Introduction

Autoimmune diseases are a group of diseases in which the immune system mounts an immune response against its own normal tissue components, often resulting in chronic tissue and organ damage, affecting approximately 7.6–9.4% of the global population [1]. The primary features of autoimmune diseases include the production of self-targeting antibodies and abnormalities in the function of immune cells. Often, the management of these conditions involves the use of immunomodulatory or immunosuppressive medications, which can result in compromised immune function and an elevated risk of infections [2]. Although retrospective analyses of autoimmune diseases have primarily associated patients with respiratory infections, it is important to highlight that the main drivers of ICU admissions and mortality in this group are severe infections [3, 4]. The evolving environmental changes brought about by societal industrialization have contributed to an increasing incidence of autoimmune diseases. Consequently, the associated risk of pathogenic infections is expected to rise as well [5, 6]. Therefore, prioritizing infection risks in individuals with autoimmune diseases is crucial for mitigating the emergence of life-threatening infectious conditions.

Sepsis is a complex, infection-induced systemic inflammatory response disorder characterized by an imbalance, often accompanied by acute organ dysfunction and a high mortality rate [7]. Despite a decline of 37.0% in the age-standardized incidence of sepsis and a 52.8% decrease in mortality, the burden of this severe condition persists, particularly in develo** countries [8, 9]. It is noteworthy that the increased overall burden could be attributed to severe infections resulting from autoimmune diseases and their associated treatments, such as the use of corticosteroids [10]. Some patients with autoimmune diseases were admitted to the intensive care unit at the initial diagnosis [11,12,13,14]. Among these cases, sepsis (severe infection) stands out as the primary cause of ICU mortality, followed by acute disease exacerbations [15]. The relationship between autoimmune diseases and sepsis has long been a subject of interest [2]. Therefore, we extracted information on patients with autoimmune diseases from the MIMIC-IV database to explore whether autoimmune diseases increase occurrence of sepsis and the 28-day mortality of sepsis. However, due to limitations in retrospective research, such as potential confounders and selection bias, a consistent conclusion regarding the relationship between autoimmune diseases and sepsis has not been reached [

Methods

Mendelian randomization

Study design and genetic instrument selection

Figure 1 shows the study design and the assumptions of MR in our study [20]. We used publicly available summary statistics from Genome-wide association study (GWAS) sources of predominantly European origin. All studies had current ethical clearance from their respective institutional review boards, including written informed consent from participants and strict quality control. As all analyses herein are based on publicly available summary data, no ethical approval from institutional review boards was required for this study. Three basic assumptions are required for the genetic variants to qualify as valid instrumental variables (IVs): (1) they should be robustly associated with the exposure; (2) they should not be associated with potential confounders of the exposure-outcome association; and (3) they should not influence the outcome by any variable other than the exposure [20]. To validate the initial MR hypothesis, we utilized independent single nucleotide polymorphisms (SNPs) that exhibited a robust association with the exposure, reaching genome-wide significance (P < 5 × 10–6). These SNPs were carefully chosen to ensure minimal linkage disequilibrium (r2 < 0.01) within a clump window larger than 5000 kb, thus ensuring their independence. If we follow the same inclusion criteria, the exposure of reverse MR analysis includes too few SNPs, so we have adopted the following criteria. The reverse MR analyses inclusion criteria for the instrumental variable SNP were as follows: P < 1 × 10–5, r2 < 0.001 within a clump window larger than 10,000 kb. To further refine the first hypothesis, we quantified the proportion of phenotypic variation explained by the entire set of SNPs and assessed the strength of our instrumental variables using the F statistic (beta2/se2). An F-statistic exceeding 10 was considered indicative of a robust instrument [21]. R2 was calculated as beta2/[beta2 + se2*(N− 2)], N being the sample size, and the genetic variability explained by each SNP was calculated [22]. Finally, after eliminating palindromic SNPs, we proceeded to utilize the remaining selected SNPs as our instrumental variables for subsequent analyses. To delve into the direct influence of distinct autoimmune diseases on sepsis, we adopted a multivariate MR approach-an extension of the conventional univariate MR. This approach duly acknowledged the inherent interplay among SNPs used in MR analyses, often manifesting shared associations across different autoimmune conditions. In our study, the SNPs utilized for multivariate MR were formulated as combinations of instrumental variables per exposure, thereby accounting for the intricate web of associations (including those associated with phenotypes of at least two autoimmune diseases). This study is reported in line with the STROBE-MR guidance, with the checklist available in the Supporting information [23].

Fig. 1
figure 1

Overview and assumptions of the Mendelian randomization study design

Full size image

Data sources for exposures, mediators and outcomes

Recognizing that utilizing diverse populations could potentially lead to biased estimates, we constrained the genetic background of the population in the MR study to individuals of European ancestry [24].

Exposures included a total of 10 distinct autoimmune diseases, and we aligned our analysis with data available from the FinnGen consortium (R9) (https://www.finngen.fi/fi.) and the MRC-IEU online database (https://gwas.mrcieu.ac.uk/). The autoimmune diseases considered for inclusion in our analysis were as follows: systemic lupus erythematosus (SLE), ankylosing spondylitis, multiple sclerosis (MS), primary biliary cholangitis, rheumatoid arthritis (RA), Crohn's disease (CD), ulcerative colitis (UC), type 1 diabetes (T1DM), celiac disease, psoriasis. We systematically reviewed and summarized the general characteristics of each autoimmune disease, and we presented these aggregated data in Additional file 3: Table S1. It is important to note that the ten autoimmune diseases from the Finngen consortium were defined using the codes of the International Classification of Diseases (ICD-9) and ICD-10. In the FinnGen consortium, individuals with undefined sex, high genotype deletion (> 5%), excess heterozygosity (± 4 standard deviations ((SDs)), and non-Finnish ancestry were excluded. All genetic association effect sizes were calculated by logistic regression, and adjusted for age, sex, and genetic principal components [25]. In the MRC-IEU database, all 10 included diseases have been previously published online. However, due to their diverse origins from different research teams, the analytical methods employed and the controlled confounding factors are not entirely uniform. For a comprehensive understanding, please refer to the cited references for detailed information [26,27,28,29,30,31,32,62]. It is well known that Neutrophil extracellular traps (NETs) are one of the major factors contributing to the severity of septic disease, and that in the early stages of sepsis, depletion of NETs does not help to prevent or contain systemic infections, and even exacerbates pathological changes [63,64,65,66]. The citrullination of nuclear histones by PAD4 leads to chromatin depolymerization, which is a key step in the formation of NETs [67]. Therefore, the abnormal structure and function of PAD4 make RA patients more susceptible to sepsis. Chitinase-3 like-protein-1 (CHI3L1, other name YKL-40) may be another important protein molecule involved in the pathological process of RA and sepsis.YKL-40 is synthesized and secreted by a wide variety of cells including macrophages, neutrophils, and chondrocytes and plays an important role in tissue injury, inflammation, tissue repair and remodeling responses [68]. It was found that YKL-40 levels were significantly increased in RA patients and induced the expression of IL-1β and TNF-α, which were involved in the inflammatory response in RA [69]. A study by Kornblit et al. found that YKL-40 levels were also significantly elevated in patients with sepsis and that YKL-40 promoted the expression of inflammatory factors [70]. Thus, RA patients may be more susceptible to sepsis due to genetic variants, and the CHI3L1 genotype (rs4950928) may be a potential locus [69, 70].

In order to further explore the potential mediation of autoimmune diseases in the occurrence of sepsis through underlying intermediary factors, we included risk factors related to autoimmune diseases associated with sepsis [36]. We found that not all autoimmune diseases lead to a decrease in blood cell counts; only SLE, celiac disease, T1DM, and reduced blood cell count showed a causal association, whereas conditions such as RA and primary biliary cholangitis had less pronounced effects. Despite an inverse causal trend between blood cell counts and sepsis, statistical significance was lacking. While some observational studies suggest a predictive relationship between changes in blood cell counts and the risk of severe infection in autoimmune diseases [71, 75,76,77,78]. Hydroxychloroquine reduces the risk of severe infections in SLE patients, while the use of glucocorticoids, especially in high doses, is closely related to severe infections [79]. Montgomery et al. found functional impairment to be a significant risk factor for severe infections in multiple sclerosis patients [17]. Therefore, patients with autoimmune diseases require closer monitoring of organ function, comorbidities, medication usage, and other factors to reduce the risk of sepsis. It is worth noting that RA is associated with an increased risk of sepsis, which can occur early in the course of the disease. This suggests that when managing patients with RA, early attention, timely treatment, and early prediction may be required to reduce the occurrence of severe infections. For example, for outpatient patients with fever or other signs of infection, infection-related markers and imaging tests should be monitored, and they should be hospitalized if necessary; for hospitalized patients, early monitoring of biomarkers and symptoms of infection, more active adjustment of antibiotics after infection, and more close monitoring for patients with high-risk factors (such as, indwelling catheters and steroid use, etc.) should be noticed.

The 28-day mortality risk in sepsis is a crucial measure of disease severity, and whether autoimmune diseases increase this risk remains inconclusive. A retrospective study by Antón et al. found that autoimmune diseases often lead to a higher mortality rate in critically ill patients [4]. However, this study primarily predicted high mortality risk without correcting for concurrent confounding factors such as SOFA score, age, underlying diseases, and had a relatively small sample size. In our analysis using two-sample MR analysis, we inferred causal relationships between autoimmune diseases and sepsis at the genetic level. We did not find a causal association between autoimmune diseases and the 28-day mortality rate in sepsis. Similarly, in the retrospective analysis from MIMIC-IV, there was no observed relationship between autoimmune diseases and the 28-day mortality rate in sepsis. Therefore, we believe that autoimmune diseases do not increase the 28-day mortality rate in sepsis. This might be related to the immune dysregulation caused by autoimmune diseases, leading to imbalanced cytokines in the sepsis inflammatory cascade [80], making it difficult to form a cascading reaction. The early mortality in sepsis is closely associated with this inflammation storm. Jorge et al.'s observational study found that the risk of death in autoimmune diseases may be related to factors such as experiencing shock upon admission to the intensive care unit, having hemoglobin levels below 8 g/dL, using immunosuppressive agents before ICU admission, and having low complement C3 levels [81]. Additionally, the quality of care provided by hospitals is a key factor influencing patient mortality risk, with more experienced hospitals often having lower mortality rates [3]. Therefore, for autoimmune disease patients admitted to the ICU, it is crucial to focus on the management of complications while enhancing diagnostic and treatment capabilities specific to autoimmune diseases to reduce the risk of mortality.

A key feature of sepsis is the immune dysfunction triggered by infections, leading to prolonged alterations in immune function such as changes in immune cell functionality and numbers. Similarly, the immunopathological mechanisms of autoimmune diseases are accompanied by disruptions in immune function [39, 82]. Furthermore, infections caused by pathogenic microorganisms can act as triggering factors for autoimmune diseases [83]. However, it remains uncertain whether the immune dysfunction triggered by severe infections caused by pathogenic microorganisms could lead to the development of autoimmune diseases. Through reverse MR analysis, we identified a causal relationship between sepsis and psoriasis, but no associations with other autoimmune diseases, and current research has also found that infection is an important trigger for the occurrence of psoriasis [84]. The specific mechanisms underlying this relationship require further investigation. MR provides a novel method to discover associations between different diseases at the genetic level, offering a new perspective for future observational studies.

This study has several limitations in the MR analysis. First, potential horizontal pleiotropy is a concern in any MR study. In our research, we did not observe significant evidence of pleiotropic effects in all exposure analyses using the MR-Egger intercept test. Additionally, the MR-PRESSO analysis detected few outliers, and associations remained consistent after removing outlier SNPs. However, the possibility of undetected outliers still exists. Second, sample overlap might be a concern as we selected a subset of autoimmune diseases from the FinnGen consortium. Nonetheless, sample overlap is unlikely to bias our results significantly, given that our IVs were selected from large-scale GWAS. Third, due to the limited number of SNPs meeting the inclusion criteria for certain autoimmune diseases (P value < 5e− 08, R2 = 0.001 with kb = 10,000), we slightly relaxed the selection criteria, which may introduce a certain level of false positives. Fourth, genetic factors are not the sole determinants of autoimmune disease onset; environmental factors also play a role in triggering disease processes. Therefore, our MR analysis lacks associations between genetically predicted autoimmune diseases and sepsis risk, but this does not exclude the potential impact of autoimmune diseases on the pathophysiology of sepsis [39]. Fifth, the genetic associations of blood cell count, inflammatory cytokines, and are based on relatively small global genomic studies, potentially leading to issues of statistical power. Sixth, univariate MR analysis may not capture the direct impact of specific biomarkers on disease outcomes, as the effect of a biomarker might be mediated by other biomarkers within a complex network. Seventh, all our analyses are based on individuals of European ancestry; generalizability to other populations requires further investigation. In real-world retrospective studies, there are also limitations. First, our results could be influenced by diagnostic bias, where the severity of autoimmune diseases and the immunocompromised state of autoimmune disease patients might lead them to be admitted to ICUs earlier than other populations, potentially resulting in better survival rates. This selection process could lead to a higher incidence of autoimmune diseases in the ICU population. Additionally, we controlled for potential influencing factors, yet the overall confounding factors included in our analysis might not be exhaustive, such as other clinical scores that were not incorporated. We believe that further research with more diverse pre-ICU admission data from intensive care units would help fully eliminate diagnostic bias. Second, the MIMIC database is derived from a single-center research institution, which may limit the generalizability of the study outcomes. Third, due to the limitations of the database, we were unable to distinguish whether patients with autoimmune diseases were diagnosed after their first ICU admission, i.e., it was difficult to determine whether the autoimmune disease was a newly diagnosed disease or a comorbidity, although this group of patients may be rare. Fourth, because Mendelian randomization uses integrated data, whereas observational studies use individual data, it is difficult for us to achieve correction for the same confounders for the two different analytical methods.

Conclusion

In this study, our aim was mainly to assess the association of autoimmune diseases with the development of sepsis and 28-day mortality through a MR and observational study. In our results, genetically predicted RA was independently associated with the development of sepsis. We did not find that none of the other autoimmune diseases predicted by genes were independently associated with the development of sepsis, including subsequent mediation analyses. In addition, neither observational studies nor MR analyses found autoimmune diseases to be associated with 28-day mortality from sepsis., Surprisingly, there was a causal relationship between genetically predicted sepsis and the development of psoriasis.