To determine via meta-analysis the diagnostic accuracy of 64-slice computed tomography coronary angiography (CTA) for assessment of significant obstructive coronary artery stenosis at different coronary artery calcium score (CACS) levels. Data of 12,053 versus 5,890 segments, 906 versus 758 arteries and 1,120 versus 514 patients in low versus high CACS subgroups from 19 eligible studies were compared. The per-patient prevalence of coronary artery disease was 48% versus 68%, respectively. Subgroups were stratified by different CACS thresholds ranging from 100 to 400. Meta-analyses of per-patient data comparing overall low versus high CACS subgroups resulted in a sensitivity of 97.5 (95.5-99)% versus 97 (94.5-98.5)%, specificity of 85 (82-88)% versus 66.5 (58-74.5)%, diagnostic odds ratio of 153 (81-290) versus 40 (20-83), positive predictive value of 85 (82-87)% versus 86 (84-88)%, negative predictive value of 97.5 (95-99)% versus 91 (88-94)% and overall accuracy of 91% versus 89% with 95% confidence interval, respectively. The drop in specificity was significant (P = 0.035), while the sensitivity and overall accuracy were insignificantly changed (P > 0.05). Meta-analyses of independent subgroups at CACS levels ≤10 and ≤100 demonstrated high specificities of 90 (94-100)% and 88.5 (81-91.5)%, whereas at CACS levels ≥400 the specificity declined significantly to 42 (28-56)% but with consistently retained high sensitivity of 97.5 (94-99)%. The specificity of CTA decreases with increasing CACS, while the sensitivity remains high independent of that. The suggested CACS thresholds are arbitrary and do not necessarily warrant cancelling angiography. Diagnostic studies are needed to explore whether a specific CACS threshold may serve as a pre-angiographic gatekeeper to prevent likely equivocal angiographies.