Metabolism and Enzyme-Mediated DDIs

In accordance with the FDA (2012) guidance, 35 NMEs approved in 2014 were evaluated in vitro as substrates, inhibitors, and/or inducers of clinically important DMEs. As substrates, the metabolic profile of 32 NMEs (91%; except 3 NMEs: namely, albiglutide, dulaglutide, and sulfer hexafluoride lipid-type A microspheres) was well characterized from in vitro studies using recombinant enzymes or human liver tissues such as human liver microsomes (HLMs) or human hepatocytes. Among those, 29 NMEs were shown to be metabolized by at least one enzyme, with the majority primarily metabolized by P450 enzymes (Fig. 1A; Table 2). Not surprisingly, CYP3A4/5 was shown to metabolize the largest number of NMEs in vitro (n = 18), although not necessarily as the major enzyme contributing to the drug’s disposition. In vivo studies further confirmed that 13 of these NMEs were indeed CYP3A substrates, with systemic exposure increases ≥25%, when coadministered with the strong CYP3A inhibitors ketoconazole (200 or 400 mg orally once daily or twice daily for 3–22 days) or itraconazole (200 mg orally once daily for 8 days), resulting in the following maximum AUC and C_max ratios, respectively: naloxegol, 12.42 and 9.12; eliglustat, 4.40 and 4.25 [CYP2D6 extensive metabolizers (EMs)]; ceritinib, 2.88 and 1.23; suvorexant, 2.79 and 1.24; olaparib, 2.59 and 1.36; netupitant, 2.42 and 1.19; vorapaxar, 1.96 and 1.93; paritaprevir, 1.84 and 1.21; idelalisib, 1.79 and 1.25; nintedanib, 1.61 and 1.79; tasimelteon, 1.45 and 1.39; dasabuvir, 1.40 and 1.16; and apremilast, 1.32 and 0.95. Interestingly, 10 of these NMEs are also substrates of P-glycoprotein (P-gp) and/or breast cancer resistance protein (BCRP) (Table 2), and inhibition of those transporters might also contribute to the observed increased exposure (details are reviewed in the following transporter section). The highest AUC and C_max ratios related to CYP3A inhibition were observed for naloxegol with concurrent use of ketoconazole (400 mg orally once daily for 5 days), confirming the primary role of CYP3A in the metabolism of the drug. In addition, coadministration of the moderate CYP3A inhibitor diltiazem increased the AUC and C_max of naloxegol by 224% and 178%, respectively. Therefore, strong CYP3A inhibitors are contraindicated with naloxegol, whereas concomitant use of moderate CYP3A inhibitors should be avoided; however, if it cannot be avoided, reduction of the naloxegol dose should be considered as indicated in the labeling (FDA, 2014t). For four of the remaining drugs with AUC ratios ≥2 in the presence of strong CYP3A inhibitors (ceritinib, eliglustat, olaparib, and suvorexant), concomitant use of strong CYP3A inhibitors is either contraindicated [suvorexant and eliglustat in CYP2D6 intermediate metabolizers (IMs) and poor metabolizers (PMs)], to be avoided [ceritinib and olaparib], or to reduce the dose [eliglustat in CYP2D6 EMs], according to the respective product labels; however, there is no such recommendation for netupitant (FDA, 2014a). As expected, all of these substrates of CYP3A were also sensitive to induction. Coadministration of rifampin (600 mg orally once daily or twice daily for 5–22 days) or carbamazepine (200 mg orally once daily or twice daily for 24 days), both strong inducers of CYP3A, significantly reduced the systemic exposure of these drugs with observed maximum AUC and C_max ratios, respectively, as follows: eliglustat, 0.04 and 0.05 (in CYP2D6 PMs); olaparib, 0.10 and 0.30; naloxegol, 0.11 and 0.26; suvorexant, 0.12 and 0.36; tasimelteon, 0.14 and 0.23; netupitant, 0.20 and 0.45; idelalisib, 0.24 and 0.43; apremilast, 0.28 and 0.57; ceritinib, 0.30 and 0.56; paritaprevir, 0.30 and 0.44; dasabuvir, 0.30 and 0.46; vorapaxar, 0.45 and 0.61; and nintedanib, 0.50 and 0.60. On the basis of these results, concomitant use of strong CYP3A inducers is contraindicated [apremilast and tasimelteon], to be avoided [ceritinib, idelalisib, nintedanib, and olaparib], not recommended [eliglustat (CY2D6 EMs, IMs, and PMs), ledipasvir, naloxegol, and netupitant], or expected to reduce efficacy [dasabuvir, paritaprevir, and suvorexant], according to the respective product labels, whereas no specific recommendation was made for vorapaxar.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Quantitation of compounds acting as substrates (NMEs) or inhibitors (NMEs and metabolites) of DMEs in vitro. (A) Phase I and II enzymes contributing to NME metabolism. (B) DMEs inhibited by NMEs (solid bars) and metabolites (striped bars). *Other enzymes include amidase, catecholamine pathway enzymes, esterases, phospholipidase, phosphatase, and proteinase. MAO, monoamine oxidase.

Other P450 isoforms, such as CYP2D6, CYP2C9, CYP2C19, CYP2C8, and CYP1A2, were also involved in the metabolism of seven, six, five, four, and four NMEs in vitro, respectively (Fig. 1A). In vivo, coadministration with specific inhibitors of CYP1A2 (fluvoxamine, 50–100 mg orally once daily for 10 days in smokers for pirfenidone or 500 mg orally once daily for 7 days for tasimelteon), CYP2C8 (gemfibrozil, 600 mg orally twice daily for 5 days), or CYP2D6 (paroxetine, 30 mg orally once daily for 10 days in CYP2D6 EMs) further identified dasabuvir (CYP2C8), eliglustat (CYP2D6), pirfenidone (CYP1A2), and tasimelteon (CYP1A2) as sensitive substrates, with AUC and C_max ratios of 9.90 and 1.91, 10.00 and 8.20, 6.81 and 1.78, and 6.87 and 2.28, respectively. In addition, several NMEs were found to be primarily metabolized by non-P450 enzymes: droxidopa, which is metabolized by the catecholamine pathway, including l-amino acid decarboxylase; tedizolid phosphate, a prodrug, which is converted by nonspecific endogenous phosphatases to tedizolid, the active moiety after oral or intravenous administration; and finally, dapagliflozin, which is primarily metabolized by UDP-glucuronosyltransferase (UGT) UGT1A9 in vitro and dapagliflozin AUC and C_max were increased by 51% and 13%, respectively, when coadministered with the strong UGT1A9 inhibitor mefenamic acid.

When NMEs were considered as perpetrators, 32 (91%; except 3 NMEs namely albiglutide, dulaglutide, and sulfer hexafluoride lipid-type A microspheres) were investigated in vitro for the potential to inhibit DMEs using HLMs or cDNA-expressed enzymes to determine the inhibitory mechanisms [e.g., reversible or time-dependent inhibition (TDI)] and inhibition potency. Twenty-four NMEs inhibited at least one P450 enzyme or UGT (Table 3), with the most affected enzymes being CYP3A4 (n = 15), CYP2C8 (n = 12), CYP2C9 (n = 11), CYP2C19 (n = 9), CYP2D6 (n = 9), CYP2B6 (n = 8), CYP1A2 (n = 6), and UGT1A1 (n = 6) (Fig. 1B). In addition, the inhibitory potential of 13 major metabolites of 8 NMEs was evaluated, and inhibition of P450 enzymes and UGT1A1 was also observed by these compounds (Table 3). With regard to the mechanism of inhibition, 10 NMEs and six metabolites were evaluated for TDI of P450 enzymes and a majority, comprising seven NMEs and five metabolites, showed TDI of one or more P450 enzyme, in particular, CYP3A4/5. In addition, both eliglustat and its metabolite Genz-120965 (N-oxide of eliglustat) inhibited CYP2D6 in HLMs in a time-dependent manner, with K_I values of 1.05 and 8.44 µM and k_inact values of 0.0151 and 0.206 min⁻¹, respectively.

Using the in vitro inhibition results, as well as plasma concentration data, the in vivo DDI risk was predicted by the sponsors by estimating intrinsic clearance values in the presence and absence of an inhibitor, and the R value was calculated utilizing a basic model according to the FDA drug interaction guidance (R₁ = 1 + [I]/K_i, for reversible inhibition) (FDA, 2012). More complex models were also used, such as PBPK modeling, which is reviewed in a subsequent section. On the basis of the R₁ values, the majority of the in vitro inhibitory interactions were not considered clinically relevant (R₁ ≤ 1.1). Among drugs with R₁ > 1.1, in vivo studies with sensitive P450 substrates found only nine NMEs with positive enzyme inhibition: idelalisib was a strong inhibitor of CYP3A (midazolam: AUC ratio = 5.15; C_max ratio = 2.31), netupitant was a moderate inhibitor of CYP3A (midazolam: AUC ratio = 2.44; C_max ratio = 1.40), eliglustat was a moderate inhibitor of CYP2D6 (metoprolol: AUC ratio = 2.33; C_max ratio = 1.72), and the combination drug paritaprevir, ritonavir, ombitasvir, and dasabuvir administered as Viekira Pak was a moderate inhibitor of UGT1A1 (raltegravir: AUC ratio = 2.26; C_max ratio = 2.27). In addition, ledipasvir (in combination with sofosbuvir) and suvorexant were weak inhibitors of CYP3A (midazolam: AUC ratio = 1.47; C_max ratio = 1.23; and atazanavir: AUC ratio = 1.33; C_max ratio = 1.06, respectively) and oritavancin was a weak inhibitor of CYP2C9 (S-warfarin: AUC ratio = 1.32; C_max ratio not available). Several drugs with R₁ values > 1.1 were not evaluated for clinical inhibition. However, for ceritinib, for example, which was shown in vitro to be a potent inhibitor of CYP2C9 (K_i = 0.24 µM; R₁ = 8.5) and CYP3A4 (midazolam: K_i = 0.16 µM; R₁ = 12.3; testosterone: IC₅₀ = 0.2 µM; R₁ = 19.0, assuming competitive inhibition), the in vivo drug interaction evaluation with sensitive probe substrates of these two enzymes was requested as a postmarketing requirement (PMR).

A significant number of NMEs (n = 12) showed some inhibition of CYP2C8 in vitro; however, based on R₁ values (R₁ ≤ 1.1), 10 of these drugs were not likely to show any clinically relevant inhibition [two NMEs, tasimelteon and vorapaxar, were still evaluated in vivo and, not surprisingly, did not affect the pharmacokinetics (PK) of the coadministered CYP2C8 substrate rosiglitazone]. For belinostat and idelalisib, R₁ was greater than 1.1; however, the clinical relevance of these inhibitory interactions was not investigated. Similarly, nine NMEs inhibited CYP2C19 in vitro, of which six had R₁ values ≤ 1.1. Of the three NMEs with R₁ > 1.1, oritavancin did not significantly increase the exposure of the CYP2C19 substrate omeprazole, whereas idelalisib and suvorexant were not evaluated.

In terms of enzyme induction potential, 29 NMEs were evaluated using human hepatocytes, and 10 were found to induce DME expression or activity as well as activate human pregnane X receptor (PXR) to some extent in vitro (Table 4): apremilast (CYP3A4), belinostat (CYP1A2), ceritinib (CYP2C9 and CYP3A4), idelalisib (CYP2B6, CYP2C8, CYP2C9, CYP3A4, UGT1A1, and UGT1A4), ledipasvir (CYP2B6, CYP3A4, and PXR), olaparib (CYP2B6), pirfenidone (CYP2C19 and CYP3A4), suvorexant (CYP1A2, CYP2B6, CYP3A4, and PXR), tasimelteon (CYP2B6, CYP2C8, and CYP3A4), and vorapaxar (CYP1A2 and CYP2B6). Nuclear receptors were not commonly investigated; only four NMEs (dapagliflozin, ceritinib, ledipasvir, and suvorexant) were evaluated for PXR activation and one (ledipasvir) for aryl hydrocarbon receptor activation together with P450 enzyme induction. As a result, ledipasvir and suvorexant showed positive PXR activation. One of the metabolites of tasimelteon, M12, also showed some induction of CYP1A2 and CYP2B6. In addition, activation of CYP2E1 was observed for vorapaxar, with a 300% increase in activity at 30 µM in HLMs. For most of the drugs, however, the in vitro induction results were observed at concentrations much higher than the expected clinically relevant concentrations (C_max values are presented in Table 4). Therefore, considering their low systemic exposure and high protein binding, these interactions are unlikely to have any clinical relevance. Indeed, in vivo, only the combination drug ledipasvir and sofosbuvir was found to be a weak CYP2B6 inducer, decreasing the AUC and C_max of the coadministered probe substrate efavirenz by 21%. Interestingly, the majority of the in vitro inducers also showed inhibition of the same P450 enzyme (Table 3). For example, suvorexant was found to increase CYP3A4 mRNA by 22.0-fold at 20 µM (42.7% of positive control rifampin) as well as activate PXR (33% of rifampin) at 10 µM in human hepatocytes (suvorexant C_max = 1.0 µM); however, it also inhibited CYP3A4 both directly (IC₅₀ = 4 µM) and in a time-dependent manner (K_I = 11.6 µM; k_inact = 0.136 min⁻¹) in HLMs. In vivo, overall inhibition of CYP3A was observed, with 47% and 23% respective increases in AUC and C_max of the coadministered CYP3A probe substrate midazolam. Similarly, ceftolozane and tazobactam, coadministered as a combination drug, as well as the tazobactam metabolite M1, were all found to reduce the expression and activity of CYP1A2, CYP2B6, and CYP3A4 in vitro, and they also inhibited these enzymes. However, when tested with in vivo probe substrates for CYP1A2 (caffeine) and CYP3A (midazolam), no significant effect was observed (the clinical impact on CYP2B6 was not evaluated).

In summary, regarding DMEs, CYP3A was involved in the metabolism of the most NMEs in vitro (17 of 35), 14 of which were further confirmed to be substrates of CYP3A in vivo. As perpetrators, 24 drugs showed positive inhibition and/or induction toward at least one enzyme in vitro; however, only one-third were found to affect the exposure of a clinical probe (AUC or C_max ratio ≥1.25 or ≤0.8), highlighting the challenge of translating inhibition and induction data from in vitro to in vivo.

Transport and Transporter-Mediated DDIs

Of the 30 NDA approval packages released by the FDA in 2014, 22 (73%) contained in vitro transporter data, either substrate assays, inhibition assays, or both. Although this is a lower percentage of overall approvals than was seen in the previous year (20 of 25 NDA approval packages from 2013 contained transporter data), the overall number of compounds (drugs and metabolites) tested against transporters increased. As a result of multiple combination drugs, there were 25 NMEs represented in the 22 NDA approval packages containing transporter assays. In addition, 17 individual metabolites were also evaluated; therefore, 42 new compounds were screened for in vitro transporter interactions in the approvals from 2014 (also screened was a pool of 10 metabolites of eliglustat). To follow up the in vitro findings, nine drugs (a total of 10 NMEs) were studied in vivo as substrates for P-gp, organic anion-transporting polypeptides OATP1B1/3, and organic anion transporter OAT3 using clinical inhibitors and/or inducers. A total of 23 clinical studies were performed, with 20 showing positive results (AUC or C_max ratio ≥1.25 or ≤0.8). As perpetrators, 10 drugs (a total of 13 NMEs) were evaluated clinically for the inhibition of P-gp, OATP1B1/3, OAT1/3, organic cation transporter OCT2, and BCRP. Of the 14 clinical studies conducted, one-half showed positive results.

In addition to an increase in the number of compounds screened in the 2014 approval packages, the number of total assays described also jumped from just over 120 assays in 2013 to over 450 assays in 2014, with more than two-thirds of the assays using the new compound as a prospective inhibitor. This was a result of not only more compounds being tested but also more transporters tested, as well as more transporters screened per compound. In the 2013 approvals, 16 transporters were evaluated, whereas experiments involving 19 transporters were described in the 2014 documentation. In addition to the seven transporters explicitly mentioned in the 2012 FDA guidance documents (P-gp, BCRP, OATP1B1, OATP1B3, OAT1, OAT3, and OCT2; FDA, 2012), assays involving the following transporters were also described: OATP2B1, OAT2, OCT1, multidrug and toxin extrusion proteins MATE1 and MATE2-K, bile salt export pump, and multidrug resistance-associated proteins MRP1–MRP5 and MRP8.

Not surprisingly, P-gp was once again the most represented transporter with regard to substrate assays screened as well as positive substrate interactions identified (Fig. 2A). A total of 33 compounds were tested, with 19 positive interactions (i.e., efflux ratio ≥2), comprising 15 NMEs and four metabolites. Interestingly, for only one of the metabolites shown to be a substrate of P-gp was the parent compound also a P-gp substrate (dasabuvir and the metabolite M1); however, in the case of netupitant (in which the metabolite M2 was a substrate of P-gp), the potential for the parent compound to be a substrate was not fully examined and further in vitro experiments have been requested as a PMR (FDA, 2014a). Of the 15 NMEs identified as P-gp substrates (apremilast, belinostat, ceritinib, dapagliflozin, dasabuvir, eliglustat, empagliflozin, idelalisib, ledipasvir, naloxegol, nintedanib, olaparib, olodaterol, ombitasvir, and paritaprevir), 10 NMEs were tested in vivo as P-gp substrates, and nine positive interactions were identified, with the largest interaction observed when ledipasvir was coadministered with simeprevir (ledipasvir: AUC ratio = 1.88; C_max ratio = 1.78). Coadministration with the P-gp inhibitor verapamil also resulted in a ledipasvir AUC ratio of 1.66 and a C_max ratio of 1.21, although no effect was observed with cyclosporine, also a P-gp inhibitor. Similarly, verapamil had no effect on the AUC of empagliflozin. Interactions of a comparable magnitude (1.25 ≤ AUC ratio < 2) were observed with coadministration of ketoconazole (also a CYP3A inhibitor) with apremilast, dasabuvir, idelalisib, nintedanib, olodaterol, and paritaprevir (all of which, with the exception of olodaterol, are also CYP3A substrates); therefore, these interactions are likely mediated by P450 enzymes in addition to possible transporter influence. In addition, the P-gp inhibitor quinidine had an effect on the PK of naloxegol (AUC ratio = 1.39; C_max ratio = 2.43). Regarding induction of P-gp, both rifampin (idelalisib: AUC ratio = 0.24; C_max ratio = 0.43; ledipasvir: AUC ratio = 0.40; C_max ratio = 0.69; and nintedanib: AUC ratio = 0.50; C_max ratio = 0.60) and carbamazepine (dasabuvir: AUC ratio = 0.30; C_max ratio = 0.46; and paritaprevir: AUC ratio = 0.30; C_max ratio = 0.44) were tested. However, because both of these compounds also induce P450 enzymes, the effects observed are likely due to the combination of transporter and P450 enzyme induction.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Quantitation of compounds acting as substrates (NMEs) or inhibitors (NMEs and metabolites) of transporters in vitro. (A) Transporters involved in transport of NMEs. (B) Transporters inhibited by NMEs (solid bars) and metabolites (striped bars). *The specific OAT isoform was not determined.

For inhibition assays, although P-gp was also the most commonly screened transporter, OATP1B1 had more positive inhibitory interactions (14 parents and seven metabolites; Fig. 2B), whereas OATP1B3 (13 parents and six metabolites), P-gp, and BCRP all had 19 positive inhibitory interactions. However, of the 16 NMEs showing positive interactions with OATP1B1 and/or OATP1B3, only five NMEs (dasabuvir, idelalisib, ledipasvir, olaparib, and paritaprevir) had C_max/IC₅₀ values greater than the FDA guidance cut-off of 0.1. Two of those NMEs (ledipasvir and olaparib) had subsequent R values less than the FDA guidance cut-off value of 1.25 and therefore did not necessitate an in vivo DDI study. For the remaining NMEs, in vivo studies with various statins, which are known substrates of OATP1B1/3, were performed (Table 5). No effect of idelalisib was observed on rosuvastatin PK, nor did empagliflozin affect the PK of simvastatin. However, the combination drug, Viekira Pak (containing paritaprevir, ritonavir, ombitasvir, and dasabuvir), did significantly affect pravastatin and rosuvastatin exposure (AUC ratio = 1.82 and 2.59, respectively; C_max ratio = 1.36 and 7.15, respectively), although the rosuvastatin interaction may be partially mediated by BCRP, because paritaprevir, ritonavir, and dasabuvir were all shown to inhibit BCRP, as well as OATP1B1 and OATP1B3, in vitro.

Nineteen compounds, comprising 13 NMEs and six metabolites, were shown to be inhibitors of P-gp in vitro. With respect to metabolites, in contrast with the substrate assays, the parent compounds for all metabolites showing inhibition of P-gp were also inhibitors (dasabuvir and metabolite M1, netupitant and metabolites M1–M3, nintedanib and metabolite M2, and suvorexant and metabolite M9). For six NMEs, in vitro inhibition was minimal and therefore no in vivo study was triggered (apremilast, netupitant, nintedanib, olodaterol, pirfenidone, and tedizolid; Table 6); however, a clinical study was still performed with netupitant and the P-gp probe substrate digoxin, but no in vivo effect was observed (digoxin: AUC ratio = 1.04; C_max ratio = 1.10). Four of the remaining compounds showed no potential systemic interactions ([I]₁/IC₅₀ was below the regulatory cut-off value of 0.1); however, the intestinal interaction potential was greater than the cut-off value ([I]₂/IC₅₀ > 10, where [I]₂ is the maximal therapeutic dose, in moles, divided by 250 milliliters). Therefore, in vivo DDI studies with digoxin were performed. Both paritaprevir and dasabuvir were inhibitors of P-gp in vitro; however, the combination drug Viekira Pak containing both compounds (as well as ritonavir, which also showed in vitro inhibition of P-gp, and ombitasvir, which did not) had no significant effect on digoxin AUC or C_max (AUC ratio = 1.16; C_max ratio = 1.14), whereas eliglustat and suvorexant increased digoxin exposure (AUC ratio = 1.49 and 1.27; C_max ratio = 1.71 and 1.21, respectively). Finally, two NMEs, idelalisib and vorapaxar, showed possible clinically relevant inhibition potential at both the systemic and intestinal levels, although neither showed a significant effect on digoxin AUC in vivo (AUC ratio = 1.00 and 1.05, respectively). However, significant effects on digoxin C_max were observed with both compounds (C_max ratio = 1.25 and 1.54, respectively). The largest in vivo effect observed was with ledipasvir and simeprevir, in which coadministration resulted in the simeprevir AUC and C_max ratios of 2.84 and 2.56, respectively; however, this interaction may be partially mediated by BCRP as well, because ledipasvir has been shown to be an inhibitor of BCRP in vitro and simeprevir is a substrate of at least the mouse homolog of BCRP (Bcrp1) (FDA, 2013a). This interaction is also particularly of interest, because both simeprevir and ledipasvir (combined with sofosbuvir) are indicated for the treatment of hepatitis C; however, based on this interaction, coadministration is not recommended (FDA, 2014k). One NME, dapagliflozin, showed no inhibitory potential of P-gp in vitro; however, the sponsor still performed an in vivo DDI study with the P-gp substrate digoxin, although no effect was observed (digoxin AUC ratio = 1.00).

In terms of BCRP inhibition, a total of 19 compounds, comprising 11 NMEs and eight metabolites, were shown to be inhibitors in vitro (Table 7). Similar to P-gp, the parent compounds for all of the metabolites showing inhibition of BCRP were also BCRP inhibitors, with the exception of tasimelteon and its M9 metabolite. Five NMEs (dasabuvir, netupitant, paritaprevir, suvorexant, and tedizolid) were predicted to be involved in possible clinical DDIs, based on both the [I]₁/IC₅₀ and [I]₂/IC₅₀ values; however, in the case of suvorexant, both values only slightly exceed the cut-off values of 0.1 (0.117) and 10 (28), respectively, and thus the sponsor did not expect clinically relevant inhibition, especially because the [I]₂/IC₅₀ value drops to 10.6 at the therapeutic dose of 15 mg (FDA, 2014c). Regarding the four remaining NMEs (dasabuvir, netupitant, paritaprevir, and tedizolid), only one in vivo study was performed using rosuvastatin and the combination drug Viekira Pak (paritaprevir, ritonavir, ombitasvir, and dasabuvir), which resulted in a rosuvastatin AUC ratio of 2.59 and a C_max ratio of 7.15, although as mentioned previously, this interaction is likely at least partially mediated by OATP1B1 and/or OATP1B3 as well. In the case of tedizolid, an in vivo drug interaction study with a BCRP substrate was recommended (FDA, 2014ad). For netupitant, although not a PMR, the reviewers noted that “since total C_max/IC₅₀ is greater than 0.1 for parent drug netupitant, a follow-up in vivo study may be recommended” (FDA, 2014a).

In summary, 15 NMEs were shown to be substrates of P-gp in vitro, nine of which were tested in in vivo DDI studies, resulting in eight NMEs showing a positive interaction (AUC ratio of the NME ≥1.25 when administered with a P-gp inhibitor). However, most of these NMEs are also CYP3A substrates, and the DDI studies were performed with the strong CYP3A inhibitor ketoconazole; therefore, the interactions observed are likely mediated by P450 enzymes as well as P-gp. Regarding inhibition, three NMEs (namely, ledipasvir, eliglustat, and suvorexant) showed inhibition of P-gp in vivo, with AUC ratios of 2.84 (substrate: simeprevir), 1.49 (substrate: digoxin), and 1.27 (substrate: digoxin), respectively, whereas dasabuvir and paritaprevir were shown to be moderate OATP inhibitors in vivo (as well as likely inhibitors of BCRP). These data are in contrast with in vitro data, where of the 25 NMEs tested, 21 had positive inhibitory interactions with one or more transporter. Of the four compounds with no positive inhibitory interactions, three (belinostat, oritavancin, and peramivir) were only tested against P-gp, whereas naloxegol was screened against multiple transporters and no interaction was observed. The same trend was observed in the 2013 NDA approval packages, in which a majority of the compounds tested were inhibitors of transporters in vitro; however, most of these interactions were not clinically relevant, indicating that further research in the transporter field is required to improve the predictive value of in vitro experiments.

Pharmacogenetic Studies

For four NMEs (dapagliflozin, eliglustat, nintedanib, and olodaterol), the effects of genetic variants of the primary enzymes for metabolic clearance (namely UGT1A9, CYP2D6, UGT1A1, UGT1A1, UGT1A7, UGT1A9, and UGT2B7, respectively) on the PK of each drug were evaluated. Eliglustat, which is primarily metabolized by CYP2D6 and CYP3A4, displayed a significant effect of CYP2D6 polymorphisms on its disposition. When eliglustat was administered at the dose regimen of 84 mg orally twice daily for 10 days, its AUC and C_max were significantly reduced by 91% and 88%, respectively in CYP2D6 ultra-rapid metabolizers (CYP2D6*1/*2 dup), and increased by 7.39- and 5.48-fold in PMs (CYP2D6*4/*4, CYP2D6*4/*5, and CYP2D6*4/*6), compared with CYP2D6 EMs (CYP2D6*1/*1 and CYP2D6*2/*2). By contrast, the PK of eliglustat was not significantly affected in CYP2D6 IMs compared with EMs when dosed at 84 mg twice daily for 52 weeks; hence, no dose adjustment is required for CYP2D6 IMs. Since eliglustat has the potential to prolong the QT interval, CYP2D6 PMs are at risk of cardiac toxicity due to the expected elevated plasma exposure; therefore, the recommended dose is reduced to 84 mg once daily in CYP2D6 PMs. On the other hand, in CYP2D6 ultra-rapid metabolizers, eliglustat may not achieve adequate concentrations for the therapeutic effect and its use is limited in this patient population. On the basis of these results, it is necessary to determine each patient’s CYP2D6 genotype prior to the administration of eliglustat to define a proper dosage regimen (FDA, 2014e).

In the case of nintedanib, which is metabolized by esterases to M1 and then subsequently glucuronidated by UGT enzymes to M2, there was no significant effect of the UGT1A1*28 allele on the exposure to nintedanib; by contrast, there was a significant decrease in both the AUC and C_max (62% and 63%, respectively) of the glucuronide metabolite M2 in individuals homozygous for the variant. However, since M2 does not appear to be biologically active, no dosing recommendations based on UGT1A1 genotype have been made. As for dapagliflozin and olodaterol, none of the tested UGT polymorphisms (UGT1A9*2 and UGT1A9*3 for dapagliflozin, and UGT1A1*28/36/37, UGT1A1*60, UGT1A1*93, UGT1A7*2, UGT1A7*3, UGT1A7*4, UGT1A7*12, UGT1A9*3, and UGT2B7*2 for olodaterol) affected the systemic exposure of the respective NME.

PBPK Modeling and Simulations

In recent years, PBPK modeling has become an important tool in drug development and is increasingly accepted by regulatory agencies in lieu of clinical studies in certain circumstances (Rowland et al., 2011; Zhao et al., 2012; Huang et al., 2013; Wagner et al., 2015). Indeed, PBPK approaches have been included in the recent regulatory guidance documents on DDIs (European Medicines Agency, 2012; FDA, 2012; Pharmaceuticals Medical Devices Agency, 2014), PGx (European Medicines Agency, 2011; FDA, 2013b), pediatrics (FDA, 2014ap), and HI (European Medicines Agency, 2005) as a tool to guide clinical study design and labeling decisions. Among the drugs approved in 2014, PBPK modeling was used in at least one clinical situation for six NMEs: belinostat, PGx; blinatumomab, DDIs; ceritinib, DDIs and absorption; eliglustat, DDIs and complex drug–gene interactions; naloxegol, DDIs; and olaparib, DDIs. The modeling results for four of these drugs (namely, ceritinib, eliglustat, naloxegol, and olaparib) were used directly to inform dose recommendations with moderate inhibitors and inducers. For ceritinib, PBPK modeling predicted that fluconazole, a moderate inhibitor of CYP3A, may increase ceritinib exposure by 37%, whereas efavirenz, a moderate CYP3A inducer, may decrease ceritinib exposure by 43%; therefore, it is not recommended to restrict concomitant use of these drugs with ceritinib. In the case of eliglustat, complex drug–PGx interaction scenarios were simulated with coadministration of moderate or strong CYP2D6 and/or CYP3A inhibitors in subjects with different CYP2D6 metabolizing status (EMs, IMs, and PMs). Alteration in the daily dose is recommended based on the predicted interaction results (FDA, 2014e). For example, PBPK simulations showed that if eliglustat (84 mg twice daily) was coadministered with paroxetine (a strong CYP2D6 inhibitor) together with ketoconazole (a strong CYP3A4 inhibitor), or terbinafine (a moderate CYP2D6 inhibitor) together with fluconazole (a moderate CYP3A4 inhibitor), its exposure would increase by 24.16- and 13.58-fold in CYP2D6 EMs, respectively, and 9.81- and 4.99-fold in CYP2D6 IMs, respectively. Therefore, administration of eliglustat with strong or moderate CYP2D6 inhibitors concomitantly with strong or moderate CYP3A4 inhibitors is contraindicated in CYP2D6 EMs and IMs. Considering the safety margins of eliglustat, performing such clinical studies may have resulted in unsafe overexposure to eliglustat, leading to possible adverse events and/or toxicities, highlighting the utility of using PBPK modeling in place of clinical evaluations in specific situations. In the case of naloxegol, PBPK simulations with moderate CYP3A4/P-gp inhibitors (erythromycin, fluconazole, and verapamil) predicted increases in naloxegol exposure of 2.21- to 4.63-fold (minimal PBPK model). Therefore, the concomitant use of naloxegol with moderate CYP3A4/P-gp inhibitors should be avoided or, if unavoidable, the dose of naloxegol should be reduced (note that the use of naloxegol with strong CYP3A4/P-gp inhibitors is contraindicated). Similarly, olaparib, which is primarily metabolized by CYP3A, was evaluated in clinical studies with the strong CYP3A inhibitor itraconazole and the strong inducer rifampin, whereas the DDI risk with concomitant use of moderate CYP3A inhibitors or inducers was predicted through PBPK modeling. It was predicted that fluconazole would likely increase olaparib AUC by 2.26-fold and efavirenz would likely decrease olaparib AUC by 59%. Therefore, a dose reduction to 200 mg twice daily (original dose of 750 mg once daily) is recommended for concomitant use of a moderate CYP3A inhibitor. On the other hand, because increasing the dose could be impractical given the number of capsules to be administered, it is recommended that concomitant use of a moderate CYP3A inducer should be avoided. If a moderate CYP3A inducer must be coadministered, it may result in reduced efficacy (FDA, 2014s).

Clinically Significant DDIs

As in our prior publication (Yu et al., 2014), the measure of the exposure to the victim drug (AUC and C_max) with and without coadministration of the perpetrator (AUC and C_max ratios) was used to evaluate the clinical significance of the DDI study results. As discussed previously, even though the concentration ratio represents only one of the factors to consider when analyzing the possible clinical impact of a drug interaction, this is a simple metric that can be applied across all studies, irrespective of the substrate evaluated. Since a 2-fold change in drug exposure will often trigger dosing recommendations, an AUC or C_max ratio ≥2 for inhibition and ≤0.5 for induction was considered in this analysis as a cut-off for further consideration. However, for completeness and to take into account the various substrates’ therapeutic ranges, drugs with changes in exposure smaller than 2-fold (1.25 ≤ AUC ratio < 2 for inhibition, and 0.5 < AUC ratio ≤ 0.8 for induction) but still triggering dosing recommendations or specific monitoring are presented in Supplemental Table 2. Overall, it was found that 17 of the 35 NMEs analyzed (46%) had at least one in vivo DDI study with a change in exposure of clinical significance, with NMEs being mainly victim drugs. Maximum clinically significant inhibition and induction results observed with NMEs as victims or perpetrators are presented in Table 8 (inhibition) and Table 9 (induction). For inhibition studies, alteration of CYP3A activity was the most common underlying mechanism, explaining one-half of the results. A majority of the NMEs (n = 12) were affected by the interaction as victims, whereas seven NMEs were perpetrators (five being both victims and perpetrators). On the other hand, induction studies were all related to NMEs as victim drugs and, in most cases, involved induction of CYP3A by the known inducers rifampin or carbamazepine.

The largest change in exposure was observed with the combination drug Viekira Pak (paritaprevir, ritonavir, ombitasvir, and dasabuvir) as a perpetrator, which drastically increased the exposure of the CYP3A and P-gp substrate tacrolimus upon coadministration (AUC ratio = 57.07; C_max ratio = 16.48). Consequently, dose reduction of tacrolimus and close monitoring of its blood concentrations are recommended when coadministered with Viekira Pak (FDA, 2014ai). On the other hand, the exposure of one of the constituents of Viekira Pak (paritaprevir), a substrate of CYP3A and of various transporters (OATPs, BCRP, and P-gp), was increased 47.43-fold when coadministered with the potent inhibitor ritonavir. Indeed, ritonavir is used in the combination drug for its boosting effect on paritaprevir concentrations (FDA, 2014ai). Paritaprevir was also sensitive to induction by carbamazepine, with decreases in AUC and C_max of 70% and 56%, respectively. Both interactions are highlighted in Viekira Pak prescribing information (FDA, 2014ai). Dasabuvir, another component of Viekira Pak, was also found to be sensitive to inhibition of CYP2C8 by gemfibrozil, with a change in AUC of almost 10-fold (interestingly, the effect of gemfibrozil on dasabuvir C_max was limited, with a change of only 1.91-fold). In addition, when the full drug combination (paritaprevir, ritonavir, ombitasvir, and dasabuvir) was tested as a perpetrator, a series of substrates were significantly inhibited: namely, amlodipine (CYP3A), buprenorphine (CYP3A), cyclosporine (CYP3A/P-gp), ketoconazole (CYP3A), norgestimate (CYP3A and UGT), raltegravir (UGT1A1), rilpivirine (CYP3A), ritonavir (CYP3A/P-gp), and rosuvastatin (OATP1B1/3 and BCRP), with AUC and C_max ratios of 2.57 and 1.26, 2.05 and 2.00, 5.80 and 15.73, 2.15 and 1.13, 2.64 and 2.46 (for the metabolite norgestrel), 2.26 and 2.27, 3.40 and 2.93, 2.78 and 2.54, and 2.60 and 7.15, respectively. The multitude of interactions observed with Viekira Pak—which mechanistically involve the constituents as both victims and perpetrators, including both inhibition and induction of several enzymes and transporters—highlight the challenges of managing drug interactions in the clinic with such complex combination drugs.

Another NME particularly sensitive to both inhibition and induction of its metabolism was eliglustat. As mentioned above, eliglustat is metabolized by CYP2D6 and to a lesser extent CYP3A4. Coadministration of the strong CYP2D6 inhibitor paroxetine or the strong CYP3A4 inhibitor ketoconazole in healthy CYP2D6 EMs increased eliglustat AUC by 10.00- and 4.40-fold, respectively, and C_max by 8.20- and 4.25-fold, respectively, whereas the strong inducer rifampin significantly decreased eliglustat plasma levels, especially in CYP2D6 PMs (AUC ratio = 0.04; C_max ratio = 0.05). On the other hand, eliglustat was also found to be an inhibitor of CYP2D6, increasing the AUC of the probe substrate metoprolol 2.33-fold and C_max 1.72-fold in CYP2D6 EMs. As previously discussed, results of extensive PBPK modeling were used to guide eliglustat dosing recommendations in complex situations of multiple impairment and/or in patients with various degrees of CYP2D6 expression (FDA, 2014e). In addition, two NMEs were found to be sensitive substrates of CYP1A2: tasimelteon and pirfenidone, with AUC ratios of 6.87 and 6.81, respectively, when coadministered with the strong CYP1A2 inhibitor fluvoxamine. Of note, fluvoxamine also inhibits CYP3A, CYP2C9, and CYP2C19, which are also involved to some extent in the metabolism of tasimelteon. Both the tasimelteon and pirfenidone labels include dosing recommendations related to cigarette smoking, which is known to induce CYP1A2 activity (FDA, 2014i,l).

Regarding transporter-based clinical interactions, there were only a few drug interactions with over 2-fold changes in substrate exposure, which could be explained purely by alteration of transport. The interactions involving ledipasvir as a victim (2.05-fold increase in AUC when coadministered with atazanavir boosted with ritonavir, and 60% decrease in AUC with rifampin coadministration) and as a perpetrator (increase in simeprevir AUC by 2.84-fold) involved mainly inhibition and induction of P-gp. Also, as discussed above, the effect of Viekira Pak on rosuvastatin disposition can be mainly explained by inhibition of OATP1B1/3 and BCRP.

In conclusion, when a cut-off of 2-fold change in drug exposure was considered for clinical relevance, approximately one-half of the drugs analyzed had clinically significant DDIs, most of which related to the NMEs as victim drugs. Not surprisingly, the underlying mechanism for a large number of these interactions was inhibition or induction of CYP3A.

Hepatic and Renal Impairment Studies

It is well recognized that organ impairment can significantly affect a drug’s plasma exposure, and, in some situations, may affect its safety and efficacy. The probability and extent of these effects in a given patient population will significantly differ depending on the severity of impairment of these eliminating organs. Therefore, the FDA recommends that sponsors conduct organ impairment studies if HI and/or RI might affect a drug and/or its active metabolites’ PK, or if the drug might be used in these respective populations (Yeung et al., 2015). For the purpose of this review, the AUC and C_max ratios (impaired/control) were considered as a standard outcome measurement of the effect of various degrees of organ impairment on the NMEs, using values observed in patients with HI or RI versus those observed in control healthy populations. Similar to the in vivo clinical significance evaluation presented in the previous section, an AUC ratio ≥2 was considered as a cut-off to systematically evaluate the NDAs for any dosing and labeling recommendations.

Overall, 24 NMEs were assessed for the effect of HI and/or RI on the drug’s PK. Among the 19 NMEs evaluated for HI studies, four demonstrated an AUC ratio ≥2 in patients with HI (mild, moderate, and severe, Child-Pugh score A, B, and C, respectively) versus healthy controls, resulting in dosing recommendations, whereas six NMEs had AUC ratios <2 but still reported dosing recommendations in these populations. In addition, dosing recommendations were given for two NMEs for which the sponsor did not conduct dedicated HI studies (Table 10). Among the four NMEs for which systemic exposure was increased by ≥2-fold, all are extensively metabolized by the liver. Three (dasabuvir, netupitant, and paritaprevir) are mainly eliminated via biliary excretion, whereas tasimelteon is primarily eliminated by renal excretion as metabolites. The highest change in exposure was observed for paritaprevir (a constituent of the combination drug Viekira Pak) in patients with severe HI, showing a 5.25-fold increase in AUC and a 2.21-fold increase in C_max, whereas the AUC and C_max ratios were 1.68 and 1.17, respectively, in patients with moderate HI. Dasabuvir, another NME in the same combination drug, demonstrated a 3.51-fold increase in AUC (C_max ratio = 1.09) in patients with severe HI and no increase in patients with mild or moderate HI. On the basis of these results, Viekira Pak is contraindicated in patients with severe HI and is not recommended in patients with moderate HI (FDA, 2014ai). The next largest change in exposure was observed for netupitant (AUC ratio = 4.18) in the population with severe HI. Of note, PK data were only available for two patients with severe HI, with individual AUC ratios of 6.06 and 2.29. The use of netupitant in patients with severe HI is to be avoided, with no dose recommendation in patients with moderate impairment (FDA, 2014a). Finally, tasimelteon showed AUC and C_max ratios of 2.67 and 1.34, respectively, in patients with moderate HI. Dose adjustment is not necessary in patients with mild or moderate HI; however, the product label states that “[tasimelteon] has not been studied in patients with severe hepatic impairment and is not recommended in these patients” (FDA, 2014l).

With regard to RI studies, six of the 22 NMEs evaluated in patients with RI demonstrated AUC ratios ≥2 in patients versus healthy controls, resulting in specific dosing recommendations; however, five NMEs (apremilast, dalbavancin, empagliflozin, olaparib, and pirfenidone) had AUC ratios <2 but still reported dosing recommendations. For two NMEs (eliglustat and netupitant), dedicated RI studies were not performed; however, dosing recommendations were provided (Table 11). Among the six NMEs for which systemic exposure was increased by ≥2-fold (albiglutide, ceftolozane, dapagliflozin, peramivir, naloxegol, and tazobactam), four are mainly eliminated via renal excretion, whereas naloxegol is mainly eliminated via biliary excretion, and albiglutide, a therapeutic protein, is eliminated via proteolysis. Peramivir showed the largest effect in patients with RI, with 4.15-, 5.27-, and 18.08-fold increases in AUC in patients with moderate, severe, and end-stage renal disease, respectively, with dose adjustment recommendations noted in the product label for these patients (FDA, 2014ac). Other changes in exposure ranged from a 2.00-fold change in tazobactam AUC when administered in patients with moderate HI to a 3.19-fold increase in AUC for naloxegol in patients with moderate RI, yielding specific labeling recommendations in all cases (Table 11). In addition, for dapagliflozin, a 2.31-fold increase in AUC was observed in patients with severe RI, whereas the AUC of the main circulating (inactive) metabolite, dapagliflozin 3-O-glucuronide, was found to increase by 3.22-fold and 1.99-fold in patients with severe and moderate RI, respectively, although the parent drug AUC was only increased by 1.60-fold in patients with moderate RI. Hence, dapagliflozin is contraindicated in patients with severe RI or ESRD, or in patients requiring dialysis (FDA, 2014j). Taken together, these data indicate the importance of assessing the PK of NMEs in impaired populations because AUC ratios reported in patients with HI or RI may be on the same order of magnitude as those observed in clinical drug interaction studies.