FDA Urges Analysis of Covariance for Randomized Control Trials

Drug sponsors can use complicated covariance analyses to help determine statistical significances or to help sharpen estimates of how well a treatment might work but they shouldn’t apply those analyses to anything in a clinical trial that might be affected by treatment, the FDA says in a new draft guidance document.

The six-page draft guidance is the FDA’s effort to round out recommendations in ICH’s E9 Statistical Principle for Clinical Trials. In it, regulators recognize that different patients have different disease prognoses and that drug sponsors would like to be able to incorporate different prognoses into statistical analyses of trial data.

Overall, analysis of covariance can “improve the power of significance tests and the precision of estimates of treatment effect,” but researchers should probably avoid factoring in any variables that could be affected by treatment, the agency says.