Limited clinical utility of CT coronary angiography in a district hospital setting

Abstract

Studies have demonstrated considerable accuracy of multi-slice CT coronary angiography (MSCT-CA) in comparison to invasive coronary angiography (I-CA) for evaluating coronary artery disease (CAD). The extent to which published MSCT-CA accuracy parameters are transferable to routine practice beyond high-volume tertiary centres is unknown. To determine the accuracy of MSCT-CA for the detection of CAD in a Scottish district general hospital. Prospective study of diagnostic accuracy. One hundred patients with suspected CAD recruited from two Glasgow hospitals underwent both MSCT-CA (Philips Brilliance 40 × 0.625 collimation, 50-200 ms temporal resolution) and I-CA. Studies were reported by independent, blinded radiologists and cardiologists and compared using the AHA 15-segment model. Of 100 patients [55 male, 45 female, mean (SD) age 58.0 (10.7) years], 59 and 41% had low-intermediate and high pre-test probabilities of significant CAD, respectively. Mean (SD) heart rate during MSCT-CA was 68.8 (9.0) bpm. Fifty-seven per cent of patients had coronary artery calcification and 35% were obese. Patient prevalence of CAD was 38%. Per-patient sensitivity, specificity, positive and negative (NPV) predictive values for MSCT-CA were 92.1, 47.5, 52.2 and 90.6%, respectively. NPV was reduced to 75.0% in the high pre-test probability group. Specificity was compromised in patients with sub-optimally controlled heart rates, calcified arteries and elevated BMI. Forty-Slice MSCT-CA has a high NPV for ruling out significant CAD when performed in a district hospital setting in patients with low-intermediate pre-test probability and minimal arterial calcification. Specificity is compromised by clinically appropriate strategies for dealing with unevaluable studies. Effective heart rate control during MSCT-CA is imperative. National guidelines should be utilized to govern patient selection and direct MSCT-CA reporter training to ensure quality control.