Introduction
Antidepressant drugs play a major role in the treatment of depression through significant
improvement of acute symptoms and reduction in the risk of relapse [1]
[2]. However, only one-third of the patients with depression respond to the first-line
antidepressant treatment [3], which indicates the need of augmentation therapy (i. e., adding another psychotropic
drug to the current regimen) for this difficult-to-treat population. Evidence indicates
the efficacy of such augmentation therapy with several psychotropic drugs, such as
lithium [4], olanzapine [5]
[6]
[7], aripiprazole [8]
[9]
[10]
[11], quetiapine [12]
[13], and risperidone [14]
[15]. Hence, recent guidelines suggest the use of augmentation therapy with another psychotropic
drug, such as antipsychotics and mood stabilizers, if series of monotherapies with
antidepressant drugs are not successful [16]
[17]
[18]. Interestingly, while many efforts have been devoted to the investigation of add-on
treatment for depression, the issue remains unaddressed as to whether and how long
such augmentation therapy should be maintained in terms of efficacy and safety. The
lack of consensus on this issue may expose patients to drugs used for augmentation
for a longer period of time than necessary. This issue is especially important to
avoid the adverse events of psychotropic drugs used for augmentation therapy, including
motor [19]
[20], metabolic [21], cognitive [22], and cardiovascular adverse events [23] caused by antipsychotics, or thyroid dysfunction caused by lithium [24]. We therefore undertook a systematic review of randomized controlled trials (RCTs)
that examined the effects of the discontinuation of drugs used for augmentation therapy
on treatment outcomes in depression and conducted a meta-analysis to address this
relevant issue in clinical practice.
Methods
A study protocol was registered at PROSPERO before commencing data collection (Registration
number: CRD42018103621). The PRISMA (Preferred Reporting Items for Systematic Reviews
and Meta-Analyses) statement was followed to ensure transparent and complete reporting
(Table S1). The study protocol is available on reasonable request. Two independent
authors (H.K. and T.K.) conducted the literature search, assessed eligibility, and
extracted data. Any discrepancies during these procedures were resolved through discussion.
Study Selection
RCTs to examine the effects of discontinuing medications that had been introduced
to augment antidepressant treatment were identified. The MEDLINE (1950 to January
2020) and EMBASE (1950 to January 2020) databases were searched by using the following
search terms: depressi*AND (withdraw*OR*OR continu*) AND (combin*OR augment*OR adjunct*OR
cotreatment*OR coadministrat*). Unpublished trials were searched using Clinical Trials.gov
(http://clinicaltrials.gov/); the following search terms were used: “depression” for “condition or disease” and
“augmentation” for “other terms.” In addition, reference lists of relevant articles
identified in the electronic search for published trials were hand-searched for additional
trials.
Inclusion Criteria
Clinical trials fulfilling the following 2 conditions were included: (1) RCTs to examine
the effects of discontinuing versus continuing medications that had been introduced
to augment antidepressant treatment after failure to pharmacotherapy with antidepressants
alone, and (2) RCTs with more than half of the participants with the diagnosis of
depression according to study-defined diagnoses. If several publications were found
from the same investigators using overlapping samples, data from the publication with
the largest number of patients were included.
Outcome Parameters
The primary outcome of interest was overall treatment failure, which was defined as
withdrawal from the study for any reason. The secondary outcomes included study withdrawal
due to relapse defined by the individual studies and changes in symptomatology scores.
When those patients who relapsed stayed in the study until the endpoint according
to the protocol, they were considered to have withdrawn from the study due to relapse
in this analysis. Additionally, study withdrawal due to adverse events was analyzed
as a measure of tolerability. All these parameters were obtained based on an intention-to-treat
basis.
Data Extraction
Outcomes in terms of withdrawal from the study due to any reason, study-defined relapse,
adverse events, changes in symptomatology scores, hospitalization, and suicide were
extracted. Information on each adverse event was also extracted when available. Demographic
and clinical characteristics of the subjects such as age, sex, durations of illness
and treatment, study design, and interventions were extracted. Information regarding
the sources of funding was also collected. We used the WebPlotDigitizer software (available
at https://automeris.io/WebPlotDigitizer/) if the included studies provided the data only in the form of graphs. In the studies
where relevant data were unreported, we contacted the authors for additional data.
Data Analysis
Prior to the meta-analysis, risk of bias of the included studies was assessed using
the Cochrane risk of bias tool [25]. Meta-analyses were performed using the Review Manager software, version 5.3 (The
Cochrane Collaboration, http://ims.cochrane.org/revman). Mean difference (MD) was estimated for continuous outcomes. We calculated pooled
estimates of standardized mean difference (SMD) to compile different symptomatology
scales. The inverse-variance statistical method and random-effects model to adjust
for study heterogeneity were used in each estimation. Two-sided 95% confidence intervals
(CIs) were used to assess significance, according to whether the CIs included the
null value. The Mantel test and random-effects model were used to calculate pooled
estimates of risk ratios (RRs) for dichotomous outcomes. These analyses were also
performed in a subgroup of studies that included a stabilization period before randomization
and studies that did not use esketamine as an augmentation therapy, since esketamine
has a unique mechanism of action and efficacy compared to other drugs. Subgroup analyses
were also conducted for studies using lithium and second-generation antipsychotics,
respectively, when relevant data were available in 2 or more studies. Adverse events
that were assessed in 2 or more studies were meta-analyzed. Study heterogeneity was
quantified using the I
2 statistics [26], with I
2≥50% indicating significant heterogeneity. The possibility of publication bias was
assessed by visual inspection of funnel plots [27]. We used 2-tailed p-values of<0.05 to assess significance.
Results
Included Studies
The systematic literature search yielded 2957 reports; 7 of these studies (n=1672)
fulfilled the inclusion criteria and were therefore included in the meta-analysis
(n=841 for continuing augmentation therapy; n=831 for discontinuing the augmentation
therapy) ([Fig. 1]) [28]
[29]
[30]
[31]
[32]
[33]
[34]. The characteristics of these studies are shown in [Table 1] and [Table 2]. In 6 studies, patients were diagnosed with study-defined treatment-resistant depression
([Table 2]). Five studies in this meta-analysis included a stabilization period in which patients
were maintained on both antidepressants and augmentation drugs, after they exhibited
remission before the RCT phase [28]
[29]
[32]
[33]
[34] ([Table 2]). The stabilization period ranged between 2 weeks and 1 year; the study by Hardy
et al. [28] had the longest stabilization period of 1 year. Antidepressants used varied depending
on the studies; however, selective serotonin reuptake inhibitors were most frequently
prescribed. Lithium (n=2), risperidone (n=2), esketamine (n=1), olanzapine (n=1),
and edivoxetine (n=1) were used for augmentation therapy. Previous use of electroconvulsive
therapy before study entry was described only in 1 study [28].
Fig. 1 PRISMA flow diagram of the literature search. PRISMA, Preferred Reporting Items for
Systematic Reviews and Meta-analyses; RCT, randomized controlled therapy.
Table 1 Characteristics of included studies at the beginning of randomized controlled trail
phase.
|
First author (Year)
|
Augmentation drug
|
Sample size
|
Male sex
|
Age, yearsa
|
Age at first episode, yearsa
|
Number of previous episodesa
|
Duration of current episode, yearsa
|
MADRS total scorea
|
CGI-S scorea
|
Presence of psychotic features
|
|
Hardy (1997)
|
Lithium
|
6
|
0%
|
79 (6)
|
63 (12)
|
N/A
|
N/A
|
5.2 (2.2)
|
N/A
|
N/A
|
|
Placebo
|
6
|
50%
|
74 (5)
|
59 (8)
|
N/A
|
N/A
|
2.3 (2.1)
|
N/A
|
N/A
|
|
Bauer (2000)
|
Lithium
|
14
|
21%
|
47.4 (16.9)
|
43.0 (15.2)
|
1.7 (2.2)b
|
0.54 (0.34)
|
N/A
|
2.4 (0.5)
|
N/A
|
|
Placebo
|
15
|
60%
|
47.4 (11.4)
|
41.3 (10.9)
|
2.2 (1.0)b
|
0.68 (0.40)
|
N/A
|
2.6 (0.5)
|
N/A
|
|
Rapaport (2006)
|
Risperidone
|
122
|
28.7%
|
47.8 (11.4)
|
29.9 (12.6)
|
N/A
|
2.0 (3.7)
|
6.8 (4.7)
|
N/A
|
0.8%
|
|
Placebo
|
119
|
43.7%
|
48.4 (12.0)
|
30.8 (14.0)
|
N/A
|
2.0 (3.8)
|
8.1 (4.6)
|
N/A
|
3.4%
|
|
Alexopoulos (2008)
|
Risperidone
|
32
|
31%
|
62.3 (7.2)
|
38.1 (13.6)
|
N/A
|
1.4 (1.9)
|
9.2 (5.0)
|
N/A
|
3.1%
|
|
Placebo
|
31
|
58%
|
62.9 (7.3)
|
40.0 (17.3)
|
N/A
|
2.4 (5.7)
|
8.7 (5.2)
|
N/A
|
9.7%
|
|
Brunner (2014)
|
Olanzapine
|
221
|
35.3%
|
44.9 (11.3)
|
31.7 (12.4)
|
3.5 (4.2)
|
1.50 (2.66)
|
5.4 (3.8)
|
1.7 (0.7)
|
0%
|
|
NA
|
223
|
31.4%
|
44.1 (12.3)
|
31.7 (13.4)
|
3.8 (9.7)
|
1.75 (3.21)
|
5.4 (4.0)
|
1.7 (0.8)
|
0%
|
|
Oakes (2015)
|
Edivoxetine
|
294
|
24.1%
|
47.5 (11.9)
|
N/A
|
N/A
|
N/A
|
4.4 (2.9)
|
1.7 (0.7)
|
N/A
|
|
Placebo
|
292
|
22.6%
|
46.9 (12.3)
|
N/A
|
N/A
|
N/A
|
4.4 (2.9)
|
1.7 (0.7)
|
N/A
|
|
Dalyc (2019)
|
Esketamine
|
90
|
35.6%
|
45.2 (12.12)
|
32.5 (11.42)e
|
N/A
|
2.15 (3.29)
|
3.7 (3.7)f
|
N/A
|
0%
|
|
Placebo
|
86
|
31.4%
|
46.2 (11.16)
|
33.4 (11.41)e
|
N/A
|
2.12 (2.83)
|
4.6 (5.3)f
|
N/A
|
0%
|
|
Dalyd (2019)
|
Esketamine
|
62
|
38.7%
|
47.2 (11.00)
|
36.2 (13.25)e
|
N/A
|
2.33 (3.72)
|
10.8 (5.5)f
|
N/A
|
0%
|
|
Placebo
|
59
|
61.3%
|
46.7 (9.76)
|
34.0 (10.54)e
|
N/A
|
2.72 (4.88)
|
10.5 (4.8)f
|
N/A
|
0%
|
a Values are shown as mean (standard deviation). b Including the current episode. c Stable remission group. d Stable response group. e Age when diagnosis of major depressive disorder was confirmed. f Extracted from the figure in the report. CGI-S, Clinical Global Impressions-Severity
of illness; MADRS, Montgomery-Asberg Depression Rating Scale; NA, not applicable;
N/A, not available; RCT, randomized controlled trial; SD, standard deviation.
Table 2 Study design of included studies.
|
First author (Year)
|
Stabilization period after remission before RCT
|
Duration of RCT phase
|
Major inclusion criteria regarding age and illness severity
|
Augmentation drug
|
Dose ofaugmentation drug
|
Antidepressants used
|
Definition of relapse
|
|
Hardy (1997)
|
1 year
|
2 years
|
-
>65 years of age
-
Diagnosis of major unipolar depressive episode according to DSM-III-R
-
“Refractory” depression (failed to show improvements after≥6 months at maximal doses
of 1 antidepressant)
-
GDRS score of<20
-
MADRS score of<15
-
MMSE score of>20
-
Responded to lithium augmentation
-
Absence of depressive symptoms for at least 1 year on lithium augmentation
|
Lithium
|
150 mg (n=1)
300 mg (n=10)
900 mg (n=1)
|
Amitriptyline (n=6)
Doxepine (n=3)
Imipramine (n=1)
Tranylcypromine (n=1)
Nortriptyline (n=1)
|
-
Hospitalization for depression
-
Need of adjustment of antidepressant medication
-
MADRS score of≥15 or change from baseline of≥10
-
GDRS score of>20 or change from baseline of≥10
|
|
Bauer (2000)
|
2–4 weeks
|
4 months
|
-
≥18 years of age
-
Diagnosis of major depressive episode according to DSM-III-R
-
“Refractory” depression (failed to respond to an adequate trial of 1 antidepressant)
-
HAMD-21 score of≤10, CGI-S score of≤3 or CGI-I score of 2 or 3 following a 6-week
lithium augmentation phase, and being judged by 2 independent psychiatrists as asymptomatic
|
Lithium
|
|
Amitriptyline (n=16)
Clomipramine (n=3)
Nortriptyline (n=3)
Dibenzepin (n=2)
Trazodone (n=1)
Paroxetine (n=1)
Clomipramine (n=1)
Imipramine (n=1)
Maprotiline (n=1)
Venlafaxine (n=1)
|
-
HAM-D-21 score of≥15
-
CGI-S score of≥4
|
|
Rapaport (2006)
|
None
|
24 weeks
|
-
18–85 years of age
-
Diagnosis of MDD, single or recurrent episode, with or without psychotic features
according to DSM-IV
-
“TRD” (resistant to respond to≥1 adequate antidepressant trials for≥6 weeks)
-
HAMD-17 score of≤7 or CGI-S score of 1 or 2
|
Risperidone
|
-
Open-label augmentation phase, 1.1±0.6 mg/day (mean±SD)
-
Double-blinded phase, 1.2±0.6 mg/day
|
Citalopram
|
|
|
Alexopoulos (2008)
|
None
|
24 weeks
|
-
≥55 years of age
-
Diagnosis of MDD, single or multiple episode, with or without psychotic features according
to DSM-IV
-
“Resistant depression” (resistant to respond to≥1 adequate antidepressant trials for≥6
weeks)
-
HAMD-17 score of≤7 or CGI-S score of 1 or 2
|
Risperidone
|
-
Acute treatment phase, 0.7±0.3 mg/day
-
Double-blind phase, 0.8±0.3 mg/day
|
Citalopram
|
|
|
Brunner (2014)
|
12 weeks
|
27 weeks
|
-
18–65 years of age
-
Diagnosis of single or recurrent unipolar MDD without psychotic features according
to DSM-IV-TR
-
“TRD” (resistant to respond to≥2 different antidepressants, for≥6 weeks for each medication)
-
Absence of psychotic features
-
Maintained≥50% improvement with augmentation therapy compared to baseline on MADRS
and CGI-S score of≤3 during 12-week stabilization phase
|
Olanzapine
|
3–18 mg/day
|
Fluoxetine
|
-
50% increase in the MADRS score from randomization with concomitant CGI-S score increase
to≥4
-
Hospitalization for depression or suicidality
-
Discontinuation due to lack of efficacy or worsening of depression or suicidality
|
|
Oakes (2015)
|
12 weeks
|
2 weeks
|
|
Edivoxetine
|
12 or 18 mg/day
|
Escitalopram ( n=167)
Citalopram (n=114)
Fluoxetine (n=107)
Sertraline (n=106)
Paroxetine (n=80)
Fluvoxamine (n=12)
|
-
MADRS score of≥14 or CGI-S score increase of≥2
-
Discontinuation due to lack of efficacy, worsening of depression, or suicidality
|
|
Daly (2019)
|
12 weeks
|
Variable durationa
|
-
18–64 years of age
-
Diagnosis of current or single- episode (≥2 years) MDD according to DSM-5
-
“TRD” (resistant to respond to≥1 different antidepressants, for≥4 weeks for each medication)
-
Absence of psychotic features
-
MADRS score of≤12 during the last 4 weeks of the maintenance phase (stable remission
group)
-
≥50% reduction from baseline in MADRS score in the last 2 weeks of the maintenance
phase but not achieving remission (stable response group)
|
Esketamine
|
-
Open-label augmentation phase, 56 mg or 84 mg for 2 times per week
-
Maintenance phase, once weekly for the first 4 weeks and then individualized to weekly
or biweekly based on the severity of depressive symptoms
-
Double-blinded phase, weekly or biweekly based on the severity of depressive symptoms
|
SNRI (n=191)
SSRI (n=86)
|
-
MADRS score of≥22 at 2 consecutive assessments with an interval of 5 to 15 days or
hospitalization due to symptom worsening
-
Suicide attempt, suicide prevention, or completed suicide
-
Another clinically relevant event suggestive of relapse
|
CGI-I, Clinical Global Impression-Improvement; CGI-S, Clinical Global Impressions-Severity
of illness; DSM, Diagnostic and Statistical Manual of Mental Disorders; GDRS, Geriatric
Depression Rating Scale; HAMD, Hamilton Depression Rating Scale; MADRS, Montgomery-Asberg
Depression Rating Scale; MDD, Major Depressive Disorder; MMSE, Mini-Mental State Examination;
SNRI, serotonin norepinephrine reuptake inhibitors; SSRI, selective serotonin reuptake
inhibitors; TR, Text Revision; TRD, treatment-resistant depression; RCT, randomized
controlled trial. a From 10.1 weeks to 19.4 weeks as a median duration in each treatment group; the intervention
was continued until the required number of relapse was achieved or until the study
was recommended to be stopped based on the results of interim analysis.
Risk of Bias
Risks of bias of the included studies are summarized in Table S2. All studies were double-blind RCTs. The methodology of random sequence generation
and allocation concealment was unclear in all the included studies. Furthermore, blinding
of outcome assessment was often unreported, leading to “unclear risk” for detection
bias in all studies. Withdrawal cases were adequately explained. Two studies (29%)
did not report full data on adverse effects and were judged to have “high risk” of
selective reporting.
Withdrawal from the Study
Overall withdrawal rates from the study were comparable between the patients who continued
the augmentation drug added on to the antidepressant and those who discontinued it
(n=7, n=1.672, RR=0.86, 95% CI=0.69–1.08, p=0.20) ([Fig. 2a]). Also, when the study by Daly et al. [34] which used esketamine for augmentation therapy was excluded, there was no significant
difference between the 2 groups (n=6, n=1.375, RR=0.94, 95% CI=0.77–1.14, p=0.52)
([Fig. 2b]). Subgroup analysis of the data from 5 studies [28]
[29]
[32]
[33]
[34] that included a stabilization period before RCT found no significant difference
between the 2 groups (n=5, n=1.368, RR=0.82, 95% CI=0.57–1.17, p=0.27) ([Fig. 2c]). This finding remained unchanged when the esketamine study [34] was excluded (n=4, n=1.071, RR=0.94, 95% CI=0.62–1.43, p=0.77) ([Fig. 2d]). Subgroup analyses for lithium and second-generation antipsychotics also failed
to find statistical differences, respectively (lithium: n=2, n=41, RR=0.37, 95% CI=0.01–15.49,
p=0.60; second-generation antipsychotics: n=3, n=748, RR=0.89, 95% CI=0.74–1.06, p=0.18).
Fig. 2 Study withdrawal due to all causes in the continuation and discontinuation groups.
The rate of study withdrawal due to relapse was significantly lower in the continuation
group than that in the discontinuation group (n=7, n=1.672, RR=0.61, 95% CI=0.40–0.92,
p=0.02) ([Fig. 3a]). When the esketamine study [34] was excluded, the difference showed a trend level without any statistical significance
(n=6, n=1.375, RR=0.63, 95% CI=0.37–1.05, p=0.08) ([Fig. 3b]). Subgroup analysis of the data from studies that included a stabilization period
found a significantly lower relapse rate in the continuation group than that in the
discontinuation group (n=5, n=1.368, RR=0.47, 95% CI=0.36–0.60, p<0.00001) ([Fig. 3c]). This finding was unchanged when the esketamine study [34] was excluded (n=4, n=1.071, RR=0.39, 95% CI=0.27–0.58, p<0.00001) ([Fig. 3d]). When the trial by Hardy et al. [28] that included the longest stabilization phase of 1 year was excluded from these
5 studies, the significant difference was not affected (n=4, n=1.356, RR=0.45, 95%
CI=0.34–0.60, p<0.00001). On the other hand, subgroup analyses for lithium and second-generation
antipsychotics did not find statistical differences between the continuation and discontinuation
groups, respectively (lithium: n=2, n=41, RR=0.32, 95% CI=0.02–6.11, p=0.45; second-generation
antipsychotics: n=3, n=748, RR=0.70, 95% CI=0.40–1.24, p=0.22).
Fig. 3 Study withdrawal due to relapse in the continuation and discontinuation groups.
No significant differences were found in the study withdrawal rates due to adverse
events between the 2 groups in the whole dataset (n=4, n=1.334, RR=1.41, 95% CI=0.84–2.36,
p=0.20) (Fig. S1a) or in a subgroup studies that included a stabilization period (n=2, n=1.030, RR=0.72,
95% CI=0.09–6.10, p=0.76) (Fig. S1b).
Symptomatology Scores
There were no significant differences in score changes in the Montgomery-Asberg Depression
Rating Scale (MADRS) (n=4, n=1.334, MD=− 0.60, 95% CI=− 3.08–1.88, p=0.64) (Fig. S2a), Hamilton Depression Rating Scale-17 items (HAMD-17) (n=2, n=304, MD=0.20, 95% CI=− 1.66–2.06,
p=0.83) (Fig. S3a), the Clinical Global Impression-Severity of Illness (CGI-S) scale (n=3, n=1.051,
MD=− 0.07, 95% CI=− 0.41–0.27, p=0.68) (Fig. S4a), Hamilton Depression Rating Scale-21 items (HAMD-21) (n=1, n=21, MD=0.30, 95% CI=− 3.02–3.62,
p=0.86) (Fig. S5a), and 2 rating scales (i. e., the MADRS and HAMD-21) combined (n=5, n=1.355, SMD=− 0.05,
95% CI=− 0.28–0.18, p=0.66) (Fig. S6a) between the continuation and discontinuation groups. With reference to the subgroup
analysis of the data from studies that included a stabilization period, no significant
differences were found in the MADRS (n=2, n=1.030, MD=− 1.77, 95% CI=− 5.54–2.01,
p=0.36) (Fig. S2b), CGI-S scale (n=3, n=1.051, MD=− 0.07, 95% CI=− 0.41–0.27, p=0.68) (Fig. S4b), HAMD-21 (n=1, n=21, MD=0.30, 95% CI=− 3.02–3.62, p=0.86) (Fig. S5b), and 2 rating scales (i. e., the MADRS and HAMD-21) combined (n=3, n=1.051, SMD=− 0.14,
95% CI=− 0.47–0.18, p=0.38) (Fig. S6b) between the 2 groups. Symptomatology scores were available in only 1 study for lithium
and 3 studies for second-generation antipsychotics. A subgroup analysis for the second-generation
antipsychotics found no significant differences in score changes of the MADRS (n=3,
n=748, MD=− 0.67, 95% CI=− 4.64–3.30, p=0.74) or HAMD-17 (n=2, n=304, MD=0.20, 95%
CI=− 1.66–2.06, p=0.83) between the continuation and discontinuation groups.
Adverse Events
Twenty-two adverse events were assessed in 2 or more studies. Among these 22 adverse
events (Fig. S7a–b), only 3 showed statistically significant differences as follows: less frequent depression
(n=2, n=1.030, RR=0.34, 95% CI=0.13–0.87, p=0.02) (Fig. S7c and frequent headache (n=6, n=1.660, RR=1.59, 95% CI=1.12–2.25, p=0.009) (Fig. S7k) and hyperhidrosis (n=2, n=615, RR=5.13, 95% CI=1.16–22.76, p=0.03) (Fig. S7l) in the continuation group than that in the discontinuation group. Subgroup analysis
of the 3 studies in which second-generation antipsychotics were used as augmentation
drugs [30]
[31]
[32] compared the incidence rates of 16 adverse events between the continuation and discontinuation
groups (Fig. S8a–p) and found no significant differences between them.
Publication Bias
A funnel plot of the included 7 studies with respect to overall withdrawal from the
study indicated the low possibility of publication bias (Fig. S9).
Discussion
This meta-analysis of double-blind RCTs found no significant differences in study
withdrawal rates due to any reason or due to the majority of adverse events between
patients with depression who continued adjunctive psychotropic drugs used for augmentation
therapy and those who discontinued these drugs. We found a significantly lower study
withdrawal rate due to relapse in the continuation group than that in the discontinuation
group; however, this statistical significance became insignificant when one trial
using esketamine as an augmentation therapy was excluded. In contrast, in a subgroup
of studies that included a stabilization period after remission before entering the
RCT phase, the rate of study withdrawal due to relapse was lower in the continuation
group than that in the discontinuation group, regardless of whether esketamine study
was included or not. These findings suggest that augmentation therapy may be discontinued,
but this may not be the case for patients who were maintained with augmentation therapy
after remission. Moreover, the results of this analysis indicate the need for continuing
esketamine for relapse prevention in the treatment of depression although the available
data are still limited.
When relapse was focused as a treatment outcome, a subgroup analysis of studies that
included a stabilization period after remission found a lower study withdrawal rate
in the continuation group than in the discontinuation group, regardless of whether
esketamine study was included or not. This finding seems reasonable since these 5
studies included the patients who benefited from such augmentation therapy in terms
of relapse prevention. Moreover, the mean number of previous depressive episodes was
up to 3.7 in these 4 studies. Since repetitive episodes of depression are characteristic
of bipolar depression [35], potential patients who could later develop bipolar disorder may have been included
in these studies. In fact, 2 of the 3 drugs used for augmentation therapy in these
4 studies are indicated for bipolar disorder. The results in this meta-analysis provide
important knowledge for further discussions with regard to similarities and differences
between treatment-resistant depression and bipolar depression.
Ketamine, an N-methyl-D-aspartate receptor-modulating anesthetic, has a rapid-onset,
strong antidepressant efficacy for patients with treatment-resistant depression [36]
[37]
[38]
[39]
[40]
[41]. However, the duration of its efficacy is only a few days [37]. Ketamine has a unique mechanism of action via the glutamatergic system, which is
different from other conventional antidepressants. Therefore, the response and remission
of depression achieved by the use of ketamine could be essentially transient and qualitatively
different from those by other antidepressants. In fact, relapse after discontinuation
of ketamine in the treatment of depression was frequently observed in the study included
in this meta-analysis [34]. On the other hand, it should be noted that there has been only one RCT that examined
the effect of discontinuing ketamine used as an augmentation therapy on relapse in
the treatment of depression. Moreover, this study included an antidepressant switch
when esketamine was introduced, whereas other studies included in this review [28]
[29]
[30]
[31]
[32]
[33] continued the same antidepressants. These issues clearly warrant further investigations
on whether and how ketamine treatment should be continued for relapse prevention in
depression.
One systematic review reported frequent adverse events associated with adjunctive
treatment with antipsychotics for depression as follows: akathisia, sedation, abnormal
metabolic laboratory results, and weight gain [42]. In contrast, we did not find any significantly different incident rates of these
symptoms between the 2 treatment strategies although the discontinuation rate due
to adverse events was numerically lower in the discontinuation group than that in
the continuation group. However, in the light of a variety of adverse events, including
motor [19]
[20], metabolic [21], cognitive [22], and cardiovascular [23] caused by antipsychotics and thyroidal and parathyroidal dysfunction caused by lithium
[24], physicians should be aware of these potential adverse events in the maintenance
treatment of depression. This is especially true for patients with mood disorders
due to increased sensitivity to antipsychotics in these patients [43].
There are several limitations of this study. First, the number of studies included
in this meta-analysis, especially those examining efficacy of ketamine, was small.
The results of our analysis highly depend on which trials are included. Moreover,
it should be noted that edivoxetine has not been approved for adjunctive treatment
for major depressive disorder because of negative findings in phase II and III trials
[44]. Second, the study design was different among the 7 studies included. The definitions
of relapse, treatment resistance, study durations, subject characteristics, durations
of adjunctive therapy, augmentation drugs, and methods of discontinuing drugs used
for augmentation therapy varied among them. Among them, various definitions of treatment
resistance as detailed in [Table 2] should be especially acknowledged since the degree of treatment resistance could
affect both the likelihood of acute response and the probability of relapse [3]. Moreover, the trial duration varied from 2 weeks to 2 years; the period of 2 weeks
may be short to evaluate long-term outcomes of this chronic illness and side effects
of drugs used for augmentation. Third, adverse events were not comprehensively or
thoroughly assessed in the majority of the trials, which clearly limits the interpretation
of the findings of this meta-analysis. Lastly, 6 of the 7 studies were funded by pharmaceutical
companies, which needs to be acknowledged when the results in favor of the continuation
strategy are interpreted.
It should be noted that no firm conclusions could be drawn in the light of the small
number of studies included. In addition, no clear answer was obtained as to how long
augmentation therapy should be continued for the maintenance treatment of depression.
Still, currently available evidence suggests that add-on drugs used for augmentation
therapy for depression may be discontinued with the exception of esketamine. However,
for patients who are maintained with augmentation therapy after remission, discontinuation
of augmentation therapy may need to be carefully considered.
