NEW YORK (GenomeWeb) – Castle Biosciences and collaborating researchers from several academic centers have published new data demonstrating the validity of the company's gene expression test, which predicts the risk of metastasis in patients with stage I or II cutaneous melanoma.
Castle's DecisionDx-Melanoma uses RT-PCR on Thermo Fisher Scientific's ABI HT7900 platform to measure the expression of 31 genes in total — 28 disease-associated genes and three controls — in RNA isolated from melanoma tissue samples. It differentiates melanoma patients into two prognostic classes. Those with a class 1 result have a very low risk of metastasis, while those with a class 2 cancer have a significantly higher risk of recurrence within five years.
In the study, published this week in Clinical Cancer Research, the researchers described the development of the gene expression signature at the heart of DecisionDx-Melanoma, and reported the results of their validation of the risk predictor in more than 250 samples.
According to the authors, the test was able to distinguish between primary melanoma samples from patients who later metastasized and those who did not with an area under the receiver operating curve of more than 0.9.
Launched earlier this year, DecisionDx-Melanoma reflects a larger shift for Castle Biosciences away from tests for niche or very low-prevalence cancers — like the uveal, or eye form of melanoma that has dominated its business thus far — and toward more prevalent diseases, including cutaneous melanoma.
According to Castle President and CEO Derek Maetzold, unlike uveal melanoma, which strikes only as many as 2,000 people each year in the US, stage I and II skin melanomas — those that have not yet measurably metastasized — occur in 65,000 to 70,000 patients per year.
Pedram Gerami a Northwestern University School of Medicine researcher and the first author of the new study, told GenomeWeb that several institutions, including Northwestern, have adopted Castle's test, although "many others, rightfully, have waited to see the published data."
"I think this is going to be impactful in [getting] other institutions involved," he said.
While many stage I and II melanoma patients do not recur, these tumors make up such a large proportion of diagnosed melanomas that the absolute number of deaths among patients with low-stage tumors actually outpaces deaths among patients with more serious stage III or IV cancers, Gerami added.
As a result, there exists significant room for improvement in predicting risk of metastasis for patients with otherwise localized disease. By better stratifying these patients according to their molecular metastasis risk, Castle hopes DecisionDx-Melanoma can help refer at-risk patients to intensive monitoring or adjuvant drug trials that could potentially improve their disease outcomes.
According to Gerami, in the group's study Castle's assay clearly demonstrated added value over current clinical staging tools. While the test did not identify every stage I or II melanoma that later recurred, and mistakenly marked some cancers that did not recur as high risk, it added significantly to metastasis prediction among patients without other clinical indicators of high-risk disease.
"We have the opportunity with this to identify patients that are classified as low-risk by staging, but actually have a high-risk of recurrence. This test has the potential to [pick out] those people and maybe give them the opportunity to see some of the treatments and go through the same protocols as those with aggressive disease based on traditional staging methods," Gerami said.
In the study, the researchers collected formalin-fixed paraffin-embedded specimens from either biopsies or full excisions of primary cutaneous melanoma. In total, the team amassed 268 stage 0-IV melanoma samples, including 107 tumors used in the initial development of the 28-gene expression panel.
Within the full cohort, 102 patients eventually developed metastatic disease, while 166 had no evidence of metastasis after five years.
The researchers then divided the group into a training set and a validation set, first using 164 cases to train the gene-expression classifier and then applying the resulting prognostic expression profile to the remaining independent set of 104 samples.
"We ran a bunch of different statistical analyses on what pattern of mRNA loss or gain would be the best criteria to distinguish the high- and low-risk groups," Garami explained.
According to the study authors, the resulting classifier had an AUC of 0.91 in the final validation set. It classified 61 cases as low-risk and 43 as high-risk. Comparing patients' known outcomes against the DecisionDx-Melanoma test results, the researchers calculated that the five-year disease-free survival rate was 97 percent for the class 1 group and only 31 percent for the class 2 cases.
The researchers reported that the negative predictive value for DecisionDx-Melanoma was 93 percent and the positive predictive value was 72 percent overall. Importantly, statistical analysis also showed that the gene expression classifier was an independent predictor of metastatic risk compared with standard clinical melanoma staging systems.
Castle hopes that further stratifying low-stage melanoma cases into high- and low-risk groups, its test can increase the number of patients followed more closely for recurrence or recruited to adjuvant clinical trials. However, it is unknown whether this will necessarily result in improved outcomes.
Gerami said that tracking this should be an important future goal to further establish the clinical utility of the assay. Another open question is whether risk prediction could be improved even further by combining molecular and other risk factors like clinical stage into a united algorithm.
"There's [already] an opportunity with the cases we have now … to look at how the data [might] be integrated with clinical staging and combined to [potentially] improve risk assessment," he said.
Besides advancing its cutaneous melanoma risk test, Castle said last year that it also plans in the early part of 2015 to more broadly launch its prognostic assay for esophageal cancer.
That test is aimed at helping clinicians predict which patients are likely to respond to preoperative chemotherapy and radiation, and which could potentially benefit more from immediate surgery.