Effects of the Serotonin Transporter Gene Promoter Polymorphism onMirtazapine and Paroxetine Efficacy and Adverse Events in Geriatric MajorDepression FREE

Genetic differences among patients may contribute to differences inmedication response and the severity of adverse effects. The 5HTTLPR (“long/short”) polymorphism in the serotonin transporter(5-HTT) gene (SLC6A4) promoter region has been studiedas a marker for selective serotonin reuptake inhibitor treatment response.Smeraldi et al¹ showed that the short formof 5HTTLPR (S allele) impairedthe efficacy of fluvoxamine maleate in a sample of 102 patients with majordepression. This effect was also reported among patients with psychotic depression.² The same effect of the 5HTTLPR polymorphismwas reported in a study with 57 geriatric patients with major depression treatedwith paroxetine hydrochloride.³ Subsequently,this finding was confirmed in a group of 60 paroxetine-treated patients withdepression.⁴ In a study of 37 patients withdepression who were receiving fluoxetine, those homozygous for theS allele of 5HTTLPR showed a higher frequencyof agitation and insomnia after treatment.⁵ Selectiveserotonin reuptake inhibitors are thought to act through the inhibition ofthe 5-HTT, so a promoter variant that affects the expression of this proteincould affect efficacy and adverse events even though the protein structureis unchanged. In some cell biology and imaging studies, the long (L allele) and short versions of the 5HTTLPR polymorphismhave been found to differentially affect the abundance of the 5-HTT.^6- 8

Other studies of the effects of the 5HTTLPR onantidepressant treatment outcomes have yielded contradictory results. Minovet al⁹ found no effect of the 5HTTLPR polymorphism on response in 104 patients receiving a varietyof antidepressant treatment regimens. Treatment heterogeneity could have maskedgenetic effects in that study. However, Kim et al¹⁰ studied102 Korean patients treated with fluoxetine and found that the S allele was associated with a better treatment response. A similarresult was obtained in a study with Japanese patients.¹¹ Anotherstudy with Japanese patients found no association between the 5HTTLPR polymorphism and fluvoxamine-induced nausea.¹² Oneexplanation for these divergent results may be that the effect of the 5HTTLPR polymorphism is dependent on ethnic background,so studies conducted with Asian patients might yield different results fromthose conducted with patients of other races. However, a study with Chinesepatients reported a better response to fluoxetine in patients with the L/L genotype,¹³ as has beenreported in European subjects.

The selective serotonin reuptake inhibitor paroxetine inhibits the 5-HTT,thus decreasing the reuptake of serotonin.¹⁴ Mirtazapineincreases serotonin release through noradrenergic stimulation of excitatory α₁-adrenergic receptors located on serotonergic neurons and by blockadeof the inhibitory α₂-adrenergic heteroreceptor located onthe serotonergic cell body.¹⁵ Mirtazapine alsoblocks serotonin types 2A (5-HT_2A), 2C, and 3 receptors and augmentsadrenergic neurotransmission. Because of the differing interactions of paroxetineand mirtazapine with the 5-HTT, we sought to determine if the5HTTLPR polymorphism predicts antidepressant efficacy in geriatricpatients with major depression treated with these agents. We also tested theeffects of the 5HTTLPR on adverse events during treatmentwith paroxetine and mirtazapine.

The study group included 255 outpatients 65 years or older with majordepressive disorder who were randomized to mirtazapine or paroxetine in adouble-blind design at 18 US centers. All patients provided written informedconsent. Details of the study design and overall clinical outcomes have beendescribed previously.¹⁶ Of the 255 randomizedpatients, 254 received the study medication; 246 (mirtazapine, n = 124;paroxetine, n = 122; all subjects–treated group) provideda DNA sample and had at least 1 postbaseline measure of efficacy, adverseevents, or compliance. At baseline, all patients scored higher than the age-adjusted25th percentile on the Mini-Mental State Examination, and all had HamiltonDepression Rating Scale (HDRS)–17¹⁷ scoresof 18 or higher. Starting doses were 15 mg/d for mirtazapine and 20 mg/d forparoxetine hydrochloride. Dosages for both medications were increased to 30mg/d after 14 days, with the option to increase to 45 mg/d of mirtazapineor 40 mg/d of paroxetine hydrochloride at subsequent visits. Exclusion criteriawere clinically significant unstable medical illness, history of treatmentresistance, principal psychiatric diagnoses other than major depression, useof other psychotropic medications, and electroconvulsive therapy within theprevious 6 months. According to self-report, 20 patients were from minoritybackgrounds. For mirtazapine, these included 1 Asian, 4 African Americans,and 2 patients from other minority backgrounds. For paroxetine, there were2 Asians, 6 African Americans, and 5 from other minority backgrounds. Allother subjects were of self-reported white European ancestry.

Genomic DNA was extracted from frozen whole blood as described previously.¹⁸ We determined 5HTTLPR genotypes by polymerase chainreaction and gel electrophoresis.¹⁹ Plasmadrug levels were obtained after 4 weeks of treatment as described elsewhere.²⁰ Efficacy outcome measures (determined at baselineand at days 7, 14, 21, 28, 42, and 56) included the HDRS-17¹⁷ andthe Geriatric Depression Scale²¹ (GDS). Measuresrelated to adverse events included the severity of adverse events, medicationcompliance, and time to discontinuation due to an adverse event. At each visit,clinicians specifically assessed patients for adverse events and rated theirseverity as mild (1), moderate (2), or severe (3). Adverse event ratings foreach patient were summed across the entire trial and were adjusted to takeinto account treatment duration and mean daily dose. The amount of medicationtaken was determined by counting the number of prescribed tablets remainingat each clinic visit. Dosage compliance was determined by the total numberof medication doses taken divided by the total number of capsules given.

Statistical methods included analysis of covariance (adjusted for baselinescores), the Cochran-Mantel-Haenszel statistic, and Cox proportional hazardsanalysis for survival data. Because discontinuations during the last 7 daysof the study were difficult to define, survival analyses were terminated atday 49. To test for interactions between the 5HTTLPR polymorphismand the 5-HT_2A 102 T-to-C variant (dbSNP database identificationnumber rs6313) that we previously found to affect paroxetine discontinuations,¹⁸ we used Cox proportional hazards analysis with aninteraction term. All analyses were performed for the entire group and againwith 20 patients from minority backgrounds omitted.

The baseline characteristics and clinical course for patients treatedwith paroxetine and mirtazapine stratified by 5HTTLPR genotype are presentedin Table 1. Overall allele frequenciesfor the entire sample were 0.55 for the L alleleand 0.45 for theS allele. Overall genotype frequenciesfor the entire sample were 32.0% for the L/L, 45.5%for the S/L, and 22.5% for the S/S genotypes. There were no significant differences between the paroxetineand mirtazapine groups in allele or genotype frequencies. For the full sampleand treatment subsamples, there were no deviations from the Hardy-Weinbergequilibrium. There were no differences among the genotype groups in mean age,numbers of men and women, numbers of people from minority backgrounds, orbaseline Mini-Mental State Examination score for either treatment group. Forthe paroxetine group, baseline body weight was significantly greater for theL/L genotype group than for theS/L (P = .02) and S/S groups(P = .002). Because plasma drug concentrationsand pharmacokinetics can potentially be affected by body weight, we includedbaseline body weight as a covariate in all subsequent analyses. There wereno significant differences in baseline body weight among genotype groups forthe mirtazapine-treated subjects. Two subjects lacked baseline body weightmeasurements and were excluded from the subsequent analyses, resulting ina final sample size of 244 subjects.

For both paroxetine and mirtazapine, when patients were stratified by5HTTLPR genotype (S/S, S/L,and L/L), there were no significant differences amongthe 3 genotype groups at any time point for either the HDRS-17 or the GDS.Prior studies combined the S/S and S/L genotypes into an “S carrier”group for analysis.^1,3 No differenceswere seen among paroxetine-treated patients on the HDRS-17 in this 2-levelanalysis (Figure 1A). For paroxetine-treatedpatients, carriers of the S allele showed significantlyhigher scores on the GDS (more depressed) at days 7 (P = .02;F_1,210 = 5.28) and 28 (P = .04;F_1,175 = 4.11) after adjusting for baseline body weight(Figure 1B). These differences betweengenotype groups were smaller than those reported previously in a sample of51 geriatric patients treated with paroxetine.³ Inthat study, at week 2 there was a 29.6% decrease in HDRS-17 scores from baselinein S carriers but a 49.3% decrease from baselinefor the L/L genotype. In our sample, with a 2-sidedanalysis and P = .01, the power to detect a differenceof this magnitude would be 0.96.

The S allele carriers treated with mirtazapineshowed lower HDRS-17 scores at day 14 (less depressed; P = .03; F_1,185 = 4.96) (Figure 1C). The GDS scores for mirtazapine-treated patients withthe S/S and S/L genotypeswere significantly lower at baseline than those for patients with theL/L genotype (P = .03;F_1,222 = 4.97) (Figure 1D). After adjusting for this difference, GDS scores for mirtazapine-treated S carriers were not significantly different from thoseof other subjects at any subsequent time point.

The 5HTTLPR polymorphism had a significanteffect on discontinuations caused by adverse events for both medications.As previously reported, discontinuations due to adverse events were more frequentamong paroxetine-treated patients than mirtazapine-treated patients.¹⁶ For paroxetine-treated patients, survival analysesshowed a significant association between the number of S alleles and the probability of discontinuation due to adverse eventsat days 14, 21, 28, 42, and 49 (Figure 2A)(Cox regression analyses; χ²₂ tests; P<.05 for all). All statistical tests were adjusted for differencesin baseline body weight. At the same time points, patients with the S/L genotype showed a significantly greater risk of discontinuationdue to adverse events than those with the L/L genotype.Patients with theS/S genotype also showed a greaterseverity of adverse events than those with the L/L genotype(F_1,201 = 5.52; P = .02)(Table 1). Paroxetine-treated patientswith the S/S genotype had a lower final daily dose(S/S vs L/L: F_1,221 = 8.63; P = .004; S/S vs S/L: F_1,221 = 7.09; P = .008), decreased dosing compliance (S/S vs L/L: F_1,221 = 19.12; P<.001; S/S vs S/L: F_1,221 = 23.06; P<.001),and lower plasma levels at day 28 (S/S vs S/L: F_1,169 = 4.30; P = .04; S/S vs S/L: F_1,169 = 7.25; P = .008) (Table1). Adverse events associated with discontinuations among paroxetine-treatedpatients with theS/S genotype included gastrointestinalcomplaints, fatigue, agitation, sweating, and dizziness.

Surprisingly, among mirtazapine-treated patients, the L allele was strongly associated with discontinuations due to adverseevents. Survival analyses showed a significant association between the numberof L alleles and the probability of discontinuationdue to adverse events at days 14, 21, 28, 42, and 49 (Figure 2B) (Cox regression analyses; χ²₂ tests; P<.05 for all). The L/L genotypewas also associated with a greater severity of adverse events (L/L vs S/S: F_1,201 = 5.18; P = .02) and a lower final daily dose (L/L vs S/S: F_1,221 = 7.17; P = .008; S/L vs S/S: F_1,221 = 4.19; P = .04) (Table 1).Among mirtazapine-treated patients, there were no significant differencesin dosing compliance among the genotype groups and no significant differencesin plasma levels at day 28 (Table 1).Adverse events associated with discontinuations among mirtazapine-treatedpatients with the L/L genotype included drowsiness,dizziness, and anxiety.

We previously showed that the 5-HT_2A 102 T-to-C polymorphismpredicted discontinuations due to adverse events for paroxetine-treated patientsin our study cohort.¹⁸ The χ² analyses(5-HT_2A 102 T-to-C genotype × 5HTTLPR genotype contingencytables) showed no significant associations between the frequencies of individual5-HT_2A 102 T-to-C genotype groups and 5HTTLPR genotype groups foreither drug (mirtazapine: χ²₄ = 1.0;paroxetine: χ²₄ = 7.1). This indicatesthat the relative numbers of individuals in each of the 5-HT_2A genotypegroups were not dependent on the 5HTTLPR genotype. Cox proportional hazardsregression analysis for paroxetine-treated patients showed significant effectsfor 5HTTLPR (χ²₁ = 6.69; P = .01; hazard ratio = 2.62; S/S vs other genotypes) and 5-HT_2A 102 T-to-C(χ²₁ = 9.14; P = .003;hazard ratio = 3.03; C/C vs other genotypes)on discontinuations due to adverse events, but the interaction term in theanalysis was not significant (χ²₁ = 0.31).This indicates that 5HTTLPR and 5-HT_2A 102T-to-C variants have additive but not interactive effects on paroxetine discontinuation.

Analyses performed with the 20 patients of minority backgrounds removedfrom the sample gave results similar to those for the full sample. For whitepatients treated with paroxetine, survival analyses showed a significant associationbetween the number of S alleles and discontinuationsdue to adverse events at days 14, 21, 28, 42, and 49 (Cox regression analysis;χ²₂ tests; P<.05 forall). Similarly, for white patients treated with mirtazapine, there was asignificant association between the number ofL allelesand discontinuations due to adverse events at days 14, 21, 28, 42, and 49(Cox regression analysis; χ²₂ tests; P<.05 for all).

These results indicate that the 5HTTLPR polymorphismis a marker for medication discontinuation due to adverse events in geriatricpatients treated with paroxetine and mirtazapine. For patients treated withparoxetine, discontinuations due to adverse events were most frequent amongthose with the S/S genotype. TheS/L genotype was associated with an intermediate frequency of discontinuations,whereas L/L carriers had the fewest discontinuations,indicating a gene-dosage effect. Patients with the S/S genotypealso had a greater severity of adverse events. These effects were significanteven after adjustment for differences in baseline body weight. Paroxetine-treatedpatients with the S/S genotype showed lower medicationcompliance, resulting in lower plasma drug concentrations at day 28. In contrast,for patients treated with mirtazapine, discontinuations were most frequentamong those with the L/L genotype.

Most prior studies of the 5HTTLPR and antidepressant treatment outcomehave focused on measures of efficacy rather than adverse effects. For paroxetine-treatedpatients, we did find that the GDS showed impaired antidepressant efficacyfor patients with the S/S genotype at 2 time points,thus confirming the results of some prior studies.^3- 4 However,the magnitude of this effect in our study was small in comparison with thatfor discontinuations due to adverse events. A recent report showed that in37 patients with depression who were treated with fluoxetine, the S/S genotype was associated with an increased frequency of agitationand insomnia.⁵ In our study, among paroxetine-treatedpatients, 1 patient discontinued the medication because of agitation, 3 discontinuedowing to fatigue, and none discontinued as a result of insomnia. It is unclearwhy our results differ from those of the fluoxetine study. Pharmacologic differencesbetween fluoxetine and paroxetine could account for variation in adverse effectprofiles associated with the S/S genotype, as couldthe differing age distributions of the 2 samples. In any case, both studiesshow that disturbances in circadian rhythms and level of alertness are associatedwith the 5HTTLPR S/S genotype during treatment withselective serotonin reuptake inhibitors. Future studies of the 5HTTLPR andantidepressant therapy should include not only efficacy outcome measures butalso measures of adverse events and discontinuation. In retrospective studiesof the 5HTTLPR and treatment outcome, if the primary outcome measure is efficacyonly, genetic effects on adverse events may be missed, particularly if thecompletion of a fixed duration of treatment is an inclusion criterion. Adverseevents, if severe, could contribute to reports of a lack of efficacy in pharmacogeneticstudies because physical or psychological discomfort may increase scores oncertain items in depression rating scales.

Differences in the mechanisms of action of mirtazapine and paroxetinemay account for the divergent effects of the 5HTTLPR on treatment efficacyand adverse events with these agents. Paroxetine directly interacts with the5-HTT to increase the availability of serotonin in the synapatic cleft.^22- 23 Mirtazapine induces the release ofnorepinephrine as well as serotonin in the brain through antagonism of α₂-adrenergic receptors on noradrenergic and serotonergic neurons.²⁴ Norepinephrine release at synapses with serotonergicneurons also augments serotonin release. Thus, whereas paroxetine has a primarilyserotonergic mode of action, mirtazapine has both serotonergic and noradrenergicactions. Although mirtazapine does not directly interact with the 5-HTT, individualdifferences in the abundance of this protein could affect the response tomirtazapine-induced changes in serotonergic neuronal activity. It is possiblethat efficient serotonin reuptake among mirtazapine-treated patients carryingthe L/L genotype alters mirtazapine-induced changesin serotonergic activity (which occur through a fundamentally different mechanismthan changes induced by paroxetine) and therefore increases the likelihoodof adverse events.

We previously demonstrated that the C/C genotypeat the 5-HT_2A 102 T-to-C polymorphism is a strong risk factor fordiscontinuation due to adverse events among paroxetine-treated patients.¹⁸ Subjects with both the 5HTTLPR S/S genotype and the 5-HT_2A 102 T-to-C C/C genotype appear to be at exceptionally high risk for discontinuation.Whereas 7 of 10 paroxetine-treated subjects with the 5HTTLPR S/S and 5-HT_2A T-to-C C/C genotypesdiscontinued the medication owing to adverse events, none of the 5 subjectswith the L/L or T/T genotypewho were receiving paroxetine discontinued. However, Cox proportional hazardsanalysis showed that the increase in risk for discontinuation due to the S/S or C/C genotype was additive,not interactive; thus, the 2 effects are independent in our sample. Notably,as the number of predictors in the model increases, cell size decreases, whichcan limit the power to detect an interaction.

Two prior studies found that among Asian patients, the effect of the5HTTLPR depends on ethnic background.^10- 11 Oursample contained only 3 Asian patients (2 treated with paroxetine, and 1 treatedwith mirtazapine). However, re-analysis of our data with all 20 subjects ofminority backgrounds removed provided results similar to those for the fullsample.

Aging affects 5-HTT function.^25- 26 Thiscould be one reason why our results differ from those performed with samplesthat included younger subjects. Another factor that could have affected outcomesin our study was the exclusion of patients with a history of treatment resistanceand those hospitalized as inpatients. We do not know the plasma drug concentrationsin patients who discontinued the medication before day 28 owing to adverseevents, which could have affected the outcomes. Personality disorders andconcurrent medical disorders could also affect outcomes, but these factorsare difficult to quantify and were not included in the analysis. Because multipleoutcomes were tested, some positive results could be due to type 1 statisticalerror.

In summary, these results show that the S alleleat the 5HTTLPR polymorphism is associated with adverseevents in geriatric patients with major depression treated with paroxetine.This effect was independent of the previously described association betweenthe 5-HT_2A 102 T-to-C polymorphism and adverse events with paroxetine.¹⁸ In addition, there was a modest effect of the 5HTTLPR on paroxetine efficacy in this patient population.The 5HTTLPR polymorphism may also be important indetermining discontinuation due to adverse events during mirtazapine therapy,but surprisingly, the L allele confers greater riskwith mirtazapine than the S variant associated withdiscontinuations among paroxetine-treated patients. Differences in the mechanismsof action of mirtazapine and paroxetine may account for the divergent effectsof the 5HTTLPR polymorphism on treatment outcomewith these agents. Replication of our findings could provide the basis forclinical use of this genetic variant in making rational medication choicesbefore treatment is initiated.

Submitted for Publication: February 11, 2004;final revision received May 3, 2004; accepted June 9, 2004.

Correspondence: Alan F. Schatzberg, MD,Department of Psychiatry and Behavioral Sciences, Stanford University Schoolof Medicine, 401 Quarry Rd, Room 300, Stanford, CA 94305-5548 (afschatz@leland.stanford.edu).

Funding/Support: This study was supported byOrganon Pharmaceuticals USA, Inc (Roseland, NJ), the National Alliance forResearch on Schizophrenia and Depression (Great Neck, NY), the Nancy PritzkerNetwork, and the Department of Veterans Affairs Sierra Pacific Mental IllnessResearch, Education, and Clinical Center (Palo Alto, Calif).

Role of the Sponsor: Funding for this trialwas provided in part by Organon Pharmaceuticals USA, Inc. Personnel from theOrganon Medical Affairs Department played a major role in the design, conduct,and analysis of the trial, which was performed in accordance with guidelinesfor clinical trials of pharmaceutical agents established by US regulatoryauthorities. The sponsor hired an independent contract research organization(Scirex, Inc, Horsham, Pa) to implement the study, oversee the collectionof blood samples, and maintain the study database. Statistical analyses weredesigned and performed by Dr Murphy and employees of Scirex, Inc. All DNAwork was performed in Dr Murphy’s laboratory at Stanford University.The research contract permitted the investigators to publish the data withoutmandatory approval by the sponsor. The sponsor reviewed the manuscript andprovided comments and suggestions. This work resulted in a patent applicationthat is pending.

Previous Presentation: This study was presentedat the Annual Meeting of the American College of Neuropsychopharmacology,San Juan, Puerto Rico, December 11, 2002.

Acknowledgment: We thank Nina Pascoe, JeremyClaassen, Clara Poon, and Yuen Ling Lee for technical assistance. ChristopherPan, Yaping Wang, Xiao Li, and Phil Smolen assisted with data analysis.