Statistical Considerations for Bias and Protocol Deviation in Medical Device Pivotal Clinical Study

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Medical Device Evaluation

Background:

• Adherence to study protocols is crucial for reliable clinical trial outcomes.
• Protocol deviations, common in medical device studies, can introduce bias.
• Selection bias, particularly in diagnostic test trials, requires careful statistical management.

Purpose of the Study:

• To discuss protocol deviation issues in pivotal medical device clinical trials.
• To provide statistical approaches for mitigating selection bias, specifically verification bias.
• To illustrate bias correction methods using real-world diagnostic test examples.

Main Methods:

• Developed bias correction methods for prescreening and verification bias.
• Applied statistical approaches to estimate device performance adjusted for enrollment enrichment and discrepant testing.
• Utilized Roche Cobas Human Papillomavirus (HPV) test and a Microarray Assay for illustrative examples.

Main Results:

• Unadjusted sensitivity and specificity for the HPV test were 83.2% and 60.4%.
• Verification bias-adjusted sensitivity and specificity, using a missing-at-random assumption, were 34.5% and 93.6% respectively.
• Demonstrated accurate estimation of sensitivity and specificity through discrepancy analysis.

Conclusions:

• Protocol deviations can significantly bias estimates of medical device clinical performance.
• Appropriate statistical methods are essential for correcting bias introduced by protocol deviations.
• Implementing bias correction techniques is recommended for accurate device performance estimation in clinical studies.