This study aimed to explore the role of radiologist-identified endoluminal pathology within the large bowel in patients without a prior history of colorectal cancer. In our cohort, 61/106 (58%) radiology-initiated referrals to the lower GI tumor board were for suspected endoluminal pathology identified on CT. CTC demonstrated a positive predictive value (PPV) of 91% for all endoluminal pathology, compared to 70% for unprepared CT. Of the true positive lesions detected, 52% were malignant on CTC compared to 67% on unprepared CT. Gastrointestinal (GI) subspecialty reporting identified 44% more endoluminal pathology on unprepared CT (95% CI 21.13–55%, p < 0.0001), without significant change in the PPV for this relatively small sample size.

CTC is widely regarded as the gold standard for detecting endoluminal pathology by imaging, The SIGGAR multi-center randomized controlled [7] trial demonstrated that CTC has a sensitivity of 96% for detecting CRC and polyps > 6 mm compared to colonoscopy, though it is primarily used when endoluminal pathology is suspected or being screened for [7]. Most CTC examinations performed at our institution were performed for symptomatic patients. In such cases, contrast-enhanced CTC is advised to improve the sensitivity of detecting extra-colonic pathology [12, 13] which accounts for the high proportion of contrast enhanced CTC in our cohort. Our local practice of red-flagging any imaging findings suggestive of malignancy to the tumor board likely reflects the true diagnostic yield of CTC in this setting. It is therefore notable that no significant difference found between CT colonography and unprepared CT in the detection of colorectal cancer (p = 0.8).

The majority of malignant lesions on unprepared CT were reported by GI subspecialists, suggesting the diagnostic performance of unprepared CT may be underestimated in general practice. Supporting this interpretation, lesions identified on unprepared CT in our study tended to be larger and more advanced than those detected on CT colonography, consistent with findings by Wahlig et al. [10].

The role of unprepared CT in detecting endoluminal pathology is less well established than that of CTC. Our findings suggest that radiologist interpretation of unprepared CT for endoluminal pathology (benign and malignant) is more accurate than previously reported. Prior studies have largely focused on sensitivity or specific patient subsets [14], and reported a lower PPV, such as the 33% reported by Colvin et al. [15] and noted that up to 20% of cancers were missed on CT before diagnostic colonoscopy [16]. The higher PPV observed in our cohort (and the consistency of our CT colonography PPV with published literature [17]) suggest that the diagnostic contribution of unprepared CT may have been underestimated.

Several factors may explain the improved PPV of unprepared CT in our study. Advances in CT acquisition protocols and post processing techniques over the past decade have likely enhanced lesion detection beyond those reported in earlier studies. A degree of referral or selection bias may also be present, as increasing awareness of colorectal cancer, particularly in younger patients, may also lower the threshold for requesting cross-sectional imaging and increase pre-test probability. Finally, institutional factors such as regular discrepancy review meetings, multidisciplinary collaboration and a strong culture of subspecialty engagement may contribute to interpretive accuracy, reflecting a broader depth of gastrointestinal radiology expertise across teams.

Our study complements Wahlig et al. [10] who demonstrated variability in radiologists’ ability to identify colorectal cancers in patients with and without known CRC. Their methodology sought to reduce bias by having radiologists review cases without prior knowledge of cancer presence. They reported individual PPVs of 0.68 and 0.67 for colorectal cancer, with combined PPVs ranging from 0.59 to 0.86, highlighting the diagnostic challenges and discrepancies often encountered in clinical practice.

In contrast, our study reflects real-world practice, and demonstrates similar PPVs between GI and non-GI radiologists. This suggests that while unprepared CT should not be used for screening, all radiologists play an important role in flagging potential endoluminal pathology. Notably, our study is the first to report outcomes of non-dedicated radiologist reporting of endoluminal pathology albeit limited by sample size, emphasizing the important role of all radiologists in identifying colorectal cancers and precursor lesions. In our data, PPV did not differ significantly between specialist and non-specialist reporters; however, GI subspecialists referred significantly more cases. These findings emphasize both the added value of subspecialty expertise and the crucial contribution of non-specialists in detecting colorectal cancer and precursor lesions.

Strengths of our study include the inclusion of both CT colonography and unprepared CT examinations referred to the lower GI tumor board, and the inclusion of non-malignant lesions. Our CTC PPV is consistent with published data, confirming the quality of GI subspecialist reporting. PPVs stated are intentionally for both malignant and benign lesions as presenting only malignant lesions undervalues the importance of detection of precursor lesions. Potential strategies to further improve diagnostic performance could include standardized reporting criteria, double-reading in equivocal cases, confirmatory CT colonography, and structured reporting templates to reduce inter-observer variability.

Our findings should reassure clinicians and radiologists that a relatively high threshold for subspecialist review and tumor board discussion is appropriate, given that a substantial proportion of cases will have genuine endoluminal pathology.

Limitations include the absence of a control cohort without suspected endoluminal disease, though the data from Wahlig et al. [10] data partly mitigates this. This single-centre design with histopathology only available for those patients who underwent resection is a further limitation, however, our center is a high-volume colorectal unit with over 600 patients reviewed annually at tumor board and approximately 100 resections performed per year. Another limitation is that only patients with abnormal CTs were included, preventing assessment of true negative and false negative rates. Future matched cohort studies could explore these metrics, but given CT’s limitations, negative predictive value (NPV) is likely poor for excluding endoluminal pathology. While identifying the specific imaging feature that trigger suspicion for endoluminal pathology was beyond the scope of this study, previous work by Wahlig et al. [9, 10] has evaluated such factors.