The impact of home nocturnal hemodialysis on end-stage renal disease therapies: A decision analysis

Transcription

1 review & 2006 International Society of Nephrology The impact of home nocturnal hemodialysis on end-stage renal disease therapies: A decision analysis PA McFarlane 1, AM Bayoumi 2, A Pierratos 3 and DA Redelmeier 4 1 St Michael s Hospital, Toronto, Ontario, Canada; 2 Department of Medicine, Inner City Health Research Unit, St Michael s Hospital, University of Toronto, Toronto, Ontario, Canada; 3 Humber River Regional Hospital, Toronto, Ontario, Canada and 4 Canada Research Chair in Medical Decision Sciences, Sunnybrook and Women s College Health Sciences Centre, Toronto, Ontario, Canada Home nocturnal hemodialysis (HNHD) is cost-effective relative to in-center hemodialysis (IHD) in short-run analyses. The effect in long-run analyses, when technique failures, declining benefits, delayed training, ation and death are considered, is unknown. We used decision analysis techniques to examine the relative cost-effectiveness of HNHD and IHD, projecting future costs and health effects over a lifetime with end-stage renal disease. We developed a Markov state-transition model comparing two strategies: only IHD or starting on IHD and subsequently transferring to HNHD. The model incorporates ation. In the base case, half the population was eligible for ation, with 1 3 of grafts from live donors. The time to was 0.75 years for live and 5 years for deceased donor s. The delay before initiation of HNHD was 5 years. Costs and outcomes were discounted at 3% per annum. Model parameters were derived from a literature review. We also conducted one-way sensitivity analyses and Monte Carlo simulations. The HNHD strategy was associated with a quality-adjusted survival estimate of 5.79 quality-adjusted life years (QALYs), with lifetime costs of $ The values for IHD were 5.31 QALYs and $ , respectively. Thus, HNHD is cost saving while improving quality of life. The incremental cost utility ratio was consistently less than $ per QALY in sensitivity and Monte Carlo analyses. Important determinants of cost-effectiveness were ation time and whether benefits declined over time. Our model suggests that HNHD improves quality-adjusted survival over IHD at an economically attractive cost effectiveness ratio. Kidney International (2006) 69, doi: /sj.ki ; published online 4 January 2006 KEYWORDS: cost-effectiveness; decision analysis; economic analysis; hemodialysis; hemodialysis, home; resource allocation Correspondence: PA McFarlane, St Michael s Hospital, 61 Queen St. E., 9th Floor, Toronto, Ontario, Canada M5C 2T2. phil.mcfarlane@ utoronto.ca Received 7 February 2005; revised 4 July 2005; accepted 8 July 2005; published online 4 January 2006 End-stage renal disease (ESRD) is a potentially devastating illness that can reduce the length and the quality of a person s life. 1 6 Mortality rates for individuals with ESRD are 20 times higher than for the general population, 7 and the average quality of life on dialysis has been reported as similar to that seen in people with hepatocellular carcinoma. 8 The treatment of choice for ESRD is renal ation, 9 which improves length and quality of life Unfortunately, the supply of kidneys for ation is less than the demand. For many, dialysis will be the only therapy for ESRD, thereby mandating attention to methods of improving quality of life on dialysis. In 1993, a program was developed in Toronto offering home nocturnal hemodialysis (HNHD). This technique combined prior innovations such as more frequent treatments, 14 long dialysis times 15,16 and treatment in the home. 17 The HNHD method as practiced in Toronto involves five to seven home-based treatments a week, with the patient dialyzing for 6 8 h overnight as they sleep. Previous research by our group has demonstrated in a 1-year analysis that HNHD is less expensive than hospital-based in-center hemodialysis (IHD). Furthermore, patients performing HNHD have a quality of life that is significantly higher than that of those performing IHD and similar to that of those who have received a kidney, 18,19 making it an economically dominant therapy (both cost saving and more effective). While the benefits of HNHD have been demonstrated prospectively in prevalent dialysis patients who remain on their original dialysis modality, the impact of HNHD over a long-run analysis is uncertain. Cross-sectional and prospective cohort analyses may overestimate both cost savings and health benefits for four reasons. First, short-run studies may exclude HNHD patients who fail the modality and return to IHD, or those who receive a kidney, which also reduces costs and improves length and quality of life Such considerations are particularly important when the time to ation is short (e.g., when a live donor is available, or when the patient has been on the waiting list for many years). Second, short-run analyses may be inaccurate when benefits decline over time, costs rise over time or when there are large initial expenses (such as capital 798 Kidney International (2006) 69,

2 PA McFarlane et al.: Home nocturnal hemodialysis: A decision analysis r e v i e w and training costs) that are realized when the patient starts on the modality, with savings accruing later. Third, as the number of centers that perform HNHD is small, the generalizability of these results has not been confirmed. Fourth, a critical assumption of short-run studies is that the quality of life of a patient who is transferred from IHD to HNHD will rise to the level of an HNHD patient in our cross-sectional study, an assumption that has not been prospectively tested. These deficiencies lead to five important questions. What is the impact of HNHD over a lifetime of ESRD therapy? Does the delay before training a patient for HNHD reduce cost-effectiveness? How long must a patient perform HNHD before the modality is cost-effective? What if costs and benefits change over time? What if there is a delay from the start of HNHD until the benefits are realized? A prospective cohort study that followed patients on IHD and HNHD over their lifetimes could answer these questions; however, timing and funding issues make this an impractical choice. Herein, we provide a decision analysis model that estimates the costeffectiveness of HNHD over a lifetime of ESRD. The decision analysis model also identifies parameters that impact on cost-effectiveness and merit priority in clinical trials. This information would be especially useful to health-care payers who are considering funding for HNHD programs. RESULTS Reference case analysis For the reference case, our model estimated that the remaining life expectancy would be 13.4 years if the patient was treated with HNHD and 13.2 years if treated with IHD. As we assumed an identical mortality rate on dialysis for both groups, this small difference in overall survival related to a higher rate of complications on dialysis for IHD. After applying quality of life weights and discounting, the health benefits associated with HNHD and IHD were 5.79 and 5.31 quality-adjusted life year (QALYs), respectively. The corresponding discounted costs were $ and $ That is, HNHD was 9.0% more effective, producing 0.48 more QALYs while saving $5508, or about 1.0% of total health expenses, and was the economically dominant option. The net monetary benefit (NMB) was positive and favored HNHD across the tested range of l (Table 1). Scenario analyses What if a patient was ineligible for? When ation was not modeled, HNHD produced 0.47 more QALYs while saving $6841. HNHD was dominant and the NMB favored HNHD across the tested range of l (Table 1). What if HNHD was selected as the initial dialysis modality for a patient with ESRD? In this scenario, HNHD generated 0.99 more QALYs, while reducing costs by $9118. Again, HNHD was dominant with a positive NMB across the tested range of l (Table 1). What if there is a delay between the start of HNHD and when the improvements in quality of life associated with HNHD are realized? When modeling the improvement in utility for HNHD as a linear function over time, HNHD was the dominant modality even when the time to full HNHD utility was longer than 20 years. Finally, we examined the impact of quality of life improvements and cost savings that declined over time, such that eventually HNHD and IHD had equivalent costs and Table 1 Results of primary and scenario-based analyses from a decision analysis examining the cost-effectiveness of HNDH QALYs generated Cost accrued Net monetary benefit at cost effectiveness ceiling Incremental cost effectiveness ratio $0 $ $ Reference case analysis IHD 5.31 $ $ $5508 $ $ HNHD 5.79 $ Difference (HNHD IHD) 0.48 $5 508 Transplant not allowed scenario IHD 2.78 $ $ $6841 $ $ HNHD 3.25 $ Difference (HNHD IHD) 0.47 $6 841 HNHD as initial dialysis modality scenario IHD 5.31 $ $9210 $9118 $ $ HNHD 6.30 $ Difference (HNHD IHD) 0.99 $9 118 Initial HNHD without ation scenario IHD 5.48 $ $ $ $ $ HNHD 7.22 $ Difference (HNHD IHD) 1.74 $ QALYs=quality-adjusted life years; IHD=in-center hemodialysis; HNHD=home nocturnal hemodialysis. Kidney International (2006) 69,

3 r e v i e w PA McFarlane et al.: Home nocturnal hemodialysis: A decision analysis effectiveness. When the differences in both costs and utility declined linearly over time, HNHD was not cost-effective (l ¼ $50 000/QALY) when the time until the costs and benefits were equivalent was less than 4 years. The model was not sensitive to costs and utility that declined individually. Sensitivity analyses HNHD was no longer cost-effective if IHD was less expensive than HNHD (i.e. annual IHD cost less than $57 000), if HNHD was more expensive than IHD (i.e. annual HNHD cost more than $75 000) or if the utility score was higher for IHD. A shorter time to HNHD training improved the costeffectiveness of HNHD; however, HNHD remained the economically dominant option even with extremely long delays (440 years). HNHD was cost-effective unless the time to deceased or live donor was less than 108 weeks. The model was not sensitive to post- variables. We performed 1000 iterations of the model in the Monte Carlo analysis (Figure 1). The estimated lifetime healthcare costs were $ $ for IHD and $ $ for HNHD. The estimated number of QALYs generated over a lifetime of ESRD by IHD was , and by HNHD was HNHD was dominant in 75.9% of iterations, and cost-effective (l ¼ $50 000/QALY) in 99.7% of iterations. Figure 2 illustrates an acceptability curve based on the Monte Carlo data graphing the probability that HNHD is cost-effective across a wide range of possible values for the societal maximum willingness to pay for a QALY. DISCUSSION Our previous work demonstrated that HNHD is costeffective in prevalent dialysis patients. This study explored the cost-effectiveness of HNHD over a lifetime of renal replacement therapy to address questions that have not been answered in clinical trials to date. This model found that HNHD remained the dominant therapy in a long-run analysis, despite the design of our analysis that deliberately biased the model against HNHD. For example, based on our previous studies, we selected for our reference case an average 5-year delay before switching from IHD to HNHD, a period similar to the average time to. As a result, less than half of the patients in the HNHD strategy of our reference analysis ever transfer to that form of dialysis. To provide another example, we assumed that home and conventional hemodialysis patients have the same mortality rate, although some data suggest that home dialysis patients may have better longevity. 20 We also assumed that many of the known health benefits of HNHD (such as better blood pressure control) do not directly lead to cost savings. Finally, we did not allow for the capital costs of HNHD to be amortized over long periods of time, and assumed that in-center facilities were already in existence and IHD capital costs included only new dialysis machines. Even with these features, our findings were robust, with our model broadly favoring HNHD in sensitivity analyses. The Monte Carlo analysis also showed the superiority of HNHD, and is likely the most compelling analysis as it accounts for the range and distribution of uncertainty in the model. However, HNHD was economically unattractive if the costs of IHD were less than HNHD, if IHD was associated with a higher quality of life or if the benefits of HNHD declined rapidly. When compared to our previous 1-year study, 18,19 this longrun analysis predicts cost savings and utility gains that are less in relative and absolute terms, but still clinically significant. 21,22 It is likely that the economic advantage of HNHD is less over long-run analyses (which account for training delays, patient death and ation) than predicted in cross-sectional or short-run studies. At this time, there is insufficient clinical trial data to allow modeling comparisons between HNHD and peritoneal dialysis or intensive forms of hemodialysis such as short daily hemodialysis. Difference in cost (HNHD IHD) $ $ $ $ $ $0 $ $ $ $ Result of Monte Carlo iteration Cost effectiveness ceiling ratio = $ Cost effectiveness ceiling ratio = $ Cost effectiveness ceiling ratio = $ $ Difference in effectiveness (HNHD IHD) Figure 1 Scatter plot of cost and effectiveness differences between home nocturnal (HNHD) and in-center hemodialysis (IHD) from 1000 iterations of a second-order Monte Carlo decision analysis model. Probability that HNHD is cost-effective 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% $0 $ $ $ $ $ $ $ $ $ $ Cost-effective ceiling ratio Figure 2 Cost effectiveness acceptability curve based on results from 1000 iterations of a second-order Monte Carlo decision analysis examining the cost-effectiveness of home nocturnal hemodialysis (HNHD). 800 Kidney International (2006) 69,

4 PA McFarlane et al.: Home nocturnal hemodialysis: A decision analysis r e v i e w Our model assumed that all capital costs for HNHD are borne at the initiation of the modality, with cost savings accruing over subsequent years. For this reason, changes in any parameter that allows patients to stay on HNHD longer improved the cost-effectiveness of this therapy. For example, a shorter time to HNHD training and a lower HNHD technique failure rate improved the cost-effectiveness of HNHD. As ation has similar effectiveness to HNHD but lower costs, ation has the effect of reducing the advantages of HNHD over IHD. Consequently, changes in parameters that reduced access to ation advantaged HNHD. For example, longer wait times (including lower access to live donors) and lower proportions of eligibility improved the cost-effectiveness of HNHD. Allowing HNHD capital costs to be amortized over 5 years substantially improved the cost-effectiveness of HNHD. Our model suggests that HNHD is no longer costeffective if the time from HNHD training to renal is less than 2 years (e.g. in jurisdictions with short cadaveric wait times, or when a potential live kidney donor is available). While clinicians may still opt to place patients on HNHD in these cases, they should be aware of the potential negative economic impact. This emphasizes the need to identify appropriate HNHD candidates as early as possible. A weakness of this model relates to the uncertainty in some of the key parameters, such as the utility of HNHD. The cost of HNHD and quality of life while on this modality were estimated from relatively small studies that were not randomized. It is possible that a randomized trial could negate the findings of this model and its supporting trials. Additionally, the values for some model parameters (e.g. graft and patient survival) were derived from unselected cohorts with characteristics different from our simulated cohort. Further research is needed to examine the costs and utility of IHD or HNHD, preferably in a prospective randomized manner. In addition, a decision analysis model is, by necessity, a simplified view of the states of health experienced by individuals. While we have modeled the major health states an individual with ESRD would experience (dialysis, ation, hospitalization and death), it is possible that the structure of our model does not sufficiently reflect the complexities of the lives of people with ESRD, and thus our results do not accurately reflect the reality of life with kidney disease. For example, we did not model the impact of renal disease on family members. The cohort that we have modeled is similar to those HNHD patients that have been studied in clinical trials; however, these patients are systematically different from the average hemodialysis patient. Patients in HNHD studies to date tend to be younger, educated and less likely to have diabetes or cardiac disease. Psychological parameters such as self-motivation may also be different in studied HNHD patients. Also, by definition, patients in HNHD studies have had access to an HNHD program. For these reasons, the results of previous clinical trials and this decision analysis should not be generalized to the broader hemodialysis population. Finally, our modeled cohort was based on studies of prevalent dialysis patients, introducing the possibility of survivorship bias. For example, patients who trained on HNHD but failed and returned to IHD or suffered an early death would not have been represented in the cost utility trials performed to date, and could have higher health-care costs and lower utility scores. Inclusion of these patients would have a tendency to reduce the benefits of HNHD. We have allowed HNHD failure in our model; however, this does not eliminate potential survivorship bias. This study suggests that HNHD is a long-term cost-effective addition to the therapeutic options for ESRD. This conclusion is robust and consistent over a wide range of sensitivity analyses; however, no large-scale randomized trials have been carried out in this area. Further modeling may be required in the future as additional clinical trial data become available. MATERIALS AND METHODS Model overview A Markov state transition decision analysis model was created to evaluate the projected outcomes over a lifetime of ESRD therapies, with and without HNHD, allowing for modality change, ation and patient death. The model was constructed in Data Pro release 10 (TreeAge Software Inc., Williamstown, MA, USA). Markov-based decision analysis models represent an illness as a series of discrete states, with the transitions between these states representing the progression of the disease. Markov models are Cadaveric donor Live donor Starting state Well on IHD Transfer to HNHD 1 Well on HNHD year 1* Live donor year 1 Cadaveric year 1 HNHD failure with return to IHD Well on HNHD subsequent years Pre period Post period Live donor subsequent years Transplant failure Cadaveric subsequent years 1 This transfer not available in the IHD arm, and patients cannot return to HNHD after HNHD failure *First year of HNHD is a composite state including both well patients and those with a dialysis complication **All states can lead to patient death Complication on IHD Complication on HNHD subsequent years Patient death** Dialysis post failure Complication on IHD post failure Figure 3 Map of states and allowed transitions for pre- and post periods. Kidney International (2006) 69,

5 r e v i e w PA McFarlane et al.: Home nocturnal hemodialysis: A decision analysis useful because they show how a disease evolves over time. Our model compared the outcomes when a patient receives only IHD as their form of dialysis, versus starting on IHD and subsequently transferring to HNHD. Live and deceased donor ations were also modeled for both groups. The start time of the model (i.e. time zero) is the first day of renal replacement therapy. One-way sensitivity analyses and Monte Carlo simulations were used to test the effect of uncertainty in the model. Our analysis followed the recommendations of the Panel on Cost-Effectiveness in Health and Medicine, 23 with the exception of the study perspective, which was that of the health-care payer. This was the perspective taken in the studies used as the basis of the cost estimates in this model. It is possible that broader perspectives, such as the societal perspective, which take into account items such as productivity effects, may produce different estimates of the economic impact of various forms of dialysis; however, little data of this type are available at this time. Supplementary material accompanying this manuscript provides further details regarding the model s structure (See Appendix S1). Costs and effectiveness In this study, QALYs were taken as the measure of effectiveness of dialysis. We calculated QALYs by multiplying the utility of a particular health state by the duration spent in the state. Utilities represent patient preferences for health states, and are ranked on a continuous scale between 0.0 (a quality of life equivalent to death) and 1.0 (the best possible quality of life). Costs were determined from the perspective of the health-care payer (such as a provincial Ministry of Health or HMO), and converted to year 2003 Canadian dollars. Costs not described in the model (such as treatments for conditions unrelated to ESRD) were assumed to be equal in both groups. This biases the model against HNHD, as we assumed that many of the benefits of HNHD (such as improvements in cardiac function and sleep apnea) could not lead to direct cost savings in our model, although they may contribute to the higher utility scores and lower hospitalization modeled for HNHD. The base estimates for costs and utility were taken from our previously published prospective studies, with ranges for testing in sensitivity analyses taken from a broader literature review. 12,18,19,24 62 Both costs and utilities were discounted at a rate of 3%/year. 23 Cost-effectiveness was calculated using the incremental cost effectiveness ratio (ICER ¼ D Cost/D Effectiveness, where D Cost- Cost HNHD Cost IHD and D Effectiveness ¼ Effectiveness HNHD Effectiveness IHD ) and the net monetary benefit (NMB ¼ (D Effectiveness l) D Costs, where l is the societal maximum Table 2 Variable baseline values and tested ranges for the pre period from a decision analysis examining the cost-effectiveness of HNDH Primary analysis Monte Carlo analysis Tested range Standard distribution Variable Base Low High Mean Error Shape Source a Transfer probabilities annual HNHD back to IHD 5% 0% 40% 5.0% 2.2% Beta Center Death on dialysis On waiting list 6.3% 5% 8% 6.3% 2.4% Beta (10) Not on waiting list 16.1% 12% 20% 16.1% 3.7% Beta (10) Complication on dialysis IHD 12.8% 3.1% 15.9% 12.8% 3.3% Beta (18, 19) HNHD 3.1% 2.3% 12.8% 3.1% 1.7% Beta (18, 19) Transfer probabilities weekly Death with dialysis complication 10% 5% 20% 5% 2.2% Beta Estimate Resolution of dialysis complication 80% 50% 95% 80% 4% Beta Estimate Annual costs (in thousands of 2003 Canadian dollars) HNHD first year $107 $75 $140 $107 $3 Gamma (18, 19, 59) HNHD postyear 1 $57 $40 $95 $57 $1 Gamma (18, 19, 59) IHD $72 $54 $74 $72 $4 Gamma (18, 19, 21, 59) Weekly costs (in thousands of 2003 Canadian dollars) Complication on dialysis IHD $7 $5 $9 $7 $2 Gamma (18, 19, 21, 59) HNHD $4 53 $5 $4 $1 Gamma (18, 19, 21, 59) Utilities Complication on dialysis Beta Estimate HNHD Beta (19, 45) IHD Beta (12, 19, 45) Other HNHD delay (weeks) Gamma Center Center=data derived from study hospital; Estimate=values derived from author s judgement. a Source refers to the reference number from the references used as basis for the parameter estimate. 802 Kidney International (2006) 69,

6 PA McFarlane et al.: Home nocturnal hemodialysis: A decision analysis r e v i e w willingness to pay for a QALY). As the value of l is unknown, we investigated a broad range of possible values. Modeled states Figure 3 provides an overview of the health states studied in this model. The model begins with patients initiating dialysis in the Well on Dialysis state. On a weekly basis, they could remain well, develop a significant complication on dialysis, receive a live or deceased donor renal or die. A complication on dialysis was defined as a nonfatal event requiring hospitalization for at least a week, and ation was not allowed while such a complication was present. In the HNHD arm, the patient began on IHD and could then transfer to HNHD. Subsequently, there was a chance of failure of HNHD, with the return of the patient to IHD. All capital costs for HNHD were borne in the first year. We assumed that the graft functioned initially following ation, but allowed for short- and long-term graft loss with subsequent return of the patient to IHD. As a simplifying assumption, we did not allow for a second following graft failure, as reation is a relatively rare event, 7 and this exclusion is unlikely to significantly bias the results of the model. The modeled costs for renal were highest in the first post year. 12 Death could occur from any state. Parameter value assignment A literature review was conducted to determine the range and distribution of likely values for parameters such as costs, utilities, and transition probabilities. To account for variability in patient populations and uncertainty and error in these estimates, the range of each variable was based on available literature (Tables 2 and 3). We based time to, proportion of population eligible for, proportion of s from live donors and time to HNHD on our hospital s experience. One-way sensitivity analyses were performed to test the robustness of the model across the probable range of variables in the model (Tables 2 and 3). A secondorder Monte Carlo simulation was performed, where the model was run 1000 times, each time randomly selecting a new value for all uncertain variables from their likely distribution (Tables 2 and 3). Table 3 Variable baseline values and tested ranges for the post- period from a decision analysis examining the cost-effectiveness of HNDH Primary analysis Monte carlo analysis Tested range Standard distribution Variable Bass Low High Mean Error Shape Source a Transfer probabilities annual Post- patient death rate (per year) Live donor first year 2.0% 10.0% 10.0% 2.0% 0.2% Beta (7, 61 63) Years % 1.5% 1.5% 1.0% 0.1% Beta (7, 61 63) Years % 1.0% 1.0% 2.0% 0.2% Beta (7, 61 63) Cadaveric donor first year 6.0% 11.0% 11.0% 6.0% 0.6% Beta (7, 6l 63) Years % 3.0% 3.0% 2.8% 0.3% Beta (7, 6l 63) Years % 2.4% 2.4% 3.0% 0.3% Beta (7, 61 63) Post- graft loss rate (per year) Live donor first year 6.0% 5.0% 5.0% 6.0% 0.6% Beta (7, 61 63) Years % 0.5% 0.8% 1.8% 0.18% Beta (7, 6l 63) Years % 2.0% 2.0% 1.0% 0.10% Beta (7, 61 63) Cadaveric donor first year 12.0% 10.0% 10.0% 12.0% 1.2% Beta (7, 6l 63) Years % 4.3% 4.3% 2.8% 0.3% Beta (7, 6l 63) Years % 0.6% 0.6% 0.4% 0.0% Beta (7, 6l 63) Annual costs (in thousands of 2003 Canadian dollars) Transplant year 1 Cadaveric donor $95 $70 $120 $95 $10 Gamma (12) Live donor $103 $75 $130 $103 $10 Gamma (12) Transplant subsequent years $40 $30 $50 $40 $4 Gamma (12) Utilities Working utility Beta (12) Others Time to (yeans) Live donor Gamma Center Deceased donor Gamma Center Proportion Eligible for 50% 0% 100% 50% 20% Beta Estimate From live donor 33% 0% 100% 33% 10% Beta Estimate Discount rate (per year) 3.0% 2.5% 5.0% Estimate Center=data derived from study hospital; Estimate=values derived from author s judgement. a Source refers to the reference number from the references used as basis for the parameter estimate. Kidney International (2006) 69,

7 r e v i e w PA McFarlane et al.: Home nocturnal hemodialysis: A decision analysis Table 4 Demographic profile of the cohort simulated in the reference case Sex (% male) 75% Diabetes 8% Coronary artery disease 21% Peripheral vascular disease 13% Congestive heart failure 21% Postsecondary education 60% Married 94% Age (years) a Duration ESRD (years) a a Mean7s.d. The estimated mortality rates while on dialysis were based on registry data, and assumed a rate of 16.1% in dialysis patients who were not on the waiting list, and 6.3% for those who were. 10 Our dialysis mortality rate was calculated based on the proportion of patients who were on the waiting listed. Mortality on dialysis following graft failure is 9% lower than for those not eligible for. 63 Following ation, weekly probabilities of death, and graft loss between 0 and 1, 1 and 5, and beyond 5 years were based on registry data assuming a linear rate change between these time points. 7,64 66 Reference case Tables 2 and 3 list the variable values used in our reference case scenario. Our reference case was based on a hypothetical cohort of patients whose demographic characteristics matched the average values from our previous studies (Table 4). 18,19 The typical patient in the reference case would be a 50-year-old man with relatively few comorbidities (e.g. diabetes, coronary artery disease), and who has been on dialysis for a long period of time. Based on the experience of home hemodialysis patients at our center, we allowed half the population to be eligible for ation, with 1 3 of grafts coming from live donors. The time to was 3 4 of a year for live donors, and 5 years for deceased donor s. In addition, the delay before initiation of HNHD was typically quite long (5 years). Of note, the reference case cohort are generally younger, have been on dialysis longer, more likely to be listed for and have fewer comorbidities than a typical hemodialysis patient. 7 Analysis We analyzed model uncertainty using sensitivity analyses. For the Monte Carlo simulation, we report the percentage of times that HNHD was cost-effective relative to IHD (l ¼ $50 000/QALY), and a cost-effectiveness acceptability curve was created where the probability that HNHD was cost-effective was calculated by determining the percentage of iterations with an incremental cost/ QALY that was equal to or less than a series of cost effectiveness ceiling ratios (which represent the societal maximum willingness to pay for a QALY, tested range $0 $ ). 67 To test the effect of ation, we changed our model so that all patients in the HNHD arm started on HNHD, with all patients in the scenario eligible for a live donor, and then varied the time to ation. ACKNOWLEDGMENTS The research was conducted at St Michael s Hospital and Humber River Regional Hospital, Toronto, Ontario, Canada. Research support for Dr McFarlane was provided by the Canadian Society of Nephrology and the Kidney Foundation of Canada. SUPPLEMENTARY MATERIAL Appendix S1. Technical Appendix. REFERENCES 1. Consensus Development Conference Panel. Morbidity and mortality of renal dialysis: an NIH consensus conference statement. Ann Intern Med 1994; 121: Krolewski M, Eggers PW, Warram JH. Magnitude of end-stage renal disease in IDDM: a 35 year follow-up study [published erratum appears in Kidney Int 1997; 51(3): 978]. Kidney Int 1996; 50: Weller JM, Port FK, Swartz RD et al. Analysis of survival of end-stage renal disease patients. Kidney Int 1982; 21: United States Renal Data System. The USRDS Dialysis Morbidity and Mortality Study: Wave 2. Am J Kidney Dis 1997; 30: S67 S United States Renal Data System. USRDS 1999 Annual Data Report. The National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases: Bethesda, MD, Campbell JD, Campbell AR. The social and economic cost of end-stage renal disease. A patient s perspective. N Engl J Med 1978; 299: Canadian Organ Replacement Registry (CORR). Canadian Organ Replacement Registry (CORR): 2001 Annual Report. Canadian Institute for Health Information: Ottawa, ON, Canada, Bell CM, Chapman RH, Stone PW et al. An off-the-shelf help list: a comprehensive catalog of preference scores from published cost utility analyses. Med Decision Making 2001; 21: Renal Physicians Association. Appropriate patient preparation for renal replacement therapy. In: Renal Physicians Association Clinical Practice Guidelines. Duke Evidence-Based Practice Center, Center for Clinical Health Policy Research: Durham, NC, 2002, p Wolfe RA, Ashby VB, Milford EL et al. Comparison of mortality in all patients on dialysis, patients on dialysis awaiting ation, and recipients of a first cadaveric. N Engl J Med 1999; 341: Hunsicker LG. A survival advantage for renal ation. N Engl J Med 1999; 341: Laupacis A, Keown P, Pus N et al. A study of the quality of life and cost utility of renal ation. Kidney Int 1996; 50: Arend SM, Mallat MJ, Westendorp RJ et al. Patient survival after renal ation; more than 25 years follow-up. Nephrol Dial Transplant 1997; 12: Buoncristiani U, Quintaliani G, Cozzari M et al. Daily dialysis: long-term clinical metabolic results. Kidney Int Suppl 1988; 24: S137 S Laurent G, Calemard E, Charra B. Long dialysis: a review of fifteen years experience in one centre Proc Eur Dial Transplant Assoc 1983; 20: Laurent G, Charra B. The results of an 8 h thrice weekly haemodialysis schedule. Nephrol Dial Transplant 1998; 13: Shaldon S, Oakley JJ. Experience with regular haemodialysis in the home. Br J Urol 1966; 38: McFarlane PA, Pierratos A, Redelmeier DA. Cost savings of home nocturnal versus conventional in-center hemodialysis. Kidney Int 2002; 62: McFarlane PA, Bayoumi AM, Pierratos A et al. The quality of life and cost utility of home nocturnal and conventional in-center hemodialysis. Kidney Int 2003; 64: Woods JD, Port FK, Stannard D et al. Comparison of mortality with home hemodialysis and center hemodialysis: a national study. Kidney Int 1996; 49: Redelmeier DA, Guyatt GH, Goldstein RS. Assessing the minimal important difference in symptoms: a comparison of two techniques. J Clin Epidemiol 1996; 49: Redelmeier DA, Goldstein RS, Min ST et al. Spirometry and dyspnea in patients with COPD. When small differences mean little. Chest 1996; 109: Gold MR, Siegel JE, Russell LB et al. Cost-Effectiveness in Health and Medicine. Oxford University Press: Oxford, Lee H, Manns B, Taub K et al. Cost analysis of ongoing care of patients with end-stage renal disease: the impact of dialysis modality and dialysis access. Am J Kidney Dis 2002; 40: McFarlane PA. Reducing hemodialysis costs: conventional and quotidian home hemodialysis in Canada. Semin Dial 2004; 17: Goeree R, Manalich J, Grootendorst P et al. Cost analysis of dialysis treatments for end-stage renal disease (ESRD). Clin Invest Med 1995; 18: Douglas RA. The costs of kidney ation and hemodialysis. Transplant Proc 1973; 5: Kidney International (2006) 69,

8 PA McFarlane et al.: Home nocturnal hemodialysis: A decision analysis r e v i e w 28. Stewart JH, Topp ND, Martin S et al. The costs of domiciliary maintenance haemodialysis. A comparison with alternative renal replacement regimens. Med J Aust 1973; 1: Buxton MJ, West RR. Cost benefit analysis of long-term haemodialysis for chronic renal failure. Br Med J 1975; 2: Hoffstein PA, Krueger KK, Wineman RJ. Dialysis costs: results of a diverse sample study. Kidney Int 1976; 9: Schippers HM, Kalff MW. Cost comparison haemodialysis and renal ation. Tissue Antigens 1976; 7: Bulgin RH. Comparative costs of various dialysis treatments. Periton Dial Bull 1981; 1: Roberts SD, Maxwell DR, Gross TL. Are home hemodialysis aides cost-effective? Cost of home hemodialysis aides. Arch Intern Med 1981; 141: Roberts SD, Maxwell DR, Gross TL. Cost-effective care of end-stage renal disease: a billion dollar question. Ann Intern Med 1980; 92: Stange PV, Sumner AT. Predicting treatment costs and life expectancy for end-stage renal disease. N Engl J Med 1978; 298: Churchill DN, Lemon BC, Torrance GW. A cost-effectiveness analysis of continuous ambulatory perioneal dialysis and hospital hemodialysis. Med Decision Making 1984; 4: Churchill DN, Torrance GW, Taylor DW et al. Measurement of quality of life in end-stage renal disease: the time trade-off approach. Clin Invest Med 1987; 10: Churchill DN, Wallace JE, Ludwin D et al. A comparison of evaluative indices of quality of life and cognitive function in hemodialysis patients. Controlled Clin Trials 1991; 12: 159S 167S. 39. Garner TI, Dardis R. Cost-effectiveness analysis of end-stage renal disease treatments. Med Care 1987; 25: Tousignant P, Guttmann RD, Hollomby DJ. Transplantation and home hemodialysis: their cost-effectiveness. J Chronic Dis 1985; 38: Croxson BE, Ashton T. A cost effectiveness analysis of the treatment of end stage renal failure. NZ Med J 1990; 103: Sesso R, Eisenberg JM, Stabile C et al. Cost-effectiveness analysis of the treatment of end-stage renal disease in Brazil. Int J Technol Assess Health Care 1990; 6: Karlberg I, Nyberg G. Cost-effectiveness studies of renal ation. Int J Technol Assess Health Care 1995; 11: Karlberg I. Cost analysis of alternative treatments in end-stage renal disease. Transplant Proc 1992; 24: Coyte PC, Young LG, Tipper BL et al. An economic evaluation of hospital-based hemodialysis and home-based peritoneal dialysis for pediatric patients. Am J Kidney Dis 1996; 27: Rodriguez-Carmona A, Perez FM, Bouza P et al. The economic cost of dialysis: a comparison between peritoneal dialysis and in-center hemodialysis in a Spanish unit. Adv Periton Dial 1996; 12: Prichard SS. The costs of dialysis in Canada [Review] [2 refs]. Nephrol Dial Transplant 1997; 12(Suppl 1): Heidenheim AP, Muirhead N, Moist L et al. Patient quality of life on quotidian hemodialysis. Am J Kidney Dis 2003; 42: Kjellstrand C, Ting G. Daily hemodialysis: dialysis for the next century. Adv Renal Replace Ther 1998; 5: Brissenden JE, Pierratos A, Ouwendyk M et al. Improvements in quality of life with nocturnal hemodialysis. J Am Soc Nephrol 1998; 9: 168A. 51. Traeger J, Galland R, Arkouche W et al. Short daily hemodialysis: a four-year experience. Dial Transplant 2001; 30: Ting G, Freitas T, Carrie BJ et al. Short daily hemodialysis clinical outcomes and quality of life. J Am Soc Nephrol 1999; 9: 228A. 53. Ting GO. The case for short daily hemodialysis, why sdhd will be the predominant modality for frequent dialysis. ASAIO J 2001; 47: Mohr PE. The economics of daily dialysis. Adv Renal Replace Ther 2001; 8: Mohr PE, Neumann PJ, Franco SJ et al. The case for daily dialysis: its impact on costs and quality of life. Am J Kidney Dis (Online) 2001; 37: Mohr PE, Neumann PJ, Franco SJ et al. The quality of life and economic implications of daily dialysis. Policy Anal Brief H Ser 1999; 1: Lacson Jr E, Diaz-Buxo JA. Daily and nocturnal hemodialysis: how do they stack up? Am J Kidney Dis 2001; 38: Kooistra MP, Vos J, Koomans HA et al. Daily home haemodialysis in The Netherlands: effects on metabolic control, haemodynamics, and quality of life. Nephrol Dial Transplant 1998; 13: Robers S. Cost of daily hemodialysis. Asaio J 2001; 47: Simoons ML, Vos J, Martens LL. Cost utility analysis of thrombolytic therapy. Eur Heart J 1991; 12: Levin LA, Jonsson B. Cost-effectiveness of thrombolysis a randomized study of intravenous rt-pa in suspected myocardial infarction. Eur Heart J 1992; 13: Kroeker A, Clark WF, Heidenheim AP et al. An operating cost comparison between conventional and home quotidian hemodialysis. Am J Kidney Dis 2003; 42: Anonymous. Mortality rate comparisons of never, previously, and currently ed ESRD patients. Am J Kidney Dis 1992; 20: Cecka JM. The UNOS renal registry. Clin Transplants 2001: Cecka JM. The UNOS scientific renal registry. Clin Transplants 1996: Cecka M. Clinical outcome of renal ation. Factors influencing patient and graft survival. Surg Clin N Am 1998; 78: Fenwick E, Claxton K, Sculpher M. Representing uncertainty: the role of cost-effectiveness acceptability curves. Health Econ 2001; 10: Kidney International (2006) 69,