Association of CYP2C19 and CYP2D6 Poor and Intermediate Metabolizer Status With Antidepressant and Antipsychotic Exposure

This systematic review and meta-analysis quantifies the difference in the antipsychotic and antidepressant exposure among patients with genetically determined CYP2C19 and CYP2D6 poor, intermediate, and normal metabolism.

Key Points

Question

What is the difference in the expected antipsychotic and antidepressant exposure between genetically associated CYP2C19 and CYP2D6 poor (PM), intermediate (IM), and normal (NM) metabolism?

Findings

A systematic review and meta-analysis of 94 unique studies and 8379 unique patients quantified the increases of risperidone, aripiprazole, and haloperidol exposure in patients with CYP2D6 PM and IM status and increases of escitalopram and sertraline exposure in patients with CYP2C19 PM and IM status as compared with patients with the NM group.

Meaning

The obtained results represent a scientific foundation for CYP2D6/CYP2C19 genotype-based dosing recommendations that could potentially lead to improved clinical outcome in drug treatment for patients with psychiatric disorders.

Abstract

Importance

Precise estimation of the drug metabolism capacity for individual patients is crucial for adequate dose personalization.

Objective

To quantify the difference in the antipsychotic and antidepressant exposure among patients with genetically associated CYP2C19 and CYP2D6 poor (PM), intermediate (IM), and normal (NM) metabolizers.

Data Sources

PubMed, Clinicaltrialsregister.eu, ClinicalTrials.gov, International Clinical Trials Registry Platform, and CENTRAL databases were screened for studies from January 1, 1990, to June 30, 2020, with no language restrictions.

Study Selection

Two independent reviewers performed study screening and assessed the following inclusion criteria: (1) appropriate CYP2C19 or CYP2D6 genotyping was performed, (2) genotype-based classification into CYP2C19 or CYP2D6 NM, IM, and PM categories was possible, and (3) 3 patients per metabolizer category were available.

Data Extraction and Synthesis

The Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines were followed for extracting data and quality, validity, and risk of bias assessments. A fixed-effects model was used for pooling the effect sizes of the included studies.

Main Outcomes and Measures

Drug exposure was measured as (1) dose-normalized area under the plasma level (time) curve, (2) dose-normalized steady-state plasma level, or (3) reciprocal apparent total drug clearance. The ratio of means (RoM) was calculated by dividing the mean drug exposure for PM, IM, or pooled PM plus IM categories by the mean drug exposure for the NM category.

Results

Based on the data derived from 94 unique studies and 8379 unique individuals, the most profound differences were observed in the patients treated with aripiprazole (CYP2D6 PM plus IM vs NM RoM, 1.48; 95% CI, 1.41-1.57; 12 studies; 1038 patients), haloperidol lactate (CYP2D6 PM vs NM RoM, 1.68; 95% CI, 1.40-2.02; 9 studies; 423 patients), risperidone (CYP2D6 PM plus IM vs NM RoM, 1.36; 95% CI, 1.28-1.44; 23 studies; 1492 patients), escitalopram oxalate (CYP2C19 PM vs NM, RoM, 2.63; 95% CI, 2.40-2.89; 4 studies; 1262 patients), and sertraline hydrochloride (CYP2C19 IM vs NM RoM, 1.38; 95% CI, 1.27-1.51; 3 studies; 917 patients). Exposure differences were also observed for clozapine, quetiapine fumarate, amitriptyline hydrochloride, mirtazapine, nortriptyline hydrochloride, fluoxetine hydrochloride, fluvoxamine maleate, paroxetine hydrochloride, and venlafaxine hydrochloride; however, these differences were marginal, ambiguous, or based on less than 3 independent studies.

Conclusions and Relevance

In this systematic review and meta-analysis, the association between CYP2C19/CYP2D6 genotype and drug levels of several psychiatric drugs was quantified with sufficient precision as to be useful as a scientific foundation for CYP2D6/CYP2C19 genotype-based dosing recommendations.

Introduction

The efficacy of psychiatric drugs is suboptimal; however, because the development of new antipsychotics and antidepressants is slow, it is of paramount importance to use the currently available drugs as effectively as possible. An important aspect of effective use is dose personalization because, owing to interindividual differences in drug metabolism, the dose required to achieve optimal blood levels of antidepressants and antipsychotics varies substantially between patients.¹ Recently published meta-analyses^2,3 focused on dose-response curves for antipsychotics and antidepressants supported the claim that the appropriate dosing is important for maximizing the efficacy and tolerability of these drugs. In addition, according to recently published data on more than 5000 patients,^4,5,6 when treated with escitalopram oxalate, 10 mg/d, sertraline hydrochloride, 100 mg/d, risperidone, 4 mg/d, or aripiprazole, 20 mg/d, more than one-third of the patients exhibit blood drug levels outside the therapeutic concentration window defined for these drugs.¹ Therefore, although these daily doses fit an average patient well, there is an apparent need to personalize the dose and maximize the treatment response beyond population-based dosing.

Most antipsychotics and antidepressants are metabolized by the polymorphic CYP2C19 and CYP2D6 enzymes,¹ and their capacity is genetically determined.^7,8 First, normal metabolizers (NM category) have normal enzymatic capacity and carry homozygous wild-type (Wt) alleles; they may also carry other genotypes if the enzymatic capacity is not significantly different compared with Wt/Wt carriers. Second, CYP2C19/CYP2D6 genotype-determined poor metabolizers (PM category) carry homozygous loss-of-function alleles and do not possess the active enzyme. Third, CYP2C19/CYP2D6 genotype-determined intermediate metabolizers (IM category) carry genotypes connected with substantially reduced but not abolished enzymatic capacity. Finally, CYP2C19/CYP2D6 genotype-determined ultrarapid metabolizers (UM category) carry genotypes connected with higher-than-normal enzymatic capacity. All these phenotypes are present in substantial proportion worldwide (Table 1).⁹

Table 1. Allele Frequencies of Variant CYP2C19 and CYP2D6 Genes Among Different Populations Worldwide^a.

Genotype-based phenotype by metabolism category	Population, %
	European	African	East Asian	South Asian	Admixed American
CYP2C19
PM	3.3	3.3	14.2	11.8	1.1
IM	21.7	21.2	45.8	35.8	16.0
PM plus IM	25.0	24.6	60.1	47.6	17.1
NM	43.4	42.5	38.1	36.4	62.8
UM	31.6	32.9	1.8	16.0	20.1
CYP2D6
PM	6.2	2.8	0.7	2.1	3.8
IM	2.6	24.5	48.6	10.0	2.6
PM plus IM	8.8	27.3	49.3	12.2	6.4
NM	88.1	64.7	49.6	85.9	91.4
UM	3.2	8.0	1.2	1.9	2.2

Abbreviations: IM, intermediate metabolizer; NM, normal metabolizer; PM, poor metabolizer; UM, ultrarapid metabolizer.

^aData are based on Zhou et al.⁹ Notable variations also exist within the global regions.

Well-replicated clinical findings indicate that the patients in the PM and IM categories exhibit a substantial increase in the exposure and adverse drug reactions of certain psychotropic drugs,^4,5,6,10,11 whereas those in the UM category most often have lower levels of response, owing to faster drug metabolism.^4,5,12,13 In addition, recent studies^4,5 found that those in the PM and UM categories are more prone to risperidone and escitalopram treatment failure, which was quantified as an increase in the incidence of switching to an alternative antipsychotic/antidepressant within 1 year. The recommended and maximum daily doses are originally designed to fit the mean genotype-weighted population. Thus, the official dosing recommendations for psychiatric drugs usually do not acknowledge the clinical relevance of CYP2C19/CYP2D6 metabolizer categories and do not distinguish between them. Investigators^4,5,6 observed, however, that the daily doses of escitalopram, sertraline, risperidone, and aripiprazole, prescribed in naturalistic settings based on clinical observations alone, were lower in individuals in the PM compared with NM categories and that the observed dose reductions were insufficient to fully compensate for the increased drug exposure. In rare cases, as with aripiprazole treatment, relevant sources such as the US Food and Drug Administration, European Medicines Agency, CPIC (Clinical Pharmacogenetics Implementation Consortium), and DPWG (Dutch Pharmacogenetics Working Group) recommend dose reduction for patients in the CYP2D6 PM category; however, these sources offer conflicting information related to the magnitude of dose adjustment. In fact, most of the recommendations are based on underpowered studies, and insufficient data are available to allow the estimation of the difference in drug exposure between metabolizer categories with sufficient precision.¹⁴

Many previous studies, often of limited sample size, have investigated the effects of CYP2C19 and CYP2D6 genotype on the exposure of antipsychotic and antidepressant drugs, and recently published reports substantially increased the number of participants undergoing genotyping.^4,5,6 Thus, the aim of this systematic review and meta-analysis of prospective and retrospective cohort studies was to quantify, with the best attainable precision, the increase of antidepressant and antipsychotic exposure in individuals in CYP2C19/CYP2D6 PM and IM categories compared with those in the NM category. Individuals in the UM category were not included in the analysis owing to the limited number of studies considering this phenotypic group.

Methods

Search Strategy and Selection Criteria

The list of antipsychotic and antidepressant drugs was based on the list of frequently used antidepressants¹⁵ and antipsychotics.¹⁶ The investigated antidepressants included escitalopram, sertraline, fluoxetine hydrochloride, fluvoxamine maleate, paroxetine hydrochloride, venlafaxine hydrochloride, amitriptyline hydrochloride, nortriptyline hydrochloride, mianserin, and mirtazapine; the antipsychotics included clozapine, quetiapine fumarate, olanzapine, risperidone, aripiprazole, and haloperidol lactate. Racemic citalopram hydrobromide was not investigated owing to stereoselective metabolism. The information on which CYP450 isoforms are involved in the metabolism of each drug were retrieved from the recent consensus guidelines.¹ The search was performed in PubMed, ClinicalTrials.gov, Clinicaltrialsregister.eu, International Clinical Trials Registry Platform, and CENTRAL databases for reports published from January 1, 1990, to June 30, 2020. An independent literature survey was performed for each drug and the search terms *NameOfTheDrug* AND CYP2C19 OR CYP2D6 were used. During the initial screening step, all studies that did not deal with drug exposure were excluded, and the remaining studies were considered for inclusion based on the following criteria: (1) participants were genotyped for all known common functional CYP2C19 or CYP2D6 variant alleles with minor allele frequency of greater than 1% according to Zhou et al⁹; (2) adequate classification of participants into CYP2C19 and/or CYP2D6 NM, IM, and PM categories was possible based on genotyping; (3) the study included at least 3 participants per experimental group; and (4) drug exposure was measured in a representative way by (a) dose-normalized steady-state plasma levels, (b) dose-normalized area under the plasma level (time) curve, or (c) apparent total clearance of the drug (reciprocal value). The screening and scanning for eligibility were performed manually by 2 independent investigators (F.M. and N.B.). The decision on study inclusion was made by consensus with a third investigator (M.M.J.), with the final checkup made by consensus among 3 (E.M., M.I.-S., and M.M.J.). Six domains were assessed by using the standardized risk of bias in nonrandomized studies of interventions tool,¹⁷ and studies with the critical risk of bias were excluded. No restrictions were made regarding the study design, participant characteristics (race, ethnicity, sex, age, smoking status, and patient vs healthy cohort), treatment duration, drug interactions, and language.

Data Extraction

The procedures of data acquisition and extraction, as well as the situations when the authors were contacted to provide the data that were inaccessible, are described in full detail in eMethods 1 in the Supplement. If a drug possesses an active metabolite, the drug exposure was calculated by pooling the parent compound and active metabolite (active moiety) exposure.¹ Participants were classified into PM, IM, and NM categories for CYP2C19 and CYP2D6 by using the previously described classification criteria (Table 1).¹⁸ Participants in the PM category were homozygous carriers of the 2 loss-of-function (null) alleles for both CYP2C19 and CYP2D6. For CYP2C19, participants in the IM category carried 1 null and 1 Wt allele, whereas those in the NM category carried the CYP2C19 Wt/Wt genotype. The CYP2D6 gene possesses alleles that reduce but do not abolish the enzymatic capacity (Red), and the CYP2D6 IM category consisted of participants carrying either CYP2D6 Red/Null or CYP2D6 Red/Red genotype, whereas the subpopulation in the CYP2D6 NM category carries 1 or 2 CYP2D6 Wt alleles. For the purpose of this study, only the individuals carrying the CYP2D6 Wt/Wt genotype represented the NM reference group, as suggested by the consensus guidelines.¹⁸

Statistical Analyses

Meta-analyses were performed in accordance with the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines,¹⁹ and the checklist is available in eMethods 2 in the Supplement. Meta-analyses for specific phenotypes/drugs were performed and represented graphically if 3 or more studies met the inclusion criteria. The effect size was the mean exposure of the PM, IM, or PM plus IM groups divided by the mean exposure of the NM group, that is, the ratio of means (RoM).²⁰ For example, an IM:NM group RoM of 1.5 means a 1.5 times higher exposure (ie, a 50% higher exposure in the IM compared with the NM group). Standard mean differences (Hedges g) were also calculated and presented in eFigure 3 in the Supplement. Weighted RoM between subgroups was used in calculation of pooling effect between studies by fixed-effects meta-analysis model. Heterogeneity across the studies was assessed using the Cochran Q test at a given significance level; the percentage of total variability attributable to heterogeneity was quantified by the I² value. A fixed-effects model was used because all the pooled studies represent the same genetic/biological construct; however, owing to considerable heterogeneity in certain analyses, a post hoc sensitivity analysis was performed by using the random-effects model, and the comparison between the fixed- and random-effects model analyses is presented in eTable 1 in the Supplement. Differences between the effect sizes of PM vs NM and IM vs NM groups were examined by using the subgroup test, and when no difference was observed, a post hoc comparison between the pooled PM plus IM and NM experimental groups was performed. For each individual study, the PM plus IM experimental group exposure was calculated by combining the PM and IM subgroups according to the Cochrane handbook formula (section 6.5.2.10 on combining groups).²¹

Small trial or publication bias was evaluated using the Egger test for funnel plot asymmetry,²² and funnel plots are presented in the eFigure 5 in the Supplement. Statistical analyses were performed with RevMan, version 5.4, software (Cochrane). Ratios of means for the individual studies were calculated using Excel, version 2013 (Microsoft Corporation), according to the previously published formula,²⁰ and subsequently entered into RevMan with the generic inverse variance option. Two-sided α < .05 indicated statistical significance.

Interpretation of Changes in Drug Exposure

If a lower boundary of the 95% CI for the drug exposure increase of the PM, IM, or PM plus IM groups compared with the NM group was greater than 1.25-fold, such an effect was considered clinically relevant. If this was not the case for a statistically significant effect, such an effect was considered preliminary or marginal. This quantitative cutoff was based on (1) the US Food and Drug Administration limits for bioequivalence (RoM, 0.80-1.25), which are based on the general consideration that the intraindividual variability in drug exposure from oral drug intake to intake is 20%,²³ and (2) the previous finding that changes of this magnitude are associated with an increased risk of therapeutic failure, measured by the drug switch rates in 2 recent studies^4,5 on patient cohorts treated with escitalopram (n = 2087) and risperidone (n = 890).

Results

Of the 2103 initially screened references, 94 unique studies^{4,5,6,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114} on 8379 unique individuals met the inclusion criteria. Reasons for exclusion are presented in Table 2 and eTable 2 in the Supplement. eFigure 1 in the Supplement gives the PRISMA¹¹⁵ flow diagram. A list of included and excluded studies are presented in eMethods 3 in the Supplement.

Table 2. Overview of Search Process and Studies Incorporated Into Meta-analyses.

Analyzed drug	Enzyme	No. of studies	No. of excluded studies			No. of included studies	No. of individuals by metabolism categorya
			Not dealing with exposure	Incorrect genotyping	No usable data		NMb	IM	PM
Antipsychotics
Aripiprazole	CYP2D6	84	58	4	10	12	814	134	90
Clozapine	CYP2D6	86	78	3	0	5	33	15	4
	CYP2C19						127	65	8
Haloperidol	CYP2D6	109	83	10	3	13	532	158	46
Quetiapine	CYP2D6	45	44	0	0	1	171	0	20
Risperidone	CYP2D6	221	163	9	26	23	1134	186	172
Antidepressants
Amitriptyline	CYP2D6	103	94	1	4	4	43	9	4
	CYP2C19						34	18	6
Escitalopram	CYP2C19	147	135	4	4	4	1110	760	152
Fluoxetine	CYP2D6	313	305	4	1	3	8	0	3
	CYP2C19						71	27	6
Fluvoxamine	CYP2D6	224	212	3	2	7	74	72	0
	CYP2C19						6	6	6
Mirtazapine	CYP2D6	70	59	2	4	5	142	14	19
Nortriptyline	CYP2D6	97	87	3	3	4	28	14	4
Paroxetine	CYP2D6	335	318	8	4	5	89	14	11
Sertraline	CYP2C19	74	68	1	2	3	565	352	40
Venlafaxine	CYP2D6	195	170	5	12	8	509	87	120
	CYP2C19						422	198	21
Total		2103	1874	57	75	94	8379b

Abbreviations: IM, intermediate metabolizer; NM, normal metabolizer; PM, poor metabolizer.

^aThe total number of patients is less than the sum of patients for all phenotypes/drugs owing to the fact that CYP2C19 and CYP2D6 genotyping was performed in certain studies.

^bIndicates reference category.

Association Between CYP2D6 Metabolizer Status and Drug Exposure

The CYP2D6 genotype was associated with significant exposure increases for aripiprazole^{5,25,26,27,28,29,30,31,32,33,115,116} (eFigure 2 in the Supplement) (PM plus IM vs NM RoM, 1.48; 95% CI, 1.41-1.57; 12 studies; 1038 patients), haloperidol^26,34,35,36 (eFigure 2 in the Supplement) (PM vs NM RoM, 1.68; 95% CI, 1.40-2.02; 9 studies; 423 patients), and risperidone^{5,26,35,37,38,39,40,41,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57} (eFigure 2 in the Supplement) (PM plus IM vs NM RoM, 1.36; 95% CI, 1.28-1.44; 23 studies; 1492 patients). Nortriptyline exposure^58,59,60 (RoM, 2.36; 95% CI, 2.10-2.65; 3 studies; 37 patients) (eFigure 2 in the Supplement) and paroxetine exposure^61,62,63 (RoM, 3.50; 95% CI, 2.52-4.85; 3 studies; 41 patients) (eFigure 2 in the Supplement) were significantly increased in the CYP2D6 IM compared with the NM groups; however, after removing the studies associated with serious risk of bias (eResults in the Supplement), these differences were based on fewer than 3 independent studies. It is uncertain whether the exposure increases observed in the fluvoxamine IM group^{64,65,66,67,68,69} and mirtazapine PM group^70,71,72,73 (eFigure 2 in the Supplement) compared with the NM groups are outside the bioequivalence (1.25) limit. Compared with the CYP2D6 NM group, marginal exposure increases were observed in the haloperidol IM group^{74,75,76,77,78,79,80,81,82} (RoM, 1.14; 95% CI, 105-125; 9 studies; 423 patients) (eFigure 2 in the Supplement) and venlafaxine IM plus PM group (RoM, 1.19; 95% CI, 1.09-1.29; 8 studies; 716 patients)^{83,84,85,86,87,88,89,90} (eFigure 2 in the Supplement). Statistically significant exposure increases based on less than 3 independent studies compared with the CYP2D6 NM group were observed in the quetiapine-treated PM (RoM, 1.32; 95% CI, 1.10-1.58; 1 study; 191 patients), amitriptyline-treated IM (RoM, 1.50; 95% CI, 1.23-1.84; 2 studies; 35 patients), mirtazapine-treated IM (RoM, 1.39; 95% CI, 1.23-1.57; 4 studies; 144 patients), paroxetine-treated PM (RoM, 5.13; 95% CI, 3.82-6.87; 2 studies; 73 patients), nortriptyline-treated PM (RoM, 3.32; 95% CI, 2.08-5.29; 1 study; 9 patients), and fluoxetine-treated PM (RoM, 2.26; 95% CI, 1.68-2.83; 1 study; 11 patients) groups (Table 3).

Table 3. Detailed Statistical Report of the Association of Metabolism Status With Antipsychotic and Antidepressant Exposure.

Drug	Enzyme	No. of studies	No. of patients by metabolism group		RoM (95% CI)	P value	I2 value, %
			Reference	Comparator
Antipsychotics
Aripiprazole	CYP2D6	5	693 NM	90 PM	1.51 (1.38-1.65)	<.001	0
	CYP2D6	9	664 NM	134 IM	1.47 (1.38-1.57)	<.001	65
	CYP2D6	12	814 NM	224 PM plus IM	1.48 (1.41-1.56)	<.001	56
Clozapine	CYP2D6	1	22 NM	4 PM	1.00 (0.43-2.32)	>.99	NA
	CYP2D6	2	33 NM	15 IM	1.22 (0.79-1.88)	.51	0
	CYP2C19	2	70 NM	8 PM	1.92 (1.32-2.79)	.008	0
	CYP2C19	4	127 NM	65 IM	1.00 (0.84-1.19)	.84	10
Haloperidol	CYP2D6	4	267 NM	46 PM	1.68 (1.40-1.91)	<.001	21
	CYP2D6	9	265 NM	158 IM	1.14 (1.05-1.25)	.003	0
Quetiapine	CYP2D6	1	171 NM	20 PM	1.32 (1.10-1.58)	<.001	NA
Risperidone	CYP2D6	13	937 NM	172 PM	1.40 (1.30-1.50)	<.001	17
	CYP2D6	11	469 NM	186 IM	1.31 (1.20-1.43)	<.001	44
	CYP2D6	23	1134 NM	358 PM plus IM	1.36 (1.28-1.44)	<.001	34
Antidepressants
Amitriptyline	CYP2D6	1	17 NM	4 PM	1.04 (0.65-1.68)	.86	NA
	CYP2D6	2	26 NM	9 IM	1.50 (1.23-1.84)	<.001	0
	CYP2C19	1	4 NM	6 PM	1.07 (0.81-1.41)	.58	NA
	CYP2C19	1	30 NM	18 IM	1.06 (0.89-1.25)	.50	NA
Escitalopram	CYP2C19	4	1110 NM	152 PM	2.63 (2.40-2.89)	<.001	84
	CYP2C19	4	1110 NM	760 IM	1.38 (1.28-1.48)	<.001	86
Fluvoxamine	CYP2D6	6	74 NM	72 IM	1.52 (1.23-1.89)	<.001	0
	CYP2C19	1	6 NM	6 IM	0.87 (0.31-2.45)	.77	NA
	CYP2C19	1	6 NM	6 PM	0.90 (0.31-2.65)	.84	NA
Fluoxetine	CYP2D6	1	8 NM	3 PM	2.26 (1.68-2.83)	<.001	NA
	CYP2C19	1	4 NM	6 PM	2.94 (2.36-3.67)	<.001	NA
	CYP2C19	2	71 NM	27 IM	1.48 (1.24-1.76)	<.001	13
Mirtazapine	CYP2D6	4	125 NM	19 PM	1.39 (1.23-1.57)	<.001	64
	CYP2D6	1	17 NM	14 IM	1.51 (1.20-1.91)	.010	NA
Nortriptyline	CYP2D6	1	5 NM	4 PM	3.32 (2.08-5.29)	<.001	NA
	CYP2D6	3	23 NM	14 IM	2.36 (2.10-2.65)	<.001	74
Paroxetine	CYP2D6	2	62 NM	11 PM	5.13 (3.82-6.87)	<.001	85
	CYP2D6	3	27 NM	14 IM	3.50 (2.52-4.85)	<.001	0
Sertraline	CYP2C19	3	565 NM	352 IM	1.38 (1.27-1.51)	<.001	0
	CYP2C19	2	537 NM	40 PM	2.70 (2.15-3.39)	<.001	0
Venlafaxine	CYP2D6	6	486 NM	120 PM	1.18 (1.04-1.33)	.01	50
	CYP2D6	3	436 NM	87 IM	1.14 (1.03-1.26)	.009	70
	CYP2D6	8	509 NM	207 PM plus IM	1.19 (1.09-1.29)	<.001	40
	CYP2C19	1	422 NM	21 PM	2.13 (1.54-2.93)	<.001	NA
	CYP2C19	1	422 NM	247 IM	1.19 (1.11-1.31)	<.001	NA

Abbreviations: IM, intermediate metabolizer; NA, not applicable; NM, normal metabolizer; PM, poor metabolizer; RoM, ratio of means.

Association Between CYP2D6 Metabolizer Status and Drug Exposure

The CYP2C19 genotype was associated with significant exposure increases for escitalopram (eFigure 2 in the Supplement) (PM vs NM RoM, 2.63; 95% CI, 2.40-2.89; 4 studies; 1262 patients)^4,91,92,93 and sertraline (eFigure 2 in the Supplement) (IM vs NM RoM, 1.38; 95% CI, 1.27-1.51; 3 studies; 917 patients).^6,94,95 Considerable heterogeneity was observed in the escitalopram meta-analyses, and the elevation in escitalopram exposure in the CYP2C19 IM group was not observed if the random-effect model was used (eFigure 4 in the Supplement). The CYP2C19 IM and NM groups did not exhibit statistically significant difference in clozapine exposure.^96,97,98,99 Statistically significant exposure increases based on less than 3 independent studies compared with the CYP2C19 NM group were observed in the clozapine-treated PM (RoM, 1.92; 95% CI, 1.32-2.79; 2 studies; 78 patients), fluoxetine-treated IM (RoM, 1.48; 95% CI, 1.24-1.76; 2 studies; 98 patients) and PM (RoM, 2.94; 95% CI, 2.36-3.67; 1 study; 10 patients), sertraline-treated PM (RoM, 2.70; 95% CI, 2.15-3.39; 2 studies; 577 patients), and venlafaxine-treated IM (RoM, 1.19; 95% CI, 1.11-1.31; 1 study; 669 patients) and PM (RoM, 2.13; 95% CI, 1.54-2.93; 1 study; 443 patients) groups (Table 3).

Heterogeneity, Small Trial or Publication Bias, and Risk of Bias Assessment

Significant heterogeneity was observed in the aripiprazole IM and IM plus PM, escitalopram PM and IM, mirtazapine PM, nortriptyline IM, and venlafaxine IM group meta-analyses. No small trial or publication bias was observed in the meta-analyses related to risperidone and aripiprazole (eResults in the Supplement), whereas asymmetry could not be assessed in other meta-analyses owing to the insufficient number of included studies (n < 10).

According to the standardized risk of bias in nonrandomized studies of interventions tool, 23 studies were associated with a serious risk of bias,^{24,32,36,38,41,44,46,49,56,58,59,61,62,70,74,79,81,82,86,103,105,107,112} and 71 studies^{4,5,6,25,26,27,28,29,30,31,33,34,35,37,39,40,42,43,45,47,48,50,51,52,53,54,55,57,60,63,64,65,66,67,68,69,71,72,73,75,76,77,78,80,83,84,85,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,104,106,108,109,110,111,113,114} were associated with moderate risk of bias (ie, the analysis is comparable with a well-performed nonrandomized study). The sensitivity analysis results performed for the studies with moderate risk of bias is presented in eTable 3 and eFigure 6 in the Supplement.

Discussion

The results obtained in this systematic review and meta-analysis provide precise quantifications of the differences in antipsychotic and antidepressant drug exposure between patients with PM or IM vs NM CYP2C19/CYP2D6 phenotypes. These results represent scientific foundations for CYP2D6/CYP2C19 genotype-based dosing recommendations, which could lead to improved clinical outcomes in drug treatment of patients with psychiatric disorders.

Although many studies show that CYP2C19 and CYP2D6 PM and IM groups exhibit a significant increase in drug exposure compared with NM groups, the power of these studies was insufficient to quantify these exposure increases with sufficient precision and to evaluate their prospective clinical relevance. The present set of meta-analyses, which incorporates 8379 CYP2C19 and CYP2D6 genotyped individuals with exposure measurements for 16 frequently used psychiatric drugs, allowed (1) validation of whether CYP2C19 and CYP2D6 PM or IM phenotypes significantly increase the drug exposure compared with the NM phenotype, (2) differentiation between marginal changes and clinically relevant drug exposure increases caused by specific phenotypes, and (3) precise estimation of the magnitude of increase in drug exposure for the clinically relevant exposure changes. High precision of clinically relevant estimates is important for the clinical implementation of appropriate dose recommendations for subpopulations defined by CYP2C19 or CYP2D6 genotype.

There is a consensus in the field about the relevance of the CYP2C19 and CYP2D6 polymorphism for interindividual variability in drug metabolism and clinical response,^117,118 and CYP2C19/CYP2D6 genotyping is already included in all currently commercially available pharmacogenetic tests.¹¹⁹ Pharmacogenomic recommendations on drug labels offer a tool by which knowledge of the specific genotype can be translated to the clinical setting in a quantitative manner. However, the dosing recommendations are usually not uniform among the relevant sources,¹⁴ and the dosing recommendations on the US Food and Drug Administration–approved drug labels^{120,121,122,123,124,125} clearly do not comply on many points with the findings summarized herein. The results suggest that there is a need to distinguish between CYP2D6 metabolism categories when deciding on aripiprazole, haloperidol, and risperidone doses and to distinguish between CYP2C19 metabolism categories when deciding on escitalopram and sertraline dose. Furthermore, unlike the PM phenotype, the IM phenotype is seldom considered a relevant factor for drug dosing and treatment, which is noteworthy in relation to results and the fact that more than half of the East Asian population and a considerable amount of other populations have the CYP2C19 or CYP2D6 IM phenotype.⁹

To approach the question of whether preemptive CYP2C19 and CYP2D6 genotyping can improve the drug treatment outcome of patients with psychiatric disorders, one must (1) demonstrate the effect of serum concentration on adverse effects and efficacy and (2) quantify the effect of genotype on serum concentration. The former has been demonstrated by a series of pharmacokinetic, clinical, and positron emission tomography studies^1,126,127 and to an extent by 2 recent meta-analyses on dose-response curves for antidepressants and antipsychotics.^2,3 The present report addresses the latter, because it quantifies the effect of PM and IM CYP2C19/CYP2D6 phenotypes on blood levels. Therapeutic drug monitoring can be used as a tool in personalized dosing because it directly measures drug blood levels and encompasses all sources of variability in drug exposure, including CYP2D6/CYP2C19 genotype. However, therapeutic drug monitoring becomes applicable only when the drug level reaches a steady state and is therefore not a suitable tool for preventing the suboptimal response or adverse effects during the initial weeks, or sometimes months, of psychiatric drug treatment. This period is critical for rapid symptom control, patients’ treatment belief, and adherence; in a therapeutic field characterized by a substantial degree of trial and error, preemptive genotyping has a potential to improve dose personalization and subsequently the drug treatment outcome as well. Overall, the optimal dose stabilization would be obtained in an ideal clinical situation, in which a psychiatrist would know the patients’ CYP2D6/CYP2C19 genotype before the drug treatment initiation to make the best possible initial dosing decisions. These decisions can be checked by therapeutic drug monitoring after the steady state is achieved. However, although several industry-sponsored clinical trials^128,129,130 advocate the advantage of genotype-guided over usual treatment in psychiatry, a well-designed trial is still necessary to validate and quantify the clinical utility of preemptive CYP2C19/CYP2D6 genotyping.

Limitations

The most important limitation of this report is the potential presence of confounding factors, which arise from the nature of the studies incorporated into meta-analyses. Most of the studies were performed in naturalistic settings, and the factors that are known to affect drug metabolism are seldom completely controlled for. Next, the inclusion and exclusion criteria were designed in a way to eliminate the possibility of erroneous classification into metabolism categories, and this revealed the apparent scarcity of representative studies for many gene-drug interactions. In addition, approximately one-third of the studies that dealt with drug exposure did not measure exposure representatively, and the data were therefore not usable. Although CYP2C19/CYP2D6 UM status may also affect the exposure of certain drugs, and although the CYP2C19 and CYP2D6 PM and/or IM status significantly affect drug exposures of most of the analyzed drugs, more studies and larger cohorts are needed to ascertain the relevance of many gene-drug interactions (eFigure 8 in the Supplement). Also, in some cases, the number of usable studies was relatively low and heterogeneity was considerable; the most notable example is the analysis of CYP2C19-escitalopram interaction with 4 representative studies for each comparison^4,91,92,93 and I² > 80%. Although the directionality of the effect is apparent, more representative studies on this interaction are needed to precisely quantify the effect size of the exposure increase. Next, it was possible to address the presence of small trial or publication bias only for several comparisons owing to the small number of studies (n < 10) for many gene-drug interactions. Although the test result was negative for the analyzed comparisons, we cannot exclude the possibility that the publication bias is present in some of the gene-drug interaction comparisons to a degree. Finally, we were able to compare the effect of ethnicity in several comparisons by the subgroup test only, and these post hoc tests are presented in the eFigure 7 in the Supplement. Although these test results were negative, we cannot completely exclude the possibility that the exposure increases of certain drugs may be ethnicity dependent to a degree.

Conclusions

In this systematic review and meta-analysis, the association between CYP2C19/CYP2D6 genotype and drug levels of aripiprazole, haloperidol, risperidone, escitalopram, and sertraline was quantified with sufficient precision as to be useful as a scientific foundation for CYP2D6/CYP2C19 genotype-based dosing recommendations. In addition, there was an indication that the CYP2C19/CYP2D6 genotype is associated with changes in drug levels of clozapine, quetiapine, amitriptyline, fluvoxamine, fluoxetine, mirtazapine, nortriptyline, paroxetine, and venlafaxine. However, more representative studies focused on these specific gene-drug associations are necessary for an adequate quantification of the magnitude of drug level changes and for representative evaluation of the relevance of these changes.

Supplement.

eMethods 1. List of Abbreviations and Procedures of Data Extraction

eMethods 2. MOOSE Checklist

eMethods 3. List of Included and Excluded Studies

eFigure 1. PRISMA Flow Diagram

eFigure 2. Original Meta-analysis Output: Ratio of Means

eFigure 3. Original Meta-analysis Output: Standardized Mean Difference

eFigure 4. Random-Effects Model

eFigure 5. Assessment of Asymmetry and Funnel Plots

eFigure 6. Assessment After the Exclusion of the Studies Associated With Serious Risk of Bias

eFigure 7. Subgroup Test: European vs East Asian Ethnicity

eFigure 8. Schematic Representation of Phenotype/Drug Interaction Classification

eTable 1. Fixed- vs Random-Effects Model Comparison Sensitivity Analysis

eTable 2. Study Characteristics

eTable 3. Sensitivity Analysis of Studies With Moderate Risk of Bias

eResults. Risk of Bias Assessment