Cardiovascular risk scores do not account for the effect of treatment: a review

OBJECTIVE: To compare the strengths and limitations of cardiovascular risk scores available for clinicians in assessing the global (absolute) risk of cardiovascular disease. DESIGN: Review of cardiovascular risk scores. DATA SOURCES: Medline (1966 to May 2009) using a mixture of MeSH terms and free text for the keywords ‘cardiovascular’, ‘risk prediction’ and ‘cohort studies’. ELIGIBILITY CRITERIA FOR SELECTING STUDIES: A study was eligible if it fulfilled the following criteria: (1) it was a cohort study of adults in the general population with no prior history of cardiovascular disease and not restricted by a disease condition; (2) the primary objective was the development of a cardiovascular risk score/equation that predicted an individual's absolute cardiovascular risk in 5–10 years; (3) the score could be used by a clinician to calculate the risk for an individual patient. RESULTS: 21 risk scores from 18 papers were identified from 3536 papers. Cohort size ranged from 4372 participants (SHS) to 1591209 records (QRISK2). More than half of the cardiovascular risk scores (11) were from studies with recruitment starting after 1980. Definitions and methods for measuring risk predictors and outcomes varied widely between scores. Fourteen cardiovascular risk scores reported data on prior treatment, but this was mainly limited to antihypertensive treatment. Only two studies reported prior use of lipid-lowering agents. None reported on prior use of platelet inhibitors or data on treatment drop-ins. CONCLUSIONS: The use of risk-factor-modifying drugs—for example, statins—and disease-modifying medication—for example, platelet inhibitors—was not accounted for. In addition, none of the risk scores addressed the effect of treatment drop-ins—that is, treatment started during the study period. Ideally, a risk score should be derived from a population free from treatment. The lack of accounting for treatment effect and the wide variation in study characteristics, predictors and outcomes causes difficulties in the use of cardiovascular risk scores for clinical treatment decision.