PDF
Citation
Print
INTRODUCTION
Sodium-glucose cotransporter 2 inhibitors (SGLT2i) improve glycemic control (GC) by selectively inhibiting glucose reabsorption in the renal proximal convoluted tubule, causing increased glucosuria [1]. This process results in the elimination of approximately 60 to 100 g of glucose per day through the urine, contributing to an average weight loss (WL) of 2 to 3 kg in patients with diabetes [2]. However, real-world evidence suggests substantial variation in the extent of GC and WL observed among individuals after SGLT2i therapy [3]. Additionally, some patients experience an increased desire for sweet foods after starting SGLT2i treatment, which can lead to weight gain or elevated blood glucose levels due to higher caloric intake [4,5].
Regarding glycemic outcomes of SGLT2i therapy, a reduction in glycated hemoglobin (HbA1c) is significantly associated with preserved renal function [6,7] and a higher baseline HbA1c level [8,9]. However, the factors influencing WL in patients taking SGLT2i are less consistent; those with a higher body mass index (BMI) tend to experience greater WL [10], with the caveat that some studies have not found a significant association [11,12]. Greater WL with SGLT2i treatment has been observed in individuals with longer durations of diabetes [13], although the reasons behind this observation were not addressed in that report. In one study, a higher HbA1c level was linked to less WL [14], while another study reported no correlation between HbA1c level and the extent of WL [15]. Notably, the weight-lowering effect of SGLT2i seems to be maintained in patients with impaired renal function, which contrasts with the potential reduction in glucose-lowering efficacy [16].
Previous research has explored the efficacy of dapagliflozin in reducing glucose levels and facilitating WL. While some patients may achieve both glycemic and weight management goals with SGLT2i, others may meet only one of these targets. Consequently, this post hoc analysis of the Effects of Dapagliflozin Compared with Glimepiride on Body Composition in Patients with Type 2 Diabetes Inadequately Controlled with Metformin (BEYOND) trial seeks to confirm the heterogeneity of treatment responses and to identify factors contributing to these differences. To this end, we concurrently investigated the effects of dapagliflozin on hyperglycemia and body weight. Moreover, this study examines changes in body composition, extending beyond weight alone as the traditional measure of obesity.
METHODS
Study design and participants
The BEYOND trial was a 52-week, multicenter, randomized, parallel-group, open-label, phase IV study conducted in Korea from January 2016 to January 2018. It assessed the effects of dapagliflozin on body composition in patients with type 2 diabetes in comparison to glimepiride [17]. Eligible participants were those between 19 and 75 years old, with HbA1c levels inadequately controlled (7.0% to <10.0%) despite being on a stable regimen of metformin monotherapy (≥1,000 mg/day) for at least 8 weeks prior to randomization. The inclusion and exclusion criteria of the BEYOND trial are detailed in Supplemental Fig. S1 [17]. A total of 124 patients were randomly assigned in a 1:1 ratio to receive either dapagliflozin (10 mg once daily) or glimepiride (1–2 mg/day), alongside metformin. Of these, 56 patients in the dapagliflozin group who completed the 52-week follow-up (constituting the full analysis set) were included in this post hoc analysis (Supplemental Fig. S2) [17].
Physical examinations, laboratory tests, and body composition assessments—including dual-energy X-ray absorptiometry (DXA) and abdominal computed tomography (CT) scans—were performed at baseline and at week 52 (Supplemental Fig. S1). Whole-body DXA operates on the principle that the differential attenuation of two distinct X-ray energies by body tissues supports a three-component model that quantifies total fat mass and lean mass. CT scans were utilized to measure the areas of abdominal visceral adipose tissue (VAT) and subcutaneous adipose tissue (SAT) (Supplemental Table S1) [17]. The level of the L3 vertebra, which most clearly displayed both transverse processes, was selected for VAT and SAT evaluation [18].
The BEYOND trial was conducted in accordance with the Declaration of Helsinki (NCT02564926; clinicaltrials.gov) and received approval from the Institutional Review Board of Kyung Hee University Hospital at Gangdong (No. 2019-08-020-001). Due to the anonymization of study information, the requirement for informed consent was waived.
Study outcomes
Adequate glucose control was defined as either a reduction in HbA1c (calculated as HbA1c at week 52 minus baseline HbA1c) of at least 1.0% or the achievement of an HbA1c level of <7.0% at week 52 [19]. Significant WL was determined by a change in body weight, calculated as [(weight at week 52–baseline weight) divided by baseline weight], of ≥−3.0% [20]. Patients were subsequently divided into four groups (Supplemental Fig. S3): group 1, those who achieved both adequate GC and significant WL (GC+/WL+); group 2, those with adequate GC without significant WL (GC+/WL−); group 3, those with significant WL but inadequate GC (GC−/WL+); and group 4, those with neither adequate GC nor significant WL (GC−/WL−).
Statistical analysis
Data that follow a normal distribution are presented as mean±standard deviation, whereas data not following a normal distribution are presented as median (interquartile range). For continuous variables, either analysis of variance or Kruskal-Wallis tests were utilized, and for categorical variables, chi-square tests were employed. To assess changes in clinical values within individuals from baseline to follow-up, we used the paired t test or—for non-normally distributed data—the Wilcoxon signed-rank test.
Logistic regression analysis was performed to identify factors associated with the efficacy of dapagliflozin therapy. In the multivariate model, we included variables that had P values less than 0.1 in the univariate analyses and a variance inflation factor below 5.0. Backward stepwise selection was employed to determine significant covariates. To assess the predictive capabilities of the models for GC, WL, or both, the area under the receiver operating characteristic curve (AUC) was utilized. Pearson correlation analysis was applied to explore the relationship between changes in HbA1c levels and body composition parameters. A two-tailed P value of less than 0.05 was considered to indicate statistical significance. All statistical analyses were performed using SPSS version 22.0 (IBM Corp., Armonk, NY, USA).
RESULTS
Characteristics of the study population
The total study population (n=56) had a mean age of 54.4 years, with 64.3% male and a mean BMI of 27.0 kg/m2 (Table 1). The mean duration of diabetes was 6.0 years, and the initial HbA1c level was 8.0% (Table 1). Among all participants, between baseline and week 52, the HbA1c level decreased by an average of 1.0%±0.8%, while body weight declined by 2.4±3.1 kg (Supplemental Table S2).
Among the participants, 69.6% (n=39) achieved GC+, and 57.1% (n=32) experienced WL+. When categorized based on glycemic and weight control over a 52-week period (Supplemental Fig. S4), certain baseline characteristics differed significantly across the four groups. The GC+/WL+ group included 42.8% (n=24) of participants, while the GC+/WL− group comprised 26.8% (n=15). The GC−/WL+ classification encompassed 14.3% (n=8), while GC−/WL− applied to 16.1% (n=9).
Significant baseline differences were observed among the groups regarding diabetes duration, lean mass, diastolic blood pressure, fasting plasma glucose (FPG) levels, high-density lipoprotein cholesterol (HDL-C) levels, and estimated glomerular filtration rate (eGFR). Group 1 exhibited the shortest duration of diabetes, differing significantly from group 4 (4.6±4.2 years vs. 9.3±5.3 years, P=0.009). Group 2 also had a shorter duration of diabetes than group 4 (4.8±4.8 years vs. 9.3±5.3 years, P=0.022), along with higher FPG levels (178.3±41.7 mg/dL vs. 147.0±30.1 mg/dL, P=0.031) and lower HDL-C levels (43.1±8.5 mg/dL vs. 53.1±8.2 mg/dL, P=0.032). Regardless of WL, participants in groups 1 and 2, who demonstrated adequate glucose control, had a significantly shorter duration of diabetes than those in groups 3 and 4, who did not achieve such control. No significant baseline differences were observed between groups 3 and 4.
Changes in glycemic status and body composition over 52 weeks
Supplemental Table S1, Fig. 1, Supplemental Fig. S2 present the final values at week 52, along with the changes in glycemic status and body composition from baseline. The final HbA1c level decreased significantly, with groups 1 and 2 (both GC+) experiencing a greater reduction from baseline compared to group 4. Individuals classified as early responders, defined by a substantial HbA1c reduction by week 12, demonstrated a consistent decrease in HbA1c over the 52-week period (Supplemental Fig. S5A). Furthermore, significant reductions in FPG levels were observed in groups 1 and 2.
No significant differences in final body weight were observed among the groups. In groups 1 and 3 (both WL+), body weight gradually decreased over the 52 weeks, while it remained stable in the other two groups (Supplemental Fig. S5B). Individuals classified as early responders, who had lost weight at both week 4 and week 12, exhibited consistent weight reduction throughout the 52-week period (Supplemental Fig. S5B). Only group 2 displayed a significantly greater final lean mass compared to group 4. Groups 1 and 3 experienced reductions in weight, BMI, fat proportion, and fat mass when compared to group 4. However, a significant reduction in waist circumference was observed exclusively in group 3. Both groups 1 and 2 (GC+) demonstrated reductions in VAT area, regardless of body weight reduction, while groups 3 and 4 (GC−) did not (Supplemental Table S2).
Clinical determinants associated with glycemic response of dapagliflozin
In the univariate logistic analysis, factors including male sex, duration of diabetes, diastolic blood pressure, and HDL-C levels were associated with adequate GC (Table 2). Upon further examination in the multivariate logistic analysis, only male sex (odds ratio [OR], 4.17; 95% confidence interval [CI], 1.45 to 33.33; P=0.016) and a shorter duration of diabetes (OR, 0.77; 95% CI, 0.64 to 0.92; P=0.003) remained significant. Male sex also exhibited a positive correlation with reductions in HbA1c levels (r=0.32, P=0.017).
The model that incorporated diabetes duration had an AUC of 0.78 (95% CI, 0.65 to 0.90; P=0.001) (Supplemental Fig. S6). A diabetes duration cut-off of 6.9 years yielded a sensitivity of 70.6% and a specificity of 74.4% for predicting the achievement of an adequate GC. When participants were stratified by this duration, those with a longer history of diabetes had 6.96 times higher odds of achieving adequate GC (95% CI, 1.96 to 24.71; P=0.030; data not shown).
Clinical determinants of weight response to dapagliflozin
In both univariate (OR, 1.07; 95% CI, 1.01 to 1.12; P=0.013) and multivariate (OR, 1.07; 95% CI, 1.02 to 1.13; P=0.014) logistic regression models, eGFR was the only factor significantly associated with achieving significant WL (Table 3).
The AUC of eGFR was 0.72 (95% CI, 0.58 to 0.85; P=0.006) (Supplemental Fig. S7). A cut-off point of 96.2 mL/min/1.73 m2 yielded a sensitivity of 62.5% and a specificity of 66.7% for predicting the achievement of WL of at least 3%. When participants were stratified by an eGFR cut-off of 96.2 mL/min/1.73 m2, the OR for achieving adequate WL was 3.33 (95% CI, 1.10 to 10.12; P=0.034; data not shown).
Clinical determinants of dual response in glycemic control and weight reduction
A shorter duration of diabetes was the only significant factor associated with achieving both GC and WL (OR, 0.81; 95% CI, 0.68 to 0.97; P=0.023) (Supplemental Table S3). The AUC of the model that included diabetes duration was 0.76 (95% CI, 0.57 to 0.94; P=0.024) (Supplemental Fig. S8). A threshold of 5.4 years yielded a sensitivity of 77.8% and a specificity of 70.8% for predicting responders. Participants with a diabetes duration of less than 5.4 years were 8.5 times more likely to be responders (95% CI, 1.40 to 51.48; P=0.020; data not shown).
Association between changes in HbA1c levels and body composition parameters
Changes in HbA1c level (%) displayed a significant positive correlation with changes in VAT alone (Table 4, Supplemental Fig. S9A). No significant correlations were observed for other body composition parameters, including weight (Table 4, Supplemental Fig. S9B).
DISCUSSION
This post hoc study revealed heterogeneous effects of dapagliflozin on HbA1c and body weight responses in patients with type 2 diabetes. Male sex and a shorter duration of diabetes were independently associated with achieving adequate GC in response to dapagliflozin therapy. Additionally, a higher eGFR was significantly associated with more pronounced WL under dapagliflozin treatment. Moreover, a diabetes duration of less than 5.5 years continued to be a significant predictor for achieving both GC and WL. Furthermore, of the various changes in body composition, only reduced VAT was positively correlated with decreased HbA1c.
Recent studies have explored the efficacy of SGLT2i regarding the identification of individuals who respond well to treatment. A longitudinal prospective study conducted in Korea pinpointed those who responded to dapagliflozin regarding GC. The study suggested a link between higher baseline HbA1c levels, shorter duration of diabetes, and the use of dapagliflozin as an additional therapy with a relative reduction in HbA1c of more than 10% from baseline [21]. The extent of HbA1c reduction is strongly correlated with baseline HbA1c levels and tends to increase with higher initial HbA1c levels [22]. However, in our study, baseline HbA1c levels did not appear to significantly correlate with HbA1c reduction, which is inconsistent with the results of the previous study. The difference between these studies in how GC is defined may have contributed to this discrepancy. Importantly, the previous study may have included some patients with very high HbA1c levels due to the lack of specific criteria for baseline HbA1c levels and the inclusion of patients on insulin or triple therapy.
In this study, male sex was independently associated with a reduction in HbA1c levels. Preclinical animal experiments indicate that male mice may experience a more pronounced decrease in HbA1c when treated with SGLT2i, possibly due to increased SGLT2 expression in the kidneys compared to females [23]. Additionally, women typically have a lower plasma volume and a higher percentage of body fat than men, factors that contribute to a larger volume of distribution for lipophilic drugs such as SGLT2i [24]. Nevertheless, large clinical trials have not demonstrated significant sex-based differences in glycemic response to SGLT2i therapy [25,26], although limited data suggest a potential benefit for men [21,27].
Another key consideration for dapagliflozin treatment is WL. The factors contributing to WL following SGLT2i use did not consistently align with those influencing GC. One retrospective cohort study identified potential contributors to WL as age over 70 years, a BMI greater than 25 kg/m2, and the concurrent use of sulfonylureas [28]. However, the study’s participants were older and had a higher prevalence of comorbidities and various medications. In patients over 70 years old, WL might be attributed to sarcopenia, especially in the absence of information on body composition [29]. Moreover, the study did not investigate significant factors affecting WL that we identified, such as eGFR or the duration of diabetes. A pooled analysis of four tofogliflozin studies suggested that baseline body weight, uric acid level, male sex, and age could influence the degree of WL [27], although none of these factors were associated with WL in our study.
SGLT2i-mediated urinary glucose excretion (UGE) is dependent on renal function and decreases with renal impairment [30]. Consequently, a higher eGFR may contribute to superior glycemic response [27,31] and promote WL by creating a calorie deficit [32] due to the increased UGE resulting from SGLT2 inhibition. However, in our study, eGFR was only associated with WL, not with GC. Subgroup analyses revealed that patients with an eGFR ≥120 mL/min/1.73 m2 experienced a more pronounced reduction in HbA1c levels after dapagliflozin treatment [21]. However, the significance of eGFR disappeared after adjusting for various covariates. While the glucose-lowering effect of SGLT2i therapy is lessened in patients with an eGFR <60 mL/min/1.73 m2 [31], the average eGFR in our study was 96.3 mL/min/1.73 m2. This high average eGFR limited our ability to observe a significant difference among our participants, who had relatively well-preserved kidney function. Currently, no definitive evidence indicates that a high eGFR directly influences WL. However, some evidence suggests that greater WL following the initiation of SGLT2i treatment may be associated with a slower decline in eGFR [33]. Further research is required to clarify the relationship between eGFR and WL.
Our analysis stands out from previous investigations by considering the impact of dapagliflozin on both hyperglycemia and obesity. Consistent with other research [6,21], we found that a shorter duration of diabetes was a key prognostic factor for achieving GC and WL with dapagliflozin treatment. Since SGLT2i function independently of insulin secretion, they can be utilized at any stage of type 2 diabetes, from a recent diagnosis to long-standing disease. However, research indicates that the duration of diabetes is inversely related to the improvement in insulin sensitivity following SGLT2i administration [34]. This implies that patients in the earlier stages of diabetes, who may have better-preserved insulin sensitivity, could derive greater benefits from the glycemic and weight-lowering effects of SGLT2i [35]. Achieving target HbA1c levels early has been linked to sustained GC in those newly diagnosed with type 2 diabetes [36]. Similarly, our study indicated that patients who responded well to dapagliflozin treatment, as evidenced by reductions in HbA1c or WL at week 12, maintained these improvements over the course of 52 weeks. Therefore, if dapagliflozin does not elicit an adequate response shortly after initiation, it may be prudent to consider lifestyle modifications or the introduction of other antidiabetic medications.
Another clinical implication is the absence of an association between weight reduction and HbA1c reduction, which aligns with the results of prior studies [27,28]. Although both GC+/WL+ and GC+/WL− groups exhibited similar reductions in HbA1c, the latter did not achieve substantial WL despite having higher baseline HbA1c levels and body weight. This highlights the importance of consistent patient education regarding healthy diet and exercise, especially for individuals with both hyperglycemia and obesity. In certain patients, dapagliflozin may effectively regulate blood glucose levels while being insufficient for weight management, and vice versa.
The strength of our research lies in its analysis of body composition changes, extending beyond assessments of body weight or BMI and utilizing abdominal CT and DXA. To our knowledge, this study is the first to suggest that a decrease in VAT, rather than overall WL, is significantly associated with GC in patients receiving SGLT2i treatment. Notably, a reduction in VAT over a period of 52 weeks was significantly linked to a decrease in HbA1c levels. Supporting this finding, a meta-analysis has shown that SGLT2i reduce VAT, SAT, and ectopic liver fat more effectively than other antidiabetic drugs or placebo [26]. This effect is likely due to the caloric deficit induced by glycosuria, which prompts a metabolic shift from glucose to lipids substrates, leading to increased lipolysis and oxidation of free fatty acids [37,38]. Excess visceral fat is known to be associated with a reduced sensitivity of glucose uptake to insulin stimulation [39], and the visceral fat is positively correlated not only with the extent of fasting hyperglycemia but also with HbA1c levels [40]. In contrast, the surgical removal of visceral fat has been shown to improve metabolic parameters without detectable changes in body weight, fat mass, or lean body mass [41].
This post hoc analysis utilized reliable data from a previous randomized controlled trial to identify independent factors associated with treatment responses to dapagliflozin. However, several limitations should be acknowledged. First, factors such as food intake and energy expenditure, which can influence glycemia and weight, were not examined. Notably, WL resulting from chronic dapagliflozin treatment may be mitigated by compensatory hyperphagia [42]. Second, UGE and plasma volume measurements were not taken concurrently. Third, we lacked data on insulin resistance or sensitivity prior to and following dapagliflozin treatment, although the triglyceride/HDL-C ratio was used as a surrogate marker [43]. Additionally, the small sample size of the study may lack the statistical power necessary to detect differences between groups. Lastly, we did not consider the potential effects of other medications, such as those affecting the renin-angiotensin-aldosterone system, on renal hyperfiltration.
In conclusion, patients with type 2 diabetes exhibit varying responses to dapagliflozin treatment regarding GC and WL. Male sex and a shorter duration of diabetes were associated with the achievement of adequate GC, while higher eGFR was linked to greater WL. Shorter duration of diabetes emerged as a key factor in attaining both GC and WL. Patients who showed early responses to dapagliflozin treatment, evidenced by reductions in HbA1c or WL, were more likely to sustain these improvements over time. Our findings suggest that a reduction in visceral fat, which is specifically targeted by SGLT2i treatment, may have a more pronounced effect on HbA1c reduction than overall WL. Therefore, our results could assist in identifying patients with type 2 diabetes who are more likely to benefit from dapagliflozin therapy in a clinical context. However, these findings require further validation in a larger population.
XML Download