INTRODUCTION
It is estimated that half of the world's population is infected with Helicobacter pylori [Reference Dorer, Talarico and Salama1]. While the majority of those infected will remain asymptomatic, up to 20% of H. pylori infections induce gastric inflammation that can lead to severe chronic disease outcomes, including chronic gastritis, peptic ulcer, gastric mucosa-associated lymphoid tissue (MALT) lymphoma, and gastric cancer [Reference Marshall and Warren2–Reference Dooley5]. Epidemiological studies of H. pylori suggest the bacterium primarily spreads person-to-person via oral–oral and faecal–oral routes; however, conflicting results and knowledge gaps have precluded a clear understanding of pathogen transmission and the role of environmental factors [Reference Azevedo, Huntington and Goodman6]. Further insight into the pathogen's routes of transmission is essential to guide public health interventions aimed at controlling its spread and lessening its health burden [Reference Brown7].
Although prospective studies have not clearly identified causal aetiological associations between exposures and infection, observational studies have correlated key risk factors with H. pylori infections. Humans are most likely to acquire H. pylori in early childhood (before age 10 years) due to intrafamilial transmission [Reference Brown7, Reference Dominici8]. Studies have observed a higher prevalence in siblings (and parents) of infected children, and results have consistently supported infected siblings as a risk factor for H. pylori infection [Reference Azevedo, Huntington and Goodman6, Reference Brown7, Reference Ford and Axon9–Reference Calvet11]. Acquisition rates during childhood are markedly higher in developing countries [Reference Pounder and Ng12], and have been associated with living in impoverished regions with overcrowding and poor sanitation and hygiene [Reference Goh10, Reference Peleteiro13, Reference Bruce and Maaroos14]. Variable infection acquisition patterns have been observed between developed and developing countries, as well as within specific geographical areas by age, race, and socioeconomic groups, which may complicate public health risk assessments [Reference Pounder and Ng12, Reference Everhart15–Reference Graham17].
Evidence suggests that H. pylori has an environmental pathogen reservoir, in which water has been implicated as the most likely medium conducive to pathogen survival, particularly in rural areas and developing countries [Reference Azevedo, Huntington and Goodman6]. When out of its natural environment, H. pylori morphologically adapts into a coccoid form that is viable and infective, but non-culturable [Reference Bruce and Maaroos14, Reference Casasola-Rodriguez18]. The bacterium forms biofilms as a protective niche in water [Reference Percival and Thomas19–Reference Watson22] and can survive ozone and chlorination treatments [Reference Casasola-Rodriguez18, Reference Giao23, Reference Baker24]. H. pylori has been detected in shallow ground water [Reference Hegarty, Dowd and Baker25], fresh and marine surface water [Reference Adams, Bates and Oliver26–Reference Twing, Kirchman and Campbell28], untreated well water [Reference Baker and Hegarty29], and untreated municipal wastewater [Reference Moreno and Ferrus30, Reference Lu31] in the United States and Europe. The bacterium has also been detected in treated municipal drinking water in Iraq, Iran, and Pakistan [Reference Bahrami, Rahimi and Ghasemian Safaei32–Reference Al-Sulami, Al-Edani and Al-Abdula35]. Although cross-sectional studies examining correlations between human infections and drinking-water source have produced inconsistent results, many suggest that contaminated drinking water is associated with a higher prevalence of H. pylori infections [Reference Azevedo, Huntington and Goodman6, Reference Brown7, Reference Ford and Axon9–11, 14, 36–Reference Queiroz43]. In addition, the US Environmental Protection Agency (EPA) considers H. pylori a Contaminate Candidate for possible regulation under the Safe Drinking Water Act (SDWA) [44–46].
Because relatively few studies have considered environmental risk factors for H. pylori transmission in the United States, and because considerable research gaps prevent targeted interventions to mitigate public health risks, nationally representative population-based data was used to examine risk factors associated with H. pylori seroprevalence. The National Health and Nutrition Examination Survey (NHANES) [47], within the National Center for Health Statistics of the Centers for Disease Control and Prevention (CDC), has conducted large-scale surveys designed to assess changes over time in the health and nutritional status of adults and children in the United States. NHANES released H. pylori serology results in HANES III (1988–1991) and the Continuous NHANES (1999–2000). Analyses of HANES III revealed that demographic factors, including socioeconomic status and race/ethnicity were significantly associated with H. pylori seroprevalence in the United States [Reference Everhart15, Reference Zajacova, Dowd and Aiello48]. The Continuous NHANES (1999–2000) data has been used to identify correlations with H. pylori seroprevalence, including positive associations with smoking [Reference Cardenas and Graham49] and iron deficiency anaemia [Reference Cardenas50], an inverse association with childhood asthma [Reference Chen and Blaser51], and mixed results for biomarkers of type 2 diabetes [Reference Chen and Blaser52, Reference Malamug53].
Grad et al. compared trends in H. pylori seroprevalence between HANES III (1988–1991) and Continuous NHANES (1999–2000) in adults aged ⩾20 years [Reference Grad, Lipsitch and Aiello54]. The authors reported disparities in prevalence by race/ethnicity, in which seroprevalence ‘preferentially’ declined for the non-Hispanic White population, yet remained constant for non-Hispanic Blacks and Mexican Americans. However, the authors noted that indicators of socioeconomic status, including income–poverty ratio, country of birth origin, and educational attainment, while significantly associated with seropositivity, did not fully explain the observed disparities by race/ethnicity. They hypothesized that other factors not considered in their analyses were probably playing a role in H. pylori incidence and prevalence in US populations.
In this paper, the Continuous NHANES (1999–2000) data was further analysed to examine whether environmental factors were associated with H. pylori seroprevalence in the population as a whole and across stratified demographic groups by age and race/ethnicity.
METHODS
A single 2-year cycle of Continuous NHANES (1999–2000) data [47] was examined using SAS v. 9.3 (SAS Institute Inc., USA). Data were collected using a complex survey design; therefore, seroprevalence estimates were weighted to represent the total US civilian non-institutionalized population, and CDC's recommended approaches for SAS survey analysis procedures were followed [47]. Available sera from eligible participants aged ⩾3 years were tested by the CDC for immunoglobulin G (IgG) antibodies against H. pylori by an enzyme-linked immunoassay (ELISA). A dichotomous seropositivity cut-off was provided by the CDC, in which an optical density (OD) ⩾1·1 = positive and OD <1·1 = negative [55]. The authors selected covariates of interest a priori, based on previous reports and biological plausibility, to examine associations with H. pylori seropositivity.
Adulthood was defined using CDC's previously established cut-point of 20 years [Reference Johnson, Paulose-Ram and Ogden56]. The two NHANES race/ethnicity categories of ‘Mexican American’ and ‘other Hispanic’ were combined into a shared category of ‘Hispanic’, and the race/ethnicity group of ‘other, including multi-racial’ was excluded from the stratified analyses due to a limited sample size. An objective measurement for current smoking status was based on serum cotinine level (non-smoker ⩽15 ng/ml and smoker >15 ng/ml), as previously reported [Reference Cardenas and Graham49].
In the Housing Characteristics (HOQ) questionnaire, participants were asked: ‘What is the source of tap water in this home? Is it a private or public water company, a private or public well, or something else?’ and ‘Are any of the water treatment devices listed on this card used in your home?’ Listed treatment devices included Brita® or other pitcher water filter, ceramic or charcoal filter, water softener, aerator, and reverse osmosis. To evaluate crowded housing, a new variable was calculated as described previously [Reference Solari and Mare57], with the total number of people living in a household as the numerator and the number of rooms in the home (excluding bathrooms) as the denominator. As described previously [Reference Solari and Mare57], because the room count includes the kitchen, a value of 1 was subtracted from homes with more than one room, for a more accurate indicator of living spaces in a home.
Because soil could be tainted by H. pylori-contaminated surface and ground waters, occupational exposures to soil were assessed by examining occupation categories reported in the Occupation Questionnaire (OCQ). Participants aged ⩾16 years who reported working in the last week were asked in what occupation their current job was. Using CDC's Occupation Group Codes [58], the authors a priori selected occupations likely to involve soil exposures and created a three-tier categorical variable to classify unemployed participants, those with occupations not likely to involve soil exposures, and occupations with likely soil exposure, for adults aged ⩾20 years to include in the adult subgroup analysis. Soil-related occupations included farm operators, managers, and supervisors; farm and nursery workers; related agricultural, forestry, and fishing occupations; construction trades; extractive and precision production occupations (e.g. miners and oil-well drillers); construction labourers; and labourers not employed in construction.
Multivariable logistic regression was used to compare the seroprevalence of IgG antibodies against H. pylori with key demographic characteristics and potential risk factors. Wald χ 2 odds ratio (OR) estimates and 95% confidence intervals (CI) were ascertained for unadjusted and adjusted regression models. Covariates with P values <0·2 in a simple logistic model were considered for inclusion in multivariable models. Multivariable models were selected using stepwise backwards elimination of covariates with P values >0·05, with the highest corresponding P value removed each step. Three models were co-developed that included the NHANES population as a whole, as well as two stratified models by age group and race/ethnicity. Selected models included all covariates with P values <0·05, as well as a priori selected demographic characteristics known to be associated with the outcome and the exposure variables of interest. To allow for direct comparisons between the final models, covariates selected for inclusion in any of the final models were retained in all three models to provide direct comparisons across strata.
RESULTS
In one 2-year cycle of continuous survey data (1999–2000), NHANES released H. pylori serology results for 7493 participants aged ⩾3 years, of which 25·4% were seropositive (Table 1). Associations between demographic characteristics and seropositivity in all participants were similar to previously published descriptive reports that included all participants [Reference Cardenas50] and only adults [Reference Cardenas and Graham49, Reference Malamug53, Reference Grad, Lipsitch and Aiello54]. Increasing age was significantly associated with H. pylori seropositivity, but there was no difference in seropositivity between men and women. Compared to non-Hispanic Whites, Hispanics [adjusted OR (aOR) 3·5, 95% CI 2·6-4·8] and non-Hispanic Blacks (aOR 4·1, 95% CI 3·3-5·2) had higher odds of seropositivity. In addition, participants born outside the United States were significantly more likely to be seropositive compared to those born in the United States (aOR 2·7, 95% CI 2·1-3·4). Members of families earning <$ 20 000 a year were more likely to be seropositive compared to families earning ⩾$20 000 (aOR 1·4, 95% CI 1·1-1·7). Similar to a report that analysed smoking in adults only [Reference Cardenas and Graham49], smoking was significantly associated with seropositivity when considering all participants (aOR 1·8, 95% CI 1·5-2·2).
OR, Odds ratio; CI, confidence interval.
a IgG ELISA optical density (OD) ⩾1·1=positive; OD <1·1=negative. b All prevalence estimates were weighted. c Wald χ 2 logistic regression used. d Bold values denote statistically significant data (P<0·05). e Covariate has some missing data. f Based on serum cotinine level (non-smoker ⩽15 ng/ml, smoker >15 ng/ml). g Types of at-home water-treatment devices included: Brita® or other pitcher water filter, ceramic or charcoal filter, water softener, aerator, reverse osmosis.
An inverse association was observed with participants' self-reported general health condition: the worse the perceived health condition, the higher odds of seropositivity, with those reporting fair or poor health having higher odds of seropositivity compared to participants reporting being in excellent health (aOR 1·4, 95% CI 1·1-1·8) (Table 1). Participants living in a more crowded home with >1 person per room, were significantly more likely to be seropositive compared to participants with less crowded living conditions (aOR 1·7, 95% CI 1·3-2·2). Having a well as the source of home tap water compared to tap water provided by a private or public water company, was positively associated with seropositivity but did not reach the threshold for statistical significance (aOR 1·5, 95% CI 0·96-2·4). Similarly, participants who did not use at-home water-treatment devices had higher odds of seropositivity, but this association was not significant in the multivariable model.
Age group
Because the existing literature suggests that H. pylori infections are more often acquired in childhood, two multivariable models stratified by age group (3–19 and ⩾20 years) were examined (Table 2). For the younger age group, in addition to demographic characteristics including age, race/ethnicity, birth origin, and family income, H. pylori seropositivity was significantly associated with living in a more crowded home (aOR 2·3, 95% CI 1·5-3·7) and having a well as the source of home tap water (aOR 1·7, 95% CI 1·1-2·6).
aOR, Adjusted odds ratio; CI, confidence interval; GED, General education diploma.
a IgG ELISA optical density (OD) ⩾1.1=positive; OD <1.1=negative. b All prevalence estimates were weighted. c Wald χ 2 logistic regression used. d Bold values denote statistically significant data (P<0.05). e Covariate has some missing data. f Based on serum cotinine level (non-smoker ⩽15 ng/ml, smoker >15 ng/ml). g Types of at-home water-treatment devices included: Brita® or other pitcher water filter, ceramic or charcoal filter, water softener, aerator, reverse osmosis. h Soil-related occupations included: farm operators, managers, and supervisors; farm and nursery workers; related agricultural, forestry, and fishing occupations; construction trades; extractive and precision production occupations; construction labourers; labourers, except construction.
Conversely, in adult participants aged ⩾20 years, H. pylori infections were not associated with family income, crowded living conditions, or well-water use (Table 2). In addition to age, race/ethnicity, birth origin, smoking, and educational attainment, occupational exposure to soil was significantly associated with seropositivity. Compared to participants with occupations not likely to involve soil exposure, workers in soil-related occupations had significantly higher odds of seropositivity (aOR 1·9, 95% CI 1·2-2·9), as did unemployed participants (aOR 1·4, 95% CI 1·1-1·7). Because exposures to both well water and soil might increase an individual's cumulative risk of infection [Reference Rahim59], we repeated the regression model for adults and examined this combined exposure. Compared to all other NHANES participants (29·9% seropositive), 43·1% of 58 adult participants with both well water and occupational soil exposure were seropositive (aOR 2·7, 95% CI 1·3-5·6).
Race/ethnicity
Because H. pylori prevalence is also known to vary by race/ethnicity [Reference Brown7, Reference Grad, Lipsitch and Aiello54], the odds of seropositivity were examined after stratification by the three major groups (Hispanic, non-Hispanic White, non-Hispanic Black) (Table 3). Well-water use had a positive association with H. pylori prevalence across all race/ethnicity groups, but was only statistically significant in non-Hispanic Blacks (aOR 2·1, 95% CI 1·02-4·4). Furthermore, not using any at-home water-treatment devices was also significantly associated with seropositivity in non-Hispanic Blacks (aOR 1·7, 95% CI 1·02-2·9), while home water treatment had no association with seropositivity for the other two groups.
aOR, Adjusted odds ratio; CI, confidence interval.
a Race/ethnicity category of ‘Other, including multiracial’ was excluded due to small sample size. b IgG ELISA optical density (OD) ⩾1·1=positive; OD <1·1=negative. c All prevalence estimates were weighted. d Wald χ 2 logistic regression used for adjusted odds ratios. e Bold values denote statistically significant data (P < 0·05). f Covariate has some missing data. g Types of at-home water-treatment devices included: Brita® or other pitcher water filter, ceramic or charcoal filter, water softener, aerator, reverse osmosis. h Multivariable model for adults only also included educational attainment. i Soil-related occupations included: farm operators, managers, and supervisors; farm and nursery workers; related agricultural, forestry, and fishing occupations; construction trades; extractive and precision production occupations; construction labourers; labourers, except construction.
Conversely, crowded housing had larger and significant associations with seropositivity in non-Hispanic Whites and Hispanics (aOR 1·9, 95% CI 1·2-3·2 and aOR 1·7, 95% CI 1·1-2·7, respectively), while in non-Hispanic Blacks, crowded housing was positively associated but not statistically significant (aOR 1·4, 95% CI 0·8-2·3).
Associations with occupational soil exposures were also considered in a race/ethnicity stratified multivariable model for adults after also controlling for educational attainment (Table 3). Working in a job likely to involve exposure to soil was only significantly associated with elevated odds of H. pylori seropositivity in non-Hispanic Whites (aOR 2·6, 95% CI 1·6-4·1). For both non-Hispanic Blacks and Hispanics, the statistically insignificant aOR was estimated at 1·2. Of non-Hispanic Whites, unemployment was also significantly associated with increased odds of seropositivity (aOR 1·5, 95% CI 1·2-2·0) compared to workers in non-soil related occupations.
Birth origin
To examine whether there were differences in odds of seroprevalence by country of birth origin, multivariable models were also stratified by country of birth origin (US-born and foreign-born) (data not shown). Overall seroprevalence in foreign-born participants was over twice that of those born in the United States (54·0% vs. 20·5%); however, adjusted odds for potential risk factors were comparable across the groups.
DISCUSSION
Given the high burden of H. pylori, it is important to understand environmental correlates of infection. This examination of 2 years of NHANES data augmented the existing literature by using a nationally representative sample. Well water was positively associated with H. pylori seropositivity in all subjects, but was stronger and statistically significant in younger respondents aged 3–19 years. In addition to demographic characteristics associated with a lower socioeconomic status, living in a crowded home also increased these younger participants' odds of being seropositive. These results support the growing evidence for intrafamilial transmission in children in impoverished and crowded areas with the potential for accompanying poor hygienic conditions. The attenuated associations between tap-water source, crowded housing, and H. pylori seropositivity in the older NHANES population suggests that H. pylori infections associated with these exposures may be important when pathogen acquisition occurs during childhood.
By contrast, in NHANES participants aged ⩾20 years who were asked about occupational exposures, occupational exposures to soil were significantly associated with seropositivity. Exposures to both well water and occupationally related soil increased the effect size of adults' odds of seropositivity, which suggests a cumulative risk associated with exposure to both contaminated water and soil.
Using occupation codes to determine soil exposure limited interpretations because of the age restriction and requirement for employment in the prior week. The elevated odds of seropositivity observed in those not working in the week prior to the interview may be attributable to underlying differences in socioeconomic status or involvement in soil-related temporary or intermittent jobs not recorded by the OCQ. Nonetheless, the increased odds of seropositivity observed in those working in soil-related occupations suggests that transmission routes other than those associated with living conditions play a role in pathogen acquisition, including exposures to soil tainted by H. pylori-contaminated surface and ground waters.
Due to the health implications associated with H. pylori in public water systems, the US EPA has included H. pylori in all three of its Contaminate Candidate Lists, released in 1998, 2005, and 2009, for possible future regulation under the SDWA [44–46]. Recently, Ryan et al. evaluated potential drinking-water guidelines at the point of ingestion for H. pylori [Reference Ryan60]. They performed a quantitative microbial risk assessment (QMRA) to identify effective guidelines that would protect human health to an acceptable level of risk while also considering sources of uncertainty. Their analyses concluded that current levels of H. pylori in drinking water might pose a potential public health risk.
This analysis only suggested an association with not using at-home water-treatment devices and odds of H. pylori seropositivity in non-Hispanic Blacks. Due to the broad NHANES question that did not distinguish which specific treatments were used, it is difficult to evaluate the impact of specific home treatments on infection risk. Water softeners would be ineffective, and most water filters are designed to reduce taste, odours, or concentrations of chemicals such as lead and chlorine. The H. pylori bacterium is 3 μm long with a diameter of ~0·5 μm. Standard Brita® filters, including the tap (0·5 μm) and pitcher (1 μm) filters, may not effectively remove the bacterium, and other filters solely designed for taste and odour improvement would be ineffective. Moreover, the effectiveness of home treatments relies on proper maintenance and upkeep.
Two published studies detected H. pylori-specific DNA in soil collected from residential areas of Japan [Reference Sasaki61] and from public playgrounds in Spain [Reference Perez62], but the bacterium's viability was not assessed. Another study conducted in Japan found no evidence of H. pylori-specific DNA in soil collected from public parks [Reference Kawaguchi63]. Although evidence for H. pylori occurring in soil has been scarce, it was worth consideration in this analysis, since the bacterium can form biofilms and has been found in ground and surface water [Reference Hegarty, Dowd and Baker25–Reference Twing, Kirchman and Campbell28].
While differences in risk factor associations across the race/ethnic groups were observed, it is difficult to conclude how socioeconomic status and the built environment are disproportionately affecting these groups because the diminished sample sizes across the strata probably limited the statistical power necessary to detect significant associations. Although there were sample size limitations, our additional stratified analyses for all NHANES participants suggests that variance patterns may be attributable to disparities in socioeconomic status and living conditions, including overcrowding and drinking well water. In addition, the survey's general health marker, which had not been considered in previous publications, had an inverse correlation with seroprevalence. Poorer perceived health was associated with higher odds of seropositivity. Still, little is known regarding the variability in H. pylori transmission and lifetime seropositivity rates in different demographic groups in the United States.
NHANES collects a variety of health and nutritional information from a nationally representative sample of the United States population, which makes it a valuable tool for hypothesis generation and exploring correlations between exposures and health outcomes; however, this analysis has some limitations. The cross-sectional design prevents us from inferring causality. In addition, the risk factor analyses is limited to what was asked of participants during the 1999–2000 NHANES. Several questionnaires were administered only to subgroups (e.g. adults aged ⩾20 years), including the occupational questions. Moreover, the survey did not collect data regarding participants' behavioural health risk factors, a crucial component when studying H. pylori transmission. Because the bacterium often spreads via faecal–oral transmission, this analysis would have benefited from consideration for participants' hygienic behaviours, including hand hygiene habits and food preparation practices. Infrequent hand-washing has been associated with H. pylori infections, [Reference Lee64] and raw food contaminated by H. pylori-infected human faeces or cleaned with H. pylori-contaminated water may be a source of faecal–oral or environmental transmission of H. pylori [Reference Atapoor, Safarpoor Dehkordi and Rahimi65]. An inherent limitation in serological tests is their inability to distinguish between current and prior infections. If an H. pylori infection is cleared, circulating anti-H. pylori IgG antibodies also decline, but complete seroreversion can take decades [Reference Graham, Qureshi, Mobley, Mendz and Hazell66]. While positive IgG serology results acceptably correlate with an active H. pylori infection [Reference Blecker67], an undetermined proportion of the seroprevalence may be attributed to previously cleared H. pylori infections. Using NHANES data also limited us from considering genetic and virulence factors that play a role in pathogen transmission and acquisition. The virulence factor CagA enhances H. pylori’s ability to evade apoptosis and cause persistent infections [Reference Zhu68–Reference Ferrand70]. In addition, studies have suggested that distinct immune mechanisms according to age may also be involved in H. pylori pathogenesis, because immunophenotypes have been shown to differ between H. pylori-positive children and adults with duodenal ulcers [Reference Figueiredo71].
Despite these limitations, this study provides risk factor data that can inform future environmental assessments and prospective human studies to better link environmental exposures to H. pylori infections. The results of our analyses using this nationally representative population further support the hypothesis that exposure to well water, particularly during childhood, may be a possible source of H. pylori infection in the United States [Reference Klein38, Reference Queiroz43, Reference Malaty72, Reference Krumbiegel73]. Having well water may also be an indicator of other environmental exposures related to rural residence and associated with H. pylori infections that were not evaluated in this analysis. As suggested by Ryan et al., future research is warranted to examine pathogen occurrence in source and finished water, treatment removal rates, and infection risks for other water sources to validate their maximum contaminant level (MCL) model [Reference Ryan60].
CONCLUSIONS
Even though a casual association cannot be demonstrated with the cross-sectional design of NHANES, a better understanding of risk factors associated with H. pylori acquisition will help identify infection prevention strategies that can target populations with increased risk and reduce the burden of disease associated with chronic infections. Our analyses indicate that associations with environmental risk factors (well-water usage and occupational exposures to soil) and H. pylori seropositivity differ by demographic characteristics such as race/ethnicity and age. Moreover, this analysis of a nationally representative dataset indicates a disproportionate burden of infection associated with poor health and residential crowding, which may be related to unhygienic living conditions. Before effective interventions can be targeted at the necessary demographic groups, transmission and acquisition pattern dynamics must be addressed. Based on the association with well water and soil, and because of the bacterium's multiple niches in the natural world, differences in H. pylori occurrence by urban/rural residence deserve further evaluation. In addition, further investigations are warranted to examine how drinking water may be contaminated with H. pylori, and what role drinking-water systems may play in maintaining and spreading the bacterium.
ACKNOWLEDGEMENTS
We thank Sarah Collier at the US Centers for Disease Control and Prevention (CDC) for her review and comments. The manuscript has been subjected to the U.S. EPA's peer review and has been approved for publication.
This project was supported in part by an appointment to the Internship/Research Participation Program at the Office of Research and Development, U.S. Environmental Protection Agency, administered by the Oak Ridge Institute for Science and Education through an inter-agency agreement between the U.S. Department of Energy and EPA.
The views expressed in this paper are those of the authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency. Mention of trade names or commercial products does not constitute endorsement or recommendation for use.
DECLARATION OF INTEREST
None.