Study population
We conducted a retrospective study of screening colonoscopies at the Center for Health Promotion of the Samsung Medical Center in Seoul, Korea, from December 2007 to June 2009. We restricted the study to male and female participants aged 40 to 79 years old as this is the recommended age range for screening colonoscopy in Korea. From a total of 14,370 consecutive colonoscopies performed at the Center for Health Promotion, we excluded 862 colonoscopies conducted in participants who were <40 or ≥80 years of age, 526 colonoscopies performed for therapeutic (non-screening) purposes, 65 colonoscopies performed outside of the colonoscopy unit, 29 colonoscopies performed in participants with a history of colorectal cancer, inflammatory bowel disease or who had undergone colonic resection, and 113 colonoscopies with incomplete examination due to poor bowel preparation. In subjects who had colonoscopy multiple times during the study period (n = 98), we chose the first colonoscopy results. Because some individuals met more than one exclusion criterion, the final sample size was 12,679 (7,975 men and 4,704 women). The study protocol was reviewed and approved by the institutional review board at Samsung Medical Center. The informed consent requirement was exempted by the Institutional Review Board because researchers only accessed retrospectively a de-identified database for analysis purposes.
Study procedures
Twenty three board-certified gastroenterologists performed the colonoscopies. Each colonoscopist had performed more than 500 examinations before the study period or before working at the Center for Health Promotion. Colonoscopies were performed after bowel preparation with 4 L polyethylene glycol solution (Colyte®, Taejun, Seoul, Korea; Colyte®-F, Taejun, Seoul, Korea; Colonlyte®, Dreampharma, Seoul, Korea).
Cecal insertion time was defined as the time from insertion into the rectum to the time when the colonoscope tip passed to a point proximal to the ileocecal valve so that the base of cecum was visible. Withdrawal time was defined as the time taken for withdrawing the colonoscope tip from the base of cecum to across the anus. Cecal insertion and withdrawal times were recorded immediately after finishing the examination by the colonoscopist. Bowel preparation was assessed as excellent (no or nearly no fecal matter in the colon; small-to-moderate amounts of clear liquid present), good (small amounts of thin liquid fecal matter visible and easily suctioned, mainly distal to the splenic flexure), fair (moderate amounts of thick liquid to semisolid fecal matter visible and suctioned, including proximal to the splenic flexure; >90% of the mucosa visible), or poor (large amounts of solid fecal matter present that preclude a satisfactory study; <90% of the mucosa visible).
Colorectal neoplasms were further classified as small single non-advanced adenomas < 5 mm, medium single non-advanced adenomas 5 to <10 mm,and multiple or advanced colorectal neoplasms (including multiple adenomas regardless of size and advanced colorectal neoplasms). Advanced adenoma was defined as a tubular adenoma with diameter ≥10 mm, an adenoma with villous component, or an adenoma with high grade dysplasia. The size of each lesion was estimated using open biopsy forceps.
A health questionnaire and a detailed physical exam were routinely completed as part of the screening program. Height and weight were measured using an Inbody 720 machine (Biospace, Seoul, Korea). Body mass index (BMI) was calculated by dividing measured weight (kg) by height squared (m2). Waist circumference was measured at the midpoint between the inferior margin of the last rib and the superior iliac crest in a horizontal plane. Information on history of colorectal polyps, diabetes mellitus, hyperlipidemia, medication use including aspirin and non-steroidal anti-inflammatory drugs (NSAIDs), alcohol drinking and smoking and family history of colorectal cancer were collected using a self-administered questionnaire before endoscopy.
Statistical analysis
Colorectal adenoma proportions were compared across the quartiles of cecal insertion time. Odds ratios (ORs) and 95% confidence intervals (CIs) for the presence of any colorectal neoplasm for the 3 highest quartiles of insertion time compared to the first quartile were calculated using fixed effects logistic regression conditioned on colonoscopists. These regression models account for data clustering by colonoscopists and eliminate confounding due to differences across colonoscopists. Separate logistic regression models were used for the different colorectal neoplasms. In multivariable models, we adjusted for age (40 to 49, 50 to 59, 60 to 69, 70 to 79 years), sex, withdrawal time (<6 minutes, ≥6 minutes), bowel preparation (excellent, good, fair and poor), BMI (<25, ≥25 kg/m2), waist circumference (<80, 80 to 89, 90 to 99, ≥100 cm), family history of colorectal cancer, history of colorectal polyps, diabetes, hyperlipidemia, aspirin use, NSAIDs use, calcium use, alcohol drinking, and smoking (current, past, and never).
Responses to health questionnaire items coded as “unknown” and items not answered were considered as missing data. We used multiple imputations with chained equations to address missing data [15]. A total of 20 imputed sets were created, each a result of 1000 iterations. Missing covariate patterns were individually explored and imputation equations refined. The residuals of the imputation regression models were graphically explored as a form of diagnostics. All analysis on imputed data accounted for the imputed nature of the datasets. Two sided p-values < 0.05 were considered statistically significant. Statistical analyses were performed using Stata version 12.0 (Stata Corp, Texas, USA).