N-Acetyl Cysteine for Depressive Symptoms
in Bipolar Disorder—A Double-Blind Randomized
Michael Berk, David L. Copolov, Olivia Dean, Kristy Lu, Sue Jeavons, Ian Schapkaitz,
Murray Anderson-Hunt, and Ashley I. Bush
Background: Treatment-resistant subthreshold depression is a major problem in bipolar disorder. Both depression and bipolar disorder
of glutathione, may improve the depressive component of bipolar disorder.
Methods: A randomized, double-blind, multicenter, placebo-controlled study of individuals (n ? 75) with bipolar disorder in the mainte-
nance phase treated with NAC (1 g twice daily) adjunctive to usual medication over 24 weeks, with a 4-week washout. The two primary
Bipolar Depression Rating Scale and 11 other ratings of clinical status, quality of life, and functioning.
[?13.16, ?2.95], p ? .002) and most secondary scales at end point. Benefit was evident by 8 weeks on the Global Assessment of
washout. There was no effect of NAC on time to a mood episode (log-rank test: p ? .968) and no significant between-group differences in
adverse events. Effect sizes at end point were medium to high for improvements in MADRS and 9 of the 12 secondary readouts.
Conclusions: NAC appears a safe and effective augmentation strategy for depressive symptoms in bipolar disorder.
Key Words: Bipolar disorder, clinical trial, depression, glutathione,
N-acetyl cysteine, neurochemistry
major psychiatric disorders (1), including schizophrenia (2,3),
bipolar disorder (4–7), and depression (8). Peripheral abnormal-
ities of the glutathione and other antioxidant metabolic pathways
associated with bipolar disorder and depression have been
reported to improve with somatic and drug treatments (6–10),
reducing oxidative stress (11–14).
Data are accumulating that mood stabilizers buffer oxidative
defenses. Valproate may protect neuronal cells from damage
related to oxidative stress (15). Chronic lithium and valproate
treatment has been found to increase expression of cellular
glutathione S-transferase M1 (16), which catalyses the glutathi-
one scavenging system, and is decreased in the postmortem
frontal cortex in bipolar disorder (17). Recently, polymorphisms
in the gene for glutamate cysteine ligase modifier (GCLM)
subunit (gclm), the key enzyme in glutathione synthesis, were
reported to suppress GCLM expression and to be linked to the
lutathione is the main antioxidant substrate in all tissue.
Increased oxidative stress with perturbed glutathione
metabolism is increasingly described as a feature of
risk for schizophrenia, where there is a decrease in brain
Glutathione production is rate-limited by its precursor, cys-
teine. Its acetylated derivative, N-acetyl cysteine (NAC) is more
efficiently bioavailable and is deacetylated by the liver (19,20).
N-acetyl cysteine is neuroprotective in a variety of neurodegen-
erative disease models (21–23). We recently found in a placebo-
controlled, randomized, double-blind clinical trial that NAC
treatment over 24 weeks significantly improved Clinical Global
Impression (CGI) and Positive and Negative Syndrome Scale
(PANSS) for schizophrenia scores in subjects with chronic schizo-
phrenia (24). N-acetyl cysteine increases blood glutathione levels
and improves auditory mismatch negativity in schizophrenia
within 8 weeks (25).
We hypothesized that add-on NAC treatment may also be of
clinical benefit in both the treatment and prevention of depres-
sive symptoms in bipolar disorder, which are characteristically
difficult to treat and represent a large unmet medical need (26).
The aim of this study was to study the efficacy and tolerability of
2 g daily of NAC compared with placebo in both treating and
preventing depressive symptoms in people with bipolar disorder
who were being maintained on treatment as usual.
Methods and Materials
Consented individuals were assigned using simple random-
ization (27) to treatment with NAC or placebo, in addition to
treatment as usual, in a double-blind fashion. The nature and
dose of the primary therapy was monitored. The person gener-
ating the randomization schedule was not involved in any aspect
of participant interview. The investigators, clinicians, and statis-
ticians were blind to treatment allocation until the data analysis
was completed. The study was registered with the Australian
Clinical Trials Registry (Protocol #12605000362695) prior to
enrollment. Participants were recruited through advertisements,
From The Mental Health Research Institute of Victoria (MB, DLC, OD, AIB),
Victoria, Australia; Department of Clinical and Biomedical Sciences (MB,
KL, IS, MA-H), The University of Melbourne, Geelong, Australia; Orygen
ton, Australia; Bendigo Health (SJ), Bendigo, Australia; Department of
Pathology (AIB), The University of Melbourne, Geelong, Australia; and
Department of Psychiatry (AIB), Massachusetts General Hospital,
Address reprint requests to Ashley I. Bush, M.D., Ph.D., The Mental Health
Research Institute of Victoria, 155 Oak Street, Parkville, Victoria 3052,
Australia; E-mail: firstname.lastname@example.org.
BIOL PSYCHIATRY 2008;xx:xxx
© 2008 Society of Biological Psychiatry
ARTICLE IN PRESS
their private psychiatrists, and database screening. All partici-
pants provided written informed consent. The trial setting was in
public and private outpatient clinics. The trial was approved by
each participating research and ethics committee (Barwon
Health, Mercy Health and Aged Care, and Bendigo Health) and
was conducted according to Good Clinical Practice guidelines.
This was, to our knowledge, the first clinical trial of NAC for
bipolar disorder. Our calculations were based on a sample size of
34 patients per group, which provides 80% power for a two-
tailed Student t test using ? ? .05 with an estimated effect size of
.7 after 24 weeks of treatment. Since there was no prior knowl-
edge of the effect of NAC, an effect size of .7 was used to estimate
a moderate clinical effect for this patient population. A total of 75
patients were planned to be enrolled and randomized in a 1:1
allocation to NAC or placebo. Assuming ?5% discontinuation
prior to the first postbaseline efficacy assessment, 70 patients
should be available to be included in the analysis of efficacy.
N-acetyl cysteine was acquired from Zambon, Italy. Purity
was 99.8% as determined by high-performance liquid chroma-
tography (HPLC). DFC Thompson, Sydney, Australia, performed
encapsulation of both the NAC and the identical placebo cap-
sules. Capsules were sealed in identical bottles, labeled as trial
medication, and both dispensed and returned by the pharmacy,
so the investigators were not exposed to their contents. Partici-
pants were seen separately and had no opportunity to compare
experiences. Pill counts for adherence were done by the phar-
macy, and an independent person confirmed the capsule audit.
All randomized participants received two NAC (500 mg)
capsules twice daily (2 g daily) or matching placebo capsules.
We selected a daily dose that was at the upper dosing range for
published clinical trials of oral NAC of 12 weeks to 12 months
duration and that reported evidence of tolerability and some
efficacy (28–31) and produced a positive result and excellent
safety profile in a phase 2 schizophrenia trial (24).
Inclusion and Exclusion Criteria
Individuals needed to meet DSM-IV criteria for bipolar disor-
der (I or II) with at least one documented episode of illness
(depressive, manic, or mixed) in the previous 6 months and be
on stable therapy for at least 1 month prior to randomization.
Participants were required to have the capacity to consent to the
study and comply with study procedures and utilize effective
contraception where indicated.
The trial was designed to be as naturalistic as possible, having
minimal inclusion and exclusion criteria, to maximize generali-
zation of results at the risk of reduced ability to find between-
group differences. Exclusion criteria included individuals with
a known or suspected clinically relevant systemic medical
disorder, including asthma, bronchospasm, or respiratory
insufficiency; recent gastrointestinal ulcers; and individuals
who were pregnant or lactating. Individuals taking greater
than 500 mg of NAC per day, 200 ?g of selenium per day, or
500 IU of Vitamin E per day were excluded, as were those with
a history of anaphylaxis with NAC or any component of the
preparation. Inability to comply with either the requirements
of informed consent or the treatment protocol was also an
Withdrawal from the study occurred if participants ceased
taking their trial medication for 7 consecutive days, stopped
effective contraception, or became pregnant. Dose changes to
existing medications or the addition or removal of an agent were
accepted and participants were allowed to continue with the
study. Participants were withdrawn from the study if they
revoked their consent or developed serious adverse events (AEs)
associated with the study drug, which could occur either at the
request of the patient or the discretion of the investigator.
Participants were assessed at baseline using a structured
clinical interview (Mini-International Neuropsychiatric Interview-
Plus [MINI-Plus]) and underwent a physical examination. Asses-
sors were all either clinical psychologists or medical practitioners
who were trained on the measures used. Clinical status was
assessed at baseline using the Montgomery Åsberg Depression
Rating Scale (MADRS), which together with time to a mood
episode were the co-primary outcome measures, examining
treatment and prevention of depression in bipolar disorder,
respectively. Secondary measures included the Bipolar Depres-
sion Rating Scale (BDRS) (32), Young Mania Rating Scale
(YMRS), CGI-Improvement and Severity scales for bipolar disor-
der, mania, and depression (CGI-I-BP, CGI-I-M, CGI-I-D, CGI-S-
BP, CGI-S-M, CGI-S-D), Global Assessment of Functioning Scale
(GAF), Social and Occupational Functioning Assessment Scale
(SOFAS), Streamlined Longitudinal Interview Clinical Evaluation
from the Longitudinal Interval Follow-up Evaluation (SLICE/
LIFE), Longitudinal Interval Follow-up Evaluation-Range of Im-
paired Functioning Tool (LIFE-RIFT), and Quality of Life Enjoy-
ment and Satisfaction Questionnaire (Q-LES-Q). These scales
were all repeated every two weeks for the first 4 weeks and one
time every four weeks thereafter for a total of 24 weeks or on the
day of study termination if the patient withdrew prior to the final
scheduled visit. A postdiscontinuation follow-up visit was con-
ducted 4 weeks (?2 weeks) after the trial completion, either at
trial end point (week 24) or premature discontinuation, to
determine any change in clinical status on washout.
Time to any intervention for mood symptoms was defined as
initiation of a new medication; initiation of emergency medical
contact, psychotherapy, hospitalization, or electroconvulsive
therapy (ECT); or discontinuation or dose adjustment of a current
agent, all in response to a clinician’s assessment of a new mood
episode. Cognitive tests including digit span (forward and back-
ward), word learning, Trail Making Test (Parts A and B), and
verbal fluency were done at baseline and end point. Assessment
of substance use was performed at baseline, week 12, and end
point, and these data are the subject of a separate report in
preparation. Adherence was monitored using pill counts of
returned clinical trial material.
Adverse events were recorded based on participant reports
throughout the trial using a checklist of 52 somatic items
(Supplement 1). This was augmented by spontaneous reports of
patients. Serious AEs were reported to all research and ethics
committees and also to the Therapeutic Goods Administration.
Randomization occurred at visit 1. Trial end point was defined
as the last postbaseline value obtained for a participant for a
given measure during the treatment phase. For participants who
completed the protocol, this corresponded to the visit 8 (week
24) assessment. All randomized participants who had at least one
postbaseline assessment were included in the analysis.
Analysis was performed by an external consultant statistician,
who was blind to treatment assignment, using SAS version 8.2 for
Windows (SAS Institute, Cary, North Carolina) on a clean and
locked database. All analyses were conducted in accordance
2 BIOL PSYCHIATRY 2008;xx:xxx
M. Berk et al.
ARTICLE IN PRESS
with the International Conference on Harmonization E9 statistical
principles (33) and are based on all randomized patients with at
least one postbaseline observation (intention-to-treat popula-
The efficacy analysis assessed average treatment group differ-
ences for each of the outcomes measured over the entire study
period and used a likelihood-based mixed-effects model re-
peated measures approach (MMRM). The MMRM model included
the fixed, categorical effects of treatment, investigator, visit, and
treatment-by-visit interaction, as well as the continuous, fixed
covariates of baseline score and baseline score-by-visit interac-
tion. The MMRM includes all available data at each time point
(34). For the MMRM outcome analysis, the residuals from the
model were examined. This included checking for normality,
independence, and homogeneity. For all models, there were no
violations of these assumptions. The covariance structure fitted
was compound symmetry based on lack of convergence of the
unstructured covariance matrix and Akaike’s information crite-
rion (AIC). In addition, Kaplan-Meier estimates and the log-rank
test were used to evaluate time to a mood episode between the
NAC and placebo groups. An assumption for the log-rank test is
that of proportional hazards. The assumption that the propor-
tional hazards stay constant over time was tested by inspecting a
graph of the logarithm of the estimated cumulative hazard
function. The lines for the two treatments were roughly parallel,
suggesting that the proportional hazards assumption was reason-
able in this case.
Results from the analysis of dichotomous data are presented
as proportions, with 95% confidence intervals (CI) and Fisher’s
exact p-value where appropriate. Effect sizes were calculated at
end point using MMRM. Applying Cohen’s guidelines (35), an
effect size of .2 to .4 is considered a small effect, .5 to .7 is
considered a medium effect, and ?.8 is considered a large effect.
For all other secondary measures (quality of life and function-
ing), the above analysis was utilized as described.
All tests of treatment effects were conducted using a two-
sided alpha level of .05, and 95% confidence intervals are
presented. The term significant in this report indicates statistical
significance (p ? .05). Adjustments for Type I errors were
performed using the Hochberg procedure (36).
Recruitment commenced in February 2004 and follow-up was
completed in December 2006. One hundred eighty-three people
were screened to take part in the trial. Of these, 108 people chose
not to consent or were not eligible and 75 were eligible and
enrolled, of which 37 were randomized into the placebo group
and 38 were randomized into the treatment (NAC) group.
Forty-eight participants completed the full 24-week trial period
and 58 (including individuals who terminated prematurely)
completed the postdiscontinuation visit (Supplements 2 and 3).
The most common reason for noncompletion in the trial was
withdrawal of consent by participants. There was no difference
between NAC and placebo groups in rate of dropout (simple
proportions, placebo 39%, NAC 35%, Fisher’s exact p ? .812,
log-rank p ? .7234, Wilcoxon-Breslow p ? .7373, Cox model p ?
.724, hazard ratio .87; .42–1.84).
In the NAC group, 31 (81.6%) of 38 participants had a
diagnosis of bipolar I disorder, and 30 (81.1%) of 37 participants
in the placebo group had a diagnosis of bipolar I disorder. In the
NAC group, 7 (18.4%) of 38 participants had a diagnosis of
bipolar II disorder, and 7 (18.9%) of 37 participants in the
placebo group had a diagnosis of bipolar II disorder (Table 1).
The mean MADRS and CGI-S-D scores at baseline (Table 2)
indicated that, on average, the cohort suffered with moderate
depressive symptoms, while the average baseline YMRS and
CGI-S-M scores indicated little hypomanic symptomatology. This
was confirmed by the clinical description of mood state at entry,
where depression was the most common (36%) pathological
category, whereas euphoric mania affected only 12% of the cohort
(Table 1). Therefore, the average profile of the cohort is consis-
tent with a higher prevalence of subthreshold depression. There
were no significant differences between the NAC and placebo
groups in these baseline clinical features or in comorbid psychi-
atric diagnoses. In the cohort, 78.7% reported regular alcohol
use, 8% reported regular cannabis use, and 45.3% were smokers,
with no between-group differences. The two groups similarly did
not differ on the baseline demographic and clinical measures
(Table 1). The mean age of the sample was 45.6 years, and there
were 45 female subjects and 30 male subjects. The average
duration of illness was 10.25 years with participants reporting a
Table 1. Patient Characteristics at Baseline
(n ? 38) n (%)
(n ? 37) n (%)
(n ? 75) n (%)
Mean Age (SD)
Public Treating Sector
Mood State at Entry
Number of Lifetime
6.5, 69 (1–70)
8.5, 69 (1–70)
22.0, 73 (1–74)
2.0, 5 (1–6)
4.5, 69 (1–70)
4.0, 69 (1–70)
7.0, 99 (1–100)
2.0, 5 (1–6)
5.0, 69 (1–70)
7.0, 69 (1–70)
20, 99 (1–100)
2.0, 5 (1–6)
There were no statistically significant differences on any comparison
between NAC and placebo groups for any of the items in this table using
Fisher’s exact test.
MINI-Plus, Mini-International Neuropsychiatric Interview-Plus; NAC, N-
M. Berk et al.
BIOL PSYCHIATRY 2008;xx:xxx 3
ARTICLE IN PRESS
mean of 2.35 admissions (median 1) over the course of their
illness. No subjects had prior exposure to NAC.
N-acetyl cysteine treatment was associated with a significant
reduction in symptoms at treatment completion (week 24) on the
MADRS primary readout (least squares [LS] mean difference [95%
CI]: ?8.05 [?13.16, ?2.95], p ? .002) (Table 2; Figure 1A).
Additionally, MADRS scores on average were significantly lower
over all visits for the NAC treatment group compared with the
placebo group (LS mean difference [95% CI]: ?3.08 [?5.99,
?.17], p ? .039). The benefit of NAC treatment appeared to be
time-dependent, with MADRS result becoming significant first at
the 20-week visit (LS mean difference [95% CI]: ?5.57 [?10.61,
?.53], p ? .031). Treatment effects varied over time for each
group for the following outcomes: MADRS, BDRS, YMRS, CGI-
S-BP, LIFE-RIFT, SLICE/LIFE, Q-LES-Q, GAF, and SOFAS (treat-
ment-by-time interaction p ? .10, Table 1 in Supplement 1). In
other words, the rates of change on these outcomes were
different between NAC and placebo treatments. Response, de-
fined as a 50% reduction in total MADRS score, at weeks 20 and
24 compared with baseline was observed in 46% and 51% of
participants in the NAC group compared with 21% and 18% in
the placebo group, respectively (p ? .036 and p ? .001,
The NAC treatment group had significantly better scores at
treatment completion on the majority of secondary symptomatic
measures used in the trial (Table 2). These included the BDRS (LS
mean difference [95% CI]: ?6.01 [?10.96, ?1.34], p ? .012)
(Table 2, Figure 1B) and the CGI-S-BP (LS mean difference [95%
CI]: ?.71 [?1.33, ?.09], p ? .026) (Table 2, Figure 1C). A
near-significant improvement in CGI-S-D at weeks 20 (LS mean
difference ?.68 [?1.37, .01], p ? .0525) and 24 (LS mean
difference ?.67 [?1.36, .02], p ? .0578) was also noted.
On scores of mania, there was a nonsignificant improvement
with NAC treatment evident on the YMRS at treatment comple-
tion (LS mean difference [95% CI]: ?1.56 [?3.31, ?.18], p ? .079)
(Table 2, Figure 1E). Note, however, that baseline YMRS scores
were low (NAC [95% CI]: 4.08 [2.72, 5.44], placebo [95% CI]: 4.03
[2.52, 5.53]). On week 12 of the trial alone, there was a greater
reduction on the YMRS for placebo treatment compared with NAC
(Figure 1E). The significance of this finding is unclear, as it was not
paralleled by other changes on the YMRS or CGI-I-M and CGI-S-M.
At no other point in the study was placebo treatment associated
with a significant improvement compared with NAC on any mea-
sure. The NAC-associated improvements on both CGI-S-D and
YMRS at treatment completion did reach significance using the
Cohen’s d effect size statistical approach (see below and Figure 2).
Quality of Life and Functional Outcomes
The symptomatic improvements associated with NAC treat-
ment were paralleled by improvements in quality of life and
functioning measures at treatment completion, including the
Q-LES-Q at week 24 (LS mean difference [95% CI]: 7.37 [2.09,
12.65], p ? .006; Figure 1F), as well as the LIFE-RIFT (LS mean
difference [95% CI]: ?2.95 [?4.79, ?1.12], p ? .002; Figure 1G)
and SLICE/LIFE (LS mean difference [95% CI]: ?3.97 [?6.96,
?.98], p ? .009; Figure 1H) (Table 2). There was a similar
improvement in the NAC-treated group on the GAF, which
separated from placebo at weeks 8 (LS mean difference [95% CI]:
6.05 [.80, 11.31], p ? .024); 20 (6.10 [.33, 11.87], p ? .038); and 24
(6.45 [.64, 12.26], p ? .030) (Figure 1I). On the SOFAS, the
NAC-treated group similarly improved compared with placebo at
weeks 8 (LS mean difference [95% CI]: 6.41 [1.18, 11.63], p ? .017)
and 24 (LS mean difference [95% CI]: 6.66 [.86, 12.47], p ? .0247
(Figure 1J) (Table 2). It should be noted that if correction of the
14 readouts for multiple testing were performed, several p-values
would not remain significant at the p ? .0036 level; however, the
improvement in MADRS primary readout would remain signifi-
cant, as would the LIFE-RIFT improvement.
Table 2. Outcome Measures at Baseline, Week 24, and Change at Posttreatment Discontinuation (Washout, Week 28)
Baseline (Mean ? SD)Week 24 (Mean ? SD)
Posttreatment to Week 28
(Mean ? SD)
Difference Between NAC and Placebo at
NAC PlaceboNACPlacebo NAC Placebo LS MeanLCL UCLp Value
that time point. Between treatment group LS means at week 24, CI, and p values are from MMRM.
BDRS, Bipolar Depression Rating Scale; CGI, Clinical Global Impression; CGI-I, CGI-Improvement; CGI-I-BP, CGI-Improvement scale for bipolar disorder;
CGI-I-D, CGI-Improvement scale for depression; CGI-I-M, CGI-Improvement scale for mania; CGI-S-BP, CGI-Severity scale for bipolar disorder; CGI-S-D,
CGI-Severity scale for depression; CGI-S-M, CGI-Severity scale for mania; CI, confidence intervals; GAF, Global Assessment of Functioning Scale; LCL, lower
ery Åsberg Depression Rating Scale; MMRM, Mini-International Neuropsychiatric Interview-Plus; NAC, N-acetyl cysteine; Q-LES-Q, Quality of Life Enjoyment
and Satisfaction Questionnaire; SLICE/LIFE, Streamlined Longitudinal Interview Clinical Evaluation from the Longitudinal Interval Follow-up Evaluation;
SOFAS, Social and Occupational Functioning Assessment Scale; UCL, upper confidence level; YMRS, Young Mania Rating Scale.
4 BIOL PSYCHIATRY 2008;xx:xxx
M. Berk et al.
ARTICLE IN PRESS
and at the postdiscontinuation visit (PDV). *p ? .05 versus placebo, **p ? .01 versus placebo, ***p ? .005 versus placebo. p values are from MMRM adjusted
others at 4 weeks. CGI, Clinical Global Impression; MMRM, mixed-effects model repeated measures; NAC, N-acetyl cysteine; PDV, postdiscontinuation visit.
M. Berk et al.
BIOL PSYCHIATRY 2008;xx:xxx 5
ARTICLE IN PRESS
Kaplan-Meier analysis did not reveal any significant differences
between the two groups for time to a mood episode (log-rank test:
p ? .968), the other primary readout. There were no differences in
new episodes of either depression or mania between the two
groups. There were no significant differences in the number of
changes to treatment as usual, with a total of 105 changes of any
kind in the NAC group and 116 medication changes in the placebo
group over the treatment period. There were no between-group
differences on any measures of cognition at end point.
An effect sizes (Cohen’s d) analysis of the benefits of NAC
treatment after 24 weeks on all rating scales was conducted. This
revealed improvements consistent with moderate to large effects
in the MADRS primary readout and in 9 out of 12 secondary
readouts (Figure 2).
The treatment benefit of NAC observed at week 24, the trial
end point, was not evident at the postdiscontinuation visit on any
of the scales included in the trial (Figure 1). This suggests that
improvements seen in the NAC group at end point had been
reversed by the discontinuation of NAC. No other postdiscon-
tinuation effects were observed.
Adverse events reported in more than 15% of the NAC group
included changed energy (21% NAC, 27% placebo), headaches
(18% NAC, 8% placebo), heartburn (16% NAC, 8% placebo), and
increased pain in joints (16% NAC, 8% placebo). No reported
event was significantly more common in the NAC group com-
pared with the placebo group (AEs reported in ?10% of partic-
ipants are described in Table 2 in Supplement 1). There were
seven serious AEs reported during the trial: three were in the
NAC group and four were in the placebo group. All were
hospitalizations, and all, except a victim of a motorcycle acci-
dent, were due to deteriorations in mental state.
There were no outcome differences between individuals on
lithium, lamotrigine, other mood stabilizers, antidepressants, or
antipsychotics on post hoc analyses, although sample sizes of the
subgroups were small. Prespecified analysis revealed no site or
investigator effects on outcome.
This first trial of NAC for bipolar disorder succeeded in one of
the two primary outcomes, as well as most of the secondary
readouts, and we believe that the results warrant further study of
this novel treatment approach. Although the trial failed to show
efficacy of NAC in the preventive readout (time to mood
episode), there was clear and marked efficacy of NAC in the
therapeutic readouts. The results indicate that adjunctive treat-
ment of bipolar disorder with 2 g/day oral NAC causes a
prominent reduction over a 6-month period in depressive symp-
toms on the MADRS (primary readout) and other secondary
readout scales, as well as improvement in secondary readout
measures of function and quality of life. There was at least a
trend for an effect of NAC on manic symptomatology, which may
have been difficult to exhibit due to the very low baseline
symptoms of mania in the cohort. It is noteworthy that the
clinical benefits of NAC on depression only emerged robustly
toward the end of the treatment period, at 20 weeks on the
MADRS, although significant benefit on functional measures, the
GAF and SOFAS, was evident at week 8. As there was no uniform
polarity requirement at baseline, the variance of many of the
symptomatic measures was large, which may have contributed to
delaying the detection of between-group differences.
While there was no overall effect on time to a mood episode
in this study (the second primary readout), with hindsight this
could be consistent with the onset of action only becoming
evident after several months of treatment. A benefit of NAC in
affecting a delay in the occurrence of a mood episode or its
prevention might be yet demonstrated in a differently designed
study. That significant differences emerged on most outcome
measures with a nonenriched naturalistic sample indicates that
the treatment benefit of NAC is robust and that further study is
warranted. Future studies using survival analysis could adopt an
enriched design, with a run-in period on active treatment,
compatible with the observed timeline of onset of action, before
such a survival analysis becomes meaningful. Future studies
could also examine more acutely ill individuals, where the time
course of action and pole specific efficacy would be further
clarified, as well as use longer follow-up periods. The naturalistic
design with minimal exclusion criteria and the multicenter,
outpatient-based nature of the current cohort increases the
potential generalizability of the findings.
Changes in background medications were allowed to increase
the generalizability of the study. The absence of any significant
between-group differences in this regard suggests that observed
effects were not a result of changes to background medications.
Nevertheless, there are high levels of comorbidity in bipolar
disorder and it is possible that effects on unidentified psychopa-
thology may have influenced these results and need to be
considered in future studies. Furthermore, dose-finding studies
are needed to determine the optimal dosing regimen of NAC for
this indication. Consideration of possible cumulative benefits
with other agents is warranted, as are trials of monotherapy.
The precise mechanism of the therapeutic benefit we ob-
served remains to be confirmed. It is hypothesized that NAC
Figure 2. Adjusted effect size (MMRM) at week 24 compared with baseline
for primary and secondary outcome measures. Data are mean effect size
(Cohen’s d statistic) ?95% confidence intervals. MMRM adjusted for base-
line score and investigator. The CGI scales are for severity. BDRS, Bipolar
Depression Rating Scale; CGI, Clinical Global Impression; CGIBP, Clinical
depression; GAF, Global Assessment of Functioning Scale; MADRS, Mont-
gomery Åsberg Depression Rating Scale; MMRM, mixed-effects model re-
peated measures; NAC, N-acetyl cysteine; Q-LES-Q, Quality of Life Enjoy-
ment and Satisfaction Questionnaire; RIFT, Longitudinal Interval Follow-up
Evaluation-Range of Impaired Functioning Tool; SLICE, Streamlined Longi-
tudinal Interview Clinical Evaluation from the Longitudinal Interval Fol-
low-up Evaluation; SOFAS, Social and Occupational Functioning Assess-
ment Scale; YMRS, Young Mania Rating Scale.
6 BIOL PSYCHIATRY 2008;xx:xxx
M. Berk et al.
ARTICLE IN PRESS
increases brain glutathione levels, restoring the oxidative imbal-
ances that are perturbed in bipolar disorder. However, without a
direct measure of glutathione levels in the brain, as might be
achieved by magnetic resonance spectroscopy (3), the status of
brain glutathione is uncertain. N-acetyl cysteine at this dose
increases blood glutathione and improves auditory mismatch
negativity in subjects with schizophrenia (25), and we recently
found that NAC at this dose as an adjunct treatment for schizo-
phrenia over a similar trial period significantly improves CGI and
PANSS scores in a double-blind placebo-controlled trial (24). The
benefit of NAC in both chronic schizophrenia and in major
affective disorder may be linked to disturbances of brain gluta-
thione metabolism found in both conditions (2–8). Buffering
brain glutathione levels may protect against neurotransmitter
defects such as glutamatergic hypofunction (37) or neurotoxicity
induced by hyperdopaminergia (38). Additionally, while we
cannot exclude a direct effect of NAC itself, this is unlikely since
the prodrug is rapidly converted into cysteine and glutathione by
the liver (39). Notably, while several antioxidants (e.g., vitamin C,
vitamin E) can preserve glutathione levels, they cannot replenish
the cysteine required for glutathione synthesis and replenish-
ment and are therefore unlikely to be clinically effective for
indications linked to glutathione deficiency.
Over time, the bulk of the burden of bipolar disorder is in the
depressive pole. The management of subthreshold depression in
the maintenance phase is a vexing clinical issue. Currently
available therapies, including antidepressants (26), are of limited
efficacy (40). The beneficial effects of NAC on depressive
symptoms are therefore of particular salience.
N-acetyl cysteine is relatively inexpensive, of established
safety (39), and available over the counter. The benefits we
observed indicate that disturbances in oxidative neurochemistry
may play a role in bipolar disorder and that augmentation of
glutathione using NAC supplementation in the maintenance
phase reduces depressive symptoms and improves functioning
and quality of life over a 6-month period.
This trial was supported by a grant from the Stanley Medical
Research Institute, as well as the Mental Health Research Institute
of Victoria. AIB is supported by the Australian Research Council
Federation Fellowship and the Commonwealth Scientific and
Industrial Research Organization Flagship Fellowship. The
Woods Family Foundation and the National Health and Medical
Research Council of Australia supported OD.
Alan Brnabic, M.A., performed payed statistical analysis. We
thank Warrick Brewer for his assistance with this project.
Dr. Berk has received grant or research support from Bristol
Myers Squibb, Eli Lilly, Glaxo SmithKline, Organon, Novartis,
Mayne Pharma, and Servier; is a consultant to Astra Zeneca,
Bristol Myers Squibb, Eli Lilly, Glaxo SmithKline, Janssen Cilag,
Lundbeck, and Pfizer; and has been a speaker for Eli Lilly, Glaxo
SmithKline, Janssen Cilag, Lundbeck, Organon, Pfizer, Sanofi
Synthelabo, Solvay, and Wyeth.
Dr. Bush is a shareholder and consultant for Prana Biotech-
nology Ltd and a shareholder of Cogstate Ltd. The other authors
reported no biomedical financial interests or potential conflicts
Supplementary material cited in this article is available
1. Ng F, Berk M, Dean O, Bush AI (2008): Oxidative stress in psychiatric
disorders: Evidence base and therapeutic implications [published on-
line ahead of print January 21]. Int J Neuropsychopharmacol.
phrenic disorders. J Neurochem 65:2652–2662.
3. Do KQ, Trabensinger MK, Lauer CJ, Dydak U, Hell D, Holsboer F, et al.
(2000): Schizophrenia: Glutathione deficit in cerebrospinal fluid and
prefrontal cortex in vivo. Eur J Neurosci 12:3721–3728.
4. Altschule MD, Goncz RM, Henneman DH, Holliday PD (1957): Carbohy-
drate metabolism in brain disease. VII. The effect of glutathione on
carbohydrate intermediary metabolism in schizophrenic and manic-
5. Andreazza AC, Cassini C, Rosa AR, Leite MC, de Almeida LM, Nardin P,
6. Abdalla DS, Monteiro HP, Oliveira JA, Bechara EJ (1986): Activities of
manic-depressive patients. Clin Chem 32:805–807.
7. Kuloglu M, Ustundag B, Atmaca M, Canatan H, Tezcan AE, Cinkilinc N
(2002): Lipid peroxidation and antioxidant enzyme levels in patients
8. Ozcan ME, Gulec M, Ozerol E, Polat R, Akyol O (2004): Antioxidant en-
9. Henneman DH, Altschule MD (1951): Immediate effects of shock thera-
pies, epinephrine and ACTH on blood glutathione level of psychotic
10. Altschule MD, Siegel EP, Henneman DH (1952): Blood glutathione level
11. Bilici M, Efe H, Koroglu MA, Uydu HA, Bekaroglu M, Deger O (2001):
perforatum extract demonstrates antioxidant properties against ele-
vated rat brain oxidative status induced by amnestic dose of scopol-
13. Eren I, Naziroglu M, Demirdas A, Celik O, Uguz AC, Altunbasak A, et al.
(2007): Venlafaxine modulates depression-induced oxidative stress in
brain and medulla of rat. Neurochem Res 32:495–505
14. Sanchez-Reus MI, Gomez del Rio MA, Iglesias I, Elorza M, Slowing K,
stress and increases gene expression of antioxidant enzymes on rote-
none-exposed rats. Neuropharmacology 52:606–616.
15. Wang JF, Azzam JE, Young LT (2003): Valproate inhibits oxidative dam-
age to lipid and protein in primary cultured rat cerebrocortical cells.
16. Wang JF, Shao L, Sun X, Young LT (2004): Glutathione S-transferase is a
novel target for mood stabilizing drugs in primary cultured neurons.
sion differences in bipolar disorder revealed by cDNA array analysis of
post-mortem frontal cortex. J Neurochem 79:826–834.
18. Tosic M, Ott J, Barral S, Bovet P, Deppen P, Gheorghita F, et al. (2006):
Schizophrenia and oxidative stress: Glutamate cysteine ligase modifier
ine on plasma cysteine and glutathione following paracetamol admin-
20. Holdiness MR (1991): Clinical pharmokinetics of N-acetylcysteine. Clin
inhibition of superoxide dismutase produces apoptotic death of spinal
22. Koppal T, Drake J, Butterfield DA (1999): In vivo modulation of rodent
23. Andreassen OA, Dedeoglu A, Klivenyi P, Beal MF, Bush AI (2000): N-
acetyl-L-cysteine improves survival and preserves motor performance
M. Berk et al.
BIOL PSYCHIATRY 2008;xx:xxx 7
ARTICLE IN PRESS
24. Berk M, Copolov D, Dean O, Lu K, Jeavons S, Schapkaitz I, et al. (2008):
N-acetyl cysteine as a glutathione precursor for schizophrenia: A dou-
ble-blind, randomized, placebo-controlled trial [published online
ahead of print April 22]. Biol Psychiatry.
25. Lavoie S, Murray MM, Deppen P, Knyazeva MG, Berk M, Boulat O, et al.
(2007): Glutathione precursor, N-acetyl-cysteine, improves mismatch
negativity in schizophrenia patients [published online ahead of print
November 14]. Neuropsychopharmacology.
26. Sachs GS, Nierenberg AA, Calabrese JR, Marangell LB, Wisniewski SR,
Gyulai L, et al. (2007): Effectiveness of adjunctive antidepressant treat-
27. Beller EM, Gebski V, Keech AC (2002): Randomisation in clinical trials.
28. Adair JC, Knoefel JE, Morgan N (2001): Controlled trial of N-acetylcyste-
ine for patients with probable Alzheimer’s disease. Neurology 57:1515–
29. Van Schooten FJ, Besaratinia A, De Flora S, D’Agostini F, Izzotti A, Cam-
oirano A, et al. (2002): Effects of oral administration of N-acetyl-L-cys-
30. Behr J, Degenkolb B, Krombach F, Vogelmeier C (2002): Intracellular
glutathione and bronchoalveolar cells in fibrosing alveolitis: Effects of
31. Demedts M, Behr J, Buhl R, Costabel U, Dekhuijzen R, Jansen HM, et al.
(2005): High-dose acetylcysteine in idiopathic pulmonary fibrosis.
32. Berk M, Malhi GS, Cahill C, Carman AC, Hadzi-Pavlovic D, Hawkins MT,
et al. (2007): The Bipolar Depression Rating Scale (BDRS): Its develop-
ment, validation and utility. Bipolar Disord 9:571–579.
33. (1998):International conference on harmonisation; guidance on statis-
tical principles for clinical trials; availability—FDA. Notice. Fed Regist
34. Mallinckrodt C, Watkin J, Molenberghs G, Carroll R (2004): Choice of
the primary analysis in longitudinal clinical trials. Pharm Stat 3:161–
35. Cohen J (1988): Statistical Power Analysis for the Behavioral Sciences,
2nd ed. Hillsdale, NJ: Lawrence Earlbaum Associates.
significance. Biometrika 75:800–802.
ment and hypofunction of NMDA receptors induced by glutathione
deficit: Relevance to schizophrenia. Neuroscience 137:807–819.
Systemic administration of N-acetylcysteine protects dopaminergic
39. Atkuri KR, Mantovani JJ, Herzenberg LA, Herzenberg LA (2006): N-ace-
tylcysteine—a safe antidote for cysteine/glutathione deficiency. Curr
40. Belmaker RH (2007): Treatment of bipolar depression. N Engl J Med
8 BIOL PSYCHIATRY 2008;xx:xxx
M. Berk et al.
ARTICLE IN PRESS