Forming a ∼390 kb cluster at chromosome 10q24, the CYP2C subfamily contains four genes: CYP2C8, CYP2C9, CYP2C18, and CYP2C19 (Goldstein and de Morais, 1994; Nelson et al., 2004). All exhibit extensive homology in both DNA and amino acid sequence, and are thus responsible for the metabolism of partially overlapping subsets of drugs (Coller et al., 2002; Koukouritaki et al., 2004; Rettie and Jones, 2005; Lai et al., 2009; Naraharisetti et al., 2010; Ohtsuki et al., 2012). Their main expression is in the liver, comprising 20% of hepatic CYP content (Shimada et al., 1994). Within the CYP2C subfamily, CYP2C9 is the most abundantly expressed, followed by CYP2C8 and CYP2C19. Psychotropic substrates for CYP2C19 include diazepam (Jung et al., 1997), phenytoin (Bajpai et al., 1996; Mamiya et al., 1998), propranolol (Otton et al., 1990), selective serotonin reuptake inhibitors (SSRIs; Hicks et al., 2015), and tricyclics (Hicks et al., 2017). CYP2C18 is distal to CYP2C19 on chromosome 10 but appears to be expressed only at the mRNA level and not at the protein level (Chen and Goldstein, 2009).

CYP2C8 is the second most important cytochrome after CYP3A4 for the conversion of buprenorphine to its active metabolite, norbuprenorphine (Picard et al., 2005), and its expression is under genetic control. Work in Asian populations has identified variants that are associated with no functional enzyme, specifically the CYP2C8^∗5.001, the CYP2C8^∗7.001, and the CYP2C8^∗11.001 at 0.006, 0.0025, and 0.003 (in Koreans; 0.01 in Vietnamese and 0.0014 in Chinese) frequency, respectively in E. Asian populations tested (Soyama et al., 2002; Hichiya et al., 2005; Yeo et al., 2011).

CYP2C9 metabolizes phenytoin. It is also relevant to drugs prescribed to treat physical comorbidities in those with chronic mental health conditions. These include anti-diabetic agents (such as tolbutamide, glimepiride, and nateglinide), angiotensin II blockers (losartan, valsartan, candesartan, and irbesartan), fluvastatin, warfarin, and nonsteroidal anti-inflammatory drugs including COX2 inhibitors (e.g., celecoxib) (Michaels and Wang, 2014). Of the reduced function variants, CYP2C9^∗2 and CYP2C9^∗3 are the most common, and have been studied in relation to the metabolism of drugs with a narrow therapeutic index, such as phenytoin, tolbutamide, and warfarin (Lee et al., 2002). The functional mutations in CYP2C9^∗2 and CYP2C9^∗3 are rs1799853 and rs1057910, leading to Arg144Cys and Ile259Leu substitutions. CYP2C9^∗2 is associated with a 10-fold lower V_max and 2-fold lower V_m for (S)-warfarin hydroxylation. Median daily warfarin dose was in one study 4.0, 2.9, 2.6, and 1 mg for individuals of CYP2C9^∗1/^∗1, CYP2C9^∗1/^∗2, CYP2C9^∗1/^∗3, and CYP2C9 homozygous mutant genotype, respectively (King et al., 2004). Individuals who are affected by two reduced function alleles have a greater chance of ADRs such as gastrointestinal bleeding from NSAIDs (Martinez et al., 2004), hypoglycemia (Holstein et al., 2005), and bleeding from warfarin (Ogg et al., 1999). In a GWAS of response to warfarin, a CYP2C9 marker was separately genotyped in addition to the array-based genomic analysis and was identified as the top signal (Takeuchi et al., 2009). Predictive modeling followed, and included a target of the drug (VKORC1), as well as CYP2C9 (Eriksson and Wadelius, 2012; Maagdenberg et al., 2018); the FDA label summarizes findings of a meta-analysis in which patients carrying at least one copy of the CYP2C9^∗2 or CYP2C9^∗3 alleles required a mean daily warfarin dose 17 or 37%, respectively less than wild-type individuals (Sanderson et al., 2005).

Like many CYPs (Kalow and Tyndale, 1992), CYP2C19 is also expressed extrahepatically in multiple tissues including in the brain (Aitchison et al., 2010). Substrates of this enzyme include: diazepam and its metabolite desmethyldiazepam, moclobemide (Roh et al., 1996), SSRIs (fluoxetine, sertraline, paroxetine, citalopram, escitalopram), tertiary amine tricyclics (e.g., amitriptyline, imipramine, and clomipramine), as well as clozapine, olanzapine, phenytoin, and propranolol to lesser extents. The SSRIs fluoxetine and fluvoxamine also inhibit CYP2C19 (Preskorn, 1997). Conversely, phenothiazines represented by perazine and promazine have been shown to induce this CYP enzyme (Wójcikowski et al., 2012). Other substrates include the anticoagulant clopidogrel, cyclophosphamide, nelfinavir, proguanil, proton pump inhibitors [omeprazole (Karam et al., 1996) and pantoprazole], thalidomide, and voriconazole (Desta et al., 2002). Of the 35 allelic variants described in the CYP Database, CYP2C19^∗2-^∗8 are the most common loss of function (poor metabolizer or PM) haplotypes.

There is substantial interethnic variation in the incidence of PMs of CYP2C19, being 2–5% in Whites, 2% in Saudi Arabians, 4% in Black Zimbabweans, 5% in Ethiopians, 13% in Koreans, 15–17% in Chinese, 21% in Indians, and 18–23% in Japanese (Evans et al., 1995; Aitchison et al., 2000c). When the square root of the PM phenotypic frequency (equal to the frequency of PM CYP2C19 alleles) is plotted versus longitude, an increase in this value versus longitude may be seen, with an increment in the value occurring between Saudi Arabia and Bombay (Evans et al., 1995; Saeed and Mayet, 2013). The increasing frequency of PMs is mainly owing to the higher frequencies of the null haplotypes CYP2C19^∗2 and CYP2C19^∗3. The most common gain-of-function haplotype is the c.−806C>T (rs12248560) defining the CYP2C19^∗17 haplotype. Of note, however, this may be found in combination with loss-of-function variants such as the c.1A>G (rs28399504) associated with the CYP2C19^∗4 haplotype, or another loss of function variant (c.463G>T) (Scott et al., 2012, 2013; Skierka and Black, 2014). It is therefore necessary to accurately characterize haplotypes with the c.−806C>T. Tables available via PharmGKB² provides further details on CYP2C19 haplotype frequencies by ethnic group.

The most common PM haplotype is CYP2C19^∗2, which accounts for about 86% of all the PMs in the White population and 69–87% in the E. Asian population. The substitution of G681A in exon 5 of the CYP2C19^∗2 haplotype creates an aberrant splice site (de Morais et al., 1994). The second most common PM haplotype is CYP2C19^∗3, which represents about 13–31% of E. Asian PMs and 1.5% of White PMs. The substitution of G636A mutation in exon 4 of the CYP2C19^∗3 creates a premature stop codon. A third variant, CYP2C19^∗4 accounts for approximately 3% of White PM alleles and contains an A → G mutation in the initiation codon (i.e., c. 1A>G). CYP2C19^∗5 accounts for 1.5% of White PM alleles and is rare in E. Asians. The CYP2C19^∗5 haplotype is a result of a c.C1297T mutation in exon 9, in which causes an Arg433Trp change in the heme-binding region. CYP2C19^∗6 (a c. G395A base substitution resulting in an Arg132Gln coding change in exon 3) and CYP2C19^∗7 (a GT → GA mutation in the donor splice site of intron 5 at c.819 +2) each account for a further 1.5% of White PM alleles. CYP2C19^∗8, a T358C substitution in exon 3 that result in a Trp120Arg change, is a less common PM allele. The products of CYP2C19^∗6 and CYP2C19^∗8 show reduced catalytic activity (2% and 9% of wild-type S-mephenytoin hydroxylase activity, respectively); the others described above are associated with failure to express active CYP2C19. CYP2C19^∗2A and CYP2C19^∗3 have both been identified in an Ethiopian population and found to account for all the PM alleles in the 114 individuals studied (Persson et al., 1996).

In CYP2C19 PMs, diazepam clearance is significantly lower than in NMs (Bertilsson, 1995). The mean clearance is lower in Chinese compared to Whites. Owing to the relatively high frequency of PMs in E. Asians, there is a greater frequency of individuals carrying one PM haplotype (i.e., heterozygous PMs). Consistent with this, “many Hong Kong physicians routinely prescribe smaller diazepam doses for Chinese than for white Whites” (Kumana et al., 1987). The main variant responsible for this effect is the G681A, which has a gene-dosage association effect on diazepam clearance (Qin et al., 1999). For recent data on antidepressants and CYP2C19, see the relevant section. Dose adjustment by CYP2C19 genotype has been published for amitriptyline, citalopram, clomipramine, imipramine, moclobemide, and trimipramine (Kirchheiner et al., 2001). CYP2C19 PMs show a significantly higher efficacy for triple therapy for Helicobacter pylori (proton pump inhibitor, clarithromycin, and amoxicillin) (Klotz, 2006, 2009).

The CYP2D subfamily consists of a gene cluster comprising CYP2D6, with two pseudogenes, CYP2D7 and CYP2D8 (Yasukochi and Satta, 2011). CYP2D6 accounts for 1.5% of microsomal CYP content in the liver (Michaels and Wang, 2014), and is involved in metabolizing the majority of psychotropic drugs (Bertilsson et al., 2002). It is also expressed in other organs, including the brain (Niznik et al., 1990; Kalow and Tyndale, 1992; Siegle et al., 2001; Aitchison et al., 2010), and has been associated with synthesis of neurotransmitters (Yu et al., 2003; Niwa et al., 2017). The CYP2D6 gene is highly polymorphic, even compared to some of the other CYPs (Pharmacogene Variation Consortium: CYP2D6). It is the most extensively studied genetically variable drug metabolizing enzyme (Bertilsson et al., 2002; Ingelman-Sundberg, 2004b), and has over 110 unique alleles identified (Kertesz et al., 2007).

These studies have revealed that there is significant variation of allelic variants between ethnic groups (Aitchison et al., 2000c)³. For example, the CYP2D6^∗4 haplotype (previously known as g.1846G>A, genomic location of NG_008376.3 (Reference SNP (refSNP) Cluster Report: rs3892097) 1847G>A, Gough et al., 1990; Hanioka et al., 1990; Kagimoto et al., 1990) has a frequency of 19% in Whites (approximately 70–90% of all the PM alleles) (Aitchison et al., 1999), and 6% in Africans and 1% in South Asians (Aitchison et al., 2000c; Mammen et al., 2018). The second most frequent PM haplotype in Whites (2–2.5%) is CYP2D6^∗5 (Aitchison et al., 1999), which represents a complete gene deletion, and occurs at a frequency of 5.3, 2.9, and 2.9% in Africans, Asians, and Hispanics, respectively (Bradford, 2002; Del Tredici et al., 2018). The CYP2D6^∗10 haplotype has key C188T and G4268C base substitutions in exons 1 and 9, respectively, that result in Pro34Ser and Ser486Thr amino acid substitutions (Yokota et al., 1993; Sakuyama et al., 2008). This haplotype is associated with reduced enzymatic activity (Johansson et al., 1994). With an allelic frequency of 0.43, it is very high in East Asians (Aitchison et al., 2000c; Mammen et al., 2018), and similar to other reduced activity metabolizers has been associated with ADRs such as tardive dyskinesia (Ohmori et al., 1998; Puangpetch et al., 2016). However, some of these apparent CYP2D6^∗10 alleles may in fact be CYP2D6^∗36 hybrid alleles. The CYP2D6^∗17 haplotype exhibits a similar reduction in enzymatic activity (Masimirembwa et al., 1996; Oscarson et al., 1997), and is found predominantly in Africans, with frequencies of 34% in Zimbabwe, 28% in Ghana, 17% in Tanzania, and 9% in Ethiopia (Bertilsson et al., 2002). CYP2D6^∗41 is the most common reduced activity (IM) haplotype in Whites, a key SNP 2989G>A (genomic position on NG_008376.3 7189G>A) occupying an intronic position leading to a splicing defect (Raimundo et al., 2000, 2004; Rau et al., 2006; Toscano et al., 2006; Hicks et al., 2013, 2016; Wang et al., 2014). Tables available via PharmGKB⁴ provides details of CYP2D6 haplotype frequencies by ethnic group.

At the opposite end of the activity spectrum are the UM allelic variants, which most commonly have extra functional copies of the CYP2D6 gene in tandem on the chromosome, seen at a frequency of 0.9–4% in Whites (Johansson et al., 1993; Aitchison et al., 1999). An apparently less common mechanism for UM alleles is upregulation of gene expression owing to SNP-related enhancer activity (Wang et al., 2015). Individuals possessing UM alleles were first identified as having lower than expected blood concentration of tricyclic antidepressants such as clomipramine (Bertilsson et al., 1993a, b; Dalen et al., 1998; Roberts et al., 2004). CYP2D6 gene duplication or multiplication events occur at rates up to 29% in Ethiopians (Aklillu et al., 1996) by old techniques such as restriction fragment length polymorphism, and remain to be accurately characterized in terms of frequency using more current approaches.

The diversity in CYP2D6 phenotype has clinical implications (Ingelman-Sundberg, 2005). Individuals with two PM haplotypes have no functional enzyme, are classified as PMs, and are more prone to ADRs for drugs with a narrow therapeutic window (Steimer et al., 2004, 2005). At the other end of the spectrum, UMs may also show more ADRs, such as tardive dyskinesia (Koola et al., 2014) or symptoms of morphine overdose on codeine (Crews et al., 2014), owing to enhanced formation of toxic metabolites (Pinto and Dolan, 2011). Variation in CYP2D6 is highly relevant to psychiatry: for most antidepressants and antipsychotics, there are clinical guidelines that state that pharmacogenomic information for CYP2D6 could or should be used in prescribing (Bousman et al., 2019a). A review with modeling found that for antidepressants metabolized by CYP2D6, normal metabolizers (NMs) would require at least double the dose required by PMs, while cost analyses have associated PM status with not only higher ADRs but also with more drop outs from treatment (Chou et al., 2003; Kirchheiner et al., 2004; Zhou et al., 2004; Tanner et al., 2020).

Haplotype functionality may be used to derive an activity score (Caudle et al., 2020), with resources provided by PharmGKB to assist with this process.³ In the most recent update, the activity score of the CYP2D6^∗10 haplotype was adjusted from 0.5 to 0.25, and the phenotype assignment for an activity score of 1 adjusted from NM to IM.

A relatively small number of allelic variants have been identified for CYP2E1, such as CYP2E1^∗2, which is associated with reduced enzyme activity (Hu et al., 1997; Mittal et al., 2015). This enzyme is produced primarily in the liver, although it is also found in the brain (García-Suástegui et al., 2017), and is responsible for metabolizing ethanol (into acetaldehyde), paracetamol/acetaminophen, and other substances into reactive intermediates, whose toxicity is enhanced in alcoholics (Cederbaum, 2012). Indeed, CYP2E1 is responsible for 20% of total ethanol metabolism (to which other enzymes such as catalases also contribute) (Heit et al., 2013). Gene transcription is induced by ethanol consumption (a moderate level of intake at 140 g ethanol per week producing an increase in expression of CYP2E1 in the intestine, but not in the liver; Liangpunsakul et al., 2005). Interestingly, a 96-bp insertion polymorphism in the CYP2E1 gene, which is associated with higher activity of the encoded enzyme, has been proposed as a possible protective factor against alcoholism (Cartmell et al., 2005). In addition to its relevance to alcohol use disorders, the role of CYP2E1 in metabolizing ethanol is a potential alcohol–drug interaction site. With occasional alcohol usage, medications such as clozapine at least partly metabolized by CYP2E1 may have their half-life increased owing to competitive inhibition with alcohol. With chronic alcohol use, the induction effect predominates, thus reducing the efficacy of CYP2E1-dependent drugs by decreasing half-life (Cederbaum, 2012).

In mice, tobacco smoke induces CYP2E1 activity in the lungs, liver and kidney (Zevin and Benowitz, 1999). In male smokers, CYPE1 clearance may be increased (Benowitz et al., 1999). There may additionally be a complicated interaction effect of smoking and alcohol at CYP2E1, whereby CYP2E1 activity (as measured by chlorzoxazone metabolism) appears to be enhanced in non-alcoholic female smokers (Girre et al., 1994), while in males (Howard et al., 2003) CYP2E1^∗1D has been associated with nicotine and alcohol co-dependence in one study (Howard et al., 2003), which was not replicated in Taiwanese (Huang et al., 2018).

Perphenazine undergoes substantial first-pass hepatic phase I and II metabolism. Serum concentrations vary widely due to polymorphisms in multiple phase I enzymes: up to 30-fold in CYP2D6 NMs (Linnet and Wiborg, 1996). Initial studies showed that after 4–5 weeks, improvement was associated with plasma perphenazine concentrations above 2 nmol/l, while extrapyramidal effects occurred at concentrations above 3 nmol/l (Hansen, 1981; Hansen et al., 1982; Hansen and Larsen, 1983). In a larger study of over 200 patients, a wider therapeutic range (2–6 nmol/l) was suggested (Hansen and Larsen, 1985). Perphenazine binds dopamine D₂ and alpha-₁/alpha-₂ receptors with 70 and 50% antagonism. The main active metabolite, 7-hydroxyperphenazine, binds dopamine D2 and alpha-1/alpha-2 receptors with 70 and 50% the antagonism of perphenazine (Hals et al., 1986). It is formed in a reaction catalyzed by CYP2D6, with other metabolites including N-dealkylated perphenazine (formed in part by other CYPs), and perphenazine sulfoxide (Dahl-Puustinen et al., 1989; Olesen and Linnet, 2000). Compared to perphenazine, the concentration of perphenazine sulfoxide is in the same range, while N-dealkylated perphenazine is approximately three times that of perphenazine (Hansen et al., 1979). At therapeutically relevant concentrations of perphenazine, CYP3A4 accounts for about 40% of the N-dealkylation, with CYP isoforms 1A2, 2C19 and 2D6 contributing 20–25% (Olesen and Linnet, 2000).

The peak serum concentration and the AUC of perphenazine for CYP2D6 PMs is about 3 and 4 times, respectively that of NMs in single dose kinetics (Dahl-Puustinen et al., 1989), and at steady state, the median concentration-to-dose ratio of perphenazine in CYP2D6 PMs is about twice that of NMs, with patients on concomitant inhibitors showing a median concentration in between the two groups (Linnet and Wiborg, 1996). Jerling et al. (1996) conducted a study of patients during treatment; CYP2D6 genotype was shown to significantly predict the oral clearance of perphenazine (patients with two CYP2D6 PM alleles having lower clearance than heterozygote PMs or NMs) (Jerling et al., 1996).

It would be expected that individuals deficient in CYP2D6 or on potent CYP2D6 inhibitors, higher perphenazine concentrations would be found and hence more adverse effects, whilst in CYP2D6 UMs, there would be lower concentrations, with less adverse effects and potentially a lower therapeutic efficacy. Consistent with this, paroxetine, a potent CYP2D6 inhibitor (Lam et al., 2002), increases the AUC of perphenazine 7-fold in NMs, which is associated with increased side effects (Ozdemir et al., 1997).

Since 1984 pimozide has been used to treat Gilles de la Tourette’s syndrome (Pringsheim and Marras, 2009), and also to treat psychotic disorders. Its use has been limited owing to an ADR of prolongation of the QT interval on the electrocardiogram, which is associated with risk for Torsades de Pointes (a type of ventricular fibrillation that may cause sudden cardiac death) (Fulop et al., 1987; Committee on Safety of Medicines-Medicines Control Agency, 1995). In an isolated rabbit heart, this effect was shown to be attributable to pimozide itself, not to metabolites (Flockhart et al., 2000); this is due to an effect of the drug on potassium channels encoded by the human ether-a-go-go-related gene (HERG, otherwise known as KCNH2), which is responsible for the delayed repolarization current in the heart.

It is important to determine which cytochromes might contribute to the pimozide concentration profile. In vitro analyses showed that the formation of the major metabolite, 1,3-dihydro-1-(4-piperidinyl)-2H-benzimidazol-2-one (DHPBI), by N-dealkylation was primarily dependent on CYP3A4, with a lesser contribution by CYP1A2 (Desta et al., 1998). CYP2D6 may also play a role, but due to it being inhibited by pimozide, it was not possible to draw a conclusion regarding this from this in vitro study.

Case reports of interactions between pimozide and CYP2D6 inhibitors such as paroxetine and fluoxetine (Ahmed et al., 1993; Horrigan and Barnhill, 1994), as well as investigation of differential interaction with clarithromycin (an inhibitor of CYP3A) by CYP2D6 status led to recognition that CYP2D6 was a major contributor to the in vivo pharmacokinetics of pimozide (Desta et al., 1998). The effect of a single dose (6 mg) on the QTc interval (QT interval corrected for heart rate) was measured over time, and showed the greatest increase within the first 20 hours, with NMs showing a larger increase (by nearly 20 ms), followed by a reduction from 20 to 50 h, and then an increase at approximately 60–100 h. The late elevation was more significant in CYP2D6 PMs, women, and clarithromycin-treated individuals, and appeared more sustained than the early increase. Owing to the more sustained nature, the late onset elevation may be more relevant to significant QTc prolongation; the early peak in NMs warrants further investigation in UMs. In CYP2D6 PMs, half-life increased from 29 ± 18 h to 36 ± 19 h, while in NMs, the corresponding values were 17 ± 7 and 23 ± 10 h. For subjects with relevant data, the pimozide induced QTc interval changes coincided with the concentration-time course of pimozide. The prescription of CYP3A inhibitors, such as valproate, is now contraindicated with pimozide. In the above study, interestingly, pimozide rapidly increased plasma prolactin concentration, the maximum increase occurring 4 hours post dose, with a sharp reduction thereafter.

Simulated steady-state pharmacokinetic profiling of pimozide in CYP2D6 PMs, IMs, and NMs led to specification in the FDA label in 2011 that CYP2D6 PMs should not be prescribed more than 4 mg, with the maximum recommended dose in CYP2D6 NMs being 10 mg (Rogers et al., 2012). In the simulated data, 4 mg/day in CYP2D6 PMs was the maximum dose that did not result in plasma concentrations in excess of those observed in CYP2D6 NMs receiving 10 mg/day (Desta et al., 1998). Pimozide is commenced at 0.05 mg/kg (Preskorn, 2012), once daily. If the patient is a CYP2D6 NM and is not on a CYP2D6 inhibitor, the dose may be increased every third day to a maximum of 0.2 mg/kg/day, to a maximum of 10 mg/day. If the CYP2D6 status is not known, CYP2D6 genotyping should be done before deciding to increase the dose above 0.05 mg/kg/d, which is the maximum dose for a CYP2D6 PM, or if on a CYP2D6 inhibitor such as paroxetine, fluoxetine, and bupropion. Paroxetine will convert 60% of CYP2D6 NMs to PMs at 20 mg daily, while at 40 mg daily, 95% will be phenocopied to PMs (Preskorn, 2003). Phenoconversion (the conversion of an individual’s genetically defined metabolizer status to another status owing to the effect of a pharmacologically active substance) to CYP2D6 PM status by the action of an enzyme inhibitor has been estimated as being 6 times more common than genetically determined CYP2D6 PM status (Preskorn, 2012, 2013).

First pass metabolism of pimozide includes both the gut and the liver as CYP3A represents 70% and 30% of the total CYP450 in the intestine and the liver, respectively (Kolars et al., 1994; Shimada et al., 1994). Metabolism will be subject to the influence of gut microbiota, diet, and other factors including hormones (CYP3A4 being subject to regulation by the PXR and CAR) (Lamba et al., 2005; Pan et al., 2009).

The drug label does not currently include dosing recommendations for CYP2D6 UMs; further research including genotyping CYP2D6 is required for pimozide, and other CYP2D6 metabolized medications.

It is suggested that CYP3A4 also be genotyped for pimozide treatment, given its association with sudden cardiac death. It has a less clear genotype–phenotype relationship (with no updated data on PharmVar), and thus has not yet been introduced into clinical guidelines. In the absence of genotyping, probe drugs such as nifedipine may be utilized to test the activity of multiple CYPs (de Andrés et al., 2014); however, such estimation of CYP3A4 phenotype is influenced by any concomitant medication and/or dietary effects.

Haloperidol (HAL) is a butyrophenone and first-generation antipsychotic (FGA) drug that acts as a dopaminergic antagonist in the mesolimbic system. It is used to treat a variety of psychiatric conditions, including psychoses (e.g., schizophrenia, schizoaffective disorder, bipolar disorder with mania or psychotic symptoms, substance-induced psychotic disorder) and other conditions with hallucinations (e.g., alcohol withdrawal, delirium, Lewy body dementia). Adverse effects may include tardive dyskinesia, neuroleptic malignant syndrome, and a prolonged QT_c interval. Two major routes of metabolism, N-glucuronidation and O-glucuronidation, are effected by UGT enzymes, specifically the former by UGT1A4, and the latter by UGT1A4, UGT1A9, and UGT2B7 (Kato et al., 2012). Various CYP isoenzymes contribute to the metabolic pathways of this medication, most notably CYP3A4, and, to a lesser extent, CYP2D6. Cytosolic carbonyl reductase catalyzes the formation of reduced HAL, which retains 10–20% of the activity of the parent compound. Reduced HAL can be further metabolized by CYP3A4 to a tetrahydropyridine. The reduced drug can also be back-oxidized by CYP3A4 to HAL (Pan et al., 1998; Kudo and Ishizaki, 1999; Tateishi et al., 2000; discussed in Aitchison et al., 1999). Owing to its lipophilicity, HAL is extensively metabolized in humans, with large interindividual variations in pharmacokinetics arising. With a proposed therapeutic range of 5.6–16.9 μg/l in serum (Ulrich et al., 1998), being able to appreciably predict pharmacokinetic parameters in individuals is of utmost importance to optimize efficacy and safety. At lower doses, CYP2D6 contributes to HAL metabolism significantly, but with higher doses, and longer term treatments, CYP3A4 back-oxidation and N-dealkylation considerably outweigh the contributions of CYP2D6 (Fang et al., 1997; Pan et al., 1998; Zhou et al., 2009). Nyberg et al. (1995) showed that CYP2D6 PMs exhibited higher plasma concentrations of HAL over a 4-week treatment period with HAL decanoate, as compared to seven NMs in the study. However, Ohnuma et al. (2003) showed that, in a large number of Japanese patients, the presence of neither an enzyme activity-reducing mutation (CYP2D6^∗10A) nor activity-increasing mutations (duplications) in CYP2D6 alone could appreciably predict HAL concentrations.

Haloperidol is a medication that is CPIC level B (for CYP2D6),¹² with a guideline currently in progressguidelines. Further, in the DPWG guidelines, there is a recommendation for initial dose to be reduced to 50% in PMs or for selection of an alternative medication based on a metabolic pathway different than CYP2D6. Possible dose adjustments are also mentioned for UMs.¹³ In a study of 70 patients in which the most commonly prescribed medication was HAL, the risk of tardive dyskinesia increased with increasing number of functional CYP2D6 genes, with UMs showing the highest risk (Koola et al., 2014).

Enzyme induction effects are also relevant for HAL metabolism. First, there is the effect of smoking. From a relevant review (Desai et al., 2001), it may be deduced that smoking increases the clearance of oral HAL (via effects including on CYP1A2), especially at doses lower than 0.5 mg/kg/day. Carbamazepine (which induces several CYPs including the CYP3As) reduces plasma HAL concentration (Hesslinger et al., 1999).

Chlorpromazine is a phenothiazine that was the first antipsychotic to be introduced (reviewed in Basu et al., 2007). Its biotransformation includes hydroxylation (by CYP2D6 and CYP1A2), N-methylation, N-N-didemethylation, N-oxidation, S-oxidation, and glutathione conjugation. The hydroxylated metabolite can undergo further oxidation leading to formation of an electrophilic quinone imine intermediate, which is capable of mediating toxic effects (by reacting with cellular proteins and DNA) or underdoing glutathione conjugation (Wen and Zhou, 2009). Muralidharan et al. (1996) confirmed the contribution of CYP2D6 to the hydroxylation pathway using quinidine, whilst also showing that CYP2D6 genetic polymorphism was not the major contributor to inter-individual variability in plasma concentrations. The latter finding was confirmed by Yoshii and colleagues (Yoshii et al., 2000), whose microsomal inhibition studies of chlorpromazine 7-hydroxylation showed that CYP1A2 may play a more important role in the hydroxylation reaction for individuals deficient in CYP2D6. Indeed, Gill et al. (1997) reported that an individual with schizophrenia who was homozygous for the CYP2D6^∗4 variant (then known as the “B”) and therefore a PM and had been intolerant and non-compliant with multiple medications settled on a very low dose (50 mg) of chlorpromazine (Gill et al., 1997).

Zuclopenthixol is a thioxanthene derivative used to treat schizophrenia, having high affinity for both D₂ and D₁ dopamine receptors (Kumar and Strech, 2005). Its metabolic pathways include sulfoxidation, N-dealkylation, and glucuronidation (Zhou et al., 2009), with metabolites not known to have antipsychotic activity.

Dahl et al. (1991) showed that clearance of zuclopenthixol was associated with debrisoquine hydroxylation, and further studies confirmed the role of CYP2D6 in zuclopenthixol metabolism (Zhou et al., 2009). Moreover, PMs had a 1.9-fold higher AUC of zuclopenthixol compared to NMs (n = 6 for each group) after a single 6 or 10 mg dose (Dahl et al., 1991). Linnet and Wiborg (1996) found similar results: investigation of phenotypic relationships to zuclopenthixol concentration showed that, in 12 psychiatric patients, CYP2D6 PMs had 60% greater concentrations than NMs, but were similar to NMs who were co-administered CYP2D6 inhibiting drugs.

Furthermore, in another study, psychiatric patients treated with zuclopenthixol who experienced adverse neurological events (tardive dyskinesia, parkinsonism) tended to have a higher frequency of non-functional CYP2D6^∗3 and ^∗4 alleles, but these results did not attain statistical significance (Jaanson et al., 2002).

Zuclopenthixol is CPIC level B,¹⁴ with a guideline in progress (and also not yet available from DPWG). One review suggests considering dose adjustment (58 and 88% for CYP2D6 PMs and IMs, respectively) or selecting an alternative medication (Stingl et al., 2013).

Aripiprazole was marketed as the first antipsychotic with dopamine and serotonin partial agonism. In Europe, aripiprazole is indicated for use in the treatment of schizophrenia and treatment of moderate to severe manic and episodes associated with bipolar I disorder and for the prevention of new manic episodes in those whose manic episodes respond to aripiprazole (Koskinen et al., 2010; Abilify 10 mg tablets; Abilify; Summary of Product Information, EMA). Other licensed indications include adjunctive treatment of major depressive disorder, Tourette’s syndrome, and irritability in autism spectrum disorder (Mailman and Murthy, 2010). Global therapeutic efficacy has been measured versus aripiprazole and dehydroaripiprazole serum concentrations, with a reported 68% response rate in those with concentrations of 150–300 ng/ml of aripiprazole, and a 57 and 50% response rate with concentrations less than 150 ng/ml or above 300 ng/ml, respectively (Kirschbaum et al., 2008). Therapeutic drug monitoring (TDM) has “recommended” (level 2 evidence) for aripiprazole by the interdisciplinary TDM group of the Arbeitsgemeinschaft für Neuropsychopharmakologie und Pharmakopsychiatrie (AGNP), with a therapeutic target range of 100–350 ng/ml for aripiprazole, or 150–500 for aripiprazole and dehydroaripiprazole (Hiemke et al., 2018).

Aripiprazole undergoes substantial first pass metabolism in the liver unless administered in a long-acting injectable (LAI) form. It is metabolized by dehydrogenation, hydroxylation, and N-dealkylation. In vitro studies show that CYP3A4 and CYP2D6 conduct the dehydrogenation and hydroxylation of aripiprazole, with CYP3A4 additionally catalyzing the N-dealkylation. Although a substrate for these enzymes, it does not appear to inhibit the activity of these enzymes. In clinical studies, 10–30 mg/day doses of aripiprazole had no significant effect on the metabolism of substrates of CYP2D6 or CYP3A4 activity as indexed by dextromethorphan; it does not appear to be an inhibitor of CYP2C9, CYP2C19, or CYP1A2 (Hjorthoj et al., 2015), nor a substrate for CYP1A enzymes, and hence no dose adjustment is required in smokers.

In a large pharmacokinetic study (N = 1288), CYP2D6 PMs and IMs had a 1.4 times increase in exposure to the active moiety compared to NMs, leading to a 15% decrease in medication dosage of aripiprazole. Switch in medication from aripiprazole was not, however, significantly associated with CYP2D6 status (Jukic et al., 2019).

The active dehydro-aripiprazole metabolite has a similar affinity as aripiprazole for dopamine D₂ receptors; at steady state it represents about 40% of the plasma concentration of aripiprazole (area under the curve or AUC; Hjorthoj et al., 2015), after oral administration or 29–33% after administration in the form of the LAI Abilify Maintena, and is therefore thought to contribute to the sustained pharmacologic effect of aripiprazole. Both aripiprazole and dehydro-aripiprazole are highly bound to plasma protein, mainly to albumin (reviewed in DeLeon et al., 2004). The average elimination half-life is oral aripiprazole is ∼75 h, but in CYP2D6 PMs, the average half-life extends to ∼146 h (Hjorthoj et al., 2015). The half-life of oral aripiprazole in CYP2D6 IMs (75.2 h) was significantly longer than that in CYP2D6 NMs (45.8 h); the systemic clearance of aripiprazole in IMs is approximately 60% that of NM subjects, with the maximum concentration being the same in IMs as in NMs (Kubo et al., 2007). At steady state, PMs have a significantly lower concentration to dose ratio than NMs, while in one report, IMs did not differ (van der Weide and van der Weide, 2015). However, van der Weide and van der Weide (2015) included in their IM group individuals who were heterozygous NMs (NM/PM genotype). In another report (Hendset et al., 2014), median serum concentrations were 1.6-fold or 1.8-fold higher in individuals of CYP2D6 PM/IM or IM/IM genotype, respectively than in those who were heterozygous NMs.

For patients who are known CYP2D6 PMs, FDA recommends administration of half of the usual dose of aripiprazole, and the DPWG guidelines recommend reducing maximum daily dose to 10 mg/day or 300 mg/month, i.e., 67–75% of the standard maximum dose.¹⁵ Given that at a dose as low as 2 mg, D₂ receptor occupancy is ∼70% (71.6 ± 5.5%, Kegeles et al., 2008), and the recommendation by consensus guidelines of doses of aripiprazole lower than those used in the initial marketing phase of the drug (Aitchison et al., 2009), it may well be recommendable to start at the lowest dose (2 mg) and to go no higher than 5 mg in CYP2D6 poor metabolizers. While there are as yet no guidelines for other CYP2D6 phenotypic groups, in the case of IMs, the Japanese data would suggest that a cautious dosing in the 2–5 mg range should be appropriate.

In addition, packaging information for aripiprazole offers additional guidelines should the medication be taken with known CYP inducers or inhibitors (Abilify - aripiprazole tablet, 2016). In the case of co-prescription of CYP3A4 or CYP2D6 inhibitors, dosage is recommended to be reduced (by 50% in the case of strong inhibitors such as ketoconazole and fluoxetine, respectively). Likewise, should aripiprazole be taken with known CYP3A4 inducers, dosage increase is recommended (doubling in the case of carbamazepine). On cessation of any inhibitors/inducers, the dose should be readjusted accordingly (Abilify).¹⁶

It has been noted that aripiprazole and 2,3-(dichlorophenyl) piperazine (2,3-DCPP), one of its metabolites, affect cholesterol biosynthesis by inhibiting 7-dehydrocholesterol reductase (DHCR7), the enzyme that converts 7-dehydrocholesterol (7-DHC) to cholesterol (Korade et al., 2010, 2016; Kim et al., 2016; Genaro-Mattos et al., 2018). Cholesterol is of critical importance to brain development; mutations in DHCR7 gene leads to Smith-Lemli-Opitz Syndrome, a neurodevelopmental condition, and exposure to DHCR7 inhibitors during the first trimester of pregnancy is associated with increased rates of fetal malformations, intrauterine death, and spontaneous abortions (Boland and Tatonetti, 2016). Thus, aripiprazole should be contraindicated during the first trimester of pregnancy; the Summary of Product Characteristics states¹⁷ “this medicinal product should not be used in pregnancy unless the expected benefit clearly justifies the potential risk to the fetus.” The most critical period for formation of the neural tube is the first six weeks of gestation, when many women do not realize they are pregnant. Therefore, it is recommended that women receiving aripiprazole in reproductive years should have a discussion of whether the woman is sexually active and of methods of contraception.

In a multiple-dose study, the mean terminal-phase elimination half-life of aripiprazole was 29.9 days and 46.5 days after 4-week injection of LAI 300 mg dose and 400 mg dose, respectively (Mallikaarjun et al., 2013). Data regarding differential half-life of the LAI by CYP2D6 genotype and/or CYP3A activity are not available. Aripiprazole lauroxil is a prodrug that undergoes bioactivation by hydroxylation and can be administered once every 6 weeks; it is similarly lacking pharmacogenetic data thus far.

Risperidone is an atypical antipsychotic used for treating schizophrenia, acting mainly on 5-HT_2A and D₂ receptors (Zhou et al., 2009); it is converted to 9-hydroxyrisperidone (9-OH-RIS) by CYP2D6, with the latter being excreted in the urine and feces. In a meta-analysis conducted by Stingl and Viviani (2015), they estimated dose adjustment of RIS to be 56 and 146% in CYP2D6 PMs and UMs, respectively, mentioning increased risk of toxicity in PMs. The DPWG note increased risk of treatment failure in CYP2D6 PMs and UMs and recommend using 67% of the standard dose in the former, and choosing an alternative drug or titrating the dose according to the maximum for the active metabolite (12 mg/day of paliperidone) in the latter. In a recent review by van Westrhenen et al. (2020), these recommendations were updated to suggest reducing the maximum dose by 33% (to 4 mg/day) in IMs as well as in PMs. For UMs, it was suggested to select an alternative medication or use therapeutic drug monitoring. It is worth noting that Cui et al. (2020) have found significant differences in recommendations of RIS dosage according to ethnicity. Specifically, adjustment in titration of this medication should be reduced in people of Asian ethnicity compared to Whites.

In a Norwegian population (Mannheimer et al., 2014, 2016), it was found that the metabolic ratio (MR) for RIS, expressed as RIS/9-OH-RIS, was, not surprisingly, associated with CYP2D6 PM status: an MR threshold of >1 predicted PM status with 91% accuracy (Mannheimer et al., 2016).

RIS metabolism by CYP2D6 is inhibited by the phenothiazine drug perazine when the two are co-administered (Paulzen et al., 2017), resulting in an increase in RIS and (RIS + 9-OH-RIS) concentrations and a reduction in the 9-OH-RIS/RIS ratio. Animal models have previously shown the role of phenothiazines in inhibiting the CYP2D family (Daniel et al., 2005).

In a study focusing on the relationship between genetic and epigenetic variation and response to RIS, three CpG sites in CYP2D6 and two to three CpG sites in CYP3A4 appeared to be more methylated in poor responders (Shi et al., 2017).

The effect of CYP2D6 genotype on RIS metabolism has been studied in young Thai autistic spectrum individuals (Nuntamool et al., 2017). Genotypes CYP2D6^∗5/^∗10, ^∗10/^∗10 and ^∗10/^∗41 showed reduced RIS metabolism, with significantly higher RIS plasma concentrations. While such an association was not seen in the CYP2D6^∗4/^∗10 genotype group, this was likely owing to the relatively low frequency of the CYP2D6^∗4 variant in this ethnicity, and the wide spread of the data in the small subgroup of CYP2D6^∗4/^∗10 genotype. The CYP2D6^∗10 variant was also associated with higher MR of RIS/9-OH-RIS. Hendset et al. (2014) genotyped for CYP2D6 ^∗3, ^∗4,^∗ 5, ^∗6, ^∗9, ^∗10 and ^∗41, and classified as “^∗1/def” (heterozygous for normal and deficient function) or “def/red” (heterozygous for deficient and reduced function); RIS serum concentration was 4.5 times higher in the def/red group compared to ^∗1/def. In addition, a 3 to 4-fold increase in the serum concentration of RIS was shown in the red/red group.

In addition to variation in metabolic activity and treatment response, Molden et al. (2016) found evidence of a relationship between genotype and discontinuation of treatment. Individuals classified as PMs for CYP2D6 had active moiety (RIS + 9-OH-RIS) concentration 1.5 times higher than NMs. Consequently, there was an over-representation of adverse events and discontinuation of treatment for PMs. Conversely, a similar study with Croatian psychiatric patients receiving RIS injections found individuals classified as UM with concentrations of RIS active moiety (RIS + 9-OH-RIS) not reaching the threshold recommended for therapeutic range (Ganoci et al., 2016). Kaur et al. (2017) reported an association between the CYP2D6^∗4 PM haplotype and treatment dropout due to poor response.

In the largest study of RIS and CYP2D6 to date (1288 patients), approximately 1.4 and 1.6-fold RIS exposure increase was observed in CYP2D6 IMs and PMs, respectively (Jukic et al., 2019). A higher incidence of RIS-associated ADRs (de Leon et al., 2005) and treatment failure (Jukic et al., 2019) is observed in CYP2D6 PMs compared with NMs, with increased treatment failure rate also being observed in CYP2D6 UMs (Jukic et al., 2019). It is possible that the latter may be exposed to subtherapeutic drug concentrations, and also possible the effect of CYP2D6 on normally minor synthesis pathways for serotonin and dopamine may at least partly relate to such associations. Recent systematic reviews and meta-analyses support the need for dosage adjustment of RIS based on CYP2D6 genotype (Cui et al., 2020; Zhang et al., 2020).

The relationship between CYP2D6 and hyperprolactinemia (a possible adverse effect of RIS) appears to be U-shaped, with a tendency (though not consistently replicated) for both extremes of CYP2D6 metabolic phenotype (i.e., PMs and UMs) to show an association with hyperprolactinemia (Troost et al., 2007; Roke et al., 2013; Youngster et al., 2014; Sukasem et al., 2016). Hyperprolactinemia has also been associated with the DRD2 Taq1A variant (Sukasem et al., 2018).

Interaction of known CYP2D6 inhibitors such as fluoxetine, bupropion, lamotrigine, sertraline, and citalopram are strongly correlated with the concentration of RIS in young male patients, compared to the available concentration of its metabolites (Calarge and Miller, 2011). A similar relationship has been described for thioridazine, and levomepromazine (Mannheimer et al., 2008). The same relationship was not found for duloxetine, another known CYP2D6 inhibitor (Hendset et al., 2010). Ishak et al. (2008) have described an association between RIS discontinuation caused by DDI from CYP2D6 inhibitors.

To a lesser extent, RIS metabolism is also mediated by CYP3A4 and DDI with inducers of this CYP enzyme are supported by the literature. Co-medication with armodafinil results in a decrease in plasma concentration of both RIS and 9OH-RIS (Darwish et al., 2015). The same relationship is true for rifampin (Mahatthanatrakul et al., 2007; Kim et al., 2008).

Olanzapine is an antipsychotic licensed for use in schizophrenia and related psychotic disorders and bipolar disorder. The main circulating metabolites are desmethylolanzapine and olanzapine-10-glucuronide (Callaghan et al., 1999; Erickson-Ridout et al., 2011; Lu et al., 2016). The conversion to desmethylolanzapine is predominantly catalyzed by CYP1A2, with lesser roles for CYP2D6, CYP2C8, and CYP2C19 (Ereshefsky, 1996; Callaghan et al., 1999; Korprasertthaworn et al., 2015; Okubo et al., 2016). Okubo and colleagues investigated the role of CYP1A2, CYP2D6, and FMO3 in individuals of varying CYP2D6 and FMO3 genotype (Okubo et al., 2016). Olanzapine N-demethylation and N-oxygenation were found to be catalyzed by CYP1A2 and CYP2D6, and by CYP2D6 and FMO3, respectively, in experiments using liver microsomes and recombinant enzymes. The effects on olanzapine oxidation activities of furafylline (which inhibits CYP1A2), quinidine (inhibits CYP2D6), and heat treatment (inhibits FMO3-mediated activities) were investigated. Each, and the combination of all three treatments suppressed the metabolic clearances of olanzapine by 28, 33, 25, and 85%, respectively. Using recombinant CYP2D6 enzymes CYP2D6.1 and CYP2D6.10, only the wild-type variant was capable of the 2-hydroxylation conversion of olanzapine into 2-hydroxymethyl-olanzapine; CYP2D6 appears to be the only enzyme catalyzing olanzapine 2-hydroxylation. Direct glucuronidation (at the 10 and 4 positions) is conducted by UGT1A4 and UGT2B10, with the UGT1A4^∗3 and UGT2B10^∗2 haplotypes being associated with increased and decreased glucuronidation, respectively (Erickson-Ridout et al., 2011).

Drug-drug (Drug Interactions Flockhart Table^TM; DrugBank) interactions are of importance in the prescription of olanzapine. CYP1A2 is induced by smoking; the plasma concentration to dose ratio of olanzapine is therefore lower in smokers (Tsuda et al., 2014). Inhibition of CYP1A2 by fluvoxamine also increases the concentration to dose ratio (Chiu et al., 2004). CYP1A2 is also inhibited by estrogens; as a result, gender (clearance is reduced in women) and body fat content influence the metabolism of olanzapine (Ereshefsky, 1996; Callaghan et al., 1999). Valproic acid co-prescription results in a decrease in olanzapine concentration (reviewed by Vella and Mifsud, 2014). In contrast, protease inhibitors used in the treatment of HIV such as ritonavir in combination with fosamprenavir induce olanzapine metabolism (via CYP1A2 and/or UGT), leading to a recommendation to increase olanzapine dose by 50% when prescribed with such (Jacobs et al., 2014).

Quetiapine is another commonly prescribed antipsychotic. While literature supports CYP3A4 being the main enzyme in the quetiapine metabolic pathway, CYP2D6 is involved in the further metabolism of its principal metabolite, N-desalkylquetiapine. In an analysis of therapeutic drug monitoring data, patients from a Norwegian psychiatric hospital were genotyped for CYP2D6, CYP3A5, and ABCB1 (3435C>T) and the associations with dose-corrected serum concentrations of quetiapine and N-desalkylquetiapine were analyzed (Bakken et al., 2015). The mean dose-corrected serum concentration (C/D) of N-desalkylquetiapine was estimated to be 33 and 22% higher in CYP2D6 PMs (P = 0.03) and heterozygous CYP2D6 NMs (P = 0.001), respectively, compared with CYP2D6 NMs. There was no significant association with ABCB1 3435C>T polymorphism or CYP3A5 genotype.

Quetiapine has, however, been observed to have a serum level 2.5 times higher in those either heterozygous or homozygous for CYP3A4^∗22 compared to those of CYP3A4 wild-type (van der Weide and van der Weide, 2014), with concentration to dose ratios that were 67% higher. The percentage of patients who had levels of quetiapine above the therapeutic range was also about five times higher in the ^∗22 carrier group (16.1 versus 2.9%). Quetiapine serum levels based on reduced CYP3A4 metabolic activity were comparable to results found with coadministered CYP3A4 inhibitors, such as ketoconazole. The frequency of the CYP3A4^∗22 haplotype is up to 10%.

In terms of DDI, valproate coadministration with quetiapine appears to result in a variable degree of increase in quetiapine plasma levels (Aichhorn et al., 2006; Winter et al., 2007), which may result in toxicity on occasion (Anderson and Fritz, 2000). In a review of therapeutic monitoring data from more than 2000 patients, it was reported that the following factors were associated with an increase in quetiapine concentration: age of at least 70 years, comedication with clozapine, fluvoxamine, and to a lesser extent citalopram/escitalopram, while, conversely, the following were associated with reduced quetiapine concentration: age under 18 years and comedication with carbamazepine or oxazepam, and to a lesser extent levomepromazine or lamotrigine (Castberg et al., 2007). The largest effect sizes were seen with fluvoxamine (+159%), clozapine (+82%), age at least 70 years (+67%), and carbamazepine (−86%). Another study found that dose-corrected quetiapine concentrations were approximately 60% lower in patients co-medicated with lamotrigine (Andersson et al., 2011).

Ziprasidone is a less commonly prescribed antipsychotic. Data indicate ziprasidone is mainly metabolized by glutathione and enzymatic reduction by aldehyde oxidase, followed by S-methylation to S-methyl-dihydroziprasidone by thiolmethyltransferase (Obach and Walsky, 2005). Approximately one third of its clearance is thought to be CYP3A4-dependent (Beedham et al., 2003). It is therefore subject to CYP3A-mediated induction (e.g., by carbamazepine, Miceli et al., 2000) and inhibition effects. As this medication has been associated with increases in the QT_c interval (Aronow and Shamliyan, 2018), inhibition effects (e.g., by fluvoxamine or ketoconazole) should be avoided. Ziprasidone is also contraindicated in the presence of other medications that also prolong QT_c (Kutcher et al., 2005; Table 4 in Hicks et al., 2017).

Imipramine, the first TCA was derived from a phenothiazine, showing improvement without serious side effects in 500 patients with severe depression (Kuhn, 1958; Hillhouse and Porter, 2015). Although TCAs are still used (e.g., second-line or with somatic symptoms) to treat depression (Uher et al., 2009b; American Psychiatric Association, 2010; Kennedy et al., 2016), the treatment of pain (e.g., migraine, neuropathic, cancer-associated) is now their more common therapeutic use (Watson, 2000; Laird et al., 2008; Baltenberger et al., 2015).

Tricyclics include tertiary and secondary amines. A tertiary amine has a nitrogen bonded to three carbons, while in the case of a secondary amine, the nitrogen is bonded to only two carbons. The tricyclics amitriptyline, clomipramine, imipramine, trimipramine, doxepin, and dothiepin are tertiary amines. Tertiary amines are demethylated to secondary amines mainly by CYP2C19, but also by CYP1A2, CYP2C9, and CYP3A4, while both tertiary and secondary undergo parallel hydroxylation reactions mainly by CYP2D6, with CYP2C19 making a lesser contribution (Bertilsson et al., 2002; Aitchison, 2003; Bertilsson, 2007). The secondary amine metabolites of amitriptyline and imipramine are nortriptyline and desipramine, respectively, each also available as licensed medications.

Using hepatic microsomes of varying CYP2C19 activity and recombinant CYPs, Koyama et al. (1997) demonstrated that imipramine N-demethylation was catalyzed by CYP2C19 and CYP1A2 (high affinity and low affinity components, respectively), imipramine 2-hydroxylation was mediated by CYP2D6 and CYP2C19 (high affinity and low affinity components, respectively), and that in individuals deficient in CYP2C19, CYP1A2, and CYP2D6 play a major role in imipramine N-demethylation and 2-hydroxylation respectively. Among the recombinant human CYPs, CYP2C19, 2C18, 2D6, 1A2, 3A4, and 2B6 in rank order catalyzed the N-demethylation, whereas CYP2D6, 2C19, 1A2, 2C18, and 3A4 catalyzed the 2-hydroxylation. In a monoclonal antibody inhibition, Yang et al. (1999) concluded similarly that imipramine was metabolized to 2-hydroxyimipramine by 2C19 and 2D6, and to desipramine by 1A2, 2C18, 2C19, and 2D6, with the contributions of the isoforms to desipramine formation varying for 2C19 (13–50%), 1A2 (23–41%), and 3A4 (8–26%).

Tricyclic antidepressants inhibit presynaptic noradrenaline (also known as norepinephrine) and serotonin reuptake via the noradrenaline and serotonin transporters, respectively, with the tertiary amines having a greater affinity for the serotonin transporter than the secondary amines, which are relatively selective for the noradrenaline transporter (Owens, 1996). The tertiary amines are therefore dual SNRIs, while the secondary amines are noradrenaline reuptake inhibitors (or NARIs). There are also contrasts in their CYP inhibition. Tertiary amines TCAs (e.g., amitriptyline, imipramine) inhibit CYP2C19 (estimated Ki of 37.7 and 56.8 μM, respectively). By contrast, the secondary amines show negligible CYP2C19 inhibition activity (Shin et al., 2002) but inhibit CYP2D6 slightly more than tertiary amine TCAs; for example, estimated K_i values for the tertiary amine TCAs amitriptyline and imipramine are 31.0, 28.6 μM, respectively, with K_i s for nortriptyline and desipramine being 7.9 and 12.5 (Shin et al., 2002). Although therapeutic plasma concentrations are less than 1 μM (Yau et al., 2007; Aitchison et al., 2010), cerebral concentrations may be higher (Weigmann et al., 2000; Aitchison et al., 2010). Further, this differential may be affected by factors such as p-gp expression at the blood–brain barrier. It is therefore possible that with repeated dosing, as the concentration of a tertiary amine TCA increases in the brain, that the level of CYP2C19 inhibition increases, and that this leads to reduction in the demethylation reaction centrally. This would be expected to be associated with a greater degree of dual reuptake inhibition and may at least partly explain the clinical observation of time for antidepressant effect to maximize. This hypothesis is consistent with a report of an inverse relationship between CYP2C19 activity and response to TCAs (mainly tertiary amines, Aitchison et al., under revision).

There are also contrasts between the tertiary and secondary amines and side effect/adverse drug reaction potential. The side effects are associated with antagonism at the following receptors: adrenergic α1 and α2 receptors, muscarinic (cholinergic) receptors, and histamine H1 receptors (Cusack et al., 1994; Owens et al., 1997; Sanchez and Hyttel, 1999). Specifically, blockade of muscarinic receptors in the parasympathetic nervous system is associated with dry mouth, blurred vision, constipation, urinary retention, and if at toxic levels, delirium; alpha adrenergic receptor antagonism is associated with orthostatic hypotension; and histamine H1 receptor blockade with sedation and weight gain. Other side effects (such as palpitations, vertigo, sweating, tremors, and interference with sexual function) may also occur (Asberg et al., 1970; Ziegler et al., 1978; Uher et al., 2009a; Hodgson et al., 2015) and may represent more than one pharmacodynamic mechanism. The tertiary amine TCAs have greater cholinergic receptor binding than the secondary amines, which in turn have greater affinity than the hydroxylated metabolites (Rudorfer and Potter, 1999). Some effects may be related to specific metabolites (e.g., N-methyl quaternary ammonium derivatives of amitriptyline, doxepin, and imipramine are antagonists at both central nervous system and cardiac muscarinic receptors) (Ehlert et al., 1990). Hydroxylated metabolite concentration has been associated with increases QTc interval (Schneider et al., 1988; Stern et al., 1991). It is therefore possible that CYP2D6 UM might have elevated hydroxy-metabolite plasma concentrations (Bertilsson et al., 1985) resulting in an increased risk of cardiotoxicity. Moreover, therapeutic drug monitoring does not usually include measuring hydroxylated metabolite plasma concentrations. In the case of a combination of CYP2C19 poor metabolizer status and CYP2D6 UM status, it might be advisable to avoid TCA prescription; this is in fact the CPIC recommendation for this combination [Table 4 in Hicks et al. (2017)].

The association between CYP2D6 and CYP2C19 genotype and clinical response to TCAs (treatment efficacy and/or side effects) has been reviewed (Hicks et al., 2017). In brief, studies support the existence of a concentration–effect relationship for TCAs and/or their active metabolites (Ziegler et al., 1977; Sjöqvist et al., 1980; Gram et al., 1984; Perry et al., 1987) (Preskorn and Jerkovich, 1990). In an early report, high concentrations of nortriptyline were linked to adverse effects, with decreased antidepressant effect (Asberg et al., 1970). Concentration-dependent side effects have been observed in individuals deficient in CYP2D6 when treated with usual doses of TCAs from accumulation of the parent drug and/or active metabolites (Bertilsson et al., 1981; Sjöqvist and Bertilsson, 1984; Balant-Gorgia et al., 1989). Ethnic groups with a higher frequency of CYP2D6 IM alleles achieve higher levels of TCAs than Whites and have a faster rate of recovery from depressive episodes (Raskin and Crook, 1975; Ziegler and Biggs, 1977; Rudorfer and Robins, 1982). An excess of CYP2D6 PM alleles has been found in amongst patients with a history of adverse reactions to TCAs and relevant SSRIs (Chen et al., 1996). CYP2D6 PMs have high levels of desipramine, associated with adverse effects necessitating dose reduction (Spina et al., 1997). An inverse correlation between the frequency of adverse drug events and number of functional CYP2D6 genes has been found, including patients on TCAs (Chou et al., 2003). Studies published since the Hicks et al. (2017) review include that of Hodgson et al. (2015).

There are guidelines by both CPIC and DPWG (for a comparison, see Bank et al., 2018). CPIC guidelines for CYP2D6 are as follows: for both PMs and UMs, it is suggested to avoid use of TCAs due to possible side effects or subthreshold concentrations, respectively. In both cases, when TCAs are still prescribed, therapeutic drug monitoring is recommended, with PMs starting at 50% regular dosage and for UMs consideration being given to use TDM to titrate up to a higher target dose. The DPWG provide specific suggested increases in the starting dosages for amitriptyline, clomipramine, doxepin, imipramine, and nortriptyline of 125, 150, 200, 170%, respectively followed by TDM (Bank et al., 2018). For CYP2D6 IMs, a 25% reduction in the initial dose is recommended by CPIC. For many drugs, evidence is still accumulating, and therefore implementation of the recommendations is “optional,” or at prescriber discretion.

In regard to CYP2C19 status, there are CPIC guidelines for the tertiary amines amitriptyline, clomipramine, doxepin, imipramine, and trimipramine. For CYP2D6 UMs, RMs, or PMs, CPIC provides an optional recommendation to substitute with medications not metabolized by CYP2C19, such as nortriptyline or desipramine. In the case of CYPC19 PMs, a 50% decrease in initiation dose is suggested, with TDM to guide titration (Hicks et al., 2017).

There are also CPIC guidelines for amitriptyline where both CYP2D6 and CYP2C19 data are available. If an individual is a CYP2D6 or CYP2C19 PM and a NM for the other enzyme, it is recommended to avoid the medication or, if warranted, consider a 50% decrease in initiation dose; for a CYP2D6 UM and CYP2C19 NM, it is recommended to avoid the medication or, if warranted, consider titrating to a higher target dose (compared to CYP2D6 NMs); and for a CYP2D6 IM and CYP2C19 NM, to consider a 25% decrease in initiation dose ((Hicks et al., 2017). No adjustments in dosage is necessary for those who are NMs for CYP2D6 and CYP2C19, or an NM for CYP2D6 and an IM for CYP2C19 (Hicks et al., 2017).

Mirtazapine acts as antagonist at adrenergic α₂-autoreceptors and α₂-heteroreceptors as well as at 5-HT₂ and 5-HT₃ receptors (Anttila and Leinonen, 2001). The α₂-autoreceptor blockade leads to increased release of noradrenaline while the blockade of α₂-heteroreceptor on serotonergic neurons increases serotonin release. Owing to antagonism of 5-HT₂ and 5-HT₃, transmission is enhanced at only 5-HT_1A (and related receptors). It is a racemic mixture of R(−) and S(+)-enantiomers, with effects on heart rate and blood pressure correlating more strongly with R (−) than with S (+) concentration, and sedation being associated with both enantiomers (Brockmöller et al., 2007). The main metabolic pathway for mirtazapine is 8-hydroxylation, catalyzed by mainly by CYP2D6 (65%) at low concentrations, reducing to 20% at higher concentrations, where CYP1A2 (50%), CYP3A4 (20%), and CYP2C9 (10%) contribute more (Dahl et al., 1997; Störmer et al., 2000). Other metabolic pathways are N-demethylation and N-oxidation. The former is conducted by mainly by CYP3A4 (50–70%), with CYP1A2 (50% at low concentrations, 5% at high concentrations), CYP2C8 (<20%), and CYP2C9 (<5%) also contributing. N-oxidation is catalyzed by CYP1A2 and CYP3A4, with the former playing a larger role (80%) at low concentrations and the latter being responsible for a greater proportion (85%) of the reaction at higher drug concentrations (Dahl et al., 1997; Störmer et al., 2000). Enzyme polymorphism may additionally affect the relative contributions of these CYPs.

The maximum concentration and area under the curve are greater in females as compared to males (Timmer et al., 2000; Sennef et al., 2003). In non-smokers and at lower concentrations of mirtazapine, CYP2D6 genotype affects the plasma levels and clearance of the S-enantiomer and its metabolites (Störmer et al., 2000; Brockmöller et al., 2007; Sirot et al., 2012; Hayashi et al., 2015). At higher concentrations (250 μM), CYP3A4 contributes to about 70%, while CYP2D6, CYP2C8, CYP2C9, and CYP1A2 each account for less than 15% of its metabolism (Störmer et al., 2000). Unlike the tricyclics, there is no clear relationship between mirtazapine plasma concentration and its efficacy. While an increase in the maximal serum concentration for coadministered amitriptyline has been described (Sennef et al., 2003), the overall inhibitory effect of mirtazapine on CYPs is not thought to be clinically significant (Anttila and Leinonen, 2001; Spina et al., 2008). In the Sennef et al. (2003) study, coadministered amitriptyline increased the maximum concentration of mirtazapine (by 36%) in only males. S-hydroxymirtazapine concentration has been reported as being elevated in individuals of CYP2B6^∗6/^∗6 genotype (Sirot et al., 2012). As yet, there are no DPWG or CPIC guidelines for mirtazapine based on genotype.

The second SSRI to be synthesized, fluoxetine was the first SSRI to enter widespread use (Wong et al., 1974, 1975). Selective serotonin reuptake inhibitors are now widely used to treat depression, escitalopram having the highest affinity for the serotonin transporter (Owens et al., 2001) and being an allosteric modulator, one molecule increasing the binding of a second at this target.

In brief, SSRIs are partly metabolized by CYP2D6 (Brosen, 1993). Demethylenation is the initial step of paroxetine metabolism (an SSRI), primarily conducted by CYP2D6 (a high affinity saturable process,Bloomer et al., 1992). Further conjugation of paroxetine results in glucuronide and sulfate conjugated metabolites (Haddock et al., 1989). Paroxetine is a potent competitive inhibitor of CYP2D6 (Bourin et al., 2001) nonetheless, differences in steady-state plasma concentration of paroxetine by CYP2D6 phenotype are seen (Gex-Fabry et al., 2008). Higher doses of paroxetine (e.g., 30 mg) are associated with a greater degree of CYP2D6 inhibition (Findling et al., 2006). In diabetic neuropathy, paroxetine has an analgesic effect, plasma concentrations greater than 300–400 nmol/l being associated with optimal response (Sindrup et al., 1990, 1991).

Fluoxetine is a racemic mixture of S(+) and R(−)-fluoxetine, with the former being metabolized predominantly by CYP2D6 to S-norfluoxetine and the latter by CYP2D6 and CYP2C9 to R-norfluoxetine; CYP3A4 and CYP2C19 make minor contributions to this demethylation reaction (Hicks et al., 2015). R/S-fluoxetine and S-norfluoxetine are all potent SSRIs, with R-norfluoxetine being 20 times less potent (Hicks et al., 2015). The strength of CYP2D6 inhibition for SSRIs is as follows in reducing order: paroxetine, fluoxetine, norfluoxetine, desmethylcitalopram, fluvoxamine, sertraline, citalopram (Baumann and Rochat, 1995). Although fluoxetine is less potent as an inhibitor of CYP2D6 than paroxetine, owing to its substantially longer half-life – 1–3 days and 4–6 days after acute and chronic administration, respectively, with the corresponding values being 4 and 16 days for norfluoxetine,¹⁸ inhibition effects may endure for weeks to months after multiple dosing (Liston et al., 2002). Fluoxetine and sertraline also inhibit CYP2C19 (Bertilsson and Dahl, 1996) while norfluoxetine is a moderate CYP3A4 inhibitor (Hemeryck and Belpaire, 2002).

The primary route of metabolism for citalopram (a racemic mixture of the R- and S-enantiomers of citalopram) and escitalopram (S-citalopram) is N-demethylation by CYP2C19, CYP2D6, and CYP3A4 (Sindrup et al., 1993; Kobayashi et al., 1997; Rochat et al., 1997; von Moltke et al., 2001). CYP2D6 then conducts the N-demethylation of N-desmethylescitalopram to N-didesmethylescitalopram (von Moltke et al., 2001). The medication and its metabolites may inhibit enzymes: citalopram and R- or S-desmethylcitalopram are weak inhibitors of CYP2C19, while R- and S-didesmethylcitalopram are moderate inhibitors, with mean IC₅₀ values of 18.7 and 12.1 μM, respectively. S-citalopram and S-desmethylcitalopram are weak inhibitors of CYP2D6 (IC₅₀ = 70–80 μM); R-desmethylcitalopram shows stronger inhibition at this enzyme (IC₅₀ 25.5 ± 2.1 μM) (von Moltke et al., 2001). Fluvoxamine is predominantly a CYP1A2 inhibitor (Christensen et al., 2002) but also inhibits other CYPs including the CYP3As (Hemeryck and Belpaire, 2002).

SSRIs are more 20 to 1500-fold more selective for inhibiting serotonin than noradrenaline. They do not stimulate the release of serotonin or norepinephrine presynaptically (Rothman et al., 2001) and have weak/no-direct pharmacological action at postsynaptic serotonin receptors (e.g., 5-HT_1A, 5-HT_2A, and 5-HT_2C) (Thomas et al., 1987; Owens et al., 1997; Sanchez and Hyttel, 1999), and minimal binding affinity for other postsynaptic receptors (adrenergic α₁, α₂, and β, histamine H₁, muscarinic, and dopamine D₂ receptors) (Thomas et al., 1987; Owens et al., 1997).

Associations between SSRI phenotypes (concentrations, efficacy, tolerability) and CYP2D6 and CYP2C19 genotypes are provided in Supplementary Tables S7–S11 in Hicks et al. (2015). In a meta-analysis of the main functional CYP2C19 variants in Whites (the CYP2C19^∗2 and the CYP2C19^∗17, plus wild-type by exclusion of these) for individuals treated with citalopram or escitalopram (in the GENDEP, STAR^∗D, GenPod, and PGRN-AMPS studies), CYP2C19 PMs had greater symptom improvement and higher remission rates compared to NMs (Fabbri et al., 2018). This is consistent with earlier data indicating that CYP2C19 PMs respond better to escitalopram if treatment is tolerated (Mrazek et al., 2011). At weeks 2–4, PMs showed increased risk of side effects (Fabbri et al., 2018). In a retrospective analysis of data from 2087 patients treated with escitalopram and genotyped for CYP2C19, PMs had an increase in exposure and a higher rate of treatment dropout compared with CYP2C19 NMs (Jukic et al., 2018). Conversely, the CYP2C19^∗17 haplotype was associated with an increase in CYP2C19 activity by approximately 20%, with those of CYP2C19^∗1/^∗17 and CYP2C19^∗17/^∗17 genotype showing a 50% increase in treatment failure rate compared with NMs (Jukic et al., 2018). Moreover, replicated findings that CYP2C19 UMs treated with escitalopram exhibit increased suicidal ideation (Jukic et al., 2017; Rahikainen et al., 2019) indicates that distinguishing between CYP2C19 NMs and UMs is clinically relevant for the escitalopram treatment. Shelton et al. (2020) using a combinatorial PGx algorithm (covering several different genes) reported a significant association with variation in the metabolism of escitalopram/citalopram.

The CPIC guidelines for SSRIs (Hicks et al., 2015) cover two medications for CYP2D6: paroxetine and fluvoxamine. The recommendation for paroxetine in the case of CYP2D6 UMs is to select an alternative drug and likewise for PMs, with implementation being optional for the latter. For fluvoxamine, in the case of CYP2D6 UMs there was insufficient data for a recommendation, with an optional recommendation to consider a 25–50% reduction in the starting dose for CYP2D6 PMs, and titrate to response, or consider using an alternative medication not metabolized by CYP2D6.

Three medications are included in the CPIC SSRI guidelines in regard to CYP2C19: citalopram, escitalopram and sertraline. A 50% reduction of the standard dosage for the three drugs is recommended for PM status, with and titration to response, or considering using an alternative medication not metabolized by CYP2D6 (strength of the recommendation being moderate for citalopram and escitalopram and moderate for sertraline). For CYP2C19 UMs, for citalopram and escitalopram, selection of a medication not metabolized by CYP2C19 is recommended, while for sertraline, initiation at the normal dose may be tried, with substitution being considered if patients do not respond to treatment. The recommendations are classified as moderate for citalopram/escitalopram and optional for sertraline (Hicks et al., 2015).

In addition to the CPIC guidelines, Stingl et al. (2013) suggest that in the case of fluoxetine (not included in the guidelines above), due to its role as both substrate and inhibitor of CYP2D6, physicians should be careful if co-prescribing it with other CYP2D6 substrates.

Venlafaxine is a SNRI, which means that like the tertiary amine tricyclics, it inhibits neurotransmitter reuptake at both the serotonin and the noradrenaline (also known as epinephrine) transporters The major metabolic route for venlafaxine is O-demethylation, which is mediated very specifically by CYP2D6 to an active metabolite, O-desmethylvenlafaxine (Otton et al., 1996). The N-demethylation is conducted by CYP3A4 and CYP2C19 (Sangkuhl et al., 2014). This means that the ratio of the O- and N-demethylated metabolites of venlafaxine may in fact be used as a biomarker of CYP2D6 activity, predicting CYP2D6 poor metabolizers with a specificity and sensitivity of >85% (Mannheimer et al., 2016). In in vitro studies, venlafaxine is a weaker inhibitor of CYP2D6 than are the SSRIs paroxetine, fluoxetine, fluvoxamine, and sertraline, and has minimal or no effect on CYP1A2, CYP2C9, and CYP3A4 (Ball et al., 1997; von Moltke et al., 1997). In a study of 1003 Scandinavians (mostly White), it was found that CYP2D6 metabolism measured as the O/N-desmethylvenlafaxine ratio was significantly lower in carriers of CYP2D6^∗41 vs. CYP2D6^∗9–10 (Jukic et al., 2019). The annotated DPWG guideline states that for CYP2D6 poor (PM) and intermediate metabolizers (IM), select an alternative to venlafaxine or reduce the dose and monitor patient’s plasma metabolite level; for CYP2D6 ultrarapid metabolizers (UM), increase dose to 150% of the normal dose or select an alternative to venlafaxine.¹⁹

Duloxetine acts as a serotonin and noradrenaline reuptake inhibitor, and a weak dopamine reuptake inhibitor (e.g., in the frontal cortex).²⁰ CYP1A2 and to a lesser extent CYP2D6 convert duloxetine into its main metabolites 4-hydroxy and 5-hydroxy duloxetine; activity (Knadler et al., 2011). CYP1A2 inducers including cigarette smoke therefore result in a reduction in duloxetine concentration (Augustin et al., 2018).