We're going to move on to consider NAFLD. And this debrief will be delivered by Dr. Kathleen Corey. Kathleen is a particularly well-known to me, a member of our division, assistant professor of medicine at Harvard Medical School, and director and founder of our Fatty Liver Clinic. Back before fatty liver clinics were fashionable, Kathleen had the vision to actually begin this many years ago. She's made important contributions to our understanding of NASH biomarkers, as well as the relationship between NAFLD and cardiovascular disease. We're looking forward to your talk, Kathleen. Thanks so much. I didn't realize that NAFLD clinics were fashionable, so that makes me feel even better about what I'm doing. I'd like to say I'm not that happy with Mark, since he really set a high bar to follow. That was a fantastic talk, and hopefully I can do NAFLD some justice. Perfect. These are my disclosures. So first off, we're going to talk about some exciting studies that were shown in this past few days about the diagnosis of NAFLD. And this is one from Wash U, as well as the Dallas Heart Study, that sought to develop a clinical prediction rule, really to be used in the primary care setting, as well as for epidemiologic studies, to determine if patients have non-alcoholic fatty liver disease. So this group did a logistic regression model to predict whether patients had steatosis, as diagnosed with more than 5.5% of steatosis on MRS from the Dallas Heart Study, and did internal validation via tenfold cross-validation. They developed this in a diverse population, so that's certainly a benefit. And what I really like about it is it used variables that we have in our own clinics, and certainly that our primary care colleagues have easy access to. So things like age and BMI, as well as sex and comorbidities. And so these can be entered, and this is available now online, into this. And similar to the NAFLD fibrosis score, you can get a score that predicts whether or not your patients have NAFLD. And you can see here that the area under the curve for the Dallas steatosis index seems to exceed those of other non-invasive scores. And especially in low-risk patients, it's incredibly valuable with a high negative predictive value of 90. The PPV, or positive predictive value, is slightly lower at 69, which is something that we see very commonly in a lot of our NITs, but still valuable. And I think that this is a score, really, that they have developed, the Dallas steatosis index, that is easy to use. It's a clinical prediction tool for NAFLD overall, for who to be screened, or who we can include in epidemiologic studies, that can be used by both patients and providers. And the only caveat I would say is, it requires external validation. So, switching gears a little bit, this is another way to diagnose NAFLD, although a bit more complex. And this is a novel gut microbiome-derived signature for predicting NAFLD cirrhosis. So this group wanted to develop a panel of biomarkers to develop Nash cirrhosis, and they did validate it in several independent cohorts. They first started out using a cohort of 163 individuals with Nash cirrhosis from UCSD, and they performed metagenomics and metabolomics on them, and identified 19 discriminatory species, which did show a robust diagnostic accuracy, especially when albumin, serum albumin, was included. Interestingly then, they profiled these with metabolites from the gut microbiome, so 19 species and 17 metabolites. And as you can see here in the orange line, with those 36 combined, they had an area under the curve of 0.93 to distinguish Nash cirrhosis, and they validated this in multiple cohorts across multiple different geographic areas, suggesting that it's valid in all different populations. So what we found here then is really a very novel, I think, interesting stool and gut microbe-based signature that accurately predicts Nash cirrhosis, is well-validated, and may give us some really interesting insight into the mechanisms of the microbiome in Nash cirrhosis. Moving on again, and my husband told me when I was practicing, I said moving on way too much, so I'm gonna try to back that down now, but I apologize in advance. We're gonna talk about the NAFLD simulator, which was something that was led by Jag Chatwal at MGH, and that I was privileged to be a part of. And this is really different than anything else that's out there, it's an active simulator that's already up and running, that I've already used in clinic, but they were fortunate enough to present it here at the liver meeting. It's an open access interactive tool for long-term risk prediction for NAFLD, and it has been validated with large observational studies. And very simply, similar to what we saw with the dallasteatosis index, users, either patients or providers, can enter in simple demographics, age, sex, and you do wanna have information on whether a patient has NAFLD, or has Nash with some degree of fibrosis, and it generates a really comprehensive report that looks something like this. This report not only gives you 10 year survival based on your fibrosis stage, it also breaks down survival to talk about 10 year liver-related mortality, and importantly, it thinks about competing risk, which is, I think, something that's really key. As we know, most of our patients will pass away from cardiovascular disease or non-liver-related cancers, and so we get a 10 year non-liver-related mortality as well, as well as a 10 year risk of decompensated cirrhosis and hepatocellular carcinoma. And this undergoes, as we get increasing data, increasing refinement. The hope is to be adding treatments as those become available, but I think it allows us to have really important discussions with our patients about the risks of their liver disease, decompensation, and non-liver-related mortality. So this is an easy-to-use clinical predictor that goes beyond just thinking about cirrhosis or decompensation, but survival overall. It's used for all phases of NAFLD, and it's already ready for prime time. Now switching gears a little bit, this is machine learning in NASH histology. And what our pathology colleagues, I have seen, been asked a lot is, when is all of this fancy artificial intelligence going to take over and start reading biopsies? And what this group sought to do was ask, can a machine learning model interpret NASH histology? So they used, this is a company that's in Boston, Path AI, has a research platform that applied to train a convolutional neural network, which I promise you I know all the details about and can tell you about later, but they used this with more than 68,000 annotations set from 75 pathologists on more than 640 images to develop a platform and a model. And their training set was from Stellar IV using 644 biopsies. These were patients screened, so it could have all degrees of fibrosis and NASH. And their validation set was 161, and they were read by three pathologists. And not surprisingly, let's see if this pointer works, we saw when they looked, only for me. I can see the mouse. Well, as you can see here, up on the far right, when you look at how the pathologists are agreeing, as we know, the agreements are low. So we're seeing low agreements in ballooning, in lobular inflammation, somewhat higher ballooning in ESHC fibrosis score. I don't think that's, oh, there we go, thank you. So not great, but what we expect. And when they did a mean of these correlations and combined them to the machine learning platform, we saw similar results indicating a degree of accuracy, where they had high correlation with the pathology reads, and even higher with steatosis. But I think more interesting, for me at least, was the fact that this machine learning program can provide really quantitation of some of these variables. And they showed an example here for ballooning. So as we know, ballooning is usually scored as a zero, a one, or a two, and it's a fairly blunt score. And as we saw here, you can see some disagreements about what stage patients fall into. Well, when you do it in a machine learning approach, you get a very precise quantitation of exactly how much ballooning there is. So these two patients, both were graded two, but this one has 8% ballooning, and this has 2.8% ballooning. Now, we don't know yet if that matters, if that's gonna change treatment, if that's gonna change outcome, but at least it provides us with more data that we can evaluate. And they did show us one example where their algorithm measured portal and lobular inflammation and looked at the ratio and found that this ratio was associated with an increased risk of disease progression. So machine learning for liver histology in this setting did show good concordance with pathologists. It did provide granular data, and it may identify novel findings relevant to clinical care. So then moving on to one of my favorite topics and a difficult topic, lifestyle intervention and weight loss for NAFLD. And I thought this was a fantastic and very bold study that this group from Australia did, looking at a modified very low energy diet, or VLED, to induce ketosis in patients with and without NAFLD. And a modified very low energy diet is one where you have two meal replacements and then a meal with very low carbohydrates, a regular meal as your third meal of the day. This was a retrospective cohort done in a specialized weight loss center, and fatty liver was defined by the fatty liver index greater than 60, and just to reinforce things we've already talked about, this is a place where you could use the NAFLD steatosis index potentially in the future. Their outcomes, which are pretty standard, are at least a 5% total body weight loss from baseline at three months, and then fortunate enough to see data at this meeting for six months. 213 individuals were included, over half had NAFLD by the fatty liver index. And what they found was that those with NAFLD, a significantly lower percentage, 49%, were able to achieve this 5% total body weight loss compared to 67% of those without NAFLD. And this persisted at six months, although it was non-significant. But the trend is certainly there of 61% versus 75%. And on multivariate analysis, only the fatty liver index was associated with a failure to achieve outcomes. So I thought this was really interesting, one, because there may be something about our patients with non-alcoholic fatty liver disease who have different weight homeostasis than those with fatty liver, and we need to be able to understand that. But as somebody who tries every day to get my patients to lose weight, even a 50% significant weight loss at three months was something that I was certainly excited about, albeit less than non-NAFLD patients. And I think it's impressive, and for highly motivated patients in a specialized setting, this is an option. The questions, of course, are what are the impact on histology, and what are the issues of weight regain? But certainly I think there's more to come in this area. This is a complex study that looked at weight loss improvement of NITs associated with collagen degradation profiles in NAFLD, and I'm really gonna focus on the relationship between NITs and collagen degradation profiles. So 200 patients with NAFLD were included, 50 were by FibroScan, and 150 by liver biopsy, and they had baseline and 15-month serum, and underwent immense testing for a number of matrix metalloproteases, degradation proteins, and fibrosis markers. And the bottom line in terms of markers for this is that only Pro-C3, so a marker of fibrogenesis, differentiated NASH from non-NASH, differentiated F01 from F2 and F34 in those who had undergone biopsy, and differentiated fibrosis stage by FibroScan. And as you can see here, the change in Pro-C3 correlated independently with an improvement in the Fib4 score, showing that Pro-C3 may be a marker of fibrosis improvement in these individuals. So overall, and we've certainly all heard more about Pro-C3, but in this study, it did distinguish NASH from non-NASH, could determine fibrosis stage, and can really help us understand what's going on with fibrosis and fibrogenesis in our patients who are undergoing, in this case, weight loss. Another thing that I think I'm increasingly asked about is bariatric endoscopy, and where are we there? And so this study sought to evaluate the efficacy and safety of duodenal mucosal resurfacing on NASH. So DMR is a mucosal ablative procedure leading to mucosal hyperplasia that impacts metabolic signaling in the duodenum, akin to what we think happens in Roux-en-Y gastric bypass. This was a phase two sham-controlled double-blind trial in sub-optimally controlled diabetics who were overweight or obese, and the primary endpoint was a 12-week decrease in fat by MRI-PDFF. What they found was that those who underwent DMR had a significant decrease in MRI fat at 12 weeks, compared to those who underwent sham at 24 weeks. They also had a significant decrease in A1c and in weight, with no serious adverse events reported. So for this, I would say DMR certainly has a place, hopefully, in the future of NAFLD. It did lead to a decrease in MRI fat compared to sham, as well as a decrease in weight and A1c. So it may be a potential future RX for weight in NAFLD, especially in our patients who have contraindications or are hesitant to undergo bariatric surgery. This study, really, I thought was fascinating. I kicked myself a little bit because I've thought about doing this study a while, but they scooped us in Paris, and good for them. They really wanted to look at, for our patients who undergo bariatric surgery who have F3 or F4 fibrosis, how many of them persist and how many of them improve? So they had 181 who underwent, they had 181 with advanced NASH, and they defined this as significant fibrosis, F3 or F4, or in SAF grade, three or four with F0 to F2 fibrosis. And for the purpose of this discussion, we're really gonna focus on, and what was the focus of the abstract was those with significant fibrosis. 66 of these 181 patients underwent follow-up liver biopsy at six years, and the main finding was that significant fibrosis persisted in 45% of these patients, despite weight loss of 24 kilos, which is akin to those who had resolution, improvement in IR, a normalization of their ALT, a decrease in their anaphylactic activity score, and NASH resolution. Folks who I would have said if they had come to my clinic, oh, you're gonna do great, you've had such fantastic metabolic improvement, of course your fibrosis will improve. They did see that sleeve gastrectomy as opposed to Roux-en-Y was independently associated with persistent fibrosis, as was older age. The good news is 19 of these 66 did have an absolutely normal liver biopsy at follow-up. So in a large percentage of patients, bariatric surgery can work. These patients tended to have a shorter disease duration, lower diabetes prevalence, and increased weight loss. And over 80% of the overall group had NASH resolution. But we really want to be mindful that about half of our patients who have stage three to four will not resolve. And it's something that I've been pushing our surgeons a lot as we're moving towards sleeve gastrectomy, which is a much faster procedure. You can do two Roux-en-Ys in the time of three sleeves, and may have less anastomotic issues. We really are moving towards sleeves, but for our patients who are very metabolically sick, especially with those with NASH and advanced fibrosis, I'm really gonna be pushing because of this for Roux-en-Y gastric bypass over sleeve. So a few insights into NAFLD pathology. This was a great study that came out of UT Southwestern that wanted to look at the benefit of a genetic risk score over just individual genes to predict fatty liver disease. So they looked at six SNPs for five genes, and they looked at 990 patients. They combined NAFLD and AFLD in this group, and they compared them to controls, about 1,900 from the Dallas Heart Study. And they calculated genetic risk score for each subject of these five genes and six alleles. Not surprisingly, P and PLA3 had the highest odds ratio associated with fatty liver of any form. And fatty liver patients had a higher cumulative burden of risk alleles versus controls with a higher mean GRS score of 6.58 versus 5.15, which has a P value that I never, unfortunately, see in my research, but certainly is impressive. And interestingly, they found that the odds of having fatty liver increased by 60%, so an odds ratio of 1.6 for each additional risk allele. So what they found is that, of course, P and PLA3 is associated with a high risk of fatty liver disease, be it alcohol or non-alcohol. And this genetic risk score of five genes, six SNPs, predicts fatty liver better than any single variant does. And it certainly may be of value in future risk prediction. Another interesting insight into the pathophysiology of NAFLD was this study to look at exercise-induced hepatic heat shock cognate 70, a biomarker of exercise and potential therapeutic treatment in NAFLD. And so this group did several things. They looked at male mice who were put in a sedentary cage for 28 days and male mice who were put in a wheeled cage, so active mice for 28 days. Then they did liver RNA-seq, gene ontology, and they identified a candidate protein. And they took this protein and they measured it in NAFLD patients who had been enrolled in an exercise program. So at baseline and then at six months. And they looked at the differences, one, at baseline and then between who had succeeded in the program and who had failed to complete the program. Finally, they incubated this protein in human hepatic stellate cells. And they identified the protein as HSC70, or hepatic heat shock cognate 70. They found that this was significantly higher in those weight-caged mice. They also found that it correlated and increased in patients who successfully completed the exercise program, correlated with increased abdominal skeletal muscle mass. And it also was associated at the end of the study with decreased fibrosis markers and steatosis. And this protein actually failed, decreased in subjects that failed the program. And when they incubated this protein with human hepatic stellate cells, they saw a decrease in TGF-beta-SMAD23 signaling. So this hepatic heat shock cognate 70 may be a biomarker of exercise. And NASH is associated with increased muscle mass and correlated with a decrease in marker of steatosis and fibrosis. And I can tell you if they can make this into a pill form for exercise, my patients will be thrilled. Now pharmacotherapy, I have to apologize. There are so many and it's an exciting time for NASH trials. I did have one slide for every trial and realized we'd be here until about midnight. And so I did have to put these down into one table and I'm highlighting just five of the many really promising trials here. We're looking at one of the important endpoints in general, what percentage of patients in these trials achieved at least a 30% risk reduction in liver fat by MRI-PDFF. So the first is cotetetide, which is a GLP-1 and glucagon receptor dual agonist for 26 weeks. And this was compared to different doses of cotetetide plus liraglutide, which as you remember is just a GLP-1 agonist alone. Now this study did not include, and it's the one I present that does not include MRI, but they did find a decrease over 26 weeks in ALT and AST and non-invasive testing scores when compared to placebo and to liraglutide. They also saw when compared to liraglutide a decreased weight, A1C and lipids. And the adverse events were fairly typical for what we see with, excuse me, the GLP-1s. A lot of GI side effects that usually over eight weeks improve. The next is glycogliflozin, which is an SGLT-1 and 2 inhibitor, so both for intestine and for kidney that was studied over 12 weeks. And over 12 weeks, 67% of those who received active drug had at least a 30% decrease in their liver fat. They decreased their weight, their A1C. No discussion of lipids. It was associated with diarrhea in about 97% of the patients who received active drug. 97%, but this was considered mild. And certainly we're seeing good metabolic effects overall. Tropifexor, which is an FXR agonist that we're all familiar with this type of drug, was tested over 12 weeks. And again, in as little as 12 weeks, we saw a decrease with 65% of people achieving this more than 30% relative reduction in liver fat. They did have a decrease in weight. They had a decrease in insulin resistance. But similar to other FXRs, they did have this concerning, although potentially manageable, increase in LDL and the AE of pruritus. Saraglitazar is a PPR-alpha-delta agonist that was studied for 16 weeks and saw 41% of patients achieving this 30% relative reduction in liver fat. They decreased A1C. They decreased lipids. It was weight neutral. And they did not have any SAEs reported. And finally, PF05221304, a liver-targeted ACCI, was studied for 16 weeks and really had a dramatic decrease in liver fat. 90% of those who received it had a 30% or more relative reduction in liver fat, as well as a decrease in A1C in those with diabetics. But the concern was the lipids. The main adverse event was hypertriglyceridemia, even to levels above 800, which is certainly significant. They did not see a change in LDL, but they saw a rise in ApoB, which always makes me think about whether we're seeing increased LDL particles or small, dense LDL particles that, as you know, are more atherogenic. That's one potential theory, but I certainly think with the cardiovascular risk associated with NAFLD, we need to be working this out. They also saw a transient increase in the ALT and AST at the beginning of drug use, which returned to baseline. So, this is an exciting time. There are many novel mechanisms of actions out there, as well as some known mechanisms of actions with a considerable amount of promise. I mean, just a delay. All right. So, we do have some pharmacotherapies, as to be expected, that didn't achieve their end point, and we thank Stephen Harrison for standing up and explaining all of these to us. We heard about solancertib, an apoptosis signal-related kinase one, or ASK1 inhibitor, that was studied both in StelR3 and StelR4 for stage three and four fibrosis, and unfortunately, as you can see, Stel did not meet its primary end point when compared to placebo in fibrosis improvement, and it did not meet any histologic end points, although they did show target engagement. A question that was raised during the presentation, which is a great one, is how much target engagement do we need, and did we achieve it? And the answer is, we don't know, and so I think that's something to be evaluated, and this may have effect in multidrug therapy. I think that's distinct from emricazan, an oral, or as Dr. Harrison said, emricazan, an oral pan-cast-based inhibitor, which suppresses apoptosis. It did not achieve any of its end points, and actually may have worsened histology, and you can see here that placebo seemed to do considerably better in almost all histologic end points than drug. So while emricazan probably is not gonna be moving on, although I have no insight into that, I do think it provides insight into the pathophysiology and the importance of apoptosis and apoptosis suppression in NASH, and what happens when we suppress, and what is happening in response to that. And I do think, in summary, much can be learned from these very large, well-done, impressive studies. So what about comorbid disease in NASH? And I thought this really was fascinating. I've often thought that my patients, a lot of my patients with NASH suffer from depression, and how much can I motivate them for lifestyle changes if they're suffering from depression? How much does depression play a role in their disease? So this group used the THIN network, the Health Improvement Network, which is a population-based cohort in the UK, and they looked at patients with new-onset depression, and then they waited a three-year washout period and looked for patients with a new diagnosis of NAFLD, alcoholic liver disease, or chronic hepatitis. And what they found, not surprising, but I think really important, was a history of depression, even before liver disease was diagnosed, was significantly and independently associated with a greater risk of alcoholic liver disease and non-alcoholic fatty liver disease. It was not with chronic hepatitis, but in those other two, it was important. Most of the antidepressants did seem to increase the risk of liver disease, and that's certainly something that we're gonna have to look into more, if that's independent or if that's associated with weight gain or other metabolic changes. But it's important to note that atypical antidepressants, like we think of as Welbutrin or Remeron in the United States, lowered the risk of alcoholic liver disease. And SSRIs were antidepressants that were associated with a lower risk of chronic hepatitis. From my clinic, we didn't see antidepressant that was associated with a lower risk of NAFLD, but I do think that there's certainly more here to be studied. So depression is a risk factor for NAFLD and alcoholic fatty liver disease. Atypical antidepressants decrease the risk of alcoholic fatty liver disease, and we really should be studying the impact of treatment of depression on fatty liver outcomes. Another study that I find fascinating was that looking at risk factors for cardiovascular disease in NAFLD. And so this group wanted to look at a group all with NAFLD to see if they could identify risk factors among patients with NAFLD for cardiovascular disease. So they looked at 285 adults with biopsy-proven NAFLD and looked over a mean of 5.2 years, excluding the first six months, for those who developed incident cardiovascular disease. And their main findings were that ALT was an independent predictor on multivariate analysis of cardiovascular disease, and fibrosis stage three or four was an independent predictor of cardiovascular disease. After adjusting for cardiovascular risk scores and after adjusting for traditional risk factors, and when fibrosis was removed from the model and NAFLD fibrosis score was placed in, it too was an independent predictor. So as time ran out, I was not able to capture these, and I thank these authors so much for sharing with me, but I did want to at least mention sarcopenia in NAFLD is increasingly important. We know it's so important in liver disease, but this group found that sarcopenia definitions perform poorly in NAFLD due to concurrent overweight and obesity, and they need a proper assessment. And in their view, and they have shown that proper assessment combines muscle volume and fat infiltration to detect low function. Dr. Villar-Gomez's group showed that ADH1B and alcohol use have an important interaction, and ADH1B2 is associated with decreased NASH and fibrosis severity in abstainers and in drinkers, and in fact, moderate drinkers with the two allele have the lowest rates of NASH and fibrosis. The harms and benefits of alcohol use in NAFLD seem to depend in part on this allele. My patients are gonna be clamoring for them to be tested for this, but I think more work is needed to be done. And finally, we did see for obatocolic acid, a study looking into the value or the harms of pruritus, and they found that while pruritus was very common in their patients receiving OCA, it did not impact patient-reported outcomes at all, which I think is reassuring. So the takeaways, really, I'll do very quickly since I'm over time. To diagnose NAFLD, the Dallas Zeatosis Index has a lot of promise for NASH cirrhosis. Stool-based microbiome signature certainly has promise to predict outcomes. I would use the NAFLD simulator, and we may be seeing machine learning to diagnose NASH on biopsy. For managing weight loss, options include a very low-calorie diet. Duodenal mucosal resurfacing needs more data but is promising. Pro-C3 can monitor fibrosis regression and weight loss, but we need to be careful with our F3s and F4s after bariatric surgery, and we consider doing Roux-en-Y and monitoring them after. Finally, there is a good NAFLD genetic risk score to best predict fatty liver. Heat shock cognate 70 is a biomarker of exercise in NAFLD. All of these drugs are promising NAFLD pharmacotherapies, and while serolonceratib and amrixin did not achieve their endpoints, we have a lot to learn from them. And finally, depression is important in liver disease, and fibrosis is important in cardiovascular disease. And with that, I thank you so much.

NAFLD

Dallas Steatosis Index

stool-based microbiome signature

management strategies

genetic risk scores

pharmacotherapies

fibrosis regression