top of page

Practical guidance from an oncologist in finding TRK and other rare mutations in gastrointestinal cancer patients


Presented by Michael Pishvaian

This event is supported by an educational grant from Bayer

About the speaker

Dr. Pishvaian is an associate professor, gastrointestinal oncologist, and co-director for clinical research at the Center for pancreatic cancer research at the University of Texas, MD Anderson Cancer Center. He is a MD and PhD in tumor biology and on both degrees at Georgetown in 2001. He stayed at Georgetown after graduation, completing his medical residency in 2004 and had fellowship in hematology and oncology in 2007. He served on the faculty at Georgetown until moving to MD Anderson in 2019.Dr Pishvaian is a translational oncologist and he focused on providing novel therapy for patients, particularly in the areas of pancreatic cancer and refractory colorectal cancer. His work has been focused on the area of precision medicine, at a special focus on therapy targeted towards homologous recombination DNA repair deficient tumors. And Dr. Pishvaian is a co-investigator on an NIH project to study mechanisms of resistance to the PARP inhibitor-based therapy.

Challenges in Treating GI Cancers

From a medical oncologist’s viewpoint, we know that advanced GI cancers are deadliest diseases as they account for the largest number of cancer-related deaths worldwide, with more than 3 million patients dying from advanced GI cancers each year.  We have made progress in many of the advanced GI cancers but some of the therapies particularly aimed towards more advanced patients. The three examples reside in colorectal cancer, pancreatic cancer, and then gastric cancer, where therapies demonstrated overall survival improvements of less than two months, leading to FDA approval though.  These nonetheless have made standard practice changing. We all learned in medical school that we should treat patients empirically- where patients are treated according to prospective randomized clinical trials that are aimed to test one therapy versus another standard therapy and it is hoped to demonstrate benefit with a newer therapy. With these parallel trial designs, well accepted as gold standard, unfortunately there are slow progress or and low threshold for success, particularly for GI cancers and diseases like pancreatic cancers. The so-call clinical significance refers to 30% response rate or three-month overall survival. The other difficulty with treating patients empirically is that while we treat all patients and have proven benefit in the patients as a group, the reality is that most patients don't benefit from the treatment that was proven to be better than the current standard. In a treatment that leads to 30% response rate, we may treat 10 patients but only three of them benefit. While all patients get the same treatment, we really don't know whom the three are that will benefit.

Folfirinox is a perfect example of this. It was not to diminish the value of this study because it was a practice-changing study. I myself use a lot of Folfirinox for patients with pancreatic cancer. However, in the pivotal study of Folfirinox versus Gemcitabine, the previous standard of care for metastatic pancreatic cancer, the response rate was only 32%, which meant that essentially 70% of patients did not show tumor shrinkage or did not respond to therapy. 1That's the reason why the promise of personalized or precision medicine comes into play where we can use predictive biomarkers to identify specific patients who will benefit from specific therapies.

Molecular Classification Versus Predictive Biomarker

1. Molecular classification

Using predictive biomarkers is not necessarily the same as using molecular classifications. The molecular classification, or molecular subtyping is to define large scale molecular characterization of a disease type usually by doing gene sequencing, often RNA sequencing, transcriptomics or sometimes even protein profiling, so that the schemata is designed to define specific molecular subtypes which often are pathways-driven, and to define prognostic groups. They might have some predictive relevance, but their goal is to define prognostic groups.

 Take pancreatic cancer as an example, there has been large-scale efforts focused on DNA sequencing, trying to identify pathways relevant to the different genetic mutations in pancreatic cancer. To summarize, four subtypes have been identified- the squamous subtype, the pancreatic progenitor classical subtype, the aberrantly differentiated endocrine/exocrine subtype, and the immunogenic subtype. In fact, these are not predictive classifications, they are only prognostic or pathway-driven subtypes.  The immunogenic subtype unfortunately, by no means, predict for response to immune-based therapy for patients with pancreatic cancer. As we know to date, there has not been a pathway proven to improve outcomes for patients with pancreatic cancer.

2. Predictive biomarker

With that background in context, precision medicine does have a different purpose- focus on using molecular testing for therapeutic purpose, with the hope that in any individual patient, there may be distinct targetable molecular abnormalities that can lead to a specific therapy. These are so called potent predictors of response. The other term that often gets thrown around is “actionable biomarker.” Now, actionable biomarker is a term that sometimes has a negative connotation.  it's important to define actionability and it is whether there is literature to support significant activities, including disease response, prolonged disease control and/or patient survival, etc. As we know, for diseases like pancreatic cancer, there has been no pancreatic cancer specific targeted therapy identified thus far.  When it comes to potential predictive biomarkers, we borrow the experience from other disease types. There is softer definition of actionability where there are implications of response to therapy based on mechanism or pathway.

While we have been using predictive biomarkers in medical oncology for more than 20 years, we just sometimes don't think of these as true predictive biomarkers. For example, ER/PR status in response to anti-estrogens in breast cancer is something that we use all the time. HER2 amplification in response to trastuzumab is a very commonly used predictive biomarker. The predictive biomarkers can be divided into molecular drivers that define a disease subtype and molecular drivers that divide disease into other subtypes.

There are relatively few predictive biomarkers that do define the disease.  This is a situation where the predictive biomarker exists as a mutation. That is the driving mutation that leads to the development of that cancer. As a result, targeted therapy often leads to a very dramatic response. The best example is the use of imatinib to target the c-KIT mutation for patients with gastrointestinal stromal tumors (GISTs).2 Prior to the early 2000s, in late 1990s, GIST was essentially just intestinal sarcoma. As it became clear that c-KIT mutations were present in the majority of these tumors and  a specific targeted therapy was defined for this c-KIT mutation, we switched from a previous standard of single agent doxorubicin that  didn't do much for this disease to a therapy that has dramatically improved outcomes and survival, and has made an impact on this disease in the past decades.2 This scenario of a defining mutation is still relatively uncommon.

Most of the time, talking about predictive biomarkers, we are looking at dividing a disease group such as lung cancer, breast cancer, colon cancer, into several different subgroups, based on the key molecular mutation in some of those patients. In those patients with molecular abnormality, targeted therapy can sometimes lead to dramatic benefit. This is the basis for many of the tests we send today. For example, in breast cancer, we start a discussion about a breast cancer patient as whether they have hormone receptor positive or negative disease, HER2-amplified or not amplified disease. Similarly, for gastric cancer, we are routinely defining patients as HER2-amplified versus not. Prior to the era of immunotherapy, melanoma patients were defined as BRAF mut versus not. And in colon cancer, the presence of a key mutation or pan-RAS mutation defines a lack of response to certain targeted therapy, anti-EGFR therapy.

Let’s look at how predictive biomarkers have divided a disease into several different subtypes, specifically in lung cancer. In 2002, there was a seminal paper presented by Dr. Schiller, studying which Platinum doublet benefited lung cancer patients the most. The study concluded that it did not matter which Platinum double was use as the median overall survival for all of the patients with advanced non-small cell lung cancer was eight months.3  We can further interpret that in the clinical setting of 2002, the median overall survival was 8 months, for 200,000 advanced lung cancer patients (or 140,000 patients per year).

Just a few years later, it was identified that EGFR activating mutations were present in about 10% of non-small cell lung cancer patients and this is a powerful predictor for response to anti-EGFR, small molecule inhibitors, such as erlotinib.4,5 Among patients who retreated with erlotinib, all the non-small cell lung cancer population did not benefit much from the drug, while the progression free survival of EGFR-mutated subgroup was tripled.4 One open-label phase 3 study showed the overall survival of this subgroup  is nearly 2 years.5 Similarly, ALK rearrangements were identified in non-small cell lung cancers.6,7 More recently, the patients with ROS1 rearrangements also responded robustly to appropriately targeted therapy demonstrating  survival that was, amazingly, measured in years, far greater than the previous eight months.

It is possible to subdivide NSCLCs into genetically discrete subsets on the basis of the activating mutations that they harbor, and each of these subsets may correspond to patient cohorts that are likely to benefit from treatment with specific inhibitors that target the products of these specific genetic lesions. 8 The pie chart (Figure 1) shows the distribution of various reported activating oncogenic mutations in a survey of 139 non-small-cell lung cancer (NSCLC)-derived cell lines. Also shown for all the activating mutations (except KRAS) are inhibitors that selectively target the activated oncoproteins, yielding growth inhibition and/or apoptosis of cancer cell lines with the corresponding mutated oncogenes. There are currently no inhibitors that target oncogenic KRAS.8 From clinical experience, if those patients receive the appropriate therapy for their predictive biomarker, their median overall survival is greater than three years. In contrast, for the other patient with KRAS for which we don't really have a targeted therapy yet, or for the non-mutated patients, the median overall survival remains only 8 to 10 months.


Figure 1 stratification of non-small-cell lung cancers on the basis of activating mutations


It has been a misconception that the mutations do not exist in GI cancers. It is not true though. The mutations do exist in every GI malignancy. Even in pancreatic cancer, about 25% of patients harbor some form of mutation.9,10 In biliary cancers, it's from 20 to 40%. In upper GI cancers, it can be up to 30% when we include HER2 amplification.11,12 In colon cancer, the actionable mutations alone that predict for response to therapy account for 10%. Meanwhile, the negative predictive marker of RAS/RAF occur in 70 to 80% of colon cancers that do benefit from predictive marker testing and therapeutic tailoring.13

Most of the data for pancreatic cancer specifically are still just anecdotal clinical data, small case series or small clinical trials. There are other proven or anecdotal data in other cancer subtypes, but they still meet definition of actionability.  There are also a handful of cases where there're pathway driven preclinical data to support the actionability. The pooled data from 8 papers incorporating around 5000 patients have shown that about 25% of patients have potentially actionable biomarkers. The majority have DNA damage repair deficiency, and some have BRAF mutations. There’s a whole slew of subtypes with <1% frequency for which specific target therapies are still developing.

Disproportional benefits in the presence or absence of predictive biomarker

It's because of data like this and also because pancreatic cancer is such a bad disease, that through 2018 and 2019, the NCCN panel put out a very forceful statement saying that first of all, germline testing is recommended for all patients with pancreatic cancer of any stage to look for hereditary cancer syndromes. In addition, tumor or somatic gene profiling is recommended for patients with locally advanced or metastatic disease (80% of them) who are candidates for anti-cancer therapy against these uncommon but actionable mutations.14

KRAS mutation is a negative predicted marker in colorectal cancer.15,16,17,18 The panitumumab and cetuximab were FDA-approved therapies in the 2000s because they did improve outcomes for patients with metastatic colorectal cancer. When we separated out KRAS mutated versus KRAS wild-type of metastatic colorectal cancer, we found a very dramatic separation of treatment responses. The patients with colorectal cancer who received single agent panitumumab did not show better than receiving best supportive care if their tumor was KRAS-mutated with the response rate of a flattened zero percent. But in the KRAS wild-type patients, the response rate approached 20% with receiving the anti-EGFR therapy alone. 15Due to these facts, the guidelines suggested patients with KRAS, NRAS, probably less common HRAS mutations, be not receive any anti-EGFR therapy16.

In terms of smaller sub-categories of predictive biomarkers, though shown in much fewer patients, the benefits sometimes can be disproportionately high. The best example is probably within the MSI-high disease. Dung TL concluded that MSI-high tumors across the spectrum benefited from immune checkpoint inhibitors such as pembrolizumab, nivolumab. These MSI-high tumors do exist in virtually every cancer type, including GI cancers, for instance, advanced colorectal cancer (3-5%), gastric cancers (20-22%), and pancreatic cancers (<1%).19 If the patients have MSI-high tumors, almost all of them do benefit to some degree with an immune checkpoint inhibitor.

The BRAF600-mutated tumors tell a similar story.  The BRAF mutations used to play a role in melanoma. Rather in colorectal cancer, the triple drug therapy-encorafenib, cetuximab, binimetinib, can significantly improve outcomes for patients with BRAF600 mutations, particularly. The Phase1b safety study showed that most patients had some degree of tumor shrinkage with a response rate of about 50% per RECIST criteria.20

A couple of other examples are illustrated showing disproportionate benefit. The HER2 amplification occurs in about 5% of all colorectal cancers. In a phase 2 study investigating dual HER2-targeted therapy with trastuzumab and lapatinib in HER2-amplified colorectal cancer, eight of 27 patients (30%) had achieved an objective response.  with a response rate of 30%. 21

The FGFR-mutated   cholangiocarcinoma has really been identified as a key subgroup. The patients usually have either FGFR fusions or amplifications, not so many mutations. The fusions and amplifications can predict response to FGFR inhibitors in patients with advanced colorectal carcinoma.22.

The largest subgroup of biomarkers in patients with pancreatic cancer is those with DNA damage repair and specifically homologous recombination DNA damage response and repair (HR-DDR) mutations. The most common one is the BRCA mutation. The ATM is probably the statistically most common DDR mutation that exists in pancreatic cancer. And then there's a whole slew of smaller subgroups of DDR mutations that can occur as well. Both in Know Your Tumor dataset, and a similar Caris database review, as well as multiple other papers, we've seen that about 17-25% of pancreatic adenocarcinoma harbor mutations in these genes.23 These mutations can have powerful predictive benefits, even for basic chemotherapy.  Dr Pishvaian’s group identified that HR-DDR-mediated pancreatic cancers benefit disproportionately from platinum-based therapy and that HR-DDR mutations alone do not have a specific prognostic value with absence of platinum-based therapy.24 They had the opportunity to look at the subset of Platinum-naive patients. These patients had advanced pancreatic cancer but were never in the course of platinum regimen. Instead, most of them, for example, started being treated with paclitaxel. When they separated out HR-DDR status, there was no survival difference between mutated and non-mutated patients.  There was no real prognostic value to the presence of this mutation alone (i.e. in the absence of platinum-based therapy). The median overall survival of the HR-DDR mutated patients who did not receive Platinum was less than one year.24 The patients with HR-DDR mutations who received platinum therapy have improved survival, and in particular, survival is greatly improved for those with advanced disease, from 1.45 (non-mutated) to 2.37 (mutated) years.24

The other class of targeted therapies that are promising for patients with pancreatic cancer are the PARP inhibitors in the presence of an HR-DDR mutation, particularly BRCA1/2 mutation. There have been a number of anecdotal cases or case series that have shown recurrently that about 20% of pancreatic cancer patients whose tumors harbor BRCA1/2 mutations benefited from single agent PARP inhibitor. Rachna Shroff’s Rucaparib study showed that not only the germline BRCA mutated patients, but at least one of somatic BRCA mutated patients had a complete response to PARP inhibitor-based therapy.25

Eileen O'Reilly has shown in a phase 1b study that cisplatin, gemcitabine, and PARP inhibitor, veliparib, can be synergistic. The response rate in the BRCA positive cohort was 77.8% with median overall survival of 23.3 months (versus 11 months in BRCA- patients).26 We are all looking forward to the results of randomized phase 2 portion of this study.

Most recently at ASCO, and also at ESMO, World GI, there was one presentation for the  phase 3 randomized study of olaparib as maintenance therapy for patients with germline BRCA 1/2 mutations, and the primary endpoint for this study was progression-free survival. And this was, by that definition, a successful study where patients who had a germline BRCA mutation, benefited from maintenance therapy with olaparib versus placebo.27 There is a hope that this will lead to the approval of the first PARP inhibitor for BRCA-mutated pancreatic cancer.

NTRK fusions in GI cancers

We have seen several examples demonstrating disproportionate benefit in GI cancers by using specific predictive biomarkers and targeted therapy. The NTRK mutations have been identified across diseases, including GIs cancer as well. In particular, NTRK fusions work as predictive biomarkers for Trk inhibitor-based therapy. The Trk family is a typical receptor tyrosine kinase known as neurotrophic receptor tyrosine kinases, TrkA/B/C, encoded by the three genes, NTRK1/2/3, respectively. From biochemistry perspective, these are typical receptor tyrosine kinase   that dimerize upon ligand activation, leading to activation of downstream pathways, including the PLCγ pathway,  MAPK pathway (RAS, RAF), PI3K pathway (AKT).28

The reason these fusions can be very potent is that these gene fusions will link the activating domain of the genes, the tyrosine kinase domains, either with or without the transmembrane domain. They'll link that to an extracellular typically dimerization domain that leads to a spontaneous dynamization of the of this fusion protein and thus constitute activation in a lagging independent fashion of the Trk kinase leading to cellular activation of the pathways.  This makes an autonomous way to activate a cell to drive rapid growth, invasion, and metastasis.  The multiple partners have been identified. Some of them come with known mechanisms for dimerization while some less known. The gene fusion events can occur spontaneously such that this can sometimes be a very powerful way to drive cancer cell growth.29

These NTRK fusions have been found in all kinds of cancers throughout the body. The certain NTRK fusions do define a couple of diseases including the MASC, as well as the secretary breast carcinomas. These are diseases in which NTRK fusions are found in the majority of patients, meaning NTRKs define these diseases. Not uncommonly, about 5- 25% of patients whose tumors were c-KIT negative, the pan-negative gastrointestinal stromal tumors, other less common tumors, and thyroid cancers can harbor these NTRK fusions. On the other hand, many other cancer types can harbor NTRK fusions at the less than 5% and often less than 1% range.29

Typically, less than 5%, but nevertheless a recurrent proportion of these patients, do exhibit NTRK fusions.23,28,30 NTRKs occur in up to 4% of metastatic colorectal cancer.28 The observation is probably derived from a biased subset because NTRK fusions are more prevalent in MSI-high tumors. When we look at a larger subset, it's probably about 1% of all colorectal cancers, about 4% of cholangiocarcinoma28, 2% of appendiceal cancers28, about 0.7% of pancreatic cancers23, 0.3% in neuroendocrine tumors.30   

Given the relatively uncommon finding of NTRK fusions across cancer types, it certainly would be difficult to screen every patient all day long for the presence of the NTRK fusions and there is a pan-Trk immunohistochemisty that can be done to  identify the subgroup of patients who might be worthy of undergoing deeper testing with next generation sequencing. This pan-Trk antibody is very sensitive while the specificity might be a little poor. But it can narrow down this subgroup of patients and concentrate the subgroup of patients that are more likely to harbor and NTRK fusion. Those patients who are pan-Trk IHC positive can then go on to have genetic testing or fusion testing to identify the presence of the fusion gene. 

The Caris group Gatalica looked at over 4000 patients and identified that the pan-Trk IHC was 75% sensitive and 95% specific and we saw what NTRK looked like in several of the specimens with the different fusion partners in different disease types.31 The ESMO guideline review came out recently. It was found that the pan-Trk IHC had about 95-100 % sensitivity and 93-100% specificity.32 This is not meant to define patients who are appropriate for therapy, but to be a good screening tool for subsequent gene fusion testing.

The approved Trk inhibitors- larotrectinib and entrectinib

It is critical to virtually test every patient with advanced cancer. In MD Anderson, we had approval for testing every patient with advanced cancer looking for the presence of an NTRK fusion. There was no way to predict which patients/disease types are going to harbor NTRK fusion based on clinical scenario or even standard pathology until we screened every patient. As disproportionate benefit was also shown in patients whose tumors harbored NTRK fusions and who received Trk inhibitor-based therapy. There are two FDA-approved Trk inhibitors, larotrectinib and entrectinib. Both of them have shown very significant benefits for patients who were treated.33,34 In the larotrectinib trial, the majority of patients had some degree of response, and many of them achieved a complete response.  Most of GI patients across different cancer types in this study except two patients, had responded to therapy.33

The other drug, entrectinib, targets ROS1/ALK on top of NTRK. Although many of the data were mixed into NTRK fusion datasets, we can see the patients who have these fusions sometimes have dramatic benefit, numerous responses to Trk inhibitor-based therapy.34

Dr Pishvaian had three pancreatic cancer patients treated with entrectinib, two of whom harbored NTRK fusions and one harbored ROS1 fusion. These three patients did benefit significantly from Trk inhibitor-based therapy and they had prolonged and occasionally dramatic benefit. One patient, who was virtually on the way to go to hospice as a result of very extensive metastatic pancreatic cancer, started on the entrectinib and within a month the patient looked like a normal patient with a PET scan that exhibited dramatic response to therapy. This patient was on therapy for more than a year, probably longer just because he was lost for follow-up in the study.

According to Dr. Pishvaian, one lesson learned from Trk inhibitor trials was that NTRK amplifications or mutations were not activated. In other words, the patients whose tumors harbored only amplifications or mutations should not be treated with these drugs, just because they would less likely benefit the patients. For example, in the data that have been presented thus far for entrectinib, four out of four NTRK fusion patients benefited from therapy. But for the other 25 NTRK abnormalities either amplifications or mutations, none of the patients benefited. Similarly, for larotrectinib, five out of six patients responded. But then the study had the opportunity to enroll five patients with other NTRK amplifications or mutations, and none of them benefited from the treatment. In short, the Trk inhibitors will benefit NTRK fusions only, but not NTRK amplifications or mutations.

These are drugs that are well tolerated. Take entrectinib as an example, there was a low percent of Grade 3 events and essentially no Grade 4 events that could be clearly attributable to the therapy. 34There was one patient on the STARTRK study ever having an eosinophilic myocarditis that was reported as “possibly related (may or may not be related).34” Most of these patients experienced some degree of fatigue. They also complained taste changes and other GI toxicities as well as arthralgias.  Most of the side effects, including the arthralgias, disappeared as long as the patient stayed on the drugs. If they can bear through it and usually for one or two months, the side effects became tolerable or gradually gone. The lantrectinib has a similar safety profile33.

The resistance to Trk inhibitors and next-generation therapeutics

The Trk inhibitors have been working for NTRK fusion patients, they do lead to disease response and disease control sometimes for very meaningfully long period of time. Unfortunately, like all cancer therapies that are not technically curative, resistance will eventually be developed. Many of the resistance mechanisms have been identified. A preclinical study done by a Japanese group identified that many of these resistance mechanisms were due to point mutations in the NTRK genes themselves and these can occur typically in the activation loop on the ATP binding pockets. There were a couple of common mutations, G595L, G595R in TrkA and G623R in TrkC.36 These are known as solvent front mutations. When mutations occurred in those nucleotides specifically, the tumor became highly resistant to larotrectinib or entrectinib. Innovatively, there are new-generation drugs being designed specifically for the solvent front mutations. TPX-000537 and LOXO-19535 have been proven to be active against all solvent front mutations.

These are two recent ASCO and AACR presentations, highlighting the progress of new-generation Trk inhibitors. For example, with TPX-0005, there was an overall response rate of 70% in the TKI naïve patients who had not received larotrectinib nor entrectinib. But in the refractory patients, there could be some benefit still. We could see at least one patient who had either of two prior Trk inhibitors and still had a very dramatic response to TPX-0005.38 There was a similar response shown with LOXO-195, presented in AACR 2019, where simply Trk-resistant cases were enrolled.  In the 29 Trk-resistant patients, the objective response rate was 34%. Further, specifically for those whose tumors harbored the solvent front mutations the response rate increased up to 45%.39 We are going on the right direction to have second and an ultimate third-generation inhibitor for these patients with these driving NTRK fusions.

There is the potential for being able to track these NTRK mutations as they develop over time. The Italian group presented an interesting case study where they tracked the developing resistance using serial circulating tumor cell DNA (ctDNA) analysis. When they put a colorectal cancer patient on a Trk inhibitor, the patient had a very significant response in terms of their copy number of the Trk fusion in the ctDNA. Over time as the cancer started to slowly grow back, and four months later, there was growth of new tumor deposits. We could concurrently see in the ctDNA the emergence of the G595R resistance mutation.40 In this sense, there may be a role for tracking these patients over time, perhaps not every two weeks as was done in this study, perhaps once a month or once every couple of months to identify the emergence of these NTRK  resistance mutations earlier on, and now with the newer therapies being able to target these patients as well. In the preclinical study mentioned earlier, there were also other non-NTRK mutation resistance mechanisms developed. Many of these appeared to be related to MAP kinase pathway activation36. At least pre-clinically, the addition of a MAP kinase inhibitor may rescue response to the Trk inhibitor itself and some of those studies had been launched and are showing some promises as well.    

 The necessity of molecular testing in GI cancers

As we all struggle with the decision as to what tests to do for these patients, and we have regular discussions between medical oncology and pathology. There has been some discussion as to whether we should be doing single gene testing, whether we should be just looking for NTRK1/2/3 fusions only, or whether we should be looking for a whole panel of genes. Dr Nathan Pennell, who was a lung oncologist, did a very interesting study where they worked on a hypothetical model, meaning these are not real patient data. But it was a hypothetical assessment of what it would cost, and how long it would take to do next- generation sequencing versus single sequential gene testing versus exclusionary testing versus hotspot panel testing for patients with non-small cell lung cancer.41 The measurements were time results, cost and efficiency overall. It was found the idea of doing sequential single testing was inferior to doing just a single NGS panel from the get-go, both for Medicare insured and commercially insured patients. We can see the cost difference was well over a million dollars for the 2000+ Medicare insured patients to just go ahead and get an NGS panel from the get-go.41 There was a significant time saving benefit for these patients as well.

Bring this to GI cancers, we have a hypothetical scenario here where the standard colorectal cancer testing panel consists of KRAS, NRAS, BRAF, the MMR genes, MSI testing, additionally HER2, NTRKs, FGFR, ALK, ROS. And if we say that each one of these tests individually costs about $400 as an individual gene test and then you're stuck with a bill of about $3200, whereas most one-time NGS panels now cost less than $3,000. There's even one company out there,doing the $900 NGS panel that captures all of these genes. It gives us the argument that we should be going ahead and doing just a broad panel for all of these patients to be able to capture all of these genes, and tailor their therapy from the outset.

One of the push backs or concerns is whether we have shown the gold standard, whether we have shown an overall survival benefit for patients who have these predictive biomarkers identified. With Know Your Tumor Pancreatic Cancer Database, Dr Pishvaian looked at over 1000 pancreatic cancer patients. All of these patients underwent molecular profiling with next generation sequencing. And they were tracking the longitudinal outcomes and therapies received by these patients. By the time when this was presented this at ASCO in June 2019, over 600 patients were identified and 189 of the patients had actionable findings by study definitions.42Out of 189 patients, 46 receive molecularly matched therapy that was tailored to the predictive biomarker that was identified while 143 received unmatched therapy. Unsurprisingly, the majority, 488 patients, did not have any specific actionable findings. And when we were able to look at the overall survival as defined from the time of diagnosis of their disease, it was pretty clear that patients who had an actionable biomarker and received appropriately enough therapy, that they had a meaning overall survival, which was one-year greater than for patients who received who either had no biomarker, no actionable biomarker, or who had an actionable biomarker but did not receive the appropriately matched therapy.42 The findings suggested that we should be doing this this testing and tailoring our therapy for the patients as early as we can.

Last but not least, these actionable mutations really are not rare. We often call them rare but they're not rare, for example, by the EU definition which is defined as less than one in 2000 patients. Even for very uncommon mutations like NTRK fusions, and MSI-high tumors, of pancreatic cancer, they should be defined as uncommon, but not rare. The molecular testing is much less expensive than standard therapies, but also that panels testing is more efficient and more cost effective than doing single sequential gene sequencing. The testing does reveal legitimately actionable biomarkers with disproportionate benefit highlighted by the 1-2% of all cancer types, and particularly GI cancer types that harbor these NTRK fusions. Ultimately, with time and with increasing data, we're going to show that actionable mutations lead to a survival benefit when patients are treated accordingly.


Questions and Discussion

Thinking of molecular testing, as both predictive and for classification biomarkers as well as the potential for targeted therapy options, that's very helpful concept. Do you think this would help with arguing for greater NGS reimbursement in GI tumors since it isn't tied to just targeted therapy, but also to the full prognostic diagnosis?

I do think so. I think that there is prognostic value for some of these standard tests  that would have been used in cancer  for 20 -years thing like ER/PR status where  clearly has the prognostic value, as well as a predictive value and certainly the breast cancer advocates had made the case for that decades ago. There's no reason that we shouldn't be doing the same for gastrointestinal cancers as well. I think it will be more and more difficult to distinguish the prognostic benefit from the predicted benefit, just because it's going to be really hard for an oncologist to identify a patient with a powerful predictive marker and not treat them with that appropriate therapy, so that that muddiness between a predictive biomarker and a  prognostic biomarker is going to remain unclear in a good way for the sake of patients because they'll get the right there.


How do you approach or address the potential value of NGS thing to your patients knowing there may be no actionable findings? Can you just talk a little bit about how you discuss that with them?

Sure. It's a conversation that I had almost every day, both in the germline setting as well as the somatic tumor testing.  From a germline perspective, especially because I see primarily pancreatic cancer patients, I discussed the fact that we are identifying that  about  7-10% of pancreatic cancer patients harbor some form of germline mutation that may have developed may  have led to the development of the pancreatic cancer, and that identifying those mutations can not only have major implications on siblings and children, because there are now screening tools to be able to identify cancers earlier. For the patient with the gBRCA data coming out for PARP inhibitor-based therapy, that there may be treatment implications for the patient themselves. As far as somatic testing, it's the same discussion, What I said to them was about 25% of these tumors harbor mutations that may lead us to a specific therapy that's very tailored towards your tumor. Even though statistically, chances are we're not going to find such a mutation at around 75% probability, it certainly is worth going through the testing for that 25% chance in that we may be able to find a mutation that leads to specific therapy.

And do you find patients are pretty receptive to that?

Very, I mean, you know, I've talked about a one in four chance, they would probably take one in 100 chance. To patients, doing tests such as MSI testing, or NTRK testing may suggest a chance or promise on potential therapy. That's quite powerful, even if the chances are quite low.


Next question has to do with right sided versus left sided CRC at your institution, do you see any difference in ordering NGS testing for right versus left- sided CRC?

That's an interesting question. We have never actually been posed that question. No, our institution does not distinguish right-sided versus left-sided in terms of authorizing testing. Certainly, we all know about the survival differences and the molecular subtyping variance seen in right versus left. We have kind of expectations of what we might see. But no, they allow us to test all patients equally.


To follow up with testing, specifically at MD Anderson, and what percentage of pancreatic patients can you estimate are getting the full NGS testing, including fusion?

That's a  hot button question. Because when I started here in May, surprisingly, MGM was not doing testing routinely on pancreatic cancer patients. And we understand, I think, to the credit of my institution, tries to be very rigorous in the justification of this kind of testing. And, and so, you know, for colon cancer testing for breast cancer, for all the testing, there could be a presentation to a pathology group to go through the data that exists and justify it. I went through a presentation for pancreatic cancer back in June and demonstrated to our entire pathology department. The value a lot of what I just showed you just now, the value of doing the testing because we can identify these patients and benefit them disproportionately. And so since June, yes, we've been testing every pancreatic cancer patient for somatic mutations, as well as for germline mutation.,  There's a variety of testing panels, and the testing panel we use tests about 150 genes, actually, we're going up to a panel that's going to test 400 genes. And then separately, they also do an RNA fusion panel to test for example NTRK fusions and others.

Okay, so it sounds like communication between oncology and pathology is crucial in this situation. Yeah, and, and I would say also that many times as center needs an oncology champion, to really say this is something that we really need it our institution.


Thinking now outside of MD Anderson, what percentage of pancreatic cancer patients are getting full NGS including fusions? And could you also estimate full NGS percentages for metastatic colorectal cancer inside and outside of MD Anderson? To summarized, the first question is percentage of pancreatic outside of MD Anderson and the next is mCRC.

I wish I knew that data. To be honest, I can only tell you experientially, especially with MD Anderson being a large referral center. We do have a   large number of our patients getting tested on the outside for when I see them. Now of course, there may be a selection bias for patients who are making the trip down to Houston that maybe they're more apt to get testing. But I do see testing and the  appetite for testing for pancreatic cancer patients growing and I think part of it is because it's a bad disease and any caring oncologist is going to want to try and do everything they can to identify treatment pathways for these patients. The other thing is that the NCCN guidelines coming out and supporting testing both germline and somatic testing for most patients has really buttressed or supported the idea that the community physicians test pancreatic cancer patients, and they rarely get pushback anymore from their pathology departments to order testing for their patients. The reimbursement is another question way beyond this discussion. Many of the tests are being done by retail testing labs. There are several labs out there, such as Foundation Medicine, Caris, Tempus, Paradigm, just to name a few, but there's many other out there. The reimbursement models for those are still very much in development. I think that's something in discussion that should be with the patient. Some labs do not balance bill at all. Some labs have some minimal copay; sometimes it's a little $50. But there should be an active discussion about that.  The patients generally are willing to, to engage in that.

As far as colorectal cancer, again, I don't think I have any hard data as to how many patients are being tested. And even in colorectal cancer gets a little bit muddier because some very proactive pathology departments and institutions have said, we are going to routinely test KRAS, NRAS, MMR, BRAF, and HER2 for every colorectal cancer patient, which is wonderful, which they're  getting a decent panel from the get-go. But unfortunately, there's sometimes more resistant to do a broader panel that's capturing other rare fusions because there are 1% of colorectal cancers whose tumors harbor NTRK fusions or 0.5%, whose tumors are ALK/ROS1.  Sometimes we are missing those smaller groups. And it depends on the willingness of the department to do a full broad panel versus just continually tailoring their own panel, what I would argue against and I think that's some of the data that I presented is, I would definitely argue against doing first KRAS testing. And if that's wild-type, tests of NRAS, BRAF, etc., are down the line. It takes too long for patients as I showed you before, and it is not cost-effective.

Talking about pancreatic cancer and mCRC, are you seeing generally people moving towards a single panel that tests everything at once? How common is it for providers to still entrench in that ladder type testing that tests a few at a time sequentially and how can that be overcome?

Down to the second question, I think the way it can be overcome is that consensus or agreement is reached between medical oncologists and pathologists. There needs a champion, both in Oncology and Pathology for the value of panel testing.

As far as how common I would say, in pancreatic cancer, most of the testing that I've seen has always been panel testing, because there really isn't a standard test, such as KRAS in pancreatic cancer that might be tested as a one-off test. When patients with pancreatic cancer that I've seen get tested, they almost always have a panel testing.

 In colorectal cancer, it's a little bit muddier because a pathology department might have a standard that they're going to test, KRAS and BRAF in all patients. It takes a next step to figure out we want a broader panel. it's probably been a little bit more mixed, maybe 50/50, in terms of patients who have just a handful of genes versus a full panel that's being tested.

You've mentioned several times the importance of having an oncology champion, in terms of communications between oncology and pathology, and really getting the right testing for patients. How much difference would there be between a place like MD Anderson and a community hospital to have an oncology champion? And what would you recommend for smaller centers or community hospitals to get that conversation started?

There are many community hospitals, which have had champions, and have had discussions and do a very broad panel of testing for every patient and are very much at the cutting edge of trying to implement precision medicine for patients. There are institutions that however still are behind the times in doing panel testing. As I mentioned, MD Anderson was not doing panel testing for pancreatic cancer eight months ago, which was a surprise to me but now they are. Just like when I had to go and present to pathology department, somebody needs to take up the charge and say this is something that we need to do for all of our patients. To initiate that spark, there are advocacy groups that are out there. For example, the Pancreatic Cancer Action Network for pancreatic cancer patients has been very vocal about encouraging panel testing. The Cholangiocarcinoma Foundation has done similarly.

There are testing labs that they have an ulterior motive. I don't want to hide that at all. A Foundation Medicine representative is going to love the idea of trying to convince a pathology lab to do panel testing with Foundation Medicine. For better for worse, or, for business or for other reasons, there are a lot of voices out there that are trying to encourage centers, academic centers, and community hospitals to adopt panel testing.


For our last question, we're going to shift gears a little bit. Thinking of your data showing NTRK fusions, more common MSI-high tumors, do you see any possibility of checkpoint inhibitors combined with Trk inhibitors?  Normally these subsets are more likely cold tumors and we are not sure when Trk tumors are inhibited by a Trk inhibitor if they get warmer, like higher numbers of immune cell infiltrate. What do you have on potentially combining Trk inhibitors with checkpoint inhibitors?

I think that's a great question and one that unfortunately I'm going to give the standard answer, but I think something should be investigated on a research basis a lot of value to that question and worthy of testing. There are examples in other disease types where a combination of a specific small molecule inhibitor, in combination with immune checkpoint inhibitors has been shown toxic and bit risky.   Thinking of melanoma, some of the earlier studies investigating the MEK inhibitors combined with immune checkpoint inhibitors, including CTLA4 or anti-PD1, anti-PDL1, had showed toxicity, surprisingly. 43,44,45,46,47 However, there are some promising data, combining small molecule inhibitors with checkpoint inhibitors in some of the earlier studies of lung cancer48 and other disease types. In general, I think it's a great question that's worthy of being asked and explored in the context of a research project.



  1. N Engl J Med. 2011 May 12;364(19):1817-25

  2. Lancet. 2004 Sep 25-Oct 1;364(9440):1127-34

  3. N Engl J Med. 2002 Jan 10;346(2):92-8

  4. J Clin Oncol. 2007 Mar 1;25(7):760-6

  5. Ann Oncol. 2015 Sep;26(9):1877-83.

  6. N Engl J Med. 2010 Oct 28;363(18):1693-703.

  7. Lancet Oncol. 2011 Oct;12(11):1004-12

  8. Nat Rev Cancer. 2010 Apr;10(4):241-53.

  9. Clin Cancer Res. 2018 Oct 15;24(20):5018-5027 pancreatic

  10. Cancer Discov. 2018 Sep; 8(9): 1096–1111.

  11. Nat Genet. 2015 Sep;47(9):1003-10.

  12. J Gastrointest Oncol. 2016 Oct;7(5):797-803

  13. Nat Med. 2015 May;21(5):449-56.

  14. NCCN guideline Pancreatic Adenocarcinoma. Version 2, 2019.

  15. J Clin Oncol. 2008 Apr 1;26(10):1626-34

  16. J Clin Oncol. 2016 Jan 10;34(2):179-85.

  17. J Gastrointest Oncol. 2015 Jun;6(3):314-21

  18. J Gastrointest Oncol. 2016 Oct;7(5):687-704

  19. Science. 2017 Jul 28;357(6349):409-413.

  20. J Clin Oncol. 2019 Jun 10;37(17):1460-1469.

  21. Lancet Oncol. 2016 Jun;17(6):738-746.

  22. J Clin Oncol. 2018 Jan 20;36(3):276-282

  23. Clin Cancer Res. 2018 Dec 15;24(24):6612

  24. JCO Precision Oncology 2019 :3, 1-10

  25. JCO Precision Oncology 2018 :2, 1-15

  26. Cancer. 2018 Apr 1;124(7):1374-1382.

  27. N Engl J Med. 2019 Oct 10;381(15):1492-1493

  28. Clin Cancer Res. 2018 Dec 1;24(23):5807-5814

  29. Nat Rev Clin Oncol. 2018 Dec;15(12):731-747. 

  30. Oncotarget. 2018 Nov 9;9(88):35809-35812

  31. Mod Pathol. 2019 Jan;32(1):147-153 

  32. Ann Oncol. 2019 Sep 1;30(9):1417-1427

  33. N Engl J Med. 2018 Feb 22;378(8):731-739

  34. Cancer Discov. 2017 Apr;7(4):400-409

  35. Cancer Discov. 2017 Sep;7(9):963-972

  36. Mol Cancer Ther. 2017 Oct;16(10):2130-2143

  37. Cancer Discov. 2018 Oct;8(10):1227-1236

  38. Journal of Clinical Oncology 2018 36:15_suppl, 2513-2513

  39. AACR; Cancer Res 2019;79(13 Suppl):Abstract CT127.

  40. Cancer Discov. 2016 Jan; 6(1): 14–16.

  41. JCO Precision Oncology 2019 :3, 1-9

  42. Journal of Clinical Oncology 37, no. 15_suppl (May 20, 2019) 4138-4138.

  43. N. Engl. J. Med 2013; 368: 1365–1366

  44. Pigment Cell Melanoma Res. 22015; 8: 611–612

  45. Nat Med. 2019 Jun;25(6):941-946.

  46. Nat Med. 2019 Jun;25(6):929-935

  47. Nat Med. 2019 Jun;25(6):936-940.

Ther Adv Med Oncol. 2018;

About the Speaker
Challanges in Treating GI Cancers
Molecular Classfication
Predictive Biomarkers
NTRK in GI Cancers
The Approved TRK Inhibitors
Next Generation Therapeutics
Necessity of Molecular Testing
Q&A with Dr. Pishvaian
bottom of page