Assessing improved risk prediction of rheumatoid arthritis by environmental, genetic, and metabolomic factors
| Autor: | Elizabeth W. Karlson, Lilia Bouzit, Susan Malspeis, Kazuki Yoshida, Karen H. Costenbader, Jeffrey A. Sparks, Jing Cui |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Oncology
medicine.medical_specialty Receiver operating characteristic business.industry Smoking Regression analysis medicine.disease Risk prediction models Article Regression Arthritis Rheumatoid Seropositive rheumatoid arthritis Blood draw Anesthesiology and Pain Medicine ROC Curve Rheumatology Risk Factors Area Under Curve Case-Control Studies Internal medicine Rheumatoid arthritis medicine Humans Genetic risk business |
| Zdroj: | Semin Arthritis Rheum |
| ISSN: | 0049-0172 |
| Popis: | Objective We sought to improve seropositive rheumatoid arthritis (RA) risk prediction using a novel weighted genetic risk score (wGRS) and preclinical plasma metabolites associated with RA risk. Predictive performance was compared to previously validated models including RA-associated environmental factors. Methods This nested case-control study matched incident seropositive RA cases (meeting ACR 1987 or EULAR/ACR 2010 criteria) in the Nurses’ Health Studies (NHS) to two controls on age, blood collection features, and post-menopausal hormone use at pre-RA blood draw. Environmental variables were measured at the questionnaire cycle preceding blood draw. Four models were generated and internally validated using a bootstrapped optimism estimate: (a) base with environmental factors (E), (b) environmental, genetic and gene-environment interaction factors (E + G + GEI), c) environmental and metabolic factors (E + M), and d) all factors (E + G + GEI + M). A fifth model including all factors and interaction terms was fit using ridge regression and cross-validation. Models were compared using area under the receiver operating characteristic curve (AUC). Results 150 pre-RA cases and 455 matched controls were included. The E model yielded an optimism-corrected AUC of 0.622. The E + M model did not show improvement over the E model (corrected AUC 0.620). Including genetic factors increased prediction, producing corrected AUCs of 0.677 in the E + G + GEI model and 0.674 in the E + G + GEI + M model. Similarly, the performance of the cross-validated ridge regression model yielded an AUC of 0.657. Conclusion Addition of wGRS and gene-environment interaction improved seropositive RA risk prediction models. Preclinical metabolite levels did not significantly contribute to prediction. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect