To evaluate the performance of computed tomography angiography (CTA) ≥64 slices for detecting coronary in-stent restenosis (ISR) and determine the influence of separate characteristics on diagnostic accuracy. We searched the PubMed, EMBASE and Cochrane databases for studies of CTA ≥64 slices in diagnosing ISR. We pooled data on bivariate modelling, and subgroup analysis was also performed. A total of 35 studies involving 4131 stents were included. The pooled positive likelihood ratio (LR+) and the negative likelihood ratio (LR-) were 14.0 and 0.10, for CTA in diagnosis-significant ISR ≥50%. LR+ and LR- were similar between CTA >64 slices versus 64 slices (both P > 0.99). LR- (0.10) was good for ruling out suspected ISR for <3-mm diameter. Time between CTA and stent implantation >6 months did not affect the ability of CTA for the high LR+ (12.3) and the LR- (0.10). Thick-strut stents ≥100 μm or bifurcation stenting demonstrated inferior accuracy, which was unfavourable for stent imaging. With the high LR+ and LR- of CTA, patients with ISR may be appropriate for non-invasive angiographic follow-up. However, CTA imaging seems unsuitable for patients with characteristics unfavourable for stent imaging, such as thick-strut stents or bifurcation stenting. • CTA may provide accurate information on characteristics of in-stent restenosis lesions. • Using CTA, ISR patients may be appropriate for non-invasive angiographic follow-up. • Stent diameter and the number of slices do not influence CTA. • CTA seems unsuitable for patients with thick-strut stents or bifurcation stenting.