Assessment of the multiple components of the variability in the adenoma detection rate in sigmoidoscopy screening, and lessons for training

Abstract

The determinants of the observed variability of adenoma detection rate (ADR) in endoscopy screening have not yet been fully explained. Between November 1999 and November 2006 13 764 people (7094 men, 6670 women; age range 55-64) underwent screening flexible sigmoidoscopy at five hospital endoscopy units in Turin. To study the determinants of the ADR for distal adenomas, accounting for patient, examiner, and hospital characteristics, we applied a multivariate multilevel regression model. Average ADRs for all adenomas and for advanced adenomas (size > or = 10 mm, villous component > 20 %, high grade dysplasia) were 13.5 % (range 5.2 %-25.0 %) and 6.4 % (3.1 %-10.7 %) for men, and 8.0 % (2.5 %-14.0 %) and 3.7 % (0.2 % - 7.4 %) for women. In multivariate analysis, increased ADR of advanced adenomas was associated with male gender (odds ratio [OR] 1.78, 95 %CI 1.49 - 2.11), self-report of one first-degree relative with colorectal cancer (CRC) (1.44, 1.11-1.86), or of recent-onset rectal bleeding (1.73, 1.24-2.40). Adjusting for these variables, a significantly lower ADR was found for endoscopists with either a lower rate of incomplete sigmoidoscopy (< 9 %; OR 0.59, 95 %CI 0.41-0.87) or a higher rate (> 12 %; 0.64, 0.45-0.91), or with low activity volume (< 85 sigmoidoscopies/year; 0.66, 0.50-0.86). Residual variability explained by the endoscopy center effect was about 1 % and statistically significant. Endoscopist performance in flexible sigmoidoscopy CRC screening is highly variable. Low volume of screening activity independently predicts lower ADR, suggesting that operators devoting more time to screening sigmoidoscopy may perform better. Variability among pathologists in adenoma classification might explain part of the residual variability across endoscopy units.