Early hemoglobin status is predictive of outcomes and long-term mortality of sepsis

Background: Hemoglobin change plays a critical role in progress of sepsis. However, the contribution of hemoglobin change to outcomes of patients with sepsis is still unknown. The aim of our study is to investigate the effect of early hemoglobin level within 48h after admission to Intensive Care Unit (ICU) on long term prognosis for sepsis. Methods: In this single centered, cohort study, we included patients from different ICU departments of Xiangya hospital from 2016 to 2018. Out of 1800 ICU patients, 199 patients with sepsis matched inclusion criteria. All 199 patients were divided into three groups according to 70, 80, and 90 g/L hemoglobin statuses, respectively. Results: Our study showed seventy-nine patients (39.7%) with ≤90g/L hemoglobin, forty-seven patients (23.6%) with hemoglobin ≤80g/L, and twenty-five (12.6%) with hemoglobin ≤70g/L at 48h after admission. Compared to survival group, there were higher rates of patients with hemoglobin ≤80g/L (33.7% vs. 15.1%, P=0.003), hemoglobin ≤70g/L (18.5% vs. 7.5%, P=0.031) in non-survival group, but similar rate of hemoglobin ≤90g/L patients (46.7% vs. 34.0%, P=0.081). Kaplan-Meier survival analysis showed that significant difference occurred among all three groups with total one-year survival rates. Furthermore, multivariate cox proportional hazards regression analysis showed that serum hemoglobin concentration ≤ 80g/L at 48h admission (Hazard Ratio HR 1.736, 95% CI 1.131-2.665, P=0.012), the occurrence of ARDS (HR 1.814, 95% CI 1.184-2.778, P=0.006), the use of CRRT (HR 1.569, 95% CI 1.030-2.390, P=0.036), and septic shock (HR 1.776, 95% transpeptidase, AKP alkline phosphatase, CK creatine kinase, CK-Mb creatine kinase isoenzyme, LDH lactate dehydrogenase, Mb myoglobin, TG triglyceride, CHOL cholesterol, LDL low density lipoprotein, HDL high density lipoprotein. group significant lower than > 90 g/L group (X 2 = 7.696, P = 0.0055), ≤ 80 g/L hemoglobin group significant lower than > 80 g/L group (X 2 = 15.29, P < 0.001), and ≤ 70 g/L hemoglobin group significant lower than > 70 g/L group (X 2 = 11.06, P = 0.0009).

3 early hemoglobin level in course of sepsis to achieve better outcomes of sepsis. Future larger researches and randomized controlled trials are needed to validate our results. Background Hemoglobin change is commonplace, but emerges silently and easy to be ignored in critical care patients (1,2). Sepsis is life-threatening organ dysfunction caused by a dysregulated host immune and inflammatory response to infection and has high incidence rate and mortality. Sepsis is also the main cause for anemia in ICU (3,4). Hemoglobin level continues to decline in the septic patients at the rate of 0.68 ± 0.66 g/dL/day but not in the nonseptic patients after the third day in ICU (5). The mechanism of hemoglobin variation in the course of sepsis is multifactorial covering inflammatory anemia, dilution of infusion, defective erythropoiesis, blunted bone marrow response to erythropoietin, preexiting conditions, hemolysis due to disseminated intravascular coagulation (DIC) or certain pathogens, adverse effect of drugs, repeated phlebotomy, nutrition deficit, et al (6,7). Pathogens, host immune system, kidney, liver and bone marrow probably have complicated crosstalks contributing to the underlying anemia in sepsis. The leading risk associated with hemoglobin decrease is the shortage of oxygen-carrying capacity of blood resulting in tissue hypoxia. Afterwards hypoxia brings about extensive organ injury, which may be latent at the early stage but further aggravates in the late period of septic course.
So maintaining early tissue oxygen supply is important to treat sepsis. Furthermore, hemoglobin plays a key role in defense against microorganism to assist leukocytes, nonspecific immunity as sequestering iron to cut off nutrition for bacteria, carring and transporting antibiotics, maintaining hemodynamic stabilization, carbon dioxide disposal, et al (8). So hemoglobin keeps internal environment homeostasis and the variation of hemoglobin is closely involved in the progress of sepsis. However, the hematologic system is often overlooked as an important organ in the treatment flowsheet of the patient with 4 sepsis, and the impact of hemoglobin change on outcomes of patients with sepsis is presently unknown.
Several high quality randomized controlled trials have focused on transfusion threshold for making clinical decision for critical ill and sepsis patients (9,10) and 70 g/L level is recommended but still arousing dispute in clinical application. While recent meta-analysis showed no difference in 28-or 30-day mortality between the liberal (a hemoglobin threshold of 9 or 10 g/dL) and restrictive transfusion (a hemoglobin threshold of 7 g/dL) strategies in patients with sepsis (11). However, hemoglobin status is distinct from transfusion for disparate impact on septic outcomes (12). Whereas few studies have explored the effect of early hemoglobin level on long term outcome of sepsis patients in ICU. Previous study has shown relationship between hemoglobin level and clinical outcome of septic patients, but whose grouping strategy including hemoglobin ≥ 7-<10 g/dl is so wide that probably missing detailed information of 8 or 9 g/dl on prognosis (12). However, the optimal hemoglobin level for good prognosis of sepsis still remains unclear. Hence, we want to expore further to determine the best fittable early hemoglobin level for better outcomes of sepsis. Therefore, in this single centered, cohort study, we evaluated the effect of early hemoglobin level of sepsis patients within 48 hours of ICU admission on long term prognosis, with the aim to investigate the most befitting early hemoglobin for better outcome of sepsis.

Study design and population
All patients who diagnosed with sepsis were admitted to ICU wards of Xiangya Hospital of Central South University for two years (January 1, 2016 to February 1, 2018). The study 5 was conducted in accordance with STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) statement. The data retrospectively collection was approved by the Ethics Committee of Xiangya Hospital. During the collection process, we adopted the basic information as hiding the patient's name and hospital number and implemented a unified de-identification to protect patient privacy. For the nature of the study, patients' informed consent was not necessary, and the Institutional Review Board waived the need for informed consent. The inclusion criteria were showed as follows: diagnosed with sepsis; ≥18 years old; ICU admission more than 48 h; with laboratory parameters within 48 h. The exclusion criteria were contained with age < 18 years, patients admitted with acute bleeding (gastrointestinal, trauma, postpartum hemorrhage, et al).
All patients underwent routine daily measure of serum level of hemoglobin. Patients with serum hemoglobin levels > 90 g/L versus ≤ 90 g/L at 48 h after admission were included in group 1, patients with serum hemoglobin levels > 80 g/L versus ≤ 80 g/L at 48 h after admission were included in group 2, and group 3 as serum hemoglobin levels > 70 g/L versus ≤ 70 g/L at 48 h after admission. We also investigated the clinical characteristics in sepsis by dividing to survival group and non-survival group according to survival state at hospital.

Variable definition
Sepsis was defined as life-threatening organ dysfunction caused by a dysregulated host response to infection according to Sepsis 3.0 criterion (infection plus an acute change in total Sequential Organ Failure Assessment, SOFA score ≥ 2 points) (13). The definition of septic shock was sepsis with persisting hypotension requiring vasopressors to maintain mean arterial pressure (MAP) ≥ 65 mmHg and having a serum lactate level > 2 mmol/L (18 mg/dL) despite adequate volume resuscitation. Acute respiration distress syndrome (ARDS) is defined by Berlin criteria of 2012 (14), and the definition of AKI is according to 6 adopted Kidney Disease: Improving Global Outcomes (KDIGO) clinical practice guideline (15).

Data source
We collected the clinical data from Electronic Hospital Records System (EHRs), including age, gender, co-morbid conditions, source of infection, vital signs at admission (heart rate, respiratory rate, and body temperature) and the clinical biochemical parameters at 48 h after admission to ICU (artery blood gas, blood routine, liver and kidney function, coagulation function, procalcitonin, myocardial enzymes, C-reaction protein, blood lipid full set), complications as septic shock, acute respiration syndrome (ARDS), and acute kidney injury (AKI) in septic patients, use of life support of the use of continues renal replication (CRRT), the use of vasoactive drug, mechanical ventilation (MV), the length of vasoactive drug usage days. All survival patients would be followed up by telephone at day 180 and 1-year after hospital discharge.
The primary outcome was defined as death from any cause within hospital, 180 days or one year after discharge. The secondary outcomes were defined as complications of sepsis (including septic shock, ARDS, AKI) and use of life support (the rate of CRRT, MV, vasopressor usage, days of using vasopressor, ICU hospitalization days, respectively).

Statistical analysis
Statistical analysis was performed using SPSS 23 (SPSS, Inc., Chicago, IL). Data were expressed as mean value (SD) for continuous variables, and as numbers (%) for categorical variables. The mean values and frequencies were compared between two groups using unpaired student t-test and chi-square or Fisher exact test variables. The survival curves of those three groups according to the levels of hemoglobin were generated using the Kaplan-Meier survival analysis and the difference in survival among 7 these groups was assessed using the log-rank test. Further, to adjust for confounding factors, the predictive values of mortality in sepsis were assessed by univariate and multivariate cox proportional hazards regression model, with hazard ratio (HR) and 95% confidence interval (CI). In all comparisons, a P value of ≤ 0.05 was considered statistically significant.

Results
From 1800 patients at ICU admission from January 1, 2016 to February 1, 2018, 199 patients were enrolled in the analysis (Fig. 1). 199 patients were included by screening through the inclusion the exclusion criteria in the Methods part. All 199 patients were

The survival curves
The Kaplan-Meier survival analysis and log-rank test were used to access survival rates among the three groups (Fig. 2). There were significant differences among all three 13 groups within total 1 year survival time, including the survival rate of ≤ 90 g/L hemoglobin group significant lower than > 90 g/L group (X 2 = 7.696, P = 0.0055), ≤ 80 g/L hemoglobin group significant lower than > 80 g/L group (X 2 = 15.29, P < 0.001), and ≤ 70 g/L hemoglobin group significant lower than > 70 g/L group (X 2 = 11.06, P = 0.0009).

Univariate and multivariate Cox hazard analysis of risk factors for mortality in septic patients
Further on, we used univariate and multiple Cox hazard analysis to investigate the independent risk factors of mortality in sepsis (

Discussion
In this cohort study, our results first show compared to > 80 g/L group, septic patients with hemoglobin ≤ 80 g/L within 48 h after admission to ICU has higher rates of inhospital, 180-d and one-year mortality. The rates of CRRT and MV usage are also higher in ≤ 80 g/L hemoglobin group. Furthermore, hemoglobin ≤ 80 g/L within 48 h on admission,the occurrence of ARDS, the use of CRRT and septic shock are the independent risk factors of one-year mortality for septic patients. While the frequency of infectious source from abdomen was the only independent protective factor for one-year mortality in septic patients.
The comparison between survival and non-survival groups demonstrates the higher mortality rate for male septic patients, which is consistent with recent experimental studies (16,17), sex-hormone-driven innate antibodies protect female from infection. And non-survival group has earlier hematologic and coagulation function turbulence, with higher rates of hemoglobin ≤ 80 g/L and ≤ 70 g/L, lower platelet, longer APTT as shown by our data. So hematologic system plays a key role at the beginning of sepsis. Serum lactic acid level is higher in non-survival group partially represents tissue hypoxia in early stage. Higher rates of complications including septic shock and ARDS, higher frequency of use of life support, covering CRRT, MV and vasopressors usage, longer days of vasopressors and higher SOFA score in non-survival group reflect more serious of sepsis.
The nearly similar baseline precludes bias affecting results. Intriguingly, there are significantly higher in-hospital, 180-d, one-year mortality in ≤ 80 g/L hemoglobin group compared to > 80 g/L group in septic patients. While ≤ 90 g/L hemoglobin group has the same in-hospital and 180-d mortality compared with > 90 g/L group. These results hint at 80 g/L hemoglobin could be the potential key cutoff to differentiate septic prognosis.
However, there are significant differences of survival rates among all three hemoglobin groups within one year shown by Kaplan-Meier survival analysis.
Furthermore, we find ≤ 80 g/L hemoglobin within 48 h after admission to ICU, the occurrence of ARDS, the use of CRRT, and septic shock are the independent risk factors of mortality of sepsis patients by multiple Cox hazard analysis to control possible cofounders. And the rate of infection site from abdomen is the only independent protective factor. Hemoglobin ≤ 90 g/L is not the independent risk factor as not included in the multiple Cox hazard model. Hemoglobin ≤ 70 g/L not incorporated into the model is possibly due to fewer cases. Albeit several large cohort shows 70 g/L is the optimal transfusion threshold for critical patients including sepsis (9, 10), these results are not in contradiction with ours. Because we try to explore the best early hemoglobin level closely affecting sepsis prognosis, instead of blood transfusion threshold. We discuss the effective early hemoglobin level for delivering oxygen to important organs in early stage of sepsis, rather than the adverse reaction nor decreased oxygen utilization ability brought by transfusing stored-blood products (18). Our conclusion is in line with early goal-directed therapy (EGDT) study which emphasizes maintaining early tissue perfusion and correcting early tissue hypoxia are the core therapy strategy for sepsis (19). Time course of hemoglobin in nonbleeding ICU patients shows that hemoglobin of sepsis patients persists to decrease in whole course in ICU compared to non-septic patients, and the change rate of hemoglobin concentration is inversely correlated to the severity of diseases (5). Several other studies results support our conclusion (20)(21)(22)(23), while recent Sung' research has shown that 90 g/L is the befitting cutoff hemoglobin level of 90-days mortality of patients with septic shock which is different from ours (24). However, there are two main differences between our study and Sung's: our study investigates much longer term, one year, prognosis of sepsis than Sung's, and our study subjects are septic patients while Sung's is the sick with septic shock. Patients with septic shock have more unstable hemodymic status thus probably requiring higher hemoglobin level to fulfill tissue oxygen need. Moreover, according our results hemoglobin ≤ 90 g/L not influences in-hospital or 180-days mortality, secondly hemoglobin ≤ 90 g/L is not independent risk factor of one- year mortality of septic patients in the process of using multivariate cox hazard model to control confounding bias. Studies with bigger data are needed to validate our results. In conclusion, we first report early hemoglobin level > 80 g/L best fits better outcome and longer survival rate, this level is also recommended by experts consensus focused on hemodynamic support of sepsis and adopted by extensive clinicians, although lack of robust evidences (25).
In our results, the higher rates of septic shock and ARDS are independent risk factors for septic mortality is reasonable. However, whether the use of CRRT is beneficial or detrimental to septic patients is always in eternal debate. There are other factors that cannot be ignored should be taken into comprehensive consideration, that is, various pharmacokinetic changes of diverse antibiotics under different renal replacement therapies among distinct individuals in sepsis (26)(27)(28). So in order to maximize therapeutic effect of antibiotics and minimize its adverse effects, dynamic monitoring of individual pharmacokinetic change of specific antibiotic is needed when accessing effect of certain kind of renal replacement therapy on sepsis. As for infection site, one study shows abdominal infection site is the independent protective factor of in-hospital mortality of sepsis associated acute lung injury, which is consistent with our results (29). To verify the exact mechanism needs further experimental studies.
We suppose the underlying rationality for our conclusion is mainly associated with relationship between decreasing hemoglobin and early tissue and organ hypoxia at the beginning of sepsis. Hypoxia works in pathophysiology of abundant diseases including sepsis (30). We propose that hypoxia in early stage of sepsis plays an important role in the progress and bad outcome of sepsis, as the only carrier for oxygen in circulation, hemoglobin level in early phase of sepsis should be paid more attention. In addition, erythrocyte has many other important functions for defending against infection and sepsis except for oxygen carrier, for instance red blood cells can help kill bacteria more efficiently (31). Changes in properties of erythrocyte are also early sensitive warning signals for monitoring progress of sepsis (32,33).
On the other hand, we want to emphasize that a simple hemoglobin cutoff should not be There are some limitations for our study. First, the type of our research is single centered, retrospective design and unavoidable for bias. We try our best to reduce confounding bias by using Cox proportional hazards regression model to prove the unique ability of early hemoglobin level predicting sepsis prognosis. What' more, we minimize cases that lose to follow-up and guaranteed follow-up equality to decrease withdraw bias; Secondly, our sample size is relatively not very big to enhance convince ability, while we complement the data to the greatest extent and do our best to decrease missing and guarantee data 19 integrity and quality to strengthen stringency. Third, the single centered design restricts the generalizability of our conclusion, but we conducte the research in different ICUs setting to narrow generalizability reduction, multiple centered plan is suggested in the future. Fourth, ≤ 70 hemoglobin not including in multivariate cox hazard model is due to the case number of ≤ 70 hemoglobin was so small, future study with larger data is needed to verify our result.
There are some strengthens of our study. First, we focus on the influence of early hemoglobin level, grouped by 70, 80 and 90 g/L, on the long term prognosis of sepsis, which was not investigated before. And we found ≤ 80 g/L hemoglobin is the independent risk factor for sepsis longer mortality, which is meaningful to raise awareness to value maintaining early hemoglobin stable in therapy for sepsis. Secondly, we suggest readers to interpret our results with discretion and call for uniting our results with actual situation of patient at the bedside to make best clinical decisions.

Conclusion
We suggest that ≤ 80 g/L hemoglobin within 48 h on ICU admission is the independent predictor for longer term mortality for patients with sepsis, meanwhile we should also make decisions according to the characteristics of each patient in specific clinical situation. Furthermore, future studies with high quality of bigger data and randomized controlled trials are called for to validate our results and experimental researches are needed to illuminate underlying mechanism of the early tissue hypoxia and its

Ethics approval and consent to participate
The study was approved by the Institutional Review Board of Xiangya Hospital Central South University which abides by the guidance of the Declaration of Helsinki. This was a retrospective study, all data from electronic medical record and we adequately protected patients' privacy, so consent requirements were waived as a result. IRB number 2017121160.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analyzed during this study are included in this published article.  Figure 1 Flow chart for participants' number at each stage of study including follow-up.