Polygenic Risk Scores Help Predict Schizophrenia Risk

The investigators suggest that polygenic risk scores derived for schizophrenia in real-world health care settings were robustly associated with the risk for a schizophrenia diagnosis.

Polygenic risk scores are a sensible measure for determining risk for schizophrenia diagnosis and can further estimate the pleiotropic effects of schizophrenia risk for related disorders, according to a study published in the American Journal of Psychiatry.  

The investigators sought to test the approach of polygenic risk scores used in a real-world clinical setting to quantitatively measure the genetic risk for schizophrenia.

Related Articles

The investigators leveraged electronic health records from 106,160 patients across 4 large health care systems (Geisinger Health System, Mount Sinai Health System, Partners HealthCare System, and Vanderbilt University Medical Center). Genetic data were assessed for the relative and absolute risk for schizophrenia; individuals with the highest level of genetic risk were used to test the clinical utility of polygenic risk scores used for risk stratification.

Polygenic risk scores were calculated using pruning single-nucleotide polymorphisms based on linkage disequilibrium and a Bayesian polygenic prediction model boasting increased accuracy. The pleiotropic effects of schizophrenia polygenic risk scores using real-world clinical data were further examined in a phenome-wide association study of 1359 disease categories. Additionally, the investigators performed a sensitivity analysis to explore whether the pleiotropic effects were mediated by the schizophrenia diagnosis itself or prescription of antipsychotic medication.

In the cross-site meta-analysis, polygenic risk scores were significantly associated with schizophrenia (odds ratio per standard deviation increase, 1.55; 95% CI, 1.4-1.7; P =4.48 x 10⁻¹⁶). Patients in the highest risk decile of the polygenic risk score distribution were 4.6 times more likely to develop schizophrenia compared with those in the bottom decile (95% CI, 2.9-7.3; P =1.37 x 10⁻¹⁰). Besides schizophrenia, polygenic risk scores (using both methods of calculation) were positively associated with other phenotypes: bipolar disorder, depression, substance use disorders, anxiety, personality disorders, neurological disorders, suicidal behavior, memory loss, viral hepatitis, urinary syndromes, and nonspecific somatic symptoms. However, polygenic risk scores were inversely associated with obesity and synovitis. These associations remained significant across all sensitivity analyses.

Limitations included restricting analyses to patients of European descent, which may prevent the generalizability of findings to other populations, and phenotype definitions disregarding variables such as the medical history of related disorders, setting of diagnosis, and treatment. Results between the different sites varied to some degree, in which biobank patients generally had a longer duration of follow-up, providing more opportunity to receive a diagnosis.   

The investigators suggest that polygenic risk scores derived for schizophrenia in real-world health care settings were robustly associated with the risk for a schizophrenia diagnosis. In addition, a range of pleiotropic relationships were revealed in phenome-wide association studies relating polygenic risk scores with other psychiatric disorders and nonpsychiatric symptoms and syndromes.

Disclosure: Several study authors declared affiliations with the pharmaceutical industry. Please see the original reference for a full list of authors’ disclosures.


Zheutlin AB, Dennis J, Linnér RK, et al. Penetrance and pleiotropy of polygenic risk scores for schizophrenia in 106,160 patients across four health care systems [published online August 16, 2019]. Am J Psychiatry. doi: 10.1176/appi.ajp.2019.18091085