Distinct Chemopreventive Effects of Aspirin in Diffuse and Intestinal-Type Gastric Cancer

Although the incidence of gastric cancer in western countries has been decreasing, it remains one of the most common cancers and is one of the leading causes of cancer mortality worldwide (1). Helicobacter pylori (H. pylori) is the most important carcinogen for gastric cancer. In the stomach, H. pylori induces chronic inflammation with a massive infiltration of neutrophils and other immune cells. This is followed by gastric atrophy and intestinal metaplasia and culminates in the development of gastric cancer through the Correa pathway, a multistep histopathologic cascade. The presence of neutrophil infiltration and intestinal metaplasia in gastric biopsy samples is an independent risk factor for gastric cancer (2, 3). Thus, the inhibition of chronic inflammation and subsequent histopathologic changes may lead to the prevention of gastric cancer.

Aspirin, including buffered and enteric-coated forms, are widely used for the treatment of various types of diseases due to their anti-inflammatory and antiplatelet functions. Aspirin and NSAIDs inhibit inflammatory signals through suppression of cyclooxygenase (Cox) activity and prostaglandin (PG) synthesis (4). Experimentally, Cox inhibition by these drugs may lead to the induction of apoptosis (5) and the inhibition of angiogenesis (6), resulting in an anticarcinogenic effect in several cancer types (7). Indeed, it has been reported that aspirin exhibits a protective effect on colorectal cancer development (8, 9), and several studies have suggested that it may also have a protective effect on gastric cancer development (10–14). However, other observational studies and randomized controlled trials reported a nonsignificant association between aspirin use and the development of gastric cancer or intestinal metaplasia; thus, the effect of aspirin or NSAIDs use on gastric carcinogenesis remains undetermined (15, 16).

In previous studies, differences in age, sex, and the presence of comorbidities, and potential confounding factors in aspirin/NSAID users and nonusers were often not appropriately adjusted, which may have led to biased results that influenced the conclusions. These confounding factors include the use of drugs that may alter the background histopathologic status of the patient. To reduce these biases and confounding factors, we performed a propensity score matching analysis to specifically evaluate the protective effect of these drugs on gastric cancer development and compared the risk of gastric cancer, intestinal metaplasia, and neutrophil infiltration between aspirin users and nonusers and between NSAID users and nonusers.

We developed the Gastric Cancer Endoscopy Database of Tokyo University (Tokyo, Japan) and the Institute for Adult Diseases, Asahi Life Foundation (Tokyo, Japan), which is a retrospectively recorded database of 53,143 patients from 1996 to 2017 who underwent upper gastrointestinal endoscopy. This database included the following information: patient characteristics, indications for endoscopy, endoscopy findings, histologic findings, gastric cancer data, H. pylori infection status, and drug use. The indications for upper gastrointestinal endoscopy included gastric cancer surveillance and observed gastrointestinal symptoms. For patients who underwent repeated endoscopies, we used the data from the first examination record.

We performed two retrospective cohort analyses using our database during 1996 to 2017. We extracted data from patients who had undergone upper endoscopy for gastric cancer surveillance without a previous 5-year diagnosis of gastric cancer.

We selected a case group who received daily aspirin (100 mg enteric-coated or 81 mg buffered aspirin) for at least 30 days prior to initial upper gastrointestinal endoscopy, and a control group who did not receive aspirin (Fig. 1A). Indications for aspirin use included a history of ischemic heart disease, cerebrovascular disease, and thromboembolism. In another analysis, we selected a case group who received daily NSAIDs, including ibuprofen, indomethacin, indomethacin farnesyl, sulindac, diclofenac sodium, naproxen, etodolac, mefenamic acid, meloxicam, or loxoprofen, for at least 30 days prior to initial upper gastrointestinal endoscopy, and a control group who did not receive NSAIDs (Fig. 1B). We calculated the duration of aspirin and NSAID use as the period between first prescription and last prescription data of drugs. Follow-up period is calculated by measuring the time between first visit and last visit to our hospitals: Median follow-up period was 4.2 years.

The human investigations were performed after approval by an Institutional Review Board at the University of Tokyo and the Institute for Adult Diseases, Asahi Life Foundation and in accordance with an assurance filed with and approved by the Japanese Ministry of Health, Labor and Welfare. Written informed consent from patients was obtained from each subject in accordance with Declaration of Helsinki.

All upper gastrointestinal endoscopies were performed using an electronic video endoscope (Olympus Optical) by experienced endoscopists. Biopsy was performed to evaluate intestinal metaplasia and neutrophil infiltration for all patients except who could not discontinue antithrombotic drugs such as warfarin, thienopyridine, and direct oral anticoagulants. Biopsy specimens were obtained from two points, the antrum and middle corpus of greater curvature. Gastric cancer was diagnosed by pathologic evaluation of biopsy and/or resected specimens. Diffuse-type gastric cancer was defined by the presence of signet-ring cancer cells. Intestinal metaplasia was diagnosed by the presence of goblet cells, absorptive cells, and cells resembling colonocytes (17). Neutrophil infiltration was diagnosed by the presence of polymorphonuclear leukocytes in the lamina propria of the gastric mucosa based on a previous report (18). All histologic findings were diagnosed by experienced pathologists. In the case of discordant diagnoses between pathologists, they discussed findings and a consensus was reached.

The primary outcome was the risk of gastric cancer. All gastric cancer cases were diagnosed by upper gastrointestinal endoscopy and confirmed by pathologic examination. The secondary outcome was the risk of diffuse-type gastric cancer, intestinal-type gastric cancer, intestinal metaplasia, and neutrophil infiltration.

We evaluated data on age, sex, comorbidities, H. pylori infection status, location of cancer, and cancer stage. Age was categorized into quintiles. Comorbidities were evaluated using the Charlson comorbidity index (19). The H. pylori infection status was categorized as “positive,” “negative,” or “unknown.” H. pylori status was defined as the latest result from serological testing, a urea breath test, or a stool antigen test.

We performed a one-to-one propensity score matching analysis between aspirin users and nonusers and between NSAID users and nonusers using estimated propensity scores from each patient (20). To estimate the propensity score, we fit a logistic regression model for aspirin and NSAID use as a function of patient demographic factors, including age, sex, 15 comorbidities, and H. pylori infection status. We calculated the c-statistic to evaluate the goodness-of-fit. Each aspirin and NSAID user was matched with a nonaspirin or NSAIDs user, respectively. Furthermore, patients were matched to those who had the most similar estimated propensity score on the logit scale within a specified range (≤0.2 of the pooled SD of estimated logits) to reduce differences between the two groups.

After propensity matching, we compared the proportions of gastric cancer, diffuse-type gastric cancer, intestinal-type gastric cancer, intestinal metaplasia, and neutrophil infiltration in the aspirin and NSAID use and nonuse groups. Categorical data comparisons between the two groups were performed using the χ² test or Fisher exact test as appropriate.

Logistic regression analysis was performed to estimate the OR and 95% confidence interval for the risk of gastric cancer, diffuse-type gastric cancer, intestinal-type gastric cancer, intestinal metaplasia, and neutrophil infiltration in the 2- to 6-year and ≥6-year groups relative to the ≤2-year group in patients with aspirin.

Statistical analyses were performed using SAS software version 9.4 (SAS Institute). P < 0.05 was considered to indicate statistical significance.

A total of 2,082 patients who received aspirin were included. The median duration of aspirin use was 4.9 years. After matching, we selected 2,082 controls who did not use aspirin from 48,026 patients. The c-statistic for goodness-of-fit was 0.932 in the propensity score model. Table 1 shows the patient backgrounds in the unmatched and propensity score–matched groups. Patient distributions were closely balanced between matched groups.

Gastric cancer was observed in 17 patients in the aspirin user group and 8 patients in the nonuser group, and the proportion of gastric cancer between the groups was not significantly different (0.82% vs. 0.38%, P = 0.071); the characteristics of gastric cancer cases are shown in Supplementary Table S1. When we analyzed the proportion of intestinal-type and diffuse-type gastric cancers separately, we found that there was an 80% lower relative risk of diffuse-type gastric cancer in the aspirin user group (0.05 %) compared with the nonuser group (0.24 %), although the difference did not reach statistically significance. In contrast, the proportion of intestinal-type gastric cancer was significantly higher in the aspirin user group (0.72%) compared with that in the nonuser group (0.14%; P = 0.004). No significant association was observed between aspirin use and cancer location and stage (Supplementary Table S2). No significant difference was observed in the proportion of intestinal metaplasia (40.44% vs. 41.61%) and neutrophil infiltration (49.71% vs. 49.86%) between the two groups (Table 2). No significant association was observed between the duration of aspirin use and gastric cancer, intestinal metaplasia, and neutrophil infiltration in patients with aspirin; however, it is noteworthy that no diffuse-type cancer was observed in patients who received aspirin for more than 2 years (Supplementary Table S3).

We included 898 patients who received NSAIDs and selected 867 controls who did not receive NSAIDs from 49,211 patients using one-to-one propensity score matching (the c-statistic of the model was 0.932). Patient distributions were closely balanced between the two groups (Table 3). The median duration of NSAID use was 0.3 years.

No significant differences were observed in the proportions of gastric cancer (0% vs. 0.33%) and intestinal metaplasia (42.81% vs. 39.79%) between NSAID users and nonusers, whereas NSAID use was significantly associated with decreased proportion of neutrophil infiltration in the antrum (30.86% vs. 36.38%, P = 0.015) compared with nonusers (Table 4). No significant association was observed between aspirin use and cancer location and stage (Supplementary Table S4).

We performed a propensity score matching analysis and found different effects by aspirin on intestinal-type and diffuse-type gastric cancers. It has been suggested that aspirin and other NSAIDs may prevent gastrointestinal cancers, including gastric cancer (14, 16, 21). Cox-2 expression, which could be suppressed by aspirin and NSAIDs, is often upregulated in inflamed gastrointestinal mucosa and may promote epithelial cell proliferation and carcinogenesis (22–24). Indeed, Cox-2 overexpression induces gastric cancer development, and anti-inflammatory drugs have been found to inhibit gastric cancer progression in several mouse models (3, 25–29). Thus, a potential chemopreventive approach to inhibit gastric cancer could involve the suppression of the Cox/PG pathway by aspirin or NSAIDs. However, our current results showed that aspirin use increased the risk of intestinal-type gastric cancer unexpectedly. In contrast, long-term (>2 years) aspirin use may suppress diffuse-type gastric cancer development, suggesting that chemopreventive effects by aspirin might be specific for this type of histologic cancer.

The pathogenesis of diffuse-type and intestinal-type gastric cancers is quite different. Intestinal-type gastric cancers usually develop through the Correa pathway and are found in the atrophic and metaplastic stomach after long-term H. pylori infection. In contrast, diffuse-type cancers sometimes arise even in younger patients without H. pylori infection. On a molecular level, approximately half of intestinal-type gastric cancer cases are characterized by catastrophic chromosomal instability, with frequent mutations in the TP53 gene (1). Microsatellite instability–related cancers and EBV-related immunologic type cancers are also part of a subset of intestinal-type cancers with characteristic gene mutation signatures. In contrast, the majority of diffuse-type gastric cancers do not have evident genetic or chromosomal alterations and are therefore classified as “genome stable” type. Thus, it is not surprising that aspirin/NSAIDs may have distinct effects on these two subtypes. In addition, although the presence of intestinal metaplasia is a significant risk factor for intestinal-type gastric cancer, it is uncertain whether intestinal metaplasia is a direct precursor of gastric cancer. Thus, it would be possible that the developmental mechanisms of metaplasia and gastric cancer are distinct; thus, the effects by aspirin can be different between these two, as seen in our current analysis.

Findings in humans and mice strongly suggest that loss of E-cadherin expression or CDH1 gene mutations are key triggers of the production of signet-ring cancer cells, but accumulating evidence suggests that chronic inflammation, with or without additional genetic events, may be required for progression to invasive diffuse-type cancer (2–4, 30, 31). Although we would propose the possibility that long-term use of aspirin may prevent diffuse-type cancer by suppressing chronic inflammatory responses in the stomach, further study is required in a larger and longer cohort (32).

Previous randomized controlled trials failed to prevent the development of gastric cancer in patients with chronic gastritis by long term use of Cox-2 inhibitors (33, 34). H. pylori sometimes disappears in stomachs with advanced atrophy, but gastric cancers, in particular intestinal-type cancers, can still develop without active H. pylori infection (35). In these cases, genetic mutations and epigenetic changes may play more important roles for cancer progression than inflammatory signals, such as the Cox/PG pathway (32). Our current results showed an inverse relationship between aspirin use and intestinal-type cancer risk. Given that aspirin has apparent effects on mucosal damage in the stomach, such mucosal injury might influence epithelial proliferation and transformation; however, further validation and elucidation of the underlying mechanisms are warranted.

Recently, a retrospective epidemiologic study reported that long-term proton pump inhibitor (PPI) was significantly associated with increased risk of gastric cancer compared with non-PPI use, in patients after H. pylori eradication (36). In fact, the proportion of PPI users in aspirin users was higher (32.7 %) than that in aspirin nonusers (4.3%). Thus, we performed additional propensity score matching analysis to evaluate the effect by PPI on gastric cancer. Unlike the previous report (36), PPI use was not associated with gastric cancer, intestinal metaplasia, and neutrophil infiltration (Supplementary Tables S5 and S6). Nevertheless, we do not exclude the possibility that the greater proportion of PPI users in aspirin use group might affect the increase of intestinal-type gastric cancer cases in this group. It is likely that the effect by PPI in gastric cancer development may be restricted in H. pylori–eradicated patients; thus, further analysis with detailed information of H. pylori status is required to evaluate an exact association between PPI use and gastric carcinogenesis.

Our study had several strengths. We are the first to perform propensity score matching for studying the association between aspirin/NSAIDs and gastric cancer, by following >50,000 patients who received upper endoscopy over 10 years in two separate hospitals. We also included detailed histopathologic findings in our analysis, such as histologic type of cancers and the presence of intestinal metaplasia and neutrophil infiltration. However, our study also had several limitations. The duration of aspirin/NSAID use was relatively short compared with previous studies. Previous observational studies reported that patients who took aspirin for more than 5 years had lower relative odds and HRs of gastric cancer compared with those that did not take aspirin (14, 16). However, the median duration of aspirin and NSAID use in our cohort was 4.9 and 0.3 years, and this may have affected our statistics to evaluate a protective effect of aspirin/NSAIDs on gastric cancer development. In addition, there is a possibility that aspirin/NSAIDs may be prescribed from other hospitals, or that these drugs are used as off-prescription drugs that we cannot include in our database. However, all types of enteric-coated aspirin are restricted to prescriptional use and not commercially available in Japan. Thus, there is limited chance of off-prescription use of aspirin. Furthermore, the number of gastric cancer events was relatively small, which may lead to statistically insignificant results. Finally, we included data from patients with unknown H. pylori status.

In conclusion, aspirin/NSAID use may affect the development of gastric cancers, but it should be validated in future comprehensive cohort or randomized control studies.

M. Fujishiro reports receiving commercial research grants from EA Pharma Co., HOYA-Pentax Co., and Takeda Pharmaceutical Co., has received speakers bureau honoraria from AstraZeneka Pharmaceuticals, Daiichi-Sankyo Co., Takeda Pharmaceutical Co., and Zeria Pharmaceutical Co. No potential conflicts were disclosed by the other authors.

The funders had no role in the design of the study, data collection and analysis, decision to publish, or preparation of the manuscript.

Conception and design: R. Niikura, Y. Hayakawa

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Hayakawa, Y. Hirata, M. Konishi, N. Suzuki, S. Ihara, T. Ushiku, M. Fujishiro, M. Fukayama

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Niikura, Y. Hayakawa, Y. Hirata, M. Konishi

Writing, review, and/or revision of the manuscript: R. Niikura, Y. Hayakawa, Y. Hirata, A. Yamada, M. Fujishiro

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Niikura, M. Fujishiro

Study supervision: Y. Hayakawa, A. Yamada, K. Koike

The authors thank Tomohiro Shinozaki of Department of Biostatistics, School of Public Health, Graduate School of Medicine, The University of Tokyo for his help with the statistical analyses. This study was supported by the KAKENHI Grant-in-Aid for Scientific Research, 17K15928 (R.Niikura), 16H06749 and 17H05081 (Y.Hayakawa.), P-CREATE from AMED, the Mitsubishi Foundation, Natural Sciences, the research grant of Astellas Foundation for Research on Metabolic Disorders, the Yasuda Medical Foundation, and the Advanced Research and Development Programs for Medical Innovation (PRIME) (Y. Hayakawa.).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.