Article Text

PDF

Original article
Association between Helicobacter pylori and mortality in the NHANES III study
  1. Yu Chen1,
  2. Stephanie Segers1,
  3. Martin J Blaser2
  1. 1Department of Population Health, New York University School of Medicine, New York, USA
  2. 2Department of Medicine, New York University School of Medicine, New York, USA
  1. Correspondence to Dr Yu Chen, Department of Population Health, New York University School of Medicine, 650 First Avenue, Room 510, New York, NY 10016, USA; yu.chen{at}nyumc.org

Abstract

Objective Persistent colonisation by Helicobacter pylori, and especially by cagA-positive strains, has been related to several health outcomes with effects in opposite directions. Thus, it is important to evaluate its influence on total and category-specific mortality.

Design We conducted prospective cohort analyses in a nationally representative sample of 9895 participants enrolled in the National Health and Nutrition Examination Survey III to assess the association of H pylori status with all-cause and cause-specific mortality. Analyses for the association of H pylori cagA positivity with mortality were conducted in 7384 subjects with data on H pylori cagA status.

Results In older people (>40.1 years), H pylori was not associated with all-cause mortality (HR 1.00; 95% CI 0.84 to 1.18). There was an inverse association of H pylori status with stroke mortality (HR 0.69; 95% CI 0.44 to 1.08), and the inverse association was stronger for H pylori cagA positivity, with the HR of 0.45 (95% CI 0.27 to 0.76). H pylori was also strongly positively related to gastric cancer mortality. After we adjusted p values using the Benjamini–Hochberg false discovery rate method to account for multiple comparisons, these associations remained, and H pylori status was not related to other outcomes.

Conclusions Our findings suggest that H pylori has a mixed role in human health, but is not a major risk factor for all-cause mortality.

  • Helicobacter Pylori
  • Epidemiology
  • Gastric Cancer
  • Cardiovascular Disease
  • Pancreatic Cancer
View Full Text

Statistics from Altmetric.com

Significance of this study

What is already known about this subject?

  • Helicobacter pylori colonisation has been associated with increased risk of gastric cancer, and reduced risks of asthma and allergy.

  • Previous studies on H pylori colonisation and risks of cardiovascular disease and lung cancer have yielded inconsistent findings.

What are the new findings?

  • Helicobacter pylori status was not related to overall all-cause mortality.

  • H pylori colonisation was associated with reduced risks of deaths due to stroke and increased risks of deaths due to gastric cancer. The data also suggest an inverse association with lung cancer.

How might it impact on clinical practice in the foreseeable future?

  • These results suggest new possibly protective effects of Helicobacter pylori colonisation.

Introduction

The gastric bacterium Helicobacter pylori has long been present in humans.1 In the 30 years since the discovery of H pylori in the human stomach, there has been substantial evidence linking gastric colonisation of H pylori to increased risks of gastric adenocarcinoma,2 ,3 peptic ulcer disease4 and lymphoma.5 In recent decades, H pylori acquisition in industrialised countries has been diminishing with each successive generation,6 and this birth cohort phenomenon parallels a decreasing incidence of gastric cancer7 and an increasing incidence of oesophageal adenocarcinoma and related diseases (gastroesophageal junction adenocarcinomas). Several studies suggest an inverse association between H pylori colonisation and risk of gastroesophageal junction adenocarcinomas.8–11 In addition, H pylori colonisation has been linked to reduced risks of asthma and allergy,12 ,13 and the risks of cardiovascular diseases14–17 and lung cancer18–23 are uncertain, although studies vary in size and design. Since H pylori has been related to risks of a variety of health outcomes, it is important to evaluate its influence on total mortality and category-specific mortality.

H pylori is acquired almost exclusively in childhood and usually persists for life unless antimicrobial therapy is given.24 ,25 Antibodies to H pylori measured in serum are considered as valid measures for long-term colonisation.26–29 When present, H pylori is the dominant species colonising the human stomach,30 ,31 and is intimately linked to gastric physiology,32 especially the cagA-positive strains that inject H pylori products into epithelial cells.33 Antibody responses to the CagA protein permit detection of such cag-positive strains,34 which are more interactive with host cells than are cag-negative strains,21 ,32 and are associated with higher risk for gastric cancer35 and peptic ulcer disease, lower risk of oesophageal reflux and sequelae,8 ,9 ,36 ,37 as well as lower risk of childhood-onset asthma.8–10 ,12 ,13 ,36 ,37 However, few studies have evaluated the association of H pylori cag positivity with all-cause and cause-specific mortality in healthy individuals.

We conducted prospective cohort analyses in a nationally representative sample of 9895 participants in National Health and Nutrition Examination Survey III (NHANES III), with status of H pylori colonisation including status of cagA strains measured at the time of enrolment in 1988–1991 and mortality data with follow-up. The goal of our study was to examine the prospective relationship of H pylori colonisation with all-cause and cause-specific mortality, focusing on health outcomes that have been previously related to H pylori.

Material and methods

Study population

NHANES III, the seventh health examination survey performed in the USA beginning in 1960,38 was conducted from October 1988 through October 1994 in two phases, each of which comprised a national probability sample. In NHANES III, 39 695 people were studied; of those, 17 464were sampled at the first phase. The first phase was conducted from 18 October 1988 to 24 October 1991, at 44 locations. All interviewed people were invited to the mobile examination centre for a medical examination. The survey protocol was approved by the Institutional Review Board of the Centers for Disease Control and Prevention. All participants gave written informed consent.

H pylori status

Antibodies to H pylori were measured in 1993 on 6–19-year-old participants from phase I (1988–1991) of the survey using an ELISA (Pylori Stat, Whittaker Bioproducts, Walkersville, Maryland, USA) on surplus serum samples.38 Participants 20 years and older from phase I were tested for H pylori IgG antibodies in 1996 using the H pylori IgG ELISA (Wampole Laboratories, Cranbury, New Jersey, USA).39 For participants 20 years and older, in addition to determining whether H pylori IgG was present, anti-CagA IgG was also measured on surplus sera, using a method developed and standardised by Vanderbilt University, as described.34 As numerous tests were conducted before the testing, H pylori status surplus serum was not available from all of the participants. Of the 13 714 participants aged 6 and above enrolled in phase I, 10 168 have data on H pylori status, of which 9966 tested positive or negative, and 202 had equivocal results. Of the 9488 individuals aged 20 and above enrolled in phase I, 7384 had data on H pylori cagA status. All data collected for the NHANES are kept in strict confidence. The results of some tests, which did not include serological testing of H pylori status, had been communicated to the participants.40 All three tests used the exact same methodology which was developed in the laboratory of one of the authors (MJB) in the 1980s.41

Participants aged 6 and older with data on H pylori IgG antibodies were classified as H pylori positive or H pylori negative. For participants 20 years and older, on the basis of H pylori and cagA results, participants were classified into three groups: H pylori positive and cagA positive, H pylori positive and cagA negative, and H pylori and cagA negative, as described.42 The H pylori-positive and cagA-positive groups included all people with a positive cagA assay, regardless of the results of the H pylori assay, based on the utility of the CagA antigen to detect true positive responses in culture-positive people in the face of negative or equivocal values in the H pylori serologic assay.43 By definition, all people in the H pylori-negative group had negative CagA assays.

Among participants aged 6 years and above who were included in phase I of the NHANES III, 9966 were found to be either clearly positive or clearly negative for H pylori. We excluded 68 adults >17 years old with missing body mass index (BMI) and three participants not eligible for mortality follow-up (in the restricted data only). The final study population for association between H pylori and mortality included 9895 participants, over 147 796 person-years of observation. Among participants ≥20 years old, cagA was tested for 7384. We excluded 28 adults with missing BMI and three participants not eligible for mortality follow-up. The final study population for association between cagA positivity and mortality included 7354 participants, over 105 930 person-years of observation.

Mortality in NHANES III

The updated NHANES III Linked Mortality Restricted-use File was used for the present study. Compared with public-use linked mortality files, the restricted-use file includes detailed mortality information for all eligible survey participants, including children, as well as more precise and detailed age and follow-up information critical for age-specific analyses.44 The data provided mortality follow-up data from the date of NHANES III survey participation (1988–1994) through 31 December 2006.45 Vital status and cause of death assignment were based on probabilistic matching of NHANES III with the National Death Index (NDI) death certificate records. Cause of death was determined based on the underlying cause listed on the death certificates. The linking of NHANES III and NDI records was conducted by probabilistic matching, similar to the standard methodology offered by the NDI. Details have been presented elsewhere.46 Briefly, National Center for Health Statistics (NCHS) conducted a new calibration study to establish the cut-off scores for determining whether an NDI match is considered a true match or a false match using the information on social security number, name, birth date, sex, race, state of residence and birth, and marital status. NCHS reviewed a subset of death certificates to verify whether the NHANES III and NDI record match was correct. Of the selected 2544 death certificates that were reviewed, 2521 were considered ‘true’ matches by the probabilistic matching process (assumed deceased), and 23 were considered alive. Among the 2521 assigned decedents, 98.8% were confirmed deceased after death certificate review.46 Among the 23 people assumed to be alive, three were found to be deceased.46 We used the International Classification of Diseases, Ninth Revision (ICD-9) to classify deaths that occurred from 1988 to 1998 and the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) for deaths that occurred from 1999 to 2006 (see online etable 1). Data collection for NHANES III was approved by the NCHS Research Ethics Review Board. Analysis of deidentified data from the survey is exempt from the federal regulations for the protection of human research participants. Analysis of restricted data through the NCHS Research Data Center has also been approved by the NCHS Ethics Review Board.

Statistical analyses

We used Cox proportional regression models to estimate HRs for all-cause and cause-specific mortality comparing participants who were H pylori positive, H pylori positive/cagA positive, H pylori positive/cagA negative, with people of H pylori-negative status. We selected diseases that have been a priori associated with H pylori in the literature; no other outcomes were examined. The assumption of proportional hazards was examined by testing the cross-product terms between covariate variables and log function of survival time, and p values for all the terms were >0.10. Analyses comparing H pylori-positive with H pylori-negative status were conducted in participants aged 6 years or older, using the data on H pylori status. Analyses comparing participants who were H pylori positive/cagA positive or H pylori positive/cagA negative with H pylori-negative participants were conducted in those ≥20 years old. Potential confounding variables included sex, race–ethnicity, age, smoking status (for participants >17 years old), BMI (for participants >17 years old), and educational attainment. Additional adjustment for health insurance status, income, poverty/income ratio, and history of hypertension and diabetes was also conducted. Multivariate analyses were not conducted for rare outcomes to avoid over-specification of the models. Sensitivity analyses were conducted excluding deaths in the first 5 years of follow-up.

Since most (94%) of the deaths occurred after the age of 40 years in the NHANES III, we focused analyses in older people (>40.1 years, which was the median age of the overall population with H pylori and cagA data). Analyses were also conducted in the overall and younger study populations separately. The number of younger participants under 40.1 years who died from gastrointestinal cancers was already limited (five people) and therefore we conducted subgroup analyses only for older participants. All analyses included sample weights that account for the unequal probabilities of selection and non-response in the NHANES III, and variance calculations incorporated the sample weights that account for the complex sample design, as specified in prior NHANES publications47 and similar to studies using mortality data from NHANES III.48–50 All significance tests were two sided using p<0.05 as the level of statistical significance. Although in the present study, p values from the various models were not independent and the work was hypothesis oriented, we also adjusted p values using the Benjamini–Hochberg false discovery rate (FDR) method to account for multiple comparisons.51 FDR adjustment was conducted using the PROC MULTTEST statement in SAS 9.2. All other analyses were conducted using commercially available software (SUDAAN, V.10.0) and were conducted in the NCHS Research Data Center at Baruch College, City University of New York due to confidentiality requirements for the restricted-use linked mortality files.

Results

H pylori status in NHANES III

Table 1 shows the distributions of participants in relation to H pylori status by demographic and chronic disease risk factors in the overall study population, including those with data on H pylori cagA status. Participants who were H pylori positive or H pylori positive/cagA positive were more likely to be older, have lower educational attainment, have larger household size, and have higher BMI than H pylori-negative or H pylori-negative/cagA-negative participants.12 Participants of race–ethnicity other than non-Hispanic white, and those who had a history of diabetes or hypertension were more likely to be H pylori positive or H pylori positive/cagA positive. There was no apparent association of H pylori positivity with sex and smoking status.

Table 1

Baseline demographic and health-related characteristics of study subjects by Helicobacter pylori status in NHANES III

H pylori status and all-cause mortality and common causes of death in older participants

Because most (94%) deaths were observed in older participants, we conducted analyses focusing on these individuals. Table 2 shows analyses of findings pertaining to subjects >40.1 years old, the median of age of the overall study population who were H pylori negative. There was no association of either H pylori positivity or cagA positivity with all-cause mortality in the population. The HRs for all-cause mortality in relation to H pylori were all around 1.0. An inverse but not statistically significant association was observed for cardiovascular disease mortality in relation to H pylori positivity and cagA positivity, with an HR of 0.89 (95% CI 0.71 to 1.11) and 0.83 (95% CI 0.67 to 1.04), respectively. The HR for stroke mortality was 0.69 (95% CI 0.44 to 1.08) comparing participants who were H pylori positive with those who were H pylori negative. Participants who were H pylori positive/cagA positive were 55% significantly less likely to die from stroke (HR 0.45; 95% CI 0.27 to 0.76) than those who were H pylori negative/cagA negative.

Table 2

Risk of all-cause, cardiovascular disease, cancer and respiratory disease mortality in relation to Helicobacter pylori status in NHANES III among participants ≥40.1 years old

The HR for lung cancer mortality was 0.61 (95% CI 0.35 to 1.05) comparing H pylori-positive with H pylori-negative participants. A significant inverse association between H pylori cagA positivity and lung cancer mortality was observed in these older participants. The HRs were 0.67 (0.35–1.29) and 0.55 (0.31–0.98) comparing H pylori-positive/cagA-negative participants and H pylori-positive/cagA-positive participants, respectively, with H pylori-negative/cagA-negative participants. There was no overall association of H pylori status with all cause mortality, overall cancer mortality, overall gastrointestinal cancer mortality or respiratory disease mortality.

After adjustment for FDR, the association of H pylori positive/cagA positive with stroke mortality remained significant (FDR-adjusted p=0.045). However, the adjusted p value for the association between H pylori-positive/cagA-positive status and lung cancer mortality was no longer significant (FDR-adjusted p=0.21) (see online etable 3).

Additional adjustment for health insurance status, family income and poverty index did not change the effect estimates appreciably (data not shown). For instance, the HRs for stroke were 0.98 (95% CI 0.57 to 1.70) and 0.43 (95% CI 0.25 to 0.73) comparing H pylori-positive/cagA-negative and H pylori-positive/cagA-positive participants, respectively, with H pylori-negative/cagA-negative participants, and the HRs for lung cancer were 0.65 (0.34–1.25) and 0.49 (0.27–0.89) comparing H pylori-positive/cagA-negative and H pylori-positive/cagA-positive participants, respectively, with H pylori-negative/cagA-negative participants. Additional adjustment for hypertension status at baseline also did not change the effect estimates appreciably (data not shown). For instance, the HRs for stroke were 1.03 (0.59–1.81) and 0.46 (0.28–0.77) comparing H pylori-positive/cagA-negative and H pylori-positive/cagA-positive participants, respectively, with H pylori-negative/cagA-negative participants. Sensitivity analyses excluding deaths in the first 5 years generated similar results (data not shown). For instance, after excluding deaths in the first 5 years, the HRs for stroke were 1.11 (0.60–2.07) and 0.47 (0.26–0.87) comparing H pylori-positive/cagA-negative and H pylori-positive/cagA-positive participants, respectively, with H pylori-negative/cagA-negative participants, and the HRs for lung cancer were 0.65 (0.30–1.40) and 0.50 (0.26–0.99) comparing H pylori-positive/cagA-negative and H pylori-positive/cagA-positive participants, respectively, with H pylori-negative/cagA-negative participants. All of the observed associations were similar in the overall study population (see online etable 2). Because of the relatively small number of deaths observed in younger participants, many effect estimates were not reliable (data not shown).

H pylori status and gastrointestinal cancer mortality

Since the associations between H pylori and the risk of gastrointestinal cancer have been investigated extensively in prior studies, we assessed the effects of H pylori colonisation on mortality due to selected gastrointestinal cancers in older participants (table 3). There was a strongly positive association between H pylori positivity and gastric cancer mortality, and positive associations of similar magnitude were observed for participants who were H pylori positive/cagA negative or H pylori positive/cagA positive compared with those who were H pylori negative/cagA negative. For pancreatic cancer, the inverse association was stronger for cagA-negative strains (HR 0.21; 95% CI 0.05 to 0.91). However, the numbers of deaths for these outcomes were limited. All of the observed associations were similar in the overall study population (see online etable 2). After adjustment for FDR, the association of H pylori positivity, H pylori-positive/cagA-negative status and H pylori-positive status with gastric cancer mortality remained significant (FDR-adjusted p=0.045). All other associations were not statistically significant (see online etable 3). Models for rare outcomes excluding deaths in the first 5 years did not converge due to limited sample size.

Table 3

Risk of mortality due to colorectal cancer, gastric cancer, oesophageal cancer, and pancreatic cancer mortality (rare outcomes) in NHANES III in participants ≥40.1 years old

Discussion

Since H pylori has been present in humans for at least 58 000 years,1 and has been rapidly declining from human populations during the past century,6 ,28 an important question is whether its presence affects human health in the aggregate. In the present study, H pylori status was not related to overall all-cause mortality. H pylori was associated with an increased risk of death due to gastric cancer, but with reduced risks of deaths due to stroke and lung cancer.

As expected, the greatest mortality in the studied population was that due to cardiovascular disease. The literature on the association between H pylori and cardiovascular disease risk is not consistent, with reports of positive52–54 or null55–59 associations. In a population-based case-control study in the Erlangen Stroke Project,60 H pylori was associated with lower risk of cardioembolic stroke (OR 0.21; 95% CI 0.06 to 0.71). More recently, a population-based German cohort of 9953 older participants showed a significant inverse association between cagA positivity and cardiovascular mortality (HR 0.62; 95% CI 0.41 to 0.94);17 the association of cagA positivity with myocardial infarction and stroke was also inverse (HR 0.71 to 0.59, respectively), but not statistically significant.17 In the present study, we observed an insignificant reduced risk of cardiovascular disease mortality and a significant reduced risk of stroke mortality, comparing H pylori-positive or H pylori-positive/cagA-positive participants with H pylori-negative/cagA-negative participants; all of the apparent association is with cagA-positive strains. While our findings indicate that H pylori or cagA are not major risk factors for cardiovascular disease or mortality, it cannot be ruled out that H pylori or specifically cagA-positive strains may be protective for stroke. Recent studies suggest that regulatory T cells (T-reg cells) may be protective against stroke risk.61 ,62 People with gastric H pylori colonisation have much more substantial gastric T-reg populations and higher gastric expression of T-reg-linked cytokines than H pylori-negative people.63 The T-reg downregulated immune system in H pylori-positive hosts may be less damaging to aging blood vessels. H pylori-associated inverse risk with asthma and allergies appears to be related to the induction of T-reg responses in animal models.64 ,65

The strong positive association of H pylori status with mortality due to gastric cancer (table 3) was consistent with the literature.2 ,3 The very high HRs indicate that nearly all gastric fatal cancers in the USA are H pylori related; the progressive decline in H pylori prevalence thus may explain nearly all of the reduction in gastric cancer mortality over the past 80 years.7 However, in the present study, there were no data on locations of the cancers and therefore we could not differentiate whether the association for gastric cancer could be due to cardia and/or non-cardia cancers. Although many studies show a heightened risk for CagA+ H pylori strain with relation to gastric cancer, other studies have shown that CagA+ and CagA– H pylori strains are similarly associated with gastric cancer.8 Prior case control studies documented an inverse association with mortality from oesophageal cancer, especially for cagA+ strains8–10 and a positive association with pancreatic cancer. Although we also observed an insignificant association with a consistent inverse direction for oesophageal cancer, the numbers are limited, and studies with larger sample size and/or longer follow-up are needed.

Several case–control studies evaluated the association between H pylori and lung cancer, with inconsistent findings and small sample sizes.18–20 ,22 ,23 More recently, a prospective study in the α-Tocopherol, β-Carotene Cancer Prevention cohort found no association between H pylori and lung cancer.21 However, the study population only included men who were smokers. In the present study, we found an overall inverse association between H pylori and lung cancer in older participants, a significant inverse association for cagA-positive strains, and a consistent trend for cagA-negative strains in older participants. However, information on subtypes and histological types of lung cancer was not available in NHANES mortality data to further investigate whether the association is due to a specific type of lung cancer. It should be noted that in the present study the adjusted p value for the association between H pylori-positive/cagA-positive strain and lung cancer mortality was no longer significant (FDR-adjusted p=0.21). Future studies are needed to confirm or refute our findings on lung cancer.

There are strengths and limitations of our study. Our study included representative samples from the general population, comprehensive data on other risk factors and objective follow-up data on mortality. However, the participants were relatively young (median age 40 among those with H pylori and cagA data) at study entry, and therefore the numbers of deaths for uncommon cancers were limited. H pylori testing was conducted on surplus serum samples from participants in certain age groups. Numerous tests were conducted on blood samples prior to the H pylori testing and therefore surplus sera were not available for everyone. However, it is unlikely that H pylori carriers who died during the follow-up were differentially excluded at the time of H pylori testing (a necessary condition for a bias). Distributions of demographic and lifestyles are similar between participants with and without H pylori testing (data not shown), except that fewer black people (7% less) had surplus sera for H pylori testing to be included in the study compared with other groups. The sampling weights that addressed the issues of oversampling and non-response may not account for the fact that many participants did not have surplus sera for H pylori testing. However, we do not believe the external validity (generalisability) of the study results would be affected, as the association between H pylori and mortality may not differ substantially by ethnic groups. Second, our results might have been influenced by death certificate errors, especially for non-cancer outcomes. Although some errors in the NHANES mortality linkage procedure are unavoidable, the process was conducted without the knowledge of the participants’ health data. Studies that have examined the validity of the death certificate diagnosis of stroke have found a high specificity and a moderate sensitivity for the diagnosis.66 ,67 Lower sensitivity or errors in the death certificate diagnosis of stroke would be expected to be non-differential by H pylori status and result in underestimation of the association. Third, the presence of H pylori may merely be a marker for other important risk factors of disease. Although test results of H pylori status were not communicated to the NHANES participants, since H pylori may be eradicated collaterally after routine antibiotic treatment of coexisting illness presumed to have been infectious, its absence could reflect the loss of other bacteria. However, there is no evidence of a positive association between H pylori and any specific bacterial infection that is also related to a reduced risk of lung cancer or stroke (positive confounding by other bacterial infection). Furthermore, the study findings did not change appreciably with additional adjustments for other indicators of socioeconomic status. Nevertheless, because of the observational nature of this study, we cannot exclude the possibility of residual and unmeasured confounding. Lastly, we have no knowledge of whether H pylori may have been lost or gained in the interval between ascertainment and mortality. However, from longitudinal studies of H pylori-positive adults, we know that rates of ‘spontaneous’ annual loss of the organism among those with established colonisation are low.29 Similarly, early childhood is the nearly exclusive time for colonisation,24 ,25 thus substantial later-in-life acquisitions are unlikely. Misclassification as a result of imperfect accuracy of the serological tests is likely to be non-differential (not related to the subsequent risk of death), which would bias the results towards no association.

In conclusion, our findings do not suggest that H pylori colonisation is a major risk factor for all-cause mortality. We observed an inverse association between H pylori colonisation and stroke mortality and a direct association with gastric cancer. While the associations observed need to be reexamined in future studies, our results provide further evidence that H pylori has a mixed role in human health68 and raise new possible protective effects of H pylori colonisation.

Acknowledgments

Correction notice This article has been corrected since it was published Online First. ‘of nearly 20 years’ has been removed from this sentence to read: We conducted prospective cohort analyses in a nationally representative sample of 9895 participants in National Health and Nutrition Examination Survey III (NHANES III), with status of H pylori colonisation including status of cagA strains measured at the time of enrolment in 1988–1991 and mortality data with follow-up. ‘p Value’ has been added to table 3, and the reference citation in this sentence has been amended ‘other studies have shown that CagA+ and CagA– H pylori strains are similarly associated with gastric cancer.8

Acknowledgments

Acknowledgements We thank Jonathan Fisher from the New York Census Research Data Center, Baruch College, City University of New York, and Nataliya Kravets from the National Center for Health Statistics for their help in the project. The findings and conclusions in this paper are those of the authors and do not necessarily represent the views of the Research Data Center, the National Center for Health Statistics, or the Centers for Disease Control and Prevention. We also thank Dr. Huilin Li for statistical advice and Mr. Ryan Shanley for helping the preparation of the manuscript.

References

View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:

Footnotes

  • Contributors Study concept and design: YC and MJB. Acquisition of data: YC and MJB. Analysis and interpretation of data: YC, SS and MJB. Drafting of the manuscript: YC and MJB. Critical revision of the manuscript for important intellectual content: YC and MJB. Administrative, technical, and material support: YC. Study supervision: YC.

  • Funding This work was supported in part by grants R01DK090989, R01GM63270, ES000260, and P30CA16087 from the National Institutes of Health, and by the Diane Belfer Program for Human Microbial Ecology.

  • Competing interests None.

  • Ethics approval Institutional Review Board of the Centers for Disease Control and Prevention.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles