Validation of serological tests for Helicobacter pylori infection in an Irish population

Transcription

1 original paper Validation of serological tests for Helicobacter pylori infection in an Irish population NP Breslin, JM Lee, MJ Buckley, E Balbirnie 1, D Rice, 1 CA O'Morain Tallaght Regional Hospital, Tallaght, Trinity Biotech Plc, 1 Southern Cross Road, Bray, Co. Wicklow, Ireland. Abstract Background Serological tests for Helicobacter pylori using laboratory and 'office' formats are commonly used, easy to perform, inexpensive and widely available. Local validation of test performance is required. Aims This study examined the performance of a laboratory and 'office' ELISA in a population of Irish dyspeptics presenting for endoscopy. Methods Consecutive patients presenting for endoscopy had blood drawn at sedation. Samples were analysed using two ELISA formats; a standard laboratory format and an 'office' ELISA test card. H. pylori infection was diagnosed by analysis of antral and corpus biopsies using the rapid urease test, culture and histology. A combination of two positive invasive tests was considered indicative of infection. Results The sensitivity and specificity of laboratory ELISA was 82.4% and 85% respectively while the values for the 'office' ELISA were 87.7% and 85.7% respectively. In patients under 45 years sensitivities and specificities of the 'office' test exceeded 90%. The two serological tests agreed in 87.5% of subjects. Conclusions Both tests performed satisfactorily. However, indeterminate results impaired the usefulness of the laboratory ELISA particularly when using a new cut-off. The 'office' ELISA performed particularly well in young patients. A simpler test using antigens from locally prevalent strains to optimise accuracy is awaited. Introduction Infection of the gastric mucosa by Helicobacter pylori induces both local and systemic immune responses. 1-3 The systemic response permits detection of immunoglobulins in blood, saliva and urine providing the clinician with minimally invasive methods of detecting exposure to the bacterium. 4-7 Of these, serological methods have been the most extensively investigated. A number of antibody test formats have been developed for diagnosing Helicobacter exposure including enzyme-linked immunoassay (ELISA), latex agglutination and immunoblotting ELISA tests can follow a standard laboratory format or be performed by the physician or practice nurse on specialised test cards; referred to as near patient tests, rapid or office ELISAs. ELISA tests for Helicobacter pylori are being increasingly used particularly in the general practice setting. 12 However, because of geographical variation in the infecting strains, and the resultant immunoglobulin profiles in infected individuals, local validation is recommended The Maastricht Consensus Report endorses the use of locally validated laboratory ELISAs. 19 Judgement was reserved on the use of 'office' ELISA's pending further validation studies. The primary aim of our study was to examine the ability of two commercially supplied serological tests for Helicobacter pylori (a laboratory ELISA and an 'office' ELISA; Trinity Biotech's IgG H. pylori ELISA and Trinity Biotech's IgG H. pylori Serocard) to accurately identify infected and non-infected adults in a population of Irish dyspeptics. Methods Consecutive dyspeptic patients referred from primary care and presenting for upper gastro-intestinal endoscopy at the Meath and Adelaide Hospitals, Dublin, Ireland were invited to take part in this validation study. Patients were excluded if they had received previous eradication therapy or had ingested antisecretory drugs within the previous 15 days or antibiotics within the previous month. All patients gave informed consent. Prior to sedation a 5ml sample of venous blood was drawn, centrifuged and stored at 20 C until analysis. Samples were analysed in batches according to two formats. The first followed a standard laboratory enzyme linked immunoabsorbent assay format and the second followed a rapid or office-type format. At the time of ELISA performance investigators were blind to the Helicobacter pylori status of individuals. Laboratory ELISA format A second generation ELISA was used. Test wells were coated with commercially supplied semi-purified antigens derived 190

2 Validation of serological tests for Helicobacter pylori infection in an Irish population from culture filtrates of Helicobacter pylori. Each sample was diluted to 1 in µl aliquots of the diluted sample were taken and added to the test well. These were incubated for 30 minutes at 25 C. The wells were washed three times using wash reagent. 100(l of enzyme conjugate (a horseradish peroxidase/anti-human globulin complex) was subsequently added to each test well and incubated for a further 30 minutes at 25 C. The test wells were washed a further four times. 100µl of enzyme substrate (tetra-methyl-benzidine) was added and incubated for five minutes at room temperature. The reaction was stopped by the addition of 100µl of 4N H 2 SO 4. Test plates were read on a spectrophotometer at 450nm and the optical density recorded. The manufacturers recommended cut-off for distinguishing positive and negative results was an optical density of 0.2. Rapid ELISA format Each test card contained two sample ports and two test ports. The first test port contained semi-purified native Helicobacter pylori antigens. The second test port did not and was designed to detect non-specific immunoglobulin binding (NSB port). The patient's specimen was added to each sample port followed by wash reagent to move the sample towards the test port. Wash reagent was then added to each test port followed by enzyme reagent. After a further washing step enzyme substrate was added. The test was read after five minutes. A sample was deemed positive for Helicobacter pylori infection when the right test port turned a blue colour when compared to no colour, or colour change of a lighter intensity, in the NSB test port. A sample was deemed negative when no colour developed in the test port. The result was deemed indeterminate if the colour change in the NSB port was of a similar or greater intensity to that in the test port. All patients considered underwent an endoscopy where macroscopic lesions were noted and biopsies for Helicobacter pylori infection were obtained. These were analysed by three methods: Histology Specimens (2 antral and 2 corpus) were placed in formalin and stained with haematoxylin and eosin. The presence of Helicobacter-like organisms and type and severity of gastritis was noted. Culture An antral biopsy was placed in transport medium and subsequently plated onto chocolate agar media. Plates were incubated for four to seven days at 37 C under microaerophilic conditions. Helicobacter pylori were identified by typical colony morphology and the presence of urease positivity. Rapid Urease Test The CLOtest R was used to detect the presence of Helicobacter derived urease in single antral biopsies (Delta West Pty Ltd, Bentley 6102, Western Australia). In all cases the CLOtest R was brought to room temperature before use and was monitored for up to 24 hours for diagnostic colour changes. 20,21 Statistical Analysis When a combination of tests for Helicobacter pylori is used the sensitivity and specificity for detection of the bacterium approach 100%. 22,23 In the primary analysis, where any two invasive tests were positive infection was deemed to be present. Persons with a single positive invasive test or all negative tests were considered uninfected. A secondary analysis was performed where a single positive test was considered indicative of infection. Standard probability analysis was performed on the data including calculation of sensitivity, specificity, positive predictive value, negative predictive value and disease prevalence. The effect of age and gender on test performance was examined. In the case of the laboratory ELISA the cut-off optical density for positive and negative results was varied. 15 A receiver operating characteristic curve then was drawn to determine the optical density which optimised test performance. 16,24-26 Results One or both ELISAs was performed in 104 patients undergoing endoscopy for the investigation of upper gastrointestinal symptoms. According to the pre-defined gold-standard criteria 58 (55.8%) of these had evidence of current Helicobacter infection. Fifty-six (53.9%), 54 (51.9%) and 36 (34.6%) individuals were Helicobacter pylori positive by rapid urease test, histology and culture respectively. Neither the rapid urease test nor culture yielded a positive result where the other two criteria were negative. Endoscopy was normal in 58 patients. There were 17 duodenal ulcers, 14 cases of gastro-oesophageal reflux, four cases of duodenitis, nine cases of gastritis and two cases of Barrett's oesophagus. The mean age of participants was 46.7 (range years). There were 63 female subjects. Laboratory ELISA The laboratory serology test was performed in 91 consecutive patients. A receiver operating characteristic curve was constructed to determine the optimum optical density cut-off point between positive and negative results in order to maximise the sensitivity and specificity (Figure 1). In the population evaluated an optical density of units was deemed to be optimum giving a sensitivity of 82.4% (95% CI: %) and specificity of 85.0% (95% CI: 73.9%-96.1%). At this cut-off the positive and negative predictive values were 87.5% (95% CI: %) and 79.6% (95% CI: %) respectively. At an optical density of 0.2 the specificity was 95%. However, the sensitivity fell to 62.8%. The optical density measurements for the 91 samples are presented in the scatterplot in Figure 2. Office ELISA The office serological test was performed in 102 patients. The sensitivity was found to be 87.7% (95% confidence interval: %) and specificity was 85.7% (95% confidence interval: %). The positive predictive value was 89.3% (95% CI: %) and the negative predictive value was 83% (95% CI: 72.7%-94.8%). If positivity in one invasive test was taken as confirmatory evidence for infection the specificity increased to 90% and positive predictive value to 92.9%. The negative predictive value showed a modest fall to 83.7%. The effect of altering acceptable gold standard criteria on test performance is illustrated in Table 1. There were three indeterminate results due to colour change 191

3 in the non-specific binding port. In all three cases the gold standard criteria for Helicobacter pylori infection were negative. In the 56 subjects less than 45 years, the sensitivity and specificity of the rapid ELISA was 93% and 92% respectively. However, these values were not significantly different from those in the over 45 age group (p= 0.2 and 0.16 respectively). The sensitivities and specificities for the laboratory ELISA were very similar in both age groups. Both ELISA tests were performed in 90 of the subjects. Agreement between the two tests was found in 77 subjects giving an inter-elisa variability of 12.5% (excluding the three indeterminate rapid results). Intra-test reliability was not assessed. 27 Discussion The reported sensitivities and specificities of ELISA tests for Helicobacter pylori vary considerably, and local validation is advised to determine test performance in the population in whom it is intended for use. 28 In our study this resulted in the selection of a new optimum cut-off level for the laboratory ELISA. However, this cut-off level has no validity outside of the population from which this sample was drawn. Furthermore, this newly defined cut-off would require confirmation in a prospective study. When compared to gold standard invasive criteria both serological test formats demonstrated satisfactory performance in classifying infected versus non-infected patients. However, owing to sample size the confidence intervals are wide and it is possible that the performance of the tests might in reality be better or indeed poorer. In addition, as illustrated in Figure 1, a number of samples tested using the laboratory ELISA had optical density readings that were close to this cut-off (0.115) and would fall into the so-called grey zone resulting in indeterminate results. 27 The rapid ELISA appeared to perform better than its laboratory counterpart although this did not reach statistical significance. In those less than 45 years of age the sensitivity and specificity exceeded 90%. Research is underway to develop a more accurate test using antigens from locally prevalent strains. In our report, the rapid ELISA test was an easy to perform seven-step procedure giving a result in 5-10 minutes. Rapid test formats have been simplified to three-step procedures that use capillary blood samples. 29 More recently these have been further simplified to one-step procedures. 30 In view of this a simpler protocol is being developed which is anticipated will give the clinician an accurate, inexpensive, easy to perform test; providing a result before the end of the consultation. Positive serological tests provide evidence of exposure to the organism in question. In the present report a proportion of individuals with positive serological tests had no evidence of current Helicobacter infection. In general practice, where serology is being increasingly employed in the assessment of dyspepsia, up to 90% of those with a positive serological test Table 1: The effect of varying acceptable 'gold-standard' criteria on test performance for laboratory and office ELISAs. Office ELISA Laboratory ELISA Number of positive One test Two tests One test Two tests invasive tests Sensitivity(%) (95% CI) ( ) ( ) ( ) ( ) Specificity(%) (95% CI) ( ) ( ) ( ) ( ) Positive predictive value(%) (95% CI) ( ) ( ) ( ) ( ) Negative predictive value(%) (95% CI) ( ) ( ) ( ) ( ) receive eradication therapy. 12 Therefore, a high positive predictive value is desirable to avoid unnecessary healthcare costs, drug toxicity and the development of bacterial resistance. The positive predictive value can be reduced where goldstandard techniques fail to diagnose true infection due to sampling errors, transport or culture difficulties or errors of interpretation. 31,32 By using a combination of diagnostic techniques we sought to minimise these effects. The acceptance of one positive test as evidence for infection improved the positive predictive value and specificity of both ELISAs. This raises the issue of what constitutes gold standard diagnostic criteria. A wide variety have been applied in reported serological validations which may account, in part, for varying test performance and hinders comparisons between tests. 27 Guidelines are required on this issue. False negative tests may have occurred due to a difference between infecting antigens in individual patients and those on the test plate. False positive results may occur if the patient has had previous eradication therapy, spontaneously lost the infection or where there is non-specific immunoglobulin binding, cross-reacting antibodies 33 or atrophic gastritis 14,17,34 In our study multiple washing steps were performed to reduce nonspecific binding and patients were excluded if they had received previous eradication therapy. Careful histopathological exami- Figure 1: Receiver operating characteristic curve for 'laboratory' ELISA according to varying optical density 192

4 Validation of serological tests for Helicobacter pylori infection in an Irish population Figure 2: Scatterplot showing optical densities obtained using the laboratory ELISA for positive and negative invasive test results. nation of multiple mucosal biopsies failed to reveal atrophic gastritis in any of the patients considered. The use of semi-purified antigens may have contributed to reduced specificity. The prevalence of Helicobacter infection in the group as a whole was 56%. This may not be representative of prevalence in primary care where a lower prevalence is likely to have a beneficial effect on the negative predictive value and detrimental effect on the positive predictive value of serological tests. 27 Recent reports of other ELISA validations suggest sensitivities and specificities of over 90%. 35 However, the performance of serological tests appears to depend on the laboratory where they are performed. 36 There is also evidence of poorer test accuracy for 'office' ELISAs when used in primary care. 37 Therefore, the practitioner needs to be aware of the performance of the serological test in his/her practice population. This report highlights the importance of validating the performance of an ELISA test for Helicobacter pylori infection in the population in whom it is to be used. Both assays had reasonable accuracy when adjustments to the cut-off for the laboratory ELISA were made. However, this introduced potential difficulties with indeterminate results. The 'office' ELISA had higher probabilities for correctly categorising infected and non-infected individuals. In those aged under 45 years sensitivities and specificities exceeded 90%. It is envisaged that a simplified single-step version with confirmation of similar test performance would provide a useful tool in the management of the young dyspeptic presenting to the primary care physician or specialist. In view of their lower cost, serological tests remain an attractive option. However, the performance of those offering optimum accuracy requires confirmation in different populations. Disclosure: Dr Balbirnie and Dr Rice are employed by Trinity Biotech, PLC References 1. Crabtree JE. Gastric mucosal inflammatory responses to Helicobacter pylori. Alim Pharmacol Ther 1996; 10 Suppl 1: Ernst, PB, Jin, Y., Reyes, VE et al. The role of the local immune response in the pathogenesis of peptic ulcer formation. Sc J Gastroenterol (Supplement) 1994; 205: Gobert B, Bene MC, Faure G et al. IgG, IgA and IgM antibodies to Campylobacter pylori detected by dot-elisa and ELISA. In: Gastroduodenal Pathology and Campylobacter pylori. Megraud, F. and Lamouliatte, H. Exerpta Medica, Amsterdam 1989; Luzza F, Maletta M, Imeneo M et al. Salivary-specific immunoglobulin G in the diagnosis of Helicobacter pylori infection in dyspeptic patients. Am J Gastroenterol 1995; 90(10): Christie JM, McNulty CA, Shepherd NA et al. Is saliva serology useful for the diagnosis of Helicobacter pylori? Gut 1996; 39(1): Patel P, Mendall, MA, Khulusi S et al. Salivary antibodies to Helicobacter pylori: screening dyspeptic patients before endoscopy. Lancet ; 344(8921): Alemohammad MM, Foley TJ, Cohen H. Detection of immunoglobulin G antibodies to Helicobacter pylori in urine by an enzyme immunoassay method. J Clin Microbiol 1993; 31(8): Danielsson D, Blomberg B, Jarnerot G et al. Heterogeneity of Campylobacter pylori as demonstrated by co-agglutination testing with rabbit antibodies. Sc J Gastroenterol (Supplement) 1988;142: Andersen LP. The antibody response to Helicobacter pylori infection and the value of serological tests to detect H. Pylori and for post treatment monitoring. In: Helicobacter pylori: Biology and Clinical Practice. Goodwin, CS and Worsley BW. CRC Press, Boca Raton 1993; Westblom TU, Madan E, Gudipati S et al. Diagnosis of Helicobacter pylori infection in adult and pediatric patients by using Pyloriset, a rapid latex agglutination test. J Clin Microbiol 1992; 30(1): Hirschl AM, Hirschl MM, Berger J et al. Evaluation of a commercial latex test for serological diagnosis of Helicobacter pylori infection in treated and untreated patients. Eur J Clin Microbiol Infect Dis 1991; 10(11): Lim AG, Martin RM, Montileone M et al. Helicobacter pylori serology and the management of young dyspeptics: a UK survey of gastroenterologists and general practitioners with an interest in gastroenterology. Alim Pharmacol Ther 1997; 11(2): Mitchell HM, Hazell SL, Kolesnikow T et al. Antigen recognition during progression from acute to chronic infection with a caga-positive strain of Helicobacter pylori. Infect Immunol 1996; 64(4): Kaldor J, Tee W, Nicolacopolous C et al. Immunoblot confirmation of immune response to Campylobacter pyloridis in patients with duodenal ulcers. Med. J. Aus ; 145(3-4): Glupczynski Y. Methodological aspects of serology for the diagnosis of Helicobacter pylori infection. Eur J Gastroenterol Hepatol 1993; 2(Supplement): van den Oever HL, Loffeld RJ, Stobberingh EE. Usefulness of a new serological test (Bio-Rad) to diagnose 193

5 Helicobacter pylori-associated gastritis. J Clin Microbiol 1991; 29(2): von Wulffen, H. An assessment of serological tests for detection of Helicobacter pylori. Eur J Clin Microbiol Infect Dis 1992; 11(7): Marchildon PA, Ciota LM, Zamaniyan FZ et al. Evaluation of three commercial enzyme immunoassays compared with the 13C urea breath test for detection of Helicobacter pylori infection. J Clin Microbiol 1996; 34(5): Anonymous. Current European concepts in the management of Helicobacter pylori infection. The Maastricht Consensus Report. European Helicobacter Pylori Study Group. Gut 1997; 41(1): Laine L, Estrada R, Lewin DN et al. The influence of warming on rapid urease test results: a prospective evaluation. Gastrointest Endoscop 1996; 44(4): Yousfi MM, El-Zimaity HM, Cole RA et al. Does using a warmer influence the results of rapid urease testing for Helicobacter pylori? Gastrointest Endoscop 1996; 43(3): Cutler AF, Havstad S, Ma CK et al. Accuracy of invasive and noninvasive tests to diagnose Helicobacter pylori infection. Gastroenterology 1995; 109(1): Thijs JC, van Zwet AA, Thijs, WJ et al. Diagnostic tests for Helicobacter pylori: a prospective evaluation of their accuracy, without selecting a single test as the gold standard. Am J Gastroenterol 1996; 91(10): Talley NJ, Kost L, Haddad A et al. Comparison of commercial serological tests for detection of Helicobacter pylori antibodies. J Clin Microbiol 1992; 30(12): Talley NJ, Newell DG, Ormand JE et al. Serodiagnosis of Helicobacter pylori: comparison of enzyme-linked immunosorbent assays. J. Clin. Microbiol ; 29(8): Hoek FJ, Noach LA, Rauws EA et al. Evaluation of the performance of commercial test kits for detection of Helicobacter pylori antibodies in serum. J Clin Microbiol 1992; 30(6): Feldman RA, Evans SJ. Accuracy of diagnostic methods used for epidemiological studies of Helicobacter pylori. Alim Pharmacol Ther 1995;9 (Supplement 2): Bodhidatta L, Hoge CW, Churnratanakul S et al. Diagnosis of Helicobacter pylori infection in a developing country: comparison of two ELISAs and a seroprevalence study. J Infect Dis 1993; 168(6): Graham DY, Evans DJJ, Peacock J et al. Comparison of rapid serological tests (FlexSure HP and QuickVue) with conventional ELISA for detection of Helicobacter pylori infection. Am. J. Gastroenterol. 1996; 91(5): Moayyedi P, Neville P, Mapstone DS et al. Validation of a new one-step near patient Helicobacter pylori test. Gut 1997; 41(Supplement 3): A Christensen AH, Gjorup T, Hilden J et al. Observer homogeneity in the histologic diagnosis of Helicobacter pylori. Latent class analysis, kappa coefficient, and repeat frequency. Sc J Gastroenterol 1992; 27(11): Morris A, Ali MR, Brown P et al. Campylobacter pylori infection in biopsy specimens of gastric antrum: laboratory diagnosis and estimation of sampling error. J Clin Path 1989; 42(7): Andersen LP, Espersen F. Immunoglobulin G antibodies to in patients with dyspeptic symptoms investigated by the western immunoblot technique. J Clin Microbiol 1992; 30(7): Granberg C, Mansikka A, Lehtonen OP et al. Diagnosis of Helicobacter pylori infection by using pyloriset EIA-G and EIA-A for detection of serum immunoglobulin G (IgG) and IgA antibodies. J Clin Microbiol 1993; 31(6): Wilcox MH, Dent, TH, Hunter JO et al. Accuracy of serology for the diagnosis of Helicobacter pylori infection-- a comparison of eight kits. J Clin Path 1996; 49(5): Feldman RA, Deeks JJ, Evans SJ. Multi-laboratory comparison of eight commercially available Helicobacter pylori serology kits. Helicobacter pylori Serology Study Group. Eur J Clin Microbiol Infect Dis 1995; 14(5): Duggan, A, Elliott, C, and Logan, R F A. How accurate is "near patient" H.pylori testing in primary care? Gut 1998; 42(Supplement 1): A82. Correspondence to: Professor Colm O'Morain, Dept. of Gastroenterology, Tallaght Regional Hospital, Tallaght, Dublin 24, Ireland. 194