Colon capsule endoscopy (CCE) has emerged as a minimally invasive alternative to colonoscopy and has in several meta-analyses shown non-inferiority in detection rates of colorectal lesions. However, CCE is a reader-dependent technique and pathology can be missed even though images of the lesions are captured. Furthermore, misclassification of pathology can lead to erroneous conclusions and possible delays in correct diagnosis and patient management. With this case-series we aim to present misclassified and non-detected cancers from the CareForColon2015 trial and discuss the underlying causes and possible solutions to reader error leading to missed pathology.

In the CareForColon2015 trial including 2,031 Danish colorectal cancer screening participants with a positive faecal immunochemical test undergoing CCE, 72 cancers were detected and verified by histopathology. Through systematic re-evaluation of all CCE investigations where a subsequent colonoscopy reported a cancer, we identified four cases of misclassification, where the lesion was reported as non-neoplastic or with a low risk profile and one case of a non-detected cancer where the lesion was captured in capsule images but not reported by the reader.

In two cases, colonic pathology with clear signs of malignancy (bleeding, central depression and ulceration) were reported as inflammation in CCE while one case had a description in the report of “a short sequenced retracted sigma-lesion with swollen mucosa and red spots” and no designation of “malignancy suspected”. The fourth case of misclassification was a cancer reported as a polyp of 12mm in size. However, a CCE frame shortly after the reported one depicted a lesion of 25mm. The reported frame only showed the border of the cancer therefore giving the impression of a medium and not high-risk finding. In the case of non-detection, two polyps were reported in CCE leading to referral for colonoscopy. A semi-circumferential malignancy suspect tumour in the right colon was found in colonoscopy and in review of the CCE, the capsule clearly visualised the malignant tumour, although not reported.

These cases prompt a discussion on the need for improvements in the CCE reading and reporting process. If readers do not have conventional endoscopy experience, their ability to classify findings correctly can be suboptimal. Additionally time constraints in overburdened healthcare systems can lead to compromised accuracy in the pursuit of efficiency. Reader training including cases of rare and malignant findings combined with a focus on concrete measures to reduce reader fatigue are important steps towards minimising the risk of missed pathology and misclassification of findings. Furthermore, implementation of artificial intelligence support is suggested by many as an essential part of the future CCE pathway to remedy the issues of reader error and ensure consistent and high quality reports to support clinical decision-making and patient management.