Interlaboratory reproducibility of semiautomated cell cycle analysis of flow cytometric DNA-histograms obtained from fresh material of 1,295 breast cancer cases

Conflicting prognostic results have been published as to the DNA variables, such as DNA ploidy, DNA index, and % S-phase cells for breast cancer patients. These variables can be obtained by interpreting DNA histograms by cell cycle analysis. Explanations for these conflicting results might be found on the level of the interpretation of the DNA histograms. In a previous study, the semi automated cell cycle analysis computer program MultiCycle (Phoenix Flow Systems, San Diego, CA) showed high intralaboratory reproducibility. However, what types of DNA histograms may cause disagreements was still unclear. The aim of this study was to determine the interlaboratory reproducibility of MultiCycle-based cell cycle analysis of 1,295 flow cytometric DNA histograms derived from fresh frozen breast cancer material and to clarify potential sources of interobserver variation when analyzing DNA histograms. DNA-ploidy classification into diploid, hyperdiploid, tetraploid, hypertetraploid, and multiploid showed an interlaboratory agreement of 94% (kappa value = 0.92). The 6% discrepancies (n = 74) were caused by tetraploid peaks, as established in one laboratory, which shifted outside the tetraploid region on reanalysis by the other laboratory (37%), shoulders sometimes interpreted as peaks (24%), small peaks not always recognized as such (24%), fitting failures (10%), and overlooking of tetraploid peaks (5%). Furthermore, the cell cycle analysis variables showed variable reproducibility. The % S-phase cells of the first, second, and third cell cycle showed overall a moderate reproducibility (0.62 less than or equal to R less than or equal to 0.79), but the average % S-phase cells and the average aneuploid % S-phase cells were more reproducible with correlation coefficients of 0.89 and 0.81, respectively. The coefficient of variation of the G0/G1 peak of the first cell cycle, the DNA indices and the % diploid cells were highly reproducible (R greater than or equal to 0.94), and the % G2/M-phase cells of the first, second, and third cell cycle were poorly reproducible (0.22 less than or equal to R less than or equal to 0.68). When a cut-point was used at the mean value of 7% for the average % S-phase cells, the number of ''threshold discrepancy cases'' was 6%. Sources of variation for cell cycle analysis were variations in the debris correction procedures, disagreement about the modes of the aneuploid peaks, disagreement about small peaks, shoulders sometimes interpreted as peaks, and overlooking of tetraploid peaks. Copyright (C) 1996 by W.B. Saunders Company