Modeling Cardiorenal Protection with Sodium-Glucose Cotransporter 2 Inhibition in Type 1 Diabetes: An Analysis of DEPICT-1 and DEPICT-2

Visual Abstract

Abstract

Key Points

• Risk modelling analysis of DEPICT trials show that dapagliflozin reduced estimated cardiovascular and kidney disease risk in T1D persons.

• Greatest reduction in estimated ESKD risk was accompanied by an expected rise in eGFR, after 4 weeks post drug discontinuation.

• Dedicated outcome trials with SGLT2 inhibitors are warranted in T1D persons with CKD or CVD for best determination of efficacy and risks.

Background

Sodium-glucose cotransporter-2 (SGLT2) inhibitors improve glycemia and reduce insulin requirements in type 1 diabetes (T1D) and type 2 diabetes. Although SGLT2 inhibitors lower cardiovascular disease (CVD) and ESKD risk in type 2 diabetes, no dedicated cardiorenal outcome trials in T1D have been conducted to date. Using validated risk prediction models, this study evaluated the effect of SGLT2 inhibition on estimated CVD and ESKD risk in a T1D cohort.

Methods

Demographics, medical history, and biomarkers were extracted from 1473 participants with T1D enrolled in the Dapagliflozin Evaluation in Patients with Inadequately Controlled Type -1 and -2 trials. Data at baseline, 24, 52, and 56 weeks (4 weeks after drug cessation) were used to estimate 10-year CVD and 5-year ESKD risk using the Steno T1 Risk Engine (SRE) and Scottish Diabetes Research Network (SDRN) risk prediction models. Risk reduction was determined on the basis of relative change in risk from baseline between participants receiving dapagliflozin (pooled 5 and 10 mg) versus placebo. Subgroup analyses were conducted by age, sex, diabetes duration, CVD risk, and CKD status at baseline.

Results

The relative change in 10-year estimated CVD risk (SRE: –6.50% [–8.04% to –4.95%] and SDRN: –6.77% [–8.40% to –5.13%]; all P < 0.001) and 5-year ESKD risk (SRE: –4.48% [–7.68% to –1.28%]; P = 0.006) were lower at the end of 24 weeks of dapagliflozin treatment compared with placebo. Furthermore, the greatest relative change in 5-year ESKD risk was observed at week 56 (SRE: –12.84% [–16.65% to –9.03%]; P < 0.001), in conjunction with an expected rise in eGFR after drug washout. Subgroup analysis revealed larger relative lowering in 10-year CVD risk in those with CKD compared with those without (SRE: –11.3% versus –5.9%, and SDRN: –11.9% versus –6.1%, respectively; all P_interaction < 0.02).

Conclusions

Dapagliflozin improves estimated CVD and ESKD risk in participants with T1D, emphasizing the need for cardiorenal outcome trials in people living with T1D.

Introduction

People with type 1 diabetes (T1D) have elevated risks of cardiovascular disease (CVD) and kidney failure compared with the general population,^1,2 even with intensive glycemic and BP management, renin-angiotensin system inhibition and statins.^3,4 This is evidenced by the 10%–30% higher lifetime risk of developing kidney failure after 40 years of living with T1D⁵ and individuals diagnosed with T1D before age 10 years facing a 300% higher risk of coronary heart disease.⁶ Identifying novel therapies that mitigate the residual cardiorenal risk in T1D therefore remains of critical importance.

Sodium-glucose cotransporter 2 (SGLT2) inhibitors have been principally investigated as glucose-lowering therapies in people with T1D in the Dapagliflozin Evaluation in Patients with Inadequately Controlled Type (DEPICT) 1 Diabetes,^7–10 Sotagliflozin as Adjunct Therapy in Adult Patients With Type 1 Diabetes Mellitus Who Have Inadequate Glycemic Control With Insulin Therapy (inTandem),^11–14 and Empagliflozin as Adjunctive to Insulin Therapy¹⁵ trials. In type 2 diabetes (T2D), SGLT2 inhibition has consistently demonstrated improvements in composite cardiovascular and kidney end points in people with atherosclerotic CVD or CKD.^16–18 However, no dedicated cardiovascular or kidney outcome trials of SGLT2 inhibitors have been conducted to date in people with T1D. Their use is limited by increased ketogenesis and risk of diabetic ketoacidosis (DKA).^19–21 Without comprehensive trials, the balance of risks and potential benefits for patients with T1D remains unclear, complicating clinical decision making.

Alternatively, validated T1D risk prediction models, such as the Steno T1 Risk Engine (SRE), estimate risk of new-onset CVD or ESKD.^22,23 Using data from the inTandem trials, these models showed that 24 weeks of sotagliflozin lowers estimated 10-year CVD and 5-year ESKD risk by approximately 5%–7% as a percentage of baseline risk compared with placebo.²⁴ More recently, the Scottish Diabetes Research Network (SDRN) developed an additional CVD risk prediction model in a T1D-specific cohort, which yielded lower 10-year CVD risk estimates compared with the SRE.²⁵ The effects of SGLT2 inhibition on these risk estimates is unknown. Risk models should account for the well-established acute dip in eGFR that occurs with SGLT2 inhibition, reflected by its transient natriuretic effects that reduce glomerular hyperfiltration and hypertension.²⁶ This eGFR dip may lead to underestimations of the kidney-protective effects observed in previous studies²⁴ because risk models would not incorporate this physiological response. To overcome this challenge, risk modeling with longer follow-up or after treatment discontinuation²⁷ can be useful to fully understand the cardiorenal benefits of this therapy.

The aim of this study was to evaluate the effects of SGLT2 inhibition on estimated CVD and ESKD risk in the DEPICT-1 and DEPICT-2 cohorts. We hypothesized that participants with T1D treated with dapagliflozin would have lower CVD and ESKD risk compared with placebo using SRE and SDRN risk models, with the larger kidney protective effects observed during secondary follow-up periods of 52 weeks and after treatment discontinuation (4 weeks after drug washout).

Methods

Study Design and Participants

This study is a post hoc analysis of the DEPICT-1 and DEPICT-2 trial data. The DEPICT-1 (Clinicaltrials.gov ID: NCT02268214) and DEPICT-2 (Clinicaltrials.gov ID: NCT02460978) trials were multicenter, phase 3, randomized, placebo-controlled, parallel-group studies investigating the safety and efficacy of dapagliflozin as an add-on to insulin in participants with T1D not reaching glycemic targets.^8,10 In brief, persons between 18 and 75 years with T1D with glycated hemoglobin A1c between 7.7% and 11.0% and body mass index (BMI) ≥18.5 kg/m² who had been treated with insulin for at least 12 months before randomization were enrolled. Participants with a calculated creatinine clearance <60 ml/min were excluded, with no eligibility criterion on the basis of urine albumin–creatinine ratio (uACR). A run-in period of 8 weeks was required to stabilize the insulin dose before randomization. Participants were randomized 1:1:1 to dapagliflozin 5 or 10 mg or placebo for 24 weeks, with a 28-week long-term extension (52 weeks) and follow-up after 4 weeks off treatment. The primary study included 758 participants in DEPICT-1 and 780 participants in DEPICT-2.

Cardiovascular and Kidney Risk Prediction Models

Two predictive models—the SRE^22,23 and the SDRN risk score²⁵—were selected to estimate the cumulative risk of a CVD event in this T1D cohort as the models have been developed and validated in people with T1D. The SRE model defined CVD as a composite outcome of fatal and nonfatal events of ischemic heart disease, ischemic stroke, heart failure, and peripheral artery disease.²³ The SDRN model defined CVD as a hospitalization event due to myocardial infarction, stroke, unstable angina, transient ischemic attack or peripheral vascular disease; coronary, carotid, or peripheral artery revascularizations; major amputations; acute coronary heart disease; or cardiovascular-related death.²⁵ The cumulative risk of ESKD was estimated using the SRE, which defined ESKD as a composite event of an eGFR <15 ml/min per 1.73 m², dialysis, or kidney transplantation.²² Risk scores were calculated at four time points: (1) baseline, (2) 24 weeks on treatment, (3) 52 weeks on treatment, and (4) 4 weeks off treatment. Participants with a history of CVD were excluded from the CVD risk models because the SRE and SDRN CVD risk models were not validated for people with T1D and preexisting CVD. Input variables for SRE^22,23 and SDRN²⁵ risk models have been described previously and were summarized in the Supplemental Methods and supplemental Table 1. Models were replicated in RStudio using published information.

Statistical Analyses

For this study, all participants randomized to dapagliflozin (5 and 10 mg) were pooled into one group and compared statistically with participants receiving placebo. Descriptive baseline characteristics, including demographics, medical history, BP, and blood and urine biochemistry, were compared between participants randomized to the placebo or dapagliflozin treatment arm using an independent sample t test or chi-squared test for continuous and categorical data, respectively. Using participant-level data, the change (absolute and relative) from baseline for each risk score was calculated. Pairwise deletion was used for handling missing data, which was very low at baseline and week 24 for all variables included in risk models (<4%) and assumed to be missing at random. At week 52 and 4 weeks off treatment, lipid profiles were missing in 55% and 93% of participants, respectively (supplemental Table 2). A two-way repeated-measure analysis of covariance was used to statistically compare CVD risk or ESKD risk between participants treated with dapagliflozin and placebo. The log-transformed change (absolute and relative) in risk from baseline was used as the dependent variable, with treatment (dapagliflozin versus placebo) as the between-participant factor and time (baseline [fixed to zero] versus 24 weeks on treatment, 52 weeks on treatment, or 4 weeks off treatment) as the within-participant factor and the log-transformed baseline risk as the covariate. In the presence of a significant treatment-by-time interaction, post hoc testing was performed using Bonferroni adjustment for multiple comparisons, to quantify the placebo-adjusted change in CVD or ESKD risk. Results are presented as both placebo-adjusted relative change from baseline and placebo-adjusted absolute change from baseline.

Secondary subgroup analyses were also performed for each risk score after 24 weeks of dapagliflozin and included (1) age on the basis of a median split (40 years); (2) sex; (3) diabetes duration on the basis of a median split (18 years); (4) baseline CVD risk being low (<10%) or moderate to high (>10%) on the basis of the average between the SRE and SDRN risk models; and (5) presence of CKD, defined as an eGFR <60 ml/min per 1.73 m² or a uACR ≥30 mg/g. Subgroup analyses were statistically compared as described above, incorporating subgroup as a second between-participant factor. Statistical significance was defined as P < 0.05. Baseline characteristics are presented as mean±SD, risk estimates as median (interquartile range), and absolute or relative changes as mean (95% confidence interval). Predictive modeling and all statistical analyses outlined in this study were reproduced using RStudio (version 2023.06.0+421) using the rstatix statistical package.

Results

Baseline Characteristics of the Study Population

Table 1 reports the baseline characteristics of the study cohort. The resultant sample size for this study in the dapagliflozin and placebo groups was 1010 and 463, respectively. The analyzed cohort was 53% female, had a mean age of 40±13 years with a T1D duration of 20±12 years, and had a BMI of 24.8±4.3 kg/m². In addition, 34% had hypertension, 45% had dyslipidemia, and 9% had a prior CVD event. CKD was observed in 17% of participants (1% had an eGFR <60 ml/min per 1.73 m², 15% had an uACR ≥30 mg/g, and 1% had both). Baseline characteristics were similar between both dapagliflozin and placebo groups (all P > 0.05; Table 1).

Table 1.

Baseline characteristics of the Dapagliflozin Evaluation in Patients with Inadequately Controlled Type study cohort randomized to dapagliflozin or placebo

Variable	Placebo (n=463)	Dapagliflozin (n=1010)	P Value for Dapagliflozin versus Placebo Group
Demographics
Age, yr	40.6±13.7	39.9±13.3	0.34
Female, No. (%)	234 (51)	540 (54)	0.32
BMI, kg/m2	24.8±4.2	24.8±4.3	0.99
Diabetes duration, yr	20.3±12.1	20.0±11.8	0.64
Medical history, No. (%)
Hypertension	163 (35)	342 (34)	0.66
Dyslipidemia	219 (47)	441 (44)	0.21
CVD	42 (9)	97 (10)	0.82
CKD	86 (19)	160 (16)	0.22
Hemodynamics, mm Hg
Systolic BP	124±15	122±14	0.12
Diastolic BP	75±10	75±9	0.72
Biochemistry
Total cholesterol, mmol/L	4.77±0.98	4.77±0.92	0.96
LDL, mmol/L	2.62±0.81	2.61±0.77	0.86
HDL, mmol/L	1.63±0.46	1.65±0.48	0.52
HbA1c, %	8.46±0.65	8.46±0.67	0.89
Creatinine, µmol/L	73.6±15.5	74.2±15.9	0.47
eGFR, ml/min per 1.73 m2	98.9±17.9	98.3±17.7	0.58
Low eGFR, No. (%)	13 (2.8)	20 (2)	0.42
Na+, mmol/L	140±3	140±3	0.38
K+, mmol/L	4.5±0.4	4.5±0.4	0.64
Uric acid, mmol/L	0.25±0.08	0.25±0.07	0.53
uACR, mg/mmola	6.02 (3.01–16.01)	6.99 (3.98–15.0)	0.37
Microalbuminuria, No. (%)	64 (14)	131 (13)	0.72
Macroalbuminuria, No. (%)	14 (3)	18 (2)	0.19

Values are reported in mean±SD unless otherwise indicated. BMI, body mass index; bpm, beats per minute; CVD, cardiovascular disease; HbA1c, glycated hemoglobin A1c; uACR, urine albumin–creatinine ratio.

^aUrine albumin–creatinine ratio reported in median (interquartile range). P value comparison between treatment groups dapagliflozin versus placebo. eGFR estimated by creatinine clearance.

SRE Estimated 10-Year CVD Risk

After 24 weeks of treatment, the within-group relative change from baseline in 10-year CVD risk was –4.74% (–5.95% to –3.52%) in the dapagliflozin group and 1.76% (–0.06% to 3.57%) in the placebo group. The placebo-adjusted relative change from baseline was –6.50% (–8.04% to –4.95%; P < 0.001; Figure 1), corresponding to a placebo-adjusted absolute change from baseline of –0.70% (–0.93% to –0.47%; P < 0.001). At 52 weeks, a large portion of participants were excluded (approximately 57%) because of missing lipid profiles. In the remaining cohort, the lower estimated 10-year CVD risk was sustained, as demonstrated by a –5.78% (–8.35% to –3.21%; P < 0.001) placebo-adjusted relative change from baseline, corresponding to a placebo-adjusted absolute change from baseline of –0.78% (–1.11% to –0.46%; P < 0.001; Figure 2). The estimated 10-year SRE CVD risk was not calculated at 4 weeks off treatment because of missing lipid profiles in 95% of participants.

Figure 1.

Placebo-adjusted relative and absolute change in estimated CVD and ESKD risk after 24 weeks of dapagliflozin or placebo. Risk scores were estimated by (1) the SRE and (2) SDRN risk score. Baseline and week 24 risk values are reported as median (IQR). Relative change and absolute change of estimated risk between dapagliflozin and placebo are reported as mean (95% CI). *P < 0.05. CI, confidence interval; CVD, cardiovascular disease; IQR, interquartile range; SDRN, Scottish Diabetes Research Network; SRE, steno T1 risk engine.

Figure 2.

Placebo-adjusted relative and absolute change in estimated CVD and ESKD risk after 24, 52, and 56 weeks of dapagliflozin or placebo. Risk scores were estimated by (1) the SRE and (2) SDRN risk score at baseline, with treatment (24 or 52 weeks) and after a 4-week drug washout (56 weeks). Relative and absolute changes of estimated risk between dapagliflozin and placebo are reported as mean (95% CI). *P < 0.05.

SDRN Risk Score Estimated 10-Year CVD Risk

After 24 weeks, the within-group relative change from baseline in 10-year CVD risk was –4.10% (–5.39% to –2.81%) in the dapagliflozin group and 2.66% (0.75%–4.58%) in the placebo group. The placebo-adjusted relative change from baseline was –6.77% (–8.40% to –5.13%; P < 0.001; Figure 1), corresponding to a placebo-adjusted absolute change from baseline of –0.74% (–1.02% to –0.47%; P < 0.001). At week 52, a large portion of participants were excluded (approximately 57%) because of missing lipid profiles. In the remaining cohort, the lower estimated 10-year CVD risk was sustained after 52 weeks of treatment, as demonstrated by a –9.07% (–12.04% to –6.10%; P < 0.001; Figure 2) placebo-adjusted relative change from baseline, corresponding to a placebo-adjusted absolute change from baseline of –0.87% (–1.26% to –0.49%; P < 0.001). The estimated 10-year SDRN CVD risk was not calculated at 4 weeks off treatment because of missing lipid profiles in 95% of participants.

SRE Estimated 5-Year ESKD Risk

After 24 weeks, the within-group relative changes from baseline in 5-year ESKD risk was 2.12% (–0.40% to 4.64%) in the dapagliflozin group and 6.60% (2.84%–10.36%) in the placebo group. The placebo-adjusted relative change from baseline was –4.48% (–7.68% to –1.28%; P = 0.006; Figure 1). However, the placebo-adjusted absolute change from baseline was 0.00% (–0.04% to 0.04%; P = 0.96). After 52 weeks, the placebo-adjusted relative change from baseline was –8.46% (–12.44% to –4.48%; P < 0.001; Figure 2) and the placebo-adjusted absolute change from baseline was –0.04% (–0.09% to 0.01%; P = 0.16). The greatest improvement in ESKD risk was seen at 4 weeks off treatment; the placebo-adjusted relative change from baseline was –12.84% (–16.65% to –9.03%; P < 0.001), corresponding to a placebo-adjusted absolute change from baseline of –0.10% (–0.17% to –0.03%; P = 0.005).

Subgroup Risk Analysis by Sex, Diabetes Duration, CKD, and Baseline CVD Risk

The placebo-adjusted relative change from baseline in the SRE 10-year CVD risk after 24 weeks of treatment was not different by age, sex, diabetes status, or CVD risk at baseline (all P_interaction ≥ 0.68; Table 2). However, the placebo-adjusted absolute change from baseline in CVD risk differed by age and CVD risk at baseline (both P_interaction < 0.004). Heterogeneity by CKD status was also observed, such that participants with CKD at baseline exhibited larger placebo-adjusted relative (–11.3% versus –5.9%) and absolute (–2.2% versus –0.5%; both P_interaction < 0.01) lowering from baseline in 10-year CVD risk compared with participants without CKD. Similarly, using the SDRN model, placebo-adjusted relative change from baseline was not different by age, sex, diabetes status, or baseline CVD risk (all P_interaction ≥ 0.70), while larger placebo-adjusted lowering of absolute risks was observed in both older individuals and those with higher baseline CVD risk (both P_interaction ≤ 0.04), and larger placebo-adjusted relative and absolute risk lowering (both P_interaction ≤ 0.02) were observed in participants with CKD (Table 3). The placebo-adjusted relative change from baseline in the SRE estimated 5-year ESKD risk after 24 weeks of treatment and 4 weeks after drug discontinuation was not different by age, sex, diabetes status, CVD risk at baseline, or CKD at baseline (all P_interaction ≥ 0.35; Table 4).

Table 2.

Subgroup analyses of the Steno T1 Risk Engine–estimated 10-year cardiovascular disease risk after 24 weeks of dapagliflozin or placebo

Subgroup	Dapagliflozin		Placebo		Placebo-Adjusted Relative Change from Baseline	P interaction	Placebo-Adjusted Absolute Change from Baseline	P interaction
	Baseline	Week 24	Baseline	Week 24
Age
>40 yr (n=681)	15.5 (11.6–20.2)	14.4 (10.8–19.0)	16.5 (11.5–23.2)	17.3 (12.0–23.5)	−6.8 (−8.9 to −4.6)	0.84	−1.1 (−1.5 to −0.8)	0.004
<40 yr (n=602)	3.7 (2.7–5.1)	3.3 (2.4–4.8)	3.9 (2.8–5.5)	3.8 (2.9–5.2)	−6.2 (−8.4 to −3.9)		−0.2 (−0.5 to 0.1)
Sex
Male (n=592)	9.6 (4.6–16.1)	9.2 (4.1–16.0)	11.5 (4.6–19.4)	11.8 (4.5–20.3)	−6.0 (−8.3 to −3.8)	0.74	−0.7 (−1.0 to −0.4)	0.78
Female (n=691)	8.6 (3.3–15.6)	8.1 (3.1–14.4)	8.0 (3.7–15.2)	7.7 (3.6–14.8)	−6.8 (−8.9 to −4.7)		−0.7 (−1.0 to −0.3)
T1D duration
>18 yr (n=624)	12.9 (6.1–18.9)	12.0 (5.8–17.8)	15.6 (6.3–22.9)	15.0 (6.8–22.3)	−7.0 (−9.2 to −4.8)	0.68	−0.8 (−1.2 to −0.5)	0.38
<18 yr (n=659)	5.0 (2.9–11.5)	4.7 (2.7–11.3)	5.5 (2.9–11.4)	5.0 (3.1–11.9)	−6.0 (−8.2 to −3.9)		−0.6 (−0.9 to −0.3)
CKD at baseline
Yes (n=188)	15.4 (7.1–24.0)	12.1 (5.8–20.7)	16.8 (8.4–26.6)	15.7 (7.7–24.6)	−11.3 (−15.2 to −7.5)	0.01	−2.2 (−2.8 to −1.6)	<0.001
No (n=1095)	8.2 (3.7–14.8)	7.9 (3.3–14.3)	7.9 (3.7–15.6)	7.9 (3.6–16.0)	−5.9 (−7.6 to −4.3)		−0.5 (−0.7 to −0.2)
CVD risk at baseline
>10% (n=603)	16.1 (13.0–21.1)	15.3 (12.0–19.9)	17.8 (13.5–24.2)	18.5 (13.7–24.0)	−7.1 (−9.3 to −4.8)	0.69	−1.2 (−1.6 to −0.9)	0.002
<10% (n=680)	4.0 (2.8–5.9)	3.7 (2.5–5.7)	4.2 (2.9–6.3)	4.1 (3.0–6.5)	−6.0 (−8.1 to −3.9)		−0.2 (−0.5 to 0.1)

Baseline and week 24 risk values presented as median (interquartile range). Placebo-adjusted relative change and absolute change from baseline presented as mean (95% confidence interval). P values represents the subgroup×treatment×visit interaction. CVD, cardiovascular disease; T1D, type 1 diabetes.

Table 3.

Subgroup analyses of the Scottish Diabetes Research Network–estimated 10-year cardiovascular disease risk after 24 weeks of dapagliflozin or placebo

Subgroup	Dapagliflozin		Placebo		Placebo-Adjusted Relative Change from Baseline	P interaction	Placebo-Adjusted Absolute Change from Baseline	P interaction
	Baseline	Week 24	Baseline	Week 24
Age
>40 yr (n=679)	14.6 (9.9–21.4)	13.7 (9.1–20.1)	15.7 (9.6–26.2)	15.7 (10.5–27.1)	−6.7 (−9.0 to −4.5)	0.87	−1.1 (−1.5 to −0.7)	0.04
<40 yr (n=602)	3.9 (2.1–6.7)	3.6 (2.0–6.2)	4.2 (2.1–7.3)	4.3 (2.1–7.2)	−6.8 (−9.2 to −4.4)		−0.3 (−0.7 to 0.1)
Sex
Male (n=592)	9.6 (4.9–16.0)	8.8 (4.5–15.9)	10.9 (4.8–22.9)	11.7 (4.7–21.8)	−7.5 (−9.9 to −5.2)	0.70	−0.9 (−1.3 to −0.5)	0.43
Female (n=689)	8.4 (3.4–15.4)	7.7 (3.1–14.5)	8.0 (4.2–14.9)	7.8 (4.2–15.0)	−6.0 (−8.3 to −3.7)		−0.6 (−1.0 to −0.2)
T1D duration
>18 yr (n=622)	13.3 (7.7–21.1)	12.4 (7.4–19.5)	15.6 (8.4–27.9)	15.8 (8.4–27.8)	−6.8 (−9.2 to −4.5)	0.89	−0.9 (−1.3 to −0.5)	0.41
<18 yr (n=659)	5.4 (2.4–10.3)	4.9 (2.3–9.6)	5.9 (2.5–9.9)	5.6 (2.6–10.8)	−6.7 (−9.0 to −4.4)		−0.6 (−1.0 to −0.2)
CKD at baseline
Yes (n=187)	17.3 (8.9–26.9)	13.8 (8.0–23.3)	24.2 (9.5–34.2)	19.2 (9.8–34.6)	−11.9 (−16.0 to −7.8)	0.02	−2.3 (−3.0 to −1.7)	<0.001
No (n=1094)	8.1 (3.8–14.0)	7.5 (3.6–13.8)	8.0 (4.1–14.7)	8.1 (3.8–15.0)	−6.1 (−7.8 to −4.3)		−0.5 (−0.8 to −0.2)
CVD risk at baseline
>10% (n=601)	16.2 (12.3–24.1)	15.8 (11.4–22.3)	19.4 (12.2–29.1)	19.0 (12.7–30.0)	−6.9 (−9.3 to −4.5)	0.92	−1.2 (−1.6 to −0.8)	0.02
<10% (n=680)	4.3 (2.3–6.9)	4.2 (2.2–6.3)	4.6 (2.4–7.0)	4.6 (2.5–7.1)	−6.7 (−8.9 to −4.4)		−0.3 (−0.7 to 0.0)

Baseline and week 24 risk values presented as median (interquartile range). Placebo-adjusted relative change and absolute change from baseline presented as mean (95% confidence interval). P values represents the subgroup×treatment×visit interaction. CVD, cardiovascular disease; T1D, type 1 diabetes.

Table 4.

Subgroup analyses of the Steno T1 Risk Engine–estimated 5-year ESKD risk after 24 weeks of dapagliflozin or placebo

Subgroup	Dapagliflozin		Placebo		Placebo-Adjusted Relative Change from Baseline	P interaction	Placebo-Adjusted Absolute Change from Baseline	P interaction
	Baseline	Week 24	Baseline	Week 24
Age
>40 yr (n=806)	0.66 (0.49–1.01)	0.63 (0.46–0.96)	0.69 (0.46–1.02)	0.70 (0.49–1.07)	−4.1 (−8.3 to 0.2)	0.95	0.00 (−0.05 to 0.05)	0.93
<40 yr (n=613)	0.49 (0.36–0.66)	0.45 (0.33–0.62)	0.45 (0.37–0.67)	0.47 (0.37–0.65)	−5.0 (−9.9 to −0.1)		0.00 (−0.06 to 0.06)
Sex
Male (n=672)	0.66 (0.50–1.00)	0.61 (0.47–0.96)	0.68 (0.49–0.99)	0.69 (0.50–0.99)	−4.4 (−9.0 to 0.2)	0.74	0.01 (−0.05 to 0.06)	0.87
Female (n=747)	0.50 (0.37–0.74)	0.46 (0.33–0.72)	0.45 (0.35–0.74)	0.47 (0.35–0.74)	−4.6 (−9.1 to −0.1)		−0.01 (−0.06 to 0.05)
T1D duration
>18 yr (n=714)	0.69 (0.51–1.04)	0.65 (0.47–0.97)	0.70 (0.47–1.14)	0.74 (0.51–1.12)	−5.4 (−10.0 to −0.9)	0.35	0.02 (−0.04 to 0.07)	0.68
<18 yr (n=705)	0.48 (0.37–0.68)	0.46 (0.34–0.64)	0.46–(0.35–0.69)	0.47 (0.37–0.65)	−3.5 (−8.1 to 1.0)		−0.01 (−0.07 to 0.04)
CKD at baseline
Yes (n=232)	0.89 (0.56–1.59)	0.80 (0.51–1.72)	1.04 (0.59–2.16)	1.07 (0.59–2.30)	−6.8 (−14.4 to 0.9)	0.59	0.01 (−0.08 to 0.11)	0.99
No (n=1187)	0.54 (0.41–0.79)	0.52 (0.38–0.77)	0.51 (0.38–0.76)	0.52 (0.40–0.77)	−4.0 (−7.5 to −0.5)		0.00 (−0.05 to 0.04)
CVD risk at baseline
>10% (n=725)	0.72 (0.53–1.12)	0.70 (0.51–1.08)	0.77 (0.52–1.18)	0.77 (0.53–1.16)	−2.6 (−7.0 to 1.9)	0.43	0.03 (−0.03 to 0.08)	0.45
<10% (n=694)	0.47 (0.35–0.60)	0.43 (0.33–0.60)	0.44 (0.35–0.59)	0.46 (0.35–0.61)	−6.5 (−11.0 to −1.9)		−0.03 (−0.08 to 0.03)

Baseline and week 24 risk values presented as median (interquartile range). Placebo-adjusted relative change and absolute change from baseline presented as mean (95% confidence interval). P values represents the subgroup×treatment×visit interaction. CVD, cardiovascular disease; T1D, type 1 diabetes.

Discussion

In our current analysis, we used two validated risk prediction models to estimate CVD and ESKD risk in 1473 participants with T1D from the DEPICT-1 and DEPICT-2 trials randomized to once-daily placebo, dapagliflozin 5 mg, or dapagliflozin 10 mg at baseline, 24, and 52 weeks of treatment and after a 4-week drug washout. Compared with placebo, dapagliflozin significantly lowered relative risk of 10-year CVD as a percentage of baseline risk by approximately 6% on the basis of the SRE and by approximately 9% on the basis of the SDRN risk prediction model. Compared with placebo, dapagliflozin significantly lowered relative risk of 5-year estimated ESKD by approximately 8% from baseline after 52 weeks of treatment on the basis of the SRE. The greatest improvement in ESKD risk was observed 4 weeks after drug discontinuation, as evidenced by a placebo-adjusted approximately 13% relative risk lowering with dapagliflozin. Subanalysis of the SRE and SDRN estimated that 10-year CVD risk score revealed larger placebo-adjusted relative risk lowering in the participants with baseline CKD compared with participants without CKD.

The SRE has previously been used to simulate the effect of SGLT2 inhibition on CVD risk in 3660 adults with T1D treated at outpatient clinics in the Steno Diabetes Center, Copenhagen, by modifying risk variables in accordance with the cohort-level results of the published DEPICT and inTandem studies. By simulating risk change within this cohort, Stougaard et al.²⁸ demonstrated an approximately 5%–7% lowering of 10-year CVD risk as a percentage of baseline risk compared with placebo. They²⁴ subsequently assessed patient-level clinical data derived from the inTandem trials and demonstrated similar improvements in 10-year CVD risk after 24 weeks of sotagliflozin compared with placebo. A more recently developed risk prediction model by the SDRN suggested that SRE overestimated 10-year CVD risk, although the extrapolation of these results to evaluate treatment effects are unknown. In line with prior work, a 6% lowering in 10-year CVD risk was observed in the current study when using the SRE risk model. Notably, the SDRN 10-year CVD risk model yielded equivalent improvements in CVD risk. Furthermore, prior studies have been limited to assessing improvements in CVD risk after 24 weeks of SGLT2 inhibition. Analyses of CVD risk after 52 weeks of SGLT2 inhibition in this work demonstrated that the favorable lowering of CVD risk is maintained. Collectively, these findings could guide modeling strategies toward providing better insight into the anticipated cardiorenal protection provoked by SGLT2 inhibition, thus addressing a pertinent knowledge gap in the T1D population.

SGLT2 inhibitors increase proximal tubular natriuresis and Na⁺ delivery to distal tubular segments, activating tubuloglomerular feedback mechanisms that decrease glomerular pressure by vasoconstriction of the glomerular afferent arteriole.^17,29 The functional consequence is an initial and sustained reduction or dip in eGFR that has been well described with SGLT2 inhibition.³⁰ Importantly, several post hoc analyses on cardiovascular outcome trials in T2D have demonstrated that a larger SGLT2 inhibitor-mediated eGFR dip is associated with greater subsequent improvements in eGFR slope over time.³¹ For example, a post hoc analysis of the inTandem trials in participants with well-preserved kidney function suggested that SGLT inhibitor therapy preserved eGFR slope compared with placebo after accounting for kidney hemodynamic effects using a drug washout (+3.0 ml/min per 1.73 m² versus placebo),³² which was similarly observed in the DEPICT trials (Supplemental Figure 1). However, current ESKD risk models that are driven by GFR are not calibrated to assess the masked treatment effect with SGLT2 inhibition, and therefore acute and reversible hemodynamic dip in eGFR may erroneously lead to underestimation in the kidney protective effect of SGLT2 inhibition. To address this limitation, this study computed ESKD risk at 56 weeks (4 weeks off treatment) to determine whether the eGFR rebound can affect ESKD risk estimates. Our work and previous risk analyses report an approximately 5% lowering of 5-year ESKD risk after 24 weeks of SGLT2 inhibitor treatment.^24,28 After a 4-week drug washout, we observed the greatest improvement in ESKD risk by 13%. By accounting for this hemodynamic eGFR dip, these results may represent a more accurate estimation of ESKD risk improvements after SGLT2 inhibitor treatment, lending support for future kidney outcome trials to investigate this predicted benefit further. Future area of research may also include creation of models that are driven by eGFR to be applied to SGLT2 inhibitor–treated patients.

Despite the consistent cardiorenal protective effects of SGLT2 inhibitors in people with T2D,^16–18 clinical use of these medications in people with T1D has been hindered by the absence of data investigating cardiorenal outcomes, in addition to the previously reported increased risk of DKA associated with SGLT2 inhibition. Although this work demonstrated that SGLT2 inhibition improved predicted CVD risk, the low-risk nature of this cohort adds difficultly discerning whether the projected therapeutic benefit observed could outweigh the DKA risk. For example, CVD risk at baseline in the DEPICT cohorts was approximately 9%, with SGLT2 inhibition resulting in a 6% relative risk lowering and a 0.7% absolute risk lowering. After 52 weeks of treatment, the DKA incidence rate for participants receiving dapagliflozin and placebo were 3.7–4.8 versus 2.2 per 100 patient-years in DEPICT-1⁸ and 4.1–4.5 versus 0.4 per 100 patient-years in DEPICT-2,¹⁰ respectively. A meta-analysis of phase 3 randomized control trials investigating SGLT2 inhibition in people with T1D (DEPICT, Empagliflozin as Adjunctive to Insulin Therapy, and inTandem trials) observed a 3.5-fold higher risk of DKA compared with placebo,³³ with larger DKA risk observed when evaluating real-world data.^34–36 By contrast, a higher risk cohort, such as participants recruited in the Preventing Early Renal Loss in Diabetes trial (i.e., 51-year-old nonsmoking man living with T1D; body weight: 90 kg, diabetes duration: 35 years, glycated hemoglobin A1c: 8.2%, Systolic BP: 126 mm Hg, eGFR of 75 ml/min per 1.73 m² with microalbuminuria) would present with a 10-year CVD risk of 31%. Considering the relative risk lowering estimated in the current work using the entire cohort and CKD-specific cohort (i.e., 6%–11%), we speculate that larger anticipated absolute risk lowering could shift the benefit-risk ratio in support of clinical use. Our secondary analyses support this notion given that older participants and those at higher CVD risk (baseline CVD risk approximately 16%) demonstrated larger placebo-adjusted absolute risk lowering. This is also supported by recent evidence that patients with T1D and CKD do not present with a higher risk of DKA during SGLT2 inhibition,³⁷ in addition to contemporary DKA mitigation strategies proposed to enhance SGLT2 inhibitor safety in T1D.³⁸ Analogous interpretations can be inferred from ESKD risk observed in this study because the 5-year ESKD risk at baseline in the DEPICT cohort was <1%. For context, the event rate of an enriched cohort akin to that in canagliflozin and renal events in diabetes with established nephropathy clinical evaluation was approximately 20% in 3.5 years.³⁹ Thus, a more favorable benefit-risk ratio in T1D patients with CKD underscores the need for cardiorenal outcome trials in this population, a hypothesis currently being evaluated in clinical trials such as effectiveness and safety of sotagliflozin in slowing kidney function decline in persons with type 1 diabetes and moderate to severe diabetic kidney disease (NCT06217302).

Several methodological considerations warrant further discussion. First, several assumptions were made in this study as a result of missing data or data privacy agreements, including age, height, weight, and BMI. While some assumptions affected variables unlikely to change with treatment (i.e., age and height), the inability to incorporate participant-level changes in weight and BMI, given the established reductions in body weight with SGLT2 inhibition, implies that cardiorenal risk lowering may have been underestimated in this work. Second, our study was limited to risk models in cohorts with low baseline cardiovascular and kidney risk and therefore was significantly underpowered to investigate treatment responses in a risk-enriched cohort, necessitating further analysis in such cohorts. Third, the SRE and SDRN models are not designed to assess change in risk. Third, retrospective analyses have reported larger clinical benefit of SGLT2 inhibitors on hospitalization, heart failure, and CKD in T1D.⁴⁰ Input parameters used for both the SRE and SDRN models might not fully capture SGLT2 inhibitor–mediated changes in cardiorenal risk, which could lead to an underestimation in the overall treatment effect. Finally, the platform used to analyze clinical trial data automatically excludes participants who did not consent to data sharing^8,10; therefore, we are unable to assess the complete DEPICT-1 and -2 cohorts.

In conclusion, our risk modeling analyses of the DEPICT trials suggest that dapagliflozin improves estimated CVD and ESKD risk in participants with T1D at low cardiorenal risk. By highlighting the predicted cardiorenal risk benefit in a large low-risk T1D cohort, our findings underline the need to investigate the effect of these agents on hard cardiovascular and kidney end points in T1D, especially in high-risk cohorts. Dedicated trials are warranted to evaluate the cardiorenal efficacy and safety of SGLT2 inhibitors in patients with T1D with established kidney and CVD, for whom the risk-benefit ratio may be more favorable.

Supplementary Material

cjasn-20-529-s001.pdf ^{(224.5KB, pdf)}

cjasn-20-529-s002.pdf ^{(1.4MB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/CJN/C174.

Funding

None.

Author Contributions

Conceptualization: David Z.I. Cherney, Vikas S. Sridhar.

Formal analysis: Massimo Nardone.

Investigation: David Z.I. Cherney, Vikas S. Sridhar.

Methodology: Frederik Persson, Vikas S. Sridhar, Elisabeth B. Stougaard.

Project administration: David Z.I. Cherney.

Supervision: Vikas S. Sridhar.

Visualization: David Z.I. Cherney, Vikas S. Sridhar.

Writing – original draft: Luxcia Kugathasan, Massimo Nardone.

Writing – review & editing: Sean Barbour, David J.T. Campbell, David Z.I. Cherney, Alessandro Doria, Pritha Dutta, Tony K.T. Lam, Anita T. Layton, Adeera Levin, Leif Erik Lovblom, Istvan Mucsi, Massimo Nardone, Bruce A. Perkins, Frederik Persson, Remi Rabasa-Lhoret, Valeria E. Rac, Peter Senior, Ronald J. Sigal, Vikas S. Sridhar, Aleksandra Stanimirovic, Elisabeth B. Stougaard.

Data Sharing Statement

Previously published data were used for this study. This publication is based on research using data from data contributors AstraZeneca that has been made available through Vivli, Inc. Dandona P, Mathieu C, Phillip M, Hansen L, Tschöpe D, Thorén F, Xu J, Langkilde AM.; DEPICT-1 Investigators. Efficacy and Safety of Dapagliflozin in Patients With Inadequately Controlled Type 1 Diabetes: The DEPICT-1 52-Week Study. Diabetes Care. 2018; 41(12):2552–2559. doi:10.2337/dc18-1087. Epub 2018 Oct 23. PMID: 30352894. Mathieu C, Rudofsky G, Phillip M, Araki E, Lind M, Arya N, Thorén F, Scheerer MF, Iqbal N, Dandona P. Long-Term Efficacy and Safety of Dapagliflozin in Patients with Inadequately Controlled Type 1 Diabetes (the DEPICT-2 Study): 52-Week Results from a Randomized Controlled Trial. Diabetes Obes Metab. 2020; 22(9):1516–1526. doi:10.1111/dom.14060. Epub 2020 May 22. PMID: 32311204; PMCID: PMC7496089.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/CJN/C173.

Supplemental Methods

Supplemental Table 1. Variables in the SRE- and SDRN-based predictive models.

Supplemental Table 2. Missing in the SRE- and SDRN-based predictive models.

Supplemental Figure 1. eGFR at baseline, week 24, week 52, and week 56 (4 weeks off treatment) in people with T1D randomized to placebo or dapagliflozin.