Diagnostic accuracy and utility of coronary CT angiography with consideration of unevaluable results: A systematic review and multivariate Bayesian random-effects meta-analysis with intention to diagnose.

Abstract

To meta-analyze diagnostic accuracy, test yield and utility of coronary computed tomography angiography (CCTA) in coronary artery disease (CAD) by an intention-to-diagnose approach with inclusion of unevaluable results. Four databases were searched from 1/2005 to 3/2013 for prospective studies that used 16-320-row or dual-source CTs and provided 3 × 2 patient-level data of CCTA (positive, negative, or unevaluable) versus catheter angiography (positive or negative) for diagnosing ≥50% coronary stenoses. A Bayesian multivariate 3 × 2 random-effects meta-analysis considered unevaluable CCTAs. Thirty studies (3422 patients) were included. Compared to 16-40 row CT, test yield and accuracy of CCTA has significantly increased with ≥64-row CT (P < 0.05). In ≥64-row CT, about 2.5% (95%-CI, 0.9-4.8%) of diseased patients and 7.5% (4.5-11.2%) of non-diseased patients had unevaluable CCTAs. A positive likelihood ratio of 8.9 (6.1-13.5) indicated moderate suitability for identifying CAD. A negative likelihood ratio of 0.022 (0.01-0.04) indicated excellent suitability for excluding CAD. Unevaluable CCTAs had an equivocal likelihood ratio of 0.42 (0.22-0.71). In the utility analysis, CCTA was useful at intermediate pre-test probabilities (16-70%). CCTA is useful at intermediate CAD pre-test probabilities. Positive CCTAs require verification to confirm CAD, unevaluable CCTAs require alternative diagnostics, and negative CCTAs exclude obstructive CAD with high certainty. • This 3 × 2 Bayesian meta-analysis included unevaluable CCTAs with intention-to-diagnose. • CCTA is currently useful at intermediate CAD pre-test probabilities. • Unevaluable CCTAs should not, generally, be treated as if they are positive. • Positive CCTAs require verification by other methods to confirm CAD. • Negative CCTAs exclude CAD with high certainty.