Research aiming at identifying the reasons for the alarming increase in the incidence of adenocarcinoma of the esophagus and gastric cardia in industrialized countries during recent decades1, 2, 3, 4 is urgent and intense. Gastroesophageal reflux,5, 6, 7 overweight6, 8, 9 and male sex3 have been established as the main known risk factors, but it remains uncertain why the incidence of these tumors is rising.10, 11 The rapidity of the increase suggests that environmental exposures are likely to be key factors.10, 11 Recently, an interesting hypothesis was proposed by La Vecchia et al., suggesting that the increasing secular trends of adenocarcinoma of the esophagus and gastric cardia, particularly among overweight men, is explained by the use of tight belts.12 Such belts have a constraining influence on the abdomen, which in turn might promote reflux, and belts did not exist in the past when suspenders were in use.12 This proposed mechanical exposure seems to fit well with all the above-presented prerequisites: increased and widespread use in industrialized populations, particularly among males, an environmental exposure and a possible combination with overweight to promote reflux, i.e., the established risk factors. As we had already included this hypothesis as part of the planning of a Swedish case–control study in the late 1990s, we investigated the relation between use of tight belts and risk of adenocarcinoma of the esophagus and gastric cardia.
It has recently been hypothesized that the general shift in use from suspenders to belts among men might be a factor that could promote reflux, particularly among overweight men, and thereby contribute to the alarmingly increasing incidence of esophageal and cardia adenocarcinoma. We addressed this hypothesis in a nationwide Swedish population-based case–control study, conducted in 1995–1997. Included were 189 patients with esophageal adenocarcinoma and 262 patients with cardia adenocarcinoma, who were compared with 167 patients with esophageal squamous cell carcinoma and 820 population-based control participants. Data were collected at structured face-to-face interviews with all study participants. Tumor classification was uniform and thorough. Odds ratios (OR) with 95% confidence intervals (CI), adjusted for potential confounders, were estimated in multivariable logistic regression models. Daily use of tight belts 20 years earlier did not entail an increased risk of adenocarcinoma of the esophagus (OR 1.1, 95% CI 0.7–1.8) or cardia (OR 0.8, 95% CI 0.5–1.3) compared to never use. Adjustment for reflux symptoms, body mass and other potentially relevant covariates did not influence the results. Similarly, in analyses restricted to overweight men, no associations were identified. No association with esophageal squamous cell carcinoma was found. In conclusion, this study provided no support for the hypothesis that use of tight belts is associated with an increased risk of developing esophageal or cardia adenocarcinoma. © 2006 Wiley-Liss, Inc.
Material and methods
Several hypotheses have previously been tested within this nationwide and prospective Swedish population-based case–control study, focusing on risk factors for esophageal and cardia adenocarcinoma, and its design has been described in detail elsewhere.5, 13 In summary, the study encompassed all native Swedish residents of ages below 80 years, living in Sweden during the period 1995 through 1997. All newly diagnosed cases of esophageal or cardia adenocarcinoma were eligible for the study. To evaluate potential recall bias, we also included a reference case group, consisting of patients with esophageal squamous cell carcinoma, born on even-numbered dates. To reduce tumor misclassification, prospective, uniform and thorough documentation of the tumors was introduced at all 195 collaborating hospital departments throughout Sweden (100% participation frequency), and finally, one pathologist reviewed the histological slides. Control persons were randomly selected from the complete Total Population Register in Sweden, and frequency matched to the age and sex distributions of the esophageal adenocarcinoma patients. All cases and control participants underwent computer-aided face-to-face interviews by professional interviewers from Statistics Sweden. The case patients were interviewed shortly after the histological diagnosis was confirmed. Our computerized structured interview included 1 question addressing the use of tight belts, referring to 20 years before the interview, i.e., allowing a plausible latency interval between the “exposure” and the studied cancers. The question was: “How often did you use a belt or corset or other garment that was tightened around your waist with force? Include only use lasting more than 3 hours,” and the participants gave an open answer. The exposure to such garments was then converted to use per week and categorized as: (1) never use, (2) low exposure level (>1 to 6 times per week) and (3) high exposure level (7 times per week to daily use the whole day). The interviews also addressed factors that we considered as potential confounders in the present study, i.e., a priori known risk factors for the 3 different cancer types. These were tobacco smoking (grouped into never, previous and current smokers, referring to the status 2 years before the interview), alcohol use (in 4 groups of total alcohol consumed, referring to 20 years before the interview), socioeconomic status (in 6 groups, based on occupational history) and dietary intake of fruit and vegetables (in 3 categories based on frequency of average intake, referring to 20 years before the interview). We also considered reflux symptoms (heartburn or regurgitation at least weekly, categorized as yes or no, referring to 5 years or more before the interview) and body mass index (BMI) (grouped into quartiles among the control participants, referring to 20 years before the interview), but these factors were addressed separately, since it was hypothesized that they act in the causal pathway. Relative risks were estimated by odds ratios (OR), with 95% confidence intervals (CI), using conditional logistic regression (SAS PHREG procedure). We analyzed the effect of use of tight belts in both crude and various adjusted models. The crude model meant estimates from a conditional logistic model conditioning on the matching variables age and sex, but not adjusted for any other variables. In the main multivariable models, we adjusted for the other potential confounders listed earlier, excluding reflux symptoms and BMI. In 2 models, we added adjustments for BMI and reflux, introduced separately and in combination. Finally, 1 adjusted model was restricted to overweight participants only and 1 to overweight men. Each exposure was evaluated using the Wald test, which considers all categories of the variable and not just pairwise comparisons with the reference category. Informed consent was obtained from each study participant individually, and all regional ethics committees in Sweden approved the study.
There were 189 study patients with esophageal adenocarcinoma, constituting 88% of all eligible patients with that cancer. The corresponding figures for patients with cardia adenocarcinoma were 262 and 84% and for esophageal squamous cell carcinoma 167 and 73%. These groups were compared with a group of 820 control persons, constituting 73% of all controls who had been primarily selected. The age and sex distributions among these groups are presented in Table I. Reflux symptoms (occurring at least weekly) and overweight were most common among patients with esophageal adenocarcinoma, followed by cardia cancer patients and least common among the patients with esophageal squamous cell carcinoma and the control participants (Table I).
|Controls||Esophageal adenocarcinoma||Cardia adenocarcinoma||Esophageal squamous cell carcinoma|
|<60 years||209 (25)2||38 (20)||75 (29)||38 (23)|
|60–79 years||611 (75)||151 (80)||187 (71)||129 (77)|
|Male||679 (83)||165 (87)||223 (85)||120 (72)|
|Female||141 (17)||24 (13)||39 (15)||47 (28)|
|No||685 (84)||76 (40)||187 (71)||142 (85)|
|Yes||135 (16)||113 (60)||75 (29)||25 (15)|
|Body mass index4|
|Quartile 1 (16.5–22.0)||205 (25)||12 (6)||50 (19)||51 (30)|
|Quartile 2 (22.1–23.7)||207 (25)||26 (14)||46 (18)||34 (20)|
|Quartile 3 (23.8–25.4)||203 (25)||53 (28)||65 (25)||35 (21)|
|Quartile 4 (25.5–40.2)||201 (25)||98 (52)||101 (39)||47 (28)|
|Total number of participants||820||189||262||167|
Daily use of tight belts the whole day (high exposure) was fairly common (about 10%) and similar among all case groups and the controls (Table II). Among persons who used tight belts daily (high exposure), the adjusted OR of esophageal adenocarcinoma was 1.1 (95% CI 0.7–1.8), and no dose–response relation was seen. In the models restricted to overweight persons, no increased risk was found, independent of sex (Table III). Inclusion or exclusion of reflux or BMI in the models did not materially change the risk estimates.
|Controls1||Esophageal adenocarcinoma||Cardia adenocarcinoma||Esophageal squamous cell carcinoma|
|Use of tight belts 20 years ago3|
|No exposure (never)||687 (86)4||152 (84)||1.0 (reference)||0.55||222 (87)||1.0 (reference)||0.94||140 (88)||1.0 (reference)||0.38|
|Low exposure (>1–6 times/week)||27 (3)||5 (3)||0.9 (0.3–2.3)5||8 (3)||0.9 (0.4–2.0)||8 (5)||1.3 (0.6–3.0)|
|High exposure (7 times/week)||83 (10)||24 (13)||1.3 (0.8–2.1)||25 (10)||0.9 (0.6–1.5)||11 (7)||0.7 (0.3–1.3)|
|Use of tight belts 20 years ago6|
|No exposure (never)||687 (86)||152 (84)||1.0 (reference)||0.75||222 (87)||1.0 (reference)||0.63||140 (88)||1.0 (reference)||0.39|
|Low exposure (>1–6 times/week)||27 (3)||5 (3)||0.7 (0.3–2.0)||8 (3)||0.8 (0.3–1.8)||8 (5)||1.1 (0.4–2.7)|
|High exposure (7 times/week)||83 (10)||24 (13)||1.1 (0.7–1.8)||25 (10)||0.8 (0.5–1.3)||11 (7)||0.6 (0.3–1.2)|
|Use of tight belts 20 years ago7|
|No exposure (never)||687 (86)||152 (84)||1.0 (reference)||0.13||222 (87)||1.0 (reference)||0.14||140 (88)||1.0 (reference)||0.39|
|Low exposure (>1–6 times/week)||27 (3)||5 (3)||0.3 (0.1–1.0)||8 (3)||0.5 (0.2–1.3)||8 (5)||1.0 (0.4–2.5)|
|High exposure (7 times/week)||83 (10)||24 (13)||0.8 (0.5–1.5)||25 (10)||0.7 (0.4–1.1)||11 (7)||0.6 (0.3–1.2)|
|Controls1||Esophageal adenocarcinoma||Cardia adenocarcinoma||Esophageal squamous cell carcinoma|
|Use of tight belts among overweight persons 20 years ago3|
|No exposure(never)||317||122||1.0 (reference)||0.23||141||1.0 (reference)||0.11||68||1.0 (reference)||0.22|
|Low exposure(>1–6 times/week)||22||4||0.4 (0.1–1.3)4||7||0.6 (0.2–1.5)||3||0.4 (0.1–1.7)|
|High exposure(7 times/week)||55||20||0.8 (0.4–1.4)||15||0.6 (0.3–1.0)||6||0.5 (0.2–1.4)|
|Use of tight belts among overweight men 20 years ago5|
|No exposure (never)||263||106||1.0 (reference)||0.44||118||1.0 (reference)||0.29||49||1.0 (reference)||0.43|
|Low exposure (>1–6 times/week)||16||4||0.6 (0.2–1.7)||6||0.8 (0.3–1.9)||2||0.8 (0.2–4.0)|
|High exposure (7 times/week)||50||17||0.8 (0.4–1.4)||15||0.6 (0.3–1.2)||5||0.5 (0.2–1.4)|
Daily users of tight belts were not at increased risk of developing cardia adenocarcinoma (adjusted OR 0.8, 95% CI 0.5–1.3), and no dose–response relation was found (Table II). No associations were seen in the models restricted to overweight persons or to overweight men (Table III).
To our knowledge, this is the first study addressing the hypothesis that use of tight garments might contribute to the increasing incidence of esophageal or cardia adenocarcinoma. Despite the several well-fitting criteria, including a predefined and plausible mechanical mechanism,12 our results indicate lack of an association.
The truly population-based design with strict random sampling of control participants and the prospective and rapid case ascertainment with high participation rates are strengths of the present study. Other advantages include the detailed information available on all plausible confounding factors. Moreover, the thorough and uniform tumor classification is an advantage in any etiologic study of esophageal and cardia adenocarcinomas, since the risk factor profiles seem to be partly distinct and the risk of misclassification is considerable.14 Exposure misclassification might influence our results, particularly since the assessment of the use of tight garments referred to use 20 years before the interview, and any such misclassification would probably be nondifferential and dilute the risk estimates of a true association.15 However, we chose to assess the exposure 20 years before the diagnosis of esophageal or cardia cancer, as we considered this a plausible biological latency interval between the exposure and the fully developed invasive cancer. Moreover, the face-to-face interviews conducted with all study participants facilitated valid exposure assessment. Recall bias is often a threat to the internal validity of case–control studies. The lack of an association should not be explained by recall bias, however, since it is unlikely that case patients, on account of their cancer, would underreport their use of tight belts as compared to the reporting by control participants or patients with esophageal squamous cell carcinoma. Furthermore, the hypothesis that use of belts might potentially affect the risk of these tumors was not known by the study participants, which should be a further argument against recall bias. Finally, chance error is an unlikely explanation for our negative results, since the exposure prevalence was rather high and the point estimates were all close to the null hypothesis. However, the limited sample size and the risk of information bias does not allow us to exclude a moderate association.
In conclusion, this nationwide, population-based case–control study provides evidence against a link between the use of tight belts and the risk of esophageal or cardia adenocarcinoma. Further studies are urgently needed to find an explanation for the alarmingly increasing incidence of esophageal adenocarcinoma.
We express our gratitude to Dr. Anders Lindgren for his review of all histopathological material, to Leila Nyrén for coordination of the project and to all doctors who acted as contact persons at the participating departments.