Skip to main content
. 2019 May 18;4(8):1131–1142. doi: 10.1016/j.ekir.2019.05.005

Table 3.

Utility of peak postoperative biomarker levels expressed as area under the receiver operating characteristics curve for the prediction of postoperative in-hospital AKI

Postoperative 6 h–POD 1
 Postoperative 6 h–POD 4
AUC (95% CI) Pa AUC (95% CI) Pa
Clinical model 1 0.77 (0.65–0.90) ref 0.77 (0.65–0.90) ref
 + Serum cystatin C 0.82 (0.70–0.93) 0.088 0.85 (0.74–0.95) 0.057
 + Urine NGAL/Cr 0.80 (0.68–0.92) 0.15 0.81 (0.69–0.93) 0.16
 + Urine KIM-1/Cr 0.80 (0.67–0.92) 0.19 0.83 (0.72–0.94) 0.064
 + Combination 1
Serum cystatin C + urine KIM-1/Cr		 0.83 (0.73–0.93) 0.049 0.83 (0.73–0.94) 0.036
 + Combination 2
Serum cystatin C + urine KIM-1/Cr		 0.80 (0.68–0.91) 0.32 0.81 (0.68–0.93) 0.24
Clinical model 2 0.56 (0.42–0.70) ref 0.56 (0.42–0.70) ref
 + Serum cystatin C 0.69 (0.56–0.82) 0.14 0.78 (0.66–0.90) 0.019
 + Urine NGAL/Cr 0.62 (0.48–0.76) 0.21 0.68 (0.55–0.81) 0.075
 + Urine KIM-1/Cr 0.63 (0.50–0.76) 0.29 0.74 (0.63–0.85) 0.023
 + Combination 1
Serum cystatin C + urine KIM-1/Cr		 0.68 (0.56–0.80) 0.006 0.68 (0.56–0.80) 0.007
 + Combination 2
Serum cystatin C + urine KIM-1/Cr		 0.64 (0.49–0.78) 0.10 0.71 (0.57–0.84) 0.010

AKI, acute kidney injury; AUC, area under the receiving operating characteristic curve; CI, confidence interval; Cr, creatinine; KIM-1, kidney injury molecule 1; NGAL, neutrophil gelatinase-associated lipocalin; POD, postoperative day.

Clinical model 1 consisted of 3 preoperative clinical parameters that either were significantly associated with AKI in univariate analysis (preoperative left ventricular ejection fraction percent and body mass index >30 kg/m2) or were clinically relevant for the model (preoperative estimated glomerular filtration rate <60 ml/min per 1.73 m2). Clinical model 2 was the Cleveland Clinic score validated for predicting AKI–renal replacement therapy following cardiac surgery.30 Duplets of biomarkers were combined as follows: (i) at least 1 biomarker above the cutoff value (combination 1), or (ii) both biomarkers above the cutoff value (combination 2). Biomarker data were log10 transformed.

a

P value denotes comparison between clinical model + biomarker(s) vs clinical model alone.