Risk Factors in Clinical Trials Software Development

In part in response to the FDA’s current good manufacturing practice (CGMP) initiative, which places risk assessment at the center of product quality regulation, DQRI established a project to investigate risk-based approaches to validating software for clinical trials. It is the hypothesis of the project that not all types of software need to be validated to the extreme levels required for critical applications, and that by identifying a defensible rational quantifiable model, resources currently used for validation activities can be better deployed elsewhere. The project defines the types of software commonly used in conducting clinical trials, and then explores approaches to identifying the risks associated with each type of software. First, a coarse-grained intuitive approach to assigning risk is taken, dividing the software types into high, medium, and low risk categories. Secondly, for a subset of high and medium risk software types, the types of failure that could occur is compiled, along with the most likely consequences of the failure. Finally, the project creates a quantifiable model to assigning risk, severity, and likelihood of occurrence, with the goal of developing tools to be used prospectively in determining the appropriate approach to validation for each software type.

White paper outlining the project’s research into risk models, and the value it adds in determining the extent of validation necessary for clinical research software. It will be submitted for the FDA’s consideration when they redesign 21 CFR Part 11 and/or additional guidance documents.

Sue Carroll

DQRI Comments to the FDA Regarding 21 CFR Part 11, Docket No. 2004N-0133 (download pdf file)

Data Quality in the Biomedical Field (more info)