Results
A total of 1,844 participants underwent EGD as part of the study; 1,263 (68%) were referred for elective EGD and 581 (32%) were patients recruited from primary care for EGD and colonoscopy. Suspected CLE was recorded in 344 (19%) subjects on the index endoscopy (Figure 1). Of these, 237 (13% of the total study population) were found to have definite BE on histopathology (IM) and the rest were CLE without IM. Among patients with definite BE, 146 (62%) had short segment (<3 cm) BE and 91 (38%) had long segment. At least one esophageal biopsy was obtained from all patients. Among patients with CLE without IM, most (101, 94%) had a short segment of CLE. Repeat endoscopy was performed on 80 subjects within 2 years of the index endoscopy. Mean duration between endoscopies was 523 days (1.46 years). The rest of the subjects declined or did not show up for follow-up endoscopy.
(Enlarge Image)
Figure 1.
Flow diagram of enrollment and outcomes of patients in the study. CLE, columnar lined esophagus; EGD, esophagogastroduodenoscopies; IM, intestinal metaplasia.
Of the 80 subjects with CLE and no IM who underwent repeat endoscopy, 96% were male, with an age distribution of 40–49 (14%), 50–59 (26%), 60–69 (46%), and > 70 years (14%). Repeat endoscopy with esophageal biopsies showed that in 24 (30%; 95% CI 20.0–40.0%) subjects with suspected CLE and no IM was not confirmed, 33 (41%; 95% CI 30.2–51.8%) subjects had endoscopically suspected BE (i.e., with CLE) that remained negative histopathology for IM, and 23 (29%; 95% CI 19.1% to 38.9%) subjects with suspected CLE had confirmed BE on histopathology (Figure 1). Of the subjects with suspected CLE and no IM, biopsies showed squamous epithelium in 20 (61%, 95% CI 50.3% to 71.7%), columnar lined epithelium in 11 (33%, 95% CI 22.7–43.3%), and gastric oxyntic mucosa in 2 (6%, 95% CI 0.8–11.2%). There was no dysplasia seen in any of the repeat biopsies in confirmed BE. Nine of the repeat endoscopies were done by the endoscopist who performed the index procedure with an 89% concordance of diagnosing suspected CLE, whereas 71 were done by a different endoscopist with a concordance of 68%. There was no significant difference in the proportions of patients with CLE and no IM or definitive BE according to the identity of the endoscopist.
Comparisons between the subjects with CLE and no IM and those with CLE and IM showed that there was no significant differences in the two groups with respect to age, gender, race, body mass index, presence or duration of GERD, length of CLE, hiatal hernia, Hill flap valve classification, or use of anti-secretory agents (P>0.1) ( Table 1). The mean length of suspected CLE on index endoscopy in patients with definite BE was slightly longer but not statistically significant (1.6 cm, s.d. 1.3) than those with suspected BE and no IM (1.5 cm, s.d. 1.4) or no CLE (1.4 cm, s.d. 1.2) on repeat EGD. The median suspected CLE length was 1 cm for both the groups. The interquartile range for the confirmed BE group was 1 cm (interquartile range 1–2 cm), and 1.5 for the no BE group (interquartile range 0.5–2 cm). Testing the medians and distributions using non-parametric tests yielded non-significant results, both for the groups as a whole (Mann–Whitney U: P=0.33) and for the smaller groups when the lengths <0.5 are excluded (P=0.51). Among patients with 0.5 cm of CLE on index exam, only 19.2% showed definitive BE on repeat EGD. There were no instances of surgical or endoscopic ablation in any of the study patients. There were no significant differences in the use of proton pump inhibitor between those with different CLE lengths. The multivariable logistic regression adjusting for source of referral (endoscopy vs. PCP) there were no significant predictor factors of definitive BE on follow-up endoscopy.