Systematic review and meta-analysis on the adverse events of rimonabant treatment: Considerations for its potential use in hepatology

Background The cannabinoid-1 receptor blockers have been proposed in the management of obesity and obesity-related liver diseases (fatty liver as NAFLD or NASH). Due to increasing number of patients to be potentially treated and the need to assess the advantage of this treatment in terms of risk/benefit, we analyze the side events reported during the treatment with rimonabant by a systematic review and meta-analysis of all randomized controlled studies. Methods All published randomized controlled trials using rimonabant versus placebo in adult subjects were retrieved. Relative risks (RR) with 95% confidence interval for relevant adverse events and number needed to harm was calculated. Results Nine trials (n = 9635) were considered. Rimonabant 20 mg was associated with an increased risk of adverse event (RR 1.35; 95%CI 1.17-1.56), increased discontinuation rate (RR 1.79; 95%CI 1.35-2.38), psychiatric (RR 2.35; 95%CI 1.66-3.34), and nervous system adverse events (RR 2.35; 95%CI 1.49-3.70). The number needed to harm for psychiatric adverse events is 30. Conclusion Rimonabant is associated with an increased risk of adverse events. Despite of an increasing interest for its use on fatty liver, the security profile and efficacy it is needs to be carefully assessed before its recommendation. At present the use of rimonabant on fatty liver cannot be recommended.


Background
As the consequence of important biologic and social modifications, in the last century new non-infectious epidemic diseases appeared as the most important causes of mortal-ity in worldwide. Among these new epidemic diseases, obesity gained a leader position since the pathologic accumulation of adipose tissue has important deleterious effect on life expectancy [1,2]. The metabolic syndrome, of which obesity is the most common cause, is considered the most important risk factor for cardiovascular diseases and other chronic diseases [3].
The algorithm for the management of obesity includes non-pharmacologic, pharmacologic and surgery-based strategies [4]. Several drugs are recommended in the pharmacological approach and in the last years the cannabinoid-1 receptor blocker rimonabant has been proposed as a potential effective therapeutic approach in the management of obesity [5].
Non-alcoholic fatty liver disease (NAFLD), the hepatic manifestation of the metabolic syndrome [6,7] has been shown to be present in more that 70% of the obese subjects [8]. There is a consensus that the key mechanism of hepatic steatosis is the insulin resistance [9]. This observation prompted to investigate new drugs which may be useful in reducing the fatty content in the liver, the cornerstone in the treatment of NAFLD [10]. Although effective, dietary modifications and increased physical activity are associated with a low rate of compliance in the real life since recurrence is almost the rule [11,12]. This explains the increased interest shown by both clinicians and industry to find pharmacological approaches. Based on the favorable results of randomized controlled trials assessing rimonabant in the management of obesity, there is an increasing interest to assess the efficacy of this drug in NAFLD. However data of ADAGIO-Lipids trial showed in NAFLD patients a significant improvement of liver function test and liver fat [13], the histological evidence was not evaluated and the two randomized controlled trials on NAFLD [14,15] were stopped prematurely due to request of national health authorities. Small, uncontrolled observations suggested that rimonabant may be an effective strategy in fatty liver [16] in spite of some adverse events [17]. The adverse events have been so far analyzed by meta-analysis of data from Rimonabant In Obesity (RIO) trials [18,19]. In this review we extend the study on the side events reported during the treatment with rimonabant by a systematic review and meta-analysis of all randomized controlled studies using this cannabinoid-1 receptor blocker in any pathological condition.

Criteria for considering studies for this review
We considered all published randomized controlled trials using rimonabant (any dosage) versus placebo for at least 12 weeks in adults for any clinical indication, regardless of the language of publication. The outcomes were the adverse events reported classified as: 1) any adverse event; 2) serious adverse events; 3) discontinuation due to adverse event; 4) discontinuation due psychiatric adverse event; and 5) discontinuation due to neurologic adverse event. Search terms were ("rimonabant" OR "Acomplia" OR ["antagonist" AND "cannabinoid" AND "receptor"]) AND ("randomized controlled trial" OR "random" OR "blinded" OR "controlled") AND "placebo". The references of all identified studies were inspected for more trials. We also checked for any missing trials by examining the references of existing reviews of this topic.

Search methods for identification of studies
Two authors (NCT, FTA) independently reviewed the search output of potentially relevant trials for inclusion and assessed the trial for inclusion. Studies not meeting the inclusion criteria were excluded. Disagreements were settled by discussion with a third co-author (CT). In an independent manner we assessed the bias risk by the following components of methodological quality of included studies: generation of allocation sequence, allocation concealment, blinding, incomplete outcome data and selective outcome reporting [20].

Statistical analysis
We use the software package RevMan 5 [21]. For dichotomous variables, we calculate the relative risk (RR) with 95% confidence interval. A fixed effect model was used throughout the review, except in the event of significant heterogeneity between the trials (P < 0.10), when the random effect model was chosen. Heterogeneity was explored by chi-squared test with significance set at Pvalue 0.10, and the quantity of heterogeneity was measured by I 2 [22]. We use a funnel plot to explore bias. Asymmetry in funnel plot of trial size against treatment effect was used to assess this bias. We perform linear regression approach described by Egger et al. to determine the funnel plot asymmetry [23]. Number needed to harm was calculated.
The characteristics and quality of studies included are reported in Table 1. All trials had a low risk of bias. Five trials were designed to assess the efficacy on weight reduction, one for the percent change of atheroma volume, one for the management of alcohol dependency, one to assess the changes in HbA 1c ; one for hyperandrogenaemia; one for change in lipid profile, and one was designed to assess insulin resistance in patients with polycystic ovary syndrome. The last one was the excluded study since the adverse events were not reported. The longest follow-up was 24 months with a range between 3 and 24 months. A pooled 2849 subjects were included in the placebo arm, 2883 in rimonabant 5 mg arm, and 3903 were assigned to rimonabant 20 mg. In the serious adverse events, discontinuations for adverse events and psychiatric adverse events analysis biases were observed in the funnel plot graph (Additional file 1).

Adverse events
The use of rimonabant at dosage of 5 mg was not associated with any adverse event. Conversely, the use of 20 mg was associated with an increased risk for adverse event vs. placebo (RR 1.35; 95%CI 1.17-1.56), with a pooled effect of increased risk (RR 1.21; 95%CI 1.09-1.34) (Figure 2). By assessing only serious adverse events, a trend for increased risk in the rimonabant groups was observed at both dosages (5 and 20 mg), although the difference did not reach statistical significance (RR 1.16; 95%CI 0.93-1.45) (Additional file 2). In the subjects receiving 5 mg rimonabant, no difference was found in the discontinuation rate for adverse events while in those receiving 20 mg rimonabant an association between drug use and discontinuation rate (RR 1.79; 95%CI 1.35-2.38) was observed. The same was found when data obtained in pooled series (5 and 20 mg of rimonabant) were considered (RR 1.47; 95%CI 1.15-1.87 for) ( Figure 3).
By performing a sub-analysis of the psychiatric adverse events forcing treatment discontinuation we observed that only rimonabant at dosage of 20 mg was associated with an increased risk of psychiatric disorders (RR 2.35; 95%CI 1.66-3.34), a difference which remains after pooling both dosages (RR 1.79; 95%CI 1.24-2.58) (Figure 4). Similarly, 20 mg rimonabant and pooled data of the two dosages were associated with increased risk for treatment discontinuation due nervous system adverse events (RR 2.35; 95%CI 1.49-3.70 and, RR 1.89; 95%CI 1.30-2.75, respectively) ( Figure 5).
The analysis of the number needed to harm (NNH) demonstrates a reduced NNH for the occurrence of any adverse event (NNH 22, for rimonabant 20 mg), and dis-Flow chart for search strategy and selection of trials  Forest plot graphic assessing the risk to be free of any adverse events at rimonabant dosages of 5 and 20 mg Figure 2 Forest plot graphic assessing the risk to be free of any adverse events at rimonabant dosages of 5 and 20 mg.

Relative risk Relative risk
Forest plot graphic assessing the risk to be free of discontinuation due to adverse events at rimonabant dosages of 5 and 20 mg Figure 3 Forest plot graphic assessing the risk to be free of discontinuation due to adverse events at rimonabant dosages of 5 and 20 mg.. continuation due to adverse events (NNH 18, for rimonabant 20 mg). It was also found that for every 30 patients treated with 20 mg rimonabant, one will discontinue the treatment due to psychiatric adverse events (Table 2).

Discussion
In this study we performed a meta-analysis on the adverse events related with the treatment with rimonabant, a cannabinoid-1 receptor blocker, used primarily in obesity and related disorders. We observed that the use of rimonabant at the dose of 20 mg per day is associated with adverse events including discontinuation due to psychiatric and neurologic adverse events. These results are similar to those reported in previous meta-analysis on the Rimonabant in Obesity (RIO) trials [33]. However our study included more studies as a trial for alcohol dependence which enrolled subjects without obesity but with greater risk to present adverse events (particularly psychiatric adverse events).
The NNH to discontinuation due to adverse in our study were similar to other meta-analysis (NNH 14). Of notice that this NNH is the lowest as compared to other drugs for obesity treatment (NNH 39, for orlistat; and NNH 500, for sibutramine). Given the large number of patients eligi-ble for the treatment with rimonabant, the NNH is a concern [33].
The results of this meta-analysis also suggest that subjects exposed to 20 mg rimonabant may have some clinically relevant adverse events, and point to the need that both patients and investigators should be alerted about the early detection of these negative events. Since there is no evidence on the clinical utility and due to the proved evidence of increased adverse events rate, rimonabant cannot be recommended as a treatment option in NAFLD. This is further supported by the finding that despite possible reduction in the hepatic fat content [13], no evidence regarding histological improvement exists also because of the premature stop of rimonabant registered trials in NALFD [14,15], and other diseases (ClinicalTrials.gov registry: NCT00547118, NCT00678483, NCT00754689, NCT00434096, NCT00412698, NCT00449605, NCT00263042, NCT00690456, NCT00228176, NCT00478972, NCT00478595, NCT00405808, NCT00458081, NCT00325650, NCT00408148).
Nowadays solid information shows that rather than a totally benign disease, NAFLD may be a cause of chronic liver disease with a potential risk to develop end-stage Forest plot graphic assessing the risk to be free of treatment discontinuation due to psychiatric adverse events at rimonabant dosages of 5 and 20 mg Figure 4 Forest plot graphic assessing the risk to be free of treatment discontinuation due to psychiatric adverse events at rimonabant dosages of 5 and 20 mg.
Forest plot graphic assessing the risk to be free of treatment discontinuation due to neurologic adverse events at rimonabant dosages of 5 and 20 mg Figure 5 Forest plot graphic assessing the risk to be free of treatment discontinuation due to neurologic adverse events at rimonabant dosages of 5 and 20 mg.

Conclusion
In conclusion the use of rimonabant at 20 mg per day is associated with clinical adverse events. At present no indication on the use of this drug on NAFLD exists and therefore it cannot be suggested.