Urinary and serum biomarkers for the diagnosis of acute kidney injury: an in-depth review of the literature*

Abstract

Background

Acute kidney injury (AKI) remains associated with high morbidity and mortality, despite progress in medical care. Although the RIFLE (Risk, Injury, Failure, Loss, End-Stage Kidney Disease) and AKIN (Acute Kidney Injury Network) criteria, based on serum creatinine and urine output, were a step forward in diagnosing AKI, a reliable tool to differentiate between true parenchymal and pre-renal azotaemia in clinical practice is still lacking. In the last decade, many papers on the use of new urinary and serum biomarkers for the diagnosis and prognostication of AKI have been published. Thus, the question arises which biomarker is a reliable differential diagnostic tool under which circumstances.

Methods

We searched Medline from inception to April 2012 using medical subject heading and text words for AKI and biomarkers [neutrophil gelatinase-associated lipocalin (NGAL), kidney injury molecule-1 (KIM-1), Cystatin C, interleukin-6 (IL-6), interleukin-8 (IL-8), interleukin-18 (IL-18), N-acetyl-glucosaminidase (NAG), glutathione transferases (GST) and liver fatty acid binding protein (LFABP)] to identify relevant papers in five different settings (paediatrics, cardiac surgery, emergency department, critically ill and contrast-induced nephropathy).

Results

We included 87 relevant papers, reporting on 74 studies. Depending upon the setting, 7–27 different definitions of AKI were used. Reported diagnostic performance of the different biomarkers was variable from poor to excellent, and no consistent generalizable conclusions can be drawn on their diagnostic value.

Conclusions

Early diagnosing of AKI in clinical conditions by using new serum and urinary biomarkers remains cumbersome, especially in those settings where timing and aetiology of AKI are not well defined. Putting too much emphasis on markers that have not convincingly proven reliability might lead to incorrect interpretation of clinical trials. Further research in this field is warranted before biomarkers can be introduced in clinical practice.

acute kidney injury, biomarkers, dialysis, renal failure, urinary

Introduction

Acute kidney injury (AKI) remains associated with high morbidity and mortality, despite progress in medical care [1–3]. In the last decade, many papers on the use of new urinary and serum biomarkers for AKI were published, mostly concluding that these biomarkers will lead to a new era with earlier diagnosis, better prognostication of outcome in terms of need for renal replacement and/or mortality, and finally better survival [4]. Nevertheless, there remains a gap between the fascinating findings at the basic science level and the clinical application of this knowledge. Objective evaluation of the available literature shows a rather disappointing picture [5]. Before we throw out the baby with the bathwater, we should evaluate what we reasonably can expect from biomarkers, what they provide today and why their performance is currently suboptimal.

Materials and methods

To identify relevant studies, we searched MEDLINE (through OvidSP) from inception to April 2012, using medical subject headings and text words for AKI and biomarkers. The full search strategy is outlined in item S1, provided as Supplementary data. To locate studies not indexed in Medline, we checked by hand the bibliography of relevant publications. We included both cohort and case–control studies evaluating the potential of biomarkers to detect AKI early or to predict the need for renal replacement therapy (RRT) in both adults and children. We excluded studies published as abstracts only and restricted to those published in English. Papers were excluded as ‘invalid intervention’ when they did not analyse AKI, or where not performed in humans, and as ‘inappropriate design’ when they did not investigate the differential diagnosis AKI versus non-AKI. Mostly, the latter studies just reported means/medians in subgroups rather than discriminatory values on the individual patient level, or insufficient data were present to calculate these. We allowed for any definition of AKI. Biomarkers included in the search strategy were neutrophil gelatinase-associated lipocalin (NGAL), kidney injury molecule-1 (KIM-1), interleukin-6 (IL-6), interleukin-8 (IL-8), interleukin-18 (IL-18), Cystatin C, N-acetyl-glucosaminidase (NAG), liver fatty acid binding protein (LFABP) and the glutathione transferases (GST), measured either in plasma, serum or urine.

Two authors (J.V. and W.V.B.) independently screened all titles and abstracts and assessed all selected full papers for eligibility. Any disagreements were resolved by face-to-face discussion. Using a structured data extraction template, we collected details on biomarker(s) studied, outcome, AKI definition, number of patients/events and diagnostic test characteristics. Results are reported in Tables 2–6. When not available, positive predictive value (PPV) and negative predictive value (NPV) were calculated if enough information was provided and if the study design allowed this.

Because of the heterogeneity of studies in different clinical settings, a meta-analysis or pooling of the data was deemed to be methodologically inappropriate.

Results

One thousand five hundred and thirty-two potentially relevant papers were identified by the search strategy. Based on the title and abstract, 149 papers were selected for full paper review. After checking the bibliography of relevant papers, six additional publications were selected for full paper review. Eighty-seven papers reporting on 74 studies were included in the review. Sixty-eight papers were excluded for reasons outlined in Figure 1.

Fig. 1.

Flowchart of the retrieval and selection process of papers included in the final review.

All biomarkers and their abbreviations as used further in the text are summarized in Table 1.

Table 1.

Biomarkers considered in this review

Acronym/abbreviation	Legend	Main source
AP	Alcaline phosphatase	Liver, bone, intestine, placenta, brush border proximal convoluted tubules
α1MG	Alpha 1 microglobulin	Liver. Reabsorption by renal proximal tubular cells
α1acidGP	Alpha 1 acid glycoprotein	Liver. Reabsorption by renal proximal tubular cells
Β2MG	Beta 2 microglobulin	All nucleated cells. Reabsorption by renal proximal tubular cells
Cystatin C	Cystatin C	All nucleated cells. Reabsorption by renal proximal tubular cells
FENA	Fractional excretion of sodium
GGTP	Gamma glutamyl transpeptidase	All cells except myocytes. Mainly liver and kidney (brush border proximal convoluted tubules and loop of Henle)
αGST	Alpha glutathione S transferase	Expressed in almost all tissues. Kidney: proximal tubular cells (cytoplasmatic)
πGST	Pi glutathione S transferase	Expressed in almost all tissues. Kidney: distal tubular cells (cytoplasmatic)
HGF	Hepatocyte Growth Factor	Mesenchymal cells
IL-6	Interleukin 6	T lymphocytes, macrophages, endothelial cells, monocytes
IL-8	Interleukin 8	Monocytes, macrophages, epithelial cells, endothelial cells
IL-10	Interleukin 10	Monocytes, lymphocytes, macrophages
IL-18	Interleukin 18	Monocytes, dendritic cells, macrophages and epithelial cells
KIM-1	Kidney injury molecule 1	Kidney: proximal tubular cells
LFABP	Liver-type fatty acid-binding protein	Hepatocytes, kidney: proximal tubular cells
NGAL	Neutrophil gelatinase-associated lipocalin	Leucocytes, loop of Henle and collecting ducts
NAG	N-Acetyl beta glucosaminidase	Several tissues (liver, brain, spleen etc.). Kidney: proximal tubular cells (lysosomal)
PAI-1	Plasminogen activator inhibitor 1	Endothelium
PCX	Podocalyxin	Podocytes
RBP	Retinol-binding protein	Liver. Reabsorption by renal proximal tubula cells
sTNFR-I	Soluble tumour necrosis factor receptor I	Most cells and tissues (cytotoxic, apoptotic and proinflammatory effects)
sTNFR-II	Soluble tumour necrosis factor receptor II	Most cells and tissues (proliferative and anti-apoptotic effects)
TNFα	Tumour necrosis factor alpha	Macrophages, lymfoïd cells, renal parenchymal cells
11k-TXB2	11-keto-Thromboxane B2	Platelets
vWF	Von Willebrand factor	Endothelium, megakaryocytes, subendothelial connective tissue

Results of the data extraction of the 87 selected papers are presented in Tables 2–6. We organized papers according to the five different clinical settings [paediatrics, cardiac surgery, emergency department, critically ill patients at intensive care unit (ICU) and contrast-induced nephropathy (CIN)].

Table 2.

Paediatric setting

Author	PY	Biomarker	U/S/P	Outcome	AKI def/outcome def	Patients/events	AURoC	PPV (%)	NPV (%)
Mishra et al. [12]	2005	NGAL	U	AKI	Screa*1.5	71/20	0.99	95	100
			S				0.91	82	89
Hirsch et al. [17]	2007	NGAL	U	CIN	Screa*1.5	91/11	0.92	100	96
			P				0.91	80	96
Zappitelli [43]	2007	NGAL	U	AKI	pRIFLE crea	140/106	0.78	NA	NA
				Persistent AKI	AKI lasting >48 h	140/?	0.79	NA	NA
Dent et al. [8]	2007	NGAL	P	AKI	Screa*1.5	120/45	0.96	84	93
Bennett et al. [6]	2008	NGAL	U	AKI	Screa*1.5	196/99	0.95	89	83
				RRT		196/4	0.86	NA	NA
Wheeler et al. [42]	2008	NGAL	S	AKI	BUN > 100 mg/dL, Screa >2 mg/dL in absence of CKD or RRT	143/22	0.68	39	94
Krawczeski et al. [10]	2011	NGAL	U	AKI	Neonates: ↑Screa >0.3 mg/dL	374/112	0.95 (neonates)	80	100
					Children: Screa*1.5 within 48 h after CPB		0.92 (others)	73	93
			P				0.95 (neonates)	78	96
							0.94 (others)	77	95
Portilla et al. [14]	2008	NGAL	U	AKI	Screa*1.5	40/21	1.00	100	100
Du et al. [18]	2011	NGAL	U	AKI	pRIFLE crea	252/18	0.66–0.67	NA	NA
Parikh et al. [13]	2011	NGAL	U	AKI	Progression of established AKI	311/53	0.71	30	93
			P				0.56	20	86
Parikh et al. [13]	2011	IL-18	U	AKI	Progression of established AKI	311/53	0.72	30	92
Parikh et al. [16]	2006	IL-18	U	AKI	Screa*1.5	55/20	0.75	65	76
Du et al. [18]	2011	IL-18	U	AKI	pRIFLE crea	252/6	0.44–0.54	NA	NA
		IL-18	U	RIFLE-I			0.48–0.64	NA	NA
Washburn [41]	2008	IL-18	U	AKI	pRIFLE crea	137/103	0.54	27	85
Portilla et al. [14]	2008	LFABP	U	AKI	Screa*1.5	40/21	0.81	71	68
Dennen et al. [7]	2010	IL-6	U	AKI	Screa*1.5	23/10	NA	60	10
Liu et al. [11]	2009	IL-6	S	AKI	Screa*1.5 within 3 days	39/18	0.76	69	87
		IL-8	S				0.74	52	78
Zappitelli et al. [15]	2011	Clinical model (CM)	S	AKI	Screa*1.5	288/121	0.77	NA	NA
							0.81	NA	NA
		Cystatin C + CM					0.83	NA	NA
		Screa + CM
Krawczeski et al. [9]	2010	Cystatin C	S	AKI	Screa*1.5 within 48 h	374/119	0.81	70	87
Du et al. [18]	2011	KIM-1	U	AKI	pRIFLE crea	252/18	0.61	NA	NA
		β2-MG	U	RIFLE-I		252/6	0.59	NA	NA
		KIM-1	U				0.73	NA	NA
		β2-MG	U				0.80	NA	NA

PY, publication year; U/S/P, urine/serum/plasma; AKI def, definition of acute kidney injury; AURoC, area under the receiver operating characteristic curve; PPV, positive predictive value; NPV, negative predictive value; NA, not available or not applicable; NGAL, neutrophil gelatinase-associated lipocalin; IL-18, interleukin 18; LFABP, liver fatty acid-binding protein; IL-6, interleukin 6; IL-8, interleukin 8; KIM-1, kidney injury molecule 1; β2MG, β2-microglobulin; Screa, serum creatinine; BUN, blood urea nitrogen; AKI, acute kidney injury; CKD, chronic kidney disease; CIN, contrast-induced nephropathy; RRT, renal replacement therapy; pRIFLE, paediatric modified Risk Injury Failure Loss of Kidney Function and End-stage renal disease classification; pRIFLE crea, pRIFLE criteria based on the creatinine criterion (omitting the diuresis criterion); CPB, cardiopulmonary bypass time.

Table 3.

Cardiac surgery

Author	PY	Biomarker	U/S/P	Outcome	AKI def/outcome def	Patients/events	AURoC	PPV (%)	NPV (%)
Parikh [67]	2011	NGAL	U	AKI	Doubling of Screa	1219/60	0.67	NA	NA
			P				0.7	NA	NA
		Screa immediate postop	P				0.72	NA	NA
Xin [68]	2008	NGAL	U	AKI	AKIN both	33/9	0.88	58	91
Wagener [69]	2007	NGAL	U	AD	Screa*1.5	81/6	0.8	4	9
Wagener [46]	2008	NGAL	U	AKI	AKIN crea	426/85	0.61	31	84
Haase-Fielitz [50]	2009	NGAL	P	AKI	Screa*1.5 within 5 days postoperatively	100/23	0.80/0.87 (ICU arr/24 h post)	52/53	93/97
		NGAL	P	AKI	RIFLE	eGFR >60 mL/min: 73/15	0.80/0.87	69/57	94/95
		NGAL	P	AKI RIFLE		100/5	0.73/0.70	73/69	69/61
		NGAL	P	RRT + M			0.95	57	99
		NGAL	P	RRT + M		eGFR >60 mL/min: 73/2	NA	NA	NA
Haase-Fielitz [19]	2009	NGAL	P	AKI	Screa* 1.25 within 48 h	100/36	0.66	58	78
					Screa* 1.25 within 72 h	100/38	0.64	55	74
					Screa*1.25 within 120 h	100/39	0.67	55	76
					Screa*1.25 within 168 h	100/40	0.64	57	72
					↑Screa 0.3 mg/dL or *1.5 48 h	100/32	0.66	51	80
					Screa*1.25 or RRT within 72 h	100/42	0.68	66	75
					Screa*1.5 within 48 h	100/20	0.78	44	92
					Screa*1.5 within 72 h	100/21	0.79	46	92
					Screa*1.5 within 120 h	100/23	0.80	51	92
					Screa*1.5 within 168 h	100/23	0.80	51	92
				RIFLE R	RIFLE both	100/31	0.72	52	84
				RIFLE I		100/13	0.79	33	97
				RIFLE F		100/6	0.80	100	98
				AKIN I	AKIN both	100/29	0.75	53	87
				AKIN II		100/11	0.78	33	97
				AKIN III		100/6	0.81	100	98
				RRT		100/4	0.83	100	99
Tuladhar [70]	2009	NGAL	U	AKI	↑Screa 0.5 mg/dL within 48 h	50/9	0.96	47	97
			P				0.85	33	93
Perry [49]	2010	NGAL	P	AKI	Screa*1.5 within 4 days	879/75	0.64	16	93
Prabhu [71]	2010	NGAL	P	AKI	RIFLE crea	30/8	0.98	79	100
McIllroy [51]	2010	NGAL	U	AKI	AKIN crea	426/85
					eGFR <30	21/9	0.34	44	60
					eGFR 30–60	101/20	0.51	22	86
					–eGFR 60–90	142/35	0.55	29	85
					eGFR 90–120	109/13	0.88	40	99
					eGFR >120	53/8	0.27	4	74
Koyner [20]	2008	NGAL	P	AKI	Screa*1.25 or RRT need within 72 h	72/34	0.54	NA	NA
		NGAL	U				0.69	64	67
Haase [72]	2009	NGAL	P	AKI	AKIN both	100/46	0.77	71	77
		NGAL + CysC	S				0.81	78	75
Koyner [73]	2012	NGAL	U	AKI	AKI progression	380/45	NA	NA	NA
			P
Koyner [45]	2010	NGAL	U	AKI Stage 1	AKIN	123/36	0.72	NA	NA
		NGAL	U	AKI	eGFR >60	74/27	0.81	NA	NA
		NGAL	U	AKI	eGFR <60	49/19	0.58	NA	NA
		NGAL	U	AKI Stage 3		123/9	0.88	NA	NA
		NGAL	U	AKI Stage 3	eGFR >60	74/4	0.97	NA	NA
		NGAL	U	AKI Stage 3	eGFR <60	49/5	0.73	NA	NA
Heise [74]	2011	NGAL	U	AKI	AKIN both	47/38	0.77	94	50
Han [44]	2009	NGAL	U	AKI	↑Screa 0.3 mg/dL or 2 to 3-fold increase in Screa within 72 h	90/36	0.59/0.65 (imm/3 h postop)	47/57	69/77
		NGAL	U	Early AKI	↑Screa 0.3 mg/dL within 24 h	90/16	0.51/0.58	NA	NA
		NGAL	U	Late AKI		90/20	0.66/0.71	NA	NA
Liangos [47]	2009	NGAL	U	AKI	Screa*1.5 within 72 h	103/13	0.50	15	90
Koyner [20]	2008	Cystatin C	P	AKI	Screa*1.25 or RRT need within 72 h	72/34	0.62	NA	NA
		Cystatin C	U				0.73	72	73
Haase-Fielitz [50]	2009	Cystatin C	S	AKI	Screa*1.5 within 5 days postoperatively	100/23	0.83/0.84	62/43	93/96
		Creatinine	S				0.68/0.86	47/56	85/93
		Ureum	S				0.60/0.79	32/53	90/88
		Cystatin C	S	AKI		eGFR >60 mL/min: 73/15	0.78/0.84	64/53	93/96
		Creatinine	S				0.69/0.86	54/49	86/97
		Ureum	S				0.61/0.79	35/36	89/100
		Cystatin C	S	AKI RIFLE	RIFLE	100/5	0.75/0.73	75/69	69/71
		Creatinine	S				0.58/0.73	56/75	53/65
		Ureum	S				0.55/0.76	54/76	55/62
		Cystatin C	S	RRT + M			0.99	36	100
		Creatinine	S			eGFR >60 mL/min: 73/2	NA	NA	NA
		Ureum	S				NA	NA	NA
		Cystatin C	S	RRT + M			NA	NA	NA
		Creatinine	S				NA	NA	NA
		Ureum	S				NA	NA	NA
Haase [72]	2009	Cystatin C	S	AKI	AKIN both	100/46	0.76	65	75
Ristikankare [75]	2010	Cystatin C	S	AKI	RIFLE both POD1	110/62	0.71/0.77 (POD1/POD2)	NA	NA
		Creatinine	P		RIFLE crea POD2		0.66/0.74 (POD1/POD2)	NA	NA
Wald [76]	2010	Cystatin C	P	AKI	Screa*1.5 or ↑0.3 mg/dL within 72 h	150/47	0.68	NA	NA
Heise [74]	2011	Cystatin C	U	AKI	AKIN both	47/38	0.59	NA	NA
		α1MG	U				0.61	NA	NA
Koyner [45]	2010	Cystatin C	U	AKI Stage 1	AKIN	123/36	0.72	NA	NA
		Cystatin C	U	AKI	eGFR >60	74/27	0.70	NA	NA
		Cystatin C	U	AKI	eGFR <60	49/19	0.70	NA	NA
		Cystatin C	U	AKI Stage 3		123/9	0.85	NA	NA
		Cystatin C	U	AKI Stage 3	eGFR >60	74/4	0.93	NA	NA
		Cystatin C	U	AKI Stage 3	eGFR <60	49/5	0.81	NA	NA
Liangos [47]	2009	CystatinC	U	AKI	Screa*1.5 within 72 h	103/13	0.5	31	91
Xin [68]	2008	IL-18	U	AKI	AKIN both	33/9	0.89	78	92
Liangos [47]	2009	IL-18	U	AKI	Screa*1.5 within 72 h	103/13	0.66	25	95
Liang [77]	2010	IL-18	U	AKI	RIFLE crea	122/30	0.62	27	79
				Progr AKI	RIFLE R → I or RIFLE ≥ I from start	Progr AKI: 122/11	0.91	17	100
Haase [48]	2008	IL-18	U	RIFLE ≥ R	RIFLE within 120 h. Urine output criterion only during first 24 h	100/50	0.61/0.57 (ICU arr/24 h post)	NA	NA
						100/19	0.52/0.58	NA	NA
				RIFLE ≥ I	Screa*1.25 within 24 h	100/?	0.58/0.56	NA	NA
				AKI	Screa*1.25 within 48 h	100/?	0.59/0.56	NA	NA
					Screa*1.25 within 72 h	100/?	0.60/0.58	NA	NA
					Screa*1.25 within 120 h	100/?	0.60/0.59	NA	NA
					Screa*1.5 within 24 h	100/?	0.56/0.57	NA	NA
					Screa*1.5 within 48 h	100/20	0.53/0.55	NA	NA
					Screa*1.5 within 72 h	100/?	0.58/0.55	NA	NA
					Screa*1.5 within 120 h	100/?	0.52/0.56	NA	NA
				AKI	AKIN ≥ Stage1	100/32	0.48/0.59	NA	NA
				Sustained AKI (at least two consecutively increased Screa levels)	Screa*1.25 within 48 h	100/?	0.58/0.56	NA	NA
					Screa*1.25 within 120 h	100/?	0.59/0.56	NA	NA
					Screa*1.5 within 48 h	100/?	0.56/0.57	NA	NA
					Screa*1.5 within 120 h	100/?	0.51/0.57	NA	NA
					RIFLE ≥ R	100/50	0.61/0.57	NA	NA
					RIFLE ≥ I	100/19	0.52/0.58	NA	NA
Han [44]	2009	KIM-1	U	AKI	↑Screa 0.3 mg/dL or 2 to 3-fold increase in Screa within 72 h	90/36	0.68/0.65	61/72	71/68
		NAG	U				0.61/0.63	50/54	75/69
		Panel of 3	U				0.75/0.78	NA	NA
		KIM-1	U	Early AKI	↑Screa 0.3 mg/dL within 24 h	90/16	0.79/0.73	NA	NA
		NAG	U				0.60/0.59	NA	NA
		Panel of 3	U				0.80/0.84	NA	NA
		KIM-1	U	Late AKI	↑Screa 0.3 mg/dL within 24–72 h	90/20	0.61/0.60	NA	NA
		NAG	U				0.62/0.65	NA	NA
		Panel of 3	U				0.72/0.74	NA	NA
Liangos [47]	2009	KIM-1	U	AKI	Screa*1.5 within 72 h	103/13	0.78	24	98
		KIM1/NAG/IL18					0.78	NA	NA
		KIM-1/CPB time					0.78	NA	NA
		KIM-1/CCF					0.88	NA	NA
Liang [77]	2010	KIM-1	U	AKI	RIFLE crea	122/30	0.88	57	96
				Progr AKI	RIFLE R → I or RIFLE ≥ I from start	Progr AKI: 122/11	0.70	19	97
		KIM-1 + IL-18	U	Progr AKI			0.90	35	90
Koyner [45]	2010	KIM-1	U	AKI Stage 1	AKIN	123/36	0.67	NA	NA
		KIM-1	U	AKI		74/27	0.68	NA	NA
		KIM-1	U	AKI	eGFR >60	49/19	0.64	NA	NA
		KIM-1	U	AKI Stage 3	eGFR <60	123/9	0.82	NA	NA
		KIM-1	U	AKI Stage 3	eGFR >60	74/4	0.80	NA	NA
		KIM-1	U	AKI Stage 3	eGFR <60	49/5	0.82	NA	NA
Jorres [78]	1994	βNAG	U	Renal Injury	Screa > 1.3 mg/dL	36/12	NA	NA	NA
		α1MG	U
		Albumine	U
		Transferrin	U
		IgG	U
		11k-TXB2	U
da Silva Magro [79]	2004	αGST	U	ARF	↓CrCl < 75 mL/min within 72 h	41/20	<0.80	NA	NA
		FENA	U				NA	NA
		Creatinine	P				NA	NA
		Urea	P				NA	NA
		CrCl					NA	NA
Eijkenboom [80]	2005	αGST	U	ARF	Screa*1.5	84/1	NA	NA	NA
		πGST	U
Wagener [46]	2008	CPB time	U	AKI	AKIN crea	426/85	0.59	NA	NA
		AXT					0.59	NA	NA
Liangos [47]	2009	NAG	U	AKI	Screa*1.5 within 72 h	103/13	0.62	26	91
		α1MG					0.62	21	94
		CPB perf time					0.67	NA	NA
		CCF score					0.83	NA	NA
Liangos [81]	2009	IL-8	P	AKI	1. Screa*1.5 or ↑ 0.3 mg/dL	143/41	0.62	NA	NA
					2. Screa*1.5within 72 h	143/18	0.72	NA	NA
Gueret [82]	2009	IL-6	P	ARD	Faint ARD: Screa*1.25	63/36	NA	NA	NA
					Moderate ARD: Screa*1.5
					Severe ARD: Screa*2 within 48 h
Koyner [45]	2010	HGF	U	AKI Stage 1	AKIN	123/36	0.53	NA	NA
		αGST	U				0.62	NA	NA
		πGST	U				0.60	NA	NA
		FENA	U				0.49	NA	NA
		FEUrea	U				0.59	NA	NA
		HGF	U	AKI	eGFR >60	74/27	0.52	NA	NA
		αGST	U				0.66	NA	NA
		πGST	U				0.61	NA	NA
		FENA	U				0.50	NA	NA
		FEUrea	U				0.52	NA	NA
		HGF	U	AKI	eGFR <60	49/19	0.54	NA	NA
		αGST	U				0.57	NA	NA
		πGST	U				0.56	NA	NA
		FENA	U				0.46	NA	NA
		FEUrea	U				0.51	NA	NA
		HGF	U	AKI Stage 3		123/9	0.65	NA	NA
		αGST	U				0.68	NA	NA
		πGST	U				0.78	NA	NA
		FENA	U				0.65	NA	NA
		FEUrea	U				0.57	NA	NA
		HGF	U	AKI Stage 3	eGFR >60	74/4	0.73	NA	NA
		αGST	U				0.75	NA	NA
		πGST	U				0.93	NA	NA
		FENA	U				0.66	NA	NA
		FEUrea	U				0.60	NA	NA
		HGF	U	AKI Stage 3	eGFR <60	49/5	0.44	NA	NA
		αGST	U				0.61	NA	NA
		πGST	U				0.56	NA	NA
		FENA	U				0.63	NA	NA
		FEUrea	U				0.35	NA	NA

Same as in Table 2. Not defined in Table 2: ARF, acute renal failure; ARD, acute renal dysfunction; Progr AKI, progressive AKI; ICU, intensive care unit; imm, immediately; ICU arr, ICU arrival; 24 h post, 24 h postoperatively; imm/3 h, immediately/3 h after surgery; POD, postoperative day; AKIN, acute kidney injury network; RIFLE crea, staging according to RIFLE based on the creatinine criterion (omitting the diuresis criterion); AKIN crea, staging according to AKIN based on the creatinine criterion (omitting the diuresis criterion); RIFLE/AKIN both, staging according to RIFLE or AKIN based on both criteria (creatinine and diuresis); CrCl, creatinine clearance; eGFR, estimated glomerular filtration rate; αGST, α-glutathione-S-transferase; πGST, π-glutathione-S-transferase; α₁MG, α₁ microglobulin; AXT, aortic cross-clamp time; βNAG, N-acetyl-β-d-glucosaminidase; CysC, cystatine C; CCF, Cleveland Clinic Foundation Score; CPB time, CPB perfusion time; IgG, immunoglobulin G; HGF, hepatocyte growth factor; 11k-TXB₂, 11-keto-thromboxane B₂; FENA, fractional excretion of sodium; FEUrea, fractional excretion of urea.

Table 4.

Emergency department

Author	PY	Biomarker	U/S/P	Outcome	AKI def/outcome def	Patients/events	AURoC	PPV (%)	NPV (%)
Nickolas et al. [21]	2008	NGAL	U	AKI	RIFLE crea	635/30	0.95	90	99.5
Shapiro et al. [23]	2010	NGAL	P	AKI	↑0.5 mg/dL or RRT need within 72 h	661/24	0.82	7	99
				≥RIFLE R	Screa*1.5	661/27	NA	7	98
				≥RIFLE I	Screa*2	661/15	NA	4	99
Nickolas et al. [22]	2012	NGAL	U	Intrinsic AKI		1635/96	0.81	23	97
		T0 creatinine > 1.4	S				0.90	28	98
		T0 creatinine > 1.1	S				NA	17	99
		IL-18					0.64	14	94
		KIM-1	U				0.71	17	95
			U
Nickolas et al. [21]	2008	NAG	U	AKI	RIFLE crea	635/30	0.71	9	98
		α1MG	U				0.89	17	99
		α1acidGP	U				0.83	10	99
		FENA	U				0.71	16	94
		Creatinine	S				0.92	35	99
Shapiro et al. [23]	2010	Creatinine	P	≥RIFLE I	Screa*1.5	661/27	NA	4	99
				AKI	Screa*2	661/15	NA	3	99
Soto et al. [24]	2010	Cystatin C	S	AKI	AKIN crea	616 AKI: 130 Prerenal azotaemia: 159 Stable CKD: 15	0.87	48	94
			U				0.59	32	84
		Creatinine	S	AKI	AKIN crea		0.9	54	95
			U				0.62	NA	NA

Same as in Tables 2 and 3. Not defined in Tables 2 and 3: α₁acidGP, α₁ acid glycoprotein; T0, at admission.

Table 5.

Critically ill patients at ICU

Author	PY	Biomarker	U/S/P	Outcome	AKI def/outcome def	Patients/events	AURoC	PPV (%)	NPV (%)
Ahlström [83]	2004	Cystatin C	S	ARF	RIFLE F both	202/54	0.89	NA	NA
Herget-Rosenthal [84]	2004	Cystatin C	U	RRT	Patients with non-oliguric ATN	73/26	0.92	75	95
Herget-Rosenthal [85]	2004	Cystatin C	S	ARF	RIFLE crea	85/44	Rday-2/-1: 0.82/0.97	92/95	66/83
							Iday-2/-1: 0.92/0.98	100/100	63/81
							Fday-2/-1:0.97/0.99	100/100	76/93
		Cystatin C	S	RRT		85/17	Rday-2/-1:0.69/0.75	45/76	86/93
Mazul-Sunko [86]	2004	Cystatin C	P	ARF	Screa ≥ 267 µmol/L or diuresis <30 mL/h in patients without CKD	29/10	NA	NA	NA
Hei [87]	2008	Cystatin	S	ARF	↑Screa to 132 µmol/L or ↑BUN to 18 mmol/L	60/10	NA	42	85
			S
Perianayagam [58]	2009	Cystatin C	S	RRT or M	Inclusion of patients with ARF defined as BL ≤ 1.9 mg/dL: ↑Screa 0.5 mg/dL	200/84	0.65	NA	NA
		Clinical model + cystatin C					0.83	NA	NA
		Clinical model + creatinine					0.83	NA	NA
		Clinical model + urea					0.84	NA	NA
		Clinical model + urine output					0.84	NA	NA
		Clinical model alone					0.82	NA	NA
Portal [88]	2010	Cystatin C	S	AKI	AKIN crea	80/30	0.78	NA	NA
		Cystatin C	S	Severe AKI	Severe AKI: ≥stage 2	80/19	0.78	NA	NA
Nejat [89]	2010	Cystatin C	P	AKI	AKIN crea	444/198	0.78	NA	NA
		Creatinine	S				0.87	NA	NA
		Cystatin C	P	AKI 7d	AKIN crea within 7 days	319/73 (no AKI on entry)	0.65	NA	NA
		Creatinine	S				0.61	NA	NA
		Cystatin C	P	AKI sust	Screa*1.5 within 7 days and ≥24 h	319/19 (no AKI on entry)	0.80	NA	NA
		Creatinine	S				0.57	NA	NA
		Cystatin C	P	RRT		319/? (no AKI on entry)	0.84	NA	NA
		Creatinine	S				0.77	NA	NA
Nejat [59]	2010	Cystatin C	U	AKI 48 h	AKIN crea	319/73 (no AKI on entry)	0.54	NA	NA
			U	AKI 48 h		51/? (sepsis, no AKI on entry)	0.71	NA	NA
						51/? (sepsis, no AKI on entry)
			P	AKI 48 h		268/? (no sepsis, no AKI on entry)	Not predictive	NA	NA
			U	AKI 48 h		268/? (no sepsis, no AKI on entry)	0.45	NA	NA
			P	AKI 48 h			NA	NA	NA
Metzger [63]	2010	Cystatin C	S	AKI	AKIN both 2/16	20/9	0.67	NA	NA
Endre [62]	2011	Cystatin C	U	AKIN48	AKIN crea	381 (no AKI on entry)/82	0.55	NA	NA
		Cystatin C	U	RIFLE 24 h	RIFLE crea, sustained for ≥24 h	381/27	0.63	NA	NA
		Cystatin C	U	AKIN48	eGFR <60 mL/min	69/20	0.64	NA	NA
					eGFR 60 to <90 mL/min	116/25	0.54	NA	NA
					eGFR 90 to <120 mL/min	128/26	0.58	NA	NA
					eGFR >120 mL/min	66/11	0.35	NA	NA
		Cystatin C	U	RRT		128/26	0.66	NA	NA
Royakkers [90]	2011	Cystatin C	U	AKI	RIFLE both	151	Day-2: 0.49	NA	NA
						No-AKI: 60	Day-1: 0.46	NA	NA
						AKI at adm.: 56
						AKI after adm.: 35
			S	AKI	RIFLE both		Day-2: 0.72	NA	NA
							Day-1: 0.62	NA	NA
			U	RRT		151/14	0.61	NA	NA
			S				0.66	NA	NA
Niemann [91]	2009	NGAL	S	AKI	RIFLE crea	59/27	NA	NA	NA
						45 (Screa baseline < 1.5 mg/dL)/24	0.79	NA	NA
Makris [92]	2009	NGAL	U	AKI	RIFLE both	31/11	0.98	91	95
		Creatinine	S				0.79	NA	NA
Aghel [93]	2010	NGAL	S	Worsening renal function	↑Screa ≥ 0.3 mg/dL within 5 days	91/35	0.70	54	86
		BUN	S				0.56	NA	NA
		eGFR					0.61	NA	NA
Bagshaw [55]	2010	NGAL	P	Worsening AKI/RRT	RIFLE	83 (sepsis: 43)/20	0.71/0.78	NA	NA
			U			RRT: 13	0.70/0.70	NA	NA
Märtensson [94]	2010	NGAL	U	AKI	RIFLE or AKIN both	45/18	0.86	100	84
			P				0.85	79	88
		NGAL	U			25(only septic shock)/18	0.86	100	58
			P				0.67	81	54
Siew [56]	2009	NGAL	U	AKI 24 h	AKIN crea	451/64	0.71	NA	NA
		Clinical model					0.81	NA	NA
		NGAL + clinical model					0.82	NA	NA
		NGAL	U	AKI 24 h	AKIN crea, eGFR ≥75 mL/min	275/18	0.77	NA	NA
		NGAL	U	AKI 48 h		451/86	0.64	NA	NA
		NGAL	U	Sust AKI 24 h	Persistent AKI 24 h post-detection	451/47	0.70	NA	NA
			U	Sust AKI 48 h	Persistent AKI 24 h post-detection	451/52	0.66	NA	NA
Constantin [33]	2010	NGAL	P	AKI	RIFLE crea	88/52	0.93	97	80
						56/20 (no AKI on admission)	0.96	94	92
						88/7
				RRT			0.79	21	98
Cruz [34]	2010	NGAL	P	AKI	RIFLE both within 5 days	301/133	O.67	26	91
				AKI 48 h	AKI within 48 h	301/?	0.78	24	97
				RRT		133/15	0.82	12	99
Yang [95]	2010	NGAL	U	Failure of recovery	Failure to return to baseline or RRT need Inclusion of AKI patients according to RIFLE crea	100/35	0.88	72	93
Endre [62]	2011	NGAL	U	AKIN48	AKIN crea	381 (no AKI on entry)/82	0.55	NA	NA
		NGAL	U	RIFLE 24 h	RIFLE crea, sustained for ≥24 h	381/27	0.68	NA	NA
		NGAL	U	AKIN48	eGFR <60 mL/min	69/20	0.71	NA	NA
			U		eGFR 60 to <90 mL/min	116/25	0.53	NA	NA
			U		eGFR 90 to <120 mL/min	128/26	0.53	NA	NA
			U		eGFR >120 mL/min	66/11	0.44	NA	NA
		NGAL	U	RRT			0.78	NA	NA
Haase [31]	2011	NGAL	U	RRT		1345/29	NA	5	99.5
			P	RRT		977/30	NA	6	99
		Creatinine	S	RRT		1345/29	NA	6	99
De Geus [57]	2011	NGAL	P	RIFLE R	RIFLE crea	632/67	0.77	40	97
				RIFLE I		632/48	0.80	NA	NA
				RIFLE F		632/56	0.86	NA	NA
			U	RIFLE R		632/67	0.80	35	95
				RIFLE I		632/48	0.85	NA	NA
				RIFLE F		632/56	0.88	NA	NA
		eGFR		RIFLE R		632/67	0.84	NA	NA
				RIFLE I		632/48	0.87	NA	NA
				RIFLE F		632/56	0.92	NA	NA
		NGAL	P	RIFLE I/F	RIFLE crea, eGFR >60 mL/min	632/104	0.75	NA	NA
		NGAL	U	RIFLE I/F	RIFLE crea, eGFR >60 mL/min	632/104	0.79	NA	NA
		Creatinine	S	RIFLE I/F	RIFLE crea, eGFR >60 mL/min	632/104	0.65	NA	NA
		eGFR		RIFLE I/F	RIFLE crea, eGFR >60 mL/min	632/104	0.67	NA	NA
		Clinical model		RIFLE F	RIFLE crea	632/56	0.95	NA	NA
		Clinical model + NGAL		RIFLE F	RIFLE crea	632/56	0.96	NA	NA
		Clinical model	P
		Clinical model + NGAL		RIFLE F	RIFLE crea	632/56	0.94	NA	NA
		NGAL		RIFLE F	RIFLE crea	632/56	0.95	NA	NA
			U
		Creatinine	U	RRT		632/28	0.89	NA	NA
		eGFR	P				0.88	NA	NA
			S	RRT			0.90	NA	NA
				RRT			0.91	NA	NA
Metzger [63]	2010	NGAL	U	AKI	AKIN crea	20/9	0.54	NA	NA
Siew [60]	2010	NGAL	U	AKI 24 h	AKIN crea	451/64	0.71	NA	NA
		NGAL + IL-18	U				0.71	NA	NA
Portal [88]	2010	NGAL	P	AKI	AKIN crea	80/30	0.79	71	81
			U				0.76	NA	NA
		NGAL	P	Severe AKI	Severe AKI: ≥Stage 2	80/19	0.87	61	91
			U				0.84	NA	NA
		NGAL	S	RRT			0.84	NA	NA
Endre [62]	2011	IL-18	U	AKIN48	AKIN crea	381 (no AKI on entry)/82	0.55	NA	NA
		IL-18	U	RIFLE 24 h	RIFLE crea, sustained for ≥24 h	381/27	0.72	NA	NA
		IL-18	U	AKIN48	eGFR <60 mL/min	69/20	0.65	NA	NA
			U		eGFR 60 to <90 mL/min	116/25	0.48	NA	NA
			U		eGFR 90 to <120 mL/min	128/26	0.57	NA	NA
			U		eGFR >120 mL/min	66/11	0.49	NA	NA
		IL-18	U	RRT			0.70	NA	NA
Siew [60]	2010	IL-18	U	AKI 24 h	AKIN crea	451/64	0.62	NA	NA
		IL-18	U	AKI 24 h	AKIN crea, eGFR ≥75 mL/min	275/18	0.67	NA	NA
		IL-18	U	AKI 48 h	AKIN crea	451/86	0.60	NA	NA
		IL-18	U	AKIN I	AKIN crea	451/61	0.59	NA	NA
		IL-18	U	AKIN II + III	AKIN crea	451/25	0.62	NA	NA
		IL-18	U	RRT		451/17	NA	NA	NA
Metzger [63]	2010	IL-18	U	AKI	AKIN crea	20/9	0.57	NA	NA
Parikh [96]	2005	IL-18	U	AKI	Screa*1.5 within 6 days	138/52	24 h before ↑Screa 0.73	62	78
							48 h before ↑Screa 0.65	NA	NA
Endre [62]	2011	KIM-1	U	AKIN48	AKIN crea	381 (no AKI on entry)/82	0.55	NA	NA
		KIM-1	U	RIFLE 24 h	RIFLE crea, sustained for ≥24 h	381/27	0.64	NA	NA
		KIM-1	U	AKIN48	eGFR <60 mL/min	69/20	0.66	NA	NA
			U		eGFR 60 to <90 mL/min	116/25	0.44	NA	NA
			U		eGFR 90 to <120 mL/min	128/26	0.65	NA	NA
			U		eGFR >120 mL/min	66/11	0.37	NA	NA
		KIM-1	U	RRT		381/12	0.63	NA	NA
Metzger [63]	2010	KIM-1	U	AKI	AKIN crea	20/9	0.71	NA	NA
Liangos [39]	2007	KIM-1	U	RRT + M	↑Screa 0.5 mg/dL	201/96	0.61	NA	NA
		NAG + KIM-1	U				0.71	NA	NA
		Apache + KIM-1					0.80	NA	NA
Westhuyzen [97]	2003	αGST	U	ARF	Screa*1.5 and ↑ ≥0.15 mmol/L	26/9	0.89	60	95
		π GST	U				0.93	67	100
		γGT	U				0.95	67	100
		AP	U				0.86	67	90
		NAG	U				0.85	50	100
		LDH	U				0.69	100	96
		Cr Cl					0.80	50	91
Iglesias [98]	2003	IL-6	U	ARF	Screa >3.5 mg/dL or RRT need	537/112	NA	NA	NA
		TNF-α	U		If baseline Screa 1.8–3 mg/dL: Screa*2 or RRT need
		sTNFR-I	U
		sTNFR-II	U
Herget-Rosenthal [84]	2004	α GST	U	RRT	Patients with non-oliguric ATN	73/26	0.64	NA	NA
		γGT	U				0.64	NA	NA
		NAG	U				0.81	55	88
		α1MG	U				0.86	72	93
		RBP	U				0.80	NA	NA
		β2MG	U				0.51	NA	NA
		LDH	U				0.59	NA	NA
		Liano score	U				0.83	63	85
Liangos [39]	2007	NAG	U	RRT + M	Inclusion of patients with ARF defined as:	201/96	0.71	NA	NA
		Screa enrollment	U				0.60	NA	NA
		Urine output			BL ≤ 1.9 mg/dL: ↑Screa 0.5 mg/dL		0.65	NA	NA
		Apache score			BL 2–4.9 mg/dL: ↑Screa 1 mg/dL		0.78	NA	NA
		Apache + NAG			BL ≥5 mg/dL: ↑Screa 1.5 mg/dL		0.79	NA	NA
Liu [99]	2007	Clinical model		AKI	Screa*1.5 within 4 days	876/209	0.66	NA	NA
		Clinical model + biomarkers PAI-1 and sTNFR-I (not included in model, because not predictive: IL-6,IL-8,IL-10,TNF alfa, vWF, sTNRF-II, IADM-1)	P				0.7	NA	NA
		Clinical model		AKI	Screa*1.5 day 1		0.72	NA	NA
		Clinical model + biomarkers					0.77	NA	NA
Chawla Lakhmir [54]	2007	IL-6	P	AKI	Screa*1.25 or ↑0.3 mg/dL	547/127	NA	NA	NA
					during first week
Hei [87]	2008	β2-MG	U	ARF	↑Screa to 132 µmol/L or ↑BUN to 18 mmol/L	60/10	NA	15	79
		NAG	U				NA	21	90
		β2-MG	S				NA	24	87
		Creatinine	S				NA	100	88
Walshe [100]	2009	αGST	U	AKI	AKIN both	38/19	Biomarker not a good predictor of AKI	NA	NA
		πGST
Portal [88]	2010	Creatinine	S	AKI	AKIN crea	80/30	0.72	NA	NA
		eGFR					0.71	NA	NA
		APACHEII					0.74	NA	NA
		SOFA					0.67	NA	NA
		Creatinine	P	Severe AKI	Severe AKI: ≥Stage 2	80/19	0.81	NA	NA
		eGFR	U				0.77	NA	NA
		APACHE II	S				0.87	NA	NA
		SOFA					0.75	NA	NA
Metzger [63]	2010	MP ICU-T	U	AKI	AKIN both 2/16	30/16	0.91	94	93
		MP ICU-V	U		AKIN crea	20/9	0.84	80	90
Matsui [101]	2010	LFABP	U	AKI	AKIN crea	25/14	0.95	1	85
		NAG	U				0.63	70	53
		Albumine	U				0.70	77	67
		PCX	U				0.42	80	50
Endre [62]	2011	γGT	U	AKIN48	AKIN crea	381 (no AKI on entry)/82	0.57	NA	NA
		AP	U				0.56	NA	NA
		γGT	U	RIFLE 24 h	RIFLE crea, sustained for ≥24 h within 7 days	381/27	0.61	NA	NA
		AP	U		eGFR <60 mL/min		0.64	NA	NA
		γGT	U	AKIN48	eGFR 60 to <90 mL/min	69/20	0.79	NA	NA
			U		eGFR 90 to <120 mL/min	116/25	0.48	NA	NA
			U		eGFR >120 mL/min	128/26	0.50	NA	NA
			U		eGFR <60 mL/min	66/11	0.56	NA	NA
		AP	U		eGFR 60 to <90 mL/min	69/20	0.71	NA	NA
			U		eGFR 90 to <120 mL/min	116/25	0.48	NA	NA
					eGFR >120 mL/min	128/26	0.54	NA	NA
						66/11	0.55	NA	NA
		γGT	U	RRT		381/12	0.63	NA	NA
		AP					0.72	NA	NA
Blasco [102]	2011	γGT	U	AKI	CrCl 25% lower than theoretical CrCl based on age and gender	100/36	0.86	80	88
		Creatinine	P				0.85	70	87
		γGT + creatinine					NA	66	100

Same as in Tables 2–4. Not defined in Tables 2–4: ATN, acute tubular Necrosis; AKI sust, sustained AKI; M, mortality; Rday-2/-1, 1 and 2 days before AKI diagnosed by serum creatinine; IDay-2/-1, 1 and 2 days before AKI diagnosed by serum creatinine; FDay-2/-1, 1 and 2 days before AKI diagnosed by serum creatinine; at adm, at admission; after adm, after admission; γGT, γ-glutamyl transferase; AP, alkaline phosphatase; LDH, lactate dehydrogenase; RBP, retinol-binding protein; PAI-1, plasminogen activator inhibitor-1; sTNFR-I and II, soluble tumour necrosis factor receptor I and II; IADM-1, intracellular adhesion molecule 1; TNFα, tumour necrosis factor alpha; vWF, von Willebrand factor; IL-10, interleukin-10; MP ICU-T, marker pattern ICU training set; MP ICU-V, marker pattern ICU validation set; pNGAL, plasma NGAL; PCX, podocalyxin; BL, baseline serum creatinine.

Table 6.

Contrast-induced nephropathy

Author	PY	Biomarker	U/S/P	Outcome	AKI def/Outcome def	Patients/events	AURoC	PPV (%)	NPV (%)
Nakamura [103]	2006	L-FABP	U	CIN	↑Screa > 0.5 mg/dL or *1.25 within 2–5 days	66/13	NA	NA	NA
Bachorzewska-Gajewska [104]	2007	NGAL	U	CIN	Screa*1.25 within 48 h	100/11	NA	NA	NA
		NGAL	S				NA	NA	NA
Ling [64]	2008	NGAL	U	CIN	↑Screa 0.3 mg/dL: or *1.25	40/13	0.73	20	97
Shaker [105]	2010	NGAL	S	CIN	Screa*1.25	30/2	NA	NA	NA
Bulent Gul [65]	2008	IL-18	U	CIN	↑Screa > 0.5 mg/dL or *1.25	51/15	NA	NA	NA
Ling [64]	2008	IL-18	U	CIN	↑Screa 0.3 mg/dL or *1.25	40/13	0.75	20	96
Kato [106]	2008	Cystatin C	P	CIN	↑Screa 0.5 mg/dL or *1.25 within 48 h	87/18	0.93	63	98
		Creatinine	S				0.83	42	89
Briguori [107]	2010	Cystatin C	S	CIN	↑Screa 0.3 mg/dL 48 h PP or RRT	410/34	0.92	39	100
Ishibashi [108]	2010	Cystatin C	S	CIN	↑Screa 0.5 mg/dL or *1.25 within 48 h	100/18	>0.50	68	96
		Creatinine	S				>0.50	NA	NA
Bachorzewska-Gajewska [104]	2007	Cystatin C	S	CIN	Screa*1.25 within 48 h	100/11	NA	NA	NA
Shaker [105]	2010	Cystatin C	S	CIN	Screa*1.25	30/2	NA	NA	NA

Same as in Tables 2–5. Not defined in Tables 2–5: PP, post-procedure.

In the paediatric setting, 16 papers reporting on 11 studies were discussed. Eleven papers reporting on seven studies were conducted post-cardiac surgery [6–16], one studied CIN [17], one was conducted in the emergency department setting [18] and three papers reporting on two studies were conducted in critically ill children. Only one study in the paediatric setting [6] explored the use of biomarkers for early prediction of need for RRT [196 patients, 4 events, area under the receiver operating characteristic curve (AURoC) 0.86, PPV and NPV not available]. All other studies considered the predictive value for AKI, which was defined according to seven different definitions. The number of patients varied from 23 to 374, and the number of events varied from 6 to121. AURoCs for the prediction of AKI varied from 0.44 to 1.00. PPVs and NPVs ranged from 27 to 100% and from 10 to 100%, respectively.

In adult cardiac surgery, 26 papers reporting on 22 studies were included. Twenty-seven different definitions for AKI were used. Patient number varied from 30 to 1219 and event number varied from 1 to 85. The AURoCs varied from 0.27 to 0.98. PPVs and NPVs ranged from 4 to 100% and from 61 to 100%, respectively. One study [19] reported on the use of biomarkers for the prediction of need for RRT (100 patients, 4 events, AURoC 0.83, PPV 100%, NPV 99%), although another study [20] also defined AKI as a 25% increase in serum creatinine (Screa) or RRT need with AURoCs between 0.54 and 0.73 and PPV 64–72%, NPV 67–73%.

In the emergency department setting, four studies were included [21–24]. The patient number varied from 616 to 1635 and the event number from 24 to 130. Three different definitions of AKI were used. AURoCs varied from 0.59 to 0.95. PPVs and NPVs varied from 4 to 90% and from 84 to 99.5%, respectively.

In critically ill patients, 33 papers reporting on 29 studies, of which 1 was a meta-analysis, were included. Twenty-one different definitions were used to define AKI. The patient number varied from 26 to 1345 and the event number varied from 4 to 209. AURoCs for the prediction of AKI varied from 0.35 to 0.99 and PPVs and NPVs ranged from 1 to 100% and from 50 to 100%, respectively. Fifteen papers reported on the adverse clinical outcomes of AKI (failure of recovery, RRT or the composite outcome of RRT and mortality). AURoCs for the prediction of adverse clinical outcomes varied from 0.51 to 0.92 (PPV 5–75% and NPV 86–99.5%).

Eight studies exploring the use of biomarkers in CIN were included. Seven different definitions of CIN were used. The patient number ranged from 30 to 410 and event number from 2 to 34. AURoCs varied from 0.73 to 0.93. PPVs and NPVs ranged from 20 to 68% and from 96 to 100%, respectively.

Discussion

This paper evaluated the state of the art on the diagnostic usefulness of biomarkers in AKI. Most striking was the wide array of definitions used for AKI. With regard to the diagnostic usefulness of biomarkers, a variety of results ranging from overtly negative to very optimistic were found.

There are different explanations for this wide range of results. First, widely different clinical settings are investigated, from paediatric post-cardiac bypass, where timing and amount of renal impact are exactly known, to patients with septic shock, where timing of renal insult is unknown. Overall, diagnostic performance of biomarkers appears to be better in situations with a known timing and aetiology of renal injury. Second, most of these markers are not only associated with kidney damage, but also with the underlying conditions causing AKI, such as sepsis, diabetic nephropathy, systemic lupus erythematosus and haemolytic uraemic syndrome [25–28]. Some of these markers are also increased in chronic kidney disease (CKD) [29, 30], blurring the differential diagnosis between CKD and AKI. The reference baseline creatinine of a given patient is not always available, so in these patients, it is unclear whether an increased creatinine is due to AKI or CKD. The behaviour of most biomarkers in patients with CKD is, however, largely unknown. The usefulness of biomarkers should thus be addressed differently for different clinical settings, as it is apparent that performance is strongly dependent on the underlying circumstances. As such, results in one setting cannot be generalized.

Third, one should be cautious for some statistical pitfalls when translating studies to clinical practice. Most studies report higher values for the biomarker in patients with AKI, but with a substantial overlap between AKI and non-AKI patients, hampering discrimination in the single case. AURoCs are mostly used to circumvent this problem; however, the use of this instrument is troublesome in conditions with rather low prevalence, so PPV and NPV should also be provided, to allow a good judgement on the accuracy of the biomarker. The use of odds ratios should be avoided in cases with low prevalence. In a study on performance of uNGAL for example, an impressive 16.4 higher odds ratio to need RRT was reported; however, still only 2.5% of the patients in this NGAL+/Screa− group end up on RRT, and consequently, just as for Screa, NGAL levels should thus not be used to decide on the start of RRT [31–33].

Fourth, many studies use circular reasoning to define AKI by new markers. They start from the premise that an increase in the biomarker without a rise in Screa indicates AKI [34, 35], thus increasing sensitivity, but at the expense of specificity and PPV. This confusion is due to the lack of a gold standard to diagnose AKI, and the difficulty to make a distinction between intrinsic renal damage (histological AKI) and physiological decrease in renal perfusion (functional AKI) in the clinical setting. Undoubtedly, there is also a grey zone, where functional and histological AKI co-exist in a varying mix. Recent work by Nejat et al. [36] indicated that even in cases of transient AKI, markers of tubular damage can be present, demonstrating that ‘functional’ and ‘intrinsic’ AKI are probably gradations of an evolving spectrum of AKI. Urinary markers will probably be more sensitive for true histological damage, whereas serum levels of markers are probably more sensitive for changes in clearance. In research conditions, e.g. to evaluate potential nephrotoxicity, the use of high-sensitivity markers is excellent. However, in clinical practice, we need to define what degree of subclinical damage as detected by biomarkers we will accept. Therefore, we need studies linking acute tubular damage to long-term outcomes. Several studies make the link between AKI, even when apparently fully recovered, and progressive CKD, but in all these, the classic definition of AKI, based on creatinine, was used, and to our knowledge, these data are lacking for biomarkers [37, 38].

Lastly, some studies advertise the use of biomarkers in situations where the outcome already seems predictable using standard parameters, such as clinical appraisal and oliguria.

In the study of Koyner et al. [20], seven patients needed RRT, five of them within 27 h after cardiac surgery, making it plausible that RRT need in these patients could have been predicted by bedside clinical appraisal. Also, the value of biomarkers to predict the need for RRT should be evaluated compared with other available standards, such as clinical appreciation, time on cardiopulmonary bypass (CPB) or urinary output [39]. Singer et al. [40] used uNGAL to separate intrinsic from pre-renal AKI, but 32 out of 104 patients were esteemed ‘unclassifiable’, and were thus not incorporated in the evaluation of the diagnostic accuracy of uNGAL. This means that only patients in whom the distinction between pre-renal azotaemia and AKI could already be established based on clinical grounds were included. From the provided graphs, it is clear that the biomarker in these ‘unclassifiable’ cases were also ‘intermediate’, and thus not discriminative.

Not taking into account all these reflections can lead to erroneous conclusions on the value of biomarkers, resulting in incorrect diagnoses, and thus incorrect therapeutic interventions.

It is clear from our survey that whereas the RIFLE (Risk, Injury, Failure, Loss, End-Stage Kidney Disease) and AKIN (Acute Kidney Injury Network) classification form a giant leap forward, the definition of AKI is still problematic and that both RIFLE and AKIN are applied differently in every individual study in clinical practice, resulting in differences in classification.

The studies of Mishra et al. [12] and Bennett et al. [6] showed excellent performance for NGAL under ideal circumstances: children without co-morbidities after CPB. When more heterogenic populations were studied such as critically ill patients, results became more ambiguous [41–43]. Septic paediatric patients have higher uNGAL and IL-18 levels than non-septic patients, indicating that the association between uNGAL and AKI in septic patients might reflect more the association between the severity of disease and AKI rather than true renal damage.

Reported results on biomarkers in cardiac surgery are disappointing and conflicting, with AURoCs for uNGAL comparable to the predictive power of clinical parameters such as the duration of CPB and aortic cross-clamp time [44–46]. KIM-1 has been reported to have the best AURoC for the prediction of AKI. However, the low PPV showed that three out of four patients are being mislabelled as having AKI [47]. Plasma NGAL and IL-18 also performed poorly [48, 49]. Serum cystatin C performed reasonably when AKI was defined as an increase in Screa of 50%, but poorly when based on RIFLE. More importantly, also Screa distinguished accurately between the AKI and the non-AKI groups within 24 h after arrival at the ICU [50]. In subjects with a baseline estimated glomerular filtration rate (eGFR) of <60 mL/min/1.73 m², uNGAL was not able to predict the development of evolving AKI [51]. The discriminatory ability of uNGAL was less with the decreasing RIFLE class [19]. The major challenge in the emergency setting is to distinguish fluid responsiveness from non-fluid responsive AKI, and AKI from CKD. As we do not have an established clinical definition of ‘pre-renal’ AKI, discrimination of this condition always remains somewhat subjective.

In a study with 635 patients, 411 had normal kidney function, 80 had pre-renal azotaemia, 30 AKI and 106 non-progressive CKD [21]. This implicates that a patient with increased Screa at the emergency department has a 6-fold higher odds to have either CKD or pre-renal kidney dysfunction than to have AKI. Despite the significant difference in median uNGAL levels between the different groups, there was a substantial overlap. Most of the patients with AKI had already very high Screa levels at presentation (mean 5.6 mg/dL, standard deviation 5.5 mg/dL). In patients presenting at the emergency department with suspected sepsis, median NGAL levels were higher in patients with versus without AKI, but with substantial overlap, making discrimination per individual patient impossible [23]. In addition, patients who died without AKI had pNGAL levels comparable to the range of the AKI group, again indicating that the discriminative power of pNGAL for AKI is more related to the fact that it is a marker of the severity of disease than of kidney injury.

In a prospective cohort study of 616 patients admitted to a tertiary care emergency department, both sCysC and Screa but not uCysC differentiated between AKI and non-AKI. The authors state that sCysC distinguishes between AKI and pre-renal azotaemia. Neither biomarker discriminated between AKI and CKD [24].

The application of biomarkers is most cumbersome in critically ill patients. As in this setting the timing of the renal insult is unknown, multiple samples will have to be taken, which all have the potential of being false-positive and may lead to logistical problems as well as being less cost-effective. The pathophysiology of AKI in sepsis patients is complex and not merely due to ischaemia/reperfusion injury, but also to inflammatory processes [52]. It is likely that the levels of many biomarkers, e.g. IL-18 or NGAL, will be influenced by these processes, irrespective of AKI [32]. Plasma markers are thus in these patients more an indication of severity of disease than of true kidney damage [33, 53–55]. Finding a biomarker or a panel of biomarkers that takes into account these different underlying pathological processes while still discriminating AKI is challenging. In most studies, there was no attempt to evaluate the discriminative power of the biomarker on top of other parameters, such as severity of sepsis, or urinary output. Although in a large cohort, uNGAL remained an independent predictor for AKI in a multivariate logistic regression model, the addition of uNGAL did not significantly improve discriminative power of the model. Overall, the clinical model performed better than uNGAL 56. In this study [56], uNGAL had a better discriminative value than eGFR (0.79 versus 0.67) at admission in the subgroup with eGFR >60 mL/min at admission; however, no multivariate analysis on the additional discriminative value of NGAL on top of other clinical parameters was provided for this subgroup.

In ICU patients, uNGAL and pNGAL did perform worse than admission eGFR, and adding pNGAL and uNGAL to a multivariate model with eGFR only improved reclassification for AKI insignificantly [57]. Adding sCysC to a clinical model did not improve the diagnostic performance of the clinical model [58]. Urinary CysC predicted AKI, with a model including uCysC, uCrea, age, hypotension and APACHE II subcategory scores, in 444 critically ill patients, but when excluding patients with overt AKI on admission based on clinical grounds and Screa, the AURoC was only 0.54 [59]. The predictive performance of urinary IL-18 in critically ill patients was poor [60], and uLFABP did not perform well to diagnose established AKI [61].

Some authors showed a better diagnostic performance of biomarkers when stratifying patients according to the presence of pre-existing CKD [60, 62].

Another strategy to optimize the performance of biomarkers in ICU patients was to use a marker pattern (20 urinary peptides) rather than individual biomarkers [63]. However, a panel of 20 peptides measured by capillary electrophoresis-mass spectrometry, with high cost and long turn-around time, might not be realistic in clinical conditions.

Also in CIN, the use of biomarkers is prone to problems. The risk for CIN increases in patients with co-morbid conditions, such as diabetes or older age, and in patients with pre-existing kidney disease, all of which influence on themselves biomarker levels. In patients without pre-existing CKD, the incidence of CIN is very low. In children, plasma and serum uNGAL were reported to be better predictors for CIN than Screa [17]. The diagnosis of AKI was made using NGAL and IL-18 in adults undergoing elective coronarography, percutanous transluminal coronary angioplasty or angioplasty, up to 24 h before diagnosis by Screa, but this was not confirmed in another study [64, 65].

Interestingly, NGAL is known to be increased in atherosclerotic plaques and might be released from these plaques during PCI without being related to kidney injury [66]. It can also be questioned whether CIN in patients with arteriography versus simple intravenous contrast administration is the same condition, as the former can also be induced by cholesterol embolism.

Conclusion

Whereas biomarkers can increase our understanding of the pathophysiology of AKI, it appears that the promising results with new biomarkers for early detection and differential diagnosis of AKI in clinical practice can only be confirmed in the setting of children without co-morbidities and with a well-defined timing of renal injury. Results are far less robust when we search for validation in adult heterogenic populations, including patients with co-morbid conditions such as diabetes mellitus, vascular disease and CKD. Whereas AURoC values sometimes look impressive, applying levels of urinary or serum biomarkers for discrimination in individual patients is hampered by wide overlap between groups, which might result in many false positives. Biomarkers reflect a general degree of severity of disease, rather than being specific for kidney injury. Before biomarkers can be advocated to diagnose AKI, further research on the implications of ‘subclinical’ AKI, i.e. diagnosed by the biomarker but not by an increase in creatinine or decrease of diuresis, should be performed

Conflict of interest statement

None declared.

(See related article by Bagshaw et al. Novel biomarkers of AKI: the challenges of progress ‘Amid the noise and the haste’. Nephrol Dial Transplant 2013; 28: 235–238.)

Author notes

*

The results presented in this paper have not been published previously in whole or part, nor in abstract format