Predictive performance of kinetic eGFR, midregional proadrenomedullin, and H3.1 nucleosomes for acute kidney disease in sepsis: a secondary analysis of a large multicentre randomized controlled trial

BACKGROUND: Sepsis-associated acute kidney injury (SA-AKI) frequently progresses to acute kidney disease (AKD) and is linked to poor outcomes. We evaluated whether combining biomarkers reflecting complementary pathophysiological domains-histone H3.1 nucleosomes (cellular injury), midregional pro-adrenomedullin (MR-proADM; endothelial dysfunction), and kinetic estimated glomerular filtration rate (kinetic eGFR; dynamic renal function)-improves prediction of AKD and renal recovery. METHODS: This secondary analysis of the multicentre randomized SISPCT trial included 690 patients with sepsis after exclusion of patients with pre-existing renal replacement therapy or missing AKI data. Biomarkers and kinetic eGFR were assessed at baseline, day 2, and day 7. Multivariable logistic regression models adjusted for age, sex, and non-renal SOFA score were used to assess associations with AKD. Predictive performance was evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC). Calibration for the combined models for each day was evaluated using bootstrap-corrected calibration plots. Clinical utility was assessed using decision curve analysis with fivefold cross-validation. RESULTS: AKD occurred in 196 patients (28.4%). At baseline, only MR-proADM was independently associated with AKD (adjusted OR 1.17, 95% CI 1.08-1.28; p < 0.001), whereas H3.1 and kinetic eGFR were not. At day 2 and day 7, only changes in kinetic eGFR were independently associated with AKD (both p < 0.001). Discriminative performance increased over time, with AUCs for the combined model of 0.65 at baseline, 0.70 at day 2, and 0.76 at day 7. At later time points, kinetic eGFR consistently showed the highest discriminative performance based on the point estimates of the AUC. An overall good calibration performance was shown. Decision curve analysis demonstrated only modest and inconsistent additional clinical net benefit of the combined model compared with kinetic eGFR alone. AKD was associated with increased 90-day mortality (55% vs. 24%; risk ratio 2.3, 95% CI 1.9-2.9). Biomarkers showed limited and inconsistent performance for prediction of renal recovery. CONCLUSIONS: In patients with sepsis, MR-proADM at baseline and dynamic changes in kinetic eGFR during the first week were independently associated with AKD, whereas histone H3.1 nucleosomes provided limited predictive value. Although multimarker models seemed to modestly improve baseline discrimination, this advantage was not sustained over time and added little clinical benefit beyond kinetic eGFR. These findings highlight the value of dynamic renal function assessment for risk stratification of SA-AKI progression to AKD. Trial registration ClinicalTrials.gov Identifier: NCT00832039.