Long-term endoscopic surveillance of Barrett's esophagus

Conio et al. (1) have made an important contribution to the literature on the incidence of adenocarcinoma (AC) of the esophagus in patients with Barrett’s esophagus (BE). One hundred fifty-five patients with BE were followed for a mean of 5.5 yr in this prospective study. Only one study, which was retrospective, has more patient-years of follow-up (2). The incidence of AC was 1/220 patient-years follow-up, or 0.45% per year. This finding is similar to those from recent studies, such as Drewitz et al. (3) and O’Connor et al. (4), as well as the Shaheen et al. funnel analysis (5). There are a number of significant strengths of this prospective cohort study. All the patients had newly diagnosed BE. The onset of BE is never really known, only the date of the first endoscopy and biopsy establishing the diagnosis. Patients with high grade dysplasia at baseline endoscopy were excluded. Thus, only patients with low grade dysplasia or no dysplasia were included. The “purity” of this welldefined population is a major strength of the study. Patients had to have a minimum of six biopsies of the Barrett’s epithelium to be included. The criteria for dysplasia were well defined, and one pathology unit read all of the biopsies. Compliance with surveillance was defined and evaluated. Only 55% of the patients were judged to be compliant in this eight-center study. Surprisingly, this is better than the 42% compliance found in a prior single-center study of 60 patients (6). As the authors suggest, this incomplete compliance might reflect “clinical practice” but also points out how far we have to go to achieve more effective surveillance. Barriers to compliance with surveillance in patients with BE need to be established and means of dealing with them explored. Three of the five cancer patients failed to follow study recommendations: case 1 refused surgery, and cases 2 and 5 had prolonged endoscopy intervals. Was the cancer found in case 2 at 35 months an unusually rapidly progressing or a missed cancer? It is not clear how many patients were excluded from surveillance because of “significant comorbid disease.” Uniform criteria were not used, and the outcome of these excluded patients is not defined. When results from Table 2 in this report are totaled, 5% of the deaths in these cancer incidence reports were due to AC. This low death rate from cancer emphasizes that most patients with BE do not die from AC of the esophagus. It is interesting that the percent of deaths from AC in the surveyed patients was 8%, compared with 4% of the deaths in those with no or “some” surveillance, a nonsignificant difference. The mean patient age in this series was 60 yr, similar to the other sited series, with the mean age ranging from 57 to 63 yr. Were the 500 endoscopies and more than 4000 endoscopies worthwhile? Were lives saved? Clearly, a large number of invasive tests were performed to the benefit of very few patients. In this series, surveillance did not prevent death from AC. This surveillance experience highlights the need for better risk stratification of patients with BE. Some risk factors are known: gender, ethnicity, age, and high grade dysplasia. We are unable to impact the first three factors, and their risk is not quantitated. Also, as demonstrated in Table 3 of the report, there is variability in the risk of high grade dysplasia progressing to AC. Mucosal irregularity is a predictor of greater likelihood of cancer (relative risk vs no nodularity, 8.95) (7) and of invasion beyond the lamina propria (patients with an endoscopically visible lesion, p 0.01) (8). It is not clear whether any patients in this series had mucosal irregularities before the development of cancer. Three biomarkers have been shown to correlate with the future development of cancer in cohort studies: abnormal flow cytometry (aneuploidy and increased 4N) (9), p17 (p53) loss of heterozygosity (10), and cyclin D1 over-expression (11). These markers have yet to be confirmed in multicenter trials. To date, defining a stratification index has been elusive. Such an index might include factors such as gender, ethnicity, age, dysplasia, body mass index, smoking status, and a panel of biomarkers. An evidence basis for specific criteria needs to be defined. The message is clear: cost-effective surveillance of BE will require more effective risk stratification of patients.