HISTOLOGIC FEATURE RELIABILITY IN CHILDHOOD NEURAL TUMORS

We studied intraobserver reproducibility in recognizing the presence or absence of 57 histologic features or patterns in a random subset of tumors (822) from the Childhood Brain Tumor Consortium database. The study protocol maximized consistency of the observer. We found that only six histologic features had high (greater than or equal to 0.75) reliability estimates while a large number had intermediate estimates of 0.50-0.74. Supratentorial or infratentorial tumor location sometimes altered reliability. Reliability estimates were unacceptable for certain histologic features often used as diagnostic criteria, descriptors of tumor characteristics, or markers of anaplasia. We hypothesize that low reliability reflects, in part, the need for more specific operational definitions, particularly those with subjective boundaries (e.g. the cellular density constituting the histologic boundary between high and intermediate cell density). Rarity of a feature in the microscopic section (e.g. granular bodies) may also contribute to low reliability. We also show that the kappa statistic, a commonly used measure of reliability, is inappropriate for very common or uncommon histologic features (e.g. features at the extremes of prevalence in the study cases) and we offer a simple empiric method for determining when an alternative measure, the Jaccard statistic, is appropriate.