Data sharing: experiences accessing individual participant data for a collaborative meta-analysis on umbilical cord management strategies at preterm birth.
Meta-analysis (MA) using individual participant data (IPD) is regarded as the gold standard, allowing more complex analyses (e.g., subgroup analysis), verification of published findings, inclusion of unpublished data, and some outcome harmonisation. Although researchers and funding agencies show strong in-principle support for data sharing required to perform such projects, rates of actual data sharing are frequently low.
Objectives: To document the experiences of obtaining IPD for a large individual participant data on Cord Management at Preterm birth (iCOMP), a large collaborative IPD and network MA in neonatology.
Methods: Lead and senior investigators for eligible trials were approached and asked to share IPD. We recorded the number (%) of trials in which the trialist responded favourably/declined or did not respond, the number of trials that actually shared their IPD, methods used, and barriers in sharing IPD.
Results: Of 118 eligible studies identified for inclusion in iCOMP, 82 (69.5%) agreed in-principle to share IPD. Three (2.5%) trials were uncontactable, and 36 trials (30.5%) did not respond even after multiple contact attempts. Among those who responded, 14 trialists (17%) declined or were unable to share data, frequently citing data access issues because of the trial’s age. Twenty-one trials (17.8%) did not share data despite initial agreement. We received data from 61 (51.7%) trials. A median of 23 (IQR 25) emails were sent from data request to receipt. The longest data acquisition period for a study was 831 days. Sixty datasets were received directly from the trialist and one dataset in the publication appendix. Data were shared in different formats with Microsoft Excel the most common. Additional barriers included lengthy processes for data sharing agreements and ethics requirements.
Conclusions: Data sharing is a resource-intensive and repetitive process which is often plagued by challenges. Streamlined processes and alternative arrangements, such as depositing into a data commons or data trust needs to be developed to better facilitate future collaborative IPD-MA. Sharing experiences in running and collecting data for IPD MA enables mitigating similar challenges in future projects. Patient, public, and/or healthcare consumer involvement: A consumer representative was part of our advisory group and authorship team.