Performance of eNose detection model 95% CI.

IntroductionAfter curative treatment for colorectal cancer (CRC), there is a 15% risk of recurrence. Early detection of an asymptomatic recurrence may lead to curative treatment options. To date, follow-up strategies do not have optimal sensitivity and specificity. In this prospective study, we aimed to assess the diagnostic performance of eNose technology to detect recurrent CRC following curative surgery.Materials and methodsA prospective evaluation study was performed to investigate whether eNose can discriminate patients with recurrent CRC following curative resection from patients without recurrent CRC based on VOC patterns during follow-up. The primary outcome measure is the diagnostic accuracy of eNose for detecting recurrence in CRC patients. With machine learning, a model was developed, and several performance metrics were used to evaluate the diagnostic performance of the eNose model.ResultsA total of 406 patients who underwent curative resection for CRC between 2018−2023, were included in the study. VOC analysis was used to detect recurrent CRC during follow-up. Although the eNose model demonstrated promising results in the train set with an AUC of 0.90 (95% CI:0.84–0.96), the corresponding accuracy of 0.56 was low. Moreover, with a corresponding sensitivity of 0.52, accuracy of 0.51, and AUC of 0.51 (95% CI:0.38–0.64), the performances in the test set, declined.ConclusioneNose technology is not able to accurately detect recurrent CRC after curative resection of the primary tumour. Larger studies are needed before clinical implementation can be realized, while the lack of reproducibility must be addressed.