Acute Kidney Injury in the Intensive Care Unit: Prediction of Severity and Outcome

Abstract

Acute Kidney Injury (AKI) is common and contributes to high mortality amongst critically ill patients. The overall aim of this project was to investigate functional and structural biomarkers for AKI detection and prediction of its severity and outcome in critically ill patients. The role of current available diagnostic tools, which include plasma creatinine, and urine output, and new emerging AKI biomarkers such as plasma and urinary Neutrophil Gelatinase-Associated Lipocalin (NGAL) and Cystatin C (CysC), urinary γ-Glutamyl Transpeptidase (GGT), Alkaline Phosphatase (AP), α- and π-Glutyamyl S-Transferase (GST), and albumin were investigated. Normalisation of urinary biomarker concentrations to urinary creatinine concentration is commonly used to account for variations in water reabsorption. The accuracy of this method is compromised by tubular secretion of creatinine and variations in urinary creatinine excretion in non-steady state when GFR changes. Alternatives include using the absolute concentration or quantifying the excretion rate. We compared these in post-hoc analysis of 484 ICU patients, in chapter 3. Absolute concentration performed best for diagnosis of AKI on admission, but normalised concentrations performed best for prediction of mortality, dialysis or subsequent development of AKI. Excretion rate did not diagnose or predict outcomes better than absolute or normalised concentration, but the total biomarker excretion over 24 hours was associated with AKI severity and poorer survival, hence may be useful as outcome variables in clinical trials of AKI. AKI biomarkers promise earlier diagnosis. However, their performances are influenced by the duration from insult until time of measurement. Earlier measurement of biomarkers following insult may better diagnose AKI compared to later measurement. This was investigated in 77 patients following cardiac arrest, sustained or profound hypotension or ruptured abdominal aortic aneurysm, in chapter 4. Almost 60% of patients had AKI. Early measurement of biomarkers in the ED did not improve their AKI diagnostic performance. Biomarker ability to predict the composite outcome of mortality or dialysis was also assessed. Of the biomarkers measured, urinary NGAL independently predicted mortality or dialysis, and modestly improved the risk prediction model. The temporal profile of plasma creatinine following insult was investigated in 77 patients, in chapter 5. Plasma creatinine increases were sufficient to diagnose AKI in almost 60% of patients, most on ED arrival within less than two hours of the event. Loss of filtration function alone could not account for this early increase in plasma creatinine. Other factors such as increased creatinine production, or redistribution need to be considered. Creatinine measurement during the early period following insult should not be utilised as a baseline as this is likely to underdiagnose AKI. In 40% of AKI patients, plasma creatinine decreased below the AKI threshold within 48 hours. Even with transient loss of filtration function, structural biomarkers; plasma and urinary NGAL, urinary CysC, and GGT were increased. Volume resuscitation is a fundamental component of the acute management of critically ill patients. However, there has been growing concern about adverse effects of positive fluid balance. Positive fluid balance may also increase volume of distribution of creatinine and dilute plasma creatinine. This was investigated in 725 ICU admissions, in chapter 6. A 48-hour positive fluid balance greater than 5% body weight was independently associated with increased mortality. Fluid dilution masked AKI diagnosis in 8.2% of patients, and underestimated AKI severity in 14.4% of AKI patients. Adjusting plasma creatinine according to fluid balance would result in better risk classification, and potentially earlier intervention. A urine output criterion has been incorporated into AKI consensus definitions, but has not been well validated. In chapter 7, the utility of urine output in prediction of clinically meaningful outcome, mortality or dialysis was investigated in 725 ICU admissions. A threshold of average urine output over a moving 6-hour period of 0.3 ml/kg/h maximised the association of urine output with mortality or dialysis. This threshold also had the same sensitivity as the current plasma creatinine threshold for this outcome. This suggests that the current AKI urine output definition may be too liberally defined. The optimal threshold of 1 to 12-hour urine output was linearly dependent on duration of urine output assessment, and may be expressed as a formula; [(0.03 x duration of assessment) + 0.11] ml/kg/h. This threshold line best detected AKI for duration of assessment from 3 to 9 hours Finally, the original aims and the results are reviewed and suggestions provided for future research in Chapter 8.