Helicobacter pylori infection leads to gastritis, ulcer disease, and gastric carcinoma. In the developing world, where colonization may occur in infancy, it has been associated with infant malnutrition (1–3). Our previous work suggested that 80% of Gambian children were colonized in their first year, although this figure depends on the [13C]-urea breath test ([13C]-UBT). Whereas this is a widely used noninvasive tool for diagnosing H pylori infection, it has proved difficult to extensively validate in infancy.
[13C]-UBT results are conventionally expressed as positive or negative, relative to empirically derived cutoffs (4), and high false-positivity rates have been reported in very young children (5–7). Enzyme immunoassays detecting bacterial antigens in stool offer an alternative noninvasive diagnostic test, which may be more accurate than [13C]-UBT in this age group (8), although comparison with endoscopic diagnosis has suggested that sensitivity may be decreased in young children (9,10). In infants, endoscopic validation of noninvasive tests may be impossible; therefore, cutoff levels need to be determined by analysis of the distribution of test results.
The distribution of results for any set of UBTs includes 2 subpopulations, corresponding to the presence or absence of H pylori. Methods have been developed to describe these subpopulations, thereby defining an appropriate cutoff value to determine H pylori status (11,12).
These methods are difficult to use in younger children because the subpopulations overlap, potentially leading to false-positive results (2,6). In this study we devised a novel means of analyzing [13C]-UBT results and used it to compare these with simultaneously performed stool antigen enzyme-linked immunosorbent assay in Gambian infants.
A prospective cohort study was undertaken at the Medical Research Council Research Station in Keneba, The Gambia, the site of previous studies into H pylori in childhood (1–3). With the approval of the Gambian Government/MRC Ethical Committee, expectant mothers from the region were invited to enroll their offspring, and participating infants underwent serial [13C]-UBT throughout the first year of life, according to a previously described protocol (2). Stool samples were collected within 24 hours of breath testing and stored at −20°C.
One hundred ninety-eight infants were enrolled, and a total of 862 [13C]-UBTs were performed during the first year of life. Kernel density estimation was used to visualize the spread of the logarithmically transformed [13C]-UBT results. As expected, 2 subpopulations were observed within the distribution of results. Normal distributions were approximated to these subpopulations so that they could be described statistically. This was achieved by arithmetically reflecting the results of the left half of the lower subpopulation about the mode of that distribution (derived visually from Kernel density estimation). A similar approach was used for the right half of the upper subpopulation. The arithmetic mean and standard deviation of the discrete subpopulations were estimated, and a cutoff of 10.9δ‰ was defined as the point at which equally defined confidence intervals met between the subpopulations (Fig. 1).
The 13C -UBT results were classified as positive, negative, or equivocal, with respect to the 95% confidence intervals of the upper and lower subpopulations (Fig. 1). A mixture of positive, negative, and equivocal results from infants 2, 5, and 12 months of age were then selected. Stool samples corresponding to these [13C]-UBTs for 52 children were thawed and tested for the presence of H pylori antigen with the use of H pylori stool antigen (HpSA) kits (Meridian Diagnostics, Cincinnati, OH), in accordance with the manufacturer's instructions.
Age-specific cutoffs for the HpSA values were evaluated as described above (Fig. 1). HpSA produced more precise discrimination than did [13C]-UBT between the H pylori–positive and H pylori–negative subpopulations (Fig. 2), with HpSA cutoffs showing greater discriminative power, lying out with the confidence intervals of the separate subpopulations. The [13C]-UBT results, classified as positive, negative, or equivocal, were compared with the binary HpSA results (Table 1).
Many of the younger infants had equivocal [13C]-UBT results, but there were no equivocal results by 12 months of age. All equivocal [13C]-UBT results were negative by HpSA. In the youngest infants concordance between the tests was poor. Agreement between the 2 tests increased with age, reaching 86% in infants at 12 months of age. Evaluation of the serial results of both tests suggested that some infants were transiently colonized, although selection bias inherent to the study design precluded quantitative evaluation of this phenomenon. In the absence of invasive test results, our approach provided an effective and straightforward tool with which to determine test cutoffs, and it gave a clear indication of test precision.
In this study, HpSA discriminated between negative and positive results more precisely than did [13C]-UBT. In infants 2 and 5 months old, there was a large overlap between the [13C]-UBT results of the positive and negative subpopulations. A significant number of results were discordant with those from HpSA in these young groups. False-positive [13C]-UBT results in young infants may result from the action of oral urease-producing bacteria, whereas false-negative HpSA results may relate to low bacterial counts in recent colonization. Our results and previous work illustrate the wide spread of the [13C]-UBT results within the negative subpopulation in the first months of life (2). By 1 year of age this spread is markedly diminished, producing a narrower “gray zone” of [13C]-UBT results. This suggests that positive [13C]-UBT results in the first 6 months of life should be interpreted cautiously. Negative breath test results are more reliable. Conversely, positive HpSA results in infancy are more useful than negative test results. Effectively, this means that the HpSA has a high positive predictive value in early childhood, whereas the 13C-UBT has a high negative predictive value. The optimum test for any given situation will depend on which result is of greater interest to clinicians or researchers. The cutoff value calculated for the entire [13C]-UBT dataset was 10.9‰, far higher than commonly used values (3.5–5.5‰) but similar to a recently reported value of 8.5‰ (2,7,11,12). These figures and the observation of a widely spread negative population in the youngest infants support an age-dependent interpretation of the [13C]-UBT result. More important, they again illustrate the need for population-specific validation of the test (7,13).
Studies supporting high prevalences of H pylori infection in infancy have been challenged by reports of false-positive breath test results, but in this study, [13C]-UBT and HpSA demonstrated an equal prevalence by 1 year of age, yielding firm evidence of H pylori infection during the first year of life. Further intervention studies are now warranted to evaluate the role of H pylori in early childhood, including its potential effects on infant growth.
1. Dale A, Thomas JE, Darboe MK, et al
. Helicobacter pylori
infection, gastric acid secretion and infant growth. J Pediatr Gastroenterol Nutr 1998; 26:393–397.
2. Thomas JE, Dale A, Harding M, et al
. Interpreting the 13
C-urea breath test
among a large population of young children from a developing country. Pediatr Res 1999; 46:147–151.
3. Thomas JE, Dale A, Bunn JEG, et al
. Early Helicobacter pylori
colonisation: the association with growth faltering in The Gambia. Arch Dis Child 2004; 89:1149–1154.
4. Klein PD, Klein ER. Applications of stable isotopes to pediatric nutrition and gastroenterology: measurement of nutrient absorption and digestion using 13C. J Pediatr Gastroenterol Nutr 1985; 4:9–19.
5. Imrie C, Rowland M, Bourke B, et al
. Limitations to carbon 13-labelled urea breath testing for Helicobacter pylori
. J Pediatr 2001; 139:734–737.
6. Kinderman A, Demmelmair H, Koletzko B, et al
. Diagnostic value and errors of the 13
C-urea breath test
(UBT) in children. Gut 1997; 41(Suppl 1):A67.
7. Kindermann A, Demmelmair H, Koletzko B, et al
. Influence of age of the 13
C-urea breath test
results in children. J Pediatr Gastroenterol Nutr 2000; 30:85–91.
8. Cardinali L, de CC, Rocha GA, et al
. Evaluation of [13C] urea breath test
and Helicobacter pylori
stool antigen test for diagnosis of H. pylori
infection in children from a developing country. J Clin Microbiol 2003; 41:3334–3335.
9. Kalach N, Nguyen V, Bergeret M, et al
. Usefulness and influence of age of a novel rapid monoclonal enzyme immunoassay stool antigen for the diagnosis of Helicobacter pylori
infection in children. Diag Microbiol Infect Dis 2005; 52:157–160.
10. Raguza D, Granato CFH, Kawakami E. Evaluation of the stool antigen test for Helicobacter pylori
in children and adolescents. Dig Dis Sci 2005; 50:453–457.
11. Mion F, Rosner G, Rousseau M, et al
C-urea breath test
for Helicobacter pylori
: cut-off point determination by cluster analysis. Clin Sci 1997; 93:3–6.
12. Shepherd AJ, Williams CL, Doherty CP, et al
. Comparison of an immunoassay for the detection of Helicobacter pylori
antigens in the faeces with the urea breath test
. Arch Dis Child 2000; 83:268–270.
13. Du JX, Watkins T, Bravo LE, et al
. 13C-urea breath test
for Helicobacter pylori
in young children: cut-off point determination by finite mixture model. Stat Med 2004; 23:2049–2060.
Keywords:© 2007 Lippincott Williams & Wilkins, Inc.
Helicobacter pylori; Urea breath test; Fecal enzyme-linked immunosorbent assay; Infants