® titled “Hepatocellular Carcinoma: A Rapidly Evolving Treatment Landscape

.” I’m Dr Richard Finn from UCLA [University of California, Los Angeles] Health. Joining me in this discussion are my colleagues Dr Tony

 Bekaii-Saab from the Mayo Clinic in Phoenix, Arizona; Dr Peter Galle from the University Medical Center in Mainz, Germany; Dr Pierre Gholam from Case Western Reserve University in Cleveland, Ohio; and Dr Katie Kelley from UCSF [University of California, San Francisco] in San Francisco, California. Today we’re going to be discussing a number of topics pertaining to the use of systemic therapy in advanced liver cancer. We’ll discuss the latest research in the field and the impact of recent clinical trials on making decisions around treatment selection. 

Let’s get started with our first topic, which is some background on managing patients with liver cancer in general and our approach to them. As we’ve talked about in other settings, the ideal management for a patient with liver cancer involves a multidisciplinary setting. This is really 2 diseases in 1. There is chronic liver disease and some degree of cirrhosis, which occurs in about 90% of our patients, and there is the tumor and the complications from having a tumor burden. We know that the only way to cure these patients is with surgery: either resection, which is difficult to do at times because of the underlying cirrhosis, or liver transplant, which has its own challenges.

To optimize outcomes for patients, patients need to be assessed at a center that offers all the latest cutting-edge approaches. This includes interventional radiology, which plays a key role; hepatology; and oncology, as well as supportive care. Today we’re fortunate to have a multidisciplinary group. We have Dr Pierre Gholam and Dr Galle, both hepatologists, who are very involved in the liver cancer space.

Pierre, can you give us an overview and discuss some of the complexities and challenges in managing a patient with liver cancer? How do you approach it from the hepatology standpoint and work with your multidisciplinary team, from diagnosis to staging to triaging patients?

Thank you, Richard. It’s a pleasure to join this distinguished group of experts and talk about the interface between chronic liver disease and HCC [hepatocellular carcinoma]. HCC is clearly a major complication of advanced liver disease, which appears in an increasing number of patients because we’re able to take better care of patients; as they live longer, they have more time to develop hepatocellular carcinoma in the background of liver fibrosis. There are multiple challenges that face us when managing someone with cirrhosis who may be at high risk for HCC. We’ve made significant strides in the past decade or more in terms of refining our ability to diagnose these patients effectively.

Perhaps the biggest advancement in the field on a diagnostic level comes from the robustly established criteria for definitive diagnosis of HCC. Our societies, including the American Liver Foundation, the European Association for the Study of the Liver, and even the Asian Pacific Association for the Study of the Liver, all agree that noninvasive diagnosis with imaging in the context of cirrhosis is sufficient and enables us to not only make the diagnosis confidently but also establish a plan of care that includes treatment. I’ll remind this audience that this includes the findings of arterial phase hyperenhanced, followed by delayed venous or later phases washout.

These findings by themselves in the appropriate person who has cirrhosis are sufficient to make the diagnosis without the need for a biopsy. We have wide consensus, including in the medical oncology community, that this is indeed the case. Another relatively recent refinement in our ability to gauge the probability of HCC and less definitive imaging comes from the LI-RADS [Liver Imaging Reporting and Data System]. This system establishes degrees of probability of liver cancer and lesser definitive lesions, which is similar to what the medical oncologist might be familiar with in terms of PI-RADS [Prostate Imaging Reporting and Data System] and in the breast imaging field. The LI-RADS criteria, which have been incorporated in all these professional society guidelines, range from LI-RADS 1, which as you know is definitely benign, to LI-RADS 5, which means definitely HCC, and everything in between.

Our societies have incorporated this into our multidisciplinary discussions, and the government has taken notice. OPTN [Organ Procurement and Transplantation Network] and UNOS [United Network for Organ Sharing], which are the agencies that govern and regulate transplant operations in the United States, now rely on these criteria to allocate priority and the end points in patients who are candidates for liver transplantation. All of these are advancements in our ability to diagnose effectively, without significant liability, on the biopsy side.

There are certainly complexities related to management that have to do with juggling the relative burden of liver disease, from a coagulation deficit standpoint, management of fluid overload, cognitive dysfunction, renal dysfunction, you name it. This is the kind of person who gets HCC, and you have to offer them effective treatment for their cancer but also must also make sure this does not in any deleterious way affect all these other very precarious situations.

The Complexity of HCC Multidisciplinary Management

 Alpha-fetoprotein [AFP] has been around for a long time, right? How do you approach the use of AFP in diagnosing patients with liver cancer?

As an AFP enthusiast, we have had a love-hate relationship with AFP in the hepatology field. AFP was, at first, an established end point, a requisite end point in our surveillance strategies endorsed by all societies. Circa 2010 we realized that AFP by itself does not add much in terms of sensitivity and specificity. In fact, the US hepatology society, AASLD [American Association for the Study of Liver Diseases], dropped it from its guidelines for a while. Clearly, AFP has now been established to be of value in terms of determining prognosis.

It is not a good measure of or means of establishing a diagnosis. We would all agree on that. In fact, if you look at various clinical trials in first- and second-line settings, you can tell that the cutoff for AFP is useful in delineating response to therapy in some patients, but it could also could potentially, to some extent, predict the degree to which people might benefit from therapy. If you ask me about whether I use AFP in my surveillance algorithm for patients with HCC [hepatocellular carcinoma], I would say that I absolutely do in every patient in combination, of course, with imaging. I don’t use dynamic imaging for surveillance, but clearly the limitations of AFP have to be factored into this.

 For the audience, we need to distinguish between screening and diagnosis, and you touched on that toward the end. Because of the complexity of underlying liver disease and tumor burden, outcomes and studies have shown that the patients with cancer who do the best are those who have better-preserved liver function. Ideally, patients who have chronic liver disease are screened on a regular basis. To your point, the guidelines recommend an ultrasound every 6 months, and whether you add AFP to that is where your desire is to incorporate it into screening vs diagnosis. You touched on the fact that AFP doesn’t necessary play a role in diagnosis; that’s dynamic imaging. When you make the diagnosis of liver cancer, how do you assess liver function? How do you assess whether this patient is a good candidate for aggressive treatments?

There are tried-and-true methods of assessing “liver function.” These were mainly derived in response to the need for patients with advanced liver disease to receive surgery. Surgeons obviously don’t like taking patients to the OR [operating room] and not having a good outcome. We had to figure out a way to understand who might fare well with surgery, and by extension, this turned out to be a pretty good method to assess survival at a finite period of time. The tried-and-true method for over 4 or 5 decades is the Child-Pugh-Turcotte score, which is a combination of 3 objective and 2 subjective criteria that you tabulate. The subjective criteria are ascites and hepatic encephalopathy. Your objective criteria are PT [prothrombin time]/INR [international normalized ratio], albumin, and bilirubin. You add those up, and if you exceed 6 points, then you have Child-Pugh B. That puts you at a reduced likelihood of survival at 1 year, although your survival is still expected to be around 60% to 70%. If you exceed 9 points, then you have Child-Pugh C cirrhosis. That’s someone whose life expectancy at 1 year probably does not exceed 50% to 60% in the modern era. That’s 1 way to do it. It’s not a perfect way, but it’s certainly a helpful way to assess liver function.

 That has been incorporated into the Barcelona staging system, right? The BCLC [Barcelona Clinic Liver Cancer] has gravitated to be the main staging algorithm for how we conduct clinical studies and therefore how we can relate to patients in the clinic. To your point, if we have a patient who is Child-Pugh C, they are going to die of liver disease before they die of cancer in general, right?

 As we will discuss, as we have more active drugs, there are probably some patients who have decompensated liver disease from a large tumor burden. Perhaps if we improve their tumor burden safely, then they might improve their liver function. In general, when patients are cerotic, that’s a competing risk for death. That’s why staging systems that we use, like TNM [tumor node metastasis] and other malignancies, don’t carry enough weight in the liver cancer space because they ignore the competing risk for outcomes, which is liver disease.

Most studies have concentrated on patients who are Child-Pugh A. The Child-Pugh B group is very heterogeneous. We’ll talk later about how we approach those patients. In treating them with systemic medications, you generally see that they’re excluded from clinical trials.

Screening Vs Diagnosis: Assessing Liver Function

 Can you give us a brief overview of the BCLC [Barcelona Clinic Liver Cancer] staging system? There’s early stage, which is 0; there is A, B, C, and then there’s D stage, which is advanced and generally involves supportive care unless the patient could have a transplant.

The Barcelona Clinic Liver Cancer staging system incorporates 3 important components of care in patients who have advanced liver disease and HCC [hepatocellular carcinoma]. There is the staging system, which factors in the extent of disease, the number of lesions, macrovascular invasion, and extrahepatic spread. Those intuitively make sense when you’re treating cancer. There is an ECOG performance status component, which we all know across different tumor types. It’s an important factor in the patient’s ability to be treated, and it carries prognostic value. There is also the Child-Pugh score, which as we’ve just outlined, is critical in determining the patient’s ability to survive based on the presence of liver disease alone as well as their potential ability to tolerate and receive treatment.

Once you factor all those together, you come up with the 0-to-D system. A patient in stage D is someone who has a very poor performance status or very advanced liver disease; therefore, best supportive care is agreed upon to be the best intervention for these patients. A patient in stage 0 is someone who has a very localized lesion for whom in-home therapy could include ablation or surgery. A patient in stage C, which is the domain of systemic therapy historically, is someone who typically has macrovascular invasion or extrahepatic spread but does not have very advanced liver disease and still has an ECOG performance status of 0 or 1. Therefore, this patient could potentially benefit from treatments that might enable them to improve quality of life, and one might argue that it could certainly extend life based on a number of studies in first- and second-line settings.

A patient with BCLC stage B is someone who may have fairly significant liver disease, but it is mostly confined to the liver. There are those we now call the patients with good stage B and bad stage B. Patients with good stage B may have sizable tumors, but they are localized such that that local-regional therapy with TACE [transarterial chemoembolization] and TARE [transarterial radioembolization] or other interventions might be of benefit. The patients with bad stage B are people who have the disease localized to the liver, but it is fairly widespread throughout the liver such that realistically, one would not be able to be expected to apply local regional-therapy for those patients.

Of course, stage A is the domain of definitive therapy in the proper context and someone who might be able to be a candidate for liver transplantation. That’s a very broad and lengthy discussion of what BCLC tells us.

The BCLC Staging System

 That’s a good transition into the approach for the intermediate group, which is a large group of patients. These are patients who have liver-confined disease and well-preserved liver function. Historically, local-regional treatments were the backbone. Peter, you’ve been very active in this space. Can you give us some background about the ideal TACE [transarterial chemoembolization] patient? It is important to recognize that TACE is not curative and that we are transitioning those patients to the next step in their treatment.

The situation in the intermediate stage is a peculiar 1. Let’s go 20 years back, which was when TACE was defined. There have been many trials that have not been able to demonstrate the beneficial effect of TACE. We then had 2 randomized controlled trials and a meta-analysis providing the background for TACE as the standard of care in patients in the intermediate stage. Why were they successful? First of all, at that time they had no competitor. There was no systemic therapy around.

This is something we should not forget: They were running against nothing. The trials were then positive. There were 2 trials because they were concentrating on a very recused tumor burden and on patients with sufficiently good liver function. They selected very carefully, and in this particular setting, the efficacy of TACE was demonstrated. What is happening afterward is that TACE is spread out. Everybody gets TACE, even patients with very large tumors, which are precisely those who have never been tested in clinical trials and those with poor liver function. In addition, what we have in these days is an armamentarium of many systemic therapeutic options. We have to reconsider whether TACE is the appropriate therapy, and we still have to consider to whom TACE is and is not applied.

What has happened has been nicely demonstrated in noninterventional trials, such as the BRIDGE trial. It was interesting to see that in a time when the standard of care in advanced-stage disease was defined by sorafenib. With the BRIDGE trial, which was a worldwide observational trial, it was becoming obvious that most patients—even those with advanced-stage disease—were treated by TACE. The availability and the lessons learned 20 years ago were being maintained, and that needs to be challenged today. I’m not saying that there is nobody who, in a perfect way, is suited for TACE. There are still patients suited for TACE: those with a limited tumor burden, and what should also not be neglected are cases where the tumor is accessible. If there is a nicely feeding artery and superselective approach, then this is a good patient for TACE.

If it is a diffuse, large tumor, and if this is not superselectively targetable, and if this is coming at the price of loss of liver function, this is precisely why we have to stop or never get started. In a good patient progressing on TACE or not benefiting from TACE, we need to make decisions in times of availability of systemic therapy. If there is no objective response, let’s say after 2 cycles of TACE in a situation where TACE can be appropriately done, then we need to switch. This is where the multidisciplinary team steps in. This is exactly what we all have to discuss: What is the best measure? What is the best therapeutic option for the patient? How can we switch on time?

That has been neglected, but I’m happy to see that, in the field, the term overexhaustive TACE is being discussed. Experts are acknowledging more and more that TACE is not for everybody and that it is time to end our endless cycles. Systemic therapy needs to define its role, and we will see next year’s trials challenging TACE in terms of intermediate-stage treatment, and systemic therapy will shift the BCLC [Barcelona Clinic Liver Cancer] scheme to the left.

You bring up some important points. One thing to remember is that, with liver cancer, you can have advanced liver cancer or someone who is appropriate for systemic treatment even though they don’t have extrahepatic spread, even though they don’t have metastases. There can be disease within the liver that makes a patient a candidate for systemic treatment. That would mean vascular invasion, meaning the tumors invading blood vessels. That’s a marker of patients who don’t do well with local-regional treatment. As you had commented, these are patients who have repeated procedures, and we’re not controlling their tumor.

Every time you reimage them, they still have residual tumor or tumor that is progressing despite repeated TACE. That’s when we should be bringing these cases to the multidisciplinary team. Before we had systemic treatments that were really improving survival, TACE filled that void. We had nothing else to offer patients. In 2008, sorafenib was approved vs placebo; it improves survival without a significant response rate. Even with that data, it still took time to convince multidisciplinary teams that systemic treatment plays a role in the management of patients who progress on TACE. For that matter, some centers use radioembolization or Y 90, but the issues are the same: These aren’t curative approaches. Patients will progress.

The TACE Approach for Intermediate-Stage Disease

 Tony, lead us off with some discussion about the impact of the 3 choices we have for frontline treatment—sorafenib, lenvatinib, and more recently atezolizumab and bevacizumab—based on the IMbrave150 trial.

Thank you, Rich. The landscape is getting more simplified. We have 3 options, but there is certainly a dominant option. The other 2 options are likely are going to be options for those who may be ineligible for immunotherapy. For the longest time, as you all recall, we had sorafenib, and most of us will remember the days before sorafenib where Megace [megestrol acetate], tamoxifen, or some other unproven agent would be used because we were so desperate for systemic agents in this disease. Then came sorafenib based on the SHARP trial, which showed an improvement in outcomes, but it was a tough drug. It was a tough drug to use.

The dose was challenging, and everyone came up with their own dosing schedule and regimen. For many years after that, there were attempts to displace sorafenib, mostly by agents with similar targets, plus or minus another drug on occasion with not much success. In fact, with sunitinib, we’ve seen the curve going in the wrong direction, until the REFLECT study with lenvatinib vs sorafenib, which was a noninferiority study. It certainly has limitations in terms of how to interpret the results, but it showed that we now have another option other than sorafenib, which appears to be noninferior for survival. It seemed to have trends favoring it for PFS [progression free survival] and maybe some response.

In the whole setting, it was just the noninferior option with the caveat that it was less toxic because it was primarily a VEGF-directed therapy, not so much as with sorafenib with all this background noise with the other kinases. It was easier to use, it got adopted, and it was becoming a favorite. We then got the IMbrave150 trial, which was designed to answer the question of immunotherapy plus a PD-L1 inhibitor, atezolizumab plus bevacizumab vs sorafenib. This couldn’t come more timely since we’ve had some setbacks with I/O [immuno-oncology], specifically with the nivolumab vs sorafenib study, which was negative. It had some trends, but overall it was a negative study.

Pembrolizumab in the second-line setting vs placebo was also negative, which was another setback. The big question was this: How can we move that field forward? There are patients who have tremendous responses to single-agent I/O, yet we’re not able to identify them. In the undiluted patient population, we’re not seeing those signals. Dr Finn presented a study at ESMO [European Society for Medical Oncology] Asia Congress, IMbrave150, which randomized atezolizumab plus bevacizumab vs sorafenib. This study was essentially powered for 2 coprimary end points, overall survival [OS] and PFS, and the secondary end point of response rate. There were a number of other end points to the study, including a PRO [patient-reported outcome] quality-of-life element.

This study came and hit every single 1 of the end points, from PFS to OS, with very impressive hazard ratios. In fact, the OS has not been reported yet, but it should be at some point later this year or maybe early next year. It seems to be tracking along close to the 2-year mark. Not knowing the results, we’ll see where they end up, but it’s quite an impressive movement in all the end points with the response rate between 25% and 30%. Most important, it preserves and improves the quality of life of the patients, so that certainly displaces sorafenib as a standard.

By default, because lenvatinib is a noninferior agent to sorafenib, and it has a similar survival, the same inference can be made for lenvatinib. This becomes our standard of care for patients with advanced disease and a Child-Pugh A score because that is what these studies have been looking at. Perhaps we can discuss later what that means for other stages, but for patients with Childs-Pugh A in the first-line setting with a good performance status, there are no contraindications for immunotherapy. This is now the standard.

 We’ve seen some evolution. Sorafenib holds its line until the REFLECT study, where we saw lenvatinib was noninferior for overall survival. But it improved response rates, and it improved progression-free survival, which raised some interesting points for us in the liver cancer space on how we can affect these secondary end points. That didn’t necessarily translate into a survival advantage.

One of the things that came up was the potential impact of treatment beyond progression: There were more patients in the sorafenib arm who were able to get investigational agents, including I/O, which were not necessarily available to the lenvatinib patients. That study also had some interesting inclusion and exclusion criteria in that it excluded patients who had main portal-vein invasion or over 50% of their liver involved with tumor. Many of us interpreted that data as lenvatinib being perhaps a more active drug with a different adverse-effect profile, and you highlighted some of that. The VEGF readout from the event is much stronger than we see with sorafenib with regard to hypertension and proteinuria, whereas GI [gastrointestinal] toxicity is fairly similar between the 2 agents.

Hand-foot skin syndrome, which can be a dose limiting toxicity for sorafenib, was not all that relevant for patients receiving lithium lenvatinib. It then comes to atezolizumab-bevacizumab, which to your point, reassuringly hit all their end points: overall survival, progression-free survival, response rate, and quality of life. Peter, you did a more detailed quality-of-life readout from IMbrave150. Can you give us a top line of what you presented earlier this year?

The background of this patient-reported outcome analysis was a very solid one: More than 90% of patients reported, and it was a thorough assessment based on 2 questionnaires going to quality of life, role functioning, physical functioning. The difference for sorafenib in terms of quality-of-life deterioration in a meaningful way was between 3.7 months for sorafenib and almost a year for atezolizumab-bevacizumab. The significant difference here makes a difference in the life of our patients. They have a much better quality of life, and they can maintain their work and activity. In a situation where we are talking about palliation, their life span is still very limited. It makes quite a difference whether the quality of life is good or impaired, and this is very in favor of this combination in terms of improvement of quality of life and other aspects.

Optimal HCC Frontline Treatments

 Katie, you’re a busy clinician and researcher in a high-volume center [at the University of California, San Francisco]. When you’re seeing patients, how would you assess the frontline treatment for your patient given these 3 options? What would you tell a clinician in our audience who hasn’t seen as many liver cancers? What are some things to pay attention to? Is there a role for TKIs [tyrosine kinase inhibitors] given the hazard ratio for IMbrave150, for which the OS [overall survival] is 0.58? How do you approach it?

That’s agreat question, and thanks again for having me today. I’ll start by saying that I definitely agree with Tony [Bekaii-Saab]: Our de facto standard and the global standard for patients with Child-Pugh A, fit advanced liver cancer patient, is the combination of atezolizumab plus bevacizumab. Based on the strong results across the board from IMbrave150 from the efficacy, the safety, and the quality of life that Peter just highlighted. The bigger question that you’re asking is this: What about patients who are not good candidates for the atezolizumab plus bevacizumab regimen? Who are those patients?

If I were asked to summarize the patients for whom we wouldn’t use atezolizumab plus bevacizumab, I would generally categorize them and maybe 3 or 4 different types of patients. First, the clear-cut exclusion criteria on IMbrave150 would be patients with active varices or those who are at higher risk for bleeding varices or other types of bleeding based on their screening, endoscopy, or other clinical factors suggesting a risk for bleeding. That’s 1 type of patient for whom we must be very cautious and should probably not use the IMbrave150 regimen: patients with high risk for bleeding. We also have to underscore the fact that we don’t have Child-Pugh B safety or efficacy data from the atezolizumab plus bevacizumab regimen. Patients staged at Child-Pugh B are not established as good candidates for IMbrave150 right now.

You touched on the fact that patients with active autoimmune disease or patients post-transplant are also patients who we don’t think are necessarily safe for checkpoint inhibition as a treatment. Those are some general categories of patients for whom, right now, we have to give pause and not use the atezolizumab-plus-bevacizumab regimen. Turning to the other options, this is a fantastic place to be in liver cancer compared with years past, when there were just 1 or 2 drugs starting after regorafenib in 2017. We have choices to choose from, depending on patients’ comorbidities and other medical features, just like in breast cancer and colon cancer. When making that choice for these groups of patients who aren’t good candidates for atezolizumab plus bevacizumab, we can look to lenvatinib and sorafenib.

How do I choose between those 2 agents? Reassuringly, for the TKIs, while they do increase the risk for bleeding, we can say that they have not shown a dramatic increase in variceal bleeding compared with placebo. If we look back to the SHARP trial of sorafenib vs placebo or the Asian Pacific Association for the Study of the Liver trial of sorafenib vs placebo, those trials did not show that TKIs increased risk for serious bleeding events dramatically when compared with placebo in a patient population that was at risk for bleeding. They were patients at stage Child-Pugh A, admittedly. I’m more likely to use lenvatinib or sorafenib with patients who are not eligible for atezolizumab-bevacizumab perhaps because of recent bleeding from varices that have now been treated.

In terms of choosing between the 2 agents in that group of patients, if it’s a patient with Child-Pugh B liver function, then it’s important to note that we don’t have prospective data for lenvatinib in patients with Child-Pugh B. Although there are some favorable points to its toxicity profile in Child-Pugh A patients overall, encephalopathy, proteinuria, and some of the other complications do not yet have prospective safety data in that liver dysfunction population. But we do have more prospective data and history, whether it’s retrospective registry data such as GIDEON, as well as some prospective studies in the phase II context for Child-Pugh B. In a patient who is staged with Child-Pugh B liver function, I’m more likely to choose sorafenib with an empiric starting dose reduction based on some of the VA [US Department of Veterans Affairs] data that have been published.

Turning to the other contraindication to atezolizumab plus bevacizumab, there is concern for patients with autoimmune disease or transplant patients. When comes to autoimmune disease, there is a huge spectrum of risk. In some patients, it may be reasonable to try a checkpoint inhibitor–based regimen with close monitoring, but in patients with active disease requiring immunosuppression and certainly in patients after liver transplant, the risk of rejection supersedes the likelihood of benefit in the advanced cancer setting with graft loss in liver transplant patients, upward of 30% in some small studies.

That’s an unacceptable toxicity from checkpoint inhibition from the limited data that we have so far. In those patients, both sorafenib and lenvatinib are reasonable choices. It’s important to note that lenvatinib does have more QTC [corrected QT interval] prolongation risk than sorafenib. In patients with complex concomitant medications, such as immunosuppressive medicines, it’s important to do drug interaction checks for both drugs, sometimes choosing sorafenib if there is a QTC liability from the combination. Those would be my answers. I’ll also comment on another question that commonly comes up in how we choose between sorafenib and lenvatinib: the question of viral status and hepatitis B vs hepatitis C.

We always have to look at the subgroup analyses for signals of interest in important populations, like patients with hepatitis B vs hepatitis C. The cautionary note is that these are not adequately powered analyses to make high-level pronouncements about which type of patient to treat in which way. It’s important to look at the REFLECT trial of sorafenib vs lenvatinib and the forest plots. The survival is similar, and there was no significant difference in the subset analyses for hepatitis B vs hepatitis C. There was a trend toward benefit for lenvatinib for time to progression and progression-free survival for both viral subgroups, perhaps it was more pronounced for hepatitis B.

Conversely for sorafenib, we’ve seen some analyses that suggest that there is perhaps more activity in the hepatitis C population than hepatitis B, again with the caveats of not overinterpreting subset analyses. Overall I would say that we don’t have enough data to make a strong decision based on viral status. Perhaps lenvatinib should be favored in patients with hepatitis B, but we can’t choose the drug based on hepatitis C status 1 way or the other. It should be based on comorbidity more than viral status.

You bring up a lot of important, practical points. IMbrave150, like other phase 3 studies, excluded patients who had a bleeding risk. The study required people to have an endoscopy within 6 months. But if we look at overall grade 3/4 bleeding events, it was fairly similar between arms. There were a few deaths related to upper-GI [gastrointestinal] bleeding: In the atezolizumab-bevacizumab arm, it was 1%. In this carefully selected population, bevacizumab appeared to be very safe.

Choosing a Frontline Therapy

 With this idea of autoimmune events, that would clearly exclude patients. Autoimmune backgrounds or phenomenon would put someone at risk for toxicity from IO [immune-oncology]. Obviously, it is pan-IO. Interestingly, the NCCN [National Comprehensive Cancer Network] guidelines left it open to use single-agent nivolumab in the frontline setting. I don’t know if you have any thoughts on that, Katie?

In the NCCN guidelines, inclusion of the nivolumab is based on a couple of considerations. There are some patients who are not good candidates for any antiangiogenic therapy, whether it’s a TKI [tyrosine kinase inhibitor] or bevacizumab. I’ve had a few patients who have nonhealing wound complications after a big surgery, for example. I have also had patients who have a significant thrombotic predisposition or have had clots that are intractable on antiangiogenic therapy. There are some patients for whom we can’t use antiangiogenic therapies, and there are also some patients with poor-risk Child-Pugh B for whom we may not feel comfortable using some of our other agents.

For nivolumab, not only is there not the risk of antiangiogenic toxicity or comorbidity, but there are also some prospective safety data with respect to the CheckMate 040 Child-Pugh B cohort. That included patients with a score of up to B8. There were also some retrospective case series data suggesting that there is no impact on baseline liver function from nivolumab and that the rate of immune-related hepatotoxicity or hepatitis is similar in patients with Child-Pugh B as it is for patients with Child-Pugh A. The NCCN guideline inclusion reflects the fact that there are safety data for nivolumab in these contexts and that, while the CheckMate 459 first-line trial of sorafenib vs nivolumab did not meet its end points and was a negative study, we also know that there was a trend toward benefit, with the caution of interpreting trends like that. But we have the second-line data from CheckMate 040 that led to FDA approval in the second-line setting.

We have quality of life end points that support the use of nivolumab in this context. Overall, in a patient population for whom there is not an alternative good therapy either because of risk of toxicity or Child-Pugh B status, it’s a reasonable thing to consider in the absence of level 1 evidence favoring any other choice.

 We can say that for the CheckMate 459 study, which was nivolumab vs sorafenib, while it was a negative study, the response rate was 20% and survival for that single-arm nivolumab group was 16 months.

Exactly. I think we all agree that there is potential for benefit with nivolumab; it just did not meet its first-line outcomes.

It’s a reasonable option; it’s more like rolling the dice though, for 2 reasons. The first is that I’m not sure that 20% of patients actually benefit more. It’s probably closer to 10% or less, which is not insignificant. The other thing is, the patients who are not eligible for atezolizumab/bevacizumab and that those you listed, Katie, are those who will not do well overall anyway. I understand that having this option is good, but the problem is that those patients are those who least benefit from therapy overall. It’s good to have that option.

At the same time, I always get concerned, especially with the NCCN guidelines listing these options, that sometimes the fine print doesn’t get read. You want to find a way to emphasize the fine print that says that this is the exception to the rule across the board. I would still think that lenvatinib or sorafenib would be preferable for patients who are eligible for it over nivolumab. I’m guilty as everyone else is, of having treated patients with nivolumab, single agent who exactly fit the patient profile that Katie described. I get concerned about the guidelines, not about the rationale to include it.

 I agree. I would definitely emphasize that I would consider it for patients who are not candidates for sorafenib or lenvatinib and don’t have a level 1 option, like the open chest wall wound from a bad surgery outcome. That is 1 example that comes to mind. I totally agree.

NCCN Guidelines: Nivolumab in Frontline Setting 

 Bleeding has become a big issue now that we have atezolizumab/bevacizumab. We never really thought of it so much with the TKIs [tyrosine kinase inhibitors], with lenvatinib or sorafenib, though theoretically it’s a VEGF class effect. Though it may be more common with bevacizumab, we thought. Does every patient with newly diagnosed liver cancer need an endoscopy? What’s your thought, Pierre? I’ll be honest, in clinic, we can write a prescription for lenvatinib or sorafenib, it gets filled within a week, and then get started on therapy. Now we’re thinking, “I want to start on atezolizumab/bevacizumab. It’s more active, and it improves survival, but is it going to take me 3 weeks to get an endoscopy?”

That would probably be one of the top 2 or 3 most asked questions when it comes to atezolizumab/bevacizumab therapy. How might one manage or mitigate the risk of bleeding in the setting of cirrhosis? An important thing to remember and consider is that the risk of having varices, or perhaps more importantly, large varices, which are those that are at-risk for bleeding, increases as the Child-Pugh score increases. Someone with Child-Pugh score C is much more likely to have large varices than patients with Child-Pugh B, and they are much more likely to have varices than patients with Child-Pugh A.

The other important point to understand, especially for a medical oncology audience, is that the risk of progression from small to intermediate to large varices per unit of time, let’s say per year, is somewhere around 7% to 10%. If you had someone who had no varices at one point, it would be very unlikely unless their disease progressed rapidly, such as is the case with uncontrolled alcohol use disorders or things like that; it would be quite likely for that patient to suddenly have large varices. There are some rational ways to address this in someone and to not delay therapy too much to mitigate the risk of bleeding.

It is certainly well established that bevacizumab can increase the risk of bleeding, and that includes life-limiting or life-threatening variceal bleeding. Would I be well-advised to screen a patient for the presence of large varices if they’ve never had screening like you’re doing, Rich? Absolutely, I would do that. If it were going to take me 3 weeks to do it? It doesn’t take me 3 weeks to do it when I do it myself.

I can schedule someone to have this done tomorrow, and it’ll take me 5 minutes to do it.

 After their COVID-19 [coronavirus disease 2019] testing, of course.

Yes, with the COVID-19 test, sure. You’re at this prestigious medical institution, UCLA [the University of California-Los Angeles Health], you should probably not need 3 weeks to schedule this for your patients either. I would invest in the effort to do that in a patient who has never been screened or surveyed for varices. It is different for someone who may have had an endoscopy a year ago that showed no varices or small varices, so-called trace varices. I’d be comfortable starting that patient on therapy and perhaps scheduling an endoscopy at some time in the future. That would be a common-sense, rational way to approach this as opposed to getting stuck on the notion that I need to have an endoscopy right now or within 6 months to set up that procedure.

 That’s a good point because getting it scheduled is variable depending on where you practice. Even still, starting 1 dose and then go on and get your endoscopy a few days later. We need to get away from this idea that bevacizumab is heparin or an anticoagulant or something with some sort of biologic effect.

You mentioned the rapidly progressive liver disease. My concern in this context may be the patient who, a year ago had small varices, but now has a new main portal vein tumor thrombus and rapidly progressive portal hypertension. We in oncology have to remember those patients. Correct me and weigh in on this: those patients probably merit a new endoscopy at the decision point.

Yes. That’s an interesting wrinkle in all this. If someone has no portal vein thrombosis burden and suddenly develops a tumor thrombus, a massive portal vein thrombosis, might one expect this sudden rise in portal pressure to lead to the development of large varices? Believe it or not, the literature is not entirely clear on that. There is some indication that acute inclusion in the portal vein can perhaps, in animal models or other settings, cause a rapid rise in HVPG [hepatic venous pressure gradient] and subsequently develop varices. In patients who have had chronic liver disease with a low-flow state in the portal vein to begin with, the impact may not be as dramatic as one might think. Your point is well taken, though. Let’s say someone has no portal vein thrombosis, not even partial portal vein thrombosis, and they suddenly get the classic centripetal expansion of the portal vein with an enhancing portal vein clot. Yes, I would absolutely invest in an EGD [esophagogastroduodenoscopy] in that patient.

Importance of Endoscopy in Newly Diagnosed HCC

 There is a lot of exciting stuff coming down the pike. We’ve talked about the rapidly changing landscape. I’ve been involved in the combination of lenvatinib and pembrolizumab, looking at this combination of IO [immune-oncology] and TKI [tyrosine kinase inhibitor], that is showing response rates of around 36% in a frontline population with Child-Pugh A. That’s moved onto a frontline study of lenvatinib/pembrolizumab vs lenvatinib. That is the LEAP-002 study, which has completed accrual, and we’re waiting for a readout. The safety from that study looks consistent with single-agent lenvatinib or single agent pembrolizumab. Katie, you’re involved in some of the ongoing studies. Can you give us a high-level overview of what we have to look forward to in the coming months?

Yes. Following upon the IMbrave150 trial and the success of the combination of antiangiogenic with IO therapy, there is now the approach of combining a TKI with a checkpoint inhibitor as you alluded to: lenvatinib plus pembrolizumab, as well as the cabozantinib plus atezolizumab study, COSMIC-312. Both of these trials are based on the premise that not only is there a potential for additive benefit or synergy when you combine antiangiogenic therapy with IO therapy, but these TKI therapies can also be immunomodulatory in the microenvironment and promote an immune response in other ways by inhibiting other kinase pathway signaling and conferring potential for additive effect, being active drugs in their own right. You mentioned pembrolizumab plus lenvatinib, that’s the LEAP-002 phase 3 trial, which has completed accrual.

The phase 1b data were presented by Andrew Zhu, [MD, PhD,] this year at ASCO 2020 [the American Society of Clinical Oncology annual meeting], and as you alluded to, it had profound results in terms of response rate of about 36% by RECIST 1.1 as well as a very high disease control rate. This again suggests both angiogenic inhibition and the IO benefit, so it’ll be interesting to see how the phase 3 data look compared to the lenvatinib control arm, bearing in mind that the IMbrave150 trial was against sorafenib in the control arm. The COSMIC-312 trial I mentioned is cabozantinib plus atezolizumab, and it is based on early phase 1b data that haven’t been presented yet. It is also based on the rationale that cabozantinib has some interesting tyrosine kinase inhibition that may be immunomodulatory in a different way, and that has shown some promising results in renal cell carcinoma, presented just recently at ESMO [the European Society for Medical Oncology Virtual Congress] 2020 as well.

Those are a different way of combining antiangiogenic therapy with IO. The other family of trials on the horizon are the CTLA-4 inhibition plus PD-1 or PD-L1 inhibition combinations, based, of course, on what we’ve seen in lung cancer and melanoma and other tumor types that show synergy in the immune response. The 2 trials that are in phase 3 and reporting on the horizon are, first, the HIMALAYA trial, which is a study of durvalumab, a PD-L1 inhibitor, plus tremelimumab, with 3 different arms in the final phase 3 design. The first is the standard arm of sorafenib, but the other control arm is durvalumab alone. The combination arm is a little different than the usual CTLA-4 inhibition. It is based on phase 2 data presented at this year’s ASCO, which is looking at a single priming dose of the CTLA-4 inhibitor tremelimumab given on day 1 of treatment, along with maintenance durvalumab monthly.

It is based on some interesting data from melanoma and lung cancer, which were born out in phase 2, and it showed that giving a single priming high dose of tremelimumab achieved a burst of proliferative activated T cells at the beginning of treatment, and it was associated with a higher response rate and lower rates of toxicity than the higher sequential dose of tremelimumab. It could potentially promote efficacy and maybe minimize downstream toxicity with a single priming dose strategy.

The other combination on the horizon is nivolumab plus ipilimumab, which is also based on phase 2 data. That has also led to regulatory approval after sorafenib, in the second-line context, which we’ll talk more about later. That ongoing trial is CheckMate 9DW, and its compared to the control arm of physician’s choice: either sorafenib or lenvatinib. Both of these will be interesting strategies to compare to the TKI or bevacizumab plus checkpoint inhibitor combination.

There are 4 ongoing phase 3 studies in the frontline setting, after the approval for atezolizumab/bevacizumab, which speaks to the fact that there’s been incredible interest and excitement in the liver cancer space. You mentioned some data from ESMO recently. There were also data with lenvatinib/pembrolizumab presented within the melanoma cohort for patients who progressed on prior IO, showing that there was some activity with this multitargeted TKI, sensitizing patients to response after progression on frontline IO. The other thing along those lines from the phase 1b study of atezolizumab/bevacizumab is, there was a single-agent atezolizumab arm and a subset of patients who progressed on single-agent atezolizumab and then had bevacizumab added to the atezolizumab.

If you look at the PFS [progression-free survival] on atezolizumab alone vs the PFS on atezolizumab and bevacizumab, the atezolizumab/bevacizumab at progression had a PFS that was twice as long as the frontline atezolizumab. There is clearly a lot of science that we still need to work out.

Promising Ongoing Frontline Trials

 The second-line setting has raised some issues for us because all the data we have from phase 3 studies were based on prior sorafenib. Here comes along progress, which we all love in the clinic: we have lenvatinib. There’s been no phase 3 study on drugs after lenvatinib. We then have IMbrave150 and atezolizumab/bevacizumab. Peter, how do you think about your second-line patients now if they’re getting atezolizumab/bevacizumab in the frontline setting?

This is a question that is frequently answered, and as there are no reliable data, it comes down to expert opinion and discussion. What is clear is that there are theoretical aspects to it. It might be an option to discuss a switch in mode of action. When having had an IO [immune-oncology] agent plus an antiangiogenic antibody, it somehow seems to be a good rationale to switch to a TKI [tyrosine kinase inhibitor], although this is just not based on evidence: it’s the theoretical option. With a pragmatic approach worldwide, it’s quite a matter of availability of the approval. If you take, for example, lenvatinib, in many countries, it is listed as first-line therapy.

What do you do now if you have had atezolizumab/bevacizumab in the first-line setting, and lenvatinib is used in the second-line setting? It’s not useful, but of course, it’s useful in theory. The good thing is that we have options. The problem here is we have to do lots of assessments. In real life, these will never mature as phase 3 trials with A against B against C. I believe what we should definitely try, and this is going back to our initial discussion when we start intermediate safe therapy, we should offer our patients systemic therapy, and we should try to offer them several lines of therapy.

In the past, this had already been demonstrated for the sequence sorafenib, regorafenib, and it was the same for lenvatinib. If you have more than 1 option, it is apparently better. In the old days, we were used to treatment beyond progression; in these days, it might be wise to reconsider the therapeutic strategy upon progression and to switch.

 Yes, that’s an important point. When we had sorafenib, and we couldn’t offer patients anything after, if they weren’t going to qualify for a clinical study, it was pretty much continued until clinical progression. This raises the issue of following our patients closely, imaging them on a regular basis, and transitioning them to their next line of treatment on radiographic progression. Certainly 2 or 3 mm, I don’t think we would call that progression, but when there is continuous growth of a tumor or new lesions, that’s a sign to start to transition patients because the runway we’re dealing with here is their underlying liver disease as well. 

Are we going to get to the point where we’re controlling patients’ tumors longer than the natural history of their liver disease? It’s an interesting thing to think about as we have all these new drugs available. Tony, a patient gets atezolizumab/bevacizumab in the frontline setting. I would say that there are several options that we know about. There is lenvatinib, which is a frontline TKI similar to sorafenib, but we then have phase 3 data with cabozantinib and regorafenib after prior sorafenib. Even with the cabozantinib data, we had 25% of patients who were in the third-line setting there. Can you review some of the data and explain how you think about incorporating these things and adverse effect profiles? What about cancer biology? How does the biology of these drugs come into play?

Add to the list the non-TKI, which is ramucirumab and the AFP [alpha-fetoprotein] level of more or equal to 400 ng/mL. At first glance, the landscape looks confusing. However, it’s gotten quite simplified. We often, as oncologists and physicians overall, try to complicate our lives, thinking about this more in terms of ACE [angiotensin-converting enzyme] inhibitors, and we have a million, how do you pick? My life has gotten easier. First of all, I think all TKIs are created equal until proven otherwise, yet no one has proved that any of these TKIs, including lenvatinib vs sorafenib, is better than the other. It’s choice, it does not matter as much. That’s 1 point. The second is the question of ramucirumab, to eliminate the non-TKI, it has become irrelevant. If you’re exposed to bevacizumab, I’m not going to waste your time on ramucirumab anymore. It’s a dud. It’s a dumb question.

Let me challenge you on that for a minute.

Well, you'll challenge me. You go with a VEGF monoclonal antibody, and you then go to another that hits the receptor. I’m not sure if you’re going to see much value out of it. It was barely active following a TKI. Yes, you get the floor.

There are 2 things, quickly, because I don’t want to get distracted. Isn’t there a precedent in colon cancer with ramucirumab after bevacizumab? Can you comment on that? You see a lot more colon cancer cases.

Yes, there is a precedent that shows that it’s barely meaningful when added to chemotherapy, not as a standalone treatment. That’s 1 point. But if you look at it, it doesn’t enhance response. It barely moves the survival by a notch. That’s not on its own; it has to work with something else, not on its own. The answer is maybe, but is it worth wasting the patient’s time on a little maybe?

I appreciate that point. Ramucirumab is well-tolerated, and for patients who have been on a TKI, it has a different adverse effect profile. I want to call out, and we’ll get back to the TKI discussion.

No, that’s different. Ramucirumab is…we’re looking at the landscape and the second-line setting.

I want to mention some data from ESMO [the European Society for Medical Oncology annual meeting] quickly. We [with the University of California, Los Angeles Health] presented a poster at ESMO where we’re looking at ramucirumab after drugs other than sorafenib. This is an expansion phase to the REACH-2 study, and it’s ongoing. We presented 24 patients, and many of them had atezolizumab/bevacizumab. There was a subset of patients who had atezolizumab/bevacizumab as well as IO, as well as lenvatinib. At least in the overall population, the PFS [progression-free survival] was fairly similar to that we saw in REACH, and the safety was certainly the same. I understand what you’re saying. We’re looking for a different mechanism, so that may be where the TKIs come in.

Alright. That question is not going to be settled anytime soon. Looking at the TKIs, most of these TKIs have similar activity across the board. However, here are the problems that I have with lenvatinib and sorafenib. In addition to what Peter was mentioning, they’re only being looked at in the first-line setting. With the data post-VEGF, assuming most of these agents hit some level of VEGF, the data for lenvatinib and sorafenib post other VEGF-directed therapy is lacking. We don’t know if they have much relevance in the second-line setting. In addition, we have 2 agents that have proven activity post-sorafenib in the second-line setting.

The patient has been somewhat exposed to a VEGF inhibitor plus other TKIs, and they have activity. That’s regorafenib and cabozantinib. Regorafenib and cabozantinib would be the natural choices because the data are there for them in the second-line setting. Albeit, it’s not post-atezolizumab/bevacizumab, but one can still argue that they at least have second-line data, post some semblance of a VEGF inhibitor in the first-line setting. The other thing I would say for cabozantinib, and I actually end up favoring cabozantinib based mostly on mechanism of action vs regorafenib, is because it hits MET and AXL, which are both known to be targets that rise in the resistance for VEGF inhibitors.

At least with bevacizumab in the first-line line setting, there is a bit of a theoretical rationale for cabozantinib. Frankly, if you pan out beyond HCC [hepatocellular carcinoma] and go to the world of RCC [renal cell carcinoma], that seems to be one of the mechanisms that supports the utilization of cabozantinib in this setting. In all frankness, they all are reasonable options. We’re sometimes wasting a lot of energy discussing which I want to do, even if it’s minus 1, or plus 1, or 0. 

At the end of the day, we know that most of these patients, once they fail the first-line treatment, about half of them will not see the second-line treatment anyway. The argument stands for the second-line treatment, and for those who fail the second-line treatment, incredibly few will ever see the third-line setting. The discussion will focus on those 2 agents in my mind: cabozantinib and regorafenib. For the reasons I’ve stated, I favor cabozantinib post-atezolizumab/bevacizumab .

I have a different take, though there is obvious expert opinion here: it’s not right or wrong. I would approach it with atezolizumab/bevacizumab as maybe line 0, and we then have front-line TKIs sorafenib or lenvatinib. To your point, we then have second-line TKIs, cabozantinib or regorafenib. Pierre, could you give us what the survey says? What are your thoughts?

There are 2 rational approaches, as I understand them. There is the amnesia approach, which is to say that I give atezolizumab/bevacizumab. I hope for the best, and then I start sequencing from there onward as if that combination therapy did not exist. I start with first-line therapy, then a second-line set of approved therapies, etc. That seems to be a fairly popular approach. The other is to follow what Tony just mentioned, which is to say that there are drugs established in the first-line setting, and there are drugs that are established in the second-line setting, following sorafenib. I’m going to use those based on that prerogative.

That is honestly my favorite approach. I would use atezolizumab/bevacizumab as a first-line treatment, and I would use cabozantinib or regorafenib as the second-line therapy. Perhaps in a small select group of patients for whom I want to spare significant adverse events, I might use ramucirumab in the subset of patients with a high AFP, and I would take it from there. That would be my rational approach to this. I would definitely not use a sequence that includes IO followed by IO. I will preface this by saying that I’m not a medical oncologist, unlike this illustrious group of people here. My understanding is that IO followed by IO has not usually been a very fruitful approach. Please correct me if I’m wrong; I’m speaking out of my league here.

We’ll get to that. To wrap up, if we look at the data, all the TKIs have similar adverse effect profiles. We knew this from cabozantinib and the CELESTIAL study of cabozantinib vs placebo. We improved overall survival. We improved PFS with mostly single-digit response rates. Cabozantinib differentiates itself, as mentioned, because it hits not only VEGF but AXL and MET as well. Regorafenib was looked at in the second-line setting vs placebo after patients who had tolerated prior sorafenib. It has a significant hazard ratio in the 0.69 range for reduction in the risk of death as well as progression.

We did this with low objective responses, and we saw at ESMO recently some real-world data with regorafenib. That showed that its use in the real world from a registry study that the dosing, adverse effects, and disease control look similar to what was seen in the RESORCE trial.

Second-Line Options After IMbrave150 Regimen

 Pierre you brought up an important idea, and Tony gave us some thoughts of using ramucirumab after prior bevacizumab. Katie, is there a role for IO [immune-oncology] after IO? We learned about ipilimumab/nivolumab in the second-line setting after prior sorafenib, and we now have lenvatinib/pembrolizumab. You presented data on durvalumab/tremelimumab, though it’s not approved.

It’s a tricky question. How do we handle IO monotherapy in the later lines of therapy now, or the nivolumab/ipilimumab? After atezolizumab/bevacizumab, we don’t yet have much level 1 evidence to work with. I agree with the comments earlier that cabozantinib had the most permissive phase 3 design in the CELESTIAL trial and has rationale there. In making the decision, there are a couple of considerations to factor in.

For atezolizumab/bevacizumab, one that we haven’t yet talked about is this report of a substantial 27% of patients having anti-drug antibodies [ADA] to atezolizumab, which don’t seem to occur with such frequency when compared to some of the other agents. They all have some potential for this. We don’t yet have a clinical test, but there is some potential concern that patients may have a lack of efficacy owing to ADAs. This is theoretical, and we’re hand-waving because we can’t test for them in the clinic yet, but it does raise the consideration for a patient who tolerated atezolizumab/bevacizumab but progressed quickly and then received a second-line TKI [tyrosine kinase inhibitor] agent, whether it was lenvatinib, cabozantinib, or whichever was chosen, and they were still fit for a third-line therapy.

It’s not unreasonable to consider nivolumab/ipilimumab as a combination in a fit patient who doesn’t have another treatment option and had progressed on atezolizumab/bevacizumab. There is that consideration under the premise that perhaps there was an interfering antibody that prohibited response. In the patients who haven’t received first-line atezolizumab/bevacizumab and received a first-line TKI either because they started treatment pre-atezolizumab/bevacizumab or because they had a contraindication at the time but have now removed that contraindication, there was also the question.

There is also a good discussion to be had about whether to go down the road of a second-line TKI, third-line agent like ramucirumab vs switching to an IO agent. This is where there’s some anecdote or clinical instinct that may or may not be valid in terms of how well the patient did on their first-line TKI. I have patients who progressed quickly on a TKI, so I might be inclined to try nivolumab or pembrolizumab in the second-line setting or nivolumab/ipilimumab if they’re fit rather than doing a second-line TKI. This is based on the gestalt of response to the first-line regimen, though we don’t have great data for this.

Another consideration, to bring up something that was presented at ESMO [the European Society for Medical Oncology Virtual Congress], is that we’ve seen recent trends that PD-L1 tumor expression seemed to have a trend toward an improved response rate with nivolumab and pembrolizumab and atezolizumab/bevacizumab. It’s preliminary, we don’t have that data yet, but at for least nivolumab and pembrolizumab, there is a meta-analysis from ESMO this year that suggests that the rate of response was almost 2-fold for patients with PD-L1 positive tumor expression, in a meta-analysis of about 900 patients. I’m not advocating to go out and start checking PD-L1 in all of our patients’ tumors, but it does raise the question of whether we can study this more prospectively.

Immunotherapy Monotherapy in Later-Line Setting

 When we talked before, Pierre mentioned that you don’t need a biopsy to make the diagnosis. Generally, most patients don’t get biopsies. I very rarely find myself sending Foundation Medicine, or some other genomic profiling, for these patients, or some peripheral ctDNA [circulating tumor DNA] because it hasn’t had an impact on treatment decisions. We talk about 

, and those are certainly very uncommon in liver cancer. To your point, PD-L1 expression is not included as an entry criterion in any of our phase 3 studies because in the phase 2 studies, it hasn’t correlated.

This ESMO [European Society for Medical Oncology Virtual Congress] data were not was based on patient-level data as much as population data. But it was certainly hypothesis-generating, as well as this idea that etiology might play a role in response to PD-1 inhibition. There have been trends about hepatitis B and hepatitis C maybe doing better than those that are nonviral, but we’re not denying patients a certain treatment based on those criteria. The 1 drug we have that has had a biomarker has been ramucirumab, and that’s elevated alpha-fetoprotein [AFP]. Peter, you’ve been extensively involved in that program with me for some time. Is there still a role for ramucirumab in high AFP patients in the modern era? What’s your take?

I believe there is a role. To quote Tony, “Ramucirumab is dead;” I think it’s still alive, and there are 3 reasons for it. First of all, it has been demonstrated in that trial to be effective with a hazard ratio of 0.7. Secondly, it is not the same as bevacizumab: it is tackling the ligand with bevacizumab and the receptor with ramucirumab. That might be quite a distinct mode of action. If you think about it, the single-agent activity for bevacizumab is quite disappointing, so it’s the combination that matters.

In that setting, I would say that is still alive and that we need to see what is coming up. There is an urgent need, as you rightly point out Rich, that biomarkers need to be developed. It’s quite disappointing that, in a field where we now have signatures that are clearly prognostic that have been around for about 20 years or so, that they have never made it into the clinic and turned out to be predictive, which is what we need for all patients. They don’t want to know how long they’ll live; they want to be helped, and this is by prediction of outcome. Here is neglect, and it’s explained by neglect of biopsy, tissue, assessment, and upfront assessment in clinical trials. That’s something that we need to get much better at in the future. Some other aspects to mention where there is a lot of uncertainty and for which we should be open would be the TKI [tyrosine kinase inhibitor] issue in addition to whatever you are using, and that includes the IO [immune-oncology] treatment.

It is immune modulation when we get away from single-agent activity. With single-agent activity, here I believe Tony is right: both are created equal, but if you combine them, then you do something to the immune system, to the microenvironment, where we have a lot of uncertainties and unknowns. That is probably very distinct. Here, there are some theoretical arguments for cabozantinib as a combination or partner. What happens with ramucirumab as a combination or partner is completely unknown.

 Those are some important ideas to build upon in regard to combinations. To wrap up the AFP issue, when we looked at the regorafenib data set, AFP didn’t seem to make a difference in the second-line population. Katie, just briefly, you also did some interesting AFP analyses with the cabozantinib data set.

Yes, we looked at the question of baseline AFP and saw that cabozantinib had benefit across baseline whether elevated greater than 400 ng/mL, or 200 to 400 ng/mL, or even less than 20 ng/mL. There were benefits in terms of the primary end point and secondary end points for the use of cabozantinib. What we also looked at was using AFP as a response marker, not just for the question of whether it change prognosis. I should backtrack and say that cabozantinib showed symmetric benefit over placebo in all of those AFP cut points that we looked at, but patients with high AFP greater than 400 ng/mL certainly have a worse prognosis, so we know that already. That was borne out in the CELESTIAL trial as well, though cabozantinib benefitted those patients equally. I’m also, like Pierre, an AFP aficionado: AFP has other value, especially in tumors that are difficult to image in many cases. What we looked at is patients with evaluable AFP with a baseline level greater than 20 ng/mL and with serial time points measured, and these were blinded results at symmetric time points on a clinical trial.

Patients who had an AFP that did not go up greater than 0 ng/mL, which was if patients had AFP that was static such that it did not increase from baseline, or better yet decreased at whatever cut point we looked at such as a 20% or 50% decrease, those patients had a substantially better PFS [progression-free survival] and OS [overall survival] than patients who had a progression of AFP of greater than 20% or 50%. The point is that AFP can also be used as a surrogate response biomarker in the carefully selected patients who have an evaluable AFP at baseline and have serial time points to look at.

 Yes. AFP is certainly prognostic. I don’t think we’re at the level of PSA [prostate-specific antigen] and using AFP as end points, but it is something reassuring to see, an AFP that isn’t rapidly rising, or staying the same, or even going down. With some of these novel combinations, we see things drop fairly quickly. I’ve now had a collection of patients with AFP in the thousands who normalize. To think that we’re seeing that when we never had drugs that did that for so long.

High AFP as Deciding Factor for Treatment

 Tony, as we start to wrap things up, there are some new drugs that we heard about in the past few months and at ASCO [the American Society of Clinical Oncology annual meeting]. One of them is apatinib, it’s a small molecule, and there was a study done in China of apatinib vs sorafenib. Can you comment on that study?

What you’re referring to is the second-line study with placebo vs apatinib, but there are also studies moving into the first-line setting. There is also dovitinib. Apatinib it is not a new drug to us. It’s been looked at in gastric and colorectal cancer. In fact, it had positive data in China that, unfortunately, were not duplicated in the Western population. It was negative in the Western population. Looking at the data, it looks similar to what you would expect for an agent in this class. Apatinib is similar to lenvatinib, regorafenib, and cabozantinib, so it’s in that class of agents. The results, unsurprisingly, showed positive elements. I’m not sure how much it will add value to our landscape, in all frankness.

That’s part of the limitation of how to take these data. Especially since it was done in China only. That’s not to say that there’s anything wrong with that; it is just to say that other studies with apatinib that were done in China only were not able to replicate the same results when moved to a larger scale in different diseases. Even if it is positive, if it moves to a Western patient population, I’m not sure it’s going to add much value to the landscape we have.

Yes. With China, certainly in liver cancer, the biology is different. A majority of it is hepatitis B vs our hepatitis C and NASH [nonalcoholic steatohepatitis]. You corrected me: there was a newer small molecule also from China that went head-to-head vs sorafenib at ASCO, and it was the first TKI [tyrosine kinase inhibitor] to beat sorafenib. Again, it was in a Chinese population.

Donafenib. You’re right, it’s donafenib. It’s again similar to the TKIs that showed superiority to sorafenib, but it is unfortunately limited to China only. The same problem is that the difference was small, and it is unlikely, at least historically, to beat an IO [immune-oncology] combination. What we’re seeing is that at this point, a single agent, as everyone has referred to, may not be the way to continue moving forward in this disease. Combination seems to be the better pathway even when you fail the first IO doublet in the first-line setting.

There are certainly, as Katie mentioned and others did as well, a lot of doublets that are being looked at in the first-line setting. Unfortunately, there are not as many in later lines of therapy, and that is when we need them. With that said, there is only 1 doublet that is exciting to us if the data pan out, and that’s the PD-1 plus CTLA-4 inhibitor, or PD-L1 plus CTLA-4 inhibitor, because of the specific issue that we just discussed with patients who may not be eligible for VEGF inhibitors or those who have issues.

If those data are positive, that’s the only one that ends up with value in the first-line setting. I want to touch a bit on the biomarker issue. We are so limited on biomarkers, and there was a study from Chinese patients that looked at biomarkers associated with hyperprogression. It has limited data; it’s limited in its scope but large in its number of patients who were included, primarily looking at alterations such as 

 and others. It seems that there may be hints, and that’s to Peter’s point: what we’ve had all these years are hints here and there about biomarkers that may have relevance but we have never consolidated the effort to the point where we integrated them into all of these. Think about how many phase 3 studies have been conducted in this space, yet we have no clue what the heck we’re doing.

All we can think about is alpha-fetoprotein [AFP]. No offense to my hepatologists. AFP is important, but if that’s the only biomarker I can get, I am not a happy dog. One more thing, the anti-drug antibodies [ADAs] are important, but they’re not good determinants of negating the activity. It’s neutralizing antibodies that are more relevant than anti-drug antibodies. In fact, anti-drug antibodies tend to correlate with a higher response rate, which could be a surrogate for antitumor activity.

The problem with atezolizumab is that it has, as Katie mentioned, the highest reported rate of anti-drug antibodies, but they have not reported in any study the neutralizing antibodies. Looking at the history, it seems that neutralizing antibodies account for about 10% of ADAs. If we do the math, it would be about 3% to 4%, just guessing, of patients on atezolizumab who will have neutralizing antibodies, and those may be the patients who may not benefit as well from these agents.

Developing Trial Regimens for Chinese Populations

Peter brought up a lot of interesting ideas about combinations, and I think combinations are thefuture. Katie, as long as we have your attention, the landscape in the next several years—and I’ll tell you I was asked this in 2007 or 2008—what is the future of liver cancer? We have sorafenib. I said, “It’s the golden age. We’re going to have all of these positive phase 3 studies.” I was off by about a decade. What’s your take?

 That’s a great question. It’s good to have more options. HCC [hepatocellular carcinoma] is such a heterogeneous patient population when you look at it from underlying organ function, degree of liver dysfunction, to etiology, the viral milieu may certainly impact the immune microenvironment in response to immunotherapy, to tumor factors. We obviously talked about PD-L1 and AFP [alpha-fetoprotein], but we don’t yet have other tumor genetic biomarkers to work with in HCC, with some early data for FGF19 and FGFR4 targets; that’s really preliminary so far. Having more drugs to choose from early on to tailor according to comorbidity and safety profile is one change in the landscape on the horizon that is a net benefit as patients are receiving more lines of therapy.

This is especially so as we start earlier with level 1 data into the intermediate space instead of waiting until patients are too advanced to receive more than 1 or 2 lines of therapy. Patients will receive more lines of therapy, and they will start earlier when they fail liver-directed therapy earlier. In terms of the combinations, we need to double down on biomarker efforts to try to understand both clinical biomarkers, like viral status, as well as tumor biomarkers. The big categories of combinations are going to be the CTLA-4 and PD-1 or PD-L1 vs the antiangiogenic combinations, and trying to discern which patients are best for which. That would be my answer.

At ASCO [the American Society of Clinical Oncology annual meeting], we saw some novel data of a triplet: nivolumab/ipilimumab/cabozantinib. My recollection of that data is that cabozantinib didn’t add that much. Is that how you rate that?

 Yes. It gets hard to flush out component contributions when you get into triplets like this, and there is certainly appreciable toxicity both from nivolumab/ipilimumab, which has a 50% immune-related toxicity requiring steroid rate, and with cabozantinib and PD-1 or PD-L1. You’re right that, in that small cohort, it didn’t show detrimental toxicity or benefit. In a small, approximately 30 patient cohort, it didn’t discriminate itself efficacy-wise beyond either doublet. It’ll be interesting to see how the COSMIC-312 data look with cabozantinib/atezolizumab because again, cabozantinib has a promising immunomodulatory mechanism, and the renal cell data support this as well. At first, we’ll start with the doublets. Before moving into triplets, it might behoove us to get more biomarker data to guide us. It will be hard to interpret those outcomes for triplets.

 Peter, another combination that has only come about with mature data fairly recently is regorafenib/pembrolizumab. Regorafenib has been looking for a PD-1 partner for some time. Can you tell us how that data look compared to the other TKI [tyrosine kinase inhibitor] combinations?

The analysis is still a bit early and is not yet mature. We have investigated 35 patients so far. First of all, it does not have an unexpected toxicity, although the dosing is going toward a low dose. The maximum tolerated dose was not 160 mg, which would be a standard dose, but it was 120 mg. In fact, this combination will be developed at a lower dose, probably in the range of 90 mg per day. It is coming to the point that, in comparison to single-agent therapy where we have a tumor cyto-effect, here we talk about immunomodulation, and we need to recalculate our dosing. What we saw with respect to efficacy was a disease control rate of 90%, which is very promising. That needs to be developed further.

 It’s an interesting idea, though, that maybe we don’t need the full dose of the TKIs to get the immunomodulatory effect. That might go for all the drugs, and we’ll have to continue to follow that story as we get the phase 3 data or mature data with that regimen.

Camrelizumab is a drug we didn’t talk about, but there were data at ESMO [the European Society for Medical Oncology Virtual Congress] in combination with apatinib. That looks again to Tony’s point about how many TKI/IO [immune-oncology] combinations do we need if they all look the same? The other thing though is that camrelizumab is being looked at in a phase 3 study that’s completed accrual vs sorafenib. This is an interesting study because it’s a noninferiority study. Unlike CheckMate 459, which was negative because it was a superiority trial, this might provide phase 3, level 1 evidence for use of single-agent IO in the frontline setting. This was a global study, and we’re waiting for the readout from that. This has been an amazing discussion.

There is a lot of interest in moving these combinations to earlier lines of treatment. There are a number of studies looking at these combinations with TACE [transarterial chemoembolization] after surgery. There is a study of cabozantinib and TACE. There is EMERALD-1, which is a phase 3 study of durvalumab, bevacizumab, and TACE, and there is the phase 3 LEAP-012 study, which is lenvatinib, pembrolizumab, and TACE. There’s EMERALD-2, which is durvalumab and bevacizumab after resection. There is a pembrolizumab after resection. There’s the CheckMate 9DX study, which is nivolumab after resection. There is truly an immense academic effort in liver cancer, and it’s exciting to see because this is not the rarest disease. It has a high morbidity and high mortality, and thanks to all of your efforts, we’re starting to see some light on the horizon for our patients.

It’s certainly an exciting time for our patients with liver cancer. Having been in the space for some time, it’s exciting to be able to sit down with a patient and tell them all the great options we have that are improving survival and that we can expect responses to their tumor. There is always a lot of exciting research to look forward to, not only in the frontline setting with all the combinations we talked about, but also moving these drugs to earlier lines of disease treatment.

With that in mind, I’d like to thank all 4 of you for your participation in this lively discussion. It was a great event, and I hope our audience found it very educational as well. Thank you very much for your attention.

Evolving HCC Treatment Landscape: Combination Therapy

Hepatocellular Carcinoma: A Rapidly Evolving Treatment Landscape

Discussion around topics pertaining to the use of systemic therapy in advanced liver cancer. Topics will include the latest research in the field and the impact of recent clinical trials on making decisions around treatment selection.

Hepatocellular Carcinoma: A Rapidly Evolving Treatment Landscape | Peer Exchange

EP 1: The Complexity of HCC Multidisciplinary Management

EP 2: Screening Vs Diagnosis: Assessing Liver Function

EP 3: The BCLC Staging System

EP 4: The TACE Approach for Intermediate-Stage Disease

EP 5: Optimal HCC Frontline Treatments

EP 6: Choosing a Frontline Therapy

EP 7: NCCN Guidelines: Nivolumab in Frontline Setting

EP 8: Importance of Endoscopy in Newly Diagnosed HCC

EP 9: Promising Ongoing Frontline Trials

EP 10: Second-Line Options After IMbrave150 Regimen

EP 11: Immunotherapy Monotherapy in Later-Line Setting

EP 12: High AFP as Deciding Factor for Treatment

EP 13: Developing Trial Regimens for Chinese Populations

EP 14: Evolving HCC Treatment Landscape: Combination Therapy