FDA updates guidance on covariate treatment in clinical trials

In a revised draft guidance, the US Food and Drug Administration (FDA) has clarified how drug developers should adjust for covariates in certain clinical trials.
 
The revision provides “more detailed recommendations for the use of linear models for covariate adjustment and also includes recommendations for covariate adjustment using nonlinear models,” according to FDA’s Federal Register notice of the newly revised draft.
 
The updates to the April 2019 draft leave untouched the other recommendations for adjusting for covariates in randomized clinical trials in drug and biologics development programs. “The main focus of the guidance is on the use of prognostic baseline factors to improve precision for estimating treatment effects rather than the use of predictive biomarkers to identify groups more likely to benefit from treatment,” wrote FDA in its introduction to the guidance. (RELATED: FDA drafts guidance on adjusting for covariates in randomized trials, Regulatory Focus 24 April 2019)
 
The use of covariates to control for confounders for trials that are not randomized and covariate adjustment in analysis of longitudinal repeated measures are both outside of the scope of the draft guidance, the agency clarified.
 
Although the International Council for Harmonization (ICH) issued a 1998 E9 guidance on statistical principles for clinical trials, FDA’s newly revised draft guidance gives additional depth to the questions of covariate adjustment using both linear and nonlinear models. The guidance provides specific recommendations on statistical techniques that can be used “to increase precision” without corrupting the study’s analysis of the treatment effect in question.
 
The draft recommends adjusting for baseline covariates when efficacy endpoints are analyzed: Although an unadjusted analysis is acceptable for the primary analysis, adjustment for baseline covariates will generally reduce the variability of estimation of treatment effects and thus lead to narrower confidence intervals and more powerful hypothesis testing,” wrote the agency.
 
Both linear and nonlinear models are addressed in the guidance. For some of the complex statistical techniques that a nonlinear model may require, FDA consultation about a specific approach is best accomplished early in the game, suggests the guidance. There is a greater risk of incorrect estimation of the treatment effect “if the model is misspecified and treatment effects substantially differ across subgroups,” wrote FDA.
 
For nonlinear regression models, FDA gives a stepwise approach for one “statistically reliable” method of covariate adjustment. The draft guidance also includes a list of references.
 
Draft FDA guidance