Abstract
Background: Generalized cognitive impairment is a core characteristic of schizophrenia. Risk profile scores (RPS) summarize an individual’s risk for developing schizophrenia based on numerous common genetic variants associated with the condition. Various work points to overlap in cognitive and schizophrenia genetics in affected individuals and families. We tested whether subgroups, defined by different patterns of premorbid vs. current IQ, showed different associations of schizophrenia RPS with cognitive performance. Methods: 550 people with schizophrenia, 432 of their unaffected siblings, and 1127 community controls completed cognitive assessments and provided blood for genetics analyses, as part of the NIMH Study of Schizophrenia Genetics. We performed cluster analyses in the schizophrenia cases to identify cognitive subgroups, using premorbid (WRAT) and current (WAIS) IQ as clustering indicators. To investigate the familiality of associations, subgroup assignments derived for schizophrenia cases were also assigned to their unaffected siblings. Schizophrenia RPS for all individuals were calculated at different thresholds based on illness-associated genetic variants identified by the Psychiatric Genetics Consortium. Across RPS thresholds, we used ordered hierarchical regression to test the association of RPS with general cognitive ability (“g”) in the derived cognitive subgroups, and in unaffected sibling subgroups, controlling for age, sex, and population stratification. Results: Cluster analyses suggested a three-subgroup scheme: one subgroup with high scores on both WRAT and WAIS (SzHH), one with low scores on both WRAT and WAIS (SzLL), and one with high scores on the WRAT and low WAIS (SzHL), as found previously. Cognitive performance was significantly impaired and RPS scores were significantly elevated in all schizophrenia groups relative to controls, and siblings were intermediate. The SzHH subgroup showed significant associations between RPS and “g” at six of 10 RPS thresholds (e.g., at RPS_0.05 n = 179; P = .002; R(2) = .047). Their siblings showed the same pattern of association, with larger effect sizes (e.g., for SiblHH at RPS_0.05 n = 94; P = .01; R(2) = .073) as did the siblings of the SzHL subgroup. However, the SzHL subgroup showed no association of schizophrenia RPS with “g” at any threshold. Conclusion: These analyses indicate that, for many schizophrenia cases (SzHH), summary scores indexing common genetic risk for the condition are also predictive of cognitive impairment. Indeed, the pattern was clearly evident even in unaffected siblings. Strikingly, for another subgroup (SzHL), and despite higher polygenic risk for schizophrenia, that risk was entirely decoupled from cognition. One hypothesis is that different patterns of IQ performance identify subgroups with different trajectories of cognitive development and distinct illness etiologies.