Major cardiovascular (CV) outcomes trials of sodium-glucose cotransporter-2 (SGLT2) inhibitors, developed as anti-hyperglycemic drugs for the treatment of type 2 diabetes, have consistently demonstrated that these drugs reduce the relative risks of CV outcomes and slow the progression of kidney function decline in patients with type 2 diabetes at high CV risk (McGuire et al., 2021). The EMPA-REG OUTCOME trial was the first published trial to demonstrate the benefit of an SGLT2 inhibitor, empagliflozin, for CV and kidney protection (Barnett et al., 2014; Zinman et al., 2015).

The efficacy of SGLT2 inhibitors in preventing CV outcomes and delaying diabetic kidney disease progression is unlikely explained by improvement in glycemic control alone. Head-to-head comparison of the SGLT2 inhibitor canagliflozin with the sulfonylurea derivative glimepiride showed that at equal glycemic control, canagliflozin reduced the rate of kidney function decline (Heerspink et al., 2017). In patients without diabetes, the SGLT2 inhibitor dapagliflozin did not reduce glycated hemoglobin (HbA1c) but significantly reduced the risk of CV events and kidney failure (Persson et al., 2021).

Numerous studies have shown that SGLT2 inhibitors exert additional effects beyond their effects on HbA1c. These effects can contribute to long-term CV and kidney protection (van Bommel et al., 2017; Vallon and Verma, 2021). For example, SGLT2 inhibitors reduce glomerular hyperfiltration, systolic and diastolic blood pressure, improve tubular oxygen tension, and reduce body weight, uric acid, and albuminuria (Heerspink et al., 2016; Heerspink et al., 2019). Experimental and clinical studies have shown anti-inflammatory and anti-fibrotic effects, reflected by reductions in interleukin-6 and monocyte chemoattractant protein-1, although such effects may be secondary to improvement in glycemic control (Vallon et al., 2014; Tahara et al., 2017; Mancini et al., 2018).

Since SGLT2 inhibitors exert multiple effects on CV and kidney risk markers, we hypothesized that integrating the effects of empagliflozin on multiple CV or kidney risk factors may be able to better predict the effect of empagliflozin on these clinical outcomes than focusing on changes in a single risk marker. We have previously developed a multivariable drug efficacy prediction algorithm (i.e., the multiple Parameter Response Efficacy (PRE) score) to predict long-term drug effects on CV and kidney outcomes. Specifically, the PRE score was used to predict the efficacy of angiotensin receptor blockers, the endothelin receptor antagonist atrasentan, and the glucagon-like peptide-1 receptor agonists exenatide and liraglutide in patients with type 2 diabetes at high risk of CV disease or with established chronic kidney disease (Smink et al., 2014a; Schievink et al., 2015; Schievink et al., 2016; Idzerda et al., 2020; Tye et al., 2021) The objective of this study was to assess the accuracy of the PRE score in predicting the effect of the SGLT2 inhibitor empagliflozin on CV and kidney outcomes in patients with established CV disease who participated in the EMPA-REG OUTCOME trial. Secondly, we used the PRE score to prospectively predict the composite kidney and CV death outcome for the ongoing EMPA-KIDNEY trial.

In the EMPA-REG OUTCOME trial, a total of 7020 patients were randomly assigned to receive empagliflozin (N = 4687) or placebo (N = 2333) and were included in the intention-to-treat population. The EMPA-REG OUTCOME trial participants were characterized by a high CV risk and generally low risk of complications due to kidney disease. At baseline, the mean HbA1c was 8.1%, SBP was 135.4 mmHg, and all participants had a CV disease history. Mean eGFR was 74.0 ml/min/1.73m² and 1819 (25.9%) participants had an eGFR <60 ml/min/1.73m². Median UACR was 17.7 mg/g, and 2811 (40.0%) patients had a UACR ≥30 mg/g. Demographic and clinical characteristics of the patients were well balanced between the treatment and placebo group (Table 1).

In this study, we integrated multiple effects of the SGLT2 inhibitor empagliflozin on cardiorenal risk markers in a response efficacy score to predict the effect of empagliflozin on CV and kidney outcomes in the EMPA-REG OUTCOME trial. In the overall population, the PRE score underestimated the effect of empagliflozin on CV and kidney outcomes. However, in the subgroup analyses among individuals with UACR ≥30 mg/g, the PRE score accurately predicted the kidney outcome which was similar to the observed RRR achieved with empagliflozin.

The underestimation of the PRE score prediction suggests that the underlying mechanisms for CV and kidney protection are not fully captured by the clinical markers included in the current PRE score. The mechanism of action of SGLT2 inhibitors is complex and multifactorial and it is plausible that the set of readily available clinical markers cannot capture the multiple mechanistic pathways targeted by SGLT2 inhibitors (Heerspink et al., 2019; Sen and Heerspink, 2021). Indeed, studies have shown that SGLT2 inhibitors improve tubular oxygen tension, promote erythropoiesis through hypoxia-inducible factors, and improve cardiac metabolism and bioenergetics amongst others (Heerspink et al., 2016; Cowie and Fisher, 2020; Packer, 2021). Addition of biomarkers that specifically reflect these mechanistic pathways could improve efficacy prediction and could possibly be used as pharmacodynamic response biomarkers for SGLT2 inhibitors (Heerspink et al., 2019; Sen et al., 2021).

SGLT2 inhibitors were initially developed as oral glucose lowering drugs and approved by the regulatory agencies based on their HbA1c lowering effects (U.S. Food and Drug Administration, 2016). However, our analyses demonstrated that the reduction in HbA1c induced by empagliflozin largely underestimated the protective effect on CV and kidney outcomes. These results are in line with the results from other trials with SGLT2 inhibitors. In the EMPEROR-Reduced and EMPEROR-Preserved trials, empagliflozin reduced the risk of heart failure hospitalizations or CV death in heart failure patients with reduced or preserved ejection fraction regardless of diabetes status (Anker et al., 2021a; Anker et al., 2021b). Similarly, the benefits of dapagliflozin on clinical outcomes in patients with heart failure or chronic kidney disease were present independent of diabetes status (McMurray et al., 2019; Heerspink et al., 2020). Finally, in a head-to-head trial with glimepiride, the SGLT2 inhibitor canagliflozin slowed the decline in eGFR at equal glycemic control (Heerspink et al., 2017). Taken together, these findings suggest that the cardiorenal effect of SGLT2 inhibitors are not related to HbA1c changes alone and indicate the involvement of other non-glycemic pathways (Tuttle et al., 2021).

Albuminuria was a strong predictor of the effect of empagliflozin on kidney protective effects. Previous studies with the PRE score have also shown that the reduction in albuminuria was a strong contributor to the kidney protective effect of angiotensin receptor blockers, endothelin receptor antagonists and the SGLT2 inhibitor dapagliflozin (Smink et al., 2014b; Schievink et al., 2016; Idzerda et al., 2019). Interestingly, we showed that the effect of empagliflozin on kidney outcomes appeared to vary by baseline albuminuria. Among individuals with higher degree of albuminuria, the effect of empagliflozin was more pronounced and this effect was adequately predicted by the PRE score. Similar results have been observed in the CANVAS trial where the mediating effects of UACR were found to be significantly higher among patients with baseline UACR ≥30 mg/g, as compared to those with baseline UACR <30 mg/g (42 vs. 7%), and this suggests that the mechanism of SGLT2 inhibitors to protect the kidney may vary across patient subgroups (Li et al., 2020a). The early increase in hemoglobin was another important predictor of the benefit of empagliflozin. This finding is consistent with previous mediation analyses demonstrating that the change in hemoglobin explains a substantial proportion of the benefit of SGLT2 inhibitors on kidney and heart failure outcomes (Inzucchi et al., 2018; Li et al., 2020a; Li et al., 2020b). An increase in hemoglobin can be explained by the contraction in plasma volume as a result of SGLT2 inhibitor induced natriuresis and diuresis, but it can also reflect effects on hematopoiesis since SGLT2 inhibitors enhance erythropoietin production (Lambers Heerspink et al., 2013).

The EMPA-KIDNEY trial is the first dedicated kidney outcome trial with empagliflozin. The trial enrolled patients with chronic kidney disease with and without diabetes across a wide spectrum of eGFR and albuminuria values (Herrington et al., 2018). We estimated using the observed biomarker changes that empagliflozin will reduce the primary kidney outcome in EMPA-KIDNEY by 25%. We note, however, that these predictions are conditional on the actual population enrolled in the EMPA-KIDNEY trial and the actual observed risk marker changes. Our simulation analyses showed that if the effects of empagliflozin on UACR in the EMPA-KIDNEY population are larger than those observed in the EMPA-REG OUTCOME trial it may possibly result in a larger RRR of the primary kidney outcome. The ongoing EMPA-KIDNEY trial will soon deliver the answer whether our prediction of the treatment effect of empagliflozin was accurate.

There are several limitations in this study. First, the selection of parameters in the PRE score depends on the availability of the risk markers in the trial dataset. Second, the majority of the patients included in the EMPA-REG OUTCOME trial had mild to moderate chronic kidney disease and only few patients met the entry criteria for the EMPA-KIDNEY trial. As a result, the prediction of the EMPA-KIDNEY endpoint is based on a limited sample size and low precision. Furthermore, we note that the EMPA-KIDNEY trial enrolled patients with and without diabetes and our predictions only apply to diabetes patients since patients without diabetes were not included in the background population nor the EMPA-REG OUTCOME trial. Nevertheless, given the consistent effects of SGLT2 inhibitors in patients with and without diabetes this should not affect our findings (Wheeler et al., 2021).

To conclude, the PRE score underestimated the treatment effect of empagliflozin on CV and kidney outcomes in patients with type 2 diabetes and established CV disease. Whether the addition of novel biomarkers, which capture mechanistic pathways targeted by empagliflozin, to the PRE score improves treatment effect predictions requires further study.