2.2.8. Statistical Analysis

Analysis of the outcomes at baseline consist of the following measures. First, descriptive analysis of all of the variables (frequencies for categorical variables; means and standard deviation for continuous variables) will be carried out. A univariate analysis (one-way ANOVA and chi-square) will be used to examine the data and test whether there are baseline differences between groups after randomisation.

Statistical analyses will be chosen based on the sub-sample size (parametric or non-parametric tests). Data collection and statistical analyses will be performed using Excel software, SPSS software (version 25.0) [117], and the R statistical software environment (version 3.6.2) [118].

Clinical effectiveness analysis: The report of the results will follow a pre-specified plan based on CONSORT guidelines [119] in order to compare the two groups. Initially, a descriptive comparison (proportions, means, or medians) will be carried out between groups for prognostic variables in order to establish their baseline comparability after randomisation. To analyse the clinical effectiveness, a repeated-measure linear regression will be conducted, including all evaluations over time. For this purpose, the main variable, SF-36 score, will be used as a continuous variable. The models will include adjustments for the baseline value of the SF-36 and for any other variable that would show differences in the baseline measurement. Possible group per time interactions will be examined using linear regression. Similar analyses will be carried out using the secondary outcomes (number and severity of persistent symptoms, Montreal Cognitive Assessment, Sit-to-Stand Test, and HAD test). To counteract the problem of multiple comparisons we will use Bonferroni correction.

Cost-efficiency and cost-utility analysis are as follows. The effectiveness of the interventions will be estimated using the difference between the SF-36 score at the baseline and at the three and six-month follow-up, and the utility will be estimated using QALYs at the three and six-month follow-up. QALYs will be calculated based on these scores using the Spanish EQ-5D tariffs [120]. Along with the number and severity of the ongoing symptoms, scores will also be used as an outcome for the analysis. Cost-efficiency will be explored through the calculation of incremental cost-efficiency ratios (ICERs) for the intervention group using the TAU group as the control. ICER is defined as the ratio between incremental costs and incremental effectiveness. In the same way, cost utility will be explored through the calculation of incremental cost-utility ratios (ICURs), which are defined as the ratio between incremental costs and incremental utilities measured on QALYs. QALYs gained in each evaluation are approximated by using the area under-the-curve technique [121]. Total costs will be calculated by adding direct and indirect costs. Direct costs will be calculated by adding the costs derived from the medication and the use of health services and clinical tests. The medication costs will be calculated by determining the price per milligram during the study period according to the Vademecum during the last year of the study, including value-added tax (VAT). The total cost of drug treatment will be calculated by multiplying the price per milligram by the daily dose in milligrams and the number of days the treatment is received. Costs derived from the use of health services will be calculated considering the data from the Oblikue database [122]. Indirect costs will be calculated based on the number of days taken on sick-leave and multiplying them by the Spanish minimum daily wage during the study period 2019–2020. We assume that data will be missing at random (MAR). Only patients with both cost and relevant outcome data at three and six-month follow-up will be included in the cost-efficiency and cost-utility analyses. Notwithstanding, sensitivity analysis imputing missing three and six-month data will test the robustness of the cost-efficiency and cost-utility results. The imputations will be performed using the package “mice” [123], freely available in cran-R [118].

Variables collected, instruments, and measures are shown in Table 1.

Study variables.