In this population-based study, we addressed the impact of SLE on the 5-year survival in critically ill septic patients using a case-control design. We found that SLE was associated with high 5-year mortality in patients with sepsis requiring ICU admission and mechanical ventilation. In addition to SLE, older age, higher number of comorbidities, and usage of glucocorticoid (≥ 5 mg/day prednisolone equivalent), methotrexate as well as immunosuppressants were also independent risks for mortality in critically ill septic patients. These findings demonstrated the prolonged impacts of SLE on septic survival and highlight the need for vigilance and risk stratification in lupus patients discharged from ICU for sepsis.
The survival of patients with SLE is improving in the past 4 decades. Yen et al., using a nationwide claim database in the USA, reported that the age-specific SMR for SLE decreased from 0.45 (95% CI, 0.42 to 0.48) per 100,000 person-years in 1968 to 0.34 (95% CI, 0.32 to 0.36) per 100,000 person-years in 2013 . However, Tselios et al. recently investigated the cause- and age-specific SMR among lupus patients in Ontario over a four-decade study period (1971–2013) and found that infection (24.5%) was the leading cause of mortality, followed by atherosclerosis (15.7%), active lupus (13.3%), and malignancy in patients with SLE (9.6%) . In detail, they found that although infection-specific SMR decreased steadily by decades, but the infection-specific SMR was extremely high (30.2, 95% CI 14.4–46.0) in lupus patients aged 19–39 compared with the relatively low SMR (3.5; 95% CI 2.5–4.5) among those older or equal to 40 years . Similarly, hospitalisation for serious infections among lupus patients in the USA increases steadily between 1996 and 2011, with nearly 12 times higher than those in non-SLE populations , and sepsis has been reported to account for nearly 80% of aetiologies for ICU admission in Thailand among patients with SLE . These evidence highlight the essential need for investigating the association between SLE and sepsis relevant mortality.
Despite of a steady decrease in ICU mortality, a number of studies including our studies have found high post-ICU mortality in the past two decades [9, 12, 22]. Therefore, the long-term outcome, instead of ICU/hospital-mortality, of critically ill patients is currently one of the leading research priorities in critical care medicine, particularly among those with sepsis, given the increasing evidence have shown the prolonged immunologic and metabolic sequelae of sepsis [10, 13, 14]. van Vught et al., employing paired analyses of the whole transcriptome in leucocytes among septic patients with and without secondary infection, found impaired gluconeogenesis and glycolysis in septic patients with a secondary infection . In addition to secondary infection, the post-sepsis altered immunologic and metabolic function may have systemic impact in patients with sepsis as shown by one Taiwanese population-based study that sepsis survivors had an increased risk of all-cause mortality (HR 2.18; 95% CI 2.14–2.22) and major adverse cardiovascular events at 1 year after discharge (HR, 1.37; 95% CI, 1.34–1.41) . Given that the immune system plays a substantial role in the long-term outcome after the septic episode, there is a crucial need to explore the long-term outcome in critically ill septic patients with autoimmune diseases including SLE . Indeed, few studies focused on investigating long-term outcome in critically ill septic patients with SLE. One Taiwanese single-hospital study, investigating 240 lupus patients with bacteremia between 2000 and 2005, reported the 30-day and 1-year mortality rate was 24% and 33%, respectively . Larcher et al., investigating 525 critically ill patients with systemic rheumatic disease including 109 lupus patients, reported that ICU-, hospital-, and 1-year-mortality rates were 23.8%, 30.5%, and 37.7%, respectively . Due to the distinct young population and the markedly changed survival of SLE in recent years, there is a crucial need to investigate the long-term outcome of critically ill patients with SLE. In the present study, we used data mainly in female whose age was approximately 50 years to provide the real-world 5-year mortality data among critically ill septic patients with and without SLE and to show the independent mortality impact of SLE in critically ill septic patients. Collectively, these findings indicate vigilance for lupus patient discharged from ICU for sepsis, and more studies are warranted to elucidate underlying biological mechanisms.
The impacts of SLE on critically ill septic patients may result from an altered immunological and metabolic response in the recovery from sepsis. Shi et al., conducting RNA-seq of monocytes in 9 lupus patients and matched control subjects, showed evidence of chronic endotoxin exposure and differentially expressed type I interferon (IFN) genes in lupus patients . Indeed, type I IFN has been implicated with a wide range of infectious diseases, including bacterial, mycobacterial, and viral infection [25,26,27,28]. Yang et al., using a mouse sepsis model with cecal ligation and puncture (CLP), recently reported that type I IFN exerted the disseminated intravascular coagulation in bacterial infection through amplifying the release of high-mobility group box 1 (HMGB1) into the extracellular space . Notably, HMGB1 was recently identified to be associated with a prolonged and impaired cognitive function in patient survived from a critical illness . Therefore, type I IFN and HMGB1 has been seen as the potential therapeutic target in sepsis [30, 31]. Type I IFN has also been implicated with a dysregulated inflammation in mycobacterial infection, and increasing studies including our previous study have found that pro-inflammatory mediators including IL-1β and type I IFN strengthen the eicosanoid pathway, which in turn modulates death patterns of infected cells in mycobacterial infection [26, 32, 33]. Notably, Clayton et al. revealed an overlapped transcriptomic signature, mainly type I IFN-associated signalling pathway, between patients with tuberculosis and SLE . Two studies further revealed that impaired type I IFN immunity, including autoantibodies against type I IFN and inborn errors of type I IFN immunity, may lead to severe coronavirus disease 19 (COVID-19) infection [27, 28].
Surprisingly, most of the studies with regard to the survival in lupus patients after sepsis/infection mainly investigate patients with SLE and healthy controls in the general population, and few studies have comparable septic controls to specify the independent impact of SLE on the long-term outcome of sepsis. Kedves et al. recently conducted a population-based claim database in Hungary with age- and sex-matched health controls to explore the long-term impact of patients with SLE . They found an increased adjHR (2.17, 95% CI 1.94–2.44) for all-cause mortality in patients with SLE compared with healthy controls and reported higher infection-related deaths in lupus patients than those in healthy control subjects . Similarly, one population-based study conducted in southern Sweden also reported a higher long-term mortality rate in patients with SLE compared with the mortality rate in the general population . One recently published study using the 2010–2015 French SLE cohort reported that 1068 lupus patients with septic shock had higher 1-year mortality than lupus patients without septic shock . These evidence highlight the crucial need for comparable non-SLE septic controls with similar age as well as sex to clarify the independent impact of SLE on the long-term outcome of sepsis.
In the present study, we found that usage of glucocorticoid, methotrexate, and immunosuppressants, but not hydroxychloroquine, as well as more comorbidity, including chronic kidney disease, were associated with high 5-year mortality. Intriguingly, previous studies including our recently published study have shown that the use of hydroxychloroquine tended to inversely be associated with incident infectious disease, particularly malaria and pneumocystis pneumonia (PCP), among patients with SLE [36, 37]. We further conducted analyses focusing on the usage of hydroxychloroquine among the 513 enrolled critically ill septic patients with sepsis; however, we did not observe the protective effect of hydroxychloroquine on post-septic 5-year mortality (Supplemental Table 5–7). We postulated the relatively low incidence of hydroxychloroquine-protected diseases in Taiwan and attributed mortality might at least partly explain the lack of association between the use of hydroxychloroquine and 5-year mortality in the present study.
We applied the sepsis-3 definition, using the SOFA score to identify patients with sepsis, to define sepsis in the present study . Compared with the sepsis-3 definition, the sepsis-2 definition, using a requisite minimum of two systemic inflammatory response syndrome criteria, might not be not fully accurate to identify patients with sepsis . As shown by Kaukonen et al., one in eight critically ill patients admitted to an ICU for infection with new organ failure is estimated not to meet the sepsis-2 definition, and these patients exhibit significant mortality and morbidity . Therefore, the Sepsis-3 definition is increasingly used to identify patients with sepsis in recent studies including our recently published study to address factors for sepsis in patients receiving tumour necrosis factor inhibitors for immune-mediated inflammatory diseases [17, 39, 40].
There are limitations to this study. First, we used ICD coding to define patient with sepsis, and sepsis could potentially be overestimated. However, we used a stringent definition by restricting septic patients admitted to the ICU and received mechanical ventilation. Therefore, we think we might underestimate, instead of overestimate, the critically ill septic patients under such a stringent definition. Second, the lack of data regarding disease activity in claim data is a limitation; however, the comprehensive information regarding medications should at least partly reflect the disease activity of SLE. Similarly, the cause of death cannot be delineated in NHIRD. Third, the concern for the accuracy of the SLE diagnosis in claims data, but the diagnosis of SLE in Taiwan was validated by at least two qualified rheumatologists through checking clinical data for the certificate of catastrophic illness. Fourth, more studies are warranted to validate our findings in other populations.