Objective: To analyze factors determining intraobserver and interobserver reproducibility of stereology in borderline and invasive ovarian tumors.
Study design: Fast and simple assessment of VPE was possible by using a highly automated interactive video overlay system suitable for application in a routine pathology laboratory. The point distance of a Weibel grid and the number of fields of vision per area of interest required to obtain good reproducibility were investigated. In addition, intraobserver and interobserver reproducibility was assessed, and the results of the improved technique were compared to those of the classical method.
Results: The experiments showed that intraobserver and interobserver variations in volume percentage of epithelium (VPE) assessments in a given case were caused mainly by high field-to-field variation and not so much by differences in the precision of assessment in a single field of vision. Therefore, many fields but only few points per field need to be measured to obtain, in a short time, a precise estimate of VPE in the measurement area of a tumor. From these results, an optimized protocol for VPE assessment was constructed. Using this protocol, nine observers independently assessed VPE in seven cases. Counting only one point in each of 100 systematically randomly sampled fields of vision (corresponding to a point distance of +/- 560 microns) yielded high intraobserver reproducibility (coefficient of variation [CV], 4%; R, .99; range, 0.98-1.00) and interobserver reproducibility (CV, 6%; R, .98; range, 0.97-1.00) in a short time. Assessment of one case took approximately three minutes, and the observers experienced the work as pleasant.
Conclusion: VPE assessed with systematic random sampling, using a grid with one point per field of vision and counting 100 hits, yields an inexpensive, fast and highly reproducible measure of an important prognostic variable in ovarian tumors. This assessment can be performed easily in a routine pathology laboratory.