Jesper Lagergren

Association between Body Mass and Adenocarcinoma of the Esophagus and Gastric Cardia

The reasons for the increase in the incidence of esophageal and gastric cardia adenocarcinoma in several industrialized countries (1-4) are largely unknown (5). In the United States, the relative increase in the incidence of esophageal adenocarcinoma exceeds that of any other type of cancer (1). Obesity, which has been found to be linked with a moderately increased risk for adenocarcinoma in some (6-9) but not all (10, 11) previous epidemiologic studies, is one factor that could explain this epidemic because the prevalence of obesity has increased concomitantly (12-14). No evidence shows a link between obesity and squamous-cell carcinoma (the other main histologic type of esophageal cancer) or adenocarcinoma of the distal stomach (9). As part of a nationwide casecontrol study of risk factors for esophageal and gastric cardia adenocarcinoma, we analyzed body measures in relation to the risk for developing these tumors. Our primary aim was to precisely estimate the strength of the association while adjusting for potential confounding factors. We also examined the effects of body mass early in life compared with its effects late in life and assessed the relative importance of physical activity and energy intake. Methods Participants Our methods have been described in detail previously (15). The study base consisted of all Swedish residents who had been born in Sweden, were younger than 80 years of age, and lived in Sweden from 1995 through 1997. All persons in the study base who developed new esophageal or cardia adenocarcinomas and persons who developed esophageal squamous-cell carcinoma and were born on even dates were eligible as case-patients. We included 189 patients with esophageal adenocarcinoma, 262 patients with cardia adenocarcinoma, and 167 patients with esophageal squamous-cell carcinoma. These persons constituted 87%, 83%, and 73%, respectively, of all eligible patients in the study base. Physical disorders, mental disorders, or early death prevented participation in 12% to 22% of the eligible patients, and 1% to 5% of patients declined to participate. The 820 population-based controls represented 73% of those originally selected; of these persons, 19% declined to participate, 6% had physical or mental disorders, and 2% could not be contacted. The proportions of men among patients with esophageal adenocarcinoma, cardia adenocarcinoma, and esophageal squamous-cell carcinoma were 87%, 85%, and 72%, respectively; 83% of controls were men. The median ages for both sexes in these four groups were 69 years, 66 years, 67 years, and 68 years, respectively. The study was approved by all regional ethics committees in Sweden, and individual informed consent was obtained from each participant before the interview. Exposure Information At face-to-face interviews conducted by trained professional interviewers, participants were asked about height and weight at 20 years of age and at 20 years before the interview. Body mass index (BMI) (16) was calculated as body weight in kilograms divided by the square of body height in meters (kg/m2). To obtain an independent measure of body fat, we asked each interviewee to choose the picture that best resembled his or her body build 20 years ago from a pictogram that showed nine somatotypes ranging from very lean to grossly obese. The Spearman correlation coefficient with BMI is 0.5 to 0.8 (17). Weight change between 20 years of age and 20 years before the interview was calculated in kilograms and in percentage of body weight at 20 years of age. Meal sizeself-reported by each participant by using photographs of seven common disheswas divided into three categories and used as a measure of energy intake. This method gives an acceptably valid representation of energy intake (18, 19). Physical activity was divided into four levels by using a combination of 12 variables, including usual activities, such as standing, walking, and climbing stairs; physical activities during leisure time; and physical exertion at work. The interviewers could not be blinded to the casecontrol status of the interviewees, but they were completely blinded to the study hypothesis and were urged to treat patients and controls equally. Clinical Information All 195 hospital departments involved in the diagnosis or management of these patients collaborated in the recruitment of patients. The regional tumor registries helped us to identify missed cases. We used a common protocol for uniform documentation and classification of the tumors. At endoscopy, we measured the distances between the gastroesophageal junction (defined as the point at which the proximal longitudinal mucosal folds begin in the stomach) and the upper and lower borders of the tumor. The protocol also prescribed that serial biopsy specimens should be taken at points 2 cm apart, from the proximal stomach, through the gastroesophageal junction, and into the esophagus until normal squamous-cell epithelium was reached. Additional specimens were to be obtained proximal, distal, and lateral to the tumor. Surgeons and pathologists were required to provide standardized, detailed descriptions of the location of the cancer in the 424 patients in whom surgery was performed. In addition, 97% of all biopsy and surgical specimens were reexamined by one pathologist. Barrett esophagus was defined as columnar-cell metaplasia of the specialized type, with goblet cells and villiform surface configuration of the mucosa resembling the features of the intestines. Cancer of the gastric cardia had to have its center within 2 cm proximal, or 3 cm distal, to the gastroesophageal junction. If Barrett esophagus was detected adjacent to the tumor, it was classified as esophageal regardless of its location. We classified squamous-cell carcinomas as esophageal even if they were located in the gastric cardia. Population-based controls were randomly selected from age and sex strata in the complete and continuously updated population register to obtain an age and sex distribution similar to that of patients with esophageal adenocarcinoma. Statistical Analysis Logistic regression was used in univariate and multivariate analyses of the relation among anthropometric measures, potential confounding variables, and cancer risk. Patients in each category of cancer were compared with all controls. Model variables were estimated by the maximum likelihood method (20) and converted to odds ratios and 95% CIs. In the baseline model, adjustments were made for age (in 5-year classes) and sex; in the multivariate analyses, we further adjusted for tobacco smoking (never-smokers, former smokers, and current smokers 2 years before the interview), alcohol use (grams of pure alcohol in four classes), years of formal education (in four classes), reflux symptoms (heartburn or regurgitation at least once per week occurring at least 5 years before the interview), intake of fruit and vegetables (in four classes), meal size, and physical activity. Variables were considered in both categorized and continuous form. Wald tests and likelihood ratio tests were used. The relation of BMI to physical activity and meal sizeadjusted for age, sex, and diagnostic groupwas analyzed by linear regression analysis. Role of the Funding Sources The funding sources had no role in the collection, analysis, or interpretation of the data or in the decision to submit the paper for publication. Results Esophageal Adenocarcinoma We found a positive association between BMI 20 years before the interview and risk for esophageal adenocarcinoma. This association was dose-dependent across the entire range of BMI values (Table 1). The multivariate-adjusted odds ratio was 7.6 (95% CI, 3.8 to 15.2) among persons in the highest BMI quartile compared with those in the lowest BMI quartile. The highest BMI quartile included both overweight and obese persons, as defined according to World Health Organization criteria (21). We therefore reanalyzed the association using more commonly applied BMI cutoff points. Obese persons (persons with BMI>30 kg/m2) had an odds ratio of 16.2 (CI, 6.3 to 41.4) compared with persons whose BMI was less than 22 kg/m2. Table 1. Body Measures and Risk for Esophageal Adenocarcinoma The association between esophageal adenocarcinoma and BMI was supported by our finding of a dose-dependent relation between esophageal adenocarcinoma and somatotype (P for trend<0.001). The adjusted odds ratio among persons who reported that their body build 20 years before the interview corresponded to one of the four obese somatotypes in our pictogram was 4.1 (CI, 2.6 to 6.7) compared with persons reporting one of the three lean somatotypes. Weight gain after 20 years of age, both in absolute values and expressed as percentage of weight at 20 years of age, increased the risk monotonically (P for trend<0.001); after adjustments for BMI 20 years before the interview, however, this effect disappeared almost entirely. Height was inversely and significantly (P=0.03) associated with risk. A 50% risk reduction was found among the tallest persons compared with the shortest persons. Stratified analyses showed a seemingly stronger association between BMI and risk for esophageal adenocarcinoma among persons 60 to 69 years of age compared with those in other age strata and among women compared with men; however, the CIs were wide and the differences were not statistically significant (Table 2). No important confounding was caused by any of the seven additional covariates. The presence of reflux symptoms at least 5 years before the interview did not affect the association between BMI and esophageal adenocarcinoma. Therefore, BMI was a risk factor independent of the occurrence of gastroesophageal reflux. The strength of the association with BMI was equally strong among the 118 patients with histologically verified Barrett esophagus and the 71 patients without this metaplasia (data not shown). Table 2. Stratified Analyses of the Association between

British Society of Gastroenterology guidelines on the diagnosis and management of Barrett's oesophagus

These guidelines provide a practical and evidence-based resource for the management of patients with Barretts oesophagus and related early neoplasia. The Appraisal of Guidelines for Research and Evaluation (AGREE II) instrument was followed to provide a methodological strategy for the guideline development. A systematic review of the literature was performed for English language articles published up until December 2012 in order to address controversial issues in Barretts oesophagus including definition, screening and diagnosis, surveillance, pathological grading for dysplasia, management of dysplasia, and early cancer including training requirements. The rigour and quality of the studies was evaluated using the SIGN checklist system. Recommendations on each topic were scored by each author using a five-tier system (A+, strong agreement, to D+, strongly disagree). Statements that failed to reach substantial agreement among authors, defined as >80% agreement (A or A+), were revisited and modified until substantial agreement (>80%) was reached. In formulating these guidelines, we took into consideration benefits and risks for the population and national health system, as well as patient perspectives. For the first time, we have suggested stratification of patients according to their estimated cancer risk based on clinical and histopathological criteria. In order to improve communication between clinicians, we recommend the use of minimum datasets for reporting endoscopic and pathological findings. We advocate endoscopic therapy for high-grade dysplasia and early cancer, which should be performed in high-volume centres. We hope that these guidelines will standardise and improve management for patients with Barretts oesophagus and related neoplasia.

Extended abdominoperineal resection with gluteus maximus flap reconstruction of the pelvic floor for rectal cancer.

Intraoperative tumour perforation, positive tumour margins, wound complications and local recurrence are frequent difficulties with conventional abdominoperineal resection (APR) for rectal cancer. An alternative technique is the extended posterior perineal approach with gluteus maximus flap reconstruction of the pelvic floor. The aim of this study was to report the technique and early experience of extended APR in a select cohort of patients.

Screening and surveillance for Barrett esophagus in high-risk groups: a cost-utility analysis.

Context Barrett esophagus, a complication of gastroesophageal reflux disease (GERD), is associated with an increased risk for esophageal cancer. Some have proposed endoscopic screening for Barrett esophagus, followed by surveillance for cancer in patients with the disorder. Contribution The authors modeled the cost-effectiveness of screening 50-year-old white men with GERD. Screening followed by surveillance in patients with Barrett esophagus and dysplasia cost

The role of tobacco, snuff and alcohol use in the aetiology of cancer of the oesophagus and gastric cardia

10 440 per quality-adjusted life-year (QALY). Surveillance of patients with Barrett esophagus but no dysplasia every 5 years would cost an additional

Adenocarcinoma of oesophagus: what exactly is the size of the problem and who is at risk?

596 000 per QALY. Implications One-time screening for Barrett esophagus in 50-year-old men with GERD is reasonably cost-effective if surveillance is limited. The Editors Current guidelines from the American College of Gastroenterology recommend surveillance to detect cancer and dysplasia in patients with Barrett esophagus, a columnar-cell metaplasia that replaces the native squamous-cell epithelium of the distal esophagus (1). The costs and benefits of surveillance have been evaluated in retrospective studies (2-5) and formal decision analyses (6, 7). Barrett esophagus alone does not cause symptoms that would prompt endoscopic evaluation, and the question of whether screening strategies to detect Barrett esophagus are reasonable and, if so, in which patients is unresolved (8). Adding to the uncertainty is emerging evidence from prospective studies that the incidence of cancer in patients with Barrett esophagus may be considerably lower than previously reported (9, 10). In the absence of randomized, controlled trials of screening and surveillance in Barrett esophagus, the costs and benefits of strategies to decrease mortality rates from cancer are unknown. We used decision analysis to determine whether current recommendations for screening of patients with gastroesophageal reflux disease (GERD) represent cost-effective care. The cohort that we modeled consisted of white men 50 years of age who have symptoms of GERD, because this group is at high risk on the basis of epidemiologic evidence demonstrating an increased risk for esophageal adenocarcinoma compared with other groups [11-14]. Specifically, we examined strategies for screening patients with symptoms of GERD for the presence of Barrett esophagus, dysplasia, or cancer and compared these strategies with no screening for Barrett esophagus. Secondary objectives were 1) to determine whether more benefit was provided by the initial screening examination or by subsequent surveillance, 2) to evaluate different intervals of surveillance in patients with Barrett esophagus, and 3) to identify clinical variables critical to the determination of cost-effectiveness. Methods Patients The hypothetical cohort consisted of white men 50 years of age who had symptoms consistent with GERD (heartburn or acid regurgitation). These patients were assumed not to have symptoms or signs that would otherwise prompt endoscopic evaluation (dysphagia, anemia, weight loss, or bleeding) and had not been previously screened for Barrett esophagus. The analysis assumed that Barrett esophagus was defined by salmon-colored mucosa of any length at endoscopy and intestinal metaplasia on histologic examination. Intestinal metaplasia of the gastroesophageal junction in the absence of endoscopic identification of Barrett esophagus was not examined. It was assumed that all patients modeled for participation in screening and surveillance would be surgical candidates for esophagectomy and would consent to undergo this operation if esophageal cancer was diagnosed. The analysis followed the cohort until 80 years of age or death. Model A decision analysis model was created by using DATA 4.0 software (TreeAge, Williamstown, Massachusetts). A Markov model was used to analyze patients with symptomatic GERD (15, 16). Figure 1 shows a simplified outline of the structure of the Markov model. The actual model contains more than 7000 nodes (branch points) that encompass the natural history of patients with GERD compared to strategies of screening and surveillance for Barrett esophagus, dysplasia, and cancer. (The Appendix provides further information on model assumptions, including structure, transitions, utilities, and sensitivity analysis.) Figure 1. Markov model. Natural History Costs and quality-adjusted life-years (QALYs) without screening and surveillance for Barrett esophagus and adenocarcinoma of the esophagus were examined. Cancer would be diagnosed only if symptoms (dysphagia, weight loss, bleeding, or pain) prompted endoscopic evaluation. Depending on the stage of cancer at the time of diagnosis, palliation (esophageal stent or chemoradiation) or surgical resection could be attempted. Patients with unresectable disease accrued costs of hospice or palliative care. Patients in whom resection was attempted could die after the procedure or survive. Patients who survived surgery could be cured of cancer or experience persistent or recurrent cancer, rates of which depended on the stage of cancer at diagnosis. Death from causes other than adenocarcinoma of the esophagus was incorporated in an age-dependent manner. Screening and Surveillance The natural history analysis was compared to screening with two surveillance strategies. Both strategies used screening endoscopy at 50 years of age and biopsy done to confirm Barrett esophagus if abnormalities were seen at endoscopy. Identification of adenocarcinoma by the initial screening examination prompted esophagectomy or palliation, depending on the stage of tumor. Palliation was associated with costs of endoscopic stenting or chemotherapy and radiation. Surgery could result in perisurgical death, persistent or recurrent cancer, or cure. The first strategy limited surveillance to patients with Barrett esophagus with dysplasia at the initial examination. White men with symptoms of GERD would undergo screening endoscopy at 50 years of age; if Barrett esophagus without dysplasia or no Barrett esophagus was diagnosed, further surveillance would not be performed. Patients with Barrett esophagus with dysplasia would undergo surveillance endoscopy every 6 months for low-grade dysplasia or every 3 months for high-grade dysplasia as long as dysplasia was noted at least once during the preceding 12 months. Detection of cancer would prompt esophagectomy or palliation, depending on stage of disease at diagnosis. The second strategy also evaluated screening of white men with symptoms of GERD at 50 years of age. Patients with Barrett esophagus with dysplasia underwent surveillance as in the first strategy; however, those with Barrett esophagus without dysplasia were separately modeled to undergo surveillance endoscopy at intervals of 2, 3, 4, or 5 years. If intestinal metaplasia was not found on two consecutive surveillance endoscopies, surveillance ceased. Cancer was managed as in the first strategy. Endoscopy and biopsy was modeled to be an imperfect test: that is, false-positive and false-negative results would occur, thereby missing some cases of dysplasia or cancer and overdiagnosing other states. Procedure-related morbidity and mortality for surgery and endoscopy were included in the analysis. Adjustment for patient preferences for the various health states (utilities) was incorporated into the model. Differential rates of cure from cancer were assigned depending on the stage at diagnosis. Death from causes other than adenocarcinoma of the esophagus was incorporated in an age-dependent manner. All costs and benefits were discounted at 3% annually (0% to 5% annually in the sensitivity analysis). Transitions Transitions between health states of the Markov model were derived from the published literature. To assess key variables, the MEDLINE and EMBASE databases from 1970 to 2001 and abstracts from major gastroenterological scientific meetings from 1999 to 2001 were searched by using the terms Barretts esophagus, esophageal neoplasms, adenocarcinoma, precancerous conditions, and gastroesophageal reflux disease. If no data existed for a specified transition, the authors reached consensus. In addition to the base-case scenario, sensitivity analyses were performed using variations around each of the transitions supported by the literature (Table 1). Table 1. Model Assumptions Costs and Utilities Costs, not charges, were the basis for inputs used in this analysis. Costs included direct costs of care for delivery of services and were derived from published literature and 2001 data from the Health Care Financing Administration. Data on inpatient resource use from the Health Care Financing Administration were based on diagnosis-related groups and Current Procedural Terminology codes, whereas outpatient data were based on ambulatory payment classification and Current Procedural Terminology codes. Costs of endoscopy included biopsies and histologic interpretation. Costs are shown in Table 1. A third-party insurer perspective was taken (58-60). Patient preferences for different health states (utilities) were derived from the published literature, and outcomes were adjusted to derive QALYs (6, 7). Utilities were derived from expert opinion by using a time-tradeoff technique (7) and from responses by patients who had undergone esophagectomy for high-grade dysplasia or cancer (6). Costs of time lost from work and support required from family or friends and intangible losses due to pain and suffering were not incorporated. Sensitivity Analysis All assumptions for the model were varied over a wide range of values in a series of one-way sensitivity analyses. Published rates from the literature were used if available. In the absence of published data, baseline rates were halved and doubled with adjustment by consensus from the authors. Monte Carlo simulation was used to create a multi-way sensitivity analysis. Surgery for High-Grade Dysplasia In the base-case strategy, the diagnosis

Lifestyle related risk factors in the aetiology of gastro-oesophageal reflux

While tobacco and alcohol are established risk factors for oesophageal squamous‐cell carcinoma, their roles in the aetiology of the increasingly common oesophageal adenocarcinoma remains uncertain. We tested the association between tobacco, snuff and alcohol use and the risk of oesophageal and cardia cancer in a nationwide, population‐based case‐control study in Sweden. Face‐to‐face interviews were conducted with 618 (81% of all eligible) patients (189 oesophageal adenocarcinoma, 262 cardia adenocarcinoma and 167 oesophageal squamous‐cell carcinoma) and 820 control subjects. Odds ratios (OR) were calculated by logistic regression with multivariate adjustments for potential confounding. The risk of oesophageal adenocarcinoma was not associated with snuff or alcohol use, and the association with smoking was weak or absent. Gastric cardia adenocarcinoma was dose‐dependently associated with smoking (OR=4.2, 95% CI=2.5–7.0 among heavy smokers compared with never‐smokers), but not with alcohol or snuff use. Oesophageal squamous‐cell carcinoma was strongly associated with tobacco, moderately with alcohol, but not with snuff use; combined use of tobacco and alcohol entailed a strongly increased risk (OR=23.1, 95% CI=9.6–56.0 among heavy users compared with never‐users). We conclude that tobacco smoking, a strong risk factor for oesophageal squamous‐cell carcinoma and cardia adenocarcinoma, does not play an important role in the aetiology of oesophageal adenocarcinoma. None of the studied exposures can explain the increasing incidence of oesophageal adenocarcinoma. Int. J. Cancer 85:340–346, 2000. ©2000 Wiley‐Liss, Inc.

Association between Medications That Relax the Lower Esophageal Sphincter and Risk for Esophageal Adenocarcinoma

The incidence of oesophageal adenocarcinoma is increasing and the prognosis is poor. There is a strong predominance of white males, and heredity plays a minor role. The established risk factors are Barrett’s oesophagus, gastro-oesophageal reflux, and obesity. Infection with Helicobacter pylori and use of non-steroidal anti-inflammatory drugs might reduce the risk. Medications that relax the lower oesophageal sphincter might contribute to increasing the risk. Among dietary factors, low intake of fruit, vegetables, and cereal fibres seem to increase the risk of oesophageal adenocarcinoma. The role of tobacco smoking is probably limited and alcohol consumption is not a risk factor. It is uncertain which factors cause the increasing incidence. Increasing prevalences of reflux and obesity, and decreasing prevalence of H pylori infection may contribute to this development; however, the sex distributions of these factors do not match the incidence trends well. Endoscopic surveillance for oesophageal adenocarcinoma among persons with reflux and obesity is discussed, but presently there is no evidence that strongly supports such a strategy.

Survival after surgery for oesophageal cancer: a population-based study

Background/Aim: The aetiology of gastro-oesophageal reflux is largely unknown. The authors’ aim was to examine the relation between lifestyle habits and gastro-oesophageal reflux symptoms. Subjects: Participants of two consecutive public health surveys in Nord-Trondelag, Norway. Methods: In a case control study within the two public health surveys, 3153 individuals who in the second survey reported severe heartburn or regurgitation during the last 12 months were defined as cases, while 40 210 people without reflux symptoms constituted the control group. The risk of reflux symptoms was estimated and multivariately calculated as odds ratios in relation to exposure to tobacco smoking, alcohol, coffee, tea, table salt, cereal fibres, and physical exercise. Results: There was a significant dose response association between tobacco smoking and reflux symptoms. Among people who had smoked daily for more than 20 years the odds ratio was 1.7 (95% confidence interval 1.5 to 1.9) compared with non-smokers. A similar positive association was found for table salt intake. The odds ratio for reflux was 1.7 (95% CI 1.4 to 2.0) among those who always used extra table salt compared with those who never did so. We found moderately strong negative associations between the risk of reflux and exposure to coffee, bread high in dietary fibre content, and frequent physical exercise. Intake of alcohol or tea did not affect the risk of reflux. Conclusions: Tobacco smoking and table salt intake seem to be risk factors for gastro-oesophageal reflux symptoms. Dietary fibres and physical exercise may protect against reflux. Alcohol, coffee, and tea do not seem to be risk factors for reflux.

Antioxidants and cancers of the esophagus and gastric cardia.

In recent decades, the annual incidence of esophageal adenocarcinoma has increased dramatically (from 2% to 17%), predominantly among men in the western world (1-7). Among white men in the United States, the incidence of this cancer is increasing more rapidly than that of any other cancer (1). The increase can be traced back at least to the early 1970s (4). A coinciding but more moderate (1% to 5%) increase in the incidence of adenocarcinoma of the gastric cardia has been reported, but some studies have found no such increase (5, 6). The incidence of esophageal adenocarcinoma among men in Sweden is 1.1 per 100 000 persons. The reasons for the increasing incidence of esophageal adenocarcinoma are unknown, but gastroesophageal reflux was recently established as a strong risk factor (8). We and others (9) hypothesized that medications that facilitate gastroesophageal reflux by decreasing the pressure of the lower esophageal sphincter (LES) may be responsible for the increase in the incidence of esophageal adenocarcinoma. In an ecological study, a general per capita increase in the purchase of LES-relaxing agents from 1957 to 1986 was found in the United States (9). We identified five groups of LES-relaxing medications that were introduced before the 1970s in Sweden and have been widely used: nitroglycerins, anticholinergics, -adrenergic agonists, aminophyllines, and benzodiazepines. The LES-relaxing effect of these types of drugs is well documented (10-22). Such LES-relaxing drugs as -adrenergic antagonists, nicotine derivatives, tricyclic antidepressants, and chlorpromazines have had more limited use in Sweden. Calcium-channel blockers were not introduced on the Swedish market until the late 1970s; assuming a 20-year latency period, such drugs would be unlikely causes of the cancers observed in our study. We sought to examine the possible association between use of LES-relaxing agents and the risk for adenocarcinoma of the esophagus and gastric cardia. Methods Design This study was part of a large population-based casecontrol investigation that was described in detail elsewhere (8). In brief, the study base consisted of all native residents of Sweden who were younger than 80 years of age and were living in Sweden from 1995 to 1997. All persons in the study base who developed esophageal or cardia adenocarcinoma and half of those who developed esophageal squamous-cell carcinoma (those born on even dates) were eligible as case-patients. All Swedish hospital departments involved in the diagnosis or management of these patients collaborated in the recruitment of new case-patients. Regional tumor registries helped us to identify missed cases. We used frequency matching (23) to assemble a control group with an age and sex distribution similar to that of the case-patients. Population-based controls were randomly selected from 10-year age- and sex-matched strata in a complete and continuously updated population register. Tumor Classification Standardized routines were used for uniform documentation and classification of the tumors, including serial biopsies of the esophagus and gastroesophageal junction; a study protocol for measurements and histopathologic sampling of surgical specimens; and reporting of tumor sites by endoscopists, surgeons, and pathologists. In addition, 97% of all biopsy and surgical specimens were reexamined by one pathologist. Adenocarcinomas with a center within 2 cm proximal or 3 cm distal to the gastroesophageal junction were classified as cancers of the gastric cardia. In seven patients in whom Barrett esophagus (24) was detected adjacent to a tumor located in the gastric cardia, the tumor was classified as esophageal. Participants The study included 189 patients with esophageal adenocarcinoma, 262 with adenocarcinoma of the gastric cardia, and 167 with esophageal squamous-cell carcinoma. These numbers constituted 87%, 83%, and 73%, respectively, of all eligible case-patients in the study base. The reasons for nonparticipation of eligible case-patients were physical or mental disorders or early death in 12% to 22% and unwillingness in 1% to 5%. The 820 controls constituted 73% of all who had been selected; 19% declined to participate, 6% had a physical or mental disorder, and 2% could not be traced. Of the patients with esophageal adenocarcinoma, those with adenocarcinoma of the gastric cardia, and those with esophageal squamous-cell carcinoma, 87%, 85%, and 72%, respectively, were men; among the controls, this proportion was 83%. The median ages for persons of either sex in these groups were 69, 66, 67, and 68 years, respectively. The study was approved by all regional ethics committees in Sweden, and individual informed consent was obtained from all participants before the interview. Exposure Information All participants were interviewed in person by specially trained professional interviewers who used a computerized questionnaire. Range checks and logic consistency controls were integrated into the questionnaire. The interviewers could not be blinded to the case or control status of the interviewees, but they were unaware of the study hypothesis and were urged to treat case-patients and controls in a strictly equal manner. Participants were questioned about regular use of medications for angina pectoris, abdominal disorders, asthma, hypertension, sleeping problems, and psychiatric diseases. They were then asked to specify the names of the selected drugs that had an LES-relaxing effect and the time period, duration, and frequency of the use. To facilitate recall of past use of such drugs, the interviewees were shown a booklet with color photographs of historical drug packages together with the brand names. The nine classes of LES-relaxing medications listed in the introduction section, representing 178 separate brands of drugs, were investigated. The generic names of the drugs in each of main five classes of LES-relaxing drugs are listed in the Appendix Table. Only persons who could remember the precise name of the medication were classified as exposed. To avoid the possibility that the indication for treatment acted as a confounder (protopathic bias [25]), we disregarded use of medications in the 5 years before the interview. We also asked about the use of some medications that do not influence LES pressure, such as long-term antibiotic therapy and medications for cough or sore throat. In addition, we sought information about several potential confounding factors discussed below. Statistical Analysis Logistic regression was used in univariate and multivariate analyses of the relation among use of LES-relaxing medications, covariates, and risk for cancer. Case-patients with each type of cancer were compared with all controls. We used PROC LOGISTIC SAS, version 6.12 (SAS Institute, Cary, North Carolina), for analyses. The measure of association in casecontrol studies is the odds ratio, which is an estimate of relative risk. In studies of rare outcomes, such as esophageal adenocarcinoma, and when the associations are moderately strong, the odds ratios are approximations of the relative risk (26). In our study, sampling of controls was distributed over the entire study period. Therefore, our odds ratios can be interpreted as estimates of incidence rate ratios (IRRs) with 95% CIs. We adjusted our estimated IRRs for known or potential risk factors for esophageal cancer because such variables may affect the association between LES-relaxing drugs and risk for cancer. In the baseline model, we adjusted for age (in 5-year groups) and sex; in multivariate analyses, we further adjusted for body mass index 20 years before the interview (quartiles), tobacco smoking (never-smokers, ex-smokers, and current smokers 2 years before the interview), alcohol use (grams of alcohol; four groups), socioeconomic status (years of formal education; four groups), and intake of fruit and vegetables (four groups). All covariate categories were entered as separate indicator variables, allowing for nonlinear effects, and all terms mentioned above were entered in the final multivariate model. We also considered physical activity and total energy intake, but because these variables seemed to be unrelated to risk for cancer in our study, they were not included in the final model. In another multivariate model, we also adjusted for symptoms of reflux, defined as recurrent heartburn or regurgitation at least once a week for 1 year or longer. Our primary goal was to test the association between LES-relaxing agents and the risk for esophageal tumors. Other drugs were considered only as negative control exposures. The number needed to treat to cause one case of esophageal adenocarcinoma (number needed to treat to harm [NNTH]) was calculated as follows. The percentage of nonexposed controls according to age group was multiplied by the accumulated population of Sweden during the study period in the same age group, yielding an estimate of the unexposed population by age group. Among case-patients, the percentage of nonexposed persons was multiplied by the total number of cases in Sweden during the study period according to age group, which produced an estimate of unexposed cases by age group. The NNTH was then calculated as 1/(IRR 1). The population attributable risk was calculated by using the method described by Rothman and Greenland (27). Role of the Funding Sources The funding sources had no role in the collection, analysis, or interpretation of the data or in the decision to submit the paper for publication. Results Esophageal Adenocarcinoma We found a statistically significant positive association between use of LES-relaxing drugs and risk for esophageal adenocarcinoma. Ever-users of any of these drugs had an adjusted estimated IRR of 1.8 (95% CI, 1.3 to 2.7) compared with never-users (Table 1). Treatment with LES-relaxing drugs for less than 5 years was associated with a moderate, statistically nonsignificant excess risk (IRR,