Plasma and urine neutrophil gelatinase-associated lipocalin in the diagnosis of new onset acute kidney injury in critically ill patients

Acute kidney injury (AKI) is a complex and heterogeneous process and is associated with high morbidity and mortality especially among critically ill patients [1]. The main pathophysiological process is renal tubular ischaemia, which is often multifactorial including sepsis, major surgery, low cardiac output, hypovolaemia and medication toxicity. Such insults frequently occur throughout a patient’s stay on the intensive care unit (ICU) [1, 2].

Consensus diagnosis of AKI depends upon the detection of oliguria and/or a rise of serum creatinine level [3]. However, changes in creatinine often lag behind reductions in the glomerular filtration rate (GFR) questioning its utility as a biomarker to pre-empt deterioration in renal function [4]. The Acute Dialysis Quality Initiative (ADQI) consensus has recommended RIFLE criteria to stratify three severity stages and two outcome stages for AKI - risk, injury, failure, loss and end-stage renal failure; relying on creatinine and urine output for their definition [3]. Despite advances in understanding the mechanisms of AKI the morbidity and mortality, especially in ICU patients, has not changed for some years. One key factor identified is a delay in diagnosis [5, 6] suggesting a need for more accurate and timely biomarkers. This would allow earlier identification of patients who may benefit from enhanced resuscitative care and pre-emptive renal support to prevent further renal injury.

Neutrophil gelatinase-associated lipocalin (NGAL) has been shown in multiple laboratory experiments to be secreted early in the development of AKI. This has been confirmed in several clinical settings including cardiac surgery, following contrast administration, in the emergency department and in paediatric intensive care [7‐9].

To the best of our knowledge, only a few studies have examined the role of NGAL to predict AKI occurrence in a general adult ICU [10‐15]. A recent systematic review showed variability in the predictive ability of NGAL [16] related to inconsistencies in study design, patient inclusion and method of measurement. Most of these studies have included patients on admission who have pre-existing renal disease, (acute or chronic) leading to uncertainty about the timing and type of renal insult. It is also unclear whether urinary NGAL (uNGAL) or plasma NGAL (pNGAL) provide the best predictive ability. This study was designed to examine the predictive ability of both pNGAL and uNGAL in a heterogeneous adult ICU population with no known pre-existing renal disease prior to admission.

Neutrophil gelatinase-associated lipocalin has been used for the diagnosis, prognosis and severity assessment of AKI and has been validated in paediatric populations (where comorbidity is low) and in conditions where the timing of the insult is clear (that is cardiac surgery and so on) [20‐24]. Its role in an adult ICU population has not been well validated due to the heterogeneity and uncertainty of the timing of the insult. The pathophysiology and causes of AKI in the ICU could be indigenous [25] and may differ from pre-ICU causes (low-volume state, inotropes, contrast injury and so on). This has led to problems in the design and interpretation of studies in the general adult ICU patient cohort. In this study, we sought to overcome these problems by excluding patients with pre-existing chronic kidney disease (CKD) and/or AKI and by looking at the predictive value of both urinary and plasma NGAL at different time points following admission.

We found the incidence of AKI was 30.4% (n = 59) with 71% (n = 42) developing it within the first 24 hours. This incidence correlated well with that anticipated from our sample size calculation but was lower than that found in previous studies due to the exclusion of those patients with pre-existing renal disease [11, 12, 26]. Both admission pNGAL and uNGAL were found to fairly predict the occurrence of AKI at 72 hours following ICU admission (AUROC 0.766 and 0.791 respectively). Our study also showed that both pNGAL and uNGAL concentrations, at the time of ICU admission and at any given point of time, were significantly higher in AKI compared to that of the non-AKI patients (Figures 2 and 3). de Geus et al. [10] found that the performance of pNGAL at admission to predict the AKI severity to be similar to our study [10]. However, utilising admission NGAL values to predict AKI up to seven days can be associated with certain confounding factors. Raised NGAL at admission may correlate tubular ischemia at that point of time. The possibility of a second or multiple insults after the initial insult has been reversed is common in an ICU setting. Therefore, prediction of AKI at a later point of time may not represent true index performance. We have overcome this issue by serial measurements of NGAL. Moreover the study by de Geus included patients with previous renal disease. This will affect the sensitivity and specificity of the test. Another study [11] showed that performance of pNGAL to predict AKI occurrence up to seven days was excellent (AUROC 0.92). The small sample (n = 88) may over-inflate the diagnostic accuracy. Cruz et al.[13] found that the predictive value of admission pNGAL to diagnose AKI up to 48 hours was fair (AUROC 0.78) but when extended to five days, pNGAL performance was poor (AUROC 0.67). This suggests that there is a wide variation of NGAL assay performance when prediction is extended beyond five days. Of note, a recent study [27] looking at admission pNGAL to predict AKI occurrence up to 72 hours found results not dissimilar to our study (AUROC 0.78), however, by excluding patients with pre-existing kidney disease we have determined the diagnostic accuracy more precisely.

Many of the preliminary studies performed to evaluate the diagnostic value of NGAL were done using urine analysis yielding promising results, however, these studies were done in a paediatric, non-ICU population, and where time of injury was clearly defined [9, 20‐23]. Siew et al.[15] found that the admission uNGAL values to predict AKI at 24 hours among ICU patients was fair (AUROC 0.71) but did not perform well when extended to 48 hours (AUROC 0.64). The study used a convenience sample from a large biomarker study in acute lung injury and also included patients with previous and existing kidney injury.

Another study performed by Makris et al.[28] showed that the uNGAL performance to predict AKI up to five days was excellent (AUROC 0.97). The study mostly included trauma patients. The generalisability of this study, however, is limited due to the small sample size (n = 31). Other ICU studies investigating uNGAL to predict AKI occurrence up to five days or more have shown variable results [10, 11, 26]. Under normal circumstances, the filtered NGAL is completely absorbed by the proximal convoluted tubule (PCT) and therefore there should be minimal or no NGAL in the urine [29]. If there is damage to the PCT, NGAL is not absorbed and accounts for the rise in NGAL levels in urine. Further evidence suggests that the NGAL production can arise from the injured distal nephron [8]. This could explain the persistent rise in uNGAL in patients with ongoing injury. When the GFR is reduced, serum creatinine increases. Similarly, systemic NGAL (pNGAL) levels increase as NGAL filtration is decreased.

Our study has looked at the occurrence of AKI up to 72 hours. By taking serial samples we were able to show that both uNGAL and pNGAL continue to have fair to good predictive ability (Table 1) in admitted patients suggesting there may be a role in the daily detection of renal injury. The 72-hour time frame is a reasonable time for the pNGAL to accumulate. Of note, the kinetics of pNGAL and uNGAL in our study suggest pNGAL gradually increases with time whereas the uNGAL reaches a plateau (Figures 2 and 3). However, this finding is clearly limited by the small number of samples at 48 hours.

A small study in a selected group of patients with septic shock [12] showed that uNGAL performed better (AUC 0.86) than the pNGAL (0.67) to predict the AKI occurrence 12 hours ahead of time. Therefore, pNGAL may not be a good predictor if the prediction time is short as it may take some time to accumulate and uNGAL may be a better marker in triage situations. Therefore, uNGAL could be a better predictor in conditions where AKI is anticipated that is radio-contrast-induced AKI, coronary artery bypass graft (CABG) and so on. However, one drawback is that sometimes in critically unwell patients, obtaining urine may be difficult and sometimes uNGAL may be elevated with urinary tract infection [14].

The threshold values we used were high compared to similar studies. The difference may be secondary to the type of assays and platform used. We chose an immunoturbidimetric assay due to the fast assay time (10 minutes), a platform compatible with most analysers available in UK laboratories [30, 31] and a wide measuring range (25 to 5000 ng/mL) that will allow easy validation in clinical practice.

Our study makes an important contribution to the current body of literature on NGAL and AKI. To the best of our knowledge, this is the first study to prospectively evaluate both plasma and urine NGAL in adult ICU patients where pre-existing renal disease has been excluded. Our study has certain important strengths. First, this is the first published UK study on NGAL to predict AKI in ICU patients; second, our strict exclusion criterion excluded pre-existing kidney disease; third, we powered our study to ensure statistical robustness; fourth, we demonstrated the value of serial sampling of pNGAL and uNGAL; finally, we used a standardised user-friendly platform, reducing laboratory measurement error and allowing other investigators to validate our findings.

Our study also has deficiencies. This is a single-centre study and hence care should be exercised when generalising the results. Although we excluded patients with AKI prior to admission, there may be a small number of patients who might have had an undetected renal insult in the few hours prior to ICU admission. This is an important limitation in any study of early biomarkers in an unselected ICU population. Second, we did not perform a sensitivity analysis to discount comorbidities that could account for the changes in NGAL. Third, there had been a decrease in the number of patients for whom the serial NGAL levels were measured with time (that is 0, 24 and 48 hours). In a general ICU setting this attrition is unavoidable. In some cases, the samples were not available from the patients due to anuresis and/or in the process of end-of-life care. Fourth, urine NGAL levels were not corrected for urinary creatinine concentration to account for possible dilution of urine. Finally, we were unable to correlate NGAL values with clinical outcomes such as mortality and need for RRT due to the small numbers in these groups.

Our study showed that both pNGAL and uNGAL levels measured at admission to ICU can predict an occurrence of AKI for up to 72 hours of ICU stay. The performance as determined by AUROC is fair. This performance was sustained when measured at serial time points throughout the ICU stay.