Article Text
PDF
Abstract
Objective This study aimed to evaluate the effectiveness of glucagon-like peptide-1 receptor agonists (GLP-1RA) in reducing body mass index (BMI) and blood glucose levels in individuals with type 2 diabetes mellitus (T2DM) using the difference-in-differences (DID) technique.
Research design and methods This retrospective cohort study included patients with T2DM, receiving GLP1-RA or other second-line antidiabetic treatments between 2010 and 2023. A linear mixed-effect regression with heterogeneous augmented inverse probability weighting DID analysis was used to compare the effectiveness of GLP-1RA and other second-line treatments in reducing BMI, fasting plasma glucose (FPG) and haemoglobin A1C (HbA1c) in patients with T2DM. An average treatment effect on the treated (ATET) for each outcome was estimated.
Results 1000 patients with T2DM (GLP-1RA=220, non-GLP-1RA=880) were included. Compared with other second-line drugs, GLP-1RA significantly reduced BMI by approximately 1.02 kg/m2 (95% CI −1.46 to –0.58) over 24 months of treatment. Additionally, GLP-1RA significantly decreased FPG and HbA1c levels, compared with other second-line treatments with overall ATETs (95% CI) of −21.34 mg/dL (−29.53 to –13.15) and −0.58% (–0.77% to –0.38%), respectively.
Conclusions Our results indicate that patients with T2DM treated with GLP-1RA had a significantly greater reduction in BMI, FPG and HbA1C levels compared with those receiving other second-line antidiabetic therapies. As such, GLP-1RA might be considered the preferred treatment for obese patients with T2DM who fail to sufficiently respond to metformin monotherapy.
- Body Mass Index
- Obesity
- Diabetes Mellitus, Type 2
Data availability statement
Data are available on reasonable request. The datasets generated during and/or analysed in the current study are available from the corresponding author on reasonable request.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
STRENGTHS AND LIMITATIONS OF THIS STUDY
The effectiveness of GLP-1RA was estimated using a real-world data cohort to increase the generalisability of the findings.
A difference-in-difference analysis was conducted to compare changes in body mass index and glucose levels in the same patients before and after the initiation of GLP-1RA treatment while accounting for individual variability.
Due to the limited number of patients receiving GLP-1RA medications, it was not possible to assess the effects of individual drugs within this class.
Certain potential confounding factors such as smoking history and alcohol consumption could not be accessed or controlled due to the lack of this information in the hospital database.
Introduction
Diabetes mellitus (DM) represents a substantial global public health problem. In 2021, 537 million people were diagnosed with DM, with a projected increase to 783 million by 2045.1 Inadequate control of blood glucose levels in individuals with DM is directly associated with an increased risk of vascular complications, which are significant contributors to premature mortality.
Obesity is a significant risk factor for type 2 DM (T2DM)2 and plays an important role in insulin resistance and hyperglycaemia.3 In addition, obesity is also a well-established risk factor for cardiovascular diseases (CVDs). Consequently, weight reduction plays a critical role in managing blood glucose levels and diminishing the risk of vascular complications among patients with T2DM.
Glucagon-like-peptide-1 (GLP-1) is an incretin hormone mainly secreted from L-cells of the small intestine. The binding of this hormone to GLP-1 receptors inhibits hepatic glucose production by inhibiting alpha-cell glucagon secretion and increases insulin secretion from pancreatic beta-cells.4 In addition, binding GLP-1 with its receptors in the central nervous system can increase satiety and decrease appetite,5 6 which consequently promote weight loss. Prior evidence from randomised controlled trials (RCT) demonstrates that glucagon-like peptide-1 receptor agonists (GLP-1 RAs, eg, albiglutide, efpeglenatide, dulaglutide, liraglutide and semaglutide) exhibit robust and effective glycaemic control7 8 while also reducing body weight and systolic blood pressure in individuals with DM.9 Consequently, GLP-1 RAs have been recommended for treating T2DM as the preferred injectable agent prior to insulin therapy.10 Furthermore, GLP-1 RAs have been approved for the treatment of obesity regardless of the presence of DM.11
While the advantages of GLP-1 RAs in reducing weight have been well established in highly screened RCTs, their effectiveness in real-world settings across a heterogeneous population is uncertain, due to the challenges of generalising RCT findings.12 Although RCTs can minimise selection bias, their stringent inclusion criteria often result in the exclusion of key patient groups, such as the elderly, non-obese or very obese. Real-world evidence has also been welcomed by both the US Food and Drug Administration and the European Medicines Agency, for its value in enhancing regulatory decision-making.13 14
Difference-in-difference (DID) design is a controlled before-and-after study that relies on longitudinal data from both treatment and control groups to establish a suitable counterfactual for estimating a causal effect. It is commonly applied to gauge the impact of interventions by comparing outcome changes over time between intervention and control groups. The DID technique lessens potential biases stemming from permanent differences between the groups as well as biases arising from temporal comparisons within the treatment group due to unrelated factors influencing the outcome trends.15
Consequently, we developed a retrospective cohort study based on real-world hospital practice data and used a DID design to evaluate the effectiveness of GLP-1RAs in reducing body mass index (BMI) and blood glucose levels when compared with other second-line antidiabetic drugs in individuals with T2DM.
Methods
Study design and setting
This retrospective cohort study drew the data from the cohort of patients with T2DM who received the treatment at Ramathibodi Hospital, Mahidol University, Bangkok, Thailand between 2010 and 2023. Patients with T2DM included in this cohort were identified by ICD-10 codes of T2DM (ie, E110–119) or fasting plasma glucose (FPG) ≥126 mg/dL (≥7 mmol/L) or haemoglobin A1C (HbA1c) ≥6.5% on at least two consecutive observations or use of any antiglycaemic drugs, as described in the previously published study.16
Study patients
Patients with type 2 diabetes aged 18 years or older were eligible for this study if they met all of the following inclusion criteria: (1) patients who received either a GLP1-RA or another second-line antiglycaemic treatment for at least 2 years and (2) patients who were regularly monitored for body weight, FPG and HbA1c every 6 months for 2 years before and after starting GLP1-RA or other second-line antiglycaemic treatments. The patients were excluded from this study if the interval between prescriptions for GLP1-RA or other second-line antidiabetic medications exceeded 6 months.
Intervention of interest and comparator
The intervention of interest was GLP-1RA. The patients were categorised into the intervention group if they had been prescribed GLP-1RA as the second-line antidiabetic treatment for at least 6 months. The GLP-1RA medications included in this study were exenatide, semaglutide, liraglutide, dulaglutide and lixisenatide, all of which had previously received approval from the Thailand FDA for treating DM. The comparator was any other second-line antiglycaemic medications including sulfonylureas, thiazolidinediones (TZD), dipeptidyl peptidase IV inhibitors (DPP-4i), sodium-glucose cotransporter-2 inhibitors (SGLT2- inhibitors) or insulin. Patients receiving other second-line treatments were randomly selected at a ratio of 1:4 (intervention to control) and were matched with patients in the intervention group based on the initiation dates of GLP-1RA treatment and the other second-line drugs. Using the initiation date of GLP-1RA as the index date, two defined periods were established: a preperiod (6–30 months before treatment) and a postperiod (6–24 months after treatment initiation).
Covariables
Demographic data (ie, age and sex), underlying diseases (ie, hypertension and dyslipidaemia), duration of diabetes and diabetes complications (ie, diabetic retinopathy (DR), chronic kidney disease (CKD), CVDs and peripheral vascular diseases (PVD)) before or at the index date were collected. Demographic data and date of diagnosis of diabetes were retrieved from the hospital information system while underlying diseases and diabetes complications were identified from ICD-10 codes. Duration of diabetes was calculated by subtracting the index date with the date of diagnosis of diabetes. Relevant medications (including dose, frequency/day, amount of prescription and special notes) were retrieved from the computerised provider order entry system and billing databases based on generic/trade names, Thai Medical Terminology and in-house codes. Laboratory data were retrieved from the hospital laboratory information system based on logical observation identifier names and codes.
Outcomes of interest
The outcomes of interest were BMI, FPG and HbA1c levels. BMI was calculated by dividing body weight in kg by height in meter square. Body weight and height were measured at every outpatient clinic visit. FPG and HbA1c levels were measured using hexokinase glucose-6 phosphate dehydrogenase and turbidimetric inhibition immunoassay, respectively. HbA1c measurement was certified by the National Glycohemoglobin Standardization Program. BMI, FPG and HbA1c values were grouped into 6-month intervals and assigned to the preperiods and postperiods. If there were multiple values of BMI, FPG and HbA1c per 6 months period, then an average value was calculated.
Statistical analysis
Baseline characteristics of participants between the GLP-1RA and non-GLP-1RA groups were described using mean and standard deviation (SD) for continuous variables, or frequency and percentage for categorical variables, and were compared using Student’s t-test or χ2 test, respectively.
Multiple imputation by chained equation with 10 replications was applied to impute missing outcome data for the whole follow-up period based on the assumption that data were missing at random. DID analysis was used to estimate the effect of GLP-1RA on clinical outcomes (ie, BMI, FPG and HbA1c) by comparing pretreatment and post-treatment periods within each treatment, and then comparing these differences between treatments. A linear mixed-effect regression with heterogeneous augmented inverse probability weighting (AIPW) DID was used to estimate an average treatment effect in the treated (ATET) by assuming that treatment effect varied over time and over treatment cohorts. Covariates (ie, age, sex, duration of diabetes, DR, CKD, PVD and CVD) were also considered but only those that were significant were retained in the final model. All analyses were performed by using STATA V.18.0. A p-value <0.05 was considered statistically significant.
Patient and public involvement
Patients or the public were not involved in the design, or conduct, or reporting, or dissemination plans of our research.
Results
Out of 71 663 patients with T2DM, 54 644 patients were not eligible (see online supplemental figure S1), resulting in 17 019 potentially eligible patients. Among these, 220 patients who received GLP-1RA were included in the treatment group, while 880 out of 16 799 patients who did not receive GLP-1RA were randomly chosen as controls, following a predetermined ratio of 1 case to 4 controls. Other second-line antidiabetic drugs prescribed in the control group were DDP-4i (34.9%), SGLT2i (18.0%), TZD (15.6%), sulfonylurea (9.4%) and insulin (2.1%).
Supplemental material
Participant characteristics at the index date are presented in table 1. Mean age was significantly lower in GLP-1RA group than in control group, while the presence of microvascular and macrovascular complications (ie, DR, CKD and CVD) was significantly higher in non-GLP-1RA group than GLP-1RA group. In addition, patients receiving GLP-1 RAs had significantly higher duration of diabetes, BMI, FPG and HbA1c levels at baseline compared with patients not receiving GLP-1RA.
Characteristics of study participants at the index date
Mean BMI, FPG and HbA1c levels were plotted by groups (ie, GLP-1RA vs non-GLP-1RA) across the pretreatment and post-treatment periods (see figure 1A–C). For the GLP-1RA group, BMI and HbA1C tended to increase during the 24-month period before the index date, but then decreased during the 24-month period after GLP-1RA treatment, while FPG continued to decrease throughout the pretreatment and post-treatment periods. On the contrary, individuals in the non-GLP-1 RA group maintained relatively stable BMI and HbA1c values throughout the entire duration of the study while experiencing an increase in FPG levels during the post-treatment period.
Trends of body mass index, fasting plasma glucose and HbA1c values during pre-index and post-index date. GLP-1RA, glucagon-like peptide-1 receptor agonists; HbA1c, haemoglobin A1C.
After adjusting for covariates in the models, the overall ATET from the AIPW-DID analysis was −1.02 kg/m2 with a 95% CI of −1.46 to −0.58. This suggests that GLP-1RA significantly reduced BMI by approximately 1.02 kg/m2 over the 24-month treatment period when compared with other second-line antidiabetic drugs (see table 2 and figure 2A). Additionally, ATETs with 95% CIs at 6, 12, 18 and 24 months postinitiation of GLP-1RA treatment were as follows: −0.65 kg/m2 (−0.99 to –0.31), −1.08 kg/m2 (−1.65 to –0.50), −1.25 kg/m2 (−1.88 to –0.61) and −1.11 kg/m2 (−1.72 to –0.50), respectively (see table 3). These findings indicate that in patients taking GLP-1RA, there was a significant decrease in BMI at 6, 12, 18 and 24 months after starting treatment, compared with those using other second-line antidiabetic medications.
Average treatment effects on body mass index, fasting plasma glucose and HbA1c values before and after receiving GLP-1RA. GLP-1RA, glucagon-like peptide-1 receptor agonists; HbA1c, haemoglobin A1C.
The average treatment effects of GLP-1 RA relative to non-GLP-1 RA on body mass index, fasting blood glucose and HbA1c
The average treatment effects on body mass index, fasting plasma glucose and HbA1c before and after receiving GLP-1 RA
In addition, GLP-1RA significantly improved blood glucose levels during 24 months of treatment, compared with other second-line antidiabetic treatments, with the overall ATETs (95% CI) of −21.34 mg/dL (−29.53 to –13.15) for FPG and −0.56% (−0.77% to –0.38%) for HbA1c level (table 2 and figure 2B,C). The ATETs (95% CIs) of FPG at 6, 12, 18 and 24 months after receiving GLP-1RA were −15.87 mg/dL (–25.46 to –6.28), −14.93 mg/dL (–26.72 to –3.14), −24.15 mg/dL (–35.38 to –12.91) and −30.41 mg/dL (–41.54 to –19.28), respectively, while the ATETs of HbA1C for these corresponding times were −0.55% (–0.75% to –0.34%), −0.50% (−0.74% to −0.26%), −0.57% (−0.82% to −0.32%) and −0.69% (−0.93% to –0.44%), respectively (table 3). Similar to the results regarding BMI, patients taking GLP-1RA experienced significant decreases in FPG and HbA1c levels after starting treatment at 6, 12, 18 and 24 months, compared with those using other second-line antidiabetic medications.
Discussion
In our study, we found that GLP-1RA treatment over 24 months provided significant benefits in reducing BMI by 1.03 kg/m2, lowering FPG by 22.80 mg/dL and improving HbA1c levels by 0.59% in individuals with T2DM compared with other second-line antidiabetic treatments. Importantly, these advantages persisted for up to 2 years after the initial prescription of GLP-1RA.
Our findings are consistent with the standard recommendation to use GLP-1RA for weight-centric pharmacological management of T2DM.17 Moreover our results align with those of other observational studies in Japan,18 19 Spain,20 Italy21–23 and Hong Kong24 which highlighted the effectiveness of GLP-1RA over other antiglycaemic drugs. Specifically, the average BMI of patients with T2DM who were administered GLP-1RA decreased by approximately 0.4–1.55 kg/m2 from baseline measures; furthermore, these studies revealed a mean decrease in FPG by approximately 12–28 mg/dL and a mean decrease in HbA1c levels by 0.77%–1.1%. The duration of follow-up in these studies varied from 12 to 39 months. Remarkably, the beneficial effects of GLP-1RA on both BMI and blood glucose were shown to persist for up to 39 months20 after patients received their initial dose of GLP-1RA.
However, previous studies have only compared the effects of GLP-1RA on BMI and blood glucose changes before and after treatment using univariate analysis methods such as the Wilcoxon rank-sum test,18 22 23 analysis of covariance19 20 and mixed-effect linear regression.19 21 This approach limits the evidence and understanding of causality. Only one other study24 used DID analysis with a control group. This retrospective cohort study was conducted in an Asian country, using real-world data from hospital databases, similar to our study. The results from this study demonstrated reductions in BMI, HbA1c and FPG of 0.64 kg/m2, 0.77% and 0.69 mmol (~12.43 mg/dL) after 12 months of treatment. In comparison, our study revealed larger reductions in BMI and FPG (1.28 kg/m2 and 20.92 mg/dL, respectively), but smaller reductions in HbA1c (0.59%). The larger reductions in BMI and FPG observed in our study may be due to the inclusion of a different control group. While this study used SGLT2 inhibitors as the control, our study included a broader range of second-line antidiabetic drugs (eg, sulfonylureas, DPP-4 inhibitors, SGLT2 inhibitors and TZDs) as comparators.
Using the DID method within a causal inference framework should provide the most valid results from our real-world dataset because the DID technique could help reduce biases when comparing the postintervention periods of the treatment and control groups. This method addresses potential biases stemming from permanent differences between the groups as well as biases arising from temporal comparisons within the treatment group due to unrelated factors influencing the outcome trends.25 Adequacy of this adjustment method can be evaluated by assessing that both groups exhibit preperiod outcome slopes (figure 2) which was demonstrated for both BMI and HbA1c.
Moreover, our findings corroborate those previously reported in RCTs. Systematic reviews and meta-analyses of RCTs have indicated that treatment with GLP-1RA resulted in a significant reduction in HbA1c levels, typically ranging from 0.2% to 0.8%, compared with active controls such as sitagliptin, pioglitazone and basal insulin glargine.26 27 Additionally, treatment with semaglutide 1 mg resulted in a significant reduction in body weight, with an average weight loss of approximately 4.11 kg compared with the placebo group. It is important to note that the treatment durations in most RCTs were shorter than in our study, typically lasting 6 months as opposed to the 2-year duration in our investigation. However, evidence from a recent RCT that followed patients for a longer period (5 years)28 also demonstrated substantial benefits of GLP-1RA, particularly liraglutide, compared with sitagliptin and glimepiride. Specifically, it highlighted a significant reduction in the incidence of HbA1c levels exceeding 7%. Additionally, this RCT suggested that patients with T2DM receiving GLP-1RA experienced more significant weight loss compared with those treated with sitagliptin, glimepiride or glargine.
Strengths and limitations
Our study used data from a real-world setting to assess the effectiveness of GLP-1RAs in improving BMI and glucose homoeostasis in patients with T2DM who failed metformin monotherapy. Using this real-world data increases the generalisability of these findings to broader populations compared with those usually recruited to RCTs. In addition, outcome values measured for 2 years before and after the GLP-1RA initiations were analysed by applying the DID method. This method helps reduce biases when comparing the postintervention period between the intervention (ie, GLP-1RA) and comparator groups (ie, non-GLP-1RA). These biases could stem from inherent differences between the groups or trends in the intervention influenced by other confounding factors affecting the outcomes over time. Our study uses the most rigorous and stringent methods of adjustment available in the causal inference arsenal to handle these biases.
However, our study does have some limitations that need to be acknowledged. First, we grouped other second-line antiglycaemic drugs together instead of analysing them individually due to the limited number of participants in some drug classes such as SGLT2i. Consequently, we were unable to directly compare the effects of GLP-1RA with specific second-line antidiabetic drugs. Additionally, our study was limited by the small number of patients who were prescribed GLP-1RA, which hindered our ability to assess the individual effectiveness of different GLP-1RA medications. Furthermore, the risk of microvascular and macrovascular complications (eg, CVD and CKD), side effects and withdrawal rate of GLP-1RAs were not assessed in our study due to the limited number of participants. Therefore, further studies with larger sample sizes are needed to explore the benefit of individual GLP-1RAs and their effects on vascular complications risk.
Lastly, our study was designed as a retrospective cohort study that retrieved data from a hospital database. This means that we could not access certain potential confounding factors, such as patients’ history of smoking, alcohol consumption and lifestyle factors like dietary habits and physical activity levels. As a result, our study results may be influenced by residual confounding. However, it is worth noting that when we compared the baseline characteristics, we observed that patients in the GLP-1RA group tended to have a worse prognosis compared with those in the non-GLP-1RA group, biasing against the GLP-1RA group and making our results more robust.
Conclusion
Our findings suggest a greater reduction in BMI, FPG and HbA1C in patients with T2DM receiving GLP-1RA compared with those receiving other second-line antidiabetic medications. Consequently, GLP-1RA might be a suitable second-line antidiabetic drug for individuals with T2DM who are overweight or obese and fail metformin monotherapy.
Data availability statement
Data are available on reasonable request. The datasets generated during and/or analysed in the current study are available from the corresponding author on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
The protocol of this study was approved by the Human Research Ethics Committee, Faculty of Medicine Ramathibodi Hospital, Mahidol University (MURA2023/890). Informed consent was not obtained because this study was a retrospective cohort using existing data from the previous cohort of patients with T2DM.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors SS, TA and AT were involved in the conception, design and conduct of the study and the analysis and interpretation of the results. PL was involved in study design, data management and data analysis. HN, GM and JA were involved in study design and result interpretation. SS and TA wrote the first draft of the manuscript, and all authors edited, reviewed and approved the final version of the manuscript. SS and TA are the guarantors of this work and, as such, had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis.
Funding This study was supported by the National Research Council of Thailand (N42A640323).
Disclaimer The funding agency had no role in study design; collection, analysis, and interpretation of data; writing the report; and the decision to submit the report for publication.
Competing interests None declared.
Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.