Risk Factors for Barrett Esophagus
We used data from a cross-sectional study conducted at the Michael E. DeBakey Veteran Affairs Medical Center (MEDVAMC) in Houston, Texas. In brief, we recruited consecutive ambulatory patients who were scheduled for a nonurgent, elective esophagogastroduodenoscopy for upper gastrointestinal symptoms at the MEDVAMC from 15 February 2008 to 13 August 2013. We concurrently recruited patients from seven primary care clinics at the MEDVAMC who were eligible for screening colonoscopy from 1 September 2008 to 8 May 2012 to also undergo an upper endoscopy at the time of their screening colonoscopy. These two populations (symptomatic and asymptomatic) represent the underlying population from which BE is diagnosed at the MEDVAMC. Study eligibility was based on the following: (1) for the primary care participants, age between 50 and 80 years (unless there was a reason for screening before age 50, then age >40), and for the elective endoscopy participants, age between 40 and 80 years; (2) no previous or current gastroesophageal surgery or cancer; (3) no active lung, colon, or breast cancer; (4) no current use of anticoagulants, which would be a relative contraindication for mucosal biopsy during endoscopy; (5) no significant liver disease indicated by platelet count <70,000, ascites, or known gastroesophageal varices; and (6) no history of major stroke or mental conditions that would limit the ability to answer questions.
This research was approved by the Institutional Review Boards for the MEDVAMC and Baylor College of Medicine.
All study participants underwent an upper endoscopy with systematic recording of suspected BE. At least one targeted biopsy specimen was taken from suspected BE areas using jumbo biopsy forceps. Using the Prague C and M criteria, BE length was defined as the highest value for circumference length (C) or the maximum (M) extent of the endoscopically visualized BE segment during the index endoscopy.
H. pylori positivity was assigned if (1) organisms were seen and active gastritis was identified on histopathology of any of the gastric biopsy specimens; (2) if H. pylori was detected on culture of gastric samples; or (3) if biopsy results were not available, review of the medical record showed a previous positive biopsy, the presence of serum antibodies, or previous treatment for H. pylori. Gastric biopsies were examined and graded for features of active and chronic gastritis and gastric atrophy according to the standardized operative link for gastritis assessment system, which uses the updated Sydney System. A score of ≥1 on any biopsy from either the antrum or corpus was considered gastritis.
All study participants completed a computerized questionnaire with guidance from a trained research assistant. Race and ethnicity (NHW, AA, Hispanic, Asian, or others) were self-reported on the questionnaire and verified by manual review of the VA Computerized Patient Record System, which contains demographic data and a photograph of each patient. The questionnaire also ascertained information on the history of alcohol and tobacco use; frequency, severity, and chronicity of GERD symptoms; and the use of proton-pump inhibitors (PPIs), H-2 receptor antagonists, aspirin, and nonsteroidal anti-inflammatory drugs (NSAID). The VA Computerized Patient Record System was manually reviewed for the use of statins. We measured each participants' waist and hip circumferences, height, and weight using a standardized protocol and calculated WHR and body mass index.
Participants, irrespective of race or ethnicity, were considered to be a BE case if intestinalized columnar epithelium with goblet cells (confirmed by Alcian-periodic acid–Schiff stain) was present in at least one biopsy sample obtained from areas of suspected BE. Dysplasia (none, low grade, and high grade) in all cases was assessed by a gastrointestinal pathologist and recorded. Subjects with endoscopic BE only (defined by the presence of lesions visibly suspicious for BE in the absence of specialized intestinal epithelium histologically) were excluded from analysis. Controls were defined as participants who underwent an elective upper endoscopy without endoscopically suspected BE.
We restricted our analyses to NHW and AA cases and controls only and compared the following: all cases vs. all controls, all NHWs vs. all AAs, NHW cases vs. NHW controls, AA cases vs. AA controls, and AA cases vs. NHW cases. We calculated the proportions and 95% confidence intervals (CIs) of those with BE among NHWs and AAs, as well as with respect to several sociodemographic and clinical factors. Sociodemographic factors included age, sex (male and female), duration of at least weekly GERD symptoms (none, 1–4, 5–9, and ≥10 years), smoking status (never, former, and current), alcohol status (never and former current), body mass index (normal <25, overweight 25–29.9, and obese≥30), and WHR (low and high). High WHR was defined as >0.9 for men and >0.85 for women, as in previous studies. Clinical factors included H. pylori infection, hiatus hernia (none, <3 cm, and ≥3 cm), active gastritis (antrum vs. corpus; none, low grade, and high grade), chronic gastritis (antrum vs. corpus; none, low grade, and high grade), PPI use (yes and no), H-2 receptor antagonists use (yes and no), NSAID use (yes and no), aspirin use (yes and no), and statin use (yes and no). Race (Caucasian and AA) was analyzed among cases only with regard to mean BE length, BE length stratified as <3 and ≥3 cm (LSBE), and dysplasia (none, low grade, and high grade). We used χ-tests to compare the above variables among the five comparison groups.
In four of the comparison groups (NHW cases vs. NHW controls, AA cases vs. AA controls, AA cases vs. NHW cases, and all cases vs. all controls), we constructed logistic regression models to calculate odds ratios (ORs) and 95% CIs for variables associated with BE. For the four multivariable models examining (1) all cases and controls, (2) NHWs only, (3) AAs only, and (4) cases only, we first found variables that were statistically significant in univariate analysis and performed logistic regression models adjusting for these variables. For multivariable models containing (1) AAs only and (2) BE cases only, we determined that at least 10 outcome events (i.e., AA cases) per independent variable are needed in each model to prevent overfitting of the models and to ensure valid statistical significance. We also created the following composite dichotomous variables for these two multivariable models: GERD or PPI use, any chronic gastritis (in the antrum or corpus), and any active gastritis (in the antrum or corpus). We then retained only those variables that remained statistically significant in the presence of other risk factors through reverse, stepwise selection (P value >0.1). In addition, for all cases and controls, each significant predictor of BE on univariate analysis was adjusted for race in individual multivariable models.
P values were calculated using two-sided statistical tests. Statistical significance was determined at α=0.05. Statistical analyses were performed using SAS 9.2 (SAS Institute, Cary, NC).
Methods
Study Population
We used data from a cross-sectional study conducted at the Michael E. DeBakey Veteran Affairs Medical Center (MEDVAMC) in Houston, Texas. In brief, we recruited consecutive ambulatory patients who were scheduled for a nonurgent, elective esophagogastroduodenoscopy for upper gastrointestinal symptoms at the MEDVAMC from 15 February 2008 to 13 August 2013. We concurrently recruited patients from seven primary care clinics at the MEDVAMC who were eligible for screening colonoscopy from 1 September 2008 to 8 May 2012 to also undergo an upper endoscopy at the time of their screening colonoscopy. These two populations (symptomatic and asymptomatic) represent the underlying population from which BE is diagnosed at the MEDVAMC. Study eligibility was based on the following: (1) for the primary care participants, age between 50 and 80 years (unless there was a reason for screening before age 50, then age >40), and for the elective endoscopy participants, age between 40 and 80 years; (2) no previous or current gastroesophageal surgery or cancer; (3) no active lung, colon, or breast cancer; (4) no current use of anticoagulants, which would be a relative contraindication for mucosal biopsy during endoscopy; (5) no significant liver disease indicated by platelet count <70,000, ascites, or known gastroesophageal varices; and (6) no history of major stroke or mental conditions that would limit the ability to answer questions.
This research was approved by the Institutional Review Boards for the MEDVAMC and Baylor College of Medicine.
Data Collection
All study participants underwent an upper endoscopy with systematic recording of suspected BE. At least one targeted biopsy specimen was taken from suspected BE areas using jumbo biopsy forceps. Using the Prague C and M criteria, BE length was defined as the highest value for circumference length (C) or the maximum (M) extent of the endoscopically visualized BE segment during the index endoscopy.
H. pylori positivity was assigned if (1) organisms were seen and active gastritis was identified on histopathology of any of the gastric biopsy specimens; (2) if H. pylori was detected on culture of gastric samples; or (3) if biopsy results were not available, review of the medical record showed a previous positive biopsy, the presence of serum antibodies, or previous treatment for H. pylori. Gastric biopsies were examined and graded for features of active and chronic gastritis and gastric atrophy according to the standardized operative link for gastritis assessment system, which uses the updated Sydney System. A score of ≥1 on any biopsy from either the antrum or corpus was considered gastritis.
All study participants completed a computerized questionnaire with guidance from a trained research assistant. Race and ethnicity (NHW, AA, Hispanic, Asian, or others) were self-reported on the questionnaire and verified by manual review of the VA Computerized Patient Record System, which contains demographic data and a photograph of each patient. The questionnaire also ascertained information on the history of alcohol and tobacco use; frequency, severity, and chronicity of GERD symptoms; and the use of proton-pump inhibitors (PPIs), H-2 receptor antagonists, aspirin, and nonsteroidal anti-inflammatory drugs (NSAID). The VA Computerized Patient Record System was manually reviewed for the use of statins. We measured each participants' waist and hip circumferences, height, and weight using a standardized protocol and calculated WHR and body mass index.
Cases and Controls
Participants, irrespective of race or ethnicity, were considered to be a BE case if intestinalized columnar epithelium with goblet cells (confirmed by Alcian-periodic acid–Schiff stain) was present in at least one biopsy sample obtained from areas of suspected BE. Dysplasia (none, low grade, and high grade) in all cases was assessed by a gastrointestinal pathologist and recorded. Subjects with endoscopic BE only (defined by the presence of lesions visibly suspicious for BE in the absence of specialized intestinal epithelium histologically) were excluded from analysis. Controls were defined as participants who underwent an elective upper endoscopy without endoscopically suspected BE.
Statistical Analysis
We restricted our analyses to NHW and AA cases and controls only and compared the following: all cases vs. all controls, all NHWs vs. all AAs, NHW cases vs. NHW controls, AA cases vs. AA controls, and AA cases vs. NHW cases. We calculated the proportions and 95% confidence intervals (CIs) of those with BE among NHWs and AAs, as well as with respect to several sociodemographic and clinical factors. Sociodemographic factors included age, sex (male and female), duration of at least weekly GERD symptoms (none, 1–4, 5–9, and ≥10 years), smoking status (never, former, and current), alcohol status (never and former current), body mass index (normal <25, overweight 25–29.9, and obese≥30), and WHR (low and high). High WHR was defined as >0.9 for men and >0.85 for women, as in previous studies. Clinical factors included H. pylori infection, hiatus hernia (none, <3 cm, and ≥3 cm), active gastritis (antrum vs. corpus; none, low grade, and high grade), chronic gastritis (antrum vs. corpus; none, low grade, and high grade), PPI use (yes and no), H-2 receptor antagonists use (yes and no), NSAID use (yes and no), aspirin use (yes and no), and statin use (yes and no). Race (Caucasian and AA) was analyzed among cases only with regard to mean BE length, BE length stratified as <3 and ≥3 cm (LSBE), and dysplasia (none, low grade, and high grade). We used χ-tests to compare the above variables among the five comparison groups.
In four of the comparison groups (NHW cases vs. NHW controls, AA cases vs. AA controls, AA cases vs. NHW cases, and all cases vs. all controls), we constructed logistic regression models to calculate odds ratios (ORs) and 95% CIs for variables associated with BE. For the four multivariable models examining (1) all cases and controls, (2) NHWs only, (3) AAs only, and (4) cases only, we first found variables that were statistically significant in univariate analysis and performed logistic regression models adjusting for these variables. For multivariable models containing (1) AAs only and (2) BE cases only, we determined that at least 10 outcome events (i.e., AA cases) per independent variable are needed in each model to prevent overfitting of the models and to ensure valid statistical significance. We also created the following composite dichotomous variables for these two multivariable models: GERD or PPI use, any chronic gastritis (in the antrum or corpus), and any active gastritis (in the antrum or corpus). We then retained only those variables that remained statistically significant in the presence of other risk factors through reverse, stepwise selection (P value >0.1). In addition, for all cases and controls, each significant predictor of BE on univariate analysis was adjusted for race in individual multivariable models.
P values were calculated using two-sided statistical tests. Statistical significance was determined at α=0.05. Statistical analyses were performed using SAS 9.2 (SAS Institute, Cary, NC).