Introduction

Histamine 2-receptor antagonists (H2RAs) and proton pump inhibitors (PPIs), common acid-suppressive agents (ASAs), are the mainstay treatments for gastroesophageal reflux disease (GERD) and peptic ulcer disease (PUD). Both classes of drugs can effectively alleviate patient symptoms and decrease the frequency and duration of gastroesophageal reflux, although through different mechanisms of action [1].

During acid-suppressive therapy, hypergastrinemia, defined as an excessive gastrin level (> 100–150 pg/mL), has been implicated as a potential factor in the pathogenesis of carcinoid, which can subsequently spread to different organs. According to a large population-based study analyzing Surveillance, Epidemiology, and End Results (SEER) data, the most frequent sites for carcinoids were the colon (35.9%); small intestine (32.9%); respiratory system, including the larynx, trachea, bronchi, and lung (25.1%); and stomach (3.2%) for more than 8000 patients with carcinoid tumors [2]. Rare carcinoids were also found in the esophagus (0.04%), liver (0.2%), gallbladder (0.2%), pancreas (0.6%), and female reproductive organs (0.6%). H2RAs and PPIs, which inhibit gastrin secretion by decreasing gastric acidity, may cause hypergastrinemia. The association between hypergastrinemia and cancer is well documented in the literature [3,4,5].

Decreased gastric acidity during acid-suppressive therapy may result in bacterial overgrowth in the gut. Studies have postulated that gastric bacterial overgrowth is predictive of several nongastrointestinal clinical outcomes, including lung and liver disease, and even cancer [6, 7]. For example, small intestinal bacterial overgrowth, defined as bacterial culture of > 105 CFU/mL in the upper jejunal aspirate, is known to be directly related to the severity of liver disease [8]. Another recent study found that the alteration of gut microbiome occurred at a higher rate in patients with lung cancer compared with that in cancer-free individuals [9].

Considering these mechanisms, ASAs may be associated with cancers, and the results of previous studies regarding this association have differed by the site of cancer [10]. A meta-analysis showed an increased risk of gastric cancer in patients using PPI or H2RA, whereas it showed a lack of association between colorectal and pancreatic cancers and long-term PPIs. However, a definitive conclusion could not be made because of the limited studies included [10,11,12]. In addition, the correlation between PPI use and chronic kidney disease and liver dysfunction has been investigated [10, 13,14,15]. Thus, pooled estimates combining hazard ratios from each study according to different types of cancer and the use of PPI/H2RA are needed. We performed a systematic review and meta-analysis of the association between ASA use and the risk of various types of cancer.

Methods

Literature search

The MEDLINE, EMBASE, and the Cochrane library core databases were searched for articles published up until the end of September 2019. We used MeSH terms and text words related to cancer (“neoplasm,” “tumor,” and “adenoma”) and ASAs (“proton pump inhibitor” and “histamine H2 antagonist”). The drug name, brand name, and chemical name of all acid-suppressive agents, including PPIs (omeprazole, esomeprazole, pantoprazole, rabeprazole, lansoprazole, dexlansoprazole, tenatoprazole, and benatoprazole) and H2RAs (azacitidine, cimetidine, famotidine, lafutidine, nizatidine, ranitidine, and roxatidine), were used in the search. The details of the search strategy are noted in Supplement Table 1.

Study selection

Only studies that met the following criteria were included: (1) the study reported the association between ASAs and the risk of cancer; (2) the study compared at least two independent groups (i.e., ASA receiving group and a nonuse group); (3) the study quantified and reported the relative risk of cancer between groups by calculating parameters, such as the risk ratio (RR), hazard ratio (HR), or odds ratio (OR); (4) the studies were randomized controlled trials, nonrandomized controlled studies, and observational studies; (5) peer-reviewed original studies; and (6) English-language studies. Two reviewers independently conducted the study selection, quality assessment, and data extraction (HJS, NJ). Disagreement between the two reviewers was resolved by consensus with the third reviewer (PS). We followed the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines [16], and the study protocol was registered to PROSPERO (CRD42019131274) prior to conducting the study.

Quality assessment

The quality assessment tool used was the Risk of Bias Assessment for Non-randomized Studies (ROBANS) since we could only include observational studies. ROBANS is a domain-based evaluation tool and is developed using 39 nonrandomized studies in 2013; it shows moderate reliability and validity [17]. It is composed of five items (selection of participants, confounding variables, measurement of intervention, blinding for outcome assessment, and incomplete outcome data) and was assessed at three levels (high, unclear, or low) for each study. We added the item of recall bias as another risk of bias since some of the included studies investigated the use of ASAs using patient-reported survey.

Data extraction

We extracted the baseline characteristics, exposures, and outcomes of included studies using the prespecified protocol. The study design, country, study period, number of participants (control/case), mean age, and percentage of male participants were collected. Interventions (PPI/H2RA) and outcomes, including relative risk and 95% confidence interval (CI), exposure/follow-up period, and covariates in regression analysis or matching, were also extracted.

Data analysis

The primary outcome was the adjusted estimates of the risk of cancer associated with ASAs. We used the best-adjusted relative risks with a 95% CI after controlling the confounding variables from each included study for the meta-analysis. In the base-case analysis, we prioritized data from groups with any use of ASAs ever, PPI use, prescription drug, long-term follow-up, and the highest cumulative defined cumulative daily drug dose (cDDD), in this order. If the study only reported the relative risk of cancer by subdivision, we used the result of the most common cancer type. For example, the studies of gastric cancer reported the results of both gastric cardia and noncardia adenocarcinoma. We used the gastric cardia adenocarcinoma data in the base-case analysis and performed a subgroup analysis for each type of gastric cancer.

The inverse-variance random effect model was used to estimate the pooled data. Each study reported a different type of relative risk, such as HR, RR, or OR. In the meta-analysis, HRs were considered as RRs [18, 19], and ORs were converted to RRs using the method described by Zhang and Yu [20]. In addition, we performed subgroup analysis according to PPI/H2RA use, types of cancer (if possible), drug uptake duration, cDDD, specific subgroup patients (e.g., different types of cancers, patients with Helicobacter pylori, patients with hepatitis B or C virus), and studies of low risk of bias of measurement of intervention (i.e., ASAs taken by both prescription and over-the-counter [OTC]). Heterogeneity was assessed using the I2 test and Q statistic, with significance of the Q-statistic test being considered at P < 0.05. Heterogeneity was considered for I2 values of more than 50% [21]. The funnel plot was used to estimate possible publication bias owing to the tendency to publish studies with positive results. We used Review Manager 5.3 software (The Nordic Cochrane Centre, The Cochrane Collaboration, 2014).

Results

Literature search

Our literature search identified 49,694 articles (Fig. 1). After removal of duplicate articles, title or abstract screening was conducted for 43,585 articles. In the title/abstract review, 39,864 articles were removed and 3682 articles were excluded from the full-text review owing to one of the following reasons: no patients with cancer, no acid-suppressant therapy, ineligible study design, no comparator group available, no outcomes of interest reported, and nonoriginal studies. Finally, 39 studies were included in the systematic review and meta-analysis [22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60].

Fig. 1
figure 1

PRISMA flow diagram of the study selection. * Some studies have included the results of associations with several different types of cancer

Full size image

General characteristics of the included studies

The 39 studies investigated esophageal cancer (n = 6), gastric cancer (n = 10), colorectal cancer (n = 7), liver cancer (n = 5), pancreatic cancer (n = 7), lung cancer (n = 2), breast cancer (n = 6), prostate cancer (n = 3), kidney cancer (n = 1), melanoma (n = 2), non-Hodgkin lymphoma (n = 1), and other cancers (n = 2). Some studies have included the results of association with more than one cancer; thus, each outcome for different types of cancer, respectively, was analyzed in the meta-analysis of each cancer. There were 30 case–control studies and 11 cohort studies in total, including two cohort studies in the study by Kao et al. and a case–control and a cohort study by Tran et al. (Table 1). The studies were from several countries: the USA, Canada, the UK, Italy, Denmark, Netherlands, Iceland, Taiwan, Hong Kong, and South Korea.

Table 1 Characteristics of included studies
Full size table

Quality assessment

The items estimating a low risk of bias with more than 75% were selection of participants, blinding for outcome assessment, incomplete outcome data, and other risk of bias (recall bias) (Fig. 2). The confounding variables and measurement of intervention were assessed as more than 50% of unclear or high risk of bias, because there were studies that only reported crude estimates, and the suitable confounding covariates for the adjusted estimates were not included. ASAs can also be bought as OTC drugs in many countries; thus, we evaluated an unclear risk of bias for the measurement of intervention if the included studies indicated the possibility that the patients assessed were taking OTCs.

Fig. 2
figure 2

Quality assessment of included studies using the Risk of Bias Assessment tool for Nonrandomized Studies (ROBANS): a ROBANS graph and b ROBANS summary. +: low risk of bias; ?: unclear risk of bias; −: high risk of bias

Full size image

Acid-suppressive agents and esophageal cancer

Five studies with 15,161 individuals reported that ASAs and the risk of esophageal cancer were not significantly associated (RR, 1.00; 95% CI, 0.77–1.29), with no significant heterogeneity (I2 = 13%, P = 0.33) (Fig. 3a). We did not include the study by Habel et al. in the meta-analysis as they reported the combined relative risk of esophageal and stomach cancer. In the subgroup analysis, both PPI use and H2RA use did not show a significant association with esophageal cancer (RR, 0.75; 95% CI, 0.55–1.03 in PPI users and RR, 0.98; 95% CI, 0.72–1.32 in H2RA users) (Table 2). The association according to the treatment duration or type of esophageal cancer (adenocarcinoma and squamous cell carcinoma) was also insignificant.

Fig. 3
figure 3

The association between acid-suppressive agent use and the risk of cancer: a esophageal cancer, b gastric cancer, c colorectal cancer, d liver cancer, and e pancreatic cancer

Full size image
Table 2 Subgroup analysis of the association between acid-suppressive agent use and the risk of cancer
Full size table

Acid-suppressive agents and gastric cancer

Nine studies including 130,074 individuals estimated that ASA users showed a 46% higher risk of gastric cancer compared with that of nonusers (RR, 1.46; 95% CI, 1.18–1.80), with slight significant heterogeneity (I2 = 51%, P = 0.04) (Fig. 3b). There was no evidence of publication bias based on the funnel plot (Fig. 4b). Both PPI use and H2RA use were associated with an increased risk of gastric cancer (RR, 1.53; 95% CI, 1.13–2.07 in PPI users and RR, 1.32; 95% CI, 1.08–1.60 in H2RA users) (Table 2). The significant association was also shown in patients with Helicobacter pylori. For the group consisting of individuals who used ASAs for 1 year or more/less than 1 year, the subgroup of cardia or noncardia cancer, a significant association with gastric cancer was not shown.

Fig. 4
figure 4

Funnel plot of included studies: a esophageal cancer, b gastric cancer, c colorectal cancer, d liver cancer, and e pancreatic cancer

Full size image

Acid-suppressive agents and colorectal cancer

In total, 605,043 individuals in seven studies showed no significant association between ASAs and colorectal cancer (RR, 1.02; 95% CI, 0.91–1.14) (Fig. 3c). We could not detect any evidence for heterogeneity (I2 = 0%, P = 0.74) or publication bias (Fig. 4c). In the subgroup analysis, the results were consistent with those of the base-case analysis: PPI/H2RA, drug intake duration of less than 1 year/1 year or more, and fewer than 60 cDDDs/60 cDDDs or more (Table 2).

Acid-suppressive agents and liver cancer

Seven cohorts from five studies of the association between ASAs and liver cancer included 809,465 individuals. ASA use was significantly associated with a 53% increased risk of liver cancer compared with nonuse (RR, 1.53; 95% CI, 1.31–1.78) (Fig. 3d). Significant heterogeneity was detected (I2 = 84%, P < 0.001) and there was no evidence of publication bias based on the funnel plot (Fig. 4d). In the subgroup analysis by type of ASAs, there was no significant association between H2RA users and the risk of liver cancer, whereas PPIs were significantly associated with liver cancer (Table 2). According to the cDDD, ASA users with 365 DDDs or more and those with less than 365 DDDs did not show a significant association with the risk of liver cancer. With regard to the type of liver cancer, ASA use associated with an increased risk of hepatocellular carcinoma (RR, 1.40; 95% CI, 1.17–1.68), but not of intrahepatic bile duct carcinoma. PPI use was also associated with the increasing risk of hepatocellular carcinoma in patients with hepatitis B or C virus (RR, 1.45; 95% CI, 1.03–2.03).

Acid-suppressive agents and pancreatic cancer

Seven studies including 554,115 individuals demonstrated that the use of ASAs was not significantly related with the risk of pancreatic cancer compared with nonuse (RR, 1.50; 95% CI, 0.92–2.45) (Fig. 3e). Significant heterogeneity was shown (I2 = 84%, P < 0.001), and there was no evidence of publication bias (Fig. 4e). The subgroup analyses of PPI or H2RA, drug intake duration, and cDDDs between ASA use and the risk of pancreatic cancer did not show a significant association (Table 2).

Acid-suppressive agents and breast cancer

In total, 209,329 individuals were included in six studies to estimate the association between ASAs and breast cancer. ASA use was not significantly associated with the risk of breast cancer (RR, 0.90; 95% CI, 0.80–1.01) with significant heterogeneity (I2 = 86%, P < 0.001) (Fig. 5a). The results of subgroup analyses were consistent with those of the base-case analysis (Table 2).

Fig. 5
figure 5

The association between acid-suppressive agent use and the risk of cancer: a breast cancer and b prostate cancer

Full size image

Acid-suppressive agents and prostate cancer

Three studies including 84,522 individuals investigated the association between ASAs and prostate cancer. We did not find a significant association between the risk of prostate cancer and ASA use (RR, 1.09; 95% CI, 0.99–1.20); no heterogeneity was found (I2 = 0%, P = 0.72) (Fig. 5b).

Acid-suppressive agents and other cancers

Two studies on lung cancer and two studies of melanoma were also included in the systematic review. ASA use was not significantly associated with the risk of lung cancer or melanoma compared with nonuse with no significant heterogeneity (RR, 1.07; 95% CI, 0.91–1.27; I2 = 43%, P = 0.18 for lung cancer and RR, 0.86; 95% CI, 0.72–1.02; I2 = 0%, P = 0.73 for melanoma).

One study reported kidney cancer, non-Hodgkin lymphoma, periampullary cancer, and all types of cancer. There was no significant association between PPIs and kidney cancer (OR, 0.99; 95% CI, 0.88–0.91) in the study by Nayan et al. and between H2RA and non-Hodgkin lymphoma (aOR, 0.68; 95% CI, 0.41–1.41) in Beiderbeck et al.’s study. Chien et al. reported that PPI use increased the risk of periampullary cancer compared with nonuse (aOR, 1.35; 95% CI, 1.16–1.17). Habel et al. studied the association between cimetidine use and all types of cancer and reported no significant association for uterine, ovarian, and kidney/bladder cancers and lymphoma/myeloma/leukemia (Table 1).

Discussion

This systematic review assessed the association between ASA use and the risk of development of each cancer. We found that ASA use was associated with a 46% increase in the risk of gastric cancer and a 53% increase in the risk of liver cancer, but it was not significantly associated with other cancers, including esophageal, colorectal, pancreatic, breast, and prostate cancer. In particular, the increase in the risk of gastric and liver cancer with PPI use was higher than that with H2RA use.

The results of our meta-analysis were similar to previous studies [10, 12]. Previous systematic review reported that long-term PPI use (at least 3 months) was significantly associated with a 78% increase in the risk of gastric cancer compared with nonuse [10], which is slightly higher than our results (36%). It may be because Islam et al. investigated the risk of gastric cancer with long-term PPI, while our study included ever use of PPIs or H2RAs. Another meta-analysis found that PPIs and H2RAs were associated with a 39% and 40% increase in gastric cancer risk [12]. In our subgroup analysis, the risk of gastric cancer in PPI users was higher than H2RA users (39% vs. 26%) when compared with nonusers. The mechanism by which ASAs relate an increased risk of gastric cancer is unknown; however, several pathways have been suggested [12]. Researchers have speculated that cancer may arise from bacterial overgrowth and nitrosamine formation caused by the suppression of gastric acid formation [61,62,63,64,65]. In contrast to this theory, other researchers have proposed that acid-suppressing medications cause hypergastrinemia, which ultimately is related to gastric polyps and carcinomas [66,67,68,69,70,71,72,73,74,75,76,77].

A previous meta-analysis reported that PPI use did not show a significant association with hepatocellular carcinoma [78], but they mentioned that their meta-analysis lacked sufficient evidence to confirm the association. On the other hand, we found a statistically significant association between ASA use and liver cancer or hepatocellular carcinoma. The risk of liver cancer was associated with PPI use, but not H2RA use. The exact pathway through which PPIs associate with an increasing risk of liver cancer is unknown; however, several mechanisms have been suggested [79]. Long-term PPI use and the associated hypergastrinemia have been implicated in carcinogenic effects on liver cells [80]. Other speculated mechanisms include the possibility that bacterial overgrowth due to decreased acid secretion in the stomach causes the transformation of primary bile acids to secondary bile acids, which subsequently exert deleterious effects on the liver, possibly leading to liver cancer [81,82,83]. In addition, it should be noted that exposure of mouse models to PPIs has been demonstrated to promote liver tumors, the progression of alcoholic liver disease, nonalcoholic fatty liver disease, and nonalcoholic steatohepatitis [84, 85]. Tran et al. explained that H2RA use generally results in weaker acid suppression and lower effects on gastrin [79, 86].

Hu et al. showed that PPI use was not associated with the risk of esophageal adenocarcinoma and/or high-grade dysplasia in patients with Barrett’s esophagus [87]. We also did not find a significant association between ASAs and esophageal cancer. This result was similar for both PPIs and H2RAs. Theoretically, PPIs and H2RAs decrease esophageal acid and bile refluxate exposure of the esophagus, thereby promoting mucosal healing and acting as a potential chemoprotective modality to mitigate esophageal cancers [87]. However, the guidelines for GERD recommend the use of ASAs for symptom control and not specifically for the prevention of esophageal adenocarcinoma [88]. It is important to note that reflux symptoms are poorly correlated with the actual amount of esophageal refluxate in patients with GERD; thus, PPI exposure may not be correlated with the incidence of esophageal cancers [89].

When Islam et al. pooled the ORs of colorectal and pancreatic cancers in PPI users and compared these values to those of nonusers, no significant association was observed [10]. These results were similar to our results: the RRs of ASAs for the risk of colorectal cancer and pancreatic cancer were 1.02 (95% CI, 0.91–1.14) and 1.50 (95% CI, 0.92–2.45). We could not find the previous systematic reviews of lung, breast, and prostate cancers and ASAs.

The results of the present study should be interpreted while considering some limitations. First, cohort and case–control studies were included in the final meta-analysis. Owing to the study designs of the included studies, we could not show a causal relationship between ASAs and cancers. However, we can describe a plausible mechanism and relative. Second, the results may include potential confounders, as the meta-analysis pooled studies that reported crude relative outcomes or adjusted outcomes with insufficient covariates. Third, ASAs can be bought OTC without a physician’s prescription in most countries, so interventions may have been misclassified. We conducted the subgroup analysis for studies including both prescription medication and OTCs and the results remained consistent. Some results changed but we could not suggest them due to the small number of studies included in the subgroup analysis.

Despite these limitations, to the best of our knowledge, this is the first systematic review and meta-analysis for the association between ASA use and multiple types of cancer. We found that the increased risk of gastric and liver cancers was associated when ASAs were used, but there was no significant association between ASA use and other cancers. Although a limited number of studies were included in this meta-analysis, the results can be the best available evidence. In particular, low heterogeneity and a consistent direction were shown in esophageal cancer and colorectal cancer. We also conducted subgroup analyses according to PPI/H2RA, duration of drug uptake, subtypes of cancer, and cumulative daily drug dose; these subgroup results can provide comprehensive and detailed information. Notably, our results showed that PPI use was associated with liver cancer, whereas H2RA use was not.

Conclusions

The results of our meta-analysis suggests that ASA use was associated with an increased risk of gastric and liver cancer, but we did not find it to be significantly associated with esophageal or colorectal cancer. There was no strong evidence for the association of lung, breast, prostate, and kidney cancer; melanoma; and lymphoma risk with ASA use. The prescription of ASAs should be carefully considered under the potential risk of gastric and liver cancer until further well-designed studies with large sample cohorts confirm the results.