Objective: The primary aim of this study was to use meta-regression techniques to compare the diagnostic accuracy of computed tomography colonography (CTC) and magnetic resonance colonography (MRC), compared with conventional colonoscopy for patients presenting with colorectal cancer (CRC).
Method: Quantitative meta-analysis was performed using prospective studies reporting comparative data between CTC and MRC individually to conventional colonoscopy. Study quality was assessed and sensitivities, specificities, diagnostic odds ratios (DOR) were calculated. Summary receiver operating characteristic (SROC) curves and sensitivity analysis were utilized. Meta-regression was used to indirectly compare the two modalities following adjustment for patient and study characteristics.
Results: Overall sensitivity and specificity for CTC (0.96, 95% CI 0.92-0.99; 1.00, 95% CI 0.99-1.00 respectively) and MRC (0.91, 95% CI 0.79-0.97; 0.98, 95% CI 0.96-0.99 respectively) for the detection of CRC was similar. Meta-regression analysis showed no significant difference in the diagnostic accuracy of both modalities (beta=-0.64, P=0.37 and 95% CI of 0.12-2.39). Both tests showed high area under the SROC curve (CTC=0.99; MRC=0.98), with high DORs (CTC=1461.90, 95% CI 544.89-3922.30; MRC=576.41, 95% CI 135.00-2448.56). Factors that enhanced the overall accuracy of MRC were the use intravenous contrast, faecal tagging and exclusion of low-quality studies. No factors improved diagnostic accuracy from CTC except studies with more than 100 patients (AUC=1.00, DOR=2938.35, 95%CI 701.84-12 302.91).
Conclusion: This meta-analysis suggested that CTC and MRC have similar diagnostic accuracy for detecting CRC. Study quality, size and intravenous/intra-luminal contrast agents affect diagnostic accuracies. For an exact comparison to be made, studies evaluating CTC, MRC and colonoscopy in the same patient cohort would be necessary.