Limited reporting of major harms in studies of initial combination antiretroviral therapy: a systematic review.

Objectives: Risk-benefit assessment of combination antiretroviral therapy (cART) requires consideration of all potential serious harms. Studies of initial cART may permit identification of associations between particular regimens and uncommon harms, but only if comprehensively reported in the public domain.

Design: Study-based, systematic review of published initial cART studies (in adult patients) for completeness of serious harms reporting.

Methods: Electronic databases, abstracts, and regulatory/sponsor reports were searched (1 January 1996 - 31 December 2012). Reporting of pre-specified harms - deaths, new/recurrent AIDS events, serious non-AIDS events (2010 INSIGHT classification) and serious adverse events (SAEs) - were assessed as the proportion of studies providing data (reporting frequency). Pharmaceutical sponsors were approached for unreported data.

Results: 103 studies (86% randomized, 54% industry-sponsored) were included. Deaths, AIDS events, serious non-AIDS events and SAEs were reported for 85 (83%), 55 (53%), 26 (25%) and 43 (42%) studies, respectively. Deaths were better reported for academic than industry-sponsored studies (91 vs. 75%; P = 0.03); the converse applied for SAEs (26 vs. 55%; P = 0.002). SAEs were better reported for randomized than cohort studies (46 vs. 14%; P = 0.03), and for phase 3 than phase 2 or 4 studies (58 vs. 32 and 29%, respectively; P = 0.02). SAE reporting increased over time [ρ = 0.704, P = 0.002 (Spearman)]. Unreported data acquired for 34 (61%) of 56 industry-sponsored studies improved ascertainment in these studies to between 82 and 100% (P < 0.001).

Conclusion: Public domain reporting of serious harms for initial cART studies is limited. Insufficient data exist to determine if particular ART drugs/regimens are associated with most serious harms.