The main result of this study is that treatment with canagliflozin after hospitalization for HF is associated with lower readmission due to HF. These results were consistent irrespective with LVEF (≤ 40% or > 40%).
The incidence of HF hospitalization was much higher in control group and it possible can not only explain with the addition of canagliflozin to their treatment. Other variables such as NT-ProBNP concentration at discharge or previous HF hospitalization might influence in our final results. However, we performed a multivariate statistical analysis and the differences were maintained.
The magnitude of the effect of canagliflozin seems to appear in the first months after discharge. Kaplan-Meir cumulative readmission curves in both groups started to dissociate at first months (Fig. 1, 2). Perhaps that swift difference is explained with hemodynamic effect and improvement of loading conditions of canagliflozin. Nevertheless, these differences were increasing almost until the first year. This evidence supports that it might exist other mechanisms which have been proposed previously (17–22).
Pivotal trials with empagliflozin, dapagliflozin and canagliflozin have shown lower rate of admission for HF in patients with T2D (7–9). Dapagliflozin is the first SGLT2 inhibitor that has demonstrated a reduction in hospitalization and urgent visit resulting in intravenous therapy for HF in patients with stablished HFrEF (10).
Little data is available about the effect of SGLT2 inhibitors after hospitalization for HF. One pilot randomized study with empagliflozin has shown a reduction in a combined endpoint of worsening HF, rehospitalization for HF or death at 60 days [4 (10%) vs. 13 (33%); p = 0.014] and also empagliflozin was safe, well tolerated, and had no adverse effects on blood pressure or renal function (25). A post hoc analysis of the EMPA-REG OUTCOME trial with patients who were hospitalized for HF after randomization has demonstrated a reduction in readmission rate at 45, 60 and 90 days. The percentages of patients with HF rehospitalization or cardiovascular death within 90 days were 22.1% in control group and 11.1% in empagliflozin group (26). The CVD-REAL 2 study (27), which examine cardiovascular outcomes in a large international cohort with 2,581,980 patients with diabetes and initiated SGLT2 inhibitors versus other glucose-lowering drugs ,observed over 441,357 person-years of follow-up, 2,646 events of HF hospitalization in the SGLT2 inhibitors group and 3,351 in control group (HR: 0.64; 95% CI: 0.50 to 0.82; p = 0.001). Although these published results are similar than those of our study, the rate of readmissions was appreciably higher, 28.9% in canagliflozin group versus 56.1% in control group (HR: 0.45; 95% CI: 0.21–0.96; p < 0.039). This fact might be explained by a longer follow-up (median 22 months) and the particular clinical characteristics of patients in the cohort. To our knowledge, this is not only the first study that quantified the benefit of canagliflozin after hospitalization for HF, but also the first study with SGLT2 inhibitors in this clinical context within a follow-up of at least one year.
Potential clinical predictors of HF readmissions are well studied .We identified coronary artery disease, NT-ProBNP concentration, previous hospitalization for heart failure and NYHA class as posible risk factors for HF readmission. However, HFrEF has not been associated with major rates of readmissions; it highlighted the poor clinical outcomes of HF with preserve ejection fraction as well. These results are aligned with those previously reported in literatura (28).
We also found that serum concentrations of NT-proBNP, with proven prognostic value for HF and diabetes (23, 29), experienced a mayor reduction in canagliflozin group and those levels were sustained over follow-up period. There are very limited data on the effects of canagliflozin on cardiovascular biomarkers. A post hoc analysis with 666 patients included in the CANVAS Program showed that NT-ProBNP concentrations increased with placebo and minimally with canagliflozin over 2-year follow-up and from a baseline median of 47 pg/ml. The effect was observed at 26 weeks and persisted over 104 weeks (nominal p < 0.05 at weeks 26 and 52, nominal p < 0.01 at week 104). In our cohort, baseline of NT-ProBNP concentrations were 5036.4 pg/ml in control group and 3763.5 pg/ml in canagliflozin group, as a consequence we can compare the real effect in cardiovascular biomarkers of canagliflozin in patients with established HF.
Several limitations of the study should be noted. First, it was not a randomized controlled trial, and it poses the limitations inherent to its retrospective design. It has become increasingly evident when we compare medical therapy. The information about the reasons why physicians did not prescribe SGLT2 inhibitors to some patients at discharge was not available. Although, we tried to adjust the potential confounding factors by multivariable analysis. Other unmeasured, residual variables as well as selection bias could not be completely controlled.
Secondly, the sample size of the cohort was small as a result of stricted inclusion criteria and a single center, but the number of events was higher than expected, even though this factor might reduce the strength of our findings.
Thirdly, difference in NT-ProBNP concentration at discharge in both groups might interact in repeated measures analysis, even after adjustment.
Finally, among the patients with HFrEF, use of sacubitril–valsartan was similar besides low in both groups, which could be a potencial factor for readmissions in those patients.