Does Rimonabant Independently Affect Free Fatty Acid and Glucose Metabolism?

Abstract

Context:

Endocannabinoid receptor 1 blockade is proposed to improve metabolic complications of obesity via central and peripheral effects.

Objective:

Our objective was to test whether rimonabant improves insulin regulation of free fatty acid and glucose metabolism after controlling for fat loss.

Design:

This was a double-blind, placebo-controlled substudy of the visceral fat reduction assessed by computed tomography scan on rimonabant (VICTORIA) trial.

Participants and Setting:

Sixty-seven abdominally obese, metabolic syndrome volunteers age 35–70 yr participated at academic medical center general clinical research centers.

Intervention:

Intervention included a 12-month lifestyle weight management program plus rimonabant 20 mg/d or placebo.

Main Outcome Measures:

Body composition and two-step euglycemic, hyperinsulinemic clamp before and after intervention were performed. Insulin sensitivity was assessed as insulin concentration needed to suppress by 50% palmitate concentration [IC_{50(palmitate)}], flux [IC_50(palmitatef], and hepatic glucose output [IC_50(HGO)] and as insulin-stimulated glucose disposal (Δ glucose disappearance per Δ insulin concentration − glucose slope).

Results:

Body fat decreased by 4.5 ± 2.9% (SD) in the rimonabant and 1.9 ± 4.5% in the placebo group (P < 0.005). The primary [improvement in IC_{50(palmitate)} and IC_{50(palmitate)}f] and secondary [improvement in IC_50(HGO) and glucose slope] outcomes were not significantly different between the rimonabant and placebo groups. Post hoc analyses revealed that 1) changes in body mass index (BMI) and IC_{50(palmitate)} were correlated (P = 0.005) in the rimonabant group; this relationship was not significantly different from placebo when controlling for greater BMI loss (P = 0.5); 2) insulin-regulated glucose disposal improved in both groups (P = 0.002) and correlated with changes in BMI.

Conclusions:

Improvements observed in insulin regulation of free fatty acid and glucose metabolism with rimonabant treatment in humans was not greater than that predicted by weight loss alone.

There are two endocannabinoid system receptor subclasses, of which cannabinoid 1 receptors (CB₁R) are important in regulating ingestive behavior. First identified within the central nervous system, CB₁R are found in numerous peripheral tissues, including adipocytes, muscle, and liver (1). Rimonabant was the first CB₁R antagonist developed for weight reduction therapy. It was briefly available for clinical use in Europe before being withdrawn over concerns regarding adverse psychiatric events (e.g. depression and suicidal ideation). Interest in CB₁R antagonists remains because data from in vitro and animal studies suggest they may have independent, beneficial effects on glucose and lipid metabolism in peripheral tissues as well as stimulating greater energy expenditure (1–8). Indirect support for this concept has come from large-scale clinical trials [Rimonabant in Obesity (RIO)-North America (9), RIO-Europe (10), RIO-Lipids (9, 11), and RIO-Diabetes (12)]. However, previous studies have not directly measured the impact of CB₁R antagonism on peripheral glucose and free fatty acid (FFA) in humans controlling for the effects of weight loss.

The goal of this study was to quantify the in vivo effects of rimonabant on glucose and fatty acid metabolism compared with placebo as a substudy of the visceral fat reduction assessed by computed tomography (CT) scan on rimonabant (VICTORIA) trial.

Subjects and Methods

Study aims and population

Main study group

The VICTORIA study was a double-blinded two-arm and placebo-controlled trial designed to measure the effect of rimonabant on visceral fat in abdominally obese adults with features of the metabolic syndrome as defined at the time of protocol development. There were 27 international study sites with data collection between February 2006 and July 2008 (inclusion and exclusion criteria are provided in Table 1). The primary aim was to compare changes in visceral fat and metabolic syndrome components in volunteers treated with lifestyle intervention (diet, activity, and behavior therapy) for 12 months and randomized to rimonabant 20 mg daily vs. placebo. Secondary aims of the main VICTORIA study were to assess changes in anthropometric measures, lipids and lipoprotein profiles, glycemia and insulinemia, and adipokine, inflammatory, and hemostatic markers. Participants were randomized for equal numbers of men and women in each group.

Table 1.

Inclusion and exclusion criteria

Inclusion criteria	Exclusion criteria
Waist circumference >102 cm men, >88 cm women	Weight management history of Significant weight loss or very-low-calorie diet (≤800 kcal/d) within 3 months of screening visit
Age 35–70 yr
At least two other components of the metabolic syndrome as defined by the NCEP/ATPIII classification:	Previous bariatric surgery
Triglyceridemia ≥150 mg/dl (or 1.69 mmol/liter)	Weight reduction medication
HDL cholesterol <50 mg/dl (1.29 mmol/liter) in women or <40 mg/dl (1.04 mmol/liter) in men	Presence of eating disorders
Blood pressure ≥130/85 mm Hg (systolic blood pressure ≥130 mm Hg and/or diastolic blood pressure ≥85 mm Hg) or treatment with antihypertensive agent(s) for this condition	Endocrine
Fasting blood glucose >110 mg/dl (or 6.1 mmol/liter)	Presence of diabetes
	Any clinically significant endocrine disorder
	Any organ dysfunction
	Lipid-lowering agents (HMG CoA reductase inhibitors allowed if dose stable for 3 months)
	Prolonged use (>10 d) of corticosteroids and neuroleptics
	Cancer diagnosis in the preceding 5 yr
	Psychological
	Any episode severe depression
	>2 episodes previous depression
	Use of antidepressants
	Pregnancy and breastfeeding

HMG CoA, 3-hydroxy-3-methyl-glutaryl coenzyme A; NCEP/ATP, National Cholesterol Education Program/Adult Treatment Panel.

Substudy

Four centers participated in a substudy with the primary aim to understand whether changes in the insulin regulation of FFA and glucose metabolism are different between those receiving rimonabant vs. placebo combined with a weight loss intervention. Those four centers enrolled 67 volunteers, one of whom did not complete the first insulin clamp study, leaving 66 volunteers with baseline data.

Lifestyle intervention, sample collection, and adverse event monitoring were the same as for the main study group. Participants in the substudy additionally underwent indirect calorimetry, two-step euglycemic-hyperinsulinemic clamp and tracer studies of glucose and FFA metabolism before and after 12 months of treatment. Only results from the substudy participants are reported herein.

Methods

Body composition

Height, weight, and waist circumference were recorded at each study visit. At baseline and 12 months, total and regional body fat was measured using dual-energy x-ray absorptiometry and visceral and abdominal sc fat areas (square centimeters) were measured from a single-slice abdominal CT at the L₄-L₅ interspace. Analysis of CT images was centralized and performed in conjunction with quality control images to ensure consistency.

Diet

Energy requirements were based on calculated basal metabolic rates and physical activity levels. A mild hypocaloric diet (600 kcal/d deficit) was prescribed, but keeping energy intake at 1200 kcal/d or higher for women and 1600 kcal/d or higher for men. Macronutrient distribution was recommended to be approximately 50% carbohydrate, 30% fat, and 20% protein with alcohol consumption below 15 g/d or below 85 g/wk. Participants also received instructions for physical activity, taking into account their medical condition (e.g. walking 30 min above the daily walking habits). Dietary and activity review were performed at each visit.

Blood samples

Blood was collected under overnight postabsorptive conditions for measurement of plasma glucose, insulin (mean of three samples), FFA-palmitate (mean of four samples), and other tests of metabolic and inflammatory parameters including hemoglobin A_1c (HbA_1c) lipid profile, and adipokines. Alcohol consumption within 48 h and intense physical activity within 24 h preceding blood sampling were discouraged. Plasma glucose concentrations were measured with a glucose analyzer (Beckman Instruments, Fullerton, CA), and plasma insulin by chemiluminescent sandwich assays (Sanofi Diagnostics, Chaska, MN). Plasma triglyceride concentrations were measured using microfluorometric assay (13), FFA-palmitate concentrations and enrichment by liquid chromatography/mass spectrometry (14), and glucose enrichment using gas chromatography/mass spectrometry (15).

Euglycemic-hyperinsulinemic clamp

A two-step euglycemic-hyperinsulinemic clamp study (Fig. 1) was used to establish 1) insulin regulation of glucose metabolism as estimated by the insulin concentration needed to suppress hepatic glucose output (HGO) by 50% [IC_50(HGO)] and glucose disposal slope [the change in glucose disposal in milligrams per kilogram fat-free mass (FFM) per minute per unit change in insulin concentration] and 2) FFA insulin sensitivity as estimated by the insulin concentration needed to suppress palmitate concentrations and flux by 50% [IC_{50(palmitate)} and IC_{50(palmitate)}f, respectively) and basal lipolysis. An iv catheter was placed for infusions of recombinant human insulin at 0.25 mU/kg FFM · min (step 1) and 1.0 mU/kg FFM · min (step 2), [U-¹³C]palmitate (Cambridge Isotope Laboratories, Andover, MA), and [6-²H₂]glucose (Isotec Inc., Miamisburg, OH) (Fig. 2). A second retrograde catheter was placed into a dorsal hand vein to collect arterialized blood using the heated box technique. Glucose disposal was calculated after reaching steady state using linear regression modeling.

Fig. 1.

Schematic of euglycemic-hyperinsulinemic clamp study. Isotopes administered were 1) [6-²H₂]glucose (constant infusion), 2) [U-¹³C]palmitate (infused over the last 60 min of each study interval), and 3) 20% dextrose labeled with [6-²H₂]glucose (to maintain euglycemia at 5 mmol/liter). Blood sampling for glucose enrichment and concentration (GEC) and palmitate enrichment and concentration (PEC) was performed over the last 30 min of each time period.

Fig. 2.

Correlations between changes in BMI and insulin sensitivity of glucose and fatty acid metabolism are shown. In the rimonabant group, there were significant correlations between changes in BMI and changes in insulin sensitivities of palmitate concentration (r = 0.61; P = 0.004) (A), hepatic glucose production (HGO) (r = 0.61; P = 0.003) (B), and glucose disposal (r = 0.45; P = 0.05) (C). Although the trends were not significant in the placebo group, the slopes of association were not different.

Indirect calorimetry

Indirect calorimetry was performed with a ventilated hood during the last 30 min of the basal period, and the low-dose and high-dose hyperinsulinemic periods and resting energy expenditure and respiratory exchange ratios (RER) were calculated.

Clinical safety

Patients were seen at months 1, 3, 5, 7, 9, and 12 for assessment of dietary compliance and dispensation of study medications and contacted by phone at months 2, 4, 6, 8, 10, and 11. There was monthly assessment of study drug compliance, concomitant medications, adverse events, and neuropsychiatric scripted questions. Physical examination and standard hematology and serum chemistry were performed at screening and months 3, 7, and 12 or at permanent premature study discontinuation.

Statistical analysis

Power calculations for the substudy were based upon the goal of detecting a difference between the rimonabant and placebo groups in the ability of insulin to enhance suppression of FFA after weight loss, which was the primary, prespecified end point of the substudy. We used previously published data (16) to estimate the mean and sd of palmitate concentrations and kinetics under basal and insulin-suppressed conditions in obese, nondiabetic, and insulin-resistant patients. We estimated that a 30% improvement in insulin suppression of palmitate would be a clinically significant improvement after active treatment. Given a sd of 36%, an α-level of 5% with a power of 80% and for a two-sided t test, the required sample size to detect a 30% difference between groups was 24 patients per arm. Within-group differences were compared using repeated-measures ANOVA. Palmitate and insulin relationships were analyzed using linear regression approaches after log transformation of both values to examine the dose-response relationships (17). Secondary, prespecified end points include between-group comparisons of changes in 1) the ability of insulin to stimulate glucose disposal (change in glucose slope) and suppress hepatic glucose production [change in IC_50(HGO)]; 2) markers of metabolic health (adiponectin and fibrinogen), serum triglycerides and total cholesterol, apolipoprotein B (Apo B) concentrations, inflammatory markers (high-sensitivity C-reactive protein, plasminogen activator inhibitor-1, and TNF-α). The remainder of the analyses we present (relationships between FFA and glucose variables and relationships between body composition variables and metabolic variables) represent post hoc analysis designed to explore associations that might relate to the heterogeneity in the data. Data are provided as mean ± sd, and statistical significance was analyzed using ANOVA and t tests for normally distributed data and Wilcoxon signed rank test for nonparametric data. Nonparametric data were log transformed if needed to allow for standard regression approaches. Correlation coefficients (r) are provided as indicators of relationship strength. Variate analyses and standard least-squares regression analyses were performed using JMP statistical software.

Results

VICTORIA study participant characteristics

The VICTORIA parent study included 254 participants. The demographic and baseline characteristics were similar in the placebo and rimonabant treatment arms, and the proportion of metabolic syndrome defining criteria was equal in the two groups.

VICTORIA substudy baseline subject characteristics (Table 2)

Table 2.

VICTORIA substudy baseline subject characteristics

Baseline characteristic	All baseline		Completed		P value between complete groups
	Placebo	Rimonabant	Placebo	Rimonabant
Number	35	31	27	21
Age (yr)	54 ± 7	55 ± 9	54 ± 8	56 ± 9	0.35
Male/female	13/22	12/20	8/17	8/12
BMI (kg/m2)	35.2 ± 4.1	33.5 ± 3.1	34.9 ± 4.2	32.8 ± 2.4	0.18
Total body weight (kg)	102.5 ± 18.6	98.2 ± 15.5	100.0 ± 17.7	97.6 ± 14.0	0.70
Waist circumference (cm)	114 ± 12	110 ± 9	113 ± 11	108 ± 8	0.45
Percent body fat	43 ± 8	44 ± 9	44 ± 2	43 ± 10	0.82
Total body fat (kg)	43.0 ± 8.7	41.2 ± 8.6	42.6 ± 1.7	40.6 ± 8.6	0.42
Visceral fat area (cm2)	193 (160–253)	184 (165–206)	191 (159–250)	184 (169–198)	0.67
sc fat area (cm2)	458 ± 104	415 ± 89	466 ± 104	407 ± 79	0.05
Deep sc fat area (cm2)	260 ± 83	227 ± 73	256 ± 78	226 ± 63	0.13
Fasting blood glucose (mg/dl)	107 ± 13	105 ± 12	104 ± 11	104 ± 13	0.93
Glucose slope (×1000) (mg · kg FFM−1 · min−1/pmol · liter−1)	7.4 (18.0–13.8)	12.0 (6.8–16.7)	8.5 (1.4–14.2)	13.2 (6.4–20.6)	0.26
IC50 hepatic glucose suppression (pmol/liter)	229 (144–257)	159 (121–208)	207 (134–263) (n = 26)	149 (121–197) (n = 21)	0.05
IC50(palmitate)f (pmol/liter)	183 (122–249)	129 (111–158)	183 (118–243) (n = 20)	140 (116–167) (n = 19)	0.05
IC50(palmitate) (pmol/liter)	152 (104–209)	113 (99–150)	148 (92–202) (n = 22)	115 (95–151) (n = 20)	0.13
Baseline RER	0.77 ± 0.06	0.78 ± 0.06	0.76 ± 0.06	0.78 ± 0.06	0.33
Low insulin rate RER	0.78 ± 0.05	0.79 ± 0.05	0.77 ± 0.05	0.79 ± 0.06	0.12
High insulin rate RER	0.82 ± 0.06	0.84 ± 0.07	0.82 ± 0.06	0.85 ± 0.07	0.16
BMR (kcal/24 h)	2747 ± 523	2616 ± 538	2715 ± 539	2591 ± 489	0.27
Adiponectin (μ g/ml)	5.13 (4.14–6.51)	6.54 (3.97–7.45)	5.43 (4.29–6.03)	6.5 (3.85–7.31)	0.69
Fibrinogen (g/liter)	3.0 ± 0.8	3.0 ± 0.7	2.9 ± 0.9	3.0 ± 0.6	0.60
Leptin (ng/ml)	32.1 (20.8–54.0)	29.4 (18.6–61.9)	36.9 (20.8–62.4)	29.5 (18.3–61.0)	0.76
Apo B (g/liter)	0.96 ± 0.29	0.97 ± 0.25	0.99 ± 0.27	0.99 ± 0.26	0.94
HDL cholesterol (mmol/liter)	1.09 (1.01–1.23)	1.13 (1.03–1.35)	1.14 (1.04–1.4)	1.17 (1.07–1.36)	0.24
LDL cholesterol (mmol/liter)	3.27 ± 0.90	3.56 ± 0.86	3.29 ± 0.87	3.62 ± 0.96	0.26
Total cholesterol (mmol/liter)	5.30 ± 1.06	5.75 ± 0.91	5.39 ± 1.00	5.85 ± 0.97	0.16
Triglycerides (mmol/liter)	1.83 (1.08–1.83)	1.97 (1.72–2.26)	1.89 (1.44–2.15)	2.00 (1.72–2.26)	0.20
HbA1c (%)	5.7 (5.5–5.9)	5.6 (5.4–5.8)	5.7 (5.5–5.9)	5.6 (5.3–5.8)	0.19

Data that are normally distributed are presented as mean ± sd, and data that are not normally distributed are presented as median (interquartile range). Data for all study completers n = 48 unless otherwise stated. HDL, High-density lipoprotein; LDL, low-density lipoprotein.

Sixty-seven participants were randomized, of which 48 reached study completion (27 placebo and 21 rimonabant). There were no statistically significant baseline differences between the two groups of study completers for any of the parameters.

For the entire group studied before randomization, we found the expected relationships between baseline IC_{50(palmitate)}f and palmitate concentration insulin sensitivity [IC_{50(palmitate)}] (r = 0.74; P < 0.001). Likewise, baseline indices of insulin-regulated FFA and glucose metabolism were interrelated: IC_{50(palmitate)}f and IC_50(HGO) were well correlated (r = 0.66; P < 0.001), as were IC_{50(palmitate)}f and the insulin/glucose disposal slope (r = −0.40; P < 0.001). IC_{50(palmitate)} was also correlated with IC_50(HGO) (r = 0.84; P < 0.001) and glucose slope (r = −0.53; P < 0.001).

Deep sc fat area was correlated with IC_{50(palmitate)}f (r = 0.48; P < 0.001) and IC_{50(palmitate)} (r = 0.35; P = 0.003). Visceral fat area was correlated with glucose disposal (r = −0.34; P = 0.005), IC_50(HGO) (r = 0.37; P = 0.002), and IC_{50(palmitate)} (r = 0.33; P = 0.007). Of note in making inferences for these relationships, deep sc and visceral fat areas in these participant were correlated with each other (r = 0.26; P = 0.03), but total sc and visceral fat areas were not significantly correlated.

Basal metabolic rate (BMR) was strongly correlated with FFM (r = 0.77; P < 0.001), suggesting to us that the dual-energy x-ray absorptiometry instruments and indirect calorimeters from each site provided comparable data.

Response to interventions (Table 2)

Body composition changes

Waist circumference and body mass index (BMI) decreased significantly in both groups at 1 yr, with greater reductions in the rimonabant group than the placebo group (Table 3). Reductions in abdominal deep sc adipose tissue and visceral fat compartments by CT were significant and correlated with overall weight loss, but not treatment type.

Table 3.

Comparison of change from baseline data for placebo and rimonabant groups

Parameter	n	Placebo group mean change ± sd	Rimonabant group mean change ± sd	P value between group mean changes
Body composition parameters
Weight change (kg)	48	−7.4 ± 9.4a	−11.0 ± 4.4a	<0.05
Waist circumference change (cm)	47	−9 ± 8a	−11 ± 7a	0.18
BMI change (kg/m2)	48	−2.5 ± 3.0a	−3.8 ± 1.6a	<0.05
Fat mass change (kg)	48	−4.6 ± 7.4b	−8.3 ± 3.4a	<0.01
% body fat change	48	−1.9 ± 4.5d	−4.5 ± 2.9a	<0.005
FFM change (kg)	48	−3.0 ± 3.8a	−2.3 ± 2.3b	0.51
Visceral fat area change (cm2)	47	−42 ± 52a	−53 ± 30a	0.11
sc fat area change (cm2)	42	−51 ± 66a	−82 ± 52a	<0.05
Deep sc fat area change (cm2)	46	−20 ± 49d	−40 ± 28a	<0.05
FFA and glucose metabolism
IC50(palmitate)f change (pmol/liter)	39	−29 ± 73b	−69 ± 75c	0.21
Glucose slope change (×1000) (mg · kg FFM−1 · min−1/pmol · liter−1)	48	6 ± 12b	9 ± 9a	0.10
IC50(HGO) change (pmol/liter)	47	−4 ± 67	16 ± 89	0.44
IC50(palmitate) change (pmol/liter)	42	−40 ± 68b	−38 ± 23a	0.94
BMR (kcal/24 h)	48	−225 ± 354	−193 ± 266b	0.38
Low insulin rate RER	47	0.045 ± 0.28	0.002 ± 0.11	0.93
High insulin rate RER	48	0.003 ± 0.09	−0.035 ± 0.11	0.23
Serum markers
Adiponectin (μg/ml)	47	3.13 ± 2.83a	3.71 ± 3.05a	0.64
Fibrinogen (g/liter)	47	0.7 ± 0.9a	0.9 ± 0.7a	0.77
Leptin (ng/ml)	47	−13.5 ± 15.0a	−14.0 ± 13.1a	0.57
Apo B (g/liter)	47	−0.1 ± 0.2	0.1 ± 0.3	<0.05
HDL cholesterol (mmol/liter)	47	−0.05 ± 0.20	−0.04 ± 0.2	0.74
LDL cholesterol (mmol/liter)	47	−0.5 ± 0.60a	−0.2 ± 1.0	0.16
Total cholesterol (mmol/liter)	47	−0.8 ± 0.80a	−0.6 ± 1.0d	0.53
Triglycerides (mmol/liter)	47	−0.4 ± 0.5a	−0.7 ± 0.8a	0.22
HbA1c (%)	43	−0.2 ± 0.3b	−0.2 ± 0.2a	0.71

Significance of difference between baseline and end changes within each group is indicated after sd. HDL, High-density lipoprotein; LDL, low-density lipoprotein.

^aP < 0.001.

^bP < 0.005.

^cP < 0.01.

^dP < 0.05.

Glucose metabolism [glucose slope and IC_50(HGO)]

Although in the nondiabetic range, HbA_1c (n = 43) decreased significantly in both placebo and rimonabant groups (Table 3). One of the prespecified secondary outcome variables, insulin-regulated glucose disposal, as measured by the glucose slope, improved in both groups (Table 3). This improvement was greater, but not statistically significantly so, in the rimonabant group. Because of the heterogeneity in the glucose slope responses to the interventions, we examined post hoc whether the inter-individual changes correlated with changes in weight or body fat compartments. Changes in BMI correlated with changes in glucose slope in the rimonabant group (r = 0.04; P = 0.05) but not placebo group (P = 0.32), whereas we found no significant correlations between changes in glucose slope and changes in CT measured visceral or sc fat areas in either group.

There were no statistically significant changes in IC_50(HGO) in either group; however, we also note significant heterogeneity in this response and also tested post hoc whether intra-individual changes in IC_50(HGO) were related to variations in weight or regional fat loss. The changes in IC_50(HGO) were correlated with changes in BMI (r = 0.61; P = 0.003) in the rimonabant group, and although there was a similar trend in the placebo group, the relationship was not statistically significant. There were no statistically significant correlations between IC_50(HGO) change and visceral fat area change in either group.

FFA metabolism [IC_{50(palmitate)}f and IC_{50(palmitate)}]

There were significant decreases (improvements) in the primary, prespecified outcome variables IC_{50(palmitate)}f and IC_{50(palmitate)} in both groups (Table 3). Post hoc analysis revealed that IC_{50(palmitate)} change was correlated with BMI change (r = 0.60; P = 0.004) in the rimonabant group. Although this correlation was not statistically significant in the placebo group, both study arms followed the same trend, and regression analysis did not reveal a significant effect of treatment arm on the relationship.

Glucose/FFA interrelationships

The changes in glucose slope were correlated with changes in insulin regulation of FFA in the placebo group (P = 0.03) but not in the rimonabant group (P = 0.8). In the placebo group, changes in IC_50(HGO) were correlated with changes in IC_{50(palmitate)} (r = 0.60; P < 0.001) and IC_{50(palmitate)}f (r = 0.50; P = 0.008). The same trend was observed in the rimonabant group, where the change in IC_50(HGO) correlated with the IC_{50(palmitate)} change (r = 0.65; P = 0.001).

Energy metabolism

BMR decreased significantly within each group (P < 0.005), but the change was not significantly different between groups. After treatment, fasting RER did not change from baseline, pretreatment values.

Serum marker and lipid data

Both groups showed improvements in markers of metabolic health (adiponectin and fibrinogen, P < 0.001) and serum triglycerides and total cholesterol. Rimonabant use was associated with significant improvements in Apo B concentrations, which were statistically accounted for by the greater weight loss (Table 3). The inflammatory markers high-sensitivity C-reactive protein, plasminogen activator inhibitor-1, and TNF-α did not change significantly from baseline in either group (data not shown).

Adverse events and study dropouts

Adverse events were considered for all randomized substudy participants (placebo n = 35; rimonabant n = 31). Adverse events directly related to the study were recognized in 57% of the placebo arm (n = 20) and 74% of the rimonabant arm (n = 23). Five participants (16%) withdrew from the rimonabant group within 2 months after psychological events.

Psychological problems were the most common, affecting 57% of the placebo group (n = 20) and 61% of the rimonabant group (n = 19). Symptoms most commonly described were depression (placebo 31%, n = 11; rimonabant 35%, n = 11), anxiety disorder (placebo 17%, n = 6; rimonabant 29%, n = 9), loss of energy or concentration (placebo 11%, n = 4; rimonabant 16%, n = 5), and sleep disorder (placebo 29%, n = 10; rimonabant 26%, n = 8). Gastrointestinal side effects were also commonly reported, affecting 14% of the placebo group (n = 5) and 39% of the rimonabant group (n = 12). Reasons for noninclusion in the final data analysis included study dropout before completion resulting from adverse events and insufficient data collection (as indicated in Table 2).

Discussion

Clarifying whether CB₁R antagonism has major peripheral actions has important clinical implications should agents that do not penetrate the central nervous system become available. Animal models support a role of CB₁R on peripheral metabolism and energy expenditure, whereas it has been inferred from clinical studies that up to half of the beneficial lipid (RIO-Lipids) (11) and glycemic improvements (RIO-Diabetes) (12) seen with rimonabant treatment are independent of weight loss. This is the first clinical study to directly assess the effects of weight loss generated by rimonabant treatment vs. placebo on the insulin regulation of adipose tissue lipolysis and glucose metabolism. Although the rimonabant treatment group lost significantly more weight (11 ± 4%) than their placebo counterparts (7 ± 9%), improvements in glucose disposal, hepatic glucose production, and fatty acid metabolism were not significantly different between groups even without accounting for the differences in weight loss. Regression analysis, controlling for the amount of weight loss, indicated that improvements in metabolic outcomes were not different between the two groups. This suggests that any direct peripheral effects of rimonabant are minimal compared with the indirect benefits resulting from weight loss.

In these volunteers studied before and after weight loss, the improvements in insulin sensitivity with respect to glucose metabolism were strongly related to those seen in fatty acid metabolism. Given that elevations in FFA are known to influence insulin action on peripheral glucose uptake and hepatic glucose production (18), the relationships we observed are perhaps not surprising. Weight loss has also been linked with lower FFA concentrations and greater glucose disposal rates in people with type 2 diabetes (19). We observed that although improvements in both glucose and fatty acid metabolism were clearly associated with reductions in BMI, the improvements were not statistically related to changes in visceral or deep sc fat.

Given that our data do not support a detectable, independent metabolic action of rimonabant yet a number of reports seem to support an effect of peripheral CB₁R antagonism in vitro and in vivo, it is useful to review possible explanations for these inconsistencies. Interest in peripheral metabolic and energy effects of CB₁R antagonists arose from observations that CB₁R antagonism in mice generated weight reductions not entirely explained by centrally mediated effects. These findings included continued weight loss despite only transient early reductions in food intake (20), a lack of parity in weight loss with pair-feeding controls (21–23), and weight loss in response to peripheral but not central administration (24). Furthermore, CB₁R knockout mice appeared protected from weight gain despite high energy intake without increased locomotor activity (25) and had greater insulin sensitivity than wild-type littermates (22). CB₁R are present in adipose tissue (25), liver (26), skeletal muscle (27), and pancreatic islets (28), providing a rational mechanism for a peripheral functional role.

Cannabinoid action in adipocytes has been related to lipoprotein lipase activity (25), mitochondrial number and function (23), oxygen consumption (4, 6), and possibly conversion of white to brown adipose tissue (29). CB₁R action in skeletal muscle has been linked with altered glucose uptake (6) and adiponectin-1 receptor gene expression (30). Islet cell CB₁R has been suggested to play a role in basal insulin secretion (28). Finally, improvements in fatty liver, dyslipidemia, and glucose homeostasis have been documented with peripheral CB₁R antagonist action (1, 8, 31, 32). Despite these findings, the role of peripheral CB₁R is not clear-cut. Similar weight loss and glucose and fatty acid metabolism changes have been observed in animals receiving CB₁R antagonists and pair-fed controls (5, 25, 31, 33). Furthermore, although CB₁R manipulation has been linked with adipose energy storage and expenditure (5), direct measurements of FFA, glycerol, and β-hydroxybutyrate did not differ between treatment and control animals despite weight loss differences (34). We also found modest reductions in basal metabolic rate with weight loss were not different between rimonabant and placebo treatment arms. This suggests rimonabant has no independent effect on resting energy expenditure.

It is possible that the metabolic changes observed in response to CB₁R antagonism in animals are a response to anxiogenic effects. Rodents treated with CB₁R antagonists have been shown to have higher anxiety scores (8) and urinary catecholamines (34). If this was associated with increase in activity, it would explain the improvements in fatty acid and glucose metabolism (35). In human studies, significant anxiogenic side effects lead to treatment withdrawal, precluding a good test of this possibility in our study. The differences observed between our study and animal data may also reflect lower clinically achievable peripheral tissue concentrations in humans. In rodents, rimonabant doses are typically 10 mg/kg · d but may be as great as 30 mg/kg · d, which are equivalent to 55–160 mg human daily doses, respectively (U.S. Food and Drug Administration drug dose conversion calculator http://www.accessdata.fda.gov/scripts/cder/onctools/animalquery.cfm. Doses above 40 mg/d were not tolerated in clinical trials, leading to a final treatment dose of 20 mg/d. Finally, it is possible that physiological differences exist between humans and these rodent models, with human peripheral CB₁R playing a negligible role in the regulation of glucose and fatty acid metabolism despite potentially adequate receptor antagonism.

In this study, it was estimated that 24 complete studies per group would provide statistical power to detect a 30% improvement in insulin suppression of FFA flux in the rimonabant arm compared with placebo. After controlling statistically for the greater weight loss induced by rimonabant, we do not find evidence for an independent, peripheral effect on energy, glucose, or lipid metabolism. Our data indicate that the centrally mediated effects on satiety were the key mechanism of weight reduction and metabolic benefits in the clinical setting.

In summary, we measured insulin regulation of glucose and FFA metabolism in rimonabant- and placebo-treated patients with metabolic syndrome undergoing a 1-yr weight loss treatment program. Weight loss was associated with improvements in insulin action in both groups, yet despite significantly greater weight loss with rimonabant, the improvements were comparable between groups. This suggests any peripheral effects of CB₁R antagonism with rimonabant are of minor importance compared with the central effects that induce weight loss.