Oral medications with significant hepatic metabolism at higher risk for hepatic adverse events†
Potential conflict of interest: Nothing to report.
Reactive metabolites generated by hepatic metabolism are thought to play an important role in the pathogenesis of drug-induced liver injury (DILI), but supporting data are limited. If this is true, then compounds with significant hepatic metabolism should cause more DILI than those without it. We conducted a study to examine the relationship between hepatic metabolism and DILI of prescription medications. We systematically extracted the metabolism characteristics of 207 of the most widely prescribed oral medications in the United States. Compounds with >50% hepatic metabolism were characterized as those with significant hepatic metabolism (n = 149). Hepatic adverse events of interest were alanine aminotransferase >3 times the upper limit of normal, jaundice, liver failure, liver transplantation, or fatal DILI. Compared with compounds with lesser hepatic metabolism, compounds belonging to the significant hepatic metabolism group had significantly higher frequency of alanine aminotransferase >3 times the upper limit of normal (35% versus 11%, P = 0.001), liver failure (28% versus 9%, P = 0.004), and fatal DILI (23% versus 4%, P = 0.001), but not jaundice (46% versus 35%, P = 0.2) or liver transplantation (9% versus 2%, P = 0.11). Twelve compounds with no hepatic metabolism had no reports of liver failure, liver transplantation, or fatal DILI. When the relationship between hepatic adverse events and combination of hepatic metabolism and daily dose was examined, compounds with both significant hepatic metabolism and daily dose >50 mg (n = 50) were significantly more hepatotoxic than compounds belonging to other groups. Compared with medications without biliary excretion, compounds with biliary excretion (n = 50) had significantly higher frequency of jaundice (74% versus 40%, P = 0.0001). Conclusion: Our study finds an important relationship between a compound's metabolism profile and reports of hepatic adverse events. (HEPATOLOGY 2009.)
Liver injury is a potential complication of many different drugs. This is not surprising, given the important role played by the liver in the metabolism and excretion of drugs from the body. Most drugs undergo some metabolism in the liver before excretion by the kidneys or through bile, though some drugs have little or no metabolism within the liver. Two major biotransformation phases have been identified for drugs metabolized in the liver.1 The hepatic metabolism of xenobiotics involves either phase I followed by phase II reactions or phase I reactions alone or rarely only phase II reactions.1 Phase I reactions are catalyzed by the cytochrome P450 enzymes (CYPs) potentially leading to formation of reactive metabolites.1 The reactive drug metabolites generated through phase I reactions can potentially lead to liver injury, but phase II reactions are important in detoxifying these reactive metabolites. These reactive metabolites or intermediates are in many instances metabolized further by phase II reactions that involve their conjugation with endogenous molecules such as glutathione, glucuronate, sulfate, or acetate in order to make them more water soluble and more easily eliminated from the body.1
Formation of toxic reactive metabolites has been suggested as potential mechanism for causing idiosyncratic drug-induced liver injury (DILI).2–5 Hepatic CYPs that generate reactive intermediates are largely concentrated in the centrilobular zone (zone III), an area that is predominantly affected in some forms of DILI (e.g., acetaminophen or halothane toxicity).6 These reactive metabolites may potentially bind to various cellular proteins and subsequently make them targets for immunomediated cell injury.5, 7 However, the role of phase II reactions in causing DILI cannot be excluded. A rodent model suggested that diclofenac-adducts generated by glucuronidation may play an important role in the pathogenesis of diclofenac-induced liver injury, although evidence directly implicating its acyl-glucuronide derivative is lacking.8 Many experts believe that reactive metabolites play an important role in the pathogenesis of DILI.2–5 If this theory was true, then compounds that are metabolized by the liver should have higher frequency of DILI than compounds without hepatic metabolism. However, some drugs without significant hepatic metabolism may cause serious DILI (e.g., ximelagatran).9, 10 We conducted a study to test the hypothesis that compounds with significant hepatic metabolism cause DILI at a greater frequency than compounds with lesser degrees of hepatic metabolism. Using two comprehensive pharmaceutical databases, we examined the relationship between hepatic metabolism of commonly prescribed medications and their reported ability to cause hepatotoxicity.
Materials and Methods
A widely available pharmaceutical database (www.drugtopics.com) was used to generate the names of the top 200 brand and top 200 generic medications by prescription volume in the United States for the year 2005.11, 12 Only oral medications were included, and compounds listed in both the brand and generic lists were considered one entry. Entries with more than one active compound (i.e., fixed drug formulations) and those containing acetaminophen compounds were excluded. These criteria identified 207 individual compounds that were considered eligible for inclusion in this study. These compounds were subsequently categorized into those with and without significant hepatic metabolism based on information contained within Thompson's Micromedex DRUGDEX System,13 PubMed, Therapeutic Drugs by Dollery,14 the AHFS Drug Information textbook,15 and FASS (the Swedish Information Medicines Engine).16 The compounds with ≥50% hepatic metabolism were categorized into a significant hepatic metabolism group, whereas compounds with <50% hepatic metabolism were categorized into a nonsignificant hepatic metabolism group (for example: allopurinol, with approximately 80% of the compound metabolized in the liver, was placed in the significant hepatic metabolism group). A 50% cutoff was chosen a priori arbitrarily based on consensus of the authors. For compounds that are pro-drugs, we searched for the metabolism of their active drugs. In addition, we further characterized each compound based on (1) whether phase I and/or II reactions were involved in its metabolism, (2) if it or its active metabolite has ≥10% biliary excretion, (3) which cytochrome P450 enzyme is predominantly involved in its metabolism, and (4) average daily dose (≤10 mg, 11–49 mg, or ≥50 mg) as defined elsewhere.17
We subsequently searched for reports of selected hepatic adverse events for each of these 207 compounds utilizing Thompson's Micromedex (DRUGDEX) System13 The DRUGDEX is a comprehensive pharmacy database, consisting of package insert data and published literature.13 To ensure completeness, each compound was cross-checked in PubMed, the U.S. Food and Drug Administration's Adverse Event Reporting System database, and the Physicians' Desk Reference.18 Hepatic adverse events of interest were alanine aminotransferase (ALT) >3 times the upper limit of normal (ULN), cholestatic jaundice, liver failure, liver transplantation, and fatal DILI. For the purposes of the study, fatal DILI was defined as drug-induced acute liver failure resulting in death. We did not take into account the total number of reported adverse events for a single compound; rather, we identified if a particular event has ever been reported for that compound. This allowed us to compute the proportion of compounds in each metabolism subgroup to cause prespecified hepatic adverse events. This study essentially represents an extension of our earlier published work that investigated the relationship between daily dose of oral medication and reports of hepatic adverse events.17
The primary data analysis compared the frequency of hepatic adverse events between compounds with and without significant hepatic metabolism. Additional analyses were conducted to examine the relationship between the reports of hepatic adverse events and (1) the type of hepatic metabolism (i.e., phase I and/or II), (2) the biliary excretion of the parent compound or its active metabolite, (3) the predominant CYP involved, and (4) the combination of hepatic metabolism and average daily dose. The results are expressed as the percentages of drugs able to cause an adverse effect rather than the frequencies of these adverse events in treated patients. Descriptive statistics such as the mean plus standard deviation and percentages were used to characterize the cohort. Categorical variables were tested for statistical significant differences by way of the chi-square or Fisher's exact test and continuous variables by way of the Student t test; P < 0.05 was considered statistically significant.
Out of 207 eligible compounds, 149 belonged to the significant hepatic metabolism group and 55 to the nonsignificant hepatic metabolism group (Supporting Table 1). There were three compounds for which the details of their metabolism could not be identified (docusate, dicyclomine, nitrofurantoin). The mean number of prescriptions written for the significant hepatic metabolism group was 7,954,705 and was not statistically different from the nonsignificant hepatic metabolism group (9,068,470, P = 0.5). Thirty-six percent of the compounds in the significant hepatic metabolism group had an average daily dose of ≥50 mg versus 51% of the compounds in the nonsignificant hepatic metabolism group (P = 0.03).
Compared with compounds without significant hepatic metabolism, compounds in the significant hepatic metabolism group were significantly more likely to have reports of ALT ≥3 times the ULN (35% versus 11%, P = 0.001), liver failure (28% versus 9%, P = 0.004), and fatal DILI (23% versus 4%, P = 0.001), but not liver transplantation (9% versus 2%, P = 0.11) or jaundice (46% versus 35%, P = 0.2) (Table 1). Compared with compounds metabolized only through phase I reactions, compounds metabolized through both phase I and II reactions did not have greater frequency of jaundice (P = 0.74), liver failure (P = 0.36), liver transplantation (P = 0.36), or fatal DILI (P = 0.56), but had significantly higher reports of ALT >3 times the ULN (45% versus 28%, P = 0.03) (Table 2). There were nine compounds with metabolism only through phase II reactions; of these, one had ALT >3 times the ULN, four had jaundice, two had liver failure, one caused liver transplantation, and two caused fatal DILI (Table 2). These nine compounds were levothyroxine, telmisartan, metoclopramide, hydralazine, prednisolone, topiramate, labetalol, and niacin.
Table 1. Relationship Between Extent of Hepatic Metabolism and Reports of Hepatic Adverse Events
|Compounds with significant hepatic metabolism (n = 149)||35%||46%||28%||9%||23%|
|Compounds without significant hepatic metabolism (n = 55)||11%||35%||9%||2%||4%|
Table 2. Relationship Between Type of Hepatic Metabolism and Reports of Hepatic Adverse Events
|Only phase I reactions (n = 117)||28%||46%||24%||6%||21%|
|Both phase I and II reactions (n = 57)||45%||49%||33%||11%||26%|
There were 50 compounds with documented biliary excretion of the parent compound or its active metabolite. When compared with those without biliary excretion, compounds with biliary excretion had significantly higher frequency of jaundice (74% versus 40%, P = 0.0001) but not other hepatic adverse events (Table 3). Table 4 shows the relationship between hepatic adverse events and metabolism through four common CYPs. There are potentially significant differences among different CYP pathways and reports of liver failure and fatal DILI. In general, CYP2C9 and CYP2C19 pathways appeared more toxic than CYP3A and CYP2D6 (Table 4).
Table 3. Relationship Between Biliary Excretion and Reports of Hepatic Adverse Events
|Compounds with biliary excretion (n = 50)||36%||74%||28%||10%||24%|
|Compounds without reported biliary excretion (n = 157)||26%||40%||24%||6%||18%|
Table 4. Relationship Between Hepatic Adverse Events and Metabolism Through Specific CYP
|CYP 3A (n = 75)||33%||40%||16%||5%||13%|
|CYP 2D6 (n = 50)||30%||46%||25%||6%||16%|
|CYP 2C19 (n = 23)||30%||56%||43%||17%||35%|
|CYP 2C9 (n = 34)||35%||53%||38%||6%||32%|
There were 12 compounds without any hepatic metabolism (Table 5). Interestingly, none of these has been reported to cause liver failure, liver transplantation, or fatal DILI, but five compounds have been reported to cause jaundice (Table 5).
Table 5. Compounds Without Hepatic Metabolism Categorized into Average Daily Dose (n = 12)
| || ||Metformin*|
| || ||Cephalexin|
| || ||Benzonatate|
| || ||Cefuroxime*|
| || ||Sotalol|
Table 6 shows the relationship between hepatic adverse events and combination of hepatic metabolism and average daily dose. It appears that compounds with both significant hepatic metabolism and average daily dose ≥50 mg (n = 50) are significantly more hepatotoxic than compounds belonging to other groups (Table 6). When compared with compounds in all other groups combined, compounds with both significant hepatic metabolism and average daily dose ≥50 mg had significantly higher frequency of liver failure (P = 0.002), liver transplantation (P = 0.002), and fatal DILI (P = 0.003). When compared with compounds in any other group separately, compounds with both significant hepatic metabolism and average daily dose >50 mg had a higher frequency of ALT >3 times the ULN (P = 0.01), liver failure (P = 0.001), liver transplantation (P = 0.08), and fatal DILI (P = 0.006) than other single group (Table 6).
Table 6. Relationship Between Hepatic Metabolism, Daily Dose, and Reports of Hepatic Adverse Events
|ALT >3× ULN||22%||22%||18%||20%||25%||46%|
The pathogenesis of idiosyncratic DILI is not well understood. Traditionally, it is thought to be unpredictable and not dose-dependent. However, in a recent study consisting of pharmaceutical databases, we have uncovered epidemiological signals to suggest that there may be a daily dose threshold (≥50 mg) beyond which oral medications have increased risk of serious DILI events.17 The current study was undertaken to examine the relationship between metabolism characteristics of medications and the risk of hepatic adverse events.
Although some drugs are metabolized into stable metabolites, many drugs are transformed into unstable and potentially reactive metabolites that can bind to and attack hepatic macromolecules.19 Although reactive metabolites are considered to be of major importance in the pathogenesis of DILI, this has not been systematically investigated previously for the overall risk for DILI. If this reactive metabolite theory is shown to be true for the overall risk for DILI, this is obviously of concern in the development of new drugs. We hypothesized that compounds with significant hepatic metabolism may potentially be more hepatotoxic due to the generation of reactive intermediaries and subsequent metabolic idiosyncrasy. Indeed, our epidemiological survey uncovered many associations between metabolic characteristics of medications and the risk of hepatic adverse events.
This study is an extension of our previously published study that systematically examined the relationship between daily dose of oral medications and hepatic adverse events. Although the present study stems from the database and consists of the same set of oral compounds as our previous study, it addressed different hypotheses and uncovered key findings that have not been reported previously.
First, a significantly higher proportion of compounds with significant hepatic metabolism had reports of ALT >3 times the ULN, liver failure, liver transplantation, and fatal DILI versus compounds with lesser degrees of hepatic metabolism. It is striking that none of the 12 compounds without any hepatic metabolism had any identifiable reports of liver failure, liver transplantation, or fatal DILI. This does not necessarily mean that compounds without hepatic metabolism are entirely immune from hepatotoxicity, but it is significantly less likely. One example of a drug in which selective hepatic metabolism did not cause serious hepatotoxicity is ximelagatran, which was not approved by the Food and Drug Administration due to cases of hepatotoxicity during its development.9–10 Because this medication was never approved for clinical use in the United States, it was not included in our study. Another example is pregabalin, which is also without hepatic metabolism but may be rarely associated with suspected severe hepatotoxicity.20 Because it was approved in December 2004, it was not included in our brand name compound category. Thus, lack of hepatic metabolism does not assure total lack of hepatotoxicity, but based on our data it indeed appears to be quite rare. According to the pharmacological interaction hypothesis, some drugs may be able to initiate an immune response through a reversible interaction with the major histocompatibility complex–T cell receptor complex.21 It has been postulated that ximelagatran leads to hepatotoxicity by evoking an immune response by binding directly but reversibly to major histocompatibility complex.10
Second, we did not find a significant relationship between the frequency of hepatic adverse events and whether a compound is metabolized by phase I and/or phase II reactions. Compounds with only phase II metabolism were not immune from hepatic adverse events. If confirmed, these observations are important, because they argue against a singular role for reactive metabolites generated by phase I reactions in causing hepatotoxicity.
Third, we found a statistically significant relationship between reports of jaundice and whether a compound has biliary excretion. Although it was not always clear from the reports contained within the DRUGDEX whether jaundice is hepatocellular or cholestatic in nature, we found that it was cholestatic in a substantial proportion. This leads us to speculate that compounds with biliary excretion may cause cholestatic jaundice in genetically predisposed individuals (e.g., defective transporters). There is increasing evidence that cholestatic liver injury associated with certain compounds results from a drug- or metabolite-mediated inhibition of hepatobiliary transport systems.22
Furthermore, we observed an additive effect of daily dose and hepatic metabolism; oral compounds with significant hepatic metabolism but also given at daily doses ≥50 mg had the highest risk of hepatic adverse drug reactions compared with other groups (Table 6). Our observation of a relationship between daily dose and idiosyncratic DILI has recently been confirmed in more than 600 prospectively collected DILI cases included in the Spanish hepatotoxicity registry.23 It is striking that 77% of their compounds belonged to the >50 mg/day category, exactly the same proportion that we reported in our earlier study.17 Our current observations extend our previous findings and may have important implications for future drug development. Based on our data, it is tempting to suggest that the pharmaceutical industry should focus on developing compounds that are administered at doses <50 mg/day and without significant hepatic metabolism.
Several aspects of this study deserve further discussion. It shares the same drawbacks as our earlier study that examined the relationship between daily dose and DILI.17 First, this is largely a systematic survey of the published literature, thus our observations should be viewed as epidemiological clues rather than confirmed facts. Second, it should be noted that the 50% cutoff that defined significant hepatic metabolism was chosen arbitrarily and was based on consensus, thus it may or may not reflect biological significance. Third, although an extensive search of multiple databases was conducted to decipher metabolism characteristics of eligible compounds, some uncertainties remain. For example, there remained three compounds (docusate, nitrofurantoin, and dicyclomine) whose metabolism profile could not be identified. Similarly, topiramate, which we classified as having only phase II metabolism, is primarily eliminated unchanged in the urine (≈ 70%). Finally, we have not considered alterations in metabolism induced by coadministered medications (e.g., antiepileptic agents or antifungal agents).
Despite the above limitations, our observations are potentially important, and we believe they are worthy of further investigation. Because DILI is a rare clinical event, it is difficult to design prospective studies that address questions of this nature. Proprietary datasets owned by the pharmaceutical industry may provide further opportunities for data mining, especially when multiple datasets are investigated in aggregate. If our findings can be reproduced by other investigations, then our observations may facilitate the development of safer medications.