Cortical morphometry might predict currently prescribed vs. discontinued medications in bipolar disorder, even after controlling for the cumulative dose effects: An ENIGMA mega-analysis

Recent research suggests that brain anatomy may help identify the most effective pharmacological treatment for each individual with bipolar disorder and reduce trial-and-error prescribing. We aimed to investigate whether brain anatomy predicts whether a medication is currently prescribed or has been discontinued, as a proxy for treatment effectiveness. The rationale is that medications that provide clinical benefit without unacceptable side effects are likely to be continued, whereas those with limited benefit or poor tolerability are typically discontinued. We used T1-weighted MRI from twelve ENIGMA-BD cohorts (n = 2462; 473 individuals with BD [61% female, age 18–73] and 1989 controls) to derive regional cortical thickness and surface area and subcortical volumes. Site differences were harmonized using ComBat models fitted on controls’ data. Within cross-validation, models were trained to first adjust for cumulative dose and other covariates and then predict medication status. On test sets, current prescription (vs. discontinuation) of lithium was predicted by greater cortical thickness and reduced surface area, whereas current prescription (vs. discontinuation) of antidepressants and atypical antipsychotics was predicted by greater cortical thickness. Predictive regions for atypical antipsychotics were generally consistent across subgroups of age, gender, illness duration, and history of psychosis, and in the largest site, and differed from those associated with cumulative effects of medication on the cortex. Predictions were poor for subcortical volumes and for antiepileptic mood stabilizers and typical antipsychotics. These findings provide preliminary support that cortical anatomy may help inform future development of biomarkers for treatment selection, pending validation in longitudinal studies.