Abstract

Background: Several studies have reported an association between gastric atrophy and upper gastrointestinal cancers. Our aim was to summarise the available information and calculate the relative risks (RRs) associated with gastric atrophy for gastric cardia adenocarcinoma (GCA), oesophageal squamous cell carcinoma (OSCC), and oesophageal adenocarcinoma (OAC) by conducting a systematic review and meta-analysis.

Methods: We searched the PubMed and ISI-Web of Science databases, as well as the reference lists of the relevant articles. Summary RRs and 95% confidence intervals (95% CI) were calculated using random-effects models for the association between gastric atrophy, defined histologically or by serum pepsinogen markers, and OSCC, OAC, and GCA.

Results: Eighteen articles were included in the meta-analysis; 13, 7, and 3 studies reported on GCA, OSCC, and OAC, respectively. The overall RRs (95% CI) for the three cancer types were: GCA, 2.89 (2.09–3.98); OSCC, 1.94 (1.48–2.55); OAC, 0.51 (0.19–1.37). Several subgroup analyses showed the robustness of the results. In the majority of the analyses, there was low to moderate heterogeneity.

Conclusions: This study found two- to threefold increased risk of OSCC and GCA but a possible reduced risk of OAC in people with gastric atrophy. Further studies are needed to establish the association with OAC and causal association with OSCC, and mechanisms of the increased risk need to be investigated for GCA.

introduction

Each year more than 1 million people die of cancers of the oesophagus and stomach [1]. The two predominant histological types of oesophageal cancer, i.e. oesophageal squamous cell carcinoma (OSCC) and oesophageal adenocarcinoma (OAC), have distinct risk factors. For example, alcohol consumption is a strong risk factor for OSCC but not for OAC [2]. Likewise, risk factors are not necessarily similar for the two main anatomic subgroups of gastric cancer, i.e. gastric cardia adenocarcinoma (GCA) and noncardia gastric adenocarcinoma (NCGA) [3]. For example, while Helicobacter pylori is a strong risk factor for NCGA, the association of this organism with GCA is less clear [4].

Atrophy of the gastric mucosa has long been known as a risk factor for NCGA [5–7]. However, the association of atrophy with OSCC, OAC, and GCA is less well studied. Until recently, GCA constituted a minority of all gastric cancers [8, 9]. Therefore in most studies, due to small numbers, the GCA results were either not reported or did not receive enough scrutiny. Higher risk of OSCC was initially reported in patients with pernicious anaemia [10], an autoimmune disease that includes atrophy of the stomach, and also in case series of other patients with atrophy [11]. However, it has received more attention in the past few years after a Swedish case–control study found a strong association [12]. OAC was rare until the 1990s [9, 13], and there are still relatively few studies that have investigated the association between atrophy and this type of cancer.

Gastric atrophy is characterised by loss of specialised glands in the stomach and their replacement by metaplastic cells and interstitial fibrosis [14]. Therefore, this condition can be diagnosed using histological examination. With the advent of atrophy, specialised gastric cells in gastric fundus that produce pepsinogen 1 (PG1) are replaced by metaplastic cells of antral type. While serum PG1 level is reduced in gastric atrophy, pepsinogen 2 (PG2) level remains stable or is even increased because PG2 is produced in the mucosa of the gastric cardia, fundus, antrum, and duodenum [15]. Therefore, atrophy can also be diagnosed with low serum PG1 or a low ratio of serum PG1 to PG2 (PG1:PG2 ratio).

In this systematic review and meta-analysis, we have examined the association between gastric atrophy, defined histologically or by serum pepsinogen markers, and OSCC, OAC, and GCA.

methods

selection of studies

We searched the PubMed and ISI-Web of Science databases for all case–control or cohort studies published in English language on the association of gastric atrophy with oesophageal cancer and GCA risk. We used the following terms to search the PubMed database (esophag* OR oesophag* OR gastric OR stomach OR cardia) AND (cancer OR carcinoma OR adenocarcinoma) AND (“gastritis, atrophic” [MeSH] OR (gastric atrophy) OR pepsinogen) AND ((case–control) OR (case–control) OR (cohort) OR (prospective)). Equivalent terms were used to search text words in the ISI-Web of Science database. All results were updated on 4 January 2010. Using this approach, 416 and 273 publications were retrieved from the PubMed and ISI-Web of Science databases, respectively. After considering the matching articles in the two databases as a single article, 552 publications remained for further evaluation. Eleven additional publications were found by manual search of the bibliographies of the relevant original and review articles and added to list. After exclusion of 542 non-relevant articles, three articles [16–18] were excluded because the original or updated information was available in 18 full-text publications that were finally included in the meta-analysis [7, 12,19–34]. More detailed information on the study selection process is presented in supplemental File 1 (available at Annals of Oncology online). To reduce the possibility of missing the published articles, at least two of the authors reviewed each publication. We contacted the authors of one study [28] as further information was required.

data extraction and statistical analysis

Where data were available, we extracted information on first author, place and year of the study, and other study characteristics such as study design and the control selection method. We also extracted the following information by cancer subtype, when applicable: the number of cases and controls; the crude and adjusted relative risks (RRs), i.e. any ratio measures of effect, including odds ratio, and 95% confidence intervals (CI); and the variables for which the results were adjusted.

We used both random-effects models (DerSimonian–Laird method) and fixed-effects models (Mantel–Haenszel method) to calculate summary RRs and 95% CIs. Because both models yielded qualitatively similar results, we only present those from random-effects models. When an individual study presented both crude and adjusted RRs and 95% CIs, we used the maximally adjusted results.

We compared gastric atrophy status among cases and controls. The status was determined by direct examination of the gastric tissue specimens or by the level of PG1 or PG1:PG2 ratio in serum samples. Different cut-off points were used for the serologic tests. When results for several cut-off points were presented, we used those that were in more concordance with the cut-off points in other studies, i.e. PG1 < 28 to ≤ 70 μg/l and PG1:PG2 ratio < 2 to ≤ 3. If the study reported both PG1 level and PG1:PG2 ratio, we used the latter, because PG1:PG2 ratio may be a more accurate indicator for gastric atrophy than PG1 level [35]. One study that used histology reported the results for gastric atrophy, intestinal metaplasia, and dysplasia separately [33]. We combined the results using a fixed-effects model.

We conducted several subgroup analyses. All the included studies were from Asia, Europe, and the United States. It has been suggested that the association between H. pylori and GCA may vary between Western and Eastern countries [36]. Since H. pylori is an important cause of atrophy, we chose to conduct the analyses by geographic region. Several studies recruited their control participants from those who referred to gastroenterology clinics. As the rate of gastric atrophy in those hospital-based controls might be different from the rate in the general population, we performed subgroup analyses that included only population-based controls. Consistent with our previous meta-analyses [37, 38], we present the results for larger studies, defined as those with standard errors < 0.5, separately. Since the association between gastric atrophy and oesophageal and cardia cancer could be confounded by other factors, we also calculated and present the summary of adjusted RRs (95% CIs) for studies in which the results were controlled (by matching or adjustments) for the main risk factors, i.e. age for all cancers, tobacco and alcohol use for OSCC, and tobacco use and body mass index or total energy intake (as an indirect indicator of body mass index) for OAC and GCA. As the sensitivity of PG1 level and PG1:PG2 ratio for detection of gastric atrophy may be different, the studies that reported PG1:PG2 ratio were also analysed separately.

We used Begg and Mazumdar’s method to calculate P for rank correlation and Egger’s weighted regression method to calculate P for publication bias. We also plotted Begg’s funnel plot to examine small study effects. To examine heterogeneity among studies, the Q-statistic (using Mantel–Haenszel weights) and the I2 statistic were calculated. Throughout the article, two-sided P < 0.05 was considered as statistically significant.

results

Table 1 shows summary characteristics of the 18 studies included in this meta-analysis, of which 13, 7, and 3 studies reported on GCA, OSCC, and OAC, respectively. Except for two studies in which histological examination of the stomach tissue was used to determine gastric atrophy, the other studies measured pepsinogen level in serum samples. The majority of the studies (n = 14) were population based.

Table 1.

Summary of studies on the association between gastric atrophy and risk of oesophageal and gastric cardia cancer

Study; country Case/control Studied cancer Method of diagnosis for atrophic gastritisa Adjusted resultsb Study design 
Parsonnet et al. [7]; United States 30/30 GCA S: PG1 < 50 μg/l No Nested CCS (in a cohort of subscribers to a medical care program) 
Hattori et al. [19]; Japan 1/4858 GCA S: PG1:PG2 ratio ≤ 3 and PG1 ≤ 70 μg/l; PG1:PG2 ratio < 3 and PG1 < 50 μg/l; PG1:PG2 ratio ≤ 2 and PG1 ≤ 30 μg/l No Screening study among steel workers, and follow-up for one year 
Fukuda et al. [20]; Japan 52/112 GCAc S: PG1:PG2 ratio < 3; PG1 < 30 μg/l No H-B CCS 
Ye et al. [12]; Sweden 75/449 OSCC S: PG1 < 28 μg/ld Yese P-B CCS 
67/449 OAC  
103/449 GCA  
Watabe et al. [21]; Japan 2/6940 GCA S: PG1:PG2 ratio ≤ 3 and PG1 ≤ 70 μg/l No P-B screening study, and follow-up for a mean period of 4.7 year 
Nomura et al. [22]; United States 29/334 GCA S: PG1:PG2 ratio ≤ 2; PG1 ≤ 30 μg/l No P-B CCS 
Oishi et al. [23]; Japan 23/NAf GCAc S: PG1:PG2 ratio ≤ 3 and PG1 ≤ 70 μg/L; PG1:PG2 ratio ≤ 2 and PG1 ≤ 30 μg/l Yes P-B Cohort 
Knekt et al. [24]; Finland 32/63 GCA S: PG1 < 49 μg/l No Nested CCS (in a P-B cohort) 
Palli et al. [25]; Several European countries 54/∼200g GCA S: PG1 < 22 μg/l No Nested CCS (in a P-B cohort) 
Hansen et al. [26]; Norway 44/132 GCA S: PG1:PG2 ratio < 2.5 No Nested CCS (in a cohort of blood donors) 
Anderson et al. [27]; Ireland 131/260 OAC S: PG1:PG2 ratio ≤ 3, and ≤7 Yese CCS; H-B cases, P-B controls 
92/260 GCA  
Derakhshan et al. [28]; Iran 19/38 OAC S: PG1:PG2 ratio < 2.5 No H-B CCS 
53/53 GCA  
Iijima et al. [29]; Japan 90/89 OSCC S: PG1:PG2 ratio < 2; PG1 < 25 μg/l Yes H-B CCS 
Yokoyama et al. [30]; Japan 90/180 OSCC S: PG1:PG2 ratio ≤ 3, and ≤ 2; PG1 ≤ 70, and ≤ 30 μg/l No P-B CCS among alcoholics 
Kamangar et al. [31]; China 125/250 Moderate to severe squamous dysplasia S: PG1:PG2 ratio ≤ 3, ≤ 4, ≤ 9, and ≤ 12; PG1 ≤ 30, ≤ 50, ≤ 100, and ≤ 130 μg/l No P-B screening study 
Ren et al. [32]; China 323/974 OSCC S: PG1:PG2 ratio ≤ 3, ≤4, ≤5, and ≤6; PG1 ≤ 50 μg/l Yes Nested CCS (in a cohort of participants in an intervention trial) 
546/974 GCA  
De Vries et al. [33]; the Netherlands 126/NAf OSCC Histology: gastric atrophy, intestinal metaplasia, and dysplasia No Nationwide gastric atrophy cohort 
Akiyama et al. [34]; Japan 253/253 OSCC Histology: gastric mucosal atrophy open-type 2 and 3 Yes H-B CCS 
Study; country Case/control Studied cancer Method of diagnosis for atrophic gastritisa Adjusted resultsb Study design 
Parsonnet et al. [7]; United States 30/30 GCA S: PG1 < 50 μg/l No Nested CCS (in a cohort of subscribers to a medical care program) 
Hattori et al. [19]; Japan 1/4858 GCA S: PG1:PG2 ratio ≤ 3 and PG1 ≤ 70 μg/l; PG1:PG2 ratio < 3 and PG1 < 50 μg/l; PG1:PG2 ratio ≤ 2 and PG1 ≤ 30 μg/l No Screening study among steel workers, and follow-up for one year 
Fukuda et al. [20]; Japan 52/112 GCAc S: PG1:PG2 ratio < 3; PG1 < 30 μg/l No H-B CCS 
Ye et al. [12]; Sweden 75/449 OSCC S: PG1 < 28 μg/ld Yese P-B CCS 
67/449 OAC  
103/449 GCA  
Watabe et al. [21]; Japan 2/6940 GCA S: PG1:PG2 ratio ≤ 3 and PG1 ≤ 70 μg/l No P-B screening study, and follow-up for a mean period of 4.7 year 
Nomura et al. [22]; United States 29/334 GCA S: PG1:PG2 ratio ≤ 2; PG1 ≤ 30 μg/l No P-B CCS 
Oishi et al. [23]; Japan 23/NAf GCAc S: PG1:PG2 ratio ≤ 3 and PG1 ≤ 70 μg/L; PG1:PG2 ratio ≤ 2 and PG1 ≤ 30 μg/l Yes P-B Cohort 
Knekt et al. [24]; Finland 32/63 GCA S: PG1 < 49 μg/l No Nested CCS (in a P-B cohort) 
Palli et al. [25]; Several European countries 54/∼200g GCA S: PG1 < 22 μg/l No Nested CCS (in a P-B cohort) 
Hansen et al. [26]; Norway 44/132 GCA S: PG1:PG2 ratio < 2.5 No Nested CCS (in a cohort of blood donors) 
Anderson et al. [27]; Ireland 131/260 OAC S: PG1:PG2 ratio ≤ 3, and ≤7 Yese CCS; H-B cases, P-B controls 
92/260 GCA  
Derakhshan et al. [28]; Iran 19/38 OAC S: PG1:PG2 ratio < 2.5 No H-B CCS 
53/53 GCA  
Iijima et al. [29]; Japan 90/89 OSCC S: PG1:PG2 ratio < 2; PG1 < 25 μg/l Yes H-B CCS 
Yokoyama et al. [30]; Japan 90/180 OSCC S: PG1:PG2 ratio ≤ 3, and ≤ 2; PG1 ≤ 70, and ≤ 30 μg/l No P-B CCS among alcoholics 
Kamangar et al. [31]; China 125/250 Moderate to severe squamous dysplasia S: PG1:PG2 ratio ≤ 3, ≤ 4, ≤ 9, and ≤ 12; PG1 ≤ 30, ≤ 50, ≤ 100, and ≤ 130 μg/l No P-B screening study 
Ren et al. [32]; China 323/974 OSCC S: PG1:PG2 ratio ≤ 3, ≤4, ≤5, and ≤6; PG1 ≤ 50 μg/l Yes Nested CCS (in a cohort of participants in an intervention trial) 
546/974 GCA  
De Vries et al. [33]; the Netherlands 126/NAf OSCC Histology: gastric atrophy, intestinal metaplasia, and dysplasia No Nationwide gastric atrophy cohort 
Akiyama et al. [34]; Japan 253/253 OSCC Histology: gastric mucosal atrophy open-type 2 and 3 Yes H-B CCS 
a

When several serological cut-offs were presented, the ones highlighted in bold font were used in the current meta-analysis.

b

Reporting relative risks controlled (by matching controls or doing adjustments in statistical models) at least for age for all cancers, as well as tobacco and alcohol use for OSCC and tobacco use and body mass index or total energy intake (as an indirect indicator of body mass index) for OAC and GCA.

c

Adenocarcinoma in proximal one-third of the stomach.

d

Participants with high PG1 level (>158 μg/l) were categorised separately. They were not included in the current analyses.

e

Also adjusted for at least one socioeconomic status indicator, e.g. education level.

f

The number of cases was compared with standardised incidence rates.

g

A total of 910 controls (approximately four controls for each case) were enrolled for a total of 233 gastric cancer cases (including noncardia cancers).

CCS, case–control study; GCA, gastric cardia adenocarcinoma; H-B, hospital based; OAC, oesophageal adenocarcinoma; OSCC, oesophageal squamous cell carcinoma; NA, not applicable; PG1, pepsinogen 1; PG2, pepsinogen 2; PG1:PG2 ratio, pepsinogen 1 to pepsinogen 2 ratio; P-B, population based; S, serological assessment.

gastric cardia adenocarcinoma

Thirteen studies were included in this analysis (Figure 1A and Table 2). The overall RR (95% CI) was 2.89 (2.09–3.98). The I2 statistic was 34%, suggesting little to moderate heterogeneity, and the P value associated with the Q-statistic was 0.11 (Table 2). Neither Begg and Mazumdar’s method (P for rank correlation = 1.00) nor Egger’s weighted regression method (P for bias = 0.65) showed evidence for publication bias. Begg’s funnel plot for the association is shown in supplemental Figure S1A (available at Annals of Oncology online).

Table 2.

Summary statistics for the association between gastric atrophy and oesophageal and gastric cardia cancers

 Number of studies I2 (%)a P valueb Random-effects, RR (95% CI) 
Gastric cardia adenocarcinoma     
    Overall analysis 13 34 0.11 2.89 (2.09–3.98) 
    Studies from Asia 41 0.13 2.72 (1.67–4.44) 
    Studies from Europe and the United States 33 0.18 3.07 (1.95–4.83) 
    Population-based studies 11 34 0.13 2.65 (1.84–3.81) 
    Large studies (standard error < 0.5) 42 0.10 2.73 (1.94–3.83) 
    Adjusted results 59 0.06 2.56 (1.54–4.25) 
    Studies using PG1:PG2 ratio 14 0.32 2.43 (1.78–3.32) 
Oesophageal squamous cell carcinoma     
    Overall analysis 49 0.07 1.94 (1.48–2.55) 
    Studies from Asia 22 0.27 1.64 (1.21–2.23) 
    Studies from Europe and the United States 63 0.10 2.71 (1.43–5.11) 
    Population-based studies 54 0.07 1.91 (1.33–2.73) 
    Large studies (standard error < 0.5) 57 0.04 1.95 (1.45–2.63) 
    Adjusted results 69 0.02 2.10 (1.19–3.70) 
    Studies using PG1:PG2 ratio 42 0.16 1.71 (1.08–2.69) 
Oesophageal adenocarcinoma     
    Overall analysis 55 0.11 0.51 (0.19–1.37) 
    Studies using PG1:PG2 ratio 0.99 0.29 (0.13–0.47) 
 Number of studies I2 (%)a P valueb Random-effects, RR (95% CI) 
Gastric cardia adenocarcinoma     
    Overall analysis 13 34 0.11 2.89 (2.09–3.98) 
    Studies from Asia 41 0.13 2.72 (1.67–4.44) 
    Studies from Europe and the United States 33 0.18 3.07 (1.95–4.83) 
    Population-based studies 11 34 0.13 2.65 (1.84–3.81) 
    Large studies (standard error < 0.5) 42 0.10 2.73 (1.94–3.83) 
    Adjusted results 59 0.06 2.56 (1.54–4.25) 
    Studies using PG1:PG2 ratio 14 0.32 2.43 (1.78–3.32) 
Oesophageal squamous cell carcinoma     
    Overall analysis 49 0.07 1.94 (1.48–2.55) 
    Studies from Asia 22 0.27 1.64 (1.21–2.23) 
    Studies from Europe and the United States 63 0.10 2.71 (1.43–5.11) 
    Population-based studies 54 0.07 1.91 (1.33–2.73) 
    Large studies (standard error < 0.5) 57 0.04 1.95 (1.45–2.63) 
    Adjusted results 69 0.02 2.10 (1.19–3.70) 
    Studies using PG1:PG2 ratio 42 0.16 1.71 (1.08–2.69) 
Oesophageal adenocarcinoma     
    Overall analysis 55 0.11 0.51 (0.19–1.37) 
    Studies using PG1:PG2 ratio 0.99 0.29 (0.13–0.47) 
a

Higgins I2 statistic for heterogeneity

b

P value for chi-square test for heterogeneity (Q-statistics).

CI, confidence interval; PG1:PG2 ratio, pepsinogen 1 to pepsinogen 2 ratio; RR, relative risk.

Figure 1.

Forest plots for the association of gastric atrophy with oesophageal and gastric cardia cancers. Studies are sorted in order of publication year. (A) Gastric cardia adenocarcinoma. (B) Oesophageal squamous cell carcinoma. (C) Oesophageal adenocarcinoma.

Figure 1.

Forest plots for the association of gastric atrophy with oesophageal and gastric cardia cancers. Studies are sorted in order of publication year. (A) Gastric cardia adenocarcinoma. (B) Oesophageal squamous cell carcinoma. (C) Oesophageal adenocarcinoma.

All subgroup analyses showed a statistically significant association between gastric atrophy and GCA risk (Table 2). The summary RR (95% CI) was 2.72 (1.67–4.44) for studies from Asia, 3.07 (1.95–4.83) for studies from Europe/the United States, 2.65 (1.84–3.81) for studies with population-based controls, 2.73 (1.94–3.83) for large studies, 2.56 (1.54–4.25) for studies with adjusted results, and 2.43 (1.78–3.32) for studies that reported PG1:PG2 ratio. There was no evidence for significant heterogeneity in the subgroup analyses, except for the adjusted analyses, which showed a marginally significant Q-statistic (P= 0.06). We also conducted an analysis after excluding the two studies with fewer than 10 cases [19, 21]; the results were the same as the overall estimates (data not shown).

oesophageal squamous cell carcinoma

Figure 1B and Table 2 show the results for the association between gastric atrophy and OSCC. Seven studies were included in this analysis. The overall RR (95% CI) was 1.94 (1.48–2.55). The I2 statistic was 49%, suggesting moderate heterogeneity, and the P value associated with the Q-statistic was 0.07, showing a marginally significant heterogeneity (Table 2). There was no evidence for publication bias using either Begg and Mazumdar’s method (P for rank correlation = 0.55) or Egger’s weighted regression method (P for bias = 0.84). Begg’s funnel plot for the association between gastric atrophy and OSCC is shown in supplemental Figure S1B (available at Annals of Oncology online).

In analyses by geographic region, P values associated with the Q-statistic were 0.27 and 0.10 for studies from Asia and from Europe/the United States, respectively, suggesting no significant heterogeneity in each region (Table 2). The summary point estimate was lower for studies from Asia (RR = 1.64; 95% CI: 1.21–2.23) than for studies from Europe/the United States (RR = 2.71; 95% CI: 1.43–5.11). However, there were only two studies from Europe/the United States. Summary RR (95% CI) was 1.91 (1.33–2.73) for studies with population-based controls, 1.95 (1.45–2.63) for large studies, and 2.10 (1.19–3.70) for studies with adjusted results. For these subgroup analyses, the heterogeneity was statistically significant or borderline (Table 2). For the studies that used PG1:PG2 ratio to detect gastric atrophy, the summary RR (95% CI) was 1.71 (1.08–2.69) and the P value associated with the Q-statistic did not suggest a significant heterogeneity (P = 0.16).

oesophageal adenocarcinoma

Only three publications reported on the association between gastric atrophy and OAC (Figure 1C and Table 2). The overall summary RR (95% CI) was 0.51 (0.19–1.37). The I2 statistic showed a moderate variation (55%) among study results, but the Q-statistic did not suggest significant heterogeneity (P = 0.11). Since very few studies were included in this subgroup, we do not report results of analyses for publication bias.

In subgroup analyses, each subgroup consisted of only one or two studies. Of note, when we considered the two studies that used PG1:PG2 ratio as the marker of atrophy, we observed a statistically significant association between gastric atrophy and OAC with no heterogeneity; the summary RR (95% CI) was 0.29 (0.13–0.47) and the P value associated with the Q-statistic was 0.99 (Table 2). Results of other subgroup analyses were similar to the overall results, with high heterogeneity in subgroups with two studies (data not shown).

discussion

The results of this meta-analysis showed a nearly twofold increased risk of OSCC and threefold increased risk of GCA associated with gastric atrophy. We found only three studies for the association between OAC and atrophy but the results of these studies showed no increased risk and perhaps they even suggested a reduced risk.

Gastric atrophy is a known risk factor for NCGA. In Pelayo Correa’s model, published over two decades ago, atrophy is in the pathway to intestinal type gastric carcinogenesis [39]. We found a considerable number of studies reporting on the association between gastric atrophy and GCA; nearly all studies showed increased risk. As gastric cardia is located between the oesophagus and stomach, misclassification of GCA as oesophageal and gastric noncardia adenocarcinoma, and vice versa, is possible. This can influence the observed association in our study. However, we do not expect a substantial influence because it is unlikely that all studies had high rates of such a misclassification. The direction and magnitude of increased risk is similar to what has been reported for NCGA [4]. Atrophy may increase the risk of GCA through the same mechanisms that it may increase the risk of NCGA. However, it has been suggested that there may be two types of GCA, one associated with H. pylori whose risk is increased with atrophy, and one that resembles OAC, on which gastric atrophy may have no or a protective effect [26]. Whereas this is an intriguing hypothesis, we were not able to examine it in the current meta-analysis because most studies did not provide information stratified jointly by H. pylori and atrophy. However, our results establish that overall atrophy substantially increases the risk of GCA.

An association between atrophy of the gastric mucosa and OSCC, a cancer outside the stomach, was not expected or extensively studied until recently. The results of our study, while indicating heterogeneity in strength of association, show that nearly all seven published studies have found a point estimate of higher than one for the risk. Several hypotheses have been suggested to explain why such association may exist. Atrophy of the mucosa leads to reduced acid production in the stomach, proliferation of the bacteria, and perhaps increased production of acetaldehyde and N-nitroso compounds [40–42]; these latter two chemicals may act as risk factors for OSCC [43, 44]. We would like to note that it is unclear that the association between atrophy and OSCC is causal. One study that found an RR of 2.16 (95% CI: 1.81–2.56) for the association of gastric atrophy, intestinal metaplasia, and dysplasia with OSCC also found an increased risk of small cell carcinoma of the lung (RR = 1.86; 95% CI: 1.65–2.09) [33]. The authors suggested that both associations may be attributable to another common factor, such as smoking. However, consistency of association across studies suggested by this meta-analysis; dose–response relationship shown by some studies [30, 31]; the presence of associations after adjusting for some potential confounders, including smoking; and proposed biological mechanisms discussed above argue for a causal relationship. Some studies [32, 33] did not find a dose–response relationship. For example, one study found that although atrophy increased the risk of OSCC, the severity of atrophy did not increase OSCC risk in a dose–response manner [33]. However, lack of a dose–response relationship does not exclude the possibility of a causal relationship as atrophy itself may be sufficient to increase OSCC risk and progression to metaplasia or dysplasia may not be consequential.

There were too few studies to establish or refute an association between gastric atrophy and OAC. However, the results of these studies point to a reduced risk of OAC in patients with atrophy. If such an inverse association is established, it may be due to reduced acid production and reduced acid reflux from the stomach to the oesophagus, and it may explain why carrying H. pylori is inversely associated with risk of OAC [37]. The association between gastric atrophy and OAC can be further studied in large prospective cancer studies [45] and in a number of epidemiological studies that have been established in the past two decades to study the risk factors of OAC [46–50].

To test the robustness of the results, we conducted several subgroup analyses. The increased risk of OSCC and GCA with atrophy was observed regardless of the geographic site of the study, design of the study (population based or hospital based), size of the study, adjustments, and method of ascertainment of atrophy. Therefore, the results seemed robust. For OSCC, the association was somewhat stronger in studies from Europe and the United States than those conducted in Asia, but this difference was mainly related to one study [32].

This meta-analysis has several strengths, including an extensive search, careful examination of the studies by at least two of the authors, the large number of studies included for OSCC and GCA, and subgroups analyses that showed robustness of the results. One limitation of the meta-analysis was the relatively small number of studies for OAC. It is also subject to limitations that may be associated with any meta-analysis, including combining heterogeneous populations and results [51]; however, as mentioned earlier, there was low to moderate heterogeneity among studies and subgroup analyses yielded results similar to the overall analysis. Histological examination of gastric biopsy specimen and using serum PG1:PG2 ratio are not perfect methods for diagnosis of gastric atrophy. The first method is subject to observer variation and sampling error [35]. There is no unanimous agreement on the best cut-off points for the second method; furthermore, this method may not allow for appropriate location of gastric atrophy without other additional tests, such as examination of gastrin-17 levels in serum [52, 53]. However, among the available tools to diagnose gastric atrophy in epidemiological studies, these two methods are the best tests and have comparable efficacy [35]. Although the large majority of the studies used PG1 or PG1:PG2 ratio as indicators of atrophy, they used different definitions for atrophy. While heterogeneity in definition may be a weakness, it may also be a strength, because it could reveal the influence of cut-off points on the associations. Within the range of cut-off points used, the large majority of the studies showed increased risk of OSCC and GCA regardless of the exact cut-off point.

In summary, we found two- to threefold increased risk of OSCC and GCA but a possible reduced risk of OAC in people with gastric atrophy. Further studies are needed to establish the association with OAC and casual association with OSCC, and mechanisms of the increased risk need to be investigated for GCA.

funding

Work by F. I. was supported by a PhD fellowship from the International Agency for Research on Cancer.

disclosure

The authors declare no conflict of interest.

We thank Dr M. H. Derakhshan for providing us with more detailed information on the results of their study.

References

1.
Kamangar
F
Dores
GM
Anderson
WF
Patterns of cancer incidence, mortality, and prevalence across five continents: defining priorities to reduce cancer disparities in different geographic regions of the world
J Clin Oncol
 , 
2006
, vol. 
24
 (pg. 
2137
-
2150
)
2.
Blot
WJ
McLaughlin
JK
Fraumeni
JF
Schottenfeld
D
Fraumeni
JF
Esophageal Cancer
Cancer Epidemiology and Prevention
 , 
2006
New York
Oxford University Press
(pg. 
697
-
706
)
3.
Forman
D
Burley
VJ
Gastric cancer: global pattern of the disease and an overview of environmental risk factors
Best Pract Res Clin Gastroenterol
 , 
2006
, vol. 
20
 (pg. 
633
-
649
)
4.
Helicobacter and Cancer Collaborative Group
Gastric cancer and Helicobacter pylori: a combined analysis of 12 case control studies nested within prospective cohorts
Gut
 , 
2001
, vol. 
49
 (pg. 
347
-
353
)
5.
Nomura
AM
Stemmermann
GN
Samloff
IM
Serum pepsinogen I as a predictor of stomach cancer
Ann Intern Med
 , 
1980
, vol. 
93
 (pg. 
537
-
540
)
6.
Stemmermann
GN
Samloff
IM
Nomura
AM
Heilbrun
LK
Serum pepsinogens I and II and stomach cancer
Clin Chim Acta
 , 
1987
, vol. 
163
 (pg. 
191
-
198
)
7.
Parsonnet
J
Samloff
IM
Nelson
LM
, et al.  . 
Helicobacter pylori, pepsinogen, and risk for gastric adenocarcinoma
Cancer Epidemiol Biomarkers Prev
 , 
1993
, vol. 
2
 (pg. 
461
-
466
)
8.
Brown
LM
Devesa
SS
Epidemiologic trends in esophageal and gastric cancer in the United States
Surg Oncol Clin N Am
 , 
2002
, vol. 
11
 (pg. 
235
-
256
)
9.
Steevens
J
Botterweck
AA
Dirx
MJ
, et al.  . 
Trends in incidence of oesophageal and stomach cancer subtypes in Europe
Eur J Gastroenterol Hepatol
 , 
2010
, vol. 
22
 (pg. 
669
-
678
)
10.
Hsing
AW
Hansson
LE
McLaughlin
JK
, et al.  . 
Pernicious anemia and subsequent cancer. A population-based cohort study
Cancer
 , 
1993
, vol. 
71
 (pg. 
745
-
750
)
11.
Rakic
S
Dunjic
MS
Pesko
P
Milicevic
M
Atrophic chronic gastritis in patients with epidermoid carcinoma of the esophagus
J Clin Gastroenterol
 , 
1993
, vol. 
17
 pg. 
84
 
12.
Ye
W
Held
M
Lagergren
J
, et al.  . 
Helicobacter pylori infection and gastric atrophy: risk of adenocarcinoma and squamous-cell carcinoma of the esophagus and adenocarcinoma of the gastric cardia
J Natl Cancer Inst
 , 
2004
, vol. 
96
 (pg. 
388
-
396
)
13.
Corley
DA
Buffler
PA
Oesophageal and gastric cardia adenocarcinomas: analysis of regional variation using the Cancer Incidence in Five Continents database
Int J Epidemiol
 , 
2001
, vol. 
30
 (pg. 
1415
-
1425
)
14.
Rugge
M
Correa
P
Di Mario
F
, et al.  . 
OLGA staging for gastritis: a tutorial
Dig Liver Dis
 , 
2008
, vol. 
40
 (pg. 
650
-
658
)
15.
Samloff
IM
Taggart
RT
Pepsinogens, pepsins, and peptic ulcer
Clin Invest Med
 , 
1987
, vol. 
10
 (pg. 
215
-
221
)
16.
Shikata
K
Kiyohara
Y
Kubo
M
, et al.  . 
A prospective study of dietary salt intake and gastric cancer incidence in a defined Japanese population: the Hisayama study
Int J Cancer
 , 
2006
, vol. 
119
 (pg. 
196
-
201
)
17.
Iijima
K
Koike
T
Abe
Y
, et al.  . 
Extensive gastric atrophy: an increased risk factor for superficial esophageal squamous cell carcinoma in Japan
Am J Gastroenterol
 , 
2007
, vol. 
102
 (pg. 
1603
-
1609
)
18.
de Vries
AC
van Grieken
NC
Looman
CW
, et al.  . 
Gastric cancer risk in patients with premalignant gastric lesions: a nationwide cohort study in the Netherlands
Gastroenterology
 , 
2008
, vol. 
134
 (pg. 
945
-
952
)
19.
Hattori
Y
Tashiro
H
Kawamoto
T
Kodama
Y
Sensitivity and specificity of mass screening for gastric cancer using the measurement of serum pepsinogens
Jpn J Cancer Res
 , 
1995
, vol. 
86
 (pg. 
1210
-
1215
)
20.
Fukuda
H
Saito
D
Hayashi
S
, et al.  . 
Helicobacter pylori infection, serum pepsinogen level and gastric cancer: a case-control study in Japan
Jpn J Cancer Res
 , 
1995
, vol. 
86
 (pg. 
64
-
71
)
21.
Watabe
H
Mitsushima
T
Yamaji
Y
, et al.  . 
Predicting the development of gastric cancer from combining Helicobacter pylori antibodies and serum pepsinogen status: a prospective endoscopic cohort study
Gut
 , 
2005
, vol. 
54
 (pg. 
764
-
768
)
22.
Nomura
AM
Kolonel
LN
Miki
K
, et al.  . 
Helicobacter pylori, pepsinogen, and gastric adenocarcinoma in Hawaii
J Infect Dis
 , 
2005
, vol. 
191
 (pg. 
2075
-
2081
)
23.
Oishi
Y
Kiyohara
Y
Kubo
M
, et al.  . 
The serum pepsinogen test as a predictor of gastric cancer: the Hisayama study
Am J Epidemiol
 , 
2006
, vol. 
163
 (pg. 
629
-
637
)
24.
Knekt
P
Teppo
L
Aromaa
A
, et al.  . 
Helicobacter pylori IgA and IgG antibodies, serum pepsinogen I and the risk of gastric cancer: changes in the risk with extended follow-up period
Int J Cancer
 , 
2006
, vol. 
119
 (pg. 
702
-
705
)
25.
Palli
D
Masala
G
Del Giudice
G
, et al.  . 
CagA+ Helicobacter pylori infection and gastric cancer risk in the EPIC-EURGAST study
Int J Cancer
 , 
2006
, vol. 
120
 (pg. 
859
-
867
)
26.
Hansen
S
Vollset
SE
Derakhshan
MH
, et al.  . 
Two distinct aetiologies of cardia cancer; evidence from premorbid serological markers of gastric atrophy and Helicobacter pylori status
Gut
 , 
2007
, vol. 
56
 (pg. 
918
-
925
)
27.
Anderson
LA
Murphy
SJ
Johnston
BT
, et al.  . 
Relationship between Helicobacter pylori infection and gastric atrophy and the stages of the oesophageal inflammation, metaplasia, adenocarcinoma sequence: results from the FINBAR case-control study
Gut
 , 
2008
, vol. 
57
 (pg. 
734
-
739
)
28.
Derakhshan
MH
Malekzadeh
R
Watabe
H
, et al.  . 
Combination of gastric atrophy, reflux symptoms and histological subtype indicates two distinct aetiologies of gastric cardia cancer
Gut
 , 
2008
, vol. 
57
 (pg. 
298
-
305
)
29.
Iijima
K
Koike
T
Abe
Y
, et al.  . 
Gastric hyposecretion in esophageal squamous-cell carcinomas
Dig Dis Sci
 , 
2009
, vol. 
55
 (pg. 
1349
-
1355
)
30.
Yokoyama
A
Omori
T
Yokoyama
T
, et al.  . 
Chronic atrophic gastritis and metachronous gastric cancer in Japanese alcoholic men with esophageal squamous cell carcinoma
Alcohol Clin Exp Res
 , 
2009
, vol. 
33
 (pg. 
898
-
905
)
31.
Kamangar
F
Diaw
L
Wei
WQ
, et al.  . 
Serum pepsinogens and risk of esophageal squamous dysplasia
Int J Cancer
 , 
2009
, vol. 
124
 (pg. 
456
-
460
)
32.
Ren
JS
Kamangar
F
Qiao
YL
, et al.  . 
Serum pepsinogens and risk of gastric and oesophageal cancers in the general population nutrition intervention trial cohort
Gut
 , 
2009
, vol. 
58
 (pg. 
636
-
642
)
33.
de Vries
AC
Capelle
LG
Looman
CW
, et al.  . 
Increased risk of esophageal squamous cell carcinoma in patients with gastric atrophy: independent of the severity of atrophic changes
Int J Cancer
 , 
2009
, vol. 
124
 (pg. 
2135
-
2138
)
34.
Akiyama
T
Inamori
M
Iida
H
, et al.  . 
Macroscopic extent of gastric mucosal atrophy: increased risk factor for esophageal squamous cell carcinoma in Japan
BMC Gastroenterol
 , 
2009
, vol. 
9
 pg. 
34
 
35.
Weck
MN
Brenner
H
Association of Helicobacter pylori infection with chronic atrophic gastritis: meta-analyses according to type of disease definition
Int J Cancer
 , 
2008
, vol. 
123
 (pg. 
874
-
881
)
36.
Dawsey
SM
Mark
SD
Taylor
PR
Limburg
PJ
Gastric cancer and H. pylori
Gut
 , 
2002
, vol. 
51
 (pg. 
457
-
458
)
37.
Islami
F
Kamangar
F
Helicobacter pylori and esophageal cancer risk: a meta-analysis
Cancer Prev Res (Phila Pa)
 , 
2008
, vol. 
1
 (pg. 
329
-
338
)
38.
Islami
F
Ren
JS
Taylor
PR
Kamangar
F
Pickled vegetables and the risk of oesophageal cancer: a meta-analysis
Br J Cancer
 , 
2009
, vol. 
101
 (pg. 
1641
-
1647
)
39.
Correa
P
A human model of gastric carcinogenesis
Cancer Res
 , 
1988
, vol. 
48
 (pg. 
3554
-
3560
)
40.
McColl
KE
Watabe
H
Derakhshan
MH
Sporadic gastric cancer; a complex interaction of genetic and environmental risk factors
Am J Gastroenterol
 , 
2007
, vol. 
102
 (pg. 
1893
-
1895
)
41.
Vakevainen
S
Mentula
S
Nuutinen
H
, et al.  . 
Ethanol-derived microbial production of carcinogenic acetaldehyde in achlorhydric atrophic gastritis
Scand J Gastroenterol
 , 
2002
, vol. 
37
 (pg. 
648
-
655
)
42.
Kodama
K
Sumii
K
Kawano
M
, et al.  . 
Gastric juice nitrite and vitamin C in patients with gastric cancer and atrophic gastritis: is low acidity solely responsible for cancer risk?
Eur J Gastroenterol Hepatol
 , 
2003
, vol. 
15
 (pg. 
987
-
993
)
43.
Secretan
B
Straif
K
Baan
R
, et al.  . 
A review of human carcinogens—Part E: tobacco, areca nut, alcohol, coal smoke, and salted fish
Lancet Oncol
 , 
2009
, vol. 
10
 (pg. 
1033
-
1034
)
44.
Kamangar
F
Chow
WH
Abnet
CC
Dawsey
SM
Environmental causes of esophageal cancer
Gastroenterol Clin North Am
 , 
2009
, vol. 
38
 (pg. 
27
-
57
vii
45.
Carneiro
F
Moutinho
C
Pera
G
, et al.  . 
Pathology findings and validation of gastric and esophageal cancer cases in a European cohort (EPIC/EUR-GAST)
Scand J Gastroenterol
 , 
2007
, vol. 
42
 (pg. 
618
-
627
)
46.
Shaheen
NJ
Green
B
Medapalli
RK
, et al.  . 
The perception of cancer risk in patients with prevalent Barrett’s esophagus enrolled in an endoscopic surveillance program
Gastroenterology
 , 
2005
, vol. 
129
 (pg. 
429
-
436
)
47.
Whiteman
DC
Sadeghi
S
Pandeya
N
, et al.  . 
Combined effects of obesity, acid reflux and smoking on the risk of adenocarcinomas of the oesophagus
Gut
 , 
2008
, vol. 
57
 (pg. 
173
-
180
)
48.
Corley
DA
Kubo
A
DeBoer
J
Rumore
GJ
Diagnosing Barrett’s esophagus: reliability of clinical and pathologic diagnoses
Gastrointest Endosc
 , 
2009
, vol. 
69
 (pg. 
1004
-
1010
)
49.
Pandeya
N
Williams
G
Green
AC
, et al.  . 
Alcohol consumption and the risks of adenocarcinoma and squamous cell carcinoma of the esophagus
Gastroenterology
 , 
2009
, vol. 
136
 (pg. 
1215
-
1222
)
50.
Hoyo
C
Schildkraut
JM
Murphy
SK
, et al.  . 
IGF2R polymorphisms and risk of esophageal and gastric adenocarcinomas
Int J Cancer
 , 
2009
, vol. 
125
 (pg. 
2673
-
2678
)
51.
Shapiro
S
Meta-analysis/Shmeta-analysis
Am J Epidemiol
 , 
1994
, vol. 
140
 (pg. 
771
-
778
)
52.
Storskrubb
T
Aro
P
Ronkainen
J
, et al.  . 
Serum biomarkers provide an accurate method for diagnosis of atrophic gastritis in a general population: the Kalixanda study
Scand J Gastroenterol
 , 
2008
, vol. 
43
 (pg. 
1448
-
1455
)
53.
Iijima
K
Abe
Y
Kikuchi
R
, et al.  . 
Serum biomarker tests are useful in delineating between patients with gastric atrophy and normal, healthy stomach
World J Gastroenterol
 , 
2009
, vol. 
15
 (pg. 
853
-
859
)