We conducted a meta-analysis evaluating the critical ratio between effective radiation dose (ED), feasibility (Fe) and diagnostic accuracy (Ac) of multidetector computed tomography (MDCT) for the detection of significant coronary artery disease. By using our predetermined criteria, we selected human studies published in English in which the ED and raw data of Ac vs. invasive coronary angiography in a segment based model were specified. Data from 31 studies including 3661 patients (mean age 61.9 ± 4.5 years, heart rate 62.5 ± 6.7 bpm) and 50,236 coronary artery segments were analysed and are reported. Overall, Fe, sensitivity, specificity, negative predictive value, positive predictive value, Ac and ED were 95%, 90%, 96%, 99%, 69%, 95% and 10.4 ± 5.4 mSv, respectively. Multivariate analysis showed that prospective ECG-gating (- 8.8 mSv CI95% - 13.4 to - 4.3 mSv, p = 0.001), dual-source (- 3.7 mSv CI95% - 7.9 to 0 mSv, p = 0.05) and BMI-adapted scanning protocols (- 4.5 mSv CI95% - 8.7 to - 2.7 mSv, p = 0.03) were independent predictors of ED reduction. In patients with low heart rate, the best compromise between ED, Fe and Ac (2.5 mSv, 97% and 98%, respectively) was obtained combining prospective ECG-gating and BMI-adapted scanning protocols, while in patients with high heart rate the strategy associated with the best results (10 mSv, 98% and 97%, respectively) was the use of dual-source MDCT with retrospective ECG gating and modulation dose. In conclusion, careful selection of CT scanning protocols according to the patient's characteristics is critical for keeping the radiation exposure "as low as reasonably achievable" (ALARA) without impairing Fe and Ac.