Patterns of Regularity Noncompliance Identified by the U.S. Food and Drug Administration and Their Effects on Meta-analyses

Authors

Craig A. Garmendia, Florida International UniversityFollow

Document Type

Dissertation

Degree

Doctor of Philosophy (PhD)

Major/Program

Public Health

First Advisor's Name

Purnima Madhivanan

First Advisor's Committee Title

Major Professor

Second Advisor's Name

Mary Jo Trepka

Second Advisor's Committee Title

Committee member

Third Advisor's Name

Emir Veledar

Third Advisor's Committee Title

Committee member

Fourth Advisor's Name

Stanislaw F. Wnuk

Fourth Advisor's Committee Title

Committee member

Keywords

Bioresearch Monitoring, Code of Federal Regulations, deviation, Form FDA 483, inspection, clinical investigator, disqualification, regulatory action, warning letter, meta-analysis

Date of Defense

9-20-2018

Abstract

The objective of this study was to determine the patterns of regulatory noncompliance, as identified by the U.S. Food and Drug Administration (FDA), and their effects on meta-analyses. In order to achieve these objective, three studies were undertaken: analysis of citations issued by FDA Investigators at the conclusion of an inspection; analysis of regulatory actions taken by the FDA towards clinical researchers based on the observations cited by FDA Investigators; and sensitivity analysis of meta-analyses based on the Agency’s determination of research misconduct, primarily the falsification of data. FDA Investigator citations were analyzed using Chi-Square analysis based on geographic location of the inspection, type of inspection, and type of violation. Temporal changes in the number of inspections and the violations cited were analyzed using bivariate Poisson regression models. Bonferroni correction was employed for temporal changes across the time period analyzed. Regulatory actions taken by the agency were analyzed via Chi-Square or Fisher’s exact test based on changes identified in previous publications, temporal changes, and differences between regulatory action types. Sensitivity analysis of meta-analyses identified through a systematic review were assessed both qualitatively and quantitatively for the effects of including publications of apixaban trials with significant FDA regulatory action, i.e. the comparison of odds ratio point estimate, upper and lower 95% confidence interval, both before and after consideration of falsified data.

Under the FDA’s Bioresearch Monitoring program from 2007-2015, the number of inspections increased, but the rate of citation issuance per inspection decreased. One third of the violations were related to adherence to investigational procedures followed by informed consent violations and violations involving study records. During this same time period, 194 clinical researchers received a regulatory action based on FDA’s review of inspection results. Since 2007, rates of significant deviations had decreased. Lack of researcher supervision and submission of false information were cited more frequently for disqualification proceedings. A systematic review found 99 statistical analyses from 22 different meta-analyses available for sensitivity analyses. Nearly one-third resulted in a change in the conclusions reported in the originally published statistical analyses.

In approximately the last decade, the number of violations cited during inspections under the Bioresearch Monitoring program has decreased; however, significant improvements can continue to be made regarding adherence to study procedures, the consenting of human subjects, and creation of adequate and accurate study documentation. Disqualification of clinical researchers is more likely to occur when researchers fail to supervise a clinical trial or false information is submitted to the FDA. Falsified data can make its way into the exploding field of meta-analyses, a study method that provides a concise and compelling method for the dissemination of medical intervention knowledge; however, this method can be highly unstable and can provide biased results. A robust sensitivity analysis that considers data quality from available sources can help ensure calculations of the best estimates.

Identifier

FIDC006987

ORCID

http://orcid.org/0000-0002-7099-1810

Recommended Citation

Garmendia, Craig A., "Patterns of Regularity Noncompliance Identified by the U.S. Food and Drug Administration and Their Effects on Meta-analyses" (2018). FIU Electronic Theses and Dissertations. 3920.
https://digitalcommons.fiu.edu/etd/3920