How covariate control can bias our insights into brain architecture and pathology

Inferential analysis of normal or pathological brain imaging data—as in brain mapping or the identification of neurological imaging markers—is often controlled for secondary variables. However, a rationale for covariate control is rarely given and formal criteria to identify appropriate covariates in such complex data are lacking. We investigated the impact and adequacy of covariate control in large-scale imaging data using the example of stroke lesion-deficit mapping. In 183 stroke patients, we evaluated control for age, sex, hypertension, or lesion volume when mapping real or simulated deficits. We examined (i) the impact of covariate control when mapping cognitive poststroke deficits, (ii) the association of covariates and brain pathology (iii) the impact of covariate control when mapping simulated deficits, and (iv) the impact of covariate control under a verified null hypothesis. We found that the impact of covariate control varies across covariates and deficits in both real and simulated data. However, simulations showed that covariate control does not necessarily improve the precision of lesion–deficit inference. Instead, it systematically reshapes statistical maps according to the associations between covariates and lesion anatomy. As a result, covariate control can bias statistical inference and, under a verified null hypothesis, may even generate spurious associations. These findings suggest that the widespread use of covariate control in clinical brain imaging—and likely other biological high-dimensional data—should be reconsidered, as it may introduce substantial analytical flexibility without necessarily improving inference.