Statisticians

This includes advising and guiding others to help ensure that the study is run in ways that will support the planned analysis.

When the study is analysed, Statisticians need to consider how participants have changed their involvement, particularly where this means study outcome data could not be collected.

They should also think, during study design, about what sorts of participation changes are most likely. They should consider what specific research question they are aiming to answer, given the possible participation changes.

The principles are coded with O for ‘overarching’ principles, D for study design and setup issues, M for data management and monitoring, and R for study analysis and reporting. See the main PeRSEVERE principles page for more details.

• The principle O1 explains how participation changes can be complex, rather than a simple matter of participants still taking part or having ‘withdrawn’. Often, there might be various ways that participants might change their involvement in a study. This is a key idea in the PeRSEVERE project, and it’s important that everyone understands it, including Statisticians and those they work with.
• Principle O3 encourages everyone involved in research studies – including participants – to bear in mind that collecting more of the planned data (particularly study outcome data) supports study quality. Among Statisticians, this might be more commonly termed ‘avoiding missing data’. It does not mean this consideration takes precedence above all others, but that it is not forgotten.
• Principle O4 says that researchers and participants losing contact with each other is not the same as participants saying they want to stop or reduce their involvement. This means research teams should consider them separately in study protocols, during study management and in study analyses. With some exceptions, it also means it should not be out of the question to try to recontact participants who previously lost contact near the end of a study, or attempt to collect more study data about them, if it is still useful to the study.
• Principle O5 supports study quality by saying that data collection continues until a participant says they want it to stop. We set out some conditions for how to apply this approach fairly and transparently.
• Principle O6 confirms the well-established idea that data already collected at the time a participant says they want data collection to stop should be retained and used in the study analysis. This usually means data already gathered by that time but not yet put onto study forms can also still be collected.
• Principle D1 says that some flexibility and resilience should be built into study protocols, allowing participants to continue taking part with reduced commitment, if they want to. Statisticians have a key role in deciding what level of flexibility is allowable without compromising study integrity. Statisticians can also help reduce participant burden by advising on the minimum data required to meet study objectives.
• Principle D2 is about study protocols being clear on how different sorts of participation change should be managed. Protocols and related documents should guide those running the study to manage participation changes in ways that will support the study analysis and any defined ‘estimands’. Protocols should also explain (along with the statistical analysis plan) how the study will be analysed to give the best chance of a robust result, even with participation changes and resulting ‘missing data’. Statisticians have a key role in ensuring the protocol is clear in these respects.
• Principle D6 covers adequate training and support to help people manage participation changes. This includes training for Statisticians in their role handling ‘missing data’ in analyses. It also includes guidance and support for all those running studies alongside Statisticians, to understand how best to support the planned analysis.
• Principle M1 says that good quality data should be collected about participation changes, including to inform the planned analysis. Good quality data should make clear how participation has changed, when it changed and – if known – why. It can also help inform Statisticians’ approach to dealing with any missing outcome data. Statistician should be involved in deciding what data to collect for a study.
• Principle M2 says that information about participation changes should be monitored during a study. Statisticians may be involved in this monitoring, reviewing or summarising data for others to review. This regular review might highlight trends in participation changes early so that suitable action can be taken to protect study integrity.
• Principle R1 says that studies should be analysed in ways that give the best chance of a robust result, even where participants have stop taking part or reduced or changed their involvement. This covers the idea of ‘handling missing data’ in study analyses.
• Principle R2 says that end of study reports should include clear data about participation changes. This broad aim is already covered by the CONSORT statement (Consolidated Standards of Reporting Trials). We suggest further improvements to this standard.

• Our guidance on:
  • Grant applications, protocol development and statistical planning
  • Risk assessment and monitoring
  • Training and support
  • Data collection about participation changes
  • Analysis and reporting

• Checklists for managing individual participation changes, to ensure all required steps are completed.
• Suggested terminology for talking about participation changes (in any context).
• Example scenarios showing how the PeRSEVERE principles can help manage – or avoid – various issues relating to participation changes.