Introduction
It is estimated that more than 50% of the world’s population are infected with
Helicobacter pylori, a gram-negative bacterium primarily colonising the human stomach [
1]. The harmful effects of
H. pylori infection on several gastrointestinal diseases have been well established [
2]. An increasing number of studies have also revealed that
H. pylori infection may have a wider ranging impact on health through its extra-gastrointestinal effects [
3]. Specifically, this includes an association between
H. pylori infection and diabetes [
4‐
6], a harmful metabolic disease with rapidly increasing prevalence worldwide, particularly over the past 30 years [
7].
A large number of published studies, reviews and book chapters have reported associations between
H. pylori infection and serum or gastric concentrations of various hormones [
2,
8‐
10], such as insulin, glucagon-like peptide 1, leptin, ghrelin, gastrin and somatostatin, some of which are involved in glucose metabolism directly or indirectly [
11‐
17]. These studies provided the preliminary pathophysiological evidence for the potential connection between
H. pylori infection and diabetes risk. Several studies have reported direct evidence for the association between
H. pylori infection and diabetes risk [
4‐
6]. However, most of these studies were based on relatively small sample sizes using either a cross-sectional or case–control design, with inconsistent results.
Whether H. pylori infection is indeed associated with diabetes risk is critically important and relevant to clinical decisions regarding future treatment of H. pylori as well as diabetes prevention strategies. Therefore, the purpose of this study was to examine the association between H. pylori infection and the long-term risk of developing diabetes based on a prospective cohort study with a relatively large sample size and multiple H. pylori test results.
Methods
Statistical analysis
Baseline characteristics of the participants are described as mean (SD) for continuous variables with a normal distribution, median (interquartile range, IQR) for continuous variables with a skewed distribution and percentages for categorical variables. Baseline characteristics of the participants with and without H. pylori seropositivity were compared using the t test, Wilcoxon rank test or χ
2 test.
To examine the association between
H. pylori seropositivity and diabetes risk (2002–2012) or persistent
H. pylori seropositivity and diabetes risk (2007–2012), modified Poisson regression was performed to estimate RR and robust standard errors to estimate the 95% CIs [
21]. Univariate analysis was performed first. Then, we performed a sex- and age-adjusted model (Model 1). We added further possible risk factors of diabetes to evaluate the association (Model 2), including BMI (continuous variable), HDL-C (continuous variable), triacylglycerol (< 1.7 or ≥ 1.7 mmol/l), FBG (continuous variable), systolic blood pressure (per 5 mmHg, continuous variable) and lipid-lowering medication (no or yes). Hs-CRP (< 3 or ≥ 3 mg/l), a sensitive biomarker of inflammation, reported to be associated with high risk of diabetes [
22], was added in Model 3. Moreover, level of education was also included in the model (Model 4), since socioeconomic status was an important factor for
H. pylori infection, and level of education was associated with socioeconomic status.
Since participants did not undergo OGTT at baseline, the study may have included participants with diabetes only detectable by OGTT, particularly among participants with impaired fasting glucose (IFG) [
23]. Thus, we further excluded participants with IFG (FBG > 6.1 mmol/l) from the sensitivity analysis.
Subgroup analyses were performed using baseline (2002) characteristics, including age (< 60 or ≥ 60 years), sex (men or women), BMI (< 25 or ≥ 25 kg/m
2), HDL-C (< 1.0 or ≥ 1.0 mmol/l), triacylglycerol (< 1.7 or 1.7 mmol/l), Hs-CRP (< 3 or ≥ 3 mg/l), hypertension (no or yes), FBG (< 5.6 or ≥ 5.6 mmol/l) and lipid-lowering therapy (no or yes) in a multivariable adjusted Poisson regression model. Alongside the grouping factors, other factors were used as adjustment variables. In addition, we performed further subgroup analyses using different baseline information (2002 and 2007) for the 5 year association between
H. pylori and risk of diabetes. RRs between subgroups were compared using
Z-test [
24].
Statistical analyses were conducted by SAS 9.2 (SAS Institute, Cary, NC, USA). Two-tailed p values < 0.05 were considered statistically significant.
Discussion
In this prospective cohort study, we carefully examined the association between H. pylori seropositivity and risk of developing diabetes, based on reliable measurements of antibody against H. pylori and a clear diagnosis of diabetes. We determined that H. pylori seropositivity was associated with lower risk of developing diabetes. To our knowledge, this is the first cohort study to evaluate the association between persistent H. pylori seropositivity and risk of developing diabetes and the first study to detect an inverse association between these factors.
The literature regarding the association between
H. pylori and diabetes risk is inconsistent, with reports of positive, null and negative associations [
4‐
6]. A published meta-analysis systematically reviewed the association between
H. pylori infection and diabetes based on observational studies and drew the conclusion that there was a trend toward more frequent
H. pylori infection in individuals with diabetes [
4]. However, cross-sectional or case–control studies could not determine the causality between
H. pylori infection and diabetes. We are aware of only one prospective cohort study evaluating the association between
H. pylori seropositivity and risk of diabetes [
6]. This study derived data from the Sacramento Area Latino Study on Ageing and found that
H. pylori seropositivity was associated with increased risk of incident diabetes (HR 2.7 [95% CI 1.1, 6.6]). However, the characteristics of our study population differed from this study in that their participants were older (60 to 101 years old) and the
H. pylori seropositivity rate was much higher than ours (91.9% vs 57.9%).
Our finding of an inverse association between
H. pylori seropositivity and diabetes risk was of physiological relevance. Several studies have reported that
H. pylori seropositivity was associated with different hormones [
2,
8‐
10], most of which were involved in glucose metabolism directly or indirectly [
11‐
17]. To our knowledge, at least three hormones supported our results, including gastrin, leptin and ghrelin. Studies have reported that long-term gastrin treatment resulted in improved metabolic control and exerted proliferative effects on pancreatic beta cells [
11]. Leptin has been reported to reduce hyperglycaemia in rodent models of type 1 diabetes and has recently been shown to normalise fasting plasma glucose concentrations in a rodent model of polygenic obesity and type 2 diabetes [
14]. In contrast, ghrelin has been shown to upregulate systemic glucose levels in both humans and rodents [
16]. Thus, studies reporting elevated gastrin and leptin and decreased ghrelin levels in individuals infected with
H. pylori support our findings. However, the regulation of these hormones is intricate. Therefore, the association between
H. pylori infection, hormones and diabetes requires further studies.
H. pylori has co-existed with humans for at least 58,000 years [
25], and only 10% of asymptomatic infected individuals develop gastrointestinal disease during their lifetime [
1]. Regarding the view of co-adaptation of humans and microorganisms, we might reconsider our definitions of
H. pylori and perhaps recognise it as a normal member of the human gastric microbiome [
26]. The Kyoto global consensus on
H. pylori gastritis, published in 2015, strongly recommended that all individuals infected with
H. pylori should be offered eradication therapy, unless there were competing considerations [
27]. If
H. pylori infection could indeed affect diabetes risk, we should reconsider this strategy. More studies are needed to explore the association between these factors.
Several limitations associated with the current study deserve mention. First, 45 participants died and 219 were lost at follow-up between 2002 and 2012, which raised a consideration for the competing risk. To test each extreme situation, we analysed the association between
H. pylori infection and risk of developing diabetes assuming that all of the participants who died or were lost at follow-up did not develop diabetes and then again, but assuming that all did develop diabetes (ESM Table
5). In both of these analyses,
H. pylori seropositivity was still associated with lower risk of diabetes. Therefore, we believe that loss and death in our study have little effect on our findings. Additionally, time-to-event analysis could not be performed in our study as we only collected information regarding the development of diabetes at 5 yearly intervals. Furthermore, we did not collect the information about
H. pylori eradication treatments in our study. However, as we knew, the awareness rate of
H. pylori was very low even among individuals with gastrointestinal symptoms in China [
28], so we could speculate that the treatment rate could also be very low. In addition, we detected
H. pylori in both 2002 and 2007 and found that the serostatus changed in only a few participants. To some extent, multi-point detection might better explain the long-term relationship between
H. pylori and humans. In addition, we did not specifically measure the concentrations of antibodies against cytotoxin-associated gene A (CagA), a major factor influencing the virulence of
H. pylori, in our population. A study has reported that FBG levels were significantly higher in uninfected mice compared with
H. pylori-infected mice, regardless of the strain of
H. pylori utilised [
29]. However, only mice infected with a Cag pathogenicity island (PAI)-negative
H. pylori strain, but not with an isogenic Cag PAI-positive strain, showed improvements in glucose tolerance. Therefore, further studies are still needed to investigate the association between antibodies against CagA and glycometabolism.
In summary, these findings provide new direct evidence for the association between H. pylori infection and diabetes risk. Analysis of the association between H. pylori and diabetes is of great public health and clinical significance given the high prevalence of H. pylori infection and significant burden of diabetes. Thus, clarification of the association between these factors might influence clinical decisions regarding future treatment of H. pylori infection as well as diabetes prevention strategies. However, extrapolation of this study requires verification by other prospective studies and clinical trials, and the underlying mechanism warrants further investigation.