Introduction
The red blood cell distribution width (RDW) represents an index of the heterogeneity of the erythrocytes (anisocytosis), which is calculated by dividing the standard deviation of erythrocyte volume by the mean corpuscular volume (MCV) and multiplying by 100 to express the result as a percentage [
1]. RDW is widely available to clinicians because it is routinely reported as part of the complete blood count.
For several decades, RDW has been typically used in combination with the MCV to differentiate the cause of underlying anemia in clinical practice [
2]. Recently, highly significant associations have been described between RDW value and all-cause, noncardiac, and cardiac mortality in patients with coronary artery disease, acute and chronic heart failure, peripheral artery disease, stroke, pulmonary embolism, and pulmonary artery hypertension [
3‐
8]. Moreover, several studies have reported that RDW shows the predictive value of all-cause mortality in critically ill or intensive care unit (ICU) patients [
9‐
12]. Although it has been postulated that systemic inflammation, malnutrition, and impaired renal function play a significant role in the underlying pathological processes [
13], the mechanism of the association between increased RDW and mortality remains unclear.
Until now, most previous studies that have investigated the relationship between RDW and clinical outcomes of various cohorts have used a single RDW measurement at initial presentation, and little is known about the potential impact of changes in RDW from baseline on survival in critically ill patients. However, RDW can be considered as a dynamic variable with rapid changes associated with acute disease states such as acute myocardial infarction and acute decompensated heart failure [
14,
15]. Thus, we hypothesized that the changes in RDW from baseline can reflect acute disease states and provide more prognostic information than the baseline RDW value alone. Therefore, we investigated whether the change in RDW value between baseline and 72 hours after hospital admission had prognostic value for clinical outcomes in patients with severe sepsis or septic shock.
Materials and methods
Patients
Eligible adult patients who were admitted to the emergency department (ED) with severe sepsis and/or septic shock between November 2007 and November 2011 were assessed for possible enrollment according to inclusion and exclusion criteria. Since November 2007, early goal-directed therapy (EGDT) has been implemented in the ICU and in the ED at our institute as part of a quality improvement initiative. If a patient presented with two or more systemic inflammatory response syndrome criteria and a suspicious sign of infection, the patient’s eligibility for EGDT was assessed. One or both of the following triggered initiation of our EGDT protocol: (a) initial systolic blood pressure <90 mmHg, despite a 20 mL/kg intravenous crystalloid fluid challenge; or (b) initial serum lactate level ≥4 mmol/L. The criteria for exclusion included: (a) age <18 years; (b) any contraindication to central venous catheterization; and/or (c) presence of a do-not-resuscitate order.
The study was carried out in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Yonsei University Health System Clinical Trial Center. We obtained informed written consent from all participants involved in our study.
Data collection
Baseline characteristics including demographic information and preexisting chronic comorbidities were collected. The Charlson Comorbidity Index (CCI) was used to assess the burden of chronic disease [
16,
17]. Moreover, both Acute Physiology and Chronic Health Evaluation (APACHE) II score and Sequential Organ Failure Assessment (SOFA) score were determined using the worst values within the initial 24 hours of ED admission for disease severity assessment. SOFA score was calculated by the parameters as follows: PaO
2/FiO
2, platelet count, bilirubin, blood pressure and the use of inotropic agent, Glasgow coma score scale, and creatinine or urine output. In addition, RDW, white blood cell (WBC) count, hemoglobin (Hb) level, hematocrit, and MCV were measured at initial presentation and at 72 hours after ED admission, using the Advia 2120 Hematology Analyzer (Siemens Healthcare Diagnostics, Deerfield, IL, USA). RDW is reported as a coefficient of variation (percentage) of red blood cell volume. The normal reference range for RDW in this hospital laboratory is 11.5 to 14.5%.
Definitions
Sepsis, severe sepsis, and septic shock were defined using the American College of Chest Physicians/Society of Critical Care Medicine consensus conference definitions [
18]. Sepsis was defined by two or more of the following conditions as a result of infection: (i) temperature greater than 38°C, (ii) heart rate greater than 90 beats/min, (iii) respiratory rate greater than 20 breaths/min or PaCO
2 less than 32 mmHg, and (iv) WBC count greater than 12,000 cells/μL or less than 4,000 cells/μL. Severe sepsis was defined as sepsis associated with organ dysfunction, hypoperfusion abnormality, or sepsis-induced hypotension. Hypoperfusion abnormalities included lactic acidosis, oliguria, and acute alteration of mental status. In addition, septic shock was defined as sepsis with hypotension despite adequate fluid resuscitation. Hypotension was defined as a systolic blood pressure of 90 mmHg or less, or a reduction of greater than 40 mmHg from baseline in the absence of other causes of low blood pressure.
Infection sites were categorized as pneumonia, peritonitis, urinary tract infection, exacerbation of chronic obstructive pulmonary disease, catheter-related infection, primary bacteremia (excluding untreated Staphylococcus epidermidis bacteremia), miscellaneous sites (mediastinitis, prostatitis, osteomyelitis, and others), or multiple sites [
19]. Effectiveness of antibiotics was assessed based on the microbial culture results, the known susceptibility of the organism to the antimicrobials used, and antimicrobial susceptibility testing [
19].
Statistical analyses
Continuous variables are expressed as means ± standard deviation or median (interquartile ranges), and categorical variables as numbers with percentages. Baseline characteristics of the groups were compared using one-way analysis of variance for continuous variables and the χ2 test for categorical variables. In the present study, all patients were followed for 12 months after admission to the ED. We evaluated 28-day and 90-day all-cause mortality as primary endpoints. We also investigated the total length of hospital stay (TLHS) as a secondary endpoint for analysis. The change in RDW between baseline and 72 hours after ED admission (ΔRDW72hr-adm) was calculated as RDW at 72 hours – RDW at baseline, and the median value of ΔRDW72hr-adm was 0.2%. Moreover, we categorized patients into four groups according to baseline RDW value and ΔRDW72hr-adm as follows: group 1, patients with RDW levels in the reference range (normal RDW level) at baseline and ΔRDW72hr-adm ≤0.2%; group 2, patients with increased RDW at baseline and ΔRDW72hr-adm ≤0.2%; group 3, patients with normal RDW at baseline and ΔRDW72hr-adm >0.2%; and group 4, patients with increased RDW at baseline and ΔRDW72hr-adm >0.2%. Based on the four groups stratified by baseline RDW value and ΔRDW72hr-adm, survival curves were designed using the Kaplan-Meier method, and comparisons were made using the log-rank test. The prognostic value of the changes in RDW on 28-day and 90-day mortality was determined using Cox proportional hazards model, and the results were presented as hazard ratios (HRs) and the 95% confidence intervals (CIs). In multivariate analysis, the Cox models were adjusted for age, sex, SOFA score, CCI, renal replacement therapy (RRT), serum albumin level, Hb, lactate, C-reactive protein (CRP), and infection sites, which were thought to plausibly interact with both RDW and mortality. Furthermore, multivariate linear regression analysis was also performed to determine the factors independently associated with TLSH. All tests were two-sided, and a P value of <0.05 was considered statistically significant. Statistical analyses were performed using SPSS for Windows version 19.0 (IBM Corp, Armonk, NY, USA).
Discussion
The present study is a prospective clinical investigation of the prognostic value of changes in RDW in patients receiving a standardized resuscitation algorithm (EGDT) for severe sepsis or septic shock. The impact of changes in RDW during the first 72 hours after ED admission on all-cause mortality was analyzed by categorizing the patients into four groups according to baseline RDW value and ΔRDW72hr-adm (RDW at 72 hours – RDW at baseline). To the best of our knowledge, this study is the first to report that an increase in RDW has a potential role in predicting mortality in patients with severe sepsis or septic shock. The main finding of the present study is that an increase in RDW from baseline during the first 72 hours after hospitalization can serve as a strong independent predictor of mortality in patients with severe sepsis or septic shock. This significant association between an increase in RDW and mortality remained unaltered even after adjusting for various confounding variables. Group 4 patients, those with increased RDW at baseline and ΔRDW72hr-adm >0.2%, exhibited the highest risks in 28-day and 90-day mortality, while group 1 patients, those with normal RDW level at baseline and ΔRDW72hr-adm ≤0.2%, had the lowest mortality risks. Thus, we speculate that RDW is a dynamic marker of risk in septic patients, and that an increase in RDW during the course of hospitalization can serve as an early indicator of adverse outcomes.
Interestingly, in multivariate analysis for 90-day mortality, a significantly higher mortality risk was observed in group 3 in which RDW increased within 72 hours after ED admission (normal RDW at baseline and ΔRDW72hr-adm >0.2%) compared to group 1, while there was an insignificant difference between group 1 and 3 in 28-day mortality. These findings demonstrate that the relationship between an increase in RDW and mortality may be more meaningful, when the follow-up duration is longer.
Although the mechanism underlying the association between higher RDW and mortality in septic patients is not yet completely understood, several plausible explanations have been suggested in previous studies. Systemic inflammation has been shown to predict progressive illness, cardiovascular mortality and death in ICU patients. Systemic inflammation response impacts bone marrow function and iron metabolism [
20,
21], and proinflammatory cytokines have been found to inhibit erythropoietin-induced erythrocyte maturation and proliferation, and to downregulate erythropoietin receptor expression, which are associated with RDW increases [
22]. Oxidative stress may also be a contributing factor of the association between RDW and mortality. High oxidative stress is present in sepsis through the generation of reactive oxygen species by activated leukocytes [
23]. Moreover, it has been proposed that oxidative stress induces an increase in RDW by reducing RBC survival and increasing the release of large premature RBCs into the peripheral circulation [
24]. Another explanation may be related to malnutrition. Nutritional markers, including total cholesterol and albumin, are believed to be significantly associated with RDW [
13]. In addition, RDW has also been noted to be associated with renal dysfunction, which is known to be closely related with inflammation and malnutrition [
13,
25]. Recently, Veeranna
et al. [
26] demonstrated the predictive ability of RDW for HbA1c in healthy nondiabetic adults, proposing the possibility of chronic hyperglycemia along with oxidative stress and inflammation as mediators of the association between RDW and adverse outcomes. All things considered, it is reasonable to assume that increased RDW may represent an integrative measure of the multiple harmful pathologic processes, including inflammation, oxidative stress, malnutrition, and renal dysfunction, that simultaneously occur in critical illness. Thus, with respect to the results of this study, we suggest that changes in RDW, especially increasing RDW, reflect the aggravation of inflammation, oxidative stress, nutritional deficiencies, and renal dysfunction, and that the combination of a baseline RDW value and an increase in RDW can be a promising independent prognostic marker in septic patients. However, inflammation, oxidative stress, nutritional deficiencies, and renal dysfunction may not entirely explain why increased RDW is associated with mortality and greater length of hospital stay, because these associations were still significant even after adjusting for total cholesterol, albumin, and SOFA score; therefore further study is needed to determine the mechanism for the association between RDW values and mortality.
There are several limitations to our study. First, we arbitrarily determined 72 hours after ED admission as the second RDW measurement and defined an increase in RDW as ΔRDW
72hr-adm >0.2%. Whether changes in RDW during the first 72 hours could represent the pathophysiologic changes and therapeutic responses in critically ill patients is not clear. Ku
et al. [
27] suggested that 72-hour RDW after the onset of bacteremia could be a predictor of all-cause mortality in patients with Gram-negative bacteremia. Therefore, based on this previous study, we investigated the clinical outcomes of the patients with severe and/or septic shock stratified by the changes in RDW values between baseline and 72 hours after admission. Second, we did not investigate the use of erythropoietin, iron or vitamin B
12 deficiency, and reticulocyte count, which could have affected RDW values and, thus, might have limited the interpretation of study results. Third, the low number of patients in group 4 might introduce significant interference in the use of statistical models with numerous covariates. We examined two aspects of our models to assure the goodness of fit of our statistical models. For one thing, we conducted multivariate logistic regression analyses with the covariates used in Adjusted Model 2 and examined the results of the Hosmer-Lemeshow tests. Therefore, we confirmed that the
P values for 28-day and 90-day mortality were 0.913 and 0.769, respectively, which demonstrated that these statistical analyses with numerous variables were acceptable. Next, we examined the proportional hazards assumption of the four groups, stratified by baseline RDW value and ΔRDW
72hr-adm, through log-minus-log-survival plot, and the parallel lines of the log-minus-log function were verified. Taken together, we suggested that these statistical analyses were tolerable, even though the number of patients in group 4 was relatively small. Furthermore, we analyzed the data with a simplified grouping in the same frame of statistical models. Patients were categorized into four groups based on whether RDW level was within the normal range or increased at baseline and at 72 hours. Group 1 was defined as patients with normal RDW at both time points, group 2 as patients with increased RDW at baseline and normal RDW at 72 hours, group 3 as patients with normal RDW and increased RDW at 72 hours, and group 4 as patients with increased RDW at both time points. In the results, we found that an increase in RDW from baseline was significantly associated with mortality. Moreover, group 4 in this subanalysis exhibited the highest 28-day and 90-day mortality rates similar to our outcomes (See Additional files
1 and
2). Finally, the sample size and the number of events were not large enough to establish statistical significance of increased risk among categorized groups. Therefore, a larger multicenter study with repeated RDW measurements is necessary to further clarify the predictive value of changes in RDW.
Competing interests
The authors declare they have no competing interests.
Authors’ contributions
CHK performed the data review/collection, the statistical analyses and developed the initial draft of the manuscript. EJK, JHH, and JSH were involved in the collection and assembly of data. JYC and YSK organized the data collection. JTP, SHH, and HJO revised the draft of manuscript and assisted in data analysis. HJO, THY, and SWK contributed to the study design and coordination of the study. All authors read and approved the final manuscript.