SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Methods and Materials
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Drug-induced liver injury (DILI) is a leading cause of drug failure in clinical trials and a major reason for drug withdrawals from the market. Although there is evidence that dosages of ≥100 mg/day are associated with increased risk for hepatotoxicity, many drugs are safe at such dosages. There is an unmet need to predict risk for DILI more reliably, and lipophilicity might be a contributing factor. We analyzed the combined factors of daily dose and lipophilicity for 164 US Food and Drug Administration–approved oral medications and observed high risk for hepatotoxicity (odds ratio [OR], 14.05; P < 0.001) for drugs given at dosages ≥100 mg/day and octanol-water partition coefficient (logP) ≥3. This defined the “rule-of-two.” Similar results were obtained for an independent set of 179 oral medications with 85% of the rule-of-two positives being associated with hepatotoxicity (OR, 3.89; P < 0.01). Using the World Health Organization's Anatomical Therapeutic Chemical classification system, the rule-of-two performed best in predicting DILI in seven therapeutic categories. Among 15 rule-of-two positives, 14 were withdrawn from hepatotoxic drugs, and one was over-the-counter medication labeled for liver injury. We additionally examined drug pairs that have similar chemical structures and act on the same molecular target but differ in their potential for DILI. Again, the rule-of-two predicted hepatotoxicity reliably. Finally, the rule-of-two was applied to clinical case studies to identify hepatotoxic drugs in complex comedication regimes to further demonstrate its use. Conclusion: Apart from dose, lipophilicity contributes significantly to risk for hepatotoxicity. Applying the rule-of-two is an appropriate means of estimating risk for DILI compared with dose alone. (HEPATOLOGY 2013)

See Editorial on Page 15

Hepatotoxicity is a major reason for drugs failing clinical trials, being withdrawn from the market after approval, and being labeled with a black box warning by the US Food and Drug Administration (FDA).1 About 1,000 drugs are suspected to be hepatotoxic,2 and drug-induced liver injury (DILI) accounts for more than 50% of acute liver failure cases in the United States alone.3

While animal studies remain the gold standard testing strategy,4, 5 a retrospective analysis revealed that such tests fail in about 45% of DILI cases in clinical trials.6 There is an unmet need to predict risk for DILI more reliably, and to overcome current limitations, the concept of a daily dose was developed.7 Specifically, many drugs prescribed at daily doses of ≥100 mg have either been withdrawn from the market or have received a black box warning due to hepatotoxicity8 to imply a significant relationship between daily dose and risk for DILI.9 In order to safeguard patients, it was recommended to avoid drug development programs that require high doses.7, 10, 11 However, many drugs given at high doses are safe with little or no risk of hepatotoxicity, suggesting that daily dose alone is not a reliable means of guiding the drug development process, regulatory application, and clinical practice.

To better predict risk for DILI, the combined factors of daily dose and lipophilicity was examined in two independent data repositories of 164 and 179 drugs labeled for their liver liabilities. Next to dose, lipophilicity is an important physicochemical property12 to affect cellular uptakes and ADMET (absorption, distribution, metabolism, excretion, toxicity) behavior13 and can be determined by the portioning of drugs between octanol and water (i.e. the logP value). Hughes et al.14 analyzed 245 preclinical compounds and observed an increased likelihood of toxic events in animal studies with highly lipophilic compounds, while Peters et al.15 reported for 213 Roche drug candidates with increased logP and increased off-target activity and found these drugs to be more toxic in in vivo studies.16 Several lines of evidence therefore suggest lipophilicity to be linked to drug toxicity, nonetheless is frequently modulated to improve bioavailability and pharmacological activity.

In the present study, we examined the relationship between daily dose, logP, and risk for DILI using two independent and publically available drug databases. As a result, a “rule-of-two” was defined that predicted hepatotoxicity more reliably when compared with dose alone. Applying the rule-of-two to the drug development process and to clinics will likely reduce risk for DILI particularly in complex comedication.

Methods and Materials

  1. Top of page
  2. Abstract
  3. Methods and Materials
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Drug Databases.

Two publicly available drug databases were used to test for the relationship between lipophilicity, daily dose, and risk for DILI. The first data set is the liver toxicity knowledge base benchmark data set (LTKB-BD) recently published by our group17 that is available from the US Food and Drug Administration's web site.18 It contains 287 drugs with the potential to cause liver injury as defined by the FDA-approved drug labels. Drugs were divided into three categories: most-DILI-concern, less-DILI-concern, and no-DILI-concern. The most-DILI-concern group contains drugs that were either withdrawn from the market due to hepatotoxicity or were assigned a black box warning for their potential to cause liver injury, or had a warnings and precautions section that specified concern about DILI that has a greater than moderate severity. The less-DILI-concern group contains drugs that specify a DILI concern in the warnings and precautions section with low severity or in an adverse reactions section, while the no-DILI-concern group comprised drugs with no DILI description mentioned in the labels. In the present study, we considered the most-DILI-concern (n = 116) and no-DILI-concern drugs (n = 48) (Supporting Table 1 and Supporting Fig. 1A).

A second data set published by Greene et al.19 was analyzed. The authors classified drugs into four groups, of which two groups were selected: those with evidence of hepatotoxicity in humans (hepatotoxic-positive) and those with no evidence of hepatotoxicity in any species (hepatotoxic-negative). Despite a difference in the approach to annotate a drug's hepatotoxic potential, the concordance between the two data sets is high (∼90%).17 After removing drugs which overlapped with the LTKB-BD and mapping to the Anatomical Therapeutic Chemical (ATC) database of the World Health Organization (WHO), a total of 179 oral drugs remained of which 115 drugs were hepatotoxic positives and 64 negatives. The flowchart for drug inclusion is shown in Supporting Fig. 1B, and a full list of drugs is given in Supporting Table 2.

Daily Dose and Lipophilicity.

Daily doses were retrieved from the WHO's ATC database (http://www.whocc.no/atc_ddd_index), supplemented with FDA-approved drug labels (http://dailymed.nlm.nih.gov/dailymed/about.cfm) and literature searches. For drugs with several reported daily doses, the average value dose was used. It has been suggested that most hepatotoxic drugs are administered at ≥100 mg/day, while few are administered at <10 mg/day.8, 11 Therefore, we defined three dose groups: daily doses <10 mg, 10-100 mg, and ≥100 mg.

A drug's lipophilicity is measured by an octanol-water partition coefficient (i.e., logP), which was calculated from the atomic-based prediction of AlogP using quantitative structure-property relationship algorithms in Pipeline Pilot (version 8.0; Accelrys Inc, San Diego, CA). Waring13 reviewed the relevant literatures and recommended that the appropriate lipoplilicity for most drugs should be in the range of 1-3. Therefore, we defined three groups: <1, 1-3, and ≥ 3.

Clinical Case Studies.

To demonstrate clinical use of the rule-of-two, information for six DILI cases was retrieved from the National Institutes of Health LiverTox database (http://livertox.nlm.nih.gov/). These cases were chosen arbitrarily. The causality assessment was performed by a panel of independent physicians/health care professionals as described in detail at http://livertox.nlm.nih.gov/.

Statistical Analysis.

The Cochran-Armitage test was applied to assess the relationship between logP and DILI in different daily dose groups. The odds ratio (OR) obtained from the logistic regression was used to measure the relative risk for DILI in a specific group. A two-sided Fisher's exact test was used to examine statistical significance of the association. The Cochran-Armitage test was performed using the “Coin” package (http://cran.r-project.org/web/packages/coin/index.html), and estimates for the OR and Fisher's exact tests were obtained using R and the “Stats” package.20

Results

  1. Top of page
  2. Abstract
  3. Methods and Materials
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Daily Dose and Lipophilicity.

First, we analyzed 164 medications of the LTKB-BD database to explore the relationship between lipophilicity, daily dose, and hepatotoxicity. As shown in Fig. 1A, at daily doses of <100 mg, no clear trend could be observed with most-DILI-concern and no-DILI-concern drugs being scattered across different logP values. In contrast, at daily doses of ≥100 mg and logP of ≥3, most-DILI-concern drugs (n = 44) were distributed into the upper right quadrant. Only two no-DILI-concern drugs appeared in this region, while no-DILI-concern drugs are associated with lower logP and daily doses, respectively.

thumbnail image

Figure 1. Relationship between daily dose, lipophilicity, and therapeutic Cmax drug levels. (A) Distribution of daily dose and logP for most-DILI-concern and no-DILI-concern drugs of the LTKB-BD data set. A total of 44 most-DILI-concern drugs appear in the upper right quadrant (i.e., the subgroup of daily doses ≥100 mg and logP ≥3), while only two no-DILI-concern drugs appeared in this region. This subgroup was associated with a significantly higher proportion of hepatotoxic drugs compared with the other subgroups (96% versus 41%; OR, 14.05; P < 0.001). (B) Correlation between daily dose and therapeutic Cmax drug levels from the LTKB-BD data set. (C) Plot of Cmax and logP for most-DILI-concern and no-DILI-concern drugs derived from the LTKB-BD data set. The Cmax data of 134 drugs were collected from the literature (Supporting Table 1). Pink circles represent most-DILI-concern drugs; green triangles represent no-DILI-concern drugs.

Download figure to PowerPoint

A Cochran-Armitage test9 was employed to assess the statistical significance of the relationship between logP, daily dose, and risk for DILI. Drugs were assigned into various subgroups defined by daily dose and logP. A summary of the prevalence of most-DILI-concern drugs for individual subgroups is given in Table 1. A statistically significant association between logP and risk for DILI was observed (P = 1.86E-7) for drugs given at daily doses of ≥100 mg. Here, 96%, 92%, and 65% were most-DILI-concern drugs with logP of either ≥3, 3-1, or <1, respectively. At daily doses of <100 mg, no statistically significant relationship between logP and hepatotoxicity was obtained.

Table 1. Distribution of Most-DILI-Concern Drugs Based on Daily Dose
 Percentage of Most-DILI-Concern DrugsP
LogP < 1 (n = 49)1 ≤ LogP < 3 (n = 47)LogP ≥ 3 (n = 68)
  1. Data were retrieved from the LTKB-BD database. Significance was determined using the Cochran-Armitage test.

Daily doses ≥100 mg (n = 106)65%92%96%1.86E-07
10 mg ≤ daily doses <100 mg (n = 43)60%56%41%0.79
Daily doses <10 mg (n = 15)60%0%20%0.13

We further assessed the relative risk for DILI associated with various doses and logP constellations. As shown in Table 2, the subgroup of daily doses ≥100 mg and logP ≥3 was associated with a significantly higher proportion of hepatotoxic drugs compared with the rest of the subgroups combined (96% versus 41%; OR, 14.05; P < 0.001). Here the false positive rate was 4% compared with 15% when daily doses of ≥100 mg were used alone. An OR of 6 was determined for drugs given at doses ≥100 mg and logP ranging from 1 to 3. LogP ≥3 alone did not yield statistical significance, and neither did a comparison of the subgroup of daily doses ≥100 mg and logP <1 (65% versus 72%). However, daily doses of ≥100 mg alone were associated with a statistically significant risk of DILI with an estimated OR of nearly 7. Conversely, for drugs with logP ≥3 and daily doses <100 mg, the OR was 0.18 (P < 0.01), suggesting reduced risk for DILI in such a constellation. It appears that the daily dose is a predominant risk factor for DILI. Nonetheless, the combination of dose and lipophilicity was associated with a significantly increased OR of 14.05.

Table 2. Distribution of Most- and No-DILI-Concern Drugs Based on LogP and Daily Dose
RuleTest ResultObserved ResultOR (95% CI)PPV%NPV%
Most-DILI-Concern (n = 116)No-DILI-Concern (n = 48)
  • Data were retrieved from the LTKB-BD database. To calculate the positive predictive value (PPV), a contingency table was prepared and true positives were divided by the sum of all positives. Likewise, the negative predictive value (NPV) was calculated from the contingency table whereby true negatives were divided by the sum of all negatives.

  • ***

    P < 0.001;

  • **

    P < 0.01;

  • *P < 0.05.

LogP ≥ 3 and daily doses ≥ 100 mgPositive44214.05*** (3.42-126.85)96%39%
Negative7246
1 ≤ LogP < 3 and daily doses ≥ 100 mgPositive2426.00** (1.29-38.45)92%33%
Negative9246
LogP < 1 and daily doses ≥ 100mgPositive22120.70 (0.29-1.70)65%28%
Negative9436
Daily doses ≥ 100 mgPositive90166.92** (3.10-15.63)85%55%
Negative2632
LogP ≥ 3Positive52161.63 (0.76-3.59)76%33%
Negative6432
LogP ≥ 3 and daily doses <100Positive8140.18** (0.06-0.51)76%64%
Negative10834
LogP < 3 and daily doses <100Positive18180.33** (0.14-0.75)75%50%
Negative9832

We also explored the relationship between logP, human therapeutic plasma concentration (i.e., Cmax), and risk for DILI for a total of 134 drugs from the LTKB-BD database. Given the good correlation between daily dose and Cmax concentration (R = 0.70) (Fig. 1B) the logP/Cmax combination should also predict risk for DILI. As depicted in Fig. 1C, most-DILI-concern drugs were associated with increased Cmax concentration and higher logP and were located in the upper-right quadrant. The OR for this subgroup (i.e., Cmax ≥1 μM and logP ≥3) was 5.68 (P = 0.002) to evidence high daily dose, systemic exposure, and high logP to be associated with increased risk for DILI.

Rule-of-Two Predicts Risk for DILI.

Our initial data analysis suggested that drugs with daily doses ≥100 mg and logP ≥3 were likely to be hepatotoxic. Therefore, the rule-of-two using daily doses of ≥100 mg and logP ≥3 was applied to an independent data set that contained 179 oral medications. As shown in Table 3, a significantly higher proportion of hepatotoxic drugs was defined by the rule-of-two positives compared with the rule-of-two negatives (85% versus 59%; OR, 3.89; P < 0.01), and the rule-of-two significantly increased the proportion of DILI drugs by reducing the false positives (i.e., six positives for the rule-of-two versus 30 positives for the >100 mg dose criteria). Likewise, as shown in Table 2, the rule-of-two performs much better, and only two compared with 16 positives are wrongly classified among no-DILI-concern drugs. Applying the rule-of-two, however, increased the false negative rate from 35% to 71% when compared with daily doses ≥100 mg alone (Table 3).

Table 3. Validation of the Rule-of-Two Using the Greene et al.19 Data Set
RuleTest ResultObserved ResultOR (95% CI)PPV%NPV%
HH (n = 115)NE (n = 64)
  • HH, evidence for hepatotoxicity in humans; NE, no evidence for hepatotoxicity in any species. The calculation of PPV and NPV is as the Table 2.

  • **

    P < 0.01;

  • *

    P < 0.05.

LogP ≥ 3 and daily doses ≥ 100 mgPositive3363.89** (1.44-11.10)85%41%
Negative8258
Daily doses ≥ 100 mgPositive75302.13* (1.09-4.16)71%44%
Negative4034
LogP ≥ 3Positive49181.90 (0.94-3.87)73%41%
Negative6646

We also analyzed 77 drugs that overlapped between the two data sets with consistent DILI annotation (Supporting Table 3). As expected, a significantly higher proportion of hepatotoxic drugs were defined by the rule-of-two positives than those of the rule-of-two negatives (95% versus 64%; OR, 11.11; P < 0.01).

Rule-of-Two Applied to Different Therapeutic Indications.

The performance of the rule-of-two was assessed among individual therapeutic categories. To this effect, the WHO's ATC classification system was used to define a drug's therapeutic categories. Drugs were classified at five different levels according to the WHO ATC system, taking therapeutic, pharmacological, and chemical properties into account.

First, drugs from the LTKB-BD and Greene et al.19 data sets were combined and mapped to the 14 main therapeutic categories as defined in the first level of the WHO's ATC therapeutic classification system. We then predicted a drug's hepatotoxic potential using the rule-of-two and assessed the performance using the correct classification rate (CCR). The CCR is an alternative measure of accuracy. In a data set with a balanced positive/negative ratio, it is equivalent to accuracy; however, it is considerably more robust than accuracy for data sets with unbalanced positive/negative ratios.21 The CCR is the average of the sensitivity and specificity of a prediction and is calculated as follows:

  • equation image

For example, consider a therapeutic category containing 10 drugs, with eight of the drugs being DILI-positive and two of the drugs being DILI-negative. If the rule-of-two predicts that all of these 10 drugs are DILI-positive, then it would have an accuracy of 80% (eight correctly classified drugs/10 total drugs), whereas it would have a CCR of only 50% ([eight correctly predicted positives/eight total positives + zero correctly predicted negatives/two total negatives]/2). As depicted in Fig. 2, the CCRs ranged from 50% to 86% depending on the therapeutic categories. Seven therapeutic categories (P, G, N, C, L, B, and A) defined by the WHO's ATC classification had a CCR of ≥ 0.6 and were therefore considered to be of high confidence defined by the rule-of-two. The other seven WHO ATC categories (D, M, S, J, R, H, and V) had lower CCRs and are therefore of low confidence. Some low confidence categories contained small numbers of drugs (<10), which might underestimate the CCR.

thumbnail image

Figure 2. The rule-of-two applied to 14 main therapeutic categories according to the WHO's ATC. The correct classification rate (CCR) plot of the WHO's ATC drugs were computed in the following ways: [true positive/(true positive + false negative) + true negative/(true negative + false positive)]/2. A, alimentary tract and metabolism; B, blood and blood-forming organs; C, cardiovascular system; D, dermatologicals; G, genitourinary system and sex hormones; H, systemic hormonal preparation, except. sex hormones, and insulins; J, anti-infectives for systemic use; L, antineoplastic and immunomodulating agents; M, musculoskeletal system; N, nervous system; P, antiparasitic product, insecticides, and repellents; R, respiratory system; S, sensory organs; V, various.

Download figure to PowerPoint

To further determine the predictive power of the rule-of-two, we analyzed the high and low confidence therapeutic categories with regard to the frequency of most- and/or no-DILI-concern drugs. We used withdrawal and/or over-the-counter (OTC) drugs, which represent two extremes in terms of safety. Withdrawal is the strongest regulatory action for a marketed drug, as a result of significant safety concern. In contrast, OTC drugs are selected by the regulatory agency on the basis of their safety with no need of a physician's prescription. There are 45 withdrawn drugs and 31 OTC drugs in the combined LTKB-BD and Greene et al. data sets (Supporting Table 4). As indicated in Table 4, the rule-of-two yielded a better estimated risk of DILI with an OR of 5.86 versus 3.81 and a lower false positive rate (18% versus 31%) than those obtained using daily doses of ≥100 mg alone. The performance is further improved if the confidence (i.e. therapeutic indication) is considered. For example, in the high confidence therapeutic categories, only one of the 15 rule-of-two positive drugs had not been withdrawn (false positive rate = 7%), compared with four of 13 rule-of-two positives in the low confidence therapeutic categories (false positive rate = 31%). The OR between withdrawn and OTC drugs for the high confidence categories is 13.00 (P < 0.01) and much higher than that for the low confidence categories (OR, 3.67; P > 0.05).

Table 4. The Rule-of-Two Applied to Withdrawn and OTC Drugs
Therapeutic CategoryRuleTest ResultObserved ResultOR (95% CI)PPV%NPV%
Withdrawn Drug (n = 45)OTC Drug (n = 31)
  • The confidence for DILI-concern drugs was evaluated in subsets of WHO ATC categories: A CCR ≥0.6 was defined as high confidence and a CCR <0.6 was defined as low confidence. To calculate the positive predictive value (PPV), a contingency table was prepared and true positives were divided by the sum of all positives. Likewise, the negative predictive value (NPV) was calculated from the contingency table whereby true negatives were divided by the sum of all negatives.

  • **

    P < 0.01;

  • *

    P < 0.05.

High confidenceLogP ≥ 3 and daily doses ≥ 100 mgPositive14113.00** (1.5-259.58)93%48%
Negative1413
Low confidenceLogP ≥ 3 and daily doses ≥ 100 mgPositive943.67 (0.68-21.02)69%62%
Negative813
AllLogP ≥ 3 and daily doses ≥ 100 mgPositive2355.86** (1.73-21.06)82%54%
Negative2226
AllDaily doses ≥ 100 mgPositive37173.81* (1.2-12.37)69%64%
Negative814
AllLogP ≥ 3Positive25141.52 (0.55-4.23)64%46%
Negative2017

Certain drugs have similar chemical structures and are in the same therapeutic category to elicit the same on-target biochemical responses. These drugs are expected to behave similarly with regard to efficacy and safety.22 We considered five drug pairs, each of which consisted of a clean and a toxic compound23 and were similar in chemical structure, displayed identical primary target activity, and exhibited no liver toxicity in preclinical studies. While the clean compound shows no sign of liver toxicity in clinical trials or postapproval, the toxic ones do.24 Discordant toxicity profiles for drug pairs represent the ultimate challenge for preclinical studies to predict reliably clinically relevant DILI. As shown in Table 5, the rule-of-two successfully identified the toxic compounds for two drug pairs that belonged to the high confidence therapeutic categories (i.e., tolcapone versus entacapone and alpidem versus zolpidem). The other three drug pairs belonged to the low confidence therapeutic categories. This emphasizes the use of the rule-of-two when considering therapeutic indication.

Table 5. Distribution of Five Published Drug Pairs Between Seven High Confidence and Seven Low Confidence Therapeutic Categories Defined by the Rule-of-Two
Drug PairTherapeutic Categories (Confidence)Hepatotoxic PotentialDaily Dose (mg)LogPRule-of-Two Test
  1. HC, high confidence; LC, low confidence.

Tolcapone/Entacapone     
 TolcaponeNervous system (HC)Toxic4503.13Positive
 EntacaponeNervous system (HC)Clean10001.658Negative
Alpidem/Zolpidem     
 AlpidemNervous system (HC)Toxic1505.468Positive
 ZolpidemNervous system (HC)Clean103.367Negative
Benoxaprofen/Ibuprofen     
 BenoxaprofenMusculoskeletal system (LC)Toxic6004.072Positive
 IbuprofenMusculoskeletal system (LC)Clean12003.607Positive
Trovafloxacin/Moxifloxacin     
 TrovafloxacinAnti-infective (LC)Toxic200−0.903Negative
 MoxifloxacinAnti-infective (LC)Clean400−0.702Negative
Bromfenac/Diclofenac     
 BromfenacMusculoskeletal system (LC)Toxic502.9Negative
 DiclofenacMusculoskeletal system (LC)Clean1004.373Positive

Application of the Rule-of-Two to Clinical Case Studies.

Information for six cases was retrieved from the National Institutes of Health LiverTox database. As summarized in Table 6, individual cases differ in the comedication regimes. Only drugs given at doses ≥100 mg/day and logP ≥3 caused severe liver injury as confirmed by an independent causality assessment of physicians and health care professionals. It is of considerable importance that none of the comedications reported in Table 6 caused liver injury, even though cases with up to eight drugs are given. Nonetheless, the comedications were given either at doses of <100 mg/day or with logP <3. In Supporting Table 5, the daily dose and logP of all comedications are summarized.

Table 6. Clinical Case Studies Whereby the Rule-of-Two Predicted Hepatotoxic Drugs in Multidrug Regimes
Sex/age* (Years)MedicationsInjury TypeClinical PresentationCausality AssessmentDaily Dose (mg)/LogPDuration of Exposure (Days)Outcome
M/74Tolcapone, levodopa, amiloride, benserazide, etilefride, hydrochlorothiazide, oxazepamHepatocellularAcute liver failureTolcapone200/3.160Death
F/65Diclofenac, furosemide, famotidine, prednisoneHepatocellularAcute liver failureDiclofenac150/4.442Death
F/45Ketoconazole, levothyroxineHepatocellularAcute liver failureKetoconazole200/3.658Liver transplantation
M/61Fenofibrate, aspirin, glibenclamide, metformin, pravastatin, nifedipine, dipyridamoleMixedAcute hepatitisFenofibrate300/3.514Recovery
M/42Disulfiram, metformin, lisinoprilHepatocellularAcute hepatitis and jaundiceDisulfiram100/5.632Recovery
M/91Ticlopidine, aspirin, prednisoneInitially hepatocellular, subsequently cholestaticAcute cholestatic hepatitisTiclopidine500/3.9120Recovery

Discussion

  1. Top of page
  2. Abstract
  3. Methods and Materials
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

To predict reliably clinically relevant DILI is an unmet challenge. We explored the relationship between daily dose and lipophilicity to improve the development of safer drugs and to avoid risk for DILI, particularly in the constellation of complex comedication regimes (see Table 6).

Notably, lipophilicity is an important physiochemical property12 and is frequently modified in an effort to optimize drug potency and ADMET behavior.12, 13, 25 The relationship of dose and lipophilicity in DILI is unknown.25 We demonstrated that drugs with high lipophilicity given at high doses likely become hepatotoxic. Strong evidence is presented for the rule-of-two to predict reliable risk for DILI across different therapeutic categories and to yield fewer false positives compared with daily dose alone.

We also explored the relationship between lipophilicity and other pharmacokinetic parameters, most notably, volume of distribution and elimination half-life. As shown in Supporting Figs. 2 and 3, a statistically significant relationship between these parameters and logP was observed, suggesting that increased lipophilicity and volume of distribution was linked to risk of DILI. Additionally, reactive metabolites contribute to risk of DILI,11, 26 and some authors have suggested that the combination of hepatic metabolism and daily dose would significantly contribute to risk for DILI.10 We therefore explored the relationship between lipophilicity and a drug's hepatic metabolism. Using the definition of Lammert et al.,10 drugs were categorized into either significant or less significant metabolism. As illustrated in Supporting Fig. 4, a statistically significant relationship between hepatic metabolism and logP was observed, suggesting that increased lipophilicity was associated with significant metabolism and risk for DILI.27 It is reasonable to assume that high lipophilicity might augment in vivo toxicological outcome based on an increased off-target activity.12, 25 Overall, high dose and increased lipophilicity is an unfavorable combination.

We further analyzed the relationship between daily dose, logP, and various types of DILI (Supporting Fig. 5). However, no clear association was observed among steatotic, cholestatic, hepatocellular, or mixed type injury classified drugs.

Research suggests that high daily dose is associated with risk of DILI. Walgren et al.8 reported that most drugs with high potential to cause severe liver injury were administrated at daily doses of ≥100 mg. Recently, Lammert et al.9 confirmed the statistically significant relationship between daily dose and poor clinical outcome for DILI, while Uetrecht11 reported that drugs at daily doses of ≤10 mg showed little risk for DILI. However, few studies have mentioned how many drugs with little or no DILI concern are also prescribed at high daily doses (≥100 mg). In the present study, we found many drugs are to be prescribed at ≥100 mg. For example, 33% of the no-DILI-concern drugs in the LTKB-BD data set and 47% of the Greene et al.19 data set had no liver liability in any species tested and are given at doses of ≥100 mg/day. Thus, daily dose alone is not a unique discriminator to predict DILI potential with many false positives that can be introduced by this criterion.

The proposed rule-of-two reduced false positives compared with daily dose alone, and this rule identified one OTC (orlistat) and 14 withdrawn hepatotoxic drugs in different high confidence therapeutic categories. For example, orlistat is used for weight loss and was approved by the FDA for OTC sale in 2006. The drug has a low bioavailability and is given at high doses.28 In May 2010, the FDA announced a safety communication to revise the OTC drug label for orlistat based on its potential to cause liver injury with 13 severe case reports.29

The study of drug pairs allowed further validation of the rule-of-two. Five drug pairs were selected, and two pairs in the high confidence therapeutic categories were classified correctly. The remaining pairs resided in the low confidence therapeutic categories. Two of these (diclofenac and ibuprofen) can be easily explained. Although it is considered safer than bromfenac, diclofenac is a most-DILI-concern drug based on drug label annotation,17 and even fatal cases have been reported for this nonsteroidal anti-inflammatory drug.30 Likewise, ibuprofen-related DILI cases have been reported.31 In contrast, the rule-of-two failed when applied to fluoroquinolones antibiotics, possibly because trovafloxacin's hepatotoxicity is mediated through its reactive metabolites and interaction with liver enriched transcription factors, most notably hepatocyte nuclear factor 4 alpha.32, 33 The metabolism of drugs is not considered by the rule-of-two.

This study has several limitations that must be considered. First, it is based on a large survey of existing literature data, and the reports might be considered as an epidemiological signal rather than factual observation. Second, we make use of the logP from the parent drugs, but not their metabolites. However, a drug's hepatotoxicity may have been mediated by its reactive metabolites. Thus, extending the rule-of-two to drug metabolites might be advantageous. Third, the thresholds used for logP and daily dose are adopted from the literature wherein its biological significance had been investigated. Although statistical significance was observed using these thresholds, the optimized cutoff of logP or daily dose should be considered in the context of specific drug classes. As shown in Table 5, the logP of bromfenac (a false negative) is 2.9, and therefore close to the threshold of logP = 3 to become a true positive. Finally, it should be noted that a rule-of-two negative is not an absolute means to measure hepatotoxic potential; its negative predictive value is relatively low (i.e., about 40%). Rather, we suggest that the rule-of-two positives are more likely hepatotoxic and could therefore be used as a supplement, but not as a replacement of safety testing strategies. It might also be applied in the decision-making of complex comedication regimes for patients treated for a variety of comorbidity. Using a drug combination based on the rule-of-two may help to avoid drug-induced toxicities, and a simple algorithm can be embedded into available software packages to alert physician when prescribing drugs.

Despite these confounding considerations, the rule-of-two is easy to implement, because logP can be calculated directly from a compound's chemical structure. Our findings bear important implications in the prioritization of drug candidates, and might assist in go/no-go decisions in the drug development candidate selection. Lastly, lipophilicity is the subject of controversial discussion in the drug discovery community; some authors argue that lipophilicity should be lowered because more lipophilic compounds are connected with higher attrition,12 whereas lipophilicity is also associated with potency on the molecular target. The present study provides evidence for high lipophilicity but low daily dose not to be associated with significant risk for DILI.

The rule-of-two can help support regulatory applications and provide guidance for clinical practice. Our findings suggest that only drugs that have both high daily dose and high lipophilicity are significantly associated with risk for liver injury. Applying the rule-of-two will significantly reduce false positives compared with daily dose alone, and may help in the causality assessment of DILI cases, especially when complicated comedication regimes are considered.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Methods and Materials
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

We thank Reagan Kelly and Hong Fang for comments and discussion. We also thank Zhichao Liu for assistance in calculating logP and Feng Qian for the graph drawing.

References

  1. Top of page
  2. Abstract
  3. Methods and Materials
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Methods and Materials
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
HEP_26208_sm_SuppFig1.tif769KSupporting Information Figure 1.
HEP_26208_sm_SuppFig2.tif779KSupporting Information Figure 2.
HEP_26208_sm_SuppFig3.tif792KSupporting Information Figure 3.
HEP_26208_sm_SuppFig4.tif767KSupporting Information Figure 4.
HEP_26208_sm_SuppFig5.tif968KSupporting Information Figure 5.
HEP_26208_sm_SuppTab1-5.doc514KSupporting Information Tables 1 to 5.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.