Outcome Prediction of Acute Kidney Injury Biomarkers at Initiation of Dialysis in Critical Units

The ideal circumstances for whether and when to start RRT remain unclear. The outcome predictive ability of acute kidney injury (AKI) biomarkers measuring at dialysis initializing need more validation. This prospective, multi-center observational cohort study enrolled 257 patients with AKI undergoing renal replacement therapy (RRT) shortly after admission. At the start of RRT, blood and urine samples were collected for relevant biomarker measurement. RRT dependence and all-cause mortality were recorded up to 90 days after discharge. Areas under the receiver operator characteristic (AUROC) curves and a multivariate generalized additive model were applied to predict outcomes. One hundred and thirty-five (52.5%) patients died within 90 days of hospital discharge. Plasma c-terminal FGF-23 (cFGF-23) had the best discriminative ability (AUROC, 0.687) as compared with intact FGF-23 (iFGF-23) (AUROC, 0.504), creatinine-adjusted urine neutrophil gelatinase-associated lipocalin (AUROC, 0.599), and adjusted urine cFGF-23 (AUROC, 0.653) regardless whether patients were alive or not on day 90. Plasma cFGF-23 levels above 2050 RU/mL were independently associated with higher 90-day mortality (HR 1.76, p = 0.020). Higher cFGF-23 levels predicted less weaning from dialysis in survivors (HR, 0.62, p = 0.032), taking mortality as a competing risk. Adding cFGF-23 measurement to the AKI risk predicting score significantly improved risk stratification and 90-day mortality prediction (total net reclassification improvement = 0.148; p = 0.002). In patients with AKI who required RRT, increased plasma cFGF-23 levels correlated with higher 90-day overall mortality after discharge and predicted worse kidney recovery in survivors. When coupled to the AKI risk predicting score, cFGF-23 significantly improved mortality risk prediction. This observation adds evidence that cFGF-23 could be used as an optimal timing biomarker to initiate RRT.


Dialysis methods
Patients who needed the inotropic equivalent (IE) of more than 15 mcg/kg/min to maintain systolic blood pressure above 120 mmHg, received continuous venovenous hemofiltration (CVVH). Hemofiltration flow and blood flows were 25mL/kg/h and 200 mL/min, respectively. Replacement fluid was bicarbonate-buffered and predilutionally administered at a dynamically adjusted rate to achieve the desired fluid therapy goals. In patients who required an IE of 5-15 mcg/kg/min, sustained low-efficiency daily dialysis (SLEDD) or diafiltration (SLEDD-f) was used with a blood flow of 200mL/min, a dialysate flow of 300mL/min, and a hemofiltration flow of 25mL/kg/h. Duration of treatment was around 6-12 h depending on the amount of ultrafiltration. Intermittent hemodialysis was performed for four h (except for the first and second session) using low-flux polysulphone hemofilter (KF-18C, Kawasumi Laboratories, Japan), with a dialysate and blood flow of 500mL/min. (5,(7)(8)(9). RRT was performed via a double-lumen central venous catheter in all patients.

cFGF-23 and iFGF-23 measurement
The results yielded by the FGF-23 C-terminal kit were denoted as "cFGF-23" in the current study, which was the combination of the C-terminal fragments and intact FGF-23 (iFGF-23).(10)

Statistical methods
Continuous data were expressed as mean ± standard deviation (SD) and group comparisons were conducted using χ 2 tests for equal proportions, t tests for normally distributed data, and Wilcoxon rank sum tests otherwise. We generated receiver-operating characteristics (ROC) curves and calculated the area under the curve (AUC) to measure the performance of candidate criteria. Multiple comparisons were analyzed using one-way analysis of variance (ANOVA).
All the relevant covariates, including characteristics, comorbidities, laboratory data, at ICU admission, etiology of AKI, indication for dialysis, dialysis modality, SOFA score, and plasma cFGF-23 at dialysis, and some of their interactions, such as interventions listed in table 1, were put on a selected variable list to predict the outcome of interest. To avoid the extremely over-fitted models, we did not put in outcome, GCS, APACHE II, MODS and kidney function markers other than urine cFGF-23. The significance levels for entry (SLE) and stay (SLS) were conservatively set at 0.15. Then, with the aid of substantive knowledge, the best candidate final logistic regression model was identified manually by dropping the covariates with p > 0.05 until all regression coefficients were significantly different from 0.
Survival curves for all-cause mortality or liberation from dialysis were plotted from adjusted Cox models. For long-term dialysis, an individual who survived at index discharge was censored at death or the end of the study period.

Assessing the performance of prediction models, decision curve analysis (DCA)
Clinical usefulness and net benefit of the cFGF-23 were estimated according with decision curve analyses (DCA) (11), in order to identify patients who will have any of the adverse events evaluated. The DCA show the clinical usefulness of each new model based on a continuum of potential thresholds for adverse events (x-axis) and the net benefit of using the model to stratify patients at risk (y-axis) relative to assuming that no patient will have an adverse event. The basic interpretation of DCA is that the strategy with the highest net benefit at a particular threshold probability has the highest clinical value. In this study, the prediction models are represented by dot lines (AKI risk prediction score) and dashed lines (cFGF-23 and AKI risk prediction score). Those models that are the farthest away from the slanted horizontal grey line (i.e., assume none adverse event) and the slanted balck line (i.e., assume all adverse events) and the horizontal black line (i.e., assume none adverse event) demonstrate the higher net clinical benefit.
All analyses were performed with R software, version 3. Decision curve analysis (DCA) plot to assess the clinical consequences of screening AKI-D patients with 90 day mortality using cFGF-23 in addition to AKI risk prediction score (12). Y-axis is the net benefit of the decision strategy. Net benefit is the net proportion of patients with 90 day mortality who would be offered predicting model, without offering predicting model to patients with good outcomes. For patient at AKI at dialysis initialization, forecasting with the AKI risk predicting model with cFGF-23 would yield no net benefit. For risk thresholds between 20 and 80% the superior strategy is forecasting with the AKI risk prediction score with cFGF-23. For moderate to high-risk thresholds (0 to 20%), using the AKI risk prediction score with Sepsis-3 model would yield no net benefit above a predicting none strategy.  Abbreviations: AKI, acute kidney injury; BUN, blood urea nitrogen; CKD, chronic kidney disease; DM, diabetic mellitus,