Article Text

PDF
SAT0563 Agreements and Discrepancies between the Food and Drug Administration (FDA) Reports and Journal Papers on Biologic Agents Approved for Rheumatoid Arthritis: a Meta-Epidemiological Study
  1. S. Tarp1,
  2. G. Amarilyo2,
  3. J. M. Woo3,
  4. W. Li3,
  5. H. Bliddal1,
  6. R. Christensen1,
  7. D. E. Furst3
  1. 1Musculoskeletal Statistics Unit, The Parker Institute, Department of Rheumatology, Copenhagen University Hospital, Bispebjerg and Frederiksberg, Denmark
  2. 2Dana-Dwek Children’s hospital, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel
  3. 3David Geffen School of Medicine, University of California, Los Angeles, United States

Abstract

Background As part of the review process of new drug applications in the USA, the Food and Drug Administration (FDA) performs an independent analysis after receiving raw data from clinical trials conducted by the sponsor. From this FDA review, a critical report is available to third parties, describing the FDA analysis and conclusions. Despite the availability of these critical data, the medical community derives much of its information from journal published papers.

Objectives To ascertain if there are differences between the FDA analysis and those data which are published, we undertook a direct comparison between FDA and published results of Biologics Agents (BAs) in RA, using ACR20 responses as the target outcome (often used in drug development).

Methods We obtained reviews of the approved BAs for RA available on the FDA Web site, and subsequently identified phase 2 and 3 trials. We then identified matching publications in scientific journals. From both sources we extracted ACR20s and calculated odds ratios (ORs) for all active trial arms vs. controls. To compare the ORs from the FDA review and the corresponding published paper, the Ratio of ORs (ROR) were calculated for all pair-wise comparisons when ACR20 was reported in both sources. The ROR was calculated as Exp{ln(OR[FDA]) - ln(OR[Pub])}. The anticipated correlation between the two lnORs measures was estimated empirically across all the pair-wise comparisons to estimate the corresponding variance of ROR.

Results RA trials were reported on 7 BAs with a total of 30 trials (2-6 trials in each FDA review) on the FDA Web site. All except one trial were published. Five trials were excluded from the analysis: in 2 trials ACR20 was reported in neither the FDA nor published paper; in 2 published papers, no ACR20 was shown and in 1 trial the FDA described no ACR20s. Out of these five, only the unpublished trial had ACR20 as the primary outcome. ACR20 was reported in 25 trials in both sources: in 21 trials the ACR20 was the primary outcome, 3 had other primary efficacy outcomes, and in 1, safety was the primary outcome. From the 25 trials, a total of 55 pair-wise comparisons were evaluated: 19 (34.6%) were in disagreement (9 were statistically significantly different, p<0.05). The FDA report showed more beneficial effect in 15 trials whereas efficacy was more favoured in 4 journal publications. The differences were related to analytic approaches (10 occasions), counting differences (1), rounding effects (4) or administrative matters (4). The differences did not change the overall outcomes.

Conclusions Discrepancies between results in published scientific journals and the FDA reviews occurred in one third of the submitted trials and comparisons. These differences can be attributed to multiple reasons. While real discrepancies occurred, there was no empirical evidence to suggest biased estimates between the FDA and published results. Increased transparency in publications would increase the credibility of published results.

Acknowledgements This study was supported by a grant from Danish Medicines Agency and unrestricted grants from The Oak Foundation.

*G. Amarilyo & S. Tarp contributed equally to this work.

Disclosure of Interest None Declared

Statistics from Altmetric.com

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.