INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Helicobacter pylori infection is usually acquired in childhood and is the main cause of active chronic gastritis and peptic ulcer disease in both adults and children (3, 8, 10, 19, 30). The infection induces cellular and humoral serum immune responses in most patients, and measurement of specific antibodies in serum has been used as a noninvasive method by which to detect H. pylori infection. Specific immunoglobulin M(IgM) antibodies can be detected shortly after the infection is acquired, but IgA and IgG titers indicate chronic infection (5).

Serological tests are commercially available, easy to perform, and inexpensive and therefore have been recommended for the diagnosis of H. pylori infection in adults (1, 19). Many serological tests, mainly IgG based, have been validated in adult populations against invasive methods with acceptable sensitivity and specificity for clinical use (11, 13, 20). Studies of children showed controversial results, with a large sensitivity range of 50 to 96% and specificity ranging from 83 to 100% (2, 5, 16, 21-23, 28). Most of the investigators used an in-house enzyme immunoassay (EIA) with a cutoff value adapted to the pediatric population under consideration. Using a commercially available EIA on 68 Brazilian children, Oliveira et al. observed a strong relationship between the age of a child and sensitivity (21). In children older than 12 years, the sensitivity was 93%, but in children between 2 and 6 years of age, this value dropped to 44%. Because of the controversial results obtained with children, the consensus statement of the European Society of Pediatric Gastroenterology, Hepatology and Nutrition and the European H. pylori Study Group considered serological testing to be less reliable for children than for adults, but further validation studies and improvement of tests are warranted (10).

The purpose of this study was to evaluate two commercially available second-generation EIAs, both for IgG and IgA, for the diagnosis of H. pylori infection in symptomatic children of different ages and nationalities living in Germany. Enzygnost II is the most widely used H. pylori enzyme-linked immunosorbent assay in Germany (according to data from the German nationwide "Instand" external quality control). Pyloriset was one of the most popular tests before Enzygnost II was introduced to the German market.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Patients. The study group included 178 children aged 9 months to 19 years (mean ± standard deviation [SD], 9.2 ± 4.3 years) who underwent upper endoscopy for evaluation of symptoms suggestive of upper gastrointestinal tract disease. Symptoms included recurrent upper abdominal pain, heartburn, regurgitation, vomiting, and hematemesis. A ¹³C-urea breath test (UBT) was performed on all patients. The following data were obtained: age, sex, nationality, previous H. pylori eradication therapy, and medication (acid-suppressive drugs, antibiotics) during the 4 weeks prior to endoscopy. Only patients without previous treatment for H. pylori infection were included.

UBT. The UBT was performed in accordance with our previously described protocol (17). In brief, after a fasting period of at least 4 h, each child drank 150 ml of refrigerated (6°C) apple juice (pH 3.4). Thereafter, the child ingested 75 mg of ¹³C-labeled urea (Eurosotop, Paris, France; 99% chemical purity) dissolved in 20 ml of apple juice and then drank another 30 ml of juice to rinse the mouth. Before (baseline) and 15 and 30 min after tracer application, the child was asked to blow into a breath bag (Medicheck, Essen, Germany). For young and disabled children, a face mask was used for breath sampling. Aliquots of expiratory air were transferred into 10-ml Vacutainers. The ratios of ¹³C to ¹²C were measured by isotope ratio mass spectrometry (Finnigan MAT delta S, Bremen, Germany). The difference between the value at 15 or 30 min and the baseline was expressed as delta over baseline (per mille). The UBT was defined as positive for H. pylori infection if the 15- and/or 30-min value was above a cutoff of 5%. This test protocol had been validated for 149 children with biopsy-based (histology, urease test, culture) H. pylori status with a sensitivity of 100% and a specificity of 93.2% (17).

Invasive diagnostic tests. During upper endoscopy, two biopsies of the antrum and two of the corpus were taken for histology, formol fixed, stained with Giemsa, and viewed for the presence of H. pylori-like microorganisms. One additional antral biopsy was obtained for a rapid urease test (HUT-Test; Astra, Wedel, Germany), which was considered to be positive when the color change occurred within 8 h, in accordance with the manufacturer's instructions. Bacterial culture was performed from an additional antral biopsy as described previously (26).

Definition of H. pylori status. A child was considered to be infected with H. pylori when at least two of the three applied methods (UBT, rapid urease test, histology) and/or culture were positive and considered to be noninfected when all of the tests gave concordant negative results. Patients not fulfilling these criteria were excluded from the analysis.

Serology. Venous blood samples (2 ml) were obtained from each child at the time of upper endoscopy. The serum was separated, divided into aliquots, and stored at 20°C before testing. All sera were tested with four EIAs: Pyloriset IgG and IgA (Orion Diagnostica, Espoo, Finland) and Enzygnost second-generation IgG and IgA (Dade Behring Marburg GmbH, Marburg, Germany). The assays were performed in accordance with the manufacturers' instructions and without knowledge of the children's H. pylori status.

Enzygnost II. The Enzygnost second-generation test is based on antigens from cytotoxin-positive strain NCTC 11639 (ATCC 43629). The serum samples were diluted 1:21 in serum dilution buffer (10 µl of serum plus 200 µl of Tris-buffer at 0.3 mol/liter). A 20-µl volume of diluted serum was added to 200 µl of sample buffer and pipetted to the plate. Two negative and positive reference sera were used on each plate; the positive reference sera were positioned before and after the patient samples. After incubation for 30 min at 37°C, plates were washed four times and 100 µl of peroxidase-IgG or -IgA conjugate (rabbit, Fab') was added. After incubation for 30 min at 37°C, plates were washed four times again. Fresh substrate solution, tetramethylbenzidine (100 µl), was added, and the plates were incubated for 30 min at room temperature. The enzyme reaction was stopped with 100 µl of 5% H₂SO₄, and the A₄₅₀ was read (correction wavelength, 650 nm). The reference values had to be within the given margins. Absorbences of the samples were multiplied with a correction factor calculated by division of the nominal reference value by using the corrected absorbence readings (DA) in the following formula ( method): log₁₀ titer (in units per milliliter) = ×DA. Lot-dependent constants  and  were given individually in tables enclosed with the tests. Titers of <10 U/ml were considered negative for IgG and IgA, and those of 10 U/ml were considered positive.

Pyloriset. Serum samples were diluted 1:201 in serum dilution buffer. Diluted serum samples (100 µl) and four prediluted calibrator serum samples were pipetted onto the strips, which were coated with inactivated H. pylori antigen. The plates were incubated for 60 min at room temperature. The wells were washed three times with washing buffer and tapped dry. Anti-human IgG/IgA enzyme conjugate (swine, 100 µl) was pipetted into each well. The plates were incubated for 60 min at room temperature and washed again. Fresh substrate solution, p-nitrophenyl phosphate (100 µl), was added, and the plates were incubated for 30 min at room temperature. The enzyme reaction was stopped with 100 µl of 1 M NaOH, and the A₄₀₅ was read. The titer of each patient's serum was read from a graph based on the standard curve obtained with semilogarithmic axes. Titers were considered positive if they were 300 U for IgG and 250 U for IgA.

Statistical analysis. Sensitivity, specificity, positive and negative predictive values, and two-sided approximate 95% confidence intervals were calculated. All analyses were performed by using the Statistical Package for Social Sciences (version 6.1.3; SPSS, Inc., Chicago, Ill.). The influence of age was calculated with the Kruskal-Wallis test for the three-group comparison. The level of significance was set at P = 0.05. Significance for comparison of differences in nationalities and age groups was calculated with the Fisher test. For adaptation of the optimal cutoff, receiver-operating curve analysis was performed and the Youden index was calculated. The Youden index is defined as follows: (sensitivity + specificity) 1. The cutoff with the highest Youden index has the lowest number of misclassified results.

Ethical approval. Informed consent was obtained from the children's parents. The study was approved by the Ethics Committee of the Ludwig Maximilians University, Munich, Germany.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Three of the 178 children were excluded from the analysis due to undetermined H. pylori status according to our strict criteria: all three had a positive UBT, but the other methods were negative. Of the remaining 175 patients, 82 (47%) were H. pylori positive and 93 (53%) were H. pylori negative. The children were divided into three age groups. The first group included 47 children <6 years old (mean ± SD, 3.6 ± 1.5 years; range, 0.9 to 6.0 years); only 12 (25.5%) were H. pylori positive. The second group consisted of 77 patients >6 and <12 years old (mean ± SD, 9.1 ± 1.6 years; range, 6.1 to 11.7 years); 43 (55.8%) were H. pylori positive. In the third age group, with children >12 years old (mean ± SD, 14.5 ± 1.8 years; range 12.1 to 19.6 years), 27 (52.9%) were H. pylori infected. H. pylori-positive children had a mean age of 10.4 years (range, 1.5 to 17.4 years) and were significantly older than H. pylori-negative patients (8.2 years; 0.9 to 19.6 years) (P = 0.001). The sex distribution did not differ significantly between H. pylori-positive (46 girls, 36 boys) and -negative patients (42 girls, 51 boys) (P = 0.11). Peptic ulcer disease was detected in 10 patients (nine duodenal ulcers, one gastric ulcer); all of them were H. pylori infected. Patients with peptic ulcer disease were significantly older than H. pylori-infected patients without ulcer disease (12.7 versus 10.1 years; P = 0.027).

Dade Behring Enzygnost II. When the cutoff values suggested by the manufacturer be used; IgG titers were above that value in 76 of 82 infected children and in 4 of 93 noninfected children (see Fig. 3). IgA results were positive in only 39 of the 82 H. pylori-positive children and in 2 of the 93 noninfected patients. The overall sensitivity, specificity, and positive and negative predictive values for all of the children and the three age groups are given in Table 1. Specificity was excellent for IgG and IgA in all three age groups. Sensitivities were slightly, but not significantly, better in older than in younger children (Table 1). Only one infected child with a negative IgG result was positive for IgA. Neither IgG nor IgA concentrations of the infected patients showed significant differences among the three age groups (P = 0.67; P = 0.58) (Table 2; Fig. 1 and 2).

View larger version (15K): [in this window] [in a new window]   FIG. 1.   Antibody titers obtained with Enzygnost II IgG in relation to the ages of 82 H. pylori (H.p.)-positive () and 93 H. pylori-negative () children with the recommended cutoff of 10 U/ml.

View larger version (15K): [in this window] [in a new window]   FIG. 2.   Antibody titers obtained with Enzygnost II IgA in relation to the ages of 82 H. pylori (H.p.)-positive () and 93 H. pylori-negative () children with the recommended cutoff of 10 U/ml.

Orion Diagnostica Pyloriset. When the cutoff value recommended by the manufacturer was used, IgG and IgA antibodies were detected in 58 and 20 of the 82 infected children and in 2 and 2 of the 93 noninfected children, respectively. The characteristics for IgG and IgA are given in Table 1. Specificity was excellent for both IgG and IgA independent of age. In contrast, sensitivity varied markedly depending on the type of antibody (IgG was better than IgA) and on the age group, with a significantly lower sensitivity (P = 0.002) for IgG in younger than older children. Only minor differences were seen in sensitivity for IgA in the different age groups (Table 1). In two infected children, determination of specific IgA was informative, giving positive results, while the IgG titers were below the cutoff value.

View larger version (17K): [in this window] [in a new window]   FIG. 3.   Antibody titers obtained with Pyloriset IgG in relation to the ages of 82 H. pylori (H.p.)-positive () and 93 H. pylori-negative () children with the recommended cutoff of 300 U/ml.

View larger version (16K): [in this window] [in a new window]   FIG. 4.   Antibody titers obtained with Pyloriset IgA in relation to the ages of 82 H. pylori (H.p.)-positive () and 93 H. pylori-negative () children with the recommended cutoff of 250 U/ml.

Optimization of cutoff values. ROC analyses were used to determine the cutoff with the best performance, defined as the highest Youden index. The Youden index would have improved in lowering the cutoff values of all of the tests except Enzygnost II IgG (Table 3).

Test performance in relation to patients' characteristics except age. When H. pylori-positive patients with (n = 10) and without (n = 72) ulcer disease were compared, no significant differences could be observed regarding test sensitivity (Pyloriset IgG, 60% versus 71.2%; Pyloriset IgA, 30% versus 23.3%; Enzygnost II IgG, 80% versus 95.6%; Enzygnost II IgA, 50% versus 46.6%, respectively). The antibody titers were not significantly different between these two groups for all four tests (P = 0.73). Sex also did not influence the accuracy of the test results.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Applying four commercially available EIAs for the diagnosis of H. pylori infection to a cohort of 175 children, we found very discrepant results. While the specificity of all four tests was very good, with values above 95%, sensitivity was very low (<50%) for the two IgA-based tests and varied markedly for the two IgG-based methods. When the manufacturer's cutoff values were used, 93% of H. pylori-infected children had a positive test result with Enzygnost II IgG, with no difference among the three age groups. In contrast, Pyloriset IgG identified only 71% of the infected children, and this percentage dropped to 25% in patients below the age of 6 years.

A lower sensitivity for serological H. pylori tests in children compared to adults has been reported. Crabtree et al. determined the optimal cutoffs for pediatric and adult patients (5). They found a much higher cutoff value for adults than for children. Application of the optimal cutoff for adults to children would have detected only 10 of the 21 infected children. Lowering of the lower cutoff allowed the identification of 18 of the 21 children with no loss of specificity. Raymond et al. found significantly lower antibody titers in children under the age of 10 years than in older infected children (P < 0.05) (22). Khanna et al. developed a pediatric in-house EIA using as the antigen whole cells from a single clinical H. pylori strain isolated from a 14-year-old girl and studied 142 children and 54 adults from the United States, Canada, Taiwan, and South Africa (16). In the same cohort of patients with H. pylori status defined by invasive methods, three commercial test kits were applied for comparison. While the pediatric in-house test showed a sensitivity of 91.4%, the three commercial tests identified only 63, 70, and 78% of the infected children. In the adult patients, all four tests showed high sensitivities, varying from 92 to 96.5%.

Only a few investigators have looked at a dependence on age or other patient characteristics with respect to the accuracy of the tests within a pediatric population as we did. To our surprise, the sensitivity of Enzygnost II IgG was very good and above 90% even in children younger than 6 years of age. In contrast, Pyloriset IgG showed significantly lower mean specific antibody titers in infected children of the youngest age groupreflected also by a low sensitivity of only 25%compared to older children and adolescents. However, only 12 children under the age of 6 years were H. pylori infected; therefore, these data have to be interpreted with caution. Lowering of the cutoff values of the tests would improve the accuracy of Pyloriset IgG but would not affect the Enzygnost II IgG results (Table 3). In agreement with our experience with Pyloriset, other studies reported a lower sensitivity in younger children. The CobasCore from Roche Diagnostic Systems was tested by Rocha et al. with a significantly higher sensitivity in older compared to younger children (P = 0.006) (24). Using the same test on 139 children, Corvaglia et al. found significantly more false-negative results in children younger than 5 years of age (P < 0.05) (4). As mentioned before, Khanna et al. tested three commercial serology tests, PyloriStat from Bio Whittaker, FlexSure from SmithKline Diagnostics, and HM-CAP from Enteric Products, and stated that the accuracy of all three tests was greatly reduced in younger children (16).

In contrast to age, we and others did not find an influence of sex on the accuracy of the test results. However, differences have been observed in children with different ethnic and geographical backgrounds. Khanna et al. (16) reported a lower sensitivity in children from developing countries and suggested that immunodeficiency due to malnutrition may explain this finding. Oliveira et al. (21) reported better sensitivity in children with peptic ulcer disease than in infected children with gastritis only, even when the results were adjusted for age. Both the sensitivity and the specificity of serological test kits depend on the antigen preparation used (15, 18). The differences in sensitivity of the same test in children of different ages (16, 21) or geographical origins (14) may be explained by different immune responses of the populations under investigation. There are differences in the antigenicity of the multiple H. pylori strains and even of the different antigens of the same strain. When immunoblotting was performed with sera of H. pylori-infected children, Rocha et al. (24) observed that the number of immunoreactive bands significantly increased with age. Reactions to the VacA and CagA antigens were more frequently seen in older children (24). Besides the differential immune responses at different ages, duration of infection may play a role. Since the primary infection occurs most commonly in infants and toddlers, younger children are expected to have a shorter duration of infection than older children and adults (25, 27). This may influence both IgG and IgA titers. Older children may also have a greater chance to be infected by more than one bacterial strain and therefore to have increased reactivity to different antigens. In contrast to age, the geographic background of the children (German versus non-German) did not significantly influence the accuracy of the four tests.

With respect to peptic ulcer disease, Oliveira et al. found 100% sensitivity in children with ulcer disease, which was significantly better than for children with nonulcer dyspepsia (21). They suggested a more robust immune response to H. pylori antigens in children with duodenal ulcer, related to a longer duration of infection or a greater bacterial load. We could not confirm their findings, since neither the sensitivity nor the antibody concentrations in serum were significantly higher in children with peptic ulcer disease. Due to the low number of patients with ulcer disease (n = 10) in our study, these findings have to be interpreted with caution.

Sensitivity was very low for the two IgA-based tests, independent of age group (Table 1), sex, and nationality. Only a quarter to a half of the infected children had positive results. This lower sensitivity for specific IgA antibodies in comparison to IgG antibodies confirms the result of previous studies of children (2, 7). Best et al. report a sensitivity of only 50% for IgA in a microsphere immunofluorescence assay, while a sensitivity of 100% was found for IgG in the same cohort in the same study (2). Measurement of specific IgA antibodies in addition to IgG antibodies hardly improved the sensitivity of the tests. With Enzygnost II, only one additional child and two with Pyloriset would have been identified if both test results (IgG plus IgA) had been considered for interpretation of infection status. However, the costs would have been doubled.

Regarding specificity of the tests, only a few false-positive results were observed in all four tests without any age dependence. These results are similar or, in some tests, better than the specificities obtained in serological EIAs of adults (11, 13, 20). Positive serology results are evidence of contact with H. pylori but do not necessarily indicate current infection. Persistent antibodies after clearance of H. pylori infection are probably the main reason for false-positive test results (6, 13, 29). After successful eradication therapy, antibody titers decrease significantly within the first months (9, 12) but may remain above the validated cutoff values for many months or even years. Cutler et al. observed antibody titers above the cutoff in 21 (72%) of 29 patients after a mean time of 3.5 years (6). In our study, patients with previous eradication therapy were carefully excluded. However, some children may resolve their infections spontaneously, i.e., during antibiotic treatment for other indications (25). False-positive results may also be due to cross-reacting antibodies against other bacteria like Campylobacter jejuni (16). It is unknown whether children have cross-reacting antibodies more frequently than adults.

All serological assays were performed by one person and without knowledge of the children's H. pylori status. Therefore, interperson and interlaboratory variability can be excluded as a cause of discrepant results (11).

In summary, the specificity of the four commercial EIAs was very high in symptomatic children. In contrast, the two different IgG-based tests differed markedly in sensitivity. We observed an age dependence for Pyloriset IgG, with a lower sensitivity in younger children. We speculate that young children may have a different immune response to H. pylori infection and may respond to certain H. pylori antigens only. Therefore, the Pyloriset test fails to recognize these specific antibodies. The better performance of Enzygnost II IgG needs to be confirmed in a pediatric population with a higher number of infected infants, toddlers, and preschool-aged children before the test can be considered to be reliable in children. The sensitivity of the two IgA-based tests is very low, and adding the IgA EIA to the IgG test doubles the cost without improving accuracy. We conclude that all commercially available EIAs for the diagnosis of H. pylori infection in children require validation in the pediatric population under consideration before the test can be adopted for clinical use.