Economic evaluation of sevelamer in patients with end-stage renal disease

Transcription

1 NDT Advance Access published June 25, 2007 Nephrol Dial Transplant (2007) 1 of 12 doi: /ndt/gfm367 Original Article Economic evaluation of sevelamer in patients with end-stage renal disease Braden Manns 1,2,3, Scott Klarenbach 3,4, Helen Lee 1, Bruce Culleton 2, Fiona Shrive 1 and Marcello Tonelli 3,4,5,6 1 Department of Community Health Sciences, 2 Department of Medicine, Division of Nephrology, University of Calgary, Calgary, 3 Institute of Health Economics, 4 Department of Medicine, Division of Nephrology, 5 Division of Critical Care Medicine and 6 Department of Public Health Sciences, University of Alberta, Edmonton, AB, Canada Abstract Background. There is uncertainty about the most costeffective way to treat hyperphosphataemia in patients with end-stage renal disease. Methods. We performed an economic analysis which compared the use of sevelamer with calcium carbonate in a simulated cohort of North American dialysis patients, using the perspective of the health care purchaser and a lifetime horizon. Outcomes considered were quality-adjusted life years (QALYs) gained and health care costs. To account for uncertainty, we considered four separate modelling strategies, obtaining data on the relative effectiveness of sevelamer from the recent Dialysis Clinical Outcomes Revisited study. Results. In the base analysis, the use of sevelamer was associated with a cost per QALY gained of CAN$ , compared with calcium carbonate. Assuming no survival or hospitalization advantage for sevelamer, use of sevelamer resulted in an incremental cost of CAN$ per patient. In alternate models which assumed sevelamer to be more effective than calcium-based phosphate, the use of sevelamer was associated with a cost per QALY gained ranging from CAN$ $ Assuming that sevelamer resulted in a differential reduction in mortality in patients 65 years of age, use of sevelamer in this subgroup was associated with a cost per QALY of CAN$ Results were similar in groups defined by age 55 or by 45 years. Since dialysis is expensive, interventions for dialysis patients that improve survival without reducing the need for dialysis will be associated with a cost-utility ratio at least as great as that of dialysis itself. As such, we repeated the primary analysis excluding the costs of Correspondence and offprint requests to: Dr Marcello Tonelli, University of Alberta, Division of Nephrology and Immunology, Clinical Science Building, Street, Edmonton, Alberta T6B 2b7 Canada. mtonelli@ualberta.ca dialysis and transplantation and found that the cost per QALY gained for sevelamer was $ Conclusions. The cost per QALY gained for treating all dialysis patients with sevelamer exceeds what would usually be considered good value for the money. While the high cost per QALY was in part due to the inclusion of the costs of dialysis and transplant in the analysis, the cost per QALY gained remained relatively unattractive even when these costs were excluded. Although a lower cost per QALY gained is realized when only patients older than 65 years are treated, this strategy remains economically unattractive, particularly given the uncertainty of clinical benefit in this group. Keywords: Sevelamer; Calcium; Hyperphosphatemia; Economic Evaluation Introduction Patients with end-stage renal disease (ESRD) have abnormal mineral metabolism and high mortality rates [1]. Observational data have linked markers of abnormal mineral metabolism such as hyperphosphataemia to increased morbidity and mortality in patients with ESRD [2 4]. The treatment for hyperphosphataemia in ESRD includes the use of oral phosphate with meals. Calcium-based phosphate have traditionally been used as first line therapy, since they correct hypocalcaemia in addition to reducing serum phosphate levels and because they are inexpensive [5]. Non-calcium-based phosphate such as sevelamer have recently been developed for treatment of hyperphosphataemia in patients with ESRD. These agents are theoretically preferable to calcium-based agents, since they might permit control of serum phosphate at lower levels of serum calcium, which in turn could reduce adverse outcomes due to calcification of cardiac and vascular tissue [1]. Current practice ß The Author [2007]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. For Permissions, please journals.permissions@oxfordjournals.org

2 2of12 Table 1. Baseline clinical effects B. Manns et al. Variable Value Reference Plausible range considered Relative risk of death for sevelamer compared with calcium-based phosphate 0.91 [6] (Table 2) Relative risk of hospitalization for sevelamer (Table 2) 0.91 [6] Annual risk of death for patients 65 treated with calcium-based phosphate binder [11] Annual risk of death for patients <65 treated with calcium-based phosphate binder [11] Annual risk of transplant for patients [11] Annual risk of transplant for patients < [11] Mortality risk of patients with transplant [19] 25% Year 1 Years 2 and on Annual risk of transplant failure 0.04 [19] 25% Annual frequency of hospitalization for ESRD patients treated with calcium-based phosphate 1.54 [24] Alternate values: 1.87 [25] 1.98 [26] 1.8 (<65 [6]) 2.9 (65 [6]) Utility (EQ-5D) score for dialysis patients [20] Alternate value: 0.57 [21] Utility (EQ-5D) score for dialysis patients < [20] Alternate value: 0.57 [21] Utility (EQ-5D) score for patients with functioning renal transplant [21] Alternate value: 0.77 [21] Discount rate for effects 0.05 [18] ESRD, end-stage renal disease. guidelines (published in 2004) recommend the use of sevelamer in many common clinical situations [1], although the impact of this strategy on hard outcomes such as mortality and hospitalization was unknown until recently. The Dialysis Clinical Outcomes Revisited (DCOR) study compared mortality and morbidity over 3 years for patients randomized to treatment with sevelamer compared with those receiving calcium-based phosphate [6]. The primary analysis showed no difference in overall mortality (RR 0.91, P ¼ 0.30) between treatment groups. However, in a pre-specified secondary analysis, sevelamer was associated with better survival in patients aged older than 65 years. Since there is a large differential cost between sevelamer and calcium-based phosphate, we conducted an economic evaluation comparing these two treatment strategies in patients with ESRD on dialysis. Methods We sought to determine whether scarce resources should be allocated to use of sevelamer in ESRD patients using costutility analysis, a special form of cost-effectiveness analysis [7], where health benefits are expressed as quality adjusted life years (QALYs). There are two broad methods of performing a cost-utility analysis; in conjunction with a clinical trial or using decision analytic modelling [8]. Our analysis combined both methods, taking the estimate of effectiveness for sevelamer from DCOR and using decision analysis to extrapolate the potential clinical benefits and costs of sevelamer over the lifetime of an ESRD patient. The objective of this analysis was to determine the cost per QALY gained for sevelamer compared with calcium-based phosphate in North American patients with dialysisdependent ESRD. Patient population In the base case analysis, we evaluated a simulated cohort of dialysis patients who were 18 years and whose characteristics were representative of typical Canadian dialysis patients. In sensitivity analyses, we considered patients more reflective of dialysis patients in the US [9], as well as patients enrolled in DCOR [6]. For the population considered in the base case analysis, we used data from the Canadian Organ Replacement Registry (CORR) which collects information on Canadian dialysis patients undergoing dialysis or transplantation [10]. For another analysis, CORR provided us with data from a random sample of 37% of patients initiating renal replacement in Canada between 1 January 1996 and 31 December 2000 (n ¼ 7034 after excluding children) [11]. This is an appropriate comparative group given that sevelamer would not have been used in the majority of these patients. Data taken from this cohort is described subsequently and in Table 1. Treatment comparators We sought to compare sevelamer with the most commonly used phosphate binder in use in patients with ESRD in

3 Economic evaluation of sevelamer 3 of 12 Treatment strategies Sevelamer Calcium-based phosphate Stratification by subgroup Age under 65 or age over 65 Dialysis Outcomes Transplanted Continue dialysis Annual mortality risk Fig. 1. Markov model, displaying the possible clinical outcomes in ESRD patients treated with either sevelamer or calcium carbonate. Canada, calcium carbonate [5]. This was also one of the calcium-based phosphate used as standard care for many of the patients enrolled in sevelamer clinical trials [12 15], including DCOR [6]. Analytical approach To determine the cost-effectiveness of sevelamer, we used a Markov model, adapted from previous analyses [16,17]. We considered yearly transitions between three clinical states, alive on dialysis, alive with a transplant and dead (Figure 1). The model was analysed over the patient s lifetime, with shorter timelines considered in sensitivity analysis. This analysis was conducted from the perspective of the health care purchaser, consistent with published guidelines [18] and reasonable given the absence of data as to whether sevelamer impacts indirect costs experienced by patients [6]. The model outputs were QALYs, life years gained, costs and the cost per QALY gained. All analyses were performed using Treeage Pro 2005 (Treeage Software, Inc., Williamstown, USA). Clinical effects of patients treated with calcium-based phosphate Baseline transition probabilities and clinical effects for patients treated with calcium-based phosphate are noted in Table 1. The probabilities of mortality and transplantation for patients treated with calcium-based phosphate were based on the observed rates within the Canadian patient cohort described above [11]. In sensitivity analysis, we considered the mortality rates observed in a national registry of dialysis patients in the US [9] and within DCOR itself [6]. Efficacy of sevelamer The DCOR study is the only randomized study with sufficient statistical power to investigate the impact of sevelamer on clinically relevant endpoints such as mortality [6]. This currently unpublished study reported efficacy in several different ways, making it uncertain how best to model the efficacy of sevelamer. We considered several alternative strategies, described in Table 2, and developed an a priori analysis plan before performing analyses. Since the primary DCOR analysis compared survival without stratifying by time on therapy, our primary analysis used this estimate of efficacy, reflecting a constant relative risk over time of 0.91 [6] for all age groups; [Model 1 (primary model)]. Since the primary analysis of the DCOR study was non-significant, we also performed a costminimization analysis [Model 2 (cost minimization model); Table 2]. Since the assumption of constant proportional hazards was not met in the primary DCOR analysis [6], we performed an exploratory analysis in which the effect of treatment on mortality differed before and after 2 years of follow-up [Model 3 (mortality over time model); Table 2]. As only 50.6% of patients completed the unblinded DCOR study [6], the balance between treatment groups introduced by randomization at study entry might no longer be present among patients remaining in the trial after 2 years. Therefore, the findings of this post hoc analysis are of

4 4of12 Table 2. Four alternate ways of modelling the efficacy of sevelamer B. Manns et al. Analysis Assumption Relative risks modelled for mortality and hospitalization Justification Model 1: primary model Model 2: cost minimization model Model 3: mortality over time model Model 4: mortality by age model RR, relative risk. Primary Constant risk over time RR death 0.91 ( ) RR hospitalization 0.91 Secondary Risk of death/hospitalization RR death 1 equivalent for sevelamer RR hospitalization 1 and calcium-based phosphate Secondary Secondary Mortality varies for Years 1 2, vs Years 3 and on Mortality varies for patients <65 and those 65 < 2 years: RR death 1.08 ( ) RR hospitalization years: RR death 0.66 ( ) RR hospitalization 0.91 Age <65 RR death 1.14 ( ) RR hospitalization 1.17 Age 65 RR death 0.78 ( ) RR hospitalization 0.72 This was the a priori planned primary analysis for the DCOR study [6] The P value was >0.05 for both death (P ¼ 0.30) and hospitalization (P ¼ 0.06) in the primary analysis of DCOR [6] The assumption of constant proportional hazards was not met in the DCOR primary analysis [6] An interaction was noted between treatment efficacy and patient age (<65 years or 65 years) in the DCOR analysis [6] uncertain clinical significance and are presented as a secondary analysis only. Finally, since the DCOR investigators reported an interaction between treatment efficacy and patient age (<65 or 65 years) [6], we also performed an exploratory analysis modelling the efficacy of sevelamer in patients aged < 65 years and 65 years separately [Model 4 (mortality by age model)]. Other clinical effects Kidney transplantation. Transplant rates for patients aged <65 and 65 years were derived from the Canadian patient cohort described before (Table 1) [11]. Survival of transplant patients and the risk of a transplant failure necessitating a return to dialysis were estimated from a contemporary cohort of North American transplant patients [19]. We assumed transplant failure and patient survival to be similar for patients treated with calcium-based phosphate and sevelamer. Health-related quality of life. A comprehensive literature search was done to identify estimates of utility scores for contemporary North American dialysis and transplant patients, favouring estimates from more recent studies which were done in unselected patient populations [20 22]. Base case analyses use the results of a contemporary cohort of ESRD patients [20]. Given that quality of life was not reported in DCOR [6] or any other sevelamer study to date, we assumed similar utility values for patients treated with calcium-based phosphate and sevelamer. Costs. Given the perspective of our analysis (i.e. publicly funded government health care in Canada or Medicare in the United States), it only included direct costs to the health care purchaser [18]. Baseline estimates of cost are reported in Table 3. Costs were inflated to 2004 values [23] and are reported in CAN$ (1US$ ¼ 1.30CAN$) except for the US scenario analyses, where costs are reported in US$. Health care costs considered were classified into one of three categories: (a) Drug costs (cost category 1): Average daily consumption of sevelamer and calcium-based phosphate in the DCOR study have not yet been reported. In the primary analysis, the cost of sevelamer and calcium carbonate was therefore estimated based on the average doses (sevelamer 6.5 g/day; calcium carbonate 4.3 g/day) consumed in another large clinical trial of sevelamer (n ¼ 200) [13]. (b) Hospitalization costs (cost category 2): The DCOR study reported that the number of hospitalizations (per patient year) was non-significantly reduced from to (P ¼ 0.06) for patients treated with sevelamer [6]. To determine the potential reduction in hospital costs associated with sevelamer use, we required information on the average cost and frequency of hospitalization for North American ESRD patients, which was determined in detail using a focused literature search (Tables 1 and 3) [24 26]. (c) Associated health care costs: Given that any therapy which extends life will increase the cost of associated health care (i.e. such as the cost of dialysis and transplant care), we also estimated the cost of ongoing dialysis and transplant therapy (cost category 3), favouring estimates from more recent studies which were done in unselected patient populations (Table 3). Some methodological controversy exists as to, which costs to include in economic evaluations that are performed within ESRD [27]. For the primary analysis, we considered all health care costs (cost categories i iii) [27]. However, since dialysis is expensive, interventions for dialysis patients that improve survival without reducing the need for dialysis, will be associated with a cost-utility ratio at least as great as that of dialysis itself [27]. Inclusion of such costs (which, in and of

5 Economic evaluation of sevelamer 5 of 12 Table 3. Baseline costs Variable Value (CAN$) Reference Plausible range considered (CAN$) Annual cost of sevelamer a $4211 Manufacturer price $ Annual cost of calcium carbonate b $85 Canadian provincial formulary - Average cost per hospitalization for $ [25] $ ESRD patients over 65 years on calcium-based phosphate binder Average cost per hospitalization for $7391 [25] $ ESRD pts under 65 on calcium-based phosphate binder Transplant cost per year [21] $ Year 1 $ $ Years 2 and on $ Cost of ongoing dialysis therapy c $ [25] Discount rate for costs 0.05 [18] a Assuming average daily dose of 6.5 g [13]. b Assuming average daily dose of 4.3 g [13]. c Includes out-patient dialysis cost and physician claims for dialysis and assumes that 78.6% of patients are on HD ($ per year, 95% CI $ ), 20.1% of patients are on PD ($ per year, 95% CI $ ) and 1.3% of patients are on home HD ($28 060, 95% CI $ ) [10]. ESRD, end-stage renal disease. itself, is methodologically correct) in economic evaluations in this area may mitigate against the acceptance of interventions which improve patient survival. In sensitivity analysis, we explored the cost per QALY gained considering only the costs within cost categories (i) and (ii). Sensitivity analysis and scenario analysis Various one-way sensitivity analyses were performed varying the values for uncertain variables within the ranges noted in Tables 1 3. As noted, we performed various analyses modelling the efficacy of sevelamer in different ways (Table 2). In the base case analysis, we assumed that the relative risk of death associated with sevelamer treatment remained constant throughout the patient s lifetime. The impact of this assumption was tested in a sensitivity analysis which assumed no survival benefit beyond 4 years (consistent with the timeline of the DCOR study). The DCOR study protocol identified six subgroups for which subgroup analyses were planned a priori and specified that stratified analyses would be performed in subgroups for which a test for interaction was statistically significant (P ¼ 0.03). Although the tests for interaction were not corrected for multiple comparisons, the nominally significant interaction between treatment and age <65 vs 5 years raises the possibility that sevelamer is more effective in older patients. In scenario analyses, we modelled the costeffectiveness of sevelamer in patients aged 65 years. In exploratory analyses, we also modelled the cost-effectiveness of sevelamer in patients aged 45 years and those aged 55 years, varying baseline mortality and hospitalization rates [11], hospitalization costs [25] and the relative risk of mortality as reported in DCOR [6]. Many of our model inputs (baseline mortality rates, cost and frequency of hospitalization) were based on Canadian data. To improve the generalizability of this analysis for US decision makers, we undertook scenario analyses that better represented a US setting. In addition, we present a scenario analysis using data on baseline mortality and hospitalization frequency taken directly from DCOR. Model inputs used in the US and DCOR-specific scenarios are described in Table 4. Probabilistic sensitivity analyses were performed as described in Appendix 1 (Figure A1-A E). Results Model validity Consistent with published guidelines [28 30], we tested for logical inconsistencies in our decision model by evaluating them under hypothetical conditions. We determined that our models had face validity and confirmed that the mathematical calculations were accurate and consistent with the specifications of the model. We also confirmed that our model had predictive validity by comparing model outputs (a function of both input variables and model structure) with observed data from the DCOR study (data not shown). This comprehensive validation increases confidence in each of these models. Base case analyses In the base case analysis, which used a lifetime horizon, Model 1 (primary model), the use of sevelamer, in comparison to calcium-based phosphate, resulted in an incremental cost of CAN$ and an increase in quality adjusted life years of 0.211, resulting in a cost per QALY gained of CAN$ overall. Since the effect of sevelamer on mortality and hospitalization was not statistically significant compared with calcium-based phosphate in the DCOR study (P ¼ 0.30 and P ¼ 0.06, respectively) [6], we repeated our analysis considering no survival or hospitalization advantage for sevelamer

6 6of12 Table 4. Clinical and costing estimates used for United States- and DCOR-specific scenario analyses B. Manns et al. Variable Base case US scenario DCOR scenario Annual risk of death for ESRD patients <65 years of age on calcium-based phosphate binder Annual risk of death for ESRD patients 65 on calcium-based phosphate binder Annual hospitalization frequency for ESRD patients on calcium-based phosphate binder (based on estimate for ages 50 59, USRDS) [9] (based on estimate for age 70, USRDS) [9] <65: : 1.54 [24] <65: : 2.01 [9] [6] [6] <65: : 2.9 [6] Annual cost of sevelamer CAN$4211 CAN$4737 (US$3644) CAN$4737 (US$3644) [36] [36] Annual cost of calcium carbonate CAN$85 CAN$200 (US $154) CAN$200 (US $154) [36] [36] Annual cost of calcium acetate NA CAN$602 (US $463) CAN$602 (US $463) [36] [36] Average cost per hospitalization for ESRD pts on calcium-based phosphate binder Transplant cost per year Year 1 Years 2 and subsequent <65: CAN$7391 >65: CAN$ [25] CAN$ CAN$ All ages: CAN$ (US $16 324) [37] CAN$ (US$84 885) CAN$ (US$14 814) [38] Cost of on-going dialysis therapy c CAN$ CAN$ (US $22 620) [37] All ages: CAN$ (US$16 324) [37] CAN$ (US$84 885) CAN$ (US$14 814) [38] CAN$ (US $22 620) [37] Many of the model inputs used in the base case (baseline mortality rates, cost and frequency of hospitalization) were based on Canadian data. To improve the generalizability of this analysis for US decision makers, this table outlines details of a scenario analysis that better represents a US setting. The table also shows details of a scenario analysis using data on baseline mortality and hospitalization frequency taken directly from the DCOR study. Costs are based on an US Canadian exchange rate of 1.30 in 2004 (Bank of Canada). (RR 1.0 for both outcomes). In this analysis [Model 2 (cost minimization model); Table 2], the use of sevelamer, in comparison with calcium-based phosphate, resulted in an incremental cost of CAN$17 000, but, as expected, no increase in quality adjusted life years. Since the DCOR study suggested differential benefit to patients based on their age and length of follow-up, we also considered other methods of modelling efficacy. In Model 3 (mortality over time model; Table 2), we modelled the effectiveness of sevelamer differently for the first 2 years, compared with 3 years and onwards, which resulted in an incremental cost per QALY gained for sevelamer compared with calciumbased phosphate overall of CAN$ (Table 5). Lastly, in Model 4 (mortality by age model; Table 2), which modelled the efficacy of patients aged <65 years and 65 years separately; the incremental cost per QALY gained overall was CAN$ (Table 5). Scenario analyses We repeated the analyses for each of the models noted above over a 4 year time horizon; the cost per QALY gained varied between CAN$ and $2.4 million for Models 1, 3 and 4. We also reanalysed Models 1, 3 and 4, excluding the related health care costs associated with dialysis and transplantation and noted that the cost per QALY gained varied from CAN$ to $ (Appendix 2, Table A1). Assuming that sevelamer resulted in a differential reduction in mortality in patients 65 years of age, the use of sevelamer in this subgroup was associated with a cost per QALY gained of CAN$ In exploratory analyses that examined the cost-effectiveness of sevelamer in subgroups of patients aged 45 years and 55 years, the cost per QALY gained was CAN$ and $ , respectively. Excluding the costs of dialysis and transplantation and including only patients 65 using Model 4, the cost per QALY gained becomes $ (Appendix 2, Table A1), though, given the uncertainty associated with subgroup analysis, this likely represents an unrealistically optimistic estimate of the cost-effectiveness of sevelamer. Given that American ESRD patients experience higher mortality rates and incur higher healthcare costs than Canadian ESRD patients, we repeated analyses using probabilities and costs that would be more reflective of the US health care system (Table 4). We also considered calcium acetate as a comparator given its frequency of use, assuming equal

7 Economic evaluation of sevelamer 7 of 12 Table 5. Base case cost effectiveness analysis, considering a lifetime time horizon (in CAN$) Strategy Cost Marginal cost Effectiveness (QALYs) Marginal effectiveness (QALYs) Incremental cost per QALY gained Model 1 (Primary) Model 2 (Cost minimization) Model 3 (Mortality over time) Model 4 (Mortality by age) Calcium-based phosphate $ Sevelamer $ $ $ Calcium-based phosphate $ Sevelamer $ $ Sevelamer dominated Calcium-based phosphate $ Sevelamer $ $ $ Calcium-based phosphate $ Sevelamer $ $ $ efficacy between calcium carbonate and calcium acetate. In the US scenario, the use of sevelamer, in comparison with calcium carbonate and calcium acetate resulted in an incremental cost per QALY gained of US$ and US$ , respectively. In the DCOR-specific scenario, the use of sevelamer, in comparison with calcium carbonate and calcium acetate, respectively, resulted in an incremental cost per QALY gained of US$ and US$ Sensitivity analysis Our results were robust to clinically plausible changes in all uncertain variables (Table 6). Excluding the impact of quality of life (but using baseline mortality rates from our Canadian cohort), the use of sevelamer, compared with calcium-based phosphate, resulted in a cost per life year gain of CAN$ If the average dose per day of sevelamer was reduced or increased by 25%, the cost per QALY gained for sevelamer was CAN$ and CAN$ , respectively. The results of the probabilistic sensitivity analysis demonstrated significant uncertainty in the true cost-effectiveness of sevelamer (Appendix 1, Figure A1-A E). As the DCOR study found no significant difference in mortality, but a trend towards a reduction in hospitalization rates (which would generate cost savings), we also sought to determine to what extent sevelamer use would have to lower hospitalization rates in order for the savings to offset the additional medication costs of sevelamer. Assuming no difference in survival, the risk of hospitalization would have to be lowered by 30% to offset the additional medication cost of sevelamer. Discussion The primary analysis of the only randomized study powered to investigate the impact of sevelamer on relevant clinical endpoints showed no statistically significant difference in mortality. Secondary analyses of this DCOR study, based on patient age and more speculatively, time on treatment, suggested a possible benefit in certain subgroups [6]. The most straightforward conclusion would be to adopt the results of the primary analysis, conclude that sevelamer does not improve clinical outcomes and perform a costminimization analysis. However, to avoid prematurely discarding the possibility that sevelamer represents an economically efficient treatment and consistent with guidelines for economic analyses [18,28], we considered several different modelling strategies. Considering only the models that assume sevelamer to be more effective than calcium-based phosphate, the use of sevelamer was associated with a cost per QALY gained ranging from CAN$ to $ This is higher than the cost per QALY gained for most therapies that are considered an efficient use of finite health care resources [31]. In part, the high cost per QALY gained for sevelamer is due to the observation that therapies which prolong life in ESRD patients result in a significant increase in related health care costs for dialysis [27]. When these costs were excluded (which would be inconsistent with accepted practice within economic evaluation), the cost per QALY gained for sevelamer decreased but still ranged between CAN$ and $ This observation does raise potential ethical issues that will need to be considered when interpreting economic evaluations of new therapies that improve survival for patients with ESRD. It is important to note that other factors besides effectiveness and cost-effectiveness influence funding decisions. For instance, dialysis is funded in all developed countries, despite the fact that it is associated with high cost per QALY gained [27]. In part, this is due to the fact that without dialysis, patients face certain death; this has been termed the rule of rescue, whereby societies tend to be willing to direct resources to therapies that avert certain death [32]. It should be noted that this is not the case with therapies like sevelamer, which, at best, may reduce the

8 8of12 B. Manns et al. Table 6. Sensitivity analysis of the cost per quality adjusted life year gained for sevelamer compared with calcium-based phosphate in the treatment of dialysis patients (in CAN$) Variables considered within sensitivity analysis Incremental cost per QALY gained Primary model (Baseline analysis) $ Assuming that sevelamer is not effective after 4 years (i.e. RR death ¼ one after 4 years) $ RR death with sevelamer: baseline 0.91 (varied within 95% CI) RR 0.77 $ RR 1.08 RR hospitalization with sevelamer: baseline 0.91 (95% CI not provided; therefore varied from 25% to RR of 1.0) RR $ RR 1.00 $ Annual risk of death of patients: baseline (age 65), (age < 65) (Varied within 95% CI) [11] Reduced risk of death ( 65), (< 65) Increased risk of death (65), (< 65) $ $ Annual transplant risk for patients: baseline (age 65), (age <65) (Varied within 95% CI) [11] Risk of transplant (65), (<65) Risk of transplant (65), (<65) $ $ Annual frequency of hospitalization for ESRD patients: baseline 1.54 Highest published risk of hospitalization per year: 1.98 [26] $ Lowest published risk of hospitalization per year: 1.54 [24] $ Utility estimates for dialysis patients: baseline (age 65 years), (age <65 years) (Varied within 95% CI) [20] Lower bound of 95% CI (65), (<65) $ Upper bound of 95% CI (65), (<65) $ Cost per life year gained analysis (assuming a utility of 1.00 for all patient groups) $ Annual cost of sevelamer: baseline $4211 Alternate Canadian formulary list price ($4569) $ Cost reduced by 25% $ Cost reduced by 50% $ Average daily dose of sevelamer: baseline 6.5 g (varied 25%) 4.88 g $ g $ Average cost per hospitalization for ESRD pts on calcium-based phosphate binder: baseline $ (age 65 years), $7391 (age <65) (varied within 95% CI) [25] Lower bound 95% CI $6336 (65), $5551 (<65) $ Upper bound 95% CI $ (65), $9231 (<65) $ Discount rate for costs and effects No discounting $ Costs 3%, effects 3% $ Costs 6%, effects 3% $ With RR >1, use of sevelamer leads to excess deaths, thus reducing the total follow up costs, making it less expensive and less effective. RR, relative risk. statistical likelihood of dying. Nonetheless, these important ethical issues deserve careful consideration by decision-makers who fund ESRD care. An interaction was noted in the DCOR study between treatment efficacy and patient age (<65 years or 65 years), with patients older than 65 years experiencing a statistically significant reduction in mortality. While the use of sevelamer in this subgroup was more attractive, particularly when the costs of dialysis and transplant were excluded, the effectiveness of sevelamer in this subgroup remains uncertain and the cost per QALY gained for this subgroup (CAN$ ) is still high and was similar when slightly different subgroups of older patients (55 or 45 years) were included. In addition, since this test for interaction was non-significant after correcting for multiple comparisons [33], it could be argued that the benefit of sevelamer in older populations requires confirmation in specifically designed trials. This is the first study to model the cost-effectiveness of sevelamer using clinical rather than surrogate endpoints. In the only other full economic evaluation of sevelamer performed in ESRD patients, Huybrechts et al. [34] used efficacy data taken from the Treat to Goal study [13] which suggested that sevelamer reduced coronary artery calcification, an unproven surrogate endpoint for mortality in ESRD patients. Based on the ability of sevelamer to reduce coronary artery calcification and assuming that reduction in such calcification would reduce subsequent mortality, cardiovascular disease and hospital costs, these authors suggested that the use of sevelamer was cost-effective.

9 Economic evaluation of sevelamer 9 of 12 However, reducing vascular calcification has not been shown to reduce mortality, and therefore, it is uncertain whether the findings of Huybrechts et al. [34] are valid. Like all economic models, our analysis depends on certain assumptions and has some limitations. First, uncertainty remains as to the most appropriate method of modelling the effectiveness of sevelamer in unselected ESRD patients. However, it is reassuring that the results of the models were generally consistent across a spectrum of sensitivity analyses. Second, the cost-effectiveness of sevelamer may be improved if its use significantly reduces the cost of hospitalization. While we modelled a reduction in the frequency of hospitalization in our analyses, the actual reduction in hospitalization costs remains speculative until further DCOR analyses based on Medicare data become available. It should be noted that a 30% reduction in the risk of hospitalization would be required to offset the increased cost of sevelamer. Third, due to the lack of data on the impact of sevelamer on patient-related costs (i.e. time off work to receive medical care, productivity losses), analyses using a full societal perspective were not presented. However, even if indirect costs are shown to be reduced by sevelamer, the magnitude of these costs are likely to be small and therefore it is unlikely that adopting a broader perspective would have qualitatively changed the results of this analysis. Finally, our analyses considered North American cohorts of dialysis patients. We did not estimate the cost-effectiveness of sevelamer in a European context as we could not obtain accurate information on important input parameters including the incidence and costs of hospitalization in representative cohorts of European patients. However, given that mortality rates in European and Canadian ESRD patients [11,35] appear similar and that the cost differential between sevelamer and calcium are similar in Europe, it is likely that the cost-effectiveness of sevelamer in Europe is similar to that observed in Canada. We did not consider scenarios where sevelamer was restricted to ESRD patients with specific abnormalities in mineral metabolism, such as concomitant hyperphosphataemia and hypercalcaemia, a criteria used by some health care purchasers [5]. There is no data indicating that such a strategy will reduce mortality or morbidity in these patients and consequently the costeffectiveness of this approach is unknown. However, since calcium-based phosphate reduce serum phosphate levels to the same extent as sevelamer, it seems unlikely that sevelamer will be more costeffective in this patient subgroup. While sensitivity and scenario analyses were relatively robust to plausible variations in model parameters, probabilistic sensitivity analysis (Appendix 1, Figure A1-A E) demonstrated significant uncertainty in the true cost-effectiveness of sevelamer. While health care funding decisions must be made using the best currently available data, it is possible that future estimates of the cost-effectiveness of sevelamer may differ, but only if several key parameters are found in future studies to differ substantially from those used in this analysis. Conclusion Even assuming that sevelamer is associated with a clinical benefit, as we did within our primary economic evaluation, we noted that the cost per QALY gained for treating all ESRD patients with sevelamer, compared with calcium-based phosphate, exceeds what would usually be considered good value for the money. While the high cost per QALY was in part due to inclusion of the costs of dialysis and transplantation in the analysis, the cost per QALY gained remained relatively unattractive even when these costs were excluded. Although lower, the cost per QALY gained for treating patients older than 65 years is still relatively high given the uncertainty of clinical effectiveness in this subgroup, suggesting that additional trials conducted specifically in this population may be useful to decision-makers and health care payers. Acknowledgements. This study was funded by the Canadian Agency for Drugs and Technologies in Health. Conflict of interest statement. None declared. References 1. National Kidney Foundation.K/DOQI clinical practice guidelines for bone metabolism and disease in chronic kidney disease. Am J Kidney Dis 2003; 42: S1 S Block G, Port FK. Calcium phosphate metabolism and cardiovascular disease in patients with chronic kidney disease. Semin Dial 2003; 16: Block GA, Hulbert-Shearon TE, Levin NW, Port FK. Association of serum phosphorus and calcium x phosphate product with mortality risk in chronic hemodialysis patients: a national study. Am J Kidney Dis 1998; 31: Block GA, Port FK. Re-evaluation of risks associated with hyperphosphatemia and hyperparathyroidism in dialysis patients: recommendations for a change in management. Am J Kidney Dis 2000; 35: Manns B, Stevens L, Miskulin D, Owen WF,Jr, Winkelmayer WC, Tonelli M. A systematic review of sevelamer in ESRD and an analysis of its potential economic impact in Canada and the United States. Kidney Int 2004; 66: Suki W, Zabaneh R, Cangiano J et al. The DCOR Trial: A prospective, randomized trial assessing the impact on outcomes of sevelamer in dialysis patients. American Society of Nephrology, Pennsylvania, Philadelphia: Gold R, Siegel J, Russell L, Weinstein M. Cost-effectiveness in Health and Medicine. Oxford University Press, New York: O Brien B. Economic evaluation of pharmaceuticals. Frankenstein s monster or vampire of trials? Med Care 1996; 34: DS99 DS U.S. Renal Database System. Morbidity and Mortality Report 2005, In USRDS ed. Minneapolis: 2005; Canadian Organ Replacement Registry. Treatment of End-Stage Organ Failure in Canada 2002 and Canadian Institutes of Health Information, Ottawa: 2005

10 10 of 12 B. Manns et al. 11. Tonelli M, Klarenbach S, Manns B et al. Residence location and likelihood of kidney transplantation. CMAJ 2006; 175: Block GA, Spiegel DM, Ehrlich J et al. Effects of sevelamer and calcium on coronary artery calcification in patients new to hemodialysis. Kidney Int 2005; 68: Chertow GM, Burke SK, Raggi P. Treat to Goal Working G. Sevelamer attenuates the progression of coronary and aortic calcification in hemodialysis patients. Kidney Int 2002; 62: Iwasaki Y, Takami H, Tani M et al. Efficacy of combined sevelamer and calcium carbonate therapy for hyperphosphatemia in Japanese hemodialysis patients. Therapeutic Apheresis & Dialysis 2005; 9: Kinugasa E, Koshikawa S. The PBSG. Effects of PB-94 (sevelamer hydrochloride), a phosphate binder, on the treatment of hyperphosphatemia in hemodialysis patients - A randomized, open label, dose titration study of PB-94 versus Caltan tablet 500 (calcium carbonate). J Am Soc Nephrol 2001; 12: A Manns BJ, Taub K, Richardson R, Donaldson C. To reuse or not to reuse? An economic evaluation of hemodialyser reuse vs conventional single use hemodialysis for chronic hemodialysis patients. Int J Technol Assess Health Care 2002; 18: Tonelli M, Winkelmayer WC, Jindal KK, Owen WF, Manns BJ. The cost-effectiveness of maintaining higher hemoglobin targets with erythropoietin in hemodialysis patients. Kidney Int 2003; 64: Canadian Agency for Drugs and Technologies in Health. Guidelines for the Economic Evaluation of Health Technologies, 3rd edn, Canada Ottawa: Wolfe RA, Ashby VB, Milford EL et al. Comparison of mortality in all patients on dialysis, patients on dialysis awaiting transplantation, and recipients of a first cadaveric transplant. N Engl J Med 1999; 341: Manns B, Johnson JA, Taub K, Mortis G, Ghali WA, Donaldson C. Quality of life in patients treated with hemodialysis or peritoneal dialysis: what are the important determinants? Clin Nephrol 2003; 60: 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: Heidenheim AP, Muirhead N, Moist L, Lindsay RM. Patient quality of life on quotidian hemodialysis. Am J Kidney Dis 2003; 42: Consumer Price Index for Canada [health care (not seasonally adjusted) ] Ottawa; Cat no Murphy SW, Foley RN, Barrett BJ et al. Comparative hospitalization of hemodialysis and peritoneal dialysis patients in Canada. Kidney Int 2000; 57: 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: Goeree R, Manalich J, Grootendorst P, Beecroft ML, Churchill DN. Cost analysis of dialysis treatments for end-stage renal disease (ESRD). Clin Invest Med 1995; 18: Manns B, Meltzer D, Taub K, Donaldson C. Illustrating the impact of including future costs in economic evaluations: an application to end-stage renal disease care. Health Econ 2003; 12: Weinstein MC, O Brien B, Hornberger J et al. Principles of good practice for decision analytic modeling in health-care evaluation: report of the ISPOR Task Force on Good Research Practices- Modeling Studies. Value Health 2003; 6: Sculpher M, Fenwick E, Claxton K. Assessing quality in decision analytic cost-effectiveness models. A suggested framework and example of application. Pharmacoeconomics 2000; 17: McCabe C, Dixon S. Testing the validity of cost-effectiveness models. Pharmacoeconomics 2000; 17: Laupacis A, Feeny D, Detsky AS, Tugwell PX. How attractive does a new technology have to be to warrant adoption and utilization? Tentative guidelines for using clinical and economic evaluations. CMAJ 1992; 146: Hadorn DC. Setting health care priorities in Oregon. Costeffectiveness meets the rule of rescue. JAMA 1991; 265: Lagakos S. The challenge of subgroup analyses reporting without distorting. N Engl J Med 2006; 354: Huybrechts KF, Caro JJ, Wilson DA, O Brien JA. Health and economic consequences of sevelamer use for hyperphosphatemia in patients on hemodialysis. Value Health 2005; 8: ERA-EDTA Registry. ERA-EDTA Registry 2004 Annual Report. Academic Medical Center, Department of Medical Informatics, The Netherlands: Medical Economics Company Incorporated Drug Topics RedBook, Montvail, New Jersey, USA: U.S. Renal Database System. USRDS 2005 Annual Report, Costs of CKD and ESRD. In USRDS ed, Minneapolis: 2005; Whiting JF, Martin JE, Cohen D et al. Economic outcome of simultaneous pancreas kidney transplantation compared with kidney transplantation alone. Transplant Proc 2001; 33: Briggs A. Uncertainty in the economic evaluation of health care technologies: the role of sensitivity analysis. Health Econ 1994; 3: Briggs A, Fenn P. Confidence Intervals or Surfaces? Uncertainty on the Cost-Effectiveness Plane. Health Econ 1998; 7: Fenwick E, Claxton K, Sculpher M. Representing uncertainty: the role of cost-effectiveness acceptability curves. Health Econ 2001; 10: Claxton K. The irrelevance of inference: a decision-making approach to the stochastic evaluation of health care technologies. J Health Econ 1999; 18: Critchfield GC, Willard KE. Probabilistic analysis of decision trees using Monte Carlo simulation. Med Decis Making 1986; 6: Doubilet P, Begg CB, Weinstein MC, Braun P, McNeil BJ. Probabilistic sensitivity analysis using Monte Carlo simulation. A practical approach. Med Decis Making 1985; 5: Received for publication: Accepted in revised form: Appendix 1 Probabilistic sensitivity analysis Methods. The standard method of expressing uncertainty in classical statistics is through use of measures of variance, such as the 95% confidence interval (CI) around a mean for normally distributed variables. This approach is problematic in economic evaluation and has led to the use of other methods designed to deal with and express analytical uncertainty [39,40]. Classical univariate sensitivity analysis has been criticized since it only considers the uncertainty present in one (or two) variables at a time and it may underestimate the uncertainty present within a cost-effectiveness ratio [39,40]. As a result, some analysts have attempted to construct 95% CIs around cost-effectiveness ratios [40]. Given the lack of independence between the incremental costs (the numerator) and incremental QALYs gained (the denominator), this is

11 Economic evaluation of sevelamer 11 of 12 A B C D E Fig. A1. Incremental cost effectiveness ratio (ICER) scatter plots (in CAN$). The ICER scatter plots plot points for all iterations in the Monte Carlo simulation. Each point represents the incremental cost and effectiveness of sevelamer relative to calcium-based phosphate for a particular iteration. The willingness-to-pay value (arbitrarily set as $ per QALY gained in these examples) is displayed as the slope of a dashed line intersecting the origin of the plot. A 95% confidence ellipse is also shown in all the graphs. (A) Model 1 (Primary model). (B) Model 2 (Cost minimization model). (C) Model 3 (Mortality over time model). (D) Model 4 (Mortality by age model). (E) 65 subgroup using Model 4. statistically problematic and as such, most recommend against the use of CIs for cost-effectiveness ratios and for the use of Monte Carlo simulation (probabilistic sensitivity analysis) as performed in this analysis [41,42]. Monte Carlo simulation enables simultaneous sensitivity analysis of all uncertain variables. It does so by replacing estimates of probabilities, utilities and costs with specific probability distributions, which are based on the reported means and variances of each variable [41,43,44]. The analysis

12 12 of 12 B. Manns et al. Table A1. Base case cost-effectiveness analysis, excluding the cost of dialysis and transplantation Strategy Cost Marginal cost Effectiveness (QALYs) Marginal effectiveness (QALYs) Incremental cost per QALY gained Primary model (includes cost of dialysis and transplant) Primary model a Model 3 a Model 4 a Model 4: Calcium-based phosphate $ Sevelamer $ $ $ Calcium-based phosphate $ Sevelamer $ $ $ Calcium-based phosphate $ Sevelamer $ $ $ Calcium-based phosphate $ Sevelamer $ $ $ Calcium-based phosphate $ Age 65 years subgroup a Sevelamer $ $ $ a Excluding the cost of dialysis and transplantation. is then repeated times, sampling different values from the appropriate distributions for each of the variables. In such a way, a statistical distribution is built up around the incremental cost-effectiveness ratio giving a better reflection of the uncertainty inherent in the analysis. With Monte Carlo simulation, one can also consider the uncertainty with respect to the maximum cost per QALY gained that decision-makers would consider acceptable (i.e. some decision-makers will choose only to fund therapies associated with cost per QALY gained < $ [31]; others, with larger budgets, may choose to fund therapies with cost per QALY gained up to $ ). This is displayed graphically as a cost-effectiveness acceptability curve demonstrating the probability that a therapy is associated with a cost per QALY gained lower than a range of displayed maximum cost-effectiveness ratios. We performed Monte Carlo simulation for our overall model and patients aged 65 years. Statistical distributions were created around all of the variables for which significant measurement uncertainty existed and for which distributions could be estimated. In general, normal distributions were used for probabilities and utilities, as appropriate and log normal distributions were used for costs. Results With the use of second-order Monte Carlo simulation, the Appendix Figures A1-A E present incremental cost-effectiveness ratio (ICER) scatter plots highlighting the uncertainty in the ICER when the uncertainty in all variables is considered simultaneously. In all models and subgroups considered, there is significant uncertainty in the 95% confidence ellipse, with a significant proportion of simulations falling into each quadrant of the cost-effectiveness plane. The probability that the use of sevelamer would be cost-effective (i.e. more effective and either cost saving or associated with a cost per QALY gained below the threshold) if a decision-maker was willing to pay only CAN$ or $ per QALY gained and considering Model 1 is 15 or 32%, respectively. Considering the subgroup of patients 65 years separately, the probability that the use of sevelamer would be cost-effective if a decision-maker was willing to pay only CAN$ or $ per QALY gained is 21 or 44%.