An evaluation of whole blood testing for Helicobacter pylori in general practice

Transcription

1 Aliment Pharmacol Ther 1998; 12: 641±645. An evaluation of whole blood testing for Helicobacter pylori in general practice N. J. TALLEY, J. R. LAMBERT*, S. HOWELL, H. H.-X. XIA, S. K. LIN* & L. AGREUS Department of Medicine, University of Sydney & Nepean Hospital, Sydney, Australia; and *PPO Pharmaco, Cambridge, UK Accepted for publication 6 April 1998 SUMMARY Background: Rapid whole blood tests for Helicobacter pylori infection were developed to assist in the management of patients with dyspepsia in general practice. However, they have not been extensively tested in this setting. Aim: To investigate the test characteristics of the BM- Test (Helisal Quick Test) when used in general practice. Method: One hundred and ten dyspeptic patients attending local general practitioners were recruited into the study. The BM-Test was administered by the general practitioner at the screening visit according to standard instructions supplied with the test kit. The patient was then referred to Nepean or Mornington Peninsula Hospitals for further assessment, including a 14 C-urea breath test. The test kit was forwarded to the appropriate hospital centre for an independent, blinded reading. The sensitivity and speci city of the BM-Test were evaluated against the results of the 14 C-UBT. Results: Based on general practitioner readings, the BM- Test had a sensitivity of 59.3% and a speci city of 90.2%. The positive and negative predictive values were 87.5% and 65.7%, respectively. When based on independent readings, sensitivity rose to 71.2% and speci- city fell to 88.2%. The BM-Test was more sensitive for older patients than for younger patients when based on both the general practitioner and independent readings. Conclusion: The BM-Test performs below the generally recommended sensitivity and speci city of 90% required for clinical practice. INTRODUCTION The evaluation and management of dyspepsia continues to be a challenge in general practice. 1 Dyspepsia has many possible causes, the most common being peptic ulcer disease, gastro-oesophageal re ux disease and functional (or non-ulcer) dyspepsia. Endoscopy remains the most reliable method of separating organic disease from functional dyspepsia. However, since relevant endoscopic pathologies are absent in the majority of patients with dyspepsia, this method results in costly and unnecessary referrals for many patients. Empirical therapy aimed at Helicobacter pylori eradication may provide a cost-effective approach for the Correspondence to: Prof. N. J. Talley, Department of Medicine, University of Sydney, Nepean Hospital, Penrith, NSW 2751, Australia. management of some patients with dyspepsia. Ulcer disease is unlikely to occur in the absence of H. pylori infection, and successful eradication of this bacterium offers excellent curative prospects for peptic ulcer disease. 1 The British Society of Gastroenterology currently recommends that serological tests for H. pylori should be used to help guide the decision to perform endoscopy in younger patients who present without sinister symptoms and are not taking non-steroidal antiin ammatory drugs. 2 Rapid whole blood tests for H. pylori infection are now available. With these, ` nger-pick' samples of blood can be tested for H. pylori antibodies in general practice of ces. The results of these tests are available after 5±10 min and subsequent patient management decisions can be made immediately. It has been estimated that in practice, testing for H. pylori infection using these Ó 1998 Blackwell Science Ltd 641

2 642 N. J. TALLEY et al. kits should reduce the demand for endoscopy by 25± 37%. 1,3, 4 To date, the Helisal Rapid Blood Test (called the BM- Test in Australia) has attracted the most attention as a viable in practice kit for assessing H. pylori status. However, the diagnostic value of the BM-Test remains controversial. Existing studies show marked variation in the accuracy of this test, with reported sensitivities ranging from 63% to 91% and speci cities from 56% to 94%. 3±10 Furthermore, dif culties in the reading and interpretation of this test have been reported; 7 equivocal results and inter-observer errors are common. Although the BM-Test was developed speci cally for use in primary care, it has not been extensively tested in this setting. 3 Our aim was to validate this test in detecting H. pylori infection in family practice. METHODS The study was conducted through the Departments of Medicine at Nepean Hospital in Sydney (NSW, Australia) and Mornington Peninsula Hospital in Melbourne (Victoria, Australia). Local general practitioners were recruited by each centre to provide patients for the study. In total, 10 general practitioners provided patients to the Nepean Hospital centre and eight general practitioners provided patients to the Mornington Peninsula Hospital centre. Eligibility was determined by participating general practitioners. Patients aged above 18 years and presenting with dyspepsia were considered for inclusion. Dyspepsia was de ned as persistent or recurrent upper abdominal pain or discomfort over the preceding 3- month period. Excluded were patients with a history of gastrointestinal bleeding or upper gastrointestinal surgery; patients taking antibiotics, proton pump inhibitors or medications containing bismuth in the preceding 4 weeks; pregnant or lactating women; and patients with any condition associated with poor compliance (e.g. drug or alcohol abuse, mental illness or dementia). Written informed consent was obtained by the general practitioner once eligibility for inclusion was con rmed. The general practitioner then administered the BM-Test according to standard instructions provided with each test kit (manufactured by CORTECS Diagnostics Ltd, UK, and distributed by Boehringer Mannheim GmbH, Germany). The test was read within 30 min of the colour reagent being drained through the test area of the kit, and the result was recorded and forwarded to the appropriate centre (Nepean or Mornington Peninsula Hospital), together with the used test kit. An appointment was then made for the patient to attend the hospital for a 14 C-urea breath test. Hospital appointments were scheduled to occur within 1 week of the BM-Test. Staff at the two research centres then provided a blinded, independent reading of the BM-Test when the test kit was received, usually 1 day after the BM-Test had been performed. Breath samples from the 14 C-UBT were collected and sent to Mornington Peninsula Hospital for the measurement of radioactivity. Differences in the prevalence of H. pylori infection between groups of patients was assessed using the chisquared test, with Yates's correction if required. Ninety- ve per cent con dence intervals were calculated for proportions. All P-values calculated were two-tailed; the alpha level of signi cance was set at 5%. The study was approved by the Institutional Ethics Committees of Nepean and Mornington Peninsula Hospitals. RESULTS Overall, 125 patients were tested with the BM-Test, but 15 did not attend the scheduled appointment for the 14 C-UBT. Thus, complete data was available for 110 patients. Of these, 65 (59.1%) were recruited through the Mornington Peninsula centre and 45 (40.9%) were recruited through the Nepean centre. The sample consisted of 50 males (45.5%) and 60 females (54.5%). There were no differences between the two centres with respect to patient gender (P ˆ 0.18). The age ranged from 25 to 85 years with a mean of 51.9 years (s.d. ˆ 14.4 years) and a median of 51 years (interquartile range: 39±64 years). Patient age did not vary as a function of gender (P ˆ 0.99) or recruitment centre (P ˆ 0.60). Forty patients (36.4%, 95% CI: 27.4±45.4%) were diagnosed as H. pylori positive from the BM-Test, as read by the general practitioner. This rose to 48 patients (43.6%, 95% CI: 34.4±52.8%) when the BM-Test was read by recruitment centre staff. A positive 14 C-UBT was returned by 54% (95% CI: 44.4±62.8%) of patients (n ˆ 59). While the Kappa statistic suggested excellent agreement between general practitioner readings of the BM-Test and readings made by the recruitment centre staff (Kappa ˆ 0.81), it is notable that disagreement occurred on eight (7.3%) of the tests. All these discrepant tests were read to be negative by the general practitioner

3 WHOLE BLOOD TESTING FOR H. PYLORI 643 but positive by the centre staff; seven tests were positive and one negative by the 14 C-UBT (Table 1). The sensitivity and speci city (and likelihood ratios) of the BM-Test are shown in Table 2, with positive and negative predictive values. Separate gures are provided for the general practitioner and centre staff readings of the BM-Test. The 14 C-UBT was used as the gold standard in all analyses. Based on general practitioner readings, the BM-Test had a sensitivity of 59.3% and a speci city of 90.2%. The positive and negative predictive values were 87.5% and 65.7%, respectively. When research centre staff readings were substituted for the general practitioner readings, the sensitivity of the BM-Test rose to 71.2%, while the speci city fell to 88.2%. Positive and negative predictive values were 87.5% and 72.6%, respectively (Table 2). The prevalence of H. pylori infection was 58.5% (95% CI: 46.5±70.5%) in Melbourne patients and 46.7% (95% CI: 32.1±61.3%) in Sydney patients (v 2 ˆ 1.48, Table 1. Correlation of the BM-Test and the 14 C-UBT in detection of H. pylori infection in general practice 14 C-UBT Melbourne Sydney Overall BM-Test + ± Total + ± Total + ± Total GP reading* ) Total Staff reading** ) Total *Result was read by general practitioner (GP) according to the manufacturer's instruction; **result was re-read by staff in the research centre. Table 2. Sensitivity and speci city of the BM-Test using 14 C-UBT as the gold standard Sensitivity Speci city Likelihood ratios PPV* NPV (%) (%) Positive Negative (%) (%) All patients GPà reading Centre reading Melbourne patients GPà reading Centre reading Sydney patients GPà reading Centre reading Patients aged 45 years or younger GPà reading Centre reading Patients aged above 45 years GPà reading Centre reading *PPV, positive predicitive value; NPV, negative predictive value. àgp, general practitioner.

4 644 N. J. TALLEY et al. P ˆ 0.22). The prevalence of H. pylori infection in males (62%, 95% CI: 48.5±75.5%) and females (46.6%, 95% CI: 34.0±59.2%) was not signi cantly different. Similarly, the prevalence of H. pylori infection was not signi cantly different between patients aged 45 years or younger (46.2%, 95% CI: 30.6±61.8%) and those older than 45 years (57.7%, 95% CI: 46.2±69.2%). The performance of the BM-Test was slightly better in Sydney than in Melbourne. Using general practitioner readings, sensitivities and speci cities were 61.9% and 91.7% amongst Sydney patients, compared to 57.9% and 88.9% amongst Melbourne patients. Similar trends were evident using research centre staff readings, with sensitivities and speci cities of 80.9% and 95.8% amongst Sydney patients, and 65.8% and 81.5% amongst Melbourne patients (Table 2). The BM-Test was more sensitive for older patients (aged above 45 years) than for younger patients (45 years or younger). Sensitivities, as calculated using general practitioner and centre staff readings, respectively, were 47.1% and 64.7% for younger patients and 65.9% and 75.6% for older patients. The corresponding speci cities were 90.5% and 85.7% for younger patients and 90.0% and 90.0% for older patients, respectively (Table 2). DISCUSSION The BM-Test was developed to facilitate detection of H. pylori infection in primary practice, its diagnostic value in this setting remains a key issue. In this study, the sensitivity and speci city of the test, as determined by general practitioner readings, were 59.3% and 90.2%, respectively. The test was more sensitive in older patients (65.9% vs. 47.1%), but was not more speci c. These results differ from an earlier study of the BM-Test in primary practice. Jones et al. 3 reported a sensitivity of 83% and a speci city of 78%, and noted that there were no age effects. However, methodological dissimilarities between these two studies may account for differences in the results. In our study, the BM-Test was evaluated against the 14 C-UBT, which is now recognized as one of the gold standard tests for existing H. pylori infection. In contrast, Jones et al. evaluated the BM-Test against the combined results of serological testing, using Helisal Serum and Helico G (ELISA) methods. In summary, while Jones' results suggest that the BM-Test compares favourably with other serological tests, our results indicate that it compares unfavourably with more reliable methods of detecting H. pylori infection. We observed that when based on centre staff readings, the sensitivity of the BM-Test improved from 59.3% to 71.2%, while the speci city fell from 90.2% to 88.2%. The positive predictive value remained unchanged, while the negative predictive value increased from 65.7% to 72.6%. This may be due to the fact that test characteristics improve with operator familiarity, as demonstrated by Jones et al. 3 Another possibility is that the test result became clearer when the test kit was sent to the Centre. Despite an improvement, the test still performed below the generally recommended 90% sensitivity and speci city for clinical practice. 11 The value of rapid whole blood testing for H. pylori infection lies in its ability to in uence referral rates in general practice. We did not examine this issue. However, data reported by Jones et al. suggest that test results had the desired effect on the management of dyspepsia, with reduced referrals amongst H. pyloripositive patients. 3 In the light of these ndings, we support the concept of rapid whole blood testing for H. pylori infection in general practice. However, the BM- Test lacks the characteristics that would make it a viable diagnostic tool in general practice (i.e. it lacks adequate sensitivity and speci city). The accuracy of a negative test result still remains a key issue. Patients with a negative test are not likely to be retested; consequently, it is essential that this outcome is associated with a high degree of certainty. 1 The negative predictive value provides an index of certainty in a negative test result, but is in uenced by the prevalence of the disease under investigation. Table 3 shows the negative predictive values and associated false negative rates for different prevalences of H. pylori infection. The estimates are based on the sensitivities and speci cities reported in this study. These results suggest that the BM-Test may provide unacceptable false negative rates when used in the general population and amongst dyspeptic patients. The population prevalence of H. pylori infection is» 25% in the relevant age group (aged < 45 years), and the negative predictive value falls below 90% at this prevalence level. Presumably, the prevalence of H. pylori infection is higher amongst patients presenting with dyspepsia than amongst the general population (estimates of 54% and 57% were obtained in this study and by Jones et al., 3 respectively). Negative predictive values fall below 75% at these prevalences. Indeed, the negative predictive value approaches an acceptable level

5 WHOLE BLOOD TESTING FOR H. PYLORI 645 Table 3. Negative predictive values and false negative rates for different prevalences of H. pylori infection (above 90%) only when the baseline prevalence of H. pylori infection is at 10%. In conclusion, we believe that rapid whole blood testing for H. pylori infection has a place in general practice. However, the BM-Test does not appear to provide the required accuracy in the relevant patient population. ACKNOWLEDGEMENTS Sensitivity Speci city NPV* FNR (%) (%) (%) (%) Prevalence of 54% GPà reading Centre reading Prevalence of 25% GPà reading Centre reading Prevalence of 10% GPà reading Centre reading *NPV, negative predictive value; FNR, false negative rate. àgp, general practitioner. This study was supported by Boehringer Mannheim, Australia. REFERENCES 1 Agreus L, Talley N. Challenges in managing dyspepsia in general practice. Br Med J 1997; 315: 1284±8. 2 British Society of Gastroenterology. Dyspepsia Management Guidelines. London: BSG, Jones R, Phillips I, Felix G, Tait C. An evaluation of nearpatient testing for Helicobacter pylori in general practice. Aliment Pharmocol Ther 1997; 11: 101±5. 4 Moayyedi P, Carter AM, Catto A, Heppell RM, Grant PJ, Axon ATR. Validation of a rapid whole blood test for Helicobacter pylori infection. Br Med J 1997; 314: Reilly TG, Poxon V, Sanders DSA, Elliot TSJ, Walt RP. Comparison of serum, salivary, and rapid whole blood diagnostic tests for Helicobacter pylori and their validation against endoscopy based tests. Gut 1997; 40: 454±8. 6 Stone MA, Mayberry JF, Wicks ACB, et al. Near patient testing for Helicobacter pylori: a detailed evaluation of the Cortecs Helisal Rapid Blood test. Eur J Gastroenterol Hepatol 1997; 9: 257±60. 7 Duggan A, Logan R. Validation of a rapid whole blood test for diagnosing Helicobacter pylori infection. Con icting results from the Helisal test. Br Med J 1997; 314: 1688±9. 8 Lahie RG, Ricar N. Validation of Helisal whole blood serum and saliva tests for the non-invasive diagnosis of Helicobacter pylori infection. Gut 1995; 37(Suppl. 1): A13(Abstract). 9 Yapp T, Kapur K, Thomas GOA, Swift J, Pugh S. Validation of the Helisal whole blood test for the detection of IgG antibodies to H. pylori. Gut 1995; 37(Suppl. 1): A56(Abstract). 10 Duggan AE, Knifton A, Logan RPH, Hawkey CJ, Logan RFA. Validation of two new rapid blood tests for H. pylori. Gut 1995; 37(Suppl. 1): A58(Abstract). 11 Lam SK, Talley NJ. Report of the 1997 Asia Paci c Consensus Conference on the Management of H. pylori Infection. J Gastroenterol Hepatol 1998; 13: 1±12.