Predicting response to cognitive training for schizophrenia using results from two studies with different outcomes

Collaborative data sharing between research groups provides an opportunity to explore the basis for the heterogeneity in cognitive training outcomes reported in the schizophrenia literature. The current analyses focused on the contribution of site and participant characteristics to these heterogeneous outcomes.

Data from two independent studies, from New York (NY) and Los Angeles (LA), were combined to yield a sample of 132 outpatient adults with schizophrenia/schizoaffective disorder. While similar treatment doses, cognitive exercises and outcome measures were used, sites differed in use of coaching, group discussion and compensation. Between-site differences in participant demographic and baseline clinical characteristics were tested. Regression examined predictors of change in cognition (MCCB) and functional capacity (UPSA) which could explain site differences in treatment effects.

Medium to large treatment effect size differences in MCCB and UPSA favored the NY site over LA. When the studies were combined, the effect of site was significant for both outcomes with a medium effect size difference. After controlling for background characteristics, the effect of site was reduced for both outcomes, but remained significant for cognition. Improvement in UPSA was associated with better baseline MCCB (p < 0.001), lower baseline UPSA (p < 0.001) and younger age (p = 0.019). The overall model with site, baseline scores, and participant background characteristics explained about 30% to 40% of the variance in outcomes.

Participant and treatment characteristics are both predictive of outcomes, but treatment characteristics may be more consequential to cognitive gain, while participant characteristics may be more consequential to change in functional capacity.