Syncope risk stratification tools vs clinical judgment: An individual patient data meta-analysis

BACKGROUND: There have been several attempts to derive syncope prediction tools to guide clinician decision-making. However, they have not been largely adopted, possibly because of their lack of sensitivity and specifi city. We sought to externally validate the existing tools and to compare them with clinical judgment, using an individual patient data meta-analysis approach. METHODS: Electronic databases, bibliographies, and experts in the field were screened to find all prospective studies enrolling consecutive subjects presenting with syncope to the emergency department. Prediction tools and clinical judgment were applied to all patients in each dataset. Serious outcomes and death were considered separately during emergency department stay and at 10 and 30 days after presenting syncope. Pooled sensitivities, specificities, likelihood ratios, and diagnostic odds ratios, with 95% confidence intervals, were calculated. RESULTS: Thirteen potentially relevant papers were retrieved (11 authors). Six authors agreed to share individual patient data. In total, 3681 patients were included. Three prediction tools (Osservatorio Epidemiologico sulla Sincope del Lazio [OESIL], San Francisco Syncope Rule [SFSR], Evaluation of Guidelines in Syncope Study [EGSYS]) could be assessed by the available datasets. None of the evaluated prediction tools performed better than clinical judgment in identifying serious outcomes during emergency department stay, and at 10 and 30 days after syncope. CONCLUSIONS: Despite the use of an individual patient data approach to reduce heterogeneity among studies, a large variability was still present. Current prediction tools did not show better sensitivity, specificity, or prognostic yield compared with clinical judgment in predicting short-term serious outcome after syncope. Our systematic review strengthens the evidence that current prediction tools should not be strictly used in clinical practice.