Deprescribing Antidepressants in Children and Adolescents: A Systematic Review of Discontinuation Approaches,Cross-Titration,and Withdrawal Symptoms

Citalopram

In a small 24-week multisite randomized placebo-controlled discontinuation study of adolescents (13–18) who responded to 12 weeks of open-label citalopram (N = 25), the time to relapse was not significantly different between groups (Cheung et al., 2016). In this trial, 75% of adolescents continuing citalopram and 62% of those who switched to placebo were without relapse during the 24-week study. Adverse effects that were significantly more common in patients who stopped citalopram included palpitations (p = 0.047) and derealization/disinterest (p = 0.019). Additional withdrawal-emergent symptoms included tension, anxiety, nausea/vomiting, and cough (Cheung et al., 2016).

Similar to escitalopram, pharmacokinetic studies of citalopram in youth suggest that its clearance is related to CYP2C19 metabolism, although this has also not been evaluated in controlled studies. Hyperbolic discontinuation (i.e., reducing dose at fixed percentages) has been proposed, although gradual discontinuation, such as that described by Cheung and colleagues (2016), represents a reasonable approach. Regarding tapering and cross-titration, when switching from citalopram to a short half-life medication, clinicians may consider initiating and titrating the new medication before discontinuing citalopram (Fig. 3).

OPEN IN VIEWER

FIG. 3.

Cross-titration strategies for children and adolescents based on pharmacokinetic and pharmacodynamic properties of the first and second antidepressant. These strategies assume good tolerability of the first medication and should be reevaluated if symptoms of serotonergic toxicity or new side effects emerge during the cross-titration. In addition, clinical, developmental, and psychosocial factors should always be considered during cross-titration and deprescribing.

In the absence of prospective data, cross-titration from citalopram to another SSRI in children and adolescents without side effects should consider the time to steady state and time to achieve a “goal dose” for the second medication and the pharmacokinetics of citalopram, as described above (Fig. 3).

Escitalopram

While few studies have addressed the discontinuation of escitalopram in children and adolescents following acute treatment (Emslie et al., 2009; Saito et al., 2023; Strawn et al., 2023b, 2020), the registration trial for generalized anxiety disorder in children and adolescents included a 1-week taper, during which time youth who were treated with 20 mg daily were downtitrated to 10 mg daily and then discontinued. During this 1-week period, 8% of youth discontinuing escitalopram and 7% of those continuing placebo experienced new adverse events. One patient in each group experienced new suicidal ideation. In contrast, abdominal pain (2% vs. 0%), dizziness (4% vs. 2%), and dysesthesias (2% vs. 1%) were more common in patients discontinuing escitalopram compared with those remaining on placebo; however, these did not significantly differ between groups (Strawn et al., 2023b).

Escitalopram, as the active s-enantiomer of citalopram, corresponds to dosages that are effectively 50% of citalopram, but the pharmacokinetic parameters of the active component (s-citalopram), including half-life, are not expected to differ between escitalopram and citalopram. Pharmacokinetic modeling studies have described strategies for discontinuing escitalopram (Strawn et al., 2023a), although these studies have not considered how discontinuation strategies should be adjusted based on differences in escitalopram metabolism. Importantly, escitalopram clearance in youth is related to CYP2C19 metabolism (Strawn et al., 2019), with faster metabolizers having greater clearance and shorter half-life. Thus, slower CYP2C19 metabolizers may tolerate more rapid reductions in dose, whereas faster metabolizers could require slower discontinuation, although these approaches have not been systematically evaluated.

Studies in adults suggest discontinuation in a hyperbolic manner (Horowitz and Taylor, 2019), although this approach has not been formally evaluated in youth. In the absence of prospective data, cross-titration from escitalopram to another SSRI in children and adolescents without side effects should consider the time to steady state and time to achieve a “goal dose” for the second medication and the pharmacokinetics of escitalopram. For example, if switching to fluoxetine, initiating and titrating fluoxetine before discontinuing escitalopram would minimize the time a patient is “subtherapeutic” for both fluoxetine and escitalopram. However, a shorter cross-titration would be needed to switch to an SSRI with a shorter half-life, which requires less time to achieve a steady state (Fig. 3).

Paroxetine

In one double-blind discontinuation study of youth with major depressive disorder who were treated with paroxetine or placebo, significantly more adverse events were reported in patients discontinuing paroxetine compared with those transitioning from placebo to placebo. In this sample, discontinuation-emergent suicidality occurred in 2 of 19 patients discontinuing paroxetine compared with none of the nine patients discontinuing placebo. In addition, 15 of 19 patients experienced discontinuation-emergent worsening psychiatric symptoms compared with one of nine patients transitioning from placebo to placebo. Finally, akathisia and “withdrawal symptoms” were noted in youth discontinuing paroxetine but not in those discontinuing placebo.

A separate study of paroxetine discontinuation in children and adolescents with social anxiety disorder found that, following 16 weeks of either paroxetine or placebo, those patients who transitioned from paroxetine to placebo (compared with those transitioning from placebo to placebo) had more discontinuation-related adverse effects (Wagner et al., 2004). In this study, youth treated with >20 mg/day discontinued by tapering medication by 10 mg/day weekly over a maximum of 4 weeks. Among patients transitioning from paroxetine to placebo, 68/144 experienced discontinuation-related adverse events compared with 42/129. The most common withdrawal-emergent adverse events included nausea (16/144 vs. 3/129 for paroxetine and placebo, respectively), dizziness (16/144 vs. 2/129 for paroxetine and placebo, respectively), and abdominal pain (10/144 vs. 2/129 for paroxetine and placebo respectively) (Wagner et al., 2004).

In another study, children and adolescents treated with paroxetine (10 mg/day for 2 weeks, then 20 mg/day for 2 weeks, and then 30 mg/day for 2 weeks), dose-related increases in plasma concentrations are supraproportional in both children and adolescents (Findling et al., 2006). In this study, clearance and volume of distribution were highly dependent on paroxetine dose and CYP2D6 genotype (as well as weight) but on not age or sex and, at low doses, ostensibly before mechanism-based inhibition dominated paroxetine kinetics, there was high intraindividual variability, but this decreased as paroxetine was titrated (Findling et al., 2006).

Paroxetine—which has a markedly shorter half-life in youth than adults (Findling et al., 1999)—has remarkably complex, nonlinear pharmacokinetics and exhibits mechanism-based inhibition of CYP2D6. Paroxetine is metabolized by CYP2D6, but during this metabolism, it forms a reactive intermediate that covalently binds to the enzyme’s active site. The binding of paroxetine reactive intermediate irreversibly inactivates CYP2D6, reducing its activity (Bertelsen et al., 2003). Thus, as paroxetine is titrated, CYP2D6 activity decreases, resulting in a lower clearance at higher doses (Findling et al., 2006). However, this process accelerates as the dose is reduced. As the paroxetine dose is reduced, its clearance also increases, including in youth, which may amplify withdrawal symptoms.

Paroxetine should be discontinued slowly in youth, given its short single-dose plasma half-life of 10 hours and long plasma half-life at steady-state dosing (Table 2, Fig. 3). The rate of elimination of paroxetine from the brain in adults estimated by fluorine magnetic resonance spectroscopy (¹⁹F MRS) is equivalent to peripheral clearance, with paroxetine plus metabolite resonance dropping by 62% after a 3-day substitution of placebo from steady state, with the magnitude of brain paroxetine decrease correlating with adverse events (Henry et al., 2000). In addition, because paroxetine clearance is related to the dose (through mechanism-based inhibition), titration may be slowed as terminal doses are reached. Also, because clearance may be increased in children and adolescents at lower doses (Findling et al., 2006), it may be necessary to administer paroxetine BID or TID during the final stages of a taper. In addition, because of its anticholinergic and antihistaminergic effects, mild cholinergic rebound is possible during paroxetine withdrawal, as is some insomnia, anxiety, and agitation related to the loss of histamine-1 blockade. If cross-titrating to other CYP2D6-metabolized medications (e.g., fluoxetine or venlafaxine), the uptitration should be slow.

Fluoxetine

The risk and timing of relapse following fluoxetine discontinuation were studied in a sample (N = 102) of treatment-responsive children and adolescents 7–18 years with major depressive disorder (mean dose = 29 mg/day), who were randomly assigned to placebo without taper versus continuation of active treatment for an additional 6 months. The relapse rate was significantly higher in the placebo group (69%; risk ratio: 2.1) versus fluoxetine (42%). The median time to relapse, as defined by significant clinical deterioration, was shorter in the placebo group (8 weeks), as was the median time to full relapse (14 weeks) compared with fluoxetine, for which time to relapse was unable to be determined as 50% of the sample did not relapse by the completion of the study (Emslie et al., 2008). Adverse events did not differ between randomized groups during the follow-up phase. In a 32-week relapse prevention phase of a double-blind, placebo-controlled, 51-week study, 20 patients continued fluoxetine at their fixed dose, and 20 were transitioned to placebo. This study switched patients directly from their fluoxetine dose (20–60 mg) to placebo without taper. Relapse occurred more quickly in patients who transitioned from fluoxetine to placebo than those who continued fluoxetine (p = 0.046) and was more common in patients who switched to placebo (34% vs. 60%). Similar to the earlier report, adverse events and tolerability were statistically similar in the F/F and F/P groups, suggesting that fluoxetine is less associated with significant discontinuation events. However, some withdrawal-emergent side effects were more common in patients switching to placebo compared with those continuing fluoxetine (headache: 15% vs. 5%, depression 10% vs. 0%; anxiety 10% vs. 0%; diarrhea 10% vs. 0%; dizziness 10% vs. 0%; myalgia 10% vs. 0%) (Emslie et al., 2004).

Although pharmacokinetic studies of fluoxetine in youth are limited, data suggest a similar profile to that of adults (Wilens et al., 2002). Of commonly prescribed antidepressants, fluoxetine has the longest half-life, ranging from 4 to 6 days, and its active metabolite, norfluoxetine, has a half-life ranging from 7 to 15 days (Wilens et al., 2002). Fluoxetine is N-demethylated to norfluoxetine by CYP2D6, and other isoenzymes, CYP2C19, CYP2C9, and CYP3A4, play a significant role in metabolism. Fluoxetine and s-norfluoxetine are potent inhibitors of CYP2D6 (competitive and noncompetitive) (Deodhar et al., 2021; Liston et al., 2002) and CYP2C19 (mechanism-based inhibition) (Deodhar et al., 2021) and moderately inhibit CYP3A4 and CYP2C9. Due to the saturation of CYP2D6, fluoxetine displays nonlinear kinetics, which accounts for increases in blood concentrations of fluoxetine that are disproportionate to dose titration. Moreover, the competitive inhibition of CYP2D6 in adults persists for 2–3 weeks following discontinuation, consistent with expected elimination (Liston et al., 2002). This is partly related to the high-affinity binding of fluoxetine and norfluoxetine to CYP2D6 and unbound fluoxetine and norfluoxetine levels circulating in the blood after discontinuation, secondary to their long elimination half-life. Using ¹⁹F MRS, brain concentrations of fluoxetine and fluorinated metabolites achieve similar levels in children as in adults, and estimates of the brain elimination half-life are similar to adults as well (16 days) (Bolo et al., 2000; Strauss et al., 2002). The slow elimination of fluoxetine from the brain and plasma contributes to the “self-tapering” of fluoxetine and has implications for the timing of cross-titration or the initiation of other psychotropic medications.

Regarding discontinuation and cross-titration, the persistent effects of fluoxetine and norfluoxetine on CYP2D6 and the long half-life generally explain the less frequent and delayed emergence of withdrawal symptoms, although not precluding their occurrence. Accordingly, in adults, headache has been reported significantly more often than placebo. At the same time, open-label studies note dizziness, lightheadedness, and dyskinesia. In general, however, discontinuation studies in adults suggest that withdrawal symptoms are mild or absent compared with other SSRIs, with abrupt discontinuation usually well-tolerated (Zajecka et al., 1998). Furthermore, in a study of adults, a 5-day interruption was not associated with discontinuation symptoms (Michelson et al., 2000). In the largest study of antidepressant discontinuation in adults, patients treated with fluoxetine 20 mg daily were discontinued by taking fluoxetine 20 mg every other day for a month and then discontinuing (Duffy et al., 2019).

In general, if no further therapy is planned, fluoxetine “self-tapers” so it can be abruptly discontinued from 20 mg/day with an observed gradual reduction in serum level over time (2–6 weeks). Pediatric studies involving patients treated with 40 mg daily have used a reduction to 20 mg and then discontinuation, and this approach appears appropriate for routine clinical practice (Fig. 3). Clinically, if changing to another medication, one approach is a washout period of 2–3 weeks followed by a recommended titration of the second antidepressant. However, a longer washout period might be appropriate if transitioning to other CYP2D6-metabolized medications, including paroxetine, venlafaxine, duloxetine, vortioxetine, and tricyclic antidepressants.

Sertraline

A 52-week multisite placebo-controlled randomized discontinuation study was conducted of treatment-responsive depressed adolescents (N = 22) aged 13–19 who completed open-label acute (12-week) and maintenance (24-week) treatment with sertraline. None of the subjects who transitioned to placebo remained without recurrence of depression, compared with 38% of those who continued sertraline. In the placebo group (N = 9), the most common adverse effects were runny nose (56%), rash/skin irritation (44%), inner tension (44%), anxiety/worry (44%), cough (33%), and less reported but still significant dry mouth (22%), nausea/vomiting (22%), diarrhea (22%), increased/decreased appetite (22%/22%), hypersomnia (11%), and abnormal movements (11%) (Cheung et al., 2008). In this study, sertraline was decreased by 25% each week for 4 weeks. Over 70% of those subjects randomized to placebo who relapsed or exited the study did so within approximately 12 weeks after substitution of placebo was begun.

In a 35-day, multiple-dosing pharmacokinetic study of sertraline in 61 children and adolescents with depression or OCD, AUC, C_max, and elimination half-life values were reported to be similar for children versus adolescents and comparable with adult data, with t_½ for sertraline of approximately 26 hours. Sertraline was titrated up to a maximum of 200 mg/day and then was openly and abruptly discontinued after 35 days. By a final pharmacokinetic sampling 9 days after discontinuation, sertraline concentrations were negligible, consistent with the empirical t_½ of 26 hours (Alderman et al., 1998). No data on possible withdrawal symptoms or clinical status were included. In children and adolescents, the half-life of sertraline is dependent upon CYP2C19 activity (Poweleit et al., 2023; Strawn et al., 2019) and CYP2B6; clearance is reduced with slower CYP2C19 metabolism (Poweleit et al., 2023). Moreover, patients with increased CYP2C19 metabolism (e.g., ultrarapid metabolizers) may be more likely to experience discontinuation symptoms due to the shortened half-life of sertraline relative to other phenotypes.

Hyperbolic discontinuation has been recommended for sertraline discontinuation in youth over several months, based on studies in adults (Strawn et al., 2023a). Alternatively, the single pediatric study to evaluate discontinuation of sertraline in depressed adolescents used reductions of 25% each week for 4 weeks (Cheung et al., 2008). Notably, the largest study of antidepressant discontinuation in adults, the Antidepressants to Prevent Relapse in Depression (ANTLER) study, reduced sertraline from 100 mg to 50 mg for 1 month and then to 25 mg every other day for a month (Duffy et al., 2019). However, this approach may not be appropriate for most adults and is likely inappropriate for youth, based on the pharmacokinetics of sertraline (Poweleit et al., 2023) and the rapid relapse in depressed youth, although with a more accelerated taper (Cheung et al., 2008).

Fluvoxamine

Fluvoxamine has been systematically evaluated in pediatric patients with OCD (Riddle et al., 2001; Riddle et al., 1996) and mixed anxiety disorders (Walkup et al., 2001). However, there are limited data related to the discontinuation of fluvoxamine in pediatric patients. In adults with panic disorder, one study evaluated fluvoxamine discontinuation and found that withdrawal symptoms were noted in nearly 86% of patients following abrupt discontinuation (Black et al., 1993). In this study of 14 adults, dizziness/incoordination, headaches, nausea, and irritability were observed, and these symptoms peaked 5 days after discontinuation. In addition, one subject had a recurrence of panic, and another developed depressive and anxiety symptoms (Black et al., 1993). Using ¹⁹F MRS, brain concentrations of fluvoxamine and metabolites in adults substituted with placebo after achieving steady state demonstrated a brain elimination half-life of 56 hours, 2.4 times longer than measured plasma elimination. Half of all subjects experienced withdrawal symptoms of sweating, headaches, and nausea approximately 72–120 hours after fluvoxamine discontinuation when brain fluvoxamine concentrations had dropped to 35% of peak levels (Strauss et al., 1998). Results suggest that brain elimination rates may be more informative than plasma pharmacokinetics regarding risks of differing discontinuation schedules for fluvoxamine. In another report, investigators noted that fluvoxamine brain concentrations at steady state were not significantly different when comparing a pediatric sample with adults after adjusting for body mass (Strauss et al., 2002).

Fluvoxamine is primarily metabolized to fluvoxamine acid (Perucca et al., 1994) via CYP2D6 (Miura and Ohkubo, 2007; Spigset et al., 2001), and CYP2D6-poor metabolizers have less clearance than normal metabolizers. Additional fluvoxamine metabolites are metabolized by CYP1A2 (Carrillo et al., 1996), and fluvoxamine itself inhibits CYP1A2. One multiple-dose pharmacokinetic study of fluvoxamine in children and adolescents has shown prominent age and gender differences. Children between 6 and 11 years were noted to demonstrate higher peak plasma concentration, higher mean AUC concentration–time curves, and lower oral clearance versus adolescents (12–17 years) (Labellarte et al., 2004). Also, female children were noted to show AUC mean plasma concentrations twice that of males and higher Cmax, with clearance rates half that of male children, even after controlling for body mass. In contrast, plasma concentration–time profiles from the adolescents (compared with adults) suggest similar pharmacokinetics for the two populations. However, clearance is approximately 50% greater in adolescents than adults (even after normalizing to body weight).

In the absence of prospective trials, discontinuing fluvoxamine in pediatric patients is often guided by empirical experience. However, given its short half-life, discontinuation in the outpatient setting may best be done over several months. In addition, because its clearance is faster in youth than adults, twice daily dosing should be used, including during tapering and even at low doses.

Vilazodone

Studies of vilazodone in children and adolescents are limited and generally do not support the efficacy of vilazodone for pediatric patients with major depressive disorder (Durgam et al., 2018a; Findling et al., 2020). Moreover, there are limited data regarding discontinuation symptoms or approaches. Vilazodone is primarily metabolized by CYP3A4 with minor contributions from CYP2C19 and CYP2D6. Patients taking concurrent CYP3A4 inhibitors or inducers have altered vilazodone exposure, but the pharmacogenetic variation in all three genes does not seem to have a substantial impact (Boinpally et al., 2014; Bousman et al., 2023).

In a double-blind discontinuation study of vilazodone in adults with major depressive disorder, newly emergent side effects were similar across treatment groups (placebo, vilazodone 20 mg/day, and vilazodone 40 mg/day) (Durgam et al., 2018b). In addition, one negative pediatric efficacy study of vilazodone as a treatment for major depressive disorder in youth did not identify significant withdrawal symptoms during a 1-week taper of the medication, in which patients treated with 30 mg/day were tapered in two steps to 15 mg, then 5 mg/day and then discontinued; patients treated with 15 mg/day were tapered in one step to 5 mg/day and then stopped (Findling et al., 2020). No tolerability or discontinuation-emergent withdrawal symptoms were reported in publicly available databases or the primary article.

In general, while withdrawal symptoms were generally not noted over short discontinuation periods, when monitoring for recrudescence of depressive or anxiety symptoms, longer discontinuation periods may be appropriate for vilazodone, and discontinuation of 8–10 weeks might be appropriate and consistent with data supporting longer discontinuation periods with other antidepressant medications.

Venlafaxine

Discontinuation symptoms associated with venlafaxine may be more severe than those associated with SSRIs or other SNRIs. In an 8-week randomized trial of depressed adults (N = 288) treated with venlafaxine extended-release (average dose: 95 mg/day) or escitalopram (average dose 12 mg/day), discontinuation symptoms were more frequent for venlafaxine compared with escitalopram (e.g., agitation, cognitive impairment, diaphoresis, dizziness, fatigue, nausea, restlessness, tremor, and imbalance). A second study of adults with generalized anxiety disorder evaluating extended-release venlafaxine (average dose 184 mg/day), duloxetine (108 mg/day), and placebo found that over a 2-week discontinuation period, discontinuation symptoms were greater for venlafaxine compared with placebo (27% vs. 15%) but were similar for duloxetine and placebo (19% and 15% of patients).

There are few prospective trials of venlafaxine discontinuation in pediatric patients. One open-label, flexible-dose study of venlafaxine ER in children and adolescents (N = 86) with MDD examined discontinuation symptoms over a 14-day taper after 6 months of active treatment (Emslie et al., 2008). Taper/poststudy-emergent adverse events were reported by 28 subjects (33%), and the most reported events were headache (12%) and dizziness (6%). One participant reported suicidal ideation 2 days after the last dose; another was hospitalized for a suicide attempt 2 days after study discontinuation; and a third reported hostility 2 days after the last dose. Contextualization of adverse event frequency is not available due to an open-label, noncomparative study design.

Venlafaxine is metabolized by CYP2D6 to the active metabolite o-desmethyl venlafaxine, and the concentration of o-desmethyl venlafaxine depends upon the CYP2D6 phenotype. Accordingly, the Clinical Pharmacogenetics Implementation Consortium (CPIC) recommends avoiding venlafaxine in poor and intermediate CYP2D6 metabolizers, and ultrarapid metabolizers may experience more noradrenergic side effects (e.g., flushing, greater increases in blood pressure or heart rate). In addition, the half-life of venlafaxine increases with age, with a t_1/2 of 8–10 hours in children 6–11 years and 11–14 hours in adolescents (Stahl and Strawn, 2024). This short half-life, relative to other antidepressants, is thought to potentiate withdrawal symptoms.

Taken together, the extant data from pediatric and adult studies suggest that discontinuation symptoms are common with venlafaxine and perhaps more severe compared with those of other antidepressants. While hyperbolic discontinuation may confer a clinically significant advantage over a linear taper for many antidepressants, this strategy is based mainly on SERT occupancy and may be inappropriate for venlafaxine. More conservative approaches may be warranted in youth, particularly given the potentially higher likelihood of treatment-emergent suicidality and shorter relative half-life. In general, discontinuation over 3 months with potentially longer tapers in some patients may be appropriate in youth, although this has not been prospectively studied.

Desvenlafaxine

An open-label, flexible-dose, extension study examining desvenlafaxine efficacy and tolerability in a sample of children and adolescents (N = 40) described discontinuation-associated adverse events. Desvenlafaxine was discontinued without taper for doses of ≤10 mg, with a 1-week taper for a dose of 25 mg daily, and with a 2-week taper for doses of ≥50 mg (Findling et al., 2014). During the taper period, one child reported headaches, and one adolescent reported depressive symptoms. In addition, in an analysis of desvenlafaxine discontinuation symptoms in adults discontinuing from fixed and flexibly dosed regimens, the most common withdrawal-related symptoms (≥5% of participants and compared with placebo) were dizziness (21.6% vs. 2.7%), irritability (9.5% vs. 4.3%), anxiety (5.6% vs. 1.6%), abnormal dreams (5.3% vs. 1.6%), fatigue (5.3% vs. 2.2%), headache (12.1% vs. 7%), nausea (14.2% vs. 4.9%), diarrhea (6.8% vs. 2.2%), and hyperhidrosis (5.3% vs. 0%). Adults taking desvenlafaxine scored significantly higher on the Discontinuation-Emergent Signs and Symptoms (DESS) checklist compared with those taking placebo (p values ≤0.028) (Montgomery et al., 2009). Notably, the discontinuation taper was relatively short (1 to 2 weeks), with adults taking 50 mg daily simply stopping the medication, while individuals taking 100 mg/day could either stop the medication or transition to 1 week of 50 mg/day and then discontinue. Patients treated with 200 mg/day decreased the desvenlafaxine dose to 100 mg for 1 week and then discontinued, and patients treated with 400 mg/day were tapered to 200 mg/day for 1 week and then 100 mg for 1 week and then discontinued the medication (Montgomery et al., 2009).

In pediatric patients, desvenlafaxine exhibits dose-related increases in C_MAX and AUC, demonstrating generally linear pharmacokinetics. However, children exhibit higher clearance rates compared with adolescents. This higher clearance, coupled with a short half-life similar to that of the parent compound venlafaxine, likely contributes to a higher risk of withdrawal symptoms. Despite this, the severity of desvenlafaxine-related withdrawal symptoms in youth appears to be less than that seen with venlafaxine, possibly due to the pharmacodynamic differences between the two drugs in terms of serotonergic versus noradrenergic activity.

Desvenlafaxine should be discontinued gradually. Although pediatric trials have discontinued desvenlafaxine at a dose of 50 mg/day, some patients may require tapering to 25 mg/day. Generally, discontinuing desvenlafaxine should take place over at least 4 weeks, and based on the withdrawal-emergent symptoms systematically evaluated in adults, patients treated with higher doses (≥200 mg/day) may require slower discontinuation (Fig. 3).

Duloxetine

In the registration trial for duloxetine in children and adolescents with GAD, a 2-week taper was included, during which patients treated with duloxetine 120 or 90 mg/day were transitioned to 60 mg/day for 1 week and then 30 mg/day for 1 week. Patients treated with 60 mg/day were transitioned to 30 mg/day for 1 week. During this period, one patient experienced suicidal ideation after discontinuing duloxetine and taking a placebo. However, during this taper, no patients treated with duloxetine or placebo reported suicidal behavior or nonsuicidal self-injurious behavior. Overall, suicidal ideation was reported in two duloxetine-treated patients and one placebo-treated patient. Other taper-emergent adverse events were headache (n = 2) and upper abdominal pain (n = 2) (Strawn et al., 2015). In adolescents with depression undergoing a 2-week taper from duloxetine (30–120 mg daily), discontinuation-related adverse events that were reported by more than one patient were headache, dizziness, insomnia/initial insomnia, paresthesia, upper abdominal pain, irritability, nausea, depression, and seasonal allergies. In contrast, only headaches were reported in more than one patient who discontinued fluoxetine in this trial (Emslie et al., 2015).

This deprescribing strategy was also used in a Phase 3 multicenter trial of children and adolescents with major depressive disorder (Emslie et al., 2014; March et al., 2009; Prakash et al., 2012). Finally, in the Japanese open-label study of adolescents with major depressive disorder, patients treated with 60 mg daily were transitioned to 40 mg daily for 1 week and then 20 mg daily for 1 week (Phase-3 clinical study of duloxetine hydrochloride in children and adolescent patients with depressive disorder: An open-label extension study, https://cdn.clinicaltrials.gov/large-docs/53/NCT03395353/Prot_000.pdf). A study of adolescents with major depressive disorder-transitioned patients who were taking duloxetine 120 mg daily to 60 mg daily for 1 week and then to 30 mg daily.

Duloxetine has a relatively short half-life (∼10 hours), and clearance is generally faster in pediatric patients than in adults. Steady-state concentrations are, on average, 30% lower in children and adolescents than in adults (Lobo et al., 2014). Moreover, duloxetine metabolism may be influenced by variations in CYP2D6 and CYP1A2, although pediatric studies have been underpowered to detect the impact of CYP2D6 genetic variation on duloxetine exposure.

Based on the discontinuation strategies used in the registration trials and the available pharmacokinetic data, we recommend slower discontinuation in children and adolescents. Specifically, we recommend 2–4-week intervals between dose reductions from 90 to 120 mg. In addition, cross-titration strategies should consider the clearance of duloxetine and the absorption and time to steady state of the second antidepressant, as described in Fig. 3.

Levomilnacipran

In two double-blind studies, children aged 7–17 with major depressive disorder were treated with placebo, fluoxetine (20 mg/d), or levomilnacipran (40–80 mg/day) for 8 weeks. In the taper period, all patients on levomilnacipran reduced their dose to 40 mg/day on days 1 and 2 and then 20 mg/day from day 3 through day 7 (Radecki et al., 2024). Withdrawal symptoms were not reported, and we could not locate the clinical study report on the Food and Drug Administration (FDA) archive. In adults, the incidence of adverse events during the double-blind down-taper period was approximately 10% in both treatment patients stopping the placebo and those discontinuing levomilnacipran (Sambunaris et al., 2014).

Pharmacokinetic data related to levomilnacipran have not been published, although a review of the FDA Multi-Disciplinary Review and Evaluation, part of the new drug application, for levomilnacipran revealed that the manufacturer updated adult population pharmacokinetic models “with the sparse [pharmacokinetic] data collected in the pediatric studies… the PPK analyses have not been extensively reviewed. Based on the Sponsor’s PPK simulations, the exposure levomilnacipran in children 7 to <12 years old and adolescents (12 to <18 years) were relatively similar but slightly lower compared to adults” (NDA 204168/S-10, accessed from: www.fda.gov/media/167755/download). The pharmacodynamics of levomilnacipran are unique among SNRIs in that it has a 15-fold greater selectivity for norepinephrine compared with serotonin reuptake inhibition (relative to duloxetine, desvenlafaxine, or venlafaxine) (Auclair et al., 2013). However, at high doses, serotonergic reuptake inhibition may occur. In addition, levomilnacipran lacks affinity for other receptors, including the dopaminergic, adrenergic, histaminic, and muscarinic receptors (Auclair et al., 2013), which suggests that withdrawal-related symptoms are primarily driven by a noradrenergic and, to a lesser extent, serotonergic mechanism.

Pediatric discontinuation data are lacking; however, based on the pharmacokinetics and approach used in registration trials, shorter discontinuation may be appropriate for levomilnacipran (e.g., discontinuation over 2–4 weeks).

Bupropion

Published data regarding bupropion discontinuation in children and adolescents were not available at the time of this review. In adults, discontinuation symptoms are uncommon (Koshino et al., 2013; Remick et al., 1982), and a report of the European Monitoring Agency’s drug monitoring database identified only 299 of 20,720 (1.44%) adverse drug reports associated with bupropion were related to withdrawal (Schifano and Chiappini, 2018). Nonetheless, two adult case reports describe discontinuation-emergent irritability, anxiety, headache, and myalgia for one patient and acute dystonia for another (Berigan, 2002; Wang et al., 2007). For both patients, symptoms resolved upon restarting bupropion.

Bupropion is primarily metabolized by CYP2B6 (Pearce et al., 2016). Bupropion and its three major metabolites are CYP2D6 inhibitors, although at present, there are no dosing alterations based on pharmacogenetic variation recommended. Bupropion displays linear pharmacokinetics in pediatric and adult patients (Daviss et al., 2005). A pharmacokinetic study of bupropion in depressed pediatric patients (N = 16) demonstrated that responders had a higher 24-hour mean area under plasma concentration time curves (AUC_0-24) for bupropion and its metabolites than nonresponders (Burleson Daviss et al., 2006). In another pharmacokinetic study of adolescents and adults, younger adolescents (12–14 years) had higher bupropion peak concentrations and AUC_0-t compared with older adolescents (15–17 years) and adults, resulting in a three- to fourfold increase in total exposure in younger adolescents, reasons for which were hypothesized to be multifactorial (Oh and Crean, 2015). These pharmacokinetic observations may be considered when predicting withdrawal symptoms and risk of relapse, although the exact clinical significance is uncertain.

In addition to pharmacokinetics, bupropion pharmacodynamics are relatively favorable from a withdrawal standpoint. Bupropion inhibits the reuptake of dopamine and norepinephrine, thus it may be discontinued more rapidly, over 1–2 weeks, compared with antidepressants with prominent serotonergic effects. However, tapering more slowly may still be helpful in terms of allowing for the detection of reemerging symptoms. As bupropion is a CYP2D6 inhibitor, cross-titration to a CYP2D6-metabolized medication (e.g., fluoxetine, paroxetine, venlafaxine) should be done cautiously.

Mirtazapine

One case report detailed a 15-year-old patient who experienced hypotension requiring vasopressor support 8 hours after mirtazapine and paroxetine were held for cardiac surgery. Upon restarting paroxetine and mirtazapine, hypotension resolved, and vasopressors were discontinued (Novak et al., 2008); however, in this report, the hypotension may have been related to other postoperative sequelae rather than discontinuation of antidepressants. Other case reports in adults report induction of hypomania as well as panic attacks upon mirtazapine discontinuation, the mechanism of which was suggested to be a resulting “noradrenergic hyperactivity” (Fauchère, 2004; Klesmer et al., 2000; MacCall and Callender, 1999).

In the ANTLER study, adults taking mirtazapine 30 mg took 15 mg daily for 1 month and then 15 mg every other day (Lewis et al., 2021). Although the study did not report discontinuation-emergent adverse effects for individual antidepressants, every-other-day dosing may increase the likelihood of withdrawal symptoms, as the mean half-life of mirtazapine is 22 hours (Davis and Wilde, 1996). Every-other-day dosing may lead to adherence challenges in children and adolescents; thus, it may be reasonable to consider dose reduction from 15 mg to 7.5 mg daily rather than 15 mg every other day. This approach may minimize the incidence of withdrawal symptoms, including rebound increases in appetite and insomnia, as well as anxiety, and allows for the reemerging symptoms of the disorder being treated to be recognized and managed. In addition, while data are limited, some patients may experience rebound insomnia as a result of discontinuation of the H₁ blockade associated with mirtazapine. Therefore, centrally acting H₁ antagonists could be considered when discontinuing mirtazapine should rebound insomnia or anxiety emerge (Table 3).

Vortioxetine

Vortioxetine discontinuation symptoms in pediatric patients have not been systematically evaluated. However, in adults, discontinuation symptoms are uncommon, unrelated to the discontinuation approach (abrupt vs. gradual), and similar to placebo (Baldwin et al., 2016; Boulenger et al., 2014; Siwek et al., 2021). A retrospective chart review of 263 vortioxetine-treated adult discontinuation symptoms occurred in 3% of patients. It included mood lability, irritability, nervousness, and agitation (Siwek et al., 2021) and were unrelated to dose or discontinuation duration (abrupt vs. gradual). In other studies, discontinuation symptoms, as measured by the DESS score, were comparable with placebo 2 weeks after abrupt discontinuation (Boulenger et al., 2014). The lack of significant withdrawal symptoms may be related to the longer half-life (66 hours) of vortioxetine and suggests that taper may not be necessary.

A multidose pharmacokinetic profile of vortioxetine in children and adolescents (N = 48) demonstrated reduced clearance and longer t_1/2 (60 hours vs. 47 hours) in children compared with adolescents (Findling et al., 2017). This may be explained, in part, by significant positive associations observed between weight and volume of distribution (p < 0.01) and age and oral clearance (p < 0.01). Vortioxetine is metabolized by several CYP450 and cytosolic enzymes to inactive metabolites. In the Findling study, mean vortioxetine clearance was lower in CYP2D6-poor metabolizers (n = 2), but clinical significance and impact on tolerability and discontinuation are unknown.

In the absence of data, vortioxetine may be decreased over 4 weeks, and there appears to be a low risk of withdrawal symptoms. In addition, it may be more rapidly discontinued when a clinician wishes to cross-titrate (Fig. 3).

Amitriptyline

The data on discontinuation of amitriptyline in pediatric patients largely derive from case reports. In one case, amitriptyline was tapered following 6 months of treatment with a dosage of 150 mg/d. Less than 120 hours after discontinuation, excessive dreaming and unrestful sleep symptoms were reported (Dilsaver et al., 1983). In another case report, abrupt discontinuation (following 4 weeks of treatment at 250 mg/d) was associated with discontinuation symptoms that persisted for 60 hours and included panic attacks and “cholinergic excess.” Importantly, reinitiation of amitriptyline at 100 mg/day ameliorated these symptoms (Gawin and Markoff, 1981). Finally, two cases of nausea and vomiting were reported following abrupt discontinuation. One case was treated with 0.8 mg of atropine (Dilsaver et al., 1983), while the other received supportive therapy (Gualtieri and Staye, 1979).

Amitriptyline is metabolized via CYP2C19 to nortriptyline, and CYP2D6 hydroxylates both nortriptyline and amitriptyline. As such, pharmacokinetic variation in these enzymes should be considered as amitriptyline is deprescribed and if cross-titrating to a medication that is metabolized by or inhibits these enzymes. In addition, the CPIC recommends that in CYP2C19-poor metabolizers, a 50% reduction of the recommended starting dose and use of therapeutic drug monitoring (TDM) to guide dose adjustments in amitriptyline-treated patients who are CYP2C19- or CYP2D6-poor metabolizers. TDM may also be relevant during deprescribing as these individuals will have markedly slower clearance of amitriptyline (Hicks et al., 2017), and monitoring may allow for faster discontinuation and ostensibly lessen the likelihood of discontinuation-emergent side effects.

While rarely used for depressive or anxiety disorders in contemporary pediatric practice, amitriptyline is commonly used for migraine prophylaxis. It is generally used at lower doses (≤50 mg/day) than typically used for the treatment of depression. At these low doses, tapering may be more abrupt, potentially over 7–10 days. However, when used for affective or anxiety disorders, at higher doses, a gradual taper may minimize the incidence of withdrawal symptoms and allow for the detection of reemerging symptoms. In addition, during this discontinuation period, cholinergic rebound and anxiety/insomnia related to loss of H₁ antagonism may be observed. These symptoms may be addressed by either resuming the previously prescribed dose and decreasing the dose more gradually or using centrally acting H₁ antagonists or peripherally acting anticholinergics (Table 3).

Imipramine

Withdrawal symptoms associated with imipramine may be more complex than for other antidepressants, including SSRIs and SNRIs. Imipramine discontinuation has been evaluated in several pediatric reports. One study systematically evaluated imipramine discontinuation in depressed adolescents who were treated for 8 weeks with imipramine (200–300 mg, n = 32), paroxetine (20–40 mg), or placebo (n = 9) (Le Noury et al., 2015). Blinded discontinuation of imipramine was associated with nine reports of cardiovascular symptoms (e.g., arrhythmia, bradycardia, chest pain, dizziness, and tachycardia) compared with no reports of cardiovascular adverse effects in the nine patients discontinuing placebo. Similarly, more gastrointestinal symptoms, including constipation, dry mouth, diarrhea, abdominal pain, nausea, and vomiting, were observed in patients discontinuing imipramine compared with those stopping placebo. In terms of psychiatric adverse effects associated with the discontinuation of paroxetine, one patient reported suicidal ideation, and one patient reported akathisia when discontinuing imipramine; however, the psychiatric adverse events that emerged upon discontinuation of imipramine were less common than those associated with paroxetine discontinuation in the same study. Finally, headaches were more common in patients discontinuing imipramine than those discontinuing placebo (seven events in 32 patients and none in the nine patients discontinuing placebo) (Le Noury et al., 2015). In a case report, following several missed doses of imipramine at 4.8 mg/day in 15 days, one boy developed symptoms of nausea, vomiting, headache, and lethargy. In another case, following abrupt cessation of imipramine dosed at 4.4 mg/d, another boy developed symptoms of abdominal distress, nausea, loss of appetite, social withdrawal, insomnia, and became increasingly agitated, aggressive, and assaultive. In both cases, after restarting imipramine, the symptoms halted. Imipramine was then decreased over a week. Symptoms reappeared but were much less severe (Petti and Law, 1981).

In an additional study, 22 children hospitalized for depression on a high dose of imipramine (mean dose = 138 mg/d) for a mean of 41.5 days were observed for withdrawal symptoms. The tapering period was 3–10 days, with a mean of 6.4 days. Symptoms were reported by mean total symptoms per 10-day period in pretreatment, treatment, and withdrawal phases. All subjects were reported to experience at least one withdrawal-emergent adverse event. The symptoms with the most significant increase in frequency during the withdrawal phase (p < 0.001) are gastrointestinal complaints (n = 141) and drowsiness/fatigue (n = 39). Other symptoms with a significant increase in frequency included decreased appetite (n = 41, p < 0.01), tearfulness (n = 75, p < .001), apathy/withdrawal (n = 25, p < 0.03), headaches (n = 13, p < 0.03), and agitation (n = 30, p < 0.05) (Law et al., 1981).

Imipramine is converted to desipramine, which inhibits norepinephrine reuptake, and both moieties are important with regard to deprescribing imipramine. Systemic exposure to imipramine is highly variable in pediatric patients (up to sixfold). CYP2D6 and CYP2C19 are involved in clearance and pharmacogenetic variation may contribute to the variability in exposure (Hicks et al., 2017; Weller et al., 1982). Response in depressed children is exposure-related and correlated with total imipramine+desipramine plasma levels, but not with plasma levels of imipramine alone (Preskorn et al., 1982). Furthermore, while the clearance of most TCAs is generally linear, the hydroxylation pathway becomes saturated at higher plasma concentrations for imipramine (and desipramine) (Rudorfer and Potter, 1999). Thus, the clearance of imipramine (and desipramine) may be more difficult to predict when crossing the threshold between “high” and “low” doses.

Discontinuation of imipramine should be gradual, as many patients experience withdrawal symptoms. Rebound anxiety and insomnia may occur due to the loss of H₁ antagonism, and gastrointestinal symptoms may emerge as a result of cholinergic rebound. Addressing these symptoms with centrally acting H₁ antagonists or peripherally acting anticholinergics can be appropriate (Table 3). In addition, if cholinergic rebound leads to diarrhea, loperamide may be helpful. In the studies described above, imipramine was discontinued over 3–10 days; however, this approach may have been too rapid. Considering imipramine’s pharmacodynamic profile, the high likelihood of histaminergic and cholinergic rebound, as well as the monoaminergic-related withdrawal symptoms, a more gradual tapering over 8–10 weeks may be better tolerated. This longer discontinuation period, based on clinical experience, could help mitigate withdrawal symptoms more effectively.

Clomipramine

In one study of adults, clomipramine was abruptly discontinued following 3 months of treatment (100–300 mg/d), and withdrawal-emergent symptoms included recrudescence of depression, suicidal ideation, increased temperature, and sweating (Diamond et al., 1989). In addition, in adults with OCD (N = 18) treated for 5–27 months with clomipramine (100–300 mg), relapse of OCD was seen in 16 of 18 patients by 7 weeks after double-blind discontinuation, with increased depressive symptoms seen in 11/18 patients. Double-blind discontinuation of clomipramine occurred over a 4-day period of 50% prior dose, and then placebo only with follow up at 1, 4, and 7 weeks postswitch (Pato et al., 1988). In this study, insomnia (67%), vivid dreams (83%), and irritability (56%) were noted in the first 4 weeks after discontinuation. Depressive symptoms began to increase by week 1 with OCD symptoms beginning to clearly rise by week 4 postclomipramine discontinuation. Treatment duration before discontinuation was unrelated to the frequency or severity of withdrawal-emergent symptoms (Pato et al., 1988). In a retrospective chart review of 352 patients treated with various serotonergic antidepressants, of whom 171 underwent drug taper and discontinuation, clomipramine was associated with the highest rate of withdrawal-emergent symptoms (31%) among all other medications, with paresthesia, irritability, and insomnia as strikingly elevated versus other serotonin reuptake inhibitors (Coupland et al., 1996). There are limited data related to clomipramine-related discontinuation symptoms in children and adolescents. In one study of children, adolescents, and adults with autism (DSM-III-R criteria) aged 6–23 years, substitution with 1 week of placebo was associated with decompensation in most patients (Gordon et al., 1993), although specific withdrawal-related symptoms are unclear.

Clomipramine is primarily metabolized by CYP2C19 (Balant-Gorgia et al., 1991; Nielsen et al., 1996; De Vos et al., 2011), in addition to CYP3A4 and CYP1A2 (Nielsen et al., 1996) while its metabolite norclomipramine and clomipramine itself are also metabolized by CYP2D6. Surprisingly, despite its frequent use in pediatric patients with OCD, pharmacokinetic data for clomipramine in youth have been difficult to locate. In adults, the elimination half-life is about 20–24 hours for clomipramine and 96 hours for norclomipramine (i.e., desmethyl clomipramine, DMCMI) (Balant-Gorgia et al., 1991; Marazziti et al. 2019).

Clomipramine’s complex pharmacology includes the parent compound’s robust serotonergic reuptake inhibition, potent noradrenergic reuptake inhibition by DMCMI, and antimuscarinic, antiadrenergic, and antihistaminic effects (Millan et al., 2001; Richelson and Nelson, 1984; Tatsumi et al., 1997). A PET study of serotonin transporter (5-HTT) occupancy of clomipramine and fluvoxamine reported that a single dose of 10 mg of clomipramine occupied 80% of 5-HTT receptors, which was equivalent to a single dose of 50 mg of fluvoxamine (Suhara et al., 2003). Coupled with these multiple pharmacologic effects, the shorter half-life of clomipramine appears to contribute to a noticeably higher risk of withdrawal-emergent effects, as evidenced by case reports of severe withdrawal reactions after discontinuation of chronic doses as low as 25 mg/day (Zemishlany et al., 1992). Therefore, discontinuation of clomipramine should be approached cautiously with more gradual and extended tapering schedules, even in the context of low-dose administration, or with more rapid cross-tapering to an alternative serotonergic antidepressant, depending on the clinical need for continued treatment. Given that therapeutic drug monitoring has received some support for determining adequate clomipramine dosing for OCD and depression (Marazziti et al., 2019; Mavissakalian et al., 1990), knowledge of prediscontinuation plasma levels may aid in determining the rapidity of dose reduction and timing of tapering plans.