Individual participant data meta-analysis tips and tricks: troubleshooting commonly encountered issues of contacting...

We present a framework to guide researchers in contacting and retrieving trial data from trialists. This framework serves 2 purposes: i) to provide a consistent and transparent approach for contacting authors, and ii) to describe how to record clearly all contact attempts and to identify trialists who have not responded to reasonable attempts at communication. This framework will help researchers identify trials that may require investigation for data integrity issues.

Individual participant data meta-analysis (IPD-MA) is considered the gold standard for evaluating clinical interventions. The popularity of IPD-MA has increased due to the potential for advanced statistical analyses and the ability to test data veracity prior to analysis. Contacting trialists and requesting data is the most time-intensive step in an IPD-MA project. Often, many datasets are not retrieved, as authors are uncontactable or do not share data. This absence of IPD can bias meta-analysis and interpretation of results. Currently, there is no framework in place to guide researchers in contacting trialists and to define a reasonable point to cease communication attempts.

The framework consists of 4 approaches: first, contacting the listed authors on the trial publication; second, contacting the trialists' associated institutions (hospitals and universities); third, contacting colleagues from similar regions within a particular country; and fourth, contacting the journal in search of trialists' contact details. If trialists do not respond to sustained communication attempts, their study should be classified as "non-responding." Depending on the trial context, non-responding trialists or those who respond with concerning reasons why data are unavailable may be subject to further review regarding trial quality and data integrity concerns.

This framework was applied in an IPD-MA comparing misoprostol with oxytocin for the prevention of postpartum hemorrhage, which included 79 randomized controlled trials. With the use of this framework, trialists from 10 trials responded to the IPD invitation and contributed data (6 of which were used in final analysis); 38 trialists responded but did not contribute data; and 31 trialists did not respond and their trials were subsequently classified as non-responding.

The proposed framework provides a uniform structure for contacting authors and requesting data for IPD-MA, which may increase the likelihood that trialists will respond to IPD-MA invitations. This framework will also help to identify trialists who do not respond to reasonable attempts at communication.