While still the third leading cause of cancer death and the fifth most common cancer worldwide, gastric cancer incidence and mortality have markedly declined over the past few decades [1]. The downward trend has paralleled the decreasing prevalence of Helicobacter pylori infection [2], a major risk factor for this malignancy. However, cancer registration data indicate unexpected increases among recent generations in many parts of the world [3-5], implying etiologic factors other than H. pylori may be responsible [6].

Inflammation is a recognized antecedent of many types of cancer. Autoimmune conditions are important causes of inflammation and appear to have increased in prevalence over recent decades [7], particularly in industrialized populations. An association with gastric cancer has been firmly established for the immune-mediated destruction of the gastric parietal cells, autoimmune gastritis, and its most common clinical manifestation, pernicious anemia [8]. Other autoimmune conditions have been variably linked with gastric cancer risk, perhaps as a consequence of their frequent co-occurrence with autoimmune gastritis or, alternatively, through some systemic effect(s) that remains to be elucidated. To better understand the role of autoimmunity in gastric carcinogenesis, we undertook a systematic review and meta-analysis of observational studies addressing associations of autoimmune diseases and gastric cancer risk.

A total of 1,134 articles were identified after removing the duplicates through the database search (S6 Fig.). Of these, 964 articles were excluded as irrelevant on the basis of title or abstract. The remaining 170 articles were reviewed for fulltext as well as their references, and finally 52 studies with relevant and extractable information were included for metaanalysis. The full list of included articles and the corresponding data items are presented in S3 Table. There were 12 prospective cohort studies, 36 retrospective cohort studies and four case-control studies. Most of the studies were conducted in Western populations (30 European studies, six in the United States and one multi-country effort). There were 15 studies from Asia. Forty-one of the 52 papers were judged high quality. A total of 45 different autoimmune conditions were reported in these studies, amongst which 30 conditions had two or more studies for pooling. However, six of these 30 included studies with zero cases and insufficient information (such as expected cancers) for deriving risk estimates.

Table 1 lists associations of 24 autoimmune conditions with gastric cancer risk. The strongest associations were observed for dermatomyositis (RR, 3.69; 95% CI, 1.74 to 7.79), pernicious anemia (RR, 2.84; 95% CI, 2.30 to 3.50), and Addison disease (RR, 2.11; 95% CI, 1.26 to 3.53). Other autoimmune conditions that were statistically significantly associated with gastric cancer risk were dermatitis herpetiformis, IgG4-related disease, primary biliary cirrhosis, diabetes mellitus type 1 (T1DM), and systematic lupus erythematosus (SLE). Among disease groups, statistically significant increases were observed for polymyositis/dermatomyositis combined with RR 2.09 (95% CI, 1.04 to 4.20), polymyalgia rheumatica/giant cell arteritis with RR 1.33 (95% CI, 1.16 to 1.52) and autoimmune thyroid disease with RR 1.31 (95% CI, 1.12 to 1.54). Multiple sclerosis was the only condition associated with statistically significant decreased risk, RR 0.64 (95% CI, 0.45 to 0.92). Excluding pernicious anemia, pooled risk estimates for alimentary tract-involving and non-alimentary autoimmune conditions were 1.18 (95% CI, 1.04 to 1.33) and 1.21 (95% CI, 1.10 to 1.34), respectively. Autoimmune diseases that involve endocrine glands, which include Addison disease, autoimmune thyroiditis and T1DM had pooled RR 1.42 (95% CI, 1.25 to 1.61), whereas autoimmune nonendocrine diseases other than pernicious anemia had pooled RR 1.15 (95% CI, 1.06 to 1.26; p_{heterogeneity}=0.008). Considering all reports combined, having any autoimmune condition was associated with statistically significant pooled RR 1.37 (95% CI, 1.24 to 1.52).

There were no reports of gastric cancer with pernicious anemia in the studies from Asia. Considering all other autoimmune diseases combined, the summary RR for gastric cancer was 1.08 (95% CI, 0.90 to 1.29) in studies from Asia compared to 1.24 (95% CI, 1.14 to 1.34) in studies from non-Asian countries (p_{heterogeneity}=0.005).

Funnel plots were symmetric for all conditions. Imputation of hypothetical missing results by the trim and fill method did not alter the estimated associations by more than 20% with the exception of immune thrombocytopenic purpura for which change in RR was 56%.

Pernicious anemia, T1DM and SLE each had more than 5 studies for analysis and are displayed separately as forest plots in Figs. 1-3.

There was marked heterogeneity among the 14 studies of the association of pernicious anemia with gastric cancer overall, indicated by I² =82.4% and p-value of < 0.001 (Fig. 1) [11,12,17-28]. Sensitivity analyses restricted to cohort studies (n=13; RR, 2.94; 95% CI, 2.37 to 3.65) and high quality studies (n=12; RR, 2.70; 95% CI, 2.15 to 3.40) were similar to our overall pooled RR.

A low heterogeneity among studies was observed in the meta-analysis of T1DM (n=9) association with I² =35.7% and p=0.13 (Fig. 2) [11,29-34]. For the four studies with sex information, the association was significantly greater in females (RR, 1.77; 95% CI, 1.51 to 2.07) compared to males (RR, 1.19; 95% CI, 1.03 to 1.39) (p_{heterogeneity} < 0.001).

Gastric cancer RR with SLE showed low heterogeneity among studies (n=6) with I² =17.6% and p=0.303 (Fig. 3) [11,12,35-38].

The 15 conditions with only one study report and the additional six with insufficient information for pooling are summarized in Table 2. Apart from those diseases mentioned in Table 1, statistically significant associations were reported for myasthenia gravis (RR, 1.38; 95% CI, 1.14 to 1.65) [11] and alopecia areata (RR, 0.46; 95% CI, 0.33 to 0.65) [39].

This is the first comprehensive review of autoimmune conditions in gastric cancer risk. Our meta-analysis identified significant associations for autoimmunity overall and for several specific conditions. We extended the known association of pernicious anemia with greater statistical power. By pooling all available evidence to date, we found novel associations for some autoimmune diseases which showed inconsistent or null results in individual studies.

There was a modest increase of gastric cancer risk across all autoimmune conditions combined. Autoimmunity could be linked with gastric carcinogenesis through different potential pathways. Many of the conditions associated with gastric cancer in this study, such as autoimmune thyroiditis, T1DM, vitiligo, and Addison disease, frequently cooccur with pernicious anemia, which has a direct mechanistic interpretation through its pathological correlate, autoimmune gastritis. Indeed, the overall association of autoimmune conditions with gastric cancer was significantly weaker in studies from Asia than from other parts of the world, consistent with the relatively lower incidence of pernicious anemia in Asian populations [40]. Alternatively, autoimmunity might exacerbate H. pylori–driven gastritis in the absence of autoimmune gastritis per se. For either of these scenarios, the etiologic contributions of other components of the gastric microbiome remains to be determined.

In the study of 4.5 million U.S. male veterans [12], autoimmune diseases with alimentary tract involvement were associated with an increased risk of alimentary tract cancers, including gastric cancer, whereas conditions without alimentary tract involvement generally were not significantly associated. Using the same definition of alimentary tract involvement and excluding pernicious anemia, we found no difference in gastric cancer risk between alimentary tractinvolving versus all other conditions. On the other hand, when we grouped the non-pernicious anemia autoimmune conditions based on endocrine gland involvement, gastric cancer risk was significantly higher for autoimmune endocrine diseases than for the remainder. Given that pernicious anemia results from destruction of secretory parietal cells, these observations reinforce the intriguing concept that the autoimmune component of gastric carcinogenesis is more closely linked to glandular targets than to the alimentary tract per se.

Epidemiologic evidence for an autoimmune contribution to gastric carcinogenesis is not limited to studies of patients with personal history of autoimmune conditions. In a study linking the Swedish Multigeneration Register and the Swedish Cancer Register, family history of autoimmune disease in siblings was associated with noncardia gastric cancer with SIR 1.36 (95% CI, 1.22 to 1.52) [41]. On the other hand, autoimmune disease in spouses was not associated. Taken together, these observations favor an underlying basis that is related to either genes or childhood environment (shared by siblings), rather than adult exposures (shared by spouses).

Although based on only a limited number of studies, multiple sclerosis and alopecia areata showed significant inverse association with gastric cancer. Both of these diseases share similar characteristics of increased T helper type 1 immune activation and altered cytokine levels which were postulated to have inhibitory effects on carcinogenesis [12,39].

Our study has some limitations and thus the results should be interpreted with caution. First, like many other studies investigating patient populations, the effect of treatments for autoimmune conditions could not be fully addressed. Immunosuppressive medications or immunomodulatory drugs, such as steroids, cytotoxic agents and nonsteroidal antiinflammatory drugs, may themselves affect gastric carcinogenesis. There may also be decreased medical surveillance in severely debilitating diseases which might explain the inverse association for multiple sclerosis. Second, despite our attempt to include all available literature, only a relatively small number of studies met inclusion criteria, and many autoimmune diseases had insufficient data for a meta-analysis. However, autoimmune diseases with small numbers of studies and comparisons that lack statistical significance should not be interpreted as evidence against association. Third, most of the studies were based in Western populations and prevalence of many autoimmune diseases varies across race/ethnicity and geographical location, limiting generalizability of the pooled risk estimates. Fourth, autoimmune diseases are underdiagnosed conditions, and some degree of misclassification of exposure is likely. Such nondifferential misclassification would tend to attenuate true associations between autoimmune disease and gastric cancer. Fifth, as in any meta-analysis, possible publication bias is a concern although we did not find evidence for missing information with trim and fill correction. Sixth, clinical assessment of autoimmune diseases may vary across the studies. Lastly, gastric cancer risk factors such as H. pylori infection or smoking were not taken into account in many of these studies.

Despite the above limitations, our study has many strengths. Our systematic review is the first attempt to comprehensively summarize current evidence associating autoimmunity with gastric cancer risk. By pooling small studies, we were able to increase power for novel findings. Most of the studies were prospective investigations of incident gastric cancer, minimizing the possibility of reverse causality. Lastly, most of the included studies were classified as good quality.

The rarity of autoimmune diseases with their characteristic predominance in females, developed countries and Western populations, combined with the opposite profile for gastric cancer burden (males, under-developed countries, Asians), may have hampered recognition of the links between these entities. Future studies are warranted to investigate these associations in different races and ethnic groups, by anatomical subsite or tumor histology, and with consideration of treatment effects. Moreover, our findings may have important clinical and public health implications. The apparent increases in autoimmune disease and the newly recognized relation to gastric cancer risk have important implications for predicting future cancer incidence and planning strategies for control and prevention.