Doi:10.1016/s0140-6736(07)61721-8

Articles
cacy and safety of the weight-loss drug rimonabant:
a meta-analysis of randomised trials
Robin Christensen, Pernelle Kruse Kristensen, Else Marie Bartels, Henning Bliddal, Arne Astrup Lancet 2007; 370: 1706–13
Background Since the prevalence of obesity continues to increase, there is a demand for eff ective and safe anti-obesity
See Comment page 1671
agents that can produce and maintain weight loss and improve comorbidity. We did a meta-analysis of all published
The Parker Institute,
randomised controlled trials to assess the effi
cacy and safety of the newly approved anti-obesity agent rimonabant.
Musculoskeletal Statistics Unit,
Frederiksberg Hospital,
Methods We searched The Cochrane database and Controlled Trials Register, Medline via Pubmed, Embase via
Frederiksberg, Denmark
WebSpirs, Web of Science, Scopus, and reference lists up to July, 2007. We collected data from four double-blind,
randomised controlled trials (including 4105 participants) that compared 20 mg per day rimonabant with placebo.
Prof H Bliddal MD);
The Department of Human
Findings Patients given rimonabant had a 4·7 kg (95% CI 4·1–5·3 kg; p<0·0001) greater weight reduction after 1 year
Nutrition, Faculty of Life
than did those given placebo. Rimonabant caused signifi cantly more adverse events than did placebo (OR=1·4;
Sciences, University of
Copenhagen, Copenhagen,
p=0·0007; number needed to harm=25 individuals [95% CI 17–58]), and 1·4 times more serious adverse events
Denmark (P K Kristensen,
(OR=1·4; p=0·03; number needed to harm=59 [27–830]). Patients given rimonabant were 2·5 times more likely to
Prof A Astrup MD); and
discontinue the treatment because of depressive mood disorders than were those given placebo (OR=2·5; p=0·01;
Copenhagen University Library,
number needed to harm=49 [19–316]). Furthermore, anxiety caused more patients to discontinue treatment in
Copenhagen, Denmark
rimonabant groups than in placebo groups (OR=3·0; p=0·03; number needed to harm=166 [47–3716]).
Interpretation Our fi ndings suggest that 20 mg per day rimonabant increases the risk of psychiatric adverse events—ie,
depressed mood disorders and anxiety—despite depressed mood being an exclusion criterion in these trials. Taken
together with the recent US Food and Drug Administration fi nding of increased risk of suicide during treatment with
rimonabant, we recommend increased alertness by physicians to these potentially severe psychiatric adverse reactions.
ast@life.ku.dk
Introduction
clinical trials on rimonabant have been published,10–13 The prevalence of obesity continues to increase and all show an increased weight loss compared with worldwide,1 and there is a demand for eff ective and safe placebo of 4–6 kg over 6–12 months, with few tolerability anti-obesity agents that can produce and maintain weight or safety concerns. In the Rimonabant in Obesity loss and improve comorbidities. Obesity is producing (RIO)-Europe study,10 the investigators concluded that several health-related consequences.2,3 Weight loss of rimonabant was generally well tolerated with mild and 5–10% of bodyweight, irrespective of how it is achieved, transient side-eff ects. However, the individual trials is associated with improvements in cardiovascular risk showed trends to increases in depressed mood, profi les and reduced incidence of type 2 diabetes.2,4,5 depression, and severe adverse events. Furthermore, Non-pharmacological treatment can be eff ective, but because patients with serious mental illness were success rate in the long term is low.6,7 Obesity guidelines excluded from the RIO programme, the estimates of recommend that anti-obesity pharmacotherapy is to be the potential psychiatric side-eff ects of the drug are regarded as a potentially important adjunctive treatment conservative.14 Rimonabant was recently assessed by the to non-pharmacological therapy for patients with a US Food and Drug Administration (FDA), and body-mass index (BMI) greater than or equal to 30 kg/m², coinciding with the submission of this paper, the FDA’s or a BMI of 27·0–29·9 kg/m² with one or more major Advisory Committee unanimously concluded that more obesity-related comorbidities.2 The anti-obesity agent detailed safety information about rimonabant in larger rimonabant (acomplia, Sanofi -Aventis, Paris, France), patient numbers over the long term was needed before has been approved by the European Agency for the the drug could be approved.15 Evaluation of Medicinal Products (EMEA) in June, 2006, We aimed to do a meta-analysis of rimonabant studies and is available in Argentina, Austria, Denmark, Finland, cacy and safety of the drug, emphasising Germany, Ireland, Norway, Sweden, Greece, and the psychiatric adverse events such as depressive disorders UK.
that could potentially lead to suicide.
Rimonabant is a selective antagonist of the cannabinoid type 1 receptor, and it is the fi rst member Methods
of a new class of compounds that targets the Search strategy and selection criteria
endocannabinoid system, which has been shown to be Five bibliographic databases (Medline from mid-1950s,
involved in the central and peripheral regulation of food
Embase from 1980, Web of Science from 1945–54, Scopus intake and the CNS rewarding system.8,9 So far, four from 1966, and the Cochrane Library) were searched up www.thelancet.com Vol 370 November 17, 2007
Articles
to November, 2006, for randomised controlled trials confi dence limits for the OR.18 We applied the Fisher’s investigating rimonabant and weight loss. Search terms exact test to calculate the exact probabilities of the were (“rimonabant” OR “Acomplia” OR [“antagonist” possible (2×2) tables, enabling us to estimate the Wald AND “cannabinoid” AND “receptor”]) AND (“obesity” test associated variance, corresponding to the ratio of its OR “weight loss” OR “overweight” OR “weight reduction” estimate (log -OR) to its standard error. Accordingly, OR “slimming”) AND “controlled”. All clinical trials these variances are applicable for subsequent relating rimonabant to weight loss were identifi ed. The mixed-eff ects meta-analysis. All analyses were based on reference lists of review articles and of included studies data reported as intention to treat, for which all RIO were searched to identify other potentially eligible studies applied the last observation carried forward studies. There was no limitation on language.
(LOCF) techniques for missing outcome data.
Only double-blind, randomised controlled trials using To combine the individual study results we did rimonabant for weight loss in overweight or obese meta-analyses using SAS software (version 9.1.3), participants were eligible for inclusion. Included studies applying a restricted maximum likelihood (REML) had to fi rst, enrol patients with BMI levels of 30 kg/m² or method to estimate the between study variance and the greater or 27 kg/m² or greater plus one or more combined effi obesity-related comorbidity, and second, include a heterogeneity between trials with a standard Q-test placebo control group. Studies assessing any of the statistic (testing the hypothesis of homogeneity),21 and following terms were included: “rimonabant”, “acomplia”, present the I² value, which can be interpreted as the “(endo) cannabinoid antagonist”, “SR141716”, and percentage of total variation across several studies due “SR141716A” (phase I and II codes for rimonabant).
to heterogeneity.22 On the basis of combined OR values, we estimated the number needed to treat and the Data extraction and quality assessment
number needed to harm, with 95% CIs, since this Two investigators (PKK, EMB) did the literature search method enables direct translation into clinical practice; and reviewed the results. Full articles were retrieved for these data were calculated on the basis of the combined further assessment if the information in the abstract OR measure, applying the overall event rate in the suggested that the study: (1) compared rimonabant with placebo group as a proxy for baseline risk.23 To investigate placebo or any other intervention; (2) included potential sources of clinical heterogeneity, we assessed participants who where overweight or obese; and (3) the extent to which study-level variables were associated assessed weight loss (in kg) as outcome. Two investigators with safety by fi tting REML-based metaregression (RC, PKK) were responsible for the assessment and models.24 See Online for webfi gure
diff erence in mean weight change and the number of individuals achieving at least 10% weight reduction handled as a dichotomous responder criterion. The Jadad assessment method (Instrument to Measure the Likelihood of Bias)16 for quality assessment of randomised 170 studies excluded because they did not include controlled trials was used as a guide to assess study quality. Quality of the included studies was assessed independently by two reviewers (RC, PKK) and any diff erences were resolved at the subsequent consensus studies identified andscreened for retrieval 22 lifestyle modification trials (not a drug study) 14 animal models Statistical analysis
2 same clinical design of already included studies We calculated the weighted mean diff erence for the 5 non-research reports 1 results of 2 year treatment of already included studies diff erence in mean weight change, standardised mean 7 duplicate, preliminary, or subgroup analyses of already diff erence for the hospital anxiety and depression scale (HADS) score, and odds ratios (OR) for dichotomous outcomes. Since asymptotic results can be unreliable when the distribution of the dichotomous data is sparse (as would be expected for serious adverse events, 1 study in progress 1 meta-analysis of trials depression, and anxiety), we used exact methods for the calculation of the confi dence intervals around the ORs. with usable information forprimary outcome (weight loss) confi dence limits is not necessarily exact on a nominal level, these confi dence limits are conservative, which is the recommended approach when handling sparse data.17 Figure 1: Flow chart of the search strategy and selection of trials
The SAS procedure PROC FREQ computes exact Webfi gure shows the full list of excluded studies.
www.thelancet.com Vol 370 November 17, 2007
Articles
Metabolic Smokers
Defi nite sample
individuals
(kg/m²)
(mmol/L)
(mmol/L)
(mmol/L)
syndrome
size (ITT)
Data are number (%) or mean (SD). ITT=intention to treat. BMI=body-mass index. TG=triglycerides. HDL-C=high-density lipoprotein cholesterol. *The number of ITT individuals is based on patients allocated to the rimonabant 20 mg and placebo groups, respectively; patients receiving rimonabant 5 mg were excluded. n and n are the number of individuals in the exposed and control group (ie, rimonabant and Table 1: Summary of baseline characteristics of all participants in the eligible trials
Favours placebo
Favours rimonabant
full access to all the data in the study and had the fi nal Rimonabant
WMD (95% CI)
responsibility for the decision to submit for mean weight
mean weight
Results
We identifi ed 227 studies in the database searches
(fi gure 1). If a study contained phases for both weight loss and weight maintenance, we included only data from the weight loss phase. After review of the identifi ed studies, 57 were identifi ed and retrieved for further scrutiny. We included only four randomised RIO-Overall
4·67 (4·06–5·27)
placebo-controlled, double-blind trials,10–13 which were of high quality (Jadad assessment score of about 5) and met Test for heterogeneity: χ²=8·02 (p=0·05), I²=62·6% the inclusion criteria. All included trials were in the RIO Test for overall effect: Z=15·07 (p<0·0001) programme (RIO-Europe,10 RIO-Lipids,11 RIO-North America,12 and RIO-Diabetes13), which investigated the cacy and safety of rimonabant for treatment of obesity, diabetes, and metabolic disorders in overweight and Rimonabant
OR (95% CI)
Table 1 shows the average baseline characteristics of the included studies. The trials were undertaken between 2001 and 2005 (published in 2005–06), and varied in pant size (table 1). All trials were multicentre studies and 4105 participants were assessed in total (focusing solely on the intention-to-treat individuals receiving either 20 mg per day or placebo). The length of the trials varied from 12 months to 24 months. However, all the studies published detailed statistics after 1 year of RIO-Overall
5·11 (3·57–7·31)
treatment. The manufacturer Sanofi -Aventis sponsored Test for heterogeneity: χ²=7·03 (p=0·07), I²=57·3% all the included studies. The studies had the same primary Test for overall effect: Z=8·92 (p<0·0001) endpoint and similar secondary endpoints (adapted to the specifi c population recruited in the study and the tests undertaken). The scheduled visits were planned at the Number of patients with >10% weight loss same time-points. After a weight maintenance diet for 4 weeks, patients who were compliant to dietary Figure 2: Effi
cacy of rimonabant compared with placebo in overweight individuals
instruction and coping with therapy were randomly (A) Mean weight change. (B) Number of participants achieving a weight loss of 10% or more during 1 year. Every assigned to placebo, rimonabant 5 mg per day, or square represents the individual study’s weight loss estimate with 95% CI indicated by horizontal lines. Square sizes are directly proportional to the precision of the estimate. WMD=weighted mean diff erence. rimonabant 20 mg per day, in addition to a hypocaloric diet (600 kcal per day defi cits).
Role of the funding source
The objectives of all the available studies were to The sponsor of the study had no role in study design, establish the long-term (1 year) effi data collection, data analysis, data interpretation, or rimonabant in the treatment of obesity and metabolic writing of the report. The corresponding author had disorders related to obesity. www.thelancet.com Vol 370 November 17, 2007
Articles
Since all analyses were based on published Rimonabant
SMD (95% CI)
intention-to-treat (LOCF) data, we extracted the exact number of patients randomly assigned and who completed the fi rst year of treatment from the studies’ Depression* (p=0·21)
trial profi les (CONSORT recommended).25 The OR values RIO-Europe†10
associated with the individual study’s 2×2 cross-table of reported adherent individuals (who completed 1 year of treatment) were consistent when rimonabant was compared with placebo (I²≤0%). There was no reason to suggest a rejection of the hypothesis of homogeneity Anxiety‡ (p=0·0009)
based on the Cochran Q test (Q=0·65; p=0·89), with the RIO-Europe†10
likelihood of completing the 1 year therapy with rimonabant or placebo being equal (OR=1·12; 95% CI 0·99–1·28). These data correspond to 1486 (59%) of the individuals randomly assigned to rimonabant for 1 year, Compared with placebo, rimonabant therapy resulted SMD=standardised mean diff erence. *Test for homogeneity for depression: χ²=6·99, I²=57·1%, (p=0·07). †Value after
1-year intervention—ie, change within group is not explicitly reported. ‡Test for homogeneity for anxiety: χ²=3·00,
in a greater reduction in bodyweight of 4·7 kg (95% CI 4·1–5·3) than did placebo (p<0·0001). All studies reported greater reductions in bodyweight in the rimonabant Table 2: Hospital anxiety and depression (HAD) scale subscores for all eligible studies
group than in the placebo group (fi gure 2). Some heterogeneity between the individual effi Favours rimonabant
Favours placebo
was evident when we compared the magnitude of weight Rimonabant
OR (95% CI)
reduction (Q=8·02; p=0·05), corresponding to a high eff ect on the combined estimate (I²=62·6%). Individuals receiving rimonabant therapy were fi ve times more likely to achieve at least 10% weight loss (OR=5·1 [95% CI 3·6–7·3]; p<0·0001) than were those taking placebo (fi gure 2). On the basis of the average number of responders within the rimonabant and placebo groups (639 [26%] and 106 [7%], respectively), this OR corresponded to a number needed to treat of six RIO-Overall
1·35 (1·13–1·60)
Mood was assessed with HADS at baseline and every Test for heterogeneity: χ²=0·58 (p=0·90), I²=0% 3 months. HADS contains seven items to assess Test for overall effect: Z=3·39 (p=0·0007) depressive symptoms and seven items to assess anxiety symptoms, with scores ranging from 0 to 3 for each item. Depressive and anxiety symptoms are separately Rimonabant
summarised. Scores of 0–7 are regarded as normal, OR (95% CI)
8–10 represent borderline symptoms, and 11 is regarded as high enough to warrant further assessment of the patient.26 Questions about suicidal thoughts are not part of the questionnaire.
All four individual studies reported HADS subscores for depression and anxiety. Since the RIO-Europe study did not report changes from baseline, we estimated the group mean diff erences as the standardised mean diff erence. There was no signifi cant diff erence between RIO-Overall
1·43 (1·03–1·98)
rimonabant and placebo in regard to the depression subscore, whereas the increase in the anxiety score was Test for heterogeneity: χ²=1·75 (p=0·63), I²=0% greater in the rimonabant group than in the placebo Test for overall effect: Z=2·15 (p=0·03) Adverse events were more likely to arise with rimonabant treatment than with placebo (fi gure 3). The
Cochran Q test for homogeneity suggested that the eff ect Figure 3: Safety of 20 mg per day rimonabant treatment compared with placebo
of rimonabant on adverse events was much the same in (A) The number of individuals who had an adverse event. (B) The number of individuals who had a serious adverse all the studies (I²=0%; Q=0·58; p=0·90). Patients event. Data based on an exact computation algorithm. www.thelancet.com Vol 370 November 17, 2007
Articles
Depressed mood disorders consisted of depression, Favours rimonabant
Favours placebo
major depression, depressive mood, and depressive Rimonabant Placebo
OR (95% CI)
symptoms. To analyse this endpoint we used the composite endpoint depressed mood disorder also for RIO-Lipids (estimated as the sum of subcategories). More psychiatric disorders were evident in the rimonabant groups than in the placebo groups leading to discontinuation (fi gure 4). Patients assigned to rimonabant were 2·5 times more likely to discontinue in the trial because of depressive mood disorders than were those receiving placebo (74 [3·0%] vs 22 [1·4%]; RIO-Overall
2·51 (1·23–5·12)
OR=2·5 [1·2–5·1]; p=0·01; number needed to harm=49 individuals [19–316]; fi gure 4). Furthermore, Test for heterogeneity: χ²=5·41 (p=0·14), I²=44·5% anxiety caused more frequent discontinuation in Test for overall effect: Z=2·53 (p=0·01) rimonabant groups than in placebo groups (26 [1·0%] vs 5 [0·3%]; OR=3·0 [1·1–8·4]; p=0·03; number needed to harm=166 individuals [47–3716]; fi gure 4). To explore the reasons for heterogeneity we undertook multiple metaregression models to reduce the Rimonabant Placebo
OR (95% CI)
between-study variance, based on the study-level covariates shown in table 1. We only focused on covariates that were able to reduce the eff ect of heterogeneity for two of the outcome measures: the log-OR values considering the number of patients with depressed mood and serious adverse events. We noted an association between the OR of depressed mood and the average triglyceride concentration at baseline, which could suggest that high concentrations of triglyceride at baseline might be a predictor for individuals who are RIO-Overall
3·03 (1·09–8·42)
likely to have depression during their use of rimonabant. Mean age at baseline seemed to be a predictor of the log Test for heterogeneity: χ²=0·61 (p=0·89), I²=0% Test for overall effect: Z=2·13 (p=0·03) OR for serious adverse events, suggesting that elderly people are more likely to have serious adverse events during treatment with rimonabant than are younger Figure 4: Number of individuals who discontinued treatment because of adverse psychiatric events
Discussion
(A) Discontinuation because of depressed mood disorders, which is a composite endpoint that consists of Four trials in the RIO programme assessed the effi depression, major depression, depressive mood, and depressive symptoms. (B) Discontinuation because of and safety of the anti-obesity agent rimonabant compared anxiety. Data based on exact computation algorithms. with placebo, and our meta-analysis has shown that receiving 20 mg per day rimonabant were 1·4 times more rimonabant therapy produced a greater weight loss than likely to report adverse events than were those receiving did placebo. Patients who were allocated to rimonabant placebo (OR=1·4 [95% CI 1·1–1·6]; p=0·0007). The were much more likely to achieve a 10% weight reduction average numbers of events within the rimonabant and after 1 year compared with those allocated to placebo. placebo groups were 2152 (86·0%) and 1310 (81·8%), These fi gures are in agreement with the outcome of a respectively, corresponding to a number needed to harm Cochrane meta-analysis,27 and suggest that rimonabant is of 25 individuals (95% CI 17–58). Serious adverse events similar to or slightly better than existing weight-loss also occurred more frequently in the rimonabant than in the placebo groups (148 [5·9%] vs 67 [4·2%]; A meta-analysis28 of 11 orlistat trials and fi ve OR=1·4 [1·0–2·0]; p=0·03), corresponding to a number sibutramine trials lasting 1 year or longer showed that needed to harm of 59 individuals (27–830; fi gure 3). patients given orlistat lost 2·7 kg (95% CI 2·3–3·1) more In the RIO-Lipids study,11 the depressive mood weight, and patients taking sibutramine lost 4·3 kg disorders defi ned as a cause of discontinuation were (3·6–4·9) more weight than did those taking placebo. subcategorised into depression, major depression, and The intensity and the enforcement of the restriction in depressed mood, whereas the RIO-Diabetes,13 calories in the RIO programme was fairly weak, and RIO-Europe,10 and RIO-North America12 only reported produced only modest weight loss of less than 2 kg in the composite group depressed mood disorders. the placebo groups. However, the more strictly the www.thelancet.com Vol 370 November 17, 2007
Articles
energy restriction is enforced and adhered to, the greater mood disorders. Participants given rimonabant were also weight loss that is seen in both treatment groups, thus at greater risk to discontinue treatment because of an leaving little room for an appetite suppressant such as increased risk of developing anxiety.
rimonabant to further reduce energy intake and produce Our study had several limitations mainly because of the weight loss. With a more eff ective dietary treatment absence of access to all available sources reporting safety programme, the additional weight loss produced by data, including unpublished phase 2 trials and several rimonabant would probably be less than that we studies in progress, but also because of an absence of reported.
consistent reporting of psychiatric severe adverse eff ects. In the RIO trials, no follow-ups were reported after Additionally, the endpoint reported in the studies was discontinuation of active treatment, thus any weight depressed mood disorders, which consisted of depres-regain could not be assessed. As with other weight-loss sion, major depression, depressive mood, and depressive drugs, relapse is expected to occur after treatment has symptoms, and these disorders have substantially ended, and to achieve weight maintenance and maintain diff erent severity and clinical implications. the improvement of the cardiovascular and diabetes risk Our analysis did not allow examination for psychiatric factors the drug needs to be taken for life. disorders other than depression and anxiety. However, In our meta-analysis, patients with type 2 diabetes according to the FDA analysis, rimonabant treatment achieved a lower placebo-subtracted weight reduction led to more adverse events than did placebo: irritability, with rimonabant than did non-diabetic patients, which is insomnia, stress, and nervousness were present in more consistent with the general observation that weight loss than 1% of the treated patients. Panic attacks, agitation, is more diffi cult to achieve in patients with type 2 nightmare, and abnormal dreams were more frequently reported by patients treated with rimonabant than with Obesity has been shown to be associated with placebo.15 Although the number of patients discontinuing depression in the general population and in clinical therapy because of adverse events seemed to be small, samples,30 especially in women and severely obese men.31 high study attrition rates raise the possibility that some Obese individuals who are seeking treatment are events were not documented. especially prone to depression. Depression has been Moreover, in all the RIO studies the enrolled patient reported to range up to 48% in these individuals, and this populations were highly selected, since patients with a proportion increases as the severity of obesity increases.32,33 past history of severe depression or those with present Obese women are 20% more likely to report suicidal severe psychiatric illness were excluded. Antidepressant ideation and 23% more likely to have made a suicide treatment was not permitted and warranted mandatory attempt in the past year compared with non-obese treatment discontinuation, as prespecifi ed by the original women.34 By contrast with the Cochrane review,27 we RIO protocols. Information about depression, depressed found it pertinent to assess the eff ect of rimonabant on mood, anxiety, etc, was self-reported, which means that psychiatric events with a focus on mood and depression. under-reporting bias could have been present. For these Because of the limitations of the trials, the risk of a severe reasons, our estimates of depressive mood disorders are psychiatric adverse event could not be examined, and the assessment could be achieved by only two diff erent Our fi ndings strongly accord with the FDA report about analyses: changes in the HADS score35 and discontinuation the safety of rimonabant that was released after the initial of treatment because of depressive mood disorder and submission of the present paper,15 although the anxiety. investigators did report endpoints other than those We noted a greater increase in anxiety reported by the reported in our analysis. The FDA reported in their HADS score in participants taking rimonabant than in meta-analysis of RIO studies that 26% of people given those taking placebo, but we failed to fi nd any eff ect on rimonabant 20 mg versus 14% of those given placebo had depression. However, HADS is generally not used as a a psychiatric symptom reported as an adverse event. They primary outcome measure in clinical trials of depression, noted that 9% of participants given rimonabant 20 mg but it is regarded as an acceptable method to screen for versus 5% given placebo reported symptoms of depression depression and anxiety primarily in non-psychiatric (depressed mood, depression, depressive symptom, or patients.35 According to the RIO protocols, an increase in major depression), and consistent with our fi ndings these the HADS score above 11 would imply that the patient incidents often led to withdrawal of the drug. The FDA should be seen by a psychiatrist for further assessment, further reported that the overall relative risk for psychiatric but none of the RIO trials reported the number of adverse events in the rimonabant 20 mg group versus participants who were discontinued from trials after placebo group was 1·9 (95% CI 1·5–2·3). The number of psychiatric consultations. Another limitation of HADS is participants needing an anxiolytic or hypnotic agent for a the absence of questions investigating suicidal thoughts. psychiatric adverse event was 185 (9%) taking rimonabant Our meta-analysis has shown that more obese people 20 mg, 102 (5%) taking rimonabant 5 mg, and 66 (4%) in the 20 mg per day rimonabant groups than in placebo taking placebo. Another 104 (5%) taking rimonabant groups were taken off treatment because of depressed 20 mg, 88 (4%) taking rimonabant 5 mg, and 46 (3%) www.thelancet.com Vol 370 November 17, 2007
Articles
taking placebo needed an antidepressant agent for a Contributors
psychiatric adverse event.
RC participated in the study conception and design, the acquisition of The OR for the incidence of suicide was examined by the data, the analysis and interpretation of data, drafting of the manuscript, revision of the manuscript, and the statistical analyses. PKK participated FDA from all available studies, including smoking in the study conception and design, the acquisition of data, the analysis cessation trials, and was found to be 1·9 (1·1–3·1) for and interpretation of data, and drafting of the manuscript. EMB 20 mg rimonabant versus placebo. When limited to seven participated in the acquisition of data, critical revision of the manuscript, obesity studies, including the unpublished and contiuning and has seen and approved the fi nal version. HB participated in the study conception and design, and interpretation of data, critical revision trials, the OR for incidence of suicide for 20 mg rimonabant of the manuscript, and supervision of the study. AA participated in the versus placebo was 1·8 (0·8–3·8). In the entire database study conception and design, interpretation of data, writing and
for rimonabant clinical trials, there have been only two revisions of the manuscript, and supervision of the study. All authors
deaths from suicide—one in RIO North America in a have seen and approved the fi nal version of the manuscript.
patient taking rimonabant 5 mg and one in a study in Confl ict of interest statement
PKK, EMB, and HB declare that they have no confl ict of interest. RC was progress in a patient taking rimonabant 20 mg.15 statistical expert/consultant in the Lantus medical expert panel for The use of pretreatment markers to identify obese Sanofi -Aventis (Denmark) in 2006. AA participates in several advisory patients at high risk for developing depressed mood boards for biotechnology and pharmaceutical companies, some of which disorders would be of great clinical importance. We are developing CB-1 antagonists for treatment of obesity. AA is president of the International Association for the Study of Obesity (IASO), which therefore looked for possible predictive baseline markers, had received funding from Sanofi -Aventis when he was president-elect. and found that high triglyceride concentrations were AA participated in the Danish rimonabant advisory board for Sanofi -
associated with an increased probability of having Aventis, until its closure in June, 2006.
depression during treatment with rimonabant. In Acknowledgments
published work there is suggestive evidence from We thank the personal and scientifi c support of Bente Danneskiold-Samsøe,
cross-sectional studies that lends support to a link between
(MD, Head of The Parker Institute) and the linguistic support of Tina Cuthbertson. This study was supported by grants from the Center high triglyceride concentrations and depression,36,37 but for Pharmacogenomics, University of Copenhagen, The Oak Foundation, any causal relation remains to be established. Furthermore, the H:S Research Foundation, and Diabesity EC-FP6 (contract number: increasing age seemed to be a predictor of serious adverse events arising from rimonabant, which could be because References
of altered pharmacokinetics. These fi ndings deserve to be WHO technical report series. Obesity: preventing and managing the global epidemic–report of a WHO consultation on obesity, further addressed, and we suggest that the available raw June 3–5, 1997. Geneva: Switzerland, 2000. data is provided from every patient in all existing trials, 2 Haslam DW, James WP. Obesity. Lancet 2005; 366: 1197–209.
enabling an individual patient data meta-analysis to be 3
Avenell A, Broom J, Brown TJ, et al. Systematic review of the undertaken.38 Such a meta-analysis is seen as a gold long-term eff ects and economic consequences of treatments for obesity and implications for health improvement. standard compared with the study-level meta-analysis.39 Health Technol Assess 2004; 8: 1–182.
This would be the simplest way to investigate the safety 4 Astrup A. Treatment of obesity. In: DeFronzo RA, Ferrannini E, concerns proposed by the present study, and by the FDA Keen H, Zimmet P, eds. International textbook of diabetes mellitus 3rd edn. Chichester: John Wiley & Sons Ltd, 2004: analyses, and also to address whether the psychiatric adverse events are associated with the magnitude of the 5 Goldstein DJ. Benefi cial health eff ects of modest weight loss. Int J Obes Relat Metab Disord 1992; 16: 397–415.
Although all the included trials excluded patients with 6 Astrup A, Rossner S. Lessons from obesity management programmes: greater initial weight loss improves long-term existing depression, or those with a history of depressed maintenance. Obes Rev 2000; 1: 17–19.
mood a priori, this selected population of obese individuals Wadden TA. Treatment of obesity by moderate and severe caloric had an increased risk of developing depressive mood restriction. Results of clinical research trials. Ann Intern Med 1993;
119: 688–93.
disorders during 20 mg per day rimonabant therapy. That Di Marso V, Matias I. Endocannabinoid control of food intake and obese patients who are prescribed rimonabant in clinical energy balance. Nat Neurosci 2005; 8: 585–89.
practice are less likely to be screened for depression 9 Rinaldi-Carmona M, Barth F, Congy C, et al. SR147778 [5-(4-bromophenyl)-1-(2,4-dichlorophenyl)-4-ethyl-N- disorders than the participants in the trials is a matter of (1-piperidinyl)-1H-pyr azole-3-carboxamide], a new potent and concern. Consequently, the number needed to harm selective antagonist of the CB1 cannabinoid receptor: biochemical could be much lower than this fi nding in clinical practice. and pharmacological characterization. J Pharmacol Exp Ther 2004;
310: 905–14.
Patients who seek treatment in clinical practice are often 10 Van Gaal LF, Rissanen AM, Scheen AJ, Ziegler O, Rossner S, for obese women with less comorbidity, and for this group the RIO-Europe Study Group. Eff ects of the cannabinoid-1 the risk of severe adverse events is less acceptable. receptor blocker rimonabant on weight reduction and cardiovascular risk factors in overweight patients: 1-year Other potential anti-obesity drugs that increase the risk experience from the RIO-Europe study. Lancet 2005; 365: 1389–97.
of depression have been withdrawn because of cases of 11 Despres JP, Golay A, Sjostrom L. Eff ects of rimonabant on suicide ideation, suicide attempts, and death from metabolic risk factors in overweight patients with dyslipidemia.
N Engl J Med 2005; 353: 2121–34.
suicide.40 The fi ndings of our meta-analysis suggest that 12 Pi-Sunyer FX, Aronne LJ, Heshmati HM, Devin J, Rosenstock J. the potential of rimonabant to induce depressive Eff ect of rimonabant, a cannabinoid-1 receptor blocker, on weight symptoms and depression in overweight patients needs and cardiometabolic risk factors in overweight or obese patients: RIO-North America: a randomized controlled trial. 295: 761–75.
www.thelancet.com Vol 370 November 17, 2007
Articles
13 Scheen AJ, Finer N, Hollander P, Jensen MD, Van Gaal LF, for the 28 Padwal R, Li S, Lau D. Long-term pharmacotherapy for obesity and overweight. Cochrane Database Syst Rev 2003; 4: CD004094.
in overweight or obese patients with type 2 diabetes: a randomised 29 Wing RR, Marcus MD, Epstein LH, Salata R. Type II diabetic controlled study. Lancet 2006; 368: 1660–72.
subjects lose less weight than their overweight nondiabetic spouses. 14 Padwal RS, Majumdar SR. Drug treatments for obesity: orlistat, Diabetes Care 1987; 10: 563–66.
sibutramine, and rimonabant. Lancet 2007; 369: 71–77.
30 Dixon JB, Dixon ME, O’Brien PE. Depression in association with 15 US Food and Drug Administration Advisory Committee. FDA severe obesity: changes with weight loss. Arch Intern Med 2003; 163:
briefi ng document: Zimulti (rimonabant) Tablets, 20 mg. Rockville: FDA, 2007. http://www.fda.gov/OHRMS/DOCKETS/AC/07/ 31 Onyike CU, Crum RM, Lee HB, Lyketsos CG, Eaton WW. Is obesity briefi ng/2007-4306b1-00-index.htm (assessed Aug 6, 2007).
associated with major depression? Results from the Third National 16 Jadad AR, Moore RA, Carroll D, et al. Assessing the quality of Health and Nutrition Examination Survey. Am J Epidemiol 2003; reports of randomized clinical trials: is blinding necessary? 158: 1139–47.
Control Clin Trials 1996; 17: 1–12.
32 Goldsmith SJ, Anger-Friedfeld K, Beren S, Rudolph D, Boeck M, 17 Agresti A. A Survey of Exact Inference for Contingency Tables. Aronne L. Psychiatric illness in patients presenting for obesity Stat Sci 1992; 7: 131–77.
treatment. Int J Eat Disord 1992; 12: 63–71.
18 Thomas DG. Algorithm AS 36: exact confi dence limits for the odds 33 Wadden TA, Sarwer DB, Womble LG, Foster GD, McGuckin BG, ratio in a 2×2 table. Appl Stat 1971; 20: 105–10.
Schimmel A. Psychosocial aspects of obesity and obesity surgery. 19 Normand SL. Meta-analysis: formulating, evaluating, combining, Surg Clin North Am 2001; 81: 1001–24.
and reporting. Stat Med 1999; 18: 321–59.
34 Fabricatore AN, Wadden TA. Psychological functioning of obese 20 van Houwelingen HC, Arends LR, Stijnen T. Advanced methods in individuals. Diabetes Spectr 2003; 16: 245–52.
meta-analysis: multivariate approach and meta-regression. Stat Med 35 Zigmond AS, Snaith RP. The hospital anxiety and depression scale. 2002; 21: 589–624.
Acta Psychiatr Scand 1983; 67: 361–70.
21 Cochran WG. The combination of estimates from diff erent 36 Chen CC, Huang TL. Association of serum lipid profi les with experiments. Biometrics 1954; 10: 101–29.
depressive and anxiety disorders in menopausal women. 22 Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring Chang Gung Med J 2006; 29: 325–30.
inconsistency in meta-analyses. BMJ 2003; 327: 557–60.
37 Almeida OP, Flicker L, Norman P, et al. Association of 23 Altman DG, Deeks JJ. Meta-analysis, Simpson’s paradox, and the cardiovascular risk factors and disease with depression in later life. number needed to treat. BMC Med Res Methodol 2002; 2: 3.
Am J Geriatr Psychiatry 2007; 15: 506–13.
24 Thompson SG, Higgins JP. How should meta-regression analyses 38 Higgins JP, Whitehead A, Turner RM, Omar RZ, Thompson SG. be undertaken and interpreted? Stat Med 2002; 21: 1559–73.
Meta-analysis of continuous outcome data from individual patients. 25 Moher D, Schulz KF, Altman DG, for the CONSOR Group. The Stat Med 2001; 20: 2219–41.
CONSORT statement: revised recommendations for improving the 39 Thompson SG, Higgins JP. Can meta-analysis help target quality of reports of parallel-group randomised trials. Lancet 2001; interventions at individuals most likely to benefi t? Lancet 2005; 365:
357: 1191–94.
26 Gadde KM, Allison DB. Cannabinoid-1 receptor antagonist, 40 Astrup A, Greenway FL, Ling W, et al. Randomized controlled trials rimonabant, for management of obesity and related risks. of the D1/D5 antagonist ecopipam for weight loss in obese subjects. Circulation 2006; 114: 974–84.
Obesity 2007; 15: 1717–31.
27 Curioni C, Andre C. Rimonabant for overweight or obesity. Cochrane Database Syst Rev 2006; 4: CD006162.
www.thelancet.com Vol 370 November 17, 2007

Source: https://email.rutgers.edu/pipermail/sneeze_l/attachments/20091022/5ee48c22/Swedishmeta-analysisofdrugAcompia-harmfuleffects-0001.pdf