Multifactor analysis of intermediate cells from the uterine cervix. The importance of slide effects on variance components

The ability to retrospectively examine cytologic material with digital image analysis is often desirable, particularly when long-term follow-up information is available for correlation with cell parameters. In an effort to characterize the magnitude of slide-age and staining-batch effects on cell parameters from digital image analysis, the cell features in 32 Papanicolaou-stained cervical smears that varied in age from 10 to 16 years were analyzed. The slides were selected from eight patients sampled on four different occasions approximately one year apart. The findings indicate that the variance of cell features from an intermediate cell population within a slide is somewhat greater than is the variance among slides from the same woman or the variance among women. Aging effects were not detectable. The significant differences observed among replications of slides from the same woman as well as between women with no evidence of cervical disease should caution other researchers to account for this potential source of random error in their statistical models.