Helicobacter pylori Infection Is Associated with Type 2 Diabetes, Not Type 1 Diabetes: An Updated Meta-Analysis

Background Extragastric manifestations of Helicobacter pylori (H. pylori) infection have been reported in many diseases. However, there are still controversies about whether H. pylori infection is associated with diabetes mellitus (DM). This study was aimed at answering the question. Methods A systematic search of the literature from January 1996 to January 2016 was conducted in PubMed, Embase databases, Cochrane Library, Google Scholar, Wanfang Data, China national knowledge database, and SinoMed. Published studies reporting H. pylori infection in both DM and non-DM individuals were recruited. Results 79 studies with 57,397 individuals were included in this meta-analysis. The prevalence of H. pylori infection in DM group (54.9%) was significantly higher than that (47.5%) in non-DM group (OR = 1.69, P < 0.001). The difference was significant in comparison between type 2 DM group and non-DM group (OR = 2.05), but not in that between type 1 DM group and non-DM group (OR = 1.23, 95% CI: 0.77–1.96, P = 0.38). Conclusion Our meta-analysis suggested that there is significantly higher prevalence of H. pylori infection in DM patients as compared to non-DM individuals. And the difference is associated with type 2 DM but not type 1 DM.


Introduction
Helicobacter pylori (H. pylori) is a gram-negative spiral bacterium, colonized in the stomach. Approximately one-half of the population over the world is infected with H. pylori [1]. Many researches have proved that H. pylori infection is highly associated with gastrointestinal diseases such as chronic gastritis, peptic ulcer disease, gastric cancer, and mucosa-associated lymphoid tissue (MALT) lymphoma since its discovery [2]. In addition, extragastric disorders associated with H. pylori infection, such as cardiovascular diseases and metabolic syndrome, have been revealed and some of them were characterized by persistent and lowgrade systemic inflammation [3]. Inflammation has been demonstrated to play an important part in the pathogenesis of diabetes mellitus (DM), especially type 2 DM (T2DM) [4]. On the other hand, Kondrashova and Hyöty reviewed that some microbes served as the risk factor participating in the trigger and the development of type 1 DM (T1DM), but some microbes such as H. pylori served as a protective factor by lowering the risk of T1DM [5]. Above all, H. pylori infection was a factor not negligible in the process of DM.
Since Simon et al. firstly reported the association between H. pylori infection and DM [6], many studies were carried out. Several case-control studies have reported a higher prevalence of H. pylori infection in DM patients [7,8]. Some cross-sectional researches also revealed a significant correlation between H. pylori infection and diabetes [9][10][11]. Moreover, a meta-analysis carried out by Zhou et al. suggested a trend toward more frequent H. pylori infection in DM patients, especially in T2DM patients [12]. However, Tamura et al. found a significantly higher DM prevalence among individuals with H. pylori infection than those without, but the difference could be mostly ascribed to older age [13]. And some studies argued that no difference in the prevalence of H. pylori infection was found between DM and non-DM individuals [14,15]. Overall, this subject remains controversial now.
The present updated meta-analysis was conducted to answer if there is a difference in the prevalence of H. pylori infection between DM and non-DM individuals. Subgroup analyses were carried out based on the types of DM, geographical regions, and methods for H. pylori detection to further investigate the relationship between H. pylori infection and DM.

Search Strategy and Selection Criteria.
Published guidelines for conducting meta-analyses were followed [16]. We searched PubMed, Embase databases, Cochrane Library, Google Scholar, Wanfang Data (Chinese), China national knowledge database (Chinese), and SinoMed (Chinese) for all relevant articles reported from January 1996 to January 2016, with combinations of the search terms "Helicobacter pylori," or "H. pylori," or "Campylobacter pylori," or "C. pylori," and "diabetes mellitus," or "diabetes," or "type 1 diabetes," or "type 1 diabetes mellitus," or "type 2 diabetes" or "type 2 diabetes mellitus".
To be eligible for inclusion, studies had to meet the following criteria: (1) they were published studies which reported H. pylori infection in DM individuals and non-DM individuals (individuals without DM, impaired glucose tolerance, or impaired fasting glucose); (2) detailed data of H. pylori infection rate in both groups was provided. Studies that did not meet the inclusion criteria were not enrolled.
Studies were excluded if they were as follows: (1) duplicate publications; (2) case report, review, meta-analysis, or guideline; (3) not reporting clinically relevant outcomes; and (4) not providing enough details.

Data Extraction and Quality
Assessment. Data were extracted by one investigator, verified by another investigator, and recorded in a well-designed form developed for this study. The data items included authors, year of publication, country, study design, methods of H. pylori detection, strains of H. pylori, types of DM, age, and sample size. The Newcastle-Ottawa scale (NOS) scoring system was used to assess the quality of the studies [17].

Statistical Analysis.
To obtain pooled effect estimates, the random effects model or fixed effects model was used for meta-analysis, according to the heterogeneity among studies. If there was no statistically significant heterogeneity (twotailed P value >0.05) among the pooled studies, the fixed effect model would be applied; otherwise, the random effect model would be applied [18]. Odds ratio (OR) with 95% confidence interval (CI) was used for the case-control and crosssectional studies, while risk ratio (RR) was for the cohort studies. The presence of between-study heterogeneity was estimated using Q-test and I 2 statistics. Sources of heterogeneity were explored by conducting subgroup analyses based on types of DM, geographical regions, and methods of H. pylori detection. The two-sided tests with significance level of 0.05 were conducted in pooled analyses and subgroup analyses using RevMan software (Version 5.3 for Windows, Cochrane Collaboration, Oxford, UK). Publication bias was evaluated graphically by the funnel plots and statistically by Begg's test and Egger's test with the STATA software (Version 14.0; STATA Corporation, College Station, TX, US). Pr and P value less than 0.05 were considered representative of no statistically significant publication bias. If publication bias was indicated, the trim and fill method procedure was performed to identify and correct the publication bias [19]. The basis of the method was to (1) "trim" (remove) the studies causing funnel plot asymmetry, (2) use the trimmed funnel plot to estimate the true "centre" of the funnel, and then (3) replace the removed studies and their missing "counterparts" around the centre (filling). An estimate of the number of missing studies was provided; an adjusted OR is derived by performing a meta-analysis including the filled studies.

Description of Studies.
A total of 783 studies were retrieved from PubMed, Embase databases, Cochrane Library, Google Scholar, Wanfang Data (Chinese), China national knowledge database (Chinese), and SinoMed (Chinese). According to the criteria for inclusion and exclusion, 79 studies were included in this meta-analysis ( Figure 1). The included study characteristics were summarized in Table 1. All of the articles were qualified to be pooled with quality score of NOS over 5. 76 studies were either casecontrol or cross-sectional studies, and 3 were prospective cohort ones.
A total of 57,397 individuals were enrolled in these studies, with a total H. pylori infection prevalence of 49.7% (28,542/57,397). The pooled H. pylori infection rate was 54.9% (9434/17,187) in DM group and 47.5% (19,108/ 40,210) in non-DM group. The OR was 1.69 (95% CI: 1.47-1.95, P < 0 001) for the two groups. There was high heterogeneity among the studies (I 2 = 86%). The forest plot for pooled prevalence is showed in Figure 2. Each study was sequentially removed from the analysis, and the adjusted ORs (1.63-1.73) were approximate to the initial ones. Especially, the study of Han et al. [20] recruited a total of 6395 patients in DM group and 24,415 in non-DM group, which accounted for nearly one-third of the enrolled individuals in this analysis. However, after removing the data of Han et al. and re-analyzing, the adjusted odds (OR = 1.71) and heterogeneity (I 2 = 83%) were still approximate to the initial ones in spite of its overweight scale.
3.2. Subgroup Analysis. We found a significant association between H. pylori infection and DM but the pooled analysis was with high heterogeneity (I 2 = 86%). Subgroup analyses based on the types of DM, geographical regions, and methods for H. pylori detection were conducted to detect the sources of heterogeneity.
(1) Types of DM 12 studies with 3175 individuals were assigned to the T1DM subgroup, while 42 studies with 41,684 individuals were to the T2DM subgroup. No significant difference was found between T1DM group and non-DM group in H. pylori infection rate (OR = 1.23, 95% CI: 0.77-1.96, P = 0 38; Figure 3). On the contrary, the pooled data indicated that the prevalence of H. pylori infection in T2DM was significantly higher than that in non-DM group (OR = 2.05, 95% CI: 1.67-2.52, P < 0 001; Figure 3). Each study including the study by Han  America, and group Africa, respectively. No significant difference of H. pylori infection rate between DM and non-DM individuals was found in group America and group Africa (P = 0 36 for America; P = 0 38 for Africa). However, in group Asia and group Europe, significantly higher H. pylori infection rate was detected in DM individuals (OR = 2.04 and OR = 1.40, resp.). But there was still high heterogeneity within these subgroups (I 2 = 68%-90%; Figure 4).
(3) Methods for H. pylori detection Methods for H. pylori detection displayed different power in accuracy, which consequently might affect the detection rate of H. pylori infection. Methods for diagnosis of H. pylori were classified as invasive tests and noninvasive tests [21]. Invasive tests included rapid urease test, histology, and culture, and the noninvasive tests included 13 C or 14 C urea breath test, stool antigen detection, and serological approaches for antibodies of H. pylori. For the serological tests of anti-H. pylori IgG or/and IgA antibody in serum, high rates of false-positive results may happen and they cannot identify the differences between the current infection and past infection [21,22]. So we typically sorted the studies with detection method of serological test into one subgroup and others into the other subgroup as they could identify the current infection precisely.
The studies of current infection group comprised of 51 articles and showed a significant higher prevalence of H. pylori infection in DM patients as compared to that in non-DM individuals with OR = 1.92 (95% CI: 1.57-2.34, P < 0 001). Similarly, by enrolling 21 articles in serological test group, we found that the infection rate was 53.7% (1956/3640) in DM group while 46.4% (4097/8829) in the non-DM one (OR = 1.40, 95% CI: 1.10-1.79, P < 0 001; Figure 5). The heterogeneities in both groups were high among studies with I 2 = 89% and I 2 = 81%, respectively ( Figure 5).  but a significant bias was detected by Egger's test with P < 0 001 ( Figure 7). As Egger's test indicated the possibility of publication bias, the trim and fill method procedure was performed to identify and correct the publication bias. There was 14 hypothetical missing studies indicated by the trim and fill procedure, and the imputed pooled estimate was 1.366 (95% CI: 1.181-1.580, P < 0 001). There still existed a statistically significant association between H. pylori infection and DM after adjusting for the publication bias, which suggested that our result was credible. Adjusted funnel plot by the trim and fill method was symmetrical and shown in Figure 8.

Discussion
DM is a chronic disease characterized by a long-term inflammation mechanism. Guo Figure 2). Moreover, we explored more databases and recruited 25 studies reported in Chinese with high-quality score of NOS (all of them were >5). In addition, in subgroup analysis, we found no significant difference in prevalence of H. pylori infection   We found that there existed an association between H. pylori infection and DM in this meta-analysis. Several possible mechanisms might explain the association.
Hyperglycemic condition in diabetic individuals could result in immune dysfunction, including damage to the neutrophil function, depression of antioxidant system, and impaired humoral immunity [25]. Moreover, abnormal enteric neuropathy caused by high blood sugar can modulate immune-cell function and stimulate proinflammatory cytokine production, resulting in neurodegeneration [26]. It leads to delay gastric emptying and lacking of acid secretion, which promotes bacterial colonization or overgrowth in gastrointestinal tract [27]. On the other hand, H. pylori infection in diabetic patients may worsen glycemic control [28], which leads to the difficulty of DM treatment, forming the vicious circle.
gondii did not show an increased rate of DM. It indicated that H. pylori infection might play an unknown role in the pathogenesis of DM, which implicated a potential step for preventing DM by eradication of H. pylori infection. Moreover, it also suggested that other pathogens such as cytomegalovirus and herpes simplex virus 1 might not have the similar effect on the DM like H. pylori. But our meta-analysis just revealed the association between H. pylori and DM, but could not suggest the effect of H. pylori on DM pathogenesis. More researches are needed to find out the actually effect of H. pylori infection on DM.
In subgroup analysis based on the types of DM, we demonstrated that 56.5% T2DM individuals were infected with H. pylori, but only 36.2% T1DM carried the bacterium (Figure 3). T2DM was more significantly prone to the infection of H. pylori. As to T2DM, insulin resistant (IR) is one of its characteristics. Aydemir et al. showed that IR was significantly higher in H. pylori infection group [33]. And Eshraghian et al. also supported that H. pylori infection was a risk factor for IR [34,35]. Furthermore, it was reported that IR in T2DM patients could be improved after successful eradication of H. pylori [4]. It might partly explain the higher H. pylori infection rate in T2DM patients. On the other hand, we found no significant difference in prevalence of H. pylori infection in comparison between T1DM patients and non-DM people (P = 0 38), consistently with the report by Candelli et al. [27]. Whether this outcome is caused by the different pathogenesis or the onset age of T1DM and T2DM remains unclear. In the T1DM group, the mean age in most studies was not over 20, except for the studies of De Block et al. [36] and Sfarti et al. [37], while in T2DM group, the mean age was usually over 50 years old (Table 1). Epidemiological studies suggested that the prevalence of H. pylori infection increases with age [34]. As T1DM mainly onsets during childhood or young age, T1DM patients probably have less chance to be exposed to H. pylori infection. Consistently, Krause et al. showed a significantly lower positive rate of antibodies against H. pylori in T1DM patients [38]. But some studies held the contrary view that T1DM individuals were also prone to H. pylori infection [39,40]. However, our meta-analysis with pooled estimate favored that T2DM rather than T1DM was associated with H. pylori infection. But the sample size of T1DM subgroup was not as large as that of T2DM. Larger sample size is needed to further verify the association between H. pylori infection and DM, especially T1DM.
The prevalence of H. pylori infection varies in different regions. We found significant higher H. pylori infection rate among DM individuals in group Asia and group Europe but not in group Africa or group America (Figure 4). Firstly, it was to be noted that there were much bigger sample size in group Asia and group Europe, respectively. This might be due to the more accurate detection methods and in group Africa and group America; the sample size might be too small to draw robust conclusion. Secondly, it might be explained by that the condition of medical care in developing countries from group Asia was too poor for DM patients to get good control of DM and prevent infectious complications. On the other hand, the epidemiology and different strains of H. pylori infection might attribute to the part of the result. Epidemiology studies revealed that almost all the Asians are infected with the strain of H. pylori carrying cytotoxinassociated gene A (CagA) but only nearly 60% of western people carried this stain [41,42]. It was reported that H. pylori infection in Asians was predominated by CagA iceA1 genotypes while Americans and Africans by CagA iceA2 genotypes [41,43]. CagA is a major virulence factor of H. pylori and has been reported to be associated with diabetic complications [44]. CagA-positive strain of H. pylori could cause poor glycemic control in T2DM and difficulty in eradication, which might result in the visible H. pylori effect among Asian but not African DM patients. However, due to the lack of data, we could not carry out the subgroup analysis based on different strains of H. pylori.
A number of testing methods are available for H. pylori detection. Serological test, namely, anti-H. pylori IgG and/ or IgA test, is not affected by acid suppression therapy or recent antibiotic use. But seropositivity could not confirm current H. pylori infection, and anti-H. pylori IgG titre usually remains elevated for a long period even after clearance or eradication. Some study using anti-H. pylori IgG as the diagnosis of H. pylori infection might overestimate the infection rate. We typically conducted the analysis of serological test group and current infection group and found that in both subgroups, DM patients had higher prevalence of H. pylori infection than non-DM people ( Figure 5). As a result, the association between H. pylori infection and DM was verified despite of different methods for H. pylori detection. Despite the robust result, there existed limitations in our study. The studies were highly heterogeneous. Variables like age, sex, race, economic status, DM prevalence, and strains of H. pylori infection in the included studies varied. For the lack of enough detailed data, subgroup analysis stratified by age, sex, different stages of DM, and strains of H. pylori, which might bring up heterogeneity, could not be carried out. Furthermore, most of the articles meeting the inclusive criteria were case-control or cross-sectional ones, and only 3 were prospective ones. More well-designed and prospective cohort studies are needed for clarifying the association between H. pylori infection and DM.
In conclusion, despite the limitations, our meta-analysis suggested that there is significantly higher prevalence of H. pylori infection in DM when compared with the non-DM individuals. And the difference is associated with type 2 DM but not type 1 DM.

Conflicts of Interest
The authors declare that they have no conflict of interest.

Authors' Contributions
Jun-Zhen Li and Jie-Yao Li contributed equally to this work.