Helicobacter pylori infection and childhood recurrent abdominal pain : Lack of evidence for a cause and effect relationship

Can J Gastroenterol Vol 13 No 7 September 1999 607 Departments of Community Health Sciences and Paediatrics, University of Calgary, Calgary, Alberta Correspondence and reprints: Dr Colin Macarthur, Department of Community Health Sciences, Health Sciences Centre, 3330 Hospital Drive NW, Calgary, Alberta T2N 4N1. Telephone 403-220-8585, fax 403-270-7307, e-mail colinmac@ucalgary.ca CANADIAN HELICOBACTER CONSENSUS CONFERENCE ON PEDIATRIC ISSUES

R ecurrent abdominal pain is a common complaint among children and adolescents.Apley's criteria -at least three discrete episodes of abdominal pain of sufficient severity to interrupt normal activities and occurring over a period of three or more months -are often used to define this chronic pain syndrome (1).Several descriptive studies have shown that the prevalence of recurrent abdominal pain ranges from 10% to 15% in schoolchildren (1)(2)(3).Peak prevalence occurs at around nine years of age, with a slightly higher prevalence in females than in males (4).Survey data also demonstrate that recurrent abdominal pain is a common presenting complaint to primary care physicians and subspecialists (5,6).In addition, the pain syndrome is associated with frequent school absences (7).
The etiology and pathogenesis of recurrent abdominal pain are, however, poorly understood.Although early reports implicated psychological factors, subsequent controlled studies failed to demonstrate a clear psychological etiology for this chronic pain syndrome (8)(9)(10).The initial focus on possible psychological causes was driven, in part, by the low frequency of identifiable 'organic' causes in children with recurrent abdominal pain.Most studies suggest that less than 10% of children with recurrent abdominal pain have an organic etiology for their chronic pain (11)(12)(13)(14).
The discovery of Helicobacter pylori and its association with antral gastritis in children stimulated interest in the possibility of an infectious etiology for recurrent abdominal pain (15).A previous review of the published literature on H pylori infection and recurrent abdominal pain, however, demonstrated inconsistent results (16).In addition, the evidence at that time was limited to case series (only one controlled study was identified), study populations were heterogeneous, and only children referred to tertiary centres were studied.Subsequent controlled studies have addressed some of these limitations (17)(18)(19)(20)(21)(22).Therefore, the objective of this critical review was to summarize the extent and quality of the current published evidence for a cause-and-effect relationship between H pylori infection and childhood recurrent abdominal pain.

MATERIALS AND METHODS
The MEDLINE bibliographic database (January 1983 to July 1998) was searched to identify pertinent studies.Only English language publications were included.Medical subject key headings included 'Campylobacter pyloridis', 'Campylobacter pylori', 'Helicobacter pylori', 'abdominal pain' and 'recurrent abdominal pain'.Current Contents and selected pediatric and gastroenterology journals were also searched.Relevant citations from the bibliographies of identified articles were also retrieved.
The search was restricted to prospective, controlled studies reporting empirical data on children (0 to 18 years of age).Editorials, commentaries, reviews, case reports, case series and letters to the editor were excluded.The published data were used to estimate the prevalence of H pylori infection in children with recurrent abdominal pain and in control children.These data were used to calculate an odds ratio for each study, with 95% confidence intervals derived using the test-based method (23).(In most situations, the odds ratio provides a good approximation of the relative risk, ie, the risk of recurrent abdominal pain in children with H pylori infection, compared with the risk in uninfected children.) To assess the quality of the evidence, information pertinent to the internal and external validity of each study was gathered.Such data included study design, setting, criteria for case definition, selection of controls, sample size, validity of measures, control for confounding variables, number of subjects lost to follow-up and blinding of investigators (where appropriate).Last, several of Hill's criteria for causal inference were applied to the published data to assess the strength of the evidence for a causal relationship between H pylori infection and recurrent abdominal pain (16,24).Criteria applied included strength of the association, consistency, temporality, biological plausibility and experimental evidence.

RESULTS
Six prospective, controlled studies that examined the association between H pylori infection and childhood recurrent abdominal pain were identified (17)(18)(19)(20)(21)(22).Five of the six were case-control studies, ie, sampling was on outcome status, whereby children with recurrent abdominal pain were compared with control children (17)(18)(19)(20)(21).The prevalence of H pylori infection in the group with recurrent abdominal pain was then compared with the prevalence of H pylori infection in control children.The remaining study assessed the effectiveness of antibiotic therapy in children with recurrent abdominal pain (22).
The characteristics of the five case-control studies are shown in Table 1.The studies were worldwide; however, only one study was truly community-based (21), with the remaining studies conducted in the tertiary care hospital setting (17)(18)(19)(20).Each of the studies used (modified) Apley's criteria to define cases (children with recurrent abdominal pain).Control children were selected from outpatient clinics (both gastroenterology and nongastroenterology), preoperative clinics, day surgery lists and schools.All the studies measured H pylori infection by using serological tests.
Table 2 shows the prevalence of H pylori infection in children with recurrent abdominal pain and in control children.As shown by the odds ratios, study results were inconsistent.Results from three studies were not significant, ie, the 95% confidence intervals included unity.Results from two studies were statistically significant but were conflicting.The study by Chong et al (19) suggested that H pylori infection was a risk factor for recurrent abdominal pain, whereas the study by Hardikar et al (20) suggested that infection was protective against recurrent abdominal pain.Of note, of all the studies, only Hardikar et al (20) controlled for potential confounding variables.In their study, multivariate analysis was used to control for age, sex, parental ethnicity and socioeconomic status.The adjusted odds ratio was 0.21 (95% CI 0.05 to 0.85).
The single treatment trial used triple therapy (amoxicil-lin, bismuth subcitrate and metronidazole) in a subset of children with both H pylori infection and recurrent abdominal pain (22).Around 20% of children were lost to followup, assessments were not blinded and the psychometric properties of the measure used to quantify abdominal pain were not provided.Last, and of most concern (as noted by the authors), there was no placebo control group.Application of Hill's criteria to the published data demonstrated that the association between H pylori infection and recurrent abdominal pain was not strong; results were inconsistent, there was no evidence for a temporal relationship, no biological plausibility and no supporting experimental evidence.

DISCUSSION
The evidence for a causal relationship between H pylori infection and recurrent abdominal pain is inconsistent.Of the five studies reviewed, the odds ratios ranged from 0.32 to 1.80.Two studies demonstrated statistically significant results; however, the findings were conflicting.The only treatment trial was limited because of methodological flaws.
Although the literature search was comprehensive, some studies (both published and unpublished) may have been missed.Furthermore, only (English language) publications were reviewed.Last, information pertinent to the review and collected by the investigators may not have been provided in the journal article.
Of the five case-control studies, results from three were not significant, whereas one demonstrated a statistically significant negative association.In other words, the weight of the evidence suggests no association between H pylori infection and recurrent abdominal pain.The study by Chong et al (19), however, suggested an increased risk of recurrent abdominal pain in children with H pylori infection.The validity and generalizability of these findings, however, may be limited.First, the study was conducted in a tertiary hospital   setting.Therefore, although endoscopy was performed on only 111 of 218 children with recurrent abdominal pain, 14 (13%) had duodenal or gastric ulcer disease.In addition, the control children were selected from those attending a gastroenterology clinic for reasons other than recurrent abdominal pain.As expected, many of these children demonstrated gastrointestinal pathology.For example, 51 of the control children had chronic liver disease, 48 had chronic inflammatory bowel disease, 24 had failure-to-thrive and 22 had gastrointestinal bleeding.In summary, the case group was not representative of all children with recurrent abdominal pain, while the control group of children was not a suitable comparison group.Last, there was no adjustment for potential confounding variables, such as age, ethnicity and socioeconomic status.
The study by Hardikar et al (20) showing a statistically significant negative association between H pylori infection and recurrent abdominal pain has several strengths (20).First, only incident cases of recurrent abdominal pain were recruited.In addition, control children (selected from day surgery lists) were in good health, with no history of recurrent abdominal pain or other gastrointestinal symptoms, and no family history of peptic ulcer disease.Also, the serological test had been validated in the specific age group and population that was studied.Moreover, the investigators controlled for potential confounding variables (age, sex, ethnicity and socioeconomic status) by using multivariate analysis.
Calculation of a summary odds ratio across all studies was not a primary aim of this review.Given the heterogeneity of study populations and the inclusion of only observational study designs in the review, cautious interpretation of any summary estimate of effect is necessary.The Mantel-Haenszel method for pooling data across studies was used (24).Of note, the c 2 test for heterogeneity of odds ratios was positive, ie, the Chong study odds ratio was significantly different from the odds ratios estimated by the remaining four studies.Calculation of a summary estimate of effect (after exclusion of the Chong study) resulted in a summary odds ratio of 0.74 (95% CI 0.50 to 1.10).
Treatment trials are an alternative approach to examining the association between H pylori infection and recurrent abdominal pain.In other words, if the association were causal, it would be expected that eradication of H pylori infection would lead to resolution of symptoms.Causal inference from treatment trials is strengthened when allocation to treatment or placebo is random, investigators and patients are blinded to group assignment, loss to follow-up is minimized, and outcome measures are standardized and valid.The single treatment trial identified by this review, however, did not meet any of these criteria (22).

CONCLUSIONS
The weight of the current evidence suggests no obvious association between H pylori infection and recurrent abdominal pain in children.The results, however, are inconsistent.Further evidence from community-based studies, using healthy children as controls, with adjustment for potential confounders, is needed.Therefore, evidence to date indicates that routine investigation for H pylori infection in children who present with the classical symptoms of recurrent abdominal pain (based on Apley's criteria) is not warranted.Similarly, treatment of the infection to resolve symptoms of recurrent abdominal pain is not justified.