Sample size
The study design was based on Burris et al. [
64] using the time to deterioration (TTD) on the EORTC QLQ-C30 with comparable population and treatment for the global HRQoL (primary objective) and fatigue symptom domain (co-primary objective).
The overall two-sided alpha of 0.05 will be split among the co-primary endpoints, with 0.04 for time to definitive deterioration of HRQoL (TTD1) and 0.01 for time to definitive deterioration of fatigue (TTD2). We estimate that about 244 randomized participants with a minimum of 189 deterioration events observed will be required to achieve a 90% statistical power to detect an improvement in median TTD1 from 8 (control) to 13 months (intervention) (HR = 0.615), using a two-sided alpha of 0.04, and assuming a 30-month accrual and 18-month follow-up, and an 1% exponential dropout rate. This sample size will permit to detect the same improvement in median TTD2 from 8 to 13 months with a 77% power and using a two-sided alpha of 0.01 (nQuery Advisor V7.0).
Statistical methods
Analyses will be performed on the intention-to-treat (ITT) population including all randomized participants analyzed according to the randomization scheme. Due to the lack of data (HRQoL) studies, a sub-population will be defined for the analysis of HRQoL, corresponding to a modified ITT population, i.e., all ITT patients with at least the available HRQoL score at baseline.
Qualitative variables will be described using frequency and percentage distributions. The number of missing data will be given, but will not be considered for the calculation of proportions.
Quantitative data will be described using the number of observations, mean, standard deviation, median, minimum and maximum values.
For the primary endpoint analysis, the time to definitive deterioration (TTD) will be defined as the time interval since randomization to the observation of the first deterioration (decrease for global HRQoL score and increase for fatigue score) of 5 points at least as compared to the baseline score, with no further improvement of 5 points at least as compared to the baseline score [
49,
64]. Patients with no deterioration observed during the study will be censored at the time of the last available score. Patients with no follow-up score will be censored 1 day after their inclusion. If a deterioration is followed by missing data, this deterioration will be considered as definitive. TTD will be estimated using the Kaplan-Meier method and compared between groups with a log-rank test, stratified by the presence of visceral metastases, the method of administration of chemotherapy and the 6MWD, at a two-sided significance level of 0.04 (TTD1) or 0.01 (TTD2). Univariate and multivariate Cox regression models will be performed to explore factors influencing the TTD.
Sensitivity analysis will be performed to assess 10-point deterioration in global HRQoL score and fatigue in comparison with baseline. Sensitivity analysis will also be performed including death as an event.
An analysis of missing data profile at baseline and during follow-up will be performed in order to determine the missing data profile. The responders versus non-responders for HRQoL questionnaire at baseline will be compared according to clinical and socio-demographics variables collected at baseline in order to determine a potential missing at random profile (i.e. missing data depending on observable variables). To investigate a potential missing not at random profile (i.e. missing data depending on non-observable HRQoL level), these profiles of participants (responders vs. non responders at baseline) will be compared according to the overall survival. A graphical analysis of HRQoL data over time until drop out will be performed in order to explore a potential missing not at random profile for missing data during follow-up.
Multiple imputations could be explored for baseline missing HRQoL data, taking into account the variables linked to the occurrence of baseline missing HRQoL data and HRQoL analyses will be repeated then after as sensibility analyses.
For secondary endpoints analysis, progression-free survival will be measured from the date of randomization until the date of event defined as the first documented progression or death from any cause. Participants with no event at the time of the analysis will be censored at the date of the last available tumor assessment. Overall survival and progression-free survival will be estimated using the Kaplan-Meier method and will be described in terms of median progression-free survival per arm, and hazard ratio for progression between the two arms. Associated 2-sided 95% confidence intervals for the estimates will be provided.
Change from baseline will be compared between the two arms: secondary HRQoL dimensions, multidimensional dimensions of fatigue, physical fitness, anthropometric measurements, PA level, sleep disorders, sarcopenia, chemotherapy toxicities, inflammation and oxidative stress, personality factor and process of change regarding PA.
The proportion of refusal (number of participants who refused to participate in the study / number of participants to whom the study was proposed) will be calculated.
Statistical analyses will be performed using SAS® software version 9.4 or later.