Renal and systemic hemodynamic effects of empagliflozin: Three randomized, double blind, placebo controlled cross-over trials

Introduction

Since the emergence of sodium-glucose cotransporter 2 inhibitors (SGLT2i) their precise mechanisms of action have been a topic of debate. According to the tubular hypothesis of hyperfiltration, SGLT2i can confer renal protection via changes in renal hemodynamic function [1]. SGLT2 is thought to play a key role in the pathophysiology of diabetic nephropathy, mediating reuptake of excessively filtered sodium and glucose thereby decreasing sodium delivery to the macula densa. This triggers a tubulo-glomerular feedback (TGF) response which dilates the afferent arteriole increasing renal blood flow (RBF) and glomerular pressure, leading to hyperfiltration and elevated glomerular filtration rate (GFR). Long term hyperfiltration can lead to glomerular injury and chronic kidney disease (CKD) [2]. Conversely, blockage of SGLT2 reduce sodium and glucose reuptake which increases distal sodium delivery, causing afferent arteriolar vasoconstriction, decreased RBF and alleviation of hyperfiltration, resulting in kidney protection [3–6]. Cherney et al. showed that treatment with empagliflozin reduced effective renal plasma flow (ERPF) and GFR in younger patients with type 1 diabetes (DM1), indicating ameliorated glomerular hyperfiltration [7] and studies in mice have demonstrated that empagliflozin can induce afferent arteriolar vasoconstriction and reduce hyperfiltration at the single nephron level [8].

In patients with DM2 and/or CKD initiation of SGLT2i is associated with an acute, but reversible, reduction in estimated GFR (eGFR) [9] whereafter it stabilizes, mitigating further decline in renal function [10, 11]. This acute eGFR reduction is not related to increased risk of renal events; on the contrary it predicts better outcomes [12, 13]. This seems to corroborate the above hypothesis, however none of the studies examining renal hemodynamic effects of SGLT2i in patients with DM2 have been able to document a decrease in RBF [14–16]. One explanation could be that the hemodynamic effects of SGLT2i differ between patients with DM1 and DM2 with post-glomerular vasodilation being more prominent in the latter [17].

Despite the extensive use of SGLT2i in patients with diabetic and non-diabetic CKD, their hemodynamic effects have not been studied in these groups. Therefore, we examined the renal and systemic hemodynamic effects of empagliflozin in patients with DM2 with and without concomitant CKD as well as in patients with non-diabetic CKD, reflecting the patient categories who are offered SGLT2i treatment. We employed a novel ⁸²Rubidium positron emission tomography/computed tomography (82Rb-PET/CT) method for estimating RBF, our primary endpoint. Secondary endpoints included GFR, 24-hour ambulatory blood pressure (BP), total peripheral vascular resistance (TVR) and renal vascular resistance (RVR). We hypothesized that empagliflozin would reduce RBF and decrease GFR and BP in all groups.

Methods

Design

We conducted three randomized, double-blind, placebo controlled, single centre cross-over studies including participants with DM2 and preserved kidney function (DM2 group), DM2 and concomitant CKD (DM2-CKD group) and non-diabetic CKD (CKD group). The design and methods of our project, the so called SiRENA project, have been previously published [18]. Here, we report the primary endpoint and selected secondary outcomes. The experimental setup was identical in the three studies (Figure 1). Briefly, participants were randomized to four weeks of treatment with either empagliflozin 10 mg daily or matching placebo and crossed over to four weeks of the opposite treatment after a wash out period of at least 2 weeks. Study endpoints were measured at an examination day at the end of each treatment period. If participants were on an SGLT2i at inclusion, it was discontinued at least two weeks prior to randomization. If necessary, it could be substituted to a different class of anti-diabetic treatment in accordance with national treatment guidelines. All other medications were kept unaltered. Four days prior to each examination, participants were encouraged to adhere to their usual diet and fluid intake was standardized.

• Download figure
• Open in new tab

Figure 1: Study design.

Schematic view of the study design

Study methods

RBF

RBF was estimated based on renal uptake of ⁸²Rb measured with ⁸²Rb-PET/CT scans. Scanning procedure and data analysis was performed as described by Langaa et al. [19, 20]. In short, a low dose CT scan was used for attenuation correction. Following bolus injection of 555 MBq of ⁸²Rb, an 8-minute dynamic PET-scan was performed with the abdominal aorta (AA) and both kidneys within the same field of view. Renal ⁸²Rb activity was estimated using image based iso-contouring of both kidneys. AA activity was defined by placing a volume of interest (VOI) within the proximal AA. Time activity curves were generated on the basis hereof (see Figure S1). We estimated RBF (ml/min/ccm) for each kidney as the single kidney K₁-value using a 1-tissue compartment model based on renal ⁸²Rb activity uptake with AA activity as input function. Mean RBF was estimated for each participant as the mean K₁-value. AA measurements were corrected for partial volume and spill-over and all measurements were β-corrected as has been previously described. Analyses were made using PMOD^© (PMOD Technologies Ltd., Zürich, Switzerland). Total RBF (ml/min) was assessed by multiplying RBF with kidney volume as estimated by the renal iso-contours using the formula:

• Download figure
• Open in new tab

Figure S1. ⁸²Rb-PET/CT renal scan.

A: Renal activity of ⁸²Rb with iso-contours of the right and left kidney (green and turquoise), coronal view. B and C: The abdominal aorta (AA) with volume of interest (VOI) (red) and a background box for partial volume correction (orange), sagittal and coronal views. D: Time activity curves of ⁸²Rb activity in the AA (red), right and left kidney (green and turquoise) and AA background (orange)

Effective renal plasma flow (ERPF) was calculated as

Results

Inclusion began in April 2021 and was completed by September 2022. Examinations were completed by December 2022. 49 participants completed the studies; 16 in the DM2 group, 17 in the DM2-CKD group and 16 in the CKD group. For further elaboration on inclusion and exclusion, see Figure 2. Baseline characteristics are described in Table 1.

• Download figure
• Open in new tab

Figure 2: Flow-chart of screening, inclusion, and exclusion for all studies.

Renal Blood Flow

Due to poor image quality, two participants were excluded from the analysis, one in the DM2 group and one in the DM2-CKD group. In the DM2-CKD group empagliflozin reduced RBF by 6% (geometric mean ratio: 0.94, 95% CI: 0.91; 0.97, p < 0.001). There were non-significant reductions of 4% in the DM2-group and 1% in the CKD-group (p = 0.29 and 0.72). Changes in RBF were not significantly different between groups (p = 0.16). There was no specific period (p = 0.53) or carry-over effect (p = 0.82). The RBF results are presented in Table 2 and Figure 3.

• Download figure
• Open in new tab

Figure 3: Renal blood flow. Median and interquartile range [IQR], black dot = geometric mean

View this table:

• View inline
• View popup
• Download powerpoint

Table 2:

Renal hemodynamics. FF: filtration fraction, RVR: renal vascular resistance, PGLO: intra-glomerular pressure R_a and R_e:afferent and efferent glomerular arteriolar resistance. P-values < 0.05 in bold.

GFR

Empagliflozin significantly reduced GFR in all groups as compared to placebo. GFR was reduced by 14% in the DM2 group (p < 0.001), by 9 % in the DM2-CKD group (p = 0.019) and by 11% in the CKD group (p = 0.018). Changes between groups did not significantly differ (p = 0.36). There was no period or carry over effect (p = 0.98). GFR results are presented in Table 2 and Figure 4.

• Download figure
• Open in new tab

Figure 4: Glomerular filtration rate (GFR). Median and interquartile range [IQR], black dot = geometric mean

Systemic hemodynamics

Empagliflozin treatment reduced 24-hour systolic blood pressure by 5 mmHg in the DM2 group (95%CI: −9; −1 mmHg, p = 0.015), as compared to placebo. The mean reduction was 7 mmHg (95%CI: −12; −2 mmHg, p = 0.0097) in the DM2-CKD group and 4 mmHg in the CKD-group (95%CI: −9; 0 mmHg, p = 0.066). Empagliflozin lowered TVR in all groups. There was no period or carry over effects (p = 0.70 and 0.79). Heart rate did not change. One patient from the DM2 groups was excluded from this analysis due to missing data. See Table 3 and Figure 5.

• Download figure
• Open in new tab

Figure 5: 24-hour systolic blood pressure (BP), and total vascular resistance (TVR) with mean and 95% confidence interval (95%CI).

View this table:

• View inline
• View popup
• Download powerpoint

Table 3:

Systemic hemodynamics. TVR: total vascular resistance. P-values < 0.05 in bold.

Discussion

Our main finding is that empagliflozin reduced RBF in patients with DM2 and concomitant CKD. There was a smaller, non-significant decrease in the DM2 group and no change in the CKD group. This is the first time SGLT2i have been shown to affect RBF in patients with DM2. Furthermore, empagliflozin decreased GFR in all groups, reflecting the known acute reduction. [11, 22]. Our findings support the hypothesis that SGLT2i alleviate hyperfiltration in diabetic nephropathy. While, per definition, no participants in the DM2-CKD group had hyperfiltration defined as elevated GFR, it could still be present at the single nephron level [23] and the decrease in RBF and GFR we observe could indicate that empagliflozin reverses the TGF mediated feedback response, normalizes afferent arteriolar tone and diminishes hyperfiltration.

Another important finding is that empagliflozin decreased TVR in all groups, regardless of DM2 or CKD status. That SGLT2i decreases TVR has been reported in DM1, but to our knowledge never in DM2 or CKD [24]. These findings suggest a novel anti-hypertensive mechanism, perhaps related to a direct vasodilatory effect [25, 26]. Furthermore, we observed a clinically meaningful decrease in ambulatory BP in all groups during empagliflozin treatment in line with the existing evidence [27]. The lack of statistical significance in the CKD group probably reflects a small sample size rather than an absence of effect in this population.

RBF decreased significantly in the DM2-CKD group only. Between-group differences should be interpreted cautiously as the project was not designed for such comparisons, and while there was no statistical difference in RBF changes between groups, the change seems most pronounced in the DM2-CKD group. Diabetic kidney disease is characterized by overexpression and over activation of SGLT2 [28, 29] and sodium reabsorption - the main driver of TGF - increases during episodes of hyperglycaemia [30, 31]. Our findings suggest that individuals with diabetes and CKD experience a more prominent TGF response to SGLT2i. It is noteworthy that while the relative effects of SLGT2i on renal outcomes are similar across different patient categories [32], there is greater absolute treatment benefit in DM2 patients with both albuminuria and reduced eGFR [33]. Interestingly, it is also in patients with both DM2 and CKD we observe a marked decrease in RBF. However, as we examined only short-term effects, we cannot say if the changes we describe persist during chronic treatment or how they relate to the long-term benefits.

In the CKD group RBF did not change. The hemodynamic effects of SGLT2i have not previously been examined in patients with non-diabetic CKD and though our limited sample size prevents us from reaching a clear conclusion, our results could indicate that SGLT2 mediated changes in RBF could be less pronounced non-diabetic CKD than in diabetic nephropathy. Similarly, SGLT2i induce TGF responses less consistently in animal models with non-diabetic CKD than in those with diabetic nephropathy [34].

GFR decreased in all groups, while RBF decreased significantly in the DM2-CKD group only, suggesting GFR changes are not solely dependent on RBF. While identifying the exact mechanisms by which SGLT2i impact GFR is beyond the scope of this paper, several potential explanations could be considered. One possibility relates to the interplay between afferent- and efferent arteriolar resistance. While GFR and RBF respond in parallel to changes in afferent arteriolar resistance, changes in efferent resistance can dissociate autoregulation of GFR from RBF [35]. It has previously been suggested that SGLT2i can alleviate hyperfiltration through dilation of the efferent arteriole, eliciting a decrease in GFR without affecting RBF [36]. However, GFR is not determined solely by glomerular pressure; intra-tubular pressure and properties of the glomerular filtration barrier could also be of importance. By inhibiting solute reabsorption from the distal tubule, SGLT2i could theoretically increase tubular pressure leading to reduced pressure gradient across the glomeruli and a decrease in GFR [37]. Additionally, animal studies have shown that dapagliflozin can directly affect podocyte function [38]. SGLT2i could thus potentially exert non-hemodynamic effects on GFR.

We calculated estimates of intra-renal hemodynamics to further explore the hemodynamic effects of SGLT2i. These calculations are indirect and can at best be indicative of the underlying physiology; the Gomez equations rely on assumptions of normal renal physiology which do not align with the characteristics of our cohort, particularly those with CKD. Furthermore, potential non-hemodynamic effects on GFR are not taken into account [21]. Nevertheless, as more direct methods in humans are lacking, they provide our best estimates. Using these estimates, we found that empagliflozin decreased FF, P_GLO and R_e in the DM2 group, which is in agreement with previous findings from van Bommel et al. who studied a comparable cohort of patients with DM2 and preserved kidney function [15]. This supports the current view that SGLT2i could exert effect in DM2 partly through dilation of the efferent glomerular arteriole. We saw a decrease in R_e in the CKD group as well, perhaps indicative of a similar mechanism.

Despite significant changes in RBF in the DM2-CKD group, we observed no change in RVR in this or any other group, nor did R_a change. This seemingly contrasts the results from Cherney et al., who found that empagliflozin decreased ERPF and increased RVR in patients with DM1 and hyperfiltration, indicating an increase in pre-glomerular resistance [7]. We do not, however, believe our findings preclude an SGLT2i induced effect the afferent arteriole. While R_a did not change, TVR decreased significantly. This is noteworthy, since we would have expected a parallel auto regulatory decrease in R_a to conserve RBF as has previously been described with for example angiotensin receptor blockers [39]. That R_a did not decrease while RBF did suggests a SLGT2i-induced relative afferent vasoconstriction. Our findings thus indicate a combined afferent and efferent arteriolar response. Similar mechanisms were described in a study in diabetic rats [40].

We measured RBF using a novel ⁸²Rb-PET/CT method. The method has low intra-assay variability and excellent inter-observer reliability, allowing for precise and reliable measurements of relative changes in perfusion [20]. When studying SGLT2i, it has the further advantage compared to para-amino-hippuric acid (PAH) of not being affected by glycosuria [41]. There are disadvantages to the method as well; ⁸²Rb-PET/CT likely underestimates perfusion at high perfusion rates [15]and the method cannot distinguish between cortical and medullar perfusion [19, 42]. Furthermore, ⁸²Rb is a potassium analogue and thus not an inert tracer and we have not taken potential urinary excretion into account [42]. Altogether, ⁸²Rb-PET/CT likely underestimates true RBF, which can explain why the measured values are lower than what have previously been described [15], in a cross over design, absolute values are of less importance as we examine changes within the same individual. Our project has other limitation; the small sample sizes make comparisons between groups difficult. Furthermore, changes in RBF were smaller and variability larger than anticipated in the power calculation, so we cannot rule out potential effects in the DM2 and CKD-groups had the sample sizes been larger. Inclusion in the DM2-CKD group did not require a biopsy verified diagnosis of diabetic nephropathy, so other aetiologies of CKD could be present. Likewise, in the CKD-group the cause of CKD was not reported in half the participants, limiting generalizability. We did not employ glycaemic clamping method nor were participants fasting. While this may add uncertainty to the results, an advantage is that our approach more accurately reflects both the patient populations and conditions in which SGLT2i would be used in a real-world setting.

In conclusion, short-term empagliflozin treatment significantly reduces RBF in patients with DM2 and concomitant CKD as compared to placebo, while GFR, BP and TVR was reduced in all groups. Our findings suggest that the renal hemodynamic effects of SGLT2i could be caused by relative pre-glomerular vasoconstriction, as well as post-glomerular vasodilation.

Funding

This project was funded by The Augustinus Foundation, The Research Foundation of the Central Denmark Region, The Medicine Fund of the Danish Regions, Gødstrup Hospital Research Fund and Boehringer-Ingelheim, who delivered the study medication. Boehringer-Ingelheim was given the opportunity to review the manuscript for medical and scientific accuracy as it relates to a Boehringer-Ingelheim substance, as well as intellectual property considerations. Boehringer-Ingelheim had no role in the design, analysis, interpretation of the results or the writing of this manuscript.

Authors contribution

All authors contributed to the manuscript. SFN, FHM, NHB and JNB designed the trials, which were conducted by SFN with support from CLD and FHM. SFN analyzed the data and interpretation was done in cooperation with FHM, JNB and NHB. SFN drafted the manuscript which was revised and accepted by all authors.

Conflict of interest disclosures

SFN has disclosed travel expenses covered by AstraZeneca. FHM disclosed advisory board participation and speaker honoraria from Boehringer-Ingelheim and AstraZeneca. JBN disclosed advisory board participation for Bayer, Boehringer-Ingelheim and AstraZeneca. No other disclosures were reported.

Data sharing

All study data will be made available upon request.