Trends in clinical characteristics and factors associated with initial prescription of SGLT2 inhibitors in Japanese patients with type 2 diabetes mellitus

Abstract

Aims

To investigate the changes in patient background and treatment lines between 2016–2019 and contributing factors when sodium-glucose co-transporter 2 inhibitors (SGLT2i) are newly prescribed for type 2 diabetes mellitus patients.

Methods

The subjects comprised patients who had attended outpatient clinics at the four Jikei University School of Medicine-affiliated hospitals. One-way analysis of variance was used to evaluate annual changes in patients’ characteristics. Logistic regression analysis was also used to explore factors contributing to the treatment lines.

Results

The age of the 1951 subjects [mean ± SD] was 59.1 ± 12.8 years; BMI 27.5 ± 4.9 kg/m²; HbA1c 8.15 ± 1.24%; eGFR 74.2 ± 25.3 ml/min/1.73m². SGLT2i was the 2.86th (± 1.22) new prescription among antidiabetic drugs, and at increasingly earlier treatment lines between 2016 and 2019 (3.28 ± 1.16 to 2.59 ± 1.19; P < 0.001). The age of initial SGLT2i prescription significantly increased over time (P < 0.001). In contrast, the patients’ BMI and eGFR values decreased over time. Again, the proportions of patients with chronic kidney disease (CKD) and cardiovascular disease-heart failure disease (CVD-HF) tended to increase over time. The patients for whom SGLT2i had been prescribed in the first line were more likely to have obesity and HF (1.64 [1.15–2.34] and 1.84 [1.12–3.02], respectively).

Conclusions

SGLT2i was more likely to be newly prescribed to patients with CVD-HF and CKD, older patients, and to be prescribed in earlier treatment lines in recent years. Obesity and HF were predictor of SGLT2i prescriptions in the first line.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13340-022-00577-y.

Introduction

Diabetes mellitus is predicted to affect 9.7 million Japanese people or 9.8% of the entire Japanese population in 2030 [1]. The most frequent cause of death among diabetes patients was malignant neoplasia (38.3%), followed by cardiovascular disease (14.9% of vascular disease and 8.7% of heart disease combined, 23.6%) [2]. In particular, type 2 diabetes (T2DM) is associated with microvascular complications and significantly increases the risk of cardiovascular disease (CVD) [3]. Therefore, the goals of treatment for T2DM are achieving glycemic control based on individual patient characteristics, preventing or delaying complications, and maintaining quality of life through treatment with diet, exercise therapy, and antidiabetic drugs (ADD). The choice to prescribe ADDs needs to be made not only in accordance with relevant guidelines but also to address the background clinical characteristics of each individual patient [6, 7]. In actual clinical practice, the attending physician should consider the following factors in the choice of ADDs: effect on CVD and chronic kidney disease (CKD), efficacy, hypoglycemia risk, impact on weight, cost, risk for side effects, and patient preference [6, 8].

Large-scale clinical studies (2015 and later) show that SGLT2i and glucagon-like peptide-1 receptor agonists (GLP-1RA) arrested the onset of CVD in patients with a prior history of CVD or those at high risk of developing CVD [9–12]. This led to the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) publishing guidelines (2018 and later) recommending the use of SGLT2i or GLP-1RA, alongside lifestyle modification and metformin (MET) in patients with a history of CVD, Heart Failure (HF) or CKD or those at high risk of developing any of the above [13]. Further, the European Society of Cardiology’s guidelines recommend that SGLT2i or GLP-1RA be used as first-choice drugs in earlier lines of treatment than previously thought [14].

Due to the above-mentioned developments, currently, SGLT2i accounts for an increasing proportion of all ADDs prescribed globally, including Japan [15, 16]. However, only a few studies have followed up on the SGLT2i prescription trend over time [16–18]. Furthermore, to date, a few studies have investigated how the background characteristics of patients (e.g., concomitant disease or prior medical history) or blood test data may affect the treatment lines for which SGLT2i is being newly prescribed [19, 20].

SGLT2i became available in April 2014 for clinical use as an ADD for patients with T2DM in Japan [21–29]. Thus, to address the above-mentioned research gap, using the Standardized Structured Medical Information eXchange2 (SS-MIX2) platform, the researcher gathered the medical records of all patients newly prescribed SGLT2i from 2016–2019 to explore the patient characteristics that contributed to the lines of treatment in which SGLT2i had been newly prescribed. It is hoped that the findings will assist physicians in newly prescribing SGLT2i in clinical practice.

Methods

Study population

The subjects comprised of patients who had attended outpatient clinics at the four Jikei University School of Medicine-affiliated hospitals and were newly prescribed SGLT2i between January 2016 and December 2019. Subjects and their associated data were identified using the SS-MIX2; this facilitated searches based on patient age, sex, laboratory data, diseases (coded using the ICD-10 [International Statistical Classification of Diseases and Related Health Programs 10th revision]), and medications and doses.

The reference date was defined for each patient as the date they were first prescribed a SGLT2i. Patient exclusion criteria were: (1) having type 1 diabetes or pregnant; (2) being younger than 20 years of age at the time the database was searched; (3) previous prescription of a SGLT2i at initial presentation; (4) being newly prescribed a SGLT2i as an inpatient; and (5) receiving insulin therapy. The reasons for exclusion were as follows: (3) the timing of the newly prescribed SGLT2i was unknown, (4) during hospitalization, ADDs were adjusted daily and the concomitant medications were unclear, and (5) a patient in an insulin-dependent state could be included. For all patients judged eligible, data on relevant physical characteristics (e.g., height, body weight, and blood pressure) were gathered from their electronic medical records.

Patient characteristics

Data were collected for all eligible patients (age, duration of diabetes, body weight, body mass index (BMI), systolic/diastolic blood pressure, relevant blood test parameters (i.e., C-peptide index [CPI = serum fasting C-peptide/fasting blood glucose × 100], HbA1c, casual blood glucose, eGFR, urinary albumin, AST, ALT, γ-GTP, triglycerides, LDL-C, and HDL-C), and prior medical history (i.e., retinopathy, neuropathy, myocardial infarction or ischemic heart disease [IHD], HF, stroke, fatty liver [FL], hypertension, and dyslipidemia)). CPI used data measured within 1 year before SGLT2i prescription. Obesity was defined in line with the WHO guidelines as a BMI ≥ 30 kg/m². All patients with eGFR < 60 mL/min/1.73 m² or urinary albumin ≥ 30 mg/gCr were defined as having CKD. All patients diagnosed with IHD, HF or stroke were collectively defined as having CVD-HF.

Data were collected on the lines of treatment for which SGLT2i were newly prescribed, as well as on the ADDs used in the study subjects. Data were also collected on the prescription trends for seven ADDs available for clinical use in Japan, with the exception of SGLT2i, which included sulfonylureas (SU), glinides (GLI), alpha-glucosidase inhibitors (AGI), thiazolidinediones (TZD), metformin (MET), dipeptidyl peptidase-4 inhibitors (DPP4i) and GLP-1RA. All combination drugs were counted in relation to the number of component ingredients in each drug.

Statistical analysis

All patients were divided by the year in which they were newly prescribed SGLT2i: 2016 (those prescribed between January 1 and December 31 2016); 2017 (those prescribed between January 1 and December 31 2017); 2018 (those prescribed between January 1 and December 31 2018); and 2019 (those prescribed between January 1 and December 31 2019). One-way analysis of variance (ANOVA), chi-square tests and trend test (Jonckheere-Terpstra trend test or Cochran-Armitage trend test) were used to explore changes in SGLT2i prescription among these periods. Factors that were significantly different by the one-way ANOVA or chi-square tests were analyzed by multiple comparisons by year using the Bonferonni correction (adjusted significance level [< 0.008]). The treatment lines of SGLT2i new prescription were analyzed as a continuous variable from the 1st to the 6th line. On the other hand, for the percentage of SGLT2i treatment lines, the 5th and 6th lines were analyzed together as 5th + .

Additionally, all patients were divided by the line of treatment for which they were newly prescribed SGLT2i into two ways: (i) those prescribed in the first line versus those prescribed in the second or later lines; and (ii) those prescribed in the second line versus those prescribed in the third or later lines; and compared by age, sex, duration of diabetes (< 10 years), year of SGLT2i prescription, presence of obesity, CPI, favorable glycemic control, prior medical history (CKD, IHD, HF, stroke or FL) to explore the factors contributing to early SGLT2i prescriptions; this was achieved using logistic regression models or multi-nominal logistic regression models. In these analyses, favorable glycemic control was defined as HbA1c < 7% and, years of SGLT2i prescription was divided at the median into 2018 or later and before 2018. The explanatory variables were selected based on the factors that showed significant differences over time, the risk of adverse effects of diabetic ketoacidosis in patients with reduced endogenous insulin secretion [20, 30]. In addition, forced entry method was adopted in in multi-nominal logistic regression models.

All statistical analyses were performed using SPSS 26.0 and Stata 15.2. All data were represented as mean ± standard deviation (SD) or patient number (%). A p value of < 0.05 (two-tailed) was considered as statistically significant (p value < 0.008 for those using the Bonferonni correction).

The present study was conducted with the approval of the Institutional Review Board (IRB) of Jikei University School of Medicine (30–273 [9294]) and in accordance with the provisions of the Declaration of Helsinki (as revised in Fortaleza, Brazil, in October 2013).

Results

Patient characteristics

Of the 4695 patients who had been prescribed SGLT2i between 2016 and 2019, 1951 were available for analysis as per the inclusion/exclusion criteria (Fig. 1). At the time of their new SGLT2i prescription, the patients had a mean age of 59.1 ± 12.8 years, a mean duration of diabetes of 8.5 ± 6.9 years, a mean body weight of 75.9 ± 16.3 kg, a mean BMI of 27.5 ± 4.9 kg/m², a mean HbA1c value of 8.15 ± 1.24%, a mean eGFR of 74.2 ± 25.3 mL/min/1.73 m², a mean urinary albumin value of 146.7 ± 503.8 mg/gCr, and a mean CPI value of 1.83 ± 1.17. Of those with concomitant diseases, those with CKD, hypertension, dyslipidemia, CVD-HF, IHD, HF, stroke, and FL accounted for 49.6%, 67.1%, 83.1%, 27.3%, 15.2%, 9.7%, 5.7%, and 54.4%, respectively (Table 1).

Fig. 1.

Flowchart of the study patients. SGLT2i: sodium-glucose co-transporter 2 inhibitors

Table 1.

Trends in patient characteristics and prescription order of initiating SGLT2i over time

	Overall	2016	2017	2018	2019	P value✝	P value✝✝
Number of patients (females)	1951 (523)	290 (78)	414 (120)	582 (151)	665 (174)	0.714	0.505
Age, years	59.1 ± 12.8	54.9 ± 11.6	57.3 ± 12.6	59.5 ± 12.5§, §§	61.7 ± 13.1§, §§, §§§	< 0.001*	< 0.001**
≥ 65 years, no. (%)	734 (37.6%)	63 (21.7%)	142 (34.3%)§	225 (38.7%)§	304 (45.7%)§, §§	< 0.001*	< 0.001**
≥ 75 years, no. (%)	223 (11.4%)	8 (2.8%)	30 (7.2%)	62 (10.7%)§	123 (18.5%)§, §§, §§§	< 0.001*	< 0.001**
Duration of diabetes, years	8.5 ± 6.9	8.1 ± 6.0	8.8 ± 6.5	8.3 ± 6.9	8.7 ± 7.4	0.430	0.536
< 10 years, no. (%)	1202 (61.6%)	188 (64.8%)	244 (58.9%)	364 (62.5%)	406 (61.1%)	0.420	0.619
Body weight (kg)	75.9 ± 16.3	80.3 ± 17.7	76.9 ± 16.5	75.2 ± 15.8§	74.1 ± 15.6§, §§	< 0.001*	< 0.001**
Body mass index, kg/m2	27.5 ± 4.9	28.9 ± 5.5	27.8 ± 4.9	27.3 ± 4.7§	26.9 ± 4.7§, §§	< 0.001*	< 0.001**
≥ 30 kg/m2, no. (%)	461 (23.6%)	94 (32.4%)	102 (24.6%)§	129 (22.2%)§	136 (20.5%)§	< 0.001*	< 0.001**
Blood pressure, mmHg
Systolic, mean ± SD	130.6 ± 14.9	129.3 ± 14.9	130.5 ± 13.8	130.5 ± 14.8	131.1 ± 15.4	0.668	0.573
Diastolic, mean ± SD	76.7 ± 11.9	76.9 ± 12.3	77.6 ± 10.7	76.6 ± 11.7	76.5 ± 12.7	0.697	0.397
HbA1c, %	8.15 ± 1.24	8.28 ± 1.23	8.16 ± 1.18	8.12 ± 1.27	8.11 ± 1.26	0.089	0.005**
< 7%, no. (%)	241 (12.4%)	34 (11.7%)	40 (9.7%)	74 (12.7%)	93 (14.0%)	0.186	0.085
Casual blood glucose level, mg/dL	183.1 ± 62.0	184.4 ± 60.3	180.7 ± 59.4	184.2 ± 63.8	183.2 ± 62.8	0.824	0.880
eGFR, ml/min/1.73m2	74.2 ± 25.3	78.6 ± 21.8	74.9 ± 21.2	75.2 ± 31.2	70.9 ± 22.9§, §§, §§§	< 0.001*	< 0.001**
Urinary albumin, mg/gCr	146.7 ± 503.8	129.0 ± 353.8	186.3 ± 685.6	121.3 ± 417.5	151.0 ± 490.6	0.353	0.063
C-peptide index	1.83 ± 1.17	1.80 ± 1.27	1.83 ± 1.05	1.83 ± 1.12	1.83 ± 1.23	0.996	0.656
AST	32.4 ± 24.1	34.5 ± 23.6	31.9 ± 22.2	32.2 ± 26.5	31.9 ± 23.3	0.447	0.038**
ALT	40.0 ± 32.4	45.8 ± 37.2	41.4 ± 31.1	38.5 ± 32.2§	37.4 ± 30.9§	0.001*	< 0.001**
γGTP	65.7 ± 72.9	70.3 ± 83.5	61.6 ± 60.1	66.3 ± 76.5	65.7 ± 71.9	0.471	0.466
Triglycerides, mg/dL	200.3 ± 165.9	185.5 ± 146.8	196.1 ± 130.6	204.3 ± 186.5	205.9 ± 174.1	0.311	0.408
High-density lipoprotein, mg/dL	52.9 ± 14.6	52.4 ± 14.0	52.5 ± 13.6	52.7 ± 13.6	53.5 ± 16.3	0.599	0.512
Low-density lipoprotein, mg/dL	111.4 ± 30.3	111.5 ± 27.8	108.4 ± 29.6	112.5 ± 29.6	112.1 ± 32.2	0.264	0.514
Diabetic retinopathy, no. (%)	240 (12.3%)	39 (13.4%)	47 (11.4%)	65 (11.2%)	89 (13.4%)	0.549	0.771
Diabetic neuropathy, no. (%)	223 (11.4%)	31 (10.7%)	45 (10.9%)	66 (11.3%)	81 (12.2%)	0.883	0.435
CKD (eGFR < 60 or uAlb > 30 or uPro > 0.15), no. %	968 (49.6%)	132 (45.5%)	206 (49.8%)	270 (46.4%)	360 (54.1%)§, §§, §§§	< 0.001*	0.023**
Hypertension, no. (%)	1310 (67.1%)	180 (62.1%)	268 (64.8%)	394 (67.7%)	468 (70.4%)	0.053	0.006**
Dyslipidemia, no. (%)	1622 (83.1%)	242 (83.4%)	353 (85.3%)	485 (83.3%)	542 (81.5%)	0.451	0.220
CVD-HF, no. (%)	532 (27.3%)	57 (19.7%)	82 (19.8%)	174 (29.9%)§, §§	219 (32.9%)§, §§	< 0.001*	< 0.001**
IHD, no. (%)	296 (15.2%)	33 (11.4%)	43 (10.4%)	91 (15.6%)	129 (19.4%)§, §§	< 0.001*	< 0.001**
HF, no. (%)	189 (9.7%)	16 (5.5%)	25 (6.0%)	77 (13.2%)§, §§	71 (10.7%)§	< 0.001*	< 0.001**
Stroke, no. (%)	112 (5.7%)	7 (2.4%)	24 (5.8%)	33 (5.7%)	48 (7.2%)§	0.035*	0.008**
FL, no. (%)	1061 (54.4%)	171 (59.0%)	233 (56.3%)	313 (53.8%)	334 (50.2%)	0.033*	0.006**
Prescription order of SGLT2i	2.86 ± 1.22	3.28 ± 1.16	3.10 ± 1.20	2.78 ± 1.21§, §§	2.59 ± 1.19§, §§, §§§	< 0.001*	< 0.001**
First line, no. (%)	291 (14.9%)	16 (5.5%)	36 (8.7%)	98 (16.8%)§, §§	141 (21.2%)§, §§	< 0.001*	< 0.001**
Second line, no. (%)	498 (25.5%)	61 (21.0%)	102 (24.6%)	152 (26.1%)	183 (27.5%)	0.191	0.035**
Third line, no. (%)	579 (29.7%)	87 (30.0%)	123 (29.7%)	163 (28.0%)	206 (31.0%)	0.722	0.757

Data were presented as mean ± Standard deviation or no. (%)

Abbreviation: SGLT2i Sodium glucose co-transporter-2 inhibitors, 2016 Patients newly prescribed SGLT2i between January and December 2016, 2017 Patients newly prescribed SGLT2i between January and December 2017, 2018 Patients newly prescribed SGLT2i between January and December 2018, 2019 Patients newly prescribed SGLT2i between January and December 2019, HbA1c Hemoglobin A1c, eGFR Estimated Glomerular Filtration Rate, CKD Chronic Kidney Disease, CVD-HF Cardiovascular Disease-Heart Failure, IHD Ischemic Heart Disease, HF heart failure, FL fatty liver

^†Data for all years were analyzed by one-way analysis of variance or chi-squared test. * P < 0.05

^††Data for all years were analyzed by Jonckheere-Terpstra trend test or Cochran-Armitage trend test. **P < 0.05

Data for each year were analyzed to multiple comparisons by Bonferroni correction. ^§ P < 0.008 vs reference (2016). ^§§ P < 0.008 vs reference (2017). ^§§§P < 0.008 vs reference (2018)

Among the patients, the age of new SGLT2i prescription increased over time from 2016 to 2019 (54.9 ± 11.6 years to 61.7 ± 13.1 years; P < 0.001), and the proportion of patients ≥ 75 years old newly prescribed SGLT2i showed an increasing trend (P < 0.001). In contrast, BMI, HbA1c, and eGFR decreased among those newly prescribed SGLT2i over time (P < 0.001, P = 0.005, and P < 0.001, respectively). Among those with concomitant diseases, the proportion of those with CKD, CVD-HF, IHD, HF, stroke, and hypertension tended to increase over time (P < 0.001, P < 0.001, P < 0.001, P < 0.001, P = 0.008, and P = 0.006, respectively), while the proportion of those with FL tended to decrease over time (P = 0.006). In contrast, there was no change over time in the proportion of those with dyslipidemia (Table 1).

Treatment lines for new SGLT2i prescriptions

Among the ADDs prescribed, SGLT2i had newly been prescribed in the third line or thereabouts (2.86 ± 1.22). Between 2016 and 2019, SGLT2i had been prescribed in increasingly earlier treatment lines (3.28 ± 1.16 to 2.59 ± 1.19; P < 0.001) (Table 1, Fig. 2), and the proportion of patients newly prescribed SGLT2i in the first line showed an increasing trend over time (P < 0.001). During the entire period, DPP4i accounted for the greatest proportion of concomitant medication for new SGLT2i prescription (67.2%), followed by MET (56.2%) and SU (27.6%). For concomitant medications at the time of new prescription of SGLT2i, the proportion of SU, TZD, MET, and DPP4i showed a decreasing trend over time, while the proportion of GLP-1RA remained unchanged. Further, an examination of changes in ADD prescription patterns over time among those newly prescribed SGLT2i in the second line (Fig. 3A) showed that MET and DPP4i had been used in about 90% of these patients; again, while patients on MET accounted for the greatest proportion in the 2016–2017 period, those on DPP4i accounted for the greatest proportion in 2018 and later. Figure 3B–D showed the results of the analysis in subjects with eGFR higher than 60 ml/min/1.73 m² or under 65 years of age. Of all ADDs prescribed prior to SGLT2i, MET accounted for the greatest proportion between 2016 and 2019 among the subjects with eGFR ≥ 60 mL/min/1.73 m² and < 65 years of age, with no change over time in the proportion of MET and DPP4i prescribed.

Fig. 2.

Trends in the lines of treatment for initiating an SGLT2i. 1st: first line; 2nd: second line; 3rd: third line; 4th: fourth line; 5th + : fifth and later line. *p values for trend < 0.05

Fig. 3.

Trends of antidiabetic drugs use by drug class among T2DM patients, 2016–2019 (initiating SGLT2 inhibitors as second-line drugs). A Overall. B Subjects with eGFR higher than 60 ml/min/1.73m². C Subjects under 65 years of age. D Subjects with eGFR ml/min/1.73m² of 60 or higher and under 65 years of age. MET Metformin, DPP4i dipeptidyl peptidase-4 inhibitors, GLP-1RA glucagon-like peptide-1 receptor agonist, Others Antidiabetic drugs including sulfonylureas, glinides, alpha-glycosidase inhibitors, and thiazolidinediones. *p values for trend < 0.05

Factors contributing to SGLT2i prescriptions in early lines (Table 2)

Table 2.

Odds ratio of selection factors for initiating SGLT2 inhibitors among antidiabetic drugs

	1st vs 2nd and later (n = 1951)		2nd vs 3rd and later (n = 1660)
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Age, y
< 65	1 (Reference)	1 (Reference)	1 (Reference)	1 (Reference)
65–74	0.92 (0.68–1.22)	1.09 (0.73–1.62)	0.99 (0.77–1.26)	1.12 (0.82–1.54)
≤ 75	0.75 (0.48–1.15)	0.73 (0.40–1.34)	1.45 (1.05–1.99)*	1.71 (1.11–2.64)*
Female (ref: male)	1.15 (0.87–1.52)	1.01 (0.70–1.45)	0.92 (0.73–1.17)	0.98 (0.74–1.32)
Period, ≥ 2018 (ref: < 2018)	2.97 (2.17–4.08)*	2.83 (1.93–4.14)*	1.49 (1.20–1.86)*	1.46 (1.12–1.90)*
Duration of diabetes, < 10 years (ref: ≥ 10 years)	4.68 (3.31–6.62)*	3.73 (2.48–5.60)*	3.10 (2.46–3.92)*	3.47 (2.61–4.61)*
BMI, ≥ 30 kg/m2(ref: < 30 kg/m2)	1.54 (1.17–2.02)*	1.64 (1.15–2.34)*	1.14 (0.89–1.47)	1.03 (0.76–1.40)
HbA1c, < 7% (ref: ≥ 7%)	3.59 (2.65–4.88)*	4.00 (2.76–5.81)*	2.49 (1.79–3.46)*	2.51 (1.70–3.70)*
CPI	1.06 (0.90–1.26)	1.10 (0.89–1.36)	1.14 (0.99–1.31)	1.09 (0.93–1.27)
CKD (yes vs no)	1.25 (0.94–1.67)	1.03 (0.74–1.45)	1.28 (1.01–1.61)*	1.32 (1.01–1.73)*
IHD (yes vs no)	1.26 (0.91–1.76)	1.09 (0.68–1.74)	1.54 (1.16–2.04)*	1.26 (0.86–1.83)
HF (yes vs no)	1.99 (1.39–2.85)*	1.84 (1.12–3.02)*	1.96 (1.38–2.78)*	1.24 (0.78–1.95)
Stroke (yes vs no)	1.10 (0.65–1.85)	1.63 (0.88–3.02)	1.10 (0.70–1.72)	0.88 (0.50–1.57)
FL (yes vs no)	1.26 (0.98–1.63)	1.06 (0.75–1.50)	0.94 (0.77–1.17)	0.94 (0.72–1.24)

Abbreviations 1st first line, 2nd second line, 3rd third line, OR adjusted odds ratios, BMI body mass index, HbA1c Hemoglobin A1c, CPI C-peptide index, CKD chronic kidney disease, IHD ischemic heart disease, HF heart failure, FL fatty liver

Data were analyzed with a logistic regression model. In multi-nominal logistic regression models, forced entry method was adopted. *P < 0.05

Patients newly prescribed SGLT2i in early (i.e., first to second) treatment lines were characterized as having (i) newly prescribed SGLT2i in 2018 or later; (ii) a short duration of diabetes; and (iii) a low HbA1c value. Those newly prescribed SGLT2i in the first line were characterized as having a prior history of obesity and HF (adjusted odds ratios [OR], 1.64 [95% CI 1.15–2.34] and 1.84 [95% CI 1.12–3.02]), while those newly prescribed SGLT2i in the second line were characterized as being elderly (≥ 75 years of age) and as having CKD (adjusted ORs 1.71 [95% CI 1.11–2.64] and 1.32 [95% CI 1.01–1.73], respectively). However, sex, CPI, IHD, stroke, and FL did not affect the treatment lines for SGLT2i prescriptions.

Discussion

This study investigated the longitudinal changes in the background factors, treatment lines, and factors affecting the treatment lines for new SGLT2i prescriptions. Patients newly prescribed SGLT2i tended to be increasingly characterized over time as being more advanced in age and having CKD or CVD-HF. SGLT2i tended to be less frequently newly prescribed in patients with FL or obesity over time but tended to be frequently newly prescribed in those with low HbA1c values. Overall, SGLT2i had newly been prescribed in the third line or thereabouts but had newly been prescribed in increasingly earlier treatment lines. A prior history of obesity or HF was identified as likely contributing to the prescription of SGLT2i in the first line; advanced age or CKD were identified as factors likely contributing to the prescription of SGLT2i in the second line.

On one hand, SGLT2i offer glucose-lowering effects due to associated urinary glucose excretion and weight-reducing effects that positively affect FL [31, 32]. Indeed, the results of the present study identified that obesity was a factor contributing to new SGLT2i prescriptions in the first line, which was thought to reflect expectations not only for favorable glycemic control but for weight loss with SGLT2i. On the other hand, SGLT2i are associated with adverse effects, such as urinary tract infections, ketoacidosis, dehydration, and skin disorders [7, 30], and place elderly patients at increased risk of sarcopenia and thus require judicious use [30]. Reflecting these considerations, SGLT2i may have been more frequently used in younger patients with obesity when they first became available for clinical use. However, as it became clear that SGLT2i may be safely used in elderly patients as well with close monitoring of associated adverse effects [33], they may have come to be increasingly prescribed in elderly patients down the years, as shown in a recent report [16]. In general, the occurrence of urinary tract infections should be carefully followed up following SGLT2i prescription. Although SGLT2i was associated with a higher risk of urinary tract infections in women [21, 22], gender was not a selection factor in the first or second choice of SGLT2i in this study, and as in previous reports [20], there was no change in the gender ratio over time.

In the current study, SGLT2i was prescribed to an increasing proportion of the patients with CKD and CVD-HF over time, particularly in 2018 and later. Similar results were reported in a retrospective nationwide study (2014–2017) of Japanese patients [20]. In previous reports, SGLT2i was a factor in the prescription of first-line therapy in patients with IHD. However, in the present study, IHD was not a significant factor in SGLT2i selection, and the analysis was extended to factors affecting second choice. A prior history of HF (of all diseases categorized under CVD-HF) and CKD was identified as a factor contributing to SGLT2i prescriptions in the first and second lines, respectively. The prescribing patterns uncovered in the present study appear to reflect the wide-ranging results of other recent large-scale clinical trials such as the EMPA-REG OUTCOME trial with empagliflozin (2015), the CANVAS program with canagliflozin (2017), and the DECLARE-TIMI 58 trial with dapagliflozin (2019) [9, 10, 34–37]. In other words, SGLT2i was prescribed in earlier treatment lines than previously thought due to expectations not only for improved glycemic control but for prevention of CVD-HF or CKD events in patients with CVD-HF and CKD and patients at high risk of developing these diseases. On the other hand, advanced age was identified as a factor contributing to SGLT2i prescriptions in the second line. A recent meta-analysis showed that SGLT2i reduced the need for hospital admission due to HF and delayed the progression of CKD among patients with T2DM, irrespective of a prior history of CVD-HF [35]. In the present study, the proportion of SGLT2i prescriptions to patients older than 75 years with CKD or CVD-HF did not change over time from 2016 to 2019 in this study. However, in all years, the percentage of SGLT2i prescriptions for patients older than 75 years with CKD or CVD-HF was higher than for patients younger than 75 years (Figure Suppl). Therefore, the observation that elderly patients are highly likely to have a prior history of CVD-HF or CKD or to be at high risk of developing these diseases [38] may have contributed to SGLT2i being prescribed in earlier treatment lines.

In the present study, SGLT2i had been newly prescribed in the third line or thereabouts across the entire population; SGLT2i had also been newly prescribed in increasingly earlier treatment lines over time. Across all ages, approximately 90% of those newly prescribed SGLT2i as a second-line treatment had received prior MET or DPP4i. Similar results were reported in a longitudinal study (2014–2017) of Japanese patients [20, 39]. Here, of all ADDs prescribed prior to SGLT2i, MET accounted for the greatest proportion in 2016 and 2017 while DPP4i accounted for the greatest proportion in 2018 and later. In the present study, logistic regression analysis on the characteristics of patients newly prescribed SGLT2i in the second line who had previously been prescribed DPP4i or MET demonstrated that those prescribed DPP4i prior to SGLT2i had an eGFR of < 60 mL/min/1.73 m² or ≥ 65 years, while those prescribed MET prior to SGLT2i had characteristics opposite to those of patients prescribed DPP4i prior to SGLT2i (Table Suppl). Less likely to be associated with hypoglycemia or weight gain in Japanese patients, DPP4i are reported to have been prescribed in about 60% of patients as first-line drugs [40]. In contrast, MET is cautiously prescribed in patients with renal dysfunction and the elderly because of the risk of dehydration and lactic acidosis, and more caution is needed when prescribed in combination with SGLT2i [30, 41]. In the current study, the proportion of new SGLT2i prescriptions to elderly and patients with reduced renal function increased over time from 2016 to 2019, which may have affected the breakdown of ADDs in prior doses.

Despite its merits, the present study has five main limitations. First, the results may have reflected the access to ADDs made available in Japan through a universal health insurance scheme, where cost considerations are less of an issue than in the healthcare systems of other countries. Second, this was a single-center study involving four university-affiliated hospitals. Third, it is unclear whether the results can be generalized because the study was conducted only in facilities certified by the Japan Diabetes Society. Fourth, the observation period of this study was relatively short. Lastly, the study excluded patients with T2DM on insulin therapy and thus included a relatively small number of patients with T2DM. Despite these limitations, however, to the best of our knowledge, this is the first study to investigate the longitudinal changes in the characteristics of T2DM patients newly prescribed SGLT2i as well as the factors affecting the treatment lines for SGLT2i prescriptions. The findings suggest that SGLT2i are being newly prescribed in T2DM patients with obesity, HF or CKD in increasingly earlier treatment lines and that this trend may further accelerate in the years to come, given that some SGLT2i are now available for non-diabetic patients with HF and CKD.

In terms of future investigations of this area, the authors plan to conduct a further study to sample a larger number of patients with a longer observation period to examine the factors contributing to the variations in SGLT2i prescriptions among T2DM patients over time. In addition, we plan to examine subjects on insulin therapy who were excluded from this study (n = 561). It is hoped that the resulting insights may help not only to spare physicians caring for T2DM patients from critical biases in the choice of ADD prescription but to formulate an optimal treatment plan for each individual T2DM patient. Additionally, these insights may also help improve diabetes treatment and ultimately result in fewer diabetic complications and longer healthy life expectancy among T2DM patients as well as in lower healthcare costs for these patients.

Supplementary Information

Below is the link to the electronic supplementary material.

Declarations

Conflict of interest

Rimei Nishimura has participated in speakers’ bureaus/advisory panels for Astellas, Boehringer Ingelheim, Eli Lilly, Kissei, Medtronic, Novo Nordisk, Sanofi, Takeda, Novartis, and MSD and served as subsidies or donatins for Ono, Boehringer Ingelheim, Takeda, and Taisho. The other authors haven no conflict of interest to declare.

Ethical standards

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (Ethics Committee of Jikei University School of Medicine/November 12, 2018/No. 30–273 [9294]) and with the Helsinki Declaration of 1964 and later versions. This study used samples and medical information collected in the past in the course of normal medical care. Individual informed consent was not taken, but documents approved by the Ethics Committee were posted on the website to make the information available to the public, and the opportunity to refuse was provided.