Triple treatment including a proton pump inhibitor plus two antibiotics (chosen between clarithromycin, metronidazole and amoxicillin) has been recommended to treat children with Helicobacter pylori infection.^1,2 The main factors for treatment failure are bacterial resistance to the antibiotics used in the regimen and low compliance with drugs.^3,4 Treatment failure is highly predictive if (1) the H pylori strain is resistant to macrolides and (2) clarithromycin is part of the regimen.^5–7 Resistance to metronidazole also affects the efficacy of the treatment if metronidazole is used as part of triple treatment, but to a much lesser extent, in both children⁸ and adults.^9,10

In the Western and Northern European countries, most of the children infected with H pylori are from immigrant families originating from African or Asian countries where the primary resistance to metronidazole is high. Resistance to clarithromycin appears to increase in children over time in some European countries, possibly because of the use of macrolides to treat respiratory tract infections.^11,12 In the US, a high resistance rate for clarithromycin in the children infected with H pylori was reported and would explain the high failure rate of clarithromycin based triple treatment.^13,14 However, most publications are from single centres or include only a small number of children.

In 1996, a European Paediatric Task Force was founded including paediatric gastroenterologists as well as microbiologists and epidemiologists from different European countries. A consensus report,¹⁵ and several multicentric studies on diagnostic or therapeutic issues in paediatric H pylori infection have been published^16,17 or are currently underway. In 1998, a databank was created, in which our group members participated. The aim of the present study was to assess the primary and secondary antibiotic resistance of H pylori strains obtained from children living in Europe over a 4-year period. Other aims were to identify risk factors for H pylori resistance against amoxicillin, clarithromycin, metronidazole and double resistance against clarithromycin plus metronidazole before the first treatment.

PATIENTS AND METHODS

Study centres and inclusion criteria

Members of the European Paediatric Task Force for H pylori infection were asked to submit data on consecutive patients ⩽18 years of age with a positive culture and who had undergone antibiotic susceptibility testing. Seventeen centres from 14 different European countries participated: three Scandinavian countries (Finland, Sweden and Denmark) combined as Northern Europe, three Eastern European countries from the former East block (Poland, Hungary and Croatia), four countries from Southern Europe (Greece, Italy, Spain and Portugal) and four from Western Europe (Belgium, France, Germany and Austria). The ethics committee of the University of Munich approved the protocol for the anonymous data collection and analysis.

Questionnaire

The study centres provided the following items on a standardised questionnaire: age, sex, country at the time of endoscopy (country of study centre), country of birth of the child and the mother, the presence of gastric or duodenal peptic ulcer disease upon endoscopy and whether the child had ever received treatment for H pylori infection in the past.

Bacterial culturing and antibiotic susceptibility testing

Cultures for H pylori and antibiotic susceptibility testing for metronidazole, clarithromycin and amoxicillin were carried out at the local sites. Minimal inhibitory concentration (MIC) breakpoints for resistance were defined as follows: metronidazole >8 μg/ml; clarithromycin ⩾1.0 μg/ml; amoxicillin ⩾0.5 μg/ml. Most centres used the epsilometer test (E test), but in three centres the agar diffusion technique was followed. A patient was considered to harbour a double resistant strain if testing for metronidazole and clarithromycin gave results above the given breakpoints.

Statistical analysis

Statistical analysis was carried out using STATA 7.0 statistical software (Strata Corporation, Texas, USA, 1997). The distribution of resistance to metronidazole or clarithromycin, and of the presence of ulcer was compared in different strata of variables. A univariate analysis was carried out on all of the patients with no missing values for all of the factors considered. The odds were calculated for the resistance to metronidazole, clarithromycin, metronidazole plus clarithromycin (double resistance) and for the presence of peptic ulcer. All variables significantly associated with the studied event (p⩽0.25) were included in a multiple logistic regression model. All models were adjusted for sex and age, as a variable divided into three age groups. All variables associated with resistance or the presence of peptic ulcer were gathered into a final model. The multivariate analysis was carried out by a logistic regression, using a backward elimination procedure of variables not significantly associated with resistance or the presence of an ulcer (p⩽0.05). Estimated odds ratio (OR) and 95% confidence intervals (95% CI) were calculated. Interaction and confusion were tested among all of the variables included in the final model.

For each calculation, two populations were considered and two multiple logistic regression analyses conducted. The first one included only the patients having no missing values for the factors included in the multivariate analysis, whereas the second one considered exactly the same population as in the univariate analysis. The aim of the first analysis was to use as many patients as possible, whereas the second analysis simplified the presentation of the results with the same samples as in the univariate analysis. As no major differences were noted between the two models, only the last one is reported in the results section.

RESULTS

Between January 1999 and December 2002, a total of 1233 patients were included in the database; 46.7% were males, 13% were <6 years of age, while equal proportions comprised the other two age groups (6–12 years: 43% and >12 years: 44%). The majority (n = 1037) were tested for upper abdominal symptoms before any specific treatment for H pylori infection (group A) was given, while 196 children were tested after, at least one failed treatment. There was a slight, but not significant decrease in registered patients over the 4-year period. The country of birth was outside Europe for 14.6% of the children, but for 41.3% of the mothers, indicating the high percentage of immigrants, particularly from Africa and the Middle East.

Table 1 gives the absolute numbers of the patients and their characteristics, including the results of susceptibility testing for the total study group, and the two subgroups, group A without and group B with previous anti-H pylori treatment.

Antibiotic resistance before the first treatment (group A)

Amoxicillin

Resistance to amoxicillin was a rare event (6/1037, 0.6%), with 2 patients reported from Greece, and one each from Croatia, Italy, Finland and Poland.

Clarithromycin

Resistance to clarithromycin was detected in 20% of the strains (199/991) obtained before the first treatment. In the univariate analysis, important risk factors were: age <6 years, male sex, living in Southern Europe and country of birth of the child and his/her mother in Western or Southern Europe (table 2). The final logistic regression model showed a 2.25 times higher risk for primary clarithromycin resistance, if the child grew up in Southern Europe (p<0.001). Young age and male sex remained independent major risk factors (table 3).

Metronidazole

Primary resistance to metronidazole was found in 23% of the strains (233/1024). In the univariate analysis, only two factors were found to be of considerable risk for primary resistant to metronidazole: country of birth of the child and the mother (table 4). Multilogistic regression models showed that patients born in Asia, Africa or Middle East had a 2.4 times higher risk (95% CI 1.61 to 3.66) of being infected with a metronidazole resistant strain than patients of equal age and sex born in Western, Northern or Southern Europe (p<0.001; table 5). No interaction was found between the different variables in the final model.

Double resistance

A primary resistance against both clarithromycin and metronidazole was found in 53 of 992 (5.3%) strains. No significant risk factors could be identified, except that in 2001 fewer children were diagnosed with a double resistant strain.

Antibiotic resistance after at least one failed treatment (group B)

When patients with treatment failure were compared with untreated children (before the first treatment), no difference was identified in the factors assessed except for antibiotic resistance. The chance to harbour a resistant strain significantly increased after failed treatment, for clarithromycin from 20% to 42% (p<0.001), for metronidazole from 23% to 35% (p = 0.001), and for resistance to both from 5.3% to 15.3% (p<0.001), although the amoxicillin resistance rate remained the same (0.6%; table 1). No other identified risk factor was linked to secondary antibiotic resistance.

Peptic ulcer disease

Peptic ulcer disease was diagnosed in 6.8% of the children (80/1180), with no difference between untreated children (before first treatment) or after treatment failure (table 1). No relation was found between antibiotic susceptibility and peptic ulcer disease. The only significant risk factor linked to peptic ulcer disease was age. Patients with peptic ulcer disease were significantly older (12.5 years, 95% CI 11.8 to 13.2) compared with patients with gastritis only (10.9 years 95% CI 10.6 to 11.1; p = 0.001). The rate of peptic ulcer disease was low in children <6 years of age (3.5%, 4/118), and between 6 and 11 years of age (4.6%, 17/390); it increased to 10.4% in patients older than 11 years. The final logistic regression model showed an adjusted OR of 3.10 (1.07 to 8.95) for children >11 years compared with the youngest age group (p = 0.036) of the same sex.

DISCUSSION

Antibiotic resistance in H pylori is a serious public health problem, because about 10–15% of infected people develop severe complications in adult life like peptic ulcer disease and gastric malignancies, and only cure of the infection prevents these sequelae. Large numbers of cases are needed to obtain reliable results for the prevalence rate of antibiotic resistance. Such studies have been carried out on H pylori strains isolated from adults in Europe (n = 1274)¹⁸ and the US (n = 347).¹⁹ However, such investigations have never been conducted on a sufficient number of children despite the fact that the context of prescribing antibiotics is different.

The main aim of our study was to obtain reliable data by including a large number of previously untreated and unselected H pylori infected children. To avoid any selection and reporting bias, every child with a successful susceptibility testing was included. The contributing centres conducted culture tests for H pylori as part of the routine work up during upper endoscopy with a success rate for culture of >80%, as previously shown.¹⁷ Although the participating centres were mostly University hospitals, a selection bias for children with a higher likelihood for a primary antibiotic resistance rate is unlikely. Most patients in this study were referred for the procedure from the caring doctor because of upper abdominal pain, some with a positive non-invasive diagnostic test indicating H pylori infection.

The multicentric character of our study caused several limitations. There was an uneven distribution of patients from the different countries, with fewer patients from Scandinavia. The results of the different countries were analysed separately, and thereafter in groups according to their geographical region. Countries from the former East block were combined, because in these countries the availability of expensive antibiotics, like second-generation macrolides, may be different and could influence the antibiotic resistance pattern.

Owing to logistic and financial constraints, cultures and antibiotic susceptibility testing were not carried out in a central bacteriology laboratory, as was the case in the European and North American studies on adults.^18,19 Two different methods were used for susceptibility testing: E test and agar diffusion. For clarithromycin, a perfect correlation was shown when these two methods were compared on the same isolates.²⁰ However, differences were observed with respect to metronidazole because there is a continuum of MICs rather that a double population. In this study, MIC values were reported from 852 isolates. As expected, the distribution curve showed a much better separation between positive and negative results for clarithromycin compared with metronidazole. Only 31 (3.6%) patients harboured isolates with MIC values for clarithromycin within 4 dilutions of the cut off value (MIC range 0.25–2.0), whereas this was the case in 73 (8.6%) patients for metronidazole (MIC range 4.0–32). Therefore, differences between the laboratories probably had only a minor effect on the resistance rates, particularly for clarithromycin.

In contrast with the European multicentre study on adults,¹⁸ reliable information was obtained with respect to previous anti-H pylori treatment. Parents tended to have reasonable recall of whether or not their child had been treated for the infection. However, they often did not remember the names of the antibiotics the child had taken in the past. Therefore, the term primary resistance applies to the absence of any previous specific anti-H pylori treatment, and not to any previous intake of amoxicillin, clarithromycin and metronidazole. In the US data base, previous treatment was recorded, but pre-treatment resistance rates were not calculated.¹⁹ However, this information is essential for the surveillance of primary resistance and to give guidelines for first line treatment.

The primary resistance rate of isolates from children was 20% for clarithromycin and 23% for metronidazole. No representative data for comparison with adults living in Europe are available in the literature. The results of the European multicentre survey represents 17 European countries, but did not distinguish between isolates from patients before or after failed treatment.¹⁸

By contrast, the European MACH2 study recruited most patients from Northern and Western Europe.²⁰ Therefore, we could only compare results of primary resistance from 515 adults in Germany during the years 1999 and 2000²¹ with the results obtained in our study from children living in Germany (before treatment n = 201). The resistance rate for clarithromycin was markedly lower in adults than in children (2.7% v 15.9%). By contrast, for metronidazole the resistance rates were very similar in adults and children (25% v 27.5%).

The presence of a resistant strain may indicate the transmission of a resistant strain or in vivo selection of resistance. The similar rate of metronidazole resistance in children and adults is in favour of the transmission of a resistant strain to the children. Children acquire the H pylori infection mainly from their infected parents, mostly their mother, in both industrialised and developing countries.²² In this study, the only risk factor for primary metronidazole resistance was the immigration from a non-European country. Metronidazole is widely used in Africa and Asia to treat parasitic diseases and gynaecological infections. Therefore, in these countries female adults have a higher risk of being exposed to this antibiotic agent.

By contrast, the higher primary clarithromycin resistance rate in isolates from children compared with adults points to an in vivo acquisition of the resistance during childhood. Macrolide resistance is based on defined point mutations in the peptidytransferase loop in both copies of the 23S rRNA gene. Monotherapy with clarithromycin induces these mutations in up to 21% of patients infected with a susceptible H pylori strain.²³ Clarithromycin has been used increasingly over the last 10 years to treat upper respiratory tract infection in children. Owing to the high price, the newer macrolides have been less frequently prescribed in countries of the former Eastern block, as well as in Africa and Asia. In fact, children born in these countries harboured considerably less clarithromycin resistant H pylori strains than children from countries with a high consumption of macrolides, such as Italy, Spain and Portugal. Other independent risk factors—young age and male sex—which were identified in the multilogistic regression analysis, also pointed to an in vivo acquisition of clarithromycin resistance in our patients. Younger compared with older children, and boys compared with girls, suffer more often from infectious diseases, particularly respiratory tract infections and therefore, are more likely to be exposed to the new macrolides. Our findings confirm a previous report in H pylori infected Spanish children, which showed a decreasing prevalence of clarithromycin resistance with age: 45.4%, 30.2% and 9.5% in the age groups 4–8 years, 9–13 years, and 14–18 years, respectively.²⁴ In Portugal, a higher clarithromycin resistance rate was also reported in children (44.8%) compared with adults (14.8%).²⁵

Peptic ulcer disease was a rare finding, with <5% in children <12 years of age. This rate increased to >10% in teenagers. Older age was the only risk factor for this complication. This could be due to the longer duration of the infection itself, but more likely, other risk factors may contribute in teenagers—for example, smoking, consumption of alcohol and ulcerogenic drugs. The low prevalence of peptic ulcer disease in symptomatic children referred for upper endoscopy has implications on the “test and treat” strategy. In a paediatric population with a prevalence of 10% for H pylori infection, 100 symptomatic teenagers need to be investigated by a reliable non-invasive diagnostic test, and 10 treated with triple treatment, in order for one with peptic ulcer disease to benefit. In children <12 years of age, >200 children need to be screened for one child with H pylori related peptic ulcer disease to benefit. In a population with a lower prevalence of H pylori infection like in most Northern and Western European countries, the cost/benefit ratio would be even lower.

Our results from more than a thousand H pylori infected children before their first treatment have major implications for the first line treatment. Despite the fact that metronidazole has a lower effect on the clinical efficacy of metronidazole-based triple treatment, when the metronidazole resistance rate reaches 40%, as is the case in children born in Africa, Asia or the Middle East, this antibiotic should not be used. Clarithromycin-based triple treatment cannot be recommended without previous susceptibility testing in children living in Southern Europe, where a third of the children are infected with a macrolide resistant strain. Even in Western and Eastern European countries, susceptibility testing should be conducted whenever possible, to obtain a high primary cure rate. This is particularly important in children <12 years of age, for whom no rescue treatment is currently available.