Corticosteroids Had no Effect on 28-Days Mortality in Adult Patients with sepsis: A Systematic Review and Meta-Analysis

Huoyan Liang Zhengzhou University First A liated Hospital Heng Song Zhengzhou University First A liated Hospital Gaofei Song Zhengzhou University First A liated Hospital Hongyi Li Zhengzhou University First A liated Hospital Xianfei Ding Zhengzhou University First A liated Hospital Quancheng Kan Zhengzhou University First A liated Hospital Tongwen Sun (  suntongwen@163.com ) Zhengzhou University First A liated Hospital https://orcid.org/0000-0001-6536-3222


Introduction
Sepsis is a life-threatening organ dysfunction, which is caused by a dysregulated host response to infection [1,2] that culminates in systematic hypoperfusion and considerable organ dysfunction. The therapy mainly are antibiotic administration and perfusion restoration in the early phrase [3]. Early and aggressive treatment is associated with a mortality rate of 30-50% in critical ill patients admitted to the ICU and induces more than 5 million deaths each year across the world [3,4]. Therefore, further investigation for the treatment of sepsis is crucial.
The pathology of sepsis is marked by a dysregulated host response to infection; therefore, immunomodulatory therapies have been used in sepsis treatment that may provide effective [5]. In fact, since the middle of the 20th century, the doctors start to use the corticosteroids as adjuvant therapy for sepsis [3]. With corticosteroids were used in sepsis, especially septic shock therapy, numerous randomized clinical trials (RCTs) were performed to evaluate the safety and e cacy of corticosteroids. However, the results of these RCTs varied. Thus, many systematic reviews have been performened to assess the safety and e cacy of corticosteroids in septic patients. However, the results of the most recent reviews remain controversial [6,7]. Subsequently, several studies have further assessed whether the combination of corticosteroids, vitamin C, and thiamine as compared with corticosteroids or placebo, improved the survival duration, increased the vasopressor-free time over 7 d, and reduced organ injury [8,9]. These results suggest that the use of corticosteroids in combination with other drugs did not affect the safety and e cacy of corticosteroids on septic patients. Hence, resolution of this controversy regarding the latest reviews that have assessed the e cacy of corticosteroids on septic patients is currently the primary problem in the treatment of sepsis. Therefore, this systematic review and meta-analysis were performed based on the latest reviews to re-integrate the relevant data to evaluate the effects and safety of corticosteroids in septic patients.

Methods
The protocol of this systematic review and meta-analysis was registered on INPLASY (ID: INPLASY2020110122). The methodology of this study was according to items of the Cochrane Collaboration, and each outcome was assessed by the Grading of Recommendations Assessment, Development and Evaluation (GRADE) guidelines [10].

Study Searches
We performed this meta-analysis based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria, and we showed the PRISMA 2020 checklist in eTable 1. We searched the PubMed, Embase, and Cochrane library databases for relevant data from inception to November 2020, March 2021 to identify RCTs that have evaluated the effect of corticosteroids on septic patients. The MeSH/Emtree and title/abstract keywords combination was used to identify the eligible articles; the keyword search terms used for the English literature included the words corticosteroids and sepsis (the detailed search strategy in eTable 2 in the Supplement). It is noteworthy that we also conducted a manual search for the references of the relevant articles (the study search owchart in Figure 1).

Study Selection
Before the potential articles were searched and screened, the eligibility criteria and excluded criteria were identi ed. Articles may be eligible accorting ro the criteria of inclusion in this study, if they meet all of the following conditions: (1) the adult patients diagnosed with sepsis, severe sepsis, or septic shock, as per the accepted criteria during the study [11][12][13]. We included studies reported the adult patients with acute respiratory distress syndrome (ARDS), and of which reported the adult septic patients; (2) the study compared the use of corticosteroids (including hydrocortisone, methylprednisolone, betamethasone, udrocortisone, and dexamethasone) with no use of corticosteroids; (3) the study measured and reported the outcomes in terms of 28-day and long-term mortality (> 90 d), ICU mortality, in-hospital mortality, short-time mortality, length of stay in hospital and ICU, vasopressor-free days, ventilation-free time, shock reversal at day 7 and day 28, time to resolution of shock, hypernatremia, hyperglycemia, superinfection, and gastroduodenal bleeding. (4) the study was a RCT and was published in English. Furthermore, the study design including case reports, case series, and observational studies were excluded. All the available articles were searched by two searchers respectively, and when the disagreements occurred during the proceed, the third investigator should be resolved these disagreements. Reviewers performed reviews in pairs to screen all relevant citations and references as per the search strategy, and the screening process included the following two stages: initial evaluation of the titles and abstracts and skimming of the full text to identify the eligible studies.

Data Extraction
Researchers conducted data extraction respectively and in duplicate based on the eligibility criteria and excluded criteria. In case of disagreements, the third reviewer resolved the issue. Relevant data, including the study title, rst author, study type, study period, the therapy in treatment group and control group, reported outcomes, sepsis de nition and so on were collected. The data only for the studies we searched including the previous review[6] was abstracted. We assessed the risk of bias for this meta-analysis by two investigators independently for every abstracted data of each articles based on the Cochrane Collaboration [14], and domains were including allocation concealment, blinding of participants and staff, blinding of outcome assessors, incomplete outcome data, selective outcome reporting, and other biases. Additionally, the GRADE framework was used to evaluate the overall evidence rank for every outcome [15].
The studies with more than six, four to six, and fewer than six items were considered high, fair, and poor quality, respectively. Importantly, we used the GRASE to assess the evidence rank of mortality and adverse events. According to the risk of bias, inconsistency, indirectness, imprecision, and publication bias to evaluated the studies as low, moderate, or high quality.

Statistical Analyses
M-H or DL methods with random-effects meta-analyses were conducted for the eligible RCTs. All the relevant data were assessed using the Review Manager (RevMan), version 5.3 (Cochrane Collaboration), STATA 16.0 (StataCorp, College Station, TX). Risk Ratio (RR) and mean difference (MD) were used to present the dichotomous outcomes and continuous outcomes, with 95% CI. Moreover, a funnel plot was used to examine the potential for some small effects if the outcome included more than 10 trials, and the possibility of publication bias was assessed using funnel plot and regression test of the Egger test [16]. The chisquare test, I 2 , and visual inspection of the forest plots were used to evaluate heterogeneity among the eligible studies; when I 2 was > 50%, the heterogeneity was considered substantial. In addition, we performened the subgroup analyses based on the following variables: sepsis subtype [sepsis, septic shock, sepsis and ARDS, sepsis and community-acquired pneumonia, severe COVID-19]; type of corticosteroids (hydrocortisone or hydrocortisone plus udrocortisone or methylprednisolone or prednisone or betamethasone or dexamethasone), searching the source of heterogeneity.

Characteristics of Eligible Studies
We initially identi ed 9,363 records, and 8971 citations remained after the duplicate trials were removed; 145 RCTs were eligible after preliminary screening by title and abstract. Finally, Fifty-three RCTs [17-69] that included 12,310 patients with sepsis were included in this meta-analysis ( Fig. 1). The characteristics of the included RCTs are listed in Table 1 Fig. 4) with low heterogeneity (I 2 = 11% and I 2 = 22%, respectively). The Funnel plot and Egger test showed no publication bias in the 28-day mortality (P = 0.15), in-hospital mortality (P = 0.01), and ICU mortality (P = 0.02) (Supplemental Fig. 1-3 in the Supplement). The results of sensitivity analysis showed that the models of the 28-day mortality, in-hospital mortality, and ICU mortality were credible (Supplemental Fig. 4-6 in the Supplement). Furthermore, L'Abbé plot reported that the mortality in the placebo group increased signi cantly than the corticosteroid group, suggesting the potential effects of corticosteroids in septic patients (Supplemental Fig. 7-9 in the Supplement).

Secondary Outcomes
Supplemental Fig. 10 The Funnel plot and Egger test showed no publication bias in the length of stay in hospital (P = 0.53), SOFA scores at day 7 (P = 0.82), hyperglycemia (P = 0.14), the shock reversal at day 7 (P = 0.05), length of stay in ICU (P = 0.15), superinfection (P = 0.03), and gastroduodenal bleeding (P = 0.97) (Supplemental Fig. 24-30 in the Supplement). Shock reversal at day 28 showed a signi cant publication bias in the Funnel plot and Egger test (P = 0.046) (Supplemental Fig. 31 in the Supplement). The results of the sensitivity analysis showed that the models of the above mentioned outcomes, including length of stay in hospital, SOFA scores at day 7, hyperglycemia, shock reversal at day 7, length of stay in ICU, superinfection, gastroduodenal bleeding, and shock reversal at day 28 were credible (Supplemental Fig. 32-38 in the Supplement).
Importantly, risk of bias were reported in the forst plot of each outcome, and the evidence rank was showed in Table 2.

Subgroup Analysis
We performed subgroup analysis based on the sepsis subtype or type of corticosteroids used for the primary outcomes or I 2 > 75% in the secondary outcomes with more than 10 trials for each outcome. The results of the subgroup analysis showed no effect on the 28-day mortality; however, the in-hospital and ICU mortality were signi cantly improved in the hydrocortisone plus udrocortisone treatment and in the patients with septic shock, sepsis, and communityacquired pneumonia (Supplemental Fig. 39-44 in the Supplement). Moreover, the result of subgroup in SOFA scores at day 7 represented that the main original heterogeneity may be from the trials with smaller samples who were given hydrocortisone treatment or trials on patients with sepsis shock (Supplemental Fig. 45-46 in the Supplement).

Discussion
This meta-analysis included 53 RCTs (12,310 patients) and demonstrated that corticosteroids failed to improve the 28-day mortality and long-time mortality; however, there was a small reduction in the in-hospital mortality and ICU mortality. To our knowledge, this systematic review and meta-analysis is the most comprehensive review on many new RCTs, the precision of the pooled effect estimates on sepsis would be increased substantially.
We found that corticosteroids therapy for sepsis increased the incidence of the vasopressor-free days, ventilation-free time, shock reversal at day 7 and day 28, and adverse events, such as hyperglycemia and hypernatremia. Corticosteroids was associated with a decreased risk of the time shock resolution and length of stay in hospital. However, our study failed to report an decrased risk of corticosteroids on the length of ICU stay and adverse events, such as superinfection and gastroduodenal bleeding. Ascertainment of the adverse events in the eligible trials was also vulnerable, which may induce the evidence rank was low.
Subgroup analyses in this reviwe showed that the results did not identify any credible effect modi cation in sepsis subtype and type of corticosteroids used.
Many evidence comes from the trials with hydrocortisone or methylprednisolone treatment. Our results of the subgroup analyses showed that the e cacy of corticosteroids on in-hospital, ICU, and short-time mortality was mainly due to the hydrocortisone plus udrocortisone.
Mechanistically, Corticosteroids could inhibit the NF-κB activation and the extensive in ammatory factors release, nally, improving the in ammatory response of sepsis or pneumonia. Our previous studies reported that corticosteroids was associated with a decreased risk of ARDS and length of the disease in CAP patients [70]. Previous reviews have assessed the e cacy and safety of corticosteroids in septic patients. Unfortunately, the conclusions were contradictory and were incretical owing to the small number of trials included. One meta-analysis included 20 RCTs and showed no reduction in the 28-day, hospital, and ICU mortality in patients with severe sepsis and sepsis shock on corticosteroids treatment [71]. Subsequently, a Cochrane systematic review further to search the effect of corticosteroids on mortality of patients with sepsis, including a total of 33 RCTs, which found a small reduction in 28 d mortality under the corticosteroids treatment [72]. Simultaneously, another study included 35 RCTs and showed a converse result that corticosteroids failed to decrease the mortality [73]. In 2018, Rochwerg et al. [7] examined 42 RCTs including 10,194 patients, wherein corticosteroids achieved no reduction in the short-term (28-31 d) mortality, and may achieve a little effect on the long-term mortality. In 2019, Fang et al. [74] included 37 RCTs; this trial suggested that corticosteroids use was associated with a decrease in the 28-day mortality, ICU mortality, and in-hospital mortality. In parallel, Annane et al.
[6] published a Cochrane systematic review on 40 RCTs and achieved a reduction in the 28-day mortality in patients with sepsis on corticosteroids therapy.
The results of this meta-analysis showed that corticosteroids treatment failed to improve the 28-day mortality, in contrast with results from previous metaanalysis. The difference in part maybe due to the result reported by Annane et al.
[6] as per which corticosteroids therapy showed an increased risk of 28-day mortality, while the CI contained the null effect line, suggesting corticosteroids had no effect on sepsis based on the statistics in fact. More importantly, we included 4 RCTs about the sever COVID-19 and showed there was no signi cant difference in 28-day mortality with corticosteroids use. The data were extracted from the latest RCTs and may have helped in reinforcing the conclusions, decreasing the heterogeneity among the studies, and improving precision with more comprehensive assess for the therapy effects of corticosteroids treatment.
This meta-analysis has several strengths. First, this study is the most comprehensive trial to assess the e cacy of corticosteroids treatment on patients with sepsis to date. Second, we performed a thorough literature search including unpublished sources, using the GRADE methodology to evaluate the evidence rank in overall RR, a prede ned illustration of potential effect variables including direction of effect and subsequent subgroup analysis to search the effect variables, and illustration include the relative and absolute effects. Third, the primary outcomes showed low or no heterogeneity among the studies, suggesting that the results were not variable. Furthermore, the heterogeneity of the SOFA scores on day 7 was high, and the subgroup analysis showed that the source of heterogeneity may be the inclusion of trials with small size on patients who were given hydrocortisone treatment. Finally, the results of the sensitivity analysis for this study suggested that this conclusions were robust and reliable.
However, this meta-analysis also has certain shortcomings, including the signi cant methodological or clinical heterogeneity among the included studies, especially with respect to the SOFA score on day 7. All the included RCTs enrolled patients with sepsis as per the previous sepsis de nition criteria; however, we do not know the e cacy and safety of corticosteroids whether change using the Sepsis-3 de nition criteria. Hence, the de ned mortality may be essential, but the certainty is limited due to the imprecision of the included studies.

Conclusions
This is the most comprehensive systematic review and meta-analysis to describe the e cacy and safety of corticosteroids for sepsis patients. The ndings demonstrate that corticosteroids failed to reduce the 28-day mortality and long-term mortality; however, they could reduce the in-hospital mortality and ICU mortality. Importantly, our subgroup analyses results indicated that this e cacy of corticosteroids in sepsis patients may be associated with the hydrocortisone plus udrocortisone treatment. Therefore, the results suggest that corticosteroids could not improve the 28-day mortality in adult patient with sepsis.
Abbreviations unite; SOFA, acute respiratory distress syndrome, ARDS; Grading of Recommendations Assessment, Development and Evaluation, GRADE; SOFA, Sequential organ failure.

Declarations
Consent for publication Not applicable.

Author Contributions
All the authors contributed importantly to the work presented in this article. TWS, HYL(Huoyan Liang) conceived the study. XFD, HYL(Hongyi Li) and GFS contributed to the data extraction. SH and XFD computed and eveluated the pooled outcomes. HYL and SH contributed to the study protocol and wrote the article. QCK and TWS revised the article. The QCK and TWS had full access to all of the data and the nal responsibility for the decision to submit this article for publication.

Availability of Data and Materials
The datasets used and analysed in the available RCTs are from the corresponding author on reasonable request.

Ethical Approval and Consent to Participate
Not applicable.

Con icts of Interest
The authors have no con icts of interest to disclose.  In-hospital mortality of patients with sepsis based on the corticosteroids treatment. The pooled effects in the forest plot was calculated by the M-H method with random-effects model.