The Effect of Hepatic Impairment on Outcomes in Phase I Clinical Trials in Cancer Subjects

Standard phase 1 clinical trials (P1CT) are conducted in cancer subjects to elucidate the safety and tolerability, pharmacokinetic and pharmacodynamic properties, and MTD or recommended phase 2 dose (RP2D) of antineoplastic drugs as single agents or in combination. Subjects with impaired renal or hepatic function are often excluded from P1CT because their altered drug metabolism and excretion may affect MTD and R2PD determinations. Liver test abnormalities secondary to hepatic metastases, underlying liver disease, or lingering effects of prior therapies may also preclude eligibility for a P1CT (1, 2). Taken together, oncologists often have little guidance on how to dose medications in subjects with varying degrees of organ dysfunction.

The FDA and the European Medicines Agency (EMA) have developed guidance on the conduct of studies addressing the optimal dose in subjects with hepatic dysfunction (3, 4). When the majority of antineoplastic agents being developed were cytotoxic with a narrow therapeutic index, only subjects with cancer could be enrolled on these trials based on ethical concerns over short- and long-term toxicities of these agents. With the development of molecularly targeted therapeutics, volunteers without cancer have been utilized in lieu of subjects with cancer for a quick assessment of whether single-dose exposure is altered by varying degrees of hepatic impairment. These trials do not assess long-term tolerability which could be altered for reasons other than drug exposure. Therefore, the Cancer Therapy Evaluation Program (CTEP) at the NCI prioritized study of these special patient populations with hepatic dysfunction phase I clinical trials (HDCT) to determine safe administration parameters of antineoplastic agents for subjects with varying degrees of liver dysfunction. HDCT sponsored by CTEP and others have provided clinically useful information on the optimal dosing of antineoplastic agents in subjects with different degrees of liver test abnormalities that have provided administration guidance in the labels for patients with abnormal organ function (5–15).

During drug development, preclinical testing and observation of clinical events in P1CT identify hepatotoxic drugs that may cause drug-induced liver injury (DILI). Hepatotoxicity is primarily identified by alterations to liver tests during the conduct of the clinical trials. CTEP maintains a large database that includes P1CT and HDCT. We compared the patterns of DILI, number of cycles administered, and reasons for discontinuation between subjects on HDCT and P1CT in a subject-level meta-analysis to gain insight into the safety and outcomes of subjects with hepatic dysfunction on clinical trials.

The records of participants in NCI-sponsored P1CT and HDCT are prospectively maintained in the Clinical Trials Monitoring Service database which is owned by the NCI and maintained by Theradex Systems under contract with the NCI. Theradex provides data management and auditing practices of the database during the conduct of the trials. The database records were abstracted and reviewed at the individual subject level for 1,896 subjects on 42 P1CT and 9 HDCT; we included clinical trials for which serial liver tests were available for review in the database. Liver tests (as described below), the total number of cycles administered of antineoplastic agents, and reasons for discontinuation were reviewed. All NCI-sponsored clinical trials utilized for this analysis were approved by the institutional Review Board with written informed consent being obtained on all subjects.

For this study, the upper limits of normal for aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were defined as 50 U/L, the upper limit of normal for total bilirubin was defined as 1 mg/dL, and the upper limit of normal of alkaline phosphatase (ALK) was defined as 150 U/L. The upper limits of normal for these tests were made independent of local laboratory definitions. Liver toxicity was assessed in multiple ways given the complexities that an individual test is not sufficiently robust: (1) Alterations in liver tests were assessed utilizing the Common Terminology Criteria for Adverse Events (CTCAE) version 4.0 to grade liver test results, (2) The NCI Organ Dysfunction Working Group (ODWG) definitions for hepatic dysfunction were used to define baseline and peak categorizations (Table 1; ref. 15), (3) A modified version of the FDA R ratio [(ALT value/ALT ULN)/(ALK value/ALK ULN)] was used to classify the pattern of liver test abnormalities (Table 1; refs. 16, 17), and (4) Hy's law was also utilized as previously described (Table 1; ref. 18). For the definition of peak, we used the most severe of whichever test. In addition, the NCI ODWG score and FDA R ratio (Table 1) use concurrent labs for the categorizations, the worse of which was used for the definition of peak.

Drugs were categorized as hepatotoxic or not, based on review of LiverTox (http://livertox.nlm.nih.gov) and their package inserts (if approved by the FDA or EMA), or available clinical trial data (if not approved by the FDA or EMA). Drugs that had black box warnings or precautions due to drug-related severe or fatal liver injury in the package inserts or labels were deemed to be hepatotoxic. If the investigational drug had not been approved by the FDA or EMA, drugs that were associated with dose-limiting toxicities or serious adverse events related to liver injury were deemed to be hepatotoxic. In the absence of data suggestive of hepatotoxicity, drugs were scored as nontoxic (Supplementary Table S1).

Descriptive statistics were used to summarize results. Data are displayed as medians with interquartile ranges unless specified otherwise. The χ² test was used to compare categorizations among groups, and the Kruskal–Wallis test was used to compare numeric data between groups. A t test was used in order to analyze the change in CTCAE grade (peak CTCAE category – baseline CTCAE category) by number of cycles received. JMP 11.2.1 (SAS Institute Inc.) was used for statistical analyses. Figures were created with Prism 6 for Mac OS X (GraphPad Software Inc.). Circos diagrams were made using an online interface (http://mkweb.bcgsc.ca/tableviewer; ref. 19) and modified with Adobe Photoshop CC 2014 for clarity.

The NCI CTEP database included 51 P1CT (n = 42) and HDCT (n = 9) with subject-level information on 1,896 subjects. Due to late determination of ineligibility or complicating intercurrent illnesses, 1,841 subjects actually received at least one dose of study treatment on trial, and further analysis was restricted to these subjects (Table 2; Supplementary Table S2). There were 1,328 subjects treated on P1CT and 513 subjects treated on HDCT. The median ages were 59.0 years (range, 49–67) on PICT and 58.6 years (range, 51–66) on HDCT (P = 0.88). There were fewer females included in HDCT (44%) than P1CT (50%; P = 0.03).

At baseline, 84% of subjects in P1CT were classified as normal by NCI ODWG criteria, but fewer subjects met this classification in HDCT (27%; Table 2; Fig. 1). Patients without hepatic dysfunction are included in HDCT for comparative pharmacokinetic analyses. Changes in subject categorization by NCI ODWG definitions between baseline and peak levels of liver tests are displayed with circos diagrams (Fig. 2). Whereas the liver tests of some subjects with mild group 1, mild group 2 or moderate classifications of liver dysfunction at baseline normalized, the liver tests of subjects with severe liver dysfunction did not (Fig. 2A). In addition, worsening categorizations or stabilization of liver dysfunction categorizations was more commonly observed than normalization in P1CT (Fig. 2B) and HDCT (Fig. 2C).

As expected, subjects on HDCT had higher baseline CTCAE grades for liver tests than subjects on P1CT (Fig. 3). There was a greater increase in peak ALT CTCAE grades for subjects on P1CT than HDCT (P = 0.04), and a greater increase in peak total bilirubin CTCAE levels for subjects on HDCT than P1CT (P < 0.01; Supplementary Table S3).

Among all subjects, there were higher peak ALT CTCAE scores for subjects receiving hepatotoxic agents than nontoxic agents (P < 0.01; Fig. 4). Similarly, there was a significant worsening from baseline of ALT and total bilirubin CTCAE grades for those receiving hepatotoxic agents than nontoxic agents (P < 0.01 and P = 0.04, respectively; Supplementary Table S4). Among subjects enrolled on P1CT, there were higher peak ALT CTCAE grades for those receiving hepatotoxic agents than nontoxic agents (P < 0.01; Supplementary Fig. S1), whereas higher peak CTCAE grades for ALK (P < 0.01) and total bilirubin (P = 0.01) were seen in subjects on HDCT receiving hepatotoxic agents than nontoxic agents (Supplementary Fig. S2).

Cholestatic peak patterns of liver impairment by the FDA R ratio were more frequent in HDCT than P1CT (P < 0.01). Criteria for Hy's law (Table 1) were met by 11 subjects on P1CT (baseline NCI ODWG normal = 8, mild group 1 = 1, mild group 2 = 1, moderate = 1) but not by any subjects on HDCT. No difference in the baseline total bilirubin CTCAE grades was observed between subjects who did and did not meet Hy's law criteria (P = 0.78).

Subjects typically received fewer cycles of therapy on HDCT [median, 2; mean, 2.9; interquartile range (IQR), 1–3; full range, 1–62] than subjects on P1CT (median, 2; mean, 4.0; IQR, 1–5; full range, 1–52; P < 0.01). Subjects with severe liver test abnormalities on HDCT typically received one cycle of treatment (median, 1; mean, 1.7; IQR, 1–2; full range, 1–12). The reasons for treatment discontinuation were similar for subjects on HDCT and P1CT with disease progression being the most common reason followed by adverse events (Table 2). There was no significant difference in the rates of discontinuation for toxicity between those who received two or fewer cycles [12.2%; 95% confidence interval (CI), 10.4–14.3%; n = 1,105] and those who received more than two cycles of therapy (11.7%; 95% CI, 9.6–14.3%; n = 726). Subjects who received more than two cycles of therapy were more likely to experience an increase in CTCAE grade of ALT (mean, 0.53; 95% CI, 0.48–0.59) than those who received two or fewer cycles (mean, 0.43; 95% CI, 0.38–0.47; P < 0.0001). Conversely, subjects who received two or fewer cycles of therapy were more likely to experience an increase in CTCAE grade of alkaline phosphatase (mean, 0.35; 95% CI, 0.31–0.40) than those who received more than two cycles (mean, 0.27; 95% CI, 0.23–0.32; P = 0.02). There were no significant differences in CTCAE grade of AST or total bilirubin based on whether patients received two or more cycles.

We reviewed the NCI CTEP experience sponsoring HDCT and P1CT and found expected differences in baseline liver test abnormalities at enrollment. Further review of the classifications of hepatic impairment revealed that subjects with severe impairment on HDCT typically do not experience improvement in their liver tests during the course of treatment, but many other subjects with less severe liver impairment do. In addition, different patterns of hepatotoxicity were observed between trial types. Higher ALT CTCAE grades were associated with hepatotoxic agents in P1CT, and higher alkaline phosphatase and total bilirubin CTCAE grades were associated with hepatotoxic agents in HDCT. This differential effect on hepatotoxicity suggests that underlying hepatic function may affect susceptibility to and patterns of DILI. Even though there was a statistically significant difference in the number of cycles received by subjects on HDCT or P1CT, we believe that this result may be due to the large sample size of our study and is not clinically relevant as patients typically received up to two cycles of treatment on either type of study.

DILI is typically classified as hepatic (hepatocellular), cholestatic, or mixed based on the pattern of liver test abnormalities (17). The FDA R ratio is one means of defining these patterns based on ratios of ALT to ALK. Hy's law is an additional measure that identifies a population at high risk of death when DILI results in hepatocellular injury with jaundice. The use of liver tests to assess hepatic function is imperfect because the elevation of transaminases may be a better indicator of injury or inflammation than dysfunction. About 3% of subjects are expected to fall outside of two standard deviations of the Gaussian distribution for liver tests (20); however, a study of 6,000 healthy individuals that did not consume excessive amounts of alcohol and did not have hepatitis C identified that 8% of patients had elevations in transaminases, possibly due to nonalcoholic steatohepatitis or metabolic syndrome (21, 22). Standard liver tests also fail to identify liver damage as proven by histologic analysis from biopsies in subjects with chronic hepatitis C or nonalcoholic steatohepatitis 16% or 13% of the time respectively (23, 24). The situation for subjects on P1CT and HDCT is confounded further by potential liver metastases, lingering effects of previous treatments, and novel combinations. At baseline, we observed that 16% of subjects who enrolled onto P1CT had abnormal liver tests, but the reasons for the liver tests abnormalities were not always certain.

Other scores have been developed to assess prognosis in patients with chronic liver disease such as the Child–Pugh (CP) score and the Model for End-Stage Liver Disease (MELD). These scores incorporate additional measures such as creatinine (MELD), serum albumin (CP), prothrombin time (both), and clinical observations such as ascites and hepatic encephalopathy (CP). In our study, very few subjects had serial testing performed that provided information on hepatic synthetic function such as prothrombin time. Also, we were not able to retrospectively determine if patients had ascites. Thus, we were not able to retrospectively determine the traditional CP Score or MELD for the patients in this analysis. It is uncertain as to how baseline CP and MELD scores affect DILI in P1CT and HDCT.

Because the P1CT and HDCT included in this meta-analysis standardly obtained AST, ALT, alkaline phosphatase, and bilirubin, we were better able to assess cellular injury, cholestasis, or duct injury than liver function per se. In addition, the currently used attribution system did not allow us to distinguish whether baseline liver test abnormalities were related to liver metastases, underlying liver disease, or lingering treatment effects as the CTCAE attributions typically do not require further clarification if the adverse event is not thought to be related to the investigational agent(s). Previous work by NCI has shown that there is a correlation between CP categories and NCI ODWG classifications (25). As the MELD is a preferred predictor for liver-related outcomes compared with CP for patients with chronic liver disease, consideration should be given for its inclusion in HDCT, and even possibly P1CT, as it may provide additional useful information on hepatic function as it relates to safe drug dosing and DILI (26–28).

Hy's law is a prognostic marker in patients who experience DILI (17), and the observation of multiple subjects who meet Hy's law criteria in a clinical trial is concerning that an investigational agent(s) would result in severe DILI if administered to a larger population (29). We unexpectedly found more subjects who met Hy's law criteria in P1CT than HDCT. This is not surprising when one considers that “preexisting or acute liver disease” excludes qualification for Hy's law; therefore, subjects with the most severe abnormalities in HDCT did not qualify as meeting Hy's law based on the attributions of their liver test abnormalities. Previous work has shown pretreatment bilirubin to be the most important predictor of meeting the criteria in Hy's law (30), but most patients in our series who met Hy's law criteria had normal bilirubin values at baseline. The few cases in our series that met Hy's law criteria limit any meaningful conclusions, but the multifactorial nature of DILI makes it challenging to predict severe DILI with a single test. In terms of P1CT and HDCT, we are not certain what role Hy's law should have beyond its prognostic capabilities in subjects who experience severe DILI.

Overall, our findings suggest that within the context of organ dysfunction trials, it is safe to treat subjects with hepatic dysfunction. Subjects on HDCT discontinue trial participation for similar reasons and at similar frequencies to those on P1CT. The differential effects of hepatotoxic agents on hepatotoxicity between subjects on HDCT and P1CT suggest that underlying hepatic function may affect susceptibility to and patterns of DILI. We suggest that it would be worthwhile to include additional scores of hepatic function, such as MELD in P1CT and HDCT, to determine if underlying hepatic dysfunction is predictive of severe DILI.

A.S. Mansfield is a consultant/advisory board member for Celegene, Genentech, Rockpointe, and Trovagene. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.S. Mansfield, M.A. Rudek, S.P. Ivy

Development of methodology: A.S. Mansfield, M.A. Rudek, S.P. Ivy

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.S. Mansfield, G.L. Smith, S.P. Ivy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.S. Mansfield, M.A. Rudek, S.P. Ivy

Writing, review, and/or revision of the manuscript: A.S. Mansfield, M.A. Rudek, G.L. Smith, P.J. Harris, S.P. Ivy

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.S. Mansfield, S.P. Ivy

Study supervision: A.S. Mansfield, M.A. Rudek

Other (management of data submitted by treating sites): D. Vulih

This work was supported by NIH grants K12CA090628 (A.S. Mansfield), UM1CA186691 (M.A. Rudek), and P30CA006973 (M.A. Rudek).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.