COST EFFECTIVENESS IN CANCER AND CHRONIC CONDITIONS
* Finalists for the Lee B. Lusted Student Prize
Methods: Study design: cost-effective and cost-utility analyses from the societal and health-care payer’s perspectives. Target population: patients nearing end-of-life (according to observed population-based setting-specific patterns of palliative services) over the last year of life (a one-year time horizon) and their primary informal caregivers. Comparators: “usual care” with current palliative services, “in-home palliative team care” with service provision at home or in long-term care home, “in-patient palliative team care”, and “comprehensive palliative team care” by a single team with care coordination across settings.
We included health care costs (2013 Canadian dollars; including costs of home care, long-term care, hospital care, outpatient services, drugs, and physician billings), and additional costs of private insurance, out-of-pocket expenses, and time lost from paid work. Effectiveness measures included days at home and percentage dying at home (base case analysis), and quality-adjusted life days (QALD; sensitivity analysis).
We developed a state-transition microsimulation model to simulate palliative care needs by the target population. Model inputs were obtained from an end-of-life cohort of Ontarian decedents (n= 256,284) assembled from linked health administration databases (2007-2009), systematic reviews of randomized controlled trials evaluating palliative team care, and published literature. We conducted one-way and probabilistic sensitivity analyses.
In-home palliative team care dominated usual care, with an expected $4,424 healthcare cost saving (via reduced ER visits and hospital admissions), approximately 5.8 more days at home, 10.3% more home deaths, and 0.47 QALDs gained [with a 72% likelihood of being cost-effective at $50k per QALY]. In-patient team care appeared to dominate usual care (corresponding values: $1,643, 0.65 days, -0.2% and 0.27 QALDs [38%]). Comprehensive palliative team care was associated with $527 additional cost, 1.4 more days at home, 1.7% more home deaths, and 2.65 QALDs gained, corresponding to an estimated cost of approximately $72k per QALY [32%]. Results from the societal perspective were similar, mainly due to a lack of data regarding the team’s effects on quality-of-life, time and costs associated with care giving.
Conclusions: In-home palliative team care is cost-effective. Firm conclusions are not possible regarding the cost-effectiveness of in-patient and comprehensive team care.
DNA ploidy analysis, a semi-automated process, has been proposed as a potential alternative for cervical screening; however, this strategy has not been evaluated economically. Our study examined the cost-effectiveness of ploidy analysis in comparison to liquid-based Papanicolaou (Pap) smear in the screening setting.
The use of ploidy was examined with five thresholds corresponding to the number (from 1 to 5) of aneuploid cells in a specimen. For example, the ploidy 3 cell strategy rendered a specimen abnormal if at least 3 aneuploid cells were found. We compared these five ploidy strategies and the liquid-based Pap smear with a no screening strategy as the reference. We developed a state-transition Markov model to simulate the natural history of HPV infection and possible progression into cervical neoplasia in a hypothetical cohort of 12-year-old females (started triennial screening from 21 years). The analysis was conducted using cost in 2012 US$ and effectiveness in quality-adjusted life-years (QALYs) from a health-system perspective throughout a lifetime horizon in the US setting. The willingness-to-pay threshold was $50,000/QALY. We calculated the incremental cost-effectiveness ratios (ICERs) for the various strategies to determine the best ploidy strategy and the overall recommended strategy. The robustness of optimal choices was examined in deterministic and probabilistic sensitivity analyses.
In the base-case analysis, the ploidy 4 cell strategy was cost-effective. It increased the quality-adjusted life expectancy by 0.083 QALY and yielded an ICER of $8,774/QALY compared to the no screening strategy. In the deterministic sensitivity analysis, the cost-effectiveness was most sensitive to the cost of the Pap smear procedure, the cost of treating high-grade squamous intraepithelial lesions, the cost of the ploidy analysis, and the ploidy strategies' operating characteristics. For most scenarios, the ploidy 4 cell strategy was cost-effective and was considered the best ploidy strategy. The cost-effectiveness acceptability curves showed that the ploidy 4 cell strategy was more likely to be cost-effective than the Pap smear strategy.
Compared to liquid-based Pap smear screening, ploidy analysis appeared less costly and comparably effective using the standard willingness-to-pay threshold. Screening for cervical neoplasia using DNA ploidy analysis may be a satisfactory alternative, particularly in low-infrastructure settings.
Figure 1. Cost-effectiveness acceptability curves comparing no screening, Papanicolaou smear screening, and the ploidy 4 cell strategy.
Methods: We updated a previously developed state-transition Markov model of CML, which evaluates seven treatment regimens including different combinations of tyrosine kinase inhibitors, chemotherapy and stem cell transplantation (SCT). For model parameters, we used published trial data, and Austrian clinical, epidemiological, and economic data. We performed a cohort simulation over a lifetime horizon, adopted a societal perspective, and discounted costs and benefits at 3% annually. For the probabilistic sensitivity analysis and the VoI analysis, we defined parameter uncertainty distributions from our source data. We calculated the expected value of perfect information (EVPI), partial perfect information (EVPPI), and the population EVPI (PEVPI). Additionally, we examined the expected value of sample information (EVSI) for different trial sizes. The goal was to estimate the expected benefit of future research and identify parameters whose further study was most valuable for resolving decision uncertainty.
Results: Three strategies are on the efficiency frontier: imatinib-->chemotherapy/SCT, nilotinib-->chemotherapy/SCT (140,000 €/QALY) and nilotinib-->dasatinib-->chemotherapy/SCT (176,000 €/QALY). The EVPI for eliminating all uncertainty results in a curve with two peaks. One peak is around a WTP threshold of 150,000 €/QALY with an EVPI of 4,600 € and another peak can be found at 180,000 €/QALY with an EVPI of 7,700 € (Figure 1). The PEVPI for Austria assuming a 10-year technology horizon was 2.5 million € (WTP 150,000 €/QALY) and 4.5 million € (WTP 180,000 €/QALY). EVPPI identified four parameters most responsible for decision uncertainty: Duration of first-line TKI-therapy, probability of progressing from chronic phase to accelerated phase of disease, probability of receiving a SCT after therapy failure, and the utility after SCT of suffering from chronic graft-versus-host disease. EVSI commented on the optimal study size for these parameters given the cost of obtaining information.
Conclusions: Acquiring additional evidence could prove valuable for determining optimal treatment regimens for chronic myeloid leukemia. If further research were funded, studies should examine a combination of natural history, treatment, and quality of life parameters, especially the effectiveness of first-line TKI treatment.
Method: We used data from two randomized clinical trials: 1)EURTAC (median age 62 years), comparing erlotinib to cisplatin plus carboplatin or gemcitabine, and 2)LUX-LUNG 3 (median age 65 years), comparing afatinib to cisplatin plus pemetrexed, for first-line EGFR-positive stage IIIB/IV NSCLC treatment. Survival probabilities of EURTAC were corrected for presence of participants with ECOG performance score >1. We developed a Markov model to simulate transition through progression-free survival (PFS), progression (overall survival [OS]-PFS), and death (1-OS) under erlotinib, afatinib, and cisplatin/pemetrexed, using a societal perspective and lifetime horizon. Survival and side-effect probabilities, obtained from the trials, costs (3% annual discounting), and utilities were modeled as distributions for probabilistic sensitivity analysis. Costs and quality-adjusted life years (QALY) were used to calculate incremental cost-effectiveness ratios (ICER). We constructed acceptability curves by plotting the proportion of simulations a treatment had the highest net benefit (NB=effectivenesness*willingness-to-pay[WTP]Cost) over a range of WTP thresholds. We calculated EVPI to estimate the expected benefit of further research to decrease decision uncertainty. We also identified parameters responsible for most of the decision uncertainty.
Result: In the base case, using societal WTP threshold of $100,000/QALY, erlotinib was cost-effective compared with afatinib, with a mean ICER of $61,809/QALY. The acceptability curve showed that erlotinib was the preferred treatment at WTP of $100,000/QALY. Uncertainty regarding decision of erlotinib versus afatinib was highest at WTP of $50,000-$70,000/QALY, and afatinib was preferred at WTP<$30,000/QALY. Population EVPI analysis showed that the maximum acceptable cost of reducing decision uncertainty is $46.5 million, assuming an effective lifetime for current treatments of 10 years at 3% discount annually. The following parameters were responsible for most of the decision uncertainty: monthly cost of erlotinib, monthly cost of afatinib, and the probability and cost of rash together for afatinib.
Conclusion: Erlotinib is the preferred treatment based on its ICER, compared with afatinib. At WTP of $60,000/QALY, further research to determine the optimal treatment is justified, given a trial of 230 participants, at $200,000/trial participant. However, further research is not justified at WTP of $100,000/QALY.
Purpose: Hepatocellular carcinoma (HCC) is the fastest growing cause of cancer-related death in the United States, and its incidence has recently increased mainly due to hepatitis C infection. Early detection of HCC through regular surveillance can improve prognosis. However due to a lack of any randomized clinical trial in hepatitis C patients, the benefit and optimal screening interval remains questionable. Our objective was to determine cost-effective patient-centered surveillance strategies, and compare them with the current recommendations.
Method: We developed a novel mathematical model to determine the optimal surveillance strategy maximizing the net benefit (NB). We allowed the screening interval to range from 3 months to 2 years. We considered one-size-fits-all policies, and policies with changing screening intervals after a long period of time, which we called the interval-switching policies. We also considered the possibility that the optimal surveillance strategy may be different depending on patients' age and stage of liver disease (advanced fibrosis [F3], compensated cirrhosis [CC], or decompensated cirrhosis [DC]). The baseline policy was a one-size-fit-all policy with the same screening interval (i.e., 6 or 12 months) as recommended by some guidelines. Our model's parameters were estimated from meta-analysis results and published clinical studies. Extensive sensitivity analyses were conducted on various model parameters, including risks of disease progression and treatment cost-effectiveness.
Result: The interval-switching policy with a cycle of 6 years had a higher NB ($560,353), in comparison with semiannual screenings (NB=$554,871) and annual screenings (NB=$553,166). We found that (1) the surveillance frequency should be more aggressive in advanced stages of the liver disease, (2) surveillance frequency should become less aggressive as the population ages, and 3) surveillance is not cost-effective after mid-70's. We also found among the policies with no change in screening intervals, the optimal policy was every 3-month screening for patients in F3, semi-annual screening for CC, and annual screening for DC.
Conclusion: Current guidelines for HCC surveillance are controversial and mostly emphasize one-size-fit-all type surveillance strategies. We also find that rather than the one-size-fits-all type policies, surveillance strategies stratified by fibrosis stages can significantly improve cost-effectiveness. Lastly, we found that HCC surveillance is not cost effective in patients older than mid-70's.
Figure 1. Optimal interval-switching policies with cycle of 6 years.
Method: A decision-analytic model was used to estimate the incremental cost-effectiveness ratio (ICER) between treatment arms of the BOLERO-2 trial according to a public payer perspective and time horizon matching the duration of the clinical trial. The population studied was composed of postmenopausal women with hormone-receptor positive (HR+) and HER2-negative metastatic breast cancer. Costs were obtained from the Center for Medicare Services drug payment table and physician fee schedule. Benefits were expressed as quality-adjusted progression-free survival weeks and quality-adjusted progression-free years, with utilities/disutilities derived from the literature. Deterministic and probabilistic sensitivity analyses were performed.
Result: Everolimus/exemestane had an incremental benefit of 11·88 QAPFW compared to exemestane or 0·22 QAPFY, and an incremental cost of $60,574. This translates into an ICER of $265,498·5/QAPFY. Key drivers of our model, by order of importance include: health utility value for stable disease, everolimus acquisition costs, and transition probabilities from the stable to the progression states. The Monte-Carlo simulation showed results that were similar to the base-case analysis.
Conclusion: Everolimus plus exemestane in postmenopausal women with hormone-receptor positive and HER2-negative metastatic breast cancer is not cost-effective compared to exemestane alone. Further research investigating the cost-effectiveness of the combination versus the monotherapy, in sub-groups of the population studied in BOLERO-2, will be beneficial.