Pharmacogenetics of antipsychotic-induced side effects.

Currently available antipsychotic drugs (APDs) carry significant though highly variable, liability to neurologic and metabolic side effects. Pharmacogenetics approaches offer the possibility of identifying patient-specific biomarkers for predicting risk of these side effects. To date, a few single nucleotide polymorphisms (SNPs) in a handful of genes have received convergent support across multiple studies. The primary focus has been on SNPs in dopamine and serotonin receptor genes: persuasive meta-analytic evidence exists for an effect of the dopamine D2 and D3 receptor genes (DRD2 and DRD3) in risk for tardive dyskinesia (TD) and for an effect of variation at the 5-HT2C receptor gene (HTR2C) for liability to APD-induced weight gain. However, effect sizes appear to be modest, and pharmacoeconomic considerations have not been sufficiently studied, thereby limiting clinical applicability at this time. Effects of these genes and others on risk for TD, extrapyramidal side effects, hyperprolactinemia, and weight gain are reviewed in this article.

Background

Schizophrenia (SCZ) is a disease with an estimated lifetime morbid risk approaching 1% worldwide,[1] and its public health consequences (mortality- and morbidity) are severe. SCZ is associated with an increase of at least 50% in mortality rates compared with the general population,[2] including a suicide rate of approximately- 5%,[3] resulting in 10-year average lifespan reduction;[4] SCZ accounts for nearly 3% of all years lived with disability[5]; amongst individuals aged 15 to 44, SCZ is the third-leading cause of disability.[6]

Despite the demonstrated efficacy of antipsychotic drugs (APDs) in short-term placebo-controlled clinical trials, long-term outcomes frequently remain unsatisfactory. The largest NIH-supported clinical trial of antipsychotic agents conducted to date revealed that both first-generation antipsychotics (FGAs) and second -generation antipsychotic (SGA) agents have limited long-term effectiveness, largely due to high rates of discontinuation (-75% discontinuation within 18 months).[7] Similar results were obtained in two large-scale European effectiveness trials.[8,9] In each of these trials, clinically significant side effects were noted in the majority of patients, and tolerability was the primary cause of at least 20% of all drug discontinuations.

The high likelihood of medication discontinuation has substantial clinical and economic implications, as treatment nonadherence is perhaps the single strongest predictor of relapse and rehospitalization.[10] Patients who have discontinued APDs may be as much as five times more likely to relapse as medicated patients.[11] Moreover, nearly half of rehospitalization costs in SCZ may be accounted for by medication nonadherence.[12] In addition to the effectiveness trials cited above, many observational studies and controlled trials have presented evidence that perceived side-effect burden frequently leads to both poor attitudes towards medications and a tendency towards discontinuation, nonadherence, and partial adherence.[13,14] Although side effects are highly prevalent, there is also substantial variability in liability to clinically significant or intolerable adverse events.[15] Consequently, understanding and predicting liability to side effects may be an effective strategy- to improve prognosis in schizophrenia.

Antipsychotic-induced side effects

FGAs were most commonly associated with neuromuscular side effects, including the potentially irreversible movement disorder, tardive dyskinesia (TD).[16] In large cohort studies, TD has been shown to affect at least one in five, and perhaps as many as one in three, patients treated chronically with FGAs.[17] New onset (incidence) of TD is approximately 3% to 5% per year of treatment, and these rates are increased as much as fivefold in elderly patients.[18] In addition to physical discomfort and social stigma, presence of TD has been associated with reduced quality of life, increased psychopathology, and increased mortality rates.[19] Even at low doses and/or intermittent treatment schedules, the high prevalence and morbidity associated with TD was the primary impetus for the promotion of SGAs as preferred firstline treatment, at least in the United States.[15,20] Although use of SGAs is not entirely free from TD risk, incidence and rates are as much as 80% lower for SGAs compared with FGAs.[21,22]

Though treatable and reversible, extrapyramidal symptoms (EPS) including Parkinsonian motor difficulties as well as akathisia, are highly prevalent with FGAs and are also associated with patient discomfort, dissatisfaction, and discontinuation of treatment.[16] Despite the initial optimism that SGAs would greatly reduce EPS burden, most SGAs still demonstrate a clinically relevant tendency to induce these symptoms.[23,24] In a large-scale effectiveness trial in chronic SCZ patients, SGAs were indistinguishable from a low-dose FGA (perphenazine) in rates of new onset of akathisia and EPS (5% to 10% each, irrespective of drug assignment).[25] However, meta-analytic reviews of the literature demonstrate that overall EPS burden may be reduced by 30% to 50% with SGAs.[26] Because the mechanism of action for all currently approved antipsychotic medications remains blockade of dopamine receptors,[27] motor and other side effects (eg, prolactin elevation) remain a concern in the treatment of SCZ.

While SGAs have moderately reduced EPS and substantially reduced TD liability relative to FGAs, these newer antipsychotics are most notable for their propensity to induce weight gain,[28] as well as related metabolic disturbances such as hypertriglyceridemia and hyperglycemia.[29] Clozapine and olanzapine are the APDs most frequently- associated with weight gain, but all APDs, even first-generation agents, seem to share these effects as a group to varying degrees.[30] For example, a largescale effectiveness trial in antipsychotic naïve patients demonstrated clinically significant weight gain (≥7% of baseline) in more than half of patients treated with haloperidol.[9] Obesity has serious implications for overall health and survival due to an increased risk for cardiovascular and malignant disorders[31]; these risks may be of particular importance in patients with SZ who often have limited access to health care and decreased motivation for weight reduction secondary to negative symptomatology.[13] Unfortunately, APD-induced weight gain is very difficult to reverse, even with sophisticated behavioral, dietary, and pharmacological interventions.[32]

Pharmacogenetic studies of antipsychotic-induced side effects

While the side effect profile of APDs is extremely burdensome in the aggregate, there is substantial interindividual variation in the degree of any particular motor or metabolic effect for a given patient.[15] Despite extensive research over the last two decades, data on clinical or biological predictors of antipsychotic side effects are limited. A few generalizations can be made, but these are not sufficient for individual-level prognosis: i) both the very old and the very young appear to be more susceptible to most APD-induced adverse events[18,32]; ii) patients experiencing extrapyramidal symptoms are twice as likely to develop TD as patients who do not exhibit EPS33; iii) olanzapine and clozapine have greater liability for metabolic effects and reduced incidence of motoric side effects compared with most other agents[7,9,26-30]; iv) APD dose may be correlated with some of these effects, but the relationship is weak and even low doses may carry substantial risk.[16,17,32] A priori identification of the patients who will be at a higher risk for development of adverse side effects could help clinicians avoid lengthy ineffective APD trials and limit patients' exposure to drug side effects.

Since the mid-1990s, the field of pharmacogenetics has offered the potential for providing readily accessible, immutable biomarkers - DNA sequence variants - that might be predictive of an individual's propensity for both positive and adverse effects of drugs. However, to date, the promise of personalized medicine has remained unfulfilled. Because academic pharmacogenetic research is often limited to small and clinically heterogeneous samples, individual studies have been unable to provide compelling results. Additionally, the modest effect sizes which are common in complex genetics present an obstacle in the quest for valid biomarkers, which require high sensitivity and specificity for individual clinical prediction. Moreover, examination of disparate polymorphisms across a wide variety of candidate genes has created an impression of scattered, unreplicated findings. Recently, however, a series of findings across multiple laboratories have begun to converge for a few genes related to serotonin and dopamine, the most prominent neurotransmitters targeted by APDs. In the subsequent sections, we will focus on the converging evidence implicating the most wellstudied candidates for pharmacogenetic predictors of antipsychotic-induced side effects. Particular emphasis will be placed on single nucleotide polymorphisms (SNPs) that have a sufficient evidence base to have permitted published meta-analytic studies.

Tardive dyskinesia

Tardive dyskinesia is the most extensively studied APDinduced side effect in the pharmacogenetics literature to date. These studies have typically been cross-sectional in nature, with ascertainment based on retrospective identification of cases with varying treatment histories and duration. The ability to study prevalence, rather than incidence in the context of a clinical trial, has permitted cumulative sample sizes in the thousands. It is important to note, however, that this ascertainment strategy may suffer from false negatives (patients with mild or reversible TD) and false positives (patients with acute motoric abnormalities that do not persist). Within this literature, variants within the genes encoding dopamine D2 and D3 receptors have been the primary focus, as detailed below.

Dopamine D2 receptor blockade is a property of all known antipsychotics, as demonstrated in vitro and in vivo,[34] yet a predictive relationship between variation in the DRD2 gene (located on chromosome 11q22) and APD-induced side effects has only been examined in a handful of studies. Most pharmacogenetic studies to date have examined the 3' Taq1A polymorphism (rs1800497), which more recently has been determined to be a nonsynonymous coding SNP in a neighboring ankyrin repeat gene (ANKK1 Glu713Lys).[35] Possibly due to linkage disequilibrium with another site (or sites) within DRD2 (, the minor (T) allele (also called the Al allele) at rsl800497 has been associated with a 40% reduction in striatal D2 receptor density based on both in vitro assays[36] and in vivo imaging studies.[37] This allele appears to be protective against TD. As shown in Table I, two recent meta-analyses (based on overlapping sets of studies) have persuasively demonstrated increased rates of TD in A2 (C) allele carriers.[38,39] The odds ratio (OR) of 1.30 indicates a 30% increase in risk for TD per allele, so that A2/A2 homozygotes are nearly 80% more likely to develop TD as A1/A1 homozygotes. Alternately, it can be said that AI/AI homozygotes have nearly half the rate of TD compared with A2/A2 homozygotes. However, it is important to note that the A2 allele is the common allele at this SNP, and A1/A1 homozygotes represent <10% of the Caucasian population (Al allele frequencies are much higher in non-white populations).

Like the D2 receptor, the dopamine D3 receptor is also selectively expressed in the basal ganglia and is considered to be a target of antipsychotic action[45]; consequently, several pharmacogenetic studies in schizophrenia have examined the DRD3 gene, located on chromosome 3q13.3. To date, only one functional SNP (rs6280), a missense variant resulting in a Ser to Gly substitution at amino acid position 9, has been validated for DRD3.[ The Gly variant has about a 35% allele frequency in non- African populations, and is actually the ancestral allele. The Gly variant has been associated with 4-fold greater dopamine binding affinity in vitro,[47] resulting in increased dopamine -mediated cAMP response and prolonged mitogen-associated protein kinase (MAPK) signal.[48] Several studies[49-52] (but not all)[53,54] have indicated that subjects carrying the Gly variant exhibit enhanced symptom response to treatment with clozapine or risperidone.

Concordant with the finding of heightened dopaminergic sensitivity for the Gly allele, multiple studies have demonstrated a significant increase in risk for tardive dyskinesia (TD) amongst Gly carriers. Despite several negative studies in the literature, three recent meta-analytic studies[40-42] indicate that this effect is detectable across a large pooled sample including patients of multiple ethnicities (Table I). Intriguingly, a recent studyindicates a strong association of the Gly allele with familial essential tremor, the most common inherited movement disorder.[48] However, the effect size for TD risk is modest (OR=1.16 in the largest meta-analysis), with diminishing effects in the more recent studies of this SNP. This pattern of diminishing effect size estimates over time, termed “the winner's curse,” is common in genetics studies and can ultimately- result in rejection of the initial finding as a false positive.[55] It is notable that this phenomenon was observed in the context of 13 published studies DRD3 Ser9Gly. Moreover, a very recent study in the large CATIE cohort (n=207 cases with TD vs 503 cases without TD), which was not included in any meta-analysis, demonstrated essentially no effects of either DRD3 Ser9Gly or DRD2 Taq1A.[56] Therefore, caution is warranted in the interpretation of other relationships reported across much smaller study sets.

A third dopamine-related gene that has been investigated in multiple pharmacogenetic studies of TD is Catechol Omethyltransferase (COMT). While subcortical dopamine activity is primarily terminated by reuptake mediated by the dopamine transporter, a secondary mechanism for dopamine clearance is metabolic degradation via COMT.[57] Additionally, COMT is the predominant mechanism of dopamine clearance in frontal cortex. The COMT gene contains a functional polymorphism that codes for a substitution of methionine (met) for valine (val) at codon 158. The met allele, which has 36% to 48% allele frequency across various ethnicities, results in a thermolabile protein that has one fourth the enzymatic activity of the val carrying protein.[58] (In other words, the val allele results in reduced synaptic dopamine due to more rapid clearance). Across five studies meta-analyzed by Bakker and colleagues,[39] the val allele was associated with modestly increased risk for TD (OR=1.19; Table I). It is unknown whether the protective effect of the met allele is a direct result of subcortical COMT activity, or is secondary to alterations (eg, upregulation) in frontostriatal circuitry. In addition to dopamine antagonism, one of the common features of many antipsychotics is near-saturation binding of serotonin (5-HT)2 receptors, which has been confirmed in vivo using PET imaging.[59,60] While 5-HT binding is often considered a hallmark of SGAs, it is important to note that serotonergic binding properties are observed for several FGAs as well.[61,62] The 5-HT2A receptor gene (HTR2A) has been examined in several pharmacogenetic studies of TD; in particular, a promoter region SNP (rs6313), which has been previously- associated with response to antipsychotics (as well as antidepressants), has been extensively studied in relation to TD. While these studies generally converge to indicate a modestly reduced effect of the C allele on symptom response,[63] this same allele has been associated with significantly increased risk for tardive dyskinesia.[43] As shown in Table I, a recent meta-analysis reported an odds ratio of about 1.6 for C allele carriers across 6 studies; effects were strongest in older patients (age >47 years), and were specifically associated with limb-truncal (but not orofacial) TD[43] Notably, this SNP is a perfect proxy for another promoter region SNP,rs6311 (also referred to as -1438G/A), which appears to affect transcription of the receptor.[64] Specifically, the G allele (a perfect proxy for the C allele at rs6313) tends to be associated with reduced expression of the receptor. It can therefore be inferred that reduced availability of the 5-HT2A receptor is a risk factor for tardive dyskinesia. Notably, 5-HT2A receptors are strongly expressed in the caudate and putamen,[65] and recent evidence obtained from dopamine-depleted rodents suggests a complex interplay of subcortical dopamine and 5-HT in the regulation of motor behavior.[66]

Two genes outside of the dopamine and 5-HT systems have received sufficient attention in the pharmacogenetics of TD to merit meta-analysis (Table I). Many commonly prescribed APDs, including FGAs (haloperidol, perphenazine, thioridazine), as well as SGAs (risperidone and aripiprazole), are metabolized in the liver by CYP2D6 (debrisoquine hydroxylase).[67] The CYP2D6 gene is highly polymorphic, with over 70 known variants (for a current classification, view the allele nomenclature at http://www.imm.ki.se/CYPalleles/). Homozygosity for null alleles gives rise to the “poor metabolizer” phenotype characterized by no enzyme activity while null allele heterozygosity gives rise to an intermediate debrisoquine hydroxylase metabolic phenotype characterized by impaired - but not absent - enzyme activity.[68] Reduced CYP2D6 activity can be expected to result in higher effect dose as measured by blood levels of active drug, with potential for increased dose-dependent side effects. Consistent with this pharmacokinetic prediction, a metaanalysis of 8 studies demonstrated a moderate effect of (any) loss of function alleles on risk for TD (OR=1.43), while homozygotes (poor metabolizers) had 1.64-fold greater odds of suffering tardive dyskinesia.[44] A recent small study further confirms these results.[69] A similar effect has been studied for SOD2, the gene encoding manganese superoxide dismutase, a mitochondrial enzyme involved in oxidative metabolism. A functional SNP (Ala9Val), affecting efficiency of MnSOD transport, has been associated with TD risk; counterintuitively, the less efficient val allele is protective.[39] Homozygotes for the Ala (T) allele are about twice as likely to develop TD compared with val carriers (Table I).

Extrapyramidal symptoms

Compared with the relative plethora of studies on tardive dyskinesia, pharmacogenetic studies of EPS are lacking. However, a few studies have reported allelic effects on acute side effects that are consistent with those reported for TD. For example, Eichammer et al[70] reported increased incidence of akathisia amongst DRD3 Gly carriers; however, two studies of extrapyramidal symptoms have been negative.[71,72] One additional study identified another DRD3 SNP (rs167771) which was associated with EPS in a study of 270 risperidonetreated patients,[73] but this result awaits replication. One small study has demonstrated an effect on EPS risk for the C allele of rs6313 in HTR2A that parallels its effect on TD.[71] Although not previously examined in TD studies, a SNP in RGS2 (rs4606) has been associated with extrapyramidal symptoms in two studies.[74,75] Although a third study was negative, this regulator of intracellular dopamine signaling merits additional research.[76]

Prolactin elevation

While prolactin elevation has also not been widely studied across most of the genes listed in Table I, there have been seven published studies examining DRD2 TaqlA.[77-83] As displayed in Table II, these studies have yielded mixed results across a variety of APDs. Notably, the three positive studies all reported that the Al allele was associated with increased risk for hyperprolactinemia, and a fourth study demonstrated the same effect in females only. This is the opposite allele that was associated with TD, which may reflect the fact that prolactin response is mediated via the tuberoinfundibular pathway (hypothalamus and pituitary).[84]

Weight gain

It has been suggested that increased 5-HT binding profiles may account for the increased liability to weight gain observed in the second-generation antipsychotics.[85] A survey of the literature of the regulation of feeding behavior points to a major role for 5-HT, with both animal and human investigations showing, in general, that increasing 5-HT results in decreased feeding, with the reverse also true.[86] Pharmacologic agonists of 5-HT2C lead to decreased feeding in animals[87] it is logical to speculate that 5-HT2C antagonists, including most secondgeneration antipsychotics, might lead to increased food intake.

Perhaps the best evidence for a specific role of serotonin-related genetic factors in antipsychotic-induced weight gain is provided by studies of the promoter region polymorphism, -759 T/C (rs3813929), in the HTR2C gene (on the X chromosome). Reynolds and colleagues[88] studied 123 adult drug-naïve Han Chinese SCZ patients treated primarily with risperidone or chlorpromazine. Subjects with the T allele at this locus gained significantly less weight than subjects with the C allele in short-term (6- and 10-week) treatment; none of the 27 subjects with the T allele met criteria for severe (>7%) weight gain after 6 weeks, as compared with 28% of the 96 subjects without the T allele. Two studies[89,90] also reported an association of the T allele to reduced weight gain in a small samples of clozapinetreated patients, although this effect was only significant in males in one of these. Ellingrod and colleagues[91] reported that the T allele is associated with less weight gain in Caucasian patients treated with olanzapine, and Templeman et al[92] reported the same for weight gain associated with a mixed group of antipsychotics in a small Spanish first-episode cohort. Recently, Lane et al[93] extended these findings to include risperidone (in 123 Han Chinese inpatients), and Ryu et al[94] demonstrated the same effect for the T allele in 84 Korean inpatients treated on various antipsychotic monotherapies. A few studies, however, have not detected significant associations between-759 T/C and clozapineinduced weight gain[95-97] which may reflect the winner's curse, but it should be noted that these studies were restricted to chronic patients with extensive prior treatment. A meta-analysis of 8 studies demonstrated a greater than 2-fold increase in risk for clinically- significant (7% to 10% or greater) weight gain from baseline associated with the C allele at this SNP.[98]

Analogous to the aforementioned role of RGS2 in EPS, one gene involved in intracellular signaling has been repeatedly with respect to APD-induced weight gain. GNB3 encodes a subunit of a heterotrimeric guanine nucleotide-binding protein (G protein), which integrates signals between receptors and effector proteins.[99] An SNP polymorphism (C825T) in this gene has been associated with essential hypertension and obesity; this SNP is also associated with relative prevalence of a high-activity splice variant of GNB3.[100] According to a recent meta-analysis, five studies have examined effects of this SNP on APDinduced weight gain; the T allele was marginally associated with increased weight gain.[101] However, this effect was consistent with its effect on BMI and other metabolic variables in the general population, so the mechanism in the context of APD treatment remains unclear.

Conclusions and future directions

As summarized in the preceding sections, pharmacogenetic studies have begun to converge on a few genetic variants that are replicably associated with the common APD-induced motor and metabolic side effects. However, three factors limit the ability of the field to deliver on the promise of personalized medicine at this time, and point to critical issues for the next generation of pharmacogenetic studies. First, a treating psychiatrist would be unable to use this information to offer a validated alternative, due to the lack of pharmacogenetic head-to-head comparisons of treatment with differing mechanisms. Second, even fairly consistent single-gene results, such as those observed for DRD3 and TD, fail to provide large enough effect sizes to make confident clinical decisions. In order to provide a clinically useful test, with sufficient sensitivity and specificity to make confident individual predictions, a combination of SNPs across different loci will be required. Third, the economics of conducting pharmacogenetic tests on a large clinical scale will need to be justified to payers, including the insurance companies and the federal government. In order to do so, pharmacogenetics researchers will need to quantify the beneficial economic impact of tailored prescription practices.[102]

Of course, any personalized clinical decision-making process will optimally include validated predictors of symptom response as well as adverse effects. The variability in symptom response ranges from patients who experience rapid symptom remission to a subset of patients often described as “treatment-refractory.”[15] Even when fully adherent with medication, as many as 40% of patients fail to demonstrate adequate response on the hallmark positive symptoms of hallucinations and delusions.[103] Unfortunately, the literature on pharmacogenetics of response is more difficult to summarize than for side effects; due to wide differences in trial methodology and definition of dependent measures, no metaanalytic studies have been published in the last decade. (One early meta-analysis of clozapine response identified an effect of HTR2C T102C, as described earlier.[61]) Finally, it should be noted that candidate gene approaches to pharmacogenetics run a dual risk of either an overly restrictive search space, or a potentially overwhelming number of candidates. While initial pharmacogenetic studies have primarily focused on dopamine and serotonin genes, the slow pace of individual candidate gene investigations has resulted in many additional scattered and isolated studies across investigators. On the other hand, the advent of genome-wide association studies (GWAS) provides a hypothesis-free method of generating candidate genes for novel complex phenotypes. Unfortunately, this method carries its own statistical concerns, most notably limitations in statistical power (due to correction for multiple comparisons) in necessarily limited clinical trial samples.

One way to enhance sample size and statistical power in the short run is to utilize a strategy that permits crosssectionally defined phenotypes. In a proof of principle study, we have recently utilized the Affymetrix 500K microarray in a sample of our retrospectively-characterized patients with schizophrenia. (Initial case-control analyses were SCZ diagnosis were published for data obtained from the first 322 Caucasian subjects.[104] All subjects self-identified as Caucasian non-Hispanic; testing of 210 ancestry informative markers (AIMs) revealed no evidence of population stratification). In this same sample, we have performed a preliminary analysis examining treatment responsiveness, using clozapine assignment as a proxy for poor response. Detailed chart reviews permitted classification of 97% of the sample. Approximately 35% of patients were assigned clozapine due to treatment nonresponsiveness, and groups were matched on key demographic variables including age, duration of illness, sex, and family history. Despite the small sample for this interim analysis, one SNP nearlyobtained genome -wide significance (P=4.3*10-7).This SNP neighbors CNTN4 (contactin-4), a neuronal membrane protein that functions as a cell adhesion molecule, and is thought to be critical for the formation of axon connections in the developing nervous system105; CNTN4 has also recently been implicated in autism.[106]

In the longer term, much larger prospective studies will be required to achieve to: i) obtain clear estimates for risk parameters; and ii) determine whether application of a pharmacogenetic risk profile is clinically and economically advantageous. Optimally, such studies mayfocus on the first episode of SCZ, which typically occurs in late adolescence or early adulthood[107] and may be the most critical period in the life of an individual with SCZ. Successful treatment of the initial psychotic episode is crucial for minimizing the cascading effects of social and vocational deterioration.[108,109] From a methodological perspective, studies of first-episode patients minimize potential confounds associated with chronic illness and variable history of prior treatment; first-episode cohorts are also marked by reduced duration of psychotic symptoms, substance abuse, and functional/social disabilities.[110] Bycontrast, studies of chronic SCZ may systematically overrepresent patients who are not fully responsive to treatment or are nonadherent to treatment (or both), and underestimate APD response. First-episode samples maybe less biased on these factors and therefore may be more informative about the spectrum of outcomes with APD treatments. While large-scale prospective trials involving first-episode cohorts are logistically challenging, such studies would hold substantial promise for advancing the field in the next decade.

Table I.

List of meta-analytic studies of single nucleotide polymorphisms (SNPs) from candidate genes for tardive dyskinesia (TD), with the associated allele and odds ratio (OR) of the association.

Gene	SNP	Allele	No of studies	N patients (with /without TD)	OR	Reference
DRD2	Taq1A (rs1800497)	A2 (C)	6	1256(507/749)	1.30	Zai et ai 2007[38]
DRD2	Taq1A (rs1800497)	A2 (C)	4	764(297/467)	1.30	Bakker et al 2008[39]
DRD3	Ser9Gly (rs6280)	Gly(C)	8	780 (317/463)	1.33	Lerer et ai 2002[40]
DRD3	SerSGly (rs6280)	Gly(C)	11	1610(695/915)	1.17	Bakker et al 2006[41]
DRD3	Ser9Gly(rs6280)	Gly(C)	13	2026(928/1098)	1.16	Tsai etal 2009[42]
COMT	Val158IMet(rs4680)	Val(G)	5	1089(382/707)	1.19	Bakker et al 2008[39]
HTR2A	T102C(rs6313)	C	6	635(256/379)	1.64	Lerer et al 2005[43]
CYP2D6		Loss of function alleles	8	569(220/349)	1.43	atsopoulos 2005[44]
SOD2	Ala9Val (rs4880)	Ala(T)	4	680(134/546)	2.04	Bakker et al 2008[39]

Table II.

List of studies of the Taq1A polymorphism (rs1800497) from the ANKK1/DRD2 locus in association with antipsychotic drug-related prolactin levels.

Reference	Drug	N patients	Allele	Significant?
Calarge et al 2009[77]	Risperidone	107	A1 (T)	Yes
Kwon et al 2008[78]	Aripiprazole	90		No
Yasui-Furukori et al 2008[79]	Risperidone	174		No
Aklillu et al 2007[80]	Perphenazine	22	A1(T)	Yes
Anderson et al 2007[81]	Rispendone	101		No
Young et al 2004[82]	Various	144	A1(T)	Yes
Mihara et ai 2000[83]	Nemonapride	25	A1(T)	Females only