Correspondence

Cat Scratch Disease

N Engl J Med 1994; 330:370-371February 3, 1994

Article

To the Editor:

We wish to point out a potentially confusing result in the work of Zangwill and colleagues (July 1 issue)1. In the Abstract and Results section they state that the positive predictive value of the indirect fluorescent-antibody test for Rochalimaea henselae was 91 percent. We suggest that it is not possible to determine the predictive value of a test in a case-control study.

The predictive value of a test is strongly influenced by the prevalence of the condition in a population2. In a case-control study, the prevalence of a condition is artificially fixed by the number of controls chosen per case patient. For example, the authors chose 45 case patients and 112 control subjects. Thirty-eight of the case patients and four of the controls had a positive test, for a positive predictive value of 91 percent (38 divided by 42). Had they chosen 45 case patients and 336 controls (an arbitrary decision), they would have observed 12 positive test results among the controls (because the specificity of the test does not change). They would then have calculated a positive predictive value of 76 percent (38 divided by 50). This result differs in a clinically important way from the positive predictive value reported by the authors and may influence a reader's evaluation of results reported for a patient if the antibody test is used in making clinical decisions.

John M. Colford, Jr., M.D., M.P.H.
Thomas B. Newman, M.D., M.P.H.
University of California, San Francisco, San Francisco, CA 94143

2 References

1. 1

   Zangwill KM, Hamilton DH, Perkins BA, et al. Cat scratch disease in Connecticut -- epidemiology, risk factors, and evaluation of a new diagnostic test. N Engl J Med 1993;329:8-13
   Full Text | Web of Science | Medline

2. 2

   Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology: a basic science for clinical medicine. 2nd ed. Boston: Little, Brown, 1991:85.
   

Author/Editor Response

The authors reply:

To the Editor: We agree with Drs. Colford and Newman that it is not possible to determine the positive predictive value of a test in a case-control study without the use of other data. As noted, the positive predictive value is directly related to the prevalence of a disease in the population tested. The most clinically relevant estimate may be one derived in a population with lymphadenopathy of unknown cause, with or without a history of exposure to cats. Jackson et al. found that persons diagnosed as having cat scratch disease accounted for 0.8 to 15.6 percent (depending on age group and season) of those admitted to hospitals for lymph-node procedures (defined as incision, excision, or biopsy).1 If our calculated sensitivity and specificity (84 percent and 96 percent, respectively) are used in combination with these data, the positive predictive value among persons hospitalized for lymph-node procedures ranges from 15 to 79 percent.

If it had been possible to restrict the data of Jackson et al. to persons with lymphadenopathy and one or more of the risk factors we described for cat scratch disease (owning a kitten, having been bitten or scratched by a kitten, and owning a kitten with fleas), we would expect the positive predictive value of this test to be substantially higher. Therefore, the use of the indirect fluorescent-antibody test to measure antibodies against R. henselae in selected populations is likely to reduce the use of antigen preparations in skin tests for cat scratch disease, reliance on clinical diagnosis alone, and the need for surgical biopsy.

Bradley A. Perkins, M.D.
Jay D. Wenger, M.D.
Centers for Disease Control and Prevention, Atlanta, GA 30333

1 References

1. 1

   Jackson LA, Perkins BA, Wenger JD. Cat scratch disease in the United States: an analysis of three national data bases. Am J Public Health 1993;83:1707-1711
   CrossRef | Web of Science | Medline

Citing Articles (1)

Citing Articles

1. 1

   C. B. Ogbunugafor, S. Basu, N. M. Morales, P. E. Turner. (2010) Combining mathematics and empirical data to predict emergence of RNA viruses that differ in reservoir use. Philosophical Transactions of the Royal Society B: Biological Sciences 365:1548, 1919-1930
   CrossRef