Cardiovascular risk prediction: balancing complexity against simple practicality

A significant decline in cardiovascular mortality and morbidity has been achieved with the improvements in detection and prevention programmes for major risk factors (such as smoking, low physical activity levels, obesity, and high blood pressure) related to cardiovascular disease (CVD).^1,2 However, CVD still remains the leading cause of mortality and disability in the world. Collectively, ischaemic heart disease and stroke were the cause of 12.9 million deaths in 2010, increasing from one in five to one in four deaths worldwide during two decades.³

Therefore, prophylaxis strategies are essential for CVD management in terms of both primary and secondary prevention. Given the large body of evidence linking dyslipidaemia, particularly increased LDL cholesterol levels with the CVD development, progression, and prognosis, pharmacological therapy, that is statins, are now the mainstay of CVD prevention.⁴

Established CVD puts patients at high risk of future adverse events and they have to undergo appropriate lifestyle interventions and pharmacological treatments (that is, secondary prevention), but no further CVD risk assessment is required. However, risk prediction in patients without known CVD (that is, primary prevention) and particularly decision making with respect to initiation of statin therapy has largely been challenging.

Substantial evidence has emerged that allows wider use of statins for primary prevention despite the data being less robust compared to its use for secondary prevention. A recent Cochrane systematic review on statins for the primary prevention of CVD, which included 18 randomised clinical trials with a total number of participants of 56 934, found that statins reduced all-cause mortality (odds ratio [OR] = 0.86, 95% confidence interval [CI] = 0.79 to 0.94); combined fatal and non-fatal CVD (relative risk [RR] 0.75, 95% CI = 0.70 to 0.81); combined fatal and non-fatal coronary events (RR 0.73, 95% CI = 0.67 to 0.80); combined fatal and non-fatal stroke (RR 0.78, 95% = CI 0.68 = 0.89); and revascularisation rates (RR 0.62, 95% CI = 0.54 to 0.72). Of note, the reduction in adverse events appeared to be cost-effective. Also, benefits of treatment outweighed possible hazards caused by statins; indeed, apart from type 2 diabetes mellitus (RR 1.18, 95% CI = 1.01 to 1.39), the risk of other possible complications (for example, cancer, myalgia, rhabdomyolysis, liver enzyme elevations, renal dysfunction, or arthritis) did not differ between patients on statins and those on placebo (overall RR 1.00, 95% CI = 0.97 to 1.03).⁵

These findings are in line with the meta-analyses performed by the Cholesterol Treatment Trialists’ Collaborators which found a reduction of the risk of vascular (RR 0.85, 95% CI = 0.77 to 0.95) and all-cause mortality (RR 0.91, 95% CI = 0.85 to 0.97) per 1.0 mmol/l LDL cholesterol lowering with statins in patients without prior history of vascular disease.⁶

Despite significant difference in drugs used and thresholds defined for statin treatment between the guidelines, the majority advocate a 10-year cardiovascular risk estimation. For example, the American College of Cardiology (ACC)/American Heart Association (AHA) recommend newly-derived pooled cohort equations and a 7.5% cut-off 10-year risk.^7,8 The European Society of Cardiology/European Atherosclerosis Society guidelines intervention strategy is determined via combination of risk according to SCORE chart and LDL cholesterol level,⁹ while the Canadian Cardiovascular Society guidelines use a modified Framingham Risk Score (which is the doubled per cent of the Framingham risk in case of family history of premature CVD) and LDL cholesterol levels.¹⁰ Finally, the National Institute for Health and Care Excellence (NICE) suggests a 10% risk over 10 years based on QRISK^®2 assessment tool as the threshold for starting preventive treatment with statins.¹¹

Thus, an accurate prediction of cardiovascular risk by identification of individuals with high probability of incident CVD seems to be of major importance for primary prevention. Risk underestimation may eventually lead to the omission of high-risk individuals for whom statin therapy is beneficial and vice versa, risk overestimation may result in unnecessary treatment and probability of side effects. In one interesting meta-analysis, primary prevention (rate ratio 1.52, 95% CI = 1.50 to 1.53) and new statin users (rate ratio 1.46, 95% CI = 1.33 to 1.61) were independent predictors of non-adherence to statin medications.¹² Consistent results were obtained in the UK during the first year of NHS Health Check programme.¹³ Precise risk calculation has become even more important since more ‘aggressive’ treatment thresholds have been proposed (for example, 7.5% for AHA/ACC and 10% for NICE guidelines).

A recent article in the BJGP by Gray et al¹⁴ addresses an important question, by comparing four CVD risk assessment tools: QRISK2, JBS2 (Joint British Societies), and two types of Framingham risk score, one laboratory-based (incorporated total and HDL cholesterol) and another office-based (incorporated body mass index instead of cholesterol levels) in 790 individuals without prior history of CVD or diabetes. Their main findings included higher risk predicted with office-based Framingham risk score and JBS2, and highlighting age as the main driver of cardiovascular risk. Unfortunately no data are available for evaluation of actual event rates against predicted risk.

The fact that among the range of available CVD risk assessment tools, there is a significant variability of results and no perfect one exists is probably not new. All of the risk tools supply clinicians with only approximate estimates of the probability of CVD development.

Performance of different scores depends on a number of factors: largely on definition of CVD — total, both hard events (myocardial infarction, stroke, and coronary heart disease death) and soft events (new angina, transient ischaemic attacks, congestive heart failure, and peripheral vascular disease) or hard events only; study cohorts they were derived from (contemporary or old, observational studies or randomised control trials, outcome definitions, duration of follow-up); quality of calibration of risk score to the target population, and incorporation of risk factors into models as continuous or categorised (or binary) variables. Importantly, with respect to statin treatment initiation no one trial used CVD risk assessment for patients’ enrolment.

The previous NICE guideline allowed choice of CVD risk prediction tool for clinicians, the updated new guideline advocates use of the QRISK2 score only. There are several reasons for this: the QRISK2 score has been derived in the UK from a large general practice database and further validated in several external UK cohorts. In addition it has been subjected to regular update and shown to have the best calibration to CVD event rate in the UK.¹⁵

Given complexity of CVD pathogenesis, including genetic predisposition, exposure to different risk factors, and increasing ethnic diversity, nation-specific tools (such as QRISK2 score in England and Wales) are likely to perform the best in terms of CVD risk evaluation allowing informed decision on preventive strategies, particularly lipid lowering with statins. Nonetheless, additional score complexity with an endless list of risk factors (that include biomarkers and genes) would be at the cost of simplicity and practicality for everyday clinical practice. Also, not all risk factors have equal weighting, nor are they yes/no phenomenon (as they represent a continuum of risk). Excessive complexity may also hinder implementation or ‘short cuts’ in use that would be to the detriment of the patient. A balance may ultimately be needed between complexity (with marginal improvement in risk prediction) and simple practicality (which allows better uptake and application). Time will tell.

Provenance

Commissioned; not externally peer reviewed.