Discordance between symptomatic, echocardiographic and renal trajectories after SGLT2 inhibitor initiation in heart failure: a landmark analysis of the prospective CaRD registry

Abstract

Objectives To quantify the concordance and stability of symptomatic, echocardiographic and renal trajectories after sodium-glucose cotransporter 2 inhibitor (SGLT2i) initiation and examine their association with subsequent events.Methods This day-210 landmark analysis included 618 adults with heart failure, baseline left ventricular ejection fraction (LVEF) <50%, and paired New York Heart Association (NYHA) class, LVEF and estimated glomerular filtration rate (eGFR) measurements on days 150–210. Domains were cross-classified rather than treated as equivalent causal components. Repeat classification was assessed by day 365. Death or heart-failure hospitalisation on days 211–365 was analysed with multivariable Cox regression; fixed-horizon Firth logistic regression and continuous, threshold, weighting and co-intervention analyses tested robustness.Results Discordant trajectories occurred in 368/618 patients (59.5%). Among 209 with repeat assessment, 181 (86.6%) retained the same classification (κ=0.74). Outcome status was complete in 576 patients, with 33 events. Non-concordance was associated with the time-to-event outcome (adjusted hazard ratio 4.27, 95% confidence interval 1.61–11.32; p = 0.004); the fixed-horizon Firth estimate was concordant (odds ratio 4.27, 95% confidence interval 1.64–11.13; p = 0.003). NYHA improvement contributed most of the prognostic signal. Adding non-concordance to a conventional baseline model increased the area under the curve from 0.752 to 0.790, but the bootstrap interval for the increment included zero.Conclusion Trajectories were frequently discordant and usually stable to day 365. Their association with events was non-causal, driven predominantly by symptomatic change and of uncertain incremental predictive value.