Blood purification therapy for severe sepsis: a multicenter, observational cohort study in northern Japan

Sepsis is associated with life-threatening organ dysfunction caused by a dysregulated host response to infection. However, no specific therapy has been shown to improve mortality in patients with sepsis. We conducted a study to clarify the utilization status of various BPTs and the clinical characteristics of patients who received BPTs in northern Japan. In addition, the association of various BPTs with clinical outcomes was examined. This is a sub-analysis of the Tohoku Sepsis Registry, a multicenter, prospective, observational cohort study. To determine whether BPT was independently associated with in-hospital mortality in patients with severe sepsis, the following analyses were performed. Differences between survivors and non-survivors were assessed using Wilcoxon rank sum tests for continuous variables and Chi-square tests for categorical variables. Univariate logistic regression analysis was used to evaluate the factors associated with in-hospital mortality. In the multivariate logistic regression analysis, adjustments were made for the variables that were significant in the univariate logistic regression analysis. Clinical factors associated with mortality were analyzed. We enrolled 616 consecutive patients (≥ 18 years) with median Sequential Organ Failure Assessment scores of 8.0. During median of 22 days hospitalization, 139 patients died (mortality 22.6%). 20.7% of patients with severe sepsis received any type of BPT (mortality 38.6%). BPT consisted of 65.1% continuous renal replacement therapy (CRRT) with renal indication (mortality 48.8%), 26.0% CRRT with non-renal indication (mortality 21.2%), 22.2% intermittent renal replacement therapy (mortality 32.1%), and 33.1% polymyxin B-immobilized fiber column-direct hemoperfusion (mortality 42.9%). Meanwhile, no BPT group (mortality 18.5%) showed a significantly lower mortality than any BPT group. Besides, in multivariate analyses, all BPT modes were not independently associated with all-cause mortality. This study suggested the clinical status of BPTs for severe sepsis patients in northern Japan. Among all types of BPT, continuous renal replacement therapy (CRRT) for renal indication was most frequently selected. Severe sepsis patients received BPT had a higher mortality and severity; however, the BPT implementation may not be associated with mortality. Trial registration UMIN-CTR, UMIN000010297, Registered on 22 March 2013, https://upload.umin.ac.jp/cgi-open-bin/ctr_e/ctr_view.cgi?recptno=R000012055).

Background Sepsis is associated with life-threatening organ dysfunction caused by a dysregulated host response to infection [1]. In the USA, mortality rates have been reported to be > 15% [2]. Radical treatments include the administration of antimicrobial agents and surgical source control. However, no specific therapy has been shown to improve mortality in patients with sepsis.
Acute kidney injury (AKI) is strongly associated with poor prognosis in patients with sepsis [3]. Renal replacement therapy is recommended for sepsis patients with AKI [4]. However, it is unclear whether renal replacement therapy improves survival or renal recovery after sepsis-induced AKI [5]. Along with renal replacement therapy, blood purification therapy (BPT) has been proposed as a treatment for modulating the inflammatory response in sepsis, even in patients without renal dysfunction [4,5]. The inhospital mortality rate in sepsis patients undergoing BPT is reported to be high, ranging from 41 to 79% in patients with renal and non-renal indications [6][7][8][9]. However, mortality rates are highly variable, and the number of reports is limited. There is little detailed information on the use of BPT modes in septic patients and mortality rates for each BPT mode.
Several studies have shown the potential of BPT in modulating the immune response. Two meta-analyses [10,11] of small, randomized controlled trials evaluating the effects of hemoperfusion, plasma exchange, and hemofiltration with hemoperfusion demonstrated favorable results. However, the quality of these studies was not adequate. Even is the focus if only on renal replacement, the optimal timing and implementation remain unclear [12]. Thus, the benefits of BPT in terms of sepsis outcomes remain unclear. Mortalityrelated factors, such as age [13] and hepatic disease [14], have been reported in sepsis patients. However, there is no literature examining whether BPT is an independent clinical factor associated with survival.
We conducted a study based on prospective and continuously collected database to clarify the utilization status of various BPTs and the clinical characteristics of patients who received BPTs in northern Japan. In addition, the association of various BPTs with clinical outcomes was examined.

Study design
This is a sub-analysis of the Tohoku Sepsis Registry, a multicenter, prospective, observational cohort study conducted across 10 sites in the Tohoku district of northern Japan, including 3 university hospitals and 7 large community hospitals with > 300 beds. The protocol of the Tohoku Sepsis Registry has been registered at the University Hospital Medical Information Network Clinical Trials Registry (UMIN000010297). It is described in detail elsewhere [15]. The study design was approved by the Institutional Review Board of each institution and conforms to the provisions of the Declaration of Helsinki (as revised in Fortaleza, Brazil, October 2013). The need for informed consent was waived due to the observational nature of the study.

Study setting and participants
The Tohoku Sepsis Registry included consecutive patients admitted to intensive care units (ICUs) with severe sepsis or those presenting with severe sepsis after admission to ICUs or general wards in the 10 included hospitals between January and December 2015.
Participants were eligible if they were diagnosed with severe sepsis according to the 2012 International Sepsis Guidelines: sepsis-induced tissue hypoperfusion or organ dysfunction, including sepsis-induced hypotension, elevated serum lactate levels, low urine output despite adequate fluid resuscitation (< 0.5 mL/kg/h for > 2 h), acute lung injury with a ratio of arterial oxygen pressure to fractional inspired oxygen (PaO 2 :FiO 2 ) of < 250.0 in the absence of pneumonia as the source of infection or a PaO 2 :FiO 2 ratio of < 200.0 in the presence of pneumonia as the source of infection, elevated serum creatinine levels (> 2.0 mg/dL), elevated serum total bilirubin levels (> 2.0 mg/dL), low platelet counts (< 10.0 × 10 4 /mm 3 ), and coagulopathy with an international normalized ratio of > 1.5 [16]. Patients aged < 18 years were excluded.

Data collection
The registered information includes that on age, sex, preexisting comorbidities, and medications before admission, in addition to the unit where sepsis was diagnosed, the presence or absence of septic shock, severity as assessed using the Acute Physiologic Assessment and Keywords: Blood purification, Hospital mortality, Multivariate analysis, Sepsis, Acute kidney injury Chronic Health Evaluation (APACHE) II and Sequential Organ Failure Assessment (SOFA) scores, and primary site of infection. Other information includes physiological data, testing results, and treatment details, including drugs, source control interventions, and BPTs. The lengths of ICU and hospital stay and outcomes at 28 days post-diagnosis and at discharge were documented.
AKI was diagnosed according to the AKI Network criteria [17], which are used for classifying the different stages of AKI (stages 0-3) based on serum creatinine levels and urine output. AKI stage 1 is defined as an elevated serum creatinine level of ≥ 0.3 mg/dL or an increase of 150.0-200.0% from the baseline serum creatinine level or a reduction in urine output of < 0.5 mL/kg/h for > 6 h. AKI stage 2 is defined as an increase of 200.0%-300.0% from the baseline serum creatinine level or a reduction in urine output of < 0.5 mL/kg/h for > 12 h. AKI stage 3 is defined as an elevated serum creatinine level of ≥ 4.0 mg/ dL with an acute increase of ≥ 0.5 mg/dL or > 300.0% from the baseline serum creatinine level or a reduction in urine output of < 0.3 mL/kg/h for 24 h or anuria for 12 h.

BPT
In this study, BPT consisted of continuous renal replacement therapy (CRRT), including CRRT with renal and non-renal indications, intermittent renal replacement therapy (IRRT), and polymyxin B-immobilized fiber column direct hemoperfusion (PMX-DHP). The indications for BPT were based on the clinical judgment of the attending physicians at each institution. The equipment used and operational settings were determined by the clinician based on resources available at the institution. No uniform protocol was followed.
Participants were divided into the no BPT and different BPT groups. Patients were included in the BPT group regardless of the frequency or duration of BPT. If multiple BPTs were administered, patients were classified into more than one BPT group.

Outcomes
The primary outcome of this study was in-hospital mortality. Secondary outcomes included 28-day mortality, ICU-free days, length of ICU stay, and length of hospital stay. In-hospital mortality and 28-day mortality were defined as all-cause mortality at discharge or 28 days after the diagnosis of sepsis, respectively. ICU-free days represented the number of days on which ICU admission was not required within a 28-day period following the diagnosis of severe sepsis. The number of ICU-free days for patients who died within the 28-day period was 0.

Statistical analysis
Baseline patient characteristics (age, sex, preexisting comorbidities, worst physiological data on day 1 of diagnosis, and treatment modalities) were compared between the two groups of BPT (BPT vs. no PBT) by using Wilcoxon rank sum tests for continuous variables and Chi-square tests for categorical variables. Continuous variables were tested for normality with the Shapiro-Wilk test, with all continuous variables showing the skewed distribution (p < 0.05), thus summarized by median and interquartile range.
To determine whether BPT was independently associated with in-hospital mortality in patients with severe sepsis, the following analyses were performed. Differences between survivors and non-survivors were assessed using Wilcoxon rank sum tests for continuous variables and Chi-square tests for categorical variables. Univariate logistic regression analysis was used to evaluate the factors associated with in-hospital mortality at a significance level of 5%. Odds ratios and 95% confidence intervals were calculated. The results of univariate logistic regression analysis were used to select variables for multivariate logistic regression analysis. In the multivariate logistic regression analysis, adjustments were made for the type of BPT, as well as variables that were significant in the univariate logistic regression analysis. Since SOFA and APACHE II scores were correlated (Spearman's rank correlation coefficient, r = 0.687; p < 0.001), only SOFA score was included in multivariable logistic regression analysis. SOFA score includes the Glasgow Coma Scale, mean arterial pressure, PaO 2 :FiO 2 ratio, and platelet counts, which were also excluded from the multivariate logistic regression analysis. Eventually, type of BPT, SOFA score, lactate level, AKI stage, primary site of infection, mechanical ventilation use within 24 h, inotropes or vasopressors use were included in the multivariate logistic regression analysis. There were six BPT types included in multivariate logistic regression analysis: BPT, CRRT with renal indications, CRRT with non-renal indications, IRRT, and PMX-DHP. We performed six multivariate logistic regression models for each of the BPT types, and clinical factors associated with mortality were analyzed. The goodness of fit was assessed using the Hosmer-Lemeshow test. All statistical analyses were conducted using Stata ® software, version 15.1 (StataCorp, College Station, Texas, USA). Significance was defined as a two-sided p value of < 0.05.

Effects of BPT type on mortality
In the multivariate logistic regression analysis of BPT type, all BPT types were not associated with in-hospital mortality. Instead, SOFA score, serum lactate levels, and AKI stage 3 were significantly associated with inhospital mortality (Table 4).

Discussion
Based on the prospective and continuously collected database of severe sepsis patients recruited from the Tohoku Sepsis Registry in northern Japan, 20.7% of patients with severe sepsis received any type of BPT. The in-hospital mortality rate for patients who received any BPT was

Table 4 Factors influencing in-hospital mortality in multivariable logistic regression analysis
ORs (95% CI) were estimated using multivariate analysis adjusted for significant covariates in the univariable analysis Due to collinearity with the APACHE II score, the SOFA score was used for all analyses Variables used for calculating the SOFA score, i.e., GCS score, MAP, PaO  Meanwhile, no BPT group (mortality 18.5%) showed a significantly lower mortality than any BPT group. According to BPT type, the CRRT for non-renal indications group and IRRT group exhibited relatively low mortality rates. Besides, any BPT was not independently associated with all-cause in-hospital mortality. To our knowledge, this is the first study based on prospective cohort focusing on the effectiveness of BPTs for severe sepsis. Compared to previous Japanese study [18] that aggregated BPT patients in critical care, our PMX-DHP utilization rate is comparatively lower (33.1% vs. 43.4%) and our CRRT utilization rate is apparently higher (84.3% vs. 46.3%). While that previous study did its patient collection in 2013, ours in 2015. At international society of intensive care and emergency medicine conference in 2014, a randomized controlled trial reported did not confirm mortality benefit of PMX-DHP in patients with septic shock due to peritonitis [19], which was published early in 2015 [20]. This turn of the tide against PMX-DHP may have affected the results. The cohort in the previous Japanese study was sepsis patients, but severe sepsis was selected in this study; thus, CRRT may be selected more frequently for unstable hemodynamics.
The mortality rates at 28 days and at discharge were significantly higher in the any BPT group than in the no BPT group (34.5% vs. 18.3% and 38.6% vs. 18.5%, respectively). We consider that this result may have been influenced by the high severity of illness in patients requiring BPT. On analysis according to BPT type, similarly high mortality rates were observed in the CRRT for renal indications and PMX-DHP groups. In contrast, the CRRT for non-renal indications group may have exhibited relatively low mortality rates and differed from other modalities in terms of other characteristics (e.g., low APACHE II scores and less renal impairment). The IRRT group also did not exhibit a significantly higher mortality rate. The IRRT group was also the only group to not exhibit significant differences in the number of ICU-free days. From this group, 50.0% of patients were also included in the CRRT for renal indications group, and in 57.1% of patients, treatment commenced > 48 h after diagnosis. It is possible that IRRT may have been administered to a patient population with stable circulatory hemodynamics. This may have been responsible for the lack of significance in mortality rates and the number of ICU-free days in the IRRT group compared with those in the no BPT group.
Our findings also suggested that none of the BPTs was independently associated with all-cause in-hospital mortality. Although the BPT groups tended to have higher mortality rates, there was no evidence to suggest that BPTs contributed to adverse effects on the outcomes of this study. There is no clear evidence that a particular BPT (CRRT for renal indications [6,7,23,24,26], CRRT for non-renal indications [8], or PMX-DHP [21]) is associated with better clinical outcomes, and our data support this. It may be difficult to expect significant improvements in prognosis with current BPTs. The establishment of a new method of BPT is eagerly awaited. Novel methods, including methods using the AN69 surface-treated hemofilter (sepXiris ® ) [28], modified AN69 surface-treated hemofilter (oXiris ® ) [29,30], and extracorporeal cytokine adsorption device (CytoSorb ® ) [31], are currently being investigated. Further, large-scale studies are needed to evaluate the prognosis associated with the use of these methods.
This study has several limitations. First, the protocols of the various types of BPTs were based on the clinical judgment of the attending physicians at each institution and were not standardized. Thus, there was potential for variability in the application of BPT (e.g., indication, blood flow rate, dialysate flow rate, replacement flow rate, ultrafiltration rate, type of anticoagulation, and type of membrane). To address this, we included the institution as a variable in the multivariate logistic regression analysis. Second, this study did not have an adequately large sample size and may not be well adjusted for confounding variables. These limitations may affect the external validity of this study. Therefore, these findings should be interpreted with caution.

Conclusions
Our study showed that 20.7% of patients with severe sepsis received any BPT, and the mortality rate of patients who received any BPT was 38.6%. Among all types of BPT, CRRT for renal indication was most frequently selected. Patients who received BPT showed higher  severity of illness, and the mortality of any BPT group was significantly higher than that of no BPT group. According to BPT type, the CRRT for non-renal indications group and IRRT group may have exhibited relatively low mortality rates. BPT may not be independently associated with all-cause in-hospital mortality, although patients with severe sepsis who were treated with BPTs exhibited higher mortality rates. More detailed analyses adjusting for potential confounding variables are needed in additional cohorts in the future.