There is a need for improved prognostic markers in melanoma. In this study, the authors tested the prognostic significance and clinicopathologic correlations of v-raf murine sarcoma viral oncogene homolog B1 (BRAF) and neuroblastoma RAS viral (v-ras) oncogene homolog (NRAS) mutations in patients with metastatic melanoma.
Clinical and pathologic data were collected retrospectively on melanoma patients who were clinically tested for BRAF (exon 15) and NRAS (exons 1 and 2) mutations at The University of Texas M. D. Anderson Cancer Center. Analyses were performed to identify significant associations of mutations with tumor and patient characteristics and with survival from the diagnosis of stage IV disease.
The genotypes of the full cohort (n = 677) were 47% BRAF mutation, 20% NRAS mutation, and 32% wild-type for BRAF and NRAS (“WT”). Tumor mutation status was associated (P = .008) with the risk of central nervous system involvement at the diagnosis of stage IV disease, with a higher prevalence observed in BRAF-mutant (24%) and NRAS-mutant (23%) patients than in WT patients (12%). Among patients with nonuveal melanoma who underwent mutation testing within 6 months of stage IV diagnosis (n = 313), patients with NRAS mutations had a median survival of 8.2 months from stage IV diagnosis, which was shorter than the median survival of WT patients (15.1 months; P = .004). Multivariate analysis of this population incorporating age, sex, metastases (M1) category, serum lactate dehydrogenase level, and mutation status confirmed that NRAS mutations are associated independently with decreased overall survival (vs WT; P = .005; hazard ratio, 2.05).
An estimated 70,230 patients will be diagnosed with cutaneous melanoma, the most lethal skin cancer, and approximately 8790 will die from this disease in 2011.1 Detailed analyses of clinical outcomes have identified several validated clinical and pathologic features that correlate with stage-specific survival in this disease.2-4 However, currently, there are no validated molecular prognostic markers in melanoma. The most common somatic event in melanoma is mutation of the serine-threonine kinase v-raf murine sarcoma viral oncogene homolog B1 (BRAF), which is a component of the RAS-RAF/mitogen-activated protein kinase/extracellular signal-regulated kinase kinase (MEK)/mitogen-activated protein kinase kinase (MAPK) signaling pathway. Overall, point mutations in the BRAF gene occur in 40% to 50% of melanomas.5 Over 90% of the mutations in BRAF result in substitution of the valine at position 600, resulting in activation of the downstream effectors of the RAS-RAF-MEK-MAPK pathway.6 This pathway also may be activated by point mutations in neuroblastoma RAS viral (v-ras) oncogene homolog (NRAS), which codes for a small guanine triphosphate (GTP)-binding protein. NRAS is mutated in 15% to 25% of melanomas, most frequently at hotspots in exon 1 (codon 12) and exon 2 (codon 61).5 The overwhelming majority of activating BRAF and NRAS mutations in melanomas have been mutually exclusive.7
Previous studies have examined associations of BRAF and NRAS mutations with primary tumor characteristics and prognosis in earlier stage disease.8-11 However, limited data are available for patients with metastatic disease.12, 13 An improved understanding of the prognostic significance of mutation status in these patients may help in the appropriate design and interpretation of clinical trials and add to the understanding of the biology of this disease. We performed a retrospective analysis in a large cohort of patients with advanced melanoma who underwent testing for both BRAF (exon 15) and NRAS (exons 1 and 2) mutations. The results of this analysis provide the first evidence that melanoma patients with an activating mutation in the NRAS gene have significantly shorter survival from stage IV diagnosis than patients without a mutation in either the BRAF gene or the NRAS gene, and the prognostic value of this marker is independent of current validated prognostic markers.
MATERIALS AND METHODS
Patient Selection and Clinical Data Collection
The results of Clinical Laboratory Improvement Amendments (CLIA)-certified testing for BRAF and NRAS mutations performed by The University of Texas M. D. Anderson Cancer Center (MDACC) Molecular Diagnostics Laboratory for melanoma patients from February 1, 2007 to September, 13, 2010 were reviewed under an Institutional Review Board-approved protocol. The specific mutations detected, patient demographics (age, sex, ethnicity), and primary tumor characteristics (date of diagnosis, anatomic location, histology, Breslow thickness, ulceration, mitotic rate) were recorded for all patients. Characteristics at the time of stage IV diagnosis (date, age, involved sites, serum lactate dehydrogenase [LDH]) also were collected; stage IV was defined according to American Joint Committee on Cancer (AJCC) Cancer Staging Manual seventh edition criteria, including M1a category (any tumor classification [T] and any lymph node status [N] with distant cutaneous and/or subcutaneous and/or lymph node metastasis [M] and nonelevated serum LDH) or M1b category (any T and any N with pulmonary metastasis without other visceral metastasis and nonelevated serum LDH) or M1c category (any T, any N with visceral metastasis or elevated serum LDH with any sites of distant metastasis).14 Treatments with the selective BRAF inhibitors vemurafenib (Roche, Indianapolis, Ind) or GSK2118436 (GlaxoSmithKline, Parsippany, NJ) or with the selective MEK inhibitors selumitinib (AstraZeneca, Wilmington, Del) or GSK1120212 (GlaxoSmithKline) also were recorded. The most recent clinical follow-up date and vital status (through January 15, 2011) were collected.
Mutation testing was performed by CLIA-certified pyrosequencing assays by the MDACC Molecular Diagnostics Laboratory. Pyrosequencing of BRAF exon 15 (inclusive of codons 595 to 601) and NRAS exon 1 (codons 12 and 13) and exon 2 (codons 60 and 61) was performed.
The association between continuous parameters and mutational status was assessed by analysis of variance (ANOVA). The Fisher exact test was used to compare the distribution of categorical variables between mutational groups. Distributions of time to event (time to stage IV and survival from stage IV) were estimated using the Kaplan-Meier method, and distributions were compared using the log-rank test. Cox proportional hazards regression analysis was used to assess the association between multiple covariates and the 2 survival parameters. For each parameter, an overall comparison was made first between the 3 mutational groups, and this was followed by each of the 3 pairwise comparisons. To adjust for multiplicity when making pairwise comparisons, the Tukey honestly significant difference method was used within the ANOVA framework for continuous variables, and the method of Holm and Bonferroni was used for categorical variables. Except as noted above, all statistical tests were performed 2-sided with a 5% Type I error rate. The software suite R (available at: http://www.r-project.org; [accessed September 1, 2011]) version 2.12.0 was used.
BRAF and NRAS Mutation Frequency
Overall, 677 patients with melanoma who had successful testing for both BRAF and NRAS were identified (Table 1). Mutations in exon 15 of BRAF (only) were identified in 320 patients (47.3%), and NRAS mutations (only) were identified in 136 patients (20.1%). The majority of BRAF mutations were represented by the valine to glutamic acid substitution at position 600 (V600E) (71.9% of BRAF mutations) and the valine to lysine substitution at position 600 (V600K) (22.5%). Substitutions at positions 60 and 61 accounted for 82.4% of NRAS mutations, most frequently a glutamine to arginine/lysine/lysine substitution at position 61 (Q61R/K/L). Four patients (0.6%) had activating mutations in both BRAF and NRAS. Two hundred seventeen patients (32.1%) did not have a mutation in either BRAF exon 15 or NRAS exon 1 or 2 and are referred to as wild type (WT) for subsequent analyses. Because of the rarity and uncertain functional significance of non-V600 BRAF exon 15 mutations (n = 6) and dual BRAF/NRAS mutations (n = 4), patients with these genotypes were excluded from further analyses.
Table 1. Frequency of BRAF and NRAS Mutations in the Study Patients
At the time of initial diagnosis of melanoma, patients with a BRAF mutation (“BRAF”) were significantly younger (median age, 49.8 years) than patients with NRAS mutation (“NRAS”) (median age, 55.7 years; P = .0008) or WT patients (median age, 59.5 years; P < .0001) (Table 2). There also was a significant association for mutation with race (P = .042), although there were very few non-Caucasian patients in this cohort (4.1%). NRAS mutations were less common in Hispanic patients (8.8%) compared with Caucasian patients (21.1%), whereas Asian/black/unknown patients had an increased WT frequency (61.1% vs 31.1% of Caucasians).
Table 2. Patient Demographics and Primary Tumor site by BRAF and NRAS Mutation Status
Patients without BRAF valine mutations at codon 600 (V600) (n = 6) and patients with double BRAF and NRAS mutations (n = 4) were excluded from this analysis. Patients who were not assessed for a specific characteristic were excluded from pair-wise and 3-group comparisons.
Includes Asian, Black, and unknown.
According to the American Joint Committee on Cancer AJCC Cancer Staging Manual, 7th edition.4
BRAF vs NRAS, P = .28; BRAF vs WT, P = 0.029; NRAS vs WT, P = .80.
Mutation status was strongly associated with the anatomic type of melanoma (P < .0001) (Table 2). Among patients who had a cutaneous primary (n = 516), 49% were BRAF mutated, 21% were NRAS mutated, and 30% were WT. Patients who had a mucosal melanoma (n = 28) had a similar prevalence of NRAS (18%) mutations, but fewer BRAF (7%) and, thus, more WT (75%). None of the uveal melanoma patients (n = 11) had a mutation in either BRAF or NRAS. Patients with an unknown primary tumor (n = 110) had a prevalence of BRAF (53%), NRAS (19%), and WT (28%) genotypes that was very similar to that of the patients with a cutaneous melanoma but differed significantly from those with a mucosal primary (P < .0001). Among the patients with documented cutaneous primary tumor characteristics, there was no significant association observed between mutation status and Breslow thickness (n = 424; P = .40), mitotic rate (n = 311; P = 0.35), or ulceration status (n = 345; P = .44) (Table 3). Mutation status was associated significantly with primary tumor site and histology. BRAF mutations were most prevalent in truncal melanomas (63.9%) and with superficial spreading morphology (63.2%); NRAS mutations were most prevalent in melanomas arising from the arm/leg (34.7%) and with nodular histology (28.3%) (Table 3).
Table 3. Primary Tumor Site and Histologic Characteristics of Cutaneous Melanomas by BRAF and NRAS Mutation Status
After excluding non-V600 BRAF mutants (n = 6), patients with concurrent BRAF and NRAS mutations (n = 4), and 1 patient with incomplete data, in total, 519 patients in the cohort who developed stage IV melanoma (48.6% BRAF, 20% NRAS, and 31.4% WT) were further analyzed. The interval from the diagnosis of melanoma to the diagnosis of stage IV disease trended toward a shorter duration for the NRAS patients, but this difference was not statistically significant using 3-group or 2-group comparisons (Table 4). BRAF patients were younger than NRAS patients (P = .0047) and WT patients (P < .0001) at stage IV diagnosis.
Table 4. Clinical Characteristics at Stage IV Diagnosis According to BRAF and NRAS Mutation Statusa
Patients who were not assessed for a specific characteristic were excluded from pair-wise and 3-group comparisons.
Of 530 patients with metastatic disease, patients with dual BRAF and NRAS mutations (n = 4) and patients who had nonvaline 600 BRAF mutations (n = 6) were excluded from further analysis. An additional patient was excluded because of incomplete clinical data.
BRAF vs NRAS, P = .52; BRAF vs WT, P = .64, NRAS vs WT, P = .95.
Metastasis classification (M) is categorized according to anatomic site of involvement only and excludes LDH. M1a indicates skin, subcutaneous, and lymph nodes; M1b, lung; M1c, any other visceral site.
BRAF vs NRAS, P = .72; BRAF vs WT, P = .054; NRAS vs WT, P = .30.
Median age, y
Interval from initial diagnosis to stage IV: Median [range], mo
Serum LDH had a nonsignificant trend toward an association with mutation status (P = .09), with NRAS (15.4%) and WT (14.7%) patients less frequently having elevated LDH than BRAF patients (22.6%). The M1 category, as defined by the anatomic sites of involvement only (ie, excluding serum LDH), at the time of diagnosis of stage IV disease was associated significantly with tumor mutation status (P = .0018). A 2-group comparison identified a significant difference (P = .0007) between BRAF and WT patients, with a higher rate of M1a and lower rate of M1b patients in the BRAF group. A pair-wise comparison within the M1 category identified no significant difference for NRAS patients versus either BRAF or WT patients.
An analysis of anatomic sites that characterize M1c disease (ie, nonpulmonary visceral) identified a significant association of mutation status with the rate of central nervous system (CNS) involvement at the time of stage IV diagnosis (P = .0076). There was a higher rate of CNS involvement among the BRAF patients (24.4%; P = .01) and the NRAS patients (23.1%; P = .056) compared with the WT patients (12.4%).
In contrast, there was no association with mutation status and the prevalence of metastases to the liver (P = .79) bone (P = .43), skin (P = .42) or lymph nodes (P = .83). There was a significant association with lung involvement (P = .049), with a lower rate of lung metastases among the BRAF (55.2%) and NRAS (56.7%) patients compared with the WT patients (66.9%) at stage IV presentation (Table 4).
Overall Survival From Stage IV
Tumor mutation status correlated with overall survival from the diagnosis of stage IV disease in the full cohort (n = 519) (Fig. 1A). NRAS patients (n = 104) had significantly shorter median overall survival (15.5 months) than WT patients (n = 163; 23.5 months; P = .02). The median overall survival of BRAF patients (n = 252; 24.2 months) did not differ from that of WT patients.
The analysis of survival by mutation status may be biased by the inclusion of patients who survived with stage IV disease for several years before the implementation of molecular testing. The relevance of this potential confounder in this cohort is supported by the atypically long survival from stage IV diagnosis detected in each of the groups.4 Therefore, an additional analysis was performed among patients who underwent mutation testing within 6 months of their stage IV diagnosis. Because recent reports support the finding that targeted therapies against the RAS-RAF-MEK-MAPK may have marked clinical activity in patients with activating BRAF mutations,15-17 the BRAF patients who had enrolled on nonrandomized clinical trials of selective BRAF or MEK inhibitors (n = 41) were analyzed separately from BRAF patients who had not (n = 112). No BRAF patients in this study were enrolled on randomized protocols of selective BRAF or MEK inhibitors. Patients with uveal melanoma (n = 11) also were excluded because of the lack of BRAF or NRAS mutations in these patients and the distinctive clinical course of this disease.
In total, 313 patients with nonuveal, metastatic melanoma who underwent molecular testing within 6 months of their stage IV diagnosis were identified (median time to testing, 43 days). The median follow-up duration from stage IV diagnosis was 12 months. NRAS patients (n = 66) had the worst outcomes, with a median survival of only 8.2 months from the diagnosis of stage IV, which was significantly shorter than the median survival of WT patients (15.1 months; P = .004) (Fig. 1B). BRAF patients who received treatment with a BRAF or MEK inhibitor had significantly longer survival (n = 41; median overall survival, not reached; median follow-up, 15.6 months) compared with WT patients (P = .0145). BRAF patients who did not receive a BRAF inhibitor (n = 112; median overall survival, 10.3 months) had a nonsignificant trend toward shorter survival compared with WT patients (P = .10). BRAF patients who received BRAF inhibitors survived longer than those who did not receive BRAF inhibitors (P = .0002).
Multivariate Cox regression modeling of these data was performed for the cohort of 313 patients (Table 5). Consistent with their status as validated prognostic markers in stage IV melanoma, elevated serum LDH (hazard ratio [HR], 2.75 vs not elevated; P < .0001), M1b category (HR, 3.29 vs M1a; P = .03), and M1c category (HR, 4.02 vs M1a; P = .007) were independent predictors of survival in this group. In contrast, age (P = .37) and sex (P = .91) did not predict survival. The observed differences in survival of NRAS patients (HR, 2.05; P = .005) and of BRAF patients who received treatment with BRAF or MEK inhibitors (HR, 0.45; P = .01), compared with the survival of WT patients, remained significant on multivariate analysis.
Table 5. Multivariate Analysis of Survival From Diagnosis of Stage IV Melanoma
M1 classification is categorized according to anatomic site(s) of involvement only and excludes LDH.
Includes treatment with inhibitors against BRAF (vemurafenib; GSK2118436; GlaxoSmithKline, Parsippany, NJ) and mitogen-activated protein kinase/extracellular signal-regulated kinase kinase (MEK) (selumitinib; GSK1120212; GlaxoSmithKline).
This study represents 1 of the largest single-institution cohorts of melanoma patients characterized for activating BRAF and NRAS mutations to date and, to our knowledge, the largest cohort of stage IV patients. The study is strengthened by the use of a clinically certified mutation-detection method for all patients and by the inclusion of testing for mutations in exon 1 (as well as exon 2) of NRAS, which was not performed in some previous studies in melanoma.12, 18 We report here the novel findings that the presence of an NRAS mutation correlates with shorter survival from the diagnosis of stage IV melanoma and that the presence of either a BRAF or NRAS mutation is associated with an increased risk of CNS involvement at initial stage IV diagnosis. Our study has also confirmed several previously reported associations with BRAF mutations, including age,19 primary tumor site,20 and improved survival with targeted therapies against the RAS-RAF-MEK-MAPK pathway.15
Studies of primary tumor characteristics in which both BRAF testing and NRAS testing were performed have been discordant. Two recent consecutive series of >200 primary melanomas both reported that primary tumors with NRAS mutations were associated with increased Breslow thickness.8, 12 Those studies, however, had conflicting findings regarding the correlation of NRAS mutation status with mitotic rate and ulceration. In our study, we did not identify any significant associations between mutation status and primary tumor Breslow thickness, mitotic rate, or ulceration (Table 3). However, because our patient population was selected by retrospectively identifying patients who had undergone clinically indicated mutation testing and, thus, consisted overwhelmingly of patients who developed distant metastases, it is possible that these results are not representative of all patients with primary melanoma. It is noteworthy that the striking similarity of the mutation profile between patients with unknown primary melanomas and those with cutaneous primary tumors supports the hypothesis that the majority of patients who have unknown primary melanoma had a regressed or occult cutaneous primary rather than an occult mucosal melanoma.
Previous analyses comparing NRAS mutations with patient survival also have reported discordant results. A recent prospective study identified no significant difference in overall survival from initial melanoma diagnosis among NRAS-mutant melanoma patients.8 Another recent prospective study of 249 Australian melanoma patients reported shorter melanoma-specific survival after the initial melanoma diagnosis for NRAS patients compared with WT patients.12 This difference was attributable to shorter relapse-free survival after initial treatment of locoregional disease among the NRAS patients, a trend also observed in our study (Table 4). The study that identified shorter relapse-free survival did not detect a significant difference in survival according to NRAS mutation status after the diagnosis of metastatic disease, but its power was limited by a low prevalence of patients with distant metastases and few relevant events for analysis.12 Our subset of 313 patients with nonuveal, metastatic melanoma who underwent molecular testing within 6 months of their stage IV diagnosis had a 45.4% mortality rate during the examined follow-up period, allowing for improved sensitivity in survival analyses. An additional retrospective study demonstrated that the NRAS-mutated tumor genotype was associated with increased overall survival (compared with the BRAF-mutated and WT tumor genotypes).13 However, that study included only 82 patients with stage IV disease. In our study, compared with WT patients, NRAS mutation was associated with shorter survival from stage IV diagnosis both in the full cohort of all patients with metastatic melanoma (n = 519) and in the cohort of patients who underwent molecular testing within 6 months of their stage IV diagnosis (n = 313). This difference remained significant on multivariate analysis using validated prognostic markers for this disease.
There is no clear etiology for the shorter survival from stage IV for the NRAS-mutant patients. The presence of an NRAS mutation did not correlate with established stage IV prognostic factors, such as elevated LDH or advanced M1 category. We did observe an increased rate of CNS involvement at stage IV diagnosis among both BRAF and NRAS patients compared with WT patients. However, survival from stage IV diagnosis for patients without initial CNS involvement was significantly shorter for NRAS patients (8.3 months) than for WT patients (14.3 months; P = .01) (Fig. 2). If an increased prevalence of CNS involvement in BRAF and NRAS patients at stage IV diagnosis is confirmed elsewhere, then it may suggest a role for increased CNS surveillance among these patients.
The trend for shorter survival from stage IV diagnosis for BRAF mutant patients who did not receive BRAF or MEK inhibitors compared with BRAF patients who did receive these agents is similar to the findings recently reported by Long et al of 197 consecutive advanced melanoma patients.21 These results are also consistent with the results from the recently reported BRIM3 randomized phase 3 clinical trial of the BRAF inhibitor vemurafenib.15
Our study does have limitations. We included only patients who attended a single tertiary cancer center and for whom mutation testing was considered clinically indicated. However, our population appears very typical genetically, because the prevalence of BRAF and NRAS mutations in this cohort is nearly identical that that reported in 2 different meta-analyses of >2000 patients each.5, 6 In addition, we note that the frequencies of V600E and V600K mutations in our cohort are similar to those reported in the Australian study of 197 patients with advanced melanoma.21
The interpretation of the shorter survival among BRAF patients who did not receive BRAF or MEK inhibitors compared with WT patients must be viewed with caution, because it was not possible to discern in this study why these patients were not enrolled on clinical trials with BRAF or MEK inhibitors. Finally, additional genetic aberrations may characterize each of the mutation-defined cohorts here, such as loss of phosphatase and tensin homolog (PTEN)22 or activating c-KIT mutations.23
The emergence of BRAF-targeted therapies that benefit only patients with activating BRAF mutations will almost certainly lead to clinical trials designed specifically for BRAF-WT patients in the near future. Identifying effective therapeutic approaches for, and improving outcomes in, patients who have melanoma with a WT BRAF gene is now one of the highest priorities in this disease. Our data support that patients who have melanoma with an NRAS mutation represent a distinct cohort with a highly aggressive disease and shorter survival with stage IV disease. Notably, the difference in overall survival from stage IV between NRAS patients and WT patients is substantial (approximately 7 months) and is comparable to the overall survival benefit observed in the recently reported phase 3 trial of ipilimumab.24 Our results support the need for stratification with respect to NRAS mutation status for the design and interpretation of future melanoma clinical trials for BRAF-WT patients. In addition, these results highlight the critical need for more effective therapies for patients who have melanoma with activating NRAS mutations.
This work was supported by The M. D. Anderson Melanoma Specialized Program in Research Excellence (SPORE) program (P50 CA93459). M.A.D. is supported by an American Society of Clinical Oncology Career Development Award. Both M.A.D. and A.J.L. are supported by the M. D. Anderson Physician-Scientist Program.
CONFLICT OF INTEREST DISCLOSURES
Jeffrey Gershenwald is on the advisory board and receives compensation from GlaxoSmithKline. Kevin Kim is on the advisory boards and receives compensation from Genetech and Roche; has received honoraria from Genetech; and receives research funding from Astra-Zeneca, Roche, Genetech, and GlaxoSmithKline. Michael Davies is a consultant and receives compensation from GlaxoSmithKline, and he receives research support from AstraZeneca, GlaxoSmithKline, Merck, and Roche. Alexander Lazar and his wife own stock in Roche and GlaxoSmithKline.