Short‐term outcomes after sodium‐glucose cotransporter‐2 inhibitor initiation in a cohort of heart failure patients

Abstract

Aims

Sodium‐glucose cotransporter‐2 inhibitors (SGLT2i) decrease mortality and risk of hospitalization in patients with heart failure with reduced ejection fraction (HFrEF). SGLT2i have a natriuretic effect shortly after initiation, followed by a lasting osmotic diuretic effect. We sought to evaluate rates of acute kidney injury (AKI) and therapy discontinuation with SGLT2i initiation in a real‐world cohort of HFrEF patients.

Methods and results

We abstracted data on 200 patients with HFrEF initiated on a SGLT2i in the outpatient setting at the University of Michigan (between 1 July 2016 and 2 July 2022). Our co‐primary endpoints were rate of AKI and discontinuation of SGLT2i. A total of 200 patients were included. The majority of patients were male (64%) with a mean left ventricular ejection fraction (LVEF) of 27%. One hundred and four (52%) patients had diabetes mellitus. Most patients exhibited New York Heart Association class II (51.5%) or III (33.5%) symptoms. The majority of patients (54%) were taking an angiotensin‐receptor neprilysin inhibitor. The mean daily furosemide equivalent diuretic dose was 93.3 mg. AKI occurred in 22 patients and 18 patients discontinued their SGLT2i. Yeast infection (n = 6), hypotension (n = 5), and AKI (n = 4) were the most common reasons for discontinuation. Using receiver operating characteristic curve analysis, the strongest models for AKI were A1C [area underneath its curve (AUC) = 75.8, empirical confidence interval (ECI) 66.5–83.5]; baseline serum creatinine (SCr) (AUC = 72.0, ECI 65.7–78.7); LVEF (AUC = 67.6, ECI 58.4–75.8); and furosemide equivalent diuretic dose (AUC = 66.0, ECI 57.5–74.6). Similarly, the strongest positive models for SGLT2i discontinuation were A1C (AUC = 81.1, ECI 74.8–87.2); baseline SCr (AUC = 67.4, ECI 58.7–75.5); LVEF (AUC = 68.7, ECI 58.9–76.5); and furosemide equivalent diuretic dose (AUC = 67.2, ECI 58.2–76.0).

Conclusions

A1C was the strongest model of AKI, and SGLT2i discontinuation in HFrEF patients started on SGLT2i. Glucosuria may be related to this effect. Patients with higher baseline SCr on higher doses of loop diuretics may be at greater risk of these outcomes. Future prospective studies will be needed to further evaluate these findings and other models of AKI and SGLT2i discontinuation to guide clinical use of SGLT2 inhibitors.

Introduction

In recent clinical trials, sodium‐glucose cotransporter‐2 inhibitors (SGLT2i) have been shown to reduce morbidity and mortality in patients with heart failure with reduced ejection fraction (HFrEF) with and without diabetes mellitus. However, few studies have focused on the safety of these medications outside of use in clinical trials. For example, little is known about the impact of loop diuretic dose on the risk of adverse events after initiating SGLT2is. SGLT2is have both osmotic and natriuretic diuretic effects that can contribute to changes in renal function from a reduction in effective circulatory volume. ¹

Aims

We aimed to evaluate the safety of SGLT2i initiation in patients with HFrEF in a real‐world setting within the Advanced Heart Failure clinics at the University of Michigan.

Methods

This was a retrospective cohort study of 200 adult patients with HFrEF initiated on a SGLT2i between 1 July 2016 and 2 July 2022 in the University of Michigan Health System outpatient Heart Failure Clinics. Adult (age ≥ 18 years) outpatients with HFrEF initiated on a SGLT2i and meeting the necessary 30 day follow‐up laboratory evaluation were included. Patients with heart failure with preserved ejection fraction and those whom SGLT2i were initiated as an inpatient were excluded. Additionally, 19 patients were excluded from the study due to lack of laboratory information within the established 30 day cut off. The co‐primary outcomes were SGLT2i discontinuation and acute kidney injury (AKI). AKI was defined as an increase in serum creatinine (SCr) ≥ 0.3 mg/dL or an increase to ≥1.5 times baseline SCr upon laboratory collection within the established time period. ² A reduction in urinary volume to <0.5 mL/kg/h for 6 h was not utilized given the retrospective data collection and absence of real time measurement. Baseline values were obtained from laboratory tests performed within a 30 day period prior to initiation of the SGLT2i. Continuous variables were reported using mean and standard deviation; categorical variables were reported as a frequency or percentage. Receiver operating characteristic (ROC) curves were used to measure each model's ability to represent an outcome. They include the ability to consider error rates simultaneously with accuracy rates, captured by sensitivity and specificity, in addition to just modelling accuracy rates. The ROC curve maps a model's sensitivity to one minus its specificity, and the area underneath its curve (AUC) provides a measure of fidelity that can be compared with other models. The point estimate used was the median of the 1000 multiple 10‐fold cross‐validation AUCs, and the 95% confidence intervals were determined empirically by the 2.5th and 97.5th percentiles of the 1000 iterations. R statistical software (4.2.2) was utilized for data analysis. ³

Results

A total of 200 patients met inclusion criteria. The majority were male (64.5%) with a mean age of 60.8 years and mean left ventricular ejection fraction (LVEF) of 27% (Table  1 ). Approximately half of the participants were initiated on empagliflozin, and the others were started on dapagliflozin. A total of 104 (52%) patients had diabetes mellitus with a mean A1C level of 7.4% (±1.62). The mean baseline SCr was 1.16 ± 0.32, and the mean 30 day SCr was 1.27 ± 0.75. The mean daily furosemide equivalent diuretic dose was 93.3 mg.

Table 1.

Baseline characteristics

Sample population	N = 200
Age (years), mean (SD)	60.82 (±13.68)
Male, n (%)	129 (64.5)
Race
Caucasian, n (%)	157 (78.5)
Black, n (%)	32 (16)
SGLT2 inhibitor
Empagliflozin, n (%)	102 (51)
Dapagliflozin, n (%)	98 (49)
Cardiomyopathy type
ICM, n (%)	71 (35.5)
NICM, n (%)	126 (63)
Diabetes, n (%)	104 (52)
Haemoglobin A1C(mg/dL), mean (SD)	7.4 (±1.62)
LVEF (%),mean (SD)	27. 2 (±8.91)
New York Heart Association class, n (%)
I	28 (14)
II	103 (51.5)
III	67 (33.5)
IV	2 (1)
Beta‐blocker use, n (%)	186 (93)
ACEi/ARB, n (%)	80 (40)
ARNi, n (%)	109 (54.5)
MRA, n (%)	161 (80.5)
Diuretic use
Bumetanide, n (%)	18 (9)
Furosemide, n (%)	101 (50.5)
Torsemide, n (%)	47 (23.5)
Furosemide equivalent (mg), mean (SD)	93.3 (±115.5)
Diuretic dose change upon SGLT2i initiation
Reduced, n (%)	19 (9.5)
Increased, n (%)	1 (0.5)
No changes, n (%)	180 (90)
Baseline serum creatine (mg/dL), mean (SD)	1.16 (±0.32)
30 day follow up serum creatinine (mg/dL), mean (SD)	1.27 (±0.75)
Baseline potassium (mEq/L), mean (SD)	4.45 (±0.44)
30 day follow up potassium level (mEq/L), mean (SD)	4.21 (±0.41)

%, percentage calculated off total cohort; ACEi, angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; ARNi, angiotensin receptor neprilysin inhibitor; ICM, ischaemic cardiomyopathy; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; NICM, non‐ischaemic cardiomyopathy; SGLT2i, sodium‐glucose cotransporter‐2 inhibitors.

From the cohort, 22 patients (11%) had AKI, and 18 patients (9%) discontinued their SGLT2i (Table  2 ). Yeast infection (33%), hypotension (28%), and AKI (22%) were the most common reasons for discontinuation. Fourteen patients (7%) had side effects following initiation of their SGLT2i though did not require discontinuation of the medication. Side effects included hypotension (4%), gout flares (0.5%), chest pain (0.5%), and hyponatremia (0.5%).

Table 2.

Outcome rates and reason for SGLT2i discontinuation

Primary outcome	N = 200
AKI, n (%)	22 (11)
SGLT2i discontinuation, n (%)	18 (9)
Reason for discontinuation	N* = 18
AKI, n (%*)	4 (22)
DKA, n (%*)	1 (5)
Hypotension, n (%*)	5 (28)
Yeast Infection, n (%*)	6 (33)
Other, n (%*)	2 (11)

%*, percentage calculated off cohort who had SGLT2i discontinuation; %, percentage calculated off total cohort; AKI, acute kidney injury; DKA, diabetic ketoacidosis; N*, patients who underwent SGLT2i discontinuation; N, total cohort, SGLT2, sodium‐glucose cotransporter‐2 inhibitors.

Using ROC curve analysis, the strongest models for AKI were A1C [AUC = 75.8, empirical confidence interval (ECI) 66.5–83.5]; baseline SCr (AUC = 72.0, ECI 65.7–78.7); LVEF (AUC = 67.6, ECI 58.4–75.8); and furosemide equivalent diuretic dose (AUC = 66.0, ECI 57.5–74.6; Figure 1 ). The strongest positive models for SGLT2i discontinuation were A1C (AUC = 81.1, ECI 74.8–87.2); baseline SCr (AUC = 67.4 ECI 58.7–75.5); LVEF (AUC = 68.7 ECI 58.9–76.5); and furosemide equivalent diuretic dose (AUC = 67.2 ECI 58.2–76.0; Figure 1 ). Ten per cent of patients had changes to their diuretic dose following SGLT2i initiation with 19 patients (9.5%) having their diuretic dose decreased and 1 patient (0.5%) with a diuretic dose increase.

Figure 1.

Models of AKI and SGLT2i discontinuation assessed with ROC AUC forest plot analysis using 1000 iterations. ACEi, angiotensin‐converting enzyme inhibitor; AKI, acute kidney injury; ARB, angiotensin receptor blocker; ARNi, angiotensin receptor neprilysin inhibitor; AUC, area underneath its curve; BB, beta blocker; DM, diabetes mellitus; LVEF, left ventricular ejection fraction; MRA, mineralocorticoid receptor antagonist; RAAS, renin‐angiotensin‐aldosterone system; ROC, receiver operating characteristic; SCr, serum creatinine; SGLT2, sodium‐glucose cotransporter‐2.

Discussion

This study revealed outpatient SGLT2i initiation was well tolerated by patients with HFrEF. Haemoglobin A1c, baseline SCr, LVEF, and furosemide equivalent diuretic dose were the strongest models for AKI and SGLT2i discontinuation.

The rate of AKI in our study was low, though difficult to compare with previous literature as randomized studies reported a composite renal endpoint. ⁴ , ⁵ In our study, it is possible that high furosemide equivalent diuretic doses contributed to AKI. Though past studies have evaluated baseline furosemide equivalents and outcomes following SGLT2i initiation, ⁶ the mean diuretic dose in this study was higher. This highlights the need to determine whether patients on higher furosemide equivalent diuretic doses should have their loop diuretic doses reduced prior to SGLT2i initiation.

Another factor that may impact the diuretic effect of SGLT2i is haemoglobin A1c. Previous literature has reported that the SGLT2i diuretic effect is glycaemia‐dependent, though this association has not been reported with haemoglobin A1C. ⁷ In those with hyperglycaemia, there may be an increase in urinary glucose excretion leading to an osmotic diuretic effect. Future studies should evaluate whether loop diuretic dose reductions are needed with SGLT2i initiation among patients with higher A1c levels.

This study should be interpreted in the context of several limitations. First, the sample size was small, and it represents a single‐centre study. Second, the sample of patients for this study was limited by the availability of electronic medical record data to confirm medication initiation and adverse events.

In conclusion, real‐world initiation of SGLT2i in patients with HFrEF was well tolerated. Models for AKI and SGLT2i discontinuation included haemoglobin A1c, baseline SCr, LVEF, and furosemide equivalent diuretic dose.

Conflict of interest

None declared.

Pérez Martínez, B. O. , Adie, S. K. , Marshall, V. D. , and Konerman, M. C. (2023) Short‐term outcomes after sodium‐glucose cotransporter‐2 inhibitor initiation in a cohort of heart failure patients. ESC Heart Failure, 10: 3223–3226. 10.1002/ehf2.14489.