Outlining Precision Prognosis and Therapy Tools Across the NET Spectrum

In a recent interview with CancerNetwork®, Jason Starr, DO, an oncologist of the Early-Onset Colorectal Cancer Group in the Department of Internal Medicine at Mayo Clinic Comprehensive Cancer Center, outlined a multidimensional framework for the management of neuroendocrine tumors (NETs), emphasizing that while small bowel NETs often lack obvious genetic drivers, pancreatic and high-grade neuroendocrine carcinomas require distinct genomic consideration.¹ A central pillar of the discussion was the transformative role of theranostics, particularly the exploitation of somatostatin receptors for both diagnostic imaging and targeted radioligand therapy.

Starr addressed the complexities of incorporating liquid biopsies into clinical practice, noting that while circulating tumor DNA (ctDNA) offers limited utility in well-differentiated disease, it serves as a valuable tool for monitoring molecular residual disease (MRD) in high-grade carcinomas. He further highlighted the critical role of multidisciplinary molecular tumor boards in distinguishing between targetable mutations and true oncogenic drivers, a distinction underscored by varying clinical responses to RET and ALK fusions.

Key safety discussions focused on mitigating adverse effects (AEs), including treatment-related myeloid neoplasms and hepatotoxicity. Starr shared ongoing research of clonal hematopoiesis of indeterminate potential (CHIP) as a predictive tool to refine patient selection for peptide receptor radionuclide therapy (PRRT). These insights may underscore a new frontier in the neuroendocrine landscape, where molecular signatures and longitudinal monitoring aim to extend the oncologic runway for patients.

CancerNetwork: What are some highlights and key takeaways from your “Decoding Neuroendocrine Tumors” session?

Starr: When you approach neuroendocrine cancers, it’s nuanced, and you [need] to have a general framework to work with. What I focused on within the discussion was the site that the disease starts, the stage of the disease, [and] genomics. Genomics don’t play as big a role with the more common NETs that we see, like the small bowel NETs. However, for pancreatic NETs, they have significance. Similarly, for high-grade neuroendocrine cancers, doing genomic profiling can be impactful for patients.

One of the focuses of the discussion was on theranostics. Theranostics is this field of medicine where you’re able to exploit some aspect of the cell, usually an antigen or a protein on the cell surface, where you can design both diagnostic imaging to target that and illuminate the area. Then you could also direct therapies against that same protein or antigen. Within neuroendocrine cancers, there’s a receptor that’s [quite] well known, the somatostatin receptor, and we exploit that receptor with therapies [and] diagnostics. There’s a lot of focus on that within the talk.

Lastly, the one thing that I touched on was blood-based biomarkers, and I caution the audience on chromogranin A [CgA] not being a very good biomarker. The sensitivity is low for that test. Then, there are some evolving biomarkers that we talked about, one called a NET test, which is a genomic signature that looks at RNA. Then we also talked about tests that can help predict benefit to [lutetium Lu 177 dotatate (Lutetium]], which is along that theme of theranostics.

How can a multi-disciplinary team best integrate longitudinal ctDNA monitoring to identify temporal heterogeneity, such as fluctuating microsatellite instability (MSI) or HER2 status, and at what point should these liquid findings prompt a surgical or medical shift in the treatment plan?

ctDNA in neuroendocrine cancers is not a mainstream approach, and the reason for that is, in small bowel NETs, for example, we usually don’t find many genetic abnormalities on the liquid biopsy. It’ll sometimes come back within that panel and you won’t find any mutations in those genes. There’s a limited utility of ctDNA for well-differentiated NETs.

That’s different for high grade neuroendocrine cancers, like grade 3 NETs, or more specifically, neuroendocrine carcinoma. [In a] study I was a part of, we looked at [approximately] 320 samples from patients with neuroendocrine cancers. The majority of the time when we got a signal it was a neuroendocrine carcinoma. There is utility with neuroendocrine carcinoma of using a blood-based assay like ctDNA.

The problem is there [are] not many targets. We don’t often find a target even within neuroendocrine carcinoma. Alternatively, MRD, that’s within the same verbiage of ctDNA. MRD can follow the patient’s disease at a molecular level within the blood. You can follow it in terms of [having] a certain starting point. You start a therapy, it goes down. You follow that longitudinally, and if it starts going up, it may suggest that progression of disease is on the horizon, or you may be seeing that on the scans.

The problem with that is, even in other diseases with ctDNA, it’s really prognostic and not necessarily predictive. That test isn’t going to predict that a therapy is going to work or not. It just tells you that it is working, if it’s not working, or if the patient’s [experiencing] progression. We’re still learning how to incorporate ctDNA technology for neuroendocrine cancers. I would stress that, for high grade neuroendocrine carcinoma or even grade 3 NETs, including small bowel or [pancreatic disease], it makes sense because those are maybe more genomically enriched patients in terms of the DNA mutations.

With your experience in managing rare genomic alterations such as RET fusions in pancreatic neuroendocrine carcinoma, how should we approach the primary resistance often seen with targeted inhibitors, and what is the role of the multi-disciplinary molecular tumor board in distinguishing between a targetable mutation and a truly actionable driver?

We have reported a case report—and I alluded to it during the lecture—of a patient with an atypical lung carcinoid. Within the spectrum of neuroendocrine cancer, and we found an EML4-ALK fusion, which is targetable. And we were able to get this patient medications that target this fusion and block this pathway. We went through the first, second, and third generations of those medications, and the patient survived over 3 years, which is incredible because we had made a referral to hospice at the time of the testing. By the time the testing came back, we disenrolled the patient in hospice and got [them] on the medication, and then [they] lived for 3 years [beyond that point].

As far as the RET fusion, that’s a case report we also published. We found a RET fusion in a patient with a pancreatic neuroendocrine carcinoma, and we got really excited, and we started the directed therapy, the targeted therapy, and unfortunately, we did not see any benefit from the therapy, which is [the converse outcome] to the patient I described just now.

The question—and what we reported in the publication—was: one, as you alluded to, is it a driver? Was this a driver mutation? And are there resistance mechanisms that don’t allow the drug to work? We hypothesized a few of those possibilities: that it wasn’t a driver, and that there was some intrinsic resistance within the cells, because not all fusions are created equal. It depends on the fusion partner. That case report highlighted the fact that, just because you find a target, it doesn’t mean that the targeted therapy is always going to work, which was disappointing for this patient.

As far as the molecular tumor board, [it is] very helpful. In the era of AI, sometimes you can do some more database searches and find [actionable drivers] quicker, because what you want to know is, is the fusion an activating fusion? Are you getting the downstream effect as a result of the genomic fusion? AI could help in that regard.

At Mayo Clinic, we do have a genomic tumor board. We have multi-disciplinary tumor boards where we can discuss this. It’s absolutely true that you can have multiple drivers. You can have what you think is a driver mutation, and it’s actually not. There’s a lot of nuance to interpreting reports and using that data in terms of whether it’s tangible and it’s going to make a difference for the patient.

How should the hematologist and nuclear medicine specialist collaborate to screen high-risk patients with GEP-NETs, and are there specific molecular signatures in the bone marrow or blood that should lead us to delay or dose-modify PRRT to prevent therapy-related myeloid neoplasms?

One of the big risks with PRRT or radioligand therapy is treatment-related myeloid neoplasm. It’s a real risk. I’ve seen it. We quote it to be around 2% to 3% in the real world, [but] it may be a little higher. We at Mayo Clinic [are] doing research where we’re trying to determine whether something called [CHIP], which is a blood test, can help predict whether patients are going to have low blood counts after the treatment. Or [are] there any signatures on the CHIP that could help us predict patients who will develop treatment-related myeloid neoplasm so that research is ongoing, because the question is, if you knew that a patient was at high risk for developing a [treatment-related myeloid neoplasm], would you recommend something first or would you not recommend PRRT. And I alluded to one of the blood-based tests, which is the RNA sequencing, and they developed the same group that did the NET test, something called a PPQ—essentially it’s another algorithm that predicts benefit from PRRT. If the PPQ was positive, those patients would almost certainly benefit from PRRT.

The question is, could you do that test at baseline? If the test was negative and it showed that the patient may not benefit, or didn’t appear to benefit from PRRT, maybe you would go with another therapy so that you wouldn’t expose the patient to the risk of developing a [treatment-related myeloid neoplasm]. One is risk stratification in terms of who’s at risk for developing [treatment-related myeloid neoplasm]. The other one is, how could you tailor your approach based on whether or not the patient’s going to benefit from the therapy, or you have a prognostic or predictive tool, that can tell you whether they’re going to benefit from the therapy?

Given the concerns regarding hepatotoxicity, how can pretreatment metabolic tumor volume and biliary health markers be used to better sequence these therapies in a way that preserves the patient’s long term ‘oncologic runway’?

One of the concerns we had with [lutetium Lu 177 dotatate] PRRT is whether you can develop hepatotoxicity based on prior treatments. For example, patients will not infrequently get a treatment called yttrium-90 [Y-90] radioembolization to the liver, [but] also the volume of disease in the liver. If the PRRT and all that therapy is going to the liver, then [we are] delivering the radiation, is the liver at risk for developing injury? The study that we published showed that even patients with extremely heavy volume of disease in the liver did not appear to develop liver failure as a result of the PRRT.² That was encouraging. It was a small sample size of patients, but the largest experience reported to date.

Is there anything that else that you would like to discuss or emphasize that we might not have touched upon already?

Theranostics have been a huge paradigm shift in neuroendocrine cancers since [lutetium Lu 177 dotatate was] FDA-approved in January of 2018. We have seen significant improvements in patients’ quality of life with this disease, we’ve seen improvements in outcomes, [and] we’ve seen better disease control. It was a big inflection point in terms of a platform also to build on.

Now we’re working on newer iterations of PRRT or radioligand therapy, and what I hope is that we’re ushering in a new frontier of these new therapies that are going to be better than the first iteration. That’s something that I’m excited about for patients, but there’s a lot of work to be done to understand these diseases more at a cellular biology level, because [for] small bowel NETs, there’s no obvious genetic driver. It’s likely that these cancers are forming based on post-transcriptional, post-translational, microenvironment changes. It’s going to require some more work to be done to understand the small bowel NETs, because that’s the majority of patients I see to understand more at a cellular level. Cancer is a genetic disease, but clearly whatever is causing the cancer is happening outside of an obvious mutation in the DNA.

References

1. Starr J. Decoding neuroendocrine tumors: precision tools for personalized prognosis and therapy. Presented at the 3rd Biennial Miami Precision Medicine Conference; April 11-12, 2026; Fort Lauderdale, FL.
2. Gococo-Benore DA, Kuhlman J, Parent EE, et al. Evaluation of hepatotoxicity from peptide receptor radionuclide therapy in patients with gastroenteropancreatic neuroendocrine tumors and a very high liver tumor burden. J Nucl Med. 2023;64(6):880-884. doi: 10.2967/jnumed.122.264533