Clinical Factors Associated with Initial Decrease in Body-Fat Percentage Induced by Add-on Sodium-Glucose Co-transporter 2 Inhibitors in Patient with Type 2 Diabetes Mellitus

Patients with T2DM who visited the Diabetes Care Center of Jinnouchi Hospital between April 2014 and December 2015 and were treated with SGLT2 inhibitors in addition to their ongoing medications were retrospectively registered. We measured body composition using a body composition analyzer (InBody770; Biospace, Seoul, Korea) both before administration of SGLT2 inhibitors and 4 weeks after treatment and calculated the change in composition. Written informed consent was obtained from all patients. The study was conducted in accordance with the Declaration of Helsinki and the Human Ethics Review Committee of Jinnouchi Hospital approved the study protocol.

Elementary body composition was measured using a direct, segmental, multi-frequency bioelectrical impedance analyzer (InBody770; Biospace, Seoul, Korea). This analyzer processes 30 impedance measurements using six different frequencies (1, 5, 50, 250, 500, and 1000 kHz) at each of the five body segments (right and left arms, right and left legs, and trunk) and 15 reactance measurements using tetrapolar 8-point tactile electrodes at three different frequencies (5, 50, and 250 kHz) at each of the same five body segments [7, 8]. Body composition analyses were conducted with the subjects clothed as specified by the hospital and performed once before starting SGLT2 inhibitors and again after 4 weeks of treatment.

Measurement of glycated hemoglobin (HbA1c) and plasma glucose is routinely performed every month in patients with diabetes mellitus (DM). Moreover, biochemical examinations of blood serum chemistry are usually performed before and after newly started medications including SGLT2 inhibitors to determine drug-induced side effects. Blood samples for HbA1c, total cholesterol, low-density lipoprotein (LDL) cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides, and estimated glomerular filtration ratio (eGFR) were collected from the antecubital vein before SGLT2 inhibitor therapy and analyzed in the Jinnouchi Hospital laboratory. In our clinic body composition in all patients starting SGLT2 inhibitors is routinely measured at initiation and after 4 weeks of SGLT2 inhibitor therapy.

The Shapiro–Wilk test was used to assess the normal distribution of continuous data. Data were expressed as mean ± standard deviation (SD), and data with skewed distributions were expressed as the median value with interquartile range. Categorical data were presented as frequencies and percentages. Differences between paired samples were analyzed by the t test or Wilcoxon signed-rank test; the differences between two groups were analyzed by the t test, Mann–Whitney U test, or Fisher’s exact test as appropriate. Add-on SGLT2 inhibitor treatment significantly reduced body weight, total fat quantity, and Fat %. However, body weight change and body fat quantity change did not show a strong correlation. Therefore, we used Fat % as the effective parameter in this study. Calculation of the Pearson product-moment correlation coefficient was used to examine the relationships between the weight change with SGLT2 inhibitors and the amount of Fat % change. Multivariate regression analysis was conducted using the significant outcomes by the Pearson product-moment correlation analysis, age, sex, and body mass index (BMI). Median Fat % reduction after add-on SGLT2 inhibitor treatment was − 0.4%. Thus we defined a greater response in fat reduction following SGLT2 inhibitors as more than 0.4% reduction in Fat %, which we analyzed using simple logistic regression analysis with the greater fat reduction as the dependent variable and the various parameters as independent variables. To investigate the clinically associated factors with the greater body fat reduction following SGLT2 inhibitor medication, multivariate logistic regression analysis was conducted among the significant factors of simple logistic regression analysis. The Hosmer–Lemeshow test was conducted to investigate the goodness of fit of the logistic regression model. Receiver-operating characteristic (ROC) curve analysis was conducted to calculate a cut-off value of pretreatment levels of HbA1c for greater body fat reduction. A p-value < 0.05 denoted statistical significance. Statistical analyses were performed using SPSS version 23 (SPSS Inc., Tokyo, Japan).

A total of 175 patients with T2DM were enrolled and all patients had body composition measured at the start of the SGLT2 inhibitor add-on therapy, but we were unable to measure body composition in 27 patients at the end of 4 weeks. Thus, 27 patients were excluded. Table 1 shows the baseline characteristics of the participants. SGLT2 inhibitors were newly added to the ongoing medications as follows: dapagliflozin (n = 46, 30.9%), tofogliflozin (n = 45, 30.4%), luseogliflozin (n = 35, 23.6%), ipragliflozin (n = 20, 13.5%), and empagliflozin (n = 2, 1.4%). Combined antidiabetic medications were insulin (n = 44, 29.7%), glucagon-like peptide-1 (GLP-1) receptor analog (n = 13, 8.8%), sulfonylurea (n = 63, 42.6%), dipeptidyl peptidase 4 inhibitor (n = 48, 32.4%), metformin (n = 71, 48.0%), thiazolidine (n = 15, 10.1%), glinide (n = 12, 8.1%), and alpha glucosidase inhibitor (n = 24, 16.2%). Among the included patients, 85 (57.4%) had hypertension and 127 (85.4%) had dyslipidemia. They were treated with calcium antagonists (n = 63, 42.6%), angiotensin II receptor blockers (n = 75, 50.7%), angiotensin-converting enzyme inhibitors (n = 2, 1.4%), beta blockers (n = 25, 26.4%), statins (n = 76, 51.4%), and ezetimibe (n = 18, 12.2%).

Add-on SGLT2 inhibitor treatment for 4 weeks significantly reduced body weight (− 1.04 ± 1.18 kg, p < 0.01), total fat quantity (− 0.62 ± 1.19 kg, p < 0.01), and Fat % (− 0.4 ± 1.4%, p < 0.01). Figure 1 shows the scatter plot diagram of the changes in total body weight and body fat. The body fat changes were significantly correlated with the changes in total body weight, but the correlation was not strong (r = 0.44, p < 0.001). In Pearson’s correlation analysis, reduction in body fat quantity was not statistically significantly correlated with baseline body weight (r = 0.07, p = 0.38) and baseline BMI (r = 0.05, p = 0.57).

Table 2 shows the results of the simple linear regression analysis conducted for changes in Fat % with various parameters. Pretreatment levels of HbA1c (r = − 0.20, p = 0.015) and estimated eGFR (r = − 0.23, p = 0.005) showed a significantly negative relationship with changes in Fat %.

Table 3 shows the results of multivariable regression analysis for changes in Fat % after add-on SGLT2 inhibitor treatment. Pretreatment levels of HbA1c and eGFR were significant and independent factors (pretreatment levels of HbA1c: β = − 0.1783, p = 0.0319; eGFR: β = − 0.2182, p = 0.0072).

We defined the greater body fat reduction group as having a reduction of > 0.4% in Fat % (median) after add-on SGLT2 inhibitor treatment. Table 4 shows the results of the simple logistic regression analysis for greater body fat reduction. Pretreatment HbA1c levels as the continuous variable [odds ratio (OR) 1.54; 95% confidence interval (CI) 1.13–2.11, p < 0.01] and smoking (OR 2.54; 95% CI 1.18–5.47, p = 0.02) were significant factors for greater body fat reduction with SGLT2 inhibitors. The multivariate logistic regression analysis demonstrated that the pretreatment HbA1c levels (OR 1.61; 95% CI 1.16–2.25; p < 0.01) and smoking (OR 2.65; 95% CI, 1.14–6.15; p = 0.02) were significantly associated factors for greater body fat reduction with SGLT2 inhibitors.

We compared the percentage changes in body fat after add-on SGLT2 inhibitor treatment between the higher HbA1c group (pretreatment levels of HbA1c ≥ 7.7%) and the lower HbA1c group (pretreatment levels of HbA1c < 7.7%). Figure 2 shows that the reduction in Fat % was significantly greater in the higher HbA1c group than the lower HbA1c group (higher HbA1c group − 0.7 ± 1.3% vs. lower HbA1c group − 0.1 ± 1.4%, p = 0.01). However, the reduction in body weight was not significantly different among the two groups, as shown in Fig. 3. In other body composition results, the reduction in water (higher HbA1c group − 0.02 ± 0.86 kg vs. lower HbA1c group − 0.56 ± 0.94 kg, p < 0.01) and skeletal muscle (higher HbA1c group −0.01 ± 1.09 kg vs. lower HbA1c group −0.68 ± 1.19 kg, p = 0.04) was significantly greater in the lower HbA1c group than the higher HbA1c group. However, the reduction in fat (higher HbA1c group − 0.85 ± 1.16 kg vs. lower HbA1c group − 0.46 ± 1.19 kg, p = 0.04) was significantly greater in the higher HbA1c group than the lower HbA1c group.

A comparison of the changes in Fat % following add-on SGLT2 inhibitor treatment between obese (BMI ≥ 25 kg/m²) and non-obese (BMI < 25 kg/m²) groups showed no significant difference between the two groups (obese group − 0.4 ± 1.5% vs. non-obese group − 0.5 ± 1.0%, p = 0.69).

In obese patients with good diabetic control (BMI ≥ 25 kg/m² and pre-treatment levels of HbA1c < 7.0%), SGLT2 inhibitor therapy significantly decreased the levels of HbA1c (from 6.40 ± 0.15% to 6.29 ± 0.22%, p = 0.03). We found no significant changes in the quantity of body fat (from 27.88 to 27.49 kg, p = 0.14) and Fat % (from 35.38 to 35.29%, p = 0.77).

The multivariate logistic regression analysis with a nominal variable of the pretreatment HbA1c levels (higher HbA1c ≥ 7.7% and lower HbA1c < 7.7%) also demonstrated that smoking and higher pretreatment levels of HbA1c (≥ 7.7%) were independently and significantly correlated with the greater body fat reduction, as shown in Table 5.

In the non-smoker group, pretreatment levels of HbA1c were also independently and significantly correlated with greater body fat reduction as shown in Table 6 (OR 1.56; 95% CI 1.05–2.37; p = 0.03).