Molecular Classification Assay Aims to Better Identify Primary Sites of NET Tumors

Andrew E. Hendifar, MD

Diagnosing the primary tumor in a patient with a metastatic neuroendocrine tumor (NET) continues to be an unmet need in the treatment landscape, according to Andrew E. Hendifar, MD. However, using a clinically validated 92-gene assay can assist practitioners in determining the origin of disease and tailoring treatments.

A 92-gene molecular cancer classifier was investigated in 75 NET samples, 44 of which were metastatic and 31 were primary. These included gastrointestinal NETs (n = 12), pulmonary (n = 22), Merkel cell (10), pancreatic (n = 10), pheochromocytoma (n = 10), and medullary thyroid carcinoma (n = 11). The assay was designed to predict tumor type or subtype based upon relative quantitative polymerase chain reaction expression measurements for 87 tumor-related genes and 5 reference genes.

In a validation study, the 92-gene classifier demonstrated a 99% accuracy for classification of neuroendocrine carcinomas (95% CI, 0.93—0.99). In addition, it correctly subtyped the tumor site of origin in 95% of cases (95% CI, 0.87–0.98). Researchers concluded that this assay be used as a tool for physicians in categorizing a patient’s NET origin.

“It is always nice when there are tools for neuroendocrine tumors, which is kind of a rare disease,” said Hendifar, medical oncology lead for the Gastrointestinal Disease Research Group and co-director of Pancreas Oncology at Cedars-Sinai Medical Center. In an interview withTargetedOncology^TMduring the 10th Annual NANETS Symposium, Hendifar discussed clinical experience with a 92-gene assay and the importance of molecular classification in the NETs landscape.

Targeted Oncology^TM: Can you provide some background to the 92-gene assay explored in patients with NETs?

Hendifar: This 92-gene assay is a molecular classification of neuroendocrine tumors. It originally had been validated in carcinomas of unknown primary, and a big subgroup of these patients had neuroendocrine tumors. The original validation set looked at over 2,000 tumors and sequenced over 22,000 genes to create this reference set of these 92 genes.

This 92 gene-expression analysis utilizes reverse transcription polymerase chain reaction, and compares a patient's gene expression to the reference database to match them up with the specific tumor diagnosis. What we did is, we looked at an Institutional Review Board-approved database of 24,000 patients who previously tested for carcinoid of an unknown primary, and we've looked at 1500 patients with NETs who were diagnosed. We're trying to understand and better characterize NETs that have a diagnosis of unknown primary to better understand how molecular classification can help their care.

It is interesting that the majority of NETs are metastatic on diagnosis and, even after exhaustive diagnostic workup, many of them don’t know what their primary [tumor] is from. It is an unmet need in the care of these patients, and knowing the anatomic location or subtype can be very important.

There is a recent approval for (avelumab [Bavencio]) in patients with Merkel cell carcinoma. There are specific treatments for medullary thyroid cancer, and many of the [drugs] that are approved are only approved in specific subtypes, so it is important information.

What information do we have on any of the genes included in this assay?

They are signal transduction and different types of gene expression, but they are not specific markers for a specific tumor like we are used to. It is basically an algorithm designed where the 92-gene expression is matched up between the different patients.

Is this something that could be implemented in multiple practices and used daily?

It is something that can be utilized in the community to help the pathologist and the practicing medical oncologist to better determine how to help their patients. Specific examples would include patients who have poorly differentiated tumors, where imaging doesn’t reveal what their anatomic subtype is. Could it be Merkel cell carcinoma? Could it be a high-grade NET of the lung? Are there differences? It is those types of situations. It can also add to the current understanding of a particular NET. If you have a well-differentiated NET but are unable to find the primary site, this might help give you an idea of where the primary site is and you can tailor your treatments accordingly.

What biomarker research is being conducted in the field?

Regarding the biomarker research for NETs, the most interesting and recent finding is that there are germline alterations that are associated with the identification of NETs, especially of the pancreas. We do have a lot of work to go in that field and identify appropriate genetic predisposition syndromes that lead to NETs, because we have a feeling that there is a lot of familial intraconal cancer. [We are] also identifying other biomarkers to help predict response and therapeutic effects. We are working on those.

Where would you like to see this assay go even further?

There has been a recent increasing interest in NETs. A lot of community oncologists and expert centers are becoming more familiar with treatment paradigms. As our understanding of this disease increases, the ability to subtype is more important. Right now, we have an assay that has been validated so we're grateful to be able to present the data here and some people could use it, and that is where we are right now. There is an extra tool available in case someone can use it that can help in the care of these patients.

Looking at the NETs landscape, what can you say about where we've been and where we're headed in the future?

The change in the treatment of NETs has been quite substantial over the last 5 years, and [we are] very fortunate to ride the wave, or jump on the backs, of a lot of hardworking physicians who have done a lot for this field. However, there has been an immense number of breakthroughs over the last several years and new treatments [have been] developed.

There are all types of treatments, from peptide-receptor radionuclide therapy, to targeted treatments, to new somatostatin analog treatments. We have a lot of things that we can do, and a lot of ways to help patients and, right now, it seems like we are heading towards this area of putting this information together to better understand how to sequence treatments. When does a patient need to see a multidisciplinary tumor team? When is the optimal [time] of surgery? Sequencing and synthesis of care, and putting these great new discoveries together in a certain treatment paradigm, is what we're going to be working on in the next few years.

It can take 5 to 7 years for a NET diagnosis, and community physicians may misdiagnosis a lung NET for non—small cell lung cancer. What steps can we take to better diagnose these patients?

That is a great point. It is very difficult to treat these patients appropriately and correctly until we have a firm understanding of what they have, and then classification [can] connect a certain diagnosis with a certain treatment plan. At NET Centers of Excellence, we already have an idea of how to do this. However, we haven’t been as successful at transmitting it to the rest of the oncology community. That is part of where the role of this unknown tumor assay comes in; we have to work on classifying these tumors accurately. Once that happens, and once we can communicate how to treat those subtypes appropriately, we can definitely improve the care of patients with NETs everywhere.

What are the key points community physicians should know about this assay?

The only other thing that is interesting for me about the assay is that it has been validated in NETs. There was a judication process where 75 NET samples were given to Centers of Excellence at Massachusetts General Hospital; University of California, Los Angeles; and Mayo Clinic. Two pathologists reviewed a specimen, gave it the diagnosis, and then 75 of the samples were sent to the CancerTYPE ID, and then the results were compared. The assay had a very high sensitivity and specificity rate—[it was] more than 95% for all of the 7 subtypes. That was very encouraging.

Reference:

1. Kerr SE, Schnabel CA, Sullivan PS, et al. A 92-gene cancer classifier predicts the site of origin for neuroendocrine tumors.Mod Pathol. 2014;27:44-54. doi:10.1038/modpathol.2013.105.