Pro 2020

A Markov decision model was developed with TreeAge Pro 2020 software (TreeAge, Williamstown, MA, USA) to simulate the disease process of patients with advanced endometrial cancer. The model structure comprised three mutually exclusive health states: PFS, progressive disease (PD), and death. All of the patients started in the PFS state and the cycle length was assumed as 1 month. Patients either stayed in the initial health status or progressed during each cycle over a lifetime horizon, as shown in Figure 1. To extrapolate the transition probabilities, the original data were extracted from the survival curves in the 309-KEYNOTE-775 trial by WebPlotDigitizer (Version:4.4; https://automeris.io/WebPlotDigitizer) (11 (link)), and these data were then used to fit parametric survival models using the algorithm derived by Hoyle et al. (13 (link)). Furthermore, we validated the results of transition probabilities using the formula: P (1 month) = 1 – (0.5) ^{(1/median time to event)}, which was derived from the equations: P = 1–e^−R and R = –ln [0.5]/(time to event/number of treatment cycles).

To understand differences in costs between each management strategy, a Markov model was developed to evaluate healthcare costs from a Medicare insurance perspective consistent with the CHEERS checklist and Second Panel on Cost-Effectiveness in Health and Medicine (13 (link)). A time horizon of 30 days (and no discount rate) was used to model immediate differences in reimbursement and to give greater weight toward immediate technical challenges and postoperative adverse events, recognizing that a longer time horizon would increasingly favor EUS-BD due to the need for tube exchange with PTBD. Cost-minimization analysis was performed using base-case assumptions to evaluate the primary outcome of cost associated with each strategy. Probabilistic sensitivity analysis was performed using a Monte Carlo of 10,000 trials to model uncertainty in overall cost estimates (reported as 95% confidence intervals for base-case outcomes). One-way sensitivity analysis was performed to assess how cost preferences might be affected by varying the expected rate of any necessary re-intervention following either EUS-BD or PTBD. Analyses were performed using TreeAge Pro 2020 (TreeAge Software, Williamstown, MA, USA).
All authors had access to the study data and had reviewed and approved the final manuscript.

We constructed a Markov model to evaluate the cost-effectiveness of FMT through colonoscopy in rCDI in a 1-year time horizon from the payer's perspective. A Markov model was used because of the high recurrence rate of CDI. We assumed the model consisted of five statuses, including first relapse, recovery, refractory CDI, recurrent CDI, and death (Figure 1) using TreeAge Pro 2020 (TreeAge Software, Williamstown, Massachusetts, USA). Fidaxomicin is administered in the event of recurrence following the initial vancomycin or after initial vancomycin failure (vancomycin arm). FMT is employed when patients experience a recurrence after the initial fidaxomicin or encounter initial fidaxomicin failure (fidaxomicin arm). In the case of FMT failure or the recurrence of CDI after the initial FMT cure (FMT arm), repeat FMT is considered.
FMT was compared to standard treatment with vancomycin (125 mg four times a day for 10 days) or fidaxomicin (200 mg twice a day for 10 days) in treating rCDI (8 (link)). In this study, the length of each cycle was 3 months. A half-cycle correction was applied to account for the fact that events in the model could happen at any point. Incremental cost-effectiveness ratio (ICER) and net monetary benefit (NMB) were used to conclude the results of the comparisons.

Given the merits of incorporating therapeutic effects over time without calculating transition probabilities,^{20 (link)} a 1-month cycle of a partitioned survival model was established using TreeAge Pro 2020 (TreeAge Software) with 3 mutually independent health states: PFS, progressive disease (PD), and death (eFigure 2 in the Supplement). In the model, the proportion of patients in each health state at each time point was determined from OS and PFS curves.²¹ The time horizon was 15 years given that more than 99% of the cohort died, and 5 to 30 years were included in the sensitivity analyses.

An economic evaluation in the form of a cost-utility analysis is conducted. Results are expressed as incremental lifetime cost per QALY gained of the MCAI compared to UC. A microsimulation Markov model developed in TreeAge (TreeAge Pro 2020, R2.1. TreeAge Software, Williamstown, MA) is used to simulate the lifetime natural history of Dutch patients with osteoporosis and is performed according to the Dutch guideline for economic evaluation and for costing, as well as the recent guideline of economic evaluations in osteoporosis from ESCEO-IOF [43 (link)–45 (link)]. Both univariate and probabilistic sensitivity analysis is conducted to handle uncertainty [53 ].

This study did not require institutional review board approval. We developed a state-transition model (Markov model) by using established modeling software (TreeAge Pro 2020, TreeAge Software, Williamstown, MA, United States). Our study was conducted from the perspective of the U.S. health care sector, and all costs were measured in United States dollars (USD) (24 (link)). Recommendations on discounting costs and outcomes at a discount rate of 3% were followed (25 (link)). Based on the WHO-CHOICE recommendations, the willingness-to-pay (WTP) thresholds were set at 1× the gross domestic product (GDP) per capita to indicate high cost-effectiveness and at 3× the GDP per capita to indicate cost-effectiveness (26 (link)). Based on 2021 data from the World Bank, the thresholds for the U.S. were $70,249 and $210,746 (27 ).