Introduction

While antidepressants (ADs) are widely prescribed for depression and anxiety disorders in children and adolescents, clinical trials suggest that only 50–60 % of patients respond to any single AD (Papakostas 2009), and clinical experience suggests even lower response rates (Kirchheiner et al. 2004; Souery et al. 2007). In addition to the difficulty in identifying an efficient treatment, AD use is frequently associated with intolerable adverse drug reactions. Inability to tolerate side-effects (SEs) is the reason for discontinuing AD therapy in 30 % of adults treated with tricyclic ADs or SSRIs (Maddox et al. 1994). A high risk of discontinuation due to SE was noted in a recent meta-analysis in young patients treated with tricyclic ADs as compared to those treated with SSRIs (Qin et al. 2014). A recent meta-analysis assessing the efficacy and tolerability of ADs in pediatric anxiety disorders found no indication of increased risk of discontinuation treatment due to side effects (Strawn et al. 2015), however, a trend towards activation symptoms was reported. Clinicians currently have no way to predict individual efficacy and SE profiles, and a trial-and-error switching paradigm often leaves these patients in psychological distress for weeks and even months. The longer depression/anxiety persists, the greater the risk of developing chronic depression, suicidal ideation, and suicidal behavior. Clearly, improved methods of patient—AD matching to minimize SEs would greatly improve treatment outcome and tolerability.

Serious adverse events (SAE) following selective serotonin reuptake inhibitors (SSRIs) treatment, including suicidality, aggression, hyperarousal, agitation and antidepressant-induced mania (AIM) have been extensively reported (Strawn et al. 2014; Cheung et al. 2005; Amitai et al. 2015; Adegbite-Adeniyi et al. 2012; Gibbons et al. 2012). A meta-analysis of adverse event reports (AERs) of suicidal thoughts and behaviors from 25 randomized controlled trials revealed an overall odds ratio (OR) of 1.78 (95 % CI 1.14–2.77), indicating that the rate of suicidal thoughts and behaviors was significantly higher in children randomized to active AD treatment relative to placebo (Bridge et al. 2007; Hammad et al. 2006).

Pharmacogenetic approach to predict AD response is a promising avenue toward individualizing AD treatment (Lohoff and Ferraro 2010; Clark et al. 2012). However, to date there is no robust and consistent evidence implicating any specific candidate gene or polymorphism that is associated with AD response or side effect (SE) profile in pediatric population (Blazquez et al. 2012; Murphy et al. 2003). Preliminary pharmacogenetic research has suggested an important role for genes related to serotonin function in AD SEs (Kato and Serretti 2010; Murphy et al. 2003; Serretti and Artioli 2004). The serotonergic system is involved in the regulation of a variety of physiological functions relevant to the therapy of depression and anxiety.

In this study, we used a pharmacogenetic approach to search for genetic variations that may affect the susceptibility for citalopram-related SEs [to note, we reported on suicidal adverse events and pharmacogenetics elsewhere (Kronenberg et al. 2007)]. Here we test the hypothesis that genetic variants in serotonin (5-HT)-related candidate genes may confer a protective or a risk effect on citalopram-related adverse events [to note: genetic variants in two serotonin-related genes, serotonin transporter (5-HTTLPR) and tryptophan hydroxylase 2 (TPH2), were described previously by our group (Kronenberg et al. 2007; Rotberg et al. 2013)]. The present study focuses on the association between candidate gene variants of three serotonergic genes (5-HTR2A, 5-HTR1Dβ, 5-HTR2C) and the emergence of adverse events during citalopram treatment. Such a relevant and clinically important analysis was not performed in our two previous studies (Kronenberg et al. 2007; Rotberg et al. 2013).

Methods

Subjects

The data presented herein were collected as part of an 8-week, open-label effectiveness trial that has been described elsewhere (Kronenberg et al. 2007; Schirman et al. 2010). Our eligibility criteria included: age between 7 and 18 years, Israeli Jewish descent and a diagnosis of major depression/dysthymia or anxiety disorder according to the diagnostic and Statistical Manual of Mental Disorders, 4th edition, text revision (DSM-IV-TR) criteria. The diagnosis had to be of at least moderate severity with a clinical global impressions-severity (GCI-S) score of ≥4, a level consistent with accepted guidelines for the use of ADs in children and adolescents (Bernstein and Shaw 1997; Birmaher et al. 1998). Exclusionary criteria included intellectual disability, organic brain syndrome, autistic spectrum disorders, history of hypomania or mania, psychosis, eating-disorders, attention deficit hyperactivity disorder (ADHD) and substance abuse. We focused on treatment-naïve pediatric population in an attempt to recruit a homogeneous sample with regard to previous exposure to ADs. In addition no co-medication was permitted throughout the study period (8 weeks). Patients, who were receiving psychotherapy at the time of study recruitment, were also excluded to evaluate the sole impact of medication.

Evaluation

Subjects were assessed at intake, 2, 4, 6, and 8 weeks. Diagnostic evaluation was conducted using the Hebrew version of the schedule for affective disorders and schizophrenia for school age children–present and lifetime version (K-SADS-PL) (Kaufman et al. 1997; Shanee et al. 1997). Overall clinical severity and improvement were assessed using the CGI-S and improvement (CGI-I) subscales, respectively (Guy 1976). Continuous measures of depression and anxiety were obtained using the children’s depression rating scale-revised (CDRS-R) (Poznanski et al. 1985) and the screen for child anxiety related emotional disorders (SCARED) (Birmaher et al. 1997, 1999). To note, diagnostic tools have been described in detail elsewhere (Kronenberg et al. 2007). In addition, SEs were rated by the clinician using the SSRI Side effect Profile Inventory (SEPI), a questionnaire devised specifically for this study to detect the possible SEs and their severity. SEPI comprises of the 24 most common SEs known for SSRIs and was described previously (Kronenberg et al. 2007). The list of side effects was developed from those reported in recent literature on the use of SSRIs in children (see Kronenberg et al. 2007). Each SE was rated on a 5-point scale from 0 (none) to 4 (very severe). For details see Table 1 in the supplementary data.

Procedure

The study was approved by the Schneider Children’s Hospital Review Board and the Israeli Ministry of Health Committee of Studies in human genetics. Informed consent was obtained from all participants and their parents. After a confirmatory diagnostic assessment, all subjects received citalopram. All subjects visited a psychiatrist (SK) once a week. SEs were evaluated weekly by direct questioning. Starting dosage for all patients was citalopram 10 mg/day for 1 week, and then increased to 20 mg/day through week 4. If the degree of improvement was minimal (CGI-I < 3) then the dosage of citalopram was increased from 20 to 40 mg/day on week 5. Both patients and clinicians were blind to the genotype.

Genetic assays

Candidate gene selection

In this study, three genes (5-HTR2A, 5-HTR1Dβ, 5-HTR2C) considered important in various neurobiological pathways associated with serotonergic activity, were chosen. Before the first dose of citalopram, 9 mL of whole blood was collected and DNA was extracted using a commercial kit (Roche Diagnostics, Basel, Switzerland). Genotyping of the most studied T102C polymorphism in 5-HTR2A gene (rs6313) was carried out using a modified version of a method described by Warren et al. (1993). Genotyping of the 5-HTR1Dβ G861C polymorphism (rs6296) was detected and determined using a modified version of a method described by Lappalainen et al. (1995a). Genotyping of the 5-HTR2C G68C nonsynonymous (Ser23Cys) polymorphism (rs6318) was carried out using a modified version of a method described by Lappalainen et al. (1995b). Research indicates that the Ser 23 allele is constitutively more active than the Cys 23 allele, with allele-specific analysis of receptor function showing functional differences in the activity of the Ser 23 minor allele relative to the Cys 23 allele (Brummett et al. 2013).

Data analysis

We examined the relationship between genotype and SEs in all the children who agreed to participate in the study. For statistical analysis, a SE is considered to be: (1) any report of at least moderate severity, or (2) at least two reports of mild severity, or (3) any SE that was reported on the last follow-up before a patient prematurely withdrew from the study, regardless of the reason for withdrawal. The relationship between genotype and SEs was assessed using standard univariate tests (χ 2). All tests were two-tailed and the significance level was set at 0.05. Bonferroni correction was used in order to minimize possible false positive results. Also, in order for an association to be considered as putative, the heterozygote genotype had to indicate an intermediate level of association as compared to the two homozygote genotypes.

Results

Study population

In all, 107 subjects met the eligibility criteria, of whom 95 agreed to participate in the trial. Of these, 87 subjects were included in the final analysis. Eight subjects were not included due to protocol breach or dropout before the first follow-up was conducted. Ten subjects discontinued/dropped-out because of SEs and were included in the final analysis. No differences were found between the subjects who were excluded from the final analysis and the subjects that were included with respect to sex, age, diagnosis and severity of diagnosis at baseline.

Demographic and clinical characteristics of the study sample are presented in Table 1. The sample comprised 44 % males. Mean age of the sample was 14.07 ± 2.67 years. Twenty-nine children (33 %) were under the age of 13. With regard to ethnic background, the proportion of subjects with 2 or more Ashkenazi grandparents was 65 %. With respect to diagnosis, 36 (41 %) had anxiety alone, 30 (35 %) had depressive disorder and 21 (24 %) had comorbid depression and anxiety disorders. Out of the children with depression, 24 had MDD, 4 had dysthymia and 2 had double depression. Anxiety here refers to any anxiety disorder (according to DSM-IV-TR): generalized anxiety disorder (N = 24), separation anxiety disorder (N = 20), social anxiety disorder (N = 14), panic disorder (N = 5), specific phobia (N = 4) and obsessive compulsive disorder (N = 4). Most of the children had more than one anxiety disorder. For forty children in the sample (47 %), the final drug dose was 40 mg/day.

Table 1 Demographic and clinical characteristics of the study sample (N = 87)
Full size table

SEs per gender, age, disorder type, drug dose

SEs were grouped into the following categories: neurologic SE, gastric SE, activation group SE, sleeping disorders and sexual SE. For the rates of the different SEs that were reported during the course of the treatment (see Table 2). Most of these side effects were reported in our previous publication (Kronenberg et al. 2007).

Table 2 The 24 side effect (SE) profile of the study population as detected by the SSRI Side effect Profile Inventory (SEPI)
Full size table

As stated in the population and demographics subsection, ten patients withdrew from the study because of SEs; four patients developed akathisia-like symptoms, i.e., severe psychological or motor agitation; three patients reported perceptual distortions or hallucinations; two patients developed hypomania; and one patient reported severe gastrointestinal discomfort and insomnia. However, most reports of SEs were mild to moderate and did not affect adherence to the drug.

Agitation was more common in boys (male:female 42.1 vs. 18.7 %, χ 2 = 5.61, df = 1, p = 0.018, see Fig. 1). No correlation was found with age or final citalopram dose.

Fig. 1
figure 1

The association between agitation and gender in the study sample. *p < 0.05—boys vs girls (agitation was significantly more common in boys)

Full size image

No other correlation was found between SE profile and gender, age, diagnosis or dose. No differences were found between the groups with concomitant MDD and anxiety disorders compared to those with a sole MDD or anxiety disorder regarding age, gender, final citalopram dose or SE profile (data not shown).

SEs by genotype

The distribution of the various SNPs assessed in the study sample is described in Table 2 in the Supplement data.

Higher rates of activation SE were observed in those with the 5-HTR1Dβ (rs6296) CC genotype compared to those with CG or GG (CC:CG:GG 71.4 vs. 39.3 vs. 22.7 %, χ 2 = 7.34, df = 2, p = 0.025). Similarly, higher rates of agitation were detected in those with the 5-HTR1Dβ CC genotype compared to those with CG or GG (CC:CG:GG 71.4 vs. 33.3 vs. 18.1 %, χ 2 = 8.99, df = 2, p = 0.011; see Fig. 2). This difference remained significant after Bonferroni correction for the three serotonergic genes.

Fig. 2
figure 2

The association between polymorphism of the 5-HTR1Dβ genotype (rs6296) and agitation. *p < 0.05 (remains significant after Bonferroni correction): Higher rates of agitation were detected in those with the 5-HTR1Dβ CC genotype compared to those with CG or GG

Full size image

All other differences between variations in genotype and SE were not significant following Bonferroni correction.

Discussion

SSRIs are among the most widely prescribed medications for the treatment of MDD and anxiety in youth, despite reports that they are associated with an increased incidence of a number of serious SEs, including suicidality, aggression, and violence, as well as AIM. Of particular, concern is the increased incidence of aggression observed in children and adolescents treated with SSRIs, since the temporal onset of these adverse effects often occurs within weeks of initiating SSRI treatment and may be related to SSRI induces suicidality (Amitai et al. 2015).

Behavioral activation was described in the majority of clinical trials studying the use of SSRIs in youth (Cheung et al. 2005). Although not all trials described all aforementioned adverse events, aggression, agitation and/or behavioral activation were reported in the majority of the papers. The rates of these adverse events in the published studies range from 1 to 8 %. Only a small proportion (less than 10 %) of these events were classified as a ‘serious adverse event’ or led to discontinuation from the trials. In case reports of behavioral activation associated with SSRIs, these adverse events occurred in patients with comorbid diagnoses of anxiety or externalizing disorders and in patients who had good initial responses to the AD (Cheung et al. 2005). Behavioral activation generally developed within days to weeks after the start of AD treatment. In the majority of these cases, the symptoms also resolved shortly (within 4 weeks) after the lowering of the dosage or the discontinuation of the medication without addition an antipsychotic or a mood stabilizer. These adverse events also occurred at lower dosages. A recent meta-analysis reported activation symptoms strongly trended toward being associated with AD treatment (OR 1.86). To note, this meta-analysis did not include citalopram as one of the studied ADs (Strawn et al. 2015).

To date, we found no published studies regarding the influence of genes on SE profile in children. Pharmacogenetic approaches are particularly relevant in pediatric populations due to several reasons: first, most studies of the pharmacodynamics of ADs have been conducted in adults, making studies of the pharmacogenetics of SSRI in children scarce (Blazquez et al. 2012). Second, pharmacogenetic approaches are particularly relevant in pediatric populations because of concerns about the suicide risk of SSRIs, resulting in a black-box warning. There is great hope that in the future it will be possible to personalize psychiatric treatments based on genetic profiles and thus moderate this high-risk in pediatric population.

Till now, there is an unresolved debate regarding the serious phenomenon of SSRI-induced suicidality and its etiology, frequency and its relation to other SSRI-induced SE’s (Amitai et al. 2015). There is controversy in the literature regarding the association between suicidality, mainly, suicidal ideation and not complete suicide, and impulsive–aggressive behaviors (IABs) associated with SSRIs; some researchers believe that the use of SSRIs is linked with outbursts of anger and aggression and these are, in fact, related to suicidality, while others believe that suicidality and aggression are two distinct phenomena. According to the first school, IABs are regarded as possible suicide intermediate phenotypes, mediating the relationship between genes and suicide outcome. Suicide is thought to result from the interaction of different factors. Among these, substantial work has pointed to the role of IABs increasing suicide risk (Zouk et al. 2007). As IABs are heritable (Seroczynski et al. 1999), exist independently from suicidal behavior, cosegregate in families with suicidal behavior and seem to increase recurrence of suicidal behavior in relatives of probands selected for high levels of IABs (Brent et al. 1996; Johnson et al. 1999; Kim et al. 2005; Popova 2006). It has been suggested that IABs are suicide intermediate phenotypes. As such, IABs would mediate either directly or indirectly, the association between genetic factors and suicide outcome (Van Praag 2000).

In adults, it was shown that the HTR2A CC genotype (rs6313) was strongly associated with discontinuation due to intolerance in patients treated with paroxetine (Murphy et al. 2003). Recently, Hofer et al. investigated polymorphisms in the 5-HTR1A and 5-HTR2A genes, including rs6313, and found no association between them and suicide risk (Hofer et al. 2016).

In this prospective open treatment trial of citalopram for pediatric anxiety and/or depression, we investigated a possible correlation between citalopram adverse events and several candidate genes related to the serotonergic system. In contrast to Murphy et al’s findings (Murphy et al. 2003), we did not find the HTR2A CC genotype related to the SE profile of citalopram in children. Our most interesting finding was that 5-HTR1Dβ polymorphism (rs6296) was positively associated with irritability, i.e., subjects with the CC genotype showed higher levels of agitation as opposed to subjects with the CG/GG genotype. We found no association between this polymorphism and elevated mood or elevated anxiety, which are also considered part of the constellation of symptoms known as behavioral activation (Amitai et al. 2015).

The 5-HTR1Dβ gene is located on the long arm of chromosome 6. A studied 5-HTR1Dβ single base G/C substitution at nucleotide 861 in the coding region (rs6296) is synonymous, i.e., this substitution does not change amino acid (Val287Val). Nevertheless, the synonymous variant may be important too, if the SNP is located in linkage disequilibrium with functional polymorphism or causes functional RNA structure changes for example (Nackley et al. 2006).

Considering the important role of 5-HTR1Dβ receptors in the control of serotonin release, and given that serotonin is one of the major neurotransmitters involved in the regulation of behavior, this polymorphism could provide a valuable tool for studying genetic linkage of the 5-HTRlDβ receptor gene to various disorders in which serotonin is postulated to be involved. Of note, several pharmacological studies exclude a direct involvement of the 5-HTRlDβ receptor in the pathogenesis of bipolar disorder (Mundo et al. 2001), which may highly relate to SSRI side effects (Strawn et al. 2014). Further investigations combining genetic and pharmacological strategies are warranted.

Thus, our study suggests an intermediate phenotype between citalopram induced suicidality and aggression, via agitation SE. Our results also indicate that pharmacodynamic effects such as receptor variation can be more important than pharmacokinetic factors. We expected patients with genotypes encoding impaired metabolism to show more severe adverse events and more discontinuations.

Limitations

This study has several limitations, shared with other pharmacogenetic investigations. Among them are the relatively small sample size, the lack of a placebo arm, the open-label nature of the study and the potential for population stratification. There was no pharmacological documentation of adherence or drug concentration, so that nonresponse could be misattributed to genotype when it could be due to lack of adherence or rapid clearance. Unfortunately, we did not measure citalopram blood concentrations and we did not assess the influence of SNPs in genes encoding for oxidative enzymes which may affect the drugs’ concentration. More SNPs should be screened to avoid false negative findings (see, for instance Altar et al. 2015). It is of note that SSRIs have a common mechanism of serotonergic action but they are different medications in terms of their pharmacodynamic and pharmacokinetic characteristics, and thus, our results are relevant only to citalopram treatment and cannot be generalized to all SSRIs.

Conclusion

Our results indicate that a variation in the 5-HTR1Dβ gene (CC genotype) may be important in the vulnerability to develop agitation/activation as SE of citalopram treatment in children and adolescents with MDD/dysthymia and/or anxiety disorders. It has been suggested that agitation/activation and suicidal behavior are intermediate phenotypes. As such, emergence of agitation/agitation may mediate either directly or indirectly, the association between genetic factors and suicidal behavior in citalopram treated children. Nevertheless, 5-HTR1Dβ polymorphism may be a risk genetic marker of aggression and agitation adverse events in children and adolescents with depression and/or anxiety treated with citalopram. If confirmed, it might be possible to identify patients who are at risk and avoid citalopram treatment or to administer lower doses in an attempt to diminish the risk of iatrogenic agitation. Moreover, if replicated, the finding may help in understanding the mechanism of agitation/activation in response to citalopram and lead to treatment strategy which may reduce the rate of this harmful adverse effect.