Tumor Thickness and Mitotic Rate Robustly Predict Melanoma-Specific Survival in Patients with Primary Vulvar Melanoma: A Retrospective Review of 100 Cases

First described by Cullen and Carswell in 1824 (1), and subsequently by Hewitt in 1861 (2), primary vulvar melanoma (PVM) is the second most common vulvar malignancy (3) with an annual incidence of 1.08 to 1.36 cases per million women per year in the United States (4, 5). PVM accounts for 4% to 10% of all gynecologic malignancies and fewer than 10% of all melanomas in women (6–8). Elderly Caucasian women are the most commonly affected population, and unilateral labial involvement is the most frequent anatomic distribution of disease, although extension across the midline and/or clitoral involvement occurs in approximately 30% of PVM (9). PVM is characterized by a high rate of local recurrence and metastasis, necessitating aggressive surgical resection (10) that often results in high morbidity and a high disease-specific mortality rate (reported to range from 71% (3) to 90% (11)), underscoring the need for effective prognostication and risk stratification at primary tumor diagnosis.

Melanomas arising in mucosal sites (including sinonasal, rectal, and vulvar) have traditionally presented unique clinical and pathologic challenges. Specifically, their derivation in anatomic locations with nonkeratinizing squamous epithelium and unique microanatomic boundaries together with their distinctive molecular–genetic changes (12) call into question whether staging criteria derived for cutaneous melanomas accurately reflect the clinical biology of mucosal melanomas. For example, staging of sinonasal melanomas remains controversial and has undergone significant modifications over the past decade (13–16). Similarly, staging of PVM has transitioned from the International Federation of Gynecology and Obstetrics (FIGO) system (where T-category is based on greatest tumor diameter) to the American Joint Committee on Cancer (AJCC) system (where T-category is based primarily on tumor thickness and secondarily on ulceration and/or mitotic rate; refs. 3, 4, 11, 17–22) as is applied to primary cutaneous melanoma (PCM; ref. 23). Further confounding the dilemma of staging and prognostication, PVMs have been historically characterized together with other melanomas arising on the female genitalia (notably, vaginal melanomas), further obscuring the prognostic features that might be specific to the vulvar origin (5, 24–28). Furthermore, most studies characterizing clinical and/or histopathologic parameters were often insufficiently powered to reveal relevant prognostic indicators, owing in large part to the relative rarity of PVM (29). Nevertheless, many different clinical and histopathologic variables have been proffered to correlate with a higher risk of disease recurrence and survival in PVM, including higher age at diagnosis (4, 20–22, 30), African American ethnicity (4, 5, 27, 31), higher number of positive regional lymph nodes (4, 22, 24, 29), central and/or multifocal vulvar involvement (21, 22, 26, 30,32, 33), increasing tumor thickness (3, 9, 20, 21, 34–37), ulceration (8, 20, 30, 34, 37), and dermal mitotic rate (8, 20, 24). Identification of the feature(s) among these that independently inform prognosis would greatly improve risk stratification and enable more personalized clinical management for patients with PVM, in particular, earlier and more aggressive intervention for those patients with highest risk disease. There is therefore a critical clinical need to define those parameters that most accurately predict the biological behavior of PVM. To address this question, we retrospectively reviewed 100 cases of PVM treated at our institution to determine the associations between clinical and conventional histopathologic parameters of the primary tumor and measurements of patient outcome. This represents the largest single institution study of primary vulvar melanoma to date. We show that increasing tumor thickness and dermal mitotic rate are independent predictors of reduced DSS and propose a novel T-category for PVM that incorporates only these two variables to stratify risk among these patients.

The study was approved by the Institutional Review Board of The University of Texas MD Anderson Cancer Center (Houston, TX). The pathology archives were searched to identify patients with PVM seen at our institution during the period from October 1951 to October 2011 to ensure a minimum of at least 4 years of clinical follow-up. Patients were included only if hematoxylin–eosin (H&E)-stained slides and carefully annotated demographic, management, and clinical follow-up data were available. It is unclear how many of the patients in the current study were also included in prior reports from our institution (10, 38).

For each patient, the following data were collected: age at diagnosis; ethnicity; presenting symptom(s) and their duration; site of involvement; date of first diagnosis; date of diagnosis and site(s) of regional and distant metastases; and date and cause of death, when applicable. H&E-stained slides of the primary vulvar melanoma were re-evaluated by at least one (and up to six) dermatopathologists (P. Nagarajan, M.T. Tetzlaff, V.G. Prieto, J.L. Curry, D. Ivan, and C.A. Torres-Cabala). Discrepancies were adjudicated by majority opinion in a consensus conference setting. The following histopathologic parameters were recorded: histologic type, tumor thickness (measured from the top of the granular layer or the base of the ulcer to the deepest point of invasion), epithelial ulceration (recorded as present or absent), dermal mitotic rate using the hotspot approach (highest number of dermal mitotic figures/mm²; ref. 39), vertical growth phase (present or absent), regression (present or absent), lymphovascular invasion (LVI, present or absent), perineural invasion (PNI, present or absent and nerve size when present), microscopic satellitosis (present or absent), tumor-infiltrating lymphocytes (absent, non-brisk, or brisk), predominant cytology (epithelioid, nevoid, or spindled), and precursor melanocytic nevus (present or absent). Regional lymph node metastasis (number of involved nodes) and distant metastasis (location) were also documented and reviewed, where applicable.

The clinical and histopathologic characteristics were summarized using descriptive statistics. Overall survival (OS) was defined as the time interval from the date of diagnosis to the date of death due to any cause. Disease-specific survival (DSS) was defined as the time interval from the date of diagnosis to the date of death directly related to the progression of melanoma. Patients that were free of PVM at the time of death were censored when analyzing for DSS. We based this conclusion on the fact that (i) none of these patients ever developed recurrence or distant metastasis (disease-free interval range: 4.5–38 years) and (ii) all these patients had been disease free for at least 4 years prior to their death. The Kaplan–Meier method estimated OS and DSS curves (40). Univariate Cox proportional hazards regression models determined the association between histopathologic and clinical characteristics and OS and DSS (41). A multivariate Cox proportional hazards model was then fitted for OS and DSS by including all statistically significant covariates from univariate Cox models; a stepwise selection was performed using a threshold of 0.05 for the significance level of the Wald χ² for covariates to stay in the final model. In this study, we employed the method by Grambsch and Therneau to evaluate the proportionality assumptions of the Cox model models (42). Fisher exact tests determined associations between clinical and/or histopathologic parameters and the propensity to develop nodal or distant metastases.

The clinical and histopathologic characteristics of the cohort are summarized in Table 1. One hundred women with PVM were identified, including 85 Caucasian, 10 Hispanic, 3 Asian, and 2 African American. The median age at diagnosis was 56 years (range: 18–87 years). The presenting symptoms included lump/mass (79% of patients), bleeding (14%), pruritus (9%), pain or discomfort (9%), irritation or discharge (4%), and ulcer (4%). The median follow-up time was 67.8 months. Twenty patients were alive at their last encounter, 4 were lost to follow-up, and 76 died. Among the deaths, 67 were attributed to melanoma, while 9 were due to other causes. Among the latter 9 patients, we did not attribute their death to melanoma because none ever developed local or distant recurrence (disease-free interval range: 4.5–38 years), and all had been free of disease for at least 4 years prior to their death. The median OS was 73.9 months, the 5-year OS was 57.2% (95% CI: 48.2%–67.8%), and the 10-year OS was 34.0% (95% CI: 25.6%–45.2%). The median DSS was 74.2 months, the 5-year DSS was 58.2% (95% CI: 49.2%–68.8%), and the 10-year DSS was 35.8% (95% CI: 27.3%–47.1%).

We first analyzed survival outcomes according to primary tumor thickness using cutoffs established by the 7th edition AJCC staging system for PCM (pT1: ≤1.00 mm; pT2: 1.01–2.00 mm; pT3: 2.01–4.00 mm; pT4: >4.00 mm; Table 1; Fig. 1A; ref. 19). Kaplan–Meier survival curves demonstrated a progressive reduction in OS (data not shown) and DSS with increasing primary tumor thickness (Fig. 1B). Univariate analysis revealed that tumor thickness >4.00 mm significantly associated with lower OS (HR = 3.97; P < 0.01) and DSS (HR = 5.77; P < 0.01), compared with melanomas whose tumor thickness was ≤1.00 mm. The differences in DSS were marginally significant for melanomas with tumor thickness 2.01–4.00 mm compared with those ≤1.00 mm (HR = 2.26; P = 0.06), but not for tumor thickness 1.01–2.00 mm versus ≤1.00 mm (Table 2).

We next stratified patients according to mitotic rate (0, 1, 2–10, and >10 mitotic figures/mm²; Table 1; Fig. 1C). Kaplan–Meier analysis demonstrated similar OS and DSS for patients whose melanomas had either 0 or 1 mitotic figure/mm², but markedly worse outcomes were observed for patients whose melanomas had either 2–10 or >10 mitotic figures/mm² (2–10/mm²: HR = 2.80; P < 0.01 and >10/mm²: HR = 2.81; P < 0.01) and DSS (2–10/mm²: HR = 3.56; P < 0.01 and >10/mm²: HR = 3.78; P < 0.01) on univariate analysis (Table 2; Fig. 1D). When the latter two categories were merged, mitotic rate ≥2/mm² significantly associated with reduced OS (HR = 3.36; P < 0.001) and DSS (HR = 4.44; P < 0.001) in comparison with those patients whose PVM had <2/mm². Kaplan–Meier estimates also clearly delineated OS and DSS in PVM patients with mitotic rate <2/mm² and ≥2/mm² (Supplementary Fig. S1).

Six additional histopathologic parameters significantly associated with patient survival outcomes on univariate Cox proportional hazards regression analysis (Table 2): (i) primary tumor histologic ulceration (present in 50% of the cases; Fig. 2A) associated with decreased OS (HR = 1.94; P = 0.01) and DSS (HR = 2.27; P < 0.01; Fig. 2B); (ii) microscopic satellitosis (identified in 9 cases; Fig. 2C) associated with decreased OS (HR = 4.20; P < 0.01) and DSS (HR = 4.38; P < 0.01; Fig. 2D); (iii) LVI (present in 23 cases; Fig. 2E) associated with decreased OS (HR = 2.76; P < 0.01) and DSS (HR = 2.83; P < 0.01; Fig. 2F); (iv) PNI (present in 15 cases; Fig. 2G) associated with significantly lower OS (HR = 2.42; P < 0.01) and DSS (HR = 2.49; P < 0.01; Fig. 2H); (v) the presence of regression (present in 25 cases; Fig. 2I) associated with longer DSS (HR = 0.54; P = 0.05; Fig. 2J) but not OS; and (vi) a precursor nevus (present in 8 cases; Fig. 2K), associated with improved OS (HR = 0.16; P = 0.01; Fig. 2L; insufficient statistical power to meaningfully evaluate association with DSS). Other histopathologic parameters considered (vertical growth phase, histologic growth pattern, predominant cell type, and tumor-infiltrating lymphocytes) did not significantly associate with patient survival (Table 2). In addition, in univariate analyses, many of these variables also significantly associated with the propensity to develop distant metastases, including: tumor thickness (P < 0.01); mitotic rate (P < 0.01); ulceration (P < 0.01); LVI (P < 0.01); PNI (P = 0.02); microscopic satellitosis (P = 0.03), and a precursor nevus (P < 0.01; Fisher exact test; Supplementary Table S1).

Two clinical parameters also associated with survival outcomes: (i) age > 56 years (the median in our cohort) at diagnosis was associated with decreased OS (HR = 1.98; P < 0.01) and DSS (HR = 1.88; P = 0.01; Table 2, Fig. 2M and N); and (ii) unilateral labial involvement was associated with longer DSS compared with tumors involving the clitoris (Fig. 2O) or with bilateral spread (HR = 0.56; P = 0.03; Fig. 2P).

To determine which parameters were independently predictive of patient survival in PVM, multivariate Cox proportional hazard models for OS and DSS were constructed including all significant covariates from the univariate analyses (P < 0.05). After stepwise selection for significance, the following parameters independently associated with OS (Table 3): tumor thickness (P = 0.03), dermal mitotic rate (P = 0.03), LVI (P = 0.04), microscopic satellitosis (P < 0.01), and associated precursor nevus (P = 0.04). Only two parameters independently associated with DSS in PVM: tumor thickness (P < 0.02) and dermal mitotic rate (P < 0.01) (Table 3).

To determine whether the T-category criteria defined by the 7th edition AJCC staging system for PCM also applied to PVM, we grouped our PVM patient cohort according to the AJCC criteria (Fig. 3A). A log-rank test demonstrated that the DSS curves were significantly different among different groups by the AJCC T-category (P < 0.001). However, while the AJCC T-category predicted patient outcome among thicker melanomas (tumor thickness >2.00 mm; pT3 and pT4), melanomas whose thickness was ≤2.00 mm (pT1a/b and pT2a/b) were less robustly distinguished from one another according to the AJCC T-categories (Fig. 3B).

In contrast, as only tumor thickness and mitotic rate independently predicted DSS in our PVM cohort, we hypothesized that a refined T-classification strategy based on these features might better predict outcomes in PVM than the AJCC T-category criteria for PCM (Fig. 3C). We defined the proposed T-category as follows: (pT1*) tumor thickness ≤ 2.00 mm and dermal mitotic rate <2/mm² versus (pT2*) tumor thickness > 2.00 mm and/or dermal mitotic rate ≥2/mm²; more detailed information regarding the selection of 2.00 mm as a cut-off is provided in Supplementary Fig. S2). Classification of PVM based on this novel T-category delineated 2 groups of patients and robustly predicted OS and DSS (OS: P = <0.001; DSS: P < 0.001; Fig. 3D).

There is a critical unmet clinical need to optimize risk stratification and enable personalized management for patients with PVM as conventional staging systems lack precision for this rare melanoma subtype. To address this need, we retrospectively analyzed 100 patients with PVM to determine which clinical and melanoma-associated histopathologic parameters correlated with patient outcome. Our study represents the largest and most comprehensive single institution study of PVM to date. We found that only tumor thickness and dermal mitotic rate (in particular, ≥2 mitotic figures/mm²) independently predicted DSS. Furthermore, we found that a novel pathologic T-category for PVM based only on these two factors robustly predicted survival.

The relationship between tumor thickness and outcomes has been historically controversial in PVM. Whereas Podratz and colleagues reported that tumor thickness correlated with prognosis in PVM better than FIGO stage (where T-category is based on greatest tumor diameter; ref. 22), other studies failed to show any association between tumor thickness and indices of patient outcome (3, 43, 21), even among thick melanomas (20). Still, others show a correlation between tumor thickness and lymph node metastasis, but not patient survival (36). In a series of 77 PVMs, Moxley and colleagues reported that increasing tumor thickness correlated with melanoma recurrence, but not OS (11). Other studies have shown that PVMs >1.75 mm thick have an increased risk for recurrence (44) and those <3.00 mm thick associated with improved OS (34, 35, 37). In an epidemiologic meta-analysis of 1,442 PVMs, Ragnarsson-Olding and colleagues demonstrated higher tumor thickness to be an adverse prognostic factor by multivariate analysis (8). Our findings significantly expand on these prior studies. First, in 100 PVMs, we confirmed increasing tumor thickness to be an independent predictor of reduced OS and DSS. Second, our data support the application of a novel T-category for PVMs in which tumor thickness functions as a bimodal predictor of survival: PVMs ≤2.00 mm exhibit improved OS and DSS compared with PVMs >2.00 mm, with dermal mitotic rate serving as a powerful modifying variable.

The prognostic significance of dermal mitotic rate has been a similarly controversial variable in PVM. This contrasts with PCM, where a number of large studies (including one with over 13,000 patients) have confirmed dermal mitotic rate to be independently predictive of melanoma-specific survival (45–48). In one of the largest studies of PCM to date, Thompson and colleagues showed that mitotic rate increased with increasing tumor thickness, and melanoma-specific survival decreased with increasing mitotic rate (48). In their study, mitotic rate emerged as the strongest indicator of melanoma-specific survival after tumor thickness, and as a consequence, dermal mitotic rate emerged as a definite staging criterion informing the prognosis of PCM ≤1.00 mm thick in the seventh edition of the AJCC staging system (19). In PVM, Bradgate and colleagues showed that PVMs with <5 mitotic figures/10 high-power fields (HPF) associated with a better prognosis than PVMs with >5 mitotic figures/10 HPF (20). In a study of 219 patients with PVM, Ragnarsson-Olding and colleagues reported that a higher mitotic rate among PVMs correlated with aggressive behavior (8, 9). Furthermore, mitotic rate (calculated by the hotspot method) independently correlated with 5-year disease-free survival (24). Nevertheless, not all studies have confirmed a correlation between dermal mitotic rate and prognosis (3). Here, we show that higher dermal mitotic rate significantly associated with reduced OS and DSS in both univariate and multivariate analyses. Of particular importance, we found no significant difference in prognosis between PVMs with 0 or 1 mitotic figure/mm², but any dermal mitotic rate ≥2/mm² independently associated with reduced OS and DSS in univariate and multivariate analyses.

The above observations led us to consider an alternative T-category for PVM. Historically, melanomas arising on mucosal sites have been viewed as inherently different from their cutaneous counterparts: (i) PVMs arise in close proximity to mucocutaneous transition zone; (ii) PVMs have distinctive external risk factors (in particular, no relationship with exposure to ultraviolet light); and (iii) PVMs arise through different molecular-genetic alterations (12, 23, 43, 49). Thus, an emerging paradigm is that a monolithic staging system derived for PCM may not apply to noncutaneous melanomas, including PVMs (14, 50). Emblematic of this controversy, studies exploring histopathologic predictors of outcome in primary sinonasal melanoma have produced distinct results from PCM. In particular, when sinonasal melanoma is restricted to mucosal and submucosal tissue (pT3), the depth of invasion or even the size of the melanoma does not accurately predict patient outcome (51). Instead, invasion into the surrounding tissues, including bone or cartilage (T4), associates with worse patient outcome. In some respects, this can be seen as a variation of Clark level as being most predictive in this subtype of mucosal melanomas.

As PVMs are often considered among gynecologic malignancies, the FIGO staging system was first applied to their prognostication. FIGO relies largely on tumor size and therefore does not reflect biological potential as a function of the depth of tumor invasion from an intraepithelial site of origin. Therefore, in contrast to other malignancies of the female genitourinary tract, staging of PVM transitioned to the AJCC system derived for PCM to capture this important aspect of melanoma biology. The AJCC T-category relies primarily on tumor thickness and secondarily on ulceration and mitotic rate (3, 4, 17–22). However, we show here that the AJCC T-category did not effectively stratify survival risk among all patients with PVM. Although the clinical outcome of patients with melanomas >2.00 mm in thickness was predicted well according to the AJCC T-category, the survival curves of patients with melanomas ≤2.00 mm thick showed considerable overlap. Therefore, in light of our observation of a bivariate clinical outcome according to tumor thickness (≤2.00 mm or >2.00 mm) and mitotic rate (<2 or ≥2/mm²), we devised a refined T-category schema based only on tumor thickness and dermal mitotic rate. This novel T-category robustly differentiated the outcome of patients with melanomas whose thickness was ≤2.00 mm depending on whether the mitotic rate was <2/mm² or ≥2/mm². Moreover, this novel T-category performed similarly to the AJCC T-category in predicting the outcome of melanomas >2.00 mm thick. Additional studies are necessary to validate our findings.

Other clinical and histopathologic variables correlated with survival indices in our study, including age, extent of disease (clitoral, midline, or bilateral involvement), ulceration, LVI, PNI, satellitosis, regression, and the presence of a precursor nevus. The peak incidence of PVM occurs in the sixth and seventh decades (4, 5, 20, 27,43, 52), and increasing age in PVM has been shown to associate with worse survival, even after adjustment for ethnicity and clinical stage, with a HR of 1.4 per decade (5). One important caveat is that increasing age also associates with thicker melanomas, higher dermal mitotic rate, and ulceration in PVMs (53). In our study, patients older than 56 years at diagnosis had decreased OS and DSS compared with those younger than 56 years.

The most common anatomic distributions of PVM are (i) involvement of a single labium majus or minus, (ii) unilateral disease involving both labia, and (iii) disease extension to midline structures such as clitoris or rarely, the introitus (8, 30). Multifocal involvement has also been described (26, 30, 33, 54). Clitoral involvement by PVM and multifocal disease have been associated with increased risk for recurrence and survival (21, 22, 30, 32), although this also remains controversial (5). In our series, 14 patients had clitoral involvement, and 10 had PVM involving bilateral labia. When grouped together, these 24 patients had significantly decreased DSS and greater risk for distant metastasis compared with patients with unilateral labial disease, solidifying an association between anatomic distribution of PVM and clinical outcome.

In our series, many histopathologic variables associated with survival indices in univariate but not multivariate analyses. Ulceration, microscopic satellitosis, PNI, and LVI each associated with reduced OS and DSS. In PCM, (i) ulceration is a well-established independent predictor of outcome (55); (ii) microscopic satellitosis associates with increased risk for regional nodal metastases (56, 57–61); (iii) PNI associates with increased local recurrence necessitating multiple resections and/or adjuvant radiotherapy; and (iv) LVI associates with decreased DSS (62–64). In PVM, ulceration also correlates with prognosis irrespective of tumor thickness (20, 30, 34, 37), although not all studies confirm this relationship (36, 65, 24). Satellitosis [incidence from 9% (current study) to 22% (9)] and PNI [incidence from <10% (9, 29) to 15% (current study)] are overall relatively rare in PVM and remain incompletely characterized as prognostic indicators. In PVM, LVI independently associates with nodal metastases (21) and adverse 5-year survival (30, 66), although this too remains controversial (24).

The prognostic significance of regression in PCM remains incompletely understood. Depending on the study, some show an association with poor prognosis (67–69); others demonstrate no relationship with prognosis (70–72); and still others demonstrate an association with favorable outcome (73–75). Histologic regression has only rarely been described in PVM (9, 29), but no association with prognosis has been described. We observed frequent histologic regression in PVM (∼25%), and this associated with improved DSS (but not OS). Precursor melanocytic nevi have been identified in association with a varying proportion of PCMs (76–79), and these tend to be thinner and are associated with a slightly better prognosis than de novo melanomas (76, 80). In contrast, PVMs only rarely arise in association with a precursor melanocytic nevus (refs. 9, 29; ∼8% in our series), and we show an association with improved OS (but not DSS). Taken together, the overall prognostic significance of these histopathologic variables in PVM remains controversial. Additional larger scale studies are required to determine their prognostic significance in PVM and will be greatly enhanced by systematic documentation of these variables.

There are important limitations to our study. First, we have not validated our proposed T-category in an independent patient cohort. Furthermore, we did not integrate nodal or distant metastatic disease status into our assessment of the relative importance of different histopathologic features of the primary lesion in our study. Although we attempted the latter, our study was insufficiently powered to distinguish prognostic differences when we further stratified patients according to relevant histopathologic parameters and either nodal or distant metastatic disease status. Additional studies are necessary using independent patient cohorts to confirm our observations and to integrate metastatic disease status into the analyses. Next, as a tertiary referral center, we recognize that focusing this report on our institutional experience with PVM might possibly present a skewed clinical experience towards more aggressive tumors of a higher stage. However, we reviewed the existing literature on PVM (included a total of 27 studies; see Supplementary Table S2) and found that among 816 patients with PVM described in the literature for whom either (i) an annotated tumor thickness was provided for each patient or (ii) the number of patients assigned to each AJCC T-category was provided, 43.5% (355/816) of these patients had a tumor thickness >4.0 mm (3, 9–12, 20, 21, 23, 25, 28, 29, 31, 32, 34, 35, 37, 44, 65, 81–89). In addition, the median tumor thickness for all of the patients for whom an annotated tumor thickness was provided (n = 237; refs. 3, 12, 23, 25, 32, 35, 44, 81–88) was 4.2 mm (range: 0.1–44 mm). In comparison, our cohort contains 32% patients with PVM with tumor thickness > 4.0 mm with a median tumor thickness of 2.3 mm (range: 0.3–35 mm), strongly suggesting that our cohort is not biased towards “thicker” melanomas. Instead, patients with PVM appear to present with a higher tumor thickness (i.e., more advanced primary), and our cohort appears to reflect PVMs encountered in general clinical practice. Finally, because our patient cohort spanned over 60 years of institutional experience, the patients were not treated with uniform surgical and/or medical management strategies. It is unclear to what extent this lack of uniformity possibly impacted our findings. However, given the rarity of PVM, it is difficult to assemble a cohort of this size treated uniformly.

In conclusion, our analysis of clinical and histopathologic parameters in 100 patients with PVM supports the key prognostic significance of tumor thickness (≤2.00 mm or >2.00 mm) and dermal mitotic rate (<2/mm² or ≥2/mm²) in this rare subtype of primary melanoma. We propose a novel T-category for PVM based exclusively on these 2 parameters as a straightforward and robust tool to stratify risk in patients with this disease. Our results support the rationale to evaluate this system in additional cohorts of patients with PVM.

M.I. Ross reports receiving speakers bureau honoraria from Amgen Inc., Merck, and Provectus Biopharmaceuticals and is a consultant/advisory board member for Amgen Inc., GlaxoSmithKline, and Merck. J.E. Gershenwald is a consultant/advisory board member for Merck. M.T. Tetzlaff is a consultant/advisory board member for Myriad Genetics and Seattle Genetics. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P. Nagarajan, V.G. Prieto, M.T. Tetzlaff

Development of methodology: P. Nagarajan, M.T. Tetzlaff

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Nagarajan, J.L. Curry, C.A. Torres-Cabala, P.P. Aung, D. Ivan, C.F. Levenback, M. Frumovitz, A. Malpica, V.G. Prieto, M.T. Tetzlaff

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P. Nagarajan, J. Ning, J. Piao, C.A. Torres-Cabala, D. Ivan, M.I. Ross, M. Frumovitz, J.E. Gershenwald, M.A. Davies, M.T. Tetzlaff

Writing, review, and/or revision of the manuscript: P. Nagarajan, J.L. Curry, J. Ning, C.A. Torres-Cabala, P.P. Aung, M.I. Ross, C.F. Levenback, M. Frumovitz, J.E. Gershenwald, M.A. Davies, A. Malpica, V.G. Prieto, M.T. Tetzlaff

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Nagarajan, M.T. Tetzlaff

Study supervision: M.T. Tetzlaff

We thank Ms. Kim Anh-Vu for her excellent assistance in preparing the figures. We thank Ms. Stephanie Deming for her excellent support in medical editing.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.