Cost-effectiveness analysis of anidulafungin versus fluconazole for the treatment of invasive candidiasis

Transcription

1 J Antimicrob Chemother 2011; 66: doi: /jac/dkr186 Advance Access publication 30 May 2011 Cost-effectiveness analysis of anidulafungin versus fluconazole for the treatment of invasive candidiasis Chin Fen Neoh 1, Danny Liew 2, Monica Slavin 3, Debbie Marriott 4, Sharon C.-A. Chen 5, Orla Morrissey 6, Kay Stewart 1 and David C. M. Kong 1 * 1 Department of Pharmacy Practice, Centre for Medicine Use and Safety, Faculty of Pharmacy and Pharmaceutical Sciences, Monash University, 381 Royal Parade, Parkville, Victoria 3052, Australia; 2 Department of Medicine, University of Melbourne, Royal Melbourne Hospital, Grattan Street, Parkville, Victoria 3050, Australia; 3 Department of Infectious Diseases, Peter MacCallum Cancer Centre, St Andrews Place, East Melbourne, Victoria 3002, Australia; 4 Department of Clinical Microbiology and Infectious Diseases, St Vincent s Hospital, 390 Victoria Street, Darlinghurst, New South Wales 2010, Australia; 5 Centre for Infectious Diseases and Microbiology, Westmead Hospital, PO Box 533, Wentworthville, New South Wales 2145, Australia; 6 Infectious Diseases Unit, Department of Medicine, Alfred Health and Monash University, Commercial Road, Melbourne, Victoria 3004, Australia *Corresponding author. Tel: ; Fax: ; david.kong@monash.edu Received 22 January 2011; returned 2 March 2011; revised 17 April 2011; accepted 17 April 2011 Background: Anidulafungin was found to be non-inferior to and possibly more efficacious than fluconazole for treatment of invasive candidiasis (IC) in a major randomized clinical trial (RCT). There are no data comparing the cost-effectiveness between azoles and echinocandins in treating IC. This economic analysis investigated the cost-effectiveness of anidulafungin compared with fluconazole for treatment of IC in an Australian setting. Methods: A decision analytic model was constructed to capture downstream consequences of using either agent for treatment of IC. The main outcomes analysed in the model were treatment success and treatment failure (observed and indeterminate). Outcome probabilities and treatment pathways were derived from a published RCT. Resources used were estimated by an expert panel and cost inputs were derived from the latest Australian resources. The analysis was based on an Australian hospital perspective. Sensitivity analyses were conducted using Monte Carlo simulation. Results: Anidulafungin (AU$74587) had a higher total cost than fluconazole (AU$60945) per successfully treated patient, primarily due to its higher acquisition cost. Hospitalization was the main cost driver for both comparators. However, when the rates of mortality in both treatment arms were considered, treatment with anidulafungin was expected to save an additional 0.53 life-years, with an incremental cost-effectiveness ratio (ICER) of AU$25740 per life-years saved, which was below the implicit ICER threshold value for Australia. The results were robust over a wide range of variables. Conclusions: This is the first economic evaluation of anidulafungin versus fluconazole in the treatment of IC in Australia. Anidulafungin appears to be a cost-effective option. Keywords: antifungals, Candida spp., modelling Introduction Despite recent advances in the antifungal armamentarium, candidaemia and other forms of invasive candidiasis (IC) remain potentially fatal infections that result in significant morbidity, particularly in immunocompromised and critically ill patients. 1,2 Candidaemia is associated with prolonged hospital stay, excess hospital costs and attributable mortality rates of up to 30% 50%. 2,3 While guidelines 1,2,4 recommend the use of either fluconazole or an echinocandin (i.e. caspofungin, anidulafungin or micafungin) as initial therapy in treating nonneutropenic patients, a recent meta-analysis indicated that echinocandins are associated with higher success rates. 5 To date, there has been only one double-blind, randomized clinical trial (RCT), by Reboli et al., 6 comparing the use of an echinocandin and an azole (anidulafungin versus fluconazole) in the treatment of IC. Anidulafungin was reported to be at least noninferior to, and possibly more efficacious than, fluconazole. 6 However, the study did not consider the cost of anidulafungin, which is considerably higher than that of fluconazole. There # The Author Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please journals.permissions@oup.com 1906

2 Cost-effectiveness study in invasive candidiasis JAC are no economic data comparing anidulafungin and fluconazole as definitive therapy in an Australian setting. The aim of this study was to determine the cost-effectiveness of anidulafungin versus fluconazole in the treatment of IC. Materials and methods The modelling was based on a double-blinded RCT by Reboli et al. 6 In the trial, 261 patients with IC were randomly assigned to receive anidulafungin (200 mg on day 1 and 100 mg daily thereafter) or fluconazole (800 mg on day 1 and then 400 mg daily). The primary outcome was determined at the end of intravenous (iv) antifungal therapy. Perspective A cost-effectiveness analysis was performed from the Australian hospital perspective. Accordingly, only direct medical costs for treating IC, such as costs of initial and alternative antifungal therapy, screening tests to diagnose IC (i.e. chest X-ray, abdominal CT scan, echocardiography, fundoscopy, non-blood and blood culture), monitoring tests to monitor adverse events of the antifungal (i.e. full blood count, renal function test, liver function test and electrolyte test), hospitalization and critical care were included. Model structure A decision analytic model (Figure 1) was constructed based on the treatment pathways and clinical outcomes described by Reboli et al. 6 The decision tree was designed to capture the downstream economic consequences of using either anidulafungin or fluconazole as primary therapy for IC. The model captured five possible outcomes depending on whether or not a patient had failure of the initial treatment, and if so, for what reason. The outcomes were: (i) treatment success; (ii) clinical success with ; (iii) clinical failure despite microbiological success; (iv) both clinical and ; and (v) indeterminate failure. Treatment was considered successful if both clinical and microbiological success, as defined in the Reboli et al. 6 trial, were achieved. Indeterminate failure comprised patients who withdrew from the study, death not due to candidaemia nor IC and receipt of less than three doses of study medication. 6 Patients who failed initial therapy for any reason other than death, clinical failure despite microbiological success or indeterminate failure were switched to another licensed antifungal therapy. The rationale was that these treatment failures were more likely to be associated with patients underlying condition rather than the efficacy of antifungal therapies. The type of alternative therapy depended on the reason for which the initial therapy was discontinued (i.e. types of fungal species attributed to observed failure and site of last positive fungal culture). All patients were followed until success of therapy or death. Model inputs The model was populated with data derived primarily from the study by Reboli et al. 6 These included outcome probabilities, mortality rates, dose and duration of initial antifungal therapies, percentage of patients who switched to oral fluconazole, duration on oral fluconazole, the reasons for failing the treatment, and mean time to failure or death for both study arms. The outcome probabilities from the clinical study are shown in Table 1. The median durations for anidulafungin and fluconazole therapy were 14.0 and 11.0 days, respectively. The mean times to failure were 12.2 and 9.3 days for anidulafungin and fluconazole, respectively. The median times to death for anidulafungin and fluconazole were 21.0 days and 14.0 days, respectively. For data inputs that were not available from the Reboli et al. 6 study, data from the Australian Candidaemia Study by Chen et al. 7 were used. These included duration of hospitalization (56.7 days) and the percentage of patients treated in the intensive care unit (ICU; 25.0%). The length of ICU stay was assumed to be 11.0 days based on other published economic studies in the candidaemia setting. 8,9 Additional data provided by an expert panel An expert panel was utilized to provide local data not available from the published literature. The expert panel comprised four infectious diseases clinicians (M. S., D. M., S. C.-A. C. and O. M.) who have extensive experience in systemic antifungal therapy use in haematology, oncology, intensive care and organ transplant settings, and are key opinion leaders on antifungal therapies in Australia. A focus group was employed to gather the required information from the expert panel. Upon completion of the focus group, the consensus opinions were then collated and returned Anidulafungin or fluconazole in IC Treatment success Treatment failure Clinical success and Clinical failure and microbiological success Clinical failure and Indeterminate failure Treatment endpoint (end of follow-up) Treatment discontinuation due to failure (switch to alternative treatment) Figure 1. Decision analytic model for treatment of IC (anidulafungin versus fluconazole). 1907

3 Neoh et al. to the experts for re-verification. The panel determined if data from Reboli et al. 6 were generalizable to the Australian setting, as well as validating the decision tree model. The panel also provided data (from the Australian perspective) related to tests for screening and diagnosis of fungal infections, tests for monitoring the response to treatment or the side effects of the antifungal prescribed and the alternative antifungal therapies used (Table 2) if initial treatment failed. Table 1. Outcome probabilities for anidulafungin and fluconazole in treatment of IC Study clinical outcome Probability of anidulafungin a, %(n¼127) Probability of fluconazole a, %(n¼118) Treatment success (n¼96) (n¼71) Treatment failure (n¼31) (n¼47) clinical success and 3.23 (n¼1) 4.26 (n¼2) clinical failure and (n¼4) (n¼6) microbiological success clinical failure and (n¼8) (n¼16) indeterminate failure (n¼18) (n¼23) a Outcome probabilities were obtained from Reboli et al. 6 The expert panel indicated that all patients underwent chest X-ray at the onset of antifungal therapy; 30% of patients received an abdominal CT scan and an ophthalmic examination at baseline, with 10% undergoing follow-up for both procedures; 5% of patients underwent echocardiography 1 2 weeks after the initiation of therapy; all patients had blood fungal culture performed at baseline and 70% had a subsequent culture; 80% of ICU patients had a non-blood fungal culture (urine, sputum, abscess fluid and catheter tips); 80% of ICU patients had central venous catheters replaced; 10 and all patients underwent liver function tests, renal function tests, full blood count and electrolyte tests 2 3 times per week (daily for haematology and ICU patients). The mean duration for alternative treatment was estimated to be 14.0 days from the last positive culture, except for abdominal infections (28.0 days) and endocarditis and endophthalmitis (6 weeks). The expert panel estimated that 10% of the patients receiving initial fluconazole therapy had a creatinine clearance of 50 ml/min and required a 50% decrease in dose. In the event of side effects, the panel advised that the initial antifungal treatment would be discontinued and adverse events would be monitored. Other model assumptions This model also assumed that: (i) patients did not pay for treatment and all costs were covered by Medicare (the Australian public health insurance scheme); (ii) patients were inpatients throughout the study period; (iii) antifungal therapy failed only once (if patients switched therapy after failing the initial therapy, the alternative therapy would be successful); and (iv) Candida albicans responsible for treatment failures was susceptible to fluconazole. Table 2. Alternative drug therapies for treatment failure with anidulafungin and fluconazole Types of initial antifungal therapy Types of treatment failure Types of fungal species Site of last positive culture found Alternative(s) Dosing regimen Anidulafungin Fluconazole clinical success and clinical failure and clinical success and clinical failure and Candida glabrata blood LAmB 3 mg/kg/day Candida albicans blood iv fluconazole or LAmB or CAmB standard a or 3 mg/kg/ day or 0.6 mg/kg/day Candida tropicalis blood iv fluconazole or LAmB or CAmB standard a or 3 mg/kg/ day or 0.6 mg/kg/day C. glabrata blood LAmB 3 mg/kg/day Candida parapsilosis blood iv fluconazole standard a unspecified fungal superinfection blood LAmB or CAmB 3 mg/kg/day or 0.6 mg/kg/day C. parapsilosis blood iv voriconazole or LAmB standard b or 3 mg/kg/day unspecified fungal superinfection blood caspofungin c standard d C. albicans blood caspofungin c standard d C. albicans peritoneum caspofungin c standard d C. glabrata blood caspofungin c standard d C. parapsilosis blood iv voriconazole or standard b or LAmB 3 mg/kg/day a 800 mg daily for 1 day (loading dose) and 400 mg daily (maintenance dose). b 6 mg/kg twice daily for 1 day (loading dose) and 4 mg/kg twice daily (maintenance dose). c In Australia, caspofungin is the only alternative echinocandin to anidulafungin. d 70 mg daily for 1 day (loading dose) and 50 mg daily (maintenance dose). 1908

4 Cost-effectiveness study in invasive candidiasis JAC Cost calculations The cost per successfully treated patient was calculated as the cost of a full course of iv anidulafungin or fluconazole therapy+oral fluconazole, the resources consumed (i.e. screening and monitoring tests) according to the number of days of therapy before successfully completing the initial therapy, and the hospitalization costs based on length of hospital stay. The cost of each treatment failure was calculated by adding the cost of initial antifungal therapy and the cost of subsequent alternative antifungal therapy. For antifungal agents that were dosed by body weight, an average of kg, as reported by the Australian Bureau of Statistics, was used. 11 All costs were in Australian dollars (AU$) for the financial year The costs of resources used are listed in Table 3. Hospitalization costs for the most common co-existing condition in both arms (i.e. septicaemia) were derived from the Australian Refined Diagnosis Related Group (AR-DRG) and were inflated according to the Australian Consumer Price Index The estimated ICU cost per day was drawn from a published study for an Australian tertiary hospital. 14 To avoid double counting, the costs of pharmacy, pathology, imaging and intensive care management were excluded from hospitalization costs. The costs of other resources were based on the Medicare Benefits Schedule Book All medication costs except for anidulafungin were obtained from the Health Purchasing Victoria (HPV) tender , 16 which lists the drug wholesale prices paid by public hospitals in Victoria. The drug acquisition cost of anidulafungin was obtained from Pfizer because it was not listed in the HPV at the time of the analysis. 12,14 16 Table 3. Costs of resources consumed Item Unit Cost (AU$) Caspofungin 70 mg/vial mg/vial Anidulafungin 100 mg/vial Fluconazole 200 mg/100 ml bag mg/cap 2.34 LAmB 50 mg/vial CAmB 50 mg/vial Voriconazole 200 mg/vial mg/tablet Posaconazole 105 ml/vial Chest X-ray 1 test CT scan (upper abdomen) 1 test Echocardiography 1 test Fundoscopy 1 test Non-blood culture 1 test Blood culture 1 test Central vein catheterization Full blood count 1 test Renal function test a 1 test Liver function test 1 test Electrolytes test b 1 test Hospitalization inpatient per day ICU per day a Consists of creatinine and urea levels in blood or urine sample. b Consists of sodium, potassium, magnesium, calcium, chloride and phosphate levels. Incremental cost-effectiveness ratio (ICER), life expectancy and life-years (LYs) saved In cost-effectiveness analysis, the ICER is used in making an informed decision about interventions that are more costly and more effective than the comparators. 17 In the current model, the incremental cost per an additional treatment success and LYs saved were determined. To determine the LYs saved, the difference in mortality rate between the anidulafungin and fluconazole arms at the 6 week follow-up 6 was multiplied by the adjusted mean life expectancy. 18,19 Given that the long-term survival of IC patients was not provided in the trial, the life expectancy of the most common co-existing condition (i.e. septicaemia) found in both study arms was estimated using the published methods. 20,21 As the life expectancy stated in the Australian life tables was age and gender specific, mean age (+range) of the patients from the anidulafungin and fluconazole arms (i.e years and years, respectively) and the proportion of males and females (i.e. 51.0% and 49.0%, respectively) from the Reboli et al. 6 study were incorporated to obtain the mean estimated life expectancy of 26.1 years (range years). The mean estimated life expectancy was then adjusted using the reported relative risk of death for sepsis survivors (0.51). 23 The adjusted mean life expectancy for a sepsis survivor was estimated at 13.3 years. The LYs saved was discounted at a rate of 5% annually. 17 Sensitivity analyses The robustness of the model was determined by sensitivity analyses investigating the impact of sequentially modifying the value of key variables. These included changes in drug acquisition costs, duration of initial antifungal treatment, cost and duration of hospitalization and ICU stays and the mean duration to failure or death. The uncertainty in adjusted life expectancy and annual discount rate were also considered. Scenario analyses included evaluation of the impact of the expert panel s estimation, variation in the percentage of patients who required fluconazole dosage adjustment due to decreased creatinine clearance, the replacement of conventional amphotericin B (CAmB) with liposomal amphotericin B (LAmB) as an alternative antifungal treatment (given that CAmB was withdrawn from the Australian market in 2010), using different dosage forms of fluconazole (i.e. oral or iv), and not stepping down from iv to oral fluconazole (when patients could tolerate it orally, had clinical improvement or had negative fungal cultures). The variables and the ranges of variation are shown in Table 4. In addition, a threshold analysis was conducted by substituting the baseline value with a range of values until it is beyond the implicit ICER threshold for Australia, AU$76000/LY saved ( value). 24 To investigate if the model was sensitive to the probability of patient distribution in each arm, a two-way exchange of outcome probabilities between anidulafungin and fluconazole was performed. A probabilistic sensitivity analysis using Monte Carlo simulation was performed 5.5 w (Palisade Corporation, Ithaca, NY, USA). 25 Input variables (i.e. outcome probabilities) were allowed to vary simultaneously according to the pre-defined probability distribution (i.e. Trigen distribution) with an uncertainty of +10% (Table 5). In the model, model simulations were run. Each outcome was then calculated and presented with the uncertainty reflected by the 95% confidence interval (CI). The cost-effectiveness acceptability curve (Figure 2) and the cost-effectiveness plane (Figure 3) were used to estimate the probability of anidulafungin being a cost-effective option in the treatment of IC over a range of maximum willingness to pay (WTP). The input variable that had the most influence on the final outcome was also determined. 1909

5 Neoh et al. Table 4. Variation range for variables in sensitivity analysis Variable Base case Variation range low high Anidulafungin cost/vial AU$ AU$0.00 AU$ iv fluconazole cost/vial AU$7.20 AU$0.00 AU$14.40 Hospitalization cost/day AU$ AU$0.00 AU$ ICU cost/day AU$ AU$0.00 AU$ Duration of initial treatment 14.0 days 0.0 day 28.0 days (anidulafungin) Duration of initial treatment 11.0 days 0.0 day 22.0 days (fluconazole) Duration of hospitalization 56.7 days 42.0 days 85.0 days (anidulafungin) Duration of hospitalization 56.7 days 26.0 day 73.0 days (fluconazole) Duration of ICU stay 11.0 days 0.0 day 22.0 days (anidulafungin) Duration of ICU stay 11.0 days 0.0 day 22.0 days (fluconazole) Mean duration to failure 12.2 days 0.0 day 24.4 days (anidulafungin) Mean duration to failure 9.3 days 0.0 day 18.6 days (fluconazole) Mean duration to death 21.0 days 0.0 day 42.0 days (anidulafungin) Mean duration to death 14.0 days 0.0 day 28.0 days (fluconazole) Adjusted life expectancy 13.3 years 5.2 years 24.7 years Annual discount rate 5.0% 0% 6.0% Counting the cost of screening yes no yes and monitoring test Percentage of patients that 10.0% 0.0% 100.0% required fluconazole dosage adjustment Replacement of alternative no no yes CAmB with LAmB Dosage form of fluconazole all iv all oral all iv (orally:iv) Step down to oral fluconazole yes no yes Results Base case analysis In the treatment of IC, anidulafungin (AU$74587) incurred an incremental cost of AU$13642 over fluconazole (AU$60945) (Table 6). The major contributing clinical outcome to the overall costs for both anidulafungin and fluconazole was treatment success. The main cost driver for both comparators was hospitalization (Figure 4). At the end of iv therapy, anidulafungin achieved better efficacy than fluconazole, with an additional 15.4% of patients having a successful response (P,0.02). 6 The ICER for an additional patient attaining successful outcome with anidulafungin treatment was AU$ When the rate of mortality was expressed as the number of LYs lost relative to life expectancy of the most common co-existing condition in the study, treatment with anidulafungin was expected to save an additional of 0.53 LYs compared with fluconazole. As such, treatment with anidulafungin was associated with the ICER of AU$25740/LY saved. Probabilistic analysis The average cost of anidulafungin therapy was AU$74587 (95% CI AU$ ). Fluconazole treatment offered a lower mean cost at AU$60943 (95% CI AU$ ). The difference in the mean percentage of success response at the end of iv therapy was 15.4% (95% CI 3.9% 27.0%). Therefore, treatment with anidulafungin was associated with a mean ICER of AU$88584/treatment success (95% CI AU$ ). The mean LYs saved with anidulafungin therapy compared with fluconazole was 0.52 (95% CI ). This translated to an average ICER of AU$25744/LY saved (95% CI AU$ ) in favour of anidulafungin. The variables that impacted on the model outcome are shown in Figure 5. The ICER per LY saved was most sensitive to the treatment success in the anidulafungin arm. According to the cost-effectiveness acceptability curve (Figure 2), the probability of anidulafungin being a cost-effective option in the treatment of IC was.99.9%. Sensitivity analyses One-way sensitivity analyses indicated that the ICER was not sensitive to changes in the parameters given in Table 4. Table 5. Input variables and uncertainty distribution used in Monte Carlo simulation Uncertainty distribution Input variables anidulafungin fluconazole Treatment success 68.03% 75.59% 83.15% 54.15% 60.17% 66.19% Treatment failure 21.97% 24.41% 26.85% 35.85% 39.83% 43.81% clinical success and 2.91% 3.23% 3.55% 3.83% 4.26% 4.69% clinical failure and microbiological success 11.61% 12.90% 14.19% 11.49% 12.77% 14.05% clinical failure and 23.23% 25.81% 28.39% 30.64% 34.04% 37.44% indeterminate failure 52.25% 58.06% 63.87% 44.05% 48.94% 53.83% 1910

6 Cost-effectiveness study in invasive candidiasis JAC 1 2.5% 95.0% 2.5% Probability of cost-effectiveness ICER per LYs saved in thousands (AU$) Figure 2. Cost-effectiveness acceptability curve of anidulafungin. Incremental cost (AU$) Incremental effectiveness (LYs saved) Figure 3. Cost-effectiveness plane. Varying a single model parameter across its corresponding range (+100% from the base case value) resulted in an ICER range of AU$ /LY saved, demonstrating an economic advantage for the anidulafungin arm. The model outcome remained robust to changes in the duration of hospitalization only to the extent of +25.9% 28.7% from the base case value. Furthermore, the model outcome was insensitive to the uncertainty in adjusted life expectancy of the most common co-existing disease among the study participants and annual discount rate. Likewise, estimations made by the expert panel did not have any impact on the model outcome. Excluding the cost of the screening and monitoring tests, as estimated by the expert panel, decreased the overall cost of anidulafungin and fluconazole to AU$73542 and AU$59884, respectively, with the ICER remaining at AU$25770/LY saved in favour of anidulafungin. The model was also robust to other scenario analyses, with the ICER remaining in the range of AU$ /LY saved. Threshold analysis showed similar observations, except the model was robust at a wider range of variables. The two-way exchange of outcome probabilities between the anidulafungin and fluconazole arms increased the ICER to AU$60279/LY saved in favour of anidulafungin. Discussion This is the first economic evaluation to examine whether anidulafungin, despite its higher acquisition cost than fluconazole, could potentially be a cost-effective option in the treatment of IC in the Australian setting. Anidulafungin was associated with an ICER of AU$25740/LY saved, which was below the implicit ICER threshold for Australia, 24 suggesting that anidulafungin is a cost-effective agent. The considerable decrease in the ICER per LY saved was due to the lower mortality rate in the anidulafungin arm. Our approach to estimate and adjust life expectancy for the most common co-existing disease has been described in the literature. 20,21 Using the life expectancy of sepsis survivors may overestimate survival, given that patients with IC might, in general, have more co-morbidities. However, septicaemia was the most common co-existing condition in both arms. While using relative risk of death to adjust for life expectancy of sepsis survivors is a limitation, sensitivity analyses demonstrated that the model outcome was robust to variations in this parameter (Table 4). Hospitalization (Figure 4) was the major cost driver in both arms, consistent with the literature. 8,9,26,27 Any variation in daily hospitalization costs or length of stay (LOS) would affect the total hospitalization cost, and thus the overall treatment cost. In the current model, the hospitalization cost of the most common co-existing condition (i.e. septicaemia) was used, an approach that has been routinely used in economic analyses involving patients with acute infections. 8,19 Importantly, the model outcome was insensitive to changes in daily hospitalization costs. Patients who succeed in the treatment or died during hospitalization may have a shorter LOS. In situations 1911

7 Neoh et al. Table 6. Proportional cost of treatment of IC using anidulafungin and fluconazole Anidulafungin Fluconazole Therapy outcome proportion (%) cost (AU$)/ patient proportional cost (AU$) proportion (%) cost (AU$)/ patient proportional cost (AU$) Treatment success Treatment failure clinical success and clinical failure and microbiological success clinical failure and microbiological failure indeterminate failure % % 1.26% 0.09% 0.34% 1.50% 2.74% 0.24% Drug cost (initial treatment) Screening & monitoring tests (initial treatment) Total cost (AU$) % 95.18% Drug cost (alternative treatment) Screening & monitoring tests (alternative treatment) Hospitalization (initial and alternative treatment) Anidulafungin Fluconazole Figure 4. Cost components of overall treatment of IC for anidulafungin and fluconazole arms. when the LOS of patients who had successful treatment was shortened, this would further add to anidulafungin s economic advantage, given that it had more successfully treated patients. In the current model, anidulafungin had a longer LOS than fluconazole for those patients who died within the trial period; the outcome remained robust when the mean durations to death 1912

8 Cost-effectiveness study in invasive candidiasis JAC ICER per LYs saved (AU$) Regression Coefficients Treatment success anidulafungin Indeterminate failure fluconazole Treatment success fluconazole Indeterminate failure anidulafungin Clinical failure and microbiological success anidulafungin Clinical failure and microbiological success fluconazole Clinical failure and anidulafungin Clinical failure and fluconazole Coefficient value Figure 5. Tornado diagram of the regression of clinical variables on the model outcome. were varied by % from the base case value (Table 4). Davis et al. 28 commented that the cost in candidaemia patients who died was, in fact, lower than those who survived. Although this would reduce the total cost of fluconazole, as deaths were more in the fluconazole arm, it would also imply more LYs saved by anidulafungin compared with fluconazole, which would favour anidulafungin. The cost of managing common side effects such as nausea and diarrhoea 6 were not included because it was impossible to obtain an accurate estimation of the resources consumed in managing these side effects, which are likely to be less severe and generally do not result in treatment discontinuation. For adverse events that led to discontinuation, the related secondary costs were not incorporated, given that most of the adverse events were signs and symptoms likely to be associated with worsening clinical condition rather than toxic effects of the antifungal agents. 6 Including the costs of managing side effects into the model would probably favour anidulafungin. As illustrated in Figure 5, treatment success in the anidulafungin arm had the greatest influence on the model outcome, followed by indeterminate failure in the fluconazole arm. This is not surprising given that these variables have the highest proportion of patient distribution among the clinical variables, which contributed primarily to the total cost incurred in each arm. Although Reboli et al. 6 indicated a possible centre effect in which the success rates changed to 73.2% for anidulafungin versus 61.1% for fluconazole when the data for the high enrolment centre were removed from the primary analysis, probabilistic analysis demonstrated that anidulafungin had a 99.9% chance of being a cost-effective option. The strength of the current model is its ability to capture the downstream economic consequences of the treatment pathways for each study arm. All patients were followed until death or success, even after the initial treatment was discontinued. In the case of invasive fungal infections (IFIs), it is extremely important to include secondary costs related to costs of therapy failure 29 to provide accurate cost-effectiveness data to guide prescribing. 30 This economic study utilized clinical data generated from an RCT 6 in which the patients reflected the clinical caseload of IC in Australia. The indication and administration of anidulafungin and fluconazole in the RCT 6 are similar to current Australian clinical practice. Another strength of the model is that it included all possible clinical outcomes as reported in the trial, 6 fully depicting the scenario of treatment of IC and thus allowing comprehensive estimation of the costs involved. This is crucial to prevent a situation in which important clinical outcomes with considerable weighted costs are ignored, leading to over- or under-estimation of the cost advantage and health benefit. The model reflected current Australian clinical practice, as the choices and the duration of alternative antifungal agents for the particular treatment failure were advised by the expert panel, based on the types of causative fungal species and the site of the last positive fungal culture. Unlike other economic studies of the treatment of IFIs, 18,19,26 which often allow the switch to only one antifungal as an alternative, our model provided a more realistic estimate of the costs incurred. There were several limitations to our study. As Reboli et al. s 6 trial included predominantly non-neutropenic patients with IC, the economic results may not be generalizable to the population of patients with neutropenic IC. Studies have shown that nonalbicans species are problematic among the non-neutropenic patients in Australia 31 and the USA. 32 The fact that fluconazole has a lower efficacy against non-albicans species with decreased susceptibility or resistance to fluconazole would further favour anidulafungin. Extrapolating the mortality rates at the 6 week follow-up of the study, where the difference between both arms was found to be not statistically significant (P¼0.13), 6 to estimate the LYs saved, while excluding the costs incurred by the survivors in the remaining years, could have potentially oversimplified the estimation. Nevertheless, this method has been described in the published literature. 18,19 As there was a higher proportion of survivors with treatment failure in the fluconazole arm, inclusion of these cost consequences would further strengthen the economic advantage of anidulafungin. Although the model was able to measure the cost of alternative 1913

9 Neoh et al. treatment, it allowed only a single switch and assumed that the alternative treatment would be successful, whereas patients may have multiple switches of antifungal treatment. However, to minimize the effect of this assumption, the alternative antifungal therapy was assigned based on fungal culture and site of infection to better reflect clinical practice. As there were more therapy discontinuations due to treatment failure with fluconazole than anidulafungin (Table 1), the additional costs associated with switching to alternative antifungal therapy would be higher for fluconazole, thus increasing the total cost in the fluconazole arm, again favouring anidulafungin. The use of expert opinion may be subject to bias, but is often used when there are no other sources of data. 17,33 To increase the generalizability of the findings, the expert panel comprised key opinion leaders in antifungal therapy drawn from five hospitals across Australia; the number of expert panel members was in accordance with the literature Importantly, scenario analysis showed that the model outcome was robust to the estimations made by the panel. While the findings cannot be directly extrapolated to other countries, the model could be used with relevant local data to provide locally applicable costeffectiveness estimations. In conclusion, this study has demonstrated that anidulafungin is a cost-effective option in the treatment of IC. This finding emphasizes that the traditional cost-containment concept, which is primarily based on superficial cost considerations, in particular drug acquisition costs, should not be the sole determinant in the choice of antifungal agent. Acknowledgements The preliminary data were presented as an oral presentation at the Twelfth Western Pacific Congress on Chemotherapy and Infectious Diseases, Singapore, 2010 (OP-3-10). We thank Lara Donovan and Jean Spinks from the Centre for Health Economics, Monash University, for their assistance with this study. We also thank Quoc Nguyen for providing data collected as part of the Australian Candidaemia Study. Funding This study was not supported by any pharmaceutical company. The study was partly funded by a scholarship from the University of Technology MARA to C. F. N. Transparency declarations D. L. has sat on advisory boards for and received honoraria from Pfizer, although not in relation to antifungal therapies. M. S. and S. C.-A. C. have sat on advisory boards for and received research funding from Pfizer, Merck, Schering-Plough and Gilead Sciences. D. M. has sat on advisory boards for Pfizer, Merck and Gilead Sciences, and has received research funding from Pfizer and Merck. O. M. has been a consultant to, received investigator-initiated grant support from and given lectures for Gilead Sciences, Pfizer, Merck, Schering-Plough and Orphan Australia. All other authors: none to declare. References 1 Pappas PG, Kauffman CA, Andes D et al. Clinical practice guidelines for the management of candidiasis: 2009 update by the Infectious Diseases Society of America. Clin Infect Dis 2009; 48: Thursky KA, Playford EG, Seymour JF et al. Recommendations for the treatment of established fungal infections. Intern Med J 2008; 38: Slavin M, Fastenau J, Sukarom I et al. Burden of hospitalisation of patients with Candida and Aspergillus infections in Australia. Int J Infect Dis 2004; 8: Slavin MA. Introduction to the updated Australian and New Zealand consensus guidelines for the use of antifungal agents in the haematology/oncology setting. Intern Med J 2008; 38: Golan Y, Harrison D, Fahrbach K. An echinocandin vs. a comparator antifungal in Candida bloodstream infections: a meta-analysis. In: Abstracts of the Twentieth European Congress of Clinical Microbiology and Infectious Diseases, Vienna, Austria, Abstract O410. European Society of Clinical Microbiology and Infectious Diseases, Basel, Switzerland. 6 Reboli AC, Rotstein C, Pappas PG et al. Anidulafungin versus fluconazole for invasive candidiasis. N Engl J Med 2007; 356: Chen S, Slavin M, Nguyen Q et al. Active surveillance for candidemia, Australia. Emerg Infect Dis 2006; 12: Cornely OA, Sidhu M, Odeyemi I et al. Economic analysis of micafungin versus liposomal amphotericin B for treatment of candidaemia and invasive candidiasis in Germany. Curr Med Res Opin 2008; 24: Sidhu MK, van Engen AK, Kleintjens J et al. Cost-effectiveness analysis of micafungin versus caspofungin for treatment of systemic Candida infections in the United Kingdom. Curr Med Res Opin 2009; 25: Marriott DJ, Playford EG, Chen S et al. Determinants of mortality in non-neutropenic ICU patients with candidaemia. Crit Care 2009; 13: R Australian Bureau of Statistics. National Health Survey: Summary of Results (2005). nsf/0/3b a042ca25711f /$file/43640_ pdf (1 August 2010, date last accessed). 12 Australian Government of Health and Ageing. National Hospital Cost Data Collection. Round 13 ( ) Cost Report Version ECCA25773B00031A09/$File/HeaderR13CWNatEst.pdf (1 August 2010, date last accessed). 13 Australian Bureau of Statistics. Consumer Price Index (2010). DB27E14A103FB951CA25776D001B5ADF/$File/64010_jun% pdf (1 August 2010, date last accessed). 14 Rechner IJ, Lipman J. The costs of caring for patients in a tertiary referral Australian intensive care unit. Anaesth Intensive Care 2005; 33: Australian Government of Health and Ageing. Medicare Benefits Schedule Book (2010). publishing.nsf/content/bbe433bf3fe34a39ca d4/$file/ MBS.pdf (1 August 2010, date last accessed). 16 Health Purchasing Victoria. Health Purchasing Victoria Tender ( ). (3 November 2010, date last accessed). 17 Petitti DB. Meta-Analysis, Decision Analysis, and Cost-Effectiveness Analysis. New York: Oxford University Press, Stam WB, Aversa F, Kumar RN et al. Economic evaluation of caspofungin versus liposomal amphotericin B for empiric antifungal 1914

10 Cost-effectiveness study in invasive candidiasis JAC treatment in patients with neutropenic fever in Italy. Value Health 2008; 11: Bruynesteyn K, Gant V, McKenzie C et al. A cost-effectiveness analysis of caspofungin versus liposomal amphotericin B for treatment of suspected fungal infections in the United Kingdom. Eur J Haematol 2007; 78: Talmor D, Greenberg D, Howell MD et al. The costs and cost-effectiveness of an integrated sepsis treatment protocol. Crit Care Med 2008; 36: Angus DC, Linde-Zwirble WT, Clermont G et al. Cost-effectiveness of drotrecogin alfa (activated) in the treatment of severe sepsis. Crit Care Med 2003; 31: Australian Bureau of Statistics. Australian Life Tables ( ). (1 August 2010, date last accessed). 23 Quartin AA, Schein RM, Kett DH et al. Magnitude and duration of the effect of sepsis on survival. Department of Veterans Affairs Systemic Sepsis Cooperative Studies Group. JAMA 1997; 277: George B, Harris A, Mitchell A. Cost-effectiveness analysis and the consistency of decision making: evidence from pharmaceutical reimbursement in Australia (1991 to 1996). PharmacoEconomics 2001; 19: O Hagan A, Stevenson M, Madan J. Monte Carlo probabilistic sensitivity analysis for patient level simulation models: efficient estimation of mean and variance using ANOVA. Health Econ 2007; 16: Rentz AM, Halpern MT, Bowden R. The impact of candidemia on length of hospital stay, outcome, and overall cost of illness. Clin Infect Dis 1998; 27: Rotstein C, Cragin L, Laverdiere M et al. Economic evaluation of voriconazole for the treatment of candidemia in Canadian adults. Can J Infect Dis Med Microbiol 2008; 19: Vandijck DM, Blot SI, Decruyenaere JM. Comment: Epidemiology, risk factors, and outcomes of Candida albicans versus non-albicans candidemia in nonneutropenic patients. Ann Pharmacother 2007; 41: ; author reply Dixon S, McKeen E, Tabberer M et al. Economic evaluations of treatments for systemic fungal infections: a systematic review of the literature. PharmacoEconomics 2004; 22: Johnson MD, Kleinberg M, Danziger L et al. Pharmacoeconomics of antifungal pharmacotherapy: challenges and future directions. Expert Opin Pharmacother 2005; 6: Playford EG, Marriott D, Nguyen Q et al. Candidaemia in nonneutropenic critically ill patients: risk factors for non-albicans Candida spp. Crit Care Med 2008; 36: Davis SL, Vazquez JA, McKinnon PS. Epidemiology, risk factors, and outcomes of Candida albicans versus non-albicans candidaemia in nonneutropenic patients. Ann Pharmacother 2007; 41: Gold M, Siegel J, Russell L et al. Cost-Effectiveness in Health and Medicine. New York: Oxford University Press, Al-Badriyeh D, Liew D, Stewart K et al. Economic impact of caspofungin as compared with liposomal amphotericin B for empirical therapy in febrile neutropenia in Australia. J Antimicrob Chemother 2009; 63: Al-Badriyeh D, Liew D, Stewart K et al. Cost-effectiveness evaluation of voriconazole versus liposomal amphotericin B as empirical therapy for febrile neutropenia in Australia. J Antimicrob Chemother 2009; 63: Wenzel R, Del Favero A, Kibbler C et al. Economic evaluation of voriconazole compared with conventional amphotericin B for the primary treatment of aspergillosis in immunocompromised patients. J Antimicrob Chemother 2005; 55: