There is already a substantial amount of guidance and best practice available regarding development of statistical analysis plans (SAPs). This includes guidelines for SAP content in general and for early phase clinical trials in particular.
Guidance on SAP content emphasises the need for:
- Adequate planning prior to beginning any trial data collection.
- Where appropriate, defining trial estimands, i.e. specifically what values the trial is looking to estimate (given the near inevitability of some participation changes or ‘intercurrent events’).
- Robust planning for how missing data will be handled in the trial analysis, in alignment with any trial estimands, underlying assumptions and proposed missingness mechanisms.
- Sensitivity analysis targeting the same estimands should be performed to assess the robustness of the statistical assumptions made in the main analysis.
As with other trial documents, we suggest that the language and terminology used in SAPs reflects the complexity of participation changes (see principle O1). The SAP could also be clear about which types of participation change are most relevant to the trial analysis.
If it has been agreed that participants who want to stop taking part can continue to contribute with less commitment via alternative follow-up schedules or methods (see principle D1) then the SAP could cover any implications for how the study outcome data will be analysed (i.e. if some data was collected in different ways).
Given the available guidance elsewhere, we will not go into further detail about statistical planning here and instead would refer the reader to those other sources, for example:
- Mallinckrodt and colleagues: choosing estimands in clinical trials with missing data.
- Cro and colleagues: a four-step strategy for handling missing outcome data in randomised trials affected by a pandemic.
- Cro and colleagues: sensitivity analysis for clinical trials with missing continuous outcome data using controlled multiple imputation: a practical guide.