371 [18F]FDG-PET/CT: cost-utility analysis alongside an RCT 6 over the first year of the EfFECTS trial or from literature, where appropriate. Yearly costs for other non-thyroid-related health care consumption, informal care, and other productivity losses were estimated from the reported first-year cost-questionnaire data in our study, using restricted linear regression analysis with age, sex, and QALYs as predictors (restricting coefficients to predict nonnegative costs) (Supplementary data, table 1). Travel expenses were estimated from the number of hospital visits for each procedure or health state, and the patient-reported travel distance. Utility parameters Utilities were calculated, starting from age and sex dependent general utilities [474], by subtracting disutilities for specific health states (Table 3). These disutilities were derived from literature or elicited from the previously mentioned expert panel based on a time-trade-off weighting. QALYs were calculated by the discounted sum of utilities over the lifelong evaluation period. Other parameters A 4% and 1.5% discount rate was applied to all future costs and utilities, respectively [477]. In addition to the base-case values, distributions were specified to account for the uncertainty in the parameters. These were either triangular parameter distributions (on a specified range, with mode equal to the base-case value) or normal distribution (with specified SD and mean equal to the base-case value). Lifelong extrapolation With each of the 100 imputed 1-year datasets, 10 sets of model parameter values were drawn at random from the specified parameter distributions. Then, for each of the 1000 parameter sets and starting from each patient’s health state at the end of the first year, the Markov model was used to simulate 1000 extrapolated patient histories. For each parameter set, the average over the extrapolated costs and QALYs was added to the 1-year costs and QALYs, as an estimate of the patients’ expected lifelong outcomes. Statistical analysis Baseline characteristics were compared between the allocated groups using Pearson’s chi-squared or Fisher’s exact tests for categorical data, and independent samples t-tests or Mann-Whitney U tests for continuous data, where appropriate. Univariate comparisons of the 1-year costs and QALYs were performed using independent unequal-variances t-tests, aggregating the 100 multiple imputation sets using Rubin’s rules (accounting for sampling and imputation uncertainty). Similarly, lifelong costs and QALYs were compared by aggregating the 1000 parameter sets using Rubin’s rules (accounting for sampling, imputation and parameter uncertainty) [576]. Unadjusted (univariate) results are presented in the Supplementary data. In the analyses presented here, we adjusted for the trial’s stratifying variables using a generalized linear model with robust estimator for observed
RkJQdWJsaXNoZXIy MTk4NDMw