Biomarkers can aid in diagnosis, prognostication, and the treatment decision-making process for patients with neuroendocrine tumors (NETs). Although current-day biomarkers have some limitations, several up-and-coming biomarkers hold promise for potential future use. This article provides information on these biomarkers, briefly touching on some trials in progress that were presented at the recent 2026 ASCO Gastrointestinal (GI) Cancers Symposium.

Following these presentations, featured expert Thorvardur R. Halfdanarson, MD, was interviewed by Conference Reporter Editor-in-Chief Tom Iarocci, MD. Clinical perspectives from Dr Halfdanarson on these findings are presented here.

I still think that SSTRs are excellent functional biomarkers for gastroenteropancreatic NETs (GEP-NETs), as they are present in most small bowel and pancreatic tumors. SSTRs, typically assessed with SSTR positron emission tomography (PET) imaging, can help with diagnosis, staging, and treatment decisions, and they can, to some extent, predict the duration of response following radioligand therapy (RLT). However, they still do not tell us how to sequence our treatments. This is where guidelines really struggle because we do not have a lot of quality sequencing data in GEP-NETs. For patients with metastatic GEP-NETs with SSTR expression, initial SSTR-targeting therapy with either SSA therapy or RLT (depending on the tumor grade, extent of disease, and presence of symptoms) is a good option. In terms of biomarkers other than SSTRs to help with treatment decisions, we have a lot of work to do.

The Ki-67 proliferative index plays a role in the diagnosis and prognostication of NETs. However, the problem with Ki-67 is that you are assigning a tumor grade based on a small biopsy from maybe 1 of 100 metastases. Further, if you do a resection of primary pancreatic NETs (PanNETs) or small bowel NETs, you often find a very low grade of Ki-67 in the primary tumor, but the nodal or liver metastases may be higher grade. There is, therefore, substantial Ki-67 heterogeneity within individuals with GEP-NETs, which can be difficult to assess with a small biopsy.

I think that this is where a combination of imaging and machine learning could be helpful. In fact, we have data suggesting that a NETPET score, which is assigned to patients based on dual fluorodeoxyglucose and SSTR PET/computed tomography imaging, may be associated with overall survival and time to progression. As an example, a patient with both fluorodeoxyglucose- and SSTR-positive lesions on imaging is more likely to have a more aggressive disease course than a patient in whom only the SSTR PET is positive. We can also use radiomics data to predict the tumor grade based on magnetic resonance imaging and/or computed tomography scans, but further studies are needed to validate this approach. I think that these types of approaches can give us a more comprehensive understanding of tumor heterogeneity across the body, not just at the site we sampled.

A biomarker that shows promise is the PRRT prediction quotient (PPQ). This is a multifaceted analysis that looks at the expression of several NET-related genes to predict response to PRRT. The PPQ has been tested in relatively small cohort studies of RLT outcomes, and it seems to be a strong biomarker. It reads as binary—either positive or negative—and is predictive of the depth of radiosensitivity. So, I think that the PPQ is an exciting biomarker that could be easily applied to a clinical trial. Ideally, the PPQ should be studied within the context of a prospective randomized clinical trial of RLT for validation before being adopted into practice. One of the attractive features of the PPQ is the simplicity of ordering it, as it is a blood test.

MGMT is another interesting biomarker that has been shown to predict the efficacy of temozolomide-based therapy for NETs. There is no question that this biomarker is valuable for sorting out who is and is not likely to respond to temozolomide-based therapy; the problem, however, is that MGMT can produce false-negative or -positive results. Because of this, we would never deny a patient temozolomide-based therapy solely because they had an MGMT test that suggested a lack of benefit.

Another thing worth mentioning is that many patients with resected well-differentiated PanNETs have mutations in ATRX or DAXX. These mutations can be detected by testing for alternative lengthening of telomeres (ALT). Positive ALT correlates strongly with the presence of mutations in ATRX or DAXX, and positive ALT and mutations in ATRX or DAXX are strong predictors of recurrence following the resection of localized PanNETs, even after adjusting for other risks of recurrence. I would not be surprised if this enters the scene soon and becomes a clinically orderable test that can be used in routine practice. We need to continue to identify additional genomic alterations that put someone at an increased risk of developing a NET. We are just starting to tap into that by exploring germline and somatic genomic alterations.

Unfortunately, studies investigating front-runner isotopes used in RLT (ie, ¹⁷⁷Lu and ²²⁵Ac) do not include much biomarker exploration, including those discussed in Trials in Progress sessions presented at the 2026 ASCO GI Cancers Symposium (posters TPS646 and TPS649). The COMPOSE trial, which is investigating ¹⁷⁷Lu-edotreotide as first- or second-line therapy for patients with advanced, well-differentiated, SSTR-positive GEP-NETs, will likely have a biomarker component, although its primary end point is progression-free survival. However, we do not know the details of that yet. Having reliable biomarkers that predict the efficacy and toxicity of RLT would be extremely valuable in the clinic when making decisions on treatment sequencing.

In the end, all these different biomarkers need to be included in prospective studies for validation; current data are frequently based on single-center studies with very heterogeneous populations, different types of testing, and/or different therapies. I wish we had a biomarker that we could follow to make early treatment decisions, but I think that we are not there yet.