Background. Incidence rates have risen rapidly for esophageal adenocarcinoma and moderately for gastric cardia adenocarcinoma, while rates have remained stable for esophageal squamous cell carcinoma and have declined steadily for noncardia gastric adenocarcinoma. We examined anthropometric risk factors in a population-based casecontrol study of esophageal and gastric cancers in Connecticut, New Jersey, and western Washington. Methods: Healthy control subjects (n = 695) and case patients with esophageal squamous cell carcinoma or noncardia gastric adenocarcinoma (n = 589) were frequency-matched to case patients with adenocarcinomas of esophagus or gastric cardia (n = 554) by 5-year age groups, sex, and race (New Jersey only) Classification of cases by tumor site of origin and histology was determined by review of pathology materials and hospital records Data were collected using in-person structured interviews. Associations with obesity, measured by body mass index (BMI), were estimated by odds ratios (ORs). All ORs were adjusted for geographic location, age, sex, race, cigarette smoking, and proxy response status. Results: The ORs for esophageal adenocarcinoma rose with increasing adult BMI. The magnitude of association with BMI was greater among the younger age groups and among nonsmokers. The ORs for gastric cardia adenocarcinoma rose moderately with increasing BMI. Adult BMI was not associated with risk of esophageal squamous cell carcinoma or noncardia gastric adenocarcinoma. Conclusions: Increasing prevalence of obesity in the United States population may have contributed to the upward trends in esophageal and gastric cardia adenocarcinomas. [J Nall Cancer Inst 1998;90:150-5].
The incidence of esophageal adenocarcinoma has been rapidly rising over the past two decades in the United States and western Europe (1–5). To a lesser extent, increases in the incidence of gastric cardia adenocarcinoma have also been reported (2,6–8). In contrast, incidence rates for squamous cell carcinoma of the esophagus have remained stable or decreased slightly, while rates for noncardia gastric adenocarcinoma have declined steadily. To identify reasons for the upward trend in esophageal and gastric cardia adenocarcinomas, we conducted a populationbased case-control study of these tumors in three areas of the United States. In the initial report from this study, cigarette smoking was found to be a risk factor (9). The present analysis evaluates the possible role of excess weight, which has been suggested as a risk factor in previous studies (10–12).
The methods for this study are described in detail elsewhere (9). Briefly, residents newly diagnosed with invasive esophageal or gastric cancers at ages 30–79 years in Connecticut (from February 1, 1993, to January 31, 1995), New Jersey (from April 1, 1993, to November 30, 1994), and western Washington (from March 1, 1993, to February 28, 1995) were identified through rapid reporting systems Population-based control subjects were selected by random digit dialing (13) for those under 65 years of age and from the Health Care Financing Administration files for those 65 years of age or older (14) Healthy control subjects and case patients with esophageal squamous cell carcinoma or noncardia gastric adenocarcinoma (comparison cases) were frequency matched to target case patients with adenocarcinomas of the esophagus or gastric cardia, including the gastroesophageal junction, in each geographic area by 5-year age group and sex and in New Jersey by race (white or non-white). Classification of cases by site of origin and histology was determined by a panel of pathologists through standardized review of pathology materials and reports from surgery, endoscopy, and radiology.
After obtaining written informed consent from each subject or next of kin of a deceased subject, an in-person, structured interview was conducted to elicit information on demographic background, tobacco and alcohol use, medication and medical histories, diet, occupation, and height and weight history up to 1 year prior to diagnosis for case patients and date of interview for control subjects. Weight history included usual adult weight (i.e., the most common weight during adulthood), highest adult weight, and usual weights during ages 20–29, 40–49 and 60–69. Interviews were obtained for 554 (81%) of 687 eligible target case patients, 589 (74%) of 795 eligible comparison case patients, and 695 (74%) of 943 eligible control subjects. Of these, information was provided by next of kin for 164 (30%) of the target case patients, 192 (33%) of the comparison case patients, and 25 (4%) of the control subjects.
Adiposity was estimated by body mass index (BMI), computed as weight in kilograms divided by height in meters squared (kg/m2). Height, weight, and BMI variables were grouped into quartiles for analysis based on sex-specific distributions among the control subjects. Anthropometric variables more finely grouped in deciles also were examined for linearity of associations. Relative risks according to anthropometric status were estimated by odds ratios (ORs) and 95% confidence intervals (CIs), using logistic regression models (15). The CIs were not adjusted for multiple comparisons. Dose response relationships were evaluated by tests of linear trend based on continuous variables. Effect modification was assessed by examination of stratum-specific results. The significance of the interaction was tested by adding a cross-product term to the model. All ORs were adjusted for geographic location, age, sex, race (white, non-white), cigarette smoking (nonsmoker, former smoker at 1 or more years prior to interview, and current smoker) and respondent status (self and next of kin). Separate analyses using more detailed cigarette smoking indicators, including pack-years of smoking, combination of packyears and smoking status, and years since smoking cessation for past smokers, did not alter the associations. Additional adjustment for other potential confounding factors, including family income, education, dietary intake (calories, fat, or fiber), level of physical activity, alcohol use, history of reflux disease, usual occupational categories, or family history of cancer, did not materially alter the risk estimates.
Initial analyses were conducted to examine the consistency of results based on all study subjects and after excluding next-of-kin interviews. Since the associations with height, usual weight, and usual BMI were similar with and without the next-of-kin interviews, results are presented for the entire study population. The findings also were unchanged when nonwhites were excluded from the analyses. In addition, the patterns of risks were similar between men and women, hence the results are presented for both sexes and all races combined. Presented in Table 1 are ORs associated with usual BMI for men and women separately as well as ORs after excluding next-of-kin interviews or non-whites.
As shown in Table 2, height tended to be inversely related to risk for all tumor types except esophageal squamous cell carcinoma, although the trend was statistically significant only for esophageal adenocarcinoma. High usual adult weight was associated with excess risks of adenocarcinomas of the esophagus and gastric cardia, reduced risk of esophageal squamous cell carcinoma, and no association with noncardia gastric adenocarcinoma. For usual BMI, the ORs for esophageal adenocarcinoma rose steadily. When compared with the first quartile, the OR increased from 1.3 (95% CI = 0.8–2.2) for the second quartile to 2.0 (95% CI = 1.3–3.3) and 2.9 (95% CI = 1.8–4.7) in the third and fourth quartiles, respectively (P for trend <.0001) Furthermore, compared with subjects in the lowest 10% of usual BMI (<21.70 for men and <20.18 for women), risk increased steadily to reach fivefold (OR = 5.4, 95% CI = 2.4–12.0) among those in the highest decile (⩾29.54 for men and ⩾31.25 for women). To a lesser extent, ORs for gastric cardia adenocarcinoma rose with usual BMI to 1.6 (95% CI = 1.1–2.6) in the highest quartile. Among those in the highest decile of usual BMI, the risk of gastric cardia adenocarcinoma increased twofold (OR = 2.1, 95% CI = 1.1–;4.1) relative to those in the lowest decile. In contrast, usual BMI was not significantly related to risk of esophageal squamous cell carcinoma or noncardia gastric adenocarcinoma.
Risk of esophageal adenocarcinoma was not related to weight gain during adulthood, except when the gain was large (⩾46 lbs or ⩾21% since age 20 years) (Table 2). Stratification by usual BMI showed excess risks associated with weight gain greater than or equal to 461bs (OR = 1.7 95% CI = 0.6–4.9) and weight gain of greater than or equal to 21% (OR = 1.5; 95% CI = 0.7–3.2) only among those in the highest quartile of usual BMI. Weight changes were not consistently associated with cardia or noncardia gastric adenocarcinoma, but risk of esophageal squamous cell carcinoma tended to increase with weight loss and decrease with weight gain since ages 20–29 years. For each tumor type, the risk pattern associated with BMI at ages 20–29, 40–49, and 60–69 years, or BMI based on the maximum adult weight, was similar to that of usual BMI (data not shown). The results were not altered substantially after excluding next-of-kin respondents, with a twofold excess risk (OR = 2.2; 95% CI = 1.2–4.0) for those who gained greater than or equal to 461bs and a 40% excess (OR = 1.4; 95% CI = 0.9–2.2) for those who gained greater than or equal to 21% of the weight during ages 20–29 years.
The joint effects of height and usual adult weight were further assessed for esophageal and gastric cardia adenocarcinomas (Table 3). Risk of esophageal adenocarcinoma generally rose with increasing weight at each height level, and declined with increasing height at each weight level. Similar but less consistent patterns were found for gastric cardia adenocarcinoma. The patterns of risk by height and usual adult weight were unremarkable for esophageal squamous cell carcinoma and non-cardia gastric adenocarcinoma (data not shown).
The positive association between risk of esophageal adenocarcinoma and usual BMI was significantly (P = .03) modified by age (at the time of diagnosis for case patients and at the time of interview for control subjects), with the greatest increase in risk seen among the youngest group (ages <50 years) and the smallest increase among the oldest group (ages 70–79 years) (Table 4). The ORs for the highest quartile relative to the lowest quartile of usual BMI in age groups less than 50, 50–59, 60–69 and 70–79 years were 33.6 (95% CI = 2.1–552), 4.5 (95% CI = 1.4–14.1), 2.3 (95% CI = 1.0–5.4), and 1.7 (95% CI = 0.8–3.8), respectively. Effect modification by age was not apparent for gastric cardia adenocarcinoma (data not shown), although no relation to usual BMI was found among those aged 70 years and older, a pattern consistent with that for esophageal adenocarcinoma. Associations between usual BMI and risk of esophageal squamous cell carcinoma or noncardia gastric adenocarcinoma were not modified by age (data not shown).
Cigarette smoking, a risk factor for each of the cancers in our study, was a significant (P = .03) effect modifier of the risk of esophageal adenocarcinoma associated with usual BMI (Table 4). The largest BMI-related increase in risk was found among nonsmokers, followed by current smokers and then former smokers. The ORs for the highest versus the lowest quartile of usual BMI were 8.7 (95% CI = 2.4–31.1) among nonsmokers, 2.1 (95% CI = 1.1–4.2) among former smokers, and 2.9 (95% CI = 1.1–7.6) among current smokers. The effect of usual BMI on risk of gastric cardia adenocarcinoma was not significantly modified by smoking, although the risks were highest among nonsmokers. No effect modification by smoking was observed for the relation of BMI to esophageal squamous cell carcinoma or noncardia gastric adenocarcinoma (data not shown). In addition, no significant effect modification by history of gastroesophageal reflux disease or by educational level was found for any of the four cancer types. The findings for usual BMI were similar for diffuse and intestinal types of adenocarcinoma within subsites of stomach.
This population-based case-control study revealed that excess weight is a strong risk factor for esophageal adenocarcinoma, with risk rising consistently with increasing BMI. The risk appeared related largely to elevated BMI per se and not to weight gain or loss during adult life. Furthermore, within each weight level, the risk tended to decrease with increasing height. To a lesser extent, excess weight increased the risk of gastric cardia adenocarcinoma, while no effect was seen for noncardia gastric adenocarcinoma or esophageal squamous cell carcinoma. The relatively large study size and standardized classification of case patients by study pathologists enabled us to assess the relation of anthropometric variables to the four types of esophageal and stomach cancers. These results argue against recall bias or differential reporting between case patients and control subjects, since recall or reporting of height and weight is unlikely to vary by case type.
Our findings provide strong support for a causal relation between adiposity and adenocarcinomas of the esophagus and gastric cardia. Limited evidence from previous population-based, case-control studies in the United States suggested a threefold increased risk of esophageal adenocarcinoma among white men with BMI greater than 26.6 (10) and a 150% excess risk of esophageal adenocarcinoma as well as a 60% excess risk of gastric cardia adenocarcinoma among subjects in the highest decile of BMI (11). In China, where the study population was relatively lean, the risk of gastric cardia adenocarcinoma was elevated 40% among women and threefold among men in the highest quartile of BMI (12). In two studies, a dose-response relation was noted between BMI and increased risk of gastric cardia adenocarcinoma (11,12). The inverse associations between height and risks of esophageal and gastric cancers also are consistent with our observation of elevated risks among the obese, since taller subjects tend to be leaner when adjusted for weight.
The mechanism by which overweight increases the risk of adenocarcinomas of the esophagus and gastric cardia is not clear. It has been suggested that obesity promotes gastroesophageal reflux disease by increasing intra-abdominal pressure (16,17). In turn, gastroesophageal reflux predisposes to Barrett's esophagus, a metaplastic precursor state for adenocarcinomas of the esophagus and gastric cardia (18,19). In our study, the magnitude of relative risk associated with BMI was similar among those with or without a self-reported history of gastroesophageal reflux, suggesting that obesity may influence cancer risk through mechanisms in addition to reflux (20,21). However, given the relatively low sensitivity of reported symptoms for the diagnosis of gastroesophageal reflux disease (22), this condition may have been substantially underreported in our study. Further investigations are needed to identify factors that may influence the cancer risks associated with obesity and gastroesophageal reflux disease, including body fat distribution, dietary practices, medications, and other conditions that may affect the frequency and severity of reflux disease and the composition of enterogastric refluxate.
If the association with usual BMI is causal and our relative risk estimates re-fleet the true magnitude of associations, attributable risk calculations indicate that individuals above the median level of BMI may account for 33% of esophageal adenocarcinoma and 22% of gastric cardia adenocarcinoma case patients occurring in the three geographic areas over the study period. Therefore, the upward trend in the incidence of these tumors may be related in part to substantial increases in the prevalence of overweight in the U.S. population. The prevalence of overweight adults [defined in (23) as BMI ⩾27.8 for men and ⩾27.3 for women, which are approximately equal to the highest quartiles of usual BMI in our study] at all ages rose from 25% in 1976 through 1980 to 33% in 1988 through 1991, with the greatest increase occurring among men age 50 years or older (23).
The only environmental factor consistently linked to adenocarcinomas of the esophagus and gastric cardia in epidemiologic studies has been cigarette smoking (9,11,24–27). In our study, increases in risk with usual BMI were greatest among nonsmokers, indicating that smoking is not a necessary cofactor for the association with overweight. Our finding that the BMI-associated risk is highest in the youngest age group suggests that obesity is particularly important for early-onset tumors, while other risk factors may assume a more prominent role for tumors developing in later years.
It is unclear why risks associated with overweight were greater for esophageal adenocarcinoma than for gastric cardia adenocarcinoma. It is possible, although unlikely, that some noncardia gastric cancers may be misclassified as cardia tumors, despite our standardized review and classification procedures, thus attenuating the association of gastric cardia adenocarcinoma with overweight. It is also possible that reflux mechanisms are more closely related to Barrett's esophagus and subsequent esophageal adenocarcinoma than to the development of gastric cardia cancer. In contrast, we found no association between BMI and noncardia gastric cancer and a slight but nonsignificant inverse relation with squamous cell carcinoma of the esophagus. The latter finding is consistent with the inverse correlation between body weight and the major risk factors (smoking, drinking, and poor nutrition) for squamous cell esophageal cancer (28), although additional adjustment for alcohol drinking or caloric intake did not affect our results.
In summary, our multicenter population-based, case-control study found that increased BMI was a strong risk factor for esophageal adenocarcinoma and a moderate risk factor for gastric cardia adenocarcinoma. The elevated risks appeared related mainly to excess weight per se and not to weight changes over time. In contrast, BMI was largely unrelated to esophageal squamous cell carcinoma and noncardia gastric adenocarcinoma. These findings suggest that the increasing prevalence of obesity in the population has contributed to the rising incidence trends for adenocarcinomas of the esophagus and gastric cardia. Further epidemiologic, clinical, and laboratory studies are needed to identify the mechanisms by which obesity increases the risk of these tumors.