Atherosclerotic cardiovascular disease (ASCVD) leads to narrowing and blockage of arteries, and is a common cause of illness and death worldwide. It can affect any artery, and therefore has multiple clinical manifestations, including angina, heart attack, stroke and compromised blood flow to the limbs. As all are influenced by common underlying disease processes, any disease manifestation is associated with higher than average risk of any future ASCVD 'event' (e.g. heart attack or stroke), including death due to ASCVD. However, ASCVD risk varies markedly between individuals. This is important as treatment benefit is related to the extent of risk. Individual treatment preferences and appropriate allocation of clinical resources will therefore also be sensitive to risk level.
Although we are currently unable to estimate individual future ASCVD event risks for those with established ASCVD, a possible method has recently been proposed. This method requires testing in settings other than that in which it was derived, before wider use is recommended. We therefore aim here to test this model among those registered within UK primary care practices. In addition to assessing the tool performance per se, the study will help determine whether it has clinical utility in UK primary care practice specifically.
Aim: An external validation of the SMART prognostic tool in predicting the 10-year risk of secondary atherosclerotic cardiovascular disease (ASCVD) events.
Study design: Historical open cohort study
Observational period: 2000-2017
Population: A UK routine primary care cohort, >18 and <80 years at baseline, diagnosed with coronary, cerebrovascular, peripheral and/or aortic ASCVD
Outcome: First occurrence of myocardial infarction, stroke (ischaemic or haemorrhagic), or CVD death, post cohort-entry
Predictors: Age, sex, diabetes, smoking status; cholesterol, blood pressure, renal function and high sensitivity C-reactive protein (hSCRP) measurements; ASCVD history (years since clinical manifestation; vascular bed(s) affected). hsCRP values (not available in CPRD) will be imputed using age*sex*ASCVD diagnosis-specific mean values observed in the derivation cohort, and missing non-hsCRP predictor values using standard methods.
Performance assessment: We will describe the risk factor distributions and outcome incidence for the validation cohort, and calculate the 10-year estimated ASCVD event risk for each individual using the SMART tool. We will assess model accuracy using standard measures of calibration and discrimination for censored data, including those reported for the derivation dataset to enable comparison. We will assess the clinical utility of the tool across a range of vascular risk estimates/potential treatment decision thresholds, in a decision curve analysis.
Myocardial infarction
Stroke (ischaemic or haemorrhagic)
Cardiovascular death
Kausik Ray - Chief Investigator - Imperial College London
Ailsa McKay - Corresponding Applicant - Imperial College London
Azeem Majeed - Collaborator - Imperial College London
Laura Gunn - Collaborator - Imperial College London
HES Admitted Patient Care;ONS Death Registration Data