Does Helicobacter pylori infection play a role in iron deficiency anemia? A meta-analysis

Abstract

AIM: To perform a meta-analysis of observational studies and randomized controlled trials (RCTs) on the association between Helicobacter pylori (H. pylori) and iron deficiency anemia (IDA).

METHODS: A defined search strategy was used to search Medline, Embase, the Cochrane Library, Clinical Trials, Cochrane Central Register of Controlled Trials, Premedline and Healthstar. Odds ratio (OR) was used to evaluate observational epidemiology studies, and weighted mean difference (WMD) was used to demonstrate the difference between control and intervention groups.

RESULTS: Fifteen observational studies and 5 RCTs were identified and used for calculation. The pooled OR for observational studies was 2.22 (95% CI: 1.52-3.24, P < 0.0001). The WMD for hemoglobin (HB) was 4.06 g/L (95% CI: -2.57-10.69, P = 0.01), and the WMD for serum ferritin (SF) was 9.47 μg/L (95% CI: -0.50-19.43, P < 0.0001). Results were heterogeneous for all comparisons.

CONCLUSION: This meta-analysis on observational studies suggests an association between H. pylori and IDA. In RCTs, eradication of H. pylori can improve HB and SF levels but not significantly.

Key Words: Helicobacter pylori, Iron-deficiency anemia, Meta-analysis, Hemoglobins, Odds ratio

INTRODUCTION

Anemia, defined as a hemoglobin concentration below established cut-off levels, is a widespread public health problem with major consequences for human health as well as social and economic development[1]. The World Health Organization (WHO) estimates that about 2 billion people in the world are suffering from this disease, and that approximately 50% of all anemia cases are diagnosed as iron deficiency anemia (IDA)[2,3]. IDA affects the work capacity of patients and may contribute to mortality, thus limiting economic development. The overall death rate from IDA has been underestimated in most surveys from many developing and developed countries[4,5]. WHO suggested that researchers and clinical doctors should investigate the etiology of IDA and develop therapeutic strategies because timely treatment restores personal health and increases national productivity by 20%[6,7].

It is known that a variety of causes such as inadequate iron intake, chronic blood loss, chronic disease, malabsorption, hemolysis, or a combination of these, can induce IDA[5,8-10]. Among possible causes, the involvement of Helicobacter pylori (H. pylori) infection remains controversial[11-13]. H. pylori is a highly prevalent microbial infection. Over 50% people in the world are infected by H. pylori. In Africa, Mexico, South America and Central America, H. pylori infection reaches 70%-90% of the population[14,15]. H. pylori has been considered as a major cause for the development of peptic ulcer disease, gastric malignancy and dyspeptic symptoms[16-19]. Recent studies have shown that H. pylori can also cause other extragastric diseases[20-22]. However, knowledge regarding any relation between H. pylori infection and IDA is limited. Moreover, studies regarding the role of H. pylori infection in IDA and the effectiveness of the eradication of H. pylori in the treatment of IDA are controversial.

This clinical research question is addressed by this meta-analysis. The aim of the study was to evaluate the association between H. pylori infection and IDA and examine the effect of H. pylori eradication on serum hemoglobin (HB) and serum ferritin (SF) levels. Observational epidemiological studies have demonstrated an association between H. pylori and IDA by comparing IDA risk between H. pylori-infected and non-infected participants. Randomized controlled trials (RCTs) have established a cause and effect relationship between

H. pylori and IDA. In this meta-analysis, we hypothesized that there is a significant difference in IDA risk between H. pylori-infected and non-infected participants, and that H. pylori eradication therapy can significantly increase HB and SF concentration, thus alleviating IDA. We tested our hypothesis by pooling the results of studies on H. pylori and IDA.

MATERIALS AND METHODS

Search strategy and identification of studies

We searched, without language restrictions, for all publications on H. pylori and IDA between January 1966, and June 2009. Searches were performed on Medline, Embase, Clinical Trials, Database of Abstracts of Reviews of Effects (DARE), Cochrane Central Register of Controlled Trials (CENTRAL), the Cochrane Database of Systematic Reviews, Premedline, Healthstar, by using the MeSH heading: “Helicobacter pylori”, “iron-deficiency anemia”, “anemia”, “iron” and “hemoglobin” and the non-MeSH terms “sideropenic refractory anemia” and “serum ferritin”. The reference lists of major textbooks, review articles, and of all the included articles identified by the search were then individually searched to find other potentially eligible studies. Information about unpublished and ongoing RCTs was sought from authors of the included RCTs, and experts in the field.

Selection criteria and validity assessment

The present meta-analysis followed the Quality of Reports of Meta-Analyses of RCTs (QUOROM) guideline for RCTs and observational studies in epidemiology [Meta-analysis of Observational Studies in Epidemiology (MOOSE) and Methodological Index for Non-Randomized Studies (MINORS)] guideline in observational studies[23-25]. To avoid selection bias, selection criteria were established before searching. Two reviewers (Qu XH and Huang XL) identified articles eligible for further review by performing an initial screen of the abstracts or titles of the search results. The second screening was based on a full-text review according to the selection criteria. The observed agreement between reviewers for eligibility of articles was 96.3%, corresponding to modest agreement (κ = 0.40). Discrepancies were resolved by discussion and consultation with other reviewers (Xiong P and Zhu CY).

Observational epidemiology studies: Observational epidemiology studies (cross-sectional, case-control, or cohort) investigating the prevalence of IDA in H. pylori-positive patients and negative controls were included in this meta-analysis. Duplicate publications and those studies in which patients had other underlying common IDA causes (e.g. aspirin/NSAID use, colonic carcinoma, gastric carcinoma, angiodysplasia) were excluded.

Randomized controlled trials: In order for an RCT to be included, its participants must have had both H. pylori infection and IDA/iron deficiency (ID). At least 2 authors independently assessed the methodological quality of included RCTs by Jadad scores[26]. In addition, for a study to be eligible for inclusion, the use of therapy to eradicate H. pylori in intervention groups and administration of oral ferrous sulfate to both intervention and control groups were required. Discrepancies in data extraction were resolved by discussion among authors (Qu XH and Huang XL).

Data abstraction

For observational epidemiology studies, we collected information on the year of publication, location of the study, age groups, number of cases and controls, country and region, number of IDA positive and negative patients, test method for H. pylori.

For RCTs, we collected information on references, year of publication, sample size, age group, treatment therapies, H. pylori testing methods and changes in mean ± SD of HB and SF in both the intervention and control groups.

Statistical analysis

For observational epidemiology studies, we recorded the prevalence of IDA in H. pylori-positive patients and controls for each study as an odds ratio (OR) and 95% CI and the weight of the studies. We used the heterogeneity χ² (Cochran Q) statistic to formally analyze heterogeneity across included studies. Meta-analysis was performed using Review Manager Version 5 (Cochrane Collaboration and Update Software) for observational studies[27].

For RCTs, we collected changes in HB and SF concentration after H. pylori eradication and evaluated them by using weighted mean difference (WMD) with 95% CI. A χ² test was used to assess heterogeneity of the studies. If the studies were found to be heterogeneous, (i.e. χ² > df), we utilized the DerSimonian and Laird random-effects model[28] rather than a fixed effects model to reassess the pooled estimates. The source of heterogeneity was investigated as described below. Publican bias was performed by both Review Manager Version 5 and STATA version 10. We also performed the Duval and Tweedie nonparametric “trim and fill” procedure[29] to further assess the possible effect of publication bias in our meta-analysis.

Subgroup analysis

Subgroup analysis was performed to assess the factors that might impact the pooled estimates and to investigate the source of heterogeneity. Sensitivity analysis was also conducted to test whether the analysis was robust by changing statistical methods, reanalyzing the data, and comparing the 2 results by the t test.

Publication bias

Funnel plots and Begg’s test[30] are thought to detect the existence of publication bias of pooled ORs within observational studies. Small studies are scattered widely at the bottom of the graph, while the spread narrows for larger studies. When a funnel plot seemed to be asymmetrical, we used Duval and Tweedie’s nonparametric “trim and filled” method as a sensitivity analysis to reassess the pooled estimates[29]. This method considers the possibility of hypothetical “missing” studies that might exist and recalculates the results with the imputed missing studies.

RESULTS

Search results

The search strategy retrieved 7969 potentially relevant references. Of these, 7689 were not relevant, e.g. animal studies, physiological or pharmacological studies. The remaining 280 references were assessed by screening their abstracts, and we excluded any references that were editorials or narrative reviews. One hundred and fifty nine studies were subjected to a full text review and excluded according to the selection criteria as described earlier. Supplementary studies were identified that had been published only as abstracts from conference proceedings of scientific meetings. We then excluded all RCTs with a Jadad score under 3 to ensure the quality of eligible trials. Fifteen observational studies[31-45] and 5 RCTs[46-49] (4 of which[46-49] provided both HB and SF data, and one study[40] provided only SF data) meeting our criteria were included in our meta-analysis (Figure 1).

Figure 1 Flow chart showing the trial flow for selection of RCTs and observational studies to be included. RCT: Randomized controlled trial; HP: H. pylori; IDA: Iron deficiency anemia; HB: Hemoglobin; SF: Serum ferritin.

Study characteristics and quality

In total, 15 183 patients from 20 studies (15 observational and 5 RCTs) were included in the meta-analysis, and the characteristics of the sample are summarized in Table 1.

Table 1 Summary characteristics of studies and participants.

Ref.	Participants	Age group (yr)	Male (%)	Prevalence of H. pylori infection (%)		H. pylori test methods
Observational studies					Sampling mode
Milman et al[31]	2264	30-60	1153 (51)	32.0	Hospital/health examination	Serum IgG
Choe et al[32]	375	10-15	205 (55)	16.8	High school/questionnaires	Serum IgG
Cuoco et al[33]	362	16-58	115 (32)	21.0	Hospital diagnosed patients	Histological examination
Choe et al[34]	660	15-17	376 (57)	29.5	High school/physical examination	Serum IgG
Nahon et al[35]	210	57.4 (21.4) (SD)	80 (38)	NA	Hospital diagnosed patients	Histological examination
Choe et al[36]	937	10-18	475 (51)	20.8	-/physical examination	Serum IgG
Choi[37]	674	9-12	344 (51)	13.6	Middle-class families/physical examination	Serum IgG
Ciacci et al[38]	55	> 17	22 (40)	60.0	Hospital diagnosed patients	Histological examination/rapid urease test
Hershko et al[39]	210	16-77	NA	NA	Hospital diagnosed patients	UBT/Serum IgG
Gessner et al[40]	690	7-11	NA	87.0	8 most populous villages/physical examination	UBT
Baggett et al[41]	668	7-11	354 (53)	86.5	10 predominantly villages/physical examination	UBT
Cardenas et al[42]	7462	≥ 3	NA	27.1	NHANES/questionnaire, laboratory, examination data	Serum IgG
Süoglu et al[43]	70	4-16	NA	50.0	Hospital diagnosed patients	Histological examination
Haghi-Ashtiani et al[44]	209	2-14	111 (45)	47.8	Hospital diagnosed patients	Histological examination
Mulayim et al[45]	117	NA	0 (0)	61.5	Hospital diagnosed patients	UBT
Randomized controlled trials					Treatment therapies
Choe et al[46]	13	10-17	NA	19.7	B+A+M	Histology
Gessner et al[40]	201	7-11	NA	87.0	L+A+C	UBT
Chen et al[47]	86	18-76	50 (58)	NA	B+A+M	UBT
Vijayan et al[48]	22	> 13	NA	NA	L+T+C	Rapid urease test/histology
Sarker et al[49]	99	2-5	46 (46)	NA	O+A+C	UBT

UBT: Urea breath test; IgG: Immunoglobulin G; NHANES: National Health and Nutrition Examination Survey; B: Bismuth; A: Amoxicillin; M: Metronidazole; L: Lansoprazole; C: Clarithromycin; T: Tinidazole; O: Omeprazole.

Of the observational studies, 4 studies with participants over 18 years old[31,35,38,39] and one study with patients aged 16-58 years old[33] were classified as adult groups. Three studies with patients aged 10-18 years old were classified as adolescent groups[32,34,36]. Two studies with patients younger than 11 years old were classified as child groups[41,49]. For the presence of H. pylori, 6 studies utilized serum immunoglobulin G (IgG)[31,32,34,36,37,42], 5 utilized histological examination[33,35,38,43,44] and 3 used the urea breath test (UBT)[40,41,45]. UBT and serum IgG were utilized by Hershko et al[39] in their studies. Of the RCTs, 2 studies with participants aged 2-11 years old were classified as child groups[40,49] and 3 were classified as adolescent and adult groups[46-48]. All the RCTs used eradication triple therapy for H. pylori as intervention.

Study methodological quality is shown in Table 2 (observational studies) and Table 3 (RCTs). In observational studies, the MINORS quality score ranged from 6 to 14 points. Only 8 (53%) of the articles included employment outcomes as part of the main study aim. Five studies satisfied the criteria for inclusion. Most data were collected according to a protocol established before the beginning of the study, and most studies have no exclusion or details about the reasons for exclusion. In RCTs, qualities of all the studies were evaluated by Jadad sore. All of the studies included had a score greater than 3. Only one study fulfilled all of the evaluated quality criteria. All studies were randomized, and for 4 of them, the generation of allocation sequence was judged adequate. Only 2 studies were designed as double-blind, but placebo was offered in only one study. Three of the 5 RCTs had no exclusions, and one of the 5 RCTs gave details on the reasons for exclusion.

Table 2 Methodological quality assessment based on MINORS¹.

Source	Aim2	Rate3	Data4	Measure5	Bias6	Time7	Loss8	Size9	Total10
Milman et al[31]	2	2	2	2	2	1	1	0	12
Choe et al[32]	2	1	1	1	2	1	0	0	8
Cuoco et al[33]	1	1	1	2	2	0	0	1	8
Choe et al[34]	2	1	1	2	2	1	0	0	9
Nahon et al[35]	1	2	2	1	2	1	1	0	10
Choe et al[36]	2	2	1	1	0	0	0	0	6
Choi[37]	2	1	1	2	2	2	1	0	11
Ciacci et al[38]	1	1	2	1	2	1	0	0	8
Hershko et al[39]	1	1	1	1	2	1	1	0	8
Gessner et al[40]	2	2	2	1	2	2	2	1	14
Baggett et al[41]	2	1	2	1	2	1	1	0	10
Cardenas et al[42]	2	2	2	1	2	1	1	1	12
Süoglu et al[43]	1	1	1	1	2	1	0	0	7

¹Assessed with the adapted Methodological Index for Non-Randomized Studies (MINORS)[25];

²Clearly stated aim (0,1,2 points);

³Inclusion of consecutive patients and response rate (0,1,2);

⁴Prospective collection of data (0,1,2);

⁵Inclusion of employment measure (0,1,2);

⁶Unbiased assessment of study end points (0 or 2);

⁷Follow-up time appropriate (0,1,2);

⁸Loss to follow-up (0,1,2);

⁹Prospective calculation of the study size (0 or 1);

¹⁰Total: minimum equals 0; maximum equals 15 points.

Table 3 Quality evaluation of the included studies.

RCT	Randomization	Blindness	Withdraw and dropout	Total
Choe et al[46]	2	2	0	4
Gessner et al[40]	2	1	1	4
Chen et al[47]	1	1	1	3
Vijayan et al[48]	2	2	1	5
Sarker et al[49]	2	0	1	3

Summary estimates

Risk of IDA for H. pylori-positive versus H. pylori-negative patients: We tested the heterogeneity of the 15 observational studies that provided information about prevalence OR, and the heterogeneity χ² statistic was 51.29 (P < 0.00001). Therefore, the pooled estimates were evaluated under a random effects model instead of a fixed effects model. The pooled OR was 2.22 and the 95% CI was 1.52-3.24 (P < 0.0001) suggesting that IDA is associated with H. pylori (Figure 2).

Figure 2 Forest plot of the observational studies. M-H, Random: Mantel-Haenszel heterogeneity random effects model. Horizontal lines = 95% CI. The rectangles represent the point estimates of the study and the size of the rectangle represents the weight given to each study in the meta-analysis. The diamond represents the summary estimate; the size of the diamond represents the CIs of the summary estimate.

H. pylori eradication effect in IDA patients: Four RCTs[46,48-49] reported blood parameter (HB levels and SF concentrations) differences and one RCT[40] reported SF concentration differences between the intervention and control groups after H. pylori eradication therapy. We pooled summary estimates to demonstrate the treatment effect and underlying connection between H. pylori and IDA. The results showed that H. pylori eradication therapy can improve IDA. Four RCTs compared the increase in HB levels and 5 in SF concentrations achieved with H. pylori eradication (plus iron) treatment and with iron administration alone in patients with IDA, and found a greater effect in the eradication group (WMD of HB: 4.06 g/L, 95% CI: -2.57-10.69, P = 0.01; WMD of SF: 9.47 μg/L; 95% CI: -0.50-19.43, P < 0.0001, Figure 3).

Figure 3 The treatment effect and underlying connection between H. pylori and IDA. A: Weighted mean difference (WMD) forest plots of HB (g/L) involved in the meta-analysis; B: Forest plots of studies estimate changes in SF (μg/L) level. IV, Random: Inverse variance heterogeneity random effects model. Horizontal lines = 95% CI. The size of the data marker corresponds to the weight of that study. The diamond represents the summary estimate. The result favors experimental groups.

Subgroup analysis

Subgroup analysis was performed to investigate the source of heterogeneity and detect the influential factors that could impact the summary estimates. The methodological and biological heterogeneity of the studies made it possible to explore the summary estimates in many different subgroups.

In observational studies, differences in the sensitivity of the H. pylori test methods could partly result in different pooled ORs. The 15 observational studies utilized 3 methods, enzyme-linked immunosorbent assay (ELISA) serum IgG, histological biopsy and UBT, to test for the presence of H. pylori. The pooled OR of ELISA serum IgG was lowest (OR = 2.16, 95% CI: 1.49-3.14, P = 0.10) and the pooled OR of UBT was highest (OR = 5.88, 95% CI: 2.27-15.23, P = 0.47). The pooled ORs were consistent with the sensitivity of these methods.

Subgroup analysis for different age groups revealed a significant difference between children, adolescents and adults in the association between H. pylori and IDA. The pooled OR of children younger than 11 years was 4.76 (95% CI: 1.73-13.08, P = 0.83). Adolescents yielded a pooled OR of 2.85 (95% CI: 1.68-4.31, P = 0.89), while adults had a pooled OR of 1.55 (95% CI: 0.67-3.62, P < 0.0001).

In RCTs, factors included in the subgroup analysis were age and therapy of each study. The summary estimate from children was 0.65 g/L (95% CI: -1.52-2.82, P = 0.39) for HB changes, significantly different from the pooled estimate of 25.03 g/L (95% CI: 9.69-40.37, P = 0.41) from adolescent and adult groups, indicating that adult IDA patients react more strongly to H. pylori eradication therapy. The WMD for SF changes was 0.70 μg/L (95% CI: -1.01-2.41, P = 0.45) in children while the WMD was 14.79 μg/L (95% CI: 2.53-27.05, P = 0.0001) in adolescent and adult patients.

To examine the method of therapy, we separated the studies into a bismuth triple therapy group and a proton pump inhibitor (PPI) triple therapy group. Bismuth triple therapy showed an obvious advantage (WMD of SF = 11.55 μg/L, 95% CI: 0.09- 23.01, P = 0.0002) over PPI triple therapy (WMD of SF = 7.15 μg/L, 95% CI: -6.45-20.75, P = 0.002), particularly when used together with oral ferrous sulfate for H. pylori patients with IDA.

We analyzed the association between H. pylori and IDA in developed areas and less developed areas, large sample subgroups and small sample subgroups, and did not find any significant difference between those parameters. Table 4 shows the summary estimates of subgroups of both RCTs and observational studies.

Table 4 Summary of subgroup analyses of both observational studies and experimental studies.

	Subjects (n)		95% CI	Heterogeneity (χ2)	P1
Observational studies		Odds ratio (OR)
H. pylori test methods
ELISA serum IgG	12372	2.16	1.49-3.14	9.27	0.10
Histological biopsy	906	2.17	0.90-5.26	28.85	< 0.0001
UBT	1475	5.88	2.27-15.23	1.53	0.47
Age
Child	1358	4.76	1.73-13.08	0.05	0.83
Adolescent	1972	2.85	1.68-4.31	0.22	0.89
Adult	3101	1.55	0.67-3.62	28.41	< 0.0001
Randomized controlled trials2		Weighted mean difference (WMD)
Hemoglobin (g/L)
Age
Child	300	0.65	-1.52-2.82	0.74	0.39
Adolescent and adult	35	25.03	9.69-40.37	0.67	0.41
Therapy
PPI	322	0.95	-2.98-4.87	3.71	0.16
Bismuth	13	32.00	9.37-54.63	-	-
H. pylori test methods
Histological biopsy	223	7.03	-9.41-23.47	2.79	0.10
UBT	35	25.03	9.69-40.37	0.67	0.41
Serum ferritin (μg/L)
Age
Child	300	0.70	-1.01-2.41	0.57	0.45
Adolescent and adult	121	14.79	2.53-27.05	17.91	0.0001
Therapy
PPI	322	7.15	-6.45-20.75	7.68	0.002
Bismuth	99	11.55	0.09-23.01	13.61	0.0002

¹P-value tested for heterogeneity of subgroups;

²Data were derived from hemoglobin changes and serum ferritin changes.

Sensitivity analysis

We chose to change the weights of every observational study involved so as to detect the stability of this meta-analysis. We then reanalyzed the data using different statistical methods. The pooled OR using a fixed effects model was 2.34 (95% CI: 1.97-2.78), which is not a significant change from the original random effects model (P = 0.74).

Publication bias

Visual inspection of the Begg’s funnel plot revealed asymmetry (P < 0.01). This raises the possibility of publication bias, so we undertook a sensitivity analysis using the trim and fill method. This method conservatively imputes hypothetical negative unpublished studies to mirror the positive studies that cause funnel plot asymmetry. The adjusted summary OR is based on the eventually filled funnel plot (4.32, 95% CI: 3.00-5.66, P < 0.001), which continued to show a statistically significant association between H. pylori and IDA (Figure 4).

Figure 4 Funnel plots without (A) and with (B) trim and fill. The pseudo 95% CI is computed as part of the analysis that produces the funnel plot, and corresponds to the expected 95% CI for a given standard error (SE). OR: Odds ratio.

DISCUSSION

Our results from a meta-analysis of 15 observational epidemiological studies revealed a correlation between H. pylori and IDA (OR, 2.22; 95% CI: 1.52-3.24, P < 0.0001), although some studies reported only a slight association. In addition, In RCTs, eradication of H. pylori can improve HB and SF levels but not significantly (WMD of HB: 4.06 g/L, 95% CI: -2.57-10.69, P = 0.01; WMD of SF: 9.47 μg/L, 95% CI: -0.50-19.43, P < 0.0001).

Dufour et al[50] first reported that H. pylori eradication had a positive effect on sideropenic refractory anemia, indicating a possible underlying association between H. pylori and IDA. A large population-based study from the USA reported that H. pylori infection was an independent risk factor for IDA in 7462 children, adolescents, and adults[42]. This research reported that H. pylori infection was associated with an increased risk of IDA (OR, 2.6; 95% CI: 1.5-4.6). Compared to former studies, our meta-analysis was a detailed and comprehensive investigation procedure. It included RCTs of highly detailed power and large-scale observational epidemiology studies.

It is reported that H. pylori infection is observed in over 50% people in the world with peaks of 70%-90% for some countries. Moreover, IDA affects 2 billion people in the world. When 2 diseases have such a high prevalence in the population they may appear to be associated with each other. Recent studies regarding the role of H. pylori infection in IDA are controversial. However, whether eradication of H. pylori prevents IDA has been widely debated. This meta-analysis was performed to clarify this issue: whether iron deficiency could specifically be related to H. pylori infection. The observational studies in our meta-analysis prove the association between H. pylori and IDA. In fact, in the RCTs of our meta-analysis, after iron replacement, HB and SF were not different between the groups with or without H. pylori eradication. However, these data should be interpreted with caution because of the marked heterogeneity among studies. We carefully performed subgroup analysis, and found age and therapies had an impact on the increase in the levels of HB and SF. These results should be investigated further in the future. Larger scale RCTs should be recommended to test the results of our meta-analysis.

As with all meta-analyses, the results we obtained could be impacted by 3 factors: heterogeneity within the studies involved, bias (including selection bias[51,52] and detection bias[53,54]), and publication bias[55-57]. The generation of heterogeneity could occur by virtue of the methodological and biological heterogeneity of the studies analyzed, such as differences in diagnostic methods, the population under study, the sample size, and language of publication. Each of the subgroups described above contributed partly to the heterogeneity of the observational studies.

We assessed the included studies with caution. For observational studies, the MINORS quality score ranged from 6 to 14 points. In RCTs, quality of all the studies were evaluated by the Jadad sore. All of the studies included had a score greater than 3. Our test for heterogeneity was significant, and hence we utilized a random effects model that accounted for inter-study variation. Compared with the fixed effects model, the random effects model evenly distributes weight among studies, minimizing the impact of heterogeneity[58].

We used age as one subset determinant for the pooled estimates. A significant association between H. pylori and IDA was found in children younger than 11 years. The common causes of IDA, such as colonic carcinoma, gastrectomy or menstruation, were usually absent in children. Thus, H. pylori infection can be treated as the only indicator for refractory IDA in children, and H. pylori infection should be considered first in children with IDA[59-61]. The same scenario occurred in adolescents, as adolescents are particularly susceptible to ID. Because of the requirement for a large amount of iron to sustain their growth, dietary deficiency, and menstrual blood loss, girls should be more strongly affected by H. pylori infection[10,34,62]. In adults, no such strong association was found. The possible explanation for this phenomenon is that H. pylori plays a smaller part in the etiology of IDA in adults[1,5,63]. In RCT subgroup analysis, eradication therapy for H. pylori did not demonstrate the same curative effect on IDA in children as in adults. This may arise from the special characteristics of children (quickly growing blood volume and large requirement)[59-61].

Different diagnostic methods for H. pylori contribute to the variation of the pooled estimates because of their different sensitivities. The pooled OR value increased with the sensitivity of the diagnostic method. Therefore, UBT, which has the highest sensitivity of the 3 tests used[64-66], obtained the highest pooled OR, while ELISA serum IgG tests yielded the lowest pooled OR. Recent guidelines have indicated that UBT is regarded as a gold standard diagnostic method and the most reliable nonendoscopic test for the existence of H. pylori[67].

The way in which RCTs chose to eradicate H. pylori can make a difference in pooled analyses. Bismuth-based triple therapy indicated a much better response to iron intake than PPI-based triple therapy. The work done by McColl and Hutchinson may explain this phenomenon[68,69]. It was reported that PPI therapy lowers the concentration of vitamin C in gastric juice and reduces the bioavailability of ingested vitamin C thus resulting in low absorption of nonheme iron. It may also retard the clinical response to iron supplementation. Vitamin C, as an essential factor in alimentary iron absorption, not only converts ferric iron to the ferrous form, which maintains solubility at the alkaline pH of the duodenum, but also chelates with ferric chloride which is also stable at a pH > 3. PPI can also reduce the absorption of vitamin B₁₂, a significant factor in iron absorption, probably by inhibiting intragastric proteolysis.

Publication bias was tested by Begg’s test and illustrated by funnel plots. The results indicated the existence of publication bias. The funnel plot showed that there were some missing small sample studies. Therefore, meta-analysis would underestimate the association between IDA and H. pylori. The “trim and fill” method helped to resolve this problem by imputing the hypothetical studies symmetrically and reassessing the pooled estimates as a sensitivity analysis. The filled funnel plot showed a strong association between IDA and H. pylori. Possible sources of asymmetry in funnel plots were explored: variations in sample size, etc, could contribute to publication bias.

Sensitivity analysis was performed in several ways to test concordance of the results by changing the statistical methods used. The sensitivity analyses that were performed did not materially change the results, increasing the confidence that can be placed in these results when applying the conclusion in practice.

Our study has limitations. We have excluded trials that studied the relationship between H. pylori and iron deficiency. However, these studies have been reviewed elsewhere[70]. Furthermore, in the experimental studies, eradication treatment without ferrous sulfate but with placebo groups were excluded because the number of studies was not adequate. This kind of design can illustrate the role that H. pylori plays in IDA in a better way, and we expect further investigations will take that design into consideration. Lastly, results were markedly heterogeneous for all comparisons. These data should be interpreted with caution because of the marked heterogeneity among studies.

In conclusion, our meta-analysis of 15 observational studies demonstrated an association between H. pylori and IDA. In addition, the meta-analysis of RCTs showed that eradication of H. pylori can improve HB and SF levels, though not significantly. From the analysis, we also concluded that IDA could not specifically be related to H. pylori infection. We do not recommend a strategy of population-based screening and treatment for H. pylori infection to prevent IDA. This concept should be discussed in the future. UBT is the most reliable nonendoscopic test for the existence of. Bismuth-based triple therapy has a better response to increase HB and SF levels than PPI-based triple therapy. There are no significant differences between less developed areas and developed areas in the association between H. pylori infection and IDA.

COMMENTS

Background

Both Helicobacter pylori (H. pylori) and iron deficiency anemia (IDA) have high prevalence worldwide. The relationship between these 2 remains controversial. Recent guidelines for H. pylori and IDA both focus on the role H. pylori plays in the process of IDA.

Research frontiers

The interaction between H. pylori and IDA is a current ’hot topic’. H. pylori infection impairs iron absorption causing a considerable decrease in the concentration of gastric juice ascorbic acid (vitamin C) that is the best promoter of nonheme iron absorption. It thus causes the decline of hemoglobin in red blood cells and directly leads to anemia. The gastric colonization by H. pylori increases lactoferrin uptake from neutrophils and increases iron demand.

Innovations and breakthroughs

To the best of the authors’ knowledge, this is the first published meta-analysis assessing the association between H. pylori infection and IDA (evaluating children, adolescents and adults) and assessing the effect of H. pylori eradication on hemoglobin (HB) and serum ferritin (SF) levels.

Applications

The research showed an association between H. pylori and IDA. Eradication of H. pylori can improve HB and SF levels but not significantly. The authors do not recommend a strategy of population-based screening and treatment for H. pylori infection to prevent IDA.

Peer review

This is a good and interesting meta-analysis.