Risk Factors in the Development of Esophageal Adenocarcinoma

Heiko Pohl MD; Katharina Wrobel MD; Christian Bojarski MD; Winfried Voderholzer MD; Amnon Sonnenberg MD; Thomas Rösch MD; Daniel C Baumgart MD


Am J Gastroenterol. 2013;108(2):200-207. 

In This Article



We conducted a case–control study among consecutive patients undergoing a standard upper endoscopy with (i) esophageal adenocarcinoma or high-grade dysplasia (HGD) in Barrett's metaplasia, (ii) Barrett's esophagus without dysplasia or with low-grade dysplasia, (iii) GERD, and (iv) absence of reflux symptoms or esophagitis (no-GERD). The study was conducted at the three hospitals of the Charité University in Berlin, Germany, between December 2005 and August 2009. All patients were identified at the time of endoscopy, from medical records, or from the endoscopic database. The sample frame for the cancer/HGD group consisted of all 244 patients who were diagnosed with esophageal adenocarcinoma or HGD at the participating medical centers between July 2002 and January 2008. Potential participants for the Barrett group included 298 patients who were diagnosed with Barrett's esophagus without HGD between December 2005 and August 2009. Patient controls with GERD and no-GERD were selected from the endoscopic database of one of the hospitals during the same time period. To select a representative sample of GERD and no-GERD patients, we contacted every fifth consecutive patient who met inclusion and exclusion criteria. If we were unable to reach the patient after two attempts, then the next following patient in the database (i.e., the 10th patient) was contacted.

The diagnosis of Barrett's esophagus was based on its endoscopic appearance and a confirmatory pathology report. All pathology evaluations were carried out by two independent expert pathologists. Given the high risk of prevalent cancer or of progression to cancer[15,16] for patients with Barrett's esophagus and HGD, it was assumed that HGD and esophageal adenocarcinoma shared the same risk factors. The two findings were grouped together as done in previous studies.[2,7,17]

Subjects in the GERD group included all patients who underwent an upper endoscopy for work-up of typical reflux symptoms or who were found to have reflux esophagitis (at least Savary-Miller stage 1 or Los Angeles classification grade A). Potential subjects for the no-GERD group included all those who underwent an upper endoscopy for reasons other than reflux disease. Exclusion criteria for the GERD and no-GERD groups included an American Society of Anesthesiologists (ASA) class >3 and a history of gastrointestinal surgery or malignancy. Patients who were identified as subjects for the no-GERD group, but who had esophagitis on upper endoscopy or who reported at least weakly reflux symptoms during the survey evaluation were subsequently adjudicated to the GERD group. Because potential study subjects were identified after completion of the endoscopy, some patients could not be contacted or had died by the time the survey was conducted. Figure 1 details patient inclusion into the study. The study protocol was approved by the Institutional Review Board of the Charité University Hospitals.

Figure 1.

Flow chart. Selection and inclusion of study participants. ASA, American Society of Anesthesiologist classification; GERD, gastroesophageal reflux disease; HGD, high-grade dysplasia.

Data Collection

Potential study subjects were contacted by phone or during a subsequent clinic visit and asked to participate. Enrolled patients completed a standardized questionnaire about the history and duration of reflux symptoms, smoking history, dietary habits (time of largest meal, fruit and vegetable intake), BMI at the age of 40, history of diabetes mellitus, and infection with Helicobacter pylori. We extracted information on Barrett length, possible esophagitis, and hiatal hernia from the endoscopy report, and information on dysplasia from the pathology report. Medical records were also reviewed with regard to history of diabetes mellitus, and H. pylori infection. A patient was considered to have had a history of H. pylori infection if supported by medical records or reported by the patient. We also collected information on medication use (proton-pump inhibitors or histamine-2 receptor blocker, aspirin and other non-steroidal anti-inflammatory drugs and statins). Details about the onset of drug use and its relationship to the time of diagnosis of Barrett's esophagus or GERD were not available.

Statistical Analysis

We examined the association between patient characteristics and different disease stages, ranging from the absence of reflux disease to cancer. We applied multivariate logistic regression analysis to calculate adjusted odds ratios (OR) with their 95% confidence interval (CI), using the statistical software Stata 11.0 (StataCorp LP, College Station, TX, USA). The regression model was developed based on the comparison between the cancer/HGD and the no-GERD groups. We included a pre-selected set of variables into the regression model (age, gender, BMI at age 40, and tobacco) and tested additional variables for significance. Only variables that maintained significance and changed the point estimates by >10% were included in the final model. We computed tests for trends after exclusion of missing data across categories of risk factors and for continuous variables using their continuous data values. Continuous variables were compared using the Student's t-test if normally distributed or the Mann–Whitney U test otherwise. Categorical variables were compared using the Chi-squared test. Based on previous epidemiologic studies, we estimated that common risk factors such as male gender or presence of hiatal hernia would rise in increments of about 20% among the four study groups. Considering an α-error of 5%, a β-error of 20%, and depending on the prevalence rate in the comparison group, it was estimated that the individual groups should contain at least 100 subjects. Patient recruitment was stopped after 100 patients had been recruited into the cancer/HGD group.