Probabilistic Sensitivity Analysis Methods for Design Under Uncertainty

To overcome the limitations of existing variance-based methods for Probabilistic Sensitivity Analysis (PSA) in design under uncertainty, a new PSA approach based on the concept of relative entropy is proposed. The relative entropy based method evaluates the impact of a random variable by measuring the divergence between two probability density functions of a response. The method can be applied both globally over the whole distribution of a performance response (called global response probabilistic sensitivity analysis–GRPSA) and in any regional range of a response distribution (called regional response probabilistic sensitivity analysis–RRPSA). The former is the most useful for studying variable impact on robust design objective, while the latter provides insight into reliability constraints. The proposed method is applicable to both the prior-design and post-design stages, for variable screening and uncertainty reduction, respectively. The proposed method is verified by numerical examples and industrial design cases.

Misclassification bias is present in most studies, yet uncertainty about its magnitude or direction is rarely quantified. The authors present a method for probabilistic sensitivity analysis to quantify likely effects of misclassification of a dichotomous outcome, exposure or covariate. This method involves reconstructing the data that would have been observed had the misclassified variable been correctly classified, given the sensitivity and specificity of classification. The accompanying SAS macro implements the method and allows users to specify ranges of sensitivity and specificity of misclassification parameters to yield simulation intervals that incorporate both systematic and random error. The authors illustrate the method and the accompanying SAS macro code by applying it to a study of the relation between occupational resin exposure and lung-cancer deaths. The authors compare the results using this method with the conventional result, which accounts for random error only, and with the original sensitivity analysis results. By accounting for plausible degrees of misclassification, investigators can present study results in a way that incorporates uncertainty about the bias due to misclassification, and so avoid misleadingly precise-looking results.

Cancer-associated venous thromboembolism (CAVTE) is a preventable, life-threatening complication with a considerable morbidity and mortality. Primary venous thromboembolism (VTE) prophylaxis is currently recommended; however, the health and economic benefits have not been evaluated and compared in China. This study aimed to assess and compare the cost-effectiveness of anticoagulants in primary CAVTE prevention among cancer patients in China. A Markov model with a 5-year horizon was established to evaluate the costs and effectiveness of direct oral anticoagulants (DOACs) compared to low-molecular-weight heparins (LMWHs) and no prevention in primary prophylaxis of CAVTE in China. Key clinical outcomes were obtained from the available clinical trials, comparing DOACs (rivaroxaban and apixaban) with LMWHs or with no thromboprophylaxis. Utility and the cost inputs were all obtained from the published literature or local data with public sources. The total costs, quality-adjusted life-years (QALYs), and incremental cost-effectiveness ratios (ICERs) were estimated as the main endpoints of the modal for each strategy. The assessment of uncertainty was performed involving deterministic sensitivity analysis and probabilistic sensitivity analysis (PSA). Impact of time horizon, generic drug price, and individual DOACs were assessed in scenario and subgroup analyses. Primary prophylaxis using DOACs were projected to yield 1.866 QALYs at a cost of $3,287.893, resulting in the ICERs of $12,895.851 (DOACs vs. no-thromboprophylaxis) and $43,613.184/QALYs (LMWHs vs. DOACs). Sensitivity analysis revealed that ICER was sensitive to the VTE and bleeding risk, drug cost of anticoagulants, self-payment ratio, and overall death rate of cancer. Probabilistic sensitivity analysis showed that DOACs and LMWHs had a 48% and 45% probability of being cost-effective at a 5-year time horizon, respectively. When the time horizon extended to 10 years, DOACs achieved a cost-effective probability of 43%. Among individual DOACs, apixaban was found to be the preferred strategy in VTE prevention due to its incremental health gain with an acceptable cost increase. Primary thromboprophylaxis with DOACs was cost-effective in cancer patients at a willing-to-pay (WTP) threshold of $37,125.24/QALY in China. Cancer death rate, risk of VTE and major bleeding, and the drug cost assumed greater relevance and importance in the decision-making process for primary thromboprophylaxis in cancer.

Background and Purpose- Our study aims to evaluate the cost-effectiveness of computed tomography angiography (CTA) for surveillance of tiny unruptured intracranial aneurysms and the impact of CTA radiation-induced brain tumor on the overall effectiveness of CTA. Methods- A Markov decision model was constructed from a societal perspective starting with patients 30-, 40-, or 50-year-old, with incidental detection of unruptured intracranial aneurysm ≤3 mm and no prior history of subarachnoid hemorrhage. Five different management strategies were assessed (1) annual CTA surveillance, (2) biennial CTA, (3) CTA follow-up every 5 years, (4) coiling and subsequent magnetic resonance imaging follow-up, and (5) annual CTA surveillance for the first 2 years, followed by every 5-year CTA follow-up. Probabilistic, 1-way, and 2-way sensitivity analyses were performed. Results- The base case calculation shows every 5-year CTA follow-up to be the most cost-effective strategy, and the conclusion remains robust in probabilistic sensitivity analysis. It remains the dominant strategy when the annual rupture risk of nongrowing unruptured intracranial aneurysms is smaller than 2.66% or the rupture risk in growing aneurysms is <57.4%. The radiation-induced brain cancer risk is relatively low, and sensitivity analysis shows that the radiation-induced cancer risk does not influence the conclusions unless the risk exceeds 663-fold of the base case values. Conclusions- Given the current literature, every 5-year CTA imaging follow-up is the cost-effective strategy in patients with aneurysms ≤3 mm, resulting in better health outcomes and lower healthcare spending. Patients with aneurysms at high risk of rupture might need more aggressive management.

PNS212 Applying a Normal Distribution to Costs in Probabilistic Sensitivity Analysis – NOT Necessarily a Negative Outcome

Tsunami evacuation risk assessment and probabilistic sensitivity analysis using augmented sample-based approach

Introduction: There are various oral anticoagulants available for stroke prevention in patients suffering from non-valvular atrial fibrillation (NVAF) with some drug-related variations in clinical profile and non-clinical attributes. Patient preferences should be taken into account in anticoagulant prescription decisions. Hypothesis: Patient valuation of different anticoagulant characteristics in stroke prevention allows for meaningful comparison of the non-VKA oral anticoagulants (NOACs; apixaban, dabigatran, edoxaban, rivaroxaban) and Vitamin K Antagonist (VKA, ie. warfarin). Methods: Multi-criteria decision analysis was developed with 5 clinical and 3 non-clinical criteria. Criteria weights were defined using results from two discrete choice experiments (DCEs). The evaluation model contained more fine-grained events than the DCEs, and therefore preference weights from DCEs needed to be distributed to the evaluation criteria. The weights were distributed according to event fatality rates, which were obtained from the RE-LY trial that compared dabigatran to warfarin. An additive model was used to combine treatment performance with the weights to estimate the overall value of each oral anticoagulant. Probabilistic and structural sensitivity analyses were performed. Results: Dabigatran obtained the highest overall value (see Figure: weighted contribution of different evaluation criteria to the overall value of five oral anticoagulants) and the highest first rank probability (88%) in the probabilistic sensitivity analysis. Rivaroxaban performed worse than the other NOACs, but better than VKA (both with 0% first rank probability). The results were insensitive to removing availability of reversal agent from the model. Conclusions: Patient preferences on treatment characteristics allows to discriminate oral anticoagulants for stroke prevention in NVAF, with dabigatran ranking highest and warfarin lowest.

BackgroundTo analyse the cost-effectiveness of traditional disease-modifying anti-rheumatic drugs (tDMARDs) compared to biological therapies from the perspective of Chinese society.Methodology/Principal FindingsA mathematical model was developed by incorporating the clinical trial data and Chinese unit costs and treatment sequences from a lifetime perspective. Hypothetical cohorts with moderate to severe RA were simulated. The primary outcome measure–quality-adjusted life years (QALYs)–was derived from disease severity (HAQ scores). Primary analysis included drug costs, monitoring costs, and other costs. Probabilistic and one-way sensitivity analyses were performed. Treatment sequences that included TNF antagonists and rituximab produced a greater number of QALYs than tDMARDs alone or TNF antagonists plus DMARDs. In comparison with tDMARDs, the incremental cost-effectiveness ratios (ICERs) for etanercept, infliximab, and adalimumab without rituximab were $77,357.7, $26,562.4 and $57,838.4 per QALY and $66,422.9, $28,780.6 and $50,937.6 per QALY, for etanercept, infliximab, and adalimumab with rituximab. No biotherapy was cost-effective under the willingness to pay threshold when the threshold was 3 times the per capita GDP of China. When 3 times the per capita GDP of Shanghai used as the threshold, infliximab and rituximab could yield nearly 90% cost-effective simulations in probabilistic sensitivity analysis.Conclusions/SignificancetDMARD was the most cost-effective option in the Chinese healthcare setting. In some relatively developed regions in China, infliximab and rituximab may be a favorable cost-effective alternative for moderate to severe RA.

Influenza is a common respiratory disease occurring in seasonal patterns, and may lead to severe complications in frail populations such as the elderly. In Poland, influenza vaccination is recommended for people aged ≥65 years; however the vaccine coverage rate in the elderly is very low. The fact that influenza vaccine is neither reimbursed by the National Health Insurance (Narodowy Fundusz Zdrowia [NFZ]) nor financed via a National Immunization Program (NIP) could be a reason for the low coverage rate. This study assessed the cost effectiveness of the full reimbursement of an influenza vaccination programme in Poland for people aged ≥65 years. A decision-analytic model was developed to compare costs and outcomes associated with the current situation in which influenza vaccination is not reimbursed and a new situation in which it would be fully covered by the NFZ. The model was parameterized to Poland using data from the literature and from the Central Statistic Office of Poland. Within the elderly population, 50% were considered to be at high risk of influenza complications. An influenza attack rate of 3.5% was used for calculation purposes. Influenza-associated hospitalizations and death rates were estimated at 439.9 per 100 000 person-years and 79.1 per 100 000 person-years, respectively. Cost estimates were derived from a cost study conducted in Poland. Costs are presented in Polish Zloty (PLN) [2009 mean exchange rate: 1 PLN = €0.232]. Only direct medical costs were included to fit to the NFZ perspective. To reflect the seasonality of influenza, a time horizon of 1 year was chosen. Life-years and quality-adjusted life-years (QALYs) accumulated over future years were discounted at a rate of 5% as recommended by Polish guidelines. Deterministic and probabilistic sensitivity analyses were conducted. In Poland, the introduction of the public funding of influenza vaccination for people aged ≥65 years would cost PLN 79 million when an increase in coverage rate from 13.5% to 40% is assumed. 23 900 cases of influenza, 1777 hospitalizations and 548 premature deaths would be averted each year due to the influenza vaccination programme. Fifty-seven persons would need to be vaccinated to prevent one case of influenza. To prevent one hospitalization and one death due to influenza, 842 and 2809 individuals would need to be vaccinated, respectively. The new strategy would be very cost effective compared with the current situation with an incremental cost-effectiveness ratio (ICER) of PLN26 118/QALY, which is below the 2009 yearly gross domestic product (GDP) per capita. Deterministic sensitivity analyses demonstrated that the most influential variables for the ICER were vaccine efficacy against death, excess hospitalization rate, utility norms, influenza attack rate, vaccine efficacy against hospitalization, and discount rates. All ICERs computed were below the threshold of 3 GDP per capita. From the probabilistic analysis, the proposed new influenza vaccination programme, if implemented, was predicted to be cost effective from the NFZ perspective with a probability of 100%, given the same threshold. Implementing a vaccination programme in Poland in which influenza vaccination would be fully reimbursed by the NFZ for people aged ≥65 years would be a very cost-effective strategy.

Objective: To assess the cost-effectiveness of prucalopride vs. continued laxative treatment for chronic constipation in patients in the Netherlands in whom laxatives have failed to provide adequate relief.Methods: A Markov model was developed to estimate the cost-effectiveness of prucalopride in patients with chronic constipation receiving standard laxative treatment from the perspective of Dutch payers in 2011. Data sources included published prucalopride clinical trials, published Dutch price/tariff lists, and national population statistics. The model simulated the clinical and economic outcomes associated with prucalopride vs. standard treatment and had a cycle length of 1 month and a follow-up time of 1 year. Response to treatment was defined as the proportion of patients who achieved “normal bowel function”. One-way and probabilistic sensitivity analyses were conducted to test the robustness of the base case.Results: In the base case analysis, the cost of prucalopride relative to continued laxative treatment was € 9015 per quality-adjusted life-year (QALY). Extensive sensitivity analyses and scenario analyses confirmed that the base case cost-effectiveness estimate was robust. One-way sensitivity analyses showed that the model was most sensitive in response to prucalopride; incremental cost-effectiveness ratios ranged from € 6475 to 15,380 per QALY. Probabilistic sensitivity analyses indicated that there is a greater than 80% probability that prucalopride would be cost-effective compared with continued standard treatment, assuming a willingness-to-pay threshold of € 20,000 per QALY from a Dutch societal perspective. A scenario analysis was performed for women only, which resulted in a cost-effectiveness ratio of € 7773 per QALY.Conclusion: Prucalopride was cost-effective in a Dutch patient population, as well as in a women-only subgroup, who had chronic constipation and who obtained inadequate relief from laxatives.

Introduction: Despite all the control efforts made to reduce the prevalence of Tuberculosis (TB) in Malaysia, TB is still a leading cause of mortality among the communicable diseases with a rate of 5.37 per 100,000 population. It has also caused a significant economic burden to the government in the order of USD15 million per year as reported by World Health Organization in 2015. However, the long-term epidemiological and economic burden of TB management has not been established. Objectives: The objectives of this study are to evaluate the health and economic burden of TB management in Malaysia, and the cost-effectiveness of TB control programme by developing a dynamic transmission model. Methodology: A retrospective observational study was conducted in Selangor, including all TB patients first diagnosed in the year 2011. With the collected data, binary logistic regression analysis was carried out to determine the risk factors for incomplete treatment in TB patients. Costing study was also conducted to estimate the cost of TB management. The total cost of TB treatment was calculated by multiplying the frequency of health services utilisation with the unit cost obtained from the Medical Fee Act, 2014. Additionally, a dynamic transmission model was constructed to simulate the natural progression of TB in Malaysia over a 10-year horizon. This model was then used to project the annual rate of TB incidence. This study was conducted from government perspective. Furthermore, we conducted the cost-effectiveness of Xpert MTB/RIF test in Malaysia based on the model built. The results were expressed in cost per disability-adjusted life years (DALYs). A discount rate of 3% was considered for both costs and outcomes. One-way sensitivity analysis and probabilistic sensitivity analysis were performed to ensure consistency and robustness of the model outcomes. Results: A total of 555 cases of TB were identified for this study. On multivariate analysis, risk factors that were significantly associated with incomplete treatment included nationality, occupation and TBIS registry. There were also several factors found, associated with incomplete treatment on univariate analysis which included age, gender, main treatment centre, comorbidities, facility utilisation, HIV/AIDS and distance to main facility. An incomplete treatment would cause the government an average of RM901.63 per patient compared to a completed treatment where RM2,242.02 was needed for a non-hospitalised uncomplicated TB case. A further analysis showed that the cost of treating a case with comorbidities and hospital stays was 4.5 times of the cost of treating an uncomplicated TB case. This cost would further escalate if MDR-TB treatments were required, whereby the cost would be 28.5 times more than treating an uncomplicated TB case. Additionally, our projection suggests that TB notification rate (NR) will increase by 57.64% after 10 years. When this burden was translated into monetary terms, a 19.96% (RM31,876,721) increase in TB budget would be required in 2021 if the control strategies remain unchanged as in 2011. In the base case analysis, the incremental cost-effectiveness ratio was RM47,266.80 for the Xpert scenario versus the current practice, which is slightly more than 1 time gross domestic product (RM41,153.62 per capita in 2014). However, with the current evidence, the probabilistic sensitivity analysis showed that at the willingness to pay (WTP) threshold value of RM30,000 per DALY averted, Xpert scenario was cost effective in more than 95% of all simulations. Thus, it is cost effective to be implemented in Malaysia, provided the WTP threshold is set at 1 time gross domestic product per DALY averted. Conclusions: This study has successfully established the cost of TB management in Malaysia. Hospitalisation appeared to be the major cost drive for TB patients receiving TB treatment, thus strategies that could reduce the length of hospital stay without compromising the quality of care in TB treatments should be considered as future research. The dynamic transmission model has projected that the TB burden would continuously escalate. Thus, the future control strategies and new public health interventions could focus on early detection and treatment of TB such as TB education and adopting new diagnostic tools for rapid detection of TB. These could be the keys to bring Malaysia a step closer to TB elimination.

Cost-effectiveness of Dabigatran Compared With Rivaroxaban for Prevention of Stroke and Systemic Embolism in Patients With Atrial Fibrillation in China

First-Line Treatment With Atezolizumab Plus Bevacizumab and Chemotherapy for US Patients With Metastatic, Persistent, or Recurrent Cervical Cancer: A Cost-Effectiveness Analysis

BACKGROUND: Patients with relapsed or refractory (R/R) acute myeloid leukemia (AML) and confirmed feline McDonough sarcoma (FMS)-like tyrosine kinase 3 gene mutations (FLT3mut+) have a poor prognosis and limited effective treatment options. Gilteritinib is the first targeted therapy approved in the United States and Europe for R/R FLT3mut+ AML with significantly improved efficacy compared with existing treatments. OBJECTIVE: To evaluate gilteritinib against salvage chemotherapy (SC) and best supportive care (BSC) over a lifetime horizon among adult patients with R/R FLT3mut+ AML from a US third-party payer's perspective. METHODS: The model structure of this cost-effectiveness analysis included a decision tree to stratify patients based on their hematopoietic stem cell transplantation (HSCT) status, followed by 2 separate 3-state partitioned survival models to predict the long-term health status conditional on HSCT status. The ADMIRAL trial data and literature were used to predict probabilities of patients being in different health states until a conservative cure point at year 3. Afterwards, living patients followed the survival outcomes of long-term survivors with AML. Model inputs for utilities, medical resource use, and costs were based on the ADMIRAL trial, published literature, and public sources. All costs were inflated to 2019 US dollars (USD). Total incremental costs (in 2019 USD), life-years (LYs), quality-adjusted LYs (QALYs), and incremental cost-effectiveness ratios (ICERs) were calculated. Deterministic sensitivity analyses and probabilistic sensitivity analyses were performed. RESULTS: Over a lifetime horizon with a 3.0% annual discount rate, the base-case model estimated that gilteritinib led to an increase of 1.29 discounted QALYs at an additional cost of $148,106 vs SC, corresponding to an ICER of $115,192 per QALY; for BSC, results were an increase of 2.32 discounted QALYs, $249,674, and $107,435, respectively. The base-case findings were robust in sensitivity analyses. The estimated probabilities of gilteritinib being cost-effective vs SC and BSC were 90.5% and 99.8%, respectively, in the probabilistic sensitivity analyses, based on a willingness-to-pay threshold of $150,000 per QALY. CONCLUSIONS: Gilteritinib is a cost-effective novel treatment for patients with R/R FLT3mut+ AML in the United States. DISCLOSURES: This work was supported by Astellas Pharma, Inc., which was involved in all stages of the research and manuscript development. Garnham, Pandya, and Shah are employees of Astellas and hold stock/stock options. Zeidan consulted and received personal fees/honoraria and research funding from Astellas. Zeidan also has received research funding from Celgene/BMS, Abbvie, Astex, Pfizer, Medimmune/AstraZeneca, Boehringer-Ingelheim, Trovagene/Cardiff Oncology, Incyte, Takeda, Novartis, Amgen, Aprea, and ADC Therapeutics; has participated in advisory boards; has consulted with and/or received honoraria from AbbVie, Otsuka, Pfizer, Celgene/BMS, Jazz, Incyte, Agios, Boehringer-Ingelheim, Novartis, Acceleron, Daiichi Sankyo, Taiho, Seattle Genetics, BeyondSpring, Cardiff Oncology, Takeda, Ionis, Amgen, Janssen, Syndax, Gilead, Kura, Aprea, Lox Oncology, Genentech, Servier, Jasper, Tyme, and Epizyme; has served on clinical trial committees for Novartis, Abbvie, Geron, Gilead, Kura, Lox Oncology, BioCryst, and Celgene/BMS; and has received travel support for meetings from Pfizer, Novartis, and Cardiff Oncology. Qi and Yang are employees of Analysis Group, Inc., which received consulting fees from Astellas for work on this study. Part of this material was presented at the American Society of Hematology (ASH) Annual Meeting; December 7-10, 2019; Orlando, FL.

We sought to determine the cost-effectiveness of Antimicrobial Stewardship Teams (ASTs) on the reduction of morbidity and mortality associated with nosocomial bacteraemia. A decision analytic model compared costs and outcomes of bacteraemic patients receiving standard treatment with or without an AST consult. Patients with a bacteraemic event during their hospital admission were included in the model. Effectiveness was estimated as quality-adjusted life years (QALYs) over the lifetime of patients. Model variables and costs, along with their distributions, were obtained from the literature and expert opinion. Incremental cost-effectiveness ratios (ICERs) were calculated to estimate the cost per QALY gained from the hospital perspective. Uncertainty in ICERs was evaluated with probabilistic sensitivity analyses. The cost-effectiveness of clinical decision support systems was evaluated as a secondary analysis. Implementing an AST for bacteraemia review cost $39,737 (95% CI $27,272-53, 017) and standard treatment cost $39,563 (95% CI $27,164-52,797). The difference in effectiveness between the two strategies was 0.08 QALYs, and the base case ICER from the probabilistic analysis was $2367 per QALY gained [95% CI dominant (less costly, more effective) to $24,379]. Results from the probabilistic sensitivity analysis demonstrated there was more than a 90% likelihood that an AST would be cost-effective at a level of $10,000 per QALY. Maintaining an AST to improve care for bacteraemia is cost-effective from the hospital perspective. The estimate of $2367 per QALY gained for the AST intervention compares favourably with many currently funded healthcare interventions and services.