Robert G. Walton, MD; Jinah Kim, MD, PhD; Cruz Velasco, PhD; Susan M. Swetter, MD
Drs. Walton, Kim, and Swetter are from the Department of Dermatology, Stanford University Medical Center, California. Dr. Kim also is from the Department of Pathology. Dr. Swetter also is from the Stanford Cancer Institute as well as the Dermatology Service, VA Palo Alto Health Care System, California. Dr. Velasco is from the Biostatics Program, School of Public Health, and the Stanley S. Scott Cancer Center, Louisiana State University Health Sciences Center, New Orleans.
The authors report no conflict of interest. Dr. Walton’s research was supported in part by a grant from the Ronald and Ann Williams Charitable Foundation.
Correspondence: Susan M. Swetter, MD, Department of Dermatology/Cutaneous Oncology, Stanford University Medical Center and Cancer Institute, 900 Blake Wilbur Dr, W3045, Stanford, CA 94305-5356 (sswetter@stanford.edu).
Univariate analysis demonstrated that Breslow depth was the best prognostic indicator of metastasis (AUC=0.946) but that tumor volume (as a continuous variable) was nearly equally predictive (AUC=0.940)(Table 3). Tumor volume alone (categorized as <250 mm3 vs >250 mm3) had lower prognostic value (AUC=0.855). Mitotic index, presence or absence of ulceration, inflammatory host reaction, and surface area also had lower prognostic values, though all were significant factors (P values ranging from <.0001 to .0077)(Table 3).
Importantly, the addition of surface area, mitotic index, presence or absence of ulceration, and inflammatory host reaction to the model to Breslow depth did not improve predictive ability for metastasis, and AUC values did not increase significantly after adding these factors (Table 4). In particular, the change in AUC for adding surface area to the model with Breslow depth was 0.023 (P=.1095). Models in Table 4 were checked for interaction of these 2 predictors, and the interaction term for thickness and surface area was not statistically significant (P=.0932)(data not shown).