Liver function effects of SGLT2 inhibitor and GLP‐1 receptor agonist combination treatment in patients with type 2 diabetes (post hoc analysis of RECAP study)

ABSTRACT

Aim

The favor effect on liver disease by odium‐glucose cotransporter inhibitor (SGLT2i) and GLP‐1 receptor agonist (GLP1Ra) was reported; however, the effect of the combination treatment of these drugs was not well known.

Methods

We retrospectively analyzed data for 643 patients with type 2 diabetes receiving SGLT2i + GLP1Ra combination treatment for at least 1 year (331 and 312 patients in the GLP1Ra‐ and SGLT2i‐preceding groups, respectively). Propensity score (PS) matching was used to compare the effects of the preceding drugs on liver function.

Results

The mean AST and ALT values at baseline, at the initiation of combination treatment, and at final observation were 29.8 ± 20.0 and 37.7 ± 29.5, 28.7 ± 17.3 and 35.3 ± 6.0, 26.0 ± 14.6 and 30.1 ± 21.6 IU/L, respectively, indicative of significant improvements in liver function (P < 0.001). Conversely, significant progress in the fibrosis‐4 (FIB‐4) index category was observed even after the combination treatment (P = 0.03). Subgroup analysis revealed that a significant decrease in ALT was observed only in patients with a baseline ALT ≥30 IU/L after the combination treatment (P = 0.005). Improvement of the FIB‐4 index category was observed in patients in the baseline FIB‐4 index ≥2.6 group and in the 1.3 ≤FIB‐4 index <2.6 group (46% and 19%, respectively). The matched model showed no significant differences in liver function after combination treatment between the SGLT2i‐ and GLP1Ra‐preceding groups.

Conclusions

SGLT2i + GLP1Ra combination treatment significantly improved liver dysfunction and prevented the progression of FIB‐4 index category among patients with an FIB‐4 index ≥1.3.

The combination treatment of SGLT2 inhibitor and GLP‐1 receptor agonist significantly improved liver function markers. This combination treatment also improved the FIB‐4 index category in patients with the baseline FIB‐4 index ≥1.3.

INTRODUCTION

Several cardiovascular outcome trials (CVOTs) using sodium‐glucose cotransporter inhibitors (SGLT2is) have demonstrated that these drugs have significantly superior cardiovascular ¹ , ² , ³ and renal ¹ , ² , ³ , ⁴ outcomes than placebos, resulting in a rise in their use in clinical practice in recent years. CVOTs have also shown that glucagon‐like peptide‐1 receptor agonists (GLP1Ras) also show superior efficacy relative to placebos in terms of cardiovascular outcomes ⁵ , ⁶ , ⁷ , ⁸ . SGLT2is and GLP1Ras also decrease body weight (BW) and blood pressure (BP). The observed improvements in cardiovascular and renal outcomes may be related to these metabolic effects. The executive summary of the Kidney Disease: Improving Global Outcomes 2022 Clinical Practice Guideline for Diabetes Management in Chronic Kidney Disease recommends metformin and SGLT2i as first‐line treatment for patients with type 2 diabetes and chronic kidney disease ⁹ and GLP1Ra as an additional therapeutic agent in patients who fail to achieve glycemic control, are unable to use first‐line drugs, or require weight loss promotion. In addition to glycemic control, BW control in patients with type 2 diabetes is challenging. Thus, combination treatment using SGLT2i and GLP1Ra has become more common in clinical practice.

In a retrospective study, Kobayashi et al. ¹⁰ reported that 52% of patients receiving GLP1Ra treatment are also administered SGLT2is. Further, the RECAP study, which focused on the renoprotective effects of SGLT2i and GLP1Ra combination treatment in patients with type 2 diabetes, especially with respect to the preceding medication, reported reduced estimated glomerular filtration rate (eGFR) decline ¹¹ , with the preceding drug showing no effect on the renoprotective activity of the second drug. This study also revealed that GLP1Ra‐preceding patients show a greater decrease in BW than their SGLT2i‐preceding counterparts ¹² . Additionally, the average body mass index (BMI) of the included patients was 29.6, and BW management was beneficial not only for glycemic control but also for controlling metabolic disorders, including liver dysfunction. The beneficial effects of SGLT2i or GLP1Ra in most common chronic liver diseases, such as metabolic dysfunction‐associated steatotic liver disease (MASLD) have been previously reported ¹³ , ¹⁴ , ¹⁵ , ¹⁶ .

However, there are still a few studies on the potential beneficial impact of combined SGLT2i and GLP1Ra treatment on liver function. Therefore, in this study, as a post hoc analysis of the RECAP study, we evaluated the effect of the combination treatment of SGLT2i and GLp1Ra on liver function and determined which subgroup of patients, SGLT2i‐preceding or GLP1Ra‐preceding patients, showed better liver function outcomes.

MATERIALS AND METHODS

Study participants and data collection

The study design and participants were as described in a previous study ¹⁰ . In brief, we conducted a retrospective survey of patients with type 2 diabetes treated with SGLT2is and GLP1Ra (Figure S1). Data for patients who visited clinics or hospitals where our research team members are affiliated between April 2010 and December 2021 were retrieved from hospital medical records. Specifically, we included patients who had been on one of the two drugs for at least 6 months and on the combination treatment for at least 12 months. The collected data included age, sex, height, BW, systolic BP (SBP), diastolic BP (DBP), eGFR, glycated hemoglobin A_1c (HbA_1c), aspartate aminotransferase (AST), alanine aminotransferase (ALT), and platelet count. Data on urinalysis (urinary albumin/creatinine ratio [mg/g Cr] or qualitative proteinuria) and concomitant medications (e.g., hypoglycemic and antihypertensive agents, statins) were also included. Further, we collected data at baseline, at the initiation of the combination treatment, and at final observation. eGFR was calculated as follows: eGFR (mL/min/1.73 m²) = 194 × age^−0.287 × serum creatinine^−1.094 (×0.739 for female patients) ¹⁷ , and qualitative proteinuria values were converted to albuminuria values using the formula reported by Sumida et al. ¹⁸ Fibrosis 4 (FIB‐4) index, which is a blood‐based diagnostic test index that is used to assess underlying fibrosis and can also be used to determine MASLD/metabolic dysfunction‐associated steatohepatitis (MASH) status, was determined from the platelet count and AST and ALT values of the patients according to Equation (1) below as previously reported ¹⁹ .

$$\mathrm{FIB}-4\text{index}=\frac{\mathrm{AST}\left(\mathrm{IU}/\mathrm{L}\right)\times \mathrm{Age}\left(\text{years}\right)}{\sqrt{\text{Plate}l\mathrm{et}\text{count}\left({10}^9\left(\mathrm{IU}/\mathrm{L}\right)\right)\times \mathrm{ALT}\left(\mathrm{IU}/\mathrm{L}\right)}}$$ (1)

We also examined the hepatic steatosis index(HIS): HSI was calculated as follows: HSI = 8 × (ALT/AST ratio) + BMI (+2, if female; +2, if diabetes) ¹⁹ .

The exclusion criteria were as follows: patients with type 1 DM or severe liver dysfunction (e.g., cirrhosis or severe infection), patients requiring chronic dialysis, patients with terminal‐stage malignancy, pregnant women, patients who discontinued treatment, and patients who opted out during the study.

Eighteen medical facilities participated in this study, and data were collected for 688 patients (Figure S2). Of these, 45 patients were excluded, while the remaining 643 patients (312 and 331 in the SGLT2i‐ and GLP1Ra‐preceding groups, respectively) were included in the full analysis set (FAS). Further, 113 patients in the FAS, who lacked essential clinical data or data on AST level, ALT level, or platelet count, were excluded, while the remaining 530 patients (276 and 254 in the SGLT2i‐preceding and GLP1Ra‐preceding groups, respectively) were analyzed as the complete case analysis set (CCA).

The duration from baseline to the initiation of the combination treatment was 29.0 ± 19.6 months, that from baseline to the final observation was 33.8 ± 17.1 months, and the total treatment duration was 61.8 ± 24.0 months.

Statistical analyses

Normally distributed data were presented as the mean ± standard deviation (SD), while skewed data were presented as the median [25th percentile, 75th percentile]. SPSS v28.0 (IBM Inc., Armonk, NY, USA) was used for statistical analyses, and statistical significance was set at P < 0.05.

Missing value analysis

As previously described ¹² , we performed multiple imputations (MI) using the chained equation method. One hundred complete filled‐in datasets were created, and each missing value was replaced. A breakdown of the missing data is shown in Figure S3.

Comparison of data corresponding to three time points (at baseline, initiation of combination treatment, and final observation)

A generalized linear mixed model was used to compare data corresponding to the three abovementioned time points. Bonferroni correction was applied for multiple comparisons between these time points. In clinical trials for MASLD/MASH diagnosis, patients with FIB‐4 score <1.3 were classified as not having advanced fibrosis, while those with FIB‐4 scores >2.67 were classified as having advanced fibrosis ²⁰ . Thus, the category of the FIB‐4 index; <1.3, 1.3–2.67, and >2.67, was used in the analysis of this study, and the Friedman test was used for the comparison between the three points. For subgroup analysis, the patients were divided into three groups based on the FIB‐4 index categories at baseline. Liver biopsies for the diagnosis of MASLD/MASH are rarely examined in clinical practice, and the FIB‐4 index may not be an appropriate surrogate marker for MASLD/MASH, especially in patients with mild liver damage. Further, it was recently reported that ALT ≥30 IU/L may be indicative of liver damage ²⁰ . Thus, our second subgroup analysis was based on ALT values at baseline according to the following groups: ALT ≥30 (IU/L), and ALT <30 (IU/L).

Propensity score matching analysis

For each dataset constructed via MI, the propensity score (PS) for the SGLT2i‐preceding group was calculated via logistic analysis using covariates, such as age, sex, height, weight, BMI, SBP, DBP, eGFR, HbA_1c, AST, ALT, platelet count, FIB‐4 index count, history of type 2 diabetes, use of concomitant medications at baseline, durations of treatment with a preceding drug (SGLT2i or GLP1Ra), and combination treatment.

As each patient's PS was calculated for each dataset constructed via MI, the mean PS was used as the representative value. Using these representative PS values, we performed PS matching via 1:1 nearest neighbor matching with a caliper value of 0.047 (0.2‐fold the SD of the PS) ²¹ without replacement. Comparisons of clinical characteristics between the two groups were performed using the unpaired t‐test for parametric variables, the Mann–Whitney rank‐sum test for nonparametric variables, and the chi‐square test for categorical data in the unmatched cohort model. Additionally, for the PS‐matched cohort model, we performed paired t‐tests for parametric variables, Wilcoxon signed‐rank test for nonparametric variables, and McNemar's test for categorical data. Using Rubin's rule, the estimated effects in each imputed dataset were averaged to obtain the overall estimated effects ²² , which account for variations in results across imputed datasets and reflect the uncertainty associated with missing data ²³ . Further, for sensitivity analysis, the same PS‐matched analysis was performed using the CCA set.

Multivariable linear regression analysis for changes in ALT and FIB‐4 index values

Using the CCA, we performed a multivariable linear regression analysis to evaluate independent predictors of changes in ALT (ΔALT) and ΔFIB‐4 index values. This analysis included the following clinical parameters as covariates: sex, history of type 2 diabetes, age, BW, mean arterial pressure (MAP), HbA_1c, eGFR, AST, ALT, FIB‐4 index at baseline, concomitant medications, the type of SGLT2is and GLP1Ras, ΔBW, ΔHbA_1c, ΔMAP, durations of treatment with the preceding SGLT2i or GLP1Ra, and the combination treatment.

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964, as revised in 2013, and were approved by the Institutional Review Board for Clinical Research, Tokai University, Japan (approval on December 6, 2021). This is a retrospective study: informed consent was waived.

RESULTS

Clinical characteristics of the included patients

The included patients (282 females and 361 males, i.e., 44% and 56%, respectively) had a mean age and height of 56.1 ± 13.1 years and 163.3 ± 9.7 cm, respectively, at baseline. The proportion of patients with a history of type 2 diabetes for >10 years was 81%. At final observation, six SGLT2is (ipragliflozin [n = 67, 10%], dapagliflozin [n = 158, 25%], tofogliflozin [n = 69, 11%], luseogliflozin [n = 32, 5%], canagliflozin [n = 67, 10%], and empagliflozin [n = 147, 23%]) and four GLP1Ras (liraglutide [n = 214, 33%], dulaglutide [n = 246, 38%], lixisenatide [n = 9, 1%], and exenatide [n = 8, 1%]) had been administered. The drug type was changed for 103 (16%) receiving SGLT2is and 166 (26%) receiving GLP1Ras.

Comparison of patient clinical characteristics at baseline, at the initiation of combination treatment, and final observation

Significant decreases in BW, BMI, SBP, DBP, MAP, and eGFR were observed after combination treatment initiation (P < 0.001, P < 0.001, P < 0.001, P < 0.05, P < 0.001, and P < 0.001, respectively; Table 1). Further, SGLT2i‐ and GLP1Ra‐only treatments did not result in any significant changes in HbA_1c, AST, and ALT levels; however, combination treatment decreased HbA_1c, AST, and ALT levels (P < 0.001), and relative to the baseline value. Regarding the FIB‐4 index category, there was a significant difference between the baseline, at addition, and final observation (P = 0.03). There was a significant difference in the HIS at baseline, at the time of additional measurement, and at the final observation (P < 0.001).

Table 1.

Comparison of the clinical characteristics of 643 patients between at baseline, at the time of the addition, and at the final observation time

	At baseline	At the time of the addition	At the time of final observation	P‐values
				Baseline vs the addition	Baseline vs the final observation	The addition vs the final observation
BW (kg)	79.4 ± 19.1	78.0 ± 18.6	74.9 ± 18.1	<0.001	<0.001	<0.001
BMI	29.6 ± 6.0	29.1 ± 5.7	27.9 ± 5.5	<0.001	<0.001	<0.001
SBP (mmHg)	133.7 ± 18.7	131.5 ± 17.7	128.8 ± 16.2	0.005	<0.001	0.001
DBP (mmHg)	77.6 ± 13.0	76.3 ± 12.2	74.7 ± 12.5	0.01	<0.001	0.005
MAP (mmHg)	96.3 ± 13.2	94.7 ± 12.2	92.7 ± 12.0	0.002	<0.001	<0.001
HbA1c (mmol/mol [%])	72.3 ± 18.0 (8.8 ± 1.6)	70.7 ± 16.5 (8.6 ± 1.5)	63.6 ± 16.2 (8.0 ± 1.5)	0.11	<0.001	<0.001
eGFR (mL/min/1.73 m2)	78.5 ± 27.4	74.0 ± 26.5	70.8 ± 6.8	<0.001	<0.001	<0.001
AST(IU/mL)	29.8 ± 20.0	28.7 ± 17.3	26.0 ± 14.6	0.30	<0.001	<0.001
ALT(IU/mL)	37.7 ± 29.5	35.3 ± 16.0	30.1 ± 21.6	0.03	<0.001	<0.001
Platelet(109/L)	24.5 ± 7.5	24.6 ± 7.3	24.4 ± 7.2	0.79 between three points
Category of FIB‐4 index; <1.3/1.3–2.67/2.67<	398 (62%)/208 (32%)/37 (6%)	389 (60%)/213 (33%)/41 (7%)	369 (57%)/232 (36%)/42 (7%)	0.03 between three points*
Hepatic steatosis index	42.4 ± 7.3	41.6 ± 7.3	39.9 ± 7.3	<0.001	<0.001	<0.001

Values are mean ± SD or n (n/total, %). P‐values were calculated by mixed general linear model with Bonferroni's correction. Missing data information is demonstrated in Figure S3.

^* P‐value was calculated by the Friedman test.

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BW, body weight; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration; FIB‐4, Fibrosis‐4; HbA_1c, glycated hemoglobin A_1c; MAP, mean arterial pressure; SBP, systolic blood pressure; SD, standard deviation.

Subgroup analysis based on the category of the FIB‐4 index

AST and ALT levels were significantly lower at final observation than at baseline for both groups (P = 0.001 for comparing AST levels in both groups for patients with FIB‐4 index ≥1.3 at baseline; other comparisons, P < 0.001; Table 2). Regarding the FIB‐4 index category, 18% of patients with FIB‐4 <1.3 and 8% of those with 1.3 ≤FIB‐4 index ≤2.67 worsened, while 46% of those with FIB‐4 >2.67, and 19% of those with 1.3 ≤FIB‐4 index ≤2.67 improved after combination treatment with SGLT2i and GLP1Ra.

Table 2.

Comparison of the clinical characteristics between at baseline, at the time of the addition, and at the final observation among three groups of patients with FIB‐4 <1.3, 1.3 ≤FIB‐4 index ≤2.67 and FIB‐4 index >2.67 at baseline

	At baseline	At the time of the addition	At the time of final observation	P‐values
				Baseline vs the addition	Baseline vs the final observation	The addition vs the final observation
FIB‐4 index <1.3 at baseline (n = 398)
BW (kg)	82.8 ± 20.4	81.7 ± 19.6	78.5 ± 19.1	0.08	<0.001	<0.001
BMI	30.3 ± 6.4	29.9 ± 6.1	28.8 ± 5.9	0.01	<0.001	<0.001
SBP (mmHg)	134.7 ± 18.6	131.7 ± 17.4	129.8 ± 16.2	0.002	<0.001	0.03
DBP (mmHg)	79.8 ± 12.6	78.5 ± 11.8	77.0 ± 11.7	0.11	<0.001	0.02
MAP (mmHg)	98.1 ± 13.1	96.2 ± 12.0	94.6 ± 11.6	0.009	<0.001	0.01
HbA1c (mmol/mol [%])	72.4 ± 18.6 (8.8 ± 1.7)	71.6 ± 16.7 (8.7 ± 1.5)	64.4 ± 17.0 (8.0 ± 1.6)	1.0	<0.001	<0.001
eGFR (mL/min/1.73 m2)	85.6 ± 27.6	81.0 ± 26.8	77.3 ± 27.0	<0.001	<0.001	<0.001
AST(IU/mL)	26.4 ± 16.2	27.3 ± 16.2	24.8 ± 14.5	0.35	<0.001	<0.001
ALT(IU/mL)	37.9 ± 29.9	37.3 ± 27.8	31.8 ± 23.5	0.33	<0.001	<0.001
Platelet(109/L)	27.6 ± 7.2	27.2 ± 7.0	26.8 ± 6.8	0.89 between three points
Category of FIB‐4 index; <1.3/1.3–2.67/2.67<	398 (100%)/0 (0%)/0 (0%)	347 (87%)/51 (13%)/0 (0%)	326 (82%)/65 (16%)/7 (2%)	<0.001 between three points*
Hepatic steatosis index	44.1 ± 7.7	43.3 ± 7.7	41.5 ± 7.6	0.002	<0.001	<0.001
1.3 ≤FIB‐4 index ≤2.67 at baseline (n = 208)
BW (kg)	75.8 ± 14.1	73.9 ± 14.3	71.2 ± 14.0	<0.001	<0.001	<0.001
BMI	28.8 ± 4.4	28.1 ± 4.4	27.1 ± 4.3	<0.001	<0.001	<0.001
SBP (mmHg)	132.4 ± 17.4	132.1 ± 17.6	128.6 ± 16.4	1.0	0.02	0.04
DBP (mmHg)	74.9 ± 12.2	73.6 ± 12.3	72.2 ± 11.6	0.52	0.01	0.45
MAP (mmHg)	94.0 ± 12.1	93.1 ± 12.3	91.0 ± 11.5	0.81	0.004	0.09
HbA1c (mmol/mol [%])	71.4 ± 14.8 (8.7 ± 1.4)	68.6 ± 14.6 (8.4 ± 1.3)	62.2 ± 14.0 (7.9 ± 1.3)	0.07	<0.001	<0.001
eGFR (mL/min/1.73 m2)	69.6 ± 20.0	64.5 ± 20.0	62.5 ± 19.7	<0.001	<0.001	0.10
AST (IU/mL)	34.2 ± 17.1	32.1 ± 17.0	30.5 ± 16.6	0.15	0.002	0.15
ALT(IU/mL)	38.7 ± 25.4	35.1 ± 25.1	33.1 ± 24.6	0.07	0.001	0.63
Platelet (109/L)	20.5 ± 3.7	20.9 ± 3.7	21.1 ± 3.6	0.23	0.03	0.95
Category of FIB‐4 index; <1.3/1.3–2.67/2.67<	0 (0%)/208 (100%)/0 (0%)	40 (19%)/151 (73%)/17 (8%)	40 (19%)/152 (73%)/16 (%)	0.002 between three points*
Hepatic steatosis index	39.8 ± 5.7	39.0 ± 5.7	37.5 ± 5.7	0.01	<0.001	<0.001
IB‐4 index >2.67 at baseline (n = 37)
BW (kg)	73.0 ± 12.1	70.7 ± 11.9	67.8 ± 11.8	0.04	<0.001	0.007
BMI	28.2 ± 3.6	27.3 ± 3.6	26.1 ± 3.5	0.02	<0.001	0.006
SBP (mmHg)	130.5 ± 18.0	131.6 ± 16.3	127.2 ± 15.7	1.0	0.88	0.62
DBP (mmHg)	71.9 ± 11.8	72.8 ± 11.1	70.3 ± 10.8	0.99	0.97	0.68
MAP (mmHg)	91.3 ± 12.0	92.4 ± 11.0	89.3 ± 10.7	0.99	0.91	0.47
HbA1c (mmol/mol [%])	72.4 ± 18.1 (8.8 ± 1.7)	73.0 ± 17.2 (8.8 ± 1.6)	61.4 ± 16.9 (7.8 ± 1.5)	1.0	0.007	0.002
eGFR (mL/min/1.73 m2)	64.8 ± 17.9	61.5 ± 17.6	61.8 ± 17.6	0.27	0.44	1.00
AST (IU/mL)	57.2 ± 27.8	49.5 ± 26.9	45.0 ± 26.8	0.14	0.01	0.70
ALT (IU/mL)	49.9 ± 25.8	42.0 ± 24.7	36.8 ± 24.6	0.13	0.005	0.49
Platelet (109/L)	15.2 ± 3.6	14.9 ± 3.4	14.1 ± 3.4	1.0	0.14	0.34
Category of FIB‐4 index; <1.3/1.3–2.67/2.67<	0 (0%)/0 (0%)/37 (100%)	2 (5%)/10 (27%)/25 (68%)	2 (5%)/15 (41%)/20 (54%)	<0.001 between three points*
Hepatic steatosis index	38.3 ± 5.9	37.3 ± 5.9	36.0 ± 5.9	0.18	<0.001	0.06

Values are mean ± SD or n (n/total, %). P‐values were calculated by mixed general linear model with Bonferroni's correction.

^* P‐value was calculated by the Friedman test.

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BW, body weight; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration; FIB‐4, Fibrosis‐4; HbA_1c, glycated hemoglobin A_1c; MAP, mean arterial pressure; SBP, systolic blood pressure; SD, standard deviation.

Subgroup analysis based on an ALT level of 30 IU/L

Patients with ALT<30 IU/L at baseline did not show any significant changes in AST and ALT levels over the three time points; however, the distribution of the category of the FIB‐4 index showed a significant difference (P = 0.005; Table 3). In contrast, patients with ALT ≥30 IU/L at baseline showed a significant decrease in AST and ALT levels after combination treatment initiation (P = 0.04 and P = 0.005, respectively) while there was no significant change in the distribution of the category of the FIB‐4 index over the treatment period.

Table 3.

Comparison of the clinical characteristics between at baseline, at the time of the addition, and at the final observation among two groups of patients with ALT <30 and those with ALT ≥30 at baseline

	At baseline	At the time of the addition	At the time of final observation	Baseline vs the addition	Baseline vs the final observation	The addition vs the final observation
ALT <30 at baseline (n = 349)
BW (kg)	75.4 ± 18.2	74.2 ± 17.2	71.4 ± 16.6	<0.001	<0.001	<0.001
BMI	28.6 ± 5.8	28.1 ± 5.4	27.0 ± 5.2	<0.001	<0.001	<0.001
SBP (mmHg)	133.1 ± 19.9	130.0 ± 8.4	126.7 ± 15.8	0.04	<0.001	0.004
DBP (mmHg)	75.7 ± 12.8	73.8 ± 11.9	72.1 ± 11.8	0.003	<0.001	0.31
MAP (mmHg)	94.9 ± 13.4	92.5 ± 12.1	90.3 ± 11.3	0.002	<0.001	0.03
HbA1c (mmol/mol [%])	71.0 ± 18.8 (8.6 ± 1.7)	70.2 ± 16.7 (8.6 ± 1.5)	63.3 ± 16.0 (7.9 ± 1.5)	0.65	<0.001	<0.001
eGFR (mL/min/1.73 m2)	76.2 ± 28.2	71.1 ± 27.0	67.9 ± 26.8	<0.001	<0.001	<0.001
AST(IU/mL)	19.2 ± 5.8	21.6 ± 8.0	21.5 ± 7.4	0.10 between three points
ALT(IU/mL)	18.7 ± 5.7	22.6 ± 11.3	22.0 ± 11.9	0.17 between three points
Platelet(109/L)	24.8 ± 7.5	24.9 ± 7.0	24.7 ± 7.0	0.87 between three points
Category of FIB‐4 index; <1.3/1.3–2.67/2.67<	218 (63%)/119 (34%)/12 (3%)	214 (61%)/117 (34%)/18 (5%)	197 (56%)/134 (39%)/18 (5%)	0.005 between three points
Hepatic steatosis index	39.5 ± 6.4	39.4 ± 6.4	38.1 ± 6.4	0.99	<0.001	<0.001
ALT ≥30 at baseline (n = 294)
BW (kg)	84.3 ± 19.2	82.6 ± 19.2	79.1 ± 18.9	0.09	<0.001	<0.001
BMI	30.9 ± 6.0	30.3 ± 5.9	29.0 ± 5.8	0.04	<0.001	<0.001
SBP (mmHg)	134.3 ± 17.3	133.3 ± 16.5	131.3 ± 16.4	0.27	0.004	0.34
DBP (mmHg)	79.8 ± 2.8	79.2 ± 11.8	77.8 ± 12.6	1.0	0.04	0.22
MAP (mmHg)	97.9 ± 12.8	97.3 ± 11.8	95.6 ± 12.3	0.67	0.007	0.17
HbA1c (mmol/mol [%])	73.8 ± 17.0 (8.9 ± 1.6)	71.3 ± 16.7 (8.7 ± 1.5)	64.0 ± 16.5 (8.0 ± 1.5)	0.13	<0.001	<0.001
eGFR (mL/min/1.73 m2)	81.3 ± 26.2	77.4 ± 25.6	74.2 ± 26.6	<0.001	<0.001	0.14
AST (IU/mL)	42.5 ± 23.2	37.2 ± 21.3	31.3 ± 18.7	0.46	<0.001	0.04
ALT (IU/mL)	60.3 ± 30.4	50.3 ± 30.3	39.8 ± 26.1	0.11	<0.001	0.005
Platelet (109/L)	24.1 ± 7.5	24.3 ± 7.6	24.1 ± 7.5	0.13 between three points
Category of FIB‐4 index; <1.3/1.3–2.67/2.67<	18 (62%)/89 (30%)/24 (8%)	175 (60%)/95 (32%)/24 (8%)	173 (59%)/97 (33%)/24 (8%)	0.70 between three points
Hepatic steatosis index	45.8 ± 7.3	44.1 ± 7.3	42.1 ± 7.3	<0.001	<0.001	<0.001

Values are mean ± SD or n (n/total, %). P‐values were calculated by the mixed general linear model with Bonferroni's correction.

^* P‐value was calculated by the Friedman test.

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BW, body weight; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration; FIB‐4, Fibrosis‐4; HbA_1c, glycated hemoglobin A_1c; MAP, mean arterial pressure; SBP, systolic blood pressure; SD, standard deviation.

PS‐matching model for comparison of patients stratified according to the preceding drug, SGLT2is or GLP1Ras

The clinical characteristics and concomitant drugs at baseline for patients in the unadjusted and PS‐matching models, which each included 201 patients in the SGLT2i‐preceding and GLP1Ra‐preceding groups, are shown in Table 4. The range of the standardized differences of the covariates was 0.0–0.13. Thus, the PS‐matched model was thought to be well balanced between the groups. The changes in liver function for patients in the PS‐matched model are shown in Figure 1 and Table 5. In this PS‐matched model with well‐balanced baseline covariates, clinical characteristics showed no significant differences with the addition of treatment. AST, ALT, and the distribution of the category of the FIB‐4 index values at final observation were not significantly different from those obtained at baseline.

Table 4.

The clinical characteristics and the concomitant drugs at baseline in the unadjusted model and in the PS‐matched model that included 165 patients in each group (SGLT2i‐preceding and GLP1Ra‐preceding group)

	Unmatched model			PS‐matching model
	GLP1Ra‐preceded group, N = 331	SGLT2i‐preceded group, N = 312	P‐value	GLP1Ra‐preceded group, N = 201	SGLT2i‐preceded group, N = 201	Standardized difference
Age (year‐old)	55.7 ± 13.5	56.5 ± 12.7	0.41	56.8 ± 14.4	56.9 ± 13.4	0.01
Sex (female [%])	152 (46%)	130 (42%)	0.28*	87 (43%)	87 (43%)	0.0
The history of type 2 diabetes >10 years (%)	281 (85%)	237 (76%)	0.006*	161 (80%)	159 (79%)	0.03
BW (kg)	79.5 ± 20.1	79.4 ± 18.1	0.95	78.6 ± 18.5	79.8 ± 18.8	0.06
BMI	29.8 ± 6.3	29.5 ± 5.6	0.51	29.5 ± 6.0	29.6 ± 5.7	0.02
SBP (mmHg)	132.0 ± 18.4	135.4 ± 18.9	0.02	133.8 ± 19.0	134.7 ± 18.9	0.05
DBP (mmHg)	76.6 ± 12.3	78.7 ± 13.6	0.04	77.4 ± 12.2	77.9 ± 13.3	0.04
MAP (mmHg)	95.0 ± 12.7	97.6 ± 13.6	0.02	96.2 ± 12.8	96.8 ± 13.6	0.05
HbA1c (mmol/mol [%])	73.6 ± 18.6 (8.9 ± 1.7)	71.0 ± 7.3 (8.6 ± 1.6)	0.07	74.3 ± 17.6 (9.0 ± 1.6)	72.0 ± 18.3 (8.7 ± 1.7)	0.13
eGFR (mL/min/1.73 m2)	78.8 ± 28.7	78.2 ± 26.0	0.79	79.5 ± 28.8	77.7 ± 26.0	0.07
AST (IU/mL)	30.1 ± 19.9	29.5 ± 20.1	0.72	29.4 ± 18.7	28.7 ± 17.5	0.04
ALT (IU/mL)	38.4 ± 29.9	36.9 ± 29.0	0.52	37.0 ± 28.6	37.1 ± 30.3	0.003
Platelet	24.6 ± 7.6	24.4 ± 7.4	0.65	24.7 ± 7.6	24.5 ± 8.2	0.03
Hepatic steatosis index	42.7 ± 7.9	42.1 ± 7.1	0.30	42.3 ± 8.0	42.4 ± 7.5	0.01
Category of the Fib‐4 index	1.27 ± 0.81	1.29 ± 0.80	0.73
FIB‐4 index<1.3	206 (62%)	192 (62%)	0.90	122 (61%)	124 (62%)	0.02
1.3 ≤FIB‐4 index ≤2.67	105 (32%)	103 (33%)		66 (33%)	65 (32%)	0.01
FIB‐4 index >2.67	20 (6%)	17 (5%)		13 (6%)	12 (6%)	0.02
Duration of the preceding treatment (month)	31.8 ± 23.1	23.9 ± 14.0	<0.001	24.7 ± 18.0	24.4 ± 14.6	0.02
Duration of the combination treatment (month)	38.8 ± 18.6	28.5 ± 13.5	<0.001	32.3 ± 14.9	32.0 ± 14.1	0.02
Total duration of the study (month)	70.6 ± 26.9	52.4 ± 5.7	<0.001	57.0 ± 9.9	56.5 ± 15.0	0.04
Concomitant medications
Sulphonyl urea	108 (33%)	91 (29%)	0.34*	61 (30%)	61 (30%)	0.0
Metformin	169 (51%)	190 (61%)	0.01*	121 (60%)	115 (57%)	0.06
Insulin	141 (43%)	140 (45%)	0.56*	92 (46%)	86 (43%)	0.06
Pioglitazone	35 (11%)	51 (16%)	0.03*	29 (14%)	27 (13%)	0.03
αGI	40 (12%)	48 (15%)	0.22*	26 (13%)	27 (13%)	0.02
Glinide	14 (4%)	14 (5%)	0.87*	10 (5%)	11 (6%)	0.02
RAS inhibitor	166 (50%)	160 (51%)	0.77*	104 (52%)	98 (49%)	0.06
CCB	128 (39%)	110 (35%)	0.37*	81 (40%)	82 (41%)	0.01
B blocker	53 (16%)	49 (16%)	0.92*	32 (16%)	32 (16%)	0.0
MRB	14 (4%)	12 (4%)	0.81*	9 (5%)	9 (5%)	0.0
Thiazide	29 (9%)	16 (5%)	0.07*	14 (7%)	15 (8%)	0.02
Loop	24 (7%)	14 (5%)	0.14*	7 (4%)	10 (5%)	0.07
Statin	160 (52%)	160 (51%)	0.46*	104 (52%)	99 (49%)	0.05

Values are mean ± SD or n/total n (%). P‐values are by unpaired t‐test or *chi‐square test.

αGI, alpha glucosidase inhibitor; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BW, body weight; DBP, diastolic blood pressure; CCB, calcium channel blocker; eGFR, estimated glomerular filtration; FIB‐4, Fibrosis‐4; GLP1Ra, glucagon‐like peptide 1 receptor agonist; HbA_1c, glycated hemoglobin A_1c; MAP, mean arterial pressure; MRB, mineral corticoid receptor blocker; PS, propensity score; RAS, renin‐angiotensin system inhibitor; SBP, systolic blood pressure; SGLT2i, sodium‐glucose cotransporter inhibitor.

Figure 1.

Changes in liver characteristics on PS‐matching model. AST, aspartate aminotransferase; ALT, alanine aminotransferase; FIB‐4, Fibrosis‐4; GLP1Ra, glucagon‐like peptide 1 receptor agonist; PS, propensity score; SGLT2i, sodium‐glucose cotransporter inhibitor.

Table 5.

Changes in the liver functions in the PS‐matched model

	At the addition		P‐value	At the final observation		P‐value
	GLP1Ra‐preceded group, N = 200	SGLT2i‐preceded group, N = 200		GLP1Ra‐preceded group, N = 200	SGLT2i‐preceded group, N = 200
BW (kg)	77.1 ± 18.1	78.8 ± 18.5	0.31	73.5 ± 18.0	76.4 ± 18.3	0.07
BMI	28.9 ± 5.8	29.2 ± 5.7	0.56	27.5 ± 5.7	28.3 ± 5.6	0.13
SBP (mmHg)	132.5 ± 17.5	130.7 ± 18.5	0.34	129.2 ± 16.7	129.3 ± 17.5	0.93
DBP (mmHg)	76.2 ± 11.8	76.2 ± 12.5	0.97	73.8 ± 12.4	74.7 ± 12.9	0.45
MAP (mmHg)	94.9 ± 11.9	94.4 ± 12.6	0.65	92.3 ± 12.4	92.9 ± 12.6	0.60
HbA1c (mmol/mol [%])	72.5 ± 16.6 (8.8 ± 1.5)	69.9 ± 17.3 (8.5 ± 1.6)	0.13	64.1 ± 16.0 (8.0 ± 1.5)	62.9 ± 15.8 (7.9 ± 1.4)	0.48
eGFR (mL/min/1.73 m2)	75.3 ± 26.8	73.0 ± 25.8	0.37	70.6 ± 27.7	71.5 ± 26.0	0.75
AST(IU/mL)	30.0 ± 18.3	27.6 ± 16.1	0.15	25.2 ± 14.2	26.1 ± 13.0	0.50
ALT(IU/mL)	37.9 ± 28.9	33.1 ± 24.3	0.09	29.1 ± 20.1	29.9 ± 18.6	0.68
Platelet	25.1 ± 7.7	23.9 ± 6.8	0.10	24.6 ± 7.6	24.6 ± 7.6	0.69
Hepatic steatosis index	41.7 ± 7.2	41.4 ± 7.2	0.70	39.6 ± 7.3	40.3 ± 7.2	0.33
Category of FIB‐4 index; <1.3/1.3–2.67/2.67<	115 (57%)/73 (36%)/13 (7%)	123 (61%)/64 (32%)/14 (7%)	0.37*	113 (56%)/73 (39%)/13 (5%)	112 (56%)/73 (36%)/16 (8%)	0.65*

Values are mean ± SD or n (n/total, %). P‐values were calculated by paired t‐test.

^* P‐value was calculated by the Friedman test.

ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; BW, body weight; DBP, diastolic blood pressure; eGFR, estimated glomerular filtration; FIB‐4, Fibrosis‐4; HbA_1c, glycated hemoglobin A_1c; MAP, mean arterial pressure; SBP, systolic blood pressure.

The sensitivity analysis results are shown in Tables S1 and S2. In the SGLT2i‐preceding group, AST, ALT, and FIB‐4 index values at final observation were not significantly different from those obtained at baseline; however, AST and ALT values at the initiation of the combination treatment were significantly lower than those obtained at baseline (P = 0.02 for both liver function parameters).

Results of the multivariable linear regression analysis of changes in ALT and FIB‐4 index values

Multivariable linear regression analysis for ΔALT indicated that ALT and BW at baseline, as well as ΔBW, were independent factors for ΔALT with the β‐coefficient values (95% confidence interval [CI]) of −0.70 [−0.76, −0.63], 0.30 [0.19, 0.41], and 0.80 [0.49, 1.11], respectively (all P‐values were <0.001). In contrast, AST at baseline, ALT at baseline, use of luseogliflozin, ΔMAP, and use of metformin were independent factors for ΔFIB‐4 index with the β‐coefficient values [95% CI] of −0.02 [−0.03, −0.02] (P < 0.001), 0.01 [0.006, 0.013] (P < 0.001), −0.35 [−0.62, −0.08] (P = 0.01), 0.01 [0.02, 0.00] (P = 0.02), and −0.12 [−0.24, −0.01] (P = 0.04), respectively.

DISCUSSION

In this study, we compared changes in liver function in 643 patients who received SGLT2i + GLP1Ra combination treatment for approximately 3 years. Combination treatment significantly improved liver function and prevented increases in FIB‐4 index values among patients with FIB‐4 index ≥1.3. The FIB‐4 index is strongly influenced by age factors, and it is expected that in a cohort study spanning approximately 4 years, the FIB‐4 index will increase compared to baseline. In a categorical analysis, it is anticipated that patients will progress to a higher FIB‐4 index category after several years of follow‐up. However, the improvement observed in some patients, particularly those with a FIB‐4 index greater than 1.3, suggests an improvement in liver function over time. Using a PS‐matched model, we noted that the preceding drug (SGLT2i or GLP1Ra) did not significantly affect liver function; however, prior SGLT2i use appeared beneficial for improving liver function. The hepatoprotective effects of SGLT2is and GLP1Ras‐only treatments have been reported ¹³ , ¹⁴ , ¹⁵ , ¹⁶ ; however, reports on the efficacy of their combined use are limited. In particular, most studies on the effects of these treatments on liver fibrosis are short‐term, and there are very few reports examining the long‐term course, as in this study. A study conducted in Japan on the administration of liraglutide over 5 years and its impact on the FIB‐4 index ²⁴ found a significant decrease only in patients with a baseline FIB‐4 index of 2.67 or higher. Here, we observed that SGLT2i + GLP1Ra combination treatment improved liver function in patients with type 2 diabetes and suppresses liver fibrosis in those with an FIB‐4 index ≥1.3. These findings suggest that these two drugs possibly have a stronger beneficial effect on MASLD and other liver conditions than their monotherapies. While there are reports of the concurrent administration of SGLT2is and GLP1Ras ¹³ , ¹⁴ , ¹⁵ , ¹⁶ the effect of the drug used before the combination treatment on liver function remains unclear. Here, we report for the first time that in clinical practice, the use of SGLT2i prior to SGLT2i + GLP1Ra combination treatment may be more beneficial to liver function.

GLP1Ras primarily improve liver function by suppressing hepatic glucose production, enhancing insulin sensitivity, alleviating local inflammation, and reducing body weight ¹³ , ¹⁶ . These effects result in a decrease in hepatic fat content and liver enzyme levels and hepatoprotective benefits, especially in conditions such as MASLD. Potential improvements in liver function dependent on weight loss owing to SGLT2i and GLP1Ra use have been reported. However, improvements in liver function independent of weight loss have also been reported ¹⁵ . SGLT2is reduce the levels of certain biomarkers of hepatocellular damage, such as cytokeratin 18‐M30 and plasma fibroblast growth factor 21 ²⁵ . Histological examinations of liver biopsies from patients with type 2 diabetes have also shown that SGLT2is improve non‐alcoholic fatty liver disease (NAFLD) activity score, steatosis, inflammation, and ballooning, and in some cases, ameliorate fibrosis and modify hepatic gene expression profiles ²⁶ .

There are insufficient reports on combination treatment in clinical practice. The strengths and novelties of this study include the following: (i) it is the first report on the assessment of the impact of different prior therapies (SGLT2i or GLP1Ra) on liver function in patients receiving SGLT2i + GLP1Ra combination treatment, (ii) it involves the analysis of data corresponding to a long period of real‐world clinical practice over 3 years, (iii) sufficient sample size, (iv) the use of PS‐based statistical analysis to adjust for potential confounding factors, and (v) the results obtained are based on trend scores obtained using actual clinical data on known background factors, enabling a more practical evaluation of treatment effects.

In daily clinical practice, injectable GLP1Ras are often used in combination with SGLT2is. SGLT2is are frequently administered first, especially in patients with type 2 diabetes and hepatic dysfunction owing to cost considerations. Our present results suggest that the order of administration may be appropriate for slowing hepatic dysfunction progression. Moreover, liver dysfunction conditions such as MASLD are independently associated with cardiovascular diseases ²⁷ , and the cardiovascular benefits of GLP1Ras and SGLT2is are well established ¹ , ² , ³ . Therefore, SGLT2i + GLP1Ra combination treatment, in addition to improving liver function, may also reduce cardiovascular risk in patients with type 2 diabetes and MASLD. Further, given that the results of this study are based on actual clinical data, the accumulation of additional evidence may lead to the establishment of this combination treatment may become more common for patients with obesity who also have type 2 diabetes.

This study has some limitations. First, it was a retrospective observational study; thus, the data collected may be associated with selection bias. The study only included subjects who continued treatment for at least 1 year, potentially excluding those who discontinued due to adverse effects or lack of efficacy, which could bias the results toward a more favorable outcome. Reportedly, among users of hypoglycemic agents, metformin and SGLT2i users show higher adherence and continuation rates, whereas users of injection therapies, for example, GLP1Ras, show lower adherence and continuation rates ²⁸ . This study also focused on patients with a BMI close to 30, which is higher than the average BMI of Japanese patients with type 2 diabetes (24.3) ²⁹ . Additionally, even though GLP1Ras have been available in Japan since 2010, their use is still limited. Thus, this study possibly included a significant number of severely obese patients with poor adherence to diet and exercise treatment who were compelled to use GLP1Ras in the long term. Second, this study was limited by a lack of data on key factors such as abdominal circumference, presence of metabolic syndrome, and albumin levels, which are critical for a more comprehensive assessment of liver function and the presence of MASLD/MASH. Consequently, more accurate assessment tools, including the Fatty Liver Index, the NAFLD Liver Fat Score, and the NAFLD‐Fibrosis Score, could not be employed. Instead, only the HSI was utilized, which may limit the thoroughness of the liver function evaluation in this study. Third, most of the GLP1RAs used in this study, including those that were changed during the course of the study, were dulaglutide or liraglutide, and semaglutide, which is considered to be relatively effective, was not used. Due to the frequent changes in treatment during the observation period, it was challenging to assess the effects of GLP1RA type and dosage on liver function. Furthermore, the dose of GLP1Ras administered in Japan is lower than that administered in other countries, for example, most of the patients included in this study did not receive the maximum doses of liraglutide and dulaglutide, which are commonly used in other countries.

Regardless of these limitations, this study provides valuable insights into liver function improvement and highlights the potential benefits of the SGLT2i + GLP1Ra combination treatment in patients with type 2 diabetes, especially in the context of MASLD. However, to fully validate these findings and better understand the long‐term benefits and mechanisms of this combination treatment, prospective studies are essential. Such studies are crucial not only for confirming observed effects but also for developing strategies to optimize the application of the treatment in clinical practice.

CONCLUSIONS

Our results indicate that in patients with type 2 diabetes, SGLT2i, and GLP1Ra combination treatment significantly improved liver function and prevented increases in FIB‐4 index value, particularly in patients with baseline FIB‐4 index ≥1.3. We also found that the preceding drug (SGLT2i or GLP1Ra) did not change the liver function outcomes. Prior SGLT2i use appeared to be more beneficial for improving liver function than prior GLP1Ra use. Further studies are warranted to validate these findings to establish this combination treatment as a standard treatment protocol for patients with type 2 diabetes.

DISCLOSURE

Kazuo Kobayashi, Kei Takeshita, Takuya Hashimoto, Moritsugu Kimura, Yoshimi Muta, Hisashi Yokomizo, and Shunichiro Tsukamoto declare no conflict of interest. Keizo Kanasaki is an Editorial Board member of the Journal of Diabetes Investigation and a co‐author of this article. To minimize bias, he was excluded from all editorial decision‐making related to the acceptance of this article for publication. Keizo Kanasaki received lecture fees from Dainippon‐Sumitomo Pharma, Astellas, Astra Zeneca, Ono, Otsuka, Taisho, Tanabe‐Mitsubishi, Eli Lilly, Boehringer‐Ingelheim, Novo Nordisk, Sanofi, Bayer, Novartis, and Kowa and received research funding from Boehringer Ingelheim, Kowa, Mitsubishi Tanabe Pharma, and Bayer. Daisuke Tsuriya received lecture fees from Eli Lilly, Novo Nordisk, and Mitsubishi Tanabe. Yuichi Takashi received lecture fees from Kyowa Kirin. Kouichi Tamura has received honoraria/lecture fees from AstraZeneca, Novartis, Bayer, Otsuka Pharmaceutical, Boehringer Ingelheim, Fuji Pharma, Kyowa Kirin, Ono Pharmaceutical, Sanwa Kagaku, Mochida Pharmaceutical, Kowa, Eli Lilly, Novo Nordisk, commissioned clinical trials, contract research, and joint research funding: AstraZeneca, Bayer, Novartis, Chinook, Otsuka Medical Devices, Novo Nordisk, Terumo, Variatris, and Kowa, and scholarship donations: Otsuka Pharmaceutical, Bayer, Mochida Pharmaceutical, and Boehringer Ingelheim. Daiji Kawanami received consulting/lecture fees from Bayer Yakuhin Ltd., Mitsubishi Tanabe Pharma Corporation, Novo Nordisk Pharma Ltd., Sanofi K.K., and Sumitomo Pharma Co., Ltd.; and grants from Bayer Yakuhin Ltd., Nippon Boehringer Ingelheim Co., Ltd., Nipro Corporation, and Sumitomo Pharma Co., Ltd. Masao Toyoda received lecture fees from Boehringer Ingelheim, Eli Lilly, Novo Nordisk, Sumitomo, and Mitsubishi Tanabe and received subsidies from Super Light Water, TAKAGI, Roche DC, and LifeScan.

Approval of the research protocol: This study was approved by the Institutional Review Board for Clinical Research, Tokai University, Japan (approval on December 6, 2021).

Informed consent: N/A.

Registry and the registration no. of the study/trial: N/A.

Animal studies: N/A.

FUNDING INFORMATION

This study received no specific grant from any funding agency in the public, commercial, or not‐for‐profit sectors.

Supporting information

Table S1 | The clinical characteristics and the concomitant drugs at baseline in unadjusted model and in the PS matching model in the CCA set that included 166 patients in each group (SGLT2i‐preceding and GLP1Ra‐preceding group).

Table S2 | Changed in the liver functions in the PS matching model in the CCA set.

Figure S1 | Schematic of the study design.

Figure S2 | Schematic of the study participants.

Figure S3 | Breakdown of the missing data.