Choice of methodology for quantifying cancer structures in tissue sections. A comparison of 2- and 3-dimensional estimators of mitotic activity, cellularity and nuclear size in breast cancer.

OBJECTIVE: To analyze the relationships between a range of 2-(2D) and 3-dimensional (3D) estimators of important tumor features (mitotic activity, cellularity and nuclear size). STUDY DESIGN: Measurements were performed in systematically sampled fields of vision of 3-microns-thick sections and in optical disectors of 40-microns-thick sections of 93 breast cancers. RESULTS: 2D mitotic profile density and frequency correlate highly with 3D mitotic density and frequency (r > or = .75); the choice of estimator was of minor importance. 2D nuclear profile density correlated quite closely with 3D nuclear volume fraction and nuclear density (r > or = .66). Cellularity estimators were, however, influenced differently by nuclear size. 2D mean nuclear profile area and 3D volume- and number-weighted mean nuclear size correlated less closely (r > or = .51). Tumors with high mitotic counts generally had large nuclei, reflecting the fact that both variables are associated with aggressiveness. CONCLUSION: 2D and 3D quantitative estimates of corresponding tumor features often correlate closely. Easily applicable and unbiased 3D techniques are recommended for obtaining the mean size of nuclei (or other particles) and of volume fractions of structures. For ranking of cellularity and mitotic activity, 2D measurements are usually sufficiently accurate and close to corresponding 3D estimates. However, if exact quantities of tissue structures are required, unbiased (3D) techniques must be used.