Audio
Melanoma continues to be a devastating disease unless diagnosed and treated early. According to the National Cancer Institute, there will be more than 76,000 new cases of invasive melanoma and nearly 10,000 melanoma-related deaths in 2014 in the United States.1 If diagnosed early, more than 93% of melanoma patients can expect to be cured, but later diagnosis of thicker melanoma is associated with a worse prognosis. Surgery remains the mainstay of therapy for cutaneous melanoma, including wide excision and sentinel lymph node (SLN) biopsy for staging of the regional nodal basins in appropriate patients. Although novel targeted therapies and immunotherapies have been associated with improved survival in metastatic melanoma, detection of cutaneous melanoma in its early phases remains the best chance for cure.
Tumor thickness, or Breslow depth, is the most important histologic determinant of prognosis in melanoma patients and is measured vertically in millimeters from the top of the granular layer (or base of superficial ulceration) to the deepest point of the tumor involvement. Increased tumor thickness confers a higher metastatic potential and poorer prognosis.2 Other histologic prognostic factors that have been incorporated into the American Joint Committee on Cancer melanoma staging system include the presence or absence of ulceration and mitotic index (measured per square millimeter), particularly for T1 melanomas (<1 mm thick), though Breslow depth greater than 0.75 mm appears to be the most reliable predictor of SLN metastasis in thin (T1) melanomas (≤1 mm).3
Tumor volume assessment may be a helpful adjunct to Breslow depth as a prognostic indicator for melanoma, particularly for predicting SLN metastasis.4 This retrospective study was designed to assess the improvement in the accuracy of Breslow depth as a prognostic factor by utilizing tumor volume combined with mitotic index, presence or absence of ulceration, and inflammatory host reaction (eg, tumor-infiltrating lymphocytes).
Methods
The study was approved by the Stanford University (Stanford, California) institutional review board. A retrospective review of invasive primary melanomas recorded in Stanford University’s pathology/dermatopathology database from January 2007 through December 2010 was conducted. Because cases included both Stanford Health Care (formerly Stanford Hospital & Clinics) and outside pathology consultations, clinical assessment of patient outcome was not possible for all cases and thus was not performed.
Assessment
Information extracted from the pathology reports included Breslow depth; estimated surface area of the primary tumor (measured by the longest vertical and horizontal dimensions recorded by the clinician prior to diagnostic biopsy and reported on the biopsy requisition form [>90% of cases] or reported by the pathologist on gross measurement of the pigmented lesion in formalin [<10% of cases]); mitotic index (measured per square millimeter); presence or absence of ulceration; and inflammatory host reaction (as noted by tumor-infiltrating response). Our method of estimating the tumor volume (lesion surface area • Breslow depth) did not take into account border irregularities in the primary tumor. This method also was limited because prebiopsy clinical measurement could differ from gross pathologic measurement of the tumor due to shrinkage of the latter ex vivo and following formalin fixation. However, when both measurements were documented, the pathological measurement was only slightly less than the clinical measurement. Metastases were defined as those in lymph nodes (microscopic or macroscopic), skin, or in distant organs, as identified through review of subsequent pathology reports.
Statistical Analysis
Statistical analyses were conducted using SAS version 9.3. Test statistics were preset at a significance level of α=.05. Using metastasis status as the outcome, univariate regression models were first fitted to assess the predictive ability of each prognostic indicator. In univariate analyses, continuous prognostic indicators (Breslow depth, tumor volume, and surface area) were included in the model while seeking the best functional form by means of fractional polynomials modeling.5,6 Predictive ability of prognostic indicators was determined by the area under the receiver operating characteristic curve (AUC).7 Using best functional form for Breslow depth, all other prognostic indicators were added to the model to assess their individual contributions to improve the predictive ability for tumor metastasis. The functional forms used for tumor volume and surface area were those determined in the univariate analysis. Multivariable models were compared aiming for an improvement of the best Breslow model indices: Schwarz criterion, Hosmer-Lemeshow goodness-of-fit test, generalized R2, and AUC.5 The added contribution of clinical predictors to the model for Breslow depth was judged by the significance of the coefficient for the added clinical predictor, the significance of the change in AUC, and the change in the model indices listed above. A check on overdispersion was carried out on the final model selected.
