Meta-analysis of the correlation between Helicobacter pylori infection and autoimmune thyroid diseases

Objective This study presents a systematic meta-analysis of the correlation between Helicobacter pylori (H. pylori) infection and autoimmune thyroid diseases (AITD). Materials and Methods Fifteen articles including 3,046 cases were selected (1,716 observational and 1,330 control cases). These data were analyzed using Stata12.0 meta-analysis software. Results H. pylori infection was positively correlated with the occurrence of AITD (OR = 2.25, 95% CI: 1.72–2.93). Infection with H. pylori strains positive for the cytotoxin-associated gene A (CagA) were positively correlated with AITD (OR = 1.99, 95% CI: 1.07–3.70). There was no significant difference between infections detected using enzyme-linked immunosorbent assay (ELISA) and other methods (χ2 = 2.151, p = 0.143). Patients with Grave’s disease (GD) and Hashimoto’s thyroiditis (HT) were more susceptible to H. pylori infection (GD: OR = 2.78, 95% CI: 1.68–4.61; HT: OR = 2.16, 95% CI: 1.44–3.23), while the rate of H. pylori infection did not differ between GD and HT (χ2 = 3.113, p = 0.078). Conclusions H. pylori infection correlated with GD and HT, and the eradication of H. pylori infection could reduce thyroid autoantibodies.


INTRODUCTION
Autoimmune thyroid diseases are familial autoimmune disorders that are more common in women than men. These diseases include Grave's disease (GD), Hashimoto's thyroiditis (HT), atrophic thyroiditis, and subacute lymphocytic thyroiditis (also known as postpartum thyroiditis, PPT), painless thyroiditis (PT), or silent thyroiditis (ST) [1]. The primary pathological features of autoimmune thyroid diseases are thyroid tissue infiltration of lymphocytes and thyroid dysfunction. Other typical hallmarks of these diseases are thyroid autoantibodies such as thyrotropin receptor antibody (TRAb), anti-thyroglobulin antibody (TGAb), and anti-thyroperoxidase antibody (TPOAb) [2]. The production of these autoantibodies can be attributed to both environmental and genetic factors [3]. Iodine overdose is the primary environmental cause [4], while bacterial and viral infections could be other causative factors [5,6]. Cross-reactive antigens can induce superantigen-activated polyclonal T cells, increase expression of human leukocyte antigen in thyroid tissue and promote other autoimmune tolerance responses [7,8]. Helicobacter pylori (H. pylori) infection is one of the most common chronic infections worldwide [9]. H. pylori specifically colonizes in the gastric mucosa, inducing chronic inflammation and promoting chronic gastritis, peptic ulcers, and gastric cancer [10]. H. pylori has been reported to be associated

Meta-Analysis
with other diseases such as diabetes [11], nonalcoholic fatty liver disease [12], iron deficiency anemia [13], and idiopathic thrombocytopenic purpura [14]. The relationship between H. pylori infection and AITD has also recently been explored. In 2013, Shi et al. [15] systematically evaluated 7 studies including 862 patients, observing that H. pylori infection was associated with the development of AITD in patients with GD but not HT. In this systematic meta-analysis, we further investigated the association between H. pylori infection and AITD.

Comparison of H. pylori infection rates
The H. pylori infection rate was reported in 15 studies. As there was heterogeneity among the studies (I 2 = 61.6%, p = 0.001), a random model was established. The H. pylori infection rates among participants in the observation and healthy control groups were 63.17% and 45.41% respectively, indicating that H. pylori infection was correlated with AITD (OR = 2.25, 95% CI: 1.72-2.93, p < 0.001). The relationship between H. pylori infection diagnostic method was further investigated by subgroup analysis. In participants diagnosed using ELISA, the infection rate was 61.64% in the observation group and 45.11% in the control group (OR = 2.28, 95 % CI: 1.47-3.55, p < 0.001). Participants diagnosed using other diagnostic methods had an infection rate of 66.32% in the observation group and 47.62% in the control group (OR = 2.16, 95% CI: 1.53-3.05, p < 0.001). Both subgroup analyses indicated that the infection rate was higher in the observation group than in the control group ( Figure 3). There was no significant difference in the rate of diagnosis via ELISA and other methods according to the χ 2 test (χ 2 = 2.151, p = 0.143).
Due to the observed heterogeneity in statistical results, a sensitivity analysis was performed to investigate the influence of specific research methods. This analysis indicated no significant difference and the conclusions were consistent, indicating that the analysis was stable (Supplementary Figure 1). Meta-regression analysis was performed to determine the heterogeneity in the results, and revealed no heterogeneities among the year and country of publication, sample size, case comparison ratio, or the method used to detect the presence of H. pylori (Supplementary Table 1).

Rates of infection with CagA-positive H. pylori strains
The rates of infection with CagA-positive H. pylori strains were reported in seven studies. Heterogeneity was observed among the rates in these studies (I 2 = 73.7%, p = 0.001). According to random model calculation, the rate of H. pylori infection was 34.70% in the observation group and 18.90% in the control group (OR = 1.99, 95% CI: 1.07-3.70, p = 0.030), indicating that infection with CagA-positive H. pylori strains was associated with AITD ( Figure 5). Consistent with this analysis, there was heterogeneity in the statistical results and sensitivity analysis. Systematic evaluation was carried out in a manner similar to that described previously. By excluding the studies by Sterzl et al. [19] and Shumuely et al. [30], the merged effect significantly changed (OR = 3.35, 95% CI: 2.42-4.64) and the heterogeneity disappeared (I 2 = 0%, p = 0.653), while the conclusion was still consistent (Supplementary Figure 2). Due to the limitations of the retrieved literature, meta-regression analysis was unable to identify the source of heterogeneity [31,32].

Eradication therapy for thyroid autoantibody levels
Five studies [20,24,[27][28][29] reported the influence of eradication therapy on thyroid autoantibodies. Patients with AITD who had H. pylori infection were selected from each study and were randomly allocated to observation and control groups. Only patients in the observation group were treated with eradication therapy. Therefore, meta-analysis could not be performed based on the data provided. The descriptive analysis is shown in Supplementary Table 2.

DISCUSSION
The correlation between AITD and H. pylori infection was studied in 15 publications reporting a total of 3,046 cases. Unlike a previous meta-analysis published by Shi et al. in 2013 [15], which included unhealthy individuals in the control group, and a previous metaanalysis by Luis et al. in 1998 [33], which included

23). GD and HT
share similar pathogenic factors, pathology, biochemistry, and clinical features such as thyroid tissue lymphocyte infiltration resulting in inflammation and TPOAb, TGAb, or increased levels of other thyroid autoantibodies [40]. Thus, H. pylori infection is expected to be associated with GD and HT, which is consistent with the results of our meta-analysis. We also observed that pharmaceutical eradication of H. pylori infection reduced levels of thyroid autoantibodies in patients with GD and HT [41]. Patients with AITD who had dysfunctional gastric acid secretion required treatment with higher thyroid hormone levels, indicating that normal gastric acid secretion was necessary for the effective absorption of oral thyroxine [35]. These studies suggest that H. pylori infection was associated with the pathogenesis and development of AITD. Our literature review identified no articles reporting the association between atrophic thyroiditis and H. pylori infection, while the conclusions from other studies are controversial. Due to the limited number of publications and limited data, the correlation between atrophic thyroiditis and H. pylori infection was not analyzed here.
Shi et al. previously reported that H. pylori infection diagnosed by ELISA was not associated with AITD. However, we found that regardless of diagnostic test, H. pylori infection was robustly associated with AITD. Although there was great heterogeneity in the merged effects, meta-regression analysis showed that the varieties of diagnostic tools were not responsible for this heterogeneity. Shi et al. attributed the lack of relationship between ELISA-diagnosed H. pylori infection and AITD to the inability of this assay to distinguish active from resolved infection, and thus included participants with resolved infection. However, most of the studies that we included did not report a history of AITD. As AITD is a chronic process, H. pylori infection may occur at any stage, which could be either the cause or the consequence of patients susceptible to the infection. Shi et al.'s and our meta-analysis evaluated only whether H. pylori infection was associated with AITD without considering whether active infection had any effect on the results.
In conclusion, H. pylori infection is associated with AITD. Our results suggest that patients with AITD are more susceptible to H. pylori infection, especially infection with CagA-positive strains of H. pylori. The high heterogeneity could be influenced by several factors. First, only the abstracts of some included publications were available in standard databases such as PubMed, Web of Science, Ovid Online, etc. [16,19], limiting the extraction of patient information. In addition, H. pylori infection could also be associated with other nongastrointestinal diseases, such as diabetes or autoimmune diseases, but we did not exclude these cases from our analyses [18,22,25,30]. Additionally, in some studies antibiotic use was not ruled out [18,26]. These three factors may contribute to the heterogeneity of these studies. Additionally, although the funnel plot suggests no bias, we cannot ignore the potential language or regional bias as we included only manuscripts published in English and Chinese.

Inclusion criteria
Articles or publications including the following information were selected: (1) Case-control, cohort, or cross-sectional studies in which the number of individuals in the observation group, control group, and number of individuals infected with H. pylori, were in either Chinese or English in the abstract or full-text of the publication. (2) The observation groups were diagnosed with AITD. GD was diagnosed according to the following criteria: hyperthyroidism, decreased TSH, increased FT3 and FT4, diffuse enlargement of thyroid tissue, and antibodies against TSH receptor (TRAb) and/ or TPOAb-and TGAb-positivity. HT was diagnosed according to the following criteria: hypothyroidism, increased TSH, decreased FT3, FT4, TPOAb-and TGAbpositivity, typical ultrasonographic features, and evaluation by fine needle aspiration thyroid cytology tests. Atrophic thyroiditis was diagnosed according to the following criteria: hypothyroidism, increased thyroid autoantibodies, and thyroid atrophy features by ultrasound evaluation. PPT was diagnosed according to the following criteria: normal thyroid function before or during pregnancy, abnormal thyroid function within one year after childbirth or abortion, no thyroid pain, low rate of iodine uptake and high blood TPOAb levels, and ultrasound examination revealing diffuse or nodular swelling of the thyroid gland. PT or ST were diagnosed according to the following criteria: sudden onset hyperthyroidism without pain, low iodine uptake rate and high blood TPOAb, with remission or development of permanent hypothyroidism in 6-12 months, and an ultrasound examination revealing diffuse swelling of the thyroid gland. Control groups included healthy individuals without any autoimmune disease. (3) H. pylori infection was determined by at least one of the following diagnostic methods: stool antigen test, urea breath test, ELISA, or Western blotting.

Exclusion criteria
Articles or publications were excluded where (1) Insufficient information was provided to support the association of H. pylori infection with AITD in the full text publication or abstract where full text manuscripts were not available; (2) The same data was reported in several publications; (3) The study did not include a control group.

Search strategy
The following databases were searched for articles published before July 2017: PubMed, Medline, Web of Science, OVID, SINOMED, VIP Chinese Science and Technology Journal Full-text Database, Wanfang Data Resources, and CNKI full-text database. English and Chinese search terms included Helicobacter pylori, H. pylori, autoimmune thyroid disease, autoimmune thyroiditis, lymphocytic thyroiditis, Grave's disease, Hashimoto's thyroiditis, atrophic thyroiditis, postpartum thyroiditis, painless thyroiditis, and silent thyroiditis. Where the same data was analyzed by multiple studies, the most recent results were used.

Quality assessment and data extraction
The quality of the retrieved articles was evaluated independently by two researchers. In case of any disagreement between the researchers, the final decision was made by a third researcher. The quality assessment was followed by the standard "NOS" evaluation recommended by Cochrane [42], including 3 items and 8 articles (population selection, comparability, exposure evaluation, and outcome evaluation) with a maximum score of nine. The results included diagnostic criteria, the case number of both H. pylori and CagA-positive   infections, the sample numbers in observation and control groups, and autoantibody titer after pharmaceutical eradication of H. pylori infection.

Statistical process
Heterogeneity analysis was performed using Stata12.0 processing software. If heterogeneity (I 2 > 50%, p < 0.10) was found between studies, a random model was used for further calculation. In the case of no heterogeneity, the fixed effects model was used. The results were presented as the odds ratio (OR) and confidence interval (95% CI). Meta-regression analysis was used to identify heterogeneity sources. The stability of the meta-analysis was assessed using sensitivity analysis. If the data provided could not be metaanalyzed, only qualitative analysis by description was provided. An χ2 test was performed using SPSS16.0 to assess differences between subgroups. The funnel plot and Begg's rank correlation method were used to identify biases in the data.

Author contributions
Yi Hou and Wen Sun carried out the systematic evaluation and drafted the manuscript. Chengfei Zhang and Tieshan Wang carried out quality assessment and data extraction. Xuan Guo participated in the quality assessment and data extraction. Lili Wu and Lingling Qin participated in the design of the study and performed the statistical analysis. Tonghua Liu conceived the study, participated in its design, and coordinated and helped draft the manuscript. All authors read and approved the final manuscript.