Polygenic risk-based prediction of AF in cancer patients: an analysis from UK biobank

Abstract Purpose Although atrial fibrillation (AF) risk is higher in cancer patients compared to non-cancer patients, predictors for AF in cancer patients are not well identified. We aimed to investigate the clinical significance of genetic predisposition for predicting AF in cancer patients. Methods The polygenic risk score (PRS) for AF was constructed from previous genome-wide association studies. From the UK Biobank cohort, genetically unrelated White participants with cancer were identified. Patients were categorized into low, intermediate, high genetic risk groups for AF according to PRS (low: bottom 10%, intermediate: 10-90%, high: top 10%). The 10-year cumulative incidence, absolute risk difference, and multivariable-adjusted risk of AF were compared among genetic risk groups in cancer versus non-cancer individuals, and in each cancer type. Results A total of 376,972 participants (mean age 56.6±8.0, male 45.6%, cancer 14.2%) were followed for a median of 12.4 years (cancer patients: 3.8 years). Compared to patients in the non-cancer group, patients in the cancer group were older and had higher proportion of comorbidities (Mean age: 60±6.9, 56.0±8.0, p<0.001; diabetes: 6.1%, 4.6%, p<0.001; hypertension: 34.8%, 27.8%, p<0.001; dyslipidemia: 24.5%, 18.0%, p<0.001; cancer and non-cancer group, respectively). AF incidence at 10-year significantly differed by genetic predisposition in both non-cancer and cancer groups (Non-cancer: 2.3%, 3.8%, 7.2%, p<0.001; Cancer: 6.1%, 9.6%, 14.3%, p<0.001, [low, intermediate, high genetic risk, respectively], p-for-interaction=0.017). The Cox-regression analysis was used to compared the risk of AF according to the genetic-risk groups, and adjusted hazard ratio (aHR) of AF were calculated using various variables including age, sex, BMI, current smoking, diabetes, hypertension, dyslipidemia, genotype array and genetic principal components 1 to 10. Compared to non-cancer patients with low genetic risk, the adjusted risk of AF increased in cancer patients with higher genetic risk (aHR [95% CI], non-cancer: intermediate 1.71 [1.58-1.84], high 3.34 [3.07-3.63]; cancer: low 2.15 [1.79-2.58], intermediate 3.47 [3.18-3.79], high 5.72 [5.03-6.50]; all p<0.001). Among the top 5 frequent cancers, absolute difference of 10-year AF incidence between low and high genetic-risk groups was prominent in prostate cancer, hematologic malignancies, and lung cancer (15.0%, 17.7%, and 17.3%) but not remarkable in breast and colorectal cancers (2.9% and 4.4%). Conclusion Cancer patients at a higher AF risk can be effectively discriminated using a PRS-based approach. The amount of absolute risk increment of AF by genetic predisposition varied across cancer types, which may be useful for individualized care in cancer survivorship.