eLetters to:

ARTICLES: Robert R. Tanz, Michael A. Gerber, William Kabat, Jason Rippe, Roopa Seshadri, and Stanford T. Shulman Performance of a Rapid Antigen-Detection Test and Throat Culture in Community Pediatric Offices: Implications for Management of Pharyngitis Pediatrics 2009; 123: 437-444 [Abstract] [Full text] [PDF]

eLetters: Submit a response to this article

eLetters published:

Spectrum Bias and Non-Independent Tests

Michael A. Kohn, Thomas B. Newman   (16 March 2009)

Management of pharyngitis: Response to Kohn and Newman

Robert R. Tanz, Micchael A. Gerber, Stanford T. Shulman   (3 June 2009)

Spectrum Bias and Non-Independent Tests 16 March 2009
Michael A. Kohn, Associate Clinical Professor University of California, San Francisco, Thomas B. Newman Send letter to journal: Re: Spectrum Bias and Non-Independent Tests michael.kohn{at}ucsf.edu Michael A. Kohn, et al.	Tanz et al (1) show that the sensitivity of both a rapid antigen detection test (RADT) and office-based throat culture for group A streptococcal (GAS) pharyngitis depend on the pre-test probability of disease as determined by the McIsaac score. We agree with their conclusions but would like to point out two problems with Table 5 and comment on the use of the term spectrum bias. The first problem with Table 5 is that the odds ratios and confidence intervals cannot be correct. One of the confidence intervals does not include the point estimate; another confidence interval is not symmetric (on a log scale) around the point estimate. The second problem is that odds ratios conditional on the presence or absence of GAS (as determined by the gold standard) are needed to demonstrate that the performance characteristics of the index tests depend on the McIsaac score. As currently written, the table reports that among all study subjects, those with a McIsaac score >2 had about 3 times higher odds of positive RADT or office-based culture, relative to those with a McIsaac score ¡Ü 2. But higher odds of having a positive test could be due simply to higher odds of having GAS, which is true of those with higher McIsaac scores. To make the point about spectrum bias, they would need to report the odds ratios among the subjects with and without GAS separately. Only in these analyses conditional on GAS would the odds ratios be predicted to be close to 1 in the absence of spectrum bias. Tanz et al follow Lachs et al (2), using the term ¡°spectrum bias¡± to refer to the dependence of test performance characteristics on history and exam findings. We find it easier think about whether the test and the clinical evaluation are independent, conditional on disease state.(3) The McIsaac score and the RADT are not conditionally independent. Among patients with GAS, a high McIsaac score makes a true-positive RADT more likely, perhaps because both the high McIsaac score and the true-positive RADT are associated with heavier growth of the bacteria. In this same group of patients, all with GAS, a low McIsaac score will be more often associated with a false-negative RADT, because of lower bacterial growth. 1. Tanz RR, Gerber MA, Kabat W, Rippe J, Seshadri R, Shulman ST. Performance of a rapid antigen-detection test and throat culture in community pediatric offices: implications for management of pharyngitis. Pediatrics. 2009 Feb;123(2):437-44. 2. Lachs MS, Nachamkin I, Edelstein PH, Goldman J, Feinstein AR, Schwartz JS. Spectrum bias in the evaluation of diagnostic tests: lessons from the rapid dipstick test for urinary tract infection. Ann Intern Med. 1992 Jul 15;117(2):135-40. 3. Newman TB, Kohn MA. Evidence-based diagnosis. Cambridge ; New York: Cambridge University Press; 2009. pp. 105-106. Conflict of Interest: None declared
Management of pharyngitis: Response to Kohn and Newman 3 June 2009
Robert R. Tanz, Professor of Pediatrics Children's Memorial Hospital, Northwestern University Feinberg School of Medicine, Micchael A. Gerber, Stanford T. Shulman Send letter to journal: Re: Management of pharyngitis: Response to Kohn and Newman rtanz{at}northwestern.edu Robert R. Tanz, et al.	Drs. Kohn and Newman identified errors in the reporting of the odds ratios and/or confidence intervals for the office BAP and the laboratory BAP in Table 5 of our article (1). We suspected the errors in the table were typographical in origin. However, in seeking to identify the source of the errors and correct them, we discovered that all of the odds ratios and the confidence intervals were incorrect, although the p-values remained the same. Table 5 should have the following odds ratios (95% CI): RADT 3.99 (3.08, 5.18), p<.001; Office BAP 3.23 (2.56, 4.07), p<.001; Laboratory BAP 2.60 (2.10, 3.22), p<.001. We regret the errors in Table 5 and we thank Drs. Kohn and Newman for leading us to identify and correct them. We have requested that the error be corrected in the on-line version of the article. Drs. Kohn and Newman also note that the greater odds of having a positive test when the McIsaac score is >2 could be due to the greater odds of having GAS when the McIsaac score is higher. We agree and we think it is enough to recognize this relationship without exploring the reasons for it. We have shown that when a patient has signs and symptoms typical of GAS infection the sensitivity of both office tests for GAS (RADT and BAP culture) is greater than when signs and symptoms are less consistent with streptococcal pharyngitis. In areas with low rates of acute rheumatic fever and other GAS pharyngitis-related complications it is appropriate to take advantage of the effect of clinical presentation on test sensitivity and obtain a throat swab only on those patients more likely to have GAS pharyngitis. Limiting testing to patients more likely to have GAS is the first step in reducing overuse of antibiotics for pharyngitis. Reference 1. Tanz RR, Gerber MA, Kabat W, Rippe J, Seshadri R, Shulman ST. Performance of a rapid antigen detection test and throat culture in community pediatric offices: Implications for management of pharyngitis. Pediatrics 2009;123:437-444. Conflict of Interest: None declared