Contribution of CYP2D6 Functional Activity to Oxycodone Efficacy in Pain Management: Genetic Polymorphisms, Phenoconversion, and Tissue-Selective Metabolism

Oxycodone is a widely used opioid for the management of chronic pain. Analgesic effects observed following the administration of oxycodone are mediated mostly by agonistic effects on the μ-opioid receptor. Wide inter-subject variability observed in oxycodone efficacy could be explained by polymorphisms in the gene coding for the μ-opioid receptor (OPRM1). In humans, oxycodone is converted into several metabolites, particularly into oxymorphone, an active metabolite with potent μ-opioid receptor agonist activity. The CYP2D6 enzyme is principally responsible for the conversion of oxycodone to oxymorphone. The CYP2D6 gene is highly polymorphic with encoded protein activities, ranging from non-functioning to high-functioning enzymes. Several pharmacogenetic studies have shown the importance of CYP2D6-mediated conversion of oxycodone to oxymorphone for analgesic efficacy. Pharmacogenetic testing could optimize oxycodone therapy and help achieve adequate pain control, avoiding harmful side effects. However, the most recent Clinical Pharmacogenetics Implementation Consortium guidelines fell short of recommending pharmacogenomic testing for oxycodone treatment. In this review, we (1) analyze pharmacogenomic and drug-interaction studies to delineate the association between CYP2D6 activity and oxycodone efficacy, (2) review evidence from CYP3A4 drug-interaction studies to untangle the nature of oxycodone metabolism and its efficacy, (3) report on the current knowledge linking the efficacy of oxycodone to OPRM1 variants, and (4) discuss the potential role of CYP2D6 brain expression on the local formation of oxymorphone. In conclusion, we opine that pharmacogenetic testing, especially for CYP2D6 with considerations of phenoconversion due to concomitant drug administration, should be appraised to improve oxycodone efficacy.


Introduction
Drug overdose deaths involving prescription opioids (natural and semi-synthetic opioids and methadone) rose from 3442 in 1999 to 14,139 in 2019, reaching a peak of 17,209 deaths in 2017 [1]. Oxycodone is a semi-synthetic opioid used for pain management, accounting for approximately 17 million prescriptions in 2018 [2]. In 2016-2017, oxycodone prescriptions represented approximately 18% of all pain management prescriptions following an emergency room visit in the United States [3]. Recently, the role of genetic polymorphisms associated with proteins involved in the pharmacokinetics and pharmacodynamics of oxycodone was reviewed by the Clinical Pharmacogenetics Implementation Consortium (CPIC), which concluded that current evidence was insufficient to make firm clinical recommendations-in the form of a clinical guideline-regarding genetic testing for oxycodone dose adjustment [4]. Some study results have clearly demonstrated and supported the value of genetic testing, though others have not found clinical value in genetic testing to improve pain management with oxycodone. This is especially the case for demonstrated that CYP3A4 and CYP2D6 isoforms mediate most of oxycodone metabolism [34,36]. A small amount is converted to α/β oxycodol via 6-keto reduction and a small fraction is also directly converted to a glucuronide conjugate ( Figure 1) [37]. Lalovic et al., using human liver microsomes, showed that CYP3A4 was the highestaffinity enzyme involved in the N-demethylation of oxycodone to noroxycodone with a mean Km of 600 ± 119 μM [34]; co-incubation with the CYP3A4 inhibitor ketoconazole decreased formation of noroxycodone by more than 90%. During the O-demethylation of oxycodone to oxymorphone, CYP2D6 was found to have the highest affinity, with a Km of 130 ± 33 μM [34]. Other enzymes contributed to less than 8% for the N-demethylation and 10-26% for the O-demethylation pathways, respectively. N-Demethylation activity in intestinal microsomes was 20-50% that of the liver microsomes, while the O-demethylation was negligible, suggesting that the liver may be primarily responsible for first-pass metabolism, especially for the O-demethylation [34]. These results are in agreement with CYP3A4 and CYP2D6 protein levels measured in the human small intestine [20,38]. Romand et al. confirmed the important roles of CYP3A4 and CYP2D6 in the phase I transformation of oxycodone, while implicating UGT2B7 (Km: 762 ± 153 μM) and to a lesser extent UGT2B4 (Km: 2454 ± 497 μM) in the phase II metabolism of oxycodone [37].
Human mass-balance and pharmacokinetic studies confirmed that oxycodone is extensively metabolized in humans with only 10% of the drug excreted unchanged in urine [39][40][41]. In agreement with in vitro studies, oral administration of oxycodone showed that CYP3A4 is involved in the major metabolic clearance pathway with about 50% of the drug Lalovic et al., using human liver microsomes, showed that CYP3A4 was the highestaffinity enzyme involved in the N-demethylation of oxycodone to noroxycodone with a mean K m of 600 ± 119 µM [34]; co-incubation with the CYP3A4 inhibitor ketoconazole decreased formation of noroxycodone by more than 90%. During the O-demethylation of oxycodone to oxymorphone, CYP2D6 was found to have the highest affinity, with a K m of 130 ± 33 µM [34]. Other enzymes contributed to less than 8% for the N-demethylation and 10-26% for the O-demethylation pathways, respectively. N-Demethylation activity in intestinal microsomes was 20-50% that of the liver microsomes, while the O-demethylation was negligible, suggesting that the liver may be primarily responsible for first-pass metabolism, especially for the O-demethylation [34]. These results are in agreement with CYP3A4 and CYP2D6 protein levels measured in the human small intestine [20,38]. Romand et al. confirmed the important roles of CYP3A4 and CYP2D6 in the phase I transformation of oxycodone, while implicating UGT2B7 (K m : 762 ± 153 µM) and to a lesser extent UGT2B4 (K m : 2454 ± 497 µM) in the phase II metabolism of oxycodone [37].
Valuable information on the relative contributions of CYP2D6 and CYP3As to the disposition of oxycodone can also be obtained from pharmacogenomic and drug-drug interactions studies. Studies performed in individuals with multiple copies of CYP2D6-which translates to an UM phenotype-have shown that average plasma levels of oxycodone are reduced while average concentrations of oxymorphone are increased (relative to NMs). The oxycodone-to-oxymorphone plasma-concentration ratio was then estimated to be~32:1 in UMs following oral administration (Table 1) [46][47][48].
In patients and healthy volunteers with non-functional CYP2D6 (PMs), the oxycodoneto-oxymorphone ratio is increased to approximately 300:1 (Table 1) [48,49]. This increased ratio is mostly explained by a significant decrease in oxymorphone plasma concentrations, as the overall contribution of CYP2D6 to the total clearance of oxycodone is limited. Still, a nominal amount of oxymorphone is detected in the plasma of these individuals, suggesting that an isoform other than CYP2D6 could also mediate the formation of oxymorphone. Similar results were obtained during in vitro drug metabolism studies [34]. Notably, much lower ratios of oxycodone/oxymorphone were observed in patients from all phenotypic groups following the administration of oxycodone intravenously [50]. This observation can be explained by a route of administration-dependent difference in the relative contributions of CYP3A4 and CYP2D6 to the clearance of oxycodone, as CYP2D6 expression is limited in the intestinal tissue [21,[51][52][53]. Hence, the relative contribution of CYP2D6 to the total clearance of oxycodone appears increased following intravenous administration.
Drug interactions studies have been conducted with potent CYP2D6 inhibitors. Heiskanen et al. and Sirhan-Daneau et al. conducted studies in normal CYP2D6 metabolizers and demonstrated-using quinidine as a non-competitive and potent inhibitor of CYP2D6-that formation of oxymorphone was almost completely impeded under potent CYP2D6 inhibition [46,47,54,55]. However, using sensitive LC-MS/MS assays, limited amounts of oxymorphone could still be detected in their plasma [54]. Similar results were obtained by Lemberg et al. using paroxetine as a mechanism-based inhibitor of CYP2D6 [56]. Under potent inhibition of CYP2D6, the oxycodone-to-oxymorphone plasmaconcentration ratio is nearly~110:1 (Table 1). CYP2D6: cytochrome P450 2D6; UM: ultra-rapid metabolizer; NM: normal metabolizer; PM: poor metabolizer; CYP3A4: cytochrome P450 3A4; *: assuming a conservative brain-to-blood unbound concentration ratio of 3.0 for oxycodone and of 1.0 for oxymorphone; **: assuming a conservative brain-to-blood unbound concentration ratio of 3.0 for oxycodone and of 1.0 for oxymorphone and a potency ratio of 1:100 for oxycodone vs. oxymorphone.
As mentioned previously, CYP3A4 is the major enzyme involved in the disposition of oxycodone. Inhibition of CYP3A4, using inhibitory agents such as itraconazole or ketoconazole, is associated with increases in both oxycodone and also oxymorphone levels. As CYP3A4 is inhibited, more oxycodone is available for its metabolism by CYP2D6 into oxymorphone. Under conditions of CYP3A4 inhibition, the oxycodone to oxymorphone plasma concentration ratios ranged from 56:1 to 21:1 ( Average plasma protein binding for oxycodone is 45%, while that for oxymorphone is 11%. Hence, for all situations described above, the relative oxycodone/oxymorphone ratios while considering free-drug concentrations in the plasma [(C avg oxycodone × F u )/(C avg oxymorphone × F u )] should be decreased by about 40% (Table 1).
Drug distribution studies have demonstrated that both oxycodone and oxymorphone can cross the blood-brain barrier. First, Bostrom et al. reported in two different studies that the brain-to-blood unbound concentration ratio of oxycodone was 3.0-6.0 [60,61]. Then, Zasshi et al. also reported that the brain-to-blood unbound concentration ratio for oxycodone was 3.0 [62]. In regards to oxymorphone, Sadiq et al. reported that the brain-toblood unbound concentration ratio for oxymorphone was 1.9 [63]. In contrast, Zasshi et al. and Lalovic et al. reported that the oxymorphone brain-to-blood unbound concentration ratio was 0.26 to 0.3 [34,44,62]. However, the discrepancy between these results could not be explained. Assuming a conservative brain-to-blood unbound concentration ratio of 3.0 for oxycodone and an average of 1.0 for oxymorphone, new oxycodone-to-oxymorphone ratios can be estimated (Table 1); noroxycodone and noroxymorphone do not have good brain penetration [44].

Oxymorphone Pharmacokinetics Following Direct Oxymorphone Administration (PK; GRADE Moderate Quality +++−)
Oxymorphone is a known potent µ-opioid receptor agonist which may contribute, totally or partially, to the overall analgesic effects observed following oxycodone oral administration [64]. Oxymorphone can be administered directly as an active drug to humans and its pharmacokinetics and pharmacodynamics can be appreciated and compared to its pharmacokinetics/pharmacodynamic relationship observed following oxycodone administration.
Following the oral administration of a single extended-release 10 mg dose of oxymorphone, peak plasma concentrations of oxymorphone, 6-hydroxy-oxymorphone, and oxymorphone 3-glucuronide reached 0.65 ng/mL, 0.37 ng/mL, and 112 ng/mL, respectively [67]. Plasma levels of oxymorphone observed under these conditions were similar to plasma concentrations observed following administration of a single 15 mg immediaterelease oxycodone dose (0.7 ng/mL) [46,47,55]. This observation strongly suggests that oxymorphone plasma levels of 0.5 to 1.0 ng/mL mediate clinically relevant analgesic effects and likely contribute significantly to oxycodone efficacy.

Unlikely Formation of Active 6-Glucuronide and Other Glucuronide Metabolites (GRADE Moderate Quality +++−)
As described by Wang et al., there is one hydroxyl group on oxycodone and two hydroxyl groups on the oxymorphone structure, making them potential targets for glucuronidation [73]. The hydroxyl group attached to carbon 14 is an alcoholic group, and this moiety is relatively stable and protected within the ring structure, creating steric hindrance ( Figure 1). Therefore, glucuronidation is unlikely to occur for either oxycodone or oxymorphone [73]. This leaves only the hydroxyl group at position 3 of the oxymorphone structure to react with glucuronic acid-the major metabolite of oxymorphone-as a carbonyl instead of a hydroxyl group is present at position 6 of both oxycodone and oxymorphone ( Figure 1). The formation of 6-hydroxy glucuronide metabolites from either oxycodone or oxymorphone is very limited as it requires prior reduction of the 6-keto moiety [35,37,44]. Contrary to codeine and morphine, which produce 6-glucuronide metabolites with known µ-opioid receptor agonist activities, oxycodone and oxymorphone 6-glucuronide metabolites are not expected to be formed to a significant extent and to contribute clinically to oxycodone activity. Studies generally use a battery of pain tests to gauge drug efficacy. Some commonly used tests include the cold pressor test (cardiovascular response and pain perception), the electrical stimulation test (detects pain from Aδ and C sensory nerves), and thermal (thermode/thermal grill) and mechanical tests. Patients often use the visual analog scale to mark their pain levels throughout the tests. When comparing studies, it is essential to compare the tests used to gauge pain thresholds, as types of testing could introduce potential confounding factors [74].
Oxycodone and oxymorphone exert their pharmacological effects mainly as agonists to µ-opioid receptors. These receptors are encoded by the OPRM1 gene located on chromosome 6. Various polymorphisms in the OPRM1 gene have been identified, which may be important in analgesic response, pathologies of addiction, and adverse events related to opioid use [75][76][77]. In this section, we review studies that included oxycodone alone or along with other opioids to assess the impact of OPRM1 polymorphisms on drug actions.
The OPRM1 A118G (rs1799971) is one of the most widely studied variants in the context of opioid response. Gong et al. reported that higher opioid doses were required 24 h post-observation in cancer patients who were carriers of the G allele [78]. This polymorphism was also associated with progressively increasing 24-h opioid doses [78]. In an outpatient setting, a study by Lostch et al. described a tendency towards increased pain (decreased opioid response) with increasing number of OPMR1 G alleles [79]. Other studies report either a reduced analgesic effect of oxycodone in healthy subjects or the requirement of increased doses of oxycodone in patients carrying the G allele compared to individuals with an OPRM1 AA genotype [80,81]. Olesen et al. showed an association between oxycodone response and OPRM1 polymorphisms using four different pain tests [74]. They found that each test was associated with different single nucleotide polymorphisms (SNPs) in the OPRM1 gene. For example, the G variant was associated with oxycodone response in the visceral pain tolerance threshold. However, it was not associated with oxycodone response in either thermal skin pain tolerance, muscle pressure tolerance, or thermal visceral tolerance thresholds [74].
Two meta-analyses studied populations of 4607 and 5902 patients and found that carriers of the OPMR1 G allele-whether it was an AG or GG variant-required higher opioid doses and experienced higher pain scores [82,83]. In a prospective study in 153 opioid users, higher opioid consumption required for pain management was observed for carriers of the G allele [84]. Ren et al. also performed a meta-analysis and reported a reduction in nausea and vomiting in G allele carriers compared to AA carriers, with an odds ratio of 1.30 [83]. Other studies found no association between the A118G polymorphism and opioidinduced side effects [85]. In a study of 268 patients, it was reported that post-operative response to intravenous oxycodone was not associated with OPRM1 G variant [86]. Other variants in the OPRM1 gene have not been extensively linked to oxycodone response in the clinic.
The catechol-O-methyl transferase (COMT) is an enzyme involved in the metabolism of neurotransmitters and acts as a key modulator of dopaminergic and adrenergic/noradrenergic neurotransmission. The COMT gene is polymorphic and the most common c.472 G > A variant, in which the amino acid valine is substituted for a methionine, results in reduced COMT enzyme activity [87]. In human studies, the COMT genotype (Val/Met) has been shown, under different settings, to affect the efficacy of opioids in acute and chronic pain [88]. Low COMT activity has been associated with increased pain sensitivity, increased opioid analgesia, and increased opioid side effects, such as nausea and vomiting [85,87,89,90]. However, there is no consensus on the real clinical impact of this polymorphism on opioid response.

Pharmacokinetic, Pharmacogenomic, and Pharmacodynamic Considerations
As indicated by the CPIC's position, it is debated whether or not CYP2D6, OPMR1, and other genetic polymorphisms contribute to the analgesic effects of oxycodone. Unfortunately, there are a limited number of well-designed studies that consider these covariables simultaneously. The CPIC guidelines recommend pharmacogenetic testing for structurally related opioid congeners, such as codeine and tramadol, which are also prodrugs activated by CYP2D6 into their respective active metabolites (morphine and O-desmethyltramadol, both of which have free hydroxyl groups at position 3). These metabolites are also more potent µ-opioid receptor agonists than their respective parent prodrugs [4]. In the next sections, we report information from pharmacokinetic, pharmacodynamic, and pharmacogenetic studies that mostly lack pharmacogenomic information for the OPMR1 or COMT polymorphisms and their impact on the analgesic efficacy of either oxycodone or oxymorphone.

Pharmacogenetic Studies Relating Oxymorphone Formation to Pain Management by
Oxycodone (PK, PD, and PGx; GRADE High Quality +) Zwisler et al. conducted a double-blinded, randomized, placebo-controlled study of healthy subjects to assess oxycodone analgesic effects in PMs and NMs of CYP2D6 [91]. They demonstrated that PMs had an attenuated response to oxycodone compared to NMs when using both the cold pressor test and the electrical stimulation tests [91]. As expected, PMs also had a higher oxycodone/oxymorphone ratio, suggesting that CYP2D6-mediated formation of oxymorphone is important for anti-nociceptive effects [91]. Samer et al. tested the efficacy of oxycodone in patients with different CYP2D6 genotypes [64]. They used a series of pain tests to show that, after administration of oxycodone, UMs had a 1.4-fold higher tolerance threshold to the cold pressor test, whereas PMs had a 20-fold lower threshold when they were compared to NMs [48]. Similar results were seen with the electrical stimulation test, where respiratory depression and sedation caused by oxycodone were also found to be significantly higher in UMs compared to NMs. Unambiguously, all UMs exhibited mild to severe side effects whereas PMs did not report any [48]. In that study, the pain thresholds were significantly associated with oxymorphone and noroxymorphone concentrations, whereas there was no association between oxycodone and noroxycodone concentrations and effects observed [64]. These authors concluded that oxycodone does not appear to be the active molecule responsible for the analgesic effects [48].
Results from these studies clearly point towards an important role for CYP2D6mediated oxycodone transformation into oxymorphone in pain control. Other evidence also comes from case reports [92][93][94]. For instance, a patient with a fractured hip and a history of codeine intolerance was prescribed a regimen of oxycodone and acetaminophen [92]. As a proper level of analgesia was not achieved, she then received several doses of tramadol, as needed. Even with the added tramadol, the patient obtained no pain relief. Later, her genotype results confirmed that she was a CYP2D6 PM [92]. This suggests that, as for codeine and tramadol, oxycodone depends on CYP2D6 for its analgesic effects. Tyndale et al. also reported that CYP2D6 PMs had a lower probability of developing dependence on oxycodone compared to NMs; however, this was mostly explained by limited exposure to opioid agonists in this subset of the population [95].

Drug-Drug Interaction Studies
Modulating CYP2D6 Activity to Assess Oxymorphone Contribution to Oxycodone Efficacy (PK, PD, and PGx Plus DDIs; GRADE Moderate Quality +++−) Drug-drug interaction studies are another way to obtain valuable insights into the contribution of oxymorphone to oxycodone actions. CYP2D6 has a limited role (10%) in the total systemic clearance of oxycodone, and modulation of CYP2D6 activity is not expected to be associated with significant changes in oxycodone plasma concentrations. However, CYP2D6 is largely responsible for converting oxycodone to oxymorphone. Hence, drugdrug interaction studies using selective and potent CYP2D6 inhibitors could help assess the oxymorphone contribution to oxycodone pain control actions. Both quinidine (which does not seem to cross the blood-brain barrier) and paroxetine (which can enter the brain) have been used as CYP2D6 inhibitors in drug interaction studies.
As mentioned previously, Samer et al. conducted a study in 10 healthy subjects to assess the impact of quinidine pretreatment on oral oxycodone and assess multiple pharmacodynamic parameters, where phenotyping for CYP2D6 activity was conducted using dextromethorphan as a probe drug [64]. Quinidine pretreatment blunted oxycodonerelated pharmacodynamic effects on pain thresholds, such as the nociceptive flexion reflex (objective NFR), the subjective pain threshold (SPT) following an electrical stimulus, and sedation, making these effects similar to measurements observed under placebo pretreatment (no oxycodone). However, quinidine pretreatment did not completely abolish all pharmacodynamic effects, particularly in the pupil size and cold pressor test assessments. These effects may be related to the local metabolism in the brain of oxycodone into oxymorphone and the fact that quinidine does not readily penetrate the blood-brain barrier and might not be efficient at blocking CYP2D6 in the CNS (compared to its high potency in the liver) [55,64]. Pharmacodynamic results obtained by Samer et al. support the notion that oxymorphone production is required for oxycodone efficacy [64].
Heiskanen et al. tested the pharmacokinetics and pharmacodynamics of oxycodone administered with either placebo or quinidine pretreatments in ten healthy volunteers that were determined to be NMs using a phenotyping test (debrisoquine as a CYP2D6 probe drug) [55]. There was only a small reduction of the oxycodone AUC 0-24 following quinidine pretreatment compared to a placebo (258.5 ng.h/mL vs. 228.8 ng.h/mL). However, oxymorphone exposure decreased by 98% when quinidine was given three hours before oxycodone and six hours after to account for extended release of oxycodone, pointing to the importance of CYP2D6 activity in oxymorphone production [55]. Several drug effects were measured through questionnaires, accompanied by pupil size measurements and psychomotor tests (like the Maddox wing test). In the placebo pretreatment arm, several of the drug effects observed including alertness, pain control, and sleepiness, correlated with both oxycodone and oxymorphone plasma concentrations [55]. Prevention of the production of oxymorphone by quinidine (oxymorphone concentrations were undetectable in 8/10 patients) did not affect the psychomotor or subjective drug effects of oxycodone, suggesting that oxycodone, not oxymorphone, was responsible for drug effects. However, the authors noted that the interpretation of their results could lead to different conclusions if a potential metabolism of oxycodone into oxymorphone in the brain was to be considered [55].
Using another CYP2D6 inhibitor, paroxetine, Kummer et al. studied the effects of another CYP2D6 inhibitor-paroxetine, on oral oxycodone-in 12 healthy participants [96]. They found that the oxycodone C max and AUC 0-∞ were unchanged after paroxetine pretreatment, as expected, since CYP2D6 does not contribute significantly to oxycodone metabolism [96]. However, they reported that paroxetine prevented the analgesic activity of oxycodone when assessed using the cold pressor test, whereas pretreatment with ketoconazole (CYP3A4 inhibition, to be discussed later) accentuated oxycodone's analgesic effects (area under the effect curve and the values for the maximal effect Emax, p < 0.05 vs. paroxetine) [96]. Similar trends were observed on pupillometry effects. Their observation supports the notion that oxymorphone plays an important role in the analgesic effects following oxycodone administration. However, the authors suggested that, as a limitation to their study, selective serotonin inhibitor drugs such as paroxetine may interfere with the interpretation of pharmacodynamic measurements such as the cold pressure test and pupillary reaction [96].
Finally, Lemberg et al. used paroxetine coadministration in a cohort of 20 chronic pain patients to determine effects of CYP2D6 blockade on oxycodone pharmacokinetics and drug effects. Notably, this study included patients who were both naïve and non-naïve to opioids [56]. Patients were included in the study if they did not need more than two rescue doses of morphine in a day for at least three days. Patients were genotyped for CYP2D6, resulting in 14 NMs, 4 IMs, and 2 UMs; no PMs were reported. Paroxetine coadministration significantly increased the C max (26%) and AUC 0-12 (19%) of oxycodone, while significantly decreasing the C max (−57%) and AUC 0-12 (−67%) of oxymorphone [56]. Though statistically significant, changes in oxycodone pharmacokinetics were less prominent than the changes in oxymorphone pharmacokinetics, as expected. Drug effect was measured as the quantitative increase in morphine dose required to manage breakthrough pain on oxycodone treatment. In this study, no differences in the additional morphine doses required as rescue medication or in the visual analog scale of pain were observed between paroxetine pretreatment vs. a placebo [56]. Hence, it was concluded that paroxetine did not affect oxycodone-associated analgesia. Note that variability in the morphine dose needed at baseline and the use of standard morphine rescue doses could potentially mask reduction in anti-nociceptive activity during paroxetine cotreatment [56].
Additional studies using quinidine or paroxetine as potent CYP2D6 inhibitor showed that oxymorphone levels were significantly reduced by the coadministration of these drugs, but discrepancies existed regarding changes in pharmacodynamic parameters when oxycodone was compared to oxycodone plus CYP2D6 inhibitors [49,97]. Assessment of pain levels and response to medication remains a semi-quantitative and somewhat subjective measure that could be influenced, either positively or negatively, by several external experimental and individual factors. Nevertheless, results from these studies must be considered in the overall determination of the relative contribution of oxymorphone to oxycodone effects in pain management [49,97]. Figure 2 provides a summary of the relationship between oxycodone and oxymorphone pharmacokinetics, CYP2D6 genetic polymorphisms, and pharmacodynamics. Figure 2. Implications of pharmacokinetics (including CYP2D6 genetic polymorphism and CYP2D6 drug-drug interactions) and pharmacodynamics as determinants of interindividual difference in drug response to oxycodone and oxymorphone. CYP2D6: cytochrome P450 2D6; DDI: drug-drug interaction; BBB: blood-brain barrier.

Drug-Drug Interactions Studies Modulating CYP3A4 Activity to Assess Oxymorphone Contribution to Oxycodone Efficacy (PK-DDI and PD; GRADE Low Quality ++−−)
Another way to try to dissect oxymorphone from oxycodone actions is to modulate CYP3A4 activity by administrating inhibitors such as ketoconazole, itraconazole, or voriconazole or inducers like rifampin [45]. Using electrical stimulation and cold pressor tests, Samer et al. demonstrated that ketoconazole coadministration increased all pharmacodynamic effects of oxycodone, including sedation and the pain threshold [64]. They also performed a pharmacokinetic-pharmacodynamic multivariate analysis to identify pharmacokinetics predictors (including oxycodone, oxymorphone, noroxycodone, and noroxymorphone) associated with outcomes (area under the drug effect time curve: AUEC90). The only positive predictor of the subjective pain threshold and of oxygen saturation was oxymorphone. CYP3A4 blockade with ketoconazole was associated with an increased risk of adverse effects, most notably in CYP2D6 UMs [64].
Using itraconazole to inhibit CYP3A4, Saari et al. demonstrated that inhibition of CYP3A4 significantly increased the concentration of intravenous oxycodone by 51% and  Another way to try to dissect oxymorphone from oxycodone actions is to modulate CYP3A4 activity by administrating inhibitors such as ketoconazole, itraconazole, or voriconazole or inducers like rifampin [45]. Using electrical stimulation and cold pressor tests, Samer et al. demonstrated that ketoconazole coadministration increased all pharmacodynamic effects of oxycodone, including sedation and the pain threshold [64]. They also performed a pharmacokinetic-pharmacodynamic multivariate analysis to identify pharmacokinetics predictors (including oxycodone, oxymorphone, noroxycodone, and noroxymorphone) associated with outcomes (area under the drug effect time curve: AUEC 90 ). The only positive predictor of the subjective pain threshold and of oxygen saturation was oxymorphone. CYP3A4 blockade with ketoconazole was associated with an increased risk of adverse effects, most notably in CYP2D6 UMs [64].
Using itraconazole to inhibit CYP3A4, Saari et al. demonstrated that inhibition of CYP3A4 significantly increased the concentration of intravenous oxycodone by 51% and of oral oxycodone by 125% [58]. This finding was expected since CYP3A4 is expressed both in the intestine and in the liver such that itraconazole inhibition can occur in both tissues following oral administration of oxycodone. It was also shown that oxymorphone plasma concentrations increased even further (159% (intravenous) and 359% (oral), respectively) [58]. Itraconazole coadministration had a significant effect on behavior (alertness, deterioration of performance, etc.) following the oral administration of oxycodone. In contrast, no significant differences in the subjective drug effect or drowsiness were observed following the coadministration of itraconazole with intravenous oxycodone [58]. A relationship between oxymorphone plasma concentrations and pain control was not observed in their study. Interestingly, oxycodone, when administered alone, did not significantly increase the heat-pain threshold in all subjects; itraconazole coadministration also did not modify this outcome. This finding underlines that pain assessment, being subjective, is challenging in healthy volunteers since pain includes psychological, behavioral, and neurological aspects [58].
Voriconazole was also used to assess effects of CYP3A4 modulation on oxycodonemediated analgesia in healthy volunteers [57]. There was a significant increase in concentrations for both oxycodone and oxymorphone (257% and 597%, respectively) [57]. Voriconazole coadministration increased some pharmacodynamic effects, including heterophoria and myosis, when compared to placebo administration. However, voriconazole did not alter the heat-pain and cold-pain thresholds or the cold-pain intensity of oxycodone. Voriconazole coadministration did not significantly alter the perceived subjective drug effect (VAS) or other observed behaviors [57]. Similar to the study by Saari et al. using itraconazole [58], effects of the variable CYP2D6 genotype were not considered while interpreting the results of CYP3A4 inhibition using voriconazole (due to a small sample size) [57]. In the studies by Kummer et al. and Gronlund et al., both conducted among healthy volunteers, behavioral effects were recorded using questionnaires, and only the cold pressor test was conducted as a pain test [96,98]. Watanabe et al. reported a case series of patients with cancer in whom oxycodone-induced adverse events including drowsiness, vomiting, and hypopnea were reported within a few days after voriconazole treatment initiation [99]. The observed side effects were related to CYP3A4-mediated inhibition of oxycodone metabolism by voriconazole; however, the contribution of oxymorphone to the observed symptoms could not be excluded [99].
Gronlund et al. investigated the effects of the inhibition of CYP2D6 by paroxetine and the dual inhibition of CYP2D6 and CYP3A4 by paroxetine and itraconazole on the pharmacokinetics of and pharmacological response to intravenous oxycodone [100]. The administration of paroxetine did not significantly alter oxycodone exposure, but oxymorphone exposure was decreased by 60% [100]. Surprisingly, even though the combined administration of paroxetine and itraconazole increased oxycodone exposure by twofold, it did not cause any changes in pharmacological response. However, this study had some limitations, as CYP2D6 genotypes were not considered, a phase with CYP3A4 inhibition alone was not conducted (only a cold-pain test was used), and the power of the study design was not sufficient for detailed pharmacodynamic analyses [100].
Together these studies clearly indicate that blockade of CYP3A4 is associated with significant increases in oxycodone and oxymorphone concentrations. However, the respective contributions of oxycodone and oxymorphone to the pharmacological response, when observed, could not be distinguished. When CYP2D6 and CYP3A4 activities were combined, effects on analgesia and side effects were observed if CYP2D6 genetic polymorphisms were considered.

Phenoconversion as a Confounding Factor (GRADE Low Quality ++−−)
In a real-world scenario, genotype-guided phenotypes can be altered by various environmental conditions such as concomitant medications and disease states, a phenomenon called "phenoconversion" [101]. Consideration of phenoconversion is essential when trying to estimate the effects of CYP2D6 genotypes on drug efficacy. For instance, an individual with a CYP2D6 *1/*1 genotype predicted to be a NM could rather behave phenotypically as a PM of CYP2D6 for some substrates if this individual is administered concomitant treatment with other CYP2D6 substrates or inhibitors [102]. Similarly, a patient with type 2 diabetes may show diminished CYP3A4 activities and decreased clearance of CYP3A substrates due to downregulation of CYP3A4 expression secondary to chronic inflammatory conditions [101,103]. Phenotypic determination by the administration of probe drug substrates or by the use of endogenous biomarkers could allow for a more precise determination of functional enzyme levels than the genotype-predicted phenotype [104][105][106].
Zwisler et al. conducted a study in which oxycodone was administered intravenously to 270 patients for 24 h post-operatively, while morphine was prescribed as rescue medication [97]. The authors reported no association between oxycodone consumption and CYP2D6 genotypes, though surgery may represent a potential trigger for systemic inflammation that could induce a phenoconversion. Notably, no phenotyping was derived and results were based only on genotype data, introducing a potentially confounding factor in the data interpretation. In addition, all patients were also treated with acetaminophen (paracetamol; 1000 mg four times a day) and diclofenac (50 mg three times daily), which may have blunted the overall appreciation of analgesic response associated with oxycodone. Patients taking a CYP2D6 inhibitor were excluded from the study; however, CYP2D6-inhibiting drugs were limited to paroxetine, fluoxetine, and terbinafine [97]. We recently published a study in which we found that 15% of patients taking CYP2D6-metabolized opioids were exposed to significant CYP2D6 drug-drug interactions [107]. It can be expected that patients with CYP2D6-induced phenoconversion may be underestimated. Furthermore, the CYP2D6 alleles panel used in the study by Zwisler et al. was limited to four alleles, and neither CYP2D6 gene deletion (*5) nor gene duplication were tested [97]. Obviously, based on new, current knowledge, some patients were likely assigned to an inappropriate phenotype, as the use of a limited PGx panel can lead to a misclassification of the predicted phenotype based on the genotype.
The same research group assessed the analgesic effects of oxycodone in 33 healthy subjects using tramadol as a CYP2D6 probe substrate [91]. In addition to the phenotypic determination based on tramadol disposition, individuals were also genotyped for CYP2D6 using the same panel test with limited CYP2D6 allele determination [91]. They found, in five individuals, a mismatch between the determined genotype and observed phenotype based on tramadol pharmacokinetics. These five subjects were re-phenotyped using the gold-standard probe-drug sparteine. Phenotyping results with sparteine confirmed their PM phenotype as in agreement with the observed tramadol phenotype test but in disagreement with the determined genotype [91]. Individuals were healthy and were not allowed to take medications (except oral contraceptives), thus excluding disease-induced and drug-induced phenoconversion [91]. The discrepancy between the genotype-predicted phenotype and the observed phenotype can be explained by the limited pharmacogenetic panels tested, which created mis-assigned subjects in studied groups.
In the study conducted by Samer et al., individuals' phenotype determinations using dextromethorphan as a probe for CYP2D6 were undertaken to assign metabolizer status [64]. In this study, 32 CYP2D6 alleles were tested and subjects were healthy and were not taking concomitant medications. There was no mismatch between the predicted phenotype based on the CYP2D6 genotype and the observed phenotype using the metabolic ratio of dextromethorphan. This study showed an association between CYP2D6 activity and oxycodone efficacy [64].
Drug-induced phenoconversion can blunt the effects of pharmacogenetics. The impact of concomitant medications on phenotype interpretation was not considered in the analyses of some studies [93,108]. For instance, in the study by Andreassen et al., a proportion of NMs were taking CYP2D6 inhibitors, which likely resulted in their phenoconversion to PMs [49]. In the analysis of their results, the authors grouped everyone based on their genotype, not their phenotype, potentially skewing the results.

Liver vs. Brain CYP2D6 Activity (GRADE Low Quality ++−−)
CYP2D6 is expressed and functional in neurons of various regions of the brain [109]. The concept of local formation of oxymorphone from oxycodone, independent of liver CYP2D6 activity, could in part explain the apparent lack of association between oxymorphone plasma levels and observed clinical response following oxycodone administration. When conducting a drug-drug interaction study using quinidine as a potent CYP2D6 inhibitor, we demonstrated phenoconversion of CYP2D6 NMs into PMs based on the pattern of oxycodone, noroxycodone, and oxymorphone plasma levels [46,47]. Partial metabolic clearances clearly demonstrated significant inhibition of CYP2D6 and oxymorphone plasma concentration values similar to those observed in genetically determined CYP2D6 PMs. However, we could not demonstrate a significant increase in response to pain stimuli (thermal cathode, cold pressure) following quinidine administration [46,47].
Similar results were observed by Heiskanen et al. [55,64]. Inhibition of P-glycoprotein by a potent inhibitor, such a tariquidar, demonstrated that the distribution of quinidine to the brain is limited by this efflux transporter (although quinidine itself could competitively inhibit P-glycoprotein) [110]. Consequently, while low doses of quinidine are able to inhibit liver CYP2D6, brain concentrations of quinidine achieved under these conditions may leave brain CYP2D6 functional and allow for local formation of oxymorphone; plasma concentrations of oxymorphone may not reflect its extracellular concentrations in the brain.
Paroxetine has also been shown to inhibit oxycodone-mediated CYP2D6 metabolism and formation of oxymorphone [56,98]. As paroxetine distributes extensively to the brain, observation of the inhibition of CYP2D6 in the liver, in addition to the brain, would be expected [111]. Gronlund et al. investigated in 11 healthy subjects the effects of CYP2D6 inhibition by paroxetine on the pharmacokinetics and pharmacological response of oral oxycodone [98]. Paroxetine had no significant effect on the pharmacological response to oxycodone compared to the placebo phase. It should be noted that CYP2D6 genotypes were not considered even though PM and IM CYP2D6 genotypes were found in three subjects [98]. The co-administration of paroxetine is not expected to have any significant impact in CYP2D6 PM or, to a lesser extent, in IM. Considering this limitation, this study likely did not have the ability to evaluate the effects of paroxetine on oxycodone. In addition, the effects of paroxetine itself on pain perception could be a confounding factor, as suggested previously [96,98,112].
Conceptually, PMs of CYP2D6 can express CYP2D6 in neither their livers nor in their brains. Strong association between oxymorphone levels and oxycodone effects in subjects with various CYP2D6 genotypes supports the notion that oxycodone actions are largely mediated by oxymorphone. As other CYP isoforms appear to be able to form a certain amount of oxymorphone from oxycodone, this may also explain some discrepancies between study results.

Conclusions
Our systematic analysis of the current knowledge pertaining to the contribution of oxymorphone to oxycodone efficacy for pain management and the relevance of CYP2D6 and OPMR1 genetic polymorphisms lead us to the following deductions: (1) In vitro characterization of the CYP450 isoenzymes involved in the transformation of oxycodone into oxymorphone clearly demonstrated the predominant, but not exclusive, role of CYP2D6 [34,36]. Grade: high quality ++++; (2) Pharmacogenomic and drug-drug interaction studies conducted in humans also confirmed the role of CYP2D6 in the formation of oxymorphone. Grade: high quality ++++; (3) Plasma concentrations of oxymorphone observed following oral administration of oxycodone, although in the low nanogram level, are of the same magnitude as those measured when oxymorphone is directly administered, and they have been shown to be efficacious for pain management in humans. This suggests that oxymorphone mediates most of the analgesic efficacy of oxycodone. Grade: high quality ++++; (4) Binding studies at the µ-opioid receptor level have clearly established the much greater agonist affinity of oxymorphone vs. oxycodone [44]. Grade: high quality ++++; (5) The estimated brain concentration ratio of oxycodone vs. oxymorphone, considering protein binding and brain distribution (tissue level) and weighted for the relative potency of both compounds as agonists on the µ-opioid receptor, suggests that oxycodone and oxymorphone contribute equally to analgesic effects. Grade: low quality ++−−; (6) As the µ-opioid receptor is on the extracellular side of the neuron membrane in the synapse, the extracellular concentrations of oxycodone and oxymorphone could be more relevant than brain tissue levels. Unfortunately, these data are not available at the moment. Grade: very low quality +−−−; (7) Brain CYP2D6 activity, capable of converting oxycodone into oxymorphone, could explain the lack of association between oxymorphone plasma levels and efficacy observed in some drug-drug interactions and clinical studies. Grade: low quality ++−−; (8) The lack of oxycodone activity in genetically characterized CYP2D6 PMs, who functionally lack CYP2D6 activity not only in their liver but also in their brain, strongly supports the concept that oxycodone activity is mediated by oxymorphone. Grade: high quality ++++; (9) Clinically, the coadministration of other CYP2D6 substrates and inhibitors, and their capacity to blunt oxycodone efficacy, could be related to both their capacity to inhibit liver CYP2D6 activity and also to their capacity to cross the blood-brain barrier, penetrate CNS, and reach neurons and inhibit brain CYP2D6 activity, thus limiting local oxymorphone formation from oxycodone. Grade: low quality ++−−; (10) Finally, mutations on OPMR1 leading to less functional µ-opioid receptors and their role in the inter-individual response to active opioid products is too often overlooked.
We support the current CPIC position as it offers stringent recommendations to establish guidelines regarding pharmacogenetic testing for oxycodone treatment. Nevertheless, clinical considerations and scientific information lead us to recommend that a phenotypic assessment of CYP2D6 activity should be considered when oxycodone is being prescribed for pain management. This position is supported, at least in part, as PGx testing is recommended for other opioid congeners (codeine, tramadol) that share several similar pharmacokinetic, pharmacodynamic, and pharmacogenomic properties. Phenotypic assessment should include the use of advanced clinical decision support systems to conduct comprehensive medication reviews and identify potential multi-drug interactions leading to CYP2D6 phenoconversion. Pharmacogenomic testing should also be considered, as it could provide a definitive response for CYP2D6 PMs, i.e., about 10% of patients with polypharmacy. Furthermore, more than 100 variant alleles have been described for CYP2D6, and the precision of the genotype results is directly related to the quality of the genotype test being performed. As directly active opioids are a clinically sound alternative to CYP2D6-activated opioids (oxycodone, codeine, hydrocodone, and tramadol), use of these agents instead of the prodrugs appears highly justifiable in patients with polypharmacy or with suspected non-functional or decreased CYP2D6 activity.
A summary of the studies included in this review study is presented in Table 2.