Introduction
Sepsis is a severe condition characterized by uncontrolled host response to infection, leading to organ dysfunction and high mortality rates [
1]. Its underlying pathogenesis is unclear, but studies have shown that bacterial endotoxins could trigger dysregulation and excessive cascade release of proinflammation, thereby inducing sepsis [
2]. Therefore, strategies aimed at the efficient removal of endotoxins and inflammatory factors are considered an essential aspect of the comprehensive approach to treating sepsis and optimizing patient outcomes [
3].
In recent years, continuous blood purification therapy has emerged as a treatment approach for sepsis. This therapy can potentially remove inflammatory cytokines from the bloodstream, thereby reducing the overall inflammatory response and promoting the restoration of immune homeostasis [
4,
5]. The oXiris hemofilter, designed for endotoxin adsorption, incorporates a novel approach by utilizing the inherent hydrogel structure of the AN69 membrane. Additionally, the surface of the oXiris filter is coated with polyethyleneimine (PEI) and heparin coatings [
6,
7]. PEI is a multilayer linear structure cationic complex that can adsorb endotoxins with negative charges on the surface. At the same time, the negative charges retained in the bottom layer can absorb inflammatory cytokines [
6,
7]. It has been reported that the application of PEI on surfaces can enhance the endotoxin adsorption and cytokines removal capacity of the AN69 membrane [
5]. Compared to other hemofilters, the oXiris hemofilter effectively liminates inflammatory mediators and adsorbs endotoxins.
Nonetheless, there is still debate on whether it brings advantages in treating patients with sepsis [
8,
9]. Therefore, all published and ongoing studies were pooled together to assess the impact of the oXiris filter on clinical outcomes of this patient population.
Methods
Protocol and registration
The methods and reports of this study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (PRISMA 2020 Checklist) (Additional file
1). The study protocol was registered in the international PROSPERO database for systematic reviews (registration number CRD42022356340). Amendments to the information provided in the protocol are shown in Additional file
2: Appendix 9.
Study inclusion and exclusion criteria
The criteria for the inclusion and exclusion of articles were as follows. The target population included adults diagnosed with sepsis undergoing CRRT. The intervention group included patients who received treatment with the oXiris filter, while the comparison group comprised patients who received other types of filter treatments. It is important to note that the oXiris group and the other filters group consisted of different individuals. Only observational studies (cohort studies and case series studies with controls) and randomized controlled trials (RCTs) were included. If studies lacked available data, they were excluded. If studies did not have a control group and used self-control before and after the intervention, they were also excluded.
Literature research and data extraction
Eleven databases (English databases: Web of Science, the Cochrane Library, PubMed, and EMBASE; Chinese databases: Wanfang Database, China National Knowledge Infrastructure, Sino-Med, and VIP Database for Chinese Technical Periodicals; and three major clinical research registration websites: International Clinical Trials Registry Platform, Chinese Clinical Trial Registry, and ClinicalTrials.gov) were systematically searched to collect all studies relevant to this research as of July 2022. To account for the possibility of new literature being published during the meta-analysis and article writing process, an additional search was conducted across 11 databases from July 2022 to November 9, 2022. The detailed search strategy is shown in Additional file
2: Appendix 1.
Study selection
The literature screening process involved three fellows who independently assessed the titles, abstracts, and full texts to determine whether the literature met the eligibility criteria. In cases of disagreements between the investigators, consensus-based decision-making was utilized. If necessary, a fourth investigator was consulted to resolve any disputes.
Four researchers independently retrieved data from the included study and used standardized forms for consistency. The data extracted included study characteristics, patient characteristics, intervention measures, and results. The method for unknown non-normal distributions was applied to transform the median and quartile into the mean and standard deviation (SD) [
10,
11]. Graphical data in the original studies were extracted using WebPlotDigitizer, a semiautomated tool that extracts underlying digital data by reverse engineering a visual image of the data [
12,
13]. Disagreements during the data extraction process were settled by consensus among the four researchers.
Outcomes
The primary outcome was the 28-day mortality, while the secondary outcomes included 7-, 14-, and 90-day mortality, length of ICU and hospital stay, ICU and hospital mortality, norepinephrine (NE) dose, lactate and interleukin-6 (IL-6) and lactate levels, and Sequential Organ Failure Assessment (SOFA) score.
Assessment of risk of bias
Three investigators, respectively, evaluated the quality of the included studies. The Cochrane Risk of Bias Tool was used to assess the quality of the included RCTs [
14]. Each area was given a risk of bias rating of low, unclear, or high. If a study had a high risk of bias for one or more key areas, it was considered at high risk of bias. If a study had an unclear risk of bias for one or more key areas and no high-risk areas, the study's risk of bias was considered unclear. If a study had a low risk of bias for all key areas, the risk of bias was considered low [
15]. The Newcastle–Ottawa Scale (NOS) was used to assess the quality of the included observational studies [
16]. Scores < 6 were classified as low quality, 6 to 7 as intermediate quality, and 8 to 9 as high quality. To evaluate the impact of a low-quality study on the results of the meta-analysis and whether the results were robust, it was included in meta-analyses but excluded for sensitivity analysis when conducting meta-analyses.
Statistical analysis
Interventions in each group, outcomes, and sample sizes of each study were summarized in a table. Revman version 5.3 was used for the statistical analysis. The primary outcome was 28-day mortality.
The Mantel–Haenszel method was used to analyze the binary variable and was represented as odds ratio (OR). The weighted mean difference (WMD) was used as the effect indicator while analyzing the continuous variables using the inverse variance approach. The standardized mean difference (SMD) was utilized as the effect indicator when the absolute values of the continuous variables differed greatly, or the units were not uniform. All results were visualized in forest plots. Statistical significance was defined as a two-tailed p < 0.05. The statistical heterogeneity in the included trials was evaluated using I2 statistics and Cochrane's Q test. Heterogeneity was deemed significant when I2 statistics were greater than 50%, and the P value for Cochrane's Q test was less than 0.1. The fixed-effects model was applied for meta-analyses if I2 was less than or equal to 50%. The random-effects model was used for meta-analyses if I2 was more than 50%. A publication bias test was not conducted for outcomes that included < 10 articles to avoid misjudging accidental errors as publication bias due to the limited number of studies.
In Xie et al.'s study, the investigators employed the inverse probability of treatment-weighting method (IPTW) to reduce selection bias and control for confounding factors. After IPTW, the total sample size increased from 76 to 149 [
17], which might inflate the weight of Xie's study in the entire meta-analysis, thus affecting the pooled results. Therefore, for 7-, 14-, and 28-day mortality, length of ICU and hospital stay, the sensitivity analysis method was used to separately include data before and after IPTW for two independent meta-analyses of each outcome. The results of the two meta-analyses were compared to judge whether the IPTW in Xie et al.'s study affected the meta-analysis results.
If the treatment period was inconsistent, the results after the longest treatment period of CRRT were included in the meta-analysis. For SOFA score, NE dose, lactate level, and IL-6 level, post-treatment data were included for meta-analyses, because the accurate mean and SD of data on changes before and after treatment could not be obtained. To explore the baseline similarity, meta-analyses on the baseline data of these four outcomes were conducted. If an included original study did not report the baseline data, its post-treatment data were excluded for sensitivity analysis to assess whether unknown baseline incomparable risk contributed to the results of the meta-analysis. If the results of the outcomes of the oXiris and control groups showed significant differences at baseline or data reported in a study could not be accurately converted into the mean and SD of the outcomes after treatment, the outcomes of the studies were not included for meta-analysis, but qualitatively described and compared to the meta-analysis of other studies.
For IL-6 level, NE dose, and SOFA score, only the post-IPTW data were used for meta-analysis when Xie et al.'s study provided data before and after IPTW [
17].
Assessment of the quality of evidence
The Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) method was used to rate the quality of the evidence using the GRADEpro GDT online tool [
18,
19]. The rating process referred to the GRADE Handbook and guidelines [
20,
21]. When studies investigating outcomes included both RCTs and observational studies, subgroup analyses were conducted according to the type of studies. The quality of the evidence from RCTs and observational studies was rated using the GRADE method.
Discussion
This meta-analysis included 695 patients with sepsis that underwent CRRT. Our results showed that compared to other filters (AN69-ST, M100, M150 and FX80), the use of the oXiris filter during CRRT treatment may be associated with lower 7-, 14- and 28-day mortalities, SOFA score, NE dose, lactate level, and shorter length of ICU stay. However, the evidence for these outcomes had a low or very low level of certainty because the original study design was mainly observational studies and the RCTs included had an unclear risk of bias and a small sample size.
In Xie et al.'s study, the investigators employed the IPTW method to reduce selection bias and control for confounding factors [
17]. To assess the potential impact of the increased virtual sample size resulting from IPTW in Xie et al.’s study on the meta-analysis results, the data before and after IPTW were separately included in two independent meta-analyses for each of the five outcomes: 7-day mortality, 14-day mortality, 28-day mortality, length of hospital stay, and length of ICU stay. The pooled results showed no significant difference using data before or after IPTW.
In the present study, we conducted a subgroup analysis of outcomes using Xie et al.'s data after IPTW. The outcomes analyzed included 28-day and hospital mortality, SOFA score, lactate level, IL-6 level, NE dose, length of hospital stay, and length of ICU stay. The subgroup analysis was performed according to the study design (i.e., RCTs vs. observational studies) if the outcomes were present in both types of studies. The certainty of the evidence was rated with the GRADE method. According to the subgroup analysis based on observational studies, oXiris was associated with a shorter length of ICU stay and lower 28-day mortality than other filters in patients with sepsis. And the certainty of the evidence was low. However, the meta-analysis of RCTs showed that oXiris did not significantly differ from other filters regarding 28-day mortality or length of ICU stay (P > 0.05). The very low certainty of the evidence may be attributed to the risk of bias, the small sample size of the RCTs, and imprecision. Given that the certainty of evidence from observational studies was higher than that from RCTs, it was inferred that oXiris was associated with lower 28-day mortality and shorter length of ICU stay in patients with sepsis. The meta-analysis of observational studies indicated that the oXiris filter showed reductions in the SOFA score, lactate level, and NE dose compared to other filters. However, no significant differences were observed when considering the findings from the meta-analysis of RCTs. The lack of significant variance during the analysis of RCTs may be attributed to the small sample sizes of the included RCTs. Evidence from RCTs or observational studies had very low levels of certainty. Based on the available evidence, the pooled results of all observational studies and RCTs suggested potential reduction effects with oXiris on the SOFA score, lactate level, and NE dose compared to other filters. However, the very low certainty of the evidence from both types of studies suggests that more high-quality RCTs are needed to confirm these findings. The meta-analysis of either RCTs or observational studies suggests a potential reduction effect of oXiris compared to other filters on the IL-6 level. However, the certainty of the evidence from both types of studies was very low. It is important to note that Guan et al.'s study, excluded from the meta-analysis due to significant baseline differences, did not find a lower IL-6 level with CRRT using oXiris for 72 h. Therefore, the effectiveness of the oXiris filter in reducing the IL-6 level remains controversial.
In three studies, the interleukin-8 (IL-8) level in the oXiris group was lower than in the control group and showed significant differences [
6,
24,
30]. Furthermore, the oXiris-CVVH group's vasoactive inotropic score (VIS) was significantly lower than the control group in three studies [
17,
24,
25]. Two of the included studies showed that the lower serum endotoxin degree was significantly associated with the oXiris filter group after treatment compared to other filters in CRRT [
6,
27].
Moreover, we found that the length of hospital stay, ICU and hospital mortality, and 90-day mortality did not differ significantly, possibly due to small sample sizes and insufficient statistical efficiency. The evidence for these outcomes had a low or very low level of certainty primarily due to the small sample size and unclear risk of bias of most observational studies and RCTs.
In vitro studies have shown that oXiris could remove both endotoxin and cytokines compared to CytoSorb and Toraymyxin [
7]. The adsorption of endotoxins and cytokines by the oXiris membrane could reduce the inflammation of the organism and rapidly alleviate the cascade response caused by inflammatory factors, thus improving the patient's outcome [
27]. Current adsorptive blood purification techniques have been observed to improve hemodynamics and the respiratory system [
34‐
36]. They involve the removal of bacterial toxins and cytokines from the patient's blood, suppressing excessive inflammatory responses, and restoring immune homeostasis [
4]. The cellular theory proposes that adsorptive blood purification techniques can interact with immune cells and the blood filter, leading to various cellular responses. These blood purification methods can potentially regulate the expression of molecules on the surface of leukocytes involved in adhesion and migration, antigen presentation, and apoptosis. Some immune cells, such as neutrophils and monocytes, can also adhere to blood purification tools, contributing to regulating the immune system [
5]. The oXiris membrane and other adsorptive blood purification techniques hold significant implications for treating sepsis patients by modulating immune cell responses and promoting immune system balance.
Low or very low quality of evidence from the present study suggested that compared with other filters (AN69-ST, M100, M150 and FX80), the use of the oXiris filter may be associated with better hemodynamic improvements (NE dose, and lactate level), improvement in organ function (SOFA score), shorter length of ICU stay, and lower short-term mortality. However, the oXiris filter did not bring benefits in reducing long-term mortality (90-day mortality), which may be attributed due to the small sample size. Overall, the oXiris filter may be promising for CRRT in sepsis patients. However, due to the small sample size and limited quality of the original studies, the role of oXiris remains to be further explored through future high-quality studies.
Strengths and limitations
The therapeutic impact of CRRT treatment with the oXiris filter in patients with sepsis was initially investigated in this meta-analysis providing the basis for the clinical use of oXiris in treating sepsis. However, the limitations of this study should be acknowledged. First, most included studies were observational studies, and only four were RCTs. However, RCTs were conducted at single centers with small sample sizes. Accordingly, there is an urgent need for high-quality RCTs with sizable sample sizes. Second, for SOFA score, NE dose, lactate level, and IL-6 level, the accurate mean and SD of data on changes in outcomes before and after treatment could not be obtained. Therefore, meta-analyses were not performed on the changes before and after treatment, but on baseline data and post-treatment data, respectively. Third, during the pooled analysis of IL-6 and lactate levels, NE dose, and SOFA score, some studies lacked accurate data conversion into the mean and SD of the outcomes after treatment, or had uneven baseline characteristics. As a result, it was not possible to pool data from all studies for a meta-analysis of these outcomes, and a qualitative description was provided instead. Fourth, there was high heterogeneity across studies on lactate level analysis (I2 > 50%). In addition, the regional limitation of the included studies should be acknowledged since most of them were from China. This limitation could not be addressed in the present study due to the lack of original studies. Nevertheless, there are ongoing RCTs in Switzerland and France (Registration nos.: NCT01948778 and NCT03426943) evaluating the efficacy of the oXiris filter in sepsis. With the publication of these studies and the subsequent updates of meta-analyses, the issue of regional bias may be addressed.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.