Helicobacter pylori - Journal of Clinical Microbiology - American ...

2 downloads 9 Views 60KB Size Report
3 (1.8%) false-positives and no false-negatives occurred with the Pyloriset, and 3 (1.8%) false-positives and 1. (0.6%) false-negative occurred with the HM-CAP.

JOURNAL OF CLINICAL MICROBIOLOGY, Oct. 1999, p. 3328–3331 0095-1137/99/$04.00⫹0 Copyright © 1999, American Society for Microbiology. All Rights Reserved.

Vol. 37, No. 10

Detection of Helicobacter pylori Antibodies in a Pediatric Population: Comparison of Three Commercially Available Serological Tests and One In-House Enzyme Immunoassay ¨ RN KJERSTADIUS,2 LILLEMOR JANSSON,2 BENGT SUNNERSTAM,1* TORBJO 3 ¨ M,4 AND JAN EJDERHAMN5 JOHAN GIESECKE, MATS BERGSTRO Department of Pediatrics1 and Department of Clinical Microbiology,2 Central Hospital, Karlstad, and Department of Epidemiology, Swedish Institute for Infectious Disease Control,3 and Department of Clinical Chemistry4 and Department of Pediatrics,5 Huddinge Hospital, Stockholm, Sweden Received 4 February 1999/Returned for modification 23 March 1999/Accepted 14 June 1999

A serum immunoglobulin G enzyme immunoassay (EIA) for Helicobacter pylori antibodies already in use in adults was evaluated with 99 pediatric serum samples to determine its usefulness for the study of H. pylori disease in children. The reference method used was either the 13C-urea breath test or a biopsy culture of gastric mucosa. In children, an EIA cutoff of 0.35 absorbancy unit yielded sensitivity, specificity, and positive and negative predictive values of 93, 97, 93, and 97%, respectively. The cutoff recommended when this EIA was published for use in adults was 0.70 absorbancy unit (H. Gnarpe, P. Unge, C. Blomqvist, and S. Ma ¨kitalo, APMIS 96:128–132, 1988). Another subset of 169 serum samples taken from children was analyzed by four serological tests in order to compare the performance of the in-house EIA with the Pyloriset, HM-CAP, and Helico-G kits. For the 169 samples, 10 (5.9%) false-positives and no false-negatives occurred with the Helico-G, 3 (1.8%) false-positives and no false-negatives occurred with the Pyloriset, and 3 (1.8%) false-positives and 1 (0.6%) false-negative occurred with the HM-CAP. For the 169 samples, 1 (0.6%) false-positive and no falsenegatives occurred with the in-house EIA. Serological detection of H. pylori antibodies with our EIA seems to be valuable in diagnosing H. pylori infection in children, but only if a lowered, specific pediatric cutoff is established. The commercial kits, particularly the Helico-G, seem to overdiagnose pediatric H. pylori infection. A positive serological test for H. pylori infection, particularly for children, needs to be confirmed by a reference method because of the possibility of spontaneous eradication of infection, with a lingering serological response. Helicobacter pylori infections are acquired in childhood (13). In the United States and in northern Europe, chronic H. pylori infection in people less than 20 years of age is rare (17). However, recent data (12) suggests that infection with H. pylori which later clears spontaneously might well occur in more than 10% of Swedish children less than 2 years of age. In developing countries, chronic H. pylori infection in children is still very common (17) and is the precursor of peptic ulcer disease (8) as well as mucosa-associated lymphoid tissue lymphoma or gastric cancer development (4, 6, 15, 16) in both children and adults. Several commercial tests detecting H. pylori immunoglobulin G (IgG) antibodies in serum or whole blood are now available for clinical use. When children are tested for H. pylori antibodies, it is important to choose a method which has already been validated in a pediatric population. Some studies have indicated that cutoff limits for serum IgG enzyme immunoassays (EIAs) for H. pylori should be set higher for adults than for children (5). The first goal of this study was to validate a whole-cell serum IgG EIA (11) for children, so that it might be used in future pediatric epidemiological studies. This EIA had previously been used only for adults, with a cutoff of 0.70 absorbancy unit. The second goal was to compare the performance of the new EIA with the performance of three commercially available kits.

MATERIALS AND METHODS Samples. For validation of the EIA, 99 blood samples collected from 66 children for two prior studies (unpublished data) of pediatric H. pylori infection were used. Repeat blood samples were collected from each individual in the aforementioned studies to assess the rate of seroconversion during a 1-year follow-up in the case of the first study and a 2-year follow-up in the case of the second. Asymptomatic and symptomatic children took part in these two studies. Reference samples consisting of 13C-urea breath tests (13C-UBT) (83 of 99 occasions) or biopsy cultures (16 occasions) were obtained on the same day as the blood samples. The children were 1 to 17.99 years old, with a median age of 12 years. A total of 46.5% of the children were girls. For comparing commercial tests with the EIA, 242 consecutive children (0 to 17.99 years old; median age, 6 years) who visited the outpatient unit of the Department of Pediatrics, Central Hospital, Karlstad, Sweden, for minor surgery or the investigation of various pediatric disorders were asked to participate in the study during the period from March through December 1994. A total of 169 valid sera were obtained and evaluated. At least 7% (13 of 169) of all participants were of immigrant origin (southern or eastern European, Middle Eastern, or South American). 13C-UBT or biopsy culturing was used as a reference method when serological results needed to be confirmed. This was the case for 21 of 169 samples, i.e., for the samples with discordant results in the four serological tests compared and for the samples with concordant positive results. For the remaining 148 of 169 samples, no reference method was used. Patients returned for the collection of endoscopy and 13C-UBT samples 0 to 28 months (mean, 16.7 months) after the initial blood samples were collected. A new in-house EIA test done at the same time as the 13C-UBT or endoscopy showed that no seroconversions or seroreversions had occurred. All blood samples were transported to the laboratory within 8 h, immediately centrifuged, and frozen at ⫺70°C until analysis. Preparation of antigen and microtiter plates for the EIA. Details of the preparation of antigen and microtiter plates for the EIA are available from the authors. The measurement interval was 0.10 to 2.00 absorbancy units. Negative, weakly positive, and strongly positive controls were defined as having absorbancy values of ⬍0.10, 0.70 to 0.90, and ⬎1.20, respectively, at 405 nm. These three controls were run with each plate. All the control serum donors were local residents. The positive control samples were verified by a biopsy

* Corresponding author. Present address: Department of Pediatrics, ¨ rebro Medical Centre Hospital, S-701 85 O ¨ rebro, Sweden. Phone: 46 O 19 15 10 00. Fax: 46 19 18 79 15. E-mail: [email protected] 3328


VOL. 37, 1999 culture; the negative control sample was taken from a healthy blood donor. All the control sera were verified by two independent laboratories (Clinical Microbiology Laboratory, University Hospital, Malmo ¨, Sweden, and Clinical Microbiology Laboratory, Sahlgrenska University Hospital, Go ¨teborg, Sweden). Patient serum was diluted 1:100 in phosphate-buffered saline (PBS)–Tween plus 1% bovine serum albumin. The microtiter plate was washed twice in PBS-Tween. A 100-␮l portion of the serum dilution was added to each of three wells for each sample; two wells were coated with the antigen suspension, while one was uncoated and served as a blank. The plate was incubated for 90 min at 37°C. Subsequently, the plate was washed four times in PBS-Tween. A 100-␮l quantity of alkaline phosphatase-IgG (35 ␮g/ml; diluted 1:300 in PBS-Tween plus 1% bovine serum albumin) was added to each well. Afterward, the plate was incubated for 120 min at 37°C and washed four times in PBS-Tween before 100 ␮l of p-nitrophenyl phosphate substrate was added to each well. The absorbancy was then read automatically every 5 minutes at 405 nm with a computer-controlled Titertek Multiscan MC apparatus (Esselte, Vienna, Austria). The absorbancies of the other samples in the plate were read when the weak control sample had reached a predetermined value of 0.70 ⫾ 0.10 absorbancy unit. The average of the results from the two test wells for each sample (minus the blank well result) was judged to be the absorbancy of the sample. To establish the pediatric cutoff value, the sensitivities and specificities of several different cutoff values were compared in a receiver-operator curve (ROC) system. The chosen pediatric cutoff value was subsequently used in the comparison study. Commercial kits. The Pyloriset (Orion Diagnostica, Espoo, Finland), HMCAP (Enteric Products, Inc., Stony Brook, N.Y.), and Helico-G (Porton Cambridge, Newmarket, United Kingdom) kits were used according to the manufacturer’s recommendations. (The Pyloriset was further developed after this study was concluded [Pyloriset EIA-G update] [22]. Our results do not apply to this new version.) 13 C-UBT. Patients fasted for 6 h before the test. No test meal or citrus juice was given. All samples were taken in triplicate and analyzed in duplicate. A baseline breath sample was taken at the inception of the test. 13C-urea was then dissolved in 25 ml of water and given orally, according to weight. Children weighing less than 25 kg were given 25 mg of 13C-urea, those weighing between 25 and 50 kg were given 50 mg, and those weighing more than 50 kg were given 100 mg. During the next 30 min, the patients rested and were allowed neither food nor drink. At the end of this 30-min period, a new breath sample was taken. The samples were stored in a refrigerator until they were mailed, unchilled, to one of the authors (M.B.) for analysis. This modified 13C-UBT, when used for an adult population, had sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 92, 95, 96, and 91%, respectively (14). In a limited series of 43 children with biopsy-verified H. pylori infection (unpublished data), the sensitivity of this 13C-UBT was 100%, and the specificity was 80%. Biopsy culturing. For each patient for whom upper endoscopy was performed, a biopsy specimen was taken from the antrum and one was taken from the greater curvature of the corpus. Within 4 h of endoscopy, the biopsy specimens were cut into sections and placed on an agar plate as described by Bernander (2). When the culture was examined for the first time after 3 to 4 days, H. pylori colonies were recognized as small, red colonies. This finding was verified by the further identification in a heat-fixed Gram stain of a gram-negative, spiral rod with urease, catalase, and oxidase activities. ROCs. An ROC curve analysis plots pairs of sensitivity and specificity values for a diagnostic method (18). These pairs are selected by randomly choosing multiple different cutoff points and calculating sensitivity and specificity for each of them. In an ROC chart, the perfect diagnostic test would appear as close as possible to the upper left-hand corner, where both sensitivity and specificity are 100%. An ROC curve can be used to choose the cutoff point that best suits the investigator’s purpose. That is, if one wishes to eliminate false-positives, one would choose a leftward cutoff that minimizes false-positives (high specificity and therefore lower sensitivity). If, on the other hand, one wishes to eliminate false-negatives, a more rightward cutoff that minimizes false-negatives would be chosen (high sensitivity and therefore lower specificity).

RESULTS Evaluation of the EIA. Of 99 reference samples (13C-UBT or biopsy culturing), 29 (29.3%) were positive and 70 (70.7%) were negative. When different cutoffs were examined in an ROC system, the previously described adult cutoff of 0.70 absorbancy unit yielded a sensitivity of 38%, a specificity of 100%, a PPV of 100%, and an NPV of 80%. A lower pediatric cutoff of 0.35 absorbancy unit yielded a sensitivity of 93%, a specificity of 97%, a PPV of 93%, and an NPV of 97%. A still lower cutoff of 0.15 absorbancy unit resulted in higher sensitivity but lower specificity (sensitivity, 97%; specificity, 70%; PPV, 64%; and NPV, 98%). We therefore decided to use the cutoff of 0.35 in the comparison study.


TABLE 1. Correlation of serum IgG with reference methods No. of samples with the following result:

Serum IgG level (absorbancy untis)

⬍0.150 0.150–0.350 ⬎0.350 Total no. of samples




Total no. of samples





1 0 9

4 2 0

0 1 18

45 17 2

50 20 29






The correlation of serum IgG levels and reference methods is shown in Table 1. Comparison of commercial tests. Results were available for 169 samples. In 152 of 169 samples, the results were concordant; i.e., all serological tests performed on the sample gave a uniformly positive or negative result in all the commercial tests and in the in-house EIA. Samples with H. pylori-positive results in all tests (4 of 169) were regarded as positive only when confirmed by the 13C-UBT or biopsy culturing. Given the low seroprevalence of H. pylori infection in Swedish children (12, 13), a false-negative result in all tests for one sample would be highly improbable. Therefore, samples that were found negative in all serological tests (148 of 169) were regarded as truly negative, and the children were not investigated by the 13CUBT. In 17 of 169 samples, the results of the serological tests were discordant; i.e., there were some positive and some negative results. The 17 patients were further evaluated with the 13 C-UBT. After 13C-UBT testing, cases with positive results were designated H. pylori positive and cases with negative results were designated H. pylori negative. In total, 16 of 17 subjects with discordant results tested negative in the 13CUBT, and 1 of 17 tested positive in the 13C-UBT. Among the 169 samples, there were 10 (5.9%) false-positives and no false-negatives with the Helico-G, 3 (1.8%) false-positives and no false-negatives with the Pyloriset, and 3 (1.8%) false-positives and 1 (0.6%) false-negative with the HM-CAP. There were 1 (0.6%) false-positive and no false-negatives with the in-house EIA. DISCUSSION Our in-house EIA uses a crude whole-cell antigen mixture (11), whereas the commercial tests evaluated are based on several types of antigens. The Pyloriset (19) and the Helico-G (21) both use an acid extract of H. pylori antigen. The HMCAP (10) uses a mixture of H. pylori high-molecular-weight cell-associated proteins. Our method of detecting H. pylori antibodies in patients by using a heterogeneous mixture of different H. pylori antigens appears to be more sensitive than methods based on purified antigens. This would be the case if the heterogeneous antigen mixture gave the EIA the capacity to recognize a broad spectrum of H. pylori antibodies. The result would be a higher sensitivity, although perhaps at the cost of a lower specificity. Several different antigen types have been used in the H. pylori serum IgG EIA methods reported in studies by other authors. These antigens have been either purified or crude. As early as 1989, Evans et al. (10) reported an EIA involving a highmolecular-weight cell-associated protein (HM-CAP) for adults. The results were excellent, with sensitivity, specificity, PPV, and NPV of 98.7, 100, 100, and 98.6%, respectively.



However, the study population of 300 persons included only 25 persons less than 20 years old. Blecker et al. (3) validated the commercial kit Malakit (Biolab, Limal, Belgium) against biopsy culturing for a symptomatic pediatric population of Belgian-born Caucasians and found sensitivity, specificity, PPV, and NPV of 96, 96, 89, and 99%, respectively. The Malakit method is reported to use a combination of a urease fraction and a structural protein whose exact composition has not been disclosed by the manufacturer. Drumm et al. (9), using a sonicated whole-cell antigen, also reported very good EIA results (sensitivity, specificity, PPV, and NPV of 96, 99, 96, and 99%, respectively). Crabtree et al. (6), using a sonicated antigen preparation from one H. pylori strain, found a sensitivity of 86% and a specificity of 98% when examining 69 children with abdominal pain. Czinn et al. (7), also using a whole-cell lysate, reported 93% sensitivity and 93% specificity for a group of 35 children. He concluded that the somewhat complex purification of H. pylori antigens used by Evans et al. (10) was not necessary for children. Thomas et al. (20) reported that when a crude H. pylori cell extract was compared to a partially purified urease antigen preparation, only 14 of 20 children had increased anti-urease IgG levels, while 19 of 20 had increased IgG levels against the crude cell extract. Whole-cell lysate assays seem to be best suited for use in pediatric populations because not all children seem to mount antibodies against urease, and the results of studies by authors using low-molecular-weight antigens have not been published in full. A very purified antigen might also underestimate seroprevalence by not recognizing all antibodies occurring in a population with a mixed geographical origin. Although not proven, different populations may harbor different H. pylori antigens (13), a fact which would support the necessity of local validation of H. pylori EIAs. Since cross-reactions have been described (1)—particularly between Campylobacter jejuni and H. pylori—some authors have recommended adsorption with C. jejuni antigen in order to increase the specificity of crude antigen assays. Another possible explanation of the occurrence of false-positives in the EIA would be accidental or spontaneous eradication of H. pylori, with a lingering serological response. This notion would be more plausible in a western European pediatric population with few intestinal infections. Because of the small number of false-positives in our results, no adsorption step was considered necessary. Lower cutoff values for children than for adults have been recommended. Crabtree et al. (5) found that 50% of all children with H. pylori gastritis would have been considered seronegative if the adult cutoff value had been used. The probable explanation is that children have not been exposed to as many different potentially cross-reacting antigens as adults and therefore have lower background activity. Blecker et al. (3), however, found no increase in normal optical densities with age and used the same cutoff value for children as for adults. Our results are in accordance with those of Crabtree et al. (5), because we found that a lower cutoff value had to be set for the pediatric group than for the adult group. Unlike Crabtree et al. (5), we arrived at this conclusion by using ROC analysis instead of standard deviations over the mean of the negative IgG values. As it was not possible to use the 13C-UBT for all 169 cases, we chose to examine the concordant seropositives and those with discordant results in the serological assays. The endoscopy and breath test samples were collected months after the initial blood samples were obtained. Transient infections, therefore, could have been missed. Our findings indicate that the Pyloriset and the HM-CAP


both yield a slightly larger number of false-positives in pediatric testing than our in-house EIA does and that the Helico-G is less specific than the other two commercial tests. It is recommended that for a pediatric population, positive results obtained with any of the commercially available H. pylori serological tests evaluated above be confirmed by a 13C-UBT so that false-positives may be identified before endoscopy is done. We believe that the in-house EIA described here serves as a useful method of screening children for H. pylori infection, provided that its antigenic contents match those of the population under study and provided that a pediatric cutoff value lower than the adult cutoff value is used. ACKNOWLEDGMENTS This study was carried out with the financial support of Riksfo ¨rbundet fo ¨r Mag-och Tarmsjuka and Sven Johanssons Minnesfond. We thank Teddy Primack for assistance in preparation of the manuscript in English. REFERENCES 1. Andersen, L. P., H. Raskov, L. Elsborg, S. Holck, T. Justesen, B. Fischer Hansen, C. Moller Nielsen, and K. Gaarslev. 1992. Prevalence of antibodies against heat-stable antigens from Helicobacter pylori in patients with dyspeptic symptoms and normal persons. APMIS 100:779–789. 2. Bernander, S. 1993. Absence of Helicobacter pylori in dental plaques in Helicobacter pylori positive dyspeptic patients. Eur. J. Clin. Microbiol. Infect. Dis. 12:282–285. 3. Blecker, U., S. Lanciers, B. Hauser, and Y. Vandenplas. 1993. Diagnosis of Helicobacter pylori infection in adults and children by using the Malakit Helicobacter pylori, a commercially available enzyme-linked immunosorbent assay. J. Clin. Microbiol. 31:1770–1773. 4. Blecker, U., T. W. McKeithan, J. Hart, and B. S. Kirschner. 1995. Resolution of Helicobacter pylori-associated gastric lymphoproliferative disease in a child. Gastroenterology 109:973–977. 5. Crabtree, J. E., M. J. Mahony, J. D. Taylor, R. V. Heatley, J. M. Littlewood, and D. S. Tompkins. 1991. Immune responses to Helicobacter pylori in children with recurrent abdominal pain. J. Clin. Pathol. 44:768–771. 6. Crabtree, J. E., J. I. Wyatt, G. M. Sobala, G. Miller, D. S. Tompkins, J. N. Primrose, and A. G. Morgan. 1993. Systemic and mucosal humoral responses to Helicobacter pylori in gastric cancer. Gut 34:1339–1343. 7. Czinn, S. J., H. S. Carr, and W. T. Speck. 1991. Diagnosis of gastritis caused by Helicobacter pylori in children by means of an ELISA. Rev. Infect. Dis. 13(Suppl. 8):S700–S703. 8. Drumm, B. 1993. Helicobacter pylori in the pediatric patient. Gastroenterol. Clin. North Am. 22:169–182. 9. Drumm, B., G. I. Perez Perez, M. J. Blaser, and P. M. Sherman. 1990. Intrafamilial clustering of Helicobacter pylori infection. N. Engl. J. Med. 322:359–363. 10. Evans, D. J., D. G. Evans, D. Y. Graham, and P. D. Klein. 1989. A sensitive and specific serologic test for detection of Campylobacter infection. Gastroenterology 97:1069–1070. 11. Gnarpe, H., P. Unge, C. Blomqvist, and S. Ma ¨kitalo. 1988. Campylobacter pylori in Swedish patients referred for gastroscopy. APMIS 96:128–132. 12. Granstrom, M., Y. Tindberg, and M. Blennow. 1997. Seroepidemiology of Helicobacter pylori infection in a cohort of children monitored from 6 months to 11 years of age. J. Clin. Microbiol. 35:468–470. 13. Lindkvist, P., D. Asrat, I. Nilsson, E. Tsega, G. L. Olsson, B. Wretlind, and J. Giesecke. 1996. Age at acquisition of Helicobacter pylori infection: comparison of a high and a low prevalence country. Scand. J. Infect. Dis. 28: 181–184. 14. Oksanen, A., B. Bergstro ¨m, A. Sjo ¨stedt, A. Gad, B. Hammarlund, and R. Seensalu. 1997. Accurate detection of Helicobacter pylori infection with a simplified 13C urea breath test. Scand. J. Clin. Lab. Investig. 57:689–694. 15. Parsonnet, J., G. D. Friedman, N. Orentreich, and H. Vogelman. 1997. Risk for gastric cancer in people with CagA positive or CagA negative Helicobacter pylori infection. Gut 40:297–301. 16. Parsonnet, J., J. P. Hansen, L. Rodriguez, A. B. Gelb, R. A. Warnke, E. Jellum, N. Orentreich, J. H. Vogelman, and G. D. Friedman. 1994. Helicobacter pylori and gastric lymphoma. N. Engl. J. Med. 330:1267–1271. 17. Pounder, R. E., and D. Ng. 1995. The prevalence of Helicobacter pylori infection in different countries. Aliment. Pharmacol. Ther. 9(Suppl. 2):33– 39. 18. Sackett, D. L., B. R. Haynes, G. H. Guyatt, and P. Tugwell (ed.). 1991. Clinical epidemiology, p. 116–119. Little, Brown & Co., Boston, Mass.

VOL. 37, 1999 19. Schembri, M. A., S. K. Lin, and J. R. Lambert. 1993. Comparison of commercial diagnostic tests for Helicobacter pylori antibodies. J. Clin. Microbiol. 31:2621–2624. 20. Thomas, J. E., A. M. Whatmore, M. R. Barer, E. J. Eastham, and M. A. Kehoe. 1990. Serodiagnosis of Helicobacter pylori infection in childhood. J. Clin. Microbiol. 28:2641–2646. 21. Uyub, A. M., S. M. Raj, R. Visvanathan, M. Nazim, S. Aiyar, A. K. Anuar,



and M. Mansur. 1994. Helicobacter pylori infection in north-eastern peninsular Malaysia. Evidence for an unusually low prevalence. Scand. J. Gastroenterol. 29:209–213. 22. van de Wouw, B. A., W. A. de Boer, A. R. Jansz, R. T. Roymans, and A. P. Staals. 1996. Comparison of three commercially available enzyme-linked immunosorbent assays and biopsy-dependent diagnosis for detecting Helicobacter pylori infection. J. Clin. Microbiol. 34:94–97.

Suggest Documents