Predictive value of iron parameters in neurocritically ill patients

Abstract Background Iron, an essential mineral for human body, has the potential to cause toxicity at high levels. Previous studies have shown inconsistent predictive value of iron parameters in critically ill patients. Thus, we aimed to evaluate the performance of iron parameters in outcome prediction of neurocritically ill patients. Methods Retrospective data were collected from patients admitted to the neurocritical care unit (NCU) of a tertiary teaching hospital between August 2016 and January 2017. The iron parameters were obtained at NCU admission. Primary endpoints were short‐term (30‐day) mortality and long‐term (6‐month) poor outcome, with the latter defined as modified Rankin Scale of 4–6. The predictive value of variables was determined with univariate and multivariate logistic analysis. A further subanalysis was conducted in patients stratified by the level of estimated glomerular filtration rate (eGFR). Results Of 103 eligible patients, the etiology included stroke (58.2%, N = 60), central nervous system infection (13.6%, N = 14), and other neurologic disorders (28.2%, N = 29). The correlation analysis showed that the increase in ferritin, as well as the reduction in transferrin and total iron‐binding capacity, had strong correlation with C‐reactive protein, procalcitonin, duration of NCU stay, Acute Physiology and Chronic Health Evaluation II score, and Sequential Organ Failure Assessment score. In a further subanalysis of 75 patients with eGFR ≥ 60 ml/min/1.73 m2, twelve (16.0%) patients died within 30 days and 39 (52.0%) patients achieved good follow‐up outcome data. In the multivariate logistic regression analysis, we identified baseline ferritin level as an independent predictor of short‐term mortality (OR: 1.002; 95% CI: 1.000–1.003; p = 0.008) and long‐term functional outcome (OR: 1.002; 95% CI: 1.000–1.004; p = 0.031). Conclusions Serum ferritin level at admission could be used as an independent predictor of short‐term mortality and long‐term functional outcome in neurocritically ill patients with eGFR ≥ 60 ml/min/1.73 m2.


| INTRODUC TI ON
Iron has many different roles in the body, acting in carrying oxygen, delivering electron, and catalyzing many biochemistry reactions.
After being absorbed from the microvilli of enterocytes, most of the iron in the body keeps a sequestered condition by binding with transferrin in the circulation and is stored in the form of ferritin in the tissues (Andrews & Schmidt, 2007). In contrast, the non-transferrin-bound iron is toxic because it can generate free radicals via participating in the Fenton/Haber-Weiss reaction. These reactive oxidative species could cause lipid peroxidation and bring damages to proteins and DNA (Brissot, Ropert, Le Lan, & Loreal, 2012;Koskenkorva-Frank, Weiss, Koppenol, & Burckhardt, 2013).
Under normal conditions, there is a balance of iron metabolism that preserves the biological function of iron while preventing an excess that would cause oxidative stress. Under pathological conditions, however, the iron metabolism shifts and the alteration could cause it to become a defense mechanism. The processes of iron metabolism involve a number of specific proteins (Malyszko, Malyszko, Pawlak, & Mysliwiec, 2006). Hepcidin, a peptide mainly synthesized by the liver, is regarded as the master regulator modulated in response to hypoxia, iron deficiency, anemia, or inflammation. Iron metabolism disturbance occurs frequently in intensive care unit (ICU) patients, whereby some of the iron parameters reported to be useful in predicting the prognosis of these patients. Nevertheless, these studies are mainly performed in surgical or general ICU. Moreover, which iron parameters could be used to predict reliably with prognosis of critical illness is controversial (Bobbio-Pallavicini et al., 1989;Leaf, Rajapurkar, Lele, Mukhopadhyay, & Waikar, 2014;Tacke et al., 2016).
In terms of neurologic disorders, several studies (Davalos et al., 1994;Millan et al., 2007;Millerot et al., 2005;Simon et al., 2015;Zuliani et al., 2006) have suggested that baseline ferritin was related to poor outcome and may represent a marker of disease severity, especially in stroke and brain trauma. However, whether iron parameters, especially ferritin, could be used to predict the prognosis in neurocritically ill patients has not been proved yet.
Here, we performed this study to evaluate the predictive value of iron parameters in neurocritically ill subjects.

| Subjects
We retrospectively collected data from a prospectively designed cohort of consecutive patients admitted to a neurocritical care unit (NCU) of a tertiary university-affiliated academic hospital (Nanfang Hospital, Southern Medical University, Guangzhou, China), from August 2016 to January 2017. The inclusion criteria were Glasgow Coma Scale (GCS) (Teasdale & Jennett, 1974) ≤12 and/or admission Acute Physiology and Chronic Health Evaluation (APACHE) II score (Knaus, Draper, Wagner, & Zimmerman, 1985) >15; and/or demand for intensive care or life-support measures. Patients younger than 18 years old, with iron-related disorders in history, pregnant, or required intensive care for <72 hr were excluded. The critical stroke patients with severe neurologic deficits, without altered consciousness, and being classified as "demand for intensive care or life-support measures," were included in the study. We excluded patients "who required intensive care for <72 hr" because our NCU and our stroke unit worked as the same unit. The patient group with acute mild to moderate intracranial hemorrhage within 24 hr and superacute ischemic stroke with thrombolysis and/or intravascular treatment that required transient close monitoring but transferred out of NCU within 72 hr was eliminated from the neurocritically ill patient group.
Several literatures have reported that iron metabolism may be affected by the renal function and renal function is connected to mortality in ICU patients (Beier et al., 2011;Cartin-Ceba, Afessa, & Gajic, 2007;Malyszko et al., 2006), indicating that the renal function might have influence on outcome in this study. Therefore, the enrolled patients were divided into two subgroups for further analysis, based on estimated glomerular filtration rate (eGFR) (Levey et al., 2009), with eGFR ≥ 60 ml/min/1.73 m 2 defined as normal (National Kidney Foundation, 2002).

| Data collection
Electronic medical records were carefully reviewed to collect the patient information of demographics, diagnoses, vital signs, GCS scores, APACHE II scores, Sequential Organ Failure Assessment (SOFA) scores (Vincent et al., 1998), laboratory values, length of NCU stay, and duration of mechanical ventilation. Laboratory data of iron parameters included serum iron, ferritin, transferrin, transferrin saturation (TS), total iron-binding capacity (TIBC), and unsaturated iron-binding capacity (UIBC). All data mentioned above were referred to the patients' baseline information within 24 hr after admission. Nobody had prior transfusion. GCS scores were extracted from the first neurological examination at NCU admission. The total scores of APACHE II and SOFA were obtained according to the corresponding parameters within the first 24 hr of NCU admission.
Primary endpoints were short-term (30-day) all-cause mortality and long-term (6-month) poor outcome, with the latter defined as modified Rankin Scale (mRS) of 4-6. The outcome information was acquired in a medical follow-up system, in which all patients admitted to NCU were followed up through face-to-face or telephone interviews by a trained personnel blinded to the present study.

| Statistical analysis
Continuous data were presented as mean ± standard deviation (SD) or median (25%-75% interquartile range) and compared by Student's t test or Mann-Whitney U test, as appropriate. Differences in proportions among categorical data were assessed using chi-squared tests and Fisher's exact tests for multiple groups. The prognostic value of iron parameters was first assessed by univariate analysis. Then, the significant variables were further included in multivariate models for adjustment.
The 95% confident intervals reported for the logistic regression odds ratios were calculated by the maximum-likelihood estimation (forward selection). Correlations between variables were determined with the Spearman's rank correlation test. p < 0.05 was considered statistically significant. All statistical analyses were performed using SPSS, version 20.0 (SPSS, Chicago, IL).

| Ethic statement
The study proposal was approved by the Medical Ethics Committee of Nanfang Hospital. Informed consents were signed by all the patients or their surrogates before data collection. For short-term mortality, higher baseline ferritin (p = 0.017) and lower baseline transferrin (p = 0.018) and TIBC (p = 0.037) showed statistical significance. For long-term prognosis, baseline transferrin was the only iron parameter showed statistical significance (p = 0.043) ( Table 1). In multivariate logistic regression model, none of the iron parameters showed statistical significance (not shown).

| Overall analysis
The correlation analysis showed that the increase in ferritin, as well as the reduction in transferrin and TIBC, was paralleled with the increase in duration of NCU stay, APACHE II scores, SOFA scores, C-reactive protein, and procalcitonin levels ( Table 2).

| Subgroup analysis
To eliminate the confounding of renal function, we performed subgroup analysis in patients with eGFR ≥ 60 ml/min/1.73 m 2 (N = 75).

| 30-day mortality of patients with eGFR ≥ 60 ml/min/1.73 m 2
Compared with short-term survivors (63 patients), the victims were more likely to have underlying diabetes, higher APACHE II and SOFA scores at admission, whereas there were no differences in any iron parameters (Supporting information Table S1). In multivariate logistic analysis, ferritin was independently associated with 30-day mortality (OR: 1.002; 95% CI: 1.000-1.003; p = 0.008) ( Table 3).

| 6-month functional outcome of patients with eGFR ≥ 60 ml/min/1.73 m 2
Increased APACHE II score, SOFA score, length of NCU stay, and serum C-reactive protein levels were observed in patients with poor functional outcome at 6 months compared to those with good functional outcome (39 patients), while there were no differences in any iron parameters (Supporting information Table S1). In multivariate logistic analysis, ferritin was independently associated with 6-month functional outcome (OR: 1.002; 95% CI: 1.000-1.004; p = 0.031) (Table 3).

| D ISCUSS I ON
In this study, we verified serum ferritin as a predictor of clinical outcome in a cohort of neurocritically ill patients and found that serum ferritin was independently associated with 30-day mortality and 6month poor functional outcome in neurocritically ill patients with eGFR ≥ 60 ml/min/1.73 m 2 .
Our results showed that neurocritically ill patients who died  In subgroup analysis, we demonstrated serum ferritin was an independent predictor of short-term mortality and long-term poor outcome, in agreement with previous researches in ICU patients (Leaf et al., 2014;Munoz et al., 2005;Simon et al., 2015;Tacke et al., 2016).
And in neurologic disorders, Millan et al. (2007) studied 134 consecutive patients with acute ischemic stroke and found that ferritin levels before thrombolytic treatment were higher in patients with poor outcome (modified Rankin scale, mRS > 2) at 90 days. It was suspected that increased ferritin could be in part the result of a neuroprotective mechanism with the aim of sequestering toxic-free iron in the ischemic brain. Another study performed in patients with intracerebral hemorrhage also demonstrated that serum ferritin at baseline was markedly higher and iron as well as transferrin was lower in patients with poor outcome (mRS ≥ 3) at 90 days (Yang et al., 2016). A prospective study enrolled 69 male patients who suffered from severe traumatic brain injury, and the results showed that higher ferritin concentrations were significantly associated with fatal outcome (Simon et al., 2015). Ferritin is known as an acute-phase protein in response to inflammation, a common pathological state in critically ill patients (Marshall, 2001). Here, we also observed that the ferritin levels were moderately correlated with the infectious biomarkers and APACHE II and SOFA scores in this study. Infectious inflammation stimulates the upregulated expression of hepcidin and ferritin, resulting in relative iron deficiency in bloodstream (Drakesmith & Prentice, 2012). Upon that the ferritin level as the hallmark of iron homeostasis fluctuation probably reflects the severity of inflammatory response and risks of multiple organ dysfunction and anemia, complications that contribute to the increased mortality in critical illness. In this study, we found for each 10 ng/ml raise in ferritin, there was 2% increase in odds of short-term mortality.
Since early evaluation of neurocritically ill patients at risk of mortality may benefit the management of their treatment strategies, the protective effect of intervention in disturbances of iron homeostasis in neurocritical illness might worthy to be further investigated (Ganz, 2013;Ma, You, & Hao, 2012;Selim, 2009 TA B L E 3 Univariate and multivariate logistic regression analysis for iron parameters on adverse outcomes in NCU patients with eGFR ≥ 60 ml/min/1.73 m 2 impact of treatment strategies for iron metabolism on the prognosis is promising.
In conclusion, our study demonstrated that elevated ferritin level was independently associated with increased short-term mortality and long-term poor outcome in neurocritically ill patients with eGFR ≥ 60 ml/min/1.73 m 2 . Future studies with larger sample size should be conducted to confirm our conclusions, and the protective effect of intervention on iron metabolism among neurocritically ill patients might deserve further investigation.

ACK N OWLED G M ENTS
None.

CO N FLI C T O F I NTE R E S T
None declared.