Helicobacter ISSN 1523-5378 doi: 10.1111/hel.12234
REVIEW ARTICLE
A Meta-Analysis of the Association between Helicobacter pylori Infection and Risk of Coronary Heart Disease from Published Prospective Studies Jing Sun, Pooja Rangan, Srinidhi Subraya Bhat and Longjian Liu Department of Epidemiology and Biostatistics, School of Public Health, Drexel University, Nesbitt Hall, 3215 Market St., Philadelphia, PA 19104, USA
Keywords Meta-analysis, Helicobacter pylori, cardiovascular risk factor. Reprint requests to: Longjian Liu, Department of Epidemiology and Biostatistics, School of Public Health, Drexel University, Nesbitt Hall, 3215 Market St., Philadelphia, PA 19104, USA. E-mail: [email protected]
Abstract Background: The association between helicobacter pylori (Hp) infection and coronary heart disease (CHD) has long been debated, and the results from previous meta-analysis are varied. Aims: The aim for this study was to identify the association between Hp and CHD using published perspective cohort studies. Materials and Methods: A systematic review and meta-analysis were performed on studies published from January, 1992 to April, 2014. All studies included used data from prospective cohort studies of CHD events or CHD deaths. Random effect models were applied in all estimations. Results: H. pylori infection increased the risk of CHD events by 11% (19 studies, n = 22,207, risk ratio (RR) = 1.11, 95% confidence interval (CI): 1.01–1.22). This effect was greater for studies that had less than 5 years’ follow-up time (RR = 1.15, 95% CI: 1.00–1.32). However, this effect was not significant for studies that had follow-up times ≥10 years (n = 5100, RR = 1.04, 95% CI: 0.87–1.24). Neither Cag-A seropositive nor Cag-A seronegative strains of H. pylori were associated with a significantly increased risk of CHD events or deaths based on the current published data. Conclusion: In conclusion, H. pylori infection increased the risk of CHD events, especially in a patient’s early life, but this association was weaker or might be masked by other CHD risk factors in long-term observations.
Atherosclerosis is the primary cause of coronary heart disease (CHD), although other factors could play a role in the development of the disease [1]. Previous studies have suggested that some microorganisms and viral infections, including Helicobacter pylori, chlamydia pneumonia, and cytomegalovirus, also contribute to the development of CHD or have been identified as stimuli contributing to ischemic events [2–5]. Helicobacter pylori (H. pylori) infection has long been associated with gastroduodenal lesions, and such chronic inflammation may trigger an immune response. In previous studies, anti-H. pylori immunoglobulin G (IgG) had been associated with an increased risk of CHD [6,7]. Several studies have also observed that CHD patients with H. pylori infection also have a higher prevalence of positive cytotoxin-associated gene-A (Cag-A) antibodies [8,9]. Cag-A-positive H. pylori is a subtype of
© 2015 John Wiley & Sons Ltd, Helicobacter 21: 11–23
H. pylori that has been long associated with higher risk of gastric cancer and greater potential to stimulate immune responses compared to the other subtypes. Although the evidence from several studies has shown a positive association between H. pylori infection and CHD, others have indicated that the association is insignificant or only due to confounding effects [10– 12]. Several previous meta-analyses [13–15] also have provided evidence supporting or opposing the potential association between H. pylori infection and CHD events. Theses controversial results leave the H. pylori – CHD association studies under debate for years. In this study, we aimed to extend previous studies by examining the associations of H. pylori infection and Cag-A strains of H. pylori infection with the risk of CHD using a standard systematic review and meta-analysis
11
Helicobacter pylori and CHD
approach on the basis of the existing prospective cohort studies that have been published since 1992.
Methods Data search approach We applied both PRISMA statement [16] and MOOSE guideline [17] to conduct the systematic literature review and analysis. The keywords “Helicobacter pylori” or “H. pylori” combined with “cardiovascular disease”, “coronary heart disease”, “coronary artery disease”, “myocardial infarction”, “ischemic heart disease”, “arteriosclerotic heart disease”, “CHD”, “CVD”, or “CAD” were used as search terms. We used three different search engines: MEDLINE via PubMed (1949 to present) CINAHL, and Cochrane Library, as our primary literature search tools. Endnote X7 software and Excel spreadsheet were used to organize the involved literature. We checked the reference list for each selected paper to identify missing studies. The studies included in the analysis had to fulfill the following criteria: (1) articles that were published in the English language between 1992 and April 2014; (2) articles that presented data on H. pylori infection diagnosed through endoscopy, plasma anti- Helicobacter pylori, physician diagnosis, or the C13 and C14 urease tests. We excluded studies that were irrelevant to our main exposure and outcomes. After searching for potentially related articles, we first reviewed individual abstracts of all found articles to identify whether they met the needs of the study exposure and outcomes of interest. We included articles that consider H. pylori infection as a primary risk factor or a confounding factor. We excluded studies that did not have quantitative analysis of the associations between H. pylori and CHD. We then conducted intensive reviews for all identified articles on the basis of their full published forms. We also identified the references listed for each selected article to identify any additional literature. To test the hypothesis that an exposure to H. pylori infection increased the risk of CHD, we included published studies that were conducted using a prospective study design only (i.e., cohort study, case–cohort study, and nested case– control study).
Data quality assessment For the reviewed studies, we applied the U.S. Preventive Services Task Force (USPSTF) quality rating criteria for randomized controlled trials and cohort studies to evaluate the quality of each identified cohort study.
12
Sun et al.
The USPSTF quality rating criteria for case–control studies was used to evaluate case cohort studies and nested case control studies [18]. We classified the studies that we calculated the risk ratio based on the raw data provided in the original studies as “fair,” as we were not able to adjust for covariate factors. To ensure the quality and consistency of our reviews, each abstract and full article was reviewed by at least two authors independently. Two authors independently assessed the quality of each study, and then, we had group discussions to decide the final grading of the studies.
Data extraction After quality assessments, all authors extracted information from each article and organized the data in Excel independently. We then compared and decided the final data being extracted. Articles that were based on the same cohort data were considered to be the same study; we only included and reviewed the latest published paper. The following data were extracted from each study: first author’s name, year the paper was published, study design, region of the cohort (by continent), cohort follow-up time, sample size, baseline population, measurement of H. pylori infection, measurements of the outcomes, covariates adjusted for in model, effect size of H. pylori positive compared to negative on CHD, effect size of H. pylori seropositive combine Cag-A positive (Hp+/Cag-A+) or combine Cag-A negative (Hp+/Cag-A ) compared to H. pylori negative on CHD, and quality of the data provided by the study. The effect size from the remaining articles was extracted and combined based on the assumption that the odds ratio and the hazard ratio were accurate approximations of the risk ratio. When the effect size was not provided in the original articles, we used conventional formulas to calculate the relative risk and the 95% confidence interval [19]. In papers that provided multiple effect sizes that varied by different covariates for which they controlled, we extracted the data from the most adjusted covariates model. In studies that provided measurements for both H. pylori IgG and IgA, we extracted the information using IgG as a measurement of H. pylori infection. In papers that only provided the effects of H. pylori Cag-A strains compared to H. pylori negative, we extracted the data from Cag-A seropositive strains.
Data analysis We used the random mixed model to analyze the results from the selected articles, as the random mixed model
© 2015 John Wiley & Sons Ltd, Helicobacter 21: 11–23
Helicobacter pylori and CHD
Sun et al.
can provide a robust estimation when combining studies that were carried out in different locations or employed different study designs. The within-study variation and between-study variation were both controlled in the random mixed model, although the random mixed model provided us a wider confidence interval compared to the fixed effect model. We executed the analysis in the statistical programming software R with the “meta” package. Forest plots were used to demonstrate the effect of each study and the summary risk ratio. Additional sensitivity analysis was performed using only studies rated as “good” quality and excluded studies that measured CHD death or mortality as an outcome. We assessed statistical heterogeneity among studies by chisquare tests and estimated the magnitude of heterogeneity using the I-square statistics, where 50% indicated severe heterogeneity of effect [20]. Funnel plots were used to address the publication bias; Egger’s test was used to test the symmetry of funnel plots [21].
Results Study characteristics Figure 1 shows the review process of our article selection. A total of 599 potential articles matched our search key words. There were a total of 24 articles from 24 unique cohorts that met the review criteria and were used in our analysis. The detail information of all studies is presented in Table 1. Sixteen of 24 studies were conducted using a population-based cohort study design. Eight studies were based on high-risk or pre-existing CHD patients. Nineteen studies included in our analysis measured major CVD/CHD events as outcomes, and nine studies measured CHD-related death or mortality as outcomes. In all included studies, H. pylori infection was diagnosed by plasma anti- Helicobacter pylori IgG. All studies except one [22] provided effect size of H. pylori seropositive compared to seronegative on CHD. Two cohorts reported effect size on Cag-A+ compared to Cag-A on CHD events or deaths [22,23]. Six studies provided effect size of Hp+/Cag-A+ compared to H. pylori seronegative on major CHD events [23–28]. Four studies provided effect size of Hp+/Cag-A compared to H. pylori seronegative on CHD [23–25,27]. Eight studies were graded as “fair” in the information given on the effect of H. pylori on CHD, while the rest were graded as “good”. Eleven studies were conducted in Europe, seven were from North America, two each from Asia and Australia, and one from Africa. Nine cohorts had a follow-up time of
REVIEW ARTICLE
A Meta-Analysis of the Association between Helicobacter pylori Infection and Risk of Coronary Heart Disease from Published Prospective Studies Jing Sun, Pooja Rangan, Srinidhi Subraya Bhat and Longjian Liu Department of Epidemiology and Biostatistics, School of Public Health, Drexel University, Nesbitt Hall, 3215 Market St., Philadelphia, PA 19104, USA
Keywords Meta-analysis, Helicobacter pylori, cardiovascular risk factor. Reprint requests to: Longjian Liu, Department of Epidemiology and Biostatistics, School of Public Health, Drexel University, Nesbitt Hall, 3215 Market St., Philadelphia, PA 19104, USA. E-mail: [email protected]
Abstract Background: The association between helicobacter pylori (Hp) infection and coronary heart disease (CHD) has long been debated, and the results from previous meta-analysis are varied. Aims: The aim for this study was to identify the association between Hp and CHD using published perspective cohort studies. Materials and Methods: A systematic review and meta-analysis were performed on studies published from January, 1992 to April, 2014. All studies included used data from prospective cohort studies of CHD events or CHD deaths. Random effect models were applied in all estimations. Results: H. pylori infection increased the risk of CHD events by 11% (19 studies, n = 22,207, risk ratio (RR) = 1.11, 95% confidence interval (CI): 1.01–1.22). This effect was greater for studies that had less than 5 years’ follow-up time (RR = 1.15, 95% CI: 1.00–1.32). However, this effect was not significant for studies that had follow-up times ≥10 years (n = 5100, RR = 1.04, 95% CI: 0.87–1.24). Neither Cag-A seropositive nor Cag-A seronegative strains of H. pylori were associated with a significantly increased risk of CHD events or deaths based on the current published data. Conclusion: In conclusion, H. pylori infection increased the risk of CHD events, especially in a patient’s early life, but this association was weaker or might be masked by other CHD risk factors in long-term observations.
Atherosclerosis is the primary cause of coronary heart disease (CHD), although other factors could play a role in the development of the disease [1]. Previous studies have suggested that some microorganisms and viral infections, including Helicobacter pylori, chlamydia pneumonia, and cytomegalovirus, also contribute to the development of CHD or have been identified as stimuli contributing to ischemic events [2–5]. Helicobacter pylori (H. pylori) infection has long been associated with gastroduodenal lesions, and such chronic inflammation may trigger an immune response. In previous studies, anti-H. pylori immunoglobulin G (IgG) had been associated with an increased risk of CHD [6,7]. Several studies have also observed that CHD patients with H. pylori infection also have a higher prevalence of positive cytotoxin-associated gene-A (Cag-A) antibodies [8,9]. Cag-A-positive H. pylori is a subtype of
© 2015 John Wiley & Sons Ltd, Helicobacter 21: 11–23
H. pylori that has been long associated with higher risk of gastric cancer and greater potential to stimulate immune responses compared to the other subtypes. Although the evidence from several studies has shown a positive association between H. pylori infection and CHD, others have indicated that the association is insignificant or only due to confounding effects [10– 12]. Several previous meta-analyses [13–15] also have provided evidence supporting or opposing the potential association between H. pylori infection and CHD events. Theses controversial results leave the H. pylori – CHD association studies under debate for years. In this study, we aimed to extend previous studies by examining the associations of H. pylori infection and Cag-A strains of H. pylori infection with the risk of CHD using a standard systematic review and meta-analysis
11
Helicobacter pylori and CHD
approach on the basis of the existing prospective cohort studies that have been published since 1992.
Methods Data search approach We applied both PRISMA statement [16] and MOOSE guideline [17] to conduct the systematic literature review and analysis. The keywords “Helicobacter pylori” or “H. pylori” combined with “cardiovascular disease”, “coronary heart disease”, “coronary artery disease”, “myocardial infarction”, “ischemic heart disease”, “arteriosclerotic heart disease”, “CHD”, “CVD”, or “CAD” were used as search terms. We used three different search engines: MEDLINE via PubMed (1949 to present) CINAHL, and Cochrane Library, as our primary literature search tools. Endnote X7 software and Excel spreadsheet were used to organize the involved literature. We checked the reference list for each selected paper to identify missing studies. The studies included in the analysis had to fulfill the following criteria: (1) articles that were published in the English language between 1992 and April 2014; (2) articles that presented data on H. pylori infection diagnosed through endoscopy, plasma anti- Helicobacter pylori, physician diagnosis, or the C13 and C14 urease tests. We excluded studies that were irrelevant to our main exposure and outcomes. After searching for potentially related articles, we first reviewed individual abstracts of all found articles to identify whether they met the needs of the study exposure and outcomes of interest. We included articles that consider H. pylori infection as a primary risk factor or a confounding factor. We excluded studies that did not have quantitative analysis of the associations between H. pylori and CHD. We then conducted intensive reviews for all identified articles on the basis of their full published forms. We also identified the references listed for each selected article to identify any additional literature. To test the hypothesis that an exposure to H. pylori infection increased the risk of CHD, we included published studies that were conducted using a prospective study design only (i.e., cohort study, case–cohort study, and nested case– control study).
Data quality assessment For the reviewed studies, we applied the U.S. Preventive Services Task Force (USPSTF) quality rating criteria for randomized controlled trials and cohort studies to evaluate the quality of each identified cohort study.
12
Sun et al.
The USPSTF quality rating criteria for case–control studies was used to evaluate case cohort studies and nested case control studies [18]. We classified the studies that we calculated the risk ratio based on the raw data provided in the original studies as “fair,” as we were not able to adjust for covariate factors. To ensure the quality and consistency of our reviews, each abstract and full article was reviewed by at least two authors independently. Two authors independently assessed the quality of each study, and then, we had group discussions to decide the final grading of the studies.
Data extraction After quality assessments, all authors extracted information from each article and organized the data in Excel independently. We then compared and decided the final data being extracted. Articles that were based on the same cohort data were considered to be the same study; we only included and reviewed the latest published paper. The following data were extracted from each study: first author’s name, year the paper was published, study design, region of the cohort (by continent), cohort follow-up time, sample size, baseline population, measurement of H. pylori infection, measurements of the outcomes, covariates adjusted for in model, effect size of H. pylori positive compared to negative on CHD, effect size of H. pylori seropositive combine Cag-A positive (Hp+/Cag-A+) or combine Cag-A negative (Hp+/Cag-A ) compared to H. pylori negative on CHD, and quality of the data provided by the study. The effect size from the remaining articles was extracted and combined based on the assumption that the odds ratio and the hazard ratio were accurate approximations of the risk ratio. When the effect size was not provided in the original articles, we used conventional formulas to calculate the relative risk and the 95% confidence interval [19]. In papers that provided multiple effect sizes that varied by different covariates for which they controlled, we extracted the data from the most adjusted covariates model. In studies that provided measurements for both H. pylori IgG and IgA, we extracted the information using IgG as a measurement of H. pylori infection. In papers that only provided the effects of H. pylori Cag-A strains compared to H. pylori negative, we extracted the data from Cag-A seropositive strains.
Data analysis We used the random mixed model to analyze the results from the selected articles, as the random mixed model
© 2015 John Wiley & Sons Ltd, Helicobacter 21: 11–23
Helicobacter pylori and CHD
Sun et al.
can provide a robust estimation when combining studies that were carried out in different locations or employed different study designs. The within-study variation and between-study variation were both controlled in the random mixed model, although the random mixed model provided us a wider confidence interval compared to the fixed effect model. We executed the analysis in the statistical programming software R with the “meta” package. Forest plots were used to demonstrate the effect of each study and the summary risk ratio. Additional sensitivity analysis was performed using only studies rated as “good” quality and excluded studies that measured CHD death or mortality as an outcome. We assessed statistical heterogeneity among studies by chisquare tests and estimated the magnitude of heterogeneity using the I-square statistics, where 50% indicated severe heterogeneity of effect [20]. Funnel plots were used to address the publication bias; Egger’s test was used to test the symmetry of funnel plots [21].
Results Study characteristics Figure 1 shows the review process of our article selection. A total of 599 potential articles matched our search key words. There were a total of 24 articles from 24 unique cohorts that met the review criteria and were used in our analysis. The detail information of all studies is presented in Table 1. Sixteen of 24 studies were conducted using a population-based cohort study design. Eight studies were based on high-risk or pre-existing CHD patients. Nineteen studies included in our analysis measured major CVD/CHD events as outcomes, and nine studies measured CHD-related death or mortality as outcomes. In all included studies, H. pylori infection was diagnosed by plasma anti- Helicobacter pylori IgG. All studies except one [22] provided effect size of H. pylori seropositive compared to seronegative on CHD. Two cohorts reported effect size on Cag-A+ compared to Cag-A on CHD events or deaths [22,23]. Six studies provided effect size of Hp+/Cag-A+ compared to H. pylori seronegative on major CHD events [23–28]. Four studies provided effect size of Hp+/Cag-A compared to H. pylori seronegative on CHD [23–25,27]. Eight studies were graded as “fair” in the information given on the effect of H. pylori on CHD, while the rest were graded as “good”. Eleven studies were conducted in Europe, seven were from North America, two each from Asia and Australia, and one from Africa. Nine cohorts had a follow-up time of
