Discussion

In this study, we analyzed how 254 orally administered drugs (i.e., P450 substrates, inhibitors, or inducers) from the LTKB-BD were associated with DILI. We discovered that drugs/substrates for P450 enzymes showed a statistically significant but dose-independent association with DILI; drugs/inhibitors for P450 enzymes would have the higher likelihood of causing DILI when they were administered at high daily doses (≥100 mg). On the contrary, drugs/inducers for P450 enzymes were not observed to be significantly associated with DILI. These findings will be helpful to identify drugs with the high likelihood of causing DILI in clinical settings.

We determined that high daily dose and the drug being a substrate of P450 enzymes, which had the significantly higher likelihood of causing DILI, were two important predictors of DILI (Table 4). In addition, we also revealed that only the integration of high daily dose and being a P450 substrate had a higher likelihood of causing DILI. A list of structural alerts associated with the formation of toxic reactive metabolites has been established (Kalgutkar et al., 2005). However, the mere presence of a structural alert in a drug will not necessarily be associated with DILI (Stepan et al., 2011). Some drugs with the same structural alerts may have different risk of DILI. For example, troglitazone, rosiglitazone, and pioglitazone contain the same thiazolidinedione scaffold, which is the structural alert to generate the toxic reactive metabolites. Troglitazone has been withdrawn because of DILI, whereas rosiglitazone and pioglitazone are currently still in use. Their daily dose may be a key difference [i.e., troglitazone (400 mg/d) versus rosiglitazone (6 mg/d) and pioglitazone (30 mg/d)]. On the contrary, the key difference to explain the different risk of DILI for ibuprofen (1200 mg/d) and ibufenac (2400 mg/d) is the formation of toxic reactive metabolites. Therefore, high daily dose and being the P450 substrate should be simultaneously considered to identify the DILI-positive drugs.

The drug pairs have similar chemical structures and are usually expected to have similar efficacy and safety. Drugs marked as toxic that were withdrawn from the market in the United States and elsewhere due to DILI (i.e., alpidem, tolcapone, trovafloxacin, benoxaprofen, and bromfenac; Table 5) were identified as DILI-positive drugs by our two predictors. Ibuprofen and diclofenac were also identified as DILI-positive drugs, and their DILI potential was confirmed by case reports of acute liver failure (Helfgott et al., 1990; Banks et al., 1995; Javier Rodriguez-Gonzalez et al., 2002). Because of the DILI potential of these two drugs, they are annotated as less-DILI and most-DILI drugs in the LTKB, respectively (Chen et al., 2011). Thus, our two predictors can identify the toxic compound in the drug pair. In addition, these two predictors could identify the drugs that caused DILI among multiple medications used by patients in clinical studies (Table 6). These results confirmed that two identified predictors (high daily dose and the drug being a P450 substrate) could have the useful clinical applications, particularly for clinicians during the decision-making process of caring for a patient suspected of having DILI. Drug developers are currently encouraged to provide metabolism data on P450 enzymes for new drug applications (U.S. Food and Drug Administration, 2012), which will facilitate a practical use of the identified two predictors in the future.

It has been reported that almost all of the drugs known to be associated with DILI cause clinically significant DDIs (Li, 2002). Many DDIs can be explained by the inhibition or induction of the expression of P450 enzymes in the liver (Tanaka, 1998). Our results suggest that clinically observed DILI may occur when P450 inhibitors are administered at high daily doses (Table 3). Based on the mechanisms of enzyme inhibition, a P450 inhibitor needs to bind to P450 enzymes and then decreases their activity. Therefore, the inhibitory effect is expected to be increased with the increasing dose of the inhibitor. Other authors also find that the inhibition of P450 enzymes is dose-dependent (Lin et al., 2000; Hollenberg, 2002). The inhibition of P450 enzymes may increase liver exposure of the parent drug, which has the potential to cause DILI. For example, clinically dose-dependent hepatotoxicity of fluconazole, which is a potent inhibitor of CYP2C19 and a moderate inhibitor of CYP2C9 and CYP3A, was observed (Wells and Lever, 1992). Phenytoin (300 mg/d) is an inhibitor of CYP2C9 with a narrow therapeutic index. It is also a substrate of CYP2C9. Therefore, the increased liver exposure of phenytoin by the inhibition of CYP2C9 might increase the risk of DILI. Phenytoin has been considered as one of the most common drugs, other than acetaminophen, reported to cause acute liver failure in the United States (Russo et al., 2004), which confirmed our findings that P450 inhibitors would have the higher likelihood of causing DILI at high daily doses. Inhibitory effects usually occur immediately; however, induction of P450 enzymes is usually both time- and dose-dependent (Markowitz et al., 2003), which will lead to a delay before enzyme activity increases depending on the duration of exposure and half-life of the inducer in the liver. Consequently, induction of hepatic P450 enzymes rarely leads to altered metabolism and toxicity in the patient, which typically occurs only if the induction is very extensive (Mohutsky et al., 2010). For example, three drugs were recorded by the U.S. FDA as potent P450 inducers (i.e., carbamazepine, phenytoin, and rifampin; (http://www.fda.gov/Drugs/DevelopmentApprovalProcess/DevelopmentResources/DrugInteractionsLabeling/ucm093664.htm). All of these drugs are considered as DILI-positive drugs in the LTKB-BD.

Because of the differences in individual susceptibility to hepatotoxic drugs, our findings will also be helpful for dose adjustment, which is important to prevent DILI. For example, females are more susceptible to DILI than males (Russo et al., 2004). This could be partly explained by the 2-fold higher activity of CYP3A4 in the female liver compared with the male liver (Wolbold et al., 2003). CYP3A4 has the highest abundance in the human liver (approximately 40%) and is the most frequently involved enzyme in the metabolism of DILI-positive drugs (Li, 2002). As indicated in this study, CYP2C would have the higher likelihood of metabolizing the DILI-positive drugs. Therefore, Hispanics with about twice the activity of CYP2C compared with Caucasians and African Americans (Parkinson et al., 2004) may have a higher risk of liver injury caused by drugs such as NSAIDs, leflunomide, troglitazone, and valproic acid, which are mainly metabolized by CYP2C (Kirchheiner and Brockmoller, 2005). NSAIDs were reported as the most frequently used drugs causing DILI in Spain (Ibanez et al., 2002).

This study has several limitations. First, we do not have a confirmation that one among the P450 enzymes involved in the metabolism of the drugs participates in the formation of toxic reactive metabolites. Second, enzymes other than P450 isoforms can also mediate the generation of toxic reactive metabolites, and their roles are not considered in this study. Third, our results are based on in vitro data on P450-mediated metabolism and their extrapolation to in vivo studies is uncertain. Fourth, the data on P450-mediated metabolism for some drugs and the data on inhibition or induction effects of some drugs on P450 enzymes are currently unavailable. Since the FDA is encouraging the industry to submit metabolism data and safety data on drug metabolites for new drug applications, we can expect to summarize more solid conclusions using the analysis we carried out in this study when those data are available in the future. In addition, authors should be encouraged by journals to publish negative findings because in these cases, it will be certain that drugs classified as “not the substrates of P450 enzymes” are not metabolized by particular P450 enzymes.

In summary, besides high daily dose, our study indicates that the drug being a substrate of P450 enzymes is another important predictor of DILI. In addition, we also learned that P450 inhibitors have a higher likelihood of causing DILI only when they are administered at high daily doses, and P450 inducers are not observed to be closely associated with DILI. Only the integration of high daily dose and being a P450 substrate have the higher likelihood of causing DILI. Our findings can serve as a simple clinical algorithm for the identification of the drug with a high likelihood of causing DILI in clinical settings.