Influence of patients' clinical features at intensive care unit admission on performance of cell cycle arrest biomarkers in predicting acute kidney injury.

ABSTRACT

Objectives: Identification of acute kidney injury (AKI) can be challenging in patients with a variety of clinical features at intensive care unit (ICU) admission, and the capacity of biomarkers in this subpopulation has been poorly studied. In our study we examined the influence that patients' clinical features at ICU admission have over the predicting ability of the combination of urinary tissue inhibitor of metalloproteinase-2 (TIMP2) and insulin-like growth factor binding protein 7 (IGFBP7). Methods: Urinary [TIMP2]•[IGFBP7] were measured for all patients upon admission to ICU. We calculated the receiver operating characteristics (ROC) curves for AKI prediction in the overall cohort and for subgroups of patients according to etiology of ICU admission, which included sepsis, trauma, neurological conditions, cardiovascular diseases, respiratory diseases, and non-classifiable causes. Results: In the overall cohort of 719 patients, 239 (33.2%) developed AKI in the first seven days. [TIMP2]•[IGFBP7] at ICU admission were significantly higher in AKI patients than in non-AKI patients. This is true not only for the overall cohort but also in the other subgroups. The area under the ROC curve (AUC) for [TIMP2]•[IGFBP7] in predicting AKI in the first seven days was 0.633 (95% CI 0.588-0.678), for the overall cohort, with sensitivity and specificity of 66.1 and 51.9% respectively. When we considered patients with combined sepsis, trauma, and respiratory disease we found a higher AUC than patients without these conditions (0.711 vs. 0.575; p=0.002). Conclusions: The accuracy of [TIMP2]•[IGFBP7] in predicting the risk of AKI in the first seven days after ICU admission has significant variability when the reason for ICU admission is considered.