Aims We hypothesized that the modified Diamond–Forrester (D-F) prediction model overestimates probability of coronary artery disease (CAD). The aim of this study was to update the prediction model based on pre-test information and assess the model’s performance in predicting prognosis in an unselected, contemporary population suspected of angina. Methods and results We included 3903 consecutive patients free of CAD and heart failure and suspected of angina, who were referred to a single centre for assessment in 2012–15. Obstructive CAD was defined from invasive angiography as lesion requiring revascularization, >70% stenosis or fractional flow reserve <0.8. Patients were followed (mean follow-up 33 months) for myocardial infarction, unstable angina, heart failure, stroke, and death. The updated D-F prediction model overestimated probability considerably: mean pre-test probability was 31.4%, while only 274 (7%) were diagnosed with obstructive CAD. A basic prediction model with age, gender, and symptoms demonstrated good discrimination with C-statistics of 0.86 (95% CI 0.84–0.88), while a clinical prediction model adding diabetes, family history, and dyslipidaemia slightly improved the C-statistic to 0.88 (0.86–0.90) (P for difference between models <0.0001). Quartiles of probability of CAD from the clinical prediction model provided good diagnostic and prognostic stratification: in the lowest quartiles there were no cases of obstructive CAD and cumulative risk of the composite endpoint was less than 3% at 2 years. Conclusion The pre-test probability model recommended in current ESC guidelines substantially overestimates likelihood of CAD when applied to a contemporary, unselected, all-comer population. We provide an updated prediction model that identifies subgroups with low likelihood of obstructive CAD and good prognosis in which non-invasive testing may safely be deferred.

These data confirm the very low diagnostic yield of non-invasive and invasive assessment of CAD in current clinical practice, particularly in women and in patients with atypical symptoms. Data call for a more rational approach to avoid unnecessary testing.