Novel Biomarkers of Acute Kidney Injury in the General Adult ICU: A Review

Abstract

Acute kidney injury is one of the most frequent problems occurring in the critically ill patients of the intensive care units and it is well established that it increases both morbidity and mortality in these patients. Moreover, despite technological and pharmaceutical advances during the last decades, the incidence as well as the mortality associated with acute kidney injury in these patients remains unchanged. Creatinine, the most common renal dysfunction biomarker in use, has many disadvantages, such as time delay in its increase and the influence by other factors on its serum concentration, such as age, gender, muscle mass, etc. Hence, the need for better renal biomarkers in order to timely intervene for acute kidney injury prevention is imperative. The lack of an early biomarker is an obstacle for the development of new acute kidney injury prevention strategies. With the incidence of acute kidney injury reaching epidemic dimensions, the need for novel markers is urgent. During the last years, the research for finding such biomarkers has been intense. The purpose of the present article is to review the studies which have tested the predictive ability of those markers (in urine and/or plasma) for early detection of acute kidney injury in the mixed adult intensive care unit population and underline the potential limitations encountered in the various studies.

• acute kidney injury
• NGAL
• cystatin C
• biomarkers
• ICU
• acute renal failure

INTRODUCTION

Although serum creatinine (sCr) is currently used for acute kidney injury (AKI) diagnosis, it is a insensitive marker during acute alterations of renal function.¹ The lack of an early biomarker is an obstacle for the development of new preventive strategies and timely interventions against AKI. Recent research for the discovery of novel biomarkers with early diagnostic and/or prognostic value has revealed several candidates including neutrophil gelatinase-associated lipocalin (NGAL), cystatin C (CysC), interleukin 18 (IL-18), etc., with NGAL being the most promising so far. The purpose of the present article is (a) to review the studies that have tested the prognostic ability of all novel renal biomarkers for early AKI development in the heterogeneous population of adult critically ill patients and (b) to underline the potential limitations encountered in the various studies.

ACUTE KIDNEY INJURY IN THE ICU

Until recently there was no clear definition of acute renal failure (ARF). A recent study found that there were at least 35 definitions in literature,² leading to a wide variation in the reported incidence and outcomes of ARF. The need for a consensus definition for AKI was evident. Therefore, the ADQI workgroup developed RIFLE classification,³ while for further refinement, the AKI Network (AKIN) proposed a modified version of the RIFLE classification, known as the AKIN criteria.⁴

There are several epidemiological studies so far which have studied the incidence of AKI in the critically ill patients. A study by Ostermann and Chang⁵ analyzed 41,972 patients admitted to 22 intensive care units in the United Kingdom and Germany between 1989 and 1999. AKI defined by RIFLE occurred in 15,019 (35.8%) patients. Bagshaw and colleagues⁶ studied 120,123 patients admitted to 57 intensive care units across Australia for at least 24 h. AKI occurred in 36.1% of patients. The most common contributing factor to ARF was septic shock (47.5%).⁷ Various studies have reported on the increased mortality of ICU patients with AKI. Ympa et al.,⁸ in a systematic review of the literature from 1970 to 2004, observed an unchanged mortality of around 50% from 80 studies. Uchino et al.⁷ reported a hospital mortality rate of 60.3%. On the other hand, AKI development is associated with a longer ICU and hospital stay.⁹

Limitations of sCr and the Need for Novel AKI Biomarkers

To date, sCr has typically been used for AKI diagnosis. However, there are several limitations to its use as an AKI biomarker. First, its release varies with age, gender, diet, muscle mass, drugs, and vigorous exercise. Second, tubular secretion accounts for 10–40% of sCr clearance, which could mask a decrease in glomerular filtration rate (GFR). Third, sCr becomes abnormal when more than 50% of GFR has been lost. Therefore, there is a delay between injury and the subsequent rise in sCr and it might require up to 24 h before sufficient increase becomes detectable.¹⁰ It is assumed that the biological damage in cellular or molecular level precedes the clinical spectrum of AKI. For instance, the injured tubular cells secrete various molecules many hours before the functional decline which is evident by sCr increase. However, so far the lack of an early AKI biomarker has hampered the development of preventive strategies against AKI.

NOVEL RENAL BIOMARKERS IN THE ADULT ICU

Neutrophil Gelatinase-Associated Lipocalin

Human NGAL was originally identified as a 25-kD protein covalently bound to gelatinase (matrix metalloproteinase-9) from human neutrophils, where it represents one of the neutrophil secondary granule proteins (Table 1).^11,12 NGAL concentrations are elevated in the serum of patients with acute bacterial infections,¹³ which is consistent with NGAL’s proposed function as an endogenous bacteriostatic protein that scavenges bacterial siderophores (small molecules which bind iron). By binding iron, NGAL could mediate various physiologic functions, such as bacteriostatic and antioxidant effect. Alternatively, it could act as a growth factor, regulating apoptosis or cell differentiation. Actually, there are data supporting a role for NGAL as a regulator of the epithelial phenotype, inducing the formation of kidney epithelia in embryo and adults.¹⁴ Supavekin et al.¹⁵ identified NGAL as one of the most upregulated genes in the early post-ischemic mouse kidney, by using cDNA microarrays technique. Mishra et al.¹⁶ used a transcriptome-wide interrogation strategy to identify renal genes that are induced early after renal ischemia and found that NGAL induction represents a novel intrinsic response of the kidney proximal tubule cells to ischemic injury. Current data suggest that induction of urine NGAL (uNGAL) under harmful conditions is a compensatory response to ameliorate oxidative stress-mediated toxicity.¹⁷ Therefore, it is speculated that synthesis of NGAL protein in the distal nephron and secretion into the urine appears to promote cell survival and proliferation. On the other hand, plasma NGAL (pNGAL) is considered to originate from distant organs (AKI results in a dramatically increased NGAL mRNA expression in liver and lungs).¹⁸ Other sources could be neutrophils, macrophages, and other immune cells (as an acute phase reactant), whereas any decrease in GFR could result in its accumulation in systemic circulation.¹⁹ The aforementioned findings have initiated a number of translational studies to evaluate NGAL as a novel biomarker in human AKI. Haase et al.²⁰ performed a meta-analysis across all settings. In a subgroup analysis in critically ill patients, they found that the diagnostic odds ratio (DOR) and AUC of NGAL were 10 and 0.73, respectively.

Table 1.  Performance of NGAL in the mixed adult ICU. Studies with AUC-ROC analysis.

Author (ref.) year	Sample size	Baseline sCr estimation if unknown	AKI definition	% AKI on entry	% Total AKI	AKI timing	Sampling/assay(s)	%AUC (95% CI) for AKI	%AUC (95% CI) for RRT	%AUC (95% CI) for death
Endre^21	529	Lowest of the admission or final ICU sCr	AKIN	28	43	48 h	Urine^i/ELISA kit	66 (60–72)^u,d	78 (61–95)	66 (57–74)
2011								55 (48–63)^u,p
Doi^22	339	MDRD equation	RIFLE	19	39	7 days	Urine^a/ELISA kit	80 (71–86)^u,d	NA	83 (69–91)
								60 (52–67)^u,p
2011								70 (63–75)^u,t
DeGeus^23	632	Admission sCr	RIFLE	16	27	7 days	Serum and urine^a/Triage Biosite	77 (72–82)^s,t	88 (82–94)^s	63 (57–69)^s
2011								80 (76–84)^u,t	89 (85–93)^u	64 (58–70)^u
DeGeus^24	510	Admission sCr	RIFLE	0	13	NA	Serum and urine^a/Triage Biosite^s,u	76 (67–85)^s,$,p	NA	NA
2011								80 (72–88)^u,a,$,p
Siew^26	451	Multiple imputation methodology	AKIN	NA	19	48 h	Urine^a/ELISA kit	64 (57–71)^u,a,t	NA	NA
2009
Cruz^25	301	MDRD equation	RIFLE	30	44	48 h	Serum/Triage Biosite	78 (65–90)^s,t	82 (70–95)	67 (58–77)
2010
Constantin^27	88	MDRD equation	RIFLE	11	59	7 days	Serum/Triage Biosite	92 (85–97)^s,t	79 (69–87)	NA
2010
Kokkoris^30	100	Lowest of the admission or final ICU sCr	RIFLE	12	36	7 days	Serum and urine^a/Triage Biosite^s and ARCHITECT platform^u	78 (68–85)^s,t	NA	NA
2012								74 (64–82)^u,a,t
Royakkers^31	140^‡	Lowest of the admission sCr or by solving MDRD equation	RIFLE	37	59	NA	Serum and urine^a,i/ELISA kit	0.45 (27–63)^s,*	0.47 (37–58)^s	NA
								0.53 (38–67)^s,§,p
								0.48 (37–58)^u,a,*^,p	0.26 (3–50)^u,a
2012								0.48 (3–50)^u,a,§,p

Note: †: two-center study; ‡: five-center study; i: indexed to urine Cr concentration; a: absolute value; d: diagnosis of established AKI on admission; p: prediction of newly diagnosed AKI after ICU admission, in patients without AKI on entry; t: total performance (d + p); s: serum or plasma specimen; u: urine specimen; $: sustained AKI (>24 h duration); §: 1 day prior to AKI diagnosis; *x: 2 days prior to AKI diagnosis; NA: not available or not applicable.

NGAL in the general adult ICU

There are a number of studies, so far, that have examined prospectively the prognostic ability of NGAL for AKI development in the mixed adult ICU population. Endre et al.²¹ performed an observational prospective study in 529 adult critically ill patients, 28% of whom had AKI on admission according to AKIN criteria. This study compared six different urine biomarkers [γ-glutamyl transpeptidase (γGT), alkaline phosphatase (AP), CysC, NGAL, IL-18, and kidney injury molecule-1 (KIM-1)], regarding their ability to diagnose AKI on admission as well as to predict AKI within 48 h post-entry. The areas under receiving operator characteristic curves (AUCs) of uNGAL for diagnosis and prediction of AKI were 0.66 and 0.55, respectively, while its AUC for renal replacement therapy (RRT) prediction was 0.79. An important finding of this study was that the performance of all biomarkers was improved by stratification for time of collection with respect to renal insult and for baseline renal function before injury. In another large prospective study, Doi et al.²² included 339 ICU patients, of whom 19% had AKI on admission according to RIFLE criteria. They compared the ability of five novel urine biomarkers [liver type-fatty acid-binding protein (L-FABP), N-Acetyl-β-D-Glucosaminidase (NAG), IL-18, NGAL, and albumin (Alb)] for AKI prediction. The AUC of uNGAL for AKI prediction 7 days after admission was 0.70. De Geus et al.²³ conducted a large prospective study including 632 consecutive patients, of whom 16% had AKI on entry. The percentage of patients with more severe AKI (RIFLE F) was 9%. The ROC-AUCs for AKI prediction 1 week after admission were 0.77 for admission pNGAL and 0.80 for admission uNGAL, while the AUCs for RRT need were 0.88 for admission pNGAL and 0.89 for admission uNGAL. In another study which included 510 ICU patients, de Geus et al.²⁴ aimed to evaluate the ability of NGAL and CyC in plasma and urine to discriminate between sustained (duration >24 h), transient, and absent AKI. They found that uNGAL was the only biomarker significantly differentiating sustained from transient AKI on ICU admission. Individually, uNGAL performed better than the other biomarkers (AUC = 0.80) for the prediction of sustained AKI. Cruz et al.²⁵ performed another prospective observational study with 301 consecutive patients, of whom 30% had AKI on admission. The AUCs for admission pNGAL regarding AKI prediction within 48 h post-admission and RRT need during ICU stay were 0.78 and 0.82, respectively. Siew et al.,²⁶ in a prospective cohort study including 451 patients, found poor performance for admission uNGAL regarding AKI diagnosis within 48 h post-admission, with an AUC of 0.64. However, uNGAL remained independently associated with AKI development after adjustment for other confounding factors included in a clinical model, but it improved only marginally the predictive performance of the clinical model alone. Moreover, admission uNGAL independently predicted RRT initiation during ICU stay [HR: 2.60 (CI 95%: 1.55–4.35), when included in a multivariate Cox proportional hazards model]. Constantin et al.²⁷ reported excellent diagnostic ability for admission pNGAL regarding AKI development within the first week after admission (AUC = 0.92). The AUC for RRT need was 0.78. The study included 88 patients, 11% of whom had AKI on admission. Nejat et al.²⁸ analyzed data from 489 ICU patients and found that uNGAL levels were not significantly increased in pre-renal AKI compared with patients with no-AKI. In contrast, Doi et al.²⁹ found that uNGAL in patients with pre-renal AKI showed modest, but significantly higher concentrations than in patients with non-AKI. Recently, we undertook a prospective observational study, including 100 critically ill patients, in order to compare the ability of four renal biomarkers [pNGAL, uNGAL, plasma CysC (pCysC), and sCr] to predict early AKI. We also sought to determine whether their combinations were superior to each individual biomarker. The AUCs for AKI prediction within 7 days were 0.78 and 0.74, for pNGAL and uNGAL, respectively.³⁰ Finally, Royakkers et al.³¹ performed a secondary analysis in a multicenter prospective observational cohort study of unselected critically ill patients in five multidisciplinary ICUs, in which they collected serial serum and urine samples and determined the first day of AKI. Both sNGAL and uNGAL, 2 days and 1 day prior to AKI development, were poor predictors of AKI.

NGAL and sepsis

The impact of sepsis on NGAL levels in plasma and urine of critically ill adult patients is controversial. Martensson et al.³² conducted a study in 65 septic patients, none of whom had AKI on entry, and evaluated the predictive ability of pNGAL and uNGAL for the early detection of AKI. The AUCs for AKI within 12 h post-admission in patients with septic shock were 0.86 for peak uNGAL and 0.67 for peak pNGAL, respectively. They also found that uNGAL was a more robust marker of AKI than pNGAL in patients with septic shock, since uNGAL levels remain within normal limits even when pNGAL rises in patients without AKI. On the other hand, Bagshaw et al.³³ performed a prospective observational study in 83 patients with AKI and either sepsis or not, and sought to determine whether there were unique patterns to pNGAL and uNGAL in septic compared with non-septic AKI. However, they did not include a septic non-AKI control group for comparison. Septic AKI was associated with significantly higher admission pNGAL and uNGAL compared with non-septic AKI. The AUCs for septic AKI prediction within 48 h after admission were 0.77 and 0.70 for peak pNGAL and peak uNGAL, correspondingly. Furthermore, the AUCs for RRT need prognosis were 0.78 and 0.70, for peak pNGAL and peak uNGAL, correspondingly. On the contrary, Cruz et al.²⁵ found no significant difference in pNGAL among patients with and without sepsis or SIRS. Lastly, Kumpers et al.³⁴ conducted a prospective study in 109 critically ill patients with AKI at inception of RRT and identified pNGAL as a strong independent predictor for 28-day survival [HR: 1.6 (95% CI: 1.15–2.23)]. The extent to which AKI per se contributes to pNGAL levels could be confounded by the release of NGAL into the bloodstream in septic conditions; therefore, larger studies in septic patients are required to elucidate the role of NGAL in sepsis.

Cystatin C

CysC is a 13-kD endogenous cysteine proteinase inhibitor and plays a major role in intracellular catabolism of various peptides and proteins (Table 2).³⁵ CysC is produced by all nucleated cells in the human body at a relatively constant rate and released into plasma.³⁶ CysC is more than 99% filtered by the glomeruli, is neither secreted into the tubular lumen nor reabsorbed into the plasma. After filtration, it is completely reabsorbed by proximal renal tubular cells, through megalin receptor-induced endocytosis, and catabolized.³⁷ There is virtually no detection of CysC in the urine. Therefore, urine CysC (uCysC) may be an important urine biomarker of renal tubular injury.³⁸ On the other hand, pCysC is primarily a sensitive marker of reduction in GFR and appears to be a good biomarker in the prediction of AKI. Similar to sCr, pCysC may also be influenced by body composition, in particular, muscle mass and adipose tissue. Its levels have also been independently associated with smoking status and alcohol consumption. In addition, CysC appears to be influenced by abnormal thyroid function, systemic inflammation, and the use of corticosteroids.³⁹ In a recent systematic review, Zhang et al.⁴⁰ found that across all settings, the DOR for pCysC level to predict AKI was 23.5, with sensitivity and specificity of 0.84 and 0.82, respectively, while its AUC to predict AKI was 0.96. Subgroup analysis showed that pCysC was of diagnostic value when measured early (within 24 h after renal insult or ICU admission). The DOR of uCysC was 2.60, with sensitivity and specificity of 0.52 and 0.70, respectively, and its AUC to predict AKI was 0.64.

Table 2.  Performance of CysC in the mixed adult ICU. Studies with AUC-ROC analysis.

Author (ref.) year	Sample size	Baseline sCr estimation if unknown	AKI definition	% AKI on entry	% Total AKI	AKI timing	Sampling/assay(s)	%AUC (95% CI) for AKI	%AUC (95% CI) for RRT	%AUC (95% CI) for death
Herget-Rosenthal^41	85	NA	RIFLE	0	52	NA	Serum/Nephelometry	82 (71–92)^s,*^,p	69 (51–84)*	NA
2004								97 (94–99)^s,§,p	75 (62–85)^§
Ahlstrom^422004	202	MDRD equation	ARF according to ADQI 2nd International Consensus Conference 2003^‡	NA	26	24 h48 h72 h	Serum/Nephelometry	88 (NA)^s,1,t89 (NA)^s,2,t90 (NA)^s,3,t	NA	62 (NA)
Nejat^432010	444^†	Lowest of the admission or final ICU sCr	AKIN	28	45	7 days	Serum/Nephelometry	78 (73–83)^s,d65 (58–71)^s,p	84 (69–99)^p	62 (53–72)
Royakkers^452011	151^#	MDRD equation	RIFLE	37	60	NA	Serum and urine^a,i/Nephelometry^s,u	72 (NA)^s,*^,p49 (NA)^u,a,*^,p50 (NA)^u,i,*^,p	66 (NA)^s,b61 (NA)^u,a,b60 (NA)^u,i,b	NA
DeGeus^242011	510	Admission sCr	RIFLE	0	13	NA	Serum and urine^a/Triage Biosite^s,u	71 (61–80)^s,$,p61 (52–70)^u,$,p	NA	NA
Kokkoris^302012	100	Lowest of the admission or final ICU sCr	RIFLE	12	36	7 days	Serum/Nephelometry	75 (65–83)^s,t	NA	NA

Note: ‡: 3-fold sCr increase OR an acute increase in plasma creatinine of at least 0.5 mg/dL to more than 4 mg/dL OR diuresis less than 0.3 mL/kg/h for 24 h OR RRT need; 1: 24 h post-entry; 2: 48 h post-entry; 3: 72 h post-entry; §: 1 day prior to AKI diagnosis; *: 2 days prior to AKI diagnosis; b: on day of AKI diagnosis; i: indexed to urine Cr concentration; a: absolute value; d: diagnosis of established AKI on admission; p: prediction of newly diagnosed AKI after ICU admission, in patients without AKI on entry; s: serum or plasma specimen; u: urine specimen; $: sustained AKI (>24 h duration); †: two-center study; #: multi-center study; NA: not available or not applicable.

CysC in the general adult ICU

Herget-Rosenthal et al.⁴¹ analyzed 85 ICU patients at high risk of developing AKI and used the RIFLE criteria to define AKI (52% of patients). pCyC was shown to detect AKI 1–2 days earlier than sCr (AUC 0.82 and 0.97 on day 2 and day 1, respectively).⁷ Ahlstrom et al.⁴² studied 202 patients in a mixed ICU, of whom 49 developed AKI according to the RIFLE criteria. In that study pCyC showed an AUC of 0.89 for AKI prognosis 48 h post-ICU entry; however, pCyC did not rise earlier than sCr. Nejat et al.⁴³ studied 444 critically ill patients. Based on the criteria used by the AKIN definition, 28% of patients had AKI on entry, 16% developed AKI over the subsequent 7 days, and 55% did not. pCyC rose prior to sCr in 66% of the patients developing AKI after entry. pCyC on entry was predictive for sustained AKI (AUC 0.80), defined as an increase in sCr of at least 50% from baseline for 24 h or longer, but not for AKI within 7 days (AUC 0.65). In another study of 845 ICU patients, pCysC was independently associated with hospital death. This association was stronger for the patients suffering from AKI.⁴⁴ In a multicenter prospective study, Royakkers et al.⁴⁵ found moderate diagnostic ability for both plasma (0.72) and urine CysC (<0.50), 2 days prior to AKI diagnosis. Moreover, their prognostic ability for the need of RRT was poor. Our team reported an AUC of 0.75 for pCysC,³⁰ while Nejat et al.²⁸ found that uCysC levels were significantly increased in pre-renal AKI compared with patients with no-AKI.

CysC and sepsis

Whether sepsis presence has had any significant effect on CysC levels remains yet unclear. Nejat et al.,⁴⁶ in a subanalysis of sepsis patients showed that only uCysC was predictive of AKI within 48 h (AUC 0.71) and not pCysC, while uCysC was not predictive of AKI in patients without sepsis (AUC = 0.45). Martensson et al.⁴⁷ conducted a prospective study with 327 ICU patients and found that sepsis presence had no effect on either pCysC levels during the first week of ICU stay or its AUC for the prediction of the composite end point (sustained or worsening AKI or death).

Interleukin-18

IL-18 is a pro-inflammatory cytokine that is constitutively expressed in the intercalated cells of the distal convoluted tubules, the connecting tubules, and the collecting ducts of the healthy human kidney (Table 3). Moreover, these cells contain three major components required for the release of the active pro-inflammatory cytokine IL-18, namely pro-IL-18, P2X7, and the intracellular cysteine protease caspase-1,⁴⁸ which converts the Proform of IL-18 to its active form. During ischemic ARF, caspase-1 mediates conversion of pro-IL-18 to active IL-18. The active IL-18 is released from the tubular cell and mediates neutrophil infiltration during ischemic ARF. Administration of neutralizing anti-IL-18 Ab’s offers protection against ischemic ARF. IL-18 plays a deleterious role in experimental ischemic ARF, perhaps in part due to increasing neutrophil infiltration into the renal parenchyma.⁴⁹

Table 3.  Performance of IL-18 in the mixed adult ICU. Studies with AUC-ROC analysis.

Author (ref.) year	Sample size	Baseline sCr estimation if unknown	AKI definition	% AKI on entry	% Total AKI	AKI timing	Sampling/assay(s)	%AUC (95% CI) for AKI	%AUC (95% CI) for RRT	%AUC (95% CI) for death
Parikh^50	138^#	NA	>50% rise in sCr	0	37	NA	Urine^a/ELISA kit	73 (NA)^§^,p	NA	NA
2005								65 (NA)*^,p
Siew^51	451	NA	AKIN	NA	19	48 h	Urine^i/ELISA kit	62 (54–69)^1,t	NA	NA
2010								60 (53–67)^2,t
Endre^212011	529^†	Lowest of the admission or final ICU sCr	AKIN	28	43	48 h	Urine^i/ELISA kit	62 (56–67)^d55 (47–62)^p	73 (59–86)	68 (60–76)
Doi^222011	339	MDRD equation	RIFLE	19	39	7 days	Urine^a/ELISA kit	78 (70–84)^d59 (51–67)^p68 (62–74)^t	NA	82 (68–91)

Note: §: 1 day prior to AKI diagnosis; *: 2 days prior to AKI diagnosis; 1: AKI within 24 h post-admission; 2: AKI within 48 h post-admission; i: indexed to urine Cr concentration; a: absolute value; d: diagnosis of established AKI on admission; p: prediction of newly diagnosed AKI after ICU admission, in patients without AKI on entry; t: total performance (d + p); #: cross-sectional study; †: two-center study; NA: not available or not applicable.

IL-18 in the general adult ICU

Parikh et al.⁵⁰ performed a nested case-control study within the Acute Respiratory Distress Syndrome (ARDS) Network trial, including 52 AKI patients and 86 controls. The data were analyzed in a cross-sectional manner and according to the time before development of AKI. The median urine IL-18 levels were significantly different at 24 and 48 h before AKI in AKI patients as compared with control patients. In multivariable analysis, after adjustment for other confounding factors, urine IL-18 (uIL-18) levels of >100 pg/mL were associated with increased odds of AKI of 6.5 (95% CI: 2.1–20.4) in the next 24 h. In diagnostic performance testing, uIL-18 demonstrated an AUC of 73% to predict AKI in the next 24 h and of 65% for AKI prediction after 48 h. In multivariable analysis, uIL-18 value on day 0 was an independent predictor of mortality. Furthermore, the presence of sepsis had no effect on uIL-18 concentration in both case and control patients. Siew et al.⁵¹ analyzed data from 451 patients, 86 of whom developed AKI within 48 h of enrolment. The AUCs for the utility of uIL-18 for AKI prediction within 24 and 48 h were 0.62 (95% CI: 0.54–0.69) and 0.60 (95% CI: 0.53–0.67), respectively. After adjustment for a priori selected clinical predictors, uIL-18 remained independently predictive of the composite outcome of death or RRT within 28 days (OR 1.86, 95% CI: 1.31–2.64). An examination of the interaction between uIL-18 and sepsis also suggested improved performance in non-septic patients for AKI prediction; however, this interaction did not reach statistical significance. Endre et al.²¹ reported that the AUCs of uIL-18 for AKI diagnosis on entry and prediction within 48 h post-admission (in patients without AKI on entry) were 0.62 and 0.55, respectively. Moreover, the AUCs of uIL-18 for prediction of dialysis and death were 0.73 and 0.68, correspondingly. According to Doi et al.²² the AUCs of uIL-18 for AKI prediction within 1 week after admission and prediction of death 14 days post-entry were 0.69 and 0.83, respectively. In another study by Doi et al.,²⁹ uIL-18 in patients with pre-renal AKI showed modest but significantly higher concentrations than in patients with non-AKI.

Liver Type-Fatty Acid-Binding Protein

FABPs are known as intracellular lipid chaperones that transport free fatty acids (FFAs) to a specific component in the cell; however, little is known about their exact biological functions and mechanisms of action (Table 4).⁵² There are several different types of FABP (nine types have been discovered so far), and their tissue distribution is rather ubiquitous.⁵³ FFAs are easily oxidized, leading to oxidative stress that can induce cellular injury. L-FABP may be an important cellular antioxidant during oxidative stress, by maintaining low levels of FFAs in the cytoplasm of tubular cells through facilitation of intracellular metabolism and excretion in urine.⁵⁴ Urine L-FABP (uL-FABP) showed ability for early and accurate detection of histological and functional deterioration in both nephrotoxin- and ischemia-induced AKI in mice.⁵⁵

Table 4.  Performance of L-FABP, KIM-1, and NAG in the mixed adult ICU. Studies with AUC-ROC analysis.

Author (ref.) year	Sample size	Baseline sCr estimation if unknown	AKI definition	% AKI on entry	% Total AKI	AKI timing	Sampling/assay(s)	%AUC (95%CI) for AKI	%AUC (95%CI) for RRT	%AUC (95%CI) for death
L-FABP
Doi^222011	339	MDRD equation	RIFLE	19	39	7 days	Urine^a/ELISA kit	80 (73–86)^d70 (62–76)^p75 (69–80)^t	NA	90 (83–94)
Matsui^57	25	NA	AKIN	NA	56	ICU stay	Urine^i/ELISA kit	95 (NA)	NA	NA
2011
KIM-1
Endre^212011	529^†	Lowest of the admission or final ICU sCr	AKIN	28	43	48 h	Urine^i/Luminex xMAP technology	66 (61–72)^d55 (47–62)^p	62 (48–76)	56 (47–64)
NAG
Doi^222011	339	MDRD equation	RIFLE	19	39	7 days	Urine^a/4-HP-NAG substrate method	70 (62–76)^d55 (47–62)^p62 (56–68)^t	NA	66 (50–80)
Matsui^57	25	NA	AKIN	NA	56	ICU stay	Urine^i/NA	63 (NA)	NA	NA
2011
Westhuyzen^632003	26	NA	>50% rise in sCr and to a concentration >0.15 mmol/L	0	15	7 days	Urine^i/Colorimetry	84 (64–95)^p	NA	NA

Note: i: indexed to urine Cr concentration; a: absolute value; d: diagnosis of established AKI on admission; p: prediction of newly diagnosed AKI after ICU admission, in patients without AKI on entry; t: total performance (d + p); †: two-center study; NA: not available or not applicable.

L-FABP in the general adult ICU

Nakamura et al.⁵⁶ studied 40 patients with septic shock, 20 patients with severe sepsis without shock, 20 ARF patients without septic shock, and 30 healthy volunteers. Polymyxin B immobilized fiber (PMX-F) hemoperfusion was performed twice in the septic shock patients. Their results suggested that uL-FABP levels might be able to reflect the severity of sepsis and also to monitor the effectiveness of treatment. However, this was not a randomized controlled trial. Matsui et al.⁵⁷ enrolled 25 critically ill patients, of whom 14 developed AKI during their ICU stay. In the AKI group, elevation of uL-FABP level occurred between 30 and 0 h before the occurrence of AKI (with an AUC of 0.95). In another prospective clinical study,⁵⁸ 145 septic shock patients who were diagnosed as AKI at the time of ICU admission (established AKI) were analyzed. uL-FABP levels were significantly higher in non-survivors than in survivors and a multiple logistic regression analysis revealed that uL-FABP was significantly associated with mortality. ROC curve analysis revealed an AUC of 0.993 for death prediction. Likewise, Doi et al.²² found an AUC of 0.90 for 14-day mortality for uL-FABP measured on ICU admission. The AUC of uL-FABP for AKI prediction 1 week post-ICU entry was 0.75. Another study found that uL-FABP in patients with pre-renal AKI showed modest, but significantly higher concentrations than in patients with non-AKI.²⁹

Kidney Injury Molecule-1

KIM-1 is a transmembrane protein that is overexpressed in dedifferentiated proximal tubule cells after ischemic or nephrotoxic AKI in animal models,⁵⁹ and a proteolytically processed domain is easily detected in the urine (Table 4).⁶⁰ The presence of KIM-1 has been demonstrated in renal biopsy specimens of patients with proven acute tubular necrosis. Similarly, urine KIM-1 (uKIM-1) was significantly elevated in patients with a clinical diagnosis of acute tubular necrosis compared with patients with normal renal function, chronic renal disease, or ARF of other etiologies including contrast nephropathy.⁶¹

KIM-1 in the general adult ICU

In a large prospective study of 529 patients, Endre et al.²¹ found that the AUCs of uKIM-1 for AKI prediction within 48 h post-admission, dialysis, and death prediction in 7 days, were 0.55, 0.62, and 0.56, correspondingly. Analyzing data from the same population, Nejat et al.²⁸ found that uKIM-1 levels were significantly increased in pre-renal AKI compared with patients with no-AKI.

N-Acetyl-β-D-Glucosaminidase

NAG is a lysosomal enzyme, found predominantly in proximal tubules so that increased activity of this enzyme in the urine suggests injury to tubular cells, and therefore can serve as a specific urinary marker for the tubular cells (Table 4). Due to its relatively high molecular weight, filtration of the enzyme is precluded by glomeruli. In the course of active kidney disease, urine NAG (uNAG) levels remain persistently elevated. The increase in uNAG activity indicates damage to tubular cells, although it can also reflect increased lysosomal activity without cellular damage.⁶²

NAG in the general adult ICU

Westhuyzen et al.⁶³ in a small prospective pilot study of 26 consecutive critically ill adult patients found that uNAG had an AUC of 0.84 for ARF prediction, while another small study by Matsui et al.⁵⁷ reported an AUC of 0.63. Doi et al.²² in a large study of 339 critically ill patients found that uNAG was a moderate predictor of both AKI within 1 week after admission (AUC = 0.62) and 14-day mortality (AUC = 0.66). Another study found that uNAG in patients with pre-renal AKI showed modest, but significantly higher concentrations than in patients with non-AKI.²⁹

Other Kidney Biomarkers

Some urine proteins (particularly enzymes) released from damaged tubular cells have been used as AKI biomarkers (Table 5). Increased urine excretion of these proteins implies cellular necrosis. The majority of the enzymes in tubular urine are brush border enzymes such as AP and γGT.^64,65 Increased excretion of these proteins implies injury to the brush border membrane with loss of microvillus structure. In a small prospective pilot study of 26 consecutive critically ill adult patients, Westhuyzen et al.⁶³ reported the following AUCs for AKI prediction: 0.89, 0.93, 0.86, and 0.95, for urine α-glutathione S transferase (α-GST), π-GST, AP, and γ-GT, respectively. In a large prospective study of 529 patients, Endre et al.²¹ found that the AUCs of urine AP and γ-GT for AKI prediction within 48 h post-admission were 0.56 and 0.57, respectively. Nejat et al.²⁸ found that γ-GT levels were not significantly increased in pre-renal AKI compared with patients with no-AKI. Lastly, urine Alb (uAlb) has also been investigated as an AKI biomarker. Matsui et al.⁵⁷ found an AUC of 0.70 for uAlb regarding AKI prediction in a small study of 25 critically ill patients, while Doi et al.²² in a large study of 339 critically ill patients reported a similar AUC for uAlb (0.69). Another study found that uAlb in patients with pre-renal AKI showed modest, but significantly higher concentrations than in patients with non-AKI.²⁹

Table 5.  Performance of γGT, AP, and Alb in the mixed adult ICU. Studies with AUC-ROC analysis.

Author (ref.) year	Sample size	Baseline sCr estimation if unknown	AKI definition	% AKI on entry	% Total AKI	AKI timing	Sampling/assay(s)	%AUC (95%CI) for AKI	%AUC (95%CI) for RRT	%AUC (95%CI) for death
γGT
Westhuyzen^632003	26	NA	>50% rise in sCr and to a concentration >0.15 mmol/L	0	15	7 days	Urine^i/Standard method	95 (79–99)^p	NA	NA
Endre^212011	529^†	Lowest of the admission or final ICU sCr	AKIN	28	43	48 h	Urine^i/G-glutamyl-p-nitroanilide rate	59 (54–65)^d57 (50–64)^p	60 (46–74)	65 (56–73)
AP
Westhuyzen^632003	26	NA	>50% rise in sCr and to a concentration >0.15 mmol/L	0	15	7 days	Urine^i/Standard method	86 (67–97)^p	NA	NA
Endre^212011	529^+	Lowest of the admission or final ICU sCr	AKIN	28	43	48 h	Urine^i/P-nitrophenol rate reactions	45 (40–50)^d56 (49–63)^p	63 (49–77)	61 (53–70)
Alb
Matsui^57	25	NA	AKIN	NA	56	ICU stay	Urine^i/NA	70 (NA)	NA	NA
Doi^222011	339	MDRD equation	RIFLE	19	39	7 days	Urine^a/Immunonephelometry	78 (72–84)^d59 (51–67)^p69 (63–74)^t	NA	72 (61–81)

Note: i: indexed to urine Cr concentration; a: absolute value; d: diagnosis of established AKI on admission; p: prediction of newly diagnosed AKI after ICU admission, in patients without AKI on entry; t: total performance (d + p); †: two-center study; NA: not available or not applicable.

LIMITATIONS AND METHODOLOGICAL ISSUES OF THE NOVEL BIOMARKERS STUDIES

The first clinical studies of the novel renal biomarkers had been conducted in relatively homogeneous populations without co-morbidities, such as pediatric and adult cardiac surgery patients, where the timing of renal insult(s) is predictable. The results of those studies were very promising for most of the biomarkers. However, the initial enthusiasm dwindled when they were investigated in the setting of the mixed adult ICU. Adult critically ill patients consist of a heterogeneous population with various disease backgrounds, great variability in disease severity, much co-morbidity, and several interventions. In such patients, AKI could be attributed to multiple insults, each one acting at different time points.

The studies performed in critically ill patients so far have several methodological differences (as shown in detail in Tables 1–5) that should be taken into consideration, such as differences in: (i) design: most studies have used the admission biomarker values, whereas few of them have performed serial measurements and used biomarker values 2 days and 1 day prior to AKI development in order to estimate their predictive ability; (ii) sample size: important issues which compromise the generalizability of their results are the small number of included patients and the participation of only one center in the majority of them; (iii) time period after admission for AKI development, ranging from 1 to 7 days; (iv) AKI definition: some studies have used the RIFLE criteria, some others the AKIN criteria, and few of them other criteria. Moreover, some studies have used both components, sCr and urine output, for AKI definition, while others have used only the sCr component; (v) the percentage of patients who had established AKI on ICU admission, ranging from 0% to 37%; and (vi) baseline sCr definition: the definition of baseline sCr value is crucial because it will determine whether a patient has AKI or not. Among studies there is great heterogeneity about this definition. In most of them, whenever a sCr value prior to ICU admission was not available, a value based on the MDRD-eGFR equation solution (taking arbitrarily a value of 75 mL/kg/m² for eGFR) was used. This resulted in a high percentage of patients having already developed AKI at admission (Tables 1–5).

Another important issue is that each study has used different primary outcome measures, thus making direct comparisons difficult. Some of them have used diagnosis of established AKI on admission, others have used prognosis of evolving AKI during the first days of their ICU stay (ranging from 1 to 7 days post-admission), while others have reported both measures (Tables 1–5). Moreover, one of them has used sustained AKI (lasting more than 24 h) as primary outcome.²⁴ Interestingly, the predictive ability of all novel biomarkers for established AKI diagnosis on ICU admission was consistently better compared to their ability for prediction of evolving AKI during ICU stay (Tables 1–5), a finding which is not very helpful in the adult ICU setting. Furthermore, another study used pre-renal AKI as primary outcome measure and found that the median concentration of uKIM-1, uCys, and uIL-18 were significantly greater in pre-renal AKI compared with no-AKI, while uNGAL and uγ-GT concentrations were not significant.²⁸ Additionally, in a recent study, urinary L-FABP, NGAL, IL-18, NAG, and Alb in patients with pre-renal AKI showed modest, but significantly higher concentrations than in patients with no-AKI.²⁹

Another methodological issue is the statistical methods used for the assessment of the diagnostic and/or predictive performance of each biomarker. To date, the vast majority of the studies have been based on AUC-ROC analyses only. However, there are modern methods considered to be complementary to AUC-ROC analysis, such as net reclassification index (NRI) and integrated discrimination improvement (IDI),^66–68 which should also be used. Moreover, few of them use proper statistical tests to compare the AUCs of different biomarkers (i.e., DeLong test⁶⁹) whereas the selection of the optimal cut-off point for sensitivity and specificity estimation is not based on appropriate statistical methods (i.e., Youden’s index⁷⁰) in most studies.

Most studies so far have focused on the performance of only one of these biomarkers to detect AKI in critically ill patients, before a sCr rise. However, none of them has determined the predictive ability of their combinations in the adult ICU setting, where the etiology of AKI is multifactorial and not well understood. Because many insults are involved, it is speculated that a single biomarker will be insufficiently sensitive and specific across the full spectrum of AKI and combinations (panels) of biomarkers with different characteristics may prove more accurate. The utility of biomarker combinations in AKI prediction has been assessed by a number of studies in cardiac surgery patients, which reported improved predictive performance for the various combinations they used.^71–75 Nevertheless, which are the optimal combinations of biomarkers, remain unclear. We have recently reported that the combinations of novel biomarkers with a conventional one (pNGAL + uNGAL + sCr) are promising in the heterogeneous population of a general adult ICU.³⁰ Some authors suggest that the ideal combination should include a biomarker with a high sensitivity and another one with a high specificity.⁷⁴ We selected our combinations after statistically comparing the AUCs of all possible combinations.³⁰ Larger studies that will identify the most efficient and cost-effective biomarker combinations and will develop and validate AKI scoring systems are required.

There is inconsistency in terms of urine biomarkers’ reported values. Throughout the literature, authors use either the absolute value of a urine biomarker concentration or its ratio to uCr concentration in a spot urine sample. To date, there are no strong arguments in favor of one of them. Therefore, future studies should report their results based on both absolute and indexed to urine creatinine values of the urine biomarkers, until a better method is discovered.

Given the great variability in the temporal pattern of the various biomarkers concentrations, another issue that should be addressed in future studies is whether serial measurements, rather than a static time-point assessment of biomarkers, might increase the likelihood of capturing the point of maximum discrimination, especially in a mixed ICU population, where neither the timing nor the severity or the multitude of renal insults are known.

Finally, biomarker studies for the diagnosis of AKI may have yielded different results had there been a true “gold standard” for AKI instead of using a change in sCr. Even a hypothetical perfect biomarker (with 100% sensitivity and specificity) could have a much lower measured sensitivity and specificity when benchmarked against an “imperfect” gold standard with low sensitivity and specificity, such as sCr.

CONCLUSIONS

Except for NGAL, all other aforementioned novel kidney biomarkers have not been extensively investigated in adult critically ill patients. On the other hand, the so far studies have reported fair to moderate results for AKI prediction in the mixed ICU setting. Investigation of various biomarker combinations in order to find the ideal panel might be prudent in order to avoid a stalemate. Moreover, future studies should ideally be multi-center and use modern statistical methods (NRI and IDI), as well as cross-validation techniques. Additionally, they should investigate the predictive ability of the biomarkers for a priory defined specific outcome measures, such us septic AKI, drug-related AKI, ischemia-reperfusion-related AKI, pre-renal azotemia, intrinsic AKI, acute-on-chronic AKI, etc. Moreover, it is strongly suggested that biomarkers be used in large randomized studies in order to validate the role of each one of them (or a panel of them) in the clinical decision-making process about earlier implementation of more aggressive interventions, i.e., continuous RRT when those markers are elevated, before irreversible renal damage has occurred.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.