Statistical analysis of cytogenetic abnormalities in human cancer cells.

It is sometimes difficult to evaluate reports of "nonrandom" chromosome involvement in certain malignant diseases, since the "random" or expected distribution is seldom defined. Therefore, we have developed methods for the statistical analysis of cytogenetic abnormalities in human cancer cells with modal karyotypes in the diploid range (35-57 chromosomes). For this analysis, it is assumed that the expected gain or loss of each chromosome will occur with equal probability and structural abnormalities will involve each chromosome in proportion to its size. To perform this analysis, the total number of numerical and structural abnormalities is determined from the modal karyotypes of a series of histologically related tumors. The maximum expected values are determined by computer simulation for different levels of significance. Then the distributions of observed and expected abnormalities of each type are compared to identify nonrandom involvement. Preferential gain or loss is analyzed for each of the 24 different chromosomes, and preferential structural rearrangement is determined for each of the 48 chromosome arms. We have analyzed two series of karyotypic data to demonstrate the utility of this method. The rationale for the assumptions made as well as alternative approaches are discussed.