Environmental Factors in Helicobacter pylori-Related Gastric Precancerous Lesions in Venezuela

Stomach cancer is the second most common cancer in the world (1) and most common in developing countries. Although its incidence and mortality are decreasing in most countries, ∼800,000 people in the world are estimated to be diagnosed with this cancer every year (1).

A significant progress in the etiology of stomach cancer during the past couple of decades is the discovery of Helicobacter pylori (2). In 1994, an IARC Working Group conducted a systematic review and concluded that there was sufficient evidence in humans for the carcinogenicity of infection with HP. However, there has been heterogeneity among populations concerning the risk of stomach cancer associated with this infection (3). Furthermore, in populations with high prevalence of HP infection, a very small fraction of HP-infected individuals is actually found to develop stomach cancer. These suggest the role of additional factors in modifying HP-related gastric carcinogenesis.

Histopathological studies of the gastric mucosa in high-risk populations have revealed a series of lesions, which apparently represent a continuum of changes from normal to carcinoma, the complete process taking at least two decades (4). This includes, in order of increasing severity, superficial gastritis (SG), chronic gastritis (CG), chronic atrophic gastritis (AG), intestinal metaplasia (IM), and dysplasia. HP has been shown to induce acute gastritis, which can progress to CG, AG, and IM (2, 5). However, the role of HP in later stage of these precursor lesions has been less clear because the density of HP in gastric mucosa decreases with advancing stage of the lesions. Thus, it is quite plausible that other environmental factors play an important role in the progression of these precursor lesions. To date, it remains to be elucidated whether and how dietary and other life-style risk factors for stomach cancer are associated with the development of gastric precursor lesions. To address these issues, we selected potential environmental risk factors based on the results of our recent case-control study on stomach cancer in Venezuela, in which a diet high in starch and low in meat, fish, and fresh vegetables, cigarette smoking, and alcohol drinking were associated with risk of gastric cancer along with low socioeconomic status indicators and family history of the disease (6). In the present study, those selected factors are examined in relation to the risk of various stages of premalignant lesions among the baseline population in a chemoprevention trial for gastric precancerous lesions in Venezuela. Because this population is known to carry an extremely high prevalence of HP infection (≈95%; 7), the major objective of the study is to identify additional environmental factors to HP infection rather than to assess interactions between HP infection and other environmental factors.

Details concerning the study design and characteristics of study participants have been published elsewhere (8). Briefly, the original study population consisted of 2200 participants in a randomized placebo-controlled trial for gastric precancerous lesions using β-carotene, α-tocopherol, and ascorbic acid. They were recruited from participants in the gastric cancer screening program of the state of Tachira, Venezuela, who were aged between 35 and 69 in general good health. At baseline, all participants underwent gastroscopic examination with biopsies, collection of blood and urine specimens, and completion of questionnaires, which were repeated at subsequent annual visits up to 3 years. Written informed consent was obtained at enrollment.

The questionnaires were administered by specially trained interviewers to obtain information concerning demographic and household characteristics, history of alcohol intake, current tobacco consumption, and consumption of 41 typical local food items recorded as numbers of servings/week. Because of the limited number of food items included in the questionnaire, nutrient intake was not estimated. Plasma specimens were assayed to determine blood micronutrient levels as described elsewhere (9).

At gastroscopic examination, five biopsies for histological assessment were taken from the lesser curvature of the antrum ∼1 cm from the pylorus, from the greater curvature of the antrum ∼1 cm above pylorus, from a midportion of lesser curvature of the antrum, from the lesser curvature of the antrum immediately below the incisura, and from the middle corpus ∼2 cm from lesser curvature. They were fixed in buffered formalin and stained with H&E and with Giemsa stain to define HP infection. Those positive for IM were also stained with periodic acid-Schiff/Alcian blue and high iron diamine/Alcian blue to determine subtypes. Subjects' slides were assigned to one of the three pathologists at the pathology laboratory of the Cancer Control Center for their review. These pathologists participated in training sessions with consultant pathologists expert in gastric pathology (Drs. P. Correa and H. Ramirez), and advanced lesions (dysplasia in particular) by Dr. Ramirez. For each biopsy, the following variables were recorded in a standard form: type and depth of mucosa biopsied, degrees of neutrophil and monocyte infiltration, active regeneration, glandular atrophy, IM and dysplasia, depth of monocyte infiltration, type of IM, and quantity of HP infection (none, difficult to find, easy to find, and abundant). Most of these lesions (except for depth and type) were graded as none, mild, moderate, or severe and scored as 0–3, respectively. The HP, monocyte, and neutrophil infiltration scores were calculated as the mean of all evaluable biopsies rated 0–3. The fundic mucosa was confirmed by the presence of oxynitic cells.

Histological diagnoses were classified as follows: SG, characterized by infiltration of lymphocytes, plasmocytes, and polymorphonuclear leukocytes in the superficial portion of the lamina propria and by focal necrosis and regenerative changes in the glandular neck region; CG, lymphoplasmocytic cells infiltrate deeper layers of the mucosa; chronic AG, characterized by loss of normal glands, which are partly replaced by connective tissue and inflammatory cells; IM: gastric glands are replaced by cells normally present only in the intestine: absorptive cells, goblet cells, argentaffin cells, and Paneth's cells; IM subtypes were classified according to Filipe and Jass (10); dysplasia: characterized by closely packed tubular glands with diminished mucus secretion, scant cytoplasm, and large, hyperchromatic, crowded, and elongated nuclei. The most advanced among multiple biopsies at entry was used as the global diagnosis for a given subject. A global diagnosis “normal” was applied when none of the above-listed lesions was detected in any biopsy.

One subject was removed from this analysis because no evaluable biopsy specimens were obtained. As a result, 2199 subjects remained in the study. Data from the baseline examination were used for the analysis, except for one subject with missing histology information at baseline who was assigned to the placebo group. For this subject, data from the second visit replaced the baseline data. Some missing values at baseline were also reconstructed by the data from the subsequent visits when appropriate.

The response variable in this study was global histological diagnosis, which was divided into four groups: dysplasia, IM, AG, and normal gastritis to CG. The last group served as the control group in this study because the combined frequency of normal gastritis and SG in this population was still <5%.

To reduce the number of multiple comparisons, we limited explanatory variables to those associated with stomach cancer risk in a case-control study in the same area of Venezuela (6). Some of those variables (e.g., nutrient intake) were not available in this study. Instead, we added four plasma nutrient variables that were closely associated with the selected food groups. Individual food items in the questionnaire were integrated into the following six food groups: (a) starchy vegetable (potatoes, yam/yuca/manioc, ripe and green plantains, breakfast potato soup, and other soups with vegetables that typically contain aforementioned types of vegetables), (b) other vegetables (celery, squash, fresh salad, avocado, and other cooked vegetables), (c) cereals/grains (tortilla/bread, rice/pasta, and grains/beans), (d) meat/fish (fresh and salted meat and poultry, processed meat, and salted/canned fish), (e) dairy (cheese and milk), and (f) fruit (any fresh fruit and fruit juice). Food/nutrients variables were grouped into quintile levels, while a fewer categories were used for other variables with low frequencies of exposure groups, such as smoking and alcohol intake.

Multinominal logistic regression analysis was employed using SAS CATMOD procedure (11) to estimate the odds ratios (OR) and 95% confidence intervals (CI) for AG, IM, and dysplasia, in comparison with controls, according to household characteristics, smoking and drinking history, and weekly intake of six food groups and levels of plasma cholesterol, β-carotene, α-tocopherol, and ascorbic acid. Basic demographic covariates (i.e., sex, age, and educational levels) were included in all models. First, household characteristics were analyzed, and those showing a significant association at the 10% level were included in the subsequent analyses for smoking and alcohol consumption. Finally, in food and plasma nutrient analysis, smoking and alcohol variables were also considered if they resulted in a significant association at least 10% level. Tests for linear trend in the logit of risk associated with ordinal categorical variables were performed using equally spaced scores to the categories. The results of global tests for the effect of each explanatory variable on all outcome levels were also obtained in the same models based on Wald χ² statistics.

To explore possible interactions between HP and other environmental factors among the control group, Spearman and Pearson correlations were calculated between histological parameters (HP, monocyte, and neutrophil infiltration scores) and selected environmental factors. In addition, multiple linear regression models were fitted with the monocyte or neutrophil infiltration score as a dependent variable stratified by the HP score (<2 versus ≥2).

The study subjects consisted of 157 with dysplasia, 551 with IM (403 type 1, 75 type 2, and 73 type 3), 337 with AG, and 1154 controls. The HP positive rates in biopsy specimens were 95% for controls, 94% for AG, 95% for IM, and 87% for dysplasia, and the corresponding mean HP scores were 1.60, 1.63, 1.55, and 1.27, respectively.

Advanced age was a strong predictor for all lesions studied, with a stronger association for advanced lesions. Levels of education were inversely associated with the risk of both IM and dysplasia (Table 1). Overall, crowded housing conditions as measured by the number of family members/bed (Table 1) or room (data not shown) were not associated with the prevalence of premalignant lesions studied, although there were limited positive association between the number of family members/bed and the risk of AG and IM. The number of years of refrigerator use was significantly inversely associated with risk of AG (P = 0.02) and marginally significantly with dysplasia (P = 0.07), while no association with IM. Family history of gastric cancer tended to increase the risk of dysplasia, but it was rather inversely associated with the risk of AG (not statistically significant).

Adjusted for demographic and household characteristics, we found smoking to be a significant predictor for the prevalence of premalignant lesions studied (P < 0.001). Compared with never smokers, ever smokers (current + past) had significantly increased risks of dysplasia (OR 2.14; 95% CI 1.48–3.08) and IM (OR 1.33; 95% CI 1.06–1.68) but not of AG (OR 0.89; 95% CI 0.67–1.19). A positive linear trend with the number of cigarettes smoked/day at the time of the examination was also evident for both dysplasia (P < 0.001) and IM (P = 0.005). Among the subjects smoking at least 10 cigarettes/day, the OR for dysplasia was 3.58 (95% CI 2.12–6.04) and that for IM was 1.77 (95% CI 1.22–2.55) compared with never smokers. When alcoholic beverage consumption was analyzed by level of ethanol intake (ml/day), there were no associations with any of the lesions studied (Table 2).

Table 3 presents the results for the six food groups of interest. The global tests indicated a strong positive association with starchy vegetable and a strong inverse association with fruit consumption. With increasing fruit and fruit juice consumption, the ORs declined for all lesions studied. The linear trend was more pronounced for AG (P = 0.005) and dysplasia (P = 0.010) than for IM. The subjects who took at least 14 servings of fruit/week had a 60% decrease in risk of dysplasia, 15% in risk of IM, and 40% in risk of AG compared with those who took <7 servings of fruit/week. On the contrary, the risk of any type of the lesions increased with increasing starchy vegetable intake. The trend was more pronounced for AG (P = 0.03) and IM (P < 0.001). The OR for at least 20 servings of starchy vegetables/week, compared with <10 servings/week, was 1.52 (95% CI 0.91–2.55) for dysplasia, 1.60 (95% CI 1.15–2.24) for IM, and 1.64 (95% CI 1.12–2.42) for AG. The global tests for the effects of dairy product and grain/cereal consumption were also marginally significant. However, the positive association with dairy consumption was only statistically significant for IM and an inverse association with grain/cereal consumption was limited to AG. No associations were observed with meat/fish or other types of vegetable.

When two-way interactions among age, education, smoking status, and fruit and starchy vegetable intakes were assessed, the only significant interaction was found between fruit consumption and education (P = 0.028), while P values for the others were >0.10. When the association of fruit intake was further analyzed by educational level, the inverse association with AG was most pronounced at the lowest educational level (0–5 years, P = 0.011) followed by the middle level (6–8 years, P < 0.108) and the highest level (≥9 years, P = 0.399). On the other hand, the association with dysplasia was most evident at the highest educational level (P = 0.038) followed by the middle level (P = 0.047) but not at the lowest level (P = 0.845).

The results from plasma nutrient analysis were presented in Table 4. Plasma cholesterol concentrations appeared to be inversely associated with the lesions studied, with a stronger association with AG (P = 0.02). Although the global test indicated a significant association between plasma ascorbic acid levels and the lesions studied, the direction of the associations was inconsistent: positive for AG and dysplasia and inverse for IM. There were no overall associations with plasma β-carotene levels, but it was positively associated with the prevalence of AG. Plasma α-tocopherol did not show associations with any of the lesions studied. Plasma cholesterol levels were positively correlated with plasma β-carotene (r = 0.28) and α-tocopherol (r = 0.61). When plasma cholesterol was simultaneously included in the models for these two lipid-soluble vitamins, both inverse and positive associations of AG with plasma cholesterol, β-carotene, or α-tocopherol were highlighted but with minimal changes for the other advanced lesions. The P values for linear trends for cholesterol and β-carotene were 0.003 and 0.007, respectively, and those for cholesterol and α-tocopherol were <0.001 and 0.008, respectively.

Among the controls, the three histological parameters (HP, monocyte, and neutrophil scores) were closely correlated each other (r = 0.54–0.65), but there were little correlations between any of the histological parameters and fruit intake, starchy vegetable intake, or smoking status (r = −0.06 to 0.04). When the data were stratified by the HP score at the approximate median point (<2 versus ≥2) in linear regression analyses for monocyte and neutrophil scores, which also included age, length of education, duration of refrigerator use, and family history of stomach cancer as covariates, the regression coefficients for fruit, starchy vegetable, and smoking between these two groups did not differ significantly, suggesting no interactions.

The stomach is one of the organs for which a series of premalignant lesions representing a continuum from normal to carcinoma has been well documented. The complete process has been considered to take more than two decades (4), which provides us with opportunities of interfering with this process. Eradication of HP would be sufficient for populations with low prevalence of this infection, but recurrence or reinfection following successful antibiotic treatment is common in developing counties (12). Furthermore, it is not unusual that HP no longer colonizes in the stomach of the individuals who have developed an advanced stage of premalignant lesions, such as dysplasia (7). Thus, it is critical to identify other environmental factors that could modify sequelae of this infection. In addition, it is of epidemiological interest to examine how such factors differ by stage of the lesions because distinctive patterns of gene expression have been recognized for individual lesions (13).

Several studies on stomach cancer including those from the pre-HP era have revealed that stomach cancer risk is lower in people who eat more fruits and vegetables (14). Alternatively, it has been suggested that diets high in starchy foods or in salty foods possibly increase the risk. A moderate positive association between stomach cancer and smoking or tobacco use has also been observed (15). The association with alcohol drinking has been more equivocal (14). Among micronutrients, vitamin C has produced the most consistent inverse association with stomach cancer risk (14). Factors associated with risk of gastric premalignant lesions, other than HP, have less extensively studied. Most of those studies focused on single types of premalignant lesions and the definitions of the precursor lesions and comparison groups varied with the studies. Nevertheless, some similarities have been observed between factors associated with stomach cancer and those with various stages of premalignant lesions.

It was not surprising that gender was not a predictor of any lesions in this study, despite male predominance in gastric cancer incidence in most populations (1). Our study population was not a random sample of the general population in the study area of Venezuela, but they voluntarily participated in stomach cancer screening. We reported previously that participants in stomach cancer screening had personal history of peptic ulcer and family history of stomach cancer more often than nonparticipants (16), which may have led to similar risk profiles among males and females in this study population. Underlying reasons for the protective effects of education are probably complex. Although the associations were weakened with simultaneous inclusion of other variables, such as duration of refrigerator use and intakes of fruit and starch vegetables that were discussed below in more detail, they remained highly statistically significant. Therefore, other factors not included in this analysis, such as personal/family hygiene habits and childhood nutritional status, may have interfered with the progression of the lesions.

There was a marginally significant inverse association between the duration of refrigerator use and the prevalence of gastric premalignant lesions in the present study. The association may be partly due to confounding from socioeconomic status because adjustment for educational levels considerably weakened the overall association. However, this protective effect has been observed in earlier studies for stomach cancer and considered to be causal indirectly through changes in dietary habits and in risk of contamination of foods with bacteria and fungi (14). In this population, an increase in fresh fruit/fruit juice consumption may be a primary factor, which accounts for this association with refrigerator use because the inclusion of fresh fruits/fruit juice consumption in the model simultaneously significantly reduced the overall association (P = 0.25).

The present study suggests the impact of cigarette smoking on the development of advanced stages of premalignant lesions, which is consistent with the results of a recent meta-analysis on tobacco smoking and gastric cancer (15). In earlier studies, the prevalence of IM or dysplasia and the risk of progression to dysplasia were positively associated with smoking status or the amount of cigarettes smoked (17–23), while others reported that the prevalence of AG increased with cigarette smoking (21, 22, 24, 25). The latter association of AG found by others may also be explained by coexistence of more advanced lesions (IM and dysplasia). Tobacco contains known gastric carcinogens in experimental animals, which are swallowed directly into the stomach by smokers (26). Smoking-related DNA adducts have been detected in gastric mucosa of smokers in greater concentrations than in that of nonsmokers (27). Smoking is known to increase bile reflux (28). Increased bile acid concentrations in gastric juice has been demonstrated in smokers compared with nonsmokers and it has been associated with the risk of AG and IM (22). In addition, smoking has been reported to increase mRNA expression of certain chemokines in gastric mucosa, which may augment inflammatory reactions (29). On the other hand, reduced gastric defense mechanisms [i.e., lower vitamin C (30) and mucus (31) contents in gastric juice] among smokers may contribute to the progression of precursor lesions. Thus, it is biologically plausible that tobacco products play a causal role in gastric carcinogenesis process.

The second significant finding is the positive association with starchy vegetables, which tended to be more pronounced for less advanced lesions. Studies have suggested that diets very high in starchy foods may increase the risk of stomach cancer, but evidence is not convincing enough without plausible biological mechanisms. It has been judged that the association may be indirect via deficiency in protective dietary elements in such monotonous diets (14). However, our observation does not seem to be ascribed to high starch intake because there was no positive association with consumption of grain/cereals, which was the major source of carbohydrate/starch. Starchy vegetables are also an important source of dietary fiber. Experimental studies have shown that dietary fiber intake modifies mucosal and systemic immune functions (32–34). High fiber diets have also been reported to increase mucosal cell apoptosis in rats treated with a carcinogen (35). Therefore, it would be possible that certain types of dietary fibers abundant in starchy vegetables could modify the course of HP infection, although some epidemiological studies have found an inverse association between total dietary fiber intake and stomach cancer risk (14). Another possible interpretation is that other aspects of diets high in starchy vegetables actually account for the increased risk. Such factors are likely to reside in traditional dietary, cooking, or sanitary habits as these vegetables are more commonly used in local traditional styles of dishes.

Higher intake of milk or dairy products have often been found among stomach cancer patients than among controls (6, 36) and it has been interpreted that the association is more likely to be a consequence of the disease (i.e., that patients took these foods to mitigate their symptoms). Although gastric precancerous lesions are generally less symptomatic, we found a marginal positive association with dairy consumption and others have reported a significant positive association between milk consumption and risk of chronic AG (25). Hamajima et al. (37) raise a possibility that milk consumption increases the prevalence of HP infection among individuals with specific genetic background, although no biological mechanisms have been provided. In addition, it is worth noting that HP has been detected in milk from cow (38) and sheep (39), which suggests possible transmission through these products.

The dietary factor found to be protective in the present study is fresh fruit consumption. This has been observed in earlier studies on IM with or without dysplasia (18, 40) as well as on HP infection in children (41) and in studies on stomach cancer (14). These associations have been primarily attributed to vitamin C and β-carotene intake (14, 41). In a rural population in China, the subjects with IM with or without dysplasia had lower serum vitamin C and β-carotene levels than those with lower stages of precancerous lesions (42). In the same population, the risk of progression to dysplasia was inversely associated with baseline serum vitamin C levels but not with β-carotene levels (23). A small randomized clinical trial in Italy reported that vitamin C supplementation after HP eradication promoted the resolution of IM (43). In a randomized placebo-controlled trial in Colombia, both vitamin C and β-carotene treatments resulted in greater probabilities of regression of IM and AG without any antibiotic treatments (44). Both nutrients have been postulated to exert their protective effects via antioxidant properties because HP infection induces the production of reactive oxygen and nitrogen species, which can cause DNA damage as well as lipid and protein molecule modifications (45–47). In addition, vitamin C has been hypothesized to reduce the formation of carcinogenic nitroso compounds in the stomach and found to possess a specific growth inhibitory effect on HP (48). Furthermore, it has been documented that the concentrations of both vitamin C and β-carotene in gastric juice are significantly reduced among patients with HP-related gastric pathology (49, 50). However, our results on plasma micronutrient analysis do not generally corroborate these observations, except for the marginal inverse association between plasma ascorbic acid and the prevalence of IM. One possible confounding factor in our population is intake of certain types of starchy vegetables (e.g., plantains and yam), which contains high levels of ascorbic acid and β-carotene and was strongly positively associated with the risk of these precursor lesions.

The inverse association between plasma cholesterol levels and severe AG has also been reported in an Italian population based on plasma pepsinogen measurements (51). This is in agreement with our observation on plasma cholesterol. These observations may be pertinent to an experimental evidence demonstrating that gastric mucin is capable of binding lipids, which in turn prevents the mucins from attack by oxygen radicals (52). Instead, lipid absorption may have been inhibited because of other high-risk dietary constituents in this study (i.e., dietary fiber). Lastly, plasma cholesterol levels may serve as a proxy to animal product intake, which was inversely associated with stomach cancer risk in our earlier study in Venezuela (6).

There are several limitations in this study. First, misclassification of histological diagnoses may be a concern. There are two sources of misclassification (i.e., sampling error and interreader variability), which have been assessed in our previous reliability study (53). The stomach is a large organ and biopsy samples just a small amount of the tissue. Thus, false-negative diagnosis is unavoidable to some extent particularly for advanced lesions (e.g., IM and dysplasia), which tend to be focal and small. On the other hand, the interreader agreement is very high for the advanced lesions (κ = 0.96 for IM and 1.00 for dysplasia), while it is moderate for less advanced lesions (κ = 0.4–0.6). Net misclassification depends on the joint effects of these two factors, but our reliability study indicates that the effect of the former (sampling error) would be larger. These misclassifications are likely to result in an underestimation of ORs.

Second, our study was not able to address interactions between HP infection and other environmental factors directly, not only because few subjects were HP negative but also because the degree of HP infection in the stomach decreases with advancing premalignant lesions. Third, whereas premalignant lesions studied here are generally asymptomatic, those who had unspecific gastric symptoms are more likely to have participated in this screening program. Such symptoms may have influenced their dietary habits. Fourth, there is significant likelihood that some of the associations observed in this study were chance findings due to multiple comparisons, although we carefully selected explanatory variables. Finally, a major limitation with the cross-sectional study design is difficulty in inferring causal relationship for the observed associations because temporal relations between exposures and outcomes are not clear. Thus, we believe that conclusions should be made in consideration of coherence to other epidemiological and laboratory data.

Despite limitations discussed above, the present study provides important implications in preventive strategies against stomach cancer in high-risk populations. Specifically, our findings suggest that even among people who have already infected with HP or developed certain premalignant lesions in the stomach, cessation of cigarette smoking and an increase in fresh fruit consumption may help slow progression to stomach cancer.