Acute kidney injury (AKI) is a common condition occurring in approximately 36% of patients admitted to the intensive care unit (ICU) [1]. The occurrence of AKI during ICU stay is associated with a prolonged hospital stay [2] and high mortality [3]. To improve patient outcomes, the prevention and early detection of AKI are important. To date, the serum creatinine (sCr) concentration and urinary output are the most widely used biomarkers to detect AKI [4]. Unfortunately, studies have highlighted delays in sCr elevation and AKI diagnosis [5, 6], which may lead to treatment delay [7]. In addition, to use sCr for AKI diagnosis and/or prediction, daily and repeated measurements are necessary. The current definition of AKI is based on the Kidney Disease: Improving Global Outcomes (KDIGO) criteria and requires advance knowledge of the patient’s sCr concentration prior to hospitalization or repeated measurements with ≥6-hr intervals, during which no therapeutic measures should be taken before AKI is diagnosed [4]. This leads to therapeutic delay, which can be deleterious in the most severely ill patients, especially in the ICU [8, 9]. Therefore, the validation of more specific biomarkers that allow earlier diagnosis of AKI is required to improve patient outcomes.

Neutrophil gelatinase-associated lipocalin (NGAL), which belongs to the lipocalin family, is a biomarker of ischemic and toxic kidney injury [10]. Mainly produced by the ascending loop of Henle and the collecting tubule, it exerts a bacteriostatic function by preventing iron uptake [10]. A recent meta-analysis showed that urinary NGAL (uNGAL) concentrations measured during the first hours after admission to the emergency department or ICU and post-cardiac surgery were associated with AKI [11]. Another study revealed an association between an elevated uNGAL concentration on admission and both the occurrence of AKI during coronary care unit stay and 5-yr mortality [12]. Liver-type fatty acid-binding protein (LFABP) of the large superfamily of lipid-binding proteins is localized in the liver [13] and is mainly involved in energy supply to the kidney proximal tubules via fatty acid uptake and protection against oxidative stress [14]. A recent meta-analysis demonstrated the high predictive performance of urinary LFABP (uLFABP) for the occurrence of AKI in various clinical settings (including the ICU) and post-surgery as well as for the occurrence of contrast-induced AKI [15]. High uLFABP concentrations have been associated with an increased risk of AKI occurrence and 90-day mortality in severely ill patients [16, 17]. Cystatin C (CysC) is a non-glycosylated protein produced by all nucleated cells [13]. A recently published prospective cohort study showed that urinary (uCysC) concentrations on admission were associated with the progression to a higher AKI stage and 7-day mortality in the ICU [18].

The predictive performances of the above biomarkers have been individually studied in ICU patients, but the results were inconsistent [19]. This study, an ancillary study of the FROG-ICU (French and euRopean Outcome reGistry in Intensive Care Units) study [20], aimed to assess the predictive performances of uNGAL, uLFABP, and uCysC measured in the early phase of ICU management and at discharge from the ICU in severely ill patients for short- and long-term outcomes.

Study design

This study was an ancillary study of the FROG-ICU cohort study that assessed the prevalence of all-cause mortality in the year following ICU discharge in 20 French and Belgian ICUs between August 2011 and June 2013 [20]. Mechanical ventilation and/or use of vasopressors and/or inotropes for ≥24 hrs were criteria for inclusion [20]. This cohort study was conducted in France and Belgium in accordance with Good Clinical Practice and the Declaration of Helsinki 2002, validated by the relevant ethics committees (Comité de Protection des Personnes - Ile de France IV, institutional review board [IRB] No. 00003835; Commission d’Éthique Biomédicale Hospitalo-Facultaire de l’Hôpital de Louvain, IRB No. B403201213352), and registered on ClinicalTrials.gov (NCT01367093).

Between August 2011 and June 2013, 2,087 patients were enrolled in the FROG-ICU cohort. After excluding patients with confirmed AKI, clinical and biological data were available for 1,759 patients. Data at discharge were available for 609 patients among survivors (Fig. 1).

Figure 1. Study flow chart.
Abbreviations: ICU, intensive care unit; AKI, acute kidney injury.

AKI definition

AKI was defined and staged according to the KDIGO criteria [4]. Given that urinary output was not recorded daily in the FROG-ICU study, AKI was defined based on the daily sCr concentration. The sCr concentration on admission was used as the reference value. Baseline sCr was defined as follows: if the estimated glomerular filtration rate (eGFR) on admission was ≥75 mL/min/1.73 m², we used the actual sCr on admission. If eGFR was <75 mL/min/1.73 m² on admission, we used the sCr calculated back from the Modification of Diet in Renal Disease Study (MDRD) equation [21] set to 75 mL/min/1.73 m².

Data collection and biomarker measurements

Clinical and biological data were collected as previously described [20]. uNGAL was measured using The NGAL Test Reagent kits (BioPorto Diagnostics, Copenhagen, Denmark) on an Architect i2000 analyzer (Abbott Laboratories, Abbott Park, IL, USA). Plasmatic NGAL (pNGAL) was measured using The NGAL Test Reagent kits (BioPorto Diagnostics) on an Architect C8000 analyzer (Abbott Laboratories). uLFABP was measured using latex-enhanced turbidimetric immunoassay kits (Sekisui Medical, Tokyo, Japan) on a TOSHIBA TBA120FR analyzer (Toshiba Medical Systems, Tochigi, Japan). This assay was performed in duplicate according to the manufacturer’s instructions. uCysC and plasmatic CysC (pCysC) were measured using CysC ELISA kits (Abbott Laboratories) on the Architect C8000 analyzer. Urinary biomarkers were measured in spot urine samples. ICU survivors were followed up until 1 yr after ICU discharge.

Outcomes

The primary outcome was the occurrence of AKI during ICU stay according to the KDIGO criteria. Secondary outcomes were 28-day mortality after ICU admission and 1-yr mortality after ICU discharge.

Statistical analysis

Results are expressed as median (interquartile range [IQR]) or as count (percentage [%]), as appropriate. Group comparisons of continuous variables were performed using the Kruskal–Wallis test. Categorical data were compared using Pearson’s chi-square test for count data. ROC curves were used to determine the optimal threshold value (based on the Youden index) for each biomarker, and the area under the ROC curve (AUC) values were calculated and compared using the DeLong test. On ICU admission and at discharge, patients were divided into two groups (high and low concentration) according to the cutoff value. Cox proportional-hazard regression was used to assess the associations between biomarker concentrations and outcomes. Urinary biomarker concentrations on ICU admission and at discharge and other variables associated with the occurrence of AKI during ICU stay in univariate analysis were entered in a multiple logistic regression model to identify factors independently associated with the outcomes. A minimum of 5–10 events for each predictor variable were considered in a logistic stepwise regression model that was based on variables including age, sex, sCr (on ICU admission for AKI and 28-day mortality, at ICU discharge for 1-yr mortality), sequential organ failure assessment (SOFA) score at ICU admission, chronic kidney disease, hypertension, diabetes mellitus, chronic heart failure, and myocardial infarction. Kaplan–Meier plots were used for survival analysis, and logistic regression analysis was performed. Statistical analysis was performed using R statistical software version 4.1.2 (http://www.r-project.org). A two-sided P<0.05 was considered for significance.

Patient characteristics

Patient characteristics are summarized in Table 1. The cohort was mainly composed of men (N=1,136 [64.6%]), with a median age of 63 (51–75) yrs. Patients were mainly admitted to the ICU for respiratory disorders (N=502 [28.5%]) and sepsis (N=204 [11.6%]) (Table 1).

Table 1

Patient characteristics

Characteristic	All patients (N=1,759)	No AKI (N=1,031)	AKI (N=728)	P
Age, yrs	63 (51–75)	60 (48–71)	68 (56–77)	<0.001
Male	1,136 (64.6)	637 (61.8)	499 (68.5)	0.004
BMI, kg/m2	26.5 (23.2–30.5)	25.4 (22.5–29.3)	27.5 (24.4–31.9)	<0.001
Underlying diseases
Chronic heart failure	129 (7.3)	51 (4.9)	78 (10.7)	<0.001
Diabetes mellitus	321 (18.3)	141 (13.7)	180 (24.8)	<0.001
Hypertension	773 (44.0)	373 (36.2)	400 (55.0)	<0.001
Myocardial infarction	65 (3.7)	38 (3.7)	27 (3.7)	>0.900
Dyslipidemia	350 (19.9)	179 (17.4)	171 (23.6)	0.002
Obesity	180 (10.2)	85 (8.2)	95 (13.1)	0.001
Coronary artery disease	155 (8.8)	59 (5.7)	96 (13.2)	<0.001
Atrial fibrillation/flutter	192 (10.9)	95 (9.2)	97 (13.4)	0.008
Chronic kidney disease	173 (9.8)	43 (4.2)	130 (17.9)	<0.001
Diagnosis on admission
Respiratory disorders	502 (28.5)	291 (28.2)	211 (29.1)	0.742
Sepsis	204 (11.6)	85 (8.2)	119 (16.4)	<0.001
Post-major non-cardiac surgery	184 (10.5)	104 (10.1)	80 (11.0)	0.583
Out-of-hospital cardiac arrest	120 (6.8)	68 (6.6)	52 (7.2)	0.713
Acute heart failure	44 (2.5)	15 (1.5)	29 (4.0)	0.001
Post-cardiac surgery	43 (2.4)	20 (1.9)	23 (3.2)	0.138
Clinical presentation
SOFA score	7 (4–10)	6 (3–8)	9 (7–12)	<0.001
Systolic blood pressure, mmHg	122 (108–139)	124 (110–141)	120 (106–136)	<0.001
Diastolic blood pressure, mmHg	61 (53–70)	63 (55–73)	58 (50–67)	<0.001
Mean blood pressure, mmHg	81 (72–92)	83 (74–94)	79 (71–88)	<0.001
Heart rate, beats/min	92 (78–106)	91 (77–105)	92 (79–110)	0.061
Temperature, °C	37.3 (36.8–37.8)	37.3 (36.8–37.9)	37.2 (36.7–37.8)	<0.001
SpO2, %	98 (96–100)	98 (96–100)	98 (96–100)	0.001
LVEF, %	45 (35–58)	43 (34–56)	45 (39–60)	0.419
Biology at inclusion
Hb, g/dL	10.0 (8.9–11.4)	10.1 (9.0–11.6)	9.8 (8.9–11.2)	0.008
White blood count, /mm3	10,900 (7,523–16,000)	10,300 (7,415–14,833)	11,720 (7,978–17,883)	<0.001
Platelets, /µL	166,000 (103,000–246,000)	182,000 (121,000–260,000)	133,000 (81,000–222,000)	<0.001
BNP, pg/mL	243 (77–991)	170 (55–576)	590 (139–1,658)	<0.001
NT-proBNP, pg/mL	2,136 (778–5,154)	2,840 (875–5,138)	2,086 (744–5,154)	0.834
Urea, mmol/L	8.5 (5.3–14.0)	6.8 (4.5–9.9)	12.9 (8.0–18.9)	<0.001
Sodium, mmol/L	140 (137–144)	140 (137–143)	141 (137–144)	0.010
Potassium, mmol/L	3.9 (3.6–4.2)	3.8 (3.5–4.2)	4.0 (3.6–4.4)	<0.001
Creatinine, µmol/L	82 (58–145)	67 (53–86)	154 (99–226)	<0.001
Lactate, mmol/L	1.3 (1.0–1.9)	1.2 (0.9–1.7)	1.5 (1.1–2.3)	<0.001
Diuresis of 24 hrs, mL	1,400 (825–2,200)	1,565 (1,084–2,395)	1,000 (500–1,848)	<0.001
Therapeutics
Inotropes or vasopressors	1,351 (76.8)	716 (69.4)	635 (87.2)	<0.001
Mechanical ventilation	1,647 (93.6)	970 (94.1)	677 (93.0)	0.411
Outcomes
28-day mortality	356 (20.3)	105 (10.2)	251 (34.6)	<0.001
1-yr mortality	647 (37.9)	270 (27.1)	377 (53.2)	<0.001
ICU stay for survivors at ICU discharge, days	12 (7–21)	11 (7–19)	14 (9–25)	<0.001
Hospital stay for survivors at hospital discharge, days	25 (15–42)	24 (14–40)	27 (17–48)	<0.001
Introduction of RRT	375 (21.3)	25 (2.4)	350 (48.1)	<0.001

Values are expressed as median (interquartile range) or as N (%), as appropriate. Group comparisons of continuous variables were performed using the Kruskal–Wallis test. Categorical data were compared using Pearson’s chi-square test. A two-sided P<0.05 was considered for significance.

Abbreviations: BMI, body mass index; SOFA, sequential organ failure assessment; SpO₂, peripheral oxygen saturation; LVEF, left ventricular ejection fraction; PaO₂, partial pressure of oxygen in arterial blood; PaCO₂, partial pressure of carbon dioxide in the arterial blood; BNP, brain natriuretic peptide; NT-proBNP, N-terminal pro brain natriuretic peptide; PCT, procalcitonin; RRT, renal replacement therapy.

Primary outcome

Seven hundred twenty-eight (41.4%) patients developed AKI during their ICU stay. uNGAL, uLFABP, and uCysC concentrations on admission were higher in patients who developed AKI during ICU stay than in patients who did not. In multivariate analysis, the adjusted odds ratios (ORs) were 2.76 (95% confidence interval [CI], 1.98–3.85; P<0.001), 1.63 (95% CI, 1.19–2.23; P=0.002), and 1.18 (95% CI, 0.86–1.64; P=0.300) for uNGAL, uLFABP, and uCysC concentrations on admission, respectively (Table 2).

Table 2

Biomarkers concentrations for outcomes

Biomarker*	All patients (N=1,760)	Occurrence of AKI				Status at 28 days				Status at 1 yr
		No AKI (N=1,031)	AKI (N=728)	P		Survivor (N=1,398)	Non-survivor (N=356)	P		Survivor (N=502)	Non-survivor (N=107)	P
uNGAL, ng/mL	147.6 (39.9–827.7)	62.6 (27.6–205.4)	697.8 (173.1–1,500)	<0.001		111.6 (34.0–581.5)	505.8 (103.5–1,500)	<0.001		39.8 (20.4–128.6)	96.1 (35.2–254.6)	<0.001
uLFABP, ng/mL	32.4 (10.5–96.0)	19.6 (7.4–60.1)	61.8 (22.6–207.1)	<0.001		28.1 (9.2–83.9)	61.3 (18.7–183.3)	<0.001		9.2 (3.5–25.8)	11.8 (5.3–45.7)	0.030
uCysC, mg/L	0.33 (0.12–2.05)	0.20 (0.10–1.03)	0.78 (0.19–3.39)	<0.001		0.30 (0.12–1.84)	0.52 (0.13–2.76)	0.002		0.11 (0.06–0.23)	0.14 (0.08–0.46)	0.029
pNGAL, ng/mL	197.0 (94.0–464.5)	125.0 (72.0–224.5)	451.0 (233.0–839.0)	<0.001		170.0 (85.0–394.0)	380.0 (176.0–799.3)	<0.001		106.0 (68.0–172.3)	180.0 (100.0–317.0)	<0.001
pCysC, mg/L	1.28 (0.86–2.02)	1.02 (0.75–1.35)	2.02 (1.41–2.88)	<0.001		1.19 (0.82–1.80)	1.80 (1.21–2.70)	<0.001		1.22 (0.91–1.84)	1.25 (0.90–1.90)	0.880

Values are expressed as median (interquartile range). Group comparisons were performed using the Kruskal–Wallis test. A two-sided P<0.05 was considered for significance.

*Biomarker concentrations for the columns “all patients,” “occurrence of AKI,” and “status at 28 days” were collected on admission and those for the column “status at 1 yr” were collected at discharge from the intensive care unit.

Abbreviations: AKI, acute kidney injury; uNGAL, urinary neutrophil gelatinase-associated lipocalin; uLFABP, urinary liver fatty acid-binding protein; uCysC, urinary cystatin C; pNGAL, plasma neutrophil gelatinase-associated lipocalin; pCysC, plasma cystatin C; AKI, acute kidney injury.

In patients who developed AKI during ICU stay, the higher the uNGAL and uLFABP concentrations on admission, the higher the severity of AKI as assessed by the KDIGO criteria (Fig. 2). Biomarker concentrations according to the eGFR category are summarized in Supplemental Data Fig. S1. For the prediction of AKI during ICU stay in severely ill patients, the AUC values of uNGAL, uLFABP, and uCysC concentrations on admission were 0.80 (95% CI, 0.77–0.81), 0.70 (95% CI, 0.67–0.72), and 0.64 (95% CI, 0.62–0.67), respectively (Table 3).

Table 3

Predictive performances of biomarkers for AKI and mortality

Biomarker	AUC (95% CI)	Cutoff value	Sensitivity	Specificity	P*	P†
Predictive value of each biomarker on admission for the occurrence of AKI during ICU stay
uLFABP	0.70 (0.67–0.72)	38.6 ng/mL	0.63	0.66	-	-
uNGAL	0.79 (0.76–0.81)	214.1 ng/mL	0.74	0.78	<0.001	-
uCysC	0.64 (0.62–0.67)	0.40 mg/L	0.61	0.64	<0.001	-
pNGAL	0.81 (0.79–0.83)	244.5 ng/mL	0.74	0.78	<0.001	0.004
pCysC	0.82 (0.80–0.84)	1.43 mg/L	0.74	0.79	<0.001	<0.001
Combination
uLFABP + uNGAL	0.80 (0.77–0.81)	-	0.74	0.73	<0.001	-
uLFABP + uCysC	0.70 (0.67–0.72)	-	0.69	0.64	0.810	-
uLFABP + uNGAL + uCysC	0.79 (0.77–0.81)	-	0.74	0.74	<0.001	-
Predictive value of each biomarker on admission for 28-day mortality
uLFABP	0.63 (0.59–0.66)	42.8 ng/mL	0.58	0.61	-	-
uNGAL	0.66 (0.63–0.70)	247.0 ng/mL	0.62	0.64	0.007	-
uCysC	0.55 (0.52-0.59)	0.37 mg/L	0.56	0.54	<0.001	-
pNGAL	0.68 (0.65–0.71)	265.5 ng/mL	0.63	0.65	0.001	0.160
pCysC	0.68 (0.64–0.71)	1.44 mg/L	0.67	0.65	0.007	<0.001
Combination
uLFABP + uNGAL	0.67 (0.64–0.70)	-	0.62	0.64	0.003	-
uLFABP + uCysC	0.59 (0.56–0.63)	-	0.58	0.58	0.004	-
uLFABP + uNGAL + uCysC	0.68 (0.65–0.71)	-	0.62	0.63	0.002	-
Predictive value of each biomarker at discharge for 1-yr mortality
uLFABP	0.57 (0.51–0.63)	8.5 ng/mL	0.63	0.49	-	-
uNGAL	0.64 (0.58–0.69)	61.3 ng/mL	0.64	0.61	0.014	-
uCysC	0.57 (0.51–0.63)	0.13 mg/L	0.56	0.55	0.990	-
pNGAL	0.66 (0.59–0.72)	163.5 ng/mL	0.58	0.72	0.014	0.520
pCysC	0.50 (0.44–0.57)	1.24 mg/L	0.52	0.51	0.130	0.160
Combination
uLFABP + uNGAL	0.61 (0.55–0.67)	-	0.53	0.66	0.038	-
uLFABP + uCysC	0.57 (0.51–0.63)	-	0.63	0.47	0.930	-
uLFABP + uNGAL + uCysC	0.61 (0.55–0.67)	-	0.55	0.63	0.050	-

*vs. AUC of uLFABP using the DeLong test.

^†vs. AUC of the same biomarker in urine using the DeLong test.

Abbreviations: AUC, area under the ROC curve; ICU, intensive care unit; AKI, acute kidney injury; uLFABP, urinary liver fatty acid-binding protein; uNGAL, urinary neutrophil gelatinase-associated lipocalin; uCysC, urinary cystatin C; pNGAL, plasma neutrophil gelatinase-associated lipocalin; pCysC, plasma cystatin C.

Figure 2. Boxplots of biomarker concentrations on ICU admission according to AKI severity as assessed by the KDIGO criteria. (A) uNGAL, (B) uLFABP, and (C) uCysC. Y-axis values are logarithmically spaced.
Abbreviations: uNGAL, urinary neutrophil gelatinase-associated lipocalin; uLFABP, urinary liver fatty acid-binding protein; uCysC, urinary cystatin C; AKI, acute kidney injury; KDIGO, Kidney Disease: Improving Global Outcomes; ICU, intensive care unit.

Secondary outcomes

Three hundred and fifty-six (20.3%) patients had died by day 28 after ICU admission. uNGAL, uLFABP, and uCysC concentrations on admission were higher in 28-day non-survivors than in survivors. The adjusted ORs for uNGAL, uLFABP, and uCysC concentrations on admission for 28-day mortality were 1.48 (95% CI, 1.03–2.14; P=0.035), 1.49 (95% CI, 1.06–2.09; P=0.021), and 1.17 (95% CI, 0.83–1.64; P=0.379), respectively. Six hundred forty-seven (37.9%) patients died during the year following ICU discharge. uNGAL, uLFABP, and uCysC concentrations at ICU discharge were higher in 1-yr non-survivors than in survivors (Table 2).

The AUC values of uNGAL, uLFABP, and uCysC concentrations on admission for 28-day mortality were 0.66 (95% CI, 0.63–0.70), 0.63 (95% CI, 0.59–0.66), and 0.55 (95% CI, 0.52–0.59) respectively. The adjusted ORs for uNGAL, uLFABP, and uCysC concentrations at discharge for 1-yr mortality were 1.99 (95% CI, 1.12–3.58; P=0.020), 1.39 (95% CI, 0.90–2.87; P=0.114), and 1.30 (95% CI, 0.73–2.32; P=0.373), respectively. The AUC values for uNGAL, uLFABP, and uCysC concentrations at discharge for the prediction of 1-yr mortality were 0.64 (95% CI, 0.58–0.69), 0.57 (95% CI, 0.51–0.63), and 0.57 (95% CI, 0.51–0.63), respectively (Table 3).

Fig. 3A shows that 1-yr mortality in patients with high uNGAL concentrations (>cutoff value) at ICU discharge was higher than that in patients with low uNGAL concentrations at discharge, with a hazard ratio (HR) of 2.54 (95% CI, 1.72–3.77; P<0.001). Additionally, we observed a higher 1-yr mortality in patients with high uLFABP or uCysC concentrations at ICU discharge than in patients with low concentrations of these biomarkers (HR 1.54 [95% CI 1.04–2.28], P=0.030 for uLFABP and HR 1.52 [95% CI 1.04–2.23], P=0.031 for uCysC, respectively) (Fig. 3B and 3C).

Figure 3. Kaplan–Meyer curves of biomarker concentrations at ICU discharge for 1-yr mortality. (A) uNGAL, (B) uLFABP, and (C) uCysC.
Abbreviations: uNGAL, urinary neutrophil gelatinase-associated lipocalin; uLFABP, urinary liver fatty acid-binding protein; uCysC, urinary cystatin C; ICU, intensive care unit.

In this ancillary FROG-ICU cohort study, we found that uNGAL, uLFABP, and uCysC concentrations in the initial phase of ICU management were associated with poor outcomes in terms of AKI occurrence during ICU stay and short-term mortality. uNGAL and uLFABP concentrations in the initial phase of ICU management were associated with the severity of AKI as assessed by the KDIGO criteria. Further, uNGAL, uLFABP, and uCysC concentrations at ICU discharge were associated with an increased risk of 1-yr mortality. However, the predictive performances of these biomarkers for AKI and short- and long-term mortality were low. To our knowledge, this study is one of the first to compare the predictive performance of uNGAL, uLFABP, and uCysC concentrations on ICU admission and at discharge.

The ability of uLFABP concentrations to predict AKI occurrence has been demonstrated in previous studies [22, 23]. However, in our international prospective cohort study, we found that the predictive ability of the uLFABP concentration measured early in the ICU in severely ill patients was significantly lower than that of uNGAL and uCysC concentrations, and the performance of pNGAL and pCysC concentrations for AKI prediction was superior to that of the urinary biomarkers, which is consistent with previous results [11, 24]. Siew, et al. [25] found that uNGAL and uLFABP concentrations were associated with patient outcomes, but the predictive performance of uNGAL, uLFABP, and uCysC concentrations for AKI was limited. In a meta-analysis published in 2015 [24], the authors found that the AKI-predictive performance of the uLFABP concentration was higher than that of the uCysC concentration, and that of the uNGAL concentration was similar to that of the uLFABP concentration, with AUC values of 0.72 (95% CI, 0.66–0.79; P<0.001) for uNGAL and 0.72 (96% CI, 0.60–0.85; P<0.001) for uLFABP. These studies did not reveal differences among the three urinary biomarkers, but their results were consistent with ours in that each biomarker showed low predictive performance, despite being associated with the occurrence of AKI.

The most recent meta-analysis [15] reported cutoff values for uLFABP and uNGAL of 28–90 ng/mL and 25–74 ng/mL, respectively, for the prediction of AKI in severely ill patients. In a multicenter prospective study, Nickolas, et al. [26] suggested a cutoff value of 104 ng/mL for uNGAL to predict AKI, with an AUC value of 0.81 (95% CI, 0.76–0.86), sensitivity of 0.68, and specificity of 0.81. These results are consistent with our findings regarding uLFABP. However, we did not find the same cutoff value for uNGAL for AKI prediction, which can be explained by the disease heterogeneity in the patients included in the FROG-ICU cohort. The association between uCysC concentrations and AKI has been poorly studied, and no cutoff value is currently available, although few studies have highlighted differences in uCysC concentrations between survivors and non-survivors [23, 26^-30].

We found that uLFABP concentrations at discharge were higher in non-survivors than in survivors at 1 yr after discharge. However, their ability to predict 1-yr mortality was poor, making this biomarker unreliable for predicting short- and long-term outcomes in severely ill patients.

Our study had several limitations. First, biological measurements of uNGAL, uLFABP, and uCysC were lacking in 401 (24.5%), 539 (33.0%), and 540 (33.0%) ICU survivors, respectively. Data from these patients were removed from the analysis of secondary outcomes, potentially affecting the accuracy of our results. Second, urinary output was not recorded daily in the FROG-ICU study, probably limiting AKI diagnosis according to the KDIGO criteria. Further studies including urinary output daily monitoring should be conducted to early detect the occurrence of AKI. Third, urinary creatinine (uCr) measurements were lacking in the FROG-ICU cohort; therefore, uNGAL and uLFABP concentrations were not adjusted by uCr, although all urinary biomarkers were measured in spot urine samples. Pan, et al. [31] reported that uLFABP concentrations were associated with prognosis (90-day mortality and induction of renal replacement therapy) regardless of adjustment by uCr. However, the lack of adjustment of the random urinary biomarkers by uCr may limit the clinical utility of our results. This study was one of the first to analyze uNGAL, uLFABP, and uCysC concentrations on admission and at discharge; therefore, our results need to be confirmed. Fourth, the FROG-ICU study was not initially designed for these primary and secondary endpoints. Our results should therefore be confirmed in larger prospective cohorts. Fifth, urinary biomarkers were measured on ICU admission and at discharge, but not after discharge. Furthermore, daily or weekly uNGAL, uLFABP, and uCysC concentrations may provide valuable information, and the dynamics of these biomarkers may predict AKI occurrence during ICU stay. Future studies should perform several measurements of uNGAL, uLFABP, and uCysC during ICU stay to evaluate the dynamics of these biomarkers.

The design of the FROG-ICU study and the long-term follow-up of survivors were the major strengths of this study. The originality and strength of our study lie in four major points. First, uNGAL, uLFABP, and uCysC assays are simple and non-invasive and can be performed at the patient’s bedside (especially on the day of discharge when patients no longer have an access route). Second, these three urinary assays were performed on ICU admission (i.e., in the early phase during ICU management) and at discharge. Most studies assessing the prognostic value of a biomarker include measurements in the early phase in the ICU, but not at discharge. Moreover, few studies include a 1-yr follow-up after ICU and hospital discharge as does the FROG-ICU study. Third, FROG-ICU is an international, prospective, multicenter study including more than 2,000 ICU patients with various pathologies and causes of admission. This strength allows us to assess several biomarkers in different pathologies and syndromes, such as AKI. Fourth, these biomarkers have been assessed individually, but, to our knowledge, this study was the first to assess the predictive performances of uNGAL, uLFABP, and uCysC concentrations individually and in combination for both short- and long-term outcomes in the same cohort.

In conclusion, our study showed that uNGAL, uLFABP, and uCysC concentrations in the initial phase of ICU management were associated with poor short-term outcomes. High uNGAL, uLFABP, and uCysC concentrations at ICU discharge were associated with an increase in 1-yr mortality. However, the predictive performance of uNGAL, uLFABP, and uCysC concentrations for short- and long-term mortality was limited. Of note, our study showed that the predictive performance of uLFABP and uCysC was limited and that the combination of uNGAL, uLFABP, and uCysC did not improve the global predictive performance nor had clinical relevance. The performance of uNGAL in predicting AKI is interesting but needs to be confirmed in large cohort studies.

Supplementary materials can be found via https://doi.org/10.3343/alm.2023.0083

The authors are grateful to Marie-Céline Fournier who coordinated the organizational aspects of the study. They also thank the Centre de Recherche Clinique (CRC) of Lariboisière University Hospital for support and the investigators of the FROG-ICU study.

Asakage A contributed to conceptualization, methodology, formal analysis, writing—original draft, and visualization. Ishihara S contributed to writing—review and editing. Boutin L, Dépret F, and Gayat E contributed to conceptualization and writing—review and editing. Sugaya T contributed to resources and data curation. Sato N contributed to resources, data curation, and project administration. Mebazaa A contributed to conceptualization, project administration, funding acquisition, and supervision. Deniau B contributed to conceptualization, methodology, writing—review and editing, visualization, and supervision. All authors reviewed and approved the final version of the manuscript.

Prof. Alexandre Mebazaa reports personal fees from Novartis, Orion, Roche, Servier, Sanofi, Otsuka, Philips; grants and personal fees from Adrenomed and Abbott; and grants from 4TEEN4. He also owns fewer than 3,000 euros shares of S-Form Pharma Company. Dr. Naoki Sato reports honoraria and consulting fees from Otsuka, Novartis, BMS, Bayer, Terumo, Boehringer-Ingelheim, Daiichi-Sankyo, Ono, AstraZeneca, Taisho, and Kowa. Dr. Benjamin Deniau was invited to a meeting in Henningsdorf, Germany by 4TEEN4 Pharmaceuticals GmbH. The other authors declare no potential conflicts of interest with respect to the research authorship and/or publication of this article.

The FROG-ICU study was funded by the Programme Hospitalier de la Recherche Clinique (AON 10-216) and by a research grant from the Société Française d’Anesthésie-Réanimation. Abbot provided unrestricted free kits to Assistance Publique-Hôpitaux de Paris to conduct biomarker analyses.