1. ¹Department of Pathology, Stanford University School of Medicine, Stanford, CA, USA
2. ²Leukemia Research Fund Immunodiagnostics Unit, Nuffield Department of Clinical Laboratory Sciences, John Radcliffe Hospital, University of Oxford, Oxford, UK

Correspondence: Dr Y Natkunam, MD, PhD, Department of Pathology, L235, Stanford University School of Medicine, 300 Pasteur Drive, Stanford, CA 94305, USA. E-mail: yaso@stanford.edu

Received 9 May 2006; Accepted 11 May 2006; Published online 19 June 2006.

Abstract

Technological advances in gene cloning and genome-wide analyses have greatly increased the number of new tumor markers that can be detected by immunohistologic techniques. While many of these have been evaluated with respect to prognosis, there is a striking discrepancy between the number of markers reported to confer prognostic information and those that are used in clinical practice. We argue that lessons learned from epidemiological studies are applicable to studies of immunohistologic markers; in particular, advances in both fields can be vitiated by non-causal associations. We suggest that the most valuable immunohistologic markers are those that reflect genetic abnormalities, that are linked to the cell of origin, or that reflect tumor infiltrating cells or stromal reactions. It should also be appreciated that a marker that is genuinely predictive of prognosis may nevertheless not find any application in clinical practice if it becomes obsolete through the introduction of newer therapies or because there is no choice of alternative treatment strategies.