Infections in Heterophile-Negative Patients

Abstract

To the Editor.—Tsaparas et al1 published an interesting study describing a cost-effective algorithm for evaluating patients with suspected infectious mononucleosis (IM) and negative heterophile antibody tests. The study has 3 problems. First, the authors assumed that all heterophile antibody–positive patients had IM caused by Epstein-Barr virus (EBV). This incorrect assumption, which unfortunately appears in other published algorithms as well,2 inflates the cost savings of their algorithm, since approximately 2% to 3% of these patients will not have EBV and will undergo treatment and monitoring for a disease they do not have.3–5 Second, the heterophile antibody test used in this study had a sensitivity of only 78% for IM. (There were 28 false-negative results detected by EBV IgM enzyme-linked immunosorbent assay. Assuming a false-positive rate of 2.5%, only 102 of the 150 positive results would be true positives. See Bayesian analysis in the Table.) Third, the authors assume a 0% false-negative rate for lymphocyte counts greater than 4 × 109/L or presence of atypical lymphocytes to detect EBV among heterophile antibody–negative patients. While probably true, this assumption was based on results from a very small control group. With only 50 subjects, the 95% confidence interval would include 0% if the actual false-negative rate was between 0% and 1.96%.The algorithm developed by Tsaparas et al might save money, although not as much as estimated since the authors assumed that all heterophile antibody–negative patients normally would have received an EBV IgM enzyme-linked immunosorbent assay. (Many physicians would look at the negative heterophile antibody result in conjunction with a complete blood count report and reassess their differential diagnosis.) A better way to achieve substantial cost savings would be to implement a clinical algorithm to differentiate IM and IM-like diseases from other diseases and use it in conjunction with the testing algorithm. Only 170 of 1921 patients (prior probability = 9%) with physician requests for heterophile antibody testing had IM or an IM-like disease. Doubling the prior probability to 18% in British Columbia would result in 5762 fewer heterophile antibody and complete blood count with differential tests, 144 fewer false-positive IM diagnoses, and (using the algorithm generated by Tsaparas et al) 47 fewer sets of viral serology tests.