How many segments are necessary to characterize delayed colonic transit time?

Abstract

Measuring colonic transit time with radiopaque markers is simple, inexpensive, and very useful in constipated patients. Yet, the algorithm used to identify colonic segments is subjective, rather than founded on prior experimentation. The aim of the present study is to describe a rational way to determine the colonic partition in the measurement of colonic transit time. Colonic transit time was measured in seven segments: ascending colon, hepatic flexure, right and left transverse colon, splenic flexure, descending colon, and rectosigmoid in 852 patients with functional bowel and anorectal disorders. An unsupervised algorithm for modeling Gaussian mixtures served to estimate the number of subgroups from this oversegmented colonic transit time. After that, we performed a k-means clustering that separated the observations into homogenous groups of patients according to their oversegmented colonic transit time. The Gaussian mixture followed by the k-means clustering defined 4 populations of patients: "normal and fast transit" (n = 548) and three groups of patients with delayed colonic transit time "right delay" (n = 82) in which transit is delayed in the right part of the colon, "left delay" (n = 87) with transit delayed in the left part of colon and "outlet constipation" (n = 135) for patients with transit delayed in the terminal intestine. Only 3.7 % of patients were "erroneously" classified in the 4 groups recognized by clustering. This unsupervised analysis of segmental colonic transit time shows that the classical division of the colon and the rectum into three segments is sufficient to characterize delayed segmental colonic transit time.