Reproducibility of nucleolar measurements in human intraocular melanoma cells on standard histologic microslides

The standard deviation of nucleolar area (SDNA), as determined by semiautomated measurement of routinely prepared histologic sections, has been shown to be an effective predictor of mortality for human intraocular melanoma. One hundred cases of this tumor underwent two independent determinations of SDNA by each of two technicians using the same device. One technician performed the two measurements within two months, the other within three weeks. Analysis of these 400 values revealed the following: the more experienced of the two technicians achieved an acceptable level of reproducibility in determining SDNA; the less-experienced technician achieved a lower level of reproducibility; a systematic bias was found upon comparing SDNA values of the two technicians; and this bias can be partially corrected by applying a linear transform. These findings support the value of SDNA as a practical clinical measure of malignant potential for intraocular melanoma but also suggest that experience can enhance the reproducibility of a technician's measurements.