Data retrieval and harmonisation for individual participant data meta-analysis in the context of a neglected tropical disease: challenges and lessons from the Infectious Diseases Data Observatory visceral leishmaniasis data platform.

INTRODUCTION: Individual participant data meta-analysis (IPD-MA) is the gold-standard approach for evidence synthesis. Poverty-related infectious diseases, a context in which standalone trials often have small sample sizes and consequently observe relatively few treatment failures, can benefit from this approach. METHODS: In collaboration with the global research community, the Infectious Diseases Data Observatory (IDDO) has developed a data platform for visceral leishmaniasis (VL)-a neglected tropical disease (NTD), primarily to facilitate IPD-MA aimed at addressing questions of public health importance. Initially, a systematic review (SR) was undertaken to identify relevant trials since 1980 and study authors were invited to collaboratively develop the platform and participate in two IPD-MAs. IPD non-retrievability (data loss) was tracked and defined as unsuccessful attempts to contact study authors, or when the authors explicitly stated data loss, or declined participation. RESULTS: The SR identified 147 VL trials (1983-2019), of which IPD was shared to the IDDO platform from 31 (21.1%). Reasons for data non-retrievability of 116 studies included: authors' retirement (n=42, 36.2%), data were lost (n=30, 25.9%), authors' non-response (n=22, 19.0%), authors were no longer contactable (n=17, 14.7%) and a lack of clarity over data ownership (n=5, 4.3%). The median sample size was 226 (IQR: 120-542, range: 30-1,143) for retrieved studies compared with 81 (IQR: 34-151, range: 7-3,126) for non-retrieved studies. No IPD were retrieved from trials published pre-2000 (0%, 0/63). Following solicitation, the median time to complete data harmonisation was 633 days (n=29 studies). CONCLUSIONS: Data solicitation and harmonisation steps remain underappreciated challenges in undertaking IPD-MA, particularly for NTDs where resources are severely limited. A critical mass of IPD from historical VL trials is no longer retrievable. It is imperative that further loss of such valuable data from ongoing and future trials is prevented.