Study design
This study is a secondary analysis of a randomized, placebo-controlled trial. The trial is registered with EudraCT (2015–005585-32) and ClinicalTrials.gov (NCT04122716). The trial was conducted at Copenhagen University Hospital Hvidovre and the University of Copenhagen, Denmark. The study protocol and statistical analysis plan have been published and are also provided with this paper11,31. Results concerning the primary endpoint (change in body weight) and a secondary endpoint (change in body fat percentage) have been previously published11.
The trial protocol was approved by the Regional Ethics Committee for the Capital Region of Denmark (H-16027082) and the Danish Medicines Agency. The study was conducted in accordance with the Declaration of Helsinki and under the oversight of ICH Good Clinical Practice. All participants provided written informed consent before enrolment. Participants who completed the trial received compensation of 3,000 Danish kroner.
All participants initially followed an 8 week LCD (800 kcal d−1). During this phase, all meals were substituted with four formula meal replacement products daily, according to the Cambridge Weight Plan. On completion of the 8 week LCD, only participants who lost at least 5% of their initial body weight were eligible for enrolment and subsequently randomized in this study. Participants were randomly allocated to one of four 52 week interventions: placebo with usual activity, exercise with placebo, liraglutide with usual activity or a combination of exercise and liraglutide, in a 1:1:1:1 ratio to one of four treatment arms using a computer-generated randomization sequence provided by Novo Nordisk (subject randomization list; see study flow, Extended Data Fig. 1). The allocation sequence was stratified by sex and age, with separate randomization lists for participants aged ≥40 years and <40 years. Within each stratum, participants were assigned sequentially according to the randomization list. Allocation was implemented by a qualified, unblinded study nurse who was not otherwise involved in trial conduct. Study participants and study personnel were blinded to the study medication. Sex was defined as assigned at birth and was registered by a physician in an electronic case report form.
We also explored the main effects of the two interventions (exercise and liraglutide) on the investigated outcomes by testing each intervention independently against its respective control condition within the factorial design of the trial: exercise versus no exercise and liraglutide versus no liraglutide.
Study participants
Participants were adults with obesity (18–65 years of age, body mass index of 32–43 kg m−2). Major exclusion criteria were any known serious chronic illness, including type 1 or 2 diabetes. A full list of inclusion and exclusion criteria is available in the protocol paper31. Use of blood pressure and lipid-lowering medication, as well as prior and current tobacco use, is reported in Supplementary Table 5.
Exercise intervention
The exercise intervention was designed to align with WHO recommendations for physical activity, aiming for at least 150 min of moderate-intensity or 75 min of vigorous-intensity aerobic exercise each week, or an equivalent combination. To achieve this target, participants were encouraged to complete two supervised group sessions and two individual exercise sessions each week. Each 45 min group session included 30 min of vigorous, interval-based indoor cycling, targeting ≥80% of maximum heart rate, followed by 15 min of circuit training. The circuit training typically comprised three rounds of five exercises, each performed for 40 s with 20 s rest intervals, combining high-intensity aerobic and resistance exercises using body weight and/or external weights. Individual sessions were self-directed and involved moderate-to-vigorous aerobic activities such as cycling, running, brisk walking or circuit training, with intensity monitored through heart rate sensors. To enhance adherence, several support strategies were implemented: monthly or bimonthly in-person weight consultations to review exercise data from sports watches and address participation barriers; monthly feedback emails providing personalized summaries and motivational prompts; and individualized adherence plans for those not meeting goals, which could include increasing non-exercise physical activity, adjusting the balance of group and individual sessions or temporarily reducing frequency with a plan to increase it gradually. Full details of the intervention, in accordance with the Consensus on Exercise Reporting Template (CERT), are provided in the supplementary appendix, methods, section D of the primary trial report11.
Pharmacological intervention
Participants received either liraglutide or volume-matched placebo, each containing 3 ml of solution (6 mg ml−1). Injections were administered once daily into the abdomen or thigh, starting with a dose of 0.6 mg d−1 and increasing weekly by 0.6 mg until a maximum of 3.0 mg d−1 was reached. If participants experienced intolerance to a given dose, they continued with the maximum tolerated dose.
Outcomes
The present analyses were conducted as part of a predefined plan to investigate cardiometabolic risk modulation during exercise and GLP-1-based obesity therapy, using imaging and blood samples. The prespecified main outcome in this analysis was the change in cIMT11 in persons with obesity after 52 weeks of weight maintenance with exercise and liraglutide treatment. Supportive outcomes were changes during the weight maintenance period in markers of general inflammation (IL-6, IL-8, IL-10, TNF and IFNγ) and endothelial function (sICAM-1, sVCAM-1, vWF and tPA), which were prespecified objectives in the trial protocol11. Safety outcomes have previously been reported11. Trial data were collected and managed using the electronic case report form REDCap (Research Electronic Data Capture).
cIMT
cIMT provides a validated, reproducible and non-invasive measure of subclinical atherogenic burden that is particularly informative in asymptomatic, middle-aged individuals with obesity and without previously established atherosclerotic disease5. Assessments were performed by trained personnel and in accordance with a previously described method12, which provides a coefficient of variation of 5.1%. Examinations were performed using a Philips CX50 ultrasound system and a high-frequency linear vascular probe, Philips L12-3 (IP X-7) with an internal electrocardiography monitor. Carotid ultrasound images were acquired in a supine position with the person at rest. The images were recorded approximately 1 cm caudally from the right carotid bulb. The right common carotid artery cIMT was measured by two blinded assessors at three evenly spaced locations on the image frame of the far carotid artery wall obtained at the R wave of the accompanying electrocardiography recording and reported as the average of these values. Blinded analyses were performed using ultrasound quantification software (QLAB v.15, Philips Healthcare).
Biomarkers and blood pressure
Blood samples were collected after a minimum fasting period of 10 h and before the ultrasound examination of cIMT. The plasma concentration of vWF antigen (%) was determined by ELISA using rabbit anti-human vWF polyclonal IgG as capture and detection antibodies (DAKO, A0082)32. Results of vWF were expressed relative to a reference plasma (Biopool) calibrated against the WHO international standard for vWF. The concentration of tPA antigen (ng ml−1) was determined by an in-house ELISA using mouse anti-human tPA monoclonal IgG as capture (clone 15-4-21) and detection (clone 15-4-6) antibodies32. sICAM-1, sVCAM-1 and high-sensitivity CRP were analysed with a V-PLEX MSD MULTI-SPOT Assay System (Vascular Injury Panel 2 (human) Kits, K15198D). Pro-inflammatory cytokines were measured in duplicates using a V-PLEX MSD Proinflammatory Panel 1 (human), K15049D, which assesses ten cytokines. However, only IL-6, IL-8, IL-10, TNF and IFNγ displayed detection rates above 75%; therefore, only these five cytokines were included in the analyses, which is typical for this assay in a population without specific inflammatory diseases33. No measurements of the five cytokines reported had measurements below the detection limit; see Supplemental Table 6 for a table of the coefficient of variation and limits of detection and quantification of the measured biomarkers. Office blood pressure was measured twice in a seated position after at least 5 min of rest with an automated sphygmomanometer, and the average of the recordings is provided.
Statistical analysis and figures
Descriptive data are presented as mean and standard deviation if normally distributed or median with interquartile range if not normally distributed. Changes in outcomes were analysed using a linear mixed model for repeated measurements with time (factorial), treatment group (placebo, exercise, liraglutide or exercise + liraglutide), time × treatment interactions, age (<40 years or ≥40 years) and sex as fixed effects, as prespecified in the statistical analysis plan for the original trial11. For serial imaging, the baseline value (at randomization) was included as a covariate. Participants were stratified by age and sex; therefore, these variables were included as covariates in the model, as recommended by the European Medicines Agency. The model assumed an unstructured covariance and a repeated effect for time. Model fit was visually inspected, and if residuals were not normally distributed, analyses were performed on log-transformed data. For data presentation, log-transformed data were back-transformed and presented as percentage change with 95% confidence intervals. All analyses were performed using SAS Enterprise Guide 8.1. The analyses were performed in the per-protocol population (defined as participants who performed at least 75% of the exercise programme and/or had taken at least 2.4 or 3.0 mg d−1 of liraglutide or placebo for at least 75% of the intervention period). Data from the intention-to-treat population are shown in Supplementary Table 7 and generally support the main findings. All missing data were assumed to be missing at random and handled implicitly in the mixed model by maximum likelihood estimation. As an exploratory analysis, the 2 × 2 factorial design was used to examine the pooled effects of exercise and liraglutide and their potential interaction. In the original trial, it was estimated that 30 participants per group needed to be included to detect a 4 kg difference in weight between the intervention groups, with 80% power at a significance level of 5% (ref. 11). Significance testing was performed on the outcomes cIMT and biomarkers (IL-6, IL-8, IL-10, TNF, IFNγ, tPA, vWF, sICAM-1 and sVCAM-1). P < 0.05 was considered statistically significant. Figures were created in R (v.4.5.1) using the ggplot2 package (v.4.0.3).
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.