Pharmacogenetics of Antidepressants, Mood Stabilizers, and Antipsychotics in Diverse Human Populations

Abstract

An increasing focus on personalized medicine is driving a renewed effort to understand the impact of ethnic and genetic background on treatment outcomes. Since responses to psychopharmacological treatments continue to be sub-optimal, there is a pressing need to identify markers of tolerability and efficacy. Pharmacogenomic studies aim to find such markers within the human genome, and have made some progress in recent years. Progress has been slower in populations with diverse racial and ethnic backgrounds.

Here we review 10 genome-wide association studies (GWAS) that assessed outcomes after antidepressant, antipsychotic, or mood stabilizer treatment. These studies used samples collected by the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE), Sequenced Treatment Alternatives to Relieve Depression (STAR*D), and Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) studies. We highlight findings from African American and European American participants since they are the largest groups studied, but we also address issues related to Asian and Hispanic groups.

None of the GWAS we reviewed identified individual genetic markers at genome-wide significance, probably due to limited sample sizes. However, all the studies found poorer outcomes among African American participants. Some of this disparity seems to be explained by psychosocial and economic disadvantages, but at least 2 studies found that widespread genetic differences between participants of European and African ancestry also play an important role. Non-European groups are underrepresented in these studies, but the differences that are evident so far suggest that poorer outcomes among African Americans are not inevitable and may be particularly suited to pharmacogenomic strategies. The vision of more personalized psychopharmacology may critically depend on larger studies in more diverse human populations.

Introduction

The prediction of response to treatment remains a crucial but daunting task for clinicians, particularly for those treating ethnically diverse patients. Outcomes for psychiatric illnesses, which are generally treated with psychotropic medications, are especially sub-optimal. Only one in three depressed patients are fully remitted after an initial course of antidepressant treatment (Rush, Trivedi et al. 2006; Hennings, Owashi et al. 2009); close to 50% of patients with bipolar disorder relapse within two years of treatment (Perlis, Ostacher et al. 2006; Perlis, Smoller et al. 2009); and the outcomes of antipsychotic drug treatments are complicated by sometimes serious neurologic and metabolic adverse events (Lieberman, Stroup et al. 2005; Meyer 2007).

Treatment outcomes are even more problematic for patients of diverse racial and ethnic backgrounds. For example, in recent large-scale treatment studies of major depression and bipolar disorder, African American (AA) participants showed poorer outcomes than others by multiple measures (Lesser, Castro et al. 2007; Gonzalez, Bowden et al. 2010). While poor treatment outcomes in non-Hispanic whites are more directly attributable to lack of efficacy or undesirable side-effects, for AA, non-while Hispanic, and other non-white groups, poor treatment outcomes are often accompanied by greater psycho-social adversity and comorbidity, lower adherence and compliance, and higher discontinuation (dropout) rates (Lesser, Castro et al. 2007; Gonzalez, Bowden et al. 2010; Arnold, Miller et al. 2013; Murphy, Kassem et al. 2013). This complicates pharmacogenetic studies that attempt to find genetic markers of drug efficacy or tolerability. These difficulties are exacerbated by lower rates of treatment-seeking and clinical trial participation among non-whites, particularly those with psychiatric illness.

Two decades ago, Lin et al (1993), lamented that psychopharmacology had neglected the importance of racial and ethnic diversity, despite increasing evidence that findings based solely on samples of white European ancestry may not generalize to other groups (Lawson 1986). In the last two decades, a growing focus on personalized medicine has led to a renewed interested in individual patient characteristics such as ancestry, race, and ethnicity. For example, one of the more recent editions,Psychopharmacology: the Fourth Generation of Progress (Bloom and Kupfer 1995), an authoritative compendium on neuropsychopharmacology that is published every 5–10 years, for the first time included a separate section on ethnicity.

Some definitions are in order. In this review, we employ the word “race” to refer to outwardly manifest, inherited traits such as skin color frequently used to distinguish between groups, e.g., “black” or “white.” “Ethnicity” is often used interchangeably with “race,” but here we use it in its more precise meaning of a purely social construct based on shared aspects of culture, language, religion, etc. (Weber 1978). By this usage, “Hispanic” is an ethnicity, but “African-American” can be either a race or an ethnicity, depending on the context. We use “ancestry” to refer to continental origin of earlier generations, with an emphasis on genetic diversity and persistent differences in allele frequencies (Risch 2006). We recognize that race, ethnicity, and ancestry are often correlated with each other and with a wide range of environmental, lifestyle, and behavioral factors including dietary habits, social privileges, attitudes toward medical care, and biological or genetic factors (Strickland, Stein et al. 1997; Chaudhry, Neelam et al. 2008) -- all of which may influence drug metabolism and efficacy. This is why the interpretation of an apparently “genetic” finding in a diverse sample may require several intermediate steps (and additional studies) before firm conclusions can be drawn about the role of genes in the trait of interest.

Fortunately, modern advances in molecular biology and statistical genetics allow the independent and joint effects of genetic ancestry to be at least partially dissociated from those of race and ethnicity, provided appropriate study designs are employed. Genome-wide SNP arrays and methods that allow for statistical estimation of genetic ancestry (e.g., through principal components analysis and multidimensional scaling) allow non-genetic and sociological contributions can be sorted out more easily than before (Price, Patterson et al. 2006; Purcell, Neale et al. 2007). Still, even the best studies can only provide estimates of genetic ancestry, and the ability to discern fine-scale patterns of ancestry in admixed populations remains quite limited.

This review focuses primarily on three large treatment studies of psychiatric disorders: the Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) (Stroup, McEvoy et al. 2003; Swartz, Perkins et al. 2003), the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study (Rush, Trivedi et al. 2006), and the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD) study (Sachs, Thase et al. 2003). We highlight some of the major findings regarding differences in treatment outcomes across racial, ethnic, or ancestry groups. All of the studies we reviewed were published in the National Human Genome Research Institute (NHGRI)’s GWAS catalog between 2008 and July 15, 2013 (Hindorff, MacArthur et al.; Hindorff, Sethupathy et al. 2009). Our additional criteria for review included total sample sizes of 500 or greater and systematic, prospective assessment of outcomes (thereby reducing the likelihood of recall bias). We focus on African and European Americans, which are the largest groups, and discuss self-identified Hispanic groups in light of the strong genetic admixture characteristic of Hispanics in North America None of the studies contained enough Asian participants for meaningful analyses.

This paper is limited in scope and not meant to be an exhaustive review. Such reviews can be found in a wide range of publications including: Pharmacogenetics of antidepressants (Aitchison, 2011; Porcelli, Drago et al. 2011); mood stabilizers (Squassina, Manchia et al., 2010), antipsychotics, (Reynolds, 2012) and other psychotropic drugs (Malhotra, Murphy et al, 2004; Tiwaki, Souza, et al, 2009).

Ethnicity, Pharmacokinetics and Pharmacodynamics of drug response

The field of pharmacogenetics arose largely out of the recognition that variation in drug response, efficacy, and side-effects profiles is the result of pharmacokinetic and pharmacodynamic processes, which in are influenced in part by genetic mechanisms (Kalow 2006). Cytochrome (CYP) p-450 is the most important enzyme system of phase I metabolism. The isoenzymes CYP2D6 (debrisoquine hydroxylase), and CYP2C19 (mephenytoin hydroxylase) are particularly associated with responses to psychotropic medications like antidepressants, benzodiazepines and antipsychotics.

Genetic polymorphisms in the CYP2D6 and CYP2C19 enzymes produce drug metabolism phenotypes that can be divided into “extensive/normal” or “poor” metabolizers, with further demarcations (e.g., “ultra-rapid” metabolizers) within each category (Lin, Poland et al. 1993; Kalow 2006; Chaudhry, Neelam et al. 2008). Slow metabolism of a drug can sometimes lead to greater therapeutic benefits but risks greater toxicity and side effects. Fast metabolism can produce fewer side effects, but lower drug efficacy. This is a form of pharmacokinetic variation.

Pharmacodynamic variation refers to individual differences at the target of drug action. The primary genetic influences on the pharmacodynamics of tricylic (TCA) and selective serotonin re-uptake inhibitor (SSRI) antidepressants are the genes which encode the serotonin transporter (SLC6A4), and the serotonin receptors ((HTR1A & HTR2A). The SLC6A4 contains a known polymorphism, the 5-HTT-LPR, which may moderate the efficacy of antidepressant drugs. Similarly, the 5-HT (1A and 2A) receptors have been shown to influence susceptibility to developing major depression, and moderate the response to SSRI treatment (McMahon, Buervenich et al. 2006; Kishi, Yoshimura et al. 2009).

The CYP metabolizer phenotypes and the genetic variants that underlie them vary across human populations (Kishi, Yoshimura et al., 2009; Lotrich, Pollack et al., 2003; McMahon, Buevernich, et al.,. 2006). However, the genotype-phenotype associations vary with race and ancestry, complicating conclusions from research findings. Several kinds of ancestry-specific differences have been found:

First, clinical differences in side effects and efficacy sometimes differ by ancestry. For example, the well-known association between the “short”(s)-allele of the 5-HTT-LPR gene with poorer antidepressant response in people of European ancestry is apparently null or even reversed in some Asian groups (Serretti, Kato et al. 2007).

Second, there is substantial within group variability, particularly among people of African descent. Early studies showed that the frequency of the poor metabolizer phenotype associated with CYP2D6 showed a wide range among groups of black African ancestry, from a low of 0–8% in sub-Saharan Africans, to 1.9% in AA, to 19% in San Bushman (Eichelbaum and Woolhouse 1985; Iyun, M.S. et al. 1986; Sommers, Moncrieff et al. 1988; Relling, Cherrie et al. 1991; Sommers, Moncrieff et al. 1991; Lin, Poland et al. 1993). This contrasts with a narrower, but still variable, frequency distribution of the poor metabolizer phenotype, ranging from 3–10% among white people in Europe and North America (Lin, Poland et al. 1993).

Third, drug outcomes may vary across groups, even when pharmacogenetic effects are similar or nil. Early studies on response to lithium salts in the treatment of bipolar disorder found that AA had higher blood concentration of lithium relative to EA. Although there were no racial differences reported in lithium-related toxicity or efficacy, there was some concern that the significantly higher rates of hypertension and renal disease among AA might call for closer monitoring (Lin et al., 1993; Gibbons & Gibbons, 1982).

Antipsychotics: The CATIE Study

Approximately 1400–1500 patients meeting criteria for schizophrenia and spanning over 50 clinical sites across the United States were recruited to participate in the multi-phase randomized CATIE trial. The CATIE study’s main objective as stated by Stroup et al (2003) was “to evaluate the effectiveness of antipsychotic drugs in typical settings and populations so that the study results will be maximally useful in routine clinical situations.” CATIE’s methods and protocol are described extensively elsewhere (Stroup, McEvoy et al. 2003; Swartz, Perkins et al. 2003; Lieberman, Stroup et al. 2005; Sullivan, Lin et al. 2008). In the original sample, about 60% reported that they were white, 35% black, and 5% other, with about 12% of the sample reporting Hispanic ethnicity (Lieberman, Stroup et al. 2005). The trial consisted of two primary phases (I and II) and an optional phase III for those who had not been successfully treated in the earlier phases. In phase I which was double-blind and randomized, four atypical (second-generation) antipsychotics were compared with a first-generation antipsychotic, perphenazine. Discontinuation of treatment (for any reason) was the main outcome of interest. In phase II, participants who were not successfully treated with their initial medication could opt to be randomized to receive a different atypical antipsychotic.

Group differences in outcome by race and ancestry

AA had the highest rate of discontinuation after phase I, relative to non-Hispanic whites, and Hispanics (Arnold, Miller et al. 2013) although they reported fewer side effects than whites or Hispanics, and fewer problems than whites with drug efficacy. The putative reasons for discontinuation among AA were non-specific (e.g., research burden, loss to follow-up) and not related to drug efficacy or side effects. Hispanics reported discontinuation due to side effects of pherphenazine and ziprasidone. There were no group differences in the drug dose levels in phase I for most drugs, but non-Hispanic whites received a higher average dose of risperdone than AA, who in turn received higher average dose than Hispanics.

Genome wide association studies of antipsychotic response

There were five GWA studies from the CATIE trial which met our criteria for inclusion in this review (see Table 1). All of these studies were based on a subsample of 738 patients who consented to give DNA samples. In this subsample, 57% reported their race as white or EA, 29% as Black or AA, and 14% as multiple races or other race. Two of the studies examined the associations between genetic variants and side effects; the other three examined the associations for drug efficacy..

Table 1.

Pharmacogenetic Genome-Wide Association Studies(GWAS) from CATIE (2008–2013).

Author & Date	Sample size Ethnic/Racial Composition	Phenotype Studied	GWAS-level significant variants	Ethnic/Racial Group differences
Aberg et al. (2012)	N=738; 57% EUR, 29% AA, 14% other	Antipsychotic-induced QTc interval prolongation for 5 atypical antipsychotics	rs4959235 on gene SLC22A23 mediated effect of Quetiapine	Similar but stronger effect among EUR than AA
Adkins et al. (2011)	N = 738; 57% EUR 29% AA, 14% other	Metabolic side effects
		Weight Gain	8 SNPS including top GWAS hit rs1568679 located in intron of MEIS2 gene	No differences between AA & EUR
		Lipids (triglycerides, HDL and total cholesterol)	10 SNPS mediating the effects of pherpenazine and clozapine on triglycerides and HDL cholesterol	Differences noted for 3 SNPS:
				i) rs17651157 and ii) rs10502661 located in an intron of FHOD3 gene (mediating effect of pherpenazine on triglycerides); low MAF for both AA & EUR; but haplotype testing improved in EUR only
				iii) rs13224682 located in an intron of PRKAR2B gene mediated effects of clozapine on triglycerides in EUR but not in AA
		Blood Glucose (A1c)	Top GWAS hits rs1967256 and rs11954387 located on intron of GPR98 gene.	No differences between AA & EUR
		Blood Pressure	None found	N/A
		Heart Rate	None found	N/A
Clarke et al. (2013)	N=738; 57% EUR, 29% AA, 14% Other	Illness Severity (Drug Efficacy)
		Clinical global impression severity	6SNPS	No differences between AA & EUR
		Patient global impression severity	7SNPS	No differences between AA & EUR for 3 SNPS on PDE4D gene mediating effects of Quetiapine.
				3 SNPS on TJPI gene mediating effect of risperdone mainly driven by EUR subsample
McClay et al. (2011a)	N= 738; 57% EUR, 29% AA, 14% other	Neurocognition in five domains
		Processing speed	rs11240594 at SLC26A9 mediating the effects of olanzapine	No differences between AA & EUR
		Verbal memory
		Vigilance	rs286913 at the EHF gene mediating effects of ziprasidone	No differences between AA & EUR
		Reasoning	None found
		Working memory	4 SNPS	Differences noted for 1 SNP:
				rs7520258 at GPR137B – MAF in EUR < .001 so signal is likely driven by AA and/or “other” groups
McClay et al. (2011b)	N= 738; 57% EUR, 29% 14% other	Response to antipsychotics
		Positive symptoms	rs17390445 on Chr 4p 15 mediating ziprasidone	Robust in EUR only
		Negative symptoms	None found	N/A

Note. GWAS data obtained from the GWAS catalogue of the National Human Genome Research Institute (NHGRI) of the National Institutes of Health (www.genome.gov/gwastudies). CATIE= Clinical Antipsychotics Trial of Intervention Effectiveness. AA= African American; EUR=European American; SNP = Single nucleotide polymorphism. Only studies with total sample sizes of 500 or more with prospectively assessed outcomes are included in this table.

The first published study on side effects (Adkins, Aberg et al. 2011) examined the effect of antipsychotics on metabolic indices such as weight gain, blood glucose, blood lipids, and blood pressure, and found eight variants/single nucleotide polymorphisms (SNPS) that met genome-wide significance. These SNPs mediated the effects of risperdone and ziprasidone on weight gain by incrEAing hip and waist circumference. There were no reported differences between AA and EA for the SNPs associated with weight gain, which included the top SNP, rs1568679 located in an intron of the Meis homeobox 2 gene (MEIS2). Other top SNPs mediated the effects of olanzapine on blood glucose levels, but no racial group differences were found for this association. This study also found 10 SNPs mediating the effects of with blood lipids. Three of the SNPs significantly associated with blood lipids were shown to be driven primarily by the EA sample, and not the African-American sample.

The second study on side effects examined QT-interval prolongation, which has been linked to cardiac arrhythmia and in rare cases, sudden death (Aberg, Adkins et al. 2012). The authors found one genome-wide significant SNP (rs4959235) near the gene SLC22A23, that seemed to mediate side effects of Quetiapine. Although the direction of the effect was similar for both racial groups, the association was stronger among EA.

One study on drug efficacy focused on indices of neurocognition (McClay, Adkins et al. 2011). A significant association was found for processing speed (rs11240594 at SLC26A9), mediating the effects of olanzapine. Another hit was found for rs286913 at the EHF gene, mediating effects of ziprasidone. No significant group differences were found for these two SNPs. Four other hits were found at or near the genes DRD2, GPR137B, CHST8, and IL1A, with appeared to mediate the effects of olanzapine in working memory. A SNP located near GPR137B was shown to have a very low frequency in the EA subsample, suggesting that the signal likely came from AA and “other” groups.

The most recently published study of drug efficacy examined patient and clinician-rated assessments of global severity of illness (Clark, Souza et al. 2013). Six SNPs were significant for clinician-rated severity, while seven SNPs were significant for patient-related severity. No group differences were found for clinician-rated severity assessments, but patent-rated assessments pointed to a group of three SNPs located near TJP1. These SNPs appear to mediate the effect of risperdone on patient-rated severity; they showed a stronger effect and greater significance for the European sample, both individually and in haplotype tests.

Antidepressant medication and the STAR*D Study

Like the CATIE study, STAR*D was a multiphase treatment trial designed to test treatment algorithms for depression that mimicked “real-world” clinical practice among about 40 primary and specialty care sites across the United States (Rush, Trivedi et al. 2006; Trivedi, Rush et al. 2006). Over 4000 patients with non-psychotic major depressive disorder enrolled. In the baseline sample (n=3,671) available for analysis, 76% described themselves as white, 17% as black/AA, and 7% as other races; 12% reported Hispanic ethnicity. STAR*D consisted of four sequenced treatment levels; all patients were treated with an SSRI (citalopram) in level I. Patients who did not remit in level 1 moved to level II where treatment options included a combination of medication and/or cognitive psychotherapy. The primary outcome was severity of depression as measured by the Quick Inventory of Depressive Symptoms (QIDS).

Group differences in outcome

AA had lower response and remission rates than non-Hispanic whites, and there was a continued trend of poorer response among AA relative to whites even after controlling for baseline clinical, social, and economic factors (Lesser, Castro et al. 2007). AA also had higher dropout rates particularly during phases 1 and II (Warden, Trivedi et al. 2007; Murphy, Kassem et al. 2013). A candidate gene study also found that AA had a substantially lower prevalence of a response-associated allele near the serotonin 2A receptor (HTR2A) (McMahon, Buervenich et al. 2006). Hispanics had intermediate outcomes, worse than whites but better than AA (Lesser, Castro et al. 2007).

Genome wide association studies of antidepressant response

Four GWA studies from STAR*D met our criteria for review. These studies were based on a subsample originally comprising approximately 2,000 patients who provided DNA samples (Laje, Perlis et al. 2009; Garriock, Kraft et al. 2010). A summary of the reviewed studies is provided in Table 2.

Table 2.

Genome-Wide Pharmacogenetic Studies from STAR*D (2008–2013).

Author & Date	Sample Ethnic/Racial Composition	Phenotype Studied	GWAS-level significant variants	Ethnic/Racial Group differences
Adkins et al. (2012)	N= 1,762; 79% EUR, 15% AA, 6% other	Citalopram-induced side effects
		Overall side effect burden	Haplotype ~30KB spanning 8 SNPS	No differences between AA and EUR
		Overall tolerability		N/A
		Sexual side effects		N/A
		Dizziness		N/A
		vision/hearing side effects	rs17135437 on EMID2	Minor allele frequency somewhat higher in AA than EUR but magnitude & direction of effect similar across both groups
Clark et al. (2012)	N=1,439	Side effects among pts who did not respond to citalopram at level 1
		Overall side effect burden	5 SNPS	No differences between AA and EUR
		Sexual side effects	11 SNPS	No differences between AA and EUR
		Dizziness	15 SNPS	N/A
		Vision/hearing side effects	2 SNPS	N/a
Garriock, et al. (2010)	N=1,491; 72% EUR, 16% AA, white Hispanics 12%	Response to & remission with citalopram
		Response (≥ 50% score reduction on QIDS-SR)	None	N/A
		Remission (≤ 5 on QIDS-SR at follow-up)	None	N/A
Hunter, et al. (2013)	N=1,116; 69% non-Hisp EUR, 11% white Hispanics, 13% non-Hisp AA, 0.4% black Hispanics, 1.4% Asians, and 5% other	Sustained (vs. unsustained) response (QIDS-C) to Citalopram	None	N/A

Note. GWAS data obtained from the GWAS catalogue of the National Human Genome Research Institute (NHGRI) of the National Institutes of Health (www.genome.gov/gwastudies). STAR*D = Sequenced Treatment Alternatives to Relieve Depression. AA= African American; EUR=European American; SNP = Single nucleotide polymorphism. Only studies with total sample sizes of 500 or more with prospectively assessed outcomes are included in this table.

The two studies examining drug efficacy found no significant GWAS hits for response or remission (Garriock, Kraft et al. 2010; Hunter, Leuchter et al. 2013). The first study on side effects found a significant variant on the EMID2 gene, mediating the effect of citalopram on vision and hearing loss. However, the minor allele frequency for this variant was low in the overall sample, although somewhat higher for AA. The magnitude and direction of the effect were similar for AA and EA. The study also found a significant haplotype spanning approximately 30KB and containing eight SNPS, significantly mediating the effect of citalopram on overall side effect burden. There were no racial group differences for the haplotype. A follow-up study examined side effects of alternate antidepressants including bupropion, buspirone, sertraline, and venlafaxine, for patients who had not responded to citalopram and had moved on to level II (Clark, Adkins et al. 2012). They found non-genome-wide significant evidence of association with 34 variants, spanning 15 genes, a cluster of which mediated the effects of buproprion on sexual side effects (10 SNPs located in the SACM1L gene). No racial group differences were found for any of these SNPs.

Recently, we considered the STAR*D data from the perspective of genetic ancestry, rather than race. We found that poorer outcomes among AA participants are driven by differences in baseline clinical and psychosocial factors, but that a significant portion of the difference that remains after correcting for these factors is explained by genetic ancestry rather than race (Figure). This suggests that among the numerous differences in allele frequencies that differentiate people of African origin from those of European origin, there are many genes that influence treatment outcomes. This also underscores the need to increase AA representation in clinical trials.

Figure 1. African Ancestry and Response to Antidepressant Treatment.

Note.

STAR*D data for individuals who were not African-American or European-American race by self-report. Bars represent proportion of African Ancestry in this subsample divided into quartiles. Fourth quartile contains individuals with 13.7% or more African ancestry.

Figures represent percent change (reduction) in depressive symptom scores from baseline to exit (e.g., 0.5 = 50% reduction in depressive symptoms). African Ancestry Estimates obtained from STRUCTURE analysis. For QIDS-C, n = 85; for QIDS-SR, n=82.

Mood Stabilizers and the STEP-BD study

The design of STEP-BD differed from those of CATIE and STAR*D in that STEP-BD involved elements of naturalistic longitudinal observation along with controlled clinical trial elements. The study was designed to assess short and long-term outcomes for patients 15 years of age and older who had bipolar disorders (Sachs, Thase et al. 2003; Perlis, Ostacher et al. 2006). Over 4,000 individuals participated in the study; about 2,000 provided DNA samples. While the initial baseline symptomatic sample had about 1,500 participants, various subsamples were used in different studies.

Ethnic group differences in outcome

A study using a STEP-BD subsample consisting of AA (n=155), Hispanics (n=152) and non-Hispanic whites (n=729) compared the three groups on response to treatment for Bipolar I and II (Gonzalez, Bowden et al. 2010). For non-psychotic manic and depressive symptoms, time to response and proportion of responders were similar across these groups, but a smaller proportion of AA reported improved global functioning. AA with psychotic symptoms were also slower to recover from depression than AA without psychotic symptoms and non-Hispanics whites with or without psychotic symptoms. There were no differences observed between Hispanics and non-Hispanic whites on time to response or proportion of responders.

Genome-wide association study of response to mood stabilizers

The only published GWAS study on drug response in STEP-BD examined a cohort of 1,177 bipolar I and bipolar II patients who took Lithium salts (Perlis, Smoller et al. 2009). The outcome of interest was recurrence of bipolar symptoms within a year of treatment. No variants met genome-wide significance.

Discussion and Conclusions

This paper reviews recent findings from GWA studies on pharmacogenetics of antipsychotics, antidepressants and mood stabilizers, focusing on racial and ethnic group differences in side effects and efficacy. We also included data from Hispanics (where available). Overall, the pharmacogenetic studies suggest that the genetic associations found so far are generally similar in African- and European-Americans. However, sample sizes have been much smaller for AA groups, raising the risk of false-negative conclusions. Moreover, large differences in clinical outcomes for EA and AA groups have not been fully explained. Two recent papers suggest that a small but significant part of these differences can be explained by differences in genetic ancestry (Murphy et al in press; Adkins, Souza et al. 2013). This underscores the fact that ancestry, race, and ethnicity are not interchangeable and that race is often too imprecise a descriptor for clinical studies.

Another area of concern is the higher discontinuation or dropout rates seen among AA participants in all of these studies. People who drop out of treatment trials may do so on account of greater side effects, reduced efficacy, poor adherence, or other unrelated reasons (Murphy et al 2013). Accordingly, the act of dropping out may itself be subject to genetic influences that are greater in particular racial and ethnic groups. These findings underscore the need for careful consideration of dropout and adherence issues when comparing AA with EA in pharmacogenetic studies.

Increased side effects can reflect broad ancestry-related differences in susceptibility and resilience. For example, the newer atypical antipsychotics are less likely to result in extrapyramidal side effects, but more likely to produce metabolic side effects such as increases in body weight, blood pressure, glucose, cholesterol, and triglyceride levels (CITE). One of the top GWAS findings in the CATIE study concerned glucose levels during olanzapine treatment. For some groups like AA, Hispanics and Native Americans who are especially prone to Type II diabetes, these side effects can be especially worrisome. Even when short term studies find no differences, the long-term consequences of antipsychotic use might be very different for different racial and ethnic groups. A genetically-informed ancestry analysis might help clarify the biological basis for such group-wise differences in long-term outcomes.

This review did not yield much pharmacogenetic information to explain group differences in responses to treatment for bipolar disorder. Recent reviews on the pharmacogenetics of mood stabilizers have not highlighted racial or ethnic group differences (Squassina, Manchia et al. 2010), most likely to due lack of published studies. The first and only GWAS study of bipolar disorder in both AA and EA found that the genes that were significantly associated with bipolar disorder for EA differed from those which were significantly associated with the disorder in AA (Smith, Bloss et al. 2009), but sample sizes were too small to make firm conclusions. STEP-BD did find that AA were more likely to suffer psychotic symptoms, and a few studies have identified genetic variants that may confer vulnerability to psychosis in bipolar disorder (Belmonte 2011; Goes, Hamshere et al. 2012). However, it is not known whether these or other genes could contribute to the greater rate of psychosis among AA with bipolar disorder.

We note some important limitations of the published work. First, the studies we reviewed did not measure blood plasma levels of medication, and relied chiefly on clinician and patient reports of adherence and symptoms. Inconsistent compliance could seriously dilute some of the effects, particularly in groups where compliance might be an issue. Second, although AA were represented in the studies we highlighted (albeit in reduced sample sizes), other groups (e.g., Asian Americans and Native Americans) were strongly underrepresented. This is particularly important in light of recent recognition of highly predictive genetic markers of serious adverse events in Asians treated with carbamezepine (Chen, Lin et al. 2011).

GWA studies on pharmacogenetics of psychotropic medication can be of considerable utility when assessing participants of diverse race, ethnicity, and ancestry. Improving tolerability and efficacy of psychotropic medication for patients from diverse backgrounds is an important public health goal. As recent studies show, the findings from modern pharmacogenetic studies can yield useful information on diverse groups when broadly considered with other clinical outcome studies, but more work is needed. As more large-scale studies are implemented, progress may depend critically on the inclusion of more diverse samples, with their greater range of genetic variation, outcome phenotypes, and better generalizability to the clinic.

Table 3.

Genome-Wide Pharmacogenetic Studies from STEP-BD (2008–2013).

Author & Date	Sample Ethnic/Racial Composition	Phenotype Studied	GWAS-level significant variants	Ethnic/Racial Group differences
Perlis, et al. (2009)	N=1,177	Recurrence of mood in after treatment lithium	None found	N/A

Note. GWAS data obtained from the GWAS catalogue of the National Human Genome Research Institute (NHGRI) of the National Institutes of Health (www.genome.gov/gwastudies). STEP-BD = Systematic Treatment Enhancement Program for Bipolar Disorder. Only studies with total sample sizes of 500 or more with prospectively assessed outcomes are included in this table.