Patent Application
20160273052
Cutaneous melanoma (CM) is an aggressive form of cancer presenting over 76,000 diagnosed cases in 2012.1 CM tumors develop through a number of discreet stages during the progression from a benign melanocytic nevus to a malignant metastatic tumor. Generally, benign nevi present as thin, pigmented lesions. After the acquisition of key genetic mutations and the initiation of cytoarchitectural modifications leading to shallow invasion of the skin, the lesions begin growing radially, a process referred to as the radial growth phase. Upon escape from growth control mediated by surrounding keratinocytes, stromal invasion to deeper regions of the dermis occurs, marking the progression to the vertical growth phase. The vertical growth phase, along with the genetic alterations that accompany this process, is thought to be the critical stepping stone in the development of metastatic melanoma.
As is common in many other cancers, if CM is detected in the early stages of tumor progression and appropriately treated, then a long-term metastasis free and overall survival following diagnosis is likely for the majority of patients.2,3 For example, subjects diagnosed with low risk, stage I CM tumors have a 5-year overall survival rate of 91-97%.3 A number of histological factors are used to stage CM and are associated with prognosis. These factors include Breslow thickness, mitotic index, ulceration, and spread of disease from the primary tumor to sentinel and/or regional lymph nodes.3-7 Tumor stage is determined based upon these histopathological parameters using the well-known TNM (T=primary tumor, N=regional lymph nodes, M=distant metastases) system that defines stages 0-IV.3 The TNM staging system is highly accurate for metastasis-free survival for stage 0 melanomas (5-year survival of 99%), and stage IV melanomas (5-year survival <10%), in which distant metastasis was detected at the time of primary diagnosis. Metastasis and short-term survival has been documented for subjects with stage I disease, with 5-10% of stage I tumors reporting metastatic activity. So, while the majority of patients with clinical stage I disease have a low chance of metastasis risk disease some patients will develop metastatic disease.
Prognosis for clinical stage II and stage III cases has poor accuracy as there is a large range within each stage and a larger overlap between the stages in 5-year survival rates. Under the current staging system, the 5-year survival rate for clinical stage II subjects is 53-82%, while the stage III 5-year survival is rate 22-68%.3,8 The distinguishing hallmark between clinical stage II and stage III tumors is the presence of localized metastasis of CM cells in the sentinel lymph node (SLN). Patients with a positive SLN are clinical stage III. However, high false negative rates and disease recurrence are associated with histological analysis of SLNs as evidenced by the wide ranges of metastatic free survival and overall survival in stage II and III patients. Immunohistochemical and genetic amplification techniques designed to improve the common hematoxylin and eosin staining methods for detection of regional disease have been developed but only provide marginal improvements9,10. Biomarkers have been identified in SLN tissue, and analyzed to improve the ability to recognize CM cells in SLNs, however, these methods have limited improvement in accuracy and are compounded by extensive tissue sampling from invasive biopsy of the lymph node.6,11 In addition, melanomas can enter the blood directly by intravasating into venous capillaries. Thus, the low sensitivity of SLN biopsy may relate to a direct hematogenous metastatic event versus an inaccurate SLN biopsy result. Inaccurate prognosis for metastatic risk has profound effects upon patients that are treated according to a population approach rather than an individual or personalized approach. For example, CM patients categorized as stage III through the use of current histological techniques, but who have an actual individual risk of metastasis that is low (false positive), are inappropriately exposed to over-treatment that includes enhanced surveillance, nodal surgery, and chemotherapy.12 Similarly, patients determined to have stage I or II disease who actually have a high risk for metastasis (false negative), are at risk of under-treatment. In addition, SLN biopsy exposes patients to significant clinical complications, such as lymphedema, and has a low positivity rate. For example, guidelines currently recommend that patients with CM staged at 1b (Breslow's thickness ≧0.75 mm but <1.00 mm or presence of ≧1 mitosis at any Breslow's thickness) are recommended to undergo SLN biopsy yet only 5% of SLN yield positivity. Meaning that of 20 patients with stage 1b melanomas who undergo SLN biopsy, 19 will be negative and exposed to a surgical complication of SLN biopsy.12 Similarly, all pathologic stage II patients (Breslow's thickness >1.0 mm are recommended to undergo SLN biopsy yet only 18% will have a positive SLN.12,13
To this end, gene expression profile (GEP) signatures have been developed and some have been shown to have powerful prognostic capabilities in a number of malignant diseases14-18. One such signature has been used for prognostication of uveal melanoma, a tumor of melanocytic origin that develops in the eye. Like cutaneous melanoma, treatment of the primary uveal tumor is highly effective. 2-4% of uveal melanoma patients present with evidence of clinical metastasis at the time of diagnosis, yet up to 50% of uveal melanoma patients develop systemic metastases within five years of diagnosis regardless of primary eye tumor treatment (radiation therapy or enucleation)19. This means that a micrometastatic event has occurred in approximately 50% of uveal melanoma patients prior to primary eye tumor treatment. A GEP signature has recently been developed that can accurately distinguish uveal melanoma tumors that have a low risk of metastasis from those that have a high risk14,20. To assess genetic expression RT-PCR analysis is performed for fifteen genes (twelve discriminating genes and three control genes) that are differentially expressed in tumors with known metastatic activity compared to tumors with no evidence of metastasis. The uveal melanoma gene signature separates cases into a low risk group that has greater than 95% metastasis free survival five years after diagnosis, and a high risk group with less than 20% metastasis free survival at the same time point. The signature has been extensively validated in the clinical setting, and has been shown to provide a significant improvement in prognostic accuracy compared to classification by TNM staging criteria20,21.
A number of groups have published genomic analysis of tumors in cutaneous melanoma22-29. While some studies have focused on the genetic alterations in malignant melanoma cells compared to normal melanocytes, others have compared benign nevi to tumors in the radial or vertical growth phases, or primary tumor to metastatic tumors. At the time the studies contained within this patent were designed and implemented (2010), no evidence could be found in the literature or other public domain sources that indicated a gene expression profile test focused solely on the primary melanoma tumor could be developed for the clinical application of predicting metastasis in patients with CM.22-24,27-29 In addition, all studies utilized fresh frozen CM samples rather than formalin fixed paraffin embedded (FFPE) tumor tissue. All of the studies related to this invention have only used FFPE primary tumor tissue.
There is a need in the art for a more objective method of predicting which tumors display aggressive metastatic activity. Development of an accurate molecular footprint, such as the gene expression profile assay encompassed by the invention disclosed herein, by which CM metastatic risk could be assessed from primary tumor tissue would be a significant advance forward for the field.
In one embodiment, the invention as disclosed herein is a method for predicting the risk of metastasis of a primary cutaneous melanoma tumor, the method comprising: (a) measuring the gene-expression levels of at least eight genes selected from the group consisting of BAP1_varA, BAP1_varB, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6, in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) determining a gene-expression profile signature comprising the gene expression levels of the at least eight genes; (c) comparing the gene-expression profile to the gene-expression profile of a predictive training set; and (d) providing an indication as to whether the primary cutaneous melanoma tumor is class 1 or class 2 of metastasis when the gene expression profile indicates that expression levels of at least eight genes are altered in a predictive manner as compared to the gene expression profile of the predictive training set.
In another embodiment, the invention as disclosed herein is a method for predicting the risk of metastasis of a primary cutaneous melanoma tumor, the method comprising: (a) measuring the gene-expression levels of at least eight genes selected from the group consisting of BAP1_varA, BAP1_varB, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6, in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) determining a gene-expression profile signature comprising the gene expression levels of the at least eight genes; (c) comparing the gene-expression profile to the gene-expression profile of a predictive training set; and (d) providing an indication as to whether the primary cutaneous melanoma tumor is class A, class B, or class C of metastasis when the gene expression profile indicates that expression levels of at least eight genes are altered in a predictive manner as compared to the gene expression profile of the predictive training set. In one embodiment, the invention as disclosed herein is a method for predicting the risk of metastasis of a primary cutaneous melanoma tumor, the method comprising: (a) measuring the gene-expression levels of at least eight genes selected from the group consisting of BAP1_varA, BAP1_varB, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6, in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) determining a gene-expression profile signature comprising the gene expression levels of the at least eight genes; (c) comparing the gene-expression profile to the gene-expression profile of a predictive training set; and (d) providing an indication as to whether the primary cutaneous melanoma tumor has a low risk to high of metastasis when the gene expression profile indicates that expression levels of at least eight genes are altered in a predictive manner as compared to the gene expression profile of the predictive training set.
In an additional embodiment, the invention as disclosed herein is a method of treating cutaneous melanoma in a patient, comprising the steps of: (a) obtaining the gene expression level of at least one of BAP1, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6 in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) comparing the gene expression level of the at least one gene to the gene-expression level the same gene or genes taken from a predictive training set; (c) making a determination as to whether the gene-expression level of the at least one gene is altered in a predictive manner; and (d) targeting the at least one gene for therapy when the determination is made in the affirmative.
Specific embodiments of the invention will become evident from the following more detailed description of certain embodiments and the claims.
The inventors reviewed the current state of the art for evidence of microarray analyses focused on distinguishing differential genetic expression profiles that characterize cutaneous melanoma tumors. Seven studies were identified that utilized microarray technology to determine genetic expression comparing various stages of cutaneous melanoma.22-27,29 The objective was to identify genes that were dysregulated between radial growth phase to vertical growth phase, or in primary melanomas compared to metastatic tumors. Mauerer, et. al., assessed gene expression differences in melanocytic nevi compared to primary melanoma, melanocytic nevi compared to metastatic melanoma, and primary melanoma compared to metastatic melanoma25. Nothing in the art was found relating to evaluation of primary cutaneous melanoma tumors relative to subsequent metastatic vs. non-metastatic outcomes. In an attempt to arrive at a putative gene expression profile set, the inventors focused on genes isolated from primary melanoma tumor samples and metastatic melanoma tumor samples that had an observed up-regulation greater than 2-fold, or down-regulation greater than 3-fold, and established those as potential mediators of metastatic progression. By these criteria, 26 up-regulated genes and 78 down-regulated genes were chosen as the basis for comparison to other expression analysis studies. This 104-gene panel was subsequently compared to expression data sets reported in Scatolini et. al., Jaeger, et. al., Winnipenninckx et. al., Haqq, et. al., Smith et. al., and Bittner et. al.22-24,26,27,29 Additionally, the expression data from Onken, et. al., which reported 74 genes that were differentially regulated in metastatic and non-metastatic uveal melanoma tumors was compared to the 104-gene panel.14
In one embodiment, the invention as disclosed herein is a method for predicting the risk of metastasis of a primary cutaneous melanoma tumor, the method comprising: (a) measuring the gene-expression levels of at least eight genes selected from the group consisting of BAP1_var_A, BAP1_var_B, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6, in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) determining a gene-expression profile signature comprising the gene expression levels of the at least eight genes; (c) comparing the gene-expression profile to the gene-expression profile of a predictive training set; and (d) providing an indication as to whether the primary cutaneous melanoma tumor is class 1 (low risk) or class 2 (high risk) of metastasis when the gene expression profile indicates that expression levels of at least eight genes are altered in a predictive manner as compared to the gene expression profile of the predictive training set.
In another embodiment, the invention as disclosed herein is a method of treating cutaneous melanoma in a patient, comprising the steps of: (a) obtaining the gene expression level of at least one of BAP1_varA, BAP1_varB, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6 in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) comparing the gene expression level of the at least one gene to the gene-expression level the same gene or genes taken from a predictive training set; (c) making a determination as to whether the gene-expression level of the at least one gene is altered in a predictive manner; and (d) targeting the at least one gene for therapy when the determination is made in the affirmative.
In an additional embodiment, the invention as disclosed herein is a method for predicting the risk of metastasis of a primary cutaneous melanoma tumor, the method comprising: (a) measuring the gene-expression levels of at least eight genes selected from the group consisting of BAP1_varA, BAP1_varB, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6, in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) determining a gene-expression profile signature comprising the gene expression levels of the at least eight genes; (c) comparing the gene-expression profile to the gene-expression profile of a predictive training set; and (d) providing an indication as to whether the primary cutaneous melanoma tumor is class A (low risk), class B (intermediate risk), or class C (high risk) of metastasis when the gene expression profile indicates that expression levels of at least eight genes are altered in a predictive manner as compared to the gene expression profile of the predictive training set.
In one embodiment, the invention as disclosed herein is a method for predicting the risk of metastasis of a primary cutaneous melanoma tumor, the method comprising: (a) measuring the gene-expression levels of at least eight genes selected from the group consisting of BAP1_varA, BAP1_varB, MGP, SPP1, CXCL14, CLCA2, S100A8, BTG1, SAP130, ARG1, KRT6B, GJA, ID2, EIF1B, S100A9, CRABP2, KRT14, ROBO1, RBM23, TACSTD2, DSC1, SPRR1B, TRIM29, AQP3, TYRP1, PPL, LTA4H, and CST6, in a sample taken from the primary cutaneous melanoma tumor, wherein measuring gene-expression levels of the at least eight genes comprises measurement of a level of fluorescence by a sequence detection system following RT-PCR; (b) determining a gene-expression profile signature comprising the gene expression levels of the at least eight genes; (c) comparing the gene-expression profile to the gene-expression profile of a predictive training set; and (d) providing an indication as to whether the primary cutaneous melanoma tumor has a low risk to high of metastasis when the gene expression profile indicates that expression levels of at least eight genes are altered in a predictive manner as compared to the gene expression profile of the predictive training set.
As used here in, “metastasis” is defined as the recurrence or disease progression that may occur locally (such as local recurrence and in transit disease), regionally (such as nodal micrometastasis or macrometastasis), or distally (such as brain, lung and other tissues). “Class 1 or class 2 of metastasis” as defined herein includes low-risk (class 1) or high-risk (class 2) of metastasis according to any of the statistical methods disclosed herein, Additionally, “cutaneous melanoma metastasis” as used herein includes sentinel lymph node metastasis, in transit metastasis, distant metastasis, and local recurrence.
As used herein, a “sequence detection system” is any computational method in the art that can be used to analyze the results of a PCR reaction. One example, inter alia, is the Applied Biosystems HT7900 fast Real-Time PCR system. In certain embodiments, gene expression can be analyzed using, e.g., direct DNA expression in microarray, Sanger sequencing analysis, Northern blot, the Nanostring® technology, serial analysis of gene expression (SAGE), RNA-seq, tissue microarray, or protein expression with immunohistochemistry or western blot technique.
As defined herein, “gene-expression profile signature” is any combination of genes, the measured messenger RNA transcript expression levels or direct DNA expression levels or immunohistochemistry levels of which can be used to distinguish between two biologically different corporal tissues and/or cells and/or cellular changes. In certain embodiments, gene-expression profile signature is comprised of the gene-expression levels of at least 28, 27, 26, 25, 24, 23, 22, 21, 20, 19, 18, 17, 16, 15, 14, 13, 12, 11, 10, 9, or 8 genes. In a further embodiments, the genes selected are KRT6B, GJA1, AQP3, TRIM29, TYRP1, RBM23, MGP and EIF1B; SAP130, ARG1, KRT6B, EIF1B, S100A9, KRT14, ROBO1, RBM23, TRIM29, AQP3, TYRP1 and CST6; GJA1, PPL, ROBO1, MGP, TRIM29, AQP3, RBM23, TACSTD2, TYRP1, KRT6B, EIF1B and DSC1; CRABP2, TYRP1, PPL, EIF1B, SPRR1B, DSC1, GJA1, AQP3, MGP, RBM23, CLCA2 and TRIM29; RBM23, TACSTD2, CRABP2, PPL, GJA1, SPP1, CXCL14, EIF1B, AQP3, MGP, LTA4H and KRT6B; S100A8, TACSTD2, BAP1_varA, KRT6B, EIF1B, TRIM29, TYRP1, CST6, PPL, RBM23, AQP3, GJA1, SPRR1B and ARG1; CST6, KRT6B, LTA4H, CLCA2, CRABP2, TRIM29, CXCL14, PPL, ARG1, RBM23, GJA1, AQP3, TYRP1, SPP1, DSC1, TACSTD2, EIF1B, and BAP1_varB.
As defined herein, “predictive training set” means a cohort of CM tumors with known clinical metastatic outcome and known genetic expression profile, used to define/establish all other CM tumors, based upon the genetic expression profile of each, as a low-risk, class 1 tumor type or a high-risk, class 2 tumor type. Additionally, included in the predictive training set is the definition of “threshold points” points at which a classification of metastatic risk is determined, specific to each individual gene expression level.
As defined herein, “altered in a predictive manner” means changes in genetic expression profile that predict metastatic risk.
In one embodiment, the cutaneous melanoma tumor is taken from a formalin-fixed, paraffin embedded wide local excision sample. In another embodiment, the cutaneous melanoma tumor is taken from formalin-fixed, paraffin embedded punch biopsy sample.
In certain embodiments, analysis of genetic expression and determination of outcome is carried out using radial basis machine and/or partition tree analysis, LRA, K-nearest neighbor, or other algorithmic approach. These analysis techniques take into account the large number of samples required to generate a training set that will enable accurate prediction of outcomes as a result of cut-points established with an in-process training set or cut-points defined for non-algorithmic analysis, but that any number of linear and nonlinear approaches can produce a statistically significant and clinically significant result. Among the advantages of use of the methods disclosed herein are relating to, e.g., the in excess of 140 samples in the training set used to cover either heterogeneity or adequately handle smaller gene expression profile changes that could not adequately predict outcomes in an independent test set. As defined herein, “Kaplan-Meier survival analysis” is understood in the art to be also known as the product limit estimator, which is used to estimate the survival function from lifetime data. In medical research, it is often used to measure the fraction of patients living for a certain amount of time after treatment. “JMP Genomics® software” provides an interface for utilizing each of the predictive modeling methods disclosed herein, and should not limit the claims to methods performed only with JMP Genomics® software.
The Examples that follow are illustrative of specific embodiments of the invention, and various uses thereof. They are set forth for explanatory purposes only, and should not be construed as limiting the scope of the invention in any way.
Formalin fixed paraffin embedded (FFPE) primary cutaneous melanoma tumor specimens arranged in 5 μm sections on microscope slides were acquired from multiple institutions under Institutional Review Board (IRB) approved protocols. All tissue was reviewed by a pathologist and sections with <60% tumor in an area compatible with dissection were excluded from the study. Tissue with >60% tumor were marked and tumor tissue was dissected from the slide using a sterile disposable scalpel, collected into a microcentrifuge tube, and deparaffinized using xylene. RNA was isolated from each specimen using the Ambion RecoverAll Total Nucleic Acid Isolation Kit (Life Technologies Corporation, Grand Island, N.Y.). RNA quantity and quality were assessed using the NanoDrop 1000 system and the Agilent Bioanalyzer 2100.
2. cDNA Generation and RT-PCR Analysis
RNA isolated from FFPE samples was converted to cDNA using the Applied Biosystems High Capacity cDNA Reverse Transcription Kit (Life Technologies Corporation, Grand Island, N.Y.). Prior to performing the RT-PCR assay each cDNA sample underwent a 14-cycle pre-amplification step. Pre-amplified cDNA samples were diluted 20-fold in TE buffer. 50 ul of each diluted sample was mixed with 50 ul of 2× TaqMan Gene Expression Master Mix, and the solution was loaded to a custom high throughput microfluidics gene card containing primers specific for 28 class discriminating genes and 3 endogenous control genes. Each sample was run in triplicate. The gene expression profile test was performed on an Applied Biosystems HT7900 machine (Life Technologies Corporation, Grand Island, N.Y.).
Mean Ct values were calculated for triplicate sample sets, and ΔCt values were calculated by subtracting the mean Ct of each discriminating gene from the geometric mean of the mean Ct values of all three endogenous control genes. ΔCt values were standardized according to the mean of the expression of all discriminant genes with a scale equivalent to the standard deviation. Three control genes were selected based upon analysis using geNorm. Various linear and non-linear predictive modeling methods, including radial basis machine, k-nearest neighbor, partition tree, logistic regression, discriminant analysis and distance scoring, were performed using JMP Genomics SAS-based software (JMP, Cary, N.C.). Kaplan-Meier curves reflecting metastasis free survival were also generated in IMP, and statistical significance was calculated according to the Log Rank method. Cox univariate and multivariate regression analysis was performed using WinSTAT for Microsoft Excel version 2012.1.
Genetic expression of the discriminant genes in the signature (Table 1) was assessed in a cohort of 273 cutaneous melanoma samples using RT-PCR (
Using JMP Genomics® software and clinical data analysis, a training set of 162 cutaneous melanoma samples was generated that could accurately predict the risk of metastasis based upon the 28 gene signature. The training set contained 13 stage 0 in situ melanomas, 61 stage I, 70 stage II, 17 stage III, and 1 stage IV melanomas. Metastatic risk was assessed using a radial basis machine predictive modeling algorithm, which reports class 1 (low risk of metastasis) or class 2 (high risk of metastasis). ΔCt values generated from RT-PCR analysis of the training set cohort were standardized to the mean for each gene, with a scale equivalent to the standard deviation. Analyses were also performed using KNN, PTA, and discriminant analysis to confirm the results from the RBM approach (as discussed below). The training set prediction algorithm was then validated using an independent cohort of 111 cutaneous melanoma samples, comprised of four stage 0 in situ melanomas, 56 stage I, 36 stage II, 12 stage III, and 2 stage IV melanomas. Samples in the validation set were standardized prior to analysis using the factorials generated during standardization of the training set. Area under the receiver operator characteristic (ROC) curve, accuracy, sensitivity (prediction of high risk metastatic event), and specificity (prediction of nonmetastatic outcome) were statistical endpoints for the analysis. In the training set cohort, ROC=0.8730, accuracy=82%, sensitivity=89%, and specificity=74% (
Kaplan-Meier survival analysis was also performed for the training and validation cohorts. In the training set, 5-year metastasis free survival (MFS) was 93% for class 1 cases, and 31% for class 2 cases (p<0.0001;
JMP® Genomics software allows for analysis using linear and non-linear predictive modeling methods. To assess whether accuracy of metastatic risk prediction for the validation cohort was limited to the RBM method, partition tree, K-nearest neighbor, logistic regression, and discriminant analysis were performed (
To assess the impact of stage 0 in situ melanoma samples on the predictive capabilities of the training set, either a) samples were removed from all stage 0 the training and validation cohorts, generating a new training set of 149 cases; or b) all stage 0 samples were included in the training set only, generating a training set comprised of 166 cases. The new cutaneous melanoma predictive training sets were used to train a 107 sample validation cohort (the 111 sample validation set tested to this point, with four stage 0 in situs removed). Radial basis machine prediction was performed as described above, and ROC, accuracy of prediction, and 5-year MFS were assessed for the training and validation sets (
Accurate prediction of high risk metastatic cases is extremely important in order to prevent those patients who are likely to metastasize from receiving a low risk treatment protocol. Thus, a measure of success for the predictive gene set is achievement of greater than 88% sensitivity in both the training and validation sets. As shown in Table 3 below, and in
A total of 29 stage III, sentinel lymph node (SNL) positive, samples were included in the study population (17 in the training set and 12 in the validation set). These results are shown below in Table 4. RBM prediction of risk for the stage III samples resulted in 25 of 29 (86%) cases classified as high risk, class 2. Metastasis beyond the SLN was documented for 23 of the 29 cases, and 21 of those were predicted to be class 2 by the algorithm. Metastatic events localized to distant lymph nodes were documents for 13 cases, 12 (92%) of which were accurately predicted to be class 2. Similarly, 6 of 7 (86%) of in transit metastasis cases were assigned to class 2, indicating that the 28-gene predictive signature accurately predicts all aggressive cutaneous melanoma tumor types, including SLN positive, distant nodal, in transit, distant visceral metastases and local recurrence.
Cox univariate and multivariate regression analysis was performed for the 107 sample validation cohort following prediction of metastatic risk (Tables 5 and 6). GEP was compared to the individual prognostic factors comprising AJCC pathologic stage, including Breslow thickness, mitotic rate, and ulceration status, and was also compared directly to AJCC stage. AJCC stage was analyzed as known in the art as Stage IA, IB, IIA vs. IIB, IIC, III, IV for the complete validation set, and as Stage IA, IB, IIA vs. IIB, IIC for analysis that included only Stage I and II. Thus, analysis was performed with inclusion of all validation set cases (Table 5), or with inclusion of only stage I and stage II cases (Table 6). Cox univariate analysis resulted in significant correlation with metastasis for all factors except mitotic rate regardless of whether all validation set cases were included, or whether inclusion was restricted to stage I and II cases only. Comparison of GEP to Breslow thickness, ulceration, and mitotic rate showed that GEP and ulceration were significantly correlated to metastatic risk in a multivariate analysis (HR=8.4 and 2.6, p=0.04 and 0.03, respectively). Direct comparison to AJCC stage by multivariate analysis indicates that both factors are significantly correlated to metastasis (GEP HR=10.3, p<0.004; AJCC HR=6.4, p<0.0002). These results indicate that the 28-gene signature is able to predict metastasis independently of AJCC TNM staging and other prognostic factors.
An important clinical application of the GEP test is to identify patients who are staged as low risk (clinical stage I and II) for metastasis but who are at actual high risk for metastasis based upon the genomic signature of their CM. The Cox univariate and multivariate regression analysis was performed for those patients in the 107 sample validation cohort following prediction of metastatic risk (Table 6) who had clinical stage I and II CM only. Again, GEP was compared to the elements comprising AJCC pathologic stage: Breslow thickness, mitotic rate, and ulceration status, and directly to AJCC stage. Under univariate analysis, Breslow's thickness, ulceration status and GEP Class 2 were significant (p<0.0005). However, under multivariate analysis, only GEP Class 2 was significant (Hazard ratio=27; p=0.01). Univariate analyses for comparison of AJCC stage to GEP Class 2 showed both to be statistically significant, but the GEP Class 2 variable had a greater Hazard ratio than AJCC stage (48 vs 15, respectively). Multivariate analyses demonstrated that both factors were independent of each other (p<0.006) but the Hazard Ratio for the GEP Class 2 was again greater than that of AJCC stage (23 vs. 4.0). Together, these two result tables (Table 5 and 6) indicate that the 28-gene signature predicts metastasis independently of AJCC TNM staging and other prognostic factors in patients of all stage (I, II, III and IV).
Of the four growth patterns described for CM tumors, superficial spreading is the most prevalent, and biologically distinct from nodular melanomas.13 Radial basis machine analysis of the superficial spreading tumors within the cohort (n=128) resulted in a highly accurate classification of metastatic and non-metastatic tumors (
Each embodiment herein described may be combined with any other embodiment or embodiments unless clearly indicated to the contrary. In particular, any feature or embodiment indicated as being preferred or advantageous may be combined with any other feature or features or embodiment or embodiments indicated as being preferred or advantageous, unless clearly indicated to the contrary.
While a bimodal (low risk Class 1 vs high risk Class 2) may be the preferred clinical reporting output, additional analysis suggest that the tri-modal or linear approach may be clinically appropriate. Analysis using a trimodal approach demonstrated a stratification outcome related to metastatic risk. Specifically, development of a Class A (low risk), Class B (medium risk) and Class C (high risk) trimodal output shows a graduated increase in metastatic risk (
Probability of the metastatic risk prediction is based upon the coefficients of the 28 variables (genes) and reported as a value between 0 and 1. For the two-class analysis, cases with a probability score between 0 and 0.5 are designated low risk, class 1 cases. Conversely, cases with a probability score between 0.5 and 1.0 are classified as class 2, high risk. The linear output of the probability score can also be used directly to assess the risk associated with the genetic signature of particular CM tumor. As shown in
All references cited in this application are expressly incorporated by reference herein.
This application claims the benefit of U.S. Provisional Application No. 61/783,755, filed Mar. 14, 2013, the disclosure of which is explicitly incorporated by reference herein in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 61783755 | Mar 2013 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 14193378 | Feb 2014 | US |
| Child | 15075133 | US |
