Welcome to the AASLD liver cancer debrief. My name is David Kaplan. I'm from the University of Pennsylvania. I will be the chair of the liver cancer SIGFOR 2024-2025. I want to thank the organizers for inviting me to present this HCC debrief. These are my disclosures. Again, I would like to thank the SPC for inviting me to share the HCC debrief. I want to thank all of the presenters who shared slides with me. I apologize in advance that due to the sheer volume of phenomenal science and limited time, many of the slides I was provided unfortunately will not be discussed in this session. Liver cancer metastasized all over TLM this year with a phenomenal symposium on systemic therapy that spilled over into two overflow rooms. A patient program focusing on living with liver cancer. Two abstract plenaries. A clinical research breakout. The second annual liver cancer shark tank. Again, standing room only for the second year in a row. One by Dr. Andrew Moon. More than 250 abstracts related to liver cancer presented at the meeting. And the piece de resistance, a mesmerizing state-of-the-art lecture given by Dr. Lewis Roberts on the state of personalized medicine and liver cancer. Liver cancer continues to grow at TLM despite novel therapies. For the purpose of this debrief, I'm going to focus only on abstracts and posters, and I broke these down into the following themes. First, risk stratification for screening patients with liver for liver cancer and at-risk populations. Second, improving diagnostic accuracy in indeterminate lesions. Third, optimizing local regional therapy for early and intermediate stage disease. Fourth, providing an update on systemic therapy, and in particular, new evidence relating to personalizing systemic therapy. And additional evidence on predicting response to systemic therapy based on pre-treatment and post-treatment predictors. Finally, I will wrap up with a discussion of some of the key abstracts on liver transplantation in hepatocellular carcinoma. First, risk stratification for liver cancer screening. The key questions in liver cancer screening currently are, what are the optimal strategies to screen, and how do we increase uptake? We're dealing with a new era of liver cancer screening in which the incidence rates for patients with viral hepatitis have been reduced but not eliminated by effective antiviral therapy. We also have a relatively low risk population among patients with non-serotic NAFLD who still may develop liver cancer and may merit screening in certain cases. However, screening patients with low incidence rates makes screening less cost-effective. And a key question is, can we identify subsets of patients within these relatively low incidence populations to focus our screening efforts? The first abstract I'm going to present asks the question, can we find sophisticated serological markers to identify risk of liver cancer among low-risk populations for screening? In this abstract from Fujiwara et al., a group from Southwestern, Mie, Mass General, Keio, Hiroshima, Tokyo, and Taiwan derived a 170-gene prognostic liver signature from 85 patients who had had an SVR-12 with antiviral therapy and who had undergone curative treatment for post-SVR HCC. This risk signature was validated then in a separate set of 39 patients who had not had HCC after SVR and were followed up over 12 years, in which individuals only with the high risk by the signature developed hepatocellular carcinoma during extended follow-up. The investigators then took this and developed a six-gene secreted signature called PLS-Seq HCV Cure, which was validated in two independent cohorts of liver cancer-naive patients, 164 patients of whom 41 developed HCC, as well as an HCC-experienced population of 146 patients. They also assessed whether or not this model could be clinically useful as a plug-in to refine a previously defined secretome-based signature that was etiology-agnostic-included alpha-fetoprotein. What they showed is that individuals that either had their etiology-agnostic PLS-Seq AFP signature or the PLS-Seq HCV Cure signature both were at fourfold increased risk of developing hepatocellular carcinoma during follow-up. However, if both of these were high risk, there was a synergistic effect and a 15-fold increase. This is early validation that serum markers can be used to enrich the screening population, and we're looking forward to additional work from this group and further validation and development in prospective studies. We're very used to stratifying our population for liver cancer screening based on clinical markers such as FIB4, but maybe there are smarter ways to do this. Lin Yip and a multinational consortium from Hong Kong, South Korea, and Bordeaux used machine learning to identify models to better predict liver cancer occurrence in a post-SVR population. The machine learning models were trained on 5,155 adult patients with various chronic liver diseases from Korea and then validated prospectively in two cohorts from Hong Kong and Europe, each with over 2,000 patients. Multiple parameters were fed into these models, and it was found, not surprisingly, that liver stiffness by VCTE ranked most important among the nine predictors. With this, they used their machine learning to develop a model that ultimately had a concordance score of 0.89 in the Hong Kong cohort and 0.91 in the French validation cohort with ROC scores for predicting five-year HCC over 0.89 in both validation cohorts. This is an excellent model. This model was called SMART-HCC, and they found that SMART-HCC outperformed other prediction models like AMAP or the Toronto HCC Risk Index and seven hepatitis B related risk scores. Using a cutoff of 0.43, they found that 82.7% and 89% of the Hong Kong and Europe validation cohorts respectively were classified as low-risk for liver cancer and could possibly be exempted from HCC surveillance with an annual incidence rate between 0.1% and 0.19%. By contrast, the high-risk group had an annual HCC incidence of 1.91% and 2.63% in both of these groups respectively and clearly would meet screening criteria. This work shows a readily available clinical variables fed into machine learning-informed models can give us excellent discrimination, particularly for identifying very low-risk patients who may not need liver cancer screening. And again, SMART-HCC is aptly named. Well, not all of us might be ready to start using machine learning models right away, but maybe we can use simpler approaches to stratify our population. In this study, Lee et al. representing a multi-center Korean consortium retrospectively assessed the predictive power of multiple risk scores, including MPageB, which was a risk score originally derived for chronic hepatitis B to predict liver cancer incidence. This was validated in 1,839 post-SVR patients, followed for a median of 2.8 years. And in this cohort, 1.5% of patients developed liver cancer. They found that the MPageB score was highly discriminative for identifying the highest-risk patients by using the MPageB greater than 13. Very few low or intermediate-risk patients in their cohort developed hepatocellular carcinoma during follow-up. Now, to recall the MPageB score, which is largely based on age, gender, and signs of liver dysfunction, to get a score of 13, one would have to have fairly severe liver dysfunction, where screening would be fairly obviously needed, or be older, generally male with mild dysfunction. And these are clinical predictors that we can start to put into our practice right away. The key point is that among low-risk HCV SVR patients, older patients, particularly men with subtle signs of liver dysfunction still should be screened for HCC due to elevated risks. To summarize this section, sophisticated serum markers or machine learning models promise to really precisely tailor our screening for hepatocellular carcinoma, although we're not quite there yet. I think the greatest benefit is that these will identify patients who could be exempted from screening due to very low risk to maximize the effectiveness and cost-effectiveness of screening. In the meantime, existing risk stratification tools like MPageB or other scores like FIV4 still should be used. I think a key point is that screening doesn't work unless it's initiated and maintained, and health systems really need to learn how to implement screening effectively. So we're going to move on and talk about some abstracts that dealt with improving the diagnostic yield of imaging tests in hepatocellular carcinoma. As you all know, many diagnostic imaging tests yield indeterminate findings, either indeterminate Lyraids 3 or 4 lesions or Lyraids M that are malignant but can't be characterized as either HCC or cholangic carcinoma. Has there been progress on modalities to improve the determination of the true clinical status of these nodules? We got a abstract from Kono et al. from UC San Diego that suggests that contrast-enhanced ultrasound really may fit the bill for this problem. This was a collaboration from 11 international sites following 646 patients who are undergoing liver cancer surveillance. Of the 646, 75 were found to either have Lyraids 3 indeterminate or LRM, not fully characterizable, malignant lesions, and underwent contrast-enhanced ultrasound to assess its additional discriminatory capacity. Eight lesions that were identified from this pool as Lyraids 1 and 2 were subsequently confirmed to be benign, so a very good negative predictive value. 39 lesions still remained indeterminate after CEUS. Eight of the lesions were upgraded to Lyraids 4 by ultrasound, of which 50% were ultimately confirmed to be HCC, and 13 were upgraded to LR5, all of which were HCC. CEUS even identified tumor in vein that had been missed by MRI. Overall, this was 95% specific and had a high positive predictive value. 85% of indeterminate cases were characterized in the true direction by the addition of CEUS, and CEUS was felt to have a clinical impact in about 37% of cases. And so this is additional evidence that this modality might help in these challenging cases. Well, what about dealing with Lyraids M lesions? So in this work from Saltzman, Kornick, and colleagues from Bond, they asked a question of whether or not they could use small extracellular vesicles as a marker for the type of malignancy in Lyraids M lesions. This was a small study of 24 patients in whom they had serum and had LRM observations, and they found three different markers were differentially expressed on cholangic carcinoma-related extracellular vesicles, CD9, CD63, and CD81. And when they looked at these three markers in isolation, none were sensitive or specific enough to be used clinically. But when combined into a five-point score, the discrimination area under the curve was 0.85. And so these data suggest that looking at extracellular vesicles in addition to other technologies might be able to improve our non-invasive diagnosis of Lyraids M lesions. So to summarize, we continue to struggle with indeterminate lesions, but CUS and serum biomarkers might get us closer to defining lesions that remain indeterminate after contrast enhanced CT or MRI. So I'm now going to shift and talk about some abstracts that dealt with optimization of local regional therapy. There have been some interest in using vitamin K for its antiproliferative effects on liver cancer for several years. Vitamin K suppresses cellular proliferation through an AKT and MEK-ERK pathway. It also inhibits HDAC6, which can affect apoptosis and autophagy in hepatocellular carcinoma. Vitamin K, its canonical effect is to gamma carboxylate pre-prothrombin. And when that does not happen, that leads to the production of descarboxyprothrombin, or DCP. It's been shown that DCP not only is a biomarker of hepatocellular carcinoma, but through its effects on MET and other receptors actually may have pro-tumorogenic effects. These potential salutary effects of vitamin K were tested in this prospective study in combination with transarterial chemoembolization by Haruna and Yakushijin. In this study, 101 patients undergoing TACE were randomized to either vitamin K, 45 milligrams a day, given on the day of TACE and for the next 28 days, and study for the outcomes of overall response rate and progression-free survival. As you can see on the right, the Kaplan-Meier curve, that there was a significant increase in the PFS from 146 to 262 days, as well as increase in overall response from 82 to 96%. I think what these data highlight is that in transarterial embolization, the main effect is ischemia, and that we have poor data to support the use of chemotherapy both in vitro and in vivo. Efforts should probably be focusing more on potentiation of ischemia after embolization, inhibiting the stress response, and using better adjuvants. Vitamin K appeared very safe in this and appears very effective, and further validation appears appropriate. Now a significant concern with all embolotherapy for HCC is that it may negatively impact liver function, and other modalities have been touted as potentially being safer to the liver and preserving liver function for longer. This was tested in a retrospective study by Sugata and colleagues from Fukui, where they evaluated two cohorts of patients, one that had undergone proton beam therapy, and the other than had undergone TACE and RFA. Now, these were relatively similar populations, but it was not a randomized trial. All the patients were trialed, Pew A and B, with three or fewer tumors, and none had uncontrolled ascites, and none had extra hepatic metastases. Those groups were similar, although there was a slightly reduced frequency of BCLC0 disease in the patients that got TACE, and it's not clear in this study if the TACE was low bar or selective. But what they found was that in the proton beam therapy versus TACE, that there was no difference in initial, in progression-free survival, although there was a very large difference in overall survival of the two groups. In corollary with this difference in survival, what they found was that while the TACE plus RFA group had increases in ALBE score immediately after their initial treatment that was sustained during follow-up, the patients that had undergone proton beam therapy had no statistically significant changes in ALBE score over their follow-up time. And these data suggest that proton beam therapy is safe. It seems to have, it's highly effective, and these are tantalizing data that it might preserve hepatic function better than embolotherapy in unresectable HCC, requiring further external validation and prospective randomized studies. Now in the United States, Yttrium-90 is rapidly supplanting chemoembolization due to better progression-free survival, even though it has much higher cost. I think the data from the vitamin K study suggests that there may be adjuvant approaches to transarterial embolization, which might mitigate those differences at a lower cost and should be explored. We have to be cognizant that these transarterial therapies may be more damaging to the background liver than we previously appreciated. This may encourage us to increase our utilization of targeted external radiotherapy, as more and more data arise that these may be similarly effective to local ablation therapies done percutaneously. It also suggests that there may be a role for earlier application of systemic therapy before we cause too much liver damage, which is an area of debate within the field. I'm gonna move on to talking about personalizing systemic therapy. Everyone watching this debrief is well aware of the massive expansion of effective therapies over the last 15 years that we've observed for hepatocellular carcinoma with multiple first and second line therapies and many new regimens in phase three studies for both the first and second line. Right now, our primary approved systemic therapies used in the first line are dual combination therapies, either using atelizumab and bevacizumab or tremolimumab induction followed by dervalumab. At the meeting, Katie Kelly from UCSF presented updated survival data from the phase three Himalaya study that evaluated single dose tremolimumab plus dervalumab versus serafinib in unresectable HCC. The initial trial had reported long-term survival to 36 months, and this is the first study to report out to 48 months. What they showed is that 25% of the patients randomized to the STRIDE regimen remained alive at 48 months compared to 15% in the serafinib arm. Notably, many more patients from the serafinib arm had crossed over to immunotherapy than STRIDE patients had crossed over to second line, suggesting that most of the survival benefit in the STRIDE arm was from the STRIDE, whereas some of the survival benefit in serafinib was from second line therapies. The initial response to the tremolimumab, dervalumab combination therapy was highly predictive of long-term survival in Himalaya. All patients who had a complete response had long-term survival, and 2 3rds of patients who had partial response and 1 1 4th of patients who had best response of stable disease had long-term survival, yielding an overall 36% long-term survival rate in patients who had initial disease control, and this was significantly higher than was observed in the serafinib arm. These data are really novel and exciting and tell us that long-term survival is something that we really can start to think about in our patients with hepatocellular carcinoma, which even five, 10 years ago was something that we really couldn't really imagine. So I just want to remind everyone of the Mbrave 150 results that showed that atezolizumab and bevacizumab were superior to serafinib in patients with unresectable HCC, both in terms of overall survival and progression-free survival. And while these results are striking, many are asking what can be done to further improve on these outcomes. The first question we can ask is whether or not patients who have a good response to atezolizumab and bevacizumab can be advanced to more aggressive therapies to improve outcome. This abstract by Professor Kudo's group in Japan showed that select patients with unresectable HCC who are treated with atezolizumab and bevacizumab who have complete or partial response can be advanced to curative therapies or more aggressive palliative therapies. In this study, they had 110 patients that they followed. 38 of those patients had either a complete or partial response. These patients were then selected either for resection, ablation, TACE, or continued on atezobev. And what they found is that of those 38 patients, 25 remained HCC-free after the curative treatments and 24 weeks off of drug. Of those 38 patients who were advanced to what was considered curative therapy, there was a five-year overall survival of 100%. And obviously, these are selected patients, but it tells us that for patients who have early, partial, or complete responses that being aggressive with local therapies may have significant long-term benefits. The next question is whether or not adjuvant therapy given at the time of atezolizumab plus bevacizumab might have benefit in certain subsets of patients. And these two groups from Korea looked at patients with portal vein thrombosis to see if there was a benefit. The first group shown on the left here from Lee et al. found no improvement of progression-free survival with the addition of radiotherapy to atezolizumab and bevacizumab. The second group by Wong et al. in a cohort of 163 patients found that there was no overall improvement with radiotherapy, but that a certain subset of patients that they deemed high-risk due to what they described as huge tumors or bile duct invasion or VP4 seemed to have a significant increase in survival. And so what these data tell us is that we really need to learn more. More prospective studies need to be done and randomized trials really need to be done to identify patients that might benefit from adjuvant radiotherapy with atezolizumab and bevacizumab. Another question that comes up with systemic therapies is what should be done when patients progress? And there's been a lot of discussion about what second-line therapy should look like. So this abstract from Hirooka et al. from AHIME looked at Linvatinib and other molecular targeted agents as second-line therapy after progression from atezolizumab and bevacizumab. This was a study of 101 patients. They were predominantly male, age around 72, but predominantly with Child-Pu-A cirrhosis. They found that overall, the median progression-free survival in Linvatinib was 4.4 months, and this was similar to other MTAs. The median overall survival was an impressive 15.7 months. And among the patients who had Child-Pu-A, the median overall survival was not reached. There didn't seem to be any difference with the other molecular target agents, which included serafinib or imiserumab or kavazatinib in this study. And the adverse effects were generally those that would be expected from Linvatinib. So this is adding to the literature about possible second-line options after atezolizumab, bevacizumab-related progression. So Tabuchi and colleagues asked a different question. They asked whether or not if patients progress on atezolizumab and bevacizumab, is there any role for continuing atezolizumab and bevacizumab versus switching to other therapies? Do recall that in ENBREG 150, radiologic progression was a reason for therapy to be stopped. So in this retrospective study of 59 patients that had progressive disease, they decided to continue treatment in 22 patients, whereas 24 patients went on to other second-line therapies and 13 were put in palliative care. I want to focus on the middle graph here, comparing continuing atezolizumab, bevacizumab versus other molecular-targeted agents, and there really was no difference in overall survival. And these data suggest that perhaps stopping at radiologic progression is not the appropriate endpoint with checkpoint inhibitor combination therapy, and that perhaps a different endpoint like treating until symptomatic progression, which was the endpoint in the SHARP study, might be more appropriate. But it muddies the water in terms of what to do as second-line therapy with progressive disease because it may be to stay on first-line therapy. I think this abstract from Lemesin et al. from France gives us a window into what the future of our practice will be like. In this study, patients who had progression on atezolizumab and bevacizumab underwent next-generation sequencing to identify potential genetic targets for drug therapy to be used in the second line. They did genomic analysis on 20 patients with analyzable result on 19, and in these, 2 3rds of them had an actionable genomic alteration. And for those, they were able to find targeted therapy for nine of them. The patients in this were predominantly HCC, although there was some mixed cholangiohepatomas. 10 of the samples had been collected prior to atezobev and 10 after. The therapies that they were able to apply included trastuzumab, aloparib, trometinib, everolimus, palpacyclib, and aloparib, and with meaningful responses seen in these individuals. And this is proof of concept that next-generation sequencing obtained early on can be particularly impactful in the second line setting. So to summarize, we're now talking about long-term survival in unrecyclable liver cancer. Some patients who have early responses may be candidates for conversion to curative approaches. We'll need some global criteria to really define what that means, but there's promising efforts ongoing. Radiotherapy may have a benefit as adjuvant therapy in select patients, but again, the population that benefits needs to be better defined. What is the optimal second line after progressive disease on combination IO therapy is a key question in the field that's being actively addressed. But again, is PD a reason to stop the therapy? Maybe symptomatic progressive disease is a better endpoint for treatment. And the future is personalized therapies based on genetic sequencing, and it appears that these might impact care in up to 60% of patients. Access to this therapy may in the future be challenging. Can we do a better job of predicting response to systemic therapy? This is an important area because only 30% of patients with unrecyclable HCC respond to dual checkpoint or CPI-Bev therapy. We really need to have pretreatment identification of those that are going to respond so that we can personalize our approaches. And early post-treatment predictors of response might facilitate earlier conversion to second line therapies. So first, can we predict which patients early on in therapy are responding? In this single center study from UC San Francisco, Lee et al. reported on the association of changes in AFP in early treatment response in 173 patients who had been treated with immune checkpoint inhibitors. These individuals had AFPs prior to treatment and within three months of starting treatment and had been followed since 2016. They found that among responders, AFP declined by 39% compared to an 11% increase seen in non-responders. The degree of reduction in alpha-fetoprotein was associated with the treatment response with an adjusted odds ratio of 1.2 for every 10% reduction in alpha-fetoprotein. They used a cutoff of 10% early AFP improvement and showed that this was associated with an improvement in overall survival as shown on the left here. And those with not surprisingly greater reductions as shown on the right up to those with an improvement of over 50% had even greater survival. The next abstract by Tanabe and colleagues from Japan specifically explored AFP response in individuals with high AFP and DCP in patients who were not AFP producers. This retrospective included 154 patients who had received a Datizolizumab and Bevacizumab by December of 2022. They found among individuals with an AFP greater than 20, an AFP decline by greater than or equal to 30% by three weeks after treatment was associated with a 5.5-fold odds of overall response and increased progression-free survival. Conversely, individuals that had extra hepatic spread and a 30% increase in AFP within three weeks were associated with early progressive disease. Among 73 AFP-low patients, those that did not have AFP-producing tumors, they found that a baseline DCP level less than 40 was an independent risk factor for objective response. However, changes in DCP did not further predict outcomes. This work that's been subsequently predicted in cancers in 2023 is starting to suggest that there may be early stopping rules. These were within three weeks of starting therapy, which might identify activity and or futility of a Datizolizumab and Bevacizumab, and in the setting of futility may indicate a benefit to moving towards second-line therapy earlier. What would be particularly helpful is if data from a pre-treatment biopsy, or even better, is from serum assays, would inform the choice of therapy at the outset. Sarah Kappans from Joseph Lovitz's group used SSRNA-Seq from 253 patients who were treated with a Datizolizumab and Bevacizumab in the context of the EMBRAVE-150 study and GO30140 clinical trial. And they used this to evaluate signatures of clinical response to these therapies. Overall, in these studies, 30% of patients were considered responders. They defined three clusters of patients, those with high expression of immunological markers that responded to therapy, patients without these markers who still responded, and non-responders. What they found is that among the immune positive group that were responders, the SSRNA-Seq was enriched with signatures associated with genes, associated with activated CD8 T cells, as well as recruitment chemokines like CXCL10. And these patients might be expected to have a good response to atezolizumab based on the anti-PD1 effect. Among patients who were immune negative responders, they found an enrichment for VEGF expression and a reduction in NRP1, which is a negative regulator of VEGF. Again, these data are interesting in that these patients, because of the signature, might be thought to be more likely to respond to bevacizumab. What they found is that individuals who had both of these signatures, both an immune positive signature and were NRP1 low, had the best overall outcomes in these cohorts. Patients that were immune high, immune positive, or NRP1 low, actually did fairly similarly. And patients who had neither of these markers had the poorest survival. These data suggest that atezolizumab are additive in this situation, addressing that one drug targets one population of patients whose tumors are immune sensitive, the other drug targeting those that are really VEGF dependent, and may explain the responses seen with this combination. A future state could be conceived of in which these data inform the use of initial therapy using the combination of checkpoint inhibitor and anti-VEGF for individuals who have immune positive or immune negative with the NRP1 low, and using either a completely different therapy or similar therapy with a third agent in the 70% of patients who are unlikely to respond to atezolizumab and bevacizumab. Some exciting data also came in out of Mount Sinai. This was a MRI-based study that suggests that MRI may be able to identify the patients who have these immune positive tumors for selection for systemic therapy. This was a study of 30 patients, and they found that several parameters of MRI were associated with the high frequency of tumor infiltrating lymphocytes on biopsies of these patients. This included tumor stiffness, what was called mean transit time, arterial flow, total plasma flow, time to peak, and upslope. And as you can see, the area under the curve for a model based on these features was highly discriminative of those that have tumor infiltrating lymphocytes. And it raises the question, in a future state, maybe we can use imaging-based modalities to identify patients who are likely to respond to checkpoint inhibitor therapy prior to initiation. One of the key regulators of immune exclusion in hepatocellular carcinoma is beta-catenin. And Paul Monga's group from Pittsburgh had several studies that they presented using mouse models of beta-catenin-immunity HCC that are beginning to better explain the mechanism of action of beta-catenin. In this work by Evan Delgado and Dr. Larrick, they found that beta-catenin inhibits interferon response factor 2, and it reduces its expression in the setting of enhanced beta-catenin activity. If they rescued IRF2 in that model, there was a reduction in overuse of beta-catenin overall tumor and disease burden, suggesting that IRF2 has anti-proliferative effects, anti-cancer effects in the model. And if they overexpressed IRF2 in that model, there was an increase in the accumulation of effector T cells within the tumors and a depletion of Tregs. And so what they have started to do is better clarify the mechanism with beta-catenin effects suppressive regulatory cells like Tregs and MDSCs within the tumor microenvironment, and actually suggest a potential role for interferon gamma therapy in some of these, although it was not synergistic with beta-catenin knockout in their model. And they also show that the beta-catenin suppresses the development of tertiary lymphoid structures necessary to mount an effective immune response against HCC. Given the critical role of tumor infiltrating lymphocytes in mediating the anti-cancer effect of atezolizumab or other checkpoint inhibitors, it's not surprising that there might be soluble factors in the blood that may help identify patients who are likely to respond. Who are likely to respond, and this work by Odagiri et al showed that in a cohort of patients, 33 patients who underwent atezolizumab and bevacizumab and had pre-treatment samples drawn, that the soluble BAT cell attenuator, BTLA, when that was highly expressed, appeared to be a marker for response to immunotherapy. And these are really exciting data that require further validation in additional cohorts. Lastly, in this section, it's important to recognize that patients with hepatocellular carcinoma have a competing risk, which is death from cirrhosis. And this group from San Francisco used LASSO regression to develop the HCC-AIM score, which stands for Adverse Outcomes in Immunotherapy. This was derived in a cohort of 200 patients, and they were able to identify which patients who were going to get immune checkpoint inhibitor therapy were going to have early hepatic decompensation with a concordance of 0.85, and in a validation cohort was 0.82. So a very good score for identifying patients likely to progress from cirrhosis and those that may not have as great a benefit from immunotherapy. So in the last section, I'm going to talk about a few aspects related to transplant for hepatocellular carcinoma. These data were presented in the MAZLED plenary, but data from Zoverian UC show that by 2022, the most common etiology underlying liver cancer at the time of transplant is MAZLED. And this increasing trend remains significant even after adjusting for age, sex, ethnicity, obesity, and the presence of diabetes. About 20% of patients who are waitlisted for transplant are going to drop out due to disease progression. The group from UCSF also developed a model to evaluate waitlist dropout, and this is called the Biomarker Integrated Dropout Gradient Estimation or BRIDGE score. They used data from four centers, included 507 patients listed for transplant with HCC exception points to derive and then validate the score. In the derivation group, they identified that tumors that required downstaging, AFP, AFPL-3, and DCP, as well as child puberty stage were associated with waitlist dropout. And they were able to develop a simple score, which they found was very predictive of high risk. People with three or more points on the score, as you can see on the right, were very likely to drop out due to progression. The C-index for this model was good at 0.76. And so one can imagine that this score might also be used in the pre-transplant population to select those for whom adjuvant therapy to keep patients from dropping out might be applied. Based on data from the UCSF group and others, UNOS implemented in 2015 a six-month waiting period for mild exception points to allow for the observation of aggressive biology and to allow dropout during waitlisting to reduce the risk of post-transplant recurrence. And Yagen and Mahmoud from the University of Pennsylvania, in this abstract, reviewed the effects of this policy, finding that, indeed, in all models, the risk of post-transplant recurrence was significantly reduced in the post-six-month waiting policy period. And the serially adjusted model suggests that the effect was likely mediated by waitlist selection, increased time to transplant, increased opportunity to perform local-regional therapy, and behavior changes by center selecting patients with more favorable biology to proceed with transplant. At present, our ability to predict post-transplant recurrence is fairly modest. We do have the retreat score. This work from Costentin and colleagues from France presented follow-up from the SILVER study, which was a randomized controlled study evaluating the impact of sirolimus on post-transplant HCC recurrence. That was a negative study. But they used this data to develop what's called the R3-AFP model. This was based on 507 patients. In this cohort, the recurrence rate overall was 18.7%. They identified several factors that were predictive of post-transplant recurrence, including having greater than four nodules, tumor size particularly larger than six centimeters, the presence of vascular invasion, and the nuclear grade differentiation on the explant. Like the retreat score, it includes the last pre-liver transplant AFP value. And they were able to divide their cohort into individuals who were very low risk. This was about 43% of the cohort. Those at low risk, which was 36%, and those at high or very high risk was an additional 21%. One can see on the right that the recurrence rate in those with very high risk was between 36% and 60%, whereas the lowest risk group was 9%. This is still, I think, higher than we would like to see. And this model still includes variables that are only available at the time of explant. And what we really need to see are models in this setting that are completely based on pre-transplant predictors that allow us to avoid transplanting patients that would have these exceedingly high recurrence rates. It's key to important that access to transplant for liver cancer remains uneven in the United States. And these data from Nephew et al further evaluated this in a prospective fashion in three hospitals in Indiana between 2019 and 2022. The study team collected several parameters associated with social burden, including race, income less than $30,000, marital status, Medicaid insurance, a health literacy score called the BRIEF, BHLS score, education less than high school, and alcohol use in the previous 90 days. All of these burdens were identified in the not... All of these burdens were higher and concentrated in the black population in their centers. And not surprisingly, all of these burdens were associated with reduced transplantation rates. What they found is that actually the greatest burden was imposed by health literacy. This had the greatest impact on overall survival. And the work highlights the need to focus resources in underserved communities to improve health literacy during the transplant education process during the waitlist period to try to maximize chances of success with transplant. So I've gone through quite a bit. There was a huge amount of science in hepatocellular carcinoma. I think that the key takeaways here are that we still have work to do on refining and validating our target population for HCC screening and defining the optimum modality. There's still room for improvement in our diagnostic modalities, particularly with indeterminate lesions. Our post-transplant recurrence rates are improving, but we still don't have a completely pre-transplant prediction model. And that really is sorely needed in the field. But on the other hand, long-term survival in liver cancer is really here. We're seeing really a dramatic improvement in long-term outcomes. We have some patients who receive our palliative systemic therapies who can actually be converted to curative therapies, and we're beginning to see some pathways in that direction. The optimal second-line therapy after our initial systemic therapies is still really undefined, and maybe we should be continuing our systemic therapies to a different endpoint. Pre-treatment biopsies, possibly pre-treatment imaging, possibly pre- treatment serum markers are likely to impact our delivery of personalized medicine for liver cancer, and we're seeing some really promising studies in that direction. And we have to remember that disparities in liver cancer care remain and need to be our priority of our efforts in addressing and in leveling the playing field and creating equity in liver cancer access, for access to transplant for liver cancer and overall liver cancer care. So again, I want to thank the Scientific Program Committee, ASLD, for giving me the opportunity to review the excellent science in liver cancer at TLM 2023, and look forward to seeing everyone at TLM 2024.

liver cancer research

AASLD conference

risk stratification

diagnostic accuracy

local regional therapy

personalized systemic therapy

early response markers

machine learning models

post-transplant recurrence