Key words reboxetine - schizophrenia - psychopathology - weight gain - meta-analysis
Introduction
Schizophrenia is a severe and chronic psychiatric disorder. Nearly 50% of schizophrenia
patients fail to respond to antipsychotic (AP) monotherapy, particularly in terms
of negative symptoms [1 ]
[2 ]
[3 ]. Augmentation strategies are commonly used to enhance AP efficacy and to reduce
adverse drug reactions (ADRs) induced by APs [2 ]
[4 ]
[5 ]
[6 ].
Most negative symptoms (e. g., anhedonia and social withdrawal) are likely to be intrinsic
to schizophrenia [7 ] and associated with neurocognitive deficits [8 ] including impaired attention, memory, and executive functions [9 ]. Certain augmentation strategies appear to be effective to improve the negative
symptoms of schizophrenia; for example, adjunctive antidepressants have a moderate
effect size for negative symptoms [7 ]
[10 ].
In the past two decades, weight gain induced by APs has attracted increased attention
[11 ]
[12 ]
[13 ]
[14 ]
[15 ]
[16 ]. Weight gain is associated with poor treatment adherence and quality of life, higher
rate of medical comorbidity particularly cardiovascular diseases, diabetes mellitus,
and osteoarthritis and mortality [12 ]
[13 ]
[17 ]
[18 ]
[19 ]
[20 ]
[21 ]. Reboxetine, a norepinephrine reuptake inhibitor, is an antidepressant and antianxiety
drug [18 ]
[22 ]. According to a recent meta-analysis, as an antidepressant, reboxetine is ineffective
and causes side effects [23 ]; however, in another network meta-analysis, reboxetine appeared to have antidepressant
effect in major depression [24 ]. Some trials have found that reboxetine is safe and efficacious in treating negative
symptoms of schizophrenia [25 ]
[26 ]
[27 ]
[28 ] and AP-induced weight gain in some [18 ]
[22 ]
[27 ] but not all [29 ]
[30 ] randomized controlled trials (RCTs) .
The efficacy of reboxetine in treating negative symptoms and/or reducing AP-induced
weight gain in schizophrenia have been examined in meta-analyses [10 ]
[12 ]
[31 ]
[32 ]
[33 ]
[34 ]. Common limitations of these meta-analyses include the small number of included
studies, resulting in insufficient power. For example, 3 meta-analyses [12 ]
[33 ]
[34 ], each with 2 RCTs (n=85) [18 ]
[22 ], found that reboxetine was superior to placebo in reducing AP-induced weight gain.
Another meta-analysis [31 ] included 9 RCTs, but one of them was an open-label study [35 ], the inclusion of which violated standard recommendations [36 ]. In addition, non-English databases were not searched for relevant RCTs [26 ].
The aim of this study was to obtain more robust evidence regarding the efficacy and
safety of adjunctive reboxetine added to APs. To this end, a comprehensive meta-analysis
was conducted involving all RCTs on reboxetine added to all types of APs in treating
schizophrenia including also recent RCTs published in Chinese, which were not included
in previous meta-analyses.
Methods
Types of studies
According to the PICOS acronym, the selection criteria were as follows: participants
(P), patients with schizophrenia diagnosed according to any criteria; intervention
(I), reboxetine plus APs; comparison (C), APs plus placebo; outcomes (O), efficacy
and safety of adjunctive reboxetine with meta-analyzable data; study design (S), randomized,
double-blind, placebo-controlled trials. A methodically sound RCT [37 ] was excluded because patients were selected from the combined sample of 2 other
RCTs [18 ]
[22 ].
Outcome measures
Clinical outcomes were recorded based on intent-to-treat (ITT) analysis, if provided.
The co-primary outcome measures were the change of negative symptoms assessed with
the Positive and Negative Syndrome Scale (PANSS) [38 ] or the Brief Psychiatric Rating Scale (BPRS) [39 ] or the total scores of the Scale for the Assessment of Negative Symptoms (SANS)
[40 ] and body weight (kg). Key secondary outcomes were the changes of total, positive,
and general psychopathology scores of the PANSS or BPRS or the total scores of the
Scale for the Assessment of Positive Symptoms (SAPS) [41 ], body mass index (BMI, kg/m2 ), cognitive functions, ADRs, and all-cause discontinuation rate.
Study selection
The PubMed, PsycINFO, EMBASE, the Cochrane Library, Chinese Journal Net, WanFang,
and the China Biology Medicine databases were independently and systematically searched
by 2 reviewers from their inception until November 3, 2016. The keywords for the search
were (reboxetine OR Edronax) AND (schizophrenic disorder OR disorder, schizophrenic
OR schizophrenic disorders OR schizophrenia OR dementia praecox). Reference lists
from review articles [10 ]
[12 ]
[31 ]
[32 ]
[33 ]
[34 ] were hand-searched for additional studies. First/corresponding authors were contacted
for missing information, whenever necessary.
Data extraction
Data were independently identified, checked, extracted, and analyzed by 2 reviewers.
Inconsistencies were resolved by consensus involving a third reviewer. If data from
the same study were reported in more than 1 RCT, only the RCT with complete data was
included in the analyses.
Statistical methods
According to the guidelines of the Preferred Reporting Items for Systematic Reviews
and Meta-Analyses (PRISMA) statement [42 ], data were summarized statistically using the Review Manager version 5.3 (http://www.cochrane.org)
if they were available and sufficiently similar. A random effects model was used in
all cases [43 ]. For the meta-analytic pooling of continuous outcomes, weighted or standardized
mean differences (WMDs or SMDs) with their 95% confidence intervals (CIs) are reported.
Summary statistics of dichotomous outcomes are presented as risk ratios (RRs)±95%
CIs. When RRs were significant, number-needed-to-harm (NNH) was calculated by dividing
1 by the risk difference. Whenever both change score and endpoint values of a continuous
outcome were available, change scores were preferred. Missing standard deviation (SD)
was replaced by the average SD of other RCTs following the suggestion of Leucht et
al. [36 ]. In cases of I2 >50% for co-primary outcomes, reasons were sought to explain the heterogeneity by
conducting a sensitivity analysis (i. e., removing 1 outlier [SMD≤− 1.3] study) [27 ]. In addition, the following 6 subgroup analyses were performed to identify the reasons
for the heterogeneity of significance: (1) Chinese versus non-Chinese studies; (2)
clozapine vs. other APs; (3) trial duration (weeks): ≥12 vs.<12 (mean splitting method
for trial duration); (4) age: ≥38.0 vs.<38.0 years (mean splitting method for age);
(5) male predominance (≥60%) vs. no sex predominance; (6) study quality: Jadad score
≥3 vs.<3. The above subgroup analyses were repeated after leaving out 1 outlying study
[27 ]. Funnel plots and Egger’s test [44 ] were used to judge publication bias. All analyses were 2 tailed, with alpha set
at 0.05.
Assessment of the studies
Cochrane risk of bias [45 ] was employed to assess the methodological quality of RCTs (Fig. 1S ). The Grading of Recommendations, Assessment, Development, and Evaluation (GRADE)
system [46 ]
[47 ] was performed to estimate the recommendation for outcome measures of adjunctive
reboxetine for schizophrenia. Following the methodology of other studies [31 ]
[48 ], the Jadad scale (range: 0–5) assessed the quality of included studies ([Table 1 ]) [49 ]. The Jadad total score of ≥3 indicated “high quality” [50 ].
Table 1 Characteristics of the studies included in the meta-analysis.
Study (country)
Number of patients
Blinding
Analyses
Trial duration (weeks)
Setting (%)
Diagnosis (%)
Diagnostic criteria
Illness severity (PANSS/BPRS)a /durationa
Agea : years (range)
Sexa : male (%)
Placebo group: dose (mg/d): mean (range)
Reboxetine group: dose (mg/d): mean (range)
Risk of Bias*
Jadad score
Hinkelmann et. al. 2013 (Germany)
T: 35 M: 16 A: 19
DB
ITT
4
NR
Sz (100)
DSM-IV
−32.9; − 12.6 years
40.4 (NR)
68.6
APs: Ø=NR (NR)
APs: Ø=NR (NR)
REB: Ø=NR (4–8)
2
3
Li et al. 2008 (China)
T: 118 M: 59 A: 59
DB
OC
12
Inpatients (100)
Sz (100)
CCMD-3
−75.2; − 27.2 years
50.4 (18–65)
100
CLZ: Ø=282.5 (NR)
CLZ: Ø=289.9 (NR)
REB: Ø=8b (4–8)
2
3
Poyurovsky et al. 2003 (Israel)
T: 26 M: 13 A:13
DB
OC
6
Inpatients (100)
Sz (100)
DSM-IV
-NR; −3.0 years
30.8 (19–49)
65.4
OLA: Ø=10 (FD)
OLA: Ø=10 (FD)
REB: Ø=4 (FD)
5
5
Poyurovsky et al. 2007 (Israel)
T: 59 M: 28 A:31
DB
ITT
6
Inpatients (100)
Sz (NR), SzD (NR)
DSM-IV
-NR; − 3.5 years
29.9 (19–48)
64.4
OLA: Ø=10 (FD)
OLA: Ø=10 (FD)
REB: Ø=4 (FD)
5
5
Schutz et al. 2001 (South Africa)
T: 30 M: 15 A:15
DB
NR
6
Inpatients (100)
Sz (100)
DSM-IV
−78.9; −8.1 years
32.5 (NR)
93.3
HAL: Ø=5 (FD)
HAL: Ø=5 (FD)
REB: Ø=8 (FD)
4
3
Shafti et al. 2015 (Iran)
T: 50 M: 25 A: 25
DB
ITT
12
Inpatients (100)
Sz (100)
DSM-IV-TR
-NR; − 8.9 years
40.4 (NR)
100
HAL: Ø=NR (5–20)
HAL: Ø=NR (5–20)
REB: Ø=4 (FD)
4
3
Sun et al. 2011 (China)
T: 135 M: 66 A: 69
DB
OC
12
In- and outpatients (NR)
Sz (100)
DSM-IV
−85.6; − 0.4 years
32.5 (18–45)
50.8
HAL: Ø=14 (10–30)
HAL: Ø=15 (10–30)
REB: Ø=6 (4–12)
2
2
Usall et al. 2014 (Spain)
T: 67 M: 33 A: 34
DB
ITT
24
NR
Sz (100)
DSM-IV
−73.5; -NR
42.1 (18–65)
73.1
OLA/RIS: Ø=NR (0.5–20)c
OLA/RIS: Ø=NR (0.5–40)c
REB: Ø=8b (4–8)
5
5
Zhao et al. 2013 (China)
T: 110 M: 55 A: 55
DB
OC
24
Inpatients (100)
Sz (100)
ICD−10
−73.6; − 21.4 years
43.1 (18–55)
100
CLZ: Ø=297.0 (NR)
CLZ: Ø=302.3 (NR)
REB: Ø=7.6 (4–8)
3
3
* Number of low risk judgements; Ø=mean; a Data were extracted from mean baseline values of each studies; b The active drug of these trials started with a low dose and reached the target dosage
within 1 week; c Doses of antipsychotic drugs are expressed in risperidone equivalent doses; A: augmentation;
BPRS: Brief Psychiatric Rating Scale; CCMD-3: China’s Mental Disorder Classification
and Diagnosis Standard, third edition; CLZ:=: clozapine; DSM-IV: Diagnostic and Statistical
Manual of Mental Disorders, fourth edition; DSM-IV-TR: Diagnostic and Statistical
Manual of Mental Disorders, fourth edition, text revision; DB: double blind; FD: fixed
dosage; HAL: haloperidol; ICD-10: the tenth revision of the International Statistical
Classification of Diseases and Related Health Problems; ITT intent-to-treat; M: monotherapy;
NR: not reported; OC: observed cases; OLA: olanzapine; PANSS: Positive and Negative
Syndrome Scale; REB: reboxetine; RIS: risperidone; Sz: schizophrenia; SzD: schizophreniform
disorder; T: total.
Results
Results of the search
The original search from the above databases yielded 339 electronic records and 1
trial retrieved by hand-search ([Fig. 1 ]). By the end of screening all papers, 9 RCTs [18 ]
[22 ]
[25 ]
[26 ]
[27 ]
[28 ]
[29 ]
[30 ]
[51 ] published in English (6 RCTs) [18 ]
[22 ]
[28 ]
[29 ]
[30 ]
[51 ] and Chinese (3 RCTs, (Table 1S ) [25 ]
[26 ]
[27 ] were eligible and analyzed.
Fig. 1 PRISMA flow diagram.
Study characteristics
All 9 RCTs (n=630) were double-blind and 4 (44.4%, n=211) used ITT analyses. The mean
study duration was 11.8±7.6 (range: 4–24, median: 12.0) weeks ([Table 1 ]). Three RCTs were conducted in China (n=363), 2 in Israel (n=85), and 1 each in
Iran (n=50), South Africa (n=30), Germany (n=35), and Spain (n=67).
Patient characteristics
The mean age of the 630 patients was 38.0±6.9 years (range: 29.9–50.4, median: 40.4
years); males accounted for 79.5±18.9% (range: 50.8–100, median: 73.1) of the sample.
The mean illness duration was 10.6±9.4 (range: 0.4–27.2, median: 8.5) years (8 RCTs
with available data) ([Table 1 ]). Six RCTs were conducted in inpatients (n=393), 1 RCT involved both in- and outpatients
(n=135), and the type of patients in 2 RCTs (n=102) was unspecified.
Treatment characteristics
Reboxetine dose was 6.2±1.8 mg/day (range: 4.0–8.0, median: 6.0 mg/day). Baseline
APs included olanzapine (2 RCTs, n=85), clozapine (2 RCTs, n=228), haloperidol (3
RCTs, n=215), and multiple APs (2 RCT, n=102) ([Table 1 ]).
Quality assessment
While 8 RCTs (88.9%) [18 ]
[22 ]
[25 ]
[26 ]
[27 ]
[28 ]
[29 ]
[51 ] were rated as low risk regarding attrition bias, 4 RCTs [18 ]
[22 ]
[29 ]
[30 ] described an adequate method of random sequence generation and 2 RCTs [29 ]
[30 ] were rated as low risk regarding the allocation concealment methods (Fig. 1S ). In addition, 5 RCTs (63%) [18 ]
[22 ]
[28 ]
[29 ]
[51 ] were rated as low risk for selective report bias. According to the GRADE approach,
the quality of evidence for 10 outcomes ranged from “low”(20%) via “moderate” (40%)
to “high” (40%) (Table 2S ). The mean score of the Jadad scale was 3.6±1.1 (range: 2–5; median: 3.0); 8 RCTs
(88.9%) were classified as high quality ([Table 1 ]).
Psychotic symptoms
Meta-analysis of negative symptoms (PANSS [6 RCTs] and SANS [3 RCTs]) showed that
adjunctive reboxetine was superior to placebo (9 RCTs, n=602, SMD: −0.47 [95% CI:
−0.87, −0.07], p=0.02; I2 =82%, [Fig. 2 ]). The significance (SMD: −0.36 [95% CI: −0.77, 0.05], p=0.09; I2 =79%) disappeared after 1 outlying (SMD ≤ −1.3) study [27 ] was removed. In subgroup analyses, the significance also disappeared in 6 out of
the 12 subgroups ([Table 2 ]). The superiority of reboxetine disappeared in non-Chinese studies (p=0.59), with
APs other than clozapine (p=0.27), trials lasting less than 12 weeks (p=0.37), male
predominance (≥60%) (p=0.10), mean age younger than 38.0 years (p=0.54), and having
a Jadad score more than 3 (p=0.10). (Table 2S ) presents subgroup analyses after leaving out 1 outlying study [27 ] and found that significance disappeared in 7 out of the 12 subgroups. The superiority
of reboxetine disappeared in non-Chinese studies (p=0.59), APs other than clozapine
(p=0.27), trials lasting less than 12 weeks (p=0.37), male predominance (≥60%) (p=0.26),
mean age younger (p=0.54) or older than 38.0 years (p=0.11), and having a Jadad score
more than 3 (p=0.26).
Fig. 2 Reboxetine for schizophrenia. Forest plot for clinical efficacy assessed with the
PANSS, BPRS, SAPS, or SANS.
Table 2 Subgroup analysis of the associations between moderating variables and change in
negative symptoms.
Variables
n (subjects)
SMDs (95% CI)
I2 (%)
p-value for each subgroup
p-value across subgroups
1. Chinese
3 (341)
−1.06 (−1.29, −0.82)
7
<0.00001
0.0002
Non-Chinese
6 (261)
−0.12 (−0.55, 0.31)
66
0.59
2. Antipsychotic class: Clozapine
2 (221)
−1.08 (−1.49, −0.66)
53
<0.00001
0.01
Other than Clozapine
7 (381)
−0.27 (−0.74, 0.21)
79
0.27
3. Trial duration (weeks)a : ≥12
5 (458)
−0.91 (−1.22, −0.61)
58
<0.00001
<0.00001
<12
4 (144)
0.16 (−0.19, 0.51)
10
0.37
4. Age (years)a : ≥38.0
5 (373)
−0.64 (−1.16, −0.11)
83
0.02
0.37
<38.0
4 (229)
−0.23 (−0.95, 0.49)
84
0.54
5. Male predominance (≥60%)
8 (482)
−0.38 (−0.84, 0.07)
82
0.1
0.03
No sex predominance
1 (120)
−1.04 (−1.42, −0.66)
N/A
<0.00001
6. Study quality: Jadad score ≥3
8 (482)
−0.38 (−0.84, 0.07)
82
0.1
0.03
Jadad score<3
1 (120)
−1.04 (−1.42, −0.66)
0
<0.00001
a Analyzed using a mean split method; CI: confidence interval; N/A: not applicable;
SMDs: standard mean differences; Bold values: p<0.05.
No group difference was found in change of overall psychopathology measured with the
PANSS (6 RCTs) (n=473, SMD: −0.50 [95% CI: −1.05, 0.06], p=0.08; I2 =88%; [Fig. 2 ]), positive symptoms (PANSS [6 RCTs] and SAPS [3 RCTs]) (n=602, SMD: −0.00 [95% CI:
−0.16, 0.16], p=0.98; I2 =0%; [Fig. 2 ]) and PANSS general psychopathology score (5 RCTs) (n=438, SMD: −0.46 [95% CI: −0.97,
0.05], p=0.08; I2 =85%, [Fig. 2 ]).
Weight change
Compared to placebo, reboxetine caused significant weight (3 RCTs, n=186, WMD: −3.83 kg
[95% CI: −7.40, −0.26], p=0.04; I2 =92%; [Fig. 3 ]) and BMI reduction (3 RCTs, n=186, WMD: −2.23 kg/m2 [95% CI: −4.35, −0.12], p=0.04; I2 =95%, [Fig. 3 ]). The results concerning weight (WMD: −1.90 kg [95% CI: −3.07, −0.72], p=0.002;
I2 =0%) was consistent even after 1 outlier (SMD<−1.0) study [27 ] was removed.
Fig. 3 Reboxetine for schizophrenia. Forest plot for the change of body weight and BMI.
Cognitive functions
Only 2 RCTs [26 ]
[27 ] assessed cognitive functions. One study [26 ] found reboxetine superior to placebo in memory quotient, recognition, and associative
learning assessed by the Wechsler Memory Scale-Revised, Chinese version. The other
study [27 ] found reboxetine outperformed placebo in attention, immediate memory, and delayed
memory assessed with the Repeatable Battery for the Assessment of Neuropsychological
Status. Because of the different scales used, meta-analysis of cognitive functions
was not possible.
Discontinuation rate and ADRs
All-cause discontinuations were similar between reboxetine and placebo (8 RCTs, n=580,
RR: 1.05 [95% CI: 0.71, 1.56], p=0.81, I2 =0%; (Fig. 2S ). Regarding ADRs, reboxetine caused more frequent dry mouth (p=0.04, NNH=14, 95%
CI: 7–50; Fig. 3S ) but was associated with less weight gain overall (p=0.01, NNH=7, 95% CI: 4–50) and
weight gain of ≥7% of the initial weight (p=0.006, NNH=3, 95% CI: 2–8; Fig. 3S ). Meta-analyses of akathisia, dizziness, insomnia, tachycardia, constipation, and
nausea/vomiting showed no significant differences between reboxetine and placebo (p=0.21–0.82;
Fig. 3S ).
Publication bias
Since a minimum of 10 RCTs are needed to conduct funnel plot or Egger’s test [52 ], publication bias was not assessed for negative symptoms (9 RCTs) and weight change
(3 RCTs).
Discussion
This meta-analysis found that reboxetine was not consistently effective in treating
negative symptoms because the significant improvement was driven by an outlying study
[27 ]. This result is consistent with previous findings [10 ]
[32 ]. In this meta-analysis, adjunctive reboxetine was effective in weight reduction
in the treatment of schizophrenia. Compared with Helfer et al.’s meta-analysis [32 ] of 5 RCTs [18 ]
[22 ]
[29 ]
[30 ]
[51 ], 4 additional RCTs [25 ]
[26 ]
[27 ]
[28 ] were included in this study, generating larger power and allowing more comprehensive
analyses. A recent meta-analysis [31 ] concluded that reboxetine may improve negative symptoms, but the evidence presented
was of a very low quality, probably due to the inclusion of an open-label study [35 ]. In the present meta-analysis, a significant effect of reboxetine on negative symptoms
was only found in the Sun et al.’s study (Jadad score<3, p<0.00001) but not in the
remaining 7 studies (Jadad score ≥3, p=0.26) after leaving out 1 outlying study [27 ].
Reboxetine administered for 12 weeks reduced weight by a mean of 3.83 kg by reducing
appetite and increasing energy expenditure [53 ]. In a recent RCT of 40 schizophrenia patients, a reboxetine-betahistine combination
caused significant weight loss compared to placebo (4.77 vs. 2.02 kg) [54 ]. In a systematic review of 40 trials, metformin achieved the greatest weight loss
(3.17 kg; 95% CI: −4.44, −1.90) compared to topiramate, sibutramine, aripiprazole,
and reboxetine [17 ]. In another meta-analysis [34 ] topiramate, aripiprazole, and sibutramine were more effective than reboxetine to
induce weight loss. To date, there has been no head-to-head study or meta-analysis
published that directly compared reboxetine and metformin on weight loss.
Adjunctive reboxetine appeared to be safe and well-tolerated. Reboxetine caused more
frequent dry mouth (NNH=14) but less weight gain (NNH=7) and weight gain of ≥ 7% of
the initial weight (NNH=3). Other ADRs and discontinuation were similar between reboxetine
and placebo.
Several studies examined the association between AP-induced weight gain and treatment
response in schizophrenia. Treatment response was positively associated with weight
gain induced by olanzapine or clozapine [19 ] and olanzapine and haloperidol [55 ]. However, only 33% (3/9) of the RCTs in the present meta-analysis explored the effect
of reboxetine on AP-induced weight gain, without investigating the association between
reboxetine and negative symptoms; thus, the association between AP-induced weight
gain and negative symptoms could not be assessed in this meta-analysis.
Three previous meta-analyses [12 ]
[33 ]
[34 ] of 2 RCTs [18 ]
[22 ] found similar advantage of adjunctive reboxetine for attenuating weight gain in
schizophrenia, but they did not analyze its effects on psychotic symptoms. The current
meta-analysis included additional 7 RCTs, allowing more robust and sophisticated analyses,
including risk of bias, Jadad scale, GRADE approach, and sensitivity analysis.
There are several limitations of this meta-analysis. First, although all included
RCTs were rated as high-quality trials according to the Cochrane risk of bias and
the Jadad scale, all 9 RCTs providing the data of co-primary outcomes had relatively
small sample size (26–135), which precluded the assessment of publication bias. Second,
meta-analyzable results for body weight (I2 =92%) were heterogeneous, although the statistical significance remained (I2 =0%) after removing an outlying study from the analysis. Third, reboxetine doses varied
across the 9 studies (4.0–8.0 mg/day); therefore, the dose-response effects of reboxetine
in reducing AP-induced weight gain could not be examined. Fourth, cognitive functions
were assessed only in 2 trials [26 ]
[27 ] with conflicting conclusions. The cognitive aspects of reboxetine treatment warrant
further investigations. Fifth, all studies had relatively short treatment duration
(4–24 weeks); thus, reboxetine’s long-term effects on body weight need further studies.
Finally, metabolic indices associated with body weight, such as lipid profile, insulin
resistance, and leptin could not be analyzed as they were not recorded in the studies.
In conclusion, in this meta-analysis, reboxetine appeared to attenuate AP-induced
weight gain in patients with schizophrenia, but it was not consistently effective
in treating negative symptoms of schizophrenia. The therapeutic effects of reboxetine
on negative symptoms remained doubtful in this meta-analysis as it was driven by an
outlying study. In the outlying study [27 ], schizophrenia patients with metabolic syndrome were recruited. Therefore, the effect
of reboxetine on negative symptoms needs to be probably replicated. In addition, high-quality
RCTs are warranted to demonstrate reboxetine’s long-term safety and efficacy, particularly
on cognitive functions.
Contributors
WZ and XHY selected studies and conducted statistical analysis. XHY and DBC extracted
the data. YTX reviewed all the data and helped mediate disagreements. WZ, XBL, and
ZMS wrote the first draft. All the authors contributed to the interpretation of data
and completing and approving the final manuscript.
