oncologist, the bronchoscopy was canceled and the patient was started on hormonal therapy. I have often complained (and heard other attending physicians complain) that medical students and residents spend more time on the computer than at the bedside with their patients. I am beginning to understand, however, that the keyboard and the mouse are powerful diagnostic tools in their own right. Clifford D. Packer, MD Correspondence: Dr Packer, Firm B Clinic, Louis Stokes Cleveland VA Medical Center, 10701 East Blvd, Cleveland, OH 44106 (clifford.packer@med.va.gov). 1. Tonelli MR. The limits of evidence-based medicine. Respir Care. 2001;46:14351440. 2. Schattner A. Simple is beautiful: the neglected power of simple tests. Arch Intern Med. 2004;164:2198-2200. 3. Fukuda M, Takashima H, Fuse H, Hirano S. Prostatic cancer with multiple pulmonary metastases treated successfully with hormonal therapy: a case report. Hinyokika Kiyo. 2002;48:499-502. Diagnostic Accuracy of Natriuretic Peptides for Heart Failure S imple and reliable tests to diagnose heart failure in elderly patients presenting with symptoms or signs of impaired left ventricular function would be of great practical importance. In recent years, brain natriuretic peptides have emerged as a promising, novel test. Enzyme-linked immunosorbent assay (ELISA), in contrast to radioimmunosorbent assay, does not require long incubation periods and would be particularly suitable to assist rapid diagnosis on site in primary and emergency care settings. The systematic review by Doust et al1 of test accuracy studies is therefore timely and of great interest, but not without limitations. We have 3 observations. First, Doust et al1 included only 3 studies that examined rapid ELISA, despite the fact that at least 4 additional studies of this test system exist. No comprehensive comparisons between the bedside ELISA and radioimmunosorbent assay can therefore be made. Second, it is unclear how the authors reached the conclusion that brain natriuretic peptide concentrations below (REPRINTED) ARCH INTERN MED/ VOL 165, MAR 28, 2005 704 a threshold of 15 pmol/L exclude heart failure in patients in whom the condition is suspected. In their metaanalysis, Doust et al1 calculated an average diagnostic odds ratio if results for more than 1 cutoff point were reported within a study. It seems inconsistent to average diagnostic odds ratios across different cutoff points while claiming that 1 cutoff point is optimal to rule out heart failure. Table 1 of their article shows the results for each study for the cutoff point nearest to 15 pmol/L, but these are not the data that were used in the meta-analysis. Indeed, we were unable to reproduce the combined diagnostic odds ratios for the various subgroups shown in Figure 1 when repeating analyses using the data presented in Table 1. It is also unclear whether the cutoff point of 15 pmol/L applies to radioimmunosorbent assay, ELISA, or both. Third, regarding the study by Hobbs et al,2 we note that the authors included a total of 630 patients from 4 different patient groups, but the total number of patients in the original publication2 is 591. This discrepancy should be clarified. In conclusion, the systematic review and metaanalysis of Doust and colleagues1 is an important contribution but also illustrates some of the problems inherent in a meta-analysis of test accuracy studies.3 We believe that further work is needed to fully elucidate the potential of the different brain natriuretic peptide tests in the diagnosis of heart failure. Markus Battaglia, MD, MPH Lucas M. Bachmann, MD, PhD Matthias Egger, MD, MScMFPHM Correspondence: Dr Egger, MRC Health Services Research Collaboration, Department of Social Medicine, University of Bristol, UK Canynge Hall, Whiteladies Road, Bristol BS8 2PR, England (Matthias.Egger@bristol .ac.uk) 1. Doust JA, Glasziou PP, Pietraz E, Dobson AJ. A systematic review of the diagnostic accuracy of natriuretic peptides for heart failure. Arch Intern Med. 2004;164:1978-1984. 2. Hobbs FD, Davis RC, Roalfe AK, Hare R, Davies MK, Kenkre JE. Reliability of N-terminal pro-brain natriuretic peptide assay in diagnosis of heart failure: cohort study in representative and high risk community populations. BMJ. 2002;324:1498-1500. 3. Egger M, Davey Smith G, Altman DG. Systematic Reviews in Health Care: Metaanalysis in Context. London, England: BMJ Books; 2001. WWW.ARCHINTERNMED.COM ©2005 American Medical Association. All rights reserved.