Introduction: Septic-shock-associated acute kidney injury (SSAKI) carries high morbidity in the pediatric population. Effective treatment strategies are lacking, in part due to poor detection and prediction. There is a need to identify novel candidate biomarkers of SSAKI. The objective of our study was to determine whether microarray data from children with septic shock could be used to derive a panel of candidate biomarkers for predicting SSAKI.
Methods: A retrospective cohort study compared microarray data representing the first 24 hours of admission for 179 children with septic shock with those of 53 age-matched normal controls. SSAKI was defined as a >200% increase of baseline serum creatinine, persistent to 7 days after admission.
Results: Patients with SSAKI (n = 31) and patients without SSAKI (n = 148) were clinically similar, but SSAKI carried a higher mortality (45% vs. 10%). Twenty-one unique gene probes were upregulated in SSAKI patients versus patients without SSAKI. Using leave-one-out cross-validation and class prediction modeling, these probes predicted SSAKI with a sensitivity of 98% (95% confidence interval (CI) = 81 to 100) and a specificity of 80% (95% CI = 72 to 86). Serum protein levels of two specific genes showed high sensitivity for predicting SSAKI: matrix metalloproteinase-8 (89%, 95% CI = 64 to 98) and elastase-2 (83%, 95% CI = 58 to 96). Both biomarkers carried a negative predictive value of 95%. When applied to a validation cohort, although both biomarkers carried low specificity (matrix metalloproteinase-8: 41%, 95% CI = 28 to 50; and elastase-2: 49%, 95% CI = 36 to 62), they carried high sensitivity (100%, 95% CI = 68 to 100 for both).
Conclusions: Gene probes upregulated in critically ill pediatric patients with septic shock may allow for the identification of novel candidate serum biomarkers for SSAKI prediction.