Pickled Food and Risk of Gastric Cancer—a Systematic Review and Meta-analysis of English and Chinese Literature

This article is featured in Highlights of This Issue, p. 873

Despite a remarkable decline in its incidence and mortality in many nations, gastric cancer remains the fourth most common malignancy and the second leading cause of cancer death in both sexes worldwide (1, 2). Distribution of gastric cancer incidence shows remarkable heterogeneity across the world. Of the nearly 1 million estimated new annual cases, approximately half occur in Eastern Asia (1).

Known risk factors of gastric cancer, most notably Helicobacter pylori (3), account for only a proportion of all cases, and do not completely explain the heterogeneity in its distribution. Only a small proportion of people who carry H. pylori in their stomachs develop gastric cancer (4), suggesting that other factors, such as eating pickled food, may be responsible in cocarcinogenesis.

Ecological studies have shown higher risks of esophageal and gastric cancers in areas with high consumption of pickled food (5, 6). In high-risk areas of China, pickled food, particularly pickled vegetables, were eaten daily for 9 to 12 months a year and constituted an important part of the family diet (5). The traditional way of preparing pickled vegetables, packing moist vegetables in a jar for a few weeks or months, allows fermentation and growth of fungi and yeasts (5, 7) and can potentially yield carcinogenic compounds such as N-nitroso compounds and mycotoxins (5, 8, 9). Indeed, a number of experimental and in vitro studies have shown mutagenicity and carcinogenicity of pickled vegetables (5, 10, 11).

Whereas ecological studies and mechanistic studies have provided support for a role of pickled food in gastric carcinogenesis, the results of epidemiologic case–control and cohort studies have not been consistent. Hence, in 1991, the Working Group of the International Agency for Research on Cancer (IARC) classified traditional Asian pickled vegetables as possibly carcinogenic to humans (7). The Working Group found limited evidence in humans and inadequate evidence in experimental animals for carcinogenicity of pickled vegetables. Since then, however, a large number of additional studies have investigated the association of pickled vegetables/foods with different cancers, including stomach cancer. In this systematic review and meta-analysis, we examine and present results of epidemiologic studies that have investigated the association of pickled vegetable/food consumption with risk of stomach cancer.

To identify publications in English or Chinese languages on the association of pickled vegetable/food consumption and risk of gastric cancer, we conducted a comprehensive search of the PubMed, Vip Chinese Periodical, and China National Knowledge Infrastructure databases. All results were updated on January 18, 2012. The following terms were used to search text words in the PubMed database: (stomach OR gastric) AND (cancer OR carcinoma OR adenocarcinoma OR neoplasm OR neoplasia OR neoplastic) AND (pickle OR pickled OR moldy OR fermented). Combinations of the following terms were used to search the Chinese databases: (⁠⁠) (diet), (⁠⁠) (pickled food), (⁠⁠) (pickled vegetable), (⁠⁠) (pickled vegetable), (⁠⁠) (pickled vegetable), (⁠⁠) (pickled vegetable), (⁠⁠) (pickled vegetable), (⁠⁠) (digestive tract), (⁠⁠) (stomach or gastric), (⁠⁠) (cancer or carcinoma), and (⁠⁠) (tumor). Using this approach, we identified 124 articles from the PubMed database and 195 from the Chinese databases. Two of the authors (F. Islami and J.-S. Ren) reviewed the publications in English language. Chinese publications were reviewed by one of the authors (J.-S. Ren).

We examined the abstracts of the selected articles and retrieved and reviewed the full text of possibly eligible articles. Furthermore, bibliographies of the relevant original and review articles were searched manually. Only case–control and cohort studies that provided risk estimates or crude numbers of cases and controls for the exposure of interest were included. Studies that reported results for cancer of the stomach in combination with other cancers were excluded. When results from an individual study were available in several publications, we used the most informative article, for example, results were reported for a higher number of participants or were adjusted for main confounding factors. For case–control studies, we limited our inclusion to those that used control subjects who did not have other upper gastrointestinal cancers. One study (12) was excluded because study subjects were atomic bomb survivors in Japan, a group heavily exposed to γ radiation, an important risk factor for gastric cancer among the exposed individuals (13). Therefore, the risk associated with other factors could be difficult to disentangle. Using these approaches, 60 full-text articles were included in this meta-analysis (14–73). A summary of the study selection process is presented in Fig. 1.

Where data were available, information on first author, place, and year of the study, the exposed and unexposed individuals by case status, the crude and adjusted OR, and 95% CI, and other study characteristics (such as study design, the method of control selection, and adjustments) were extracted. If several adjusted results were presented, we used the maximally adjusted ones. When results on multiple categories of consumption were reported, we collapsed multiple exposure categories into a single category and calculated the combined risk estimate using fixed-effects models. Depending on the available data, the reference groups included study subjects with no or the lowest amount of pickled vegetable/food consumption.

One cohort study reported results separately for all cancer cases and also after excluding cases diagnosed during the first 3 years of follow-up (69); only the latter results were included in this meta-analysis. A study conducted in Japan (17) reported data from 2 regions, one affected by atomic bomb and one not affected; only the results from the latter were considered in our analysis. Two studies reported the results separately using different groups of controls: population- and hospital-based controls (22), or including and excluding volunteers (35). In these instances, we only considered results in which population-based controls were considered and volunteers were excluded. Five studies reported their results for men and women separately (22, 23, 39, 68, 73). As the risk estimates were generally similar in both sexes, we combined the risk estimates for men and women and used those estimates. We also combined risk estimates for subgroups that were studied in only 1 or 2 studies. These included different types of pickled vegetables (18, 53), controls with and without gastric atrophy (65), individuals aged <20 and between 20 and 39 years (21), and individuals with and without family history of gastric cancer (30).

Summary risk estimates and 95% CIs were derived with random-effects models (DerSimonian-Laird method). For all analyses, we dichotomized data and compared gastric cancer risk in pickled vegetable/food users versus nonusers (11 studies) or versus those who were in the lowest reported category of use (49 studies).

We conducted several subgroup analyses including: (i) by study design (case–control and cohort); (ii) for studies with population-based controls, as the pattern of exposure might be different in population-based and hospital-based controls; (iii) for studies that presented results adjusted for at least the major potential confounding factors (age, sex, and tobacco use); (iv) by geographic area, as the method of producing pickled foods may be different in various regions; and (v) by type of exposure (pickled vegetables and pickled foods). Subgroup analyses by tumor subsite (gastric cardia and noncardia) and by excluding studies that had imprecise estimates, defined as those with SE > 0.5 in logarithmic scale (74), were also carried out. However, results of the latter 2 analyses are not reported because only 5 studies (29, 33, 40, 63, 73) reported their results by subsite [one study (63) was only on cardia] and only 2 studies (51, 70) had SE > 0.5.

Heterogeneity among studies was estimated by the I² statistic. We plotted Begg funnel plot and used Egger weighted regression method to examine publication bias. All statistical analyses were conducted with Stata (StataCorp LP, version 11) statistical software. Throughout the article, 2-sided P < 0.05 was considered as statistically significant.

Summary characteristics of the 60 studies that were included in this analysis are presented in Table 1. There were 50 case–control and 10 prospective studies. The majority of studies (n = 49) were from Eastern Asia, including 29 studies from mainland China, 13 from Japan, 6 from Korea, and one study from Taiwan. Twenty studies were retrieved from Chinese literature. The total number of cases in case–control and cohort studies was 16,448. The number of controls in case–control studies was 87,622.

Table 2 shows the results of overall and subgroup analyses. For the overall association, the OR (95% CI) was 1.52 (1.37–1.68). The I² statistic was 80% (P < 0.001), suggesting high heterogeneity. A forest plot and Begg funnel plot for overall association are shown in Figs. 2 and 3, respectively. Of the 50 case–control studies, 44 showed increased risk (30 being statistically significant) and 6 showed decreased risk (1 being statistically significant) of gastric cancer associated with use of pickled vegetables/foods. Among the 10 prospective studies, 9 showed increased risk (2 being statistically significant) and 1 showed decreased risk (not statistically significant). The distribution of the dots on the Begg funnel was fairly symmetric, and P for bias in Egger weighted regression method was 0.40, suggesting little evidence for publication bias.

Similar to the overall association, all subgroup analyses showed a slight but statistically significant association between consumption of pickled vegetables/foods and risk of gastric cancer (Table 2). The association was slightly stronger in case–control studies (OR, 1.56; 95% CI, 1.39–1.75) than in prospective studies (OR, 1.32; 95% CI, 1.10–1.59). The OR (95% CI) was 1.55 (1.36–1.76) for studies with population-based controls and 1.38 (1.22–1.56) for studies with results adjusted at least for age, sex, and smoking. When the results were stratified by country, the strongest association was observed in studies from Korea (OR, 1.89; 95% CI, 1.29–2.77) based on 6 case–control studies and the lowest OR was in studies from countries other than Korea, mainland China, and Japan (OR, 1.14; 95% CI, 0.96–1.35) based on 9 case–control and 3 prospective studies. There was no substantial difference in estimated risk between intake of pickled vegetables and other pickled food. There was high heterogeneity in all subgroup analyses.

The combined result from all studies showed an approximately 50% increase in risk of gastric cancer in those who consumed pickled vegetables/foods versus those who did not consume pickled vegetables/foods or those who used it minimally (lowest consumption category).

There are reasons to conclude that the results might be biased or confounded, and not causal. Mainly, subgroups of studies that had stronger designs showed smaller increases in risk, and these small increases in risk could still be due to bias and confounding. Case–control studies, which showed stronger associations, may be subject to interviewer bias, recall bias, and other methodologic problems (75). Cohort studies, which have been argued to generate more valid results, particularly in nutritional studies (75), showed less strong associations. Among case–control studies, those that used population-based controls or adjusted for potential confounders showed smaller increases in risk.

There are also reasons to believe that the increased risk is causal. The large majority of the studies (53 of 60) showed point estimates that were above one and only 7 were below one. All subgroup analyses showed statistically significant increased risk. The results are consistent with ecological studies (5, 6), animal studies (5), and laboratory findings (5, 10, 11) pointing toward a possible role of pickled vegetable/food in alimentary tract carcinogenesis. The results are also supported by findings from a recent meta-analysis that showed a doubling of esophageal cancer risk with consumption of pickled vegetables (74). A study from China—not included in this meta-analysis because results were reported for a combination of gastric (501 cases) and esophageal (250 cases) cancers—also showed a 3-fold increase in risk of those cancers associated with pickled food consumption (76). Although point estimates in our meta-analysis did not indicate strong associations, nutritional studies are highly subject to measurement errors that bias results toward null. Therefore, the unbiased relative risks might be substantially higher than those observed in individual studies. In particular, it has been argued (74) that some studies conducted in high-risk areas of China that have shown null results [e.g., the study by Tran and colleagues (73)] may suffer from misclassification.

We conducted multiple subgroup analyses to test the robustness of the results and to explore the reasons for heterogeneity. As discussed earlier, all subgroup analyses showed a statistically significant increased risk. Part of the heterogeneity was due to study design; studies adjusted for confounders showed smaller increases in risk. Except for studies from Japan, the results within each geographic region remained heterogeneous. For example, Chinese studies showed an I² index of 78%.

We studied the possibility of publication bias using funnel plots and Egger weighted regression method. Neither method provided substantial evidence for publication bias.

Another recent meta-analysis of this same subject included 14 Japanese and Korean studies has reported an overall 28% (95% CI, 6%–53%) increase in risk of gastric cancer associated with high intake of pickled vegetables (77). Although both studies have shown an increased risk, our study included a much larger number of studies, and included studies from several others parts of the world, most notably China. In our analysis, we observed higher relative risks in studies conducted in Korea and China than those conducted in Japan and other countries. This difference between regions may be due to the variant ingredients, processing, or consumption habits. For example, in Korea, pickled food (kimchi) is more popular and consumed in greater amount than other countries. Major types of pickled vegetables used in Japan are salted products, such as shinzuke-takana (with 3 or 4% salt) and furuzuke-takana (with approximately 10% salt), and nonsalted products, such as Sunki, which is prepared without salt but undergoes lactic acid fermentation (7). The average daily per capita consumption of various salt-fermented vegetables in Japan (37 g/d) is substantially lower than of kimchi used by Koreans (200–300 g/d; ref. 7). In some areas of China, pickled vegetables were eaten daily for 9 to 12 months a year and were an integral part of the diet in some families (5), traditionally prepared by keeping tightly packed moist vegetables in a jar packing for a few weeks or months, allowing fermentation and growth of fungi and yeasts (5, 7). On the other hand, in the United States and northern Europe, vinegar is the most commonly used preservative in the manufacture of cucumber pickles (7).

Pickled food intake could increase gastric cancer risk due to their nitrate and nitrite content, which can react to form gastric carcinogens, N-nitroso compounds (78). Other factors may also increase gastric cancer risk. For example, extra salt is consumed with intake of pickled food, and high salt intake may increase the risk of gastric cancer (71). Moreover, increased intake of pickled food may reduce consumption of fresh vegetables, which may again lead to increased risk of gastric cancer.

This meta-analysis has several strengths. We searched both English and Chinese literature and included a large number of studies. We also conducted several preplanned subgroup analyses. However, because all studies included in the meta-analysis had observational design, the results may be confounded or distorted by biases affecting observational studies.

In summary, our results suggest a potential 50% higher risk of gastric cancer associated with intake of pickled vegetables/foods and perhaps stronger associations in China and Korea. Results should be interpreted considering limitations of observational studies and in the light of analysis of dose–response effect, laboratory and animal studies, and perhaps human studies of other gastrointestinal tract organs.

No potential conflicts of interest were disclosed.

Conception and design: F. Kamangar, F. Islami

Development of methodology: F. Kamangar, D. Forman, F. Islami

Acquisition of data: J.S. Ren, F. Kamangar, F. Islami

Analysis and interpretation of data: J.S. Ren, F. Islami

Writing, review, and/or revision of the manuscript: J.S. Ren, F. Kamangar, D. Forman, F. Islami

Administrative, technical, or material support: F. Islami

Study supervision: F. Islami

J.-S. Ren was awarded a Postdoctoral Fellowship by the International Agency for Research on Cancer while conducting main part of this work.

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.