Interexaminer agreement of histopathologic features in pediatric brain tumors: a biostatistical perspective.

Pages 108-117 of the current issue of Pediatric and Developmental Pathology contain a landmark interobserver investigation of some of the vagaries and vicissitudes involved in the nosological assessment of 6 histological features of childhood brain tumors. Doctor Gilles and his clinical research colleagues are to be commended for their conceptualization, reliability research design, execution, and interpretation of their findings. The piece is also uncommonly sophisticated concerning matters biostatistical. As such, it stands as a model for future nosological studies in childhood brain pathology, and can be applied more generally as well. The purpose of this Context article is to present some additional conceptual and methodologic issues that are relevant to the future design of histopathological studies in particular and of biobehavioral and biomedical diagnostic studies in general. In order to accomplish this goal, I shall discuss interexaminer reliability issues from the following viewpoints: type of reliability assessment, scales of measurement, reliability statistics, diagnostic accuracy, and the need to always differentiate levels of statistical significance from those that are both statistically and clinically meaningful, as Gilles and colleagues have also successfully accomplished.