Eligibility criteria

The design of the present study was based on the methodology of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline [32]. This review is registered in the international prospective register of systematic reviews (PROSPERO) with registration number CRD42020167730 (Available: https://www.crd.york.ac.uk/prospero/display_record.php?RecordID=167730).

Definition

Inclusion/Exclusion criteria

Inclusion criteria included: 1) Case-control, cross-sectional, suitable cohort, and brief-report studies that contain evidence about the relationship between H. pylori and CeD, 2) Human clinical study, 3) Having at least one parameter of the relative risk effect sizes such as risk ratio (RR), odds ratio (OR), and hazard ratio (HR) or can be calculated based on the information of H. pylori status in both case and control groups.

The status of H. pylori should be detected by urea breath test (UBT), rapid urease test (RUT), culture, enzyme-linked immunosorbent assay (ELISA), histology, immunohistochemistry (IHC), and Polymerase chain reaction (PCR) methods. Furthermore, the CeD should be confirmed by endoscopy and biopsy ± serology tests and/or HLA DQ2/DQ8 genotyping. Articles without any of this information were excluded from further analysis.

Sources and search strategy

All relevant articles, published from 1990 to the end of the second month of 2021 were collected from the electronic database of PubMed, ProQuest, Scopus, and Web of Science. The ProQuest database was also reviewed for related dissertations, manually. Other related protocols were manually searched in PROSPERO. Besides, to expand the scope of the search and find any additional studies, we reviewed the different types of grey literatures such as meeting and conference abstracts for relevant articles (including Digestive Disease Week® (DDW), United European Gastroenterology (UEG), and American College of Gastroenterology). Moreover, we searched manually some key journals in the CeD and H. pylori fields, including "World Journal of Gastroenterology", "Journal of Pediatric Gastroenterology and Nutrition", "Gut", "Gastroenterology", and "Helicobacter". In the end, we performed a manual search in the references of the selected articles. We did not impose geographical or linguistic restrictions on our search, and non-English studies with English abstracts were also included.

The Medical subject heading (MeSH) database was used to find various terms of CeD and H. pylori. Two main keywords were “Helicobacter pylori” and “Celiac”. For better searching in the databases, we made syntaxes from a combination of free-text method, MeSH terms, the keywords, and Boolean operators (AND/OR/NOT). Moreover, calculating NNR (number need to read) helped us in evaluating the output of syntax and dedicating. The following syntax was applied in PubMed and adjusted for each search engine based on its search guidelines (Celiac OR “Celiac disease” OR Coeliac OR “Coeliac disease” OR (Disease AND Celiac) OR (Disease AND Coeliac) OR “Gluten Enteropathy” OR (Enteropathy AND Gluten) OR "Gluten Enteropathies" OR (Enteropathies AND Gluten) OR “Gluten-Sensitive Enteropathy” OR (Enteropathy AND Gluten-Sensitive) OR “Gluten-Sensitive Enteropathies” OR (Enteropathies AND Gluten-Sensitive) OR (Sprue AND Celiac) OR (Sprue AND Coeliac) OR (Sprue AND Nontropical) OR “Nontropical Sprue” OR “Celiac Sprue” OR “Coeliac Sprue” OR “Sprue” “Endemic Sprue” OR “Gluten Intolerance”) AND ("Helicobacter pylori" OR "Campylobacter pylori" OR "Campylobacter pyloridis" OR "Campylobacter pyloris" OR "Helicobacter nemestrinae" OR "Helicobacter infection" OR "HP infection" OR "HP infections" OR "Helicobacter infections" OR "Helicobacter colonization" OR "HP colonization" OR "Helicobacter colonization" OR "H Pylori" OR "enterohepatic Helicobacter" OR "EHS" OR "Helicobacter") AND 1990/01/01:2021/02/20[dp].

The search outputs were exported into the Endnote software (Version X7; Thompson Reuters Corporation, Toronto, ON, Canada) to remove any duplicate and screen articles. The first step of screening for determining the eligible primary articles was done based on the titles and abstracts. Two independent reviewers (A.A. and A.F.) evaluated the selected full-text articles and separately classified them into three relevant, irrelevant, and unsure groups, based on the eligibility criteria. Any discrepancy was supervised by a third reviewer (S.B.) and resolved via a consensus. If there was still a doubt in the selection of a study, the whole team made the final decision.

Quality and risk of bias assessment

The quality and risk of bias assessment of the selected articles were independently evaluated by two reviewers (Z.N. and N.G.), using the standard checklist of the Newcastle-Ottawa Scale (NOS) form [33]. The studies were classified into three poor, fair, and good categories based on getting a score between zero and eight in the selection, comparability, and outcome/exposure domains based on the standard guideline. Consensus and the opinion of the third reviewer (S.B) resolved any disagreements between the two reviewers.

Data extraction

Two authors (Z.N, F.A) independently performed data extraction from each paper based on a defined protocol. The data were divided into three categories and included 1) General information: the first author’s names, year of publication, journal names, country, and region; 2) The risk of bias assessment; 3) Study setting: study design, study duration, sample size and total population recruited in each study, inclusion and exclusion criteria, age group and age range, sex, the definition of CeD, source of the data, diagnostic methods, OR and 95% CI, and the ethical approval. In cases of any missing or additional necessary data, we further sent an e-mail to each author, in charge of the related study.

Data synthesis and analysis

We used the R software environment, version 4.0.2, and the “meta” package to calculate pooled OR and 95% CIs for each study. Based on the methodological heterogeneity between the studies, the random-effect model (REM) was used for the combination method. The size of the combined effect was calculated and displayed in a forest plot. The standard chi-square test (Q Cochrane test) and I² scale evaluated the heterogeneity between studies. Combinations with an I² scale equal or more than 70% assumed as a severe heterogeneity. In addition, a Drapery plot was drawn to show the P-value functions for the included studies, as well as pooled estimates, and to provide two-sided confidence intervals for all possible alpha levels (confidence interval function) [34,35]. For finding the cause(s) of heterogeneity, subgroup and sensitivity analyses were used. The studies were stratified based on each factor, including the median of publication year, overall age group of the participants, continent and regions of the studies, H. pylori detection method, the risk of bias assessment, as well as sampling quality. In the last subgroup, sampling, the articles were categorized into two groups (I) “Appropriate sampling” or (II) “Not appropriate sampling” groups. The studies were located in the “Not appropriate sampling” group, if the number of participants in each case and control group was less or equal than 50 people or the ratio of control to the cases was more than four times or less than 90%.

The funnel chart, followed by Begg’s test analyzed possible publication bias. Baujat and L’Abbé plots were used to show the effect of each study on the heterogeneity and overall influence of the results [36–38]. Finally, a one-out remove method was used to detect the influence of each study on the overall pooled OR.

Search results

In total, the general searching step obtained 1732 papers. After removing duplicates, 1106 studies were screened, based on their titles and abstracts. Among them, 1014 articles were excluded due to not meeting the inclusion criteria. Therefore, the full-texts of 92 remained articles were completely evaluated. Finally, 26 studies were included in the systematic review and also meta-analysis (kappa (κ) = 0.84) [11,13–18,21,24–27,39–52], and the others were excluded due to a reason mentioned in Fig 1 (Fig 1). In the end, there was no suitable cohort study in the included studies.

Fig 1

Flow diagram of literature search, screening, and selection of studies for review and meta-analysis.

Characteristics of the included articles

The designs of all the included studies were case-control, with a total of 6001 cases and 135512 control people. Among the studies Lebwohl et al. had the highest study population with 130308 participants (included 2689 cases and 127619 controls) [14]. Most studies were from Turkey (N = 6) [15–18,51,52] and Italy (N = 5) [11,27,44,45,48]. Some studies had been performed only on children [13,15–17,21,44,47,51,52] or adults [18,25,27,39,43,45,48], the others evaluated both age groups [11,14,42,49,50]. Table 1 shows the details of the included articles. The NOS assessment showed that only 19.2% of the studies had good quality.

Table 1

Main characteristics of the included studies in the meta-analysis on the Helicobacter pylori infection and celiac disease (CeD).

First author	Year	Journal title	Country	Definition of CeD	Risk of bias assessment	Case-number	Control—number	Laboratory tests of H. pylori
Karttunen	1990	Journal of Clinical Pathology	Finland	Registry	Fc	27	27	IHCf
Crabtree	1992	Journal of Clinical Pathology	England	Endoscopy + Serology + Positive response to GFDb	Pd	99	250	ELISAg
Feeley	1998	Journal of Clinical Pathology	Ireland	Registry	F	70	70	IHC
Diamantti	1998	Intestinal Disorders (poster)	Argentina	Endoscopy	P	108	113	IHC
Diamantti	1999	The American Journal of Gastroenterology	Argentina	Endoscopy + Serology + Positive response to GFD	P	104	75	IHC
Luzza	1999	Journal of Pediatric Gastroenterology and Nutrition	Italy	Endoscopy+ Positive response to GFD	Ge	81	81	IHC + RUTh + ELISA
Konturek	2000	The American Journal of Gastroenterology	Germany	Endoscopy + Serology	F	91	40	ELISA
Ciacci	2000	European Journal of Gastroenterology and Hepatology	Italy	Endoscopy	P	187	76	IHC + ELISA
Dettori	2001	Digestive and Liver Disease (Poster)	Italy	Endoscopy + Serology	G	152	306	UBTi + IHC
Aydogdu	2008	Scandinavian Journal of Gastroenterology	Turkey	Endoscopy + Serology	P	96	235	IHC
Rostami-Nejad	2009	Revista espanola de enfermedades digestivas	Iran	Endoscopy + Serology	P	28	422	IHC
Rostami-Nejad	2011	Archives of Iranian Medicine	Iran	Endoscopy + Serology	P	24	226	IHC
Lebwohl	2013	American Journal of Epidemiology	American	Endoscopy	F	2689	127619	IHC
Simondi	2015	Clinics and Research in Hepatology and Gastroenterology	Italy	Endoscopy + Serology + HLA DQ2/DQ8 genotyping	F	73	404	UBT + IHC
Jozefczuk	2015	European Review for Medical and Pharmacological Science	Poland	Endoscopy + Serology	F	74	296	UBT
Lasa	2015	Arquivos de Gastroenterologia	Argentina	Endoscopy + Serology	G	72	240	IHC
Jozefczuk	2016	European Review for Medical and Pharmacological Science	Poland	Endoscopy + Serology	F	76	84	UBT + ELISA
Uyanikoglu	2016	Euroasian Journal of Hepato-Gastroenterology	Turkey	Endoscopy	F	31	592	IHC
Narang	2017	Journal of Gastroenterology and Hepatology	India	Endoscopy + Serology + Positive response to GFD	G	324	322	IHC + RUT
Broide	2017	AGA abstracts	Germany	Endoscopy + Serology	F	50	50	UBT + IHC
Lucero	2017	Frontiers in Cellular and Infection Microbiology	Chile	Endoscopy + Serology + HLA DQ2/DQ8 genotyping	G	66	50	IHC + RUT
Agin	2019	Archives of Medical Science	Turkey	Endoscopy + Serology	P	256	1012	IHC
Cam	2018	EHMSG a	Turkey	Endoscopy	P	141	150	UBT + IHC
Dore	2018	Helicobacter	Italy	Endoscopy + HLA DQ2/DQ8 genotyping	F	270	127	UBT + IHC
Bayrak	2020	Helicobacter	Turkey	Endoscopy + Serology	F	482	2060	UBT + IHC
Bayrak	2021	Digestive Diseases	Turkey	Endoscopy + Serology	F	278	505	UBT + IHC

^aXXXIst International Workshop on Helicobacter & Microbiota in Inflammation & Cancer;

^bgluten-free diet;

^cFair;

^dPoor;

^eGood;

^fImmunohistochemistry;

^gEnzyme-linked immunosorbent assay;

^hRapid Urease Test;

ⁱUrease Breath Test.

Association between H. pylori and CeD

By analyzing a combination of 26 studies that addressed the association between H. pylori colonization and CeD, a negative association was found (pooled OR = 0.56; 95% CI = 0.45 − 0.70; P ≤ 0.001). The REM and Chi-squared method showed a significant statistical heterogeneity between the studies I² = 81%, P ≤ 0.001, and X² = 130, P ≤ 0.001, respectively (Fig 2).

Fig 2

The Forest plot of the meta-analysis on the association of the Helicobacter pylori and celiac disease.

The random-effects model was used to calculate pooled ORs with 95% CI. CI, Confidence interval; OR, Odds ratio.

Fig 3A shows a Drapery plot, presenting that the estimated effect is less than 1 in the majority of studies, similarly as shown in the Funnel plot. As the horizontal dashed lines show the CIs for common alpha levels (0.1, 0.05, 0.01, 0.001), at any level of prediction interval (90, 95, 99, 99.9%), the main triangle line (the red-bolded line in the online version) did not intersect the null effect line (Incidence Rate Ratio (IRR) = 1). According to the results of the Funnel plot and Begg’s tests, there was no significant publication bias (P = 0.825, Fig 3B). The Baujat plot (Fig 3C) showed that the study of Narang et al. contributes more than the other studies on the overall heterogeneity and influences the pooled effect size. The L’Abbé plots (Fig 3D) also showed a trend of having more H. pylori colonization in the control group.

Fig 3

(A) The drapery plot shows p-value functions for the included studies as well as pooled estimates and the two-sided confidence intervals for all possible alpha levels (confidence interval function). Each number represents a study. The red-bolded line (in the online version; bold line in the printed version) indicates the range of pooled odds ratio (OR) in each alpha level. (B) The Funnel chart shows no significant publication bias, as all studies are in a symmetric scheme. (C) The Baujat plot shows the effect of each study on the heterogeneity and overall influence of the results. As can be seen in the graph, the study of Narang et al. contributes more than the other studies to the overall heterogeneity. Each circle represents a study. (D) The L’Abbé plot shows more positive results for H. pylori colonization in the control group. The diagonal (x = y) oblique line represents the odds ratio (OR) equal to one. The dashed line indicates the OR of the studies. The size of each circle indicates the assigned random weight of each study. As the picture shows, most studies show an OR of less than one.

Subgroup and sensitivity analyses

The heterogeneity between studies was evaluated by multiple subgroup analyses in the six subgroups. Based on the continents of the studies, there were few studies in Asia, which made the results unreliable for the statistical analysis [13,49,50]. Europe has a higher OR (OR = 0.63, 95% CI = 0.49 − 0.82) than America (OR = 0.55, 95% CI = 0.40 − 0.76) and Asia (OR = 0.23, 95% CI = 0.10 − 0.52). However, further evaluation is needed due to the low rate of studies, especially in Asia. Other subgroups could not determine the source of heterogeneity (Table 2).

Table 2

Different subgroup analysis by stratifying the data.

Subgroups		Na of articles	N of cases	N of controls	ORb	P-Value	Chi-Square	I2	Tau2
Combination of all studies		26	6001	135512	0.56	<0.001	130.80	81	0.2452
Year	Before 2014	13	3756	129540	0.56	0.001	32.47	63	0.1350
	2014 & after 2014	13	2245	5972	0.56	<0.001	98.17	88	0.3804
Overall age	Children (Less than 18)	10	1910	4875	0.57	<0.001	92.64	90	0.4505
	Adult (More than 18)	9	894	1919	0.60	0.004	22.59	65	0.1785
	All age	7	3197	128718	0.52	0.08	11.41	47	0.0734
Continent	Europe	18	2586	6445	0.63	<0.001	75.43	77	0.2138
	America	5	3039	128097	0.55	0.15	6.81	41	0.0516
	Asia	3	376	970	0.23	0.06	5.67	65	0.3481
Sampling	Appropriate sampling	14	1929	3794	0.57	<0.001	97.94	87	0.4459
	Inappropriate sampling	12	4072	131718	0.55	0.001	30.2	64	0.0983
H. Pylori detection tests	IHCc & ELISAd	20	5307	134275	0.58	<0.001	77.88	76	0.1598
	UBTe & RUTf	6	694	1234	0.52	<0.001	45.88	89	0.9442
Risk of bias	Poor	9	1043	2559	0.72	0.009	20.23	60.0	0.1231
	Fair	12	4263	131954	0.55	0.001	31	65.0	0.0873
	Good	5	695	999	0.39	<0.001	31.84	87.0	0.6152

^aNumber;

^bOdds ratio;

^cImmunohistochemistry;

^dEnzyme-linked immunosorbent assay;

^eUrease Breath Test;

^fRapid Urease Test.

Based on the one-out removed method, the removal of any of the studies could not significantly affect the pooled results. The sensitivity analysis was used to find possible outlier studies with more influence on the results. In this regard, the studies with poor design or not appropriate sampling (the most influential factors in the methodology) were omitted, and the overall effect size and level of heterogeneity were recalculated. The overall OR decreased from 0.56 (95% CI = 0.45 − 0.76) to 0.40 (95% CI: 0.26 − 0.62) but no change in the heterogeneity was observed. The results of the sensitivity analysis confirmed that the correct methodology was significantly associated with a lower amount of overall OR (Table 3).

Table 3

Subgroups sensitivity analysis.

Models	Subgroups	Na of articles	N of cases	N of controls	ORb	P-Value	Chi-Square	I2	Tau2
All studies		26	6001	135512	0.56	<0.001	130.80	81	0.2452
Without “poor” quality studies		17	4958	132953	0.49	<0.001	75.27	79	0.1973
	Europe	13	1807	4722	0.57†	<0.001	38.58	67	0.1433
	America	3	2827	127909	0.47	0.40	1.84	0	0
	Asia	1	324	322	0.13	NA	NA	NA	NA
	Appropriate sampling	9	1229	2034	0.40	<0.001	38.76	79	0.3288
	Inappropriate sampling	8	3729	130919	0.62	<0.001	26.81	74	0.1195
Without “not appropriate sampling” studies		14	1929	3794	0.54	<0.001	97.94	87	0.4459
	Poor	5	700	1760	0.98	0.98	0.4	0	0
	Fair	5	600	1085	0.45	0.32	4.67	14	0.0200
	Good	4	629	949	0.34	<0.001	26.14	89	0.6170
	Europe	11	1425	3119	0.66	<0.001	37.91	74	0.1925
	America	2	180	353	0.56	0.04	4.33	77	0.3658
	Asia	1	324	322	0.13	NA	NA	NA	NA
Without “poor” quality and “not appropriate sampling” studies		9	1229	2034	0.40	<0.001	38.76	79	0.3288
	Fair	5	600	1085	0.45	0.32	4.67	14	0.0200
	Good	4	629	949	0.34	<0.001	26.14	89	0.6170
	Europe	7	833	1472	0.48	0.11	10.3	42	0.0691
	America	1	72	240	0.33	NA	NA	NA	NA
	Asia	1	324	322	0.13	NA	NA	NA	NA

^anumber;

^bodds ratio;

†The bolded items show a statistical difference in OR of region subgroup.