[1964] Mitotic Rate Cut-Point Reproducibility in Gastrointestinal Stromal Tumor (GIST) Grading

Michael Bonert, Shaikh A Mortuza, Stephen S Raab. Memorial University of Newfoundland/Eastern Health, St. John's, Newfoundland and Labrador, Canada

Background: Mitotic rate is an important histomorphologic predictor of tumor behavior that is routinely estimated and recorded in surgical pathology reports. Increasingly, mitotic counting procedures are being standardized, as it has been recognized as a source of variability. In gastrointestinal stromal tumors (GISTs), the sample area is standardized to 5 mm2 and 1 mitosis/mm2 has been chosen as a cut-point for grading. The reproducibility of this clinically important cut-point is not known. In breast pathology, the binomial distribution has been used to examine the mitotic rate cut-points.
Design: Mitotic counting was modeled as a random sampling problem using the binomial distribution. Using this model, the sample standard deviation and 95 percent confidence interval (CI) were calculated to the nearest mitosis, and the mis-classification rates around the cut-point were calculated using the software GNU Octave (http://www.gnu.org/software/octave/).
Results: The standard deviation is 2-7 mitoses/5 mm2 (0.4-1.4 mitoses/mm2), for the standardized mitotic cut-point (5 mitoses) and sampling area (5 mm2). Doubling the sampling area to 10 mm2 reduces the standard deviation to 6-13 mitoses/10 mm2 (0.6-1.3 mitoses/mm2). If the sample area is 25 mm2, the standard deviation is 0.76-1.20 mitoses/mm2. The 95 percent confidence interval is 0-10 mitoses for 5 mm2 (0-2 mitoses/mm2), 3-17 mitoses for 10 mm2 (0.3-1.7/mm2) and 15-35 mitoses for 25 mm2 (0.60-1.40 mitoses/mm2). The confidence interval decreases less with larger sample areas. With the standardized area (5 mm2), if the true mitotic rate is 2, 3, 4, or 5 mitoses/5 mm2, then 1.7%, 8.4%, 11.1%, and 38.4% of the cases would be over-graded, respectively, by chance alone. On the other hand, if the true mitotic rate is 6, 7, 8 or 9 mitoses/5 mm2 for a 5 mm2 area, then 44.6%, 30.1%, 19.1% and 11.6% of cases would be under-graded, respectively, by chance alone.
Conclusions: The mitotic grade in general and for GISTs in particular is not very reproducible if the mitotic count falls within a standard deviation of the cut-point. We suggest that quantification of measurement variability, such as the standard deviation, be included in the mitotic rate estimation, to alert users of this information about its level of uncertainty. A triage approach with a low initial sample area could be used to identify cases in close proximity to the mitotic rate cut-point; in such cases, a larger area could be used to increase mitotic grade accuracy.
Category: Quality Assurance

Tuesday, March 5, 2013 9:30 AM

Poster Session III # 299, Tuesday Morning


Close Window