Risk Assessment for Hepatobiliary Toxicity Liabilities in Drug Development

Abstract

Risk assessment of hepatobiliary toxicities represents one of the greatest challenges and, more often than not, one of the most rewarding activities in which toxicologic pathologists can partake, and often times lead. This is in part because each liver toxicity picture is a bit different, informed by a broad range and diversity of relevant data, and also in part because the heavily relied upon animal models are imperfect regarding predictivity of hepatic effects in humans. Following identification and characterization of a hepatotoxicity hazard, typically in nonclinical toxicology studies, a holistic and integrated assessment of liver-relevant endpoints is conducted that typically incorporates ADME (absorption, distribution, metabolism, and excretion) information (ideally, including extensive transporter data, exposure margins, and possibly concentration of parent/metabolite at region of injury), target expression/function, in silico prediction data, in vitro hepatocyte data, liver/circulating biomarkers, and importantly, species specificity of any of these data. Of course, a thorough understanding, developed in close partnership with clinical colleagues, of the anticipated liver disease status of intended patient populations is paramount to hepatic risk assessment. This is particularly important since the likelihood of translatable determinant hepatic events observed in nonclinical models to occur in humans has been reasonably well established.

Hepatotoxicity continues to be a leading cause of attrition in drug discovery, development, and post-marketing, and importantly is a major contributor to drug-induced safety-related events in patients, not only for small molecules but also for biopharmaceuticals, such as immunomodulatory monoclonal antibodies and therapeutic oligonucleotides.^{2,10,11,17,45}

Fundamental mechanisms of hepatotoxicity across species are generally due to disruption of highly conserved functions, and as such, it is not surprising that many of the most well-studied hepatotoxicants cause liver toxicity in many species, including humans, for example, acetaminophen, α and β amanitin (ingestion of Amanita phalloides), carbon tetrachloride, and aflatoxin B1, and therefore, hepatic findings observed in the nonclinical study package anywhere in discovery or development warrant attention with respect to potential for translatability to patients, recognizing nonclinical liver models represent imperfect predictors of clinical outcomes.^{12,24,26,32,36,38} Classic toxicologic pathology of the liver is often a key component of the initial stages of the hepatotoxicity risk assessment process, namely, hazard identification and characterization, typically occurring in the in vivo toxicology study stages of a project. Given the ultimately limited hepatotoxicity phenotypes that exist, representing disruption of myriad hepatic functional processes, a thorough and accurate description of liver findings is paramount. And of course, any morphologic findings must be carefully integrated with any alterations in serum liver markers that are standard on most toxicology studies. As well, morphologic findings should be assessed in the context of the drug candidate/metabolite exposure data, including, when possible, drug candidate/metabolite levels at the toxicologic site of action (hepatic cell (s)/zone/region).

A wide array of mechanisms underpins hepatotoxic responses ranging from immune mediated mechanisms, to mitochondrial injury, oxidative stress, and disruption of bile acid and other hepatocellular transporters. Each hepatic event often reflects disruption of multiple mechanisms, and as such, it is imperative to fully and carefully describe and characterize the morphology of the liver injury to include all involved hepatic cell types, and any lobular zonality of effects.^27,46 Doing so, enables dose-response analysis and cross-species comparisons, along with comparisons against findings known for other drug/drug candidates with similar chemical structure/moieties and/or therapeutic mechanisms of action.

In general, a risk assessment approach comprises a series of iterative steps typically beginning with hazard identification, followed by hazard characterization, then dose-response analysis, risk characterization, risk communication, and risk management. It is important to define, and for the key team members to be aware of, which component of the risk assessment paradigm one is working in, but to not let the risk assessment process itself overshadow the science, decision-making and project progression. This approach framework is certainly amenable to risk assessment of nonclinical hepatotoxicity findings.

Given the diversity of mechanisms underpinning hepatotoxic events, a team-oriented approach to understanding and resolving these mechanisms is of paramount importance as these various mechanisms, coupled with the approaches and techniques needed to further understand them, typically require unique, deep expertise, and experience to offer complementary insights and interpretation of the totality of the relevant data sets, potentially reflecting a variety of cellular, subcellular, molecular, and biochemical pathways. Although considerable work has been done through efforts of the Society of Toxicologic Pathology (STP) and other pathology organizations focusing on nonclinical pathology, there remains differences in liver findings terminology and the general lexicon when it comes to informal and diagnostic communications.^27,46 As such, communication across the team between various disciplines requires at least a working knowledge of the relevant disciplines, and importantly, a good understanding of key terms. For example, clinical colleagues and hepatologists may use terms, such as “transaminitis,” which simply means elevation in alanine aminotransferase (ALT) and aspartate aminotransferase (AST), or “toxic hepatitis”/“hepatitis” when referring to drug-induced liver injury. “LFTs” is still used and mistakenly refers to ALT, AST +/– bilirubin as liver function tests and should be avoided in nonclinical and clinical settings; it is always preferable, particularly when characterizing or describing a hepatic effect, to specify which hepatic marker(s) one is referring to, or use a collective term that does not imply that changes in transaminases reflect hepatic function. Other terms frequently used in a clinical setting of potential drug-related hepatotoxicity, include “DILI” for drug-induced liver injury, or “iDILI” for idiosyncratic drug-induced liver injury.

An essential member of the hepatotoxicity risk assessment team is the drug metabolism and pharmacokinetics (DMPK) scientist(s), wherein they contribute to the following aspects of the risk assessment; metabolite identification, elucidating, and predicting intermediary structures that may be reactive/known hepatotoxic precursors or moieties, conveying the relevance of transporter data in the context of in vivo data (liver concentration; parent and key metabolites), understanding the impact of any species differences in metabolism (including human), understanding the importance of metabolic enzyme induction or inhibition in the context of hepatocellular hypertrophy/liver weight increases, and providing guidance on drug candidate/metabolite concentrations at site of action (liver < lobular zone < cell type of initial event), that is, drug at target (toxicologic target if known or suspected). In addition, the DMPK scientist is a key partner with the clinical team, in particular the clinical pharmacologist (i.e., for quantitative systems pharmacology [QSP], pharmacokinetics/pharmacodynamics [PK/PD] modeling). While there are many examples of ADME (absorption, distribution, metabolism, and excretion)-related species-specific hepatotoxicities, often owing to a particular, often reactive metabolite uniquely formed in a given species used in a nonclinical toxicology study, the elucidation of such often requires extensive investigative research, and even then, qualitative, rather than definitive quantitative differences are typically all that can be uncovered.^{14,29,31,34,39}

Upon identification of a concerning, potentially translatable liver effect in the nonclinical study data, deciding whether to undertake investigative and mechanistic studies is a matter of judgment and experience and there are many factors that underpin the decision, such as likelihood of success, anticipated monitorability of identified liver effect in clinics, health status of the liver in the intended patient population, risk benefit ratio, exposure margins, and program and portfolio strategies. There are myriad investigative tools and assays available, many of them commercially available, and as such, many fundamental questions regarding the mechanism of action can often be readily answered, however, the time and cost investment must be factored into these decisions. An important consideration in understanding hepatotoxicity mechanisms of action is, of course, determining the primary target cell in the liver and determining the drug/metabolite activities and functional disruptions occurring therein. These attributes can often be assessed using in vitro systems as surrogates coupled with matrix-assisted laser desorption-imaging mass spectrometry (MALD-IMS) applied to in vivo study derived tissue samples.^3,6,15 Linking morphologic effects with biochemical and molecular pathway alterations is a powerful approach for understanding mechanism of action for hepatotoxicity. ⁹

Monitorability of hepatic effects is a critical component of the risk assessment, particularly as pertains to adverse hepatic effects identified in nonclinical studies. Fortunately, many of the same hepatic biomarkers used in nonclinical studies are also used for assuring patient safety in clinical trials, such as ALT, AST, bilirubin, alkaline phosphatase (ALP), and gamma-glutamyl transferase (GGT). Glutamate dehydrogenase (GLDH), which is an additional indicator of hepatocellular injury (leakage enzyme), is seeing more widespread use in nonclinical toxicology studies, both in good laboratory practice (GLP) and non-GLP studies. However, use in clinical settings remains less common. Limited use of GLDH in clinics is likely due to the fact that, in general, familiarity with this hepatic biomarker is less than it is with ALT and AST, ALT has a longer half-life than GLDH, and liver monitoring guidelines are predicated on extensive experience and research, and enabled by large data sets with ALT, AST, as well as ALP and bilirubin. This is unfortunate as GLDH is more specific and sensitive than ALT and AST for hepatocellular injury, at least in commonly used nonclinical species (personal experience).^37,43 In general, even in the absence of morphologic changes, significant increases in ALT, AST, and GLDH typically indicate hepatocellular injury although other causes of increases have been reported, such as decreased elimination via Kupffer cells (at least as pertains to ALT and AST increases) or induction (e.g., steroids, fibrates or other drugs or situations that alter cellular energy or protein metabolism).^{23,25,28,40,44} As a general rule of thumb, given the range of similarities of hepatic functions and mechanisms of injury across species, as a default in terms of translatability, hepatocellular injury is considered to be likely to translate to humans in the absence of mechanistic and/or drug/metabolite data supporting nonclinical species specificity of an observed effect. However, this scenario, of course, would be considered readily monitorable as these same hepatic biomarkers are used as part of the safety monitoring in clinical trials. Conversely, morphologic changes observed in the liver of nonclinical studies in the absence of alterations in corresponding hepatic biomarker changes, requires judgment as to the mechanism underpinning the morphologic change, as well as consideration of expected near-term, and long-term functional impacts. If there is significant uncertainty regarding adversity and functional impact for a given microscopic finding, then investigative studies may be indicated.

For example, a frequently observed nonclinical finding of biliary hyperplasia can often be seen in the absence of significant hepatic biomarker elevations and can be attributed to a variety of potential causes, some perhaps more concerning than others. Biliary hyperplasia can be a result of hepatocellular injury or inflammation in, or near the portal tract, a reaction to degeneration or necrosis of cholangiocytes, disrupted bile flow or due to alterations in biliary content.^13,21 Biliary hyperplasia can also result from direct mitogenic effects of a drug/metabolite, or indirectly, for example, due to endocrine or reproductive hormone modulation. ²¹ These latter mechanisms could arguably be considered less concerning. Aspects of biliary hyperplasia that may increase the concern level regarding potential translatability are evidence of hepatocellular injury, the presence of immature, poorly organized biliary epithelium comprising the hyperplastic response, and/or the presence of inflammation, fibrosis and disruption of the limiting plate. ²² Additional considerations that should be factored into the risk assessment of biliary hyperplasia and other hepatic effects without attendant biomarker changes are rapidity of onset, steepness of the dose-response curve, and whether the finding is limited to one or more nonclinical species. Taken together, the considerations for possible mechanisms highlight the importance of a weight of evidence approach. The example of biliary hyperplasia also reinforces the importance of terminology and specifically what precisely the morphologic changes that are observed relative to communications with clinicians/hepatologists as biliary hyperplasia in humans is more commonly referred to as ductular reaction by hepatopathologists, even though, in general, both terms refer to similar processes, that is, proliferation of cholangiocytes, and reflect similar underlying pathophysiological mechanisms. ⁴² In the author’s opinion, biliary hyperplasia in nonclinical studies, particularly with small molecules, is often due directly or indirectly to altered canalicular transporter function, including in some cases, transporters that are not typically assessed in nonclinical studies (e.g., such as aquaporins [personal observations]).

After a liver hazard is identified and characterized in the nonclinical studies, an important component of the risk assessment is considering the liver disease context of the intended patient population relative to what is known about the mechanism of the observed nonclinical liver effect. It is important to consider the influence of co-medications, dietary supplements, and/or any underlying liver disease’s impact on drug-metabolizing enzymes (cytochrome P450s (CYPs)/Phase I, Phase II) or excretion (transporters), particularly those that are thought to play a role in the liver effect observed in the nonclinical study package. The default assumption is that patients with underlying disease will potentially be more susceptible to drug-induced liver injury but in some liver disease scenarios, there is evidence of alterations in drug-metabolizing enzyme activity that could, at least theoretically, decrease susceptibility of DILI in patients with certain types of underlying liver disease. For example, in metabolic (dysfunction)-associated fatty liver disease or “MAFLD” (formerly called “NAFLD”), there is evidence of modulated CYP activity, specifically, decreased CYP3A4 activity, as well as altered Phase II enzyme activity, in particular increased glucuronosyltransferase activity, which could lead to decreased formation of a reactive/toxic metabolite and/or increased detoxification of a toxic metabolite.^4,5,8,18,30 In the context of what is known about the drug candidate’s mechanism of action for the observed liver effect in the nonclinical studies, consideration should also be given to the target patient population’s liver disease status with respect to mitochondrial function, protective mechanisms (e.g., antioxidant pathways, such as anticipated glutathione [GSH] levels), adaptive/reparative mechanisms (e.g., regenerative capacity), metabolism, and excretion (e.g., are metabolic enzyme and/or key transporter functions altered in a manner that would potentially increase the concentration of parent or putative reactive metabolite).^4,5,8,18,30 It is also worth noting that the patient’s race, ethnicity, sex, gender, and age can have an influence on susceptibility to drug induced liver injury and should be factored into the discussion and risk assessment activities.^1,20 These considerations again highlight the importance of intensive interdisciplinary team discussions, such as between clinicians, clinical pharmacologists, toxicologists, toxicologic pathologists, and DMPK scientists.

Idiosyncratic hepatotoxicity is often raised as a concern when a project team is contemplating potential translatability of nonclinical hepatic findings, particularly if it is thought that a reactive metabolite is formed, and especially if the reactive metabolite is likely contributing to the observed hepatic effects in one or more nonclinical species. Given it is generally accepted that an idiosyncratic response occurring in the liver has direct hepatic injury as an initiating event, understanding the mechanism of the direct effect in nonclinical species should enable minimization of the risk of direct hepatotoxicity in patients, ergo, reducing the risk of idiosyncratic hepatotoxicity.^7,47 Indeed, it is rare to observe idiosyncratic hepatotoxicity in patients in the absence of nonclinical or clinical liver signals, especially if the totality of the liver relevant data are taken into account in the risk assessment process. An example of an idiosyncratic hepatotoxic drug is the peroxisome proliferator-activated receptor gamma (PPARγ) agonist, troglitazone that, prior to being discontinued, was developed for, and used to treat type 2 diabetic patients. ³⁵ It should be noted that troglitazone caused biliary hyperplasia in nonhuman primates (NHPs) at multiple-dose levels, including in all dose groups in male animals, and mild to moderate ALT increases in a subset of patients, and of course, more severe clinical hepatotoxicity in rare instances. ⁴¹ Potentially, at least in the nonclinical NHP studies, and more clearly so in a subset of patients with ALT excursions, disruption of hepatic function and hepatic injury was induced. Following the identification of troglitazone as an idiosyncratic hepatotoxicant, much mechanistic work was carried out and it was subsequently shown to be a bile salt export pump (BSEP) inhibitor as well as mitochondrial toxicant.^16,19,33

In summary, with the broad range of mechanisms of hepatotoxicity coupled with the complexity of hepatotoxicity, risk assessment of liver effects identified in the nonclinical data is best served by a multidisciplinary team of nonclinical and clinical scientists.