Positive Predictive Value for Colorectal Lesions at CT Colonography: Analysis of Factors Impacting Results in a Large Screening Cohort

OBJECTIVE. The purpose of this study is to evaluate factors affecting the positive predictive value (PPV) for detecting colorectal lesions at CT colonography (CTC), using optical colonoscopy (OC) as the reference standard for concordance. MATERIALS AND METHODS. Consecutive CTC studies from a single screening program interpreted as positive for at least one detected colorectal lesion 6 mm or larger and sent for subsequent OC were analyzed according to per-polyp and per-patient results. Univariable and multivariable analysis of multiple input factors was performed. RESULTS. Of 1650 studies (median patient age, 59.7 years; 877 men and 773 women) with 2688 total CTC-detected lesions 6 mm or larger, the overall PPVs were 88.8% (2386/2688) by polyp and 90.8% (1499/1650) by patient. The by-polyp PPV was significantly higher for polypoid (91.2%; 1793/1965) versus flat or nonpolypoid (79.4%; 459/578) lesions (p < 0.0001). Overall per-patient PPVs were 72.3% (1193/1650) for any neoplasia 6 mm or larger and 38.8% (641/1650) for advanced neoplasia PPVs for advanced neoplasia increased by CTC Reporting and Data System category: 5.8% (45781) for C2, 671% (511/762) for C3, and 794% (85/107) for C4. PPVs for cancer also increased by CTC Reporting and Data System category: 0% (0/781) for C2, 2.2% (17/762) for C3, and 52.3% (56/107) for C4. On multivariable regression analysis, polyp morphologic type (flat vs polypoid) and diagnostic confidence were the strongest predictors of CTC-OC concordance. CTC PPV results are somewhat underestimated because 28.8% (87/302) of CTC-OC-discordant results were categorized as likely OC false-negatives at consensus review. CONCLUSION. Concordance between CTC and OC is high for relevant colorectal polyps and masses. Unlike stool-based tests that provide only a binary positive or negative result, CTC can specify the nature of the positive findings, resulting in much greater specificity and risk stratification for patient management decisions.