Biomarkers – Making the Appropriate Selection for Clinical Trials

Biomarkers are used in the diagnosis, screening, monitoring and prognosis of health or disease. They are used to predict outcome and to detect recurrence of disease. Use of imaging biomarkers raises issues of quality of the image under various conditions or in various hands (operators, readers). It is important to identify important (and sometimes even less important) sources of variation and to measure them carefully.

Prognostic biomarkers define patient subgroups with better or worse prognosis regardless of treatment. Predictive biomarkers define subgroups with different responses to treatment. Companion biomarkers define subgroups with better or worse prognosis relative to particular treatment(s). They are in vitro diagnostic medical devices that are linked to a therapeutic treatment. So, “companion biomarker” has a usage in the specific context of regulatory approvals.

The first problem, of course, is to determine which biomarker(s) will be useful in subsetting the patient population. Biological considerations usually determine the candidates. Difficulties often arise because prognostic and predictive factors are numerous and often confounded in the population under study. For this case, a biomarker discovery model was developed using COXEN (COeXpression ExtrapolatioN) and subsequently validated in a separate study. COXEN is a computer algorithm that predicts the effectiveness of a chemotherapeutic agent for a particular cell type based on the known in vitro results of the agent on a battery of cancer cell types. It is still considered a research vehicle and not to be used prospectively to choose drugs for particular patients. It doesn’t appear that this algorithm can predict the effectiveness of a totally new agent, only that of a known agent on a new cell type.

The FDA requires direct evidence of clinical utility and benefit from the test for a specific therapy. In terms of pre-analytic consistency and reproducibility, the sample collection time, sample mixture, and technical variability must be specified. There must be a thorough evaluation of analytic variability and consistency. A clinical trial must examine drug-specific predictability of the biomarker.

A future article will focus on companion biomarkers, tests that predict whether or how much more or less a specific treatment will benefit a specific patient.

The author is a Pharma consultant in The CECON Group network and has extensive experience in regulatory statistics in both clinical and preclinical areas. For 25 years, he was a statistical reviewer, supervisor, and manager in the Center for Drug Evaluation and Research (CDER) of the FDA. His expertise includes drug quality, stability, clinical trials, and data integrity. Click here to read more about this regulatory statistician consultant and expert witness.