Characteristics and Incidence of Gastric Cancer Among Patients with Autoimmune and Non-autoimmune Atrophic Gastritis in a U.S. Population.

Introduction
Gastric cancer remains a leading cause of cancer-related death globally [1]. In the U.S., rates of gastric cancer are rising in adults under 50 years old [2]. Recent studies have shown that the rising rates of non-cardia gastric cancer may be due to an increase in rates of gastric neuroendocrine tumors (NETs) which has nearly doubled in incidence in the U.S. from 2000 to 2018 [3].
Chronic atrophic gastritis and gastric intestinal metaplasia (GIM) are the main precursor lesions of gastric cancer [4]. Chronic atrophic gastritis is divided into two main underlying etiologies: autoantibody-mediated autoimmune gastritis (AIG), characterized by the presence of anti-parietal cell and/or anti-intrinsic factor antibodies which target oxyntic gastric mucosa resulting in mucosal atrophy; and non-autoimmune gastritis (NAIG), mediated most commonly by Helicobacter pylori (H. pylori) infection [5]. AIG is the primary known risk factor for gastric NET; however, there are limited studies detailing the clinical profiles and risk factors for AIG [5-7]. This is further complicated by the difficulty in differentiating AIG from NAIG on histopathology by pathologists, often requiring review by an expert gastrointestinal pathologist [8, 9].
Furthermore, the risk of gastric malignancy, particularly NETs, is not well defined in patients with AIG [10-13]. While AIG is a leading risk factor for the development of NETs, its association with adenocarcinoma risk is less clear [11, 14, 15]. One meta-analysis of 27 studies showed an increased risk of gastric adenocarcinoma in AIG patients [10]. However, most of the studies included in the meta-analysis did not account for H. pylori infection among AIG patients making it difficult to differentiate the risk attributable to AIG versus H. pylori for gastric adenocarcinoma. One study with strict exclusion of H. pylori infection did not find an increased risk of gastric adenocarcinoma in patients with AIG [13].
The purpose of our study was to identify and compare the demographic and clinical risk factors in patients with AIG compared to NAIG as well as compared to normal controls without either form of atrophic gastritis or GIM in a multiethnic U.S. population. In addition, we determined the incidence of gastric malignancy, including adenocarcinoma and NETs, among patients with AIG and NAIG.

Methods

Patient Population
For this retrospective cohort study, we included patients who underwent an esophagogastroduodenoscopy (EGD) at Ben Taub Hospital from March 2006 to September 2021. Ben Taub Hospital is a safety-net county hospital within the Harris Health System (Houston, Texas) providing medical care to uninsured and underinsured residents of Harris County. We identified potential cases with AIG, cases with NAIG, and normal controls without AIG, NAIG, or GIM from two sources: the Ben Taub clinical database of histopathology results from gastric biopsies and the EPIC electronic medical record using the Slicer Dicer search tool (EPIC Hyperspace; Verona, WI). This study was approved by the Institutional Review Board at Baylor College of Medicine.

Determination of AIG Cases, NAIG Cases, and Normal Controls
Potential patients with AIG were identified from two sources: 1) keyword search for “autoimmune” and “enterochromaffin-like cell (ECL) hyperplasia” combined with keyword “atrophic gastritis” from the histopathology database which includes all patients who had gastric biopsy for any indication from 7/2016 to 7/2021, and 2) EPIC Slicer Dicer search for either positive anti-parietal cell or positive anti-intrinsic factor antibody in addition to procedure code for EGD with biopsy from 1/2012 to 9/2021. All potential AIG cases identified using either source were verified using manual review of the gastric biopsy histopathology report in the electronic medical record and review of the histopathology slide in a subset. Cases with AIG were defined as having at least 1 gastric biopsy demonstrating autoimmune atrophic gastritis or atrophic gastritis with ECL cell hyperplasia on histopathology, regardless of the presence of autoantibodies. Histopathology reports for all potential AIG cases were reviewed a second time for the purposes of the research study by an expert GI pathologist (S.J.) and a gastroenterologist (J.G.) simultaneously, not independently, to confirm the presence of AIG. In cases where AIG was difficult to determine based on the histopathology report, the histopathology slides were re-examined to confirm the AIG diagnosis by the expert GI pathologist. All gastric biopsies were obtained using the 5-biopsy updated Sydney protocol [16], which is standard procedure at Ben Taub Hospital.
We compared AIG cases to two groups: NAIG cases and normal controls. NAIG cases and normal controls (i.e., without AIG, NAIG, or GIM on gastric biopsy) were identified using the hospital histopathology database from March 2006 to November 2016 as previously reported [17]. Patients with NAIG were defined by the presence of atrophic gastritis or GIM on at least one gastric biopsy and without evidence of AIG on histopathology report. Normal controls were identified as patients who underwent EGD for any indication and had no evidence of GIM, NAIG, or AIG on any gastric biopsy site per review of the histopathology report. Two reviewers (K.C. and I.L.) manually reviewed the histopathology reports in the electronic medical record of all patients with suspected NAIG and normal controls to confirm status as a NAIG case or normal control; those with discrepancy between two reviewers were excluded.

Data Collection
We abstracted demographic (age at endoscopy, sex, race/ethnicity), lifestyle (tobacco use, alcohol use), laboratory (H. pylori infection), endoscopy, and histopathologic data from the electronic medical record. H. pylori infection was defined as a positive result on stool antigen, IgG serology, urea breath test, and/or gastric biopsy histopathology at any time. The presence of autoantibodies (i.e., anti-intrinsic factor ab > 1.1 AU/mL, anti-parietal cell ab ≥ 25 units) and gastrin levels (< 200 or ≥ 200 pg/mL; Siemens Immulite 2000 Immunochemiluminometric assay) were abstracted for AIG and NAIG cases. Vitamin B12 levels, iron levels, and the presence of co-existing autoimmune diseases (type 1 diabetes, primary biliary cirrhosis, rheumatoid arthritis, systemic lupus erythematosus, vitiligo, Sjogren syndrome, systemic sclerosis, interstitial lung disease, and hypothyroidism) were abstracted for AIG cases.

Cancer Outcomes
All AIG cases and NAIG controls were reviewed up to June 1, 2023 to determine incident gastric malignancies within the Harris Health System. Gastric malignancies were defined as adenocarcinoma, NET, lymphoma, gastrointestinal stromal tumor (GIST), or Kaposi’s sarcoma. When possible, we determined incidence rates of gastric adenocarcinoma and gastric NET separately. Prevalent malignancies (diagnosed at the same time or prior to AIG or NAIG diagnosis) were not included as incident malignancies.

Statistical Analysis
Demographic, clinical, and biomarker risk factors were compared among AIG cases with NAIG cases and normal controls using chi-squared tests for categorical variables and Student’s t-tests for continuous variables. We conducted logistic regression analyses comparing AIG cases to NAIG cases to determine demographic and clinical risk factors for AIG. Potential demographic, clinical, and biomarker risk factors included sex, age (< 40, 40–60, > 60 years), race/ethnicity (non-Hispanic [NH] white, NH black, Hispanic, NH Asian), tobacco use, alcohol use, H. pylori infection, the presence of anti-parietal cell or anti-intrinsic factor antibodies, and gastrin level. Variables with statistical significance p < 0.05 on univariate analysis (tobacco use, H. pylori) and the biomarkers were sex-, age-, and race/ethnicity-adjusted on multivariate analysis.
In a sensitivity analysis, we conducted logistic regression analysis comparing risk factors for AIG cases compared to normal controls (no atrophic gastritis or GIM). Demographic and clinical risk factors were retained in the multivariate model due to statistical significance on univariate analysis (tobacco use, H. pylori) and sex-, age-, and race/ethnicity-adjusted.
We determined crude- and age-adjusted incidence rates of all gastric malignancies. A time-to-event analysis was performed using Kaplan–Meier curves to determine the probability of cancer-free survival, stratified by AIG and NAIG status. All patients were censored time at one of the following outcomes: gastric malignancy diagnosis or most recent EGD in those without cancer outcome up to June 1, 2023.
All analysis was performed using SAS version 9.4 (SAS Institute, Cary, NC) with two-sided statistical tests, and significant defined as p < 0.05.

Results
Our cohort included 158 AIG cases, 445 NAIG controls, and 2389 normal controls without atrophic gastritis or GIM (Table 1). The mean age at index EGD was 54.7 years (standard deviation [SD] 11.3 years) among AIG cases compared to 58.1 years (SD 11.4 years) for NAIG cases and 51.5 years (SD 12.4 years) for normal controls. Compared to NAIG cases, AIG cases were more likely to be female (79.1% vs. 50.1%, p < 0.001), Hispanic (72.2% vs 47.6%, p < 0.001), never tobacco users (72.8% vs. 49.0%, p < 0.001), and never alcohol users (77.9% vs. 68.8%, p = 0.039). Similarly, significant differences in sex, race, and tobacco use were seen in AIG cases when compared to normal controls without atrophic gastritis or GIM. H. pylori infection was significantly less frequent among AIG cases than NAIG cases (30.4% vs 56.0%, p < 0.001) and normal controls without atrophic gastritis or GIM (30.4% vs 50.0%, p < 0.001). Among AIG cases who underwent antibody testing, 70 of 81 subjects (86.4%) tested positive for anti-parietal cell antibody, and 52 of 104 subjects (50.0%) tested positive for anti-intrinsic factor antibody. AIG cases were more likely to have gastrin level ≥ 200 pg/mL (25.9% vs 1.1%, p < 0.001) compared to NAIG cases.
In the AIG cohort, 135 patients (85.4%) had a serum vitamin B12 level documented, and 136 patients (86.1%) had serum iron studies documented in the electronic medical record. Of these, 53 patients (39.3%) had a vitamin B12 deficiency (< 200 pg/dL) and 110 patients (80.9%) had an iron deficiency (serum ferritin < 45 ng/mL), respectively. Among 158 AIG cases, 47 (29.7%) had other co-existing autoimmune diseases; the most common of which was hypothyroidism.
When evaluating demographic and clinical risk factors for AIG compared to NAIG cases, Hispanic race/ethnicity was associated with risk of AIG (odds ratio [OR] 2.69: 95% CI 1.24–5.86), while male sex (OR 0.27: 95% CI 0.18–0.41), current tobacco use (OR 0.20; 95% CI 0.10–0.40), former tobacco use (OR 0.54; 95% CI 0.34–0.84), and H. pylori infection (OR 0.35; 95% CI 0.24–0.51) were inversely associated with AIG (Table 2). Anti-parietal cell antibody positivity (OR 7.88; 95% CI 1.66–37.52) and gastrin levels > 200 pg/ml (OR 17.61; 95% CI 4.11–75.38) were associated with increased risk of AIG, while anti-intrinsic factor antibody positivity was not. In the multivariate model adjusted for sex, age, and race/ethnicity, male sex (adj OR 0.33; 95% CI 0.21–0.51), current tobacco use (adj OR 0.29; 95% CI 0.14–0.61), and H. pylori positivity (adj OR 0.35; 95% CI 0.24–0.53) remained inversely associated with AIG risk, while anti-parietal cell antibody positivity (adj OR 5.42; 95% CI 1.07–27.36) and gastrin > 200 pg/ml (adj OR 10.90; 95% CI 2.42–49.19) remained associated with AIG risk when compared to NAIG.
We also examined demographic and clinical risk factors between 158 AIG cases and 2389 normal controls (Table 3). In the multivariate model adjusted for sex, age, and race/ethnicity, older age (40–60 years vs. < 40 years adj OR 1.87; 95% CI 1.05–3.33; > 60 years adj OR 2.80; 95% CI 1.52–5.16), Hispanic race/ethnicity (vs. NH white adj OR 3.67; 95% CI 1.77–7.62), and NH Asian race/ethnicity (adj OR 4.18; 95% 1.31–13.37) were independently associated with increased risk of AIG compared to normal controls. Male sex (vs. female sex adj OR 0.48; 95% CI 0.32–0.71), current tobacco use (vs. never use adj OR 0.40; 95% CI 0.20–0.80), and H. pylori infection (adj OR 0.38; 95% CI 0.27–0.54) were inversely associated with the risk of AIG compared to normal controls.
Among 158 AIG cases, there were 10 incident gastric malignancies (all NETs) over 3832 person-months follow-up time (median 34 months). There were 13 gastric malignancies diagnosed at the time of AIG diagnosis on index endoscopy (11 NETs, 2 adenocarcinoma) which we classified as prevalent cancers and excluded from incidence calculation. Among 445 NAIG cases, 6 patients developed incident gastric malignancy (4 adenocarcinoma, 1 GIST, 1 EBV-associated carcinoma) over 6221 person-months follow-up time (median 47 months). The age-adjusted incidence rate of all gastric malignancies was 2.6 (95% CI 1.4–4.8) per 1000 person-years among AIG cases and 0.9 (95% CI 0.4–2.1) per 1000 person-years among NAIG cases (incidence rate ratio 2.7; 95% CI 0.9–9.1). Specifically for incidence of NET among AIG cases, the incidence was 2.6 (95% CI 1.4–4.8) per 1000 person-years (unable to calculate NET incidence in NAIG due to lack of outcome). Among NAIG cases, the incidence of gastric adenocarcinoma was 0.6 (95% CI 0.2–1.7) per 1000 person-years (unable to calculate adenocarcinoma incidence in AIG due to lack of outcome). On the Kaplan–Meier curve of malignancy-free probability for AIG cases compared to NAIG cases, the risk of gastric malignancy diverges at 30 months follow-up with increased risk of gastric malignancy in AIG cases compared to NAIG cases (p = 0.252) (Fig. 1).

Discussion
In a multiethnic U.S. population, we found female sex, older age, and Hispanic and NH Asian race/ethnicity were associated with elevated AIG risk compared to normal controls, whereas current tobacco use and H. pylori infection were inversely associated with the risk of AIG. Even when compared to NAIG patients (non-autoimmune but with precancerous atrophic gastritis and/or GIM) also at higher risk of gastric cancer, male sex, current tobacco use, and H. pylori infection remained inversely associated with the risk of AIG. For patients who underwent antibody testing, positive anti-parietal cell antibody and gastrin level ≥ 200 pg/mL were strongly associated with increased AIG risk, while positive anti-intrinsic factor antibody was not. Furthermore, patients with AIG had a 2.7-times higher risk of gastric malignancy (all incident NET) compared to NAIG cases.
Our study also found female sex to be a consistent risk factor for AIG when compared to NAIG cases and normal controls. Previous studies showed female sex to be a known risk factor for AIG [5, 18], consistent with the propensity of autoimmune disease in women. While AIG was historically thought to affect predominantly NH white populations of Northern European descent, more recent studies suggest AIG affects other racial and ethnic groups at similar rates, with potentially higher rates in Hispanic populations [5, 19-22]. Our study showed that non-white racial groups (Hispanic, NH Asian) were at increased risk for AIG compared to NH white individuals. Because non-white race is a shared risk factor for both AIG and NAIG, it should not be used to differentiate risk between AIG or NAIG.
AIG was previously reported to affect predominantly elderly population [5, 19, 23]. However, more recent epidemiological data suggest a younger age of onset, particularly in Hispanic and NH black populations [7, 24-26]. Our study found older age to be a risk factor for AIG compared to normal controls but not compared to NAIG cases, although a majority of our AIG population was under 60 years old. There is speculation that an early age of onset for AIG may be masked by a relatively asymptomatic early disease course, which could explain the varying age of onset observed across studies [5].
The presence of H. pylori infection was inversely related to AIG in our cohort, which is at baseline a cohort at high-risk for H. pylori (50% of controls positive for H. pylori consistent with our prior studies [27, 28]). It is unclear whether H. pylori infection plays a protective role against the development of AIG or if AIG-related changes to the gastric mucosa prohibit effective colonization by H. pylori. H pylori is an acid-tolerant neutrophile [29] and uses urease to neutralize gastric acid and create an environment optimal for its survival [30]. It relies on the pH gradient present within the stomach for optimal navigation and acid-responsive adherence to gastric mucosa [31]. The achlorhydria resulting from AIG limits colonization by H. pylori by impairing its chemotactic and adherence mechanisms [31, 32]. Additionally, patients with AIG have increased microbial diversity compared to those with H. pylori-induced gastritis, which may further prevent effective colonization by H. pylori in these individuals [33].
While our findings do not support an etiological role for H. pylori in AIG, other studies suggest H. pylori infection could be a trigger for AIG development. Cross-reactivity has been demonstrated between H. pylori antigens and the H + /K + ATPase found on parietal cells [34, 35]. This possibility of molecular mimicry raises concern that H. pylori may play a role in the pathogenesis of AIG in pre-disposed populations [35-38]. Some populations likely exhibit a genetic predisposition to the development of AIG as evidenced by increased frequency of certain HLA haplotypes observed in AIG patients [6, 39, 40]. It is likely that a complex interplay of both genetic and environmental factors exists, leading to disease development.
In our study, we found 2.7-times increased risk of incident gastric malignancies (all NETs during median 3-year follow-up time) in patients with AIG compared to those with NAIG; the true risk of NETs in AIG is likely even higher as no NETs occurred in the NAIG cohort, preventing calculation of an incidence rate ratio. The incidence rate of gastric malignancy in our AIG cohort (2.6 per 1000 person-years) was higher than those observed for all gastric malignancies in the U.S. in recent years (7.3 per 100,000 person-years) [41], highlighting the elevated risk of gastric malignancy in this population [3]. The 2.7-fold higher cancer risk in patients with AIG compared to those with NAIG is particularly concerning since NAIG is itself a precancerous condition with a 5- to 18-fold higher risk of gastric cancer relative to the general population [42-44]. In our study, patients with NAIG who went on to develop cancer were more likely to develop adenocarcinoma rather than NET (no incident NETs among NAIG patients with median follow-up of 4 years). In U.S. studies, the incidence of gastric cancer in the setting of NAIG or GIM ranges from 1 to 3 per 1000 person-years [45-47], consistent with the incidence rate (0.9 per 1000 person-years) reported in our cohort.
Our study included a diverse, multiethnic, U.S. population of patients with AIG, NAIG, and normal controls, which allowed us to examine race/ethnicity as a risk factor which previous studies of primarily NH White populations were not able to do [13]. We included a population at high-risk for AIG in addition to H. pylori infection (50% of normal controls had H. pylori). We had strict criteria for defining AIG cases, determined by review of the gastric biopsy histopathology report simultaneously by an expert GI pathologist and gastroenterologist with re-review of the histopathology slides by the GI pathologist if determination of AIG or NAIG status could not be made from the histopathology report alone. This is a particular strength due to the significant interobserver variability among pathologists in determining histopathological features of AIG [8, 9, 48]. Furthermore, all gastric biopsies were obtained using the updated Sydney protocol (antrum and corpus) [16], which allowed for increased accuracy of AIG and NAIG determination.
Limitations to our study include the retrospective study design which limited the availability of biomarker data on all AIG cases, NAIG cases, and controls for comparison. Because of the retrospective design of the study, the majority of NAIG cases were missing autoantibody testing, which limited our ability to fully assess associations between autoantibodies and AIG. Our study was conducted in a semiclosed safety-net hospital system; but some patients may have received surveillance endoscopies for AIG or NAIG at other medical centers, potentially leading to missed incident cancers. Additionally, the specific subtype of NET and staging information for cancer cases was not collected in our study, limiting information about disease extent and prognosis. Because we used the presence of autoantibodies to identify potential cases of AIG, which were then confirmed with gastric biopsy histopathology, the association of the anti-parietal cell antibody with AIG may have been overestimated. Additionally, not all cases in the AIG cohort had serologic testing with IgG antibodies for H. pylori to help distinguish between current and previous infection, which could be helpful in elucidating whether previous infection predisposes individuals to developing AIG.
In this study, we found cases with AIG were more likely to be female, never smokers, and negative for H. pylori infection when compared with cases with NAIG (non-autoimmune atrophic gastritis or GIM). When compared with normal controls, older age and non-white race (NH black, Hispanic, NH Asian) were associated with the risk of AIG. Furthermore, AIG patients had 2.7-fold higher incidence of gastric malignancies, all NETs, compared to NAIG patients which already has a heightened risk of gastric cancer. Further investigation is needed to identify AIG patients earlier and optimize gastric cancer screening programs for this population.