An independent external validation of the QRISK3 cardiovascular risk prediction model applied to UK Biobank participants

BackgroundThe QRISK3 cardiovascular disease (CVD) risk prediction model was derived using primary care data; however, it is frequently used outside of clinical settings. The use of QRISK3 in epidemiological studies without external validation may lead to inaccurate results, however it has been used multiple times on data from UK Biobank. We aimed to externally evaluate the performance of QRISK3 for predicting 10-year risk of cardiovascular events in the UK Biobank cohort. MethodsWe used data from the UK Biobank, a large-scale prospective cohort study of 403,370 participants aged 40-69 years recruited between 2006 and 2010 in the United Kingdom (UK). We included participants with no previous history of CVD or statin treatment and the outcome was the first occurrence of coronary heart disease, ischaemic stroke or transient ischaemic attack, derived from linked hospital episode statistics (HES) and death registration data (DRD). ResultsOur study population included 233,233 females and 170,137 males, with 9295 and 13,028 incident cardiovascular events, respectively. The overall median follow-up time after recruitment was 11.7 years. The discrimination measure of QRISK3 in the overall population was reasonable (Harrells C-Index 0.722 in females and 0.697 in males), this was poorer in older participants (<0.62 in all participants aged 65 or older). QRISK3 had systematic over-prediction of CVD risk in UK Biobank, particularly in older participants, by as much as 20%. ConclusionsQRISK3 had reasonable overall discrimination for the whole study population, which was best in younger participants. The observed CVD risk in UK Biobank participants was lower than that predicted by QRISK3, particularly for older participants. The UK Biobank cohort is known to be healthier than the general population and therefore it is necessary to recalibrate QRISK3 before using it to predict absolute CVD risk in the UK Biobank cohort.