Colonoscopic ultrasound is associated with a learning phenomenon despite previous rigid probe experience.

Colonoscopic ultrasound (CUS) enables total colonoscopic examination combined with staging of tumor. Rigid probe transrectal ultrasound (TRUS) is reliable in assessing rectal cancer. Both the modalities are associated with an initial learning curve. We evaluated the predictability CUS in preoperative staging of rectal cancer during the learning curve, despite experience with TRUS. Forty-four patients with non-obstructing rectal cancer were assessed by colonoscopy and colonic ultrasound using a 7.5 MHz rotating transducer. Accuracy of ultrasound staging was compared with pathological staging. Tumor staging and nodal staging at pathology and ultrasound were named pT, pN and uT, uN, respectively. The pathological staging was pT1 in two (4.5%), pT2 in 16 (36%), pT3 in 21 (48%) and pT4 in five (11.5%) rectal cancer specimens. CUS understaged the tumor in 11 cases and overstaged it in 10 cases. Overall, the positive predictive value was 61%, negative predictive value 73%, sensitivity 61%, and specificity 73%. Lymph nodes were not visualized in 14. The overall un-weighted kappa of CUS staging of RC was 0.18 (poor). The predictive value in tumor staging of CUS is suboptimal in the learning phase, despite previous experience with TRUS.