The Risk of Diffuse-type Gastric Cancer Following Diagnosis with Gastric Precancerous Lesions: A Systematic Review and Meta-Analysis

Abstract

Purpose:

Gastric cancers are classified as diffuse-type (DTGC) or intestinal-type (ITGC). DTGCs have distinct clinical and histopathologic features, and carry a worse overall prognosis compared to ITGCs. Atrophic gastritis (AG) and intestinal metaplasia (IM) are known precursors to ITGC. It is unknown if AG and IM increase risk for DTGC.

Methods:

We performed a systematic review to identify studies reporting on the association of AG/IM and DTGC. We extracted the odds ratio (OR) of the association from studies, and performed pool analysis. Subgroup analysis was performed on studies reporting histologic severity (using operative link systems) to assess if histologic severity of AG/IM was associated with higher risk.

Results:

We identified six case-control and eight cohort studies for inclusion. Both AG (pooled OR=1.9, 95% CI 1.5 to 2.4, p<0.001) and IM (pooled OR=2.3, 95% CI 1.9 to 2.9, p<0.001) demonstrated an association with DTGC. High AG severity was associated with increased risk for DTGC compared to low AG severity (OR=1.7, 95% CI 1.2 to 2.3, p=0.002). Similarly, high IM severity was associated with increased risk compared to low IM severity (OR=1.9, 95% CI 1.3 to 2.7, p=0.001).

Conclusions:

Both AG and IM are associated with DTGC. Increasing histologic severity of both AG and IM increases risk for DTGC. There may exist a common pathway between ITGC and some DTGCs mediated through mucosal precursor lesions. These data may inform future strategies of cancer risk attenuation and control.

Introduction

Gastric cancer is the fifth-leading cause of cancer incidence, and fourth-leading cause of cancer death worldwide (1). Gastric cancer can be classified into two major histologic subtypes by Lauren’s classification, with clear differences between the two subtypes with regards to epidemiology and pathogenesis (2). Intestinal-type gastric cancer (ITGC) is most classically associated with chronic infection by Helicobacter pylori (Hp), and a terminal event in the neoplastic progression termed Correa’s cascade (3). In Correa’s cascade, chronic mucosal perturbation by Hp colonization may lead to the development of intermediate, precancerous lesions including atrophic gastritis (AG) and intestinal metaplasia (IM) (4). AG and IM are clearly-established risk factors for ITGC (5, 6). Other characteristics of ITGCs include male-predominance, older age, and location in the distal stomach (where Hp infection originates) (7).

In contrast, diffuse-type gastric cancer (DTGC) occurs at younger age, equally afflicts males and females, and more-commonly originates in the proximal stomach (7). Far less is understood about the pathogenesis of DTGCs, whose development may have a stronger genetic contribution than ITGC (7). An enhanced understanding of risk factors for DTGC hold particular public-health relevance. While the incidence of ITGC has fallen over the last half-century, the incidence of DTGC is stable and even increasing in certain groups (8, 9). Moreover, diagnosis with DTGC portends significantly worse prognosis than ITGC (10), due in part to more advanced stage at time of diagnosis (11).

There have been studies to suggest an increased risk of development of DTGC following Hp detection (12). However, the relationship between gastric precancerous lesions (AG and IM) and the subsequent development of DTGC is poorly defined. An enhanced understanding of risk factors and histologic markers for DTGC may inform future early detection and prevention efforts. With this motivation, we conducted a systematic review and meta-analysis to evaluate the association between gastric precancerous lesions (AG and IM) and the development of DTGC in the published English-language literature.

METHODS

Information Sources and Study Selection

A comprehensive literature search was designed and performed using the following databases: Excerpta Medica (EMBASE, 1947–2020), the National Library of Medicine’s Medical Literature Analysis and Retrieval System Online (MEDLINE) database (1946–2020), and the Cochrane database of systematic reviews (Wiley). Only studies in the English language were included in the analysis. Members of the research team (JEW and RJH) developed a search strategy in consultation with a professional librarian (KK) using English Medical Subject Headings (MeSH) and non-MeSH terms (Supplementary Table 1). The overall search strategy involved first identifying all previously-published English-language reports on the association between AG/IM and gastric cancer (of any histologic sub-type), followed by manual review to identify studies which separated DTGCs from ITGCs. To achieve this, the search was filtered to include systematic reviews, meta-analyses, and health technology assessments in order to identify previous systematic evaluations of the association between gastric precancerous lesions (AG and IM) and the development of any histologic type of gastric cancer. Figure 1 summarizes the PRISMA flow diagram of the studies screened and reviewed for this study. This search strategy resulted in 320 unique systematic reviews, meta-analyses, or health technology assessments being identified. The title and abstract for these 320 studies were then independently reviewed by three reviewers (JEW, SEK, and BEL), and 312 excluded for not informing the research question. The full text of the eight remaining systematic reviews, meta-analyses, or health technology assessments were then reviewed. From these sources, 81 individual studies were identified which provided data on the association between AG/IM and gastric cancer (regardless of histology). The full text of these 81 individual studies were then manually reviewed, and only studies which contained original data regarding the association between AG/IM and DTGC were included for final analysis. While gastric cancer can be broadly inclusive of multiple cancer types (e.g. stromal tumors, metastatic lesions), only gastric adenocarcinomas were included in our analysis. Studies which did not report histologic classification of cancers were excluded. In addition to the search strategy, four additional studies which contained original data were identified through other sources (including a priori knowledge by authors). Fourteen studies were included in the final analysis. Potential publication bias was evaluated through funnel plot analysis.

Figure 1.

Flow diagram of the literature search and study inclusion.

Data Extraction

Two investigators (JEW and RJH) independently extracted data from each included study. For each included study, the following information was extracted: name of first author, publication year, study design, location of study, sample size of patients with precancerous lesions (AG or IM) and controls, number of DTGCs, mean age, frequency of sexes, and frequency of Hp infection. For cohort studies, the number at risk within each exposure group (AG/IM or no AG/IM) and the number of events within each exposure group were extracted. For case-control studies, the number of cases (DTGC), the number of controls, and the number of exposed (AG/IM) for cases and controls were extracted. No clinical trial data was captured.

If studies reported either the operative link for gastric atrophy (OLGA) (13) or operative link for gastric intestinal metaplasia (OLGIM) (14) score, this score was recorded. OLGA/OLGIM scores range from 0 (none) to 4 (most severe), and reflect both the histologic severity of AG/IM as well as the anatomic distribution of AG/IM (antrum, body, or in multiple areas). Discrepancy between reviewers was resolved by case conference.

Outcomes and Statistical Analysis

The primary variable of exposure was the presence of gastric precancerous lesions (both AG and IM), and the primary outcome was the development of DTGC. In certain studies, the effect estimate for DTGC was not reported and was calculated by the investigators from the extracted data. For case-control studies, the measure of association was the odds ratio (OR). For cohort studies, multiple measures of association could be utilized (incidence rate ratios, risk ratios). Because even in the presence of precancerous lesions the absolute risk of DTGC is relatively low (<10%), we chose to use ORs to describe the measure of association in cohort studies as well, such that pooled analysis could be performed. 95% confidence intervals (CI) were used to measure imprecision. We performed analysis based on both fixed- and random-effects assumptions. As markers of heterogeneity were relatively modest for both AG (I²=0.0%, p=0.7 per chi-squared test) and IM (I²=43.7%, p=0.08), we chose to retain the fixed-effects model for final analysis.

Subgroup analysis was performed comparing subjects with different histologic severity grades of AG and IM. We chose to dichotomize histologic severity into two groups: low severity (defined as OLGA or OLGIM scores of 1 or 2) and high severity (defined as OLGA or OLGIM scores of 3 or 4). Additional planned sensitivity analysis was performed, including; 1) excluding the four studies which were known a priori by the authors, 2) restricting analysis to reports with a DTGC case count of 10 or greater, and 3) studies from East Asia. To evaluate if population-level differences in demographics (age, sex geographic location) or Hp prevalence confounded or mediated the relationship between AG/IM and DTGC, meta-regression was performed. Analysis was performed using Comprehensive Meta-Analysis version 3.3 (Biostat, Inc.; Englewood, NJ, USA).

RESULTS

The characteristics of the fourteen studies included in the final analysis are depicted in Table 1 (15–28). All fourteen studies were observational in nature, six as case-control studies and eight as cohort studies. Two studies appeared to use the same retrospective database for analysis (26, 28). These two studies were treated as a single study in subsequent analysis. Ten studies originated from East Asia, whereas four originated from Western populations. Eight studies included data on AG, whereas eleven included data on IM. Four studies incorporated OLGA scoring, whereas five incorporated OLGIM scoring. Mean or median age of the populations ranged from 45 to 60 years, male frequency ranged from 46 to 100%, and the Hp prevalence ranged from 37 to 100%. The studies included data on a total of 625 DTGCs. Funnel plots comparing the effect measures (ORs) of both AG and IM with the standard error from the studies (Supplementary Figure 1) demonstrated no evidence of publication bias. Table 2 depicts the counts of incident DTGCs by exposure status (AG vs no AG, IM vs no IM) for cohort studies, and the number of individuals exposed (AG vs no AG, IM vs no IM) by case status (either DTGC or control).

Table 1:

Characteristics of Studies Included in the Systematic Review and Meta-Analysis

Study	Reference	Study Design	Location	No. of DTGCs	Age, years	Male, %	H. pylori, %	Data on IM / AG	Data on OLGA / OLGIM
Wu, 1998	(15)	Case-Control	East Asia	57	58.4	52.6%	59.6%	IM	-
Cassaro, 2000	(16)	Case-Control	Western	18	45.5	49.6%	85.9%	IM	-
Uemura, 2001	(17)	Cohort	East Asia	13	52.4	56.9%	81.7%	IM and AG	-
Ohata, 2004	(18)	Cohort	East Asia	15	49.5	100.0%	78.6%	AG	-
Watabe, 2005	(19)	Cohort	East Asia	9	48.9	68.5%	46.1%	AG	-
Yanaoka, 2009	(20)	Cohort	East Asia	20	49.8	100.0%	100.0%	AG	-
Gonzalez, 2010	(21)	Cohort	Western	2	50.0	46.7%	67.2%	IM	-
Rugge, 2011	(22)	Cohort	Western	6	55.1	45.8%	37.3%	IM and AG	OLGA and OLGIM
Sakitani, 2011	(23)	Case-Control	East Asia	20	56.6	53.5%	54.8%	IM	-
Cho, 2013	(24)	Case-Control	East Asia	214	52.8	50.5%	79.70%	IM and AG	OLGA and OLGIM
Shinchijo, 2015	(25)	Cohort	East Asia	4	60.1	51.0%	64.0%	IM	-
Yun, 2018; Baek, 2020	(26, 28)	Case-Control	East Asia	246	57.3	58.8%	77.3%	IM and AG	OLGA and OLGIM
den Hollander, 2019	(27)	Cohort	Western	1	58.0	49.1%	25.1%	IM and AG	OLGIM

DTGC, diffuse-type gastric cancer; AG, atrophic gastritis; IM, intestinal metaplasia; OLGA, operative link on gastritis assessment; OLGIM, operative link on gastric intestinal metaplasia. As Yun 2018 and Baek 2020 appeared to be based on the same retrospective data, they were treated as a single study for analytic purposes.

Table 2:

Counts of DTGCs by Exposure Status

Cohort Studies
Study	Number DTGC / Number Exposure Group
	AG	No AG	IM	No IM
Uemura, 2001	10 / 875	3 / 651	8 / 469	5 / 1057
Ohata, 2004	7 / 1347	8 / 3308	-	-
Watabe, 2005	6 / 1525	3 / 5458	-	-
Yanaoka, 2009	8 / 1484	12 / 2645	-	-
Gonzalez, 2010	-	-	2 / 192	0 / 286
Rugge, 2011	3 / 1585	3 / 2967	3 / 1534	3 / 3018
Shinchijo, 2015	-	-	0 / 232	4 / 497
Case-Control Studies
Study	Number AG / Number no AG		Number IM / Number no IM
	DTGCs	Controls	DTGCs	Controls
Wu, 1998	-	-	19 / 38	49 / 86
Cassaro, 2000	-	-	9 / 9	26 / 41
Sakitani, 2011	-	-	10 / 10	380 / 995
Cho, 2013	192 / 22	182 / 32	143 / 71	99 / 125
Yun, 2018; Baek, 2020	155 / 91	355 / 410	146 / 100	265 / 500

For cohort studies, the number of incident diffuse-type gastric cancers (DTGCs) is reported by exposure group (AG, atrophic gastritis; IM, intestinal metaplasia). For case-control studies, the number of individuals exposed (AG vs no AG, IM vs no IM) by case status (either DTGC or control) is reported. As Yun 2018 and Baek 2020 appeared to be based on the same retrospective data, they were treated as a single study for analytic purposes.

Association of AG and DTGC

Seven studies (five cohort, two case-control) evaluated the association between AG and DTCG (Figure 2). Six of these studies reported data from East Asian populations, and one study reported data from a Western population. The reported ORs ranged from 1.2 to 7.2, and no study reported an inverse association between AG and risk for DTGC. When aggregating the seven studies, the pooled OR was 1.9 (95% CI 1.5 to 2.4) which was highly significant (p<0.01). There existed only mild heterogeneity in the effect estimates between studies (I²=0.0%, 95% CI 0.0–68.1%, p=0.5). In subgroup analysis, the association between AG and DTGC remained robust among both case-control studies (OR=1.9, 95% CI 1.4 to 2.4, p<0.01) and cohort studies (OR=2.1, 95% CI 1.3 to 3.5, p = <0.01). In sensitivity analysis, the association between AG and DTGC remained robust when excluding a priori author-known studies (OR 1.8, CI 1.2–2.7, p<0.01, Supplementary Figure 2), when restricting to reports with case count ≥ 10 (OR 1.8, CI 1.5–2.3, p<0.01, Supplementary Figure 3), and when restricting to East Asian studies (OR 2.0, CI 1.5–2.4, p<0.01, Supplementary Figure 4).

Figure 2.

Forest plot demonstrating the association between presence of atrophic gastritis and risk for diffuse-type gastric cancer. Odds ratios >1.0 indicate increased risk. ■ Indicates effect sizes for individual studies, and ♦ indicates summary statistic. Horizontal bars depicts 95% confidence intervals.

Association of IM and DTGC

Nine studies (four cohort, five case-control) evaluated the association between IM and DTGC (Figure 3). Six of these studies were derived from East Asian populations, and three were derived from Western populations. Of the nine studies, four reported a significant positive association between IM and risk for DTGC (p<0.05). The other five studies reported an effect estimate with confidence interval crossing the null. The pooled estimate from these nine studies was an OR=2.3 (95% CI 1.9 to 2.9) which was highly significant (p<0.001). There existed modest heterogeneity in the effect estimates between studies (I²=43.7%, 95% CI 0.0–73.9%, p=0.08). In subgroup analysis, a significant and positive association was seen among both case-control (OR=2.3, 95% CI 1.9 to 2.9, p<0.001) and cohort (OR=2.6, 95% CI 1.1 to 6.0, p=0.03) studies. In sensitivity analysis, the association between IM and DTGC was preserved when excluding a priori author-known studies (OR 2.1, CI 1.6–2.7, p<0.01, Supplementary Figure 2), and when restricting to reports with case count ≥ 10 (OR 2.4, CI 1.9–2.9, p<0.01, Supplementary Figure 3). When analyzed by region, the six studies from East Asia demonstrated a strong association (OR=2.4, 95% CI 1.9 to 2.9, p<0.01, Supplementary Figure 4), whereas the three Western studies suggested an association which failed to reach statistical significance (OR=1.9, 95% CI 0.8 to 4.4, p=0.14).

Figure 3.

Forest plot demonstrating the association between presence of intestinal metaplasia and risk for diffuse-type gastric cancer. Odds ratios >1.0 indicate increased risk. ■ Indicates effect sizes for individual studies, and ♦ indicates summary statistic. Horizontal bars depicts 95% confidence intervals.

Risk of DTGC by OLGA/OLGIM

Three studies reported on the association between OLGA stage and DTGC (Figure 4). These comprised of two East Asian studies and one Western study. Compared to low OLGA score (defined as scores of 1 or 2), high OLGA score (defined as scores of 3 or 4) was associated with an increased risk of DTGC (OR=1.7, 95% CI 1.2 to 2.3, p=<0.01). There was moderate heterogeneity between studies (I²=59.8%, 95% CI 0.0–88.5%, p=0.08).

Figure 4.

Forest plot demonstrating the association between operative link on gastritis assessment (OLGA, panel A) score or operative link on gastric intestinal metaplasia (OLGIM, panel B) score and risk for diffuse-type gastric cancer. Low OLGA/OLGIM scores defined as 1 or 2, and high OLGA/OLGIM scores defined as 3 or 4. Odds ratios >1.0 indicate increased risk. ■ Indicates effect sizes for individual studies, and ♦ indicates summary statistic. Horizontal bars depicts 95% confidence intervals.

Four studies reported on the association between OLGIM and DTGC, two East Asian studies and two Western studies. Compared to low OLGIM score (defined as scores of 1 or 2), high OLGIM score (defined as scores of 3 or 4) was associated with an increased risk of DTGC (OR=1.9, 95% CI 1.3 to 2.7, p=<0.01). There was minimal heterogeneity between studies (I²=0.0%, 95% CI 0.0–72.3%, p = 0.7).

Meta-Regression

Meta-regression was performed to evaluate the association between population characteristics (geographic location, age, male proportion, and Hp prevalence) and the magnitude of the effect sizes (shown in Supplementary Table 2). In meta-regression, no population attribute demonstrated significant association with the effect estimate for either AG or IM.

DISCUSSION

Hp-induced precancerous lesions AG and IM have been recognized as significant risk factors for ITGC (5, 6). However, their role as potential precursors to DTGC have been controversial and incompletely described. In this systematic review and meta-analysis, we demonstrate that AG and IM are robustly associated with DTGC. Moreover, we find that increasing lesion severity (in the form of OLGA/OLGIM scores) is associated with risk for DTGC.

Most prior research on prevention of DTGC has involved understanding germline genetic or familial syndromes which elevate risk in the absence of mucosal abnormality. Hereditary diffuse gastric cancer is an inherited syndrome most often caused by mutations in the E-cadherin gene (CDH1) (29). Patients with a detected pathogenic CDH1 mutation (for instance on a multi-panel gene test) are believed at increased risk for DTGC (30, 31). Prior to gastrectomy, these elevated-risk patients are followed by intensive endoscopic surveillance with multiple biopsies (>30), as it is believed that early signet-ring cells invading into the lamina propria are not visible from surface examination (29). While some DTGCs (particularly the familial variety) may develop through these germline mechanisms, our study suggests that at least a subset of DTGCs develop through epithelial precursors. These precursor-based DTGCs may demonstrate a somatic cancer genomic pathway common to ITGCs, mediated through Hp infection.

With enhanced understanding of the genomic underpinnings of gastric cancer, the distinction between ITGC and DTGC is beginning to blur. For instance, CDH1 alterations (mutations, copy number variations, or epigenetic changes) are found in 26% of ITGCs (32). In addition, approximately 25% of gastric cancers are considered mixed-type tumors (33, 34). In these mixed-type cancers, loss of E-cadherin expression is seen in only one component of the tumor, suggesting that a diffuse-type clone may arise from an intestinal-type tumor (35). It is also important to note that since Lauren’s initial publication (2), additional classification schema for gastric cancer have been developed. One example is the World Health Organization classification which recognizes four major histologic patterns: tubular, papillary, mucinous and poorly cohesive (including signet ring cell carcinoma), and uncommon histologic variants (36). The classification is based on the dominant histologic subtype present in a tumor, while acknowledging the co-existence of other elements. Recent molecular studies such as the Cancer Genome Atlas may allow for future molecular-based cancer classifications which have greater relevance for prognostication and treatment compared to histologic classifications (37).

Current guidelines from the United States (38) and Europe (39) recommend a personalized approach to surveillance in patients diagnosed with AG/IM, with surveillance decisions based upon individual patient-level factors (race, ethnicity, family history) and histologic characteristics (extent of disease, severity of disease). Our data suggest that surveillance of AG/IM may not only improve early ITGC detection, but also DTGC detection. Furthermore, histologic severity demonstrated a direct association with subsequent risk for DTGC in our study. While OLGA and OLGIM staging is not routinely used in Western nations, these data emphasize the importance of adopting standardized protocols of endoscopic biopsy and histologic interpretation in order to risk stratify patients and choose appropriate surveillance intervals.

This study is to our knowledge the first systematic review and meta-analysis of the association between precancerous lesions and DTGCs. The study incorporated studies from diverse populations with regards to Hp prevalence, cancer incidence, and racial/ethnic constitution. This diversity of inclusion adds to the robustness of the findings. The study findings must also be interpreted in light of notable limitations. As only English-language studies were included in the analysis, there may be incomplete capture of data which inform the research question. Our search strategy relied on prior systematic reviews to identify original reports, and this strategy may introduce bias into the results. Moreover, certain studies were included based on a priori knowledge. One reason for this choice of search strategy was that in many cases, obtaining the subtype data required detailed hand review of the manuscript text, tables, and Supplementary Material. This type of intensive, manual review would not have been feasible with a more-traditional search strategy. We also note that there is precedence for this type of search strategy, which was incorporated in our methods (40, 41). We also note that our funnel-plot analysis did not show evidence for bias, and that the association between AG/IM remained significant in multiple sensitivity analyses. While some included reports specifically excluded cardia cancers, not all did. As such, some cardia cancers were likely included in our analysis. As cardia and non-cardia cancers have different etiologic bases, this is a limitation to this study. Misclassification of exposure (presence of AG or IM) may occur, as studies differed in their biopsy practice, histologic interpretation, and indications for endoscopy. This misclassification should however be non-differential. There may exist misclassification of outcome, as categorization of cancer phenotype (DTGC vs ITGC) was made by histologic interpretation in all studies (and not genomic sequencing).

In summary, in this systematic review and meta-analysis we demonstrate that both AG and IM are associated with DTGC in diverse populations. We hope these data will both improve our understanding of the mechanisms of carcinogenesis, and lead to enhanced clinical strategies for cancer detection and prevention.

Abbreviations:

AG

atrophic gastritis

EMBASE

Excerpta Medica

DTGC

diffuse-type gastric cancer

Hp

Helicobacter pylori

IM

intestinal metaplasia

ITGC

intestinal-type gastric cancer

MEDLINE

Medical Literature Analysis and Retrieval System Online

MeSH

Medical Subject Headings

OLGA

operative link for gastritis assessment

OLGIM

operative link for gastric intestinal metaplasia

OR

odds ratio

Availability of Data and Material:

All data from this systematic review and meta-analysis were extracted from publically-available data sources indexed on Excerpta Medica (EMBASE, 1947–2020), the National Library of Medicine’s Medical Literature Analysis and Retrieval System Online (MEDLINE) database (1946–2020), and the Cochrane database of systematic reviews (Wiley).