Circulating retinol binding protein 4 in critically ill patients before specific treatment: prognostic impact and correlation with organ function, metabolism and inflammation

From the medical ICU at the University Hospital Aachen, Germany, a tertiary care university hospital, patients were prospectively enrolled at the time of ICU admission. Patients who were expected to stay < 72 hours at the ICU (e.g., postinterventional observation, acute intoxication) were not asked to participate in this study [16]. Written informed consent was obtained from each participant or his or her spouse, and the study was approved by the local ethics committee (ethics committee of the University Hospital Aachen, RWTH-University, Aachen, Germany, reference number EK 150/06). Patient data, blood samples and clinical information were collected prospectively. Owing to lack of sufficient material, 123 of initially 170 consecutively recruited patients were included in the current study. Median duration of stay at the ICU was 8 days (range, 1-137 days), and median duration of stay at the hospital was 26.5 days (range, 2-151 days). The clinical course of the patients after discharge was followed by contacting the patients and/or their primary care physician up to 3 years after entry into this study (median observation time, 591 days; range, 29-884 days). Patients were categorized as having severe sepsis (acute organ dysfunction secondary to infection) and septic shock (severe sepsis plus hypotension not reversed with fluid resuscitation), hereafter called sepsis, according to the definitions in current guidelines [17], or as being nonsepsis. Sepsis patients were subdivided into cohorts of sepsis of pulmonary and nonpulmonary origin on the basis of the clinical and radiological classification performed upon admission.

Forty-two healthy, nondiabetic blood donors (30 male, 12 female; median age, 53 years; range, 24-68 years) who had normal blood counts, normal liver enzymes and normal C-reactive protein levels served as controls.

The ICU patients were compared by age, sex and severity of illness using the score on the Acute Physiology and Chronic Health Evaluation (APACHE) II obtained at admission and the simplified acute physiology score (SAPS) 2 [16]. Laboratory parameters were routinely assessed at admission, recorded and analyzed.

Peripheral venous blood samples were obtained at admission before therapeutic intervention, immediately placed on ice, centrifuged and stored at -80°C. All measurements were performed in a blinded fashion by the same investigator. RBP4 serum concentrations were measured using an enzyme-linked immunosorbent assay according to the manufacturer's instructions (Immundiagnostik AG, Bensheim, Germany) [18]. Serum resistin, ghrelin, adiponectin and C-peptide were measured as described previously [19‐21]. Insulin sensitivity was assessed by the homeostasis model assessment (HOMA) index using the HOMA Calculator V. 2.2.2 from the University of Oxford, UK [20, 22].

Analyses were performed with SPSS 12.0 software (SPSS, Munich, Germany). Owing to skewed distributions of most variables, the median and range are given. Differences between two groups were assessed by the Mann-Whitney U test or between more than two groups by Kruskal-Wallis analysis of variance and the Mann-Whitney U test for post hoc analysis [18]. Comparisons between subgroups are illustrated with boxplot graphics, where the black bold line indicates the median per group, the box represents 50% of the values and horizontal lines show minimum and maximum values of the calculated nonoutlier values; asterisks and open circles indicate outlier values. The correlations between variables were analyzed using the Spearman correlation tests. Values of P < 0.05 were considered statistically significant. The prognostic value of the variables was tested by univariate and multivariate analyses in the Cox regression model. After significant results from the uni- and multivariate Cox regression analyses, Kaplan-Meier curves and log-rank test calculations were performed subsequently for different cutoff values for RBP4 (9, 10, 11, 12, 13, 14 and 15 mg/L). The threshold of 12 mg/L RBP4 yielded the highest log-rank values. Kaplan-Meier curves were plotted to display the impact on survival [23].

The median serum concentration of RBP4 in healthy controls was 28.1 mg/L (range, 18.8-255.6 mg/L, 90% interval, 20.0-53.8 mg/L) as anticipated from previous studies using the same assay [13, 18]. In our cohort of nondiabetic healthy controls, serum RBP4 did not correlate with the body mass index (BMI), and there was no difference between male and female volunteers. Critically ill ICU patients had significantly reduced serum RBP4 concentrations compared with healthy controls (median, 16.1 vs. 28.1 mg/L, P < 0.001; Figure 1a, Table 1). Again, male and female patients did not differ in serum RBP4.

As inflammatory disorders could considerably impact serum RBP4 concentrations, we next compared patients with sepsis (n = 85) with patients without sepsis (n = 38). In the sepsis group, the focus of infection was either pulmonary (n = 45), abdominal (n = 19) or other (n = 21, e.g., catheter-associated, urogenital or unknown). In the nonsepsis group, the etiology of critical illness was decompensated liver cirrhosis (n = 11), cardiopulmonary (n = 16) or other diseases (n = 11). Interestingly, serum RBP4 did not differ significantly between patients with sepsis (median 15.2 mg/L) and without sepsis (median 20.0 mg/L; Figure 1b). With respect to several other clinical parameters (APACHE II score, mechanical ventilation, vasopressor demand), patients with or without sepsis did not differ, but patients with sepsis had a higher ICU and overall mortality as well as significantly longer ventilation duration (Table 2).

A recent study reported reduced RBP4 serum levels in patients with sepsis of pulmonary origin [15]. To investigate the regulation of RBP4 in patients with different etiologies of sepsis and septic shock, we performed extensive subgroup analyses. However, there was no difference in RBP4 serum concentrations between patients with sepsis of pulmonary or nonpulmonary origin (Figure 2a). Furthermore, patients with pulmonary and nonpulmonary sepsis did not differ in baseline clinical parameters or in inflammatory markers such as CRP (Figure 2b, and data not shown).

Interestingly, ICU patients with liver cirrhosis displayed the lowest RBP4 serum levels among all subgroups (Figure 2a). Liver function could be identified as a strong predictor of serum RBP4, as RBP4 levels directly correlated with parameters indicating the liver's biosynthetic capacity, namely, albumin (r = 0.404, P < 0.001; Figure 3a), total protein (r = 0.272, P = 0.003), pseudocholinesterase activity (r = 0.491, P < 0.001; Figure 3b), IGF-1 concentrations (r = 0.403, P < 0.001), prothrombin time (r = 0.524, P < 0.001) and inversely with bilirubin (e.g., conjugated bilirubin; r = -0.434, P < 0.001). Of note, RBP4 serum concentrations remained significantly (P < 0.001, Mann-Whitney U test) elevated in controls (n = 42, median 28.1; range, 18.8-255.6 mg/L) compared to ICU patients without cirrhosis (n = 112, median 16.5, range 0.3-147.6 mg/L).

Serum RBP4 concentrations were also correlated with markers of renal failure, specifically increasing creatinine (r = 0.493, P < 0.001; Figure 3c), urea (r = 0.389, P < 0.001), cystatin C (r = 0.381, P < 0.001) and decreasing glomerular filtration rate (GFR; r = -0.526, P < 0.001).

Although there was no significant difference between septic and nonseptic patients, systemic inflammatory markers were associated with serum RBP4 concentrations. Classical biomarkers of systemic inflammatory responses were inversely associated with serum RBP4, such as C-reactive protein (r = -0.204, P = 0.025) or interleukin-6 levels (r = -0.272, P = 0.019). RBP4 thereby displayed characteristics of a negative acute phase reactant [24]. However, this association was observed only in sepsis patients.

Initial studies linked elevated serum RBP4 to overt or impending insulin resistance in lean, obese and type 2 diabetic subjects [13]. To understand potential pathophysiological consequences of reduced RBP4 levels in ICU patients, we analyzed associations between serum RBP4 and insulin resistance as well as diabetes. Patients with a preexisting type 2 diabetes (n = 42) did not differ from patients without diabetes (n = 81; Figure 4a) in serum RBP4 concentrations. Consequently, serum RBP4 was not correlated to the glycosylated hemoglobin (HbA1c) levels in ICU patients either. Furthermore, serum RBP4 did not correlate with the patients' BMI (data not shown), and obese patients (defined as BMI > 30, n = 30) did not show higher serum RBP4 than patients without severe obesity (BMI ≤ 30, n = 75; Figure 4b).

Although serum RBP4 did not correlate with serum glucose on admission to the ICU, serum RBP4 correlated with the C-peptide (r = 0.305, P = 0.001; Figure 4c), with the insulin resistance as calculated by the HOMA index (HOMA-IR; r = 0.248, P = 0.009; HOMA-S, r = -0.248, P = 0.009; Figure 4d). This association of serum RBP4 to endogenous insulin secretion and insulin resistance was in sharp contrast to other adipocytokines. RBP4 serum levels did not correlate with adiponectin, resistin or ghrelin serum concentrations, nor did adiponectin, resistin or ghrelin correlate with insulin resistance (data not shown).

We next analyzed whether serum RBP4 concentrations upon ICU admission may predict clinical outcome and mortality. Interestingly, patients who died during ICU hospitalization (33/123) displayed significantly lower serum RBP4 concentrations than the patients who were surviving (Figure 5a). The beneficial prognostic impact of high RBP4 levels on ICU survival was confirmed by Cox regression analysis (P = 0.031). By multivariate Cox regression analysis, the effect of RBP4 on ICU survival was independent of age, BMI, renal function (creatinine, GFR) and CRP, but indistinguishable from liver function. Two models were tested in multivariate Cox regression analyses, using ICU outcome (survival/death) as the endpoint. The first model included RBP4, age, BMI, creatinine, GFR (cystatin C-based) and CRP, and RBP4 was the only parameter that remained independently significant (P = 0.041) to predict ICU outcome in this combination. The second model combined RBP4 and markers of hepatic dysfunction (albumin, international normalized ratio, pseudocholinesterase). In this model, pseudocholinesterase remained the only independently significant (P < 0.001) parameter predicting ICU survival. If all parameters were combined, pseudocholinesterase was the dominant predictor for ICU outcome.

Using Kaplan-Meier survival curves with a cutoff for serum RBP4 concentrations of 12 mg/L (Figure 5b), high RBP4 levels were a significant positive prognostic marker for ICU survival (log-rank test, 10.96; P = 0.0009). On the other hand, serum RBP4 levels did not predict the overall survival of the ICU patients in the approximately 3-year follow-up (Figure 5c), indicating that the prognostic relevance and potentially also the pathogenic involvement of RBP4 is restricted to the acute illness.