Factors associated with weight loss response to GLP-1 analogues for obesity treatment: a retrospective cohort analysis

Abstract

Objectives

The response to glucagon-like peptide-1 (GLP-1) analogues for weight loss varies significantly. We investigated the anthropometric, demographic and clinical characteristics associated with total body weight loss (TBWL) from subcutaneous GLP-1 analogue therapy in patients with obesity in a real-world setting.

Design

Retrospective cohort analysis.

Setting

An urban, multidisciplinary obesity community clinic in Vancouver, Canada, from November 2018 to April 2021.

Participants

483 adults with a body mass index (BMI) of ≧30 kg/m² who had filled a new prescription for subcutaneous semaglutide or liraglutide, with at least 6-month follow-up, were included (mean follow-up: 17.3 months). Individuals with prior bariatric surgery were excluded.

Outcomes

The primary outcome was the %TBWL over a mean follow-up period of 520 days. Participant’s TWBL was categorised as non-response (<5% TBWL), moderate response (5%–15% TBWL) or hyper-response (>15% TBWL).

Results

The average %TBWL in the cohort was 12.2%. Among the participants, 17.8% had a non-response, 48.4% had a moderate response and 33.8% had a hyper-response. In the multivariable regression analysis, being a woman was associated with hyper-response (adjusted OR 1.92, CI 1.01 to 3.65, p=0.048). Age, diabetes status, baseline BMI, being sedentary, anxiety and depression were not independently associated with TBWL in response to GLP-1 analogue therapy.

Conclusions

In a real-world setting, female sex was found to be associated with a hyper-response to GLP-1 analogue therapy for obesity management. Other clinical factors evaluated, including diabetes status, were not associated with the response. Future research should assess additional variables and support the development of novel biomarkers that are associated with weight loss response.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Being retrospective in nature, these data reflect real-world obesity management practices to a greater extent than randomised controlled trials.

• The sample size was larger than most recent studies looking at factors associated with weight loss with glucagon-like peptide-1 receptor analogues, allowing greater power to detect associations.

• A small proportion of participants were using liraglutide instead of semaglutide, which may have reduced the overall weight loss effect size.

• Although we evaluated multiple demographic and clinical factors, there remains a possibility of residual confounding. We did not have data on other lifestyle factors that may have contributed to weight loss, such as coexisting dietary or exercise interventions but these are not likely to be significant.

Introduction

Obesity is becoming increasingly common worldwide, posing a global health threat.¹ The prevalence of obesity in Canada was 27.2% in 2018 and is expected to reach 38.5% by 2030.^{1 2} In addition to increasing the risk of cardiovascular and metabolic diseases, obesity increases the rates of COVID-19 infection, COVID-related morbidity and mortality, osteoarthritis, functional decline and early mortality.¹ The stigma experienced by patients living with obesity raises the risk of anxiety and depression and is linked to worse clinical care.³ The total cost of overweight and obesity-related disease in Canada was over $C20 billion in 2020 and is expected to continue to rise.⁴ Given the significant impact obesity continues to pose, there is an urgent need to deploy effective interventions to reduce obesity and prevent its associated morbidity and mortality.

Although health behaviour optimisation is an important component of weight loss, recent trials have identified that glucagon-like peptide-1 receptor analogues (GLP-1 RAs) provide significant and longer-term weight loss.⁵ Semaglutide 1 mg per week was associated with a 7% total body weight loss (TBWL) in the STEP 2 trial in those with diabetes and obesity.⁶ Similarly, smaller cohort studies showed 5.1%–9.2% TBWL in those without diabetes.^{7 8} Higher doses, up to 2.4 mg per week, have demonstrated even greater weight loss benefits, reaching up to 17.6% TBWL in the STEP 3 trial.^{56 9},¹² However, large phase 3 clinical trials demonstrated a wide range in TBWL achieved in those treated with semaglutide. At the 2.4 mg dose, 50%–55% of participants achieved >15% TBWL, while 10%–30% of participants achieved <5% TBWL (non-responders) over long-term follow-up.^{56 9},¹² The variation in achieved TBWL is clinically relevant, as a minimum threshold of 5% TBWL is associated with improvement in lipids, glycaemic control and systolic blood pressure.^{13 14} Weight loss greater than 15% leads to further improvement in these metabolic markers, metabolic dysfunction-associated steatotic liver disease and markers of inflammation^{13 14}

Although GLP-1 analogues are effective and tolerated by the majority of patients, a significant proportion experience adverse effects or do not respond with sufficient weight loss.¹⁵ Gastrointestinal side effects such as nausea occur in up to 25% of patients, and up to 10% experience vomiting or diarrhoea.¹⁵ Other side effects include gastroparesis, injection site reactions and less frequently, gallstone disease.¹⁵ Further, GLP-1 analogues are quite costly and, in many regions, are in limited supply.¹⁶ Patients who do not respond incur burdensome personal financial costs for these treatments, are unnecessarily exposed to adverse effects and draw significant healthcare resources with multiple healthcare visits.¹⁶ Lengthy courses of treatment with GLP-1 analogue therapy among patients who ultimately fail to respond may delay starting more effective surgical treatments.¹⁷ Conversely, hyper-response to medical therapy may influence decisions to proceed with bariatric surgery.¹⁷

Identifying subpopulations that would benefit from GLP-1 analogue therapy would enable practitioners to better allocate these treatments. Unfortunately, data evaluating associated factors of non-response and hyper-response to GLP-1 RAs are sparse.¹⁸,²⁰ Retrospective analyses of phase III clinical trials demonstrated that both sex and diabetes status were associated with GLP-1 analogue response.⁶²¹,²⁹ Dedicated analyses designed to identify additional factors associated with weight loss with GLP-1 RAs have been limited by conflicting data, lack of data on hyper-response, and small sample sizes (generally fewer than 50 participants), raising the risk of bias.^{26 30 31} We aimed to investigate whether demographic, anthropometric or clinical factors were associated with non-response or hyper-response to weight loss with subcutaneous GLP-1 analogue therapy, with either semaglutide or liraglutide, in a real-world setting using a large cohort of patients referred to an obesity specialty clinic for obesity treatment.

Methods

Study design, setting and participants

This retrospective longitudinal cohort study included adult patients with a body mass index (BMI) of 30 kg/m² or greater attending an obesity specialist community clinic in Vancouver, Canada, from November 2018 to April 2021. The clinic receives referrals from across the province of British Columbia (population 5.1 million) for medical weight loss or bariatric surgical preparation. The clinic includes internal medicine specialists trained in obesity medicine, dietitians and specialised obesity nursing care. All patients receive education on the biological underpinnings of obesity, nutrition, exercise, pharmacological therapy and if they meet indications, bariatric surgery at their initial visit. After March 2020 during the COVID-19 pandemic, medical and nursing support consultations were conducted virtually by video conferencing or telephone.

The database included 530 patients with a BMI of 30 kg/m² or greater, aged 19 years or older and with a minimum of 6-month follow-up. The database only included patients who had not undergone bariatric surgery. Patients were excluded from the database if they had undergone bariatric surgery prior to starting a GLP-1 analogue or during the observation period. There were no patients prescribed oral GLP-1 analogues. For this study, we only included patients from the database who had been prescribed a subcutaneous GLP-1 analogue (483/530 patients). Patients with coexisting binge-eating disorder were also included if they received a prescription for a GLP-1 analogue.

Patient and public involvement

Due to feasibility constraints related to the timing of data collection and study design, patients and the public were not involved in the development of the study methods. Patient consent was waived for this study given its retrospective nature. All patient data were handled and reported anonymously.

Data collection procedures and factors associated with weight loss

Data were abstracted from electronic medical records for all consecutive patients meeting the inclusion and exclusion criteria until the final sample size of 483 was reached. Data were entered into a REDCap research database. Data collectors underwent training and validity checks. Inter-rater reliability between data collectors was >95%.

Data on the type of GLP-1 analogue (subcutaneous semaglutide and liraglutide), the initial start date of GLP-1 analogue and the highest dose of semaglutide or liraglutide prescribed were collected.

Outcome measures

For the primary endpoint, baseline and follow-up weight were measured at the initial and subsequent physician visits using a platform bariatric weighing scale. During the pandemic, patients self-reported their weight using home scales. %TBWL was calculated by subtracting the lowest body weight during the follow-up period from the baseline index visit weight, divided by the baseline index weight (×100%).

Independent variables

Baseline variables were chosen based on previous data indicating associations with GLP-1 analogue weight changes (history of or presence of diabetes, defined as glycated haemoglobin (HbA1c)>6.5% or use of diabetes medications, but excluding those with pre-diabetes, and sex)⁶²¹,²⁹ and other factors associated with weight loss or known barriers to weight loss: age, baseline clinic or self-reported body weight, height and baseline BMI (calculated as measured weight in kilograms divided by measured height in meters squared); anxiety, depression, self-reported sedentariness and smoking (current, never smoker or previous smoker).^{32 33}

Analysis

Patients were categorised into groups based on their maximal %TBWL over the follow-up period from baseline: non-responders (<5% TBWL), moderate responders (5%–15% TBWL) and hyper-responders (>15% TBWL). Baseline characteristics and independent variables were compared across TBWL categories using the Kruskal-Wallis equality-of-populations rank test (2 df) or one-way analysis of variance with Tukey’s honestly significant difference post hoc test. Multinomial regression models were constructed with the three percent TBWL categories and independent variables using forced entry modelling with age, sex, baseline BMI, type 2 diabetes mellitus (T2DM), depression, sedentariness, anxiety and depression as covariates. There were missing data on baseline BMI (3.1%), sex (2%), maximum semaglutide dose used (1%) and 25% of patients had missing weight measurements in the follow-up period. If weight at the 12-month period was not available, the last recorded weight was carried forward (1% were missing a weight category). Patients with missing data were excluded from multivariable analysis. Stata software (V.12) was used for all analyses (StataCorp. Stata Statistical Software: Release 12. College Station, Texas, USA: StataCorp LP, 2011).

Results

A total of 483 participants met the inclusion criteria and were included in the database. Participants were followed for a mean of 520 days (IQR 420–609 days).

Baseline characteristics

Baseline characteristics of the cohort are highlighted in table 1. The average baseline BMI of patients in this cohort was 43.5 kg/m², and the average age was 47.2 years. Most participants were women (74.7%), and there was a high baseline prevalence of hypertension (43.2%) and dyslipidaemia (45.6%). Most patients (86%) were prescribed subcutaneous semaglutide, and 14% were prescribed liraglutide. Among those prescribed semaglutide, 1 mg/week was the most frequently prescribed dose and the use of 2 or 2.4 mg doses per week was more common among those with a greater weight loss category.

Table 1. Baseline characteristics among those prescribed glucagon-like peptide-1 analogues according to per cent total body weight loss, N (%) (total n=483)^*.

Characteristic	Per cent total body weight loss			P value
	<5%n=86 (17.8)	5%–15%n=234 (48.4)	>15%n=163 (33.7)
Age, years mean (SE)	44.8 (1.3)	48.5 (0.8)	46.4 (1)	0.03
Women versus men	59 (68.6)	169 (72.5)	133 (81.6)	0.04
Sedentariness versus active	16 (18.6)	42 (18)	33 (20.3)	0.85
Smoking
Smoker	11 (20.8)	23 (14.8)	12 (12.6)	0.24
Ex-smoker	13 (24.5)	30 (19.4)	19 (20)
Non-smoker	29 (54.7)	102 (65.8)	64 (67.3)
ETOH				0.1
None or rare	58 (69.1)	134 (57.8)	108 (66.7)
1–6 drinks/week	21 (25)	80 (34.5)	43 (26.5)
>6 drinks/week	5 (6)	18 (7.8)	11 (6.8)
BMI mean (SE)	45.4 (1.3)	43.6 (0.6)	42.6 (0.7)	0.1
Type 2 diabetes	16 (18.6)	68 (29.1)	34 (20.9)	0.07
HbA1c (SE)	6.2 (0.2)	6.2 (0.1)	5.9 (0.1)	0.14
Hypertension	32 (37.2)	113 (48.3)	63 (38.7)	0.08
Dyslipidaemia	41 (47.7)	118 (50.4)	61 (37.4)	0.03
MASLD	10 (11.6)	17 (7.3)	10 (6.1)	0.29
Anxiety	7 (8.1)	16 (6.8)	9 (5.5)	0.72
Depression	32 (37.2)	70 (29.9)	66 (40.5)	0.08
Use of diuretics	11 (12.8)	46 (19.7)	28 (17.2)	0.36
Use of SGLT2 inhibitor	4 (4.7)	17 (7.3)	8 (4.9)	0.53
Use of sulfonylureas	2 (2.3)	4 (1.7)	5 (3.1)	0.67
Use of insulin	0 (0)	1 (0.4)	0 (0)	0.99
Use of bupropion/naltrexone	37 (43)	78 (33.3)	49 (30.1)	0.12
Highest ozempic dose, mg/week				0.0001
≤0.5 mg	26 (30)	19 (8.1)	6 (3.7)
1 mg	31(36)	87 (37.2)	59 (36.2)
2 mg	9 (10.5)	55 (23.5)	49 (30.1)
2.4 mg	14 (16.3)	59 (25.2)	44 (27)

^*Categorical and continuous variables were compared byacross weight categoryies using the Kruskal-wWallis equality-of-populations rank test (2 of freedomdf) or one -way ANOVA analysis of variance with Tukey’s HSDhonestly significant difference post hoc test, respectively.

BMIbody mass indexETOHalcoholHbA1cglycated haemoglobinMASLDmetabolic dysfunction-associated steatotic liver diseaseSGLT2isodium-glucose transport protein 2 inhibitor

Average weight loss across all patients was 12.2%. A total of 17.8% of participants were non-responders, 48.4% were moderate responders, and 33.8% were hyper-responders. As shown in table 1, being a woman versus man (p=0.04), increasing age (p=0.03) and absence of dyslipidaemia (p=0.03) were associated with %TBWL. Tukey’s HSD post hoc test revealed that patients with greater weight loss (5%–15%) were older than non-responders.

Factors associated with GLP-1 receptor analogue response: multivariable analysis

The results of the multivariable analysis are displayed in table 2. No independent variables were associated with moderate response compared with non-response. Being a woman was significantly associated with hyper-response compared with non-response (OR 1.92, CI 1.01 to 3.65, p=0.048). None of the other variables analysed were associated with hyper-response versus non-response.

Table 2. Multivariable analysis of factors associated with per cent total body weight loss response to glucagon-like peptide-1 analogues (total n=483) versus non-response^*.

Independent variables	<5% vs 5%–15% TBWL		<5% vs >15% TBWL
	Adjusted OR (95% CI)	P value	Adjusted OR (95% CI)	P value
Age (years)	1.02 (1.00 to 1.04)	0.08	1.01 (0.98 to 1.03)	0.52
Sex (male vs female)	1.22 (0.68 to 2.17)	0.50	1.92 (1.01 to 3.65)	0.05
Baseline BMI	0.99 (0.96 to 1.01)	0.33	0.97 (0.94 to 1.00)	0.06
Diabetes versus no history of diabetes	1.81 (0.94 to 3.47)	0.08	1.27 (0.63 to 2.58)	0.51
Depression versus no diagnosis of depression	0.84 (0.47 to 1.50)	0.56	1.38 (0.76 to 2.49)	0.29
Anxiety versus no diagnosis of anxiety	1.01 (0.37 to 2.71)	0.99	0.59 (0.20 to 1.74)	0.34
Sedentary versus no sedentariness	0.97 (0.49 to 1.94)	0.94	1.45 (0.71 to 2.97)	0.31

^*Multinomial logistic regression of total body weight loss categories compared with non-response in weight loss.

BMIbody mass indexTBWLtotal body weight loss

Discussion

Determining clinical and demographic factors associated with weight loss response to GLP-1 analogues is essential for identifying subpopulations that may either not respond or experience a pronounced hyper-response to treatment. Older age, being a woman and the absence of dyslipidaemia were significantly associated with %TBWL in the univariate analysis. After adjusting for confounding variables in the multivariable analysis, only being a woman remained significantly associated with a hyper-response; however, the CI was quite broad with a p value of 0.048. Surprisingly, none of the other variables, including diabetes status and baseline BMI, were associated with weight loss response in the multivariable analysis.

To our knowledge, the baseline clinical and demographic associations of age and dyslipidaemia with weight loss response have not been demonstrated in previous studies. This association with age could be mediated by changes in hormonal signalling that occur with increasing age. For example, previous analyses identified a positive relationship between age and postprandial endogenous GLP-1 release.³⁴ This analysis agrees with multiple existing studies that have demonstrated that women have a greater weight loss response to GLP-1 analogue therapy. This is thought to be partially due to a lower average body mass, resulting in higher drug concentrations at any given fixed dose. Controlling for drug concentrations reduces but does not completely eliminate this association, suggesting additional mechanisms.²⁸ These have not yet been elucidated but may be due to sex-related physiological differences in obesity and hormonal appetite regulation, including leptin, ghrelin and insulin signalling.²¹

The lack of association of diabetes with weight loss response contrasts with existing randomised controlled trials (RCTs).²⁷,²⁹ This discrepancy may be mediated by differences in prescribing practices, adherence and tolerance that may not be accounted for by RCTs. In this real-world study, patients with diabetes had relatively controlled glycaemic indices. The average HbA1c in this study was 6.1%, compared with 8.1% in the STEP 2 trial, the only phase III RCT dedicated to examining weight loss with semaglutide in those with T2DM. In our study, patients had weight-promoting oral hypoglycaemic agents such as sulfonylureas discontinued, whereas in the STEP 2 trial, sulfonylureas were used by 26% of participants.⁶ Insulin, which is also weight promoting, was prescribed to only one person in our cohort, and patients taking insulin were excluded from the STEP 2 trial.

The lack of statistically significant demographic, clinical and anthropometric factors identified in this study suggests that future research may need to shift towards biochemical and genetic markers of therapeutic response. Several proposed clinical predictors of GLP-1 analogue response, including sex, age and T2DM, have all been associated with measurable differences in the hormones involved in energy homeostasis.³¹³⁴,³⁶ Potential targets for further investigation include hormonal, genetic and microbiome biomarkers.

Leptin and ghrelin are hormones that are heavily involved in energy homeostasis, and interindividual variability in these hormonal systems is implicated in the pathophysiology of obesity.³⁷ Those with higher baseline levels of ghrelin and greater postprandial ghrelin reduction have a greater response to GLP-1 analogue therapy.³¹ Data regarding leptin and GLP-1 analogue response has been mixed; however, one study has shown that lower baseline leptin levels were associated with a greater response in those without T2DM.^{30 31} Greater baseline levels of GLP-1 have been associated with a positive response to GLP-1 analogue therapy in those with obesity; however, it is unclear whether this holds true for all subgroups.^{30 31} Because GLP-1 is rapidly metabolised, its circulating levels are unstable, which poses a challenge for use as a biomarker. Single-nucleotide polymorphisms (SNPs) in the GLP-1 receptor gene have been shown to affect GLP-1 binding affinity and downstream signalling.³⁸ Some of these have been implicated in the pathogenesis of obesity and weight loss.³⁹ Particular SNPs in the GLP-1 receptor gene, as well as other genes such as β-arrestin, are associated with variable responses to GLP-1 analogues.^{40 41} GLP-1 analogues are known to cause changes in gut microbiota, and glycaemic response to GLP-1 analogues has been associated with unique gut microbiome signatures.⁴²

There are several factors to consider when interpreting the findings of this analysis. Being retrospective in nature, these data reflect real-world obesity management practices to a greater extent than RCTs. Our sample size was larger than most recent studies examining correlates of weight loss with GLP-1 RAs. However, the study did have some limitations. It was conducted at a single obesity centre that receives regional referrals, and some data from the cohort were missing. Most of the patients with diabetes were well controlled, which may limit the ability to generalise our findings to populations with more advanced or poorly controlled diabetes. Additionally, a small proportion of participants were taking liraglutide rather than semaglutide, potentially lowering the observed weight loss effect size. While we evaluated multiple demographic and clinical factors, there remains a possibility of residual confounding, as we did not have data on other lifestyle factors, such as coexisting dietary or exercise interventions, which may have influenced weight loss outcomes, but these interventions are known to have little impact on long-term weight loss.

Conclusions

Our analysis revealed that, overall, in a real-world setting, response to GLP-1 analogue therapy is highly variable. Being a woman may be associated with a favourable response, but other key clinical factors like diabetes, age, anthropometric characteristics, depression, anxiety and sedentary behaviour do not correlate with response to GLP-1 analogue therapy. Due to the heterogeneity of weight loss with GLP-1 analogue therapy, practical and accessible predictors of response are needed to help guide the selection of patients who may benefit from these medications. Given the lack of reliable clinical or demographic indicators of response, we propose that future research should be targeted towards biomarkers. The identification of such markers would provide a tool for customising the treatment of obesity, thereby minimising the waste of resources, reducing the burden of obesity, and preventing its downstream morbidity and mortality.

Data availability statement

Data are available on reasonable request.