In this large population-based case–control study, we found site-specific associations between family history of cancer or gastroesophageal disorders and risk of EA, GCA, and DGA. Family history of cancer in the prostate was associated with an increased risk of EA; family history of esophageal cancer was associated with an increased risk of GCA; and family history of gastrointestinal cancer and particularly gastric cancer was associated with an increased risk of DGA. In addition, family history of hiatal hernia was associated with an increased risk of both EA and GCA, an effect that was more pronounced among individuals with a personal history of hiatal hernia. To our knowledge, this is the first report of an association between family history of hiatal hernia and risk of EA and GCA.
The rapid increase in the incidence of EA over the past few decades in Western countries may indicate a strong contribution of environmental factors to the etiology of EA , but genetic factors may also play a role. To date, there is a paucity of genetic association studies of EA . Given that Barrett’s esophagus is an established risk factor for EA [22, 23] and Barrett’s esophagus patients have a more than 30 times greater risk of developing EA , most previous genetic studies compared patients with Barrett’s esophagus and/or EA with controls and found evidence of familial aggregation of these conditions . It has been suggested that polymorphisms in genes involved in the detoxification of xenobiotics and luminal toxic agents (e.g. GSTM1, GSTT1, GSTP1) as well as those involved in the regulation of cell cycle progression (e.g. CCND1) may play a role in individual susceptibility to EA . Recently, a genome-wide association study of Barrett’s esophagus  has identified two susceptibility loci on chromosomes 16q24 and 6p21. The closest protein-coding gene to the 16q24 locus is FOXF1, a gene implicated in esophageal development and structure. This finding suggests that structural factors related to the development of the esophagus may play a role in the etiology of Barrett’s esophagus. It is consistent with the fact that most of individuals affected by Barrett’s esophagus have a history of hiatal hernia in their lower esophagus. We found that EA risk was higher among individuals with a family history of hiatal hernia (OR = 2.05) than those without this history and that risk is highest among those with both a personal and a family history of hiatal hernia (OR = 10.75). The observation that EA risk increases with an increasing number of family members (participant’s family plus participant him/herself) affected by hiatal hernia suggests that genetic factors that are involved in the development of hiatal hernia and Barrett’s esophagus play a role in the development of EA. In addition, even though in our study personal and family history of hiatal hernia was obtained though self-report, our results are consistent with the strong evidence of familial aggregation of Barrett’s esophagus and EA from previous studies [4, 25, 26], in which patients were ascertained through clinical diagnosis.
We also found an increased risk of EA among individuals with a family history of prostate cancer but lower risk among those with a family history of breast cancer and these associations did not substantially change after controlling for gender-specific sibship size. Such results were unexpected and require confirmation. Reports of associations between family history of breast or prostate cancer and risk of gastroesophageal cancer are sparse and have been inconsistent [11, 12, 14]. A multicenter, population-based case–control study conducted in the U.S. during 1993-1995  found that family history of breast cancer was associated with increased risks of EA and non-cardia gastric adenocarcinoma, while the association of family history of prostate cancer with lower risk of these two types of cancers was not statistically significant. In a case–control study of patients seen at the Memorial Sloan-Kettering Cancer Center from 1992 to 1994, family history of breast or prostate cancer was not associated with an increased risk of EA and GCA . In a study based on the Swedish Family-Cancer Database , maternal breast cancer was associated with an increased risk of gastric cancer in the offspring (standardized incidence ratio = 1.84, 95% CI = 1.02-3.04). Despite these sparse observations, our results are consistent with the strong male dominance in EA . Among white Americans, males have 7.7 times higher risk of getting EA than females . In addition, this gender difference cannot be adequately explained by differences in known risk factors for EA including gastroesphageal reflux diseases, obesity, and tobacco consumption , suggesting that other unknown factors such as sex hormonal factors and related signaling axes may play important roles in the development of EA [27, 29]. Experimental studies have shown that estrogen may have an inhibitory effect  and androgens may have growth-enhancing effects  on the carcinogenic process of the esophagus. Compared to individuals without a family history of prostate cancer, those with a positive family history may have a higher exposure to androgens  or enhanced susceptibility to the effect of androgens, leading to a higher risk of esophageal cancer. Similarly, individuals with a family history of breast cancer may have higher exposure to estrogen  or enhanced susceptibility to the effect of estrogen, therefore resulting in a reduced risk of esophageal cancer.
EA and GCA display some similar descriptive epidemiological features  and risk factors. Low intake of fiber, hiatal hernia, and Barrett’s esophagus [17, 22, 23] all have been associated with increased risks of both EA and GCA. Therefore, it is not surprising that family history of esophageal cancer was associated with an increased risk of GCA, which might be explained by common family dietary habits such as low fiber intake, or by a shared genetic susceptibility to hiatal hernia and Barrett’s esophagus. In the current study, additional adjustment for fiber intake did not explain the association between family history of esophageal cancer and risk of GCA, suggesting that shared genetic susceptibility may play a more important role.
The observed association between family history of a gastrointestinal cancer or of gastric cancer and an increased risk of DGA could be explained by shared susceptibility to both genetic and environmental factors. Family studies [33, 34] have shown that shared and non-shared environmental factors largely accounted for the variation in gastric cancer whereas genetic factors accounted for only a small proportion of gastric cancer susceptibility. This observation is consistent with the sharp decline in gastric cancer incidence over the past 40 years, owing to improvements in diet and food storage methods as well as a decline in the prevalence of H. pylori infection . Common exposure to H. pylori may explain the positive associations of GCA risk with family history of gastric cancer and family history of other gastrointestinal cancers, as H. pylori infection has been shown to increase the risk of not only gastric cancer, but also pancreatic cancer . Data on history of H. pylori infection was not available for majority of our study participants; therefore we were unable to investigate whether history of H. pylori infection was a potential effect modifier of the association between family history of gastric cancer and risk of GCA. However, the associations between family history of these gastrointestinal cancers and increased risk of GCA may also be mediated by shared genetic susceptibilities . For example, candidate gene studies have consistently found that polymorphisms in the IL-1β and MTHFR genes were associated with individual susceptibility to both intestinal-type gastric cancer and pancreatic cancer [4, 38, 39].
The strengths of this study include the population-based design, relatively large sample size, separation of gastroesophageal adenocarcinoma by anatomical sites, and data on family history of cancer and non-malignant gastroesophageal disorders. Results from our study are compatible with the limited evidence to date and known etiological mechanisms of EA, GCA, and DGA. Our study also has a number of limitations. First, even though we excluded NOK interview data from our analyses, our data on family history were reported by study participants rather than by family member themselves; therefore they may be subject to recall bias, under-ascertainment of family history, and misclassification of the primary tumor site. The lack of excessive reporting of family history of esophageal cancer by cases with EA indicates that our results were most likely not affected by selective recall bias. In addition, misclassification of the primary tumor site of a relative was lessened by only collecting data on first-degree relatives . Nevertheless, it is not possible to collect information on histology (e.g. adenocarcinoma vs. squamous cell carcinoma) or specific tumor subsite (gastric cardia vs. non-cardia) among those who reported esophageal/gastric cancers in first-degree relatives. Second, our sample size was insufficient to evaluate very rare cancers in family members, to examine risk patterns separately for parents vs. siblings or to assess whether the number of affected relatives refined risk estimates. In particular, there was only one EA patient and seven controls with a positive family history of EA, our study was not sufficient powered to examine the association between family history of EA and risk of EA. However, sparse-data bias was unlikely to have occurred, because: a) reducing over-stratification in our regression analyses, i.e. removing additional adjustment for known risk factors (Additional file 1: Tables S1 and S2 versus Tables 2 and 3), did not change our results substantially; and b) our results were very similar with and without Firth’s correction  for sparse data (data not shown). A pooled analysis of data from all relevant population-based studies is needed for these additional analyses. Third, we recognize that if the effect of family history of cancer or gastroesophageal disorders may be exerted through common family dietary habits or by a shared genetic susceptibility, adjusting for known risk factors for gastroesophageal cancer may underestimate the actual strength of the association between family history of cancer/gastroesophageal disorders and risk of gastroesophageal cancer. However, overadjustment is unlikely because, as mentioned above, removing additional adjustment for known risk factors did not change our results substantially. Fourth, BMI was categorized using sex-specific quartiles rather than the World Health Organization classification to avoid sparse data for some categories. In addition, we did not control for history of Helicobacter pylori (H. pylori) infection in multivariate analyses, because serum IgG antibodies to H. pylori whole-cell antigens (Helico-G) and CagA were not measured for majority of study participants . Furthermore, we were unable to distinguish shared genetics from shared early-life environmental exposures among first-degree relatives, as environmental exposure histories of participants’ relatives were not available.