993; 43:S120–S127. 6. Mendelssohn DC, Kreiger F, Winchester J. A comparison of dialysis in the US and Canada. Contemp Dial Neph 1993; 20:27–31. 7. Mendelssohn DC, Kua BT, Singer PA. Referral for dialysis in Ontario. Arch Int Med 1995; 155:2473–2478. 8. Laupacis A, Keown P, Pus N, et al. A study of the quality of life and cost-utility of renal transplantation. Kidney Int 1996; 50:235–242. 9. Moran JE. Changes in the dose of peritoneal dialysis: Have these independently affected outcomes? Am J Kidney Dis 1998; 32(suppl 4):552–557. 10. U.S. Renal Data System, USRDS 1997 Annual Data Report. National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases. Be- thesda, MD. April 1997. IV. The USRDS Dialysis Mor- bidity and Mortality Study: Wave 2. Am J Kidney Dis 1997; 30(suppl 1): S67–S85. 11. Churchill ND, Thorpe KE, Vonesh EF, Keshaviah PR. Lower probability of patient survival with continuous peritoneal dialysis in the United States compared with Canada. J Am Soc Nephrol 1997; 8:965–971. 12. Posen G, Arbus G, Hutchison T, Jeffery J. Survival comparison of adult non-diabetic patients treated with either hemodialysis or CAPD for end-stage renal failure. Perit Dial Bull 1987; 7:78–79. 13. Burton PR, Walls J. Selection adjusted comparison of life expectancy of patients on continuous ambulatory peritoneal dialysis, haemodialysis and renal transplantation. Lancet 1987; 1:1115–1119. 14. Serkes KD, Blagg CR, Nolph KD, Vonesh EF, Shapiro F. Comparison of patient and technique survival in continuous ambulatory peritoneal dialysis (CAPD) and hemodialysis: a multicenter study. Perit Dial Int 1990; 10: 15–19. 15. Maiorca R, Vonesh EF, Cavilli P, et al. A multicenter, selection adjusted comparison of patient and technique survivals on CAPD and hemodialysis. Perit Dial Int 1991; 11:118–127. 16. Held PJ, Port FK, Turenne MN, et al. Continuous ambulatory peritoneal dialysis and hemodialysis: comparison of patient mortality with adjustment for comorbid conditions. Kidney Int 1994; 5:1163–1169. 17. Bloembergen WE, Port FK, Mauger EA, Wolfe RA. A comparison of mortality between patients treated with hemodialysis and peritoneal dialysis. J Am Soc Nephrol 1995; 6:177–183. [...]... in the light of the preceding dialysis modality Finally, the benefits of an ‘‘integrated-care’’ approach to ESRD patients will be addressed (3) II STUDIES COMPARING HD AND PD DIALYSIS IN EUROPE European data on take-on rate of patients on dialysis, utilization of different dialysis modalities, and the impact of socio-economic factors are discussed in Chapter 43 Studies comparing outcome of patients on... hours of dialysis three times a week The most prevalent practice in the developing countries is to give two 3- to 4-hour sessions of hemodialysis (HD) every week The dialysis time often has to be cut short to accommodate more patients The decision on frequency of dialysis is often based on patient symptomatology and financial considerations Dialysis frequency is increased only if the patient develops complications. .. 25 80 1,500–2,000b 80 200 4,000–5,000b 4,800b 7,200b 10, 000b a Total population: 940 million; per capita GNP: $340; budgeted health-care expenditure: $6 per year b Annual expenditure of dialysis because of financial reasons, leading ultimately to discontinuation of dialysis or death from complications of underdialysis Data on the adequacy of such dialysis schedules are not available Some nephrologists... with a population of 11 million, has only two dialysis centers We conducted a survey of dialysis units in India that revealed that there are only about 0.2 dialysis centers pmp (8) About 68% of these centers are in the private sector and are accessible only to upper-income patients Most of the dialysis centers have only two to four dialysis stations, grossly insufficient for the number of patients needing... countries Long-term maintenance dialysis remains out of reach for most people in low- or middle-income level countries and mass-based maintenance dialysis programs are almost nonexistent In recent years, increase in government funding for dialysis has allowed coverage of the entire population in some South American countries like Brazil and Uruguay (4,5) However, in most of the developing world, dialysis. .. hemodialysis and peritoneal dialysis J Am Soc Nephrol 1999; 10: 354–365 Collins AJ, Hao W, Xia H, et al Mortality risks of peritoneal and hemodialysis Am J Kidney Dis 1999; 34 :106 5 107 4 Jindal KK, Hirsch DJ Excellent technique survival on home peritoneal dialysis: results of a regional program Perit Dial Int 1994; 14:324–326 Kiernan L, Kliger A, Gorban-Brennan N, Juergensen P, Finkelstein F Comparison of. .. censored by transfer to another dialysis modality, at transplantation, loss of follow-up, or at the end of observation In the intention-to-treat survival analysis, survival time is considered as the sum of the time on HD and the time on PD Death is considered as the final event, and patients are censored at the moment of transplantation, at loss of follow-up, or at the end of the observation, but not when... A vast majority of the population cannot derive the benefit of modern dialysis therapy of the standard practiced in the Dialysis in Developing Countries 821 developed world In recent years, active support of dialysis programs by some governments has led to substantial improvement in the quality and quantity of dialysis being provided in a number of countries However, in the majority of developing countries,... probability of being alive at a moment 806 Van Biesen et al Fig 3 Survival of patients (UH Gent) maintained on PD (dotted line) or HD (full line) for more than 48 months; p = 0.01 Number of surviving patients at 48 months is taken at 100 % ti, can be expressed as the product of the probabilities of surviving to time point t(i-x), under the condition that one has survived to time point t(i-x-1) [notation S(t(i-x)/t(i-x-1)]... the ratio of the hazards of two different subjects, can be calculated If this is done, and the two subjects differ only for one comorbid con- Fig 4 Survival of patients (UH Gent) who were treated with ‘‘integrated care’’ (dotted line) compared to the survival of patients who were started and remained on hemodialysis The two groups of patients were ‘‘matched’’ for the same duration of time of dialysis . correcting for the number of comorbid conditions, by re- doing the HD-versus-PD comparison using both treatment-received and intention-to-treat methodologies, and by varying both the time of. on take-on rate of patients on dialysis, utilization of different dialysis modalities, and the impact of socio-economic factors are discussed in Chapter 43. Studies comparing outcome of patients. product of the probabilities of surviving to time point t(i-x), under the condition that one has survived to time point t(i-x-1) [notation S(t(i-x)/t(i-x-1)]. At any time point, S (t/t-1) can