Validation of western Helicobacter pylori IgG antibody assays in Korean adults

Journal of Medical Microbiology (2015), 64, 513 518 DOI 10.1099/jmm.0.000050 Validation of western Helicobacter pylori IgG antibody assays in Korean adults Sun-Young Lee, 1 Hee-Won Moon, 2 Mina Hur 2 and Yeo-Min Yun 2 Correspondence Hee-Won Moon hannasis@kuh.ac.kr Received 5 October 2014 Accepted 4 March 2015 1 Department of Internal Medicine, Konkuk University School of Medicine, Seoul, Korea 2 Department of Laboratory Medicine, Konkuk University School of Medicine, Seoul, Korea Helicobacter pylori infection is endemic in Korea, and serology testing is widely performed. The aim of this study was to validate and compare the diagnostic accuracy of Korean and Western serological assays for H. pylori detection in Korean adults. The 114 Korean adults who visited our centre over a 6-month period for the evaluation of H. pylori infection using the urea breath test (UBT) were enrolled in this prospective study. Anti-H. pylori IgG was measured using three commercially available immunoassays: Genedia H. pylori ELISA (Green Cross Medical Science), Chorus helicobacter IgG (DIESSE Diagnostica Senese) and Vidas H. pylori IgG (biomérieux). Positive UBT findings were obtained in 40.6 % of included subjects. The sensitivities and the specificities of Vidas, Chorus and Genedia were 89.7 %, 100 % and 100 % and 85.5 %, 75.4 % and 80.7 %, respectively. We found no differences in sensitivity between the Vidas and Chorus (P50.125), Chorus and Genedia (P50.125) and Vidas and Genedia (P51.000) assays. There were also no differences in specificity between the Vidas and Chorus (P50.070), Chorus and Genedia (P50.508) and Vidas and Genedia (P50.549) assays. In Korean adults, the Genedia H. pylori ELISA, Chorus helicobacter IgG and Vidas H. pylori IgG assays exhibited a high concurrence rate with similar diagnostic accuracy. Thus, both the Korean and Western noninvasive assays are reliable for serodiagnosis of H. pylori in Korean individuals. INTRODUCTION Several non-invasive tests, which do not require endoscopy, have been developed to establish whether a patient is infected with Helicobacter pylori. The serum anti-h. pylori immunoglobulin G (IgG) antibody test, which is both simple and inexpensive, is one of these tests (Burucoa et al., 2013; Faigel et al., 2000). Although it cannot distinguish between active and recently treated infections, serology is more accurate than biopsy in endemic areas of H. pylori infection and gastric cancer. This is due to the increased likelihood of false negative results with invasive tests in patients having severe gastric atrophy and/or intestinal metaplasia (Korstanje et al., 2008). Various commercial enzyme-immunoassays (EIAs), including ELISA kits, are now available and are considered to be a reliable and valid method for detection of gastric colonization by H. pylori. However, there is some debate on the overall performance of H. pylori diagnostic kits from particular geographical regions. For example, a Chinese study found that Western commercial serological tests performed poorly when used in Chinese patients (Leung et al., 1999). Despite the high accuracy of these tests reported in the West, they yielded variable diagnostic performance in other regions, Abbreviations: AU, arbitrary units; AUC, area under ROC curve; CI, confidence interval; RFV, relative fluorescent value; TV, test value; UBT, urea breath test. possibly due to strain-specific differences in serum anti-h. pylori IgG antibody responses in different populations (Marchildon et al., 2003). Conversely, a recent Chinese study showed that a Western serology test was reliable for screening H. pylori infection in Chinese populations (Ren et al., 2005). Serology tests that were developed in Western countries are widely performed in Korea. However, it has been shown that Koreans and their Western counterparts differ in regard to the types of H. pylori they harbour, with Koreans showing a high prevalence of the East Asian CagA type (Lee, 2012, 2014; Seo et al., 2011). Therefore, serological assays should be validated in each population and a consensus report also stated that only validated commercial tests should be used (Malfertheiner et al., 2012). Despite this, most serological assays used in Korea are not validated in Asian populations and have not been directly compared to Korean assays for detection of H. pylori antigen in Korea, where the prevalence of gastric cancer is the highest in the world. Therefore, we aimed to validate and compare the diagnostic accuracy of Korean and Western serological assays for H. pylori detection in Korean adults. METHODS Sample size calculation. We calculated the minimal number of infected cases and non-infected cases required to ensure a 10 % precision of the sensitivity and specificity estimate (x50.1). We 000050 G 2015 The Authors Printed in Great Britain 513

S.-Y. Lee and others estimated that the sensitivity (or specificity) of the evaluated kits is 90 % (P50.9) with a 95 % confidence interval (Burucoa et al., 2013). Using the equation [cases5(1.96) 2 p(12 p)/x 2 ](Banoo et al., 2010), we required 35 infected and non-infected patients. With 50 % estimated prevalence, it was necessary to recruit a total sample of 75 patients. Subjects. Korean adults who visited our centre between January and June of 2014 for the evaluation of H. pylori infection using the urea breath test (UBT) were included in this prospective study. The inclusion criteria were Koreans over 16 years old who had completed UBT and blood sampling on the same day at our centre. Subjects were excluded when they had a recent history of medication, including H. pylori eradication, within the previous 6 months. Other exclusion criteria were the presence of significant disease that required treatment, pregnancy, lactation or incomplete data in any of the performed tests. Informed consent was obtained and the study was approved by the institutional review board (IRB) of Konkuk University School of Medicine, which confirmed that it was in accordance with the ethical guidelines of the Helsinki Declaration. After the approval, this study was registered as ClinicalTrials.gov ID: KCT0001187 (https://cris.nih.go.kr). Serum anti-h. pylori antibody test. Blood samples were obtained from each participant after 8 h of fasting. These were subjected to centrifugation, and serum was stored at 270 uc prior to analysis. Serum levels of anti-h. pylori IgG antibodies were measured using three commercially available immunoassays: Genedia H. pylori ELISA (Green Cross Medical Science), Chorus H. pylori IgG (DIESSE Diagnostica Senese) and Vidas H. pylori IgG (biomérieux). All three tests were performed using the same blood sample, according to the manufacturer s instructions. Genedia H. pylori ELISA test. The Genedia kit is an ELISA that uses peroxidase-conjugated anti-human IgG. This assay uses H. pylori antigen obtained from Korean H. pylori strains. The cut-off value for this kit was defined as mean absorbance of the negative control +0.4 OD at 450 nm, as specified in the manufacturer s instructions. The sensitivity and specificity of this assay in Korean adults have been reported as 97.8 % and 92.0 %, respectively (Kim et al., 1998). Chorus H. pylori IgG test. For the Chorus H. pylori IgG assay, H. pylori antigen-specific immunoglobulins are bound to solid phase and detected by peroxidase-conjugated anti-human IgG monoclonal antibodies. The titre was expressed in AU (arbitrary units); sample IgG concentrations.12 AU were considered positive, and those,8 AU were considered negative. The samples between these cut-off levels were considered equivocal. The manufacturer s reported sensitivity and specificity are were 89 % (95 % confidence interval [CI]: 77 95 %) and 100 % (95 % CI: 92 100 %), respectively. A study from Italy found that, in Italian adults, this assay displayed a sensitivity and specificity of 95 % and 100 %, respectively (Figura et al., 1994). Vidas H. pylori IgG test. The Vidas H. pylori IgG assay, which utilizes the enzyme immunoassay sandwich method, is combined with a final fluorescent detection and thus is described as an enzymelinked fluorescent assay. During the final detection step a relative fluorescent value (RFV) was calculated, and the test value (TV) was obtained by dividing the sample RFV by standard RFV. The result was considered positive when TV 1.00 and negative when TV,0.75. The samples between these cut-off levels were considered equivocal. The manufacturer s claimed sensitivity and specificity for the Vidas H. pylori IgG test are 98.1 % (95 % CI: 93.1 99.8 %) and 90.82 % (95 % CI: 83.3 95.7 %), respectively. Definitive diagnosis for H. pylori infection. The gold standard for H. pylori infection in this study was the UBT. It was conducted in the fasting state on the same day as the blood sampling for the serological assays, according to the manufacturer s instructions. Briefly, 100 mg of C 13 urea packed into a capsule was ingested by each patient, and exhaled breath samples were analysed using Heliview mass spectrometry (Medichems). Breath samples at baseline and 30 min after 13 C urea administration were collected. A difference in 13 CO 2 concentration between the baseline and 30 min samples of over 2.4 was recorded as a positive finding for H. pylori infection. The results were expressed as positive or negative, with a difference in 13 CO 2 concentration shown as parts per thousand. If all three serology tests and the UBT were positive, the subject was classified as having a definite H. pylori infection. If all of the tests were negative, the subject was classified as having no infection. In subjects who showed discrepancies between the three serology tests, those with a positive UBT were defined as likely to have an infection and those with a negative UBT were classified as unlikely to have an infection. Statistical analysis. Sensitivity, specificity, positive predictive values and negative predictive values were calculated based on confirmation of H. pylori infection status using the UBT. Values were expressed as a percentage, with 95 % CI, and a P-value,0.05 was considered statistically significant. Statistical analysis was performed using MedCalc Statistical Software (version 11.2.1, MedCalc Software). McNemar s test was used to calculate the statistical difference between assays. Continuous variables were summarized as mean± SD, and categorical variables were expressed as frequency (%). Agreement between assays was assessed using Cohen s kappa coefficient. Poor agreement was reported when the kappa value was,0.4, while excellent agreement was reported when the kappa value was.0.75. Kappa values between 0.4 and 0.75 were reported as fair to good agreement (Malfertheiner et al., 2012). Equivocal results of Vidas and Chorus were considered negative for analysis. We determined that 100 samples with estimated prevalence of 50 % were required to achieve 80 % power, in order to detect a difference of 0.1000 between a diagnostic test with an area under the ROC curve (AUC) of 0.8000 and another diagnostic test with an AUC of 0.9000 using a twosided z-test at a significance level of 0.0500. The data were discrete (rating scale) responses. The AUC was computed between false-positive rates of 0.000 and 1.000. The ratio of the SD of the responses in the negative group to the SD of the responses in the positive group for diagnostic test 1 was 1.000 and for diagnostic test 2 was 1.000. The correlation between the two diagnostic tests was assumed to be 0.600 for both positive and negative groups (Obuchowski & McClish, 1997). All analyses were conducted using PASS (http://www.ncss.com/software/pass). RESULTS Subject characteristics Of the 114 subjects who visited our centre for evaluation, 18 were excluded due to a recent history of H. pylori eradication or other medication. Positive UBT findings were obtained for 39 (40.6 %) of the 96 included subjects (Table 1). Thirty-five subjects (36.5 %) showed positive findings in all three serological tests and UBT, whereas 39 subjects (40.6 %) showed negative findings in all three serological tests and UBT. There were four subjects (34-, 36-, 42- and 61-year-old males) who tested negative by UBT but gave positive findings in all three serological tests. Comparison of serological assays using UBT as the predefined gold standard The sensitivities and the specificities of Vidas, Chorus and Genedia were 89.7 %, 100 % and 100 % and 85.5 %, 75.4 % 514 Journal of Medical Microbiology 64

H. pylori serology test in Korean adults Table 1. Baseline characteristics of the included subjects Subjects (n596) Age (years, mean± SD) 48.1±12.6 Gender (male, %) 61 (63.5 %) Status of H. pylori infection by UBT 39 (40.6 %) Genedia (OD, median with ranges) 0.4805 (0.0020 3.3630) Chorus (AU, median with ranges) 36.9 (5.0 200.0) Vidas (TV, median with ranges) 0.88 (0.01 5.13) UBT, Urea breath test; AU, arbitrary units; TV, test value. and 80.7 %, respectively (Table 2). There were no differences in sensitivity between the Vidas and Chorus (P50.125), Chorus and Genedia (P50.125) and Vidas and Genedia (P51.000) assays. In addition, there were no differences in specificity between the Vidas and Chorus (P50.070), Chorus and Genedia (P50.508) and Vidas and Genedia (P50.549) assays. Weighted kappa values of Vidas, Chorus and Genedia for concordance with UBT were 0.741, 0.714 and 0.773, respectively, and did not differ significantly (Table 2). Concordance rates among the three serological tests Weighted kappa values for concordance rates between the Vidas and Chorus, Chorus and Genedia and Genedia and Vidas assays were 0.747 (0.617 0.878), 0.812 (0.695 0.929) and 0.683 (0.537 0.828), respectively. Pairwise comparison of ROC curves showed that there were no statistically significant differences between Vidas and Chorus (P50.659), Genedia and Vidas (P50.810) and Chorus and Genedia (P50.483), as shown in Fig. 1. Subjects showing discrepancy between the diagnostic tests Only 18 of the 96 subjects included in this study showed discrepancy in the results of the three serology tests (Table 3). Of the three tests, only Vidas gave a false-negative finding in four subjects who tested positive by UBT. The titres measured by Vidas were between 0.7 and 0.99 (one negative and three equivocal) in these four subjects. Of the 14 samples that tested negative by UBT, falsepositives were obtained by Chorus in ten cases, by Genedia in seven cases and by Vidas in four cases. All four Vidas false-positives had low titres, measuring between 1.02 and 1.15 TV. Similarly, the ten samples that gave false-positive results in the Chorus assay exhibited low titres, giving values of between 18.0 and 108 AU, and the seven Genedia positives showed variable OD values. DISCUSSION In this study, we found that the diagnostic accuracy of the three serological assays for detection of H. pylori did not differ relative to the UBT findings. Notably, there were no differences in sensitivity and specificity among Genedia, Chorus and Vidas assays. In most of the subjects the three serology tests showed results consistent with UBT, suggesting that all of three of these non-invasive assays are reliable for serodiagnosis in Korean individuals. Here we measured only IgG antibodies, as it has been shown that IgM and IgA antibodies do not perform as well in the detection of H. pylori (Laheij et al., 1998; Talley et al., 1991). We recently showed that serum anti-h. pylori IgG antibodies are also useful in assessing the risk of gastric neoplasia and predicting pathology type in Korean adults (Choi et al., 2014). Of the 3328 participants included in this previous study, the incidence of gastric neoplasia was highest in the gastric atrophy (+)/H. pylori (2) group (4.17 %; odds ratio 25.8, P50.009), but the neoplasm exhibited the least advanced histology using the IgG serology test. Importantly, in this previous study we could not confirm whether there was a difference between Western and Korean serology kits (Choi et al., 2014) because the Genedia H. pylori ELISA test, which utilizes H. pylori antigen obtained from Korean strains, was not tested. Table 2. Performance of three different serological assays based on UBT findings Sensitivity Specificity PPV NPV Efficiency Weighted (overall accuracy) kappa value Vidas H. pylori IgG 89.7 % 85.5 % 81.4 % 92.2 % 87.2 % 0.741 (75.8 97.1 %) (73.3 93.5 %) (66.6 91.6 %) (81.1 97.8 %) (0.605 0.877) Chorus Helicobacter 100 % 75.4 % 73.6 % 100 % 85.4 % 0.714 IgG (91.0 100.0 %) (62.2 85.9 %) (59.7 84.7 %) (91.8 % 100.0 %) (0.581 0.847) Genedia H. pylori 100.0 % 80.7 % 78.0 % 100.0 % 88.5 % 0.773 ELISA (91.0 100.0 %) (68.1 90.0 %) (64.0 88.5 %) (92.3 100.0 %) (0.649 0.896) PPV, positive predictive value; NPV, negative predictive value. All values except the overall accuracy are expressed as a percentage with 95 % confidence interval. Equivocal results were considered negative in the analysis. http://jmm.sgmjournals.org 515

S.-Y. Lee and others Sensitivity 100 80 60 40 20 AUC of VIDAS: 0.952 AUC of CHORUS: 0.964 AUC of Genedia: 0.945 VIDAS CHORUS Genedia VIDAS vs. CHORUS: P=0.659 Genedia vs. VIDAS: P=0.810 CHORUS vs. Genedia: P=0.483 0 0 20 40 60 80 100 100-Specificity Fig. 1. Pairwise comparison of ROC curves. Area under the curves of Vidas, Chorus and Genedia were 0.952, 0.964, and 0.945, respectively. There was no significant difference between the groups. In our current study, we found a high concurrence rate among the three serological tests with each displaying similar diagnostic accuracies. Our findings are consistent with another recent report which demonstrated that Western serology testing can provide a reliable method for screening of H. pylori infection in Chinese populations (Ren et al., 2005). Although a meta-analysis also found that a similar overall accuracy of common commercial serological kits (Loy et al., 1996), studies from China and Japan have shown poor performance of Western commercial serological tests when used in Asian subjects (Leung et al., 1999; Marchildon et al., 2003; Miwa et al., 2000). These results may due to the use in Western kits of H. pylori antigens that are less prevalent in East Asians (Miwa et al., 2000), as antigen composition will vary according to the different strains utilized in the generation of a particular serological assay (Marchildon et al., 2003). Therefore, Western serological assays should be validated in Asians where the prevalence of H. pylori infection and gastric cancer is high. Importantly, prior to this study, the Vidas and Chorus assays had not been validated in Asians while various Western kits were evaluated in previous studies. Also, these assays had not been directly compared to Korean serological assay in Korean individuals. The Vidas H. pylori IgG test uses a pool of native purified antigens derived from five different strains (ATCC 53721, 53722, 53725, 53726, and 53727), and the Chorus H. pylori IgG test also uses H. pylori G21 and G39 cells as antigens. Antigens in these Western kits are prepared from diverse strains and represent various antigens, including both commo and strain-specific. Moreover, the prevalence of the East Asian CagA subtype was found to be 94.4 % in our population according to a previous report from our institution, suggesting that the majority of Koreans are infected with the East Asian CagA subtype of H. pylori (Seo et al., 2011). Less diversity in our population may partly explain this discrepancy in results among studies. It is notable that the three serological assays showed different diagnostic trends as compared with UBT. For example, Vidas had a tendency to give more false-negative findings, whereas Chorus had a tendency to give more false-positives. Adjustment of the cut-off level could enhance the performance of each assay according to test purposes. For example, all four Table 3. Detailed results of subjects showing discrepancies among the three serological tests UBT Subject Genedia (OD) Chorus (AU) Vidas (TV) Probably infected (n54) Positive (n54) M/29 P, 1.0930 P, 83.6 E, 0.99 M/72 P, 0.4630 P, 109 E, 0.76 M/41 P, 0.7570 P, 132 N, 0.70 F/48 P, 3.0660 P, 178 E, 0.99 Probably not infected (n514) Negative (n54) M/36 P, 0.5830 P, 18.0 N, 0.65 M/57 P, 0.7990 P, 43.8 E, 0.85 F/61 P, 0.4980 P, 99.1 N, 0.58 F/51 P, 1.7200 P, 108 E, 0.91 Negative (n53) F/58 P, 0.4200 N, 6.6 N, 0.20 M/29 P, 0.8680 N, 7.8 E, 0.76 M/32 P, 0.5560 E, 9.6 N, 0.75 Negative (n53) M/57 N, 0.2790 P, 18.2 P, 1.12 M/39 N, 0.2210 P, 24.3 P, 1.15 M/64 N, 0.0380 P, 40.4 P, 1.11 Negative (n53) F/48 N, 0.2720 P, 14.8 N, 0.37 M/36 N, 0.0330 P, 33.3 N, 0.57 M/59 N, 0.3930 P, 61.0 N, 0.43 Negative (n51) M/63 N, 0.2840 E, 10.2 P, 1.02 UBT, Urea breath test; M, male; F, female; AU, arbitrary unit; TV, test value; P, positive; N, negative; E, equivocal. 516 Journal of Medical Microbiology 64

H. pylori serology test in Korean adults subjects who showed false-negative (negative or equivocal) findings with the Vidas test had antibody titres between 0.7 and 0.99, values close to the positive cut-off ( 1.00 TV). Additionally, while there is a wide equivocal zone for the Chorus and Vidas tests, retesting is recommended for samples within 10 % of the cut-off values for the Genedia test due to the absence of an equivocal zone. For the Chorus and Vidas tests, using a higher cut-off value would decrease the overall H. pylori seroprevalence as compared with that obtained with the lower cut-off values for these tests. In the present study, equivocal results were considered negative when we compared the results to UBT. This approach is preferable for Chorus assay, but use of a low limit for the equivocal zone (0.75 TV) can increase sensitivity in the Vidas assay, which shows a sensitivity of 97.4 % and specificity of 78.2 %. The differential sensitivities and specificities obtained in various studies with the same kit indicate that it is more important to test a kit in a particular study population than to choose any specific kit (Jensen et al., 1993). Of the 14 UBT negative subjects, ten tested positive with the Chorus assay, indicating that among the three serology tests, the false positive rate was highest in Chorus. Of these ten subjects, three showed false-positive results with the Vidas test, albeit with relatively low titres between 1.11 and 1.15. Since these three subjects also showed low titres in the Chorus assay, between 18.2 and 40.4 AU, past infection or spontaneous regression of H. pylori might be the reason for false-positives in these cases. Conversely, serum from the four subjects who showed false-positives in both the Chorus and Genedia tests showed a wide range of titres, from 0.498 to 1.720 OD for Genedia, and from 18.0 to 108 AU for Chorus. Taken together, these data suggest that recommendations for the optimal use of different interpretation criteria should be made for each serology test. 13 C-UBT was used as the gold standard in this study because there is no single test that suffices as a criterion standard to detect H. pylori infection. Although UBT is considered to be one of the most reliable tests, it is not always accurate due to the various optimal cut-off values of 13 C-UBT. The presence of four subjects who tested negative by UBT but gave positive findings in all three serological tests in our study suggests a high probability of falsenegative findings by UBT under cryptic conditions. Our data suggest that the cut-off values for each test are the most important factor when evaluating the correlation and diagnostic accuracy of commonly used tests in Koreans. In conclusion, the Genedia H. pylori ELISA, Chorus helicobacter IgG and Vidas H. pylori IgG assays exhibited a high concurrence rate, with similar diagnostic accuracy, in Korean adults. Therefore, all three of these non-invasive assays are reliable for serodiagnosis in the Korean population. ACKNOWLEDGEMENTS This paper was supported by Konkuk University in 2015. The authors declare no competing interest. REFERENCES Banoo, S., Bell, D., Bossuyt, P., Herring, A., Mabey, D., Poole, F., Smith, P. G., Sriram, N., Wongsrichanalai, C. & other authors (2010). Evaluation of diagnostic tests for infectious diseases: general principles. Nat Rev Microbiol 8 (Suppl), S17 S29. Burucoa, C., Delchier, J. C., Courillon-Mallet, A., de Korwin, J. D., Mégraud, F., Zerbib, F., Raymond, J. & Fauchère, J. L. (2013). Comparative evaluation of 29 commercial Helicobacter pylori serological kits. Helicobacter 18, 169 179. Choi, H. S., Lee, S. Y., Kim, J. H., Sung, I. K., Park, H. S., Shim, C. S. & Jin, C. J. (2014). Combining the serum pepsinogen level and Helicobacter pylori antibody test for predicting the histology of gastric neoplasm. J Dig Dis 15, 293 298. Faigel, D. O., Magaret, N., Corless, C., Lieberman, D. A. & Fennerty, M. B. (2000). Evaluation of rapid antibody tests for the diagnosis of Helicobacter pylori infection. Am J Gastroenterol 95, 72 77. Figura, N., Oderda, G. & Verdiani, S. (1994). Evaluation of a commercial ELISA kit for the serological diagnosis of Helicobacter pylori infection. New Microbiol 17, 319 325. Jensen, A. K., Andersen, L. P. & Wachmann, C. H. (1993). Evaluation of eight commercial kits for Helicobacter pylori IgG antibody detection. APMIS 101, 795 801. Kim, S. Y., Ahn, J. S., Ha, Y. J., Doh, H. J., Jang, M. H., Chung, S. I. & Park, H. J. (1998). Serodiagnosis of Helicobacter pylori infection in Korean patients using enzyme-linked immunosorbent assay. J Immunoassay 19, 251 270. Korstanje, A., van Eeden, S., Offerhaus, J. A., Waltman, F. L., Hartog, G., Roelandse, F. W., Souverijn, J. H., Biemond, I. & Lamers, C. B. (2008). Comparison between serology and histology in the diagnosis of advanced gastric body atrophy: a study in a Dutch primary community. J Clin Gastroenterol 42, 18 22. Laheij, R. J., Straatman, H., Jansen, J. B. & Verbeek, A. L. (1998). Evaluation of commercially available Helicobacter pylori serology kits: a review. J Clin Microbiol 36, 2803 2809. Lee, S. Y. (2012). Future candidates for indications of Helicobacter pylori eradication: do the indications need to be revised? J Gastroenterol Hepatol 27, 200 211. Lee, S. Y. (2014). Current progress toward eradicating Helicobacter pylori in East Asian countries: differences in the 2013 revised guidelines between China, Japan, and South Korea. World J Gastroenterol 20, 1493 1502. Leung, W. K., Ng, E. K., Chan, F. K., Chung, S. C. & Sung, J. J. (1999). Evaluation of three commercial enzyme-linked immunosorbent assay kits for diagnosis of Helicobacter pylori in Chinese patients. Diagn Microbiol Infect Dis 34, 13 17. Loy, C. T., Irwig, L. M., Katelaris, P. H. & Talley, N. J. (1996). Do commercial serological kits for Helicobacter pylori infection differ in accuracy? A meta-analysis. Am J Gastroenterol 91, 1138 1144. Malfertheiner, P., Megraud, F., O Morain, C. A., Atherton, J., Axon, A. T., Bazzoli, F., Gensini, G. F., Gisbert, J. P., Graham, D. Y. & other authors (2012). Management of Helicobacter pylori infection the Maastricht IV/ Florence Consensus Report. Gut 61, 646 664. Marchildon, P. A., Sugiyama, T., Fukada, Y., Peacock, J. S., Asaka, M., Shimoyama, T. & Graham, D. Y. (2003). Evaluation of the effects of strain-specific antigen variation on the accuracy of serologic diagnosis of Helicobacter pylori infection. J Clin Microbiol 41, 1480 1485. Miwa, H., Kikuchi, S., Ohtaka, K., Kobayashi, O., Ogihara, A., Hojo, M., Nagahara, A. & Sato, N. (2000). Insufficient diagnostic accuracy of imported serological kits for Helicobacter pylori infection in Japanese population. Diagn Microbiol Infect Dis 36, 95 99. http://jmm.sgmjournals.org 517

S.-Y. Lee and others Obuchowski, N. A. & McClish, D. K. (1997). Sample size determination for diagnostic accuracy studies involving binormal ROC curve indices. Stat Med 16, 1529 1542. Ren, Z., Borody, T., Pang, G., Dunkley, M., Clancy, R., Xia, H. H., Chu, K. M., Wong, J. & Wong, B. C. (2005). Evaluation of anti-helicobacter pylori IgG2 antibody for the diagnosis of Helicobacter pylori infection in western and Chinese populations. Aliment Pharmacol Ther 21, 83 89. Seo, T. H., Lee, S. Y., Uchida, T., Fujioka, T., Jin, C. J., Hwang, T. S. & Han, H. S. (2011). The origin of non-h. pylori-related positive Giemsa staining in human gastric biopsy specimens: A prospective study. Dig Liver Dis 43, 23 27. Talley, N. J., Newell, D. G., Ormand, J. E., Carpenter, H. A., Wilson, W. R., Zinsmeister, A. R., Perez-Perez, G. I. & Blaser, M. J. (1991). Serodiagnosis of Helicobacter pylori: comparison of enzyme-linked immunosorbent assays. J Clin Microbiol 29, 1635 1639. 518 Journal of Medical Microbiology 64