Helicobacter pylori prevalences and risk factors among school beginners in a German urban center and its rural county.

In 1998, the Helicobacter pylori [(13)C]urea breath test was offered to all school beginners (birth cohort 1991/1992) in the city of Leipzig and in Leipzig County, Germany, to determine the colonization prevalence and potential transmission pathways of the bacterium. A total of 3,347 school beginners participated in the test, and 2,888 parents completed the detailed, self-administered questionnaire. The H. pylori prevalence was 6.5% [95% confidence interval (CI), 5.3-7.6] in the city and 5.7% (CI, 4.2-7.0) in the county. Using cluster analysis (WARD's method, Euclidean distances), we identified different sets of variables (confirmed by multivariate logistic regression analyses [odds ratios (ORs)] that are signficantly associated with H. pylori positivity. Among city children, the risk is significantly increased with contact to a pet hamster (OR = 2.4; 95% CI, 1.2-4.7; p < 0.015) and travels to Asian countries (OR = 3.7; 95% CI, 1.6-8.7; p < 0.002). Among county children, H. pylori positivity increased significantly with drinking of water from nonmunicipal sources (OR = 16.4; 95% CI, 3.1-88.5; p < 0.001), more than 3 children living in a household (OR = 4.2; 95% CI, 1.2-14.6; p < 0.02), and contact with pet hamsters (OR = 2.4; 95% CI, 1.0-5.7; p < 0.04). These data suggest that, in a general population sample, indirect fecal-oral transmission and living conditions are important risk factors in the spread of H. pylori infection. However, clinical symptoms do not necessarily indicate H. pylori positivity.

Helicobacter pylori is one of the most common pathogens. Since the observations of Warren and Marshall in 1983Marshall in /1984, its significance as a causal agent for peptic ulcers and public health problems is evident. Numerous papers have been published discussing potential risk factors of transmission and infection.
In a recent review of H. pylori epidemiologic studies, the European Helicobacter pylori Study Group (2) concluded that "The route of transmission still remains unclear." The authors observed problems in the interpretation of some past studies in relation to study design, selection and size of the study population, and the statistical methods applied. The recommendation was that future epidemiologic studies take these aspects into consideration.
The large number of cases and the high cost of medical treatment have now turned the focus of epidemiologic research toward certain risk populations. Preschool children are thought to be the main risk population (3). Generally accepted risk factors for the aquisition of H. pylori infection are socioeconomic status, childhood living conditions (4), and large family size (5). In a recent German study, family history of gastric disease correlated with aquisition of the bacterium (6), whereas pets have not been considered a risk factor (7). As far as the routes of transmission are concerned, the results have been contradictory (2). There is no doubt that the bacterium gains entry to the stomach via the mouth. Oral-oral, fecal-oral, and iatrogenic spread, as well as several vectors such as domestic cats, nonhuman primates, and the domestic housefly, have all been discussed as possible routes of transmission; however, a major route has not been identified (8).
Because this organism appears to be ubiquitous and may be picked up anywhere, we hypothesized that indirect fecal-oral transmission could possibly be a dominant pathway in its acquisition. Thus, we conducted this study to determine potential sources of H. pylori in the environment of young children. The data are based on a detailed parent-completed questionnaire designed to explore a wide range of possible risk factors and an H. pylori colonization screening test administered to the entire 1998 school entry population in Leipzig, Germany, as well as its rural region, the County of Leipzig.

Materials and Methods
The Leipzig Helicobacter pylori study was a joint project of the UFZ (Centre for Environmental Research Leipzig-Halle, GmbH), the University Children's Hospital Leipzig, and the Departments of Public Health, City and County of Leipzig, conducted in the fall and winter of 1997/1998. The investigation included all children eligible to enter grade one in the fall of 1998. We used the epidemiologic questionnaire to elicit information on education, employment, and medical history of the proband's parents and siblings, past and present home address, living conditions, water supply, leisure-time garden plots, and a detailed medical history of the child. Information was sought on the family's nationality and travels to foreign countries. Further questions focused on the child's past and present contacts to various pets and different sources of drinking water. Further details regarding the study's methodology have been described elsewhere (9). The criterion for a child's participation in the study was the parents' written, informed consent. Only children with a consent form signed by at least one parent could take the H. pylori breath test. A total of 3,347 of the 3,919 eligible school beginners undergoing the medical examination participated in the voluntary H. pylori [ 13 C]urea breath test (response rate 85.4%) (9). Mean age of the children was 6.1 years (range 5.2-7.0 years).
The life-time prevalences and potential risk factors in the transmission of the infection, and thus in association with an H. pylori positive test, were analyzed using STA-TISTICA for Windows (10). Because more children participated in the H. pylori test than parents completed questionnaires, this assessment is based on the data of only those children with an H. pylori test and a completed questionnaire (responses to all variables under investigation). Thus, this analysis includes the data of 1,844 (73.8%) of the 2,498 city children and 1,044 (73.5%) of the 1,421 county children.
Initially, we selected for the analysis 47 variables assumed to be possible risk factors for acquiring the H. pylori bacterium. Thirty of these were tentative sources of infection, such as domestic crowding, water and food, pets, hygienic and sanitary living conditions of residence, weekend cottage, travels to foreign countries. The other 17 variables were clinically potential diagnostic indicators. To identify the most important variables in terms of risk associated with the outcome variable (H. pylori positivity), we performed a cluster analysis [Ward's method (11), Euclidean distances]. The cluster analysis was carried out using a subset of all cases to avoid overestimation of the cluster variables entering the multivariate logistic regression. We used the Ward's method to derive small clusters. This process was repeated with differing case data sets and continued until stable clusters were obtained. All variables identified in this way were entered into the logistic regression. We began the multivariate logistic regression calculations with an initial model. The model was varied by adding and eliminating variables until we reached a stable model (assessment based on p-value). We calculated all stepwise logistic regressions with identical, complete sets of variables in order to be able to compare the results. To avoid an overestimation in the regression analysis, the cluster analysis was carried out on successive, arbitrary 75% subsamples of the available population data. The model was defined as stable when the calculated odds ratio value of the significant variables no longer changed during the process.
Authorities of the city of Leipzig granted permission for participation of preschool children in the study, and the study was approved by the Ethics Committee of the University of Leipzig.   Table 1. Acronyms are used in Figure 1.

Results
in the H. pylori survey. Nonresponders could not be analyzed in detail because parents could not be questioned about the reasons for not letting their children participate in the study. Our data are based on an approximately 73% subsample comprised of city and rural children who participated in the test and returned the parent-completed questionnaire with responses to all questions. Among this preschool population, a total of 179 children (119 city and 60 county) were found to be infected with H. pylori. The prevalences of H. pylori colonization are presented in Table 1.
The risk factors assumed to play a role in the transmission of H. pylori infection and their frequency of occurrence among these probands are summarized in Tables 2 and 3.
Results of the cluster analysis of the total population (city and county) are presented in Figure 1. The main cluster, the variables close to the target HP test (H. pylori positivity), includes the varibles with the least geometric distance from the target. These variables were entered into the multivariate logistic regression analysis to determine odds ratios and significance levels. When applied separately to the data sets (Leipzig city and Leipzig County), cluster analyses of the same 47 selected variables showed that the distance of some cluster variables to the target, H. pylori positivity, varied for the city and county, indicating differences in their importance in the association. The variable "contact with pet hamsters" was associated more closely with H. pylori positivity in the city children than in the county children. In contrast, the variables "use of human excrement as garden fertilizer" and the "drinking of water from nonmunicipal sources" appear closer linked to the target variable in the county. The odds ratios (Table  4) show that not all variables clustering close to the target are significant.
Although we included reported clinical symptoms in a child or a family history of recurrent abdominal pain, dyspepsia, gastric/peptic ulceration, etc. in the model to test their general and specific importance as risk factors, this analysis indicated that these host factors did not significantly contribute to the prediction of H. pylori positivity. For example, inclusion of the symptoms "recurrent abdominal pain" and "recurrent belching" into the regression model for the total population (Table 5) only weakened the association. The p-value of the model, in spite of remaining significant, increased from 0.0008 (without the clinical indicators) to 0.0028.

Discussion
Compared to similar studies (4-7), this investigation is, to our knowledge, the largest study of a homogeneous cohort population at risk for H. pylori (Table 1). Moreover, this study allowed the comparison of an almost complete birth cohort of urban and rural children (city and its surrounding county).    Table 2. The shorter the distance between a variable and the target HP test (positive H. pylori test), the closer the association and, thus, the more important as a risk factor.
Also, 95.8% of the children were white, which further indicates the homogeneity of the cohort. We found no significant difference in the prevalence of H. pylori colonization between city (6.5%) and county (5.7%) children (a total of 6.2%; Table 1). The prevalences are within the norm expected for a developed country, where an acquisition of the bacterium among children is estimated to be less than 1% per year (12). In a recent study of preschool children in a southern German city with a high proportion of foreigners, the total mean prevalence was 11.3%, with 4.9% among German children, and thus comparable to our findings (13).
The risk factor analysis showed some unexpected results. The analyses, cluster analysis, and odds ratio calculations indicated four significant risk factors suggestive of being routes for the acquisition of H. pylori bacteria (Table 4, Figure 1): • "Travels to Asian countries" as a possible source of infection; the prevalence of H. pylori is much higher in Asian countries than in Central Europe (14-16) • "Large family size," also a known risk factor (16) • "Contact with pet hamsters" • "Drinking water from nonmunicipal sources." The latter two are less known risk factors and require further discussion.
In a recent paper, Bode et al. (7) found no significant association with cats, guinea pigs, birds, and rabbits. Our results for these types of pets support their findings. However, the multivariate logistic regression analysis in our study population confirmed the cluster analy-sis, and indicated that "contact with pet hamsters" significantly increased the odds (OR = 2.1) of infection with H. pylori (Table  4). Furthermore, among rural children the risk of being infected with H. pylori increased with the "drinking of water from nonmunicipal sources" (OR = 16.4; Table 4). This may occur at allotment gardens, where well water may be used for human consumption. The result of the cluster variable analysis of "use of human excrement as fertilizer" (top-dressing) on vegetation in gardens was closely associated with the target H. pylori positivity among county children, even though the regression analysis showed a nonsignificant association. These findings point to the possibility that handling and/or eating vegetables (as well as grass and flowers), which may be contaminated with (human) feces, could be a route of acquiring the infection. H. pylori has been isolated from human feces (17) and has been found to be associated with the consumption of uncooked vegetables (18). Furthermore, consuming water from sources other than public supplies or drinking contaminated water is a risk factor in developing countries (19), and Carballo et al. (20) reported that untreated water is a possible source of H. pylori infection. Early results of an epidemiologic and microbiologic investigation now being carried out by our team suggest that, even in Central Europe, well water could be a source of H. pylori infection (21). The H. pylori bacterium appears to survive in water in a coccoid form, even in very cold water (22). All of these studies lend support to the hypothesis that indirect fecal-oral transmission may be an important pathway by which the bacterium gains entrance into the human body, with children being especially at risk.
The clinical predictors investigated did not seem to be significant in this population (Tables 3 and 5). No significant associations were found for "gastric/duodenal disease," "recurrent abdominal pain," or "recurrent heartburn" in children or in their families.
Measurement errors associated with the low within-and between-population occurrence ( Table 3) are balanced by the odds ratio analysis that is based on the total population (city and county) with 2,715 complete data sets. These odds ratios and confidence intervals suggest that the errors were minimized ( Table 5). The role of clinical symptoms as indicators of H. pylori infection remains controversial (23), but similar findings have been reported indicating that "chronic abdominal pain" in childhood (24,25) may not necessarily indicate H. pylori positivity.

Conclusion
In this study we demonstrated that H. pylori colonization in this preschool population is significantly related to risk factors such as "contact with pet hamsters," "travels to Asian countries," and "drinking of water from nonmunicipal sources." This suggests that indirect fecal-oral transmission may play a key role in the spread of this infection. These factors appear to be as important as household density, which was found to increase the risk of H. pylori infection when more than three children live in a household. However, relative to these risks, the clinical symptom indicators investigated do not seem to play the same significant role in predicting H. pylori colonization.
Although, at present, the recommendations of the European Society for Primary Care Gastroenterology do not support a "test and treat" strategy (26), our goal was to show that epidemiologic screenings of population-based, nonsymptomatic children offer information about current prevalences and risk factors, thus, contributing to future recommendations.