Helicobacter pylori seroprevalence and its association with prevalent and incident blood pressure and haemoglobin A1c in the US-based National Longitudinal Study of Adolescent to Adult Health

Abstract

Introduction

Chronic Helicobacter pylori (H. pylori) infection may increase the risk of hypertension (HTN) and type 2 diabetes mellitus (T2D), but association studies have produced mixed results. This cross-sectional and longitudinal study examined whether H. pylori seropositivity was associated with elevated blood pressure (BP), haemoglobin A1c (HbA1c), prevalent and incident HTN, and T2D over 10 years.

Methods

Add Health, a US-based cohort study, has followed participants from adolescence in the mid-1990s (wave I) through early midlife (wave V) in 2016–2018. Wave IV (n=15 701) participants were tested for immunoglobulin antibodies to H. pylori with a binary seropositivity cut-off (≥13.217 U/mL). Systolic BP (SBP), diastolic BP (DBP) and HbA1c were measured at waves IV and V. Prevalent HTN was defined as SBP ≥140 mm Hg and/or DBP ≥90 mm Hg, self-reported HTN or HTN medication use at wave IV. Prevalent T2D was defined as HbA1c ≥6.5%, self-reported T2D or T2D medication use at wave IV. Incident HTN and T2D at wave V were defined among participants without HTN or T2D at wave IV. Multivariable linear and Poisson regression models determined prevalence differences (PD) and incidence rate ratios, respectively, with 95% CIs.

Results

The median (IQR) ages in waves IV and V were 28 (27, 30) and 37 (36, 39) years, respectively, with 54% female in 4600 participants; 958 (20.8%) were seropositive for H. pylori. In adjusted models, seropositive participants had lower mean SBP (PD −1.62, 95% CI −3.06 to –0.18) than seronegative participants. H. pylori seropositivity was not associated with DBP, HbA1c, prevalent and incident HTN, or T2D.

Conclusions

In a cohort of younger adults, H. pylori seropositivity was not associated with prevalent or incident HTN and T2D. More studies are required to understand the interaction between infectious disease and chronic cardiometabolic disorders and how it can change as people age.

WHAT IS ALREADY KNOWN ON THIS TOPIC.

WHAT THIS STUDY ADDS

• In a representative sample of younger US adults, the association between H. pylori seroprevalence and HTN and T2D was examined at two time points.

• H. pylori seroprevalence was associated with lower mean systolic blood pressure, but was not associated with diastolic blood pressure, haemoglobin A1c, and prevalent or incident HTN and T2D.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our study highlights the complexity by which chronic inflammation may affect chronic disorders and agrees with studies that have not noted an association between H. pylori and HTN and T2D.

• This provides the first step in a life course approach in examining how the relationship between H. pylori infection and cardiometabolic outcomes may change over time in a diverse population.

Introduction

Helicobacter pylori (H. pylori) is an enteric gram-negative bacterium found in the human stomach that infects more than half of the global population.¹ Acute infection is typically asymptomatic; thus, H. pylori often remains undetected, and infection may become chronic.² Chronic H. pylori infection can cause a systemic inflammatory state that can lead to endothelial dysfunction and insulin resistance.3,7 These are mechanisms associated with the development of hypertension (HTN) and type 2 diabetes mellitus (T2D), which portend significant morbidity and mortality.8,16 Identifying modifiable risk factors for HTN and T2D, particularly in early adulthood, may prevent or delay their onset and avert undue suffering.1117,22

The relationship between H. pylori infection, HTN and T2D is controversial.34 23,28 Studies on H. pylori infection and HTN have yielded mixed results depending on the study design, population and exposure ascertainment.26,29 Community-based cross-sectional studies have noted that blood pressure (BP) was not associated with H. pylori infection, measured by a urea breath test.²⁷ In a meta-analysis, H. pylori infection increased the odds of HTN and was associated with higher levels of systolic BP (SBP) and diastolic BP (DBP) compared with H. pylori-negative participants.²⁸ Studies of diabetic patients have shown that they are more likely to have a current or past H. pylori infection than non-diabetic patients.^{4 30} A meta-analysis noted that H. pylori-positive participants with T2D did not have significantly higher haemoglobin A1c (HbA1c) levels than H. pylori-negative participants.²⁴ Finally, a cohort of older diabetes-free Latinx individuals demonstrated that H. pylori seropositivity was associated with incident T2D.³¹ Nevertheless, possible pathways by which H. pylori may directly or indirectly affect the risk of HTN and T2D are plausible.34 6 25 26 29 32,36

Our study used BP and HbA1c, measured at two time points during young adulthood and early midlife, to assess the association of HTN and T2D with H. pylori infection using data from the National Longitudinal Study of Adolescent to Adult Health (Add Health). Add Health is a US-based, representative cohort study designed to observe health and health-related behaviours in adolescence and outcomes in adulthood.^{37 38} Add Health’s younger cohort, longitudinal design and rich dataset allowed us to understand changes in BP and HbA1c over time, which may indicate clinical and subclinical HTN and T2D.

Materials and methods

Data source

Add Health is a longitudinal, nationally representative cohort study designed to understand adolescent health and behaviour and investigate health outcomes in adulthood.³⁸ Details of recruitment and sampling structure have been published previously.³⁸ Briefly, the Add Health study population, comprised of students in grades 7–12 from randomly selected schools throughout the USA in 1994, has been followed over time through five in-home interviews called ‘waves’ until 2018.³⁸

Study population

Add Health study participants have been followed through adolescence (ages 12–18), their transition to adulthood (ages 18–26), young adulthood (ages 24–32) and early midlife (ages 33–43) unless they passed away, dropped out of the study or were lost to follow-up.^{37 38} Participants in waves IV and V, completed in 2008–2009 and 2016–2018, respectively, were used in this study. There were 15 701 young adult participants in wave IV and 12 300 early midlife participants in wave V.^{38 39} Among wave IV participants with dried capillary whole blood spots (DBS), a random sample of 5021 individuals was selected for testing of infectious disease markers and cytokines; two samples were not viable, resulting in an analytic sample of 5019 participants.⁴⁰ Participants missing a school identification (n=99), cross-sectional sampling weights for waves IV (n=198) or V (n=1661) or census region (n=14) were excluded from the study analyses (online supplemental file 1, online supplemental figure 1).

Patient and public involvement

No patients or members of the public were involved in the design, conduct or dissemination of this study.

Measures

The primary exposure was the presence of immunoglobulin G antibodies to H. pylori measured by a commercially available ELISA (Abnova, KA0220, Taipei City, Taiwan) from DBS at wave IV.⁴⁰ As the kit was designed for serum samples collected from venous blood and not DBS, positive and negative DBS control samples were used in the assays along with the kit’s standards and controls.⁴⁰ Furthermore, matched serum and DBS samples collected from 37 individuals were compared with determine the seropositivity cut-off for DBS samples.⁴⁰ According to the test kit specifications, 20 U/mL was the threshold for seropositivity in serum, corresponding to a cut-off of 13.217 U/mL in DBS; this threshold was used to create a binary variable.⁴⁰ The correlation between DBS and serum in positive samples was linear, with a correlation coefficient of 0.95.⁴⁰ Intra-individual reliability of H. pylori concentrations was determined to be high (0.99, 95% CI 0.97 to 1.00) in a race/ethnicity- and sex-stratified sample (n=22).⁴⁰ Participants missing a value for H. pylori (n=122) were kept as missing. In seropositive individuals, antibody levels were analysed as a continuous variable to avoid a right-skewed distribution in the total cohort.

During waves IV and V, BP was measured using oscillometric BP monitors; SBP and DBP were reported in mm Hg.⁴¹ Per cent glycosylated haemoglobin or HbA1c was determined from DBS at waves IV and from a venous blood draw at wave V.⁴² In 80 paired DBS and venous blood samples tested for HbA1c, values were strongly correlated (Pearson r=0.99) in a linear fashion.⁴² At each wave, participants self-reported if they had ever been diagnosed with HTN or T2D and, if so, at what age they were diagnosed and if they were taking any medications for their HTN or T2D. Participants were also asked to record all prescription medications used in the last 4 weeks.

SBP and DBP were analysed as continuous measures in waves IV and V. They were corrected for HTN medication use by adding 9 mm Hg to SBP and 6 mm Hg to DBP to all treated participants.⁴³ BP classification was based on guidelines from the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation and Treatment of High Blood Pressure.¹⁰ Prevalent HTN was defined as SBP ≥140 mm Hg and DBP ≥90 mm Hg, self-reported HTN from questionnaires at wave IV, or self-reported use of HTN medications. Incident HTN was defined as no evidence of HTN at wave IV and SBP ≥140 mm Hg and DBP ≥90 mm Hg at wave V, self-reported HTN or self-reported use of HTN medications at wave V.

HbA1c was also analysed as a continuous and categorical variable. Participants who reported T2D medication use at wave IV (n=69) and wave V (n=270) were excluded from analyses using HbA1c as a continuous measure. T2D classification was based on guidelines from the American Diabetes Association.⁴⁴ Prevalent T2D at wave IV was defined as HbA1c ≥6.5%, self-reported T2D from questionnaires at wave IV, or self-reported T2D medication use. At wave V, incident T2D was defined as no evidence of T2D at wave IV and HbA1c ≥6.5%, self-reported T2D or self-reported use of T2D medications at wave V. To measure the annualised rate of change in SBP, DBP and HbA1c separately between wave IV and V, we calculated the difference in each measure from waves IV to V and divided it by the years between the two visits.

Sociodemographic, family history, behavioural and medication information were assessed using validated questionnaires.³⁹ Biological sex, immigrant generation, nativity and social origins score, a constructed socioeconomic status variable of parental education, occupation, household income and receipt of public assistance, were extracted from wave I.⁴⁵ Age, educational attainment, health insurance status, body mass index (BMI), self-reported diagnosis of high cholesterol, household income, alcohol use, C reactive protein (CRP) and physical activity were gathered from waves IV and V. Self-classified race and ethnicity and smoking were collected from wave V constructed variables.

High sensitivity-CRP concentrations were measured from DBS at wave IV using an adapted sandwich ELISA with quality control samples.⁴⁶ The classification of CRP was based on the American Heart Association/Centers for Disease Control clinical and public health guidelines.⁴⁶ Daily smokers were defined as participants who answered ‘30’ to the number of days they smoked cigarettes in the past 30 days at Wave IV. Alcohol use was a categorical variable from 0 to 7, as the number of days they drank alcohol in the past 30 days; participants who never drank were included in the zero category. Physical activity was defined as any walking for exercise in the past 7 days versus none. Covariates were identified using a directed acyclic graph based on a literature review of the relationship between exposure and outcomes.23 6 10 14 44 47,65

The variables for race, ethnicity, educational attainment, household income and health insurance were collapsed to avoid small cell counts. The six-category constructed race and ethnicity variable was condensed to four categories: non-Hispanic (NH) White, NH Black, Hispanic and Other, which included NH Asian, NH Pacific Islander, Native American and Other. Race and ethnicity are social constructs and do not reflect biological differences; the ‘Other’ category could not be differentiated further due to sample size. A minimally sufficient set of confounders to control for in the models included age, sex, race and ethnicity, household income, health insurance, social origins, immigrant generation, and either BP or HbA1c, depending on the outcome. Educational attainment was not adjusted for in the models, as it is likely on the pathway between H. pylori infection and HTN or T2D.

Statistical analyses

Unweighted total and weighted means, SD and proportions were calculated from bivariate analyses of H. pylori seroprevalence with the covariates to present descriptive statistics. Non-parametric Wilcoxon rank-sum and adjusted Wald χ² tests were used to compare means and proportions between seropositive and seronegative participants. The number of missing values and the percentage missing of the total group for each covariate were recorded in table 1. Weighted proportions and means are calculated from the non-missing participants.

Table 1. Add Health participant characteristics at wave IV by H. pylori seroprevalence (N=4600).

	N	H. pylori negative (n=3642)	N	H. pylori positive (n=958)	P value
Age, mean (SE)	3642	28.2 (0.13)	958	28.4 (0.14)	0.03
Biological sex, n (%)	3642		958		0.65
Female		2056 (54.1)		546 (55.2)
Male		1586 (45.9)		412 (44.8)
Race/ethnicity groups, n (%)	3642		958		<0.0001
Non-Hispanic White		2403 (77.9)		379 (52.9)
Non-Hispanic Black		521 (9.3)		298 (24.1)
Hispanic		423 (8.0)		196 (17.7)
Non-Hispanic Asian/Pacific Islander/Native American/Other		295 (4.8)		85 (5.3)
Born in USA* (yes), n (%)	3594	3389 (96.1)	939	862 (94.3)	0.14
Missing (refused/don’t know)†		48 (1.3)		19 (2.0)
Social origins, mean (SE)	3477	0.10 (0.06)	857	−0.42 (0.10)	<0.0001
Missing†		165 (4.5)		101 (11.0)
Immigrant generation, n (%)	3594		939		0.004
First		205 (3.9)		77 (5.7)
Second		451 (9.2)		183 (16.1)
Third		2938 (86.8)		679 (78.1)
Missing (refused/don’t know)†		48 (1.3)		19 (2.0)
Education, n (%)	3642		958		<0.0001
High school or less		235 (7.9)		118 (14.3)
High school graduate		588 (17.3)		195 (21.0)
Vocational/technical training		357 (9.3)		105 (10.4)
Some college		1340 (36.5)		342 (36.0)
College graduate		710 (18.9)		133 (11.6)
Some grad school or more		412 (10.1)		65 (6.7)
Household income, n (%)	3455		880		0.002
<$5000		84 (2.6)		39 (4.9)
$5000–24999		449 (14.0)		172 (21.3)
$25 000–49 999		968 (28.8)		252 (30.3)
$50 000–79 999		1385 (40.0)		302 (33.1)
≥$100 000		569 (14.5)		115 (10.4)
Missing (refused/don’t know)†		187 (5.1)		78 (8.1)
Health insurance (none), n (%)	3616	734 (21.8)	947	249 (29.2)	0.0004
Missing (refused/don’t know)†		26 (0.7)		11 (1.1)
Walk for exercise (any), n (%)	3638	2128 (56.3)	958	571 (57.7)	0.59
Missing (refused/don’t know)†		4 (0.1)		0 (0)
Systolic BP (mm Hg), mean (SE)	3553	125 (0.35)	924	125 (0.67)	0.39
Missing (refused/invalid)†		89 (2.4)		34 (3.5)
Diastolic BP (mm Hg), mean (SE)	3553	79.5 (0.26)	924	80.1 (0.45)	0.43
Missing (refused/invalid)†		89 (2.4)		34 (3.5)
BP classification, n (%)	3553		924		0.53
Normal		1193 (32.9)		306 (34.7)
Pre-HTN		1640 (45.2)		396 (41.1)
HTN stage 1		584 (17.8)		180 (19.5)
HTN stage 2		136 (4.2)		42 (4.8)
Missing (refused/invalid)†		89 (2.4)		34 (3.5)
HTN medications (yes), n (%)	1476	137 (10.1)	342	37 (8.7)	0.53
No reported medication		2166 (59.5)		616 (64.3)
BMI (kg/m2), mean (SE)	3596	29.1 (0.21)	950	30.3 (0.39)	0.002
Missing†		46 (1.3)		8 (0.8)
Daily smoker (yes), n (%)	3622	896 (28.4)	944	234 (28.9)	0.80
Missing (refused/don’t know)†		20 (0.6)		14 (1.5)
Alcohol use in last 30 days (never drank), n (%)	3358	643 (18.4)	876	233 (25.9)	0.13
Missing (refused/don’t know)†		284 (7.8)		82 (8.6)
HbA1c (%), mean (SE)	3638	5.5 (0.02)	954	5.6 (0.04)	0.01
Missing (refused/invalid)†		4 (0.1)		4 (0.4)
HbA1c classification, n (%)	3638		954		0.007
Normal		2574 (73.7)		588 (66.5)
Prediabetic		931 (23.3)		315 (29.1)
Diabetic		133 (3.1)		51 (4.3)
Missing (refused/invalid)†		4 (0.1)		4 (0.4)
Anti-T2D medications (yes), n (%)	3642	53 (1.3)	958	16 (1.0)	0.58

All estimates, except N, account for the complex survey design of Add Health. Wilcoxon rank-sum tests were used for continuous variables and adjusted Wald χ2 tests were used for categorical variables. 122 participants were missing H. pylori data.

P values less than the alpha of 0.05 are in bold text.

^*Weighted proportions and means are calculated from the non-missing participants.

^†Missing percentages are calculated from the whole group for each covariate.

BMI, body mass index; BP, blood pressure; HbA1c, hemoglobin A1c; H. pylori, Helicobacter pylori; HTN, hypertension; NH, non-Hispanic; SE, standard error; T2D, type 2 diabetes.

Multivariable generalised linear models with a Gaussian-identity and log-binomial (quasibinomial in R) link were used to produce prevalence difference and ratio estimates, respectively. Given the known associations of sex, age, race and ethnicity with HTN and T2D, models were minimally adjusted for these factors. Forward stepwise models were conducted where one covariate from the minimally sufficient adjustment set was added at a time to the minimally adjusted model to examine the model fit. Nested models were compared by changes in the Akaike information criterion and through likelihood ratio tests (LRT) with an alpha of 0.10. Correlation analyses between the quantity of IgG antibodies to H. pylori and SBP, DBP and HbA1c were performed separately among the seropositive individuals to determine if H. pylori titre was associated with increased SBP, DBP and HbA1c, using Pearson’s correlation as the statistical test.

Differences in annualised rates of change in SBP, DBP and HbA1c by H. pylori seroprevalence were calculated from multivariable linear regression models. Incidence rates and rate ratios were determined from Poisson regression models using the quasipoisson family and log link. Sex, race and ethnicity are linked to H. pylori, HTN and T2D separately. We hypothesised that any association between H. pylori and HTN/T2D would either vary by sex, race and ethnicity or be conditional on sex, race and ethnicity. Therefore, stratified analyses were conducted for each sex, racial and ethnic group. Tests for interaction were assessed using an LRT. For all models, 95% CIs were calculated to determine the precision of estimation. H. pylori serology values were left-skewed, with many values close to the cut-off. To ensure robustness of the results presented, a sensitivity analysis varying the seropositivity threshold by ±10% was conducted for each association (online supplemental file 2). All analyses were weighted to reflect Add Health’s complex sampling design using cross-sectional weights from wave IV or wave V.⁶⁶ To reduce differences between the analytic biosample and the wave IV cohort, the participants not missing survey weight variables (clustering and stratification) contributed these variables to the weighted estimate. Their cross-sectional weights were set to 0.0001 to not contribute to the sampling weight. All analyses were completed using R software, V.4.3.1 (R Foundation for Statistical Computing).

Results

Demographic characteristics

After exclusions (n=419), 4600 Add Health participants from the wave IV biosample subset were included in the analysis (online supplemental file 1, online supplemental figure 1). The median age (IQR) among these participants was 28 (27, 30) years, and 54.2% were female. Of 4600 participants, 958 (20.8%) tested positive for H. pylori antibodies. By racial and ethnic groups, 13.4%, 36.9%, 33.4% and 20.2% of participants who self-reported as NH White, NH Black, Hispanic and NH Asian/Pacific Islander/Native American/Other had H. pylori antibodies, respectively.

Wave IV characteristics by H. pylori seroprevalence are presented in table 1 and online supplemental file 3. Participants seropositive for H. pylori were more likely to self-report as NH Black (24.1% vs 9.3%) and Hispanic (17.7% vs 8.0%) than seronegative participants. A greater proportion of H. pylori seropositive participants were first- or second-generation immigrants and had a lower mean social origin score than seronegative participants. Seropositive participants were less likely to be college graduates, self-report a higher household income and were more likely to report having no health insurance (table 1) than participants seronegative for H. pylori. H. pylori seropositive participants also had significantly higher BMI than seronegative participants, although the mean BMI was high for both groups. There were no differences in biological sex, walking for exercise, daily smoking, alcohol use or CRP levels by H. pylori seroprevalence (online supplemental file 3, table 1).

In this cohort, 33.2% (n=1499) were considered to have normal BP, 44.4% (n=2036) were classified as prehypertensive, 18.1% (n=764) had stage I HTN and 4.3% (n=178) had stage 2 HTN at wave IV. 510 participants self-reported ever being diagnosed with HTN, and 174 presented anti-HTN medications during their home interview. By H. pylori seroprevalence, there was no difference in SBP, DBP, BP classification, self-reported HTN diagnosis or HTN medication use at wave IV (table 1).

3162 (72.4%) participants had normal levels of HbA1c, 1246 (24.3%) were classified as prediabetic and 184 (3.30%) were considered to have T2D. 123 participants self-reported ever being diagnosed with T2D, and 69 reported taking anti-T2D medications. H. pylori seropositive participants had higher HbA1c levels (5.6% vs 5.5%) and were more likely to be classified as prediabetic and diabetic than seronegative participants (table 1). There was no difference in self-reported diagnosis of T2D or use of T2D medications by H. pylori seroprevalence.

Prevalence analyses

Continuous SBP, DBP and HbA1c

In unadjusted linear regression models examining the association between H. pylori seroprevalence and continuous SBP at wave IV, there was no mean difference in SBP between seropositive and seronegative participants (online supplemental file 1, online supplemental table 1). After adjusting for age, sex, race, ethnicity, household income, social origins, HbA1c, BMI, smoking and alcohol use, the mean SBP among H. pylori seropositive participants was 1.62 mm Hg (95% CI −3.06 to –0.18) lower on an absolute scale than among seronegative participants (online supplemental table 1; figure 1A). There was no mean difference in DBP between seroprevalence groups in unadjusted or adjusted (figure 1C) models. H. pylori seropositive participants had higher mean percentages of HbA1c on the absolute scale in unadjusted models (prevalence differences (PD): 0.07%, 95% CI 0.01 to 0.14; online supplemental table 1). These results were attenuated after adjusting for age, sex, race, ethnicity, SBP, DBP, BMI and CRP (figure 1E).

Figure 1. Estimated adjusted systolic blood pressure (SBP), diastolic blood pressure (DBP) and haemoglobin A1c (HbA1c) by Helicobacter pylori seroprevalence and serology. Estimated (A) SBP, (C) DBP and (E) HbA1c values by H. pylori seroprevalence from fully adjusted linear regression models. The boxplots present the median, quartile 1 and quartile 3 for each group distribution. The adjusted prevalence differences (PD) with 95% CI are stated for each outcome. Among individuals seropositive for H. pylori infection, estimated (B) SBP (in blue), (D) DBP (in green) and (F) HbA1c (in orange) values from fully adjusted linear regression models were correlated against H. pylori titre (serology). Values from seronegative participants are shown in grey; the LOESS line for each outcome is based only on seropositive individuals. Pearson’s correlation coefficients (r) are shown for each outcome.

The relationship between H. pylori seroprevalence and cardiometabolic markers was also evaluated using continuous variables among seropositive participants. There was no association in adjusted correlation plots (figure 1). Pearson’s correlation coefficient (r) between H. pylori titre and SBP, DBP and HbA1c was 0.05 (figure 1B), 0.18 (figure 1D) and 0.04 (figure 1F), respectively.

In subgroup analyses among female participants, H. pylori seropositive patients had lower mean SBP (PD: −3.05 mm Hg, 95% CI −5.08 to –1.02) and DBP (PD −1.61, 95% CI −3.21 to –0.01) levels on an absolute scale than seronegative participants (online supplemental table 2). However, there was no interaction by biological sex for SBP (online supplemental figure 2A), DBP (online supplemental figure 2C) and HbA1c (online supplemental figure 2E, online supplemental table 2). Among NH White participants, those seropositive for H. pylori had lower mean SBP (PD: −2.14 mm Hg, 95% CI −3.91 to –0.37) and HbA1c (PD: −0.04%, 95% CI −0.08 to –0.007) on an absolute scale than seronegative participants. Still, there was no interaction by race and ethnicity for SBP (online supplemental figure 2B), DBP (online supplemental figure 2D) and HbA1c (online supplemental figure 2F, online supplemental table 2).

Binary HTN and T2D

During wave IV, 1183 and 316 Add Health participants had evidence of HTN and T2D, respectively (online supplemental table 17). In multivariable models, H. pylori seropositivity was not associated with prevalent HTN (online supplemental table 3A) or T2D (online supplemental table 3B). There were no differences in strata-specific estimates nor significant interactions for prevalent HTN or T2D by sex, race and ethnicity (online supplemental table 4).

Incidence analyses

Annualised rates of change in SBP, DBP and HbA1c

In wave V, 3174 Add Health participants who also completed wave IV were included in incidence analyses with a median age (IQR) of 37 (36, 39; online supplemental file 1, online supplemental figure 1). H. pylori seropositive participants were less likely to be college graduates and report household income ≥$50 000 than seronegative participants (online supplemental table 5). At wave V, 658 and 200 participants self-reported ever being diagnosed with HTN and T2D, respectively. Furthermore, 346 and 120 participants reported taking HTN and T2D medications, respectively (online supplemental table 5).

Among participants in wave V, 1129 individuals consented and completed a home visit where anthropometric measures and blood spots were collected (online supplemental figure 1). SBP, DBP and HbA1c acquired from these visits were used to calculate annualised rates of change from wave IV to V after a median of 10 years. On average, annualised rates of change in SBP and DBP demonstrated increased BP over time, while HbA1c percentages decreased slightly by wave V (online supplemental table 6). However, the mean annualised rates of change were not different by H. pylori seroprevalence at wave IV in unadjusted and adjusted models (figures2A,D 3A, online supplemental table 6).

Figure 2. Estimated annualised rates of change in systolic blood pressure (SBP) and diastolic blood pressure (DBP) by Helicobacter pylori seroprevalence and subgroup analyses by sex and race and ethnicity. Estimated annualised rates of change in (A) SBP and (D) DBP values by H. pylori seroprevalence from fully adjusted models. Subgroup analyses are presented for estimated SBP and DBP by biological sex (B and E, respectively) and race and ethnicity (C and F, respectively) from fully adjusted models. The boxplots present the median, quartile 1 and quartile 3 for each group distribution. NH, non-Hispanic.

Figure 3. Estimated annualised rates of change in haemoglobin A1c (HbA1c) by Helicobacter pylori seroprevalence and subgroup analyses by sex and race and ethnicity. Estimated annualised rates of change in (A) HbA1c values by H. pylori seroprevalence from fully adjusted models. Subgroup analyses are presented for estimated HbA1c by biological sex and race and ethnicity (B and C, respectively) from fully adjusted models. The boxplots present the median, quartile 1 and quartile 3 for each group distribution. NH, non-Hispanic.

In subgroup analyses, among Hispanic participants, those seropositive for H. pylori infection had a lower annualised rate of change in SBP (PD: −1.04 mm Hg/year, 95% CI −2.05 to –0.02) over time on the absolute scale compared with seronegative participants (figure 2C, online supplemental table 7). However, there was no interaction by biological sex or race and ethnicity for annualised rates of change in SBP, DBP and HbA1c. There was also no difference in stratum-specific estimates by sex for SBP (figure 2B), DBP (figure 2E) or HbA1c (figure 3B) and race and ethnicity for DBP (figure 2F) or HbA1c (figure 3C; online supplemental table 7).

Incident HTN and T2D at wave V

In wave V, 403 (16.8%) participants had incident HTN (online supplemental table 17). The adjusted incidence rate ratio (IRR) between H. pylori seroprevalence and incident HTN was 1.04 (95% CI 0.73 to 1.47; table 2). There was no interaction by sex, race or ethnicity (online supplemental table 8). 153 (5.2%) Add Health participants had incident T2D at wave V (online supplemental table 17). There was no association between H. pylori seropositivity and incident T2D (table 2) in adjusted models. In subgroup analyses, those who self-reported as NH Asian, Pacific Islander, Native American or Other race and ethnicity and were seropositive for H. pylori infection had a decreased rate of incident T2D (IRR: 0.10, 95% CI 0.02 to 0.42) as seronegative participants (online supplemental table 8). However, there was no interaction by sex, race or ethnicity for incident T2D (online supplemental table 8).

Table 2. Final model estimates in association between H. pylori seroprevalence and incident hypertension and type 2 diabetes at wave V.

	Final model (no interaction)		Final+interaction (biological sex)		Final+interaction (race and ethnicity)
	IRR (95% CI)	P value	IRR (95% CI)	P value	IRR (95% CI)	P value
Outcome: incident hypertension
H. pylori seroprevalence
Negative	1 (Reference)		1 (Reference)		1 (Reference)
Positive	1.04 (0.73 to 1.47)	0.83	1.32 (0.82 to 2.12)	0.26	1.32 (0.88 to 1.99)	0.18
Age	1.00 (0.92 to 1.08)	0.94	1.00 (0.92 to 1.08)	0.94	1.00 (0.92 to 1.08)	0.99
Biological sex
Female	1 (Reference)		1 (Reference)		1 (Reference)
Male	1.76 (1.33 to 2.32)	<0.0001	1.93 (1.41 to 2.65)	<0.0001	1.78 (1.35 to 2.36)	<0.0001
Race and ethnicity
NH White	1 (Reference)		1 (Reference)		1 (Reference)
NH Black	1.18 (0.84 to 1.65)	0.35	1.19 (0.85 to 1.67)	0.32	1.38 (0.97 to 1.96)	0.07
Hispanic	0.65 (0.37 to 1.16)	0.15	0.67 (0.38 to 1.19)	0.17	0.79 (0.42 to 1.47)	0.45
Other*	1.58 (1.00 to 2.49)	0.05	1.60 (1.00 to 2.57)	0.05	1.76 (1.05 to 2.95)	0.03
Household income
≥$100 000	1 (Reference)		1 (Reference)		1 (Reference)
$40 000–99 999	1.56 (0.95 to 2.59)	0.08	1.56 (0.94 to 2.58)	0.08	1.57 (0.95 to 2.59)	0.08
≤$30 000	1.67 (1.05 to 2.65)	0.03	1.65 (1.04 to 2.62)	0.03	1.64 (1.04 to 2.60)	0.03
HbA1c	1.08 (0.93 to 1.25)	0.31	1.07 (0.93 to 1.24)	0.33	1.09 (0.94 to 1.26)	0.28
BMI	1.04 (1.03 to 1.06)	<0.0001	1.04 (1.03 to 1.06)	<0.0001	1.04 (1.03 to 1.06)	<0.0001
Interaction (H. pylori_male)*			0.58 (0.28 to 1.21)	0.14
Interaction (H. pylori_NH Black)*					0.47 (0.21 to 1.07)	0.07
Interaction (H. pylori_Hispanic)*					0.38 (0.11 to 1.32)	0.13
Interaction (H. pylori_Other)*					0.57 (0.18 to 1.84)	0.35
Outcome: incident type 2 diabetes
H. pylori seroprevalence
Negative	1 (Reference)		1 (Reference)		1 (Reference)
Positive	0.93 (0.51 to 1.69)	0.80	1.34 (0.70 to 2.58)	0.37	1.06 (0.48 to 2.32)	0.88
Age	1.02 (0.89 to 1.16)	0.78	1.02 (0.90 to 1.16)	0.74	1.02 (0.90 to 1.16)	0.74
Biological sex
Female	1 (Reference)		1 (Reference)		1 (Reference)
Male	1.23 (0.80 to 1.89)	0.35	1.42 (0.87 to 2.31)	0.16	1.21 (0.78 to 1.88)	0.4
Race and ethnicity
NH White	1 (Reference)		1 (Reference)		1 (Reference)
NH Black	1.24 (0.70 to 2.20)	0.45	1.22 (0.69 to 2.15)	0.49	1.30 (0.70 to 2.44)	0.4
Hispanic	1.60 (0.76 to 3.37)	0.21	1.70 (0.82 to 3.51)	0.15	1.51 (0.56 to 4.08)	0.41
Other*	1.89 (0.71 to 5.01)	0.20	1.90 (0.71 to 5.04)	0.2	2.45 (0.89 to 6.73)	0.08
Household income
≥$100 000	1 (Reference)		1 (Reference)		1 (Reference)
$40 000–99 999	1.44 (0.56 to 3.74)	0.45	1.44 (0.55 to 3.73)	0.45	1.47 (0.58 to 3.76)	0.41
≤$30 000	2.78 (1.09 to 7.06)	0.03	2.71 (1.06 to 6.92)	0.04	2.80 (1.11 to 7.05)	0.03
SBP	1.01 (1.00 to 1.03)	0.08	1.01 (1.00 to 1.03)	0.06	1.01 (1.00 to 1.03)	0.08
BMI	1.09 (1.06 to 1.11)	<0.0001	1.09 (1.06 to 1.11)	<0.0001	1.09 (1.06 to 1.11)	<0.0001
Interaction (H. pylori_male)*			0.39 (0.14 to 1.13)	0.08
Interaction (H. pylori_NH Black)*					0.79 (0.24 to 2.55)	0.69
Interaction (H. pylori_Hispanic)*					1.12 (0.19 to 6.63)	0.90
Interaction (H. pylori_Other)*					0.13 (0.02 to 0.82)	0.03

Final model incidence rate ratios, 95% CIs and p values for the linear regression models examining the associations between H. pylori seroprevalence at wave IV and incident hypertension at wave V.

Significant estimates (p values less than the alpha of 0.05) are in bold text.

^*This category includes participants self-classified as non-Hispanic Asian, Pacific Islander, Native American and Other.

BMI, body mass index; HbA1c, hemoglobin A1c; H. pylori, Helicobacter pylori; IRR, incidence rate ratio; NH, non-Hispanic; SBP, systolic blood pressure.

Sensitivity analyses

In a sensitivity analysis that varied the H. pylori seropositivity threshold, there was little difference in estimates across the different cutoffs, except in two subgroup analyses, where CIs gained or lost significance with small changes (online supplemental tables 9–16). These variations emphasise the exploratory nature of the subgroup analyses and highlight that we may not have the power for these analyses. Still, this sensitivity analysis strengthens confidence in the presented associations.

Discussion

In a large representative sample of US adults in young adulthood to early midlife, H. pylori seroprevalence was marginally associated with lower mean SBP. There were no differences in mean DBP or HbA1c or prevalent and incident HTN or T2D. Over 20% of the analysed Add Health participants tested positive for H. pylori antibodies, suggesting that even in early midlife, participants who are primarily non-immigrants continue to be exposed to H. pylori infection. Signs of H. pylori infection were particularly apparent among NH Black and Hispanic participants, those with lower levels of education and income and those who reported having no health insurance.

The prevalence of H. pylori in our study population was 20.8%; this proportion falls within the range of 18.9%–87.7%, as presented in a meta-analysis on the global prevalence of H. pylori.¹ However, the 18.9% estimate is from 175 participants in Switzerland with an average age of 44.⁶⁷ It is notable that the prevalence of H. pylori in this comparatively young cohort, with few recent immigrants, is relatively high, given medical care access to treatment in the USA. Within the USA, Hooi et al used eight studies to arrive at an overall prevalence estimate of 35.6% among mostly older individuals.¹ One study in Texas found that in 96 participants aged 21–40 years, 16% were positive for H. pylori antibodies.⁶⁸ Estimates vary by sociodemographic variables, as seen in our study. Still, the difference in race and ethnicity percentages between the seroprevalence groups was striking, as 77.9% of the seronegative participants were NH White compared with 52.9% of the seronegative respondents. The prevalence of H. pylori is higher among lower socioeconomic regions, which was supported in our study as seropositive participants tended to have lower educational attainment and a lower social origins score, which is a factor score of parental education, occupation, household income and household receipt of public assistance.^{1 2 45}

Prevalent and incident HTN

Studies between H. pylori infection and hypertension have shown mixed results depending on the population, study design and H. pylori detection method.26,2969 These studies have also tended to be cross-sectional and examine middle-aged and older individuals. In our study, H. pylori seropositivity was only associated with lower mean SBP. Subgroup analyses showed some trends in the association between H. pylori seroprevalence and SBP by sex, but the CIs were close to 0. Lower SBP levels among those positive for H. pylori infection have been noted previously.^{27 78} In a cross-sectional community-based study of 10 537 English participants, Harvey et al found that after accounting for age, sex, BMI, smoking, high alcohol intake and use of anti-HTN medications, SBP, but not DBP, levels were lower in patients positive for H. pylori infection from urea breath tests compared with negative patients (PD −1.51, 95% CI −2.47 to –0.55).²⁷ Harvey et al. postulated that although the difference in SBP was significant, it was unlikely to be ‘clinically important’ and may be explained by residual confounding.²⁷

In another study, Kopácová et al found that subjects under 25 years and positive for H. pylori by urea breath test had significantly lower SBP and DBP levels than those negative for H. pylori infection.⁷⁸ Interestingly, there was no difference in SBP and DBP among participants aged 25–64. Participants aged 65 years or older and positive for H. pylori infection had markedly increased SBP and DBP levels compared with participants negative for H. pylori infection. However, they could not replicate these results in their 2014 study, where there was no difference in SBP or DBP levels due to H. pylori infection.⁶⁹

In a cross-sectional study of 5168 participants in China with a mean age of 42.58 years, H. pylori infection, measured by urea breath test, was associated with higher DBP (PD: 0.74, 95% CI 0.10 to 1.37) but not SBP.⁷⁰ Further, H. pylori infection was associated with prevalent HTN in models adjusted for age, sex, clinical biomarkers, T2D and BMI.⁷⁰ Notably, in subgroup analyses by age, there was only an association between H. pylori and prevalent HTN among older (≥42 years) participants. The odds of prevalent HTN were also higher in female than male participants.⁷⁰ Meta-analyses focused on prevalent HTN have stated that H. pylori infection increases the odds of HTN.^{26 28 29} However, these meta-analyses have noted significant heterogeneity among their selected studies and variations by study design and exposure ascertainment.^{26 29}

We noted no association between H. pylori seropositivity and incident HTN at wave V after a median of 10 years of follow-up. In a German cohort of 3307 participants with a mean age of 56.7 years, Wawro et al found that after 6–9 years of follow-up, H. pylori seroprevalence was not associated with incident HTN (RR 1.10, 95% CI 0.86 to 1.42).⁷⁵ Yue et al published a meta-analysis of 55 studies that included incidence studies; they stated that H. pylori infection increased the odds of HTN (OR 1.32, 95% CI 1.15 to 1.52) and that participants with H. pylori infection had higher levels of SBP and DBP than H. pylori-negative participants.²⁸

In our study, H. pylori infection was measured by antibody levels, which can only reflect a prior infection and may allow for false-positive and false-negative participants to be included in the seropositive group. Moreover, we tested for antibodies in blood spots, which may not be as accurate as serum testing. Our study also examined the association with BP in a young cohort with a median age of 28. It may be too soon to identify subclinical markers of HTN, as HTN tends to appear later in life.^{10 14} As Harvey et al noted in their study, residual confounding may also be biasing our prevalence results.²⁷ Notably, this is the first US study examining the association between H. pylori seroprevalence and HTN. Future studies are needed to parse out this association as participants in Add Health move into middle age.

Prevalent and incident T2D

We found no association between H. pylori seropositivity and HbA1c, prevalent or incident T2D. Similarly to HTN, the relationship between H. pylori infection and T2D is controversial.^{3 4 23 24} A meta-analysis of 14 cross-sectional studies noted that H. pylori-positive participants with T2D did not have significantly higher HbA1c levels than H. pylori-negative participants.²⁴ Conversely, in a cohort of Chinese retirees, Xiong et al. determined that H. pylori-positive participants were significantly more likely to have higher fasting plasma glucose levels and self-report prevalent T2D than H. pylori-negative participants.⁷⁴

In the USA, Chen and Blaser used the National Health and Nutrition Examination Survey (NHANES) to demonstrate that H. pylori seropositivity was associated with increased HbA1c after adjusting for age, sex, smoking, education, ethnic background and BMI.⁷⁹ In our study, 316 (6.9%) participants had prevalent T2D at wave IV. While this prevalence agrees with reported estimates from NHANES, it is possible that the sample size was too small to identify a relationship between H. pylori seroprevalence and T2D.⁷⁹ H. pylori seropositive participants had higher levels of HbA1c at wave IV in unadjusted compared with seronegative participants that were attenuated after adjustment. As with HTN, this cohort may also be too young to note subclinical changes in HbA1c levels due to H. pylori infection.

Studies on H. pylori infection and incident T2D have also found mixed results.^{31 72 80} In 782 diabetes-free Latinx individuals >60 years of age, Jeon et al determined that H. pylori-seropositive participants at baseline were at an elevated risk of developing incident T2D.³¹ In a retrospective study of 16 091 Korean participants without T2D at baseline, H. pylori seropositivity was not associated with the cumulative incidence of T2D after 108 614 person-years of follow-up.⁷² Notably, Yu et al found that gastric atrophy or chronic inflammation of gastric mucosa and not H. pylori seroprevalence was associated with incident T2D in a cohort of 1379 T2D-free individuals followed for a mean of 3.4 years in Taiwan.⁸¹

Our study agreed with investigations that found no association between H. pylori seroprevalence and incident T2D. Importantly, H. pylori antibodies were measured from leftover blood spots of wave IV participants during or after wave V was underway. While efforts were made to ensure that these participants had also completed wave V, including a home visit, only about 30% of the participants with wave V data had anthropometric and biological samples. Therefore, incident T2D and HTN at wave V were primarily based on self-reported data in this subset. HTN and T2D are both asymptomatic diseases that can go unnoticed for years without surveillance.^{10 44} In participants without biological markers, we could not identify undiagnosed HTN or T2D. Still, as shown in online supplemental table 17, even among the biosamples, the percentage of H. pylori seropositivity was lower than among the self-reported data (online supplemental file 2). The proportion of H. pylori seropositivity in the composite incident HTN and T2D is slightly closer to the self-reported data. Future studies are needed in this population to continue following trends in the relationship between chronic inflammation and cardiometabolic outcomes.

While there remains disagreement over whether H. pylori infection increases the risk of HTN and T2D, several mechanisms have been presented to connect the factors.34 6 25 26 29 32,36 As a chronic pathogen, researchers hypothesise that H. pylori infection leads to a persistent inflammatory response from the body that may cause long-term damage to the vasculature.32,35 Dyslipidaemia has also been linked to HTN and T2D.2629 82,87 A meta-analysis of 27 studies found that H. pylori infection was positively associated with low-density lipoprotein cholesterol, total cholesterol and triglycerides and negatively associated with high-density lipoprotein cholesterol.⁸⁵ Disruption of the gut microbiome by H. pylori infection is also linked to metabolic syndrome and insulin resistance.^{4 88} We noted no difference in self-reported high cholesterol. Still, participants seropositive for H. pylori had higher BMI than seronegative participants at waves IV and V.

There were several limitations to this study. We measured antibodies to H. pylori and could not ascertain how long the pathogen had been in the body. As H. pylori infection tends to present as asymptomatic, it is rarely captured in a medical record unless there is a reason for the doctor to test for the pathogen. Moreover, although we examined medication records at both waves for any antibiotics used to clear H. pylori, we may have missed eradication therapies among seropositive individuals. This may have affected their risk of incident HTN or T2D at wave V. We also measured H. pylori antibodies from blood spots instead of serum, which is more commonly tested for H. pylori in the literature. This difference in sample type may have contributed to the differences we noted in this study compared with other studies on H. pylori seroprevalence.

There may also be misclassification bias in exposure and outcomes, particularly for H. pylori infection, as ascertainment is based on antibody values. HTN and T2D are partly based on self-reported data, which may also have presented misclassification. To address some of this bias, definitions of HTN and T2D included both self-reported data and results from biosamples. Moreover, blood spots are not as accurate as serum testing; therefore, there may be measurement errors. There is a risk of type 1 errors with multiple comparisons in subgroup analyses; therefore, these results should be interpreted cautiously. There is also the possibility of selection bias between the wave IV cohort and the analytic cohort used in this study, given differences in biological sex, race, ethnicity, education, lifestyle factors and biological markers between the two groups (online supplemental table 18). The inclusion of sampling weights and other survey weighting variables reduced some of this selection bias in the results. Other limitations of this study include having no information on the Cag A gene in H. pylori. This gene signifies a more pathogenic strain of H. pylori.² Our incidence analyses also relied heavily on self-reported data, which can be biased and underestimate the true incidence of HTN and T2D in this population. Finally, there may be uncontrolled and residual confounding that could have biased our results.

The strengths of this study lie in its diversity and life course approach, and it provides the first step in examining how the relationship between H. pylori infection and cardiometabolic outcomes may change over time. Future studies will be able to build on this work by assessing whether the association between H. pylori and cardiometabolic outcomes changes as participants age and if Cag A modifies these associations. Add Health recruited and continues to follow a large and diverse population of individuals in the USA, which allowed assessment of H. pylori infection and chronic disease among a younger cohort. Most studies on cardiometabolic disorder primarily include older participants who are more likely to have other comorbidities, making it hard to focus on BP or HbA1c.^{4 29} Still, because our participants were younger and, therefore, less likely to be symptomatic and diagnosed with these health conditions, it is possible that we were underpowered to detect associations during this period in the life course.

In our study, H. pylori seropositivity was not associated with prevalent or incident HTN, T2D or changes in SBP or DBP over time. Our study highlights the complexity by which chronic inflammation may affect chronic disorders and agrees with studies that have not noted an association between H. pylori and HTN or T2D. Our study examined the interaction between H. pylori and cardiometabolic outcomes at two timepoints in a representative US population to provide clarity on an issue riddled with mixed results. Our results mostly agreed with null findings, which may change as participants age. More studies, particularly in this population, are required to understand the interaction between infectious and chronic disorders and how it can change as people age into mid to late life.

Data availability statement

Data may be obtained from a third party and are not publicly available.