Robin T Vollmer1. 1. Laboratory Medicine 113, VA Medical Center, 508 Fulton St, Durham, NC 27705; Robin. Vollmer@med.va.gov.
Abstract
OBJECTIVES: Mitotic counts in melanoma are important and have now become part of the staging of this tumor. Yet, this change was largely based on studies that evaluated the mitotic counts in a limited fashion. Because counts of things with a microscope are often distributed as a Poisson random variable, the major goal of this study was to uncover the probabilistic nature of mitotic counts in melanoma. METHODS: Specifically, a general double Poisson model was applied to mitotic counts in 53 cutaneous melanomas representing both thin and thick tumors. RESULTS: The general double Poisson probability model fit the data well. A single Poisson function was sufficient for 46 of the 53 study cases, and two Poisson functions were required for seven cases because of tissue heterogeneity. Furthermore, the success of the model implied that there is a high probability for false-negative mitotic counts, especially in thin melanomas, and that the "hot" spot methodology introduces bias. CONCLUSIONS: Mitotic counts in melanomas are a probabilistic phenomenon closely related to the Poisson probability distribution, and this factor needs to be considered when using mitotic counts for staging and prognosis in melanoma.
OBJECTIVES: Mitotic counts in melanoma are important and have now become part of the staging of this tumor. Yet, this change was largely based on studies that evaluated the mitotic counts in a limited fashion. Because counts of things with a microscope are often distributed as a Poisson random variable, the major goal of this study was to uncover the probabilistic nature of mitotic counts in melanoma. METHODS: Specifically, a general double Poisson model was applied to mitotic counts in 53 cutaneous melanomas representing both thin and thick tumors. RESULTS: The general double Poisson probability model fit the data well. A single Poisson function was sufficient for 46 of the 53 study cases, and two Poisson functions were required for seven cases because of tissue heterogeneity. Furthermore, the success of the model implied that there is a high probability for false-negative mitotic counts, especially in thin melanomas, and that the "hot" spot methodology introduces bias. CONCLUSIONS: Mitotic counts in melanomas are a probabilistic phenomenon closely related to the Poisson probability distribution, and this factor needs to be considered when using mitotic counts for staging and prognosis in melanoma.