Dr. Navarro, Dr. Rubin, SIG members and guests, I want to thank you for the opportunity to present this talk entitled, Dose Duration, Drug Characteristics as Risk Factors for Liver Injury. Again, I am Craig Lambert. I'm an associate professor of medicine at Indiana University. I have no disclosures to provide. I'd like to take you a little bit on a story focused on these objectives and just briefly touch on the overview and explanation of drug-induced liver injury, also be identified as DILI for the rest of the talk, with direct focus on idiosyncratic DILI as well. We'll look at downstream consequences of drug metabolism that contribute to DILI, as well as identify key drug factors that have been associated with DILI, at least to date. We'll look at pieces such as dose, lipophilicity, solubility, as well as duration, not forgetting though to consider drug-drug interactions in the setting of polypharmacy, metabolism, and also the generation of reactive metabolites. So DILI, we would all probably agree, is a very important clinical problem for the treating hepatologist and some would say most diagnostically challenging of all liver-related conditions. So why would that be? Well, it may be the actual unexpected or expected variable phenotypes and outcomes that come from DILI. Generally, it is rare and also we generally believe that this has not necessarily been dose-related. However, more on that later. We know DILI can be a great masquerader and mimic all acute and chronic liver disease, but really it is important that we identify these compounds in the pre-marketing era for whether they are safe or liable, just for patient safety, and in fact, in clinical or animal models that are associated with bringing drug to market, we know that 45% of these trials are unpredictable based on animal models. Therefore, it is a challenge to both clinical and basic research scientists as we start to embark to try to dissect this mosaic of DILI risk, which has generally been unknown. We know environmental factors, host factors, and drug characteristics make up part of this Venn diagram of explanation of drug-induced liver injury risk. Current lines of evidence suggest that environmental factors are important and, in fact, this is only the beginning in understanding these, but alcohol, tobacco, and diet, just to name a few, are likely critically important, as well as the interaction with particular host factors such as genetic risks, age, and even microbiome or chronic liver disease. Ultimately, though, drug characteristics, which we will focus on the rest of the talk, are critically important as we have seen from prior studies. A prerequisite for intrinsic and idiosyncratic DILI is the metabolism of drugs in the liver for which drug characteristics are likely important. In the liver, the generation of reactant metabolites, which leads to initial consequences, such as covalent binding of proteins, increased oxidative stress, activation of signal transduction pathways like C and terminal June kinase, as well as increased organelle stress, such as in the endoplasmic reticulum, mitochondria as well, but also perturbations in the conicular export systems, such as the multidrug-resistant proteins, as well as biosalt-X-borne pumps are probably important in ultimately establishing cellular death via necrosis or apoptosis, but also even sublethal stress, which can provide a co-stimulatory signal via the innate immune system. Finally, as Dr. Stoltz had talked about, the adaptive immune response is also likely critically important as well. A really important paradigm is that idiosyncratic drug reactions are independent of dose. Now, such reactions may appear to be independent of dose, primarily because most patients do not have idiosyncratic drug reactions, and the usual dose range is usually quite narrow for the compounds that we have, yet even with these confounding factors, a relationship between the dose and the incidence of idiosyncratic drug reactions have been observed. Incidence of the lupus syndrome induced by hydralazine was determined in the longitudinal study in approximately 281 patients. Data on the duration of treatment and the maximum dose achieved were reviewed as well. We saw patients that took zero or took 50 milligrams had no formation of the lupus syndrome over the course of three years, whereas 5.4% that took 100 milligrams and 10.4% with 200 milligrams were observed. A very clinically hepatology important observation was made by Dr. Utrecht in 1999 when he noticed that less than 10 milligram daily dose was, if ever, associated with rare daily, but also compounds more than 50 milligrams were commonly linked with black box warnings for hepatotoxicity. This gets me to a meta-analysis that was just completed in the past few years that looked at worldwide compounds removed for adverse drug reactions. Between 1950 and 2014, there were 81 compounds worldwide that were removed for hepatotoxicity. This is a list of 17 compounds that have been taken away from 1915 to 2014 in the U.S. On the left, the compound is listed, year withdrawn, and daily doses on the right. As you can see, almost all of these compounds have averaged daily doses well above 50 milligrams per day. Now the relationship between daily dose of oral medications and idiosyncratic dilly was first clinically clarified in 2008 by a very bright second year medical student, obviously though guided by his mentor. We targeted two pharmaceutical databases of top brand name and generic prescriptions in the U.S. We then annotated these based on the average daily dose, less than 10 milligrams, 10 to 49 milligrams, and more than 50 milligrams per day. And what we found? We found doses with more than 50 milligrams seemed to be at higher risk for liver failure, drug-induced liver injury-related death, as well as transplant. Dr. Bjornsson, a co-author in this paper, took this list and also annotated his Swedish dilly death and transplant cases from 66 to 02, as well as the Swedish dilly jaundice cases up into 2004. What we see here is the same example of increased frequency of dilly cases among doses more than 50 milligrams, whereas only 9% were present in the less than 10 milligram group. Also, so there was no difference in the pattern of injury and also noted the adverse liver outcomes were also linked with high dose. This was then codified in the Russo cohort as well of U.S. dilly cases and transplant from 1990 to 2002, whereas 91% of those compounds were occurring at average daily doses more than 50 milligrams. A question remained after this epidemiologic study, though. Is idiosyncratic dilly associated with higher dosage or is it the ovary representation of drugs associated with idiosyncratic dilly in the more than 50 milligram per day group? As you can see on this slide from the dilly network perspective study, drug class is critically important and antimicrobials are clearly the most frequent cause of dilly in the United States. Another paradigm became clear from the study. There are many drugs that are given at high dose that have minimal, if any, risk of hepatotoxicity. Thus dose alone is not a reliable means of assessing dilly risk. Next to dose, lipophilicity also will be identified as a log P value is an important physicochemical property that affects cellular uptakes and ADMET or absorption, distribution, metabolism, excretion, and possible toxicity. Log P value also is known measure of liver or the compound's affinity for lipids and is determined by the partitioning of drugs between actinol and water. Lines of evidence also showed increased risk of off-drug target binding in compounds with high lipophilicity to date. Lipophilicity and daily dose were examined as risk factor for dilly using two independent drug database labels with presence of absence or presence of liver injury. Each compound was assigned to most, less, and no dilly concern based on risk of dilly from withdrawn status or assignment of a black box or warning from liver toxicity. Lipophilicity was calculated using atomic-based prediction modeling. Drugs were assigned to groups between less than one, one to three, and more than three. In prior literature, it's shown that the appropriate lipophilicity for drugs are typically in the range of one to three. This figure shows the distribution of compounds according to daily dose and log P. The most concern are shown with pink circles and no dilly concern are with green triangles. Forty-four most concern dilly appeared in the upper right quadrant and only two no dilly concern appeared in this region. Thus, daily dose more than 50 milligrams and log P more than three was strongly associated with dilly concern with an odds ratio of 14. Interestingly, drug dose less than 100 milligrams per day and log P less than three had the lowest dilly concern with an odds ratio of 0.33. A Cochrane-Armitage test was employed to assess the statistical significance of relationship between log P, daily dose, and the risk for dilly. A summary of the prevalence of the most dilly concern drugs for individual subgroups are found in this table. Individual investigators observed that 96% to 92% and 65% were most dilly concern drugs with log P of either more than three, three to one, or less than one, respectively. At daily doses less than 100 milligrams, interestingly, no statistical significant relationship between log P and dilly existed. In the study, the authors also noted that a significant relationship between the extent of hepatic metabolism and log P was present and this made sense because lipophilic drugs are cleared by the liver and generally require biotransformation to be eliminated. Therefore, log P may simply be a surrogate for extensive biotransformation as well as hepatocyte exposure to reactive metabolites. In order to better understand metabolism risk then, the extent of hepatic metabolism as a risk factor for dilly was examined by us as well in 2010 using a relatively arbitrary definition of hepatic metabolism where more than 50% of hepatically metabolized compounds were identified as significant or non-significant, less than 50%. This definition was applied to drugs from a previously published database as highlighted in 2005. We observed compounds with significant hepatic metabolism had increased report of more than three times upper limit normal ALT, liver failure, and fatal dilly compared to those without significant metabolism. Further, there was no relationship between frequency of hepatic adverse events and metabolism via phase one or phase two reactions. We also observed an additive effect of dilly dose in hepatic metabolism. The right column includes drugs with extensive metabolism according to average dilly dose. The left column includes drugs with non-extensive hepatic metabolism according to average dilly dose and what we see is oral compounds with significant hepatic metabolism, but also given at daily doses more than 50 milligrams, had the highest risk of hepatic adverse drug reactions compared to other groups. The field has continued to seek better models beyond these drug factors such as dose, lipophilicity, and metabolism in order to improve dilly prediction. Covalent binding of proteins by reactive metabolites of drug is a frequently cited mechanism and has been shown to cause liver injury through direct toxicity or in the endorsement of immune reactions. However, the reactive metabolites role in dilly has been relatively inconclusive, partly due to the fact that protein adducts seen with drug are not necessarily always associated with liver injury. In this 2016 study, authors interrogated four publicly available drug databases, providing 479 oral and FDA approved compounds and looked to hope to better understand dilly risk. After the assessment of nearly 354 reactive metabolite data drugs, they compiled it in this table and this highlights the odds of dilly per rules listed in the first column. As you can see, 83% of most dilly concerned drugs and 21% of no dilly drugs had discoverable reactive metabolites and in fact, reactive metabolite formation alone was insufficiently powered per the authors to address dilly prediction, but also had a relatively low specificity. However, adding reactive metabolites to the previously identified rule of two improved the prediction considerably as reactive metabolite addition correctly identified three false positives for what was most dilly concerned. Subsequently, the association between dose, log P, and reactive metabolite formation and dilly risk was assessed by logistic regression. The multivariate regression suggested all three parameters contributed significantly to this model and they used the coefficients to calculate a dilly score. The dilly score was then used to interrogate three published cohorts in order to assess its prediction according to human hepatotoxicity. Because daily dose is not considered systemic exposure, the authors also investigate maximum concentration of a drug as an appropriate measure for daily dose. Excuse me. It can be seen in panes B, D, and F. Oops. And what they found was among all these groups, these three independently were able to predict dilly, but accordingly the dilly score itself predicted well and identified the differences among the cases with high risk of dilly and no risk of dilly. Yet these properties of drug may be highly related as drugs with higher lipophilicity typically result in an increased permeability and therefore uptake by hepatocytes. Therefore, these compounds may be more likely to undergo biotransformation. Conversely, drugs with low lipophilicity will have low permeability and undergo less hepatic metabolism. It could be at lower risk of dilly. However, some drugs will have low permeability, but higher absorption. And this is likely due to the presence of intestinal transporters as identified here. Thus, collaborators within the dilly network sought to consider dilly drug characteristics according to the biopharmaceutics drug disposition classification system, also known as the BDDSC. The BDDCS characterizes drugs according to metabolism and solubility. Importantly, BDDCS is not impacted by drug. Using dill in cases, the authors observed and concluded that among approximately 100 drugs with BDCS available, class 2 compounds that are at higher risk of dilly, but the singular value of this classification was not higher than the dilly predictors on their own. Further, the groups were different according to latency, as well as the pattern of injury. And you can see here that class 1 had the highest proportion of hepatocellular injury in the longest latency period. Also interesting is that they found class 4 compounds with low solubility and poor metabolism can also be associated with dilly. Idiosyncratic dilly with prolonged latency usually reflects the inadaptive immune attack. This is consistent with the finding that it is typical for the liver injury to occur promptly if the patient is re-challenged with an offending drug. The prolonged latency to initial onset can be in part attributed to the time required for antigen-specific lymphocytes to be activated and proliferate to the numbers needed to mediate a dilly event. Many examples in the literature of instances of dilly with short or less than seven days or after long latency have been reviewed, but systemic evaluation hadn't been completed until this paper by the Dilly Network as well in 2015. Although no relatively strong demographic or select characteristic risks were found, the authors observed that classic autoimmune dilly drugs like nitrofarnitone and minocycline were two dominant causes of dilly with long latency. But a number of other meds were represented such as 6-mercaptopurine, statins, as well as amiodarone. However, quantitative systems pharmacology modeling has also made a strong case that some idiosyncratic dilly-causing drugs can have prolonged latency without involving the adaptive immune system. Idiosyncratic dilly associated with troglitazone was the first example of this where modeling based solely on alterations of bile acid homeostasis accurately predicted not only the incidence of serum ALT elevations, but also the latency to peak. The latency in the model resulted from several factors including the gradual accumulation of sulfate metabolites within hepatocytes, but also the compensatory regulation of bile acid transporters through FXR and the fact that the mechanism of the bile salt export pump inhibition was competitive as opposed to non-competitive. In patients that are polymedicated prior to the dilly episode, it is often possible to determine the causative agent just because of our understanding of temporal relationships, as well as symptom development. However, we should keep in mind that concomitant medications are not always innocent bystanders, but can impact dilly susceptibility through drug-drug interactions. Concomitant drugs are capable of modulating the metabolism of other drugs and this could be via induction, inhibition, or substrate competition among the SIP reactions. Alterations of proportion of drug metabolized by other minor pathways could also be affected, which induces more cellular stress, results as increased dilly risk beyond that drug just on its own. And one of the classic examples is rifampicin, which we know is a SIP inducer and has been demonstrated to increase the risk of dilly when given together with isoniazine for antituberculosis treatment. In conclusion, dilly remains a result of complex interplay among drug-specific characteristics, host factors, as well as environmental variables. There is unmet need to predict risk for dilly more reliably. ALT increases in clinical trials is no longer an adequate approach to survey for dilly in pre-marketing period. Therefore, vigilance in development of agents to identify entities that have entirely safe and which have potentially liable prior to human application is critical. Now using compound-specific risks such as dose, lipophilicity, extensive liver metabolism, and formation of reactive metabolites are just a few aspects of drugs and screening. And we know safe drugs with these actually do exist. And therefore, modeling dilly risks, such as what has been done by Dr. Watkins with dilly sim, remains important as we can use comprehensive exposure in vitro and clinical data to better predict and model dilly aspects, including risks, mechanisms, and dosing paradigms. Thank you.

Dr. Craig Lambert

drug-induced liver injury

DILI risk factors

reactive metabolites

predictive models