Content uploaded by Florian Naudet
Author content
All content in this area was uploaded by Florian Naudet on Sep 22, 2017
Content may be subject to copyright.
Pharmacologically controlled drinking in the treatment
of alcohol dependence or alcohol use disorders:
a systematic review with direct and network meta-
analyses on nalmefene, naltrexone, acamprosate,
baclofen and topiramate
Clément Palpacuer
1
, Renan Duprez
2
, Alexandre Huneau
1
, Clara Locher
1,3,4
,RémyBoussageon
5
,
Bruno Laviolle
1,3,4
& Florian Naudet
1,6
Inserm, CIC 1414 Clinical Investigation Centre, Rennes, France,
1
Fondation Saint Jean de Dieu, Centre Hospitalier Dinan/St Brieuc, Dinan, France,
2
Rennes University
Hospital, Department of Biological and Clinical Pharmacology and Pharmacovigilance, Rennes, France,
3
Rennes 1 University, Laboratory of Experimental and Clinical
Pharmacology, Rennes, France,
4
Département de Médecine Générale, Faculté de Médecine et de Pharmacie, Université de Poitiers, Poitiers, France
5
and Meta-
Research Innovation Center at Stanford (METRICS), Stanford University, Palo Alto, CA, USA
6
ABSTRACT
Background and Aims Pharmacologically controlled drinking in the treatment of alcohol dependence or alcohol use
disorders (AUDs) is an emerging concept. Our objective was to explore the comparative effectiveness of drugs used in this
indication. Design Systematic review with direct and network meta-analysis of double-blind randomized controlled tri-
als (RCTs) assessing the efficacy of nalmefene, naltrexone, acamprosate, baclofen or topiramate in non-abstinent adults di-
agnosed with alcohol dependence or AUDs. Two independent reviewers selected published and unpublished studies on
Medline, the Cochrane Library, Embase, ClinicalTrials.gov, contacted pharmaceutical companies, the European Medicines
Agency and the Food and Drug Administration, and extracted data. Setting Thirty-two RCTs. Participants A total of
6036 patients. Measurements The primary outcome was total alcohol consumption (TAC). Other consumption out-
comes and health outcomes were considered as secondary outcomes. Findings No study provided direct comparisons
between drugs. A risk of incomplete outcome data was identified in 26 studies (81%) and risk of selective outcome
reporting in 17 (53%). Nalmefene [standardized mean difference (SMD) = 0.19, 95% confidence interval (CI) = 0.29,
–0.10; I
2
= 0%], baclofen (SMD = 1.00, 95% CI = 1.80, 0.19; one study) and topiramate (SMD = 0.77, 95%
CI = 1.12, –0.42; I
2
= 0%) showed superiority over placebo on TAC. No efficacy was observed for naltrexone or
acamprosate. Similar results were observed for other consumption outcomes, except for baclofen (the favourable outcome
on TAC was not reproduced). The number of withdrawals for safety reasons increased under nalmefene and naltrexone.
No treatment demonstrated any harm reduction (no study was powered to explore health outcomes). Indirect compari-
sons suggested that topiramate was superior to nalmefene, naltrexone and acamprosate on consumption outcomes, but
its safety profile is known to be poor. Conclusions There is currently no high-grade evidence for pharmacological treat-
ment to control drinking using nalmefene, naltrexone, acamprosate, baclofen or topiramate in patients with alcohol de-
pendence or alcohol use disorder. Some treatments show low to medium efficacy in reducing drinking across a range of
studies with a high risk of bias. None demonstrates any benefit on health outcomes.
Keywords Alcohol dependence, alcohol use disorders, controlled drinking, meta-analysis, pharmacotherapy,
randomized controlled trial.
Correspondence to:Clément Palpacuer, Centre d’Investigation Clinique INSERM 1414, Hôpitalde Pontchaillou, 2 rue Henri leGuilloux, 35033 Rennes cedex9,
France. E-mail: clement.palpacuer@gmail.com
Submitted 18 March 2017; initial review completed 16 May 2017; final version accepted 28 July 2017
INTRODUCTION
Pharmacologically controlled drinking in the treatment of
patients suffering from alcohol dependence or alcohol use
disorders (AUDs) is an emerging concept [1,2]. In Europe,
nalmefene, an opioid antagonist that acts on the urge to
consume alcohol, was the first treatment approved by the
European Medicines Agency (EMA) for this indication [3].
© 2017 Society for the Study of Addiction Addiction
REVIEW doi:10.1111/add.13974
Nevertheless, the evidence for its approval was
contested [4,5]. As no difference was observed on health
outcomes, which is the aim of a harm reduction strategy,
the approval was based on small differences versus placebo
on alcohol consumption outcomes, onlyobserved in a sub-
group defined a posteriori in studies with a possible attri-
tion bias (nalmefene tolerance was poor) [6–8]. In
addition, nalmefene has not been compared with other ac-
tive treatments, in particular naltrexone [9], another opi-
oid antagonist that has previous approval in maintaining
abstinence and that has already been used off-label in re-
ducing consumption, with some evidence suggesting a pos-
sible interest for this indication [10].
Other drugs acting on the gamma-aminobutyric acid
(GABA)ergic system have also been studied for this indica-
tion [11]. All have already been used, more or less officially,
to help reduce alcohol consumption: (1) baclofen has been
used in France under a ‘temporary recommendation for
use’; (2) acamprosate is the calcium salt of N-acetyl-
homotaurine and has had approval for post-withdrawal
maintenance of alcohol abstinence and (3) topiramate is
an anti-epileptic drug that is considered as an emerging
alternative.
While widely diffused guidelines [12,13] and major re-
views [14] now promote this approach as part of a harm
reduction strategy, there is a need to assess the evidence
that supports the use of these drugs.
The objective of this study was to test the efficacy of
treatments (1) from head-to-head comparisons, and (2)
from indirect comparisons using a network meta-
analysis.
METHODS
Design
A standard meta-analysis protocol was developed and reg-
istered before the beginning of the study in the PROSPERO
database (systematic review registration—PROSPERO
2015: CRD42015019841). Double-blind randomized con-
trolled trials (RCTs) assessing the efficacy of nalmefene, nal-
trexone, acamprosate, baclofen or topiramate against each
other or against placebo were considered for inclusion in
the meta-analysis.
Eligibility criteria
We reviewed studies involving adults (aged 18 years and
over) with a diagnosis of alcohol use disorders (AUDs)
and/or alcohol dependence. We focused on non-abstinent
patients. As detoxification is generally obtained after 5–7
days, we included patients with fewer than 5 days’absti-
nence before the beginning of the study [15]. Therefore,
studies were not included when longer abstinence (or de-
toxification) was an explicit inclusion criterion. In addition,
studies with patients presenting systematic physicalor psy-
chological comorbidity were not included.
Studies were eligible if they focused on the comparison
of oral formulations of (1) nalmefene, (2) naltrexone, (3)
acamprosate, (4) baclofen or (5) topiramate with each
other or with a placebo. Only monotherapies were consid-
ered. In case of multiple-dose studies, only the dose closest
to the recommended dose was taken into account (see
Supporting information, Appendix S1, p. 2). We reviewed
RCTs systematically without any limitation in terms of du-
ration. Only study reports in English, French, German and
Spanish were considered.
Search strategy and study selection process
Eligible studies were identified from PubMed/Medline, the
Cochrane library and Embase, including congress
abstracts. The searches were conducted in April 2015.
An update of the search was performed for the Pubmed
database in June 2016. The same algorithm was used for
all electronic databases: ‘Nalmefene OR Baclofen OR
Acamprosate ORTopiramate OR Naltrexone ANDAlcohol’
with the filter ‘Clinical Trial’. Two reviewers (A.H. and R.D.)
reviewed independently the titles and abstracts of all
citations identified by the literature search. The full text
of relevant studies was examined. All disagreements were
resolved by consensus or by consultation with another
reviewer (F.N.). To avoid omission of any studies, all the
references identified were compared to (1) references in-
cluded in a previous comprehensive meta-analysis on
pharmacotherapy of alcohol use disorders [16] that were
shared by the author and (2) references included in a
previous indirect comparison of nalmefene and naltrex-
one [17].
Unpublished studies were also searched for by consult-
ing the ClinicalTrials.gov website and key organizations
such as the Food and Drug Administration (FDA) and
EMA. Pharmaceutical companies (Novartis, Janssen Cilag,
Merck Sereno, Bristol Myers Squibb and Lundbeck) were
also contacted to obtain information concerning their
studies. When needed, the authors of the abstracts were
contacted for further information and were asked for refer-
ences of the studies. If no response was obtained to a first
request, they were contacted a second time. We also
contacted authors of recent meta-analyses on the pharma-
cological management of AUDs and asked them if they
would agree to share their data.
Assessment of methodological quality
Two reviewers (C.P. and R.D.) assessed independently the
risks of bias in each study included in the meta-analysis
using the Cochrane Collaboration tool for assessing risk of
bias [18].
2Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
Data collection
Two reviewers (C.P. and R.D.) collected data from the stud-
ies included using a data extraction sheet, based on the
Cochrane Handbook for Systematic Reviews of Interventions
guidelines [19]. Disagreements were first resolved by con-
sensus and then by consulting a third reviewer for arbitra-
tion (F.N.). Studies appearing to duplicate authors,
treatment comparisons, sample sizes and outcomes were
checked one against the other to avoid any duplicates
and to avoid integrating data from several reports on the
same study. For each study included, information was ex-
tracted on (1) characteristics of the study [year, country,
co-treatments (psychological interventions and their type),
number of arms, funding], (2) characteristics of trial partic-
ipants (age, gender, number of patients included in the
analysis, population analysed), (3) type of intervention
(treatments and comparators, duration) and (4) outcome
measures as detailed below.
Outcome measures
Our primary outcome was total alcohol consumption
(TAC). The choice of TAC was based on our previous
meta-analysis on nalmefene [8], and on the fact that the
rationale for using pharmacologically controlled drinking
stems from the idea that if individuals reduce their total al-
cohol consumption, they reduce their levels of risk accord-
ingly. Other consumption outcomes were considered as
secondary outcomes, namely: (1) the number of heavy
drinking days (HDD), defined as a day with a high con-
sumption level, (2) the number of non-drinking days, (3)
the number of drinking days and (4) the number of drinks
per drinking day. We also extracted safety outcome data:
(1) adverse events, (2) serious adverse events, (3) with-
drawals from the study, (4) withdrawals for safety reasons
and (5) mortality.
Data analysis
For consumption outcomes, the treatment effects were
expressed as standardized mean differences (SMD, Hedge’s
g) between arms, because these outcomes were likely to be
assessed in different ways in the different studies. The
efficacy index used for safety outcomes was the odds ratio
(OR). All outcomes were presented with their 95%
confidence interval (CI). The analytical plan was divided
into two main steps: we first carried out conventional
meta-analyses focusing on direct, pair-wise comparisons.
We then performed a network meta-analysis of all studies
included to estimate the effects of each active treatment
as compared to all the others. All analyses were performed
using R statistical software [20] and the meta [21],
netmeta [22] and metafor [23] libraries. Results were pre-
sented according to PRISMA (Preferred Reporting Items
for Systematic Reviews and Meta-Analyses) format [24]
and its extension for network meta-analyses [25].
Direct comparisons
An estimate of the overall effect (summary measure) was
calculated using a random-effects model (DerSimonian &
Laird method). For mortality, serious adverse events and
withdrawals for safety reasons, which are rare outcomes,
we used Peto’s OR method [26], as recommended by the
Cochrane Collaboration for handling sparse data [19].
Heterogeneity was evaluated using a visual inspection
of the forest plot, the Qstatistic and the I-squared (I
2
)
index.
Indirect comparisons
Network meta-analyses enabling indirect comparisons
were performed using the graph-theoretical method [27],
a frequentist approach. We used a design-basedbreakdown
of Cochrane Qand the I
2
index for assessing the heteroge-
neity in the whole network, the homogeneity within de-
signs and, if applicable, the homogeneity/consistency
between designs. A random-effects model was used for all
indirect comparisons.
In addition to the estimation of summary measures and
their 95% CI, we estimated the P-scores enablinga ranking
of all treatments which mostly follow that of the point esti-
mates, but take precision into account. These P-scores
measure the mean extent of certainty that a treatment is
better than the competing treatments [28].
Risk of bias across studies
Publication bias was investigated graphically using funnel
plots for each direct meta-analysis of our principal outcome
when there were at least four studies. Funnel plot asymme-
try was tested using the rank correlation test when there
were at least 10 studies [29].
Minor changes to the initial protocol
While we had planned initially to focus on consump-
tion outcomes, we added safety outcomes before data
extraction because (1) these outcomes are necessary
to estimate the risk/benefit ratio of the drugs consid-
ered and (2) a reduction of mortality and/or in serious
adverse events are expected benefits of a harm reduc-
tion approach.
We assessed the quality of evidence contributing to
each network estimate for total alcohol consumption
(TAC) using the Grading of Recommendations Assessment,
Development and Evaluation (GRADE) framework [30]. As
it appeared that data on HDDs were more complete than
for TAC, we also evaluated the qualityof the evidence with
the GRADE method for this outcome.
Pharmacologically controlled drinking 3
© 2017 Society for the Study of Addiction Addiction
When substantial heterogeneity (I
2
>50%) was found
in direct comparisons, meta-regressions were performed, a
posteriori, to explore whether the year of study, the defini-
tion of non-abstinence (no eligibilitycriteria requiring a pe-
riod of abstinence before inclusion versus presence of a
criterion specifying four or fewer abstinent days before in-
clusion) or the sex ratio (studies including men only versus
other studies) were possible effect modifiers. When effect
modifiers were suspected, an analysis excluding studies
that might have introduced heterogeneity was performed
as a post-hoc sensitivity analysis.
Role of the funding source
The funder of the study had no role in study design, data
collection, data analysis, data interpretation or writing of
the report. The corresponding author had full access to
all the data in the study and had final responsibility for
the decision to submit for publication.
RESULTS
Aflow-chart detailing the study selection process is given
in Fig. 1. After adjusting for duplicates, the searches pro-
vided a total of 2680 citations. Of these, 158 studies were
assessed for eligibility. Thirty-two RCTs involving 6036 pa-
tients met our eligibility criteria and were included in the
analysis [31–60]. The EMA provided us with access to all
study reports concerning nalmefene (including two un-
published eligible studies), because the drug had received
approval in Europe. We did not have access to study reports
either from the EMA for other drugs or from the FDA for all
drugs. For one study [54], the main analysis was presented
as part of a systematic review [61].
Study characteristics
All 32 RCTs included in the meta-analysis were placebo-
controlled. They were published between 1994 and
Figure 1 Flow diagram
4Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
2015. Their main characteristics are summarized in
Table 1. The studies compared the effects of oral nalmefene
(n= 9), naltrexone (n= 14), acamprosate (n= 1), baclofen
(n=4)ortopimarate(n= 4) against placebo. No study pro-
vided direct comparisons between active treatments. Pla-
cebo was the most frequently studied intervention, with
2780 patients. Nalmefene and naltrexone were the most
widely studied drugs with, respectively, 1693 and 850 pa-
tients receiving one of the two medications in the trials.
The other drugs were studied less frequently with, respec-
tively, 258, 106 and 349 patients for acamprosate, baclo-
fen and topiramate. Generally, study durations were
longer for nalmefene studies [median: 24 weeks, interquar-
tile range (IQR) = 12–28] than for naltrexone (median:
12 weeks, IQR = 10–16), baclofen (median: 12 weeks,
IQR = 4–12) and topiramate (median:12 weeks, IQR = 12–
12). The study duration of the acamprosate trial was
24 weeks. There were no major differences across studies
in terms of subject age (except for one naltrexone study
where the adults included were younger than in the others
[49]) and sex ratio (except for three naltrexone studies con-
ducted in men only [38,47,48]). Except for one nalmefene
trial (CPH-101-0701) and one naltrexone trial [51], all pa-
tients received a psychological co-intervention (including
medical management programmes) during the study
periods.
Risk of bias within studies
The risk of bias assessment is reported graphically in Fig. 2.
Twenty-six studies (81%) were classified as having an un-
clear or a high risk of incomplete outcome data due to
the large number of withdrawals. Extractable data were
based on a complete-case population for 26 trials (81%).
Therefore, an attrition bias cannot be excluded. Seventeen
studies (53%) were considered to present an unclear or a
high risk of selective outcome reporting, as they did not
mention a protocol registration number that allowed us
to check whether all outcomes were reported. Because
we had access to study reports, we were more confident
in the exhaustiveness of the nalmefene studies in compari-
son with other drugs. All these studies lacked consensus on
the ways of measuring and/or reporting outcomes, as de-
tailed in the Supporting information, Appendix S1 (p. 4).
Therefore, the analyses of the different outcomes were
based on a different number of studies and participants.
Data analysis
Direct comparisons
For all outcomes, the results of direct analyses (number of
studies, efficacy indices, I
2
index) are summarized in
Table 2 and presented graphically with individual study re-
sults in the Supporting information, Appendix S1 (p. 8). In
general, substantial heterogeneity was shown for the nal-
trexone and topiramate studies. In particular, for naltrex-
one, while some studies were in favour of a superiority of
the drug, one study found a significant superiority of pla-
cebo [41].
Regarding the primary outcome (TAC), it was reported
in only seven (78%) nalmefene studies, five (36%) naltrex-
one studies, one (100%) acamprosate study, one (25%)
baclofen study and two (50%) topiramate studies.
Nalmefene, topiramate and baclofen showed a greater de-
crease in TAC than placebo. Data on HDD were more com-
plete. The same results were observed, except for baclofen,
which did not show any superiority over placebo. Com-
pared to placebo, topiramate increased significantly the
number of non-drinking days and reduced the number of
drinking days. Nalmefene reduced significantly the num-
ber of drinks per drinking days. For nalmefene, estimated
effect sizes were always small. Conversely, the effect size
for baclofen was large on TAC but inconsistent on other
criteria. Topiramate consistently showed medium to large
effect sizes on all consumption outcomes.
No difference was found for any drug on mortality and
serious adverse events. More adverse events and with-
drawals for safety reason were evidenced for naltrexone
and nalmefene, and more withdrawals for nalmefene.
Indirect comparisons
For all outcomes, evidence of differences between drugs
was derived from indirect comparisons, as the networks
were star-shaped. Heterogeneity ranged from minor to
substantial (range of I
2
=0–61.5, I
2
greater than 50%
found for three outcomes). The consistency hypothesis
could not be ascertained because there was no direct com-
parison with active treatments. For TAC (Table 3), based on
the evidence detailed for direct comparisons, topiramate
was superior to nalmefene, naltrexone and acamprosate.
Similarly, baclofen showed superiority over naltrexone
and acamprosate on TAC. There was no evidence for a dif-
ference between nalmefene and naltrexone. The superior-
ity of topiramate seemed to be consistent on all the other
consumption outcomes while the superiority of baclofen
was not. No significant difference was found across drugs
on the safety outcomes (see Supporting information, Ap-
pendix S1, p. 62), except for withdrawals for safety reasons,
as presented in Table 3. For all outcomes, detailed contrast
tables are provided in the Supporting information, Appen-
dix S1 (p. 62). Network geometry is presented in Fig. 3
for TAC and the withdrawals for safety reasons, and in
the Supporting information, Appendix S1 (p. 59) for the
other outcomes.
Results are summarized visually with the ranking
according to P-scores in Fig. 4. Overall, the higher the
P-scores for the consumption outcomes, the lower the
Pharmacologically controlled drinking 5
© 2017 Society for the Study of Addiction Addiction
Table 1 Main characteristics of studies included.
Study Country
Industry
funding
a
Treatment and dose Abstinence criterion
Primary outcome measures
of the study
b
Study
duration
(weeks)
Medical
condition
Number of
patients
randomized
Age (years)
mean ± SD
Sex
(women)
(%)
Anton et al. [31]
2004
North
America
Yes Nalmefene
20 mg/day
At least 3 days prior
to randomization
Heavy drinking days 12 AD NLM: 66 NLM:
46.5 ± 10.9
NLM: 27.3
PLB: 68 PLB:
45.1 ± 11.1
PLB: 22.1
CPH-101-0399
2002
c
Europe Yes Nalmefene
40 mg/day
No Heavy drinking days 16 AUD NLM: 50 49.2 ± 8.6
d
NLM: 12
PLB: 50 PLB: 22
CPH-101-0701
2004
c
Europe Yes Nalmefene 10 to
40 mg/day (as
needed)
No more than 14
consecutive abstinent
days during the 12 weeks
preceding the first screening
visit
Heavy drinking days 28 AUD NLM: 85 NLM:
45.8 ± 8.6
NLM: 38.8
PLB: 82 PLB:
44.8 ± 10.4
PLB: 37.8
Gual et al. [32]
2013
Europe Yes Nalmefene
20 mg/day (as
needed)
No more than 14 consecutive
abstinentdaysinthe4weeks
preceding screening
Heavy drinking days
Total alcohol consumption
24 AD NLM: 358 NLM:
45.1 ± 10.7
NLM: 25.7
PLB: 360 PLB:
44.4 ± 10.7
PLB: 28.9
Karhuvaara
et al. [33] 2007
Europe Yes Nalmefene 10 to
40 mg/day (as
needed)
No more than 14 consecutive
abstinent days during the
12 weeks preceding inclusion
Heavy drinking days 28 AUD NLM: 242 NLM:
49.5 ± 9.1
NLM: 19
PLB: 161 PLB:
48.8 ± 8.4
PLB: 18
Mann et al. [34]
2013
Europe Yes Nalmefene
20 mg/day (as
needed)
No more than 14 consecutive
abstinentdaysinthe4weeks
preceding screening
Heavy drinking days
Total alcohol consumption
24 AD NLM: 306 NLM:
51 ± 10.1
NLM: 33.3
PLB: 298 PLB:
52.1 ± 9.1
PLB: 32.2
Mason et al. [35]
1994
North
America
Yes Nalmefene
20 mg twice daily
No Relapse
Abstinent days
Drinks per drinking day
Levelofalcoholcraving
Retention in treatment
12 AD NLM: 7 42 ± 9.4
d
28.6
d
PLB: 7
Mason et al. [36]
1999
North
America
Yes Nalmefene 10 or
40 mg twice daily
No Rate of relapse to heavy
drinking
Percentage of abstinent
days
12 AD NLM: 70 NLM:
41.9 ± 8.2
NLM: 31.4
PLB: 35 PLB:
41.7 ± 9.9
PLB: 37.1
(Continues)
6Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
Tabl e 1. (Continued)
Study Country
Industry
funding
a
Treatment and dose Abstinence criterion
Primary outcome measures
of the study
b
Study
duration
(weeks)
Medical
condition
Number of
patients
randomized
Age (years)
mean ± SD
Sex
(women)
(%)
Drinks per drinking day
Van de n Br in k
et al. [37] 2014
Europe Yes Nalmefene 20 mg/day
(as needed)
No more than 14 consecutive
abstinentdaysinthe4weeks
preceding screening
Heavy drinking days
Total alcohol consumption
52 AD NLM: 509 NLM:
44.3 ± 11.2
NLM: 22.8
PLB: 166 PLB:
44.3 ± 12
PLB: 23.5
Ahmadi et al.
[38] 2004
Asia No Naltrexone 50 mg/day At least 3 days and no more
than 30 days before study entry
Maintenance of abstinence
Relapse to drinking
36 AD NTX: 58 NTX:
42.8 ± 9.6
NTX: 0
PLB: 58 PLB:
43.2 ± 8.8
PLB: 0
Anton et al. [39]
2011
North
America
No Naltrexone 50 mg/day At least 4 days prior to
randomization
Time to relapse to drinking 16 AD NTX: 50 NTX:
44.4 ± 10.1
NTX: 20
PLB: 50 PLB:
46.6 ± 9.0
PLB: 18
Anton et al. [40]
2011
g
North
America
No Naltrexone 50 mg/day 4 days Drinks per drinking day 16 AD NTX: 21 NTX:
47.2 ± 11.3
NTX: 33.3
PLB: 23 PLB:
47.0 ± 9.6
PLB: 30.4
Davidson et al.
[41] 2004
North
America
NA Naltrexone 50 mg/day No Drinks per drinking day
Percentage of heavy drinking
days
Percentage of abstinent days
10 AUD NTX: 22 NTX:
46.5 ± 10.5
NA
PLB: 19 PLB:
50.8 ± 7.0
Heinala et al.
[42] 2001
Europe No Naltrexone 50 mg daily
during the first 12 weeks
then as needed
No Relapse to heavy drinking 32 AD NTX: 63 45.5 ± 7.8
d
28.9
d
PLB: 58
Killeen et al. [43]
2004
North
America
No Naltrexone 50 mg/day No Percentage of drinking days
Drinks per drinking day
Drinks per day
Heavy drinking days
Time to first heavy drinking
day
12 AUD NTX: 54 NTX:
37.9 ± 7.8
NTX: 41
PLB: 43 PLB:
35.5 ± 7.8
PLB: 44
Kranzler et al.
[44] 2000
North
America
No Naltrexone 50 mg/day Not explicitly mentioned 11 AD NTX: 61 NTX:
39.7 ± 8.4
NTX: 19.7
(Continues)
Pharmacologically controlled drinking 7
© 2017 Society for the Study of Addiction Addiction
Tabl e 1. (Continued)
Study Country
Industry
funding
a
Treatment and dose Abstinence criterion
Primary outcome measures
of the study
b
Study
duration
(weeks)
Medical
condition
Number of
patients
randomized
Age (years)
mean ± SD
Sex
(women)
(%)
At least 3 days and no more
than 28 days prior to
randomization
PLB: 63 PLB:
41.8 ± 8.1
PLB: 25.4
Kranzler et al.
[45] 2003
North
America
Yes Naltrexone 50 mg/day No Not explicitly mentioned 8 AUD NTX: 35 47.3 ± 8.6
d
42
d
PLB: 39
Kranzler et al.
[46] 2009
North
America
No Naltrexone 50 mg/day No Drinks per day 12 AUD NTX: 45 49.1 ± 9.6
d
41.7
d
PLB: 41
Morgenstern
et al. [47] 2012
North
America
No Naltrexone 100 mg/day No Sum of standard drinks
Heavy drinking days
12 AUD NTX: 102 NTX:
38.5 ± 10.8
NTX: 0
PLB: 98 PLB:
42.3 ± 11.1
PLB: 0
Morris et al. [48]
2001
Australia Yes Naltrexone 50 mg/day At least 3 days and no more
than 30 days before study
entry
Maintenance of abstinence
Relapse to drinking
12 AD NTX: 55 NTX:
47.0 ± 8.0
NTX: 0
PLB: 56 PLB:
48.0 ± 8.0
PLB: 0
O’Malley et al.
[49] 2015
North
America
No Naltrexone 25 mg/day
+25 mg/day as needed
No Percentage of heavy drinking
days
Percentage of abstinent days
8 AUD NTX: 70 NTX:
21.6 ± 2.1
NTX: 29
PLB: 70 PLB:
21.3 ± 2.1
PLB: 33
Oslin et al.[50]
2008
North
America
No Naltrexone 100 mg/day At least 3 days prior to the
start of the study medication
Number of days per time block
of abstinence and of heavy
drinking
Binary categories of any
drinking in the time block or
any heavy drinking
24 AD NTX: 126 NTX:
43.3 ± 10.4
NTX: 26.7
PLB: 122 PLB:
44.2 ± 11.3
PLB: 27.5
Tidey et al.[51]
2008
North
America
Yes Naltrexone 50 mg/day No Not explicitly mentioned 3 AUD NTX: 88
e
NTX:
29.9 ± 12.0
NTX: 42
PLB: 85
e
PLB:
28.1 ± 10.7
PLB: 40
Mason et al. [52]
2006
North
America
Yes Acamprosate 1 g twice
daily
No Percentage of alcohol-free days 24 AD ACP: 258 ACP:
44.9 ± 10.5
ACP: 30
PLB: 260 PLB: 36
(Continues)
8Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
Tabl e 1. (Continued)
Study Country
Industry
funding
a
Treatment and dose Abstinence criterion
Primary outcome measures
of the study
b
Study
duration
(weeks)
Medical
condition
Number of
patients
randomized
Age (years)
mean ± SD
Sex
(women)
(%)
PLB:
44.5 ± 10
Addolorato et al.
[53] 2002
Europe No Baclofen 30 mg/day in
threeseparatedoses
No Not explicitly mentioned 4 AD BAC: 20 BAC:
45.8 ± 10.6
NA
PLB: 19 PLB:
48.8 ± 10.4
Addolorato et al.
[54] 2011
Europe No Baclofen 30 mg/day in
threeseparatedoses
At least 3 days and no
more than 10 days prior
to randomization
Heavy drinking days
Abstinent days
Craving score
12 AD BAC: 14 BAC: 45.6
(30.0–57.0)
f
BAC: 14
PLB: 14 PLB: 43.1
(32.0–60.0)
f
PLB: 22
Garbutt et al.
[55] 2010
North
America
No Baclofen 30 mg/day 3 days prior to
randomization
Percentage of heavy drinking
days
12 AD BAC: 40 BAC:
47.5 ± 7.6
BAC: 45
PLB: 40 PLB:
50.3 ± 7.2
PLB: 45
Ponizov sky et al.
[56] 2015
Asia No Baclofen 50 mg/day in
two separate doses
Nomorethan6total
abstinent days per
monthonaverage
Percentage of heavy drinking
days
Percentage of abstinent days
12 AD BAC: 32 BAC:
42.6 ± 9.6
BAC: 25
PLB: 32 PLB:
44.7 ± 8.7
PLB: 25
Johnson et al.
[57] 2003
North
America
Yes Escalating doses
from 25 to 300 mg/day
in two separate doses
No Drinks per day
Drinks per drinking day
Percentage of heavy drinking
days
Percentage of abstinent days
Plasma γ-glutamyl transferase
12 AD TPM: 78 TPM:
41.5 ± 8.8
TPM: 30.6
PLB: 80 PLB:
42.1 ± 8.8
PLB: 26.6
Johnson et al.
[58] 2007
North
America
Yes Escalating doses
from 25 to 300 mg/day
in two separate doses
No Percentage of heavy drinking
days
14 AD TPM: 183 TPM:
46.7 ± 9.4
TPM: 25.7
PLB: 188 PLB:
47.8 ± 8.7
PLB: 28.1
Knapp et al.[59]
2015
North
America
No Escalating doses
from 25 to 300 mg/day
in two separate doses
No Drinks per day 12 AD TPM: 21 TPM:
46.8 ± 10.5
TPM: 42.9
PLB: 24 PLB:
46.8 ± 7.3
PLB: 45.8
(Continues)
Pharmacologically controlled drinking 9
© 2017 Society for the Study of Addiction Addiction
P-scores for the safety outcomes. We also noticed that the
safety profiles of nalmefene and naltrexone were similar.
Risk of publication bias and selective outcome reporting
We only investigated the existence of a publication bias for
our principal outcome. When applicable, funnel plots are
presented in the Supporting information, Appendix S1
(p. 8). In general, the small number of studies per outcome
prevented any meaningful use of this strategy. Therefore,
we consider that small study effects cannot be ruled out.
For example, concerning topiramate on the outcome
HDD, it is noticeable that the largest study against placebo
in terms of numbers of patients included found the smallest
effect size [standardized mean difference (SMD) = 0.2]. In
addition, asstated in the risk assessment paragraph, we es-
timated that selective outcome reporting was possible for
the studies included.
Investigation of heterogeneity
We performed meta-regression analyses for outcomes asso-
ciated with substantial heterogeneity (I
2
>50%). The def-
inition of abstinence and the year of study were not
identified as possible effect modifiers. We identified the sex
ratio as a possible effect modifier, as it was associated signif-
icantly with the number of drinks per drinking day for nal-
trexone studies. A sensitivity analysis was performed after
having removed studies including only men. Overall, re-
sults were not changed, and some heterogeneity was still
unexplained.
Quality of evidence contributing to each network estimate
According to the GRADE method, the quality of evidence
was very low for the overall ranking of treatments for
TAC and HDD (see Supporting information, Appendix S1,
p. 66).
DISCUSSION
Statement of principle findings
Regarding our primary outcome (TAC), nalmefene, baclo-
fen and topiramate showed superiority over placebo. Fur-
thermore, indirect comparisons showed superiority of
topiramate over nalmefene, naltrexone and acamprosate.
It is noteworthy that the reporting of TAC (our primary
end-point) was less consistent than the reporting of other
consumption outcomes such as HDD. However, similar
trends were observed for the other consumption outcomes,
except in the case of baclofen (the positive results for baclo-
fen on TAC were based on a single small study). For all
treatments except topiramate, effect sizes were small or in-
consistent. Furthermore, nalmefene and naltrexone were
associated with a significant increase in withdrawals from
Tabl e 1. (Continued)
Study Country
Industry
funding
a
Treatment and dose Abstinence criterion
Primary outcome measures
of the study
b
Study
duration
(weeks)
Medical
condition
Number of
patients
randomized
Age (years)
mean ± SD
Sex
(women)
(%)
Kranzler et al.
[60] 2014
North
America
No Escalating doses
from 25 to 200 mg/day
in two separate doses
No Heavy drinking days
Abstinent days
12 AUD TPM: 67 TPM:
49.3 ± 9
TPM: 32.8
PLB: 71 PLB:
52.8 ± 7.4
PLB: 42.3
a
We considered that there was an industry funding in case of partial or total funding by a pharmaceutical company, or when a pharmaceutical company supplied the drug;
b
the primary outcome measures in several studies were not clearly
stated, particularly in early trials;
c
unpublished studies provided by the EMA;
d
only overall data available;
e
these values correspond to the number of patients analysed, as the number of patients randomized was not extractable from the paper;
f
median (min–max);
g
study presented at a congress, results available on Clinicaltrials.gov (NCT00667875); AD = alcohol dependence; AUD = alcohol use disorder; NA = not available; NLM = nalmefene; NTX = naltrexone; ACP = acamprosate;
BAC = baclofen; TPM = topiramate; PLB = placebo; SD = standard deviation.
10 Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
the study and withdrawals for safety reasons, which raises
concerns about a plausible attrition bias.
Overall, there is no evidence for a significant reduction
in serious adverse events or in mortality. As for studies on
substances aiming to maintain abstinence [16], no study
was designed with sufficient power, and study durations
were inadequate to investigate these health outcomes. In
addition, any pharmacological approach that might benefit
patients by reducing their alcohol consumption might also
harm them because of possible safety issues. Therefore, if
we consider the reduction of consumption as a valid surro-
gate outcome, it is not certain that its validity for a given
drug can be transposed to another drug with a distinct
safety profile. For these reasons, we have advocated long-
term mega-trials exploring health outcomes (including
large, randomized, controlled cluster trials) in targeted pop-
ulations [7]. One could argue that adequately designed tri-
als on health outcomes would be unfeasible (too large and
too long) and suggest that this evidence could be obtained
from observational studies on large administrative data-
bases. This is even more challenging. In our opinion,
non-randomized observational evidence of this kind might
not be robust enough to resist the major conflicts of interest
inherent in the evaluation of pharmaceuticals. Weaknesses
of these study designs include indication bias (especially
when comparing drugs with absence of treatment) and
non-measured confounders. Without entering further into
this debate, there is a great need for consensus to move the
field forward. Otherwise, we will continue to produce low-
quality evidence that maintains the unacceptable status
quo, as illustrated by our meta-analysis.
Strengths and weaknesses of the study
Regarding the strengths and weaknesses of the study, one
major strength of network meta-analyses is that they pro-
vide a broad picture of the evidence for specific conditions,
and make itpossible understand how much evidence exists
for each treatment [62]. In this case, the network meta-
analysis enabled us to appraise critically the literature on
treatments used commonly to reduce alcohol consumption
in alcohol dependence and AUDs. It showed how the
emerging concept of pharmacologically controlled drinking
is based currently onresults from studies with a high risk of
bias (including attrition bias due to the high rates of with-
drawal). Pinpointing these shortcomings is a crucial result
Figure 2 Quality evaluation of included studies according to the Cochrane Collaboration tool for assessing risk of bias. SG = sequence generation;
AC = allocation concealment; B = blinding; IOD = incomplete outcome data; SOR = selective outcome reporting; OB = other bias. For each item,
the risk of bias was classified as ‘low risk’(LR), ‘high risk’(HR) or ‘unclear risk’(UR), with the last category indicating either lack of information or un-
certainty over the potential for bias [Colour figure can be viewed at wileyonlinelibrary.com]
Pharmacologically controlled drinking 11
© 2017 Society for the Study of Addiction Addiction
Table 2 Results of direct comparisons against placebo.
Nalmefene Naltrexone Acamprosate Baclofen Topiramate
n TE (95%CI) I
2
nTE(95%CI) I
2
nTE(95%CI) I
2
n TE (95%CI) I
2
n TE (95%CI) I
2
Consumption outcomes
Total alcohol
consumption
70.19 (0.29, 0.10) 0.0 5 0.11 (0.40, 0.18) 75.6 1 0.04 (0.21, 0.14) NA 1 1.00 (1.80, 0.19) NA 2 0.77 (1.12, 0.42) 0.0
Heavy
drinking days
70.22 (0.32, 0.12) 0.0 8 0.03 (0.21, 0.16) 51.0 0 NA NA 2 0.03 (0.33, 0.39) 0.0 4 0.59 (0.96, 0.22) 73.8
Non-drinking
days
80.09(0.01, 0.19) 0.0 3 0.28 (0.95, 0.40) 82.8 0 NA NA 2 0.08 (0.44, 0.27) 0.0 3 0.45 (0.15, 0.75) 62.9
Drinking days 0 NA NA 6 0.16 (0.35, 0.04) 46.4 1 0.05 (0.22, 0.12) NA 0 NA NA 1 0.75 (1.46, 0.05) NA
Drinks per
drinking day
30.26 (0.48, 0.05) 0.0 8 0.04 (0.31, 0.23) 66.1 0 NA NA 0 NA NA 2 0.40 (0.88, 0.09) 79.1
Safety outcomes
Adverse
events
62.00 (1.52, 2.65) 42.8 4 2.21 (1.36, 3.59) 0.0 0 NA NA 1 0.56 (0.12, 2.57) NA 1 1.92 (0.58, 6.32) NA
Serious
adverse
events
8
a
(7) 0.87 (0.50, 1.50) 33.9 4
a
(1) 1.41 (0.46, 4.31) NA 1
a
(0) NA NA 2
a
(0) NA NA 3
a
(1) 1.03 (0.25, 4.17) NA
Withdrawals
from the
study
91.38 (1.05, 1.80) 55.1 11 1.12 (0.74, 1.71) 49.0 1 1.75 (1.24, 2.48) NA 4 0.78 (0.25, 2.40) 66.2 4 1.31 (0.68, 2.51) 65.2
Withdrawals
for safety
reasons
93.32 (2.52, 4.37) 0.0 7
a
(6) 2.63 (1.27, 5.45) 0.0 1 0.84 (0.25, 2.76) NA 4
a
(1) 7.58 (0.47, 123) NA 3 2.02 (0.57, 7.13) 73.9
Mortality 9
a
(4) 0.41 (0.08, 2.11) 0.0 6
a
(0) NA NA 1
a
(0) NA NA 4
a
(0) NA NA 3
a
(1) 0.14 (0.00, 7.01) NA
Treatment effects (TE) are expressed as standardizes mean difference (SMD) for consumption outcomes and odds ratio (OR) for safety outcomes; n= number of studies reporting the outcome; NA = not applicable; CI = confidence interval.
a
Some
studies present no event in either treatment arm, and were therefore not included in the model. The number of studies used to estimate the treatment effect is indicated in parentheses; bold values are statistically significant.
12 Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
in an emerging field under pressure from both financial
conflicts of interest (e.g. the initial intensive, targeted mar-
keting of nalmefene) and non-financial conflicts of interest
(e.g. widespreadpublic support, in part via internet forums,
for the use of baclofen).
Furthermore, our study denotes a lack of a coherent
research agenda (placebo is the most widely studied inter-
vention and no study has compared two active treatments
directly) implementing consensual methods to assess and
measure outcomes [63].
We performed network meta-analyses which were
based completely on indirect comparisons. These analyses
should not replace head-to-head RCTs, but appeared neces-
sary as someof the treatments explored have been used off-
label for several years. Major shortcomings need to be
discussed to avoid any overinterpretation of the results.
First, indirect comparisons rely on a similarity assumption
(their validity depends on the distribution of treatment ef-
fect modifiers) [64], and systematic differences throughout
studies might bias the results. For instance, we observed
that study durations were generally longer for nalmefene
trials than for the other drugs. Therefore, these indirect
comparisons are supported in part by the assumption of a
constant effect of the treatment over time. It is noticeable
that the same hypothesis is made when a surrogate out-
come is measured in the short or medium term to deduce
the long-term effect of pharmacologically controlled drink-
ing on health outcomes. Secondly, we found considerable
heterogeneity (especially for naltrexone and topiramate),
which reduces the validity of the indirect comparisons.
We id entified the sex ratio as a possible explanation for this
heterogeneity in explanatory analyses. In post-hoc sensitiv-
ity analyses, the results appeared robust, but heterogeneity
was still present. Other sources of heterogeneity are
possible, such as the great variety of psychological
co-interventions used in the different trials, ranging from
simple medical management to standardized interventions
such as BRENDA. In addition, although we carefully
selected studies including non-abstinent patients, different
treatment goals across studies could explain some of this
heterogeneity. It might be hypothesized that patients en-
gaged in abstinence-orientated treatment may be likely to
Table 3 Results of indirect comparisons.
Total alcohol consumption, SMD (95% CI)
Placebo 0.18 (0.02, 0.35) 0.09 (0.11, 0.29) 0.04 (0.32, 0.39) 1.00 (0.13, 1.86) 0.79 (0.36; 1.21)
3.32 (2.48; 4.45) Nalmefene 0.09 (0.35, 0.17) 0.15 (0.54, 0.24) 0.81 (0.07, 1.69) 0.60 (0.15; 1.06)
2.64 (1.26; 5.51) 0.79 (0.36, 1.76) Naltrexone 0.06 (0.46, 0.35) 0.90 (0.01, 1.79) 0.69 (0.22; 1.16)
0.84 (0.25; 2.82) 0.25 (0.07, 0.88) 0.32 (0.08, 1.32) Acamprosate 0.96 (0.02, 1.89) 0.75 (0.20; 1.3)
7.58 (0.46; 125) 2.28 (0.14, 38.11) 2.88 (0.16, 51.96) 9.05 (0.43, 191) Baclofen 0.21 (1.17; 0.76)
2.77 (1.58; 4.86) 0.84 (0.44, 1.57) 1.05 (0.42, 2.66) 3.31 (0.87, 12.62) 0.37 (0.02, 6.35) Topiramate
Withdrawals for safety reasons, OR (95% CI)
Efficacyindices between treatments(in rows) and treatment (in columns) are presented with their 95% confidence interval(CI); for total alcohol consumption
(top right corner), a negative value is in favour of the treatment in the row; for the withdrawals for safety reasons (lower-left corner), a value below 1isin
favour of the treatment in the row; bold values are statistically significant. SMD = standardized mean difference; OR = odds ratio; CI = confidence interval.
Figure 3 Network geometry for the total alcohol consumption outcome (a) and the withdrawals for safety reasons outcome (b). Each node is
proportional to the number of patients for whom the outcome is available. Each edge is proportional to the number of studies for which the com-
parison is available [Colour figure can be viewed at wileyonlinelibrary.com]
Pharmacologically controlled drinking 13
© 2017 Society for the Study of Addiction Addiction
stop taking the drug when they begin drinking because
they see thisas treatment failure. The exact treatment goals
were reported inconsistently in the different studies, and it
was not possible to take this information into account.
A publication bias and a selective outcome reporting
bias might also account for some of the effects we observed.
Finally, because nalmefene has approval in Europe, the
EMA granted us access to unpublished studies and study
reports, and we included what we believe to be all available
relevant data from completed studies on nalmefene that
met our inclusion criteria. This was not the case for the
other treatments, and this could have biased our results to-
wards an underestimation of the comparative effectiveness
of nalmefene.
While comparative effectiveness is central to drug eval-
uation, this evidence was lacking at the time nalmefene
was approved. The main concerns relate to the lack of
meaningful comparisons with naltrexone, which is a very
similar compound. When the NICE (the National Institute
for Health and Care Excellence) evaluated nalmefene, the
pharmaceutical company Lundbeck (the compound patent
holder) stated that all the naltrexone studies had limita-
tions in the data reported, meaning that an indirect com-
parison could not be performed [65]. Indeed, our
independent appraisal found that the naltrexone studies in-
cluded were highly heterogeneous and had an unclear or
high risk of selective outcome reporting bias, compromis-
ing a definitive answer to the question ofcomparative effec-
tiveness between the two compounds. Direct comparisons
are therefore needed.
In disagreement with the above statement, reported to
NICE, a Lundbeck-sponsored indirect comparison of
nalmefene and naltrexone was published recently and
found a superiority of nalmefene [17]. Inclusion criteria
in this meta-analysis were close to ours, except that we
were more stringent concerning the duration of the absti-
nence period allowed before entering the trial. However,
this study compared subgroup analyses on nalmefene RCTs
with naltrexone RCTs as a whole. The validity of these a
posteriori-defined subgroup analyses has been criticized
Figure 4 P-scores estimations. P-scores are values comprised between 0 and 1 that measure the mean extent of certainty that a treatment is better
than the competing treatments. The closer to 1 the P-score, the better the treatment. TAC = Total alcohol consumption; HDD = Heavy drinking
days; NDD = Non drinking days; DPDD = Drinks per drinking day; SAE = Serious adverse events; AE = Adverse events [Colour figure can be viewed
at wileyonlinelibrary.com]
14 Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
[6,66], and in the context of an indirect comparison it chal-
lenges the similarity assumption dramatically. While we did
not find any significant difference between naltrexone and
placebo, indirect comparisons showed no evidence of a
superiority of nalmefene over naltrexone. In addition, both
drugs were linked to more adverse events than placebo.
Perspectives
As shown previously, nalmefene has small effects on sur-
rogate outcomes (with moderate confidence in the evi-
dence) and no demonstrated effect on health outcomes
[8]. No evidence was found of superiority against another
drug. Among the currently ongoing registered post-
marketing studies, none is to address these issues (i.e.
lack of power for health outcomes and/or lack of
comparisons) [7].
No efficacy was found here for either naltrexone or
acamprosate. Regarding baclofen, we found very poor evi-
dence for its use (very low confidence in evidence that was
very inconsistent). We found only small studies involving
low doses (ranging from 30 to 50 mg/day). These doses
were well below the doses that can be prescribed in day-
to-day practice. Indeed, the prescription of high doses (up
to 300 mg/day) has increased rapidly among alcohol spe-
cialists [67], while the marketing authorization for spastic-
ity forbids the use of more than 40 mg/day and warns
against associated alcohol use [68]. These practices were
based initially on case reports, case series or rudimentary
cohort studies [69]. A 1-year efficacy clinical trial in this
specific indication was completed in France (BACLOVILLE
study, NCT01604330) and was partially presented at a
congress. Only the results concerning mortality and the
primary outcome were released. They suggested efficacy
on consumption outcomes, with an absolute difference in
likelihood of success (defined as abstinence or a low level
of consumption) of 20.3% (95% CI = 7.3, 33.3) in favour
of baclofen. No effect was found on mortality with, respec-
tively, seven of 162 deaths in the baclofen group (five due to
alcohol) and three of 158 in the placebo group (one due to
alcohol). Although not significant (the study was powered
to evidence benefit on consumption outcomes, and not on
mortality) this result suggests that the effect observed on
consumption outcomes does not necessarily reflect efficacy
on health outcomes. As in other studies, attrition was high
(32% lost to follow-up). Data on other outcomes, including
the outcomes relevant for our meta-analysis, have still
been not released, despite a primary completion date in
September 2014 and a database lock in October 2015. It
is estimated that tens of thousands of patients with alcohol
use disorders are targeted [70] in France by a very specific
supervision procedure (‘temporary recommendation for
use’) [68]. Therefore, the complete results, enabling the
possible benefits to be weighed against the known safety
issues [71], especially in the case of deliberate acute baclo-
fen poisoning [72], are required urgently [73].
Interestingly, despite the major limitations we have
raised, our data suggests that topiramate could be consid-
ered as the most effective treatment, with medium to large
effect sizes on most consumption outcomes (but with very
low confidence in the evidence). This result, which should
be considered exploratory, is nonetheless interesting to an-
alyse. Indeed, it would be tempting to promote the use of
topiramate in alcohol dependence and AUDs, as itis associ-
ated with a significant decrease in alcohol consumption
and its safety profile does not appear to differ from that of
placebo for adverse events, serious adverse events or mor-
tality. Previous evidence does not support this idea, how-
ever, as the use of topiramate is known to result, for
example, in negative cognitive side effects such as deterio-
ration in verbal fluency, language comprehension, working
memory and visual block tapping [74]. The absence of dif-
ference between topiramate and the other drugs could be
explainedbyinsufficient study duration, insufficient power
and/or by poor quality in the reporting ofharmful effects or
events [75]. In our opinion, this example illustrates the fact
that surrogate outcomes should not be used alone. We feel
that the EMA should revise their standards and support the
implementation of adequately designed trials on health
outcomes. In any event, there is an urgent need to provide
policymakers with evidence as to which of these pharma-
cologically controlled drinking approaches can be trans-
lated effectively into a real harm reduction strategy.
To conclude, our results suggest that no treatment cur-
rently has high-grade evidence for pharmacologically con-
trolled drinking in the treatment of patients suffering from
alcohol dependence or alcohol use disorders. At best, some
showed low to medium efficacy in reducing drinking, but
across a range of studies with a high risk of bias. Although
based on all available data in the public domain, this meta-
analysis found no evidence of any benefit of the use of
drugs aiming for a controlled drinking strategy on health
outcomes. We invite researchers and stakeholders to set
up a coherent agenda to demonstrate that pharmacologi-
cally controlled drinking can be translated into genuine
harm reduction for patients. From the clinical perspective,
while this new approach is often presented as a ‘paradigm
shift’in terms of therapeutics, doctors and patients should
be informed that the critical examination of the pros and
cons of the evidence clearly questions the current guide-
lines that promote drugs in this indication.
Declaration of interests
None. All authors completed the Unified Competing Inter-
est form at http://www.icmje.org/coi_disclosure.pdf (avail-
able on request from the corresponding author) and
declare that C.P. was a trainee in Servier (pharmacokinetics
Pharmacologically controlled drinking 15
© 2017 Society for the Study of Addiction Addiction
department) for 6 months in 2013; R.D., A.H., C.L., R.B.
and B.L. have had no relationships with any company that
might have an interest in the submitted work in the previ-
ous 3 years; F.N. has had relationships (travel/accommoda-
tion expenses covered/reimbursed) with Servier, BMS,
Lundbeck and Janssen, who might have an interest in the
work submitted in the previous 3 years.
Acknowledgements
We would like to thank Dr Jonas, who shared the biblio-
graphic database of a previous systematic review, all the
authors who agreed to provide additional information
concerning their studies, Dr Rücker for the information
she provided regarding the use of the netmeta library,
Angela Swaine Verdier for revising the English and Karima
Hammas for her help in formatting the manuscript. This
study was funded by Rennes CHU (CORECT : Comité de la
Recherche Clinique et Translationnelle).
References
1. Klingemann J. Acceptance of reduced-risk drinking as a ther-
apeutic goal within the polish alcohol treatment system.
Alcohol Alcohol 2016; 51:436–41.
2. Rosenberg H., Davis A. K. Differences in the acceptability of
non-abstinence goals by type of drug among American
substance abuse clinicians. J Subst Abuse Treat 2014; 46:
214–8.
3. European Medicines Agency. Assessment report: Selincro—
international non-proprietory name: nalmefene. EMA/
78844/2013. Dec 13, 2012. Available at: http://www.ema.
europa.eu/docs/en_GB/document_library/EPAR_-_Public_as
sessment_report/human/002583/WC500140326.pdf (acce-
ssed 22 Aug2016) (Archived at http://www.webcitation.org/
6sw9M3XxC).
4. Braillon A. Nalmefene in alcohol misuse: junk evaluation by
the European medicines agency. BMJ 2014; 348: g2017.
5. Spence D. Bad medicine: nalmefene in alcohol misuse. BMJ
2014; 348: g1531.
6. FitzgeraldN., Angus K., Elders A., de Andrade M., Raistrick D.,
Heather N. et al. Weak evidence on nalmefene creates
dilemmas for clinicians and poses questions for regulators
and researchers. Addiction 2016; 111:1477–87.
7. Naudet F., Palpacuer C., Boussageon R., Laviolle B. Evaluation
in alcohol use disorders—insights from the nalmefene experi-
ence. BMC Med 2016; 14:119.
8. Palpacuer C., Laviolle B., Boussageon R., Reymann J. M.,
Bellissant E., Naudet F. Risks and benefits of nalmefene in
the treatment of adult alcohol dependence: a systematic liter-
ature review and meta-analysis of published and unpublished
double-blind randomized controlled trials. PLOS Med 2015;
12: e1001924.
9. Stevenson M., Pandor A., Stevens J. W., Rawdin A., Rice P.,
Thompson J. et al. Nalmefene for reducing alcohol consump-
tion in people with alcohol dependence: an evidence review
group perspective of a NICE single technology appraisal.
PharmacoEconomics 2015; 33:833–47.
10. Garbutt J. C. Efficacyand tolerability of naltrexone in the man-
agement of alcohol dependence. Curr Pharm Des 2010; 16:
2091–7.
11. Teng Goh E., Morgan M. Y. Pharmacotherapy for alcohol
dependence: the why, the what and the wherefore. Aliment
Pharmacol Ther 2017; 45:865–82.
12. Rolland B., Paille F., Gillet C., Rigaud A., Moirand R., Dano C.
et al. Pharmacotherapy for alcohol dependence: the 2015
recommendations of the French alcohol society, issued in
partnership with the European Federation of Addiction
Societies. CNS Neurosci Ther;2016:25–37.
13. Soyka M., Kranzler H. R., Hesselbrock V., Kasper S., Mutschler
J., Moller H. J. Guidelines for biological treatment of substance
use and related disorders, part 1: Alcoholism, first revision.
World J Biol Psychiatry 2017; 18:86–119.
14. Burton R., Henn C., Lavoie D., O’Connor R., Perkins C.,
Sweeney K. et al. A rapid evidence review of the effectiveness
and cost-effectiveness of alcohol control policies: an English
perspective. Lancet 2017; 389:1558–80.
15. Hall W., Zador D. The alcohol withdrawal syndrome. Lancet
1997; 349:1897–900.
16. Jonas D. E., Amick H. R., Feltner C., Bobashev G., Thomas K.,
Wines R. et al. Pharmacotherapy for adults with alcohol use
disorders in outpatient settings: a systematic review and
meta-analysis. JAMA 2014; 311: 1889–900.
17. Soyka M., Friede M., Schnitker J. Comparing Nalmefene and
naltrexone in alcohol dependence: are there any differences?
Results from an indirect meta-analysis. Pharmacopsychiatry
2016; 49:66–75.
18. Higgins J. P., Altman D. G., Gotzsche P. C., Juni P., Moher D.,
Oxman A. D. et al. The Cochrane Collaboration’s tool for
assessing risk of bias in randomised trials. BMJ 2011; 343:
d5928.
19. Higgins J. P. T., Green S., editors. Cochrane handbook for sys-
tematic reviews of interventions. Cochrane Collaboration.
Version 5.1.0. Available at: http://handbook-5-1.cochrane.
org/ (accessed 23 August 2017) (Archived at http://www.
webcitation.org/6sw9vRlcB).
20. R Development Core Team. R: a Language and Environm ent for
Statistical Computing. Vienna: R Foundation for Statistical
Computing; 2009.
21. Schwarzer G. Meta: general package for meta-analysis,
version 3.6–0. 27 May 2014. Available at: https://cran.r-pro-
ject.org/web/packages/meta/ (accessed 23 August 2017)
(Archived at http://www.webcitation.org/6swAAXySQ).
22. Rücker G., Schwarzer G., Krahn U., König J. netmeta: Net-
work meta-analysis using frequentist methods. R package
version 0.8-0, 2015. (Available at: https://cran.r-project.
org/web/packages/netmeta/ (accessed 23 August 2017) (Ar-
chived at http://www.webcitation.org/6swALYuOx).
23. Viechtbauer W. Conducting meta-analyses in R with the
metafor package. JStatSoftw2010; 36:1–48. Available at:
http://www.jstatsoft.org/v36/i03/ (accessed 26 August 2016)
(Archived at http://www.webcitation.org/6swAYx7cG).
24. Liberati A., Altman D. G., Tetzlaff J., Mulrow C., Gotzsche P. C.,
Ioannidis J. P. et al. The PRISMA statement for reporting
systematic reviews and meta-analyses of studies that evaluate
health care interventions: explanation and elaboration. JClin
Epidemiol 2009; 62:e1–34.
25. Hutton B., Salanti G., Caldwell D. M., Chaimani A., Schmid
C. H., Cameron C. et al. The PRISMA extension statement
for reporting of systematic reviews incorporating network
meta-analyses of health care interventions: checklist and
explanations. Ann Intern Med 2015; 162:777–84.
26. Yusuf S., Peto R., Lewis J., Collins R., Sleight P. Beta blockade
during and after myocardial infarction: an overview of the
randomized trials. Prog Cardiovasc Dis 1985; 27:335–71.
16 Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction
27. Rucker G. Network meta-analysis, electrical networks and
graph theory. Res Synth Meth 2012; 3:312–24.
28. Rucker G., Schwarzer G. Ranking treatments in frequentist
network meta-analysis works without resampling methods.
BMC Med Res Methodol 2015; 15:58.
29. Sterne J. A., Sutton A. J., Ioannidis J. P., Terrin N., Jones D. R.,
Lau J. et al. Recommendations for examiningand interpreting
funnel plot asymmetry in meta-analyses of randomised con-
trolled trials. BMJ 2011; 343:d4002.
30. Salanti G., Del Giovane C., Chaimani A., Caldwell D. M.,
Higgins J. P. Evaluating the quality of evidence from a network
meta-analysis. PLOS ONE 2014; 9: e99682.
31. Anton R. F., Pettinati H., Zweben A., Kranzler H. R., Johnson
B.,BohnM.J.et al. A multi-site dose ranging study of
nalmefene in the treatment of alcohol dependence. JClin
Psychopharmacol 2004; 24:421–8.
32. Gual A., He Y., Torup L., van den Brink W., Mann K. A
randomised, double-blind, placebo-controlled, efficacy
study of nalmefene, as-needed use, in patients with
alcohol dependence. Eur Neuropsychopharmacol 2013; 23:
1432–42.
33. Karhuvaara S., Simojoki K., Virta A., Rosberg M., Loyttyniemi
E., Nurminen T. et al. Targeted nalmefene with simple medical
management in the treatment of heavy drinkers: a random-
ized double-blind placebo-controlled multicenter study.
Alcohol Clin Exp Res 2007; 31:1179–87.
34. Mann K., Bladstrom A., Torup L., Gual A., van den Brink W.
Extending the treatment options in alcohol dependence: a
randomized controlled study of as-needed nalmefene. Biol
Psychiatry 2013; 73:706–13.
35. Mason B. J., Ritvo E. C., Morgan R. O., Salvato F. R., Goldberg
G., Welch B. et al. A double-blind, placebo-controlled pilot
study to evaluate the efficacy and safety of oral nalmefene
HCl for alcohol dependence. Alcohol Clin Exp Res 1994; 18:
1162–7.
36. Mason B. J., Salvato F. R., Williams L. D., Ritvo E. C., Cutler
R. B. A double-blind, placebo-controlled study of oral
nalmefene for alcohol dependence. Arch Gen Psychiatry
1999; 56:719–24.
37. van den Brink W., Sorensen P., Torup L., Mann K., Gual A.
Long-term efficacy, tolerability and safety of nalmefene as-
needed in patients with alcohol dependence: a 1-year,
randomised controlled study. J Psychopharmacol 2014; 28:
733–44.
38. Ahmadi J., Babaeebeigi M., Maany I., Porter J.,
Mohagheghzadeh M., Ahmadi N. et al. Naltrexone for
alcohol-dependent patients. Ir J Med Sci 2004; 173:34–37.
39. Anton R. F., Myrick H., Wright T. M., Latham P. K., Baros A.
M., Waid L. R. et al. Gabapentin combined with naltrexone
for the treatment of alcohol dependence. Am J Psychiatry
2011; 168:709–17.
40. Anton R., Baros A., Latham P., Randall P., Stewart S., Vergne
D. et al. Naltrexone plus aripiprazole compared to naltrexone
alone and placebo in the treatment of alcohol dependence—
a double blind pilot study. Neuropsychopharmacology 2011;
36: S234–5.
41. Davidson D., Saha C., Scifres S., Fyffe J., O’Connor S., Selzer C.
Naltrexone and brief counseling to reduce heavy drinking in
hazardous drinkers. Addict Behav 2004; 29:1253–8.
42. Heinala P., Alho H., Kiianmaa K., Lonnqvist J., Kuoppasalmi
K., Sinclair J. D. Targeted use of naltrexone without prior
detoxification in the treatment of alcohol dependence: a
factorial double-blind, placebo-controlled trial. JClin
Psychopharmacol 2001; 21:287–92.
43. Killeen T. K., Brady K. T., Gold P. B., Simpson K. N., Faldowski
R. A., Tyson C. et al. Effectiveness of naltrexone in a commu-
nity treatment program. Alcohol Clin Exp Res 2004; 28:
1710–7.
44. Kranzler H. R., Modesto-Lowe V., Kirk J. Naltrexone vs.
nefazodone for treatment of alcohol dependence. A
placebo-controlled trial. Neuropsychopharmacology 2000;
22:493–503.
45. Kranzler H.R., Armeli S., Tennen H., Blomqvist O., Oncken C.,
Petry N. et al. Targetednaltrexone forearly problem drinkers. J
Clin Psychopharmacol 2003; 23:294–304.
46. Kranzler H. R., Tennen H., Armeli S., Chan G., Covault J.,
Arias A. et al. Targeted naltrexone for problem drinkers. JClin
Psychopharmacol 2009; 29:350–7.
47. Morgenstern J., Kuerbis A. N., Chen A. C., Kahler C. W., Bux
D. A. Jr., Kranzler H. R. A randomized clinical trial of
naltrexone and behavioral therapy for problem drinking
men who have sex with men. J Consult Clin Psychol 2012;
80:863–75.
48. Morris P. L., Hopwood M., Whelan G., Gardiner J., Drummond
E. Naltrexone for alcohol dependence: a randomized con-
trolled trial. Addiction 2001; 96:1565–73.
49. O’Malley S. S., Corbin W. R., Leeman R. F., DeMartini K. S.,
Fucito L. M., Ikomi J. et al. Reduction of alcohol drinking in
young adults by naltrexone: a double-blind, placebo-
controlled, randomized clinical trial of efficacy and safety. J
Clin Psychiatry 2015; 76: e207–13.
50. Oslin D. W., Lynch K. G., Pettinati H. M., Kampman K. M.,
Gariti P., Gelfand L. et al. A placebo-controlled randomized
clinical trial of naltrexone in the context of different levels of
psychosocial intervention. Alcohol Clin Exp Res 2008; 32:
1299–308.
51. Tidey J. W., Monti P. M., Rohsenow D. J., Gwaltney C. J.,
Miranda R. Jr., McGeary J. E. et al. Moderators of naltrexone’s
effects on drinking, urge, and alcohol effects in non-
treatment-seeking heavy drinkers in the natural environ-
ment. Alcohol Clin Exp Res 2008; 32:58–66.
52. Mason B. J., Goodman A. M., Chabac S.,Lehert P. Effect of oral
acamprosate on abstinence in patients with alcohol depen-
dence in a double-blind, placebo-controlled trial: the role of
patient motivation. J Psychiatr Res 2006; 40:383–93.
53. Addolorato G., Caputo F., Capristo E., Domenicali M.,
Bernardi M., Janiri L. et al. Baclofen efficacy in reducing
alcohol craving and intake: a preliminary double-blind
randomized controlled study. Alcohol Alcohol 2002; 37:
504–8.
54. Addolorato G., Leggio L., Ferrulli A., Cardone S., Bedogni G.,
Caputo F. et al.Dose–response effect of baclofen in reducing
daily alcohol intake in alcohol dependence: secondary analy-
sis of a randomized, double-blind, placebo-controlled trial.
Alcohol Alcohol 2011; 46:312–7.
55. Garbutt J. C., Kampov-Polevoy A. B., Gallop R., Kalka-Juhl L.,
Flannery B. A. Efficacy and safety of baclofen for alcohol
dependence: a randomized, double-blind, placebo-controlled
trial. Alcohol Clin Exp Res 2010; 34: 1849–57.
56. Ponizovsky A. M., Rosca P., Aronovich E., Weizman A.,
Grinshpoon A. Baclofen as add-on to standard psychosocial
treatment for alcohol dependence: a randomized, double-
blind, placebo-controlled trial with 1 year follow-up. JSubst
Abuse Treat 2015; 52:24–30.
57. Johnson B. A., Ait-Daoud N., Bowden C. L., DiClemente C. C.,
Roache J. D., Lawson K. et al. Oral topiramate for treatment of
alcohol dependence: a randomised controlled trial. Lancet
2003; 361:1677–85.
Pharmacologically controlled drinking 17
© 2017 Society for the Study of Addiction Addiction
58. Johnson B. A., Rosenthal N., Capece J. A., Wiegand F., Mao
L., Beyers K. et al. Topiramate for treating alcohol depen-
dence: a randomized controlled trial. JAMA 2007; 298:
1641–51.
59. Knapp C. M., Ciraulo D. A., Sarid-Segal O., Richardson M. A.,
Devine E., Streeter C. C. et al. Zonisamide, topiramate, and le-
vetiracetam efficacy and neuropsychological effects in alcohol
use disorders. J Clin Psychophar macol 2015; 35:34–42.
60. Kranzler H. R., Covault J., Feinn R., Armeli S., Tennen H.,
Arias A. J. et al. Topiramate treatment for heavy drinkers:
moderation by a GRIK1 polymorphism. Am J Psychiatry
2014; 171:445–52.
61. Addolorato G., Leggio L. Safety and efficacy of baclofen in the
treatment of alcohol-dependent patients. Curr Pharm Des
2010; 16:2113–7.
62. Mills E. J., Thorlund K., Ioannidis J. P. Demystifying trial net-
works and network meta-analysis. BMJ 2013; 346: f2914.
63. Gargon E., Williamson P. R., Altman D. G.,Blazeby J. M., Tunis
S., Clarke M. The COMET initiative database: progress and
activities update (2015). Trials 2017; 18:54.
64. Jansen J. P., Naci H. Is network meta-analysisas valid as stan-
dard pairwise meta-analysis? It all depends on the distribution
of effect modifiers. BMC Med 2013; 11:159.
65. National Institute for Health and Care Excellence. Nalmefene
for reducing alcohol consumption in people with alcohol de-
pendence: evaluation report. 2014. Available at: https://
www.nice.org.uk/guidance/ta325?unlid=3629887952016
628192516 (accessed 24 August 2016) (Archived at http://
www.webcitation.org/6swAnhc4N).
66. Naudet F. Comparing nalmefeneand naltrexonein alcohol de-
pendence: is there a spin? Pharmacopsychiatry 2016; 49:
260–1.
67. Rolland B., Paille F., Fleury B., Cottencin O., Benyamina A.,
Aubin H. J. Off-label baclofen prescribing practices among
French alcohol specialists: results of a national online survey.
PLOS ONE 2014; 9: e98062.
68. Braillon A. Recommendations of French alcohol society and
European Federation of Addiction Societies. CNS Neurosci Ther
2016; 22:535–6.
69. Rolland B., Auffret M., FranchittoN. Safety reports on the off-
label use of baclofen for alcohol-dependence: recommenda-
tions to improve causality assessment. Expert Opin Drug Saf
2016; 15:747–51.
70. Degrassat-Theas A., Bocquet F., Sinegre M., Peigne J., Paubel
P. The ‘temporary recommendations for use’: a dual-purpose
regulatory framework for off-label drug use in France. Health
Pol ic y 2015; 119: 1399–405.
71. Harm Reduction International. What is harm reduction? A
position statement from harm reduction international.
2016. Available at: https://www.hri.global/what-is-
harm-reduction (accessed 23 August 2017) (Archived
at http://www.webcitation.org/6swBVmgy4).
72. Pommier P., Debaty G., Bartoli M., Viglino D., Carpentier F.,
Danel V. et al. Severity of deliberate acute baclofen poisoning:
a nonconcurrent cohort study. Basic Clin Pharmacol Toxicol
2014; 114:360–4.
73. Braillon A., Naudet F. Baclofen and alcohol use disorders:
from miracle to mirage. Eur Neuropsychopharmacol 2017;
27:691–2.
74. Fritz N., Glogau S., Hoffmann J., Rademacher M., Elger C. E.,
Helmstaedter C. Efficacy and cognitive side effects of tiagabine
and topiramate in patients with epilepsy. Epilepsy Behav 2005;
6:373–81.
75. Ioannidis J. P., Contopoulos-Ioannidis D. G. Reporting of
safety data from randomised trials. Lancet 1998; 352:
1752–3.
Supporting Information
Additional Supporting Information may be found online in
the supporting information tab for this article.
Appendix S1 Recommended doses for each treatment in
adults.
Appendix S2 List of outcomes reported for each study.
Appendix S3 Results of direct comparisons against placebo.
Appendix S4 Results of indirect comparisons: network
plots.
Appendix S5 Results of indirect comparisons: contrast
tables.
Appendix S6 Evaluation of the quality of evidence using
Grading of Recommendations Assessment, Development
and Evaluation (GRADE) framework.
Appendix S7 Preferred Reporting Items for Systematic
Reviews and Meta-Analyses (PRISMA) Checklist.
Appendix S8 Protocol.
18 Clément Palpacuer et al.
© 2017 Society for the Study of Addiction Addiction