Cancer history as a predictor in cardiovascular risk scores: a primary care cohort study.

Abstract

Cardiovascular risks are raised in cancer survivors but cancer history is not included in cardiovascular risk scores that inform preventive decisions. To assess whether cancer diagnosis should be included in cardiovascular risk scores. Cohort study using data from English general practices linked to hospital, cancer registration, and death registration data from 1990 to 2015. Adults alive 1 year after a first cancer diagnosis and age, sex, general practice, and calendar- time matched cancer-free individuals were included. Individuals with <2 years of follow-up before index, recent statin prescriptions, or pre-existing coronary heart or cerebrovascular disease were excluded. Cox proportional hazard models used to develop QRISK3 scores were replicated with added cancer history variables. Whether independent hazard ratios for these variables met thresholds for inclusion in QRISK3 (>10% relative difference with <i>P</i><0.01) was assessed. In total, 81 420 cancer survivors and 413 547 cancer-free individuals were followed for a median 5.2 years (interquartile range [IQR] 2.8- 9.1) and 6.3 years (IQR 3.5-10.2), respectively. Including a 1-year cancer survivorship variable in a QRISK3-based model met the threshold for inclusion for males (independent hazard ratio [iHR] 1.16, 95% confidence interval [CI] = 1.11 to 1.20, <i>P</i><0.001) but not females (iHR 1.07, 95% CI = 1.01 to 1.14, <i>P</i> = 0.02). When including cancer type, the threshold was met for both sexes with history of haematological cancer (males: iHR 1.27, 95% CI = 1.16 to 1.40, <i>P</i> <0.001; females: iHR 1.59, 95% CI = 1.32 to 1.91, <i>P</i><0.001) and for males but not females with history of solid cancers (males: iHR 1.13, 95% CI = 1.08 to 1.18, <i>P</i> <0.001; females: iHR 1.04, 95% CI = 0.98 to 1.10, <i>P</i> = 0.19). Developers should consider including cancer history variables in future cardiovascular risk models.