Prognostic impacts of repeated sepsis in intensive care unit on autoimmune disease patients: a retrospective cohort study

To the best of our knowledge, this is the first retrospective cohort study focusing on the impact of repeated in-ICU sepsis on the survival of AD patients using the MIMIC database. We found that repeated in-ICU sepsis acted as a risk factor for 1-year overall-cause mortality and septic shock of AD patients. Furthermore, higher scores in Max SOFA, CCI and SAPS-II were also closely associated with poor prognosis among AD patients with sepsis.

Firstly, repeated in-ICU sepsis episode was significantly associated with higher 1-year overall-cause mortality among AD population, along with robust adjusted HRs in sensitivity analysis by the adjustment of confounders. The findings confirm that repeated in-ICU sepsis served as an independent risk factor for the long-term survival of AD individuals and its effect was less influenced by other covariates. Partly similar to our study, Pandolfi et al. [9] found that hospital readmission due to recurrent sepsis was associated with an increased risk of 1-year mortality. Likewise, some previous studies showed that readmission for recurrent sepsis was indeed a risk factor related to a higher mortality rate [18, 19]. Whereas, these studies on repeated sepsis rarely involved AD population. Other studies on ADs found that sepsis was an important factor influencing the mortality of ADs patients [5, 20], but they did not give a specific definition of whether the sepsis they studied was new-onset or recurrent. These studies provided some clues of the impact of sepsis on ADs, and our findings regarding the long-term prognosis of ADs in relation to recurrent sepsis helps made up for the limitations of these studies.

It remains unclear about the underlying pathogenesis of the increased vulnerability of AD with repeated sepsis to a poorer long-term prognosis; however, immune dysfunction may provide a potential explanation. Sepsis is a disorder as a result of the dysregulated host response to an infection, beginning with inflammation in the early phase and converting into an immunosuppression in the late phase [6, 7]. Previous studies showed that recurrent sepsis or post-sepsis state could induce the exhaustion and impaired response of CD4⁺ T cell and memory CD8⁺ T cell [21‐23], along with the increase of regulatory T cells and myeloid derived suppressor cells that had T cell inhibiting capabilities [22, 24, 25]. Another study also identified HLA-DR^lowS100A^highmonocytes as a cell subtype enriched in the sepsis patients, which was also correlated to the immunosuppression [26]. In addition to the alteration of cell populations, immunosuppressive cytokine secretion and increased expression of negative costimulatory molecules were also found in the long-term immunosuppression status of sepsis [27]. The alteration of immunity might successively cause infection susceptibility and recurrent sepsis episodes. As such, the prolonged immunosuppression status and the recurrence of sepsis formed a bidirectional causality, perpetuating a detrimental loop and eventually causing organ dysfunctions even deaths. For AD populations in our study, the sepsis-induced immunosuppression might be also dysregulated and exacerbated by the ADs; in the meanwhile, the AD-associated multi-organ damage might coexist with the organ dysfunction induced by the sepsis. The underlying pathogenesis on the association between ADs and sepsis episodes needs more studies in the future.

Another important finding was that tools reflecting severe physical dysfunction were closely linked to AD patients with in-ICU sepsis. SOFA, CCI and SAPS-II, three commonly-used quantified assessment tools in ICU, were considered useful to assess organ dysfunction, disease severity, comorbidity status and prognosis of sepsis [12, 13, 28]. In our analysis of 1-year overall-cause mortality, we observed that higher scores in Max SOFA, CCI, and SAPS-II were risk factors for ADs patients. This association remained consistent even when conducting subgroup analyses where the repeated group showed a correlation between high scores of quantified assessment tools and mortality, resembling the findings observed within the single group (all P < 0.05). Moreover, within the repeated group, scores of the tools were higher during the last ICU stays in the comparison between the first admissions and the last ones (all P < 0.05), suggesting that sepsis developing repeatedly were more severely ill than those occurred at the first time. These findings are partly similar to other published literature reporting a correlation between Max SOFA, CCI, and SAPS-II scores and ADs [5, 29‐31]. To summarize, the higher scores of three quantified assessment tools were long-term mortality-associated risk factors for AD patients whether with recurrent sepsis episodes or with single one. Those subjected to repeated sepsis had a propensity for higher scores after the previous hits of sepsis, indicating more severe physical dysfunction in the recurrence. As such, it is crucial to comprehensively assess the physical conditions and multiorgan involvements among AD patients with sepsis. Given a greater possibility of worse conditions, AD patients with repeated sepsis may require enhanced attention and long-term follow-up following a sepsis episode.

As for secondary outcomes, compared to those with single sepsis episode, individuals with repeated in-ICU sepsis had an increased risk of septic shock, while no significant differences were observed between the repeated and the single group on in-ICU death (or not) (Table 4). Nevertheless, the growth of Max SOFA scores and SAPS-II scores were correlated with both two secondary outcomes. The findings suggested that the septic shock was a vital outcome for AD patients with repeated sepsis. In terms of in-ICU death, one form of mortality in hospital, our patients had an ICU stay duration of one week or less (Table 1), suggesting in-ICU deaths were all within acute phases. During the acute phase, the occurrence of in-ICU death was associated with the severity of diseases (reflecting by SOFA or SAPS-II) instead of the frequency of sepsis episodes. In despite of the irrelevance, we reckon that the underlying mechanisms of in-ICU deaths were different between the single and the repeated group. For AD patients with once-only sepsis episode, early hyperinflammation might manifest in their initial hit, during which organ failure even deaths might happen due to uncontrolled inflammation as well as inadequate medical interventions [32‐35]. On the other hand, for AD survivors with repeated sepsis, the prolonged immunosuppression made them lose the ability to resist the secondary infection and repeated sepsis, rendering them vulnerable to deaths during the acute phase [36, 37]. However, very few is still known about the association between in-ICU deaths and speiss episodes among AD population.

Respiratory failure, with acute ones occupying the majority (90.3% of overall respiratory failure before PSM), also emerged as another risk factor for long-term mortality, which was in line with previous study finding that respiratory failure was associated with the mortality of rheumatoid arthritis with sepsis [38]. On the other hand, the factors for a good prognosis in our study included longer length of ICU stay and the use of immunosuppressants. In the published literature, there were conflicting findings about the association between the length of ICU stay and mortality of AD patients. One study reported no significant relationship between the length of ICU stay and mortality of SLE patients [39], but it was restricted by small size of patient sample. Another study with a relatively large sample found that an extended ICU stay (2 ~ 14 days) was a risk factor for 3-year mortality for rheumatoid arthritis with sepsis [38], which was inconsistent with our results. But differently, the maximum ICU stay did not exceed 1 week in the baseline of our AD patients. The decreased risk of mortality observed among our AD patients who had a relatively longer ICU stay might be due to the supporting critical care and sufficient anti-sepsis therapy in an appropriate length of ICU stay [40]. As for immunosuppressants or biologics, a current study showed that the occurrence of sepsis among AD was not relevant to the use of immunosuppressants in the ICU [41]. but another small-scale study found biologics or corticosteroids could improve the 30-day survival of SLE patients with sepsis [39]. The contradiction of the results on ICU staying and drugs for ADs with sepsis needed more studies to resolve.

Our study had some strengths and limitations. A major strength was that our study was conducted based on a large patient sample and its design of a retrospective cohort study can better indicate the association between repeated sepsis and ADs. Also, variables of interest were adjusted by other confounders through post-PSM multivariate analysis and sensitivity analysis so as to make the results as reliable as possible. However, our study is subject to limitations imposed by the MIMIC database. Detailed causes of death were not available, which hindered further analysis of specific causes of mortality. In the meanwhile, due to privacy protection principles followed by the MIMIC database, the maximum follow-up period for each patient is limited to one year after the last discharge. Consequently, we were only able to focus on the 1-year overall-cause mortality as one of the outcomes. Secondly, we found it difficult to exclusively evaluate the disease activity and immunity of ADs due to the lack of some AD-specific measurements from the database, including autoantibody titer, cytokines, complements, the count of lymphocyte subtypes, etc. We could only utilize SOFA, CCI and SAPS-II, some quantified assessment tools commonly used in ICU, to assess the conditions of disease severity. As shown in our results, the tools indeed worked. Future studies should consider incorporating additional disease activity-related indices to further investigate the relationship between ADs and sepsis. Thirdly, this study utilized an open-access database derived from in-hospital database systems that were collected from routine clinical practice. It is possible that inaccuracies or non-standardization may exist in the documentations.