Liraglutide pharmacokinetics and exposure‐response in adolescents with obesity

Summary Background Obesity in adolescence presents a major public health challenge, often leading to obesity in adulthood with associated chronic disease. Objectives This study aimed to perform a population pharmacokinetic and exposure‐response analysis of liraglutide by meta‐analysis of data from trials conducted in children, adolescents and adults with obesity. Methods The population pharmacokinetic analysis investigated the effect of covariates body weight, age group (children, adolescents and adults) and sex on liraglutide exposure in adolescents compared with previous results in adults. The exposure‐response relationship of liraglutide for the change from baseline in body mass index standard deviation score (BMI SDS) was evaluated in adolescents and compared to that in adults. Results Body weight was the main covariate affecting liraglutide exposure, with lower exposures at higher body weights, whereas age group was of no importance and sex was of little importance. An exposure‐response relationship was demonstrated for liraglutide in both adolescents and adults as the decrease in BMI SDS from baseline increased in an exposure‐dependent manner with increasing liraglutide exposure. Conclusions The population pharmacokinetic analysis supported similar liraglutide exposures in adolescents and adults; body weight was the most important covariate affecting exposure. An exposure‐response relationship was established for liraglutide.


| INTRODUCTION
The prevalence of obesity over the past 30 years has more than doubled in children and tripled in adolescents, reaching epidemic proportions in the United States. 1,2 While the overall prevalence of obesity is lower in children than in adults, rates of increase in childhood obesity are now higher than those seen in adults in many countries. 3 Childhood obesity presents a major public health challenge since obesity in childhood and adolescence often leads to obesity in adulthood. [4][5][6] Paediatric obesity is associated with a wide range of chronic diseases, including type 2 diabetes, hypertension, hyperlipidaemia, polycystic ovary syndrome and sleep apnea. [7][8][9][10] Prolonged obesity continuing into adulthood may lead to complications such as cardiovascular disease, 11,12 which was responsible for more than twothirds of adult deaths associated with high body mass index (BMI) worldwide in 2015. 3 Childhood and adolescent obesity may also adversely affect quality of life, resulting in adverse psychosocial problems such as low self-esteem, depression and reduced educational achievement. 4,8 The economic consequences of obesity and its related complications are also high. 13,14 Current treatments for children and adolescents with obesity tend to favour lifestyle modifications that target diet and exercise; however, such programs often have limited effect on reducing BMI in adolescents because weight loss and maintenance of weight loss are hard to achieve. 15,16 Bariatric surgery may be an effective alternative for individuals with morbid obesity, 16 Glucagon-like peptide-1 (GLP-1), an incretin hormone secreted by intestinal L cells in response to food intake, 18 stimulates insulin secretion and inhibits glucagon secretion in a glucose-dependent manner. 18,19 Liraglutide is a GLP-1 receptor agonist that promotes weight loss through reduced appetite and a subsequent reduction in energy intake. 20 As an adjunct to a reduced-calorie diet and increased physical activity, liraglutide 3.0 mg (Saxenda) once daily is approved in the United States, European Union and elsewhere (65 countries) for chronic weight management in adults with obesity (BMI ≥30 kg/m 2 ), or overweight (≥27 to <30 km/m 2 ) in the presence of at least one weight-related comorbidity.
The efficacy and safety of liraglutide in adults were evaluated in the Satiety and Clinical Adiposity -Liraglutide Evidence (SCALE) program. The safety, tolerability, pharmacokinetics and pharmacodynamics of liraglutide in adolescents aged 12-17 years 21 and in children aged 7-11 years 22 have previously been investigated in short-term (<10 weeks) randomized, controlled trials. In phase 3 randomized, controlled 56-week trial investigating the effects of liraglutide for weight management in pubertal adolescents with obesity, liraglutide, in addition to lifestyle therapy, significantly reduced the BMI standard-deviation score (SDS) compared with placebo (estimated treatment difference: À0.22; 95% confidence interval [CI], À0.37, À0.08; P = 0.002). 23 A previous population pharmacokinetic analysis performed with liraglutide using adult data from the 56-week phase 3a SCALE Obesity and Prediabetes trial (1839) 24 and SCALE Diabetes trial (1922) 25 trials found that sex and body weight were the only intrinsic factors that influenced the exposure (ie, the average steady-state plasma concentration) of liraglutide 3.0 mg. 26 Moreover, in a previous exposureresponse analysis using data from the aforementioned SCALE trials and a phase 2, 20-week trial (1807), 27 greater weight loss was observed with higher liraglutide exposures. 28 This report aimed to perform population pharmacokinetic and exposure-response analyses of data from the recent 56-week trial in adolescents (trial 4180) in a meta-analysis including data from previous trials conducted in children, adolescents and adults with obesity, to support the 3.0 mg liraglutide dose for weight management in an adolescent population. The population pharmacokinetic analysis will provide additional information including adolescent data from a long-term (52-week) phase 3 trial, to support previous findings. The exposure-response analysis is the first to evaluate phase 3 data from both adults and adolescents together.

| Population pharmacokinetic analysis
The objectives of the population pharmacokinetic analysis were (1) to investigate body weight, age group (children aged 7-11 years, adolescents aged 12-17 years and adults aged ≥18 years) and sex as covariates to determine whether their impact on liraglutide exposure was in accordance with previous results in adults 26 and (2)  and 3630 (adults) were included in the pharmacokinetic assessment of liraglutide (Table 1). In trial 4180, Tanner staging (stages 2-5) of adolescents was assessed by site staff trained in pubertal assessments.
For females, assessments of breast and pubic hair development were made. For males, assessments of testicular volume (by orchidometer), penis development and pubic hair development were made. Prepubertal patients were not permitted in the trial. In all trials, liraglutide concentration in plasma was determined using a validated enzyme-linked immuno-sorbent assay (ELISA) with a lower limit of quantification (LLOQ) of 0.03 nmol/L. 29 The ELISA was a sandwich immunoassay with two monoclonal antibodies directed against different epitopes on liraglutide and was according to guidance regarding recovery, accuracy, precision, sensitivity and stability.
A standard one-compartment model with first-order absorption and elimination was the starting point for the description of liraglutide pharmacokinetics, and the model was developed and validated according to the US FDA and European Medicines Agency guidelines. 30,31 The structural model was parameterized in terms of the following parameters: • k a (absorption rate constant) • CL/F (apparent clearance) The first-order conditional estimation with interaction (FOCE-I) method was used for the population pharmacokinetic analysis, implemented in the non-linear mixed effects modelling (NONMEM) software. Due to pharmacokinetic sampling in steady-state conditions, exposure levels were expected to be much higher than the LLOQ (mean values 500-to 1000-fold above LLOQ). Observed values close to or below the LLOQ were therefore believed to be due to missed doses. Data records with missing concentration values or concentration values below the LLOQ were excluded from the population pharmacokinetic analysis in order not to overestimate CL/F. Stricter exclusion criteria were tested in sensitivity analyses but were found not to be relevant.  An analysis of the influence of covariates on exposure was carried out including all tested covariates in one step, using a confirmatory approach, 32  The covariate effects of baseline body weight, sex and age group were investigated for CL/F. Due to the sparseness of data, as well as the focus on average exposure, only the effect of baseline body weight was investigated with respect to V/F. The CL/F and V/F were parameterized as follows for the i th subject (shown for CL/F only): where TVCL and TVV are "typical values (TV)" of apparent clearance (CL/F) and volume of distribution (V/F), respectively, for a reference subject (an adult female with a body weight of 100 kg) and the θ values are the estimated covariate effect parameters.
Specific steady-state exposures for individual participants were based on the full model, including all covariates, and derived from the subject-specific post-hoc CL/F estimates, the maintenance dose and the dose interval (24 hours): The CIs for full model parameters were estimated using bootstrapping. For each investigated covariate, differences were considered relevant if the 90% CI of the estimated mean of the relative exposure fell outside the standard bioequivalence limits (0.80-1.25).
Between-subject variability was included for CL/F and V/F, assuming log-normal distributions without correlation between parameters. Furthermore, CL/F and V/F were estimated using a full variance-covariance matrix. No between-subject variability was included for k a . Within-subject variability (residual) was described by a proportional error model. Standard graphical quality analyses, including goodness-of-fit plots ( Figure S1), were made during qualification of the pharmacokinetic model.

| Exposure-response analysis
The objectives of the exposure-response analysis were: (1) to investigate the exposure-response relationship of liraglutide in adolescents with respect to the change from baseline in BMI SDS and (2) to determine whether the exposure-response relationship for the change from baseline in BMI SDS was similar in adults and adolescents.
An evaluation of the exposure-response of liraglutide in adults has been published previously using data from the phase 2 trial (1807), the phase 3a SCALE Obesity and Prediabetes trial (1839) and the phase 3a SCALE Diabetes trial (1922). 28 Data from the phase 3a trial 4180 (in adolescents) were additionally included in the present exposure-response assessment of liraglutide ( Table 2). The analyses were conducted using response data at week 20 (phase 2 trial 1807), week A modified exposure-response model based on the previous meta-analysis 28 was used in the development of the present exposureresponse model. The starting models included the effect of all investigated covariates on the placebo effect E 0 , and of selected covariates on the E max , as identified based on previous analyses. 28  The final models were parameterized as follows: where CFB denoted the "change from baseline" in BMI SDS, E 0 was the placebo effect, E max was the maximal drug effect, EC 50 the exposure leading to half-maximum effect, E cov covariate effects on the overall response (ie, placebo and treatment-related responses) and e the normal distributed residual error. Sex, baseline response parameter, age group and trial factors were considered as covariates. Baseline BMI SDS and age group were included as covariates in the final model.
The exposure-response evaluation for changes from baseline in body weight, BMI and waist circumference were evaluated using similar models as for the BMI SDS: I adolescent was the covariate representing a lower E max in adolescents compared to adults, where adolescent was an indicator variable for adolescents. I male was the covariate representing a lower E max in males compared to females, where male was an indicator variable for males.
ɣ was the Hill coefficient. In the final models, E cov covariate effects for body weight and BMI included age group, while for waist circumference the covariate effects were baseline waist circumference (cm), age group and diabetic state (ie, diabetes vs non-diabetes). The "nondiabetes" also included individuals with prediabetes. The other expressions were as defined above.
The exposure-response relationship for the proportion of participants with at least 5% reduction in body weight (body weight responder rate [BW.RR]) was estimated using logistic regression analyses parameterized as: The expressions were as defined above. In the final model, age group was included as a covariate effect.  Table S1 and Table 3, respectively.

| Covariate analysis
The effects of intrinsic covariates on liraglutide exposure are shown in Figure 1. All covariates were tested simultaneously; thus for a given covariate effect, the effects of the other covariates were accounted for. In accordance with previous findings in adults, body weight was the main intrinsic covariate affecting liraglutide exposure, with lower exposure at higher body weights. Age group was of no importance and sex was of little importance.
The inverse relationship between body weight and exposure was apparent across trials and age groups (Figure 2).
Individual and mean C avg values appeared to be similar in adolescents and adults when adjusted to the 3.0 mg dose, whereas children had slightly higher liraglutide concentrations ( Figure 3A). When adjusted for differences in body weight, however, exposures were similar across all age groups ( Figure 3B).

| Demographics and datasets
In the exposure-response full dataset, there were more females than males in most of the trials, except for trial 1922 in which sex was evenly distributed (   dosing. [23][24][25]36,37 Furthermore, weight loss has been shown to vary according to initial body weight, 28 and a clinically relevant weight loss is achieved across baseline BMI categories, 38 supporting that weightadjusted dosing is not necessary.

| Liraglutide exposure-response relationship in adolescents and adults
A previous population pharmacokinetic analysis of liraglutide dosed up to 3.0 mg also identified body weight as a key covariate affecting exposure in adults with overweight or obesity with or without type 2 diabetes mellitus; sex was additionally identified as a key covariate in this large phase 3 dataset. 26 The effect of sex was less clear in the present smaller adolescent dataset.
The primary response variable used for the present exposureresponse analysis of liraglutide for weight management was BMI SDS, which is a measure of the number of SDs from the population mean BMI, matched for age and sex. 23 The exposure-response relationship with respect to change from baseline in BMI SDS was successfully described by an E max model with baseline BMI SDS and age group as covariates on the placebo response. As the BMI SDS tends to be skewed at higher BMI SDS values, 33 it was important to evaluate results together with other weight-related parameters. Supportive exposure-response analyses were therefore also conducted for the percentage change from baseline in body weight and BMI, change from baseline in waist circumference as well as for the proportion of participants achieving at least 5% weight loss. The exposure-response relationships for the additional response variables were all successfully described by E max models of exposure-response.
The difference in the weight-loss responses observed between adults and adolescents in the placebo group, where adults achieved a greater mean weight loss than adolescents, could have been due to lower adherence to active treatment and/or the diet and exercise program in some adolescents, although linear growth (height increase) in a part of the adolescent population might also have contributed. There were several lower-than-expected individual liraglutide concentration measurements in trial 4180 in adolescents, which could also potentially indicate a reduced adherence to the trial medication for some trial participants. Low liraglutide concentrations were also observed in four children in trial 4181, although the reason for this was unknown. 22 Lower adherence in adolescents compared with adults is also a significant concern in diabetes management. 39 Clinically, there is an ongoing focus to reinforce adherence to both medication and behavioural strategies such as diet and exercise in both diabetes and obesity management. [39][40][41] In contrast, for adolescents with the highest liraglutide exposures, there was a tendency for weight-loss responses being comparable to the adult responses. Although the tendency for a steeper exposureresponse in adolescents was not significant in this small data set, the exposure estimates in trial 4180 were based on sparse pharmacokinetic sampling over a long time period, and the estimates could be influenced by samples obtained during periods with reduced or varying adherence to treatment. Thus, the individual exposure estimates also could reflect the adherence level.
Limitations of the present exposure-response analysis include a relatively low number of paediatric participants as well as some variability in the populations included in terms of inclusion criteria and background lifestyle therapy. Trial design differed between the trials and included counselling in healthy nutrition and physical activity for all participants, but for the adult participants in trials 1839 and 1922 a calorie-restricted diet was also included. Moreover, the reduced adherence to treatment in adolescents compared with adults as speculated above could not be confirmed in practice. In the exposure-response analysis, the value for a 19-year-old female or male was used as the adult value for the BMI SDS response variable, 33 assuming that these growth charts would be representative also for older adults. Strengths of the analysis include an overall high rate of completion among participants across trials. Long-term exposure data were available through sparse pharmacokinetic sampling, and the availability of data in participants receiving placebo allowed for the described exposure-response analyses. Finally, the availability of large datasets in adults allowed for prior model development, joint analyses and outcome comparisons.
In summary, the population pharmacokinetic analysis indicated similar exposures in adolescents and adults, with similar body weight ranges between the studied adolescents and adults. Body weight was identified as the most important covariate affecting exposure.
An exposure-response relationship with a larger response with increasing liraglutide exposure was established for BMI SDS change from baseline as well as the percentage change from baseline in body weight and BMI, change from baseline in waist circumference, and for the proportion of participants with at least 5% weight loss.
While a lower mean response was observed in adolescents compared to adults, consistent with the placebo response across the exposure range, the added benefits of increasing liraglutide exposures were similar in adolescent and adults with overweight or obesity.