Introduction

Overweight/obesity confer additional comorbidities to individuals with type 2 diabetes mellitus, and most patients with type 2 diabetes are overweight or obese [1]. Recently, the American Medical Association recognised obesity as a complex disease associated with comorbidities, including type 2 diabetes and cardiovascular disease (CVD) [2]. In this regard, achieving weight loss of 5–10% can improve glycaemic control and CVD risk factors, including BP, HDL-cholesterol and triacylglycerol in patients with type 2 diabetes [1, 3–5]. In addition, clinically meaningful improvements in glycaemic control can be seen even with weight loss of 2–5% [5]. Unfortunately, it is difficult to both achieve and maintain weight loss, particularly as some anti-hyperglycaemic agents ([AHAs] e.g. sulfonylureas, insulin and pioglitazone) are associated with weight gain [1]. Even when lifestyle modifications lead to initial weight loss, weight regain is common [6–8]. Therefore, clinical strategies that help achieve and maintain weight loss remain unmet needs in type 2 diabetes management.

Newer AHAs, such as sodium glucose co-transporter 2 (SGLT2) inhibitors, have been proposed to address unmet clinical needs (e.g. weight loss) beyond improving glycaemic control. Canagliflozin is an SGLT2 inhibitor recently approved for the treatment of patients with type 2 diabetes that lowers the renal threshold for glucose and increases urinary glucose excretion (UGE). These actions result in decreased plasma glucose in patients with hyperglycaemia and a mild osmotic diuresis [9–12]. Canagliflozin as monotherapy and in combination with other AHAs improves glycaemic control and reduces body weight and BP in patients with type 2 diabetes [9–12].

Given the observed weight loss with canagliflozin, an unanswered question relates to the weight-loss-associated mechanisms with this therapy. We used a pooled dataset of four placebo-controlled Phase 3 studies to characterise the changes in body weight, HbA1c and systolic BP (SBP) observed with canagliflozin treatment and estimated the relative contributions of weight-loss-associated mechanisms to the overall reductions in HbA1c and SBP in patients with type 2 diabetes.

Methods

Study design, patient population and treatments

Data were pooled from patients with type 2 diabetes enrolled in four randomised, double-blind, placebo-controlled Phase 3 studies with similar designs. Each study had a primary assessment at 26 weeks and evaluated canagliflozin 100 and 300 mg vs placebo as monotherapy, add-on to metformin, add-on to metformin plus sulfonylurea, or add-on to metformin plus pioglitazone [9–12]. Eligible patients were aged 18–80 years, and generally had HbA1c ≥7.0% (53 mmol/mol) and ≤10.5% (91 mmol/mol) and estimated glomerular filtration rate ≥55 ml min−1 1.73 m−2 at screening.

All studies were conducted in accordance with ethical principles that comply with the Declaration of Helsinki and are consistent with Good Clinical Practice and applicable regulatory requirements. Study protocols and amendments were approved by ethical committees or institutional review boards at participating institutions. All patients provided written informed consent prior to participation.

Study assessments

Change in body weight, HbA1c and SBP were assessed at several time points between baseline and week 26 in each study [9–12]. The proportion of patients achieving >0%, ≥5% and ≥10% weight loss and changes in body weight by quartiles of weight loss at week 26 were assessed. The last observation carried forward (LOCF) approach was used to impute any missing data at week 26.

Statistical analyses

Mean per cent change from baseline in body weight and mean changes from baseline in HbA1c and SBP at week 26, as well as corresponding cumulative distribution plots, are reported for each variable. For the analysis of weight-loss-independent and weight-loss-associated effects of canagliflozin on HbA1c and SBP, patients within each treatment group were divided into deciles based on weight change, and mean change in HbA1c and SBP were calculated within each decile. ANCOVA models were used with HbA1c or SBP change as the response variable and body-weight change as a covariable. The slope of the relationship between body-weight change and HbA1c or SBP change was determined and the differences in slopes between each treatment group were calculated. If the slopes were not statistically different (p < 0.05) between treatment groups, a single slope was used for all groups. Weight-loss-independent components of HbA1c and SBP changes were calculated as placebo-subtracted least squares (LS) mean differences from ANCOVA models (i.e. between-group differences observed at the same weight loss). All analyses were performed in Matlab, v8.4 (MathWorks, Natick, MA, USA).

Results

Patients

Baseline demographic and disease characteristics are summarised in electronic supplementary material (ESM) Table 1. Patients had mild to moderate hyperglycaemia at baseline (mean HbA1c, 8.0% [64 mmol/mol] across groups). Of the patients included in these analyses, ≥30% of patients were overweight (BMI ≥25 and <30 kg/m2) and ≥50% of patients in each study were obese (BMI ≥ 30 kg/m2).

Changes in body weight, HbA1c and SBP

Dose-dependent reductions were seen in body weight, HbA1c and SBP with canagliflozin compared with placebo (Fig. 1a–c). Most patients treated with canagliflozin 100 and 300 mg had a reduction in HbA1c: 86% and 92%, respectively, compared with 55% of patients treated with placebo (Fig. 1d). Similarly, most canagliflozin-treated patients experienced weight loss with canagliflozin 100 and 300 mg: 82% and 85%, respectively, compared with 55% of placebo-treated patients (Fig. 1e and ESM Fig. 1). More patients treated with canagliflozin 100 and 300 mg vs placebo had ≥ 5% weight loss (25%, 33% and 6%, respectively) and ≥ 10% weight loss (3%, 3% and 1%, respectively; ESM Fig. 1). Body-weight reductions were seen in the highest three quartiles of weight loss with canagliflozin 100 and 300 mg compared with a small increase in the lowest quartile (ESM Fig. 2). Decreases in body weight vs placebo were seen with both canagliflozin doses across quartiles. The proportions of patients experiencing some reduction in SBP were also greater with canagliflozin 100 and 300 mg compared with placebo (64%, 67% and 50%, respectively; Fig. 1f).

Fig. 1
figure 1

Mean (SE) reduction in (a) HbA1c, (b) body weight and (c) SBP. White bars, placebo; grey bars, canagliflozin 100 mg; black bars, canagliflozin 300 mg. (d–f) The cumulative distribution of change from baseline values for each measure. In each of these figures, the y coordinate represents the percentage of patients whose change from baseline value is lower than the change from baseline value represented by the x coordinate. Light grey dashed line, placebo; medium grey dashed–dotted line, canagliflozin 100 mg; solid black line, canagliflozin 300 mg. To convert values for change in HbA1c in % into mmol/mol, multiply by 10.929. BW, body weight

Full size image

Effect of body-weight change on HbA1c and SBP

Patients with greater weight loss had greater reductions in HbA1c and SBP; slopes for the effect of weight loss on HbA1c and SBP were similar across groups (Fig. 2). Each 1% reduction in body weight was associated with a 0.045% (0.5 mmol/mol) reduction in HbA1c and a 0.62 mmHg reduction in SBP. The weight-loss-independent and weight-loss-associated effects of canagliflozin were evaluated using the ANCOVA models, as illustrated in Fig. 2. Approximately 85% of the placebo-subtracted reductions in HbA1c with canagliflozin were weight-loss-independent, and 15% were weight-loss-associated (Fig. 2). Approximately 58% of the placebo-subtracted reductions in SBP with canagliflozin were weight-loss-independent, with 42% being weight-loss-associated.

Fig. 2
figure 2

Relationship between changes in body weight and (a) HbA1c and (b) SBP at week 26. Larger open symbols represent the mean change for the entire treatment group, and smaller symbols represent mean changes within each decile of weight change; white triangles, placebo; grey squares, canagliflozin 100 mg; black circles, canagliflozin 300 mg. To convert values for change in HbA1c in % into mmol/mol, multiply by 10.929. BW, body weight

Full size image

Discussion

Most AHA therapies are considered to be either weight-loss neutral (e.g. biguanides, dipeptidyl peptidase-4 inhibitors, α-glucosidase inhibitors, bile acid sequestrants, bromocriptine) or to potentially increase body weight (e.g. sulfonylureas, insulin, glinides, pioglitazone) [1]. Thus, prior to the introduction of SGLT2 inhibitors, glucagon-like peptide-1 receptor agonists were the only class consistently associated with body-weight reduction in patients with type 2 diabetes [1]; however, patient-to-patient variability in weight loss has been observed [1, 13]. In this pooled analysis, we demonstrate that canagliflozin treatment provided dose-dependent reductions in body weight in patients with type 2 diabetes. Of note, patients in the highest weight-loss quartile had weight loss of ~7.4–8.0%, demonstrating some variability in canagliflozin-associated weight loss. In addition, consistent HbA1c reductions were observed, whereas SBP changes were more variable.

HbA1c and SBP reductions have previously been associated with weight loss in patients with type 2 diabetes [14]. In this analysis, greater HbA1c and SBP reductions with canagliflozin were seen in patients with greater body-weight reductions over 26 weeks and occurred through both weight-loss-associated and weight-loss-independent mechanisms. Most of the HbA1c-lowering effect of canagliflozin (~85%) was independent of weight loss, and is likely attributable to increased UGE. Weight loss contributed ~40% to the overall reduction in SBP seen with canagliflozin. The estimated effects of weight loss on HbA1c and SBP in this analysis are slightly lower than corresponding estimates obtained following intensive diet/exercise modification for 1 year in the Look AHEAD study (estimated mean reductions of 0.074% [0.8 mmol/mol] in HbA1c and 0.79 mmHg in SBP per each 1% reduction in body weight [7]). Differences in estimated effects of weight loss on HbA1c and SBP may be attributable to different study durations; a comparable analysis was not performed for canagliflozin at 52 weeks because the placebo groups in most of the canagliflozin studies were switched to an active comparator for the 26 week extension. In addition, these differences may also be related to potential study effects to reduce HbA1c and SBP independent of weight loss in the Look AHEAD study.

The exact mechanism of weight-loss-independent BP reduction is not completely understood, but may partially be related to the mild osmotic diuresis or alterations in sodium reabsorption. If osmotic diuresis was considered as a major mechanism, it might be expected that the BP-lowering effect would be less in patients with renal impairment. Interestingly, we have observed that in patients with stage 3 chronic kidney disease, canagliflozin was associated with less weight loss and HbA1c lowering, but similar SBP lowering. However, the contribution of weight loss to HbA1c and SBP reductions was similar to the present analysis (unpublished data).

In summary, we have provided novel observations regarding the clinical effect of SGLT2 inhibition. As previously reported, canagliflozin treatment was associated with reductions in body weight, HbA1c and SBP. However, in this report, we provide data that show reductions in HbA1c and SBP with canagliflozin occurred by both weight-loss-independent and weight-loss-associated mechanisms, with a greater proportion of the reduction in HbA1c, relative to SBP, determined to be independent of weight loss.