You are on page 1of 13

Journal of Affective Disorders 308 (2022) 268–280

Contents lists available at ScienceDirect

Journal of Affective Disorders


journal homepage: www.elsevier.com/locate/jad

Review article

Lithium in the treatment of acute bipolar depression: A systematic review


and meta-analysis
Jeffrey J. Rakofsky *, Michael J. Lucido, Boadie W. Dunlop
Emory University, Atlanta, GA, USA

A R T I C L E I N F O A B S T R A C T

Keywords: Objectives: To evaluate lithium in the treatment of acute bipolar depression.


Bipolar depression Methods: We conducted a systematic literature review for: 1) cross-over or parallel-group design studies
lithium comparing lithium response in bipolar versus unipolar depressed patients, and 2) parallel group studies of bipolar
meta-analysis
depressed patients comparing lithium versus placebo or other psychotropics. Meta-analyses using response rate
as the primary outcome were conducted to evaluate lithium's efficacy.
Results: The literature search yielded 947 records. Ultimately, 17 studies were included, totaling 1545 patients,
including 676 who received lithium. The overall summary effects reveal that there were no statistically signif­
icant differences between lithium versus antidepressants or placebo, however, lithium performed numerically
worse than antidepressants (RR = 0.61; 95%CI, 0.37–1.02; p = 0.06) but better than placebo (RR = 1.18; 95%CI,
0.99–1.41; p = 0.07). The specificity of lithium for bipolar versus unipolar depression was not supported in the
primary analysis of all trials, though an analysis limited to double-blinded, monotherapy, cross-over studies
revealed a statistically significant result supporting lithium's efficacy for those with bipolar depression.
Limitations: Limitations include study selection rules, the use of response rates rather than remission rates or
continuous score outcomes, and the small number of studies included in each meta-analysis.
Conclusions: These meta-analyses do not support lithium as a first-line treatment for acute bipolar depression.
However, the bipolar vs. unipolar sensitivity analysis and the modest, though non-significant advantage over
placebo suggest lithium may still be a viable treatment option. Larger and more rigorously-designed studies are
needed to determine lithium's full range of efficacy relative to placebo and other psychotropics.

1. Introduction not approved for the treatment of depressive episodes for bipolar type I
or type II patients (Roxane Laboratories, 2011). Doubts surrounding
Bipolar disorder types I and II together affect 2.1% of the population lithium's ability to treat bipolar depression originated from John Cade's,
(Merikangas et al., 2007) and are associated with a markedly impaired 1949 uncontrolled study reporting no antidepressant effect of lithium on
quality of life, (Michalak et al., 2005) significant disability, (Murray and a small group of depressed patients (Cade, 1949). The controversy
Lopez, 1997) and early mortality (Crump et al., 2013). Patients with continues today with ten treatment guidelines across the globe recom­
bipolar disorder experience depressive symptoms for 32–50% of the mending lithium as a first-line treatment for bipolar depression and five
time, (Judd et al., 2003; Judd et al., 2002) and these symptoms are not recommending lithium at all (Kelly, 2019). This review of the
associated with social impairment (Judd et al., 2005) and an increased guidelines argued that “woozle effects” (evidence based on repeated
risk of death by suicide (Baldessarini et al., 2019). There are few US citations over time), reference inflation (overstating findings of the
Food and Drug Administration (FDA) approved treatments for bipolar original studies), and belief preservation (maintaining beliefs despite
depression, leaving patients with limited options and residual depressive contradictory evidence) explain why lithium continues to be favored by
symptoms that predict poor functional recovery in those with multiple some guidelines in spite of relatively limited evidence of lithium's effi­
previous episodes (Rosa et al., 2012). cacy. Unfortunately, there has only been one large, double-blind, ran­
Lithium carries an FDA marketing approval for the treatment of domized controlled trial comparing lithium to placebo (Young et al.,
manic episodes and maintenance treatment for bipolar I disorder but is 2010) and there are no such studies currently listed on trial registries,

* Corresponding author at: 12 Executive Park Drive, Atlanta, GA 30329, USA.


E-mail address: jrakofs@emory.edu (J.J. Rakofsky).

https://doi.org/10.1016/j.jad.2022.04.058
Received 26 December 2021; Received in revised form 4 April 2022; Accepted 10 April 2022
Available online 13 April 2022
0165-0327/© 2022 Elsevier B.V. All rights reserved.
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

such as Clinicaltrials.gov and the European Union Clinical Trials Reg­ cyclic depression’ which are synonymous with unipolar depression
ister. Consequently, addressing this controversy requires careful (World Health Organization, 1968). For the lithium versus psychotropic
consideration of the existing evidence base. analyses, included studies must have enrolled two or more bipolar pa­
Several narrative reviews (Bhagwagar and Goodwin, 2002; Johnson, tients in each treatment arm, and patients must have been randomized
1987; Mendels et al., 1979; Srisurapanont et al., 1995; Zornberg and to receive the different treatments. However, the bipolar versus unipolar
Pope, 1993) and four meta-analyses (Bahji et al., 2020; Selle et al., 2014; studies were not required to use randomization as the two study groups
Taylor et al., 2014; Vieta et al., 2010) of lithium monotherapy in the were categorized by disease rather than by treatment.
treatment of acute bipolar depression have been published. Two other
meta-analyses examined the effect of lithium versus placebo for aug­ 2.2. Information sources
menting tricyclic antidepressants in mixed (i.e., unipolar and bipolar)
depressed patient populations (Austin et al., 1991; Bauer and Dopfmer, Sources included OVID/Medline, PubMed, Embase, Google Scholar,
1999). All four of the monotherapy meta-analyses included several the ClinicalTrials.gov registry and the European Union Clinical Trials
treatments for bipolar depression in addition to lithium and included no Register (clinicaltrialsregister.eu). The database search occurred on
more than 1–5 lithium trials depending on the publication. Selle et al. October 10th, 2021. Additionally, reference lists within bipolar disorder
had the most lithium studies and was the only one to show an advantage treatment guidelines (American Psychiatric Association, 2010; Goodwin
over placebo. They had five placebo cross-over trials and conducted two et al., 2016; Grunze et al., 2010; Yatham et al., 2018), bipolar disorder
separate analyses of four trials each. One of their analyses compared textbooks (Goodwin and Jamison, 1990; Goodwin et al., 2007), previous
lithium to placebo among hospitalized patients and the other compared meta-analyses (Austin et al., 1991; Bahji et al., 2020; Bauer and
lithium's effect in patients with bipolar depression versus unipolar Dopfmer, 1999; Selle et al., 2014; Taylor et al., 2014; Vieta et al., 2010),
depression. These analyses revealed overall significantly larger effect and relevant review articles (Altshuler et al., 2003; Bhagwagar and
sizes for lithium over placebo and in bipolar depressed patients over Goodwin, 2002; Johnson, 1987; Kelly, 2019; Mendels et al., 1979; Sri­
unipolar depressed. However, these results were considered to be sec­ surapanont et al., 1995; Zornberg and Pope, 1993) were also identified
ondary outcomes in the work by Selle et al. and the analyses suffered for possible inclusion.
from several concerns, including little explanation to justify the selec­
tion criteria for the analyzed studies and the lack of comparison of 2.3. Search
lithium versus other active treatments. Since the publication of this
meta-analysis, two additional randomized controlled trials of lithium in The following terms were used to search the databases, [Bipolar
the treatment of bipolar depression studies have been reported (Altsh­ disorder OR Manic-depression OR depression] AND [lithium] AND [All
uler et al., 2017; Amsterdam et al., 2016), only one of which was clinical trials].
included in the most recent meta-analysis (Bahji et al., 2020).
Given the differences in methodology between previously published 2.4. Study selection
lithium bipolar depression meta-analyses, the ongoing controversy
surrounding lithium's antidepressant efficacy, and the lack of well- The first author (JJR) screened the titles and abstracts of potential
powered, high-quality lithium bipolar depression trials, additional studies to determine inclusion. Eligible studies were subsequently
meta-analyses focused exclusively on lithium and utilizing a compre­ confirmed by the second author (MJL) who independently checked the
hensive search strategy are warranted. We undertook a systematic re­ full text of all retrieved articles. Disagreement was resolved through
view of the literature with the aim of conducting meta-analyses to discussion and consensus between JJR, MJL, and BWD.
evaluate the efficacy of lithium versus: 1) placebo; 2) antidepressants; 3)
other mood stabilizers; and 4) antipsychotics. Additionally, we meta- 2.5. Data collection process
analyzed the specificity of lithium's efficacy in acutely depressed pa­
tients, evaluating studies that compared outcomes in bipolar versus One reviewer (JJR) extracted the following data from included
unipolar depressed patients. studies while the second author (MJL) checked the extracted data,
including author, year of publication, study design, sample size, name/
2. Methods dose of interventions, lithium blood level, study duration, outcome
scale, and results. Disagreements were resolved by discussion between
This systematic review followed the Preferred Reporting Items for JJR, MJL, and BWD.
Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Moher
et al., 2009). 2.6. Risk of bias in individual studies

2.1. Eligibility criteria Two reviewers (JJR, MJL) independently rated each eligible study,
using the modified Downs and Black quality assessment scale, (Downs
All cross-over or parallel-group design studies comparing lithium and Black, 1998) consisting of 27 questions that address study reporting,
response in bipolar patients versus unipolar depressed patients, and all external validity, and internal validity (bias and confounding). Each
parallel group studies of bipolar depressed patients comparing lithium question is worth one point. Studies with total scores of 14 or less are
versus placebo pill or other psychotropics were included. Included considered to be of poor quality, 15–19 of fair quality, and 20 or greater
studies had to have enrolled bipolar I or II patients in the depressed of good quality. Any disagreements in ratings were resolved by a third
phase, initiated lithium during the course of the study, treated patients reviewer (BWD).
in the lithium arm for at least one week, provided the raw numbers or
percentages of patients responding or improving, and must have re­ 2.7. Synthesis of results and risk of bias across studies
ported results separately for bipolar patients if the study included mixed
(bipolar and unipolar) depressed patient samples. Included studies must For the lithium versus placebo/psychotropic studies, results were
also have been prospective in design and written in English. Studies that organized by drug class (placebo, antidepressants, mood stabilizers,
enrolled patients diagnosed with ‘manic-depressive psychosis- circular antipsychotics). Meta-analyses comparing the number of responders in
type’ were included as this term is synonymous with bipolar disorder each treatment or diagnostic group were conducted when two or more
according to the ICD-8 (World Health Organization, 1968). This is in eligible publications were identified. Response rate was selected as the
distinction to ‘manic-depressive psychosis- depressed type’ and ‘non- primary outcome because it was the most frequently reported outcome

269
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

(i.e., more common than remission rate or standardized mean differ­ Comprehensive Meta-Analysis was used for the meta-regression and
ence). For all meta-analyses, sensitivity analyses were conducted by publication bias analyses (Borenstein et al., 2013).
including only similarly designed studies. Specifically, for the lithium
versus placebo comparisons and the lithium versus antidepressant 3. Results
comparisons, only monotherapy, double-blind, randomized-controlled
studies were included in the sensitivity analyses, and for the lithium 3.1. Overall results
studies comparing bipolar and unipolar depressed patients, only double-
blind, monotherapy, cross-over studies were included. When there were The literature search yielded a total of 947 records. Of those records,
insufficient number of studies to conduct a meta-analysis, narrative 50 were fully reviewed, and 33 were removed due to the following
analyses were used to summarize response rates. reasons: did not provide responder rates for individual patients (n = 1),
Studies varied in their definition and reporting of response and cross-over study with no unipolar depression comparison group (n = 1),
improvement. For the analyses, all individuals who responded or did not break out results by mood disorder diagnosis (n = 6), did not
improved, regardless of degree and as defined by the study, were enroll bipolar patients (n = 5), less than one week of lithium treatment
counted as responders. However, in a few of the placebo-controlled, (n = 1), too few bipolar patients (n = 4), did not indicate the number of
cross-over studies, responders were classified as ‘unequivocal’ or patients with bipolar disorder (n = 1), did not identify how many bipolar
‘equivocal/probable’ based on whether they relapsed during placebo patients were depressed at baseline (n = 2), case series (n = 9), retro­
phases after improving during lithium phases. For these studies, only spective (n = 2), review article (n = 1). See Fig. 1. Ultimately, there were
those who relapsed (unequivocal) were counted as responders. 17 studies that met criteria, including a total of 1545 patients with 676
Mantel Haenszel random-effects models were used to conduct all of those receiving lithium. Detailed descriptions of the included studies
meta-analyses and determine the mean effect size across studies (Deeks are provided in Tables 1–5.
et al., 2021). Results for each study were summarized as risk ratios (RR).
The overall mean effect size and 95% confidence intervals (CI) were 3.2. Lithium vs placebo
reported for each meta-analysis. To assess heterogeneity, T2, I2, Q, and P
values were reported. T2 is a measure of between-study variance and The search identified 3 publications (Arieli and Lepkifker, 1981;
when greater than zero, can be used to calculate prediction intervals Ebert et al., 1995; Young et al., 2010) that compared lithium to placebo
(PI). Prediction intervals include the range of effect sizes that have a and met the inclusion criteria. See Table 1. Two of these studies (Arieli
95% likelihood of falling within the population of all comparable and Lepkifker, 1981; Young et al., 2010) were monotherapy trials while
studies. I2 demonstrates the proportion of variance that is due to true the other (Ebert et al., 1995) studied combination treatment: initiating
study effects rather than sampling error. The Q statistic is a test for the an antidepressant plus lithium simultaneously and comparing that to
presence of heterogeneity and is accompanied by a p-value that indicates simultaneous initiation of an antidepressant plus placebo. Young et al.
statistical significance (Borenstein et al., 2009). (2010) was the only study that recruited outpatients, excluded those
For meta-analyses with ten or more studies, a random-effects meta- who failed antidepressants from 2 or more classes, and required the
regression analysis was conducted to look for moderators that might duration of participants' major depressive episodes to be no less than 1
explain the heterogeneity in effect sizes (Borenstein et al., 2009). month and no longer than 1 year. On the other hand, Ebert et al. (1995)
Similarly, measures of publication bias were conducted only for meta- enrolled only male patients, between 30 and 45 years old with a
analyses containing ten or more studies (Ioannidis and Trikalinos, melancholic depressive subtype. Mean lithium levels ranged from 0.61
2007). A funnel plot was created to investigate the presence of publi­ mEq/L in two studies (Ebert et al., 1995; Young et al., 2010) to 1.27
cation bias. If present, the “trim and fill” method was used to determine mEq/L in the other (Arieli and Lepkifker, 1981). Attrition was only re­
the impact of the bias (Duval and Tweedie, 2000). RevMan 5.4 was used ported in one of the studies and was equal between the treatment groups
for all meta-analysis calculations and to create forest plots for the main (Young et al., 2010). Study duration ranged from 3 to 8 weeks. Only one
and sensitivity analyses (The Cochrane Collaboration, 2020), while study (Young et al., 2010) explicitly defined response (50%
Idenficaon

844 Records idenfied 103 Studies idenfied through


through database/registry addional hand search
search
Screening

864 Records excluded due


947 Records screened by to duplicaon or
tle and abstract irrelevance

33 arcles excluded for:


Case series (9)
Eligibility

Results not reported by mood disorder (6)


50 Full text arcles
No bipolar paents (5)
assessed for eligibility Too few bipolar paents (4)
Number of bipolar paents depressed at baseline
not reported (2)
Retrospecve (2)
No responder rates for individual paents (1)
17 Arcles included in the
Included

Cross-over study with no MDD group (1)


qualitave/quantave Less than one week of lithium treatment (1)
analysis (meta-analysis) Number of paents with bipolar disorder not
reported (1)
Review arcle (1)

Fig. 1. Flow Diagram for Literature Search.

270
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

Table 1
Lithium versus placebo in the treatment of bipolar depression.
Author, year Sample Dose Lithium level Duration Scale Results Notes
sizes

Young et al., Lithium: Lithium: 600–1800 Mean median in ITT was 0.61 8 weeks MADRS, Response rate (at least 50% -Double blind, parallel
2010 N = 136 mg mEq/L; 34.9% below 0.6 HAMD, reduction in baseline MADRS): group, RCT
Placebo: mEq/L; 25% had a median CGI Lithium- (85/136) 62.5%, -Bipolar I and II
N = 129 level of 0.8 mEq/L or greater Placebo- (72/129) 55.8% -Multicenter
-Outpatients
-Monotherapy
-Randomized 2:2:1:1
-MDE duration less than
1 year and 4 weeks or
more
-had not failed 2 or more
classes of antidepressants
Bias = 23 (Good)
Arieli and Lithium N Lithium: Mean = 1.27 mEq/L 3 weeks HAMD, Response rate (defined by JJR, -Double blind, parallel
Lepkifker, =3 2000–2500 mg CGI MJL, BWD as at least 50% group, RCT
1981 Placebo N reduction in baseline HAMD): -Hospitalized
=3 Lithium- (2/3) 66% -Single-site
Placebo- (1/3) 33% -Included MDD patients
as well
-One week placebo
washout
-Reported %
improvement on HAMD
Bias = 19 (Fair)
Ebert et al., Lithium N Lithium: 900 mg Mean (SD) at week 1, 2, 3 = 5 weeks HAMD, Improved or much improved -Double blind, parallel
1995 = 20 0.65 (0.16) mEq/L CGI (CGI): Lithium- (17/20) 85% group, RCT
Placebo N Amitriptyline: 225 -Augmentation
= 20 mg dosed to 200 ng/ Placebo- (12/20) 60% -Hospitalized
mL 0.66 (0.17) mEq/L -Single site
-All male
0.61 (0.15) mEq/L -30-45 years old
-Melancholic subtype
-All started on
Amitriptyline at the start
of trial
-1st 20 randomized, 2nd
20 matched on severity
and duration
Bias = 16 (Fair)

Key: CGI = Clinical Global Impression Scale, HAMD = Hamilton Depression Scale, ITT = Intent to treat sample, MADRS = Montgomery Asberg Depression Scale, MDE
= Major depressive episode, RCT = Randomized controlled trial, SD = Standard deviation.

improvement on the Montgomery Asberg Depression Rating Scale mEq/L and 0.64 mEq/L in Altshuler et al. (2017) and Amsterdam and
(MADRS) (Montgomery and Asberg, 1979), while one provided Clinical Shults (2008), respectively, to highs of 0.94 mEq/L (range = 0.3–2.4),
Global Impression (CGI) Improvement scores, (Ebert et al., 1995) and and 1.27 mEq/L in Amsterdam et al. (2016) and Arieli and Lepkifker
the third provided percentage of improvement on the 17-item Hamilton (1981), respectively. Antidepressants included venlafaxine, (Amsterdam
Depression Rating Scale (HAMD-17) (Hamilton, 1960) for each subject et al., 2016; Amsterdam and Shults, 2008) sertraline, (Altshuler et al.,
(Arieli and Lepkifker, 1981). Bias scores were 16–23 and fell within the 2017) and clomipramine (Arieli and Lepkifker, 1981). Attrition was
‘Fair’ to ‘Good’ range depending on the study. uneven in most of the studies with more patients dropping out of the
The meta-analysis found non-statistically significant higher response lithium group than the comparator. The attrition rate difference be­
rates with lithium versus placebo (3 studies, N = 311; RR = 1.18; 95% tween groups ranged from 12.9% to 41.6% depending on the study.
CI, 0.99–1.41; p = 0.07) There was no evidence of heterogeneity (T2 = 0, Study duration ranged from as short as three weeks to as long as 16
I2 = 0%, Q = 1.39, df = 2, p = 0.5). See Fig. 2A. When only monotherapy weeks. To define response, all studies required a 50% score reduction in
studies were included, the sensitivity analysis revealed similar results (2 the depression severity scale with two requiring additional items (a
studies, N = 275; RR = 1.13; 95% CI, 0.92–1.38; p = 0.24; heteroge­ decrease of </= 2 points on the (CGI)- Bipolar Depression Severity score
neity: T2 = 0, I2 = 0%, Q = 0.4, df = 1, p = 0.53). See Fig. 2B. for at least two consecutive weeks (Altshuler et al., 2017); a final clinical
CGI-Severity score of 1 (not at all), 2 (borderline ill), or 3 (mildly ill)
3.3. Lithium vs. antidepressants (Amsterdam et al., 2016)). Bias scores were 19–23 which fell in the ‘Fair’
to ‘Good’ range.
The search identified four publications (Altshuler et al., 2017; The meta-analysis revealed that patients receiving lithium were less
Amsterdam et al., 2016; Amsterdam and Shults, 2008; Arieli and Lep­ likely to respond than those receiving antidepressants although the
kifker, 1981) comparing lithium to antidepressants that met the inclu­ result did not reach statistical significance (4 studies, N = 314; RR =
sion criteria. See Table 2. All studies except for one (Arieli and Lepkifker, 0.61; 95% CI, 0.37–1.02; p = 0.06), with significant heterogeneity
1981) included bipolar II depressed patients and were outpatient trials, present (T2 = 0.19, I2 = 77%, Q = 13.3, df = 3, p = 0.004, PI = 0.07,
and all but one (Amsterdam and Shults, 2008) were double-blind. Three 5.37). See Fig. 3A. A sensitivity analysis limited to the three double-
studies (Altshuler et al., 2017; Amsterdam et al., 2016; Amsterdam and blind, randomized-controlled studies revealed a similar result (3
Shults, 2008) excluded patients with a history of nonresponse to lithium studies, N = 231; RR = 0.72; 95% CI, 0.44–1.16; p = 0.18; heteroge­
or the antidepressant. Mean lithium levels ranged from lows of 0.63 neity: T2 = 0.12, I2 = 72%, Q = 7.14, df = 2, p = 0.03, PI = 0.003,

271
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

Table 2
Lithium versus antidepressants in the treatment of bipolar depression.
Author/year Sample sizes Dose Lithium level Duration Scale Results Notes

Arieli and Bipolar I Lithium: Mean = 1.27 mEq/ 3 weeks HAMD, CGI Response rate (defined by -Double blind, parallel group,
Lepkifker, depressed 2000–2500 mg L JJR,MJL, BWD as at least RCT
1981 patients 50% reduction in baseline -Hospitalized
Lithium N = 3 Clomipramine: NR HAMD): Lithium- (2/3) = -Single-site
66% -1 week placebo washout
Clomipramine Clomipramine- (4/5) = -Reported percent improvement
N=5 80% on HAMD
Bias = 19 (Fair)
Amsterdam Bipolar II Lithium mean (SD) Mean (SD) = 0.64 12 HAMD-28, Response rate (HAMD-17): -Open label
and Shults, depressed dose: 966.24 mg (0.265) mEq/L weeks HAMD-17, Lithium- (8/40) 20% -Single site
2008 patients (410.9) HAMD-17R, -Outpatients
CGI-S, CGI-C Venlafaxine- (25/43) -Monotherapy
Lithium N = 40 Venlafaxine mean Range = 0.29–1.5 58.1% -No hx of non-response to either
Venlafaxine N (SD) dose: 185.6 mg mEq/L drug WITHIN CURRENT MDE
= 43 (92.04) -Lithium grp had more prior
hypomanic episodes, prior MDEs,
earlier onset, longer current MDE
duration
-Response: 50% reduction in
HAM-D
− 62.5% drop out in li grp vs.
20.9% drop out in Ven grp
Bias = 20 (Good)
Amsterdam Bipolar II Lithium mean max Mean maximum 12 HAMD, CGI- Response rate (at least 50% -Double blind, parallel group,
et al., 2016 depressed dose (SD): 1180.85 (SD) = 0.94 (0.38) weeks S reduction in baseline RCT
patients mg (399.14) mEq/L HAMD): Lithium- (22/64) -Single center
34.4% -Outpatients
Lithium N = 64 Venlafaxine mean Range: 0.3–2.4 Venlafaxine- (44/65) -No hx of non-response to either
Venlafaxine N max dose (SD): mEq/L 67.7% drug WITHIN CURRENT MDE
= 65 265.55 mg (101.07) -Response: >50% reduction in
baseline HAMD AND a final CGI-
S of 1,2,or 3.
− 42% were rapid cyclers
− 43.8% (li) vs. 15.4% (ven)
withdrew prematurely
Bias = 23 (Good)
Altshuler Bipolar II Lithium: minimum Mean = Overall: 16 IDS-C, CGI- Response rate (at least 50% -Double blind, parallel group,
et al., 2017 depressed target dose 900 mg 0.63 mEq/L weeks BP reduction in baseline IDS- RCT
patients C): -Multicenter
Lithium N = 49 Sertraline: Responders: 0.61 Lithium- (33/49) 67.4% -Outpatients
minimum target mEq/L -Primary outcome was switch
dose 100 mg rate
Sertraline N = Nonresponders: Sertraline- (33/45) 73.3% -no mixed symptoms allowed
45 0.51 mEq/L -Response = Decrease of 50% or
more OR a decrease of 2 or more
points on CGI-BP depression x 2
consecutive visits over 2 weeks;
could not be due to hypomanic
switch.
− 41.6% were rapid cyclers
Bias = 22 (Good)

Key: CGI-BP = Clinical Global Impression Scale for Bipolar Disorder, CGI-I = Clinical Global Impression Improvement Scale, CGI-S = Clinical Global Impression
Severity Scale, HAMD = Hamilton Depression Scale, IDS-C = Inventory of Depression Symptomatology-Clinician Rated, MDE = Major depressive episode, NR = Not
reported, RCT = Randomized controlled trial, SD = Standard deviation.

Table 3
Lithium versus other mood stabilizers in the treatment of bipolar depression.
Author/ Sample sizes Dose Lithium Duration Scale Results Notes
year level

Suppes Bipolar II Lithium: 900 mg Median = 16 HAMD, Response rate (at least -Open label
et al., depressed over two weeks 0.8 mEq/L weeks MADRS, CGI- 50% reduction in baseline -Outpatients
2008 patients BP, GAF HAMD): -Single blind (raters blind)
Lithium = 54 Lamotrigine: 200 Lithium- (30/54) 55.1% − 2 studies with combined analysis
subjects mg over eight weeks -More rapid cyclers in the Lamotrigine
group
Lamotrigine = Lamotrigine- (30/44) -item analysis shows significant
44 subjects 67.5%, difference Li > LTG for HAM-D psychic
anxiety and MADRS pessimistic thoughts
Bias = 20 (Good)

Key: CGI-BP = Clinical Global Impression Scale for Bipolar Disorder, HAMD = Hamilton Depression Scale, MADRS = Montgomery Asberg Depression Scale.

272
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

Table 4
Lithium versus second generation antipsychotics in the treatment of bipolar depression.
Author, Sample sizes Dose Li level Duration Scale Results Notes
year

Young Bipolar I and Lithium: Mean in ITT was 0.61 mEq/L; 34.9% 8 weeks MADRS, Response rate (at least -Double blind, parallel
et al., II patients 600–1800 mg below 0.6 mEq/L; 25% had a median HAMD, CGI 50% reduction in group, RCT
2010 level of 0.8 mEq/L or greater baseline MADRS): -Multicenter
Lithium N = QTP: 300 mg Lithium- (85/136) 62.5% -Outpatients
136 or 600 mg -Monotherapy
-Randomized 2:2:1:1
QTP 300 mg QTP 300 mg- (175/255) -MDE duration less than 1
N = 255 68.6%, year and 4 weeks or more
QTP 600 mg QTP 600 mg- (183/263) -had not failed 2 or more
N = 263 69.6% classes of antidepressants
Bias = 23 (Good)

Key: CGI = Clinical Global Impression Scale, HAMD = Hamilton Depression Scale, ITT = Intent to treat sample, MADRS = Montgomery Asberg Depression Scale, MDE
= Major depressive episode, QTP = Quetiapine, RCT = Randomized controlled trial.

155.3). See Fig. 3B. greater.


Response, which was based on a reduction of 50% or greater on the
3.4. Lithium vs. mood stabilizers MADRS score was 68.6% for quetiapine 300 mg (vs. placebo, p < 0.05),
69.6% for those receiving quetiapine 600 mg (vs. placebo, p < 0.01),
Only one publication comparing lithium to another mood stabilizer 62.5% for those receiving lithium (vs. placebo, p = 0.279), and 55.8%
was identified. See Table 3. In that study, Suppes et al. (2008) combined for those receiving placebo. The study was underpowered for a non-
the results of two similarly-designed, randomized, 16-week studies inferiority comparison between quetiapine and lithium, and the au­
comparing lithium monotherapy to lamotrigine monotherapy in bipolar thors explicitly stated that no such power calculations were made.
II depressed outpatients. Forty-four patients were randomized to receive Attrition from the study was similar for all treatment groups with
lamotrigine and 54 were randomized to receive lithium. Lamotrigine 72–75% of patients completing the study depending on the treatment
was titrated over eight weeks to 200 mg/day and could be increased to a group. The bias score for this study was 23 which was in the ‘Good’
maximum dose of 400 mg/day (median dose was 250 mg/day). Lithium range.
was titrated to 900 mg/day by week two and was dosed to blood levels
from 0.8 to 1.2 mEq/L (median lithium level was 0.8 mEq/L). Patients 3.6. Bipolar depression vs. unipolar depression
and physicians were unblinded while raters were blinded.
Response rates, defined as a 50% reduction on the HAMD-17, were The search identified ten publications (Baron et al., 1975; Donnelly
67.5% (30/44) for the lamotrigine group and 55.1% (30/54) in the et al., 1978; Goodwin et al., 1969; Goodwin et al., 1972; Johnson, 1974;
lithium group, revealing no significant difference (no p-values pro­ Kramlinger and Post, 1989; Mendels, 1976; Mendels and Frazer, 1973;
vided). The same was true of the primary outcome comparing change in Noyes et al., 1974; Price et al., 1986) that compared outcomes in pa­
HAMD-17 scores between both treatment groups (F(7,402) = 0.54, p = tients with bipolar depression versus those with unipolar depression
0.95). Limitations to this study included the lack of a placebo group, the receiving lithium. See Table 5. All except for two were monotherapy
single-blind design, and a modest sample size. The study was signifi­ studies (Kramlinger and Post, 1989; Price et al., 1986). All were double-
cantly underpowered as indicated by the authors' statement that over blinded except for one that was single-blinded (Johnson, 1974), one that
10,000 subjects were needed to find a significant difference between the was blinded for only some of the patients (Price et al., 1986), and two
groups. Additionally, there was a large number of participant drop-outs that provided no information about rater blinding (Mendels and Frazer,
with more individuals in the lithium group leaving the study prema­ 1973; Noyes et al., 1974). All included hospitalized patients except for
turely (completer rate = lamotrigine 51% vs. lithium 39%, p = 0.29). one which included a mix of inpatients and outpatients (Price et al.,
The bias score for this study was 20 which was in the ‘Good’ range. 1986), and all were cross-over, placebo-controlled studies except for two
that were parallel-group treatment studies (Mendels and Frazer, 1973;
3.5. Lithium vs. antipsychotics Price et al., 1986). For many of these studies, the target range or actual
range of lithium levels was provided rather than a mean lithium level for
Only one publication comparing lithium to an antipsychotic was all subjects. Only one study (Price et al., 1986) had a range that was
identified. See Table 4. In that study (known as EMBOLDEN I), Young lower than 0.7 mEq/L while seven studies (Donnelly et al., 1978;
et al. (2010) compared two separate doses of quetiapine monotherapy to Goodwin et al., 1969; Johnson, 1974; Mendels, 1976; Mendels and
lithium monotherapy and to placebo in an eight week, multisite study of Frazer, 1973; Noyes et al., 1974; Price et al., 1986) had a range that was
bipolar I and II outpatients with depression. As part of the inclusion higher than 1.2 mEq/L. Duration of time on lithium ranged from ‘at least
criteria, participants could not have failed two or more classes of anti­ 10 days’ (Price et al., 1986) to ‘up to 28 days’ (Donnelly et al., 1978;
depressants during the current episode and must have been depressed Mendels and Frazer, 1973). The number of placebo phases in the cross-
for at least four weeks and no longer than one year. Patients were ran­ over trials varied from as few as one (Donnelly et al., 1978; Johnson,
domized in a 2:2:1:1 manner resulting in 255 patients receiving que­ 1974; Kramlinger and Post, 1989) to as many as eight (Goodwin et al.,
tiapine 300 mg/day, 263 receiving quetiapine 600 mg/day, 136 patients 1969). In three of those with multiple placebo phases (Baron et al., 1975;
receiving lithium, and 129 patients receiving placebo. Quetiapine was Goodwin et al., 1969; Goodwin et al., 1972), response required
titrated from a starting dose of 50 mg to target doses by day four (300 improvement during the lithium phase and then worsening during
mg) and day eight (600 mg), while lithium was started at 600 mg/day subsequent placebo phases. All of the other studies (i.e., those with
and increased to 900 mg/day by day four. Lithium was dosed to reach a multiple-placebo phases, single placebo phase, open-label) defined
target lithium level between 0.6 and 1.2 mEq/L; however, the mean response as improvement relative to baseline. The bias scores for these
median serum concentration was 0.61 mEq/L in the intent to treat studies were 10–16 which fell within the ‘Poor’ to ‘Fair’ range
population and 34.9% had median levels below 0.6 mEq/L. Only a depending on the study.
quarter of the patients had a median lithium level that was 0.8 mEq/L or The meta-analysis showed a non-statistically significant benefit for

273
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

Table 5
Lithium in the treatment of bipolar depression vs major depressive disorder.
Author, year Sample sizes Dose Lithium level Duration Scale Results Notes

Goodwin 13 Manic- Lithium: Initial dose Target Range NR for Lithium or Bunney- Manic-depression: -Cross-over study (placebo- >
et al., 1969 depressive, 900–1800 mg = 0.8–1.3 placebo phase Hamburg Global (3/13) 23% Lithium- > 1–8 additional
depressed, mEq/L Rating Scale placebo periods)
Type-1 = 6, -Hospitalized
Type-II = 6 -Monotherapy
Recurrent Unequivocal/ -Complete and unequivocal =
MDEs = 1 complete response complete remission of
5 Non-cyclic Non-cyclic depressive symptoms within 2
depression depression: (0/5) weeks of Li and return of
0% symptoms during placebo
period
Unequivocal/ Bias = 13 (Poor)
complete response
Goodwin 40 Bipolar NR NR Lithium: At least 2 Bunney- Bipolar: (12/40) -Cross-over study (placebo- >
et al., 1972 weeks Hamburg 30% Unequivocal Lithium- > placebo +/−
response additional lithium period)
12 Unipolar Placebo: At least 6 Global Rating Unipolar: (2/12) -Hospitalized
days Scale 17% Unequivocal -Monotherapy
response -Unequivocal response = 3
point improvement on lithium
and relapse on placebo
Bias = 14 (Poor)
Mendels and Manic Lithium: mean dose Range = 4 weeks HAMD, BDI, Bipolar: (6/9) 60% -Open label study
Frazer, depression, 1200–2589 mg 0.78–1.37 nurses Improved -Hospitalized
1973 (bipolar) = 9 mEq/L observation -Biological study
Recurrent rating Unipolar: (1/4) -Monotherapy
depressive 25% Improved -Does not indicate whether
illness evaluators were blind to
(unipolar) = 4 diagnosis
-No comorbidities allowed
-One patient treated for two
separated episodes and
improved both times. We only
counted one of these episodes
in this analysis
-Improved = 70–75%
improvement on 2/3 rating
scales and ability to be
discharged home
Bias = 10 (Poor)
Johnson, 2 Manic- Lithium: initial dose Range = Lithium: at least 3 HAMD, Bipolar: (1/2) 50% -Cross-over study (placebo- >
1974 depressive 1000–1500 mg 0.7–1.9 mEq/L weeks psychiatrist's Marked response Lithium)
circular, clinical global but then became -Hospitalized pts
depressed phase impression hypomanic -Single Blind design
(bipolar) Bias = 14 (Poor)
6 Manic- Placebo: 9 days Unipolar and
depressive, depressive
depressed neurotics: (4/9)
(unipolar) 44% Marked
3 Depressive response
neurotics
Noyes et al., 6 Manic Lithium: 26 mg/kg Mean = 1.42 Lithium: at least 2 HAMD Bipolar: (6/6) -Cross-over study (placebo- >
1974 depressive, mEq/L weeks + time to 100% Responded Lithium- > placebo)
depressed phase reach plateau in -Hospitalized pts
improvement -Monotherapy
16 Unipolar Range = Placebo: 3–5 days Unipolar: (7/16) -Does not indicate whether
depressed 0.93–2.15 initially 44% Responded evaluators were blinded.
mEq/L -Does not define response
Bias = 16 (Fair)
Baron et al., 8 Bipolar Type- NR Mean = 0.87 Lithium: 19 days Bunney- Bipolar: (4/8) 50% -Cross-over study (placebo- >
1975 1 = 7 Type-II = mEq/L Hamburg Global Responded Lithium- > placebo)
1 Rating scale unequivocally -Hospitalized pts
-Monotherapy
10 Unipolar Range = Placebo: 1st − 1-2 Unipolar: (1/10) -Unequivocal response=
0.8–1.0 mEq/L weeks 2nd − 2 weeks 10% Responded 2 point improvement on
unequivocally lithium and 2 point worsening
on placebo
Bias = 16 (Fair)
Mendels, 13 Bipolar Lithium: Up to max Maximum = Lithium: 3 weeks HAMD, BDI, Bipolar: (9/13) -Cross-over study (placebo- >
1976 depressed blood level, clinical 1.5 mEq/L Bunney- 69% Responded Lithium- > placebo)
effect, or toxicity Hamburg -Hospitalized pts
8 Unipolar Placebo: 1st − 7-15 Global Rating Unipolar: (4/8) -Response = sufficient relief of
days, 2nd − 7-22 Scale 50% Responded symptoms that did not require
days any other form of active
(continued on next page)

274
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

Table 5 (continued )
Author, year Sample sizes Dose Lithium level Duration Scale Results Notes

treatment and able to return to


premorbid level of function
Bias = 13 (Poor)
Donnelly 33 Bipolar Lithium: median dose Range = Lithium: 4 weeks Scale created by Bipolar: (21/33) -Cross-over study (placebo- >
et al., 1978 Type-1 = 17 1500 mg 0.9–1.3 mEq/L authors 64% Responded Lithium)
Type-II = 16 -Hospitalized pts
-Monotherapy
20 Unipolar Placebo: 5 days Unipolar: (8/20) -Responder = t value >2.26
40% Responded and a rating decrease of at
least 2 points on Lithium vs.
placebo
Bias = 16 (Fair)
Price et al., 11 Bipolar Lithium: initial dose Target range Lithium: at least 10 SCRS global Bipolar: (4/11) -Open label, augmentation
1986 900–1500 mg = 0.5–1.3 days depression item 36% Responded following antidepressant cross
mEq/L over (placebo <− >
Antidepressant)
73 Unipolar Antidepressant: 4–6 HAMD Unipolar: (43/73) -Hospitalized and outpatients
weeks 59% -Antidepressant: TCAs, SSRIs,
atypicals
Placebo: 2–3 weeks, Responded -Lithium augmentation for
those who did not respond to
antidepressant
-Only 15 patients were
blinded to Lithium
augmentation
-Response = global functional
improvement to permit
discharge from hospital or
research clinic and no major
med changes for inpatients 4
weeks after discharge
Bias = 12 (Poor)
Kramlinger 13 Bipolar Lithium: Mean (SD) Mean (SD) Lithium: 3 weeks Bunney- Bipolar: (6/13) -Cross-over following
and Post, Type-1 = 9 Responders = 975 Responders = Hamburg Global 46% Responded carbamazepine non-response
1989 Type-II = 4 (139) mg 0.67 (0.06) Rating Scale (placebo- > Lithium)
mEq/L -Hospitalized
2 Unipolar Non-responders = Non- Placebo: 1 week Unipolar: (2/2) -Continued carbamazepine
793 (248) mg responders = 100% Responded -Compared to historical
Carbamazepine: 0.69 (0.28) controls of mixed sample
mean dose 888 mg mEq/L treated with lithium alone
Target range -Good response = Mean
= 0.7–1.2 weekly decrease of 2 points or
mEq/L more from baseline placebo
week
Bias = 16 (Fair)

Key: SCRS = BDI = Beck Depression Inventory, HAMD = Hamilton Depression Scale, NR = Not reported, SD = Standard deviation, Short Clinical Rating Scale, SSRI =
Selective serotonin reuptake inhibitor, TCA = tricyclic antidepressant.

Fig. 2. Forest Plot for Studies of Lithium versus Placebo in the Treatment of Acute Bipolar Depression. Key: A = Main analysis, B = Sensitivity analysis limited to
double-blinded, monotherapy, randomized controlled trials, Events = Number of patients who met response criteria.

275
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

Fig. 3. Forest Plot for Studies of Lithium versus Antidepressants in the Treatment of Acute Bipolar Depression. Key: A = Main analysis, B = Sensitivity analysis
limited to double-blinded, monotherapy, randomized controlled trials, Events = Number of patients who met response criteria.

lithium in bipolar depression compared to unipolar depression patients 95% CI, − 0.14 to − 0.04; p = 0.0005). See Fig. 5. The publication bias
(10 studies, N = 307; RR = 1.33; 95% CI, 0.89–1.97; p = 0.16; hetero­ assessment, also conducted on the original set of ten studies, revealed
geneity: T2 = 0.13, I2 = 37%, Q = 14.4, df = 9, p = 0.11, PI = 0.51, 3.48). that the effect size was larger in the smaller studies. See Fig. 6. The “trim
See Fig. 4A. When only double-blinded, monotherapy, cross-over studies and fill” method was used to determine the effect size based on the
were included, the sensitivity analysis revealed a statistically significant included studies and those that could have been missing, resulting in a
greater response rate to lithium in bipolar depressed patients compared smaller overall effect size that still lacked statistical significance (RR =
to unipolar depression patients (5 studies, N = 162; RR = 1.65, 95% CI, 1.20; 95% CI, 0.82–1.75).
1.08–2.55; p = 0.02; heterogeneity: T2 = 0, I2 = 0%, Q = 1.7, df = 4, p =
0.79). See Fig. 4B. 4. Discussion
A meta-regression analysis was performed on the original set of ten
studies and only one covariate, date of publication, was selected given This systematic review and meta-analysis is the first to focus exclu­
this small number of studies. This analysis revealed that date of publi­ sively on lithium in the treatment of acute bipolar depression. The
cation had a negative impact on the overall effect size (B = − 0.0876; overall summary effects reveal that lithium performed numerically

Fig. 4. Forest Plot for Studies of Lithium in the Treatment of Bipolar Depressed versus Unipolar Depressed Patients. Key: A = Main analysis, B = Sensitivity analysis
limited to double-blinded, monotherapy, cross-over studies, BD = Bipolar disorder, UD = Unipolar depression Events = Number of patients who met
response criteria.

276
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

3.00

2.50

2.00

1.50

1.00
Log risk ratio

0.50

0.00

-0.50

-1.00

-1.50

-2.00

-2.50

1963 1965 1968 1970 1973 1975 1978 1980 1983 1985 1988 1990 1993 1995

Date of Publication

Fig. 5. Meta-regression of Log Risk Ratio on Date of Publication for Studies of Lithium in the Treatment of Bipolar Depressed versus Unipolar Depressed Patients.

Funnel Plot of Standard Error by Log risk ratio


0.0

0.5
Standard Error

1.0

1.5

2.0

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0

Log risk ratio

Fig. 6. Funnel Plot for Studies of Lithium in the Treatment of Bipolar Depressed versus Unipolar Depressed Patients.

worse than antidepressants but better than placebo, although neither of increases the likelihood that the true overall mean effect favors anti­
these findings were statistically significant. Although the comparison of depressants and that a statistically significant result may well have been
lithium's efficacy for bipolar versus unipolar depression did not reveal achieved with greater statistical power. The sensitivity analysis limited
lithium to be statistically superior in bipolar patients, the sensitivity to double-blind randomized controlled trials resulted in a 95% confi­
analysis limited to double-blinded, monotherapy, cross-over studies dence interval (0.44–1.16) that was shifted rightward but still with a
revealed a statistically significant result supporting lithium's superior majority of values below one, reinforcing the likelihood that the true
efficacy for those with bipolar depression. Too few studies were avail­ overall mean effect favors antidepressants. However, the majority of
able to conduct a meta-analysis of lithium versus mood stabilizers or these studies exclusively enrolled bipolar II patients and were conducted
lithium versus second generation antipsychotics. in an outpatient setting, thereby limiting the ability to generalize these
Regarding the comparison with antidepressants, although the result results to either bipolar I patients or those severe enough to require
of the meta-analysis was not significant, nearly all of the values included inpatient treatment. It is possible that lithium is inherently less effective
within the 95% confidence interval (0.37–1.02) were less than one. This for bipolar II than bipolar I depression, but this remains to be proven.

277
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

The lithium levels were potentially inadequate for treatment of bipolar lamotrigine was compared and only one publication evaluating this
depression, with mean levels <0.8 mEq/L in two of the four studies. comparison was identified, thus precluding a meta-analysis. No signifi­
However, as of yet, there is no established range of lithium levels that cant difference was seen between the treatment groups, although there
predict antidepressant response. Attrition rates were strikingly uneven were a number of methodologic elements that might have led to this null
between treatment groups in most of the studies, with more lithium result. Regarding lithium versus second generation antipsychotics, only
treated patients withdrawing prematurely. Although intent to treat ef­ one antipsychotic, quetiapine was compared and only one publication
ficacy analyses take discontinuation into account, the higher attrition evaluating this comparison was found, although there were two que­
rate may mask the degree of improvement achievable among patients tiapine treatment arms. The two quetiapine groups outperformed the
who are able to tolerate lithium adequately to complete the trial. lithium group, but the sample sizes were uneven and only 25% of the
Completer analyses including patients who achieved effective lithium lithium-treated patients had a median lithium level of 0.8 mEq/L or
blood levels may provide a clearer answer, but these data are not greater.
available. Taken together, these meta-analyses do not support the use of
In addition, differences in manic induction and cycle acceleration by lithium as a first-line treatment for acute bipolar depression. On the
the drug are a very important consideration when making judgments other hand, the sensitivity analysis demonstrating lithium's specificity
about the relative value of lithium versus antidepressants. Although the for bipolar depression over unipolar depression, and the modest though
included studies found no differences in (hypo)manic symptom emer­ non-significant advantage over placebo suggest lithium may still be a
gence or cycle acceleration between groups, confidence in these results viable treatment option. Larger and more rigorously-designed studies
is limited for several reasons: 1) nearly all of the subjects had bipolar are needed to understand lithium's efficacy more definitively versus
disorder type II, not type I (Altshuler et al., 2006), 2) the duration of the placebo and its position relative to other psychotropics within the
trials was too brief to capture cycle acceleration, and 3) those patients treatment guidelines. Studies that explore the role of moderators such as
with a history of antidepressant-induced mood switching likely would episode number (Swann et al., 1999), mixed features (Swann et al.,
have declined participation in these studies. Only three antidepressants 1997), and depression polarity proneness (Popovic et al., 2012), may
were included in this meta-analysis, limiting generalizations to the many shed further light on lithium's antidepressant effect as has been the case
other FDA-approved antidepressants available. It should be noted that regarding its anti-manic and prophylactic abilities.
the current meta-analyses excluded one study (Fieve et al., 1968) that is Limitations to this work include the study selection rules, which
commonly included in reviews, which showed a stronger antidepressant excluded publications that might have resulted in different overall
effect for imipramine than for lithium. Exclusion of this study was summary effects. However, these rules were established a priori to
justified because it enrolled patients with manic-depressive psychosis- identify publications that would be reasonably similar enough to
depressed type, a diagnosis which is synonymous with unipolar combine in a meta-analysis. The decision to focus on response rates
depression and not bipolar disorder. rather than remission rates or continuous score outcomes might have
Compared to placebo, lithium showed a modest benefit though the influenced the results; however, response rates were more commonly
results were not statistically significant as the 95% confidence interval reported, allowing the authors to increase the number of included
(0.99–1.41) just barely crossed over one. The sensitivity analysis limited studies. The number of studies included in each meta-analysis were
to studies evaluating monotherapy treatment resulted in a 95% confi­ small, potentially leading to type I and II errors and precluding the au­
dence interval (0.93–1.38) that was shifted leftward only slightly, thors from using publication bias tests for most of the analyses. Within-
reinforcing the likelihood that the true overall mean effect favors study bias ratings were low (Poor-Fair) for the studies included in the
lithium. Unfortunately, both analyses were comprised of only a few bipolar versus unipolar depression comparisons but high (Fair to Good)
studies and the majority of patients came from a single study. One study for the rest. Although there was evidence of heterogeneity in a few of the
(Stokes et al., 1971) that is often included in reviews and that showed meta-analyses, the PI's for the lithium versus antidepressant studies
that lithium was not significantly more effective than placebo was suggest that the majority of studies within the population of comparable
excluded from this meta-analysis because it was a cross-over study and studies would have effect sizes that favor antidepressants and align with
presented results in terms of percentage of treatment periods with the overall summary effect. The PI's for the bipolar depression versus
improvement. unipolar depression studies included effect sizes favoring both bipolar
Regarding the comparison between bipolar depression patients and and unipolar depression groups, however, there was no significant
those with unipolar depression receiving lithium, the meta-analysis heterogeneity in the sensitivity analysis which included the more
favored bipolar depressed patients although the result was not statisti­ rigorously conducted studies.
cally significant. On the other hand, the sensitivity analysis limited to In summary, these meta-analyses demonstrated that lithium was
the more rigorous trials did find a significant difference in response preferentially beneficial for patients with bipolar depression, was
rates, likely resulting from greater homogeneity and methodological numerically more efficacious than placebo, and numerically less effi­
quality of the included studies. This is reflected in I2 scores that dropped cacious than antidepressants for bipolar II depressed outpatients. More
from 35% in the meta-analysis to 0% in the sensitivity analysis. That studies and additional meta-analyses are needed to confirm these results
lithium has greater efficacy in patients with bipolar depression over and determine the full range of lithium efficacy in the treatment of acute
those with unipolar depression replicates the findings of an earlier meta- bipolar depression.
analysis (Selle et al., 2014), and the conclusion of a review article
(Mendels et al., 1979), and two editions of a textbook (Goodwin and CRediT authorship contribution statement
Jamison, 1990; Goodwin et al., 2007) that addressed this question.
The bipolar depression versus unipolar meta-analysis was the only JJR and MJL performed the literature search and data extraction.
one in this manuscript that contained ten or more studies, making a All authors (JJR, MJL, BWD) participated in the interpretation of the
meta-regression analysis and publication bias assessments appropriate. results.
The meta-regression suggested that date of publication might explain JJR wrote the first draft of the manuscript and all authors (JJR, MJL,
some of the heterogeneity in individual study effect sizes which might be BWD) contributed equally to subsequent revisions and have approved
due to differences in lithium dosing or study designs over time. The the final draft.
publication bias assessment identified a risk of missing studies but the
trim and fill method did not change the overall result in a meaningful Role of the funding source
way.
Regarding lithium vs. mood stabilizers, only one mood stabilizer, This research did not receive any specific grant from funding

278
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

agencies in the public, commercial, or not-for-profit sectors. Downs, S.H., Black, N., 1998. The feasibility of creating a checklist for the assessment of
the methodological quality both of randomised and non-randomised studies of
health care interventions. J. Epidemiol. Community Health 52, 377–384.
Conflict of Interest Duval, S., Tweedie, R., 2000. Trim and fill: a simple funnel-plot-based method of testing
and adjusting for publication bias in meta-analysis. Biometrics 56, 455–463.
The authors have no affiliation with any organization with a direct or Ebert, D., Jaspert, A., Murata, H., Kaschka, W.P., 1995. Initial lithium augmentation
improves the antidepressant effects of standard TCA treatment in non-resistant
indirect financial interest in the subject matter discussed in the depressed patients. Psychopharmacology 118, 223–225.
manuscript. Fieve, R.R., Platman, S.R., Plutchik, R.R., 1968. The use of lithium in affective disorders.
JJR- research support from Compass Pathways, Otsuka, consulting I.Acute endogenous depression. Am. J. Psychiatry 125, 487–491.
Goodwin, F.K., Jamison, K.R., 1990. Manic-depressive Illness. Oxford University Press,
fees from Eleven10, and honoraria from FOCUS and SMI: Clinical New York.
Advisor. Goodwin, F.K., Murphy, D.L., Bunney Jr., W.E., 1969. Lithium-carbonate treatment in
MJL- none. depression and mania. A longitudinal double-blind study. Arch. Gen. Psychiatry 21,
486–496.
BWD- research support from Acadia, Compass, Aptinyx, NIMH, Sage, Goodwin, F.K., Murphy, D.L., Dunner, D.L., Bunney Jr., W.E., 1972. Lithium response in
Otsuka, and Takeda, and has served as a consultant to Greenwich Bio­ unipolar versus bipolar depression. Am. J. Psychiatry 129, 44–47.
sciences, Myriad Neuroscience, Otsuka, Sage, and Sophren Goodwin, F.K., Jamison, K.R., Ghaemi, S.N., 2007. Manic-depressive Illness: Bipolar
Disorders And Recurrent Depression, 2nd ed. Oxford University Press, New York, N.
Therapeutics. Y.
Goodwin, G.M., Haddad, P.M., Ferrier, I.N., Aronson, J.K., Barnes, T., Cipriani, A.,
Acknowledgments Coghill, D.R., Fazel, S., Geddes, J.R., Grunze, H., Holmes, E.A., Howes, O.,
Hudson, S., Hunt, N., Jones, I., Macmillan, I.C., McAllister-Williams, H.,
Miklowitz, D.R., Morriss, R., Munafo, M., Paton, C., Saharkian, B.J., Saunders, K.,
None. Sinclair, J., Taylor, D., Vieta, E., Young, A.H., 2016. Evidence-based guidelines for
treating bipolar disorder: revised third edition recommendations from the British
References Association for Psychopharmacology. J. Psychopharmacol. 30, 495–553.
Grunze, H., Vieta, E., Goodwin, G.M., Bowden, C., Licht, R.W., Moller, H.J., Kasper, S.,
WFSBP Taskforce on Treatment Guidelines for Bipolar Disorders, 2010. The World
Altshuler, L.L., Frye, M.A., Gitlin, M.J., 2003. Acceleration and augmentation strategies Federation of Societies of Biological Psychiatry (WFSBP) guidelines for the biological
for treating bipolar depression. Biol. Psychiatry 53, 691–700. treatment of bipolar disorders: update 2010 on the treatment of acute bipolar
Altshuler, L.L., Suppes, T., Black, D.O., Nolen, W.A., Leverich, G., Keck Jr., P.E., Frye, M. depression. World J. Biol. Psychiatry 11, 81–109.
A., Kupka, R., McElroy, S.L., Grunze, H., Kitchen, C.M., Post, R., 2006. Lower switch Hamilton, M., 1960. A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23,
rate in depressed patients with bipolar II than bipolar I disorder treated adjunctively 56–62.
with second-generation antidepressants. Am. J. Psychiatry 163, 313–315. Ioannidis, J.P., Trikalinos, T.A., 2007. The appropriateness of asymmetry tests for
Altshuler, L.L., Sugar, C.A., McElroy, S.L., Calimlim, B., Gitlin, M., Keck Jr., P.E., Aquino- publication bias in meta-analyses: a large survey. CMAJ 176, 1091–1096.
Elias, A., Martens, B.E., Fischer, E.G., English, T.L., Roach, J., Suppes, T., 2017. Johnson, G., 1974. Antidepressant effect of lithium. Compr. Psychiatry 15, 43–47.
Switch rates during acute treatment for bipolar II depression with lithium, sertraline, Johnson, G.F., 1987. Lithium in depression: a review of the antidepressant and
or the two combined: a randomized double-blind comparison. Am. J. Psychiatry 174, prophylactic effects of lithium. Aust. N. Z. J. Psychiatry 21, 356–365.
266–276. Judd, L.L., Akiskal, H.S., Schettler, P.J., Endicott, J., Maser, J., Solomon, D.A., Leon, A.C.,
American Psychiatric Association, 2010. Practice Guideline for the Treatment of Patients Rice, J.A., Keller, M.B., 2002. The long-term natural history of the weekly
With Bipolar Disorder, 2nd ed. symptomatic status of bipolar I disorder. Arch. Gen. Psychiatry 59, 530–537.
Amsterdam, J.D., Shults, J., 2008. Comparison of short-term venlafaxine versus lithium Judd, L.L., Akiskal, H.S., Schettler, P.J., Coryell, W., Endicott, J., Maser, J.D.,
monotherapy for bipolar II major depressive episode: a randomized open-label Solomon, D.A., Leon, A.C., Keller, M.B., 2003. A prospective investigation of the
study. J. Clin. Psychopharmacol. 28, 171–181. natural history of the long-term weekly symptomatic status of bipolar II disorder.
Amsterdam, J.D., Lorenzo-Luaces, L., Soeller, I., Li, S.Q., Mao, J.J., DeRubeis, R.J., 2016. Arch. Gen. Psychiatry 60, 261–269.
Short-term venlafaxine v. lithium monotherapy for bipolar type II major depressive Judd, L.L., Akiskal, H.S., Schettler, P.J., Endicott, J., Leon, A.C., Solomon, D.A.,
episodes: effectiveness and mood conversion rate. Br. J. Psychiatry 208, 359–365. Coryell, W., Maser, J.D., Keller, M.B., 2005. Psychosocial disability in the course of
Arieli, A., Lepkifker, E., 1981. The antidepressant effect of lithium. Curr. Dev. bipolar I and II disorders: a prospective, comparative, longitudinal study. Arch. Gen.
Psychopharmacol. 6, 165–190. Psychiatry 62, 1322–1330.
Austin, M.P., Souza, F.G., Goodwin, G.M., 1991. Lithium augmentation in Kelly, T., 2019. Lithium and the Woozle effect. Bipolar Disord. 21, 302–308.
antidepressant-resistant patients.A quantitative analysis. Br. J. Psychiatry 159, Kramlinger, K.G., Post, R.M., 1989. The addition of lithium to carbamazepine.
510–514. Antidepressant efficacy in treatment-resistant depression. Arch. Gen. Psychiatry 46,
Bahji, A., Ermacora, D., Stephenson, C., Hawken, E.R., Vazquez, G., 2020. Comparative 794–800.
efficacy and tolerability of pharmacological treatments for the treatment of acute Mendels, J., 1976. Lithium in the treatment of depression. Am. J. Psychiatry 133,
bipolar depression: a systematic review and network meta-analysis. J. Affect. Disord. 373–378.
269, 154–184. Mendels, J., Frazer, A., 1973. Intracellular lithium concentration and clinical response:
Baldessarini, R.J., Tondo, L., Pinna, M., Nunez, N., Vazquez, G.H., 2019. Suicidal risk towards a membrane theory of depression. J. Psychiatr. Res. 10, 9–18.
factors in major affective disorders. Br. J. Psychiatry 1–6. Mendels, J., Ramsey, T.A., Dyson, W.L., Frazer, A., 1979. Lithium as an antidepressant.
Baron, M., Gershon, E.S., Rudy, V., Jonas, W.Z., Buchsbaum, M., 1975. Lithium Arch. Gen. Psychiatry 36, 845–846.
carbonate response in depression. Prediction by unipolar/bipolar illness, average- Merikangas, K.R., Akiskal, H.S., Angst, J., Greenberg, P.E., Hirschfeld, R.M.,
evoked response, catechol-O-methyl transferase, and family history. Arch. Gen. Petukhova, M., Kessler, R.C., 2007. Lifetime and 12-month prevalence of bipolar
Psychiatry 32, 1107–1111. spectrum disorder in the National Comorbidity Survey replication. Arch. Gen.
Bauer, M., Dopfmer, S., 1999. Lithium augmentation in treatment-resistant depression: Psychiatry 64, 543–552.
meta-analysis of placebo-controlled studies. J. Clin. Psychopharmacol. 19, 427–434. Michalak, E.E., Yatham, L.N., Lam, R.W., 2005. Quality of life in bipolar disorder: a
Bhagwagar, Z., Goodwin, G.M., 2002. The role of lithium in the treatment of bipolar review of the literature. Health Qual. Life Outcomes 3, 72.
depression. Clin. Neurosci. Res. 2, 222–227. Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., Group, P., 2009. Preferred reporting
Borenstein, M., Hedges, L., Higgins, J., Rothstein, H., 2009. Introduction to Meta- items for systematic reviews and meta-analyses: the PRISMA statement. Ann. Intern.
analysis. John Wiley & Sons Ltd, West Sussex. Med. 151, 264–269. W264.
Borenstein, M., Hedges, L., Higgins, J., Rothstein, H., 2013. Comprehensive Meta- Montgomery, S.A., Asberg, M., 1979. A new depression scale designed to be sensitive to
analysis Version 3. Biostat, Englewood, NJ. change. Br. J. Psychiatry 134, 382–389.
Cade, J.F., 1949. Lithium salts in the treatment of psychotic excitement. Med. J. Aust. 2, Murray, C.J., Lopez, A.D., 1997. Global mortality, disability, and the contribution of risk
349–352. factors: Global Burden of Disease Study. Lancet 349, 1436–1442.
<collab>Roxane Laboratories, collab, 2011. Lithium Carbonate (Lithium Carbonate) Noyes Jr., R., Dempsey, G.M., Blum, A., Cavanaugh, G.L., 1974. Lithium treatment of
[package insert]. Roxane Laboratories, Inc., Columbus, Ohio. depression. Compr. Psychiatry 15, 187–193.
Crump, C., Sundquist, K., Winkleby, M.A., Sundquist, J., 2013. Comorbidities and Popovic, D., Reinares, M., Goikolea, J.M., Bonnin, C.M., Gonzalez-Pinto, A., Vieta, E.,
mortality in bipolar disorder: a Swedish national cohort study. JAMA Psychiatry 70, 2012. Polarity index of pharmacological agents used for maintenance treatment of
931–939. bipolar disorder. Eur. Neuropsychopharmacol. 22, 339–346.
Deeks, J.J., Higgins, J.P.T., Altman, D.G., 2021. Chapter 10: analysing data and Price, L.H., Charney, D.S., Heninger, G.R., 1986. Variability of response to lithium
undertaking meta-analyses. In: Higgins, J.P.T., Thomas, J., Chandler, J., augmentation in refractory depression. Am. J. Psychiatry 143, 1387–1392.
Cumpston, M., Li, T., Page, M.J., Welch, V.A. (Eds.), Cochrane Handbook for Rosa, A.R., Gonzalez-Ortega, I., Gonzalez-Pinto, A., Echeburua, E., Comes, M., Martinez-
Systematic Reviews of Interventions Version 6.2. Aran, A., Ugarte, A., Fernandez, M., Vieta, E., 2012. One-year psychosocial
Donnelly, E.F., Goodwin, F.K., Waldman, I.N., Murphy, D.L., 1978. Prediction of functioning in patients in the early vs. late stage of bipolar disorder. Acta Psychiatr.
antidepressant responses to lithium. Am. J. Psychiatry 135, 552–556. Scand. 125, 335–341.

279
J.J. Rakofsky et al. Journal of Affective Disorders 308 (2022) 268–280

Selle, V., Schalkwijk, S., Vazquez, G.H., Baldessarini, R.J., 2014. Treatments for acute The Cochrane Collaboration, 2020. Review Manager (RevMan) [Computer Program].
bipolar depression: meta-analyses of placebo-controlled, monotherapy trials of Version 5.4. The Cochrane Collaboration.
anticonvulsants, lithium and antipsychotics. Pharmacopsychiatry 47, 43–52. Vieta, E., Locklear, J., Gunther, O., Ekman, M., Miltenburger, C., Chatterton, M.L.,
Srisurapanont, M., Yatham, L.N., Zis, A.P., 1995. Treatment of acute bipolar depression: Astrom, M., Paulsson, B., 2010. Treatment options for bipolar depression: a
a review of the literature. Can. J. Psychiatr. 40, 533–544. systematic review of randomized, controlled trials. J. Clin. Psychopharmacol. 30,
Stokes, P.E., Shamoian, C.A., Stoll, P.M., Patton, M.J., 1971. Efficacy of lithium as acute 579–590.
treatment of manic-depressive illness. Lancet 1, 1319–1325. World Health Organization, 1968. International Classification of Diseases. Geneva.
Suppes, T., Marangell, L.B., Bernstein, I.H., Kelly, D.I., Fischer, E.G., Zboyan, H.A., Yatham, L.N., Kennedy, S.H., Parikh, S.V., Schaffer, A., Bond, D.J., Frey, B.N.,
Snow, D.E., Martinez, M., Al Jurdi, R., Shivakumar, G., Sureddi, S., Gonzalez, R., Sharma, V., Goldstein, B.I., Rej, S., Beaulieu, S., Alda, M., MacQueen, G., Milev, R.V.,
2008. A single blind comparison of lithium and lamotrigine for the treatment of Ravindran, A., O'Donovan, C., McIntosh, D., Lam, R.W., Vazquez, G., Kapczinski, F.,
bipolar II depression. J. Affect. Disord. 111, 334–343. McIntyre, R.S., Kozicky, J., Kanba, S., Lafer, B., Suppes, T., Calabrese, J.R., Vieta, E.,
Swann, A.C., Bowden, C.L., Morris, D., Calabrese, J.R., Petty, F., Small, J., Dilsaver, S.C., Malhi, G., Post, R.M., Berk, M., 2018. Canadian Network for Mood and Anxiety
Davis, J.M., 1997. Depression during mania. Treatment response to lithium or Treatments (CANMAT) and International Society for Bipolar Disorders (ISBD) 2018
divalproex. Arch. Gen. Psychiatry 54, 37–42. guidelines for the management of patients with bipolar disorder. Bipolar Disord. 20,
Swann, A.C., Bowden, C.L., Calabrese, J.R., Dilsaver, S.C., Morris, D.D., 1999. 97–170.
Differential effect of number of previous episodes of affective disorder on response to Young, A.H., McElroy, S.L., Bauer, M., Philips, N., Chang, W., Olausson, B., Paulsson, B.,
lithium or divalproex in acute mania. Am. J. Psychiatry 156, 1264–1266. Brecher, M., Investigators, E.I., 2010. A double-blind, placebo-controlled study of
Taylor, D.M., Cornelius, V., Smith, L., Young, A.H., 2014. Comparative efficacy and quetiapine and lithium monotherapy in adults in the acute phase of bipolar
acceptability of drug treatments for bipolar depression: a multiple-treatments meta- depression (EMBOLDEN I). J. Clin. Psychiatry 71, 150–162.
analysis. Acta Psychiatr. Scand. 130, 452–469. Zornberg, G.L., Pope Jr., H.G., 1993. Treatment of depression in bipolar disorder: new
directions for research. J. Clin. Psychopharmacol. 13, 397–408.

280

You might also like