Evidence Summary

What is it?

Modified cover letter sent with trial questionnaires. The letter is created as a behavioural change intervention using psychological theory, the behaviour to change being return of the questionnaire.

Does it work?

Sending a theory-based cover letter probably increases retention.

How big is the effect?

An increase of 4% (95% confidence interval = -1% to 10%).

How certain are we?

GRADE Moderate certainty.


We recommend that trialists use theory-based cover letters to increase retention in trials that use questionnaires.

How can I use this straight away?

See Resource bundle below for the theory-based cover letter template and other support materials

Practical Impact

Imagine initial retention is 65% of those approached. You have a trial with 100 participants that needs responses from 80 to meet its statistical power calculations.  Retention of 65% means that you will be 15 responses short (see chart below).


Now imagine using theory-based cover letters. The chart below shows the impact of an absolute increase of 4% (95% CI = -1% to 10%). Retention is now 69%, which means our best estimate is that you would now only be 11 responses short.

Cumulative Meta-Analysis*

*Random effects model done using Comprehensive Meta-Analysis v3 (www.meta-analysis.com).
Differences >0% favour the intervention. The GRADE assessment is moderate because of imprecision.

Resource Bundle


How to Cite

Citation: Treweek S, Gillies K. Evidence pack– Retention: Theory-based cover letter (Ret1), 2020, https://www.trialforge.org/retention-sector/theory-based-cover-letter/.


More Information

  1. This summary is from the Cochrane review of strategies to improve retention in randomised trials (https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.MR000032.pub3/full).
  2. The ‘Does it work?’ statement is structured according to effect size and GRADE certainty as per GRADE Guidelines 26 (https://doi.org/10.1016/j.jclinepi.2019.10.014). The statement is for moderate effect size and High GRADE certainty
  3. The IQuaD and INTERVAL data are from Goulao et al 2020 (https://doi.org/10.1016/j.jclinepi.2020.01.018). AMBER and OPAL are presented in Duncan et al 2019 (PS1A– https://doi.org/10.1186/s13063-019-3688-6).
  4. The recommendation statement is the consensus view of the authors of this summary based on the GRADE certainty and features of the trials contributing to the evidence.
  5. If you have any questions contact info@trialforge.org.
Version: v1.0 - 26/3/2020

Start typing and press Enter to search