The detrimental effects of intestinal injury mediated by inflammation are limited in cardiac arrest patients: A prospective cohort study

Background Ischaemic intestines could be a driver of critical illness through an inflammatory response. We have previously published reports on a biomarker for intestinal injury, plasma Intestinal Fatty Acid Binding Protein (IFABP), and inflammatory biomarkers after out-of-hospital cardiac arrest (OHCA). In this post-hoc study we explored the potential indirect effects of intestinal injury mediated through the inflammatory response on organ dysfunction and mortality. Methods We measured IFABP and twenty-one inflammatory biomarkers in 50 patients at admission to intensive care unit after OHCA. First, we stratified patients on median IFABP and compared biomarkers between “low” and “high” IFABP. Second, by causal mediation analysis, we assessed effects of IFABP through the two most important inflammatory biomarkers, interleukin (IL)-6 and terminal complement complex (TCC), on day two circulatory variables, Sequential Organ Failure Assessment (SOFA)-score, and 30-day mortality. Results Cytokines and complement activation were higher in the high IFABP group. In mediation analysis, patients on the 75th percentile of IFABP, compared to the 25th percentile, had 53% (95% CI, 33–74; p < 0.001) higher risk of dying, where 13 (95% CI, 3–23; p = 0.01) percentage points were mediated through an indirect effect of IL-6. Similarly, the indirect effect of IFABP through IL-6 on SOFA-score was significant, but smaller than potential other effects. Effects through IL-6 on circulatory variables, and all effects through TCC, were not statistically significant and/or small. Conclusion Effects of intestinal injury mediated through inflammation on organ dysfunction and mortality were limited. Small, but significant, effects through IL-6 were noted. Trial registration: ClinicalTrials.gov: NCT02648061.


Introduction
After being initially successfully resuscitated from cardiac arrest, twothirds of patients suffer from severe multiple organ dysfunction.2][3] While two-thirds die of neurological injury overall, cardiovascular dysfunction is the most common cause of death the first three days. 4The mechanisms leading from whole-body ischaemia to multiple organ dysfunction and death have not been determined, but the immune system is suggested to play a central role. 5Moreover, intestinal ischaemia might be a "driver" of the subsequent inflammatory response and multiple organ dysfunction. 6,7Indeed, when and how intestinal injury leads to multiple organ dysfunction are highlighted as research questions by the European Society of Intensive Care Medicine. 8][17][18][19] Previously, we have reported that both IFABP and some of the https://doi.org/10.1016/j.resplu.analysed inflammatory biomarkers, including interleukin (IL)-6 and the terminal complement complex (TCC), were associated with multiple organ dysfunction and mortality after out-of-hospital cardiac arrest (OHCA). 20,21ausal mediation analysis is an emerging statistical method which contributes to new insights into pathophysiologic processes. 22he method combines conventional regression models and enables a decomposition of the total effect of an exposure on outcome into an indirect effect through a mediator and a direct effect which is not attributed to the mediator.To draw firm causal conclusions, rigorous assumptions are required. 23,24he aim of this study was to explore, by causal mediation analysis, the potential indirect effects of intestinal injury mediated through the inflammatory response on organ dysfunction and mortality after out-of-hospital cardiac arrest (OHCA).

Study design and setting
This is a post hoc analysis of a prospective cohort included at St. Olav's University Hospital, Norway.The inclusion period was January 2016 to November 2017.In addition to the inflammatory biomarkers (including IL-6 and TCC) and intestinal injury (IFABP), reports on circulatory characteristics in this cohort have been published previously [20][21][25][26] . The AGeMA Statement for mediation analysis reporting published in JAMA, guided this study report.27

Participants
Inclusion criteria were successfully resuscitated patients admitted to the Intensive Care Unit (ICU) after OHCA.Exclusion criteria were age < 18 years, pregnancy, sepsis or anaphylaxis as assumed causes of arrest, transferal from other hospitals, decision upon arrival to limit life-sustaining therapy, or the following before admission to the ICU: application of extracorporeal membranous oxygenation (ECMO), or a ventricular assist device (VAD), or cardiothoracic surgery.Patients were followed for five days or until they were transferred to ward or died, ECMO or VAD was applied, cardio-thoracic surgery was performed, or withdrawing of life-prolonging therapies was decided.Day "zero" lasted from admission until next morning at 06:00, and thus had variable length.

Initial management
Targeted temperature management (TTM) at 36 °C for 24 hours was the standard treatment of comatose patients.Patients presenting with clinical signs of hypoperfusion and/or hypotension were given fluids, vasopressors and/or inotropic drugs.If indicated, percutaneous coronary intervention was done.Comprehensive information regarding the clinical care has been reported earlier. 28

Data sources and measurement
We gathered data from the pre-hospital report in accordance with the Utstein cardiac arrest template, and information on assessment and treatment from the hospital record. 29We scored Sequential Organ Failure Assessment (SOFA) daily. 30Lactate was obtained from the first arterial blood gas sample after admission.
In all comatose patients without contraindications, we inserted a pulmonary artery catheter (Swan-Ganz CCOmbo, Edwards Lifesciences, Irvine, CA) for continuous haemodynamic measurements.We obtained all circulatory variables and medications from the elec-tronic critical care information system (Picis CareSuite, Optum Inc. Peabody, MA).The circulatory variables were mean arterial pressure (MAP), cardiac output (CO), systemic vascular resistance (SVR), dose of noradrenaline (norepinephrine), and amount of fluid infusion.MAP, CO, SVR and dose of noradrenaline presented are the mean values over the period beginning 30 minutes before each blood sampling and ending 30 minutes after.Amount of fluid given is the mean for the previous 24 hours.We obtained 30-day mortality from the medical records and the Norwegian death registry.

Blood sampling and plasma preparation
Blood samples were drawn at inclusion and every morning, gently mixed, placed vertical in ambient temperature for 30 minutes and then centrifuged at 2200g for 10 minutes.Within 1 hour from sampling, EDTA-plasma was frozen to À80 °C.

Intestinal Fatty Acid Binding Protein (IFABP), endothelium and platelet markers
Plasma levels of IFABP, syndecan-1, vascular endothelial (VE)cadherin and von Willebrand factor (vWF) were measured in duplicate by enzyme immunoassays (EIA) using commercially available antibodies (Cat# DY3078, R&D Systems, Minneapolis, MN) in a 384-format using a combination of a CyBi-SELMA pipetting robot (Analytik Jena, Germany) and an automatic dispenser/washer (Bio-Tek, Winooski, VN).Absorption was registered at 450 nm with wavelength correction set to 540 nm applying an ELISA plate reader (BioTek).

Statistical methods
Descriptive statistics are presented as mean (standard deviation [SD]), median (interquartile range [IQR]), or proportion (%).Of twenty-one biomarkers in the previous report, biomarkers consistently associated with outcome were included in the present (TCC, C3bc, IL-1ra, IL-8, IL-6, RANTES, Syndecan-1, VE-cadherin and vWF), and only the two variables associated with mortality in the multivariable analysis (IL-6 and TCC) were included in the mediation analyses. 20We used admission values of all biomarkers.IL-6 had one outlier which was changed to "missing" and omitted in the analyses.
To assess the associations between inflammatory biomarkers and intestinal injury, we divided the cohort in "low" and "high" by the median IFABP.Biomarkers were compared between the two groups using the Student t-test or the Wilcoxon rank-sum test.We displayed survival and the values of biomarkers first five days graphically.

Mediation analyses
To assess the potential effects of IFABP through inflammatory biomarkers on organ dysfunction and death, we applied causal mediation analysis. 22,24This approach is based on the potential outcomes framework and is implemented in Stata (StataCorp.2023.Stata 18 Causal mediation analysis Manual.Stata Press, College Station, TX).The potential total effect of an exposure on outcome is the sum of the natural indirect effect through a mediator and the natural direct effect, in the text simply named the "indirect" and "direct" effect.To make valid causal conclusions, these assumptions also must be fulfilled: There must be no unmeasured confounding in the exposure-outcome, the exposure-mediator, or the mediatoroutcome relationships.In addition, there must be no confounders in the mediator-outcome relationship caused by the exposure.The mediation analyses, with assumed causes and effects, are presented in Fig. 1.
We applied two mediation models using organ dysfunction (Model 1) and 30-day mortality (Model 2) as the outcomes.The indirect effect was our main interest.In both models IFABP was set as the exposure, and IL-6 and TCC were log 2 transformed and entered one by one as mediators.Organ dysfunction at start of day two was evaluated by six outcome variables separately (SOFA-score, SVR, MAP, CO, dose of noradrenaline and amount of fluid given).Linear regression was applied for both the mediator and outcome functions in Model 1.In Model 2, we applied a probit-function for the dichotomous outcome.The effects (total, direct and indirect) were expressed as the regression coefficients (Model 1) and risk difference (RD) (Model 2) between the 75th percentile of IFABP, compared to the 25th percentile, with 95% confidence intervals (CI).
In Model 2, with IL-6 as the mediator, we included covariates that were plausible confounders (i.e. could affect both the exposure-mediator association and the mediator-outcome association). 24These covariates were initial non-shockable rhythm, time to return of spontaneous circulation (ROSC) and initial lactate at hospital admission.We added the covariates one at a time because of the low sample size.The Stata commands are displayed in Supplementary Table 1.

Sensitivity analysis
We performed several sensitivity analyses.First, we exchanged admission levels for day one levels of IL-6.Second, we applied two logistic regression models with IFABP, IL-6 and TCC as indepen-dent variables.In the first regression model, 30-day mortality was evaluated with confounding factors added one at a time, and in the second by excluding patients who were considered to die of irreversible brain injury. 21his was a descriptive study, and no sample size was calculated. 28All tests were two-sided and statistical significance was defined as p < 0.05.We did not adjust for multiple testing.Data were extracted with the software Matlab (Mathworks Inc., Natick, MA), and the statistical analyses were performed with Stata 18.0 (StataCorp LCC, Collage Station, TX).

Ethics
The Regional Committee for Medical and Health Research Ethics, Central Norway Health Region (REK Midt, No. 2015/1807) approved the study.Participants or their proxies provided written consent.The study is registered in ClinicalTrials.gov(Identifier: NCT02648061).

Participants
Sixty-five consecutive patients were assessed for eligibility; seven were excluded due to immediate withdrawal of life-support, two had septic etiology, two were not in need of ICU admission, three received VAD or ECMO and one underwent immediate surgery.Fifty patients were included in the study.Patient characteristics are shown in Table 1.Detailed demographic data have been published previously. 21,26At start of day two (mean 36 hours after admission), seven patients had died, all in the "high" IFABP group (Fig. 2), and six patients had been transferred to ward, all in the "low" IFABP group.

Association between IFABP and inflammatory biomarkers at admission
Median of IFABP was 31.4 (IQR, 21-38) ng/mL at admission, rapidly declining thereafter (Fig. 2). 21The high IFABP group Fig. 1 -The two mediation models.IFABP is the exposure and inflammatory biomarkers are mediators in both models.Organ dysfunction is the outcome in Model 1 and 30-day mortality is the outcome in Model 2. Potential confounding factors such as whole-body ischaemia could affect all variables.IFABP: Intestinal Fatty Acid Binding Protein.IL: Interleukin; TCC: terminal complement complex.
The estimated indirect effects of IFABP through IL-6 and TCC on circulatory variables were not statistically significant and/or small (Supplementary Table 2).
The total effect was 0.53 (RD; 95% CI, 0.33-0.74;p < 0.001) and the direct effect was 0.41 (RD; 95% CI, 0.19-0.62;p < 0.001).The effects were reduced when we added the covariates "time to ROSC" and "non-shockable rhythm" to Model 2 (Supplementary Table 3).When lactate was added as a covariate, estimates could not be computed due to non-convergence of the calculations.

Sensitivity analyses
When IL-6 at day one was exchanged for IL-6 at admission, the indirect effect through IL-6 on mortality was lower and not significant.In multivariable logistic regression, using death within 30 days as the dependent variable, IFABP was associated with outcome in both models, but IL-6 and TCC in none (Supplementary Tables 4-6).

Discussion
In this post-hoc study, IFABP was associated with cytokines and complement activation at admission, but not with other inflammatory biomarkers.Only the two biomarkers most consistently associated with mortality, IL-6 and TCC, were included in the causal mediation analyses. 20,24Compared with the direct effects of intestinal injury on SOFA-score and mortality, the indirect effects mediated through IL-6 were significant, but substantially smaller than the direct effects.
The indirect effects through IL-6 on circulatory variables, as well as all indirect effects through TCC on circulatory variables, SOFAscore and mortality, were not statistically significant and/or small.
We applied causal mediation analysis to explore if detrimental effects of intestinal injury could be mediated through inflammation.Thus, the direct effect of the exposure on outcome should not be interpreted as a causal effect, rather potential effects not attributed  to the mediator.Further, we cannot firmly conclude on causal indirect effects, since the assumption of no unmeasured confounders is not likely to hold. 24Conversely, after analysing twenty-one inflammatory biomarkers and selecting the two biomarkers most consistently associated with outcome, few indirect effects were significant, and these were substantially smaller than the direct effects. 20When we added possible confounders, the indirect effects were reduced.Therefore, an important inflammatory response caused by intestinal injury seems unlikely in this cohort.Intestinal ischemia may be a clinical concern related to low-flow and high-dose epinephrine given during resuscitation. 9,324][35] The presented findings and our previous paper explore the relationship between intestinal ischemia and organ failure and mortality. 21This relationship is not fully understood and may indicate that more underlying information is needed before we start new clinical trials.This is supported by the fact that clinical trials investigating the use of the IL-6 inhibitor tocilizumab and prehospital methylprednisolone had no effect on survival compared to placebo. 36,37ausal mediation analysis combines regression models, and the limited mediated effects could be because of weak associations between IFABP and inflammatory biomarkers, between inflammatory biomarkers and outcome, or both. 24To answer this was not part of the study, but we found associations between IFABP, IL-6 and TCC.0][41][42][43][44] Importantly, IFABP and inflammatory biomarkers could merely increase in parallel with the degree of whole-body ischaemia. 20,21here are a few caveats with this study.First, small indirect effects through separate biomarkers could add up to a large effect in total, although IL-6 and TCC are central in the inflammatory response. 5,20Second, perhaps only the most severely injured intestines cause a sustained inflammatory response. 45However, after excluding patients who died of irreversible brain injury, leaving patients with extra-cerebral organ dysfunction as assumed cause of death, inflammatory biomarkers were still not associated with mortality, while IFABP was.4][46][47][48][49] In the sensitivity analysis, the effect through IL-6 at day one was less pronounced than through IL-6 at admission, suggesting against a delayed inflammatory response.
The concept of "gut origin sepsis" denotes intestinal injury leading to a "sepsis-like" syndrome in the critically ill patient. 6The term "sepsis-like syndrome" has also been applied on patients after cardiac arrest who share some features with septic patients, such as vasodilation, intravascular volume depletion, and endothelial injury. 5,49We found limited effects of IFABP through inflammatory biomarkers on SOFA-score and variables reflecting vasodilation (SVR, dose of noradrenaline and MAP) and volume depletion (fluid infusion), while markers of endothelial injury were not consistently associated with outcome and excluded from the analysis. 20Thus, Fig. 3 -Effects of IFABP and IL-6 on 30-day mortality in causal mediation analysis (Model 2).The total effect (TE) of IFABP on mortality is the sum of the indirect effect (IE) of IFABP through IL-6 and the direct effect (DE).The reported effects are risk differences between the 75th percentile of IFABP (38 ng/mL), compared to 25th percentile (21 ng/ mL).IFABP: Intestinal Fatty Acid Binding Protein; IL: Interleukin; CI: Confidence interval.
our findings indicate that intestinal injury did not lead to a "sepsis like syndrome" in this cohort, and contrast with the concept of "gut origin sepsis".
In summary, a limited mediated effect of intestinal injury through IL-6 on organ dysfunction and death is possible.Our findings suggest, however, that intestinal injury is not a "driver" of critical illness acting through an inflammatory response after cardiac arrest. 7,8

Limitations
The main limitations of this study are the post hoc design, single center study, and the small sample size.Specific features of this cohort, such as a high proportion of bystander CPR, could limit the generalizability.Further, the validity of IFABP has not been established in the cardiac arrest population, and although lL-6 is extensively studied, different analytical methods can make comparisons challenging.Lastly, the immune system is complex, there may be an interplay between different organ failures, and other factors not studied in this analysis may be important.

Conclusion
To our knowledge, this is the first report on the relationship between intestinal injury and inflammation after cardiac arrest.By causal mediation analysis, effects of intestinal injury mediated through the inflammatory response on organ dysfunction and mortality were limited.Small, but significant, effects through IL-6 were noted.

Fig. 2 -
Fig. 2 -Survival and trajectories of selected biomarkers.Biomarkers are expressed as means by "low" or "high" IFABP, at the start of the first five days.Only patients still treated in ICU are included (n = 50, 45, 38, 31, 27 and 25 for day 0-5, respectively).Day one started mean 11 hours after admission.OHCA: Out-of-hospital cardiac arrest; IFABP: Intestinal Fatty Acid Binding Protein; IL: interleukin; RANTES: regulated on activation normal T-cell expressed and secreted; TCC: terminal complement complex.

Table 1 -
Characteristics of successfully resuscitated OHCA patients.

Table 2 -
Biomarkers on admission of successfully resuscitated OHCA patients.