Mortality of Patients with Rheumatoid Arthritis Requiring Intensive Care: a Single-Center Retrospective Study

Clinical Rheumatology (2019) 38:3015–3023 https://doi.org/10.1007/s10067-019-04651-w

ORIGINAL ARTICLE

Mortality of patients with rheumatoid arthritis requiring intensive care: a single-center retrospective study

Yael Haviv-Yadid1 & Yulia Segal2 & Amir Dagan3,4,5 & Kassem Sharif6 & Nicola Luigi Bragazzi7 & Abdulla Watad6 & Howard Amital6,8,9 & Yehuda Shoenfeld8,9,10,11 & Ora Shovman6,8,9

Received: 20 February 2019 /Revised: 10 June 2019 /Accepted: 18 June 2019 /Published online: 26 June 2019 # International League of Associations for Rheumatology (ILAR) 2019

Abstract Background Patients with rheumatoid arthritis (RA) are at a high risk for life-threatening conditions requiring admission to the intensive care unit (ICU), but the data regarding the outcomes of these patients is limited. The present study investigated the clinical characteristics and outcomes of RA patients admitted to an ICU. Methods This retrospective cohort study included RA patients admitted to the general ICU of the Sheba Medical Center during 2002–2018. The main outcome was 30-day mortality. Using Student’s t test, χ2, and multivariable analyses, we compared the demographic, clinical, and laboratory parameters of the survivors and the non-survivors. Figures with p value < 0.05 were considered statistically significant. Results Forty-three RA patients were admitted to the ICU during the study period (mean age, 64.0 ± 13.1 years; 74.4% female). The leading causes of ICU admission were infection (72.1%), respiratory failure (72.1%), renal failure (60.5%), and septic shock (55.8%). The 30-day mortality rate was 34.9%, with infection (9/15, 60%) as the most frequent cause. The mean Acute Physiology and Chronic Health Evaluation (APACHE) II and Sequential Organ Failure Assessment (SOFA) scores were 19.7 ± 12.5 and 7.0 ± 4.5, respectively. Multivariable analysis showed that heart failure (p = 0.023), liver failure (p = 0.012), SOFA score (p = 0.007), and vasopressor treatment in ICU (p = 0.039) were significantly associated with overall mortality. SOFA score was linked with overall mortality (area under the curve (AUC) = 0.781 ± 0.085, p = 0.003) and mortality from respiratory failure (AUC = 0.861 ± 0.075, p = 0.002), while APACHE II score was only correlated with mortality from infection (AUC = 0.735 ± 0.082, p = 0.032). Conclusions Our study demonstrated a relatively high mortality rate among RA patients who were admitted to the general ICU. RA patients with risk factors such as heart failure, liver failure, elevated SOFA score, and vasopressor treatment in ICU should be promptly identified and treated accordingly.

Key Points • The 30-day mortality rate of patients with RA that were admitted to the general ICU of a tertiary hospital was 34.9%. • The most common causes of ICU admission among patients with RA were infections and respiratory failure. Infections were the most common cause of death among these patients. • Patients with RA that present to the ICU with heart failure, liver failure, elevated SOFA score, and/or require vasopressor treatment in ICU should be promptly identified and treated accordingly.

* Ora Shovman 7 School of Public Health, Department of Health Sciences (DISSAL), [email protected] University of Genoa, Genoa, Italy

1 Intensive Care Unit, Sheba Medical Center, Ramat Gan, Israel 8 Zabludowicz Center for Autoimmune Diseases, Sheba Medical Center, Tel Hashomer, 52621 Ramat Gan, Israel 2 Department of Internal Medicine ‘T’, Sheba Medical Center, Ramat Gan, Israel 9 Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel 3 Department of Internal Medicine ‘B’, Assuta Ashdod Medical Center, Ashdod, Israel 10 Past incumbent of the Laura Schwarz-Kipp Chair for Research of 4 Rheumatology Unit, Assuta Ashdod Medical Center, Ashdod, Israel Autoimmune Diseases, Tel Aviv University, Tel Aviv, Israel 5 Ben-Gurion University of the Negev, Beer Sheva, Israel 11 I.M. Sechenov First Moscow State Medical University of the 6 Department of Internal Medicine ‘B’, Sheba Medical Center, Ramat Ministry of Health of the Russian Federation (Sechenov University), Gan, Israel Moscow, Russia 3016 Clin Rheumatol (2019) 38:3015–3023

Keywords Acute physiology and chronic health evaluation (APACHE) II score . Intensive care unit . Rheumatoid arthritis mortality . Sequential organ failure assessment (SOFA) score

Introduction Material and methods

Patients with autoimmune rheumatic disorders (ARDs) are at Population high risk for complications requiring admission to the intensive care units (ICUs) [1, 2]. The prevalence of different This is a single-center observational retrospective study con- ARDs in ICUs has changed in the past decades, and systemic ducted in the general ICU of the Sheba Medical Center in lupus erythematosus (SLE) has surpassed rheumatoid arthritis Tel-Hashomer, a tertiary referral hospital in Israel. Using com- (RA) as the leading autoimmune disease necessitating admis- puterized ICU discharge summaries, we analyzed the medical sion to ICU [1]. However, RA remains the second most prev- records of 43 patients who have fulfilled the American College alent ARD in ICUs and represents a challenge for the physi- of Rheumatology (ACR) 1987 or ACR 2010 criteria for RA cians. The majority of previous studies on ARDs patients in and were admitted to the general ICU during 2002–2018. RA ICUs analyzed mixed cohorts of patients with different dis- patients who were admitted to the other ICUs of the hospital eases that have distinct courses, mortality rates, and mortality- such as the cardiac, neurological, or pediatric ICUs were not associated factors. This heterogeneity in cohorts may partially included. The study fulfilled the ethical guidelines of the most explain the significant variability in the reported mortality recent declaration of Helsinki (Edinburgh, 2000) and received rates that range from 17 to 55%, and the wide spectrum of approval (No. 3233-16-SMC) by the local ethical committee. variables that may be associated with increased mortality [1]. Data regarding patient demographics (age and gender) and To date, few retrospective studies have investigated the clinical and laboratory parameters were obtained retrospec- clinical aspects and the prognostic features of RA patients in tively from patient charts and laboratory records. Among ICU [3–6]. One of these studies is a large population-based those who were admitted twice to the ICU, only the first ad- study from Canada which demonstrated that RA patients had a mission was considered. In addition, data regarding the overall higher risk for ICU admission in comparison with the general mortality rate in the general ICU was extracted from the com- population (HR 1.65; 95% CI 1.50–1.83) [3]. According to puterized records of Sheba Medical Center in Tel-Hashomer. this study, during 2000–2010, the yearly ICU admission rate The severity of the disease was assessed within 24 hours of RA patients was 1%, with a 10-year cumulative rate of 8%, after ICU admission using the “Acute Physiology and a figure that was 60% higher compared with that in the general Chronic Health Evaluation II” (APACHE II) [11] and the population. The most frequent reasons for ICU admission “Sequential Organ Failure Assessment” (SOFA) [12]scores. were ischemic heart disease (IHD) followed by infections [3]. The APACHE II score is calculated based on the age of the In the last decades, several therapeutic advancements have patient and 12 routine physiological measurements, including been achieved in the management of RA, and this led to a alveolar-arterial oxygen difference (AaDO2) or partial pressure decreasing mortality rate for the last 50 years [7]. Despite this, of oxygen in arterial blood (PaO2), depending on the fraction of RA patients still have a higher mortality risk compared with the inspired oxygen (FiO2), rectal temperature, mean arterial that in the general population [7]. With regard to the ICU pressure, arterial pH, heart rate, respiratory rate, serum sodium mortality rate, the data is relatively limited. In the Canadian and potassium, creatinine, hematocrit, white blood cell count, study [3], the mortality rate found in the RA cohort was not and Glasgow coma scale [11]. A score of 25 represents a pre- different from that in the general population admitted to the dicted mortality of 50% in the general ICU. The SOFA score is ICU. Another work from Israel investigated the 30-day sur- calculated using objective measures of organ dysfunction or vival rate of patients hospitalized in the ICU following sepsis failure in major organ systems, including respiratory, renal, car- and found that the mortality rate within the RA subgroup was diovascular, neurologic, liver, and coagulation systems. The 51.2%, similar to non-RA patients [6]. Previous studies have SOFA score ranges from 0 to 24, and higher scores predict described the course and the outcome of dermatomyositis, worse outcomes and higher mortality rates [12]. scleroderma, and vasculitis patients admitted to the general Mortality was defined as death within 30 days of ICU ad- ICU of a tertiary hospital in Israel [8–10]. The objective of mission. Survival was defined as being successfully trans- the current study is to analyze the clinical characteristics of ferred from the ICU into another department and either sur- RA patients admitted to the ICU between 2002 and 2018, to viving until day 30 in the hospital or being released from the investigate their clinical course and outcome, and to identify hospital and then having another medical entry recorded into factors that may correlate with the 30-day mortality. the patient’s chart after the 30-day mark. Clin Rheumatol (2019) 38:3015–3023 3017

Table 1 The main characteristics of the study population (n =43) Variable Value

Age, years (mean ± standard deviation; median, range) 64.0 ± 13.1; 66 [24–84] Gender (female), n (%) 32 (74.4%) RF positivity* 20 (62.5%) Anti-CCP positivity* 12 (37.5%) Comorbidities, n (%) Essential hypertension 28 (65.1%) Ischemic heart disease 15 (34.9%) Chronic obstructive lung disease 15 (34.9%) Ever smoker 14 (32.6%) Chronic renal failure 13 (30.2%) Cerebral vascular accident 11 (25.6%) Liver cirrhosis 4 (9.3%) Reason for admission, n (%) Infection 31 (72.1%) Respiratory failure 31 (72.1%) Renal failure 26 (60.5%) Septic shock 24 (55.8%) Thrombocytopenia 13 (30.2%) Liver failure 10 (23.3%) Heart failure 7 (16.3%) Neurological symptoms 7 (16.3%) Hemorrhage 5 (11.6%) Pancytopenia 2 (4.7%) Days from hospital admission to transfer to ICU, days (median, range) 2 [0–53] APACHE II (mean ± standard deviation; median, range) 19.7 ± 12.5; 22 [0–43] SOFA (mean ± standard deviation; median, range) 7.0 ± 4.5; 8 [0–17] Treatment before ICU admission, n (%) Steroids 27 (62.8%) Cyclophosphamide 1 (2.3%) IVIG 1 (2.3%) Synthetic DMARDS** 23 (53.5%) Methotrexate 19 (44.2%) Hydroxychloroquine 12 (27.9%) Leflunomide 2 (4.7%) Azathioprine 1 (2.3%) Biologic DMARDS** 10 (23.3%) Anti-TNF agents 7 (16.3%) Rituximab 4 (9.3%) Abatacept 1 (2.3%) Positive culture during first week, n (%) Fungi 9 (20.9%) Gram-positive 5 (11.6%) Gram-negative 6 (14%) Mixed 8 (18.6%) Site of infection, n (%) Lungs 20 (46.5%) GI 10 (23%) Urine 7 (16.3%) Blood 5 (11.6%) 3018 Clin Rheumatol (2019) 38:3015–3023

Table 1 (continued) Variable Value

Skin 2 (4.7%) CNS 2 (4.7%) Treatment in ICU, n (%) Steroids 36 (83.7%) Mechanic ventilation 33 (76.7%) Vasopressor support 30 (69.8%) Dialysis 8 (18.6%) Plasmapheresis 1 (2.3%) IVIG 1 (2.3%) Length of admission, days (median, range) 7 [1–37] Overall mortality, n (%) 15 (34.9%) Reason of death, n (%) Infection 9 (60%) Respiratory failure 8 (53.3%) Hemorrhage 3 (20%)

Statistical analysis MedCalc Statistical Software version 16.4.3 (MedCalc Software bvba, Ostend, Belgium). All variables were collected into a single database. Before com- mencing any statistical processing and data analysis, figures were visually inspected for potential outliers. Continuous vari- Results ables were computed as mean ± standard deviation, whereas categorical variables were recorded as percentages. The patients The study included 43 RA patients that were admitted to the were then classified according to the outcome—survivors and ICU during the study period. The main characteristics of the non-survivors—and various parameters were compared be- study population are summarized in Table 1. The mean age tween the 2 groups. In order to investigate the impact of differ- was 64.0 ± 13.1 years (median 66, range 24–84 years), and 32 ent variables on mortality, univariable analyses were performed, (74.4%) of the patients were females. Data regarding anti- including Student’s t test and χ2. Multivariable logistic regres- cyclic citrullinated peptide (anti-CCP) antibodies and rheuma- sion analysis was carried out adjusting for confounding factors, toid factor (RF) were available for 32 out of 43 patients, and such as gender and age. The variables included in the multivar- they were found in 12 (37.5%) and 20 (62.5%) patients with iable analysis were the statistically significant variables that RA, respectively. The major comorbidities in our RA popula- were found in the univariable analysis. The ability of the tion were essential hypertension (28 patients, 65.1%), ische- SOFA and APACHE II scores to predict the 30-day mortality mic heart disease (15 patients, 34.9%), chronic renal failure from all causes was assessed using the receiver operating char- (13 patients, 30.2%), and liver cirrhosis (4 patients, 9.3%). acteristic (ROC) analysis. In addition, ROC analysis was used The majority of patients had several reasons for ICU ad- to evaluate the power of the SOFA and APACHE II scores in mission. The leading causes of ICU admission were infection predicting the 30-day mortality stratified according to the cause (72.1%), respiratory failure (72.1%), renal failure (60.5%), of mortality. and septic shock (55.8%). The three most frequently co- Figures with p value < 0.05 were considered statistically occurring reasons for ICU admission were infections, renal significant. All statistical analyses were carried out with the failure, and septic shock. No patients were hospitalized due commercial software “Statistical Package for Social Sciences” to the manifestations of the disease itself. (SPSS for Windows, version 23.0.0; Chicago, IL, USA). The median number of days from the admission to the hos- Graphs were generated using the commercial software pital to the transfer to the ICU was 2 days (range, 0–53). The Clin Rheumatol (2019) 38:3015–3023 3019

Table 2 Univariable analysis of the relationship between the parameters investigated and the mortality of patients with rheumatoid arthritis admitted at intensive care unit (n =43)

Variable Non-survivor (n = 15) Survivor (n = 28) Statistical significance (p value)

Age 66.71 ± 10.61 62.67 ± 14.02 0.361 Gender (female) 11 (73.3%) 21 (75%) 1 RF positivity* 7 (58.3%) 13 (65%) 0.724 Anti-CCP positivity* 5 (41.7%) 7 (35%) 0.724 Comorbidities Essential hypertension 9 (60%) 19 (67.9%) 0.740 Ischemic heart disease 6 (40%) 9 (32.1%) 0.740 Chronic obstructive lung disease 6 (40%) 9 (32.1%) 0.740 Ever smoker 6 (40%) 8 (28.6%) 0.507 Chronic renal failure 6 (40%) 7 (25%) 0.324 Cerebral vascular accident 4 (26.7%) 7 (25%) 1 Liver cirrhosis 2 (13.3%) 2 (7.1%) 0.602 Infection 12 (80%) 19 (67.9%) 0.492 Respiratory failure 13 (86.7%) 18 (64.3%) 0.164 Renal failure 12 (80%) 14 (50%) 0.101 Septic shock 11 (73.3%) 13 (46.4%) 0.116 Thrombocytopenia 7 (46.7%) 6 (21.4%) 0.162 Liver failure 7 (46.7%) 3 (10.7%) 0.019*** Heart failure 5 (33.3%) 2 (7.1%) 0.040*** Neurological symptoms 4 (26.7%) 3 (10.7%) 0.215 Hemorrhage 3 (20%) 2 (7.1%) 0.324 Pancytopenia 2 (13.3%) 0 (0%) 0.116 Days from hospital admission to transfer to ICU, days (median) 1 2.5 0.308 APACHE II (mean ± standard) 21.4 ± 12.3 18.8 ± 12.4 0.528 SOFA (mean ± standard) 9.80 ± 4.71 5.57 ± 3.63 0.003*** Treatment before ICU admission Steroids 10 (66.7%) 17 (60.7%) 0.752 Cyclophosphamide 1 (6.7%) 0 (0%) 0.349 IVIG 1 (6.7%) 0 (0%) 0.349 Synthetic DMARDS** 7 (46.7%) 16 (57.8%) 0.540 Methotrexate 6 (40%) 13 (46.4%) 0.755 Hydroxychloroquine 3 (20%) 9 (32.1%) 0.492 Leflunomide 0 (0%) 2 (7.1%) 0.535 Azathioprine 0 (0%) 1 (3.6%) 1 Biologic DMARDS** 3 (20%) 7 (25%) 1 Anti-TNF agents 2 (13.3%) 5 (17.9%) 1 Rituximab 1 (6.7%) 3 (10.7%) 1 Abatacept 0 (0%) 1 (3.6%) 1 Positive culture during first week Fungi 4 (26.7%) 5 (17.9%) 0.696 Gram-positive 1 (6.7%) 4 (14.3%) 0.643 Gram-negative 1 (6.7%) 5 (17.9%) 0.403 Mixed 3 (20%) 5 (17.9%) 1 Site of infection Lungs 10 (66.7%) 10 (35.7%) 0.064 GI 5 (33.3%) 5 (17.9%) 0.281 Urine 1 (6.7%) 6 (21.4%) 0.391 3020 Clin Rheumatol (2019) 38:3015–3023

Table 2 (continued)

Variable Non-survivor (n = 15) Survivor (n = 28) Statistical significance (p value)

Blood 3 (20%) 2 (7.1%) 0.324 Skin 0 (0%) 2 (7.1%) 0.535 CNS 0 (0%) 2 (7.1%) 0.535 Treatment in ICU Steroids 14 (93.3%) 22 (78.6%) 0.391 Mechanic ventilation 13 (86.7%) 20 (71.4%) 0.451 Vasopressor support 14 (93.3%) 16 (57.1%) 0.017*** Dialysis 4 (26.7%) 4 (14.3%) 0.419 Plasmapheresis 1 (6.7%) 0 (0%) 0.349 IVIG 0 (0%) 1 (3.6%) 1 Length of admission 7 7.5 0.743

APACHE II, Acute Physiology And Chronic Health Evaluation II; CNS, central nervous system; DMARDs, disease-modifying antirheumatic drugs; GI, gastro-intestinal tract; ICU, intensive care unit; IVIG, intravenous immunoglobulin; SOFA, sequential organ failure assessment; TNF,tumornecrosis factor; RF, rheumatoid factor; Anti-CCP, Anti-cyclic citrullinated peptide *Data regarding anti-CCP and RF was available for 32 out of 43 patients **Synthetic DMARDs include Methotrexate, Hydroxychloroquine, Leflunomide, and Azathioprine. Biologic DMARDs include Anti-TNF agents, Rituximab, and Abatacept ***Statistically significant with a p value < 0.05 mean ± SD APACHE II and SOFA scores were 19.7 ± 12.5 and reveal a correlation between liver failure and liver cirrhosis 7.0 ± 4.5, respectively. Before ICU admission, patients were or synthetic or biologic DMARDs. treated mainly with steroids (62.8%), synthetic DMARDs The univariable analysis showed that the following vari- (53.5%) and biological DMARDs (23.3%). During the hospital- ables were significantly different between non-survivors and ization in the ICU, 33 patients (76.7%) required mechanic ven- survivors: heart failure (5 patients [33.3%] versus 2 patients tilation, 30 (69.8%) required vasopressor support, and 8 subjects [7.1%], p = 0.040), liver failure (7 patients [46.7%] versus 3 (18.6%) were on dialysis. Cultures were performed during the patients [10.7%], p = 0.019), elevated SOFA (9.80 ± 4.71 ver- first week of the ICU hospitalization, and it was found that 9 sus 5.57 ± 3.63, p = 0.003), and vasopressor support in ICU patients (20.9%) had fungal infection, 6 patients (14%) had (14 patients [93.3%] versus 16 patients [57.1%], p =0.017). Gram-negative infection, 5 patients (11.6%) had Gram-positive Further details are reported in Table 2. infection, and 8 patients (18.6%) had several pathogens on cul- The multivariable analysis demonstrated that heart failure ture. The major site of infection was the lungs (46.5%). During (OR 10.68 [95%CI 1.39 to 81.74], p = 0.023), liver failure (OR the stay in the ICU, most patients (83.7%) received steroids. The 8.59 [95%CI 1.61 to 45.76], p = 0.012), SOFA score (OR 1.41 median length of ICU admission was 7 days. [95%CI 1.10 to 1.80], p = 0.007), and vasopressor treatment in The 30-day mortality rate among the whole RA cohort was ICU (OR 10.24 [95%CI 1.13 to 93.02], p = 0.039) remained 34.9%, higher than the overall 30-day mortality rate for all statistically significant predictors of mortality (Table 3). patients admitted to our general ICU (23%). Nine patients ROC analysis revealed that an elevated SOFA score, but not (60%) died from infection, 8 patients (53.3%) died from re- APACHE II score, was significantly associated with overall spiratory failure, and 3 patients (20%) died from hemorrhage. mortality (AUC = 0.781 ± 0.085, p =0.003)(Tables 4 and 5, Sub-analysis of 10 patients with acute liver failure did not Fig. 1). Stratifying on the cause of mortality, SOFA score was

Table 3 Multivariable analysis of the relationship between the Variable Coefficient Standard error Wald p value OR 95% CI parameters investigated and the mortality of patients with Heart failure 2.37 1.04 5.20 0.023* 10.68 1.39 to 81.74 rheumatoid arthritis admitted at Liver failure 2.15 0.85 6.36 0.012* 8.59 1.61 to 45.76 intensive care unit, adjusted for SOFA 0.34 0.13 7.33 0.007* 1.41 1.10 to 1.80 age and gender (n =43) Vasopressor treatment 2.33 1.13 4.27 0.039* 10.24 1.13 to 93.02

CI, confidence interval; ICU, intensive care unit; OR, odds ratio; SOFA, sequential organ failure assessment *Statistically significant with a p value < 0.05 Clin Rheumatol (2019) 38:3015–3023 3021

Table 4 Receiver operating characteristic (ROC) analysis of the predic- mortality) of patients with rheumatoid arthritis admitted at intensive care tive power of the Sequential Organ Failure Assessment (SOFA) score, in unit (n =43) predicting the mortality (overall and stratified according to the cause of

Variable AUC ± SE 95% CI Significance level p (area = 0.5)

Mortality (overall) 0.781 ± 0.085 0.615–0.947 0.003* Mortality due to respiratory failure 0.861 ± 0.075 0.715–1.000 0.002* Mortality due to infection 0.686 ± 0.100 0.490–0.882 0.089

AUC, area under the curve; SE, standard error *Statistically significant with a p value < 0.05 associated with mortality from respiratory failure (AUC = 0.861 However, a relatively high 30-day mortality rate (more than ± 0.075, p = 0.002) but not from infection (AUC = 0.686 ± 50%) was seen in 124 RA patients admitted with the diagnosis 0.100, p = 0.089). The APACHE II score was associated only of sepsis to the ICUs of 7 tertiary hospitals in Israel during with mortality from infection (AUC = 0.735 ± 0.082, p =0.032). 2002–2012 [6]. Sub-analysis of non-survivors who had died from infection In our study, infections (72.1%) and respiratory failure showed that 33% of these patients had a SOFA score that was (72.1%) were the leading causes of admission to the general equal to, or lower than the median SOFA score for all patients, ICU. Furthermore, infections, predominantly of the pulmo- while having an APACHE II of which score was equal to, or nary tract, were the most common cause of death (9/15, higher than the respective median APACHE II score. 60%). This data supports the findings from previous reports on mixed cohorts of patients with ARDs [1], as well as reports on isolated cohorts of SLE [14, 15] or vasculitis patients [10]. Discussion In contrast, an earlier large population-based study has indi- cated that ischemic heart disease was the most common reason In our study, the 30-day mortality rate of RA patients admitted to for ICU admission in the RA population (45.8%) followed by the general ICU was 34.7%, higher than the 30-day mortality infections (19.8%) [3]. This discrepancy may be explained by rate for all patients in our ICU (23%). This mortality rate for the the fact that the earlier study included RA patients admitted to general population in our ICU was slightly higher than the rate different high-intensity ICUs, while our study focused on RA reported in literature (13–19%) [13], likely due to the tertiary patients admitted to the general ICU. nature of our hospital. To the best of our knowledge, only one Several factors may be associated with mortality in patients previous study analyzed RA patients admitted to the high- with ARDs admitted to the ICU, including APACHE II score, intensity adult ICU due to all causes [3]. This was a large SOFA score, length of stay, shock, comorbidities, old age, population-based study from Canada that has reported similar cytopenia, and vasopressors [1, 16]. In our multivariable anal- 30-day mortality rates in the RA cohort (16.5%) and in the ysis, heart failure was linked with an increased risk for mor- general population (14.2%). tality. It has been shown that RA patients have twice the risk of Other studies focused on RA patients admitted to the ICU developing heart failure compared with the general population due to infection only [5, 6], and data regarding the mortality rate [17]. Moreover, heart failure appears to be an important con- of this cohort is controversial. For instance, a case-control study tributor to excess overall mortality among RA patients [18]. from Austria that included RA patients admitted to ICU due to Interestingly, the increased risk of heart failure is not elucidat- infection have reported a low mortality rate of 10.5% [5]. ed by an increased frequency or effect of either IHD or

Table 5 Receiver operating characteristic (ROC) analysis of the predic- according to the cause of mortality) of patients with rheumatoid arthritis tive power of the Acute Physiology And Chronic Health Evaluation admitted at intensive care unit (n =43) (APACHE) II score, in predicting the mortality (overall and stratified

Variable AUC ± SE 95% CI Significance level p (area = 0.5)

Mortality (overall) 0.589 ± 0.094 0.405–0.773 0.339 Mortality due to respiratory failure 0.670 ± 0.111 0.453–0.886 0.138 Mortality due to infection 0.735 ± 0.082 0.574–0.896 0.032*

AUC, area under the curve; SE, standard error *Statistically significant with a p value < 0.05 3022 Clin Rheumatol (2019) 38:3015–3023

Fig. 1 Receiver operating characteristic (ROC) analysis of the predictive power of SOFA and APACHE II scores, in predicting the mortality (overall and stratified according to the cause of mortality) of patients with rheumatoid arthritis

cardiovascular risk factors [19]. There is substantial evidence studies that included SLE or vasculitis patients that were admit- that persistent systemic inflammation may play a pivotal role tedtoanICU,APACHEIIwasnotidentifiedasanaccurate in promoting the development and progression of heart failure predictor of mortality [10, 14]. In the present study, although through several mechanisms, including myocardial endotheli- APACHE II was not found to have a predictive value for overall al dysfunction, increased arterial stiffness, cardiac muscle fi- morality, it was demonstrated that APACHE II scores may pre- brosis, and predominantly diastolic dysfunction [17, 20–22]. dict ICU mortality due to infections only. Thus, it may be con- Acute liver failure was found to be an additional predictor of cluded that in general, both scores are required to recognize RA mortality in our RA cohort, but not in other previous studies [1]. patients who are at high risk for mortality. Interestingly, in our study, acute liver failure was not correlated The present study has several limitations. Firstly, as this is a with either liver cirrhosis or synthetic or biologic DMARDs. single-center study at a tertiary hospital, the sample size is The use of vasopressor treatment has been previously rec- relatively small (although comparable to other similar stud- ognized as a predictor of poor outcome in the ARDs popula- ies), and the generalizability of these results is therefore lim- tion in ICU [14, 16, 23]. In line with these results, our study ited. In addition, this small sample size has resulted in a rela- identified vasopressor treatment as a risk factor for increased tive lack of power, characterized by wide 95% confidence mortality in the RA cohort in ICU. intervals. Secondly, the prevalence of organ and system- Different studies have evaluated the ability of SOFA and specific dysfunction may have been underestimated due to APACHE II scores to predict mortality in patients with ARDs the inclusion of RA patients that were only admitted to the which were hospitalized in ICU [1]. SOFA was found to be a general ICU. predictor of mortality in patients with vasculitis admitted to the ICU [10, 24, 25], but no conclusive literature exists regarding other ARDs. In the present study, SOFA score was identified as Conclusion a predictor of overall mortality in our RA population, and also a predictor of mortality due to respiratory failure. Our study demonstrated a relatively high mortality rate among APACHE II score remains the most common variable asso- RA patients who required admission to the general ICU. ciated with mortality in ARDs patients [1]. However, in several Infections and respiratory failure were the leading causes of Clin Rheumatol (2019) 38:3015–3023 3023 admission, and infections were the most common cause of 12. Vincent JL, Moreno R, Takala J, Willatts S, De Mendonça A, death. Several predictors for higher mortality rate were recog- Bruining H et al (1996) The SOFA (sepsis-related organ failure assessment) score to describe organ dysfunction/failure. On behalf nized, including heart failure, liver failure, elevated SOFA of the working group on sepsis-related problems of the European score, and vasopressor treatment in ICU. Clinical trials involv- Society of Intensive Care Medicine. Intensive Care Med 22:707– ing larger cohorts are required to investigate the outcomes of 710 patients with RA that are admitted to the general, cardiac, and 13. Garland A, Olafson K, Ramsey CD, Yogendran M, Fransoo R (2015) A population-based observational study of intensive care neurologic ICUs. unit-related outcomes. With emphasis on post-hospital outcomes. Ann Am Thorac Soc 12:202–208 Compliance with ethical standards 14. Namendys-Silva SA, Baltazar-Torres JA, Rivero-Sigarroa E, Fonseca-Lazcano JA, Montiel-López L, Domínguez-Cherit G The study fulfilled the ethical guidelines of the most recent declaration of (2009) Prognostic factors in patients with systemic lupus erythema- Helsinki (Edinburgh, 2000) and received approval (No. 3233-16-SMC) tosus admitted to the intensive care unit. Lupus. 18:1252–1258 by the local ethical committee. 15. Shen HN, Yang HH, Lu CL (2013) Temporal trends in characteristics and outcome of intensive care unit patients with systemic lupus Disclosures None. erythematosus in Taiwan: a national population-based study. Lupus. 22:644–652 16. Bernal-Macías S, Reyes-Beltrán B, Molano-González N, Augusto Vega D, Bichernall C, Díaz LA, Rojas-Villarraga A, Anaya JM References (2015) Outcome of patients with autoimmune diseases in the intensive care unit: a mixed cluster analysis. Lupus Sci Med 2:e000122 1. Quintero OL, Rojas-Villarraga A, Mantilla RD, Anaya JM (2013) 17. Nicola PJ, Maradit-Kremers H, Roger VL, Jacobsen SJ, Crowson Autoimmune diseases in the intensive care unit. An update. CS, Ballman KVet al (2005) The risk of congestive heart failure in Autoimmun Rev 12:380–395 rheumatoid arthritis: a population-based study over 46 years. 2. Bell A, Tattersall R, Wenham T (2016) Rheumatological conditions Arthritis Rheum 52:412–420 in critical care. BJA Education 16:427–433 18. Nicola PJ, Crowson CS, Maradit-Kremers H, Ballman KV, Roger 3. Peschken CA, Hitchon CA, Garland A, Bernstein CN, Chen H, VL, Jacobsen SJ, Gabriel SE (2006) Contribution of congestive Fransoo R, Marrie RA (2016) A population-based study of inten- heart failure and ischemic heart disease to excess mortality in rheu- sive care unit admissions in rheumatoid arthritis. J Rheumatol 43: matoid arthritis. Arthritis Rheum 54:60–67 26–33 19. Crowson CS, Nicola PJ, Kremers HM, O’Fallon WM, Therneau 4. Marrie RA, Hitchon CA, Peschken CA, Chen H, Bernstein CN, TM, Jacobsen SJ, Roger VL, Ballman KV, Gabriel SE (2005) How Garland A (2017) Health care utilisation before and after intensive much of the increased incidence of heart failure in rheumatoid ar- care unit admission in rheumatoid arthritis. Clin Exp Rheumatol 35: thritis is attributable to traditional cardiovascular risk factors and 975–982 ischemic heart disease? Arthritis Rheum 52:3039–3044 5. Pieringer H, Hintenberger R, Pohanka E, Steinwender C, Meier J, 20. Davis JM 3rd, Roger VL, Crowson CS, Kremers HM, Therneau Gruber F, Auer-Hackenberg L (2017) RABBIT risk score and ICU TM, Gabriel SE (2008) The presentation and outcome of heart admission due to infection in patients with rheumatoid arthritis. failure in patients with rheumatoid arthritis differs from that in the – Clin Rheumatol 36:2439 2445 general population. Arthritis Rheum 58:2603–2611 6. Barrett O, Abramovich E, Dreiher J, Novack V, Abu-Shakra M 21. Paulus WJ, Tschöpe C (2013) A novel paradigm for heart failure (2017) Short- and long-term mortality due to sepsis in patients with – with preserved ejection fraction: comorbidities drive myocardial rheumatoid arthritis. Rheumatol Int 37:1021 1026 dysfunction and remodeling through coronary microvascular endo- 7. Dadoun S, Zeboulon-Ktorza N, Combescure C, Elhai M, thelial inflammation. J Am Coll Cardiol 62:263–271 Rozenberg S, Gossec L, Fautrel B (2013) Mortality in rheumatoid 22. Marti CN, Gheorghiade M, Kalogeropoulos AP, Georgiopoulou arthritis over the last fifty years: systematic review and meta-anal- VV,Quyyumi AA, Butler J (2012) Endothelial dysfunction, arterial ysis. Joint Bone Spine 80:29–33 stiffness, and heart failure. J Am Coll Cardiol 60:1455–1469 8. Sherer Y, Shepshelovich D, Shalev T, Haviv Y, Segal E, Ehrenfeld M, Levy Y, Pauzner R, Shoenfeld Y, Langevitz P (2007) Outcome 23. Camargo JF, Tobón GJ, Fonseca N, Diaz JL, Uribe M, Molina F, of patients having dermatomyositis admitted to the intensive care Anaya JM (2005) Autoimmune rheumatic diseases in the intensive unit. Clin Rheumatol 26:1851–1855 care unit: experience from a tertiary referral hospital and review of – 9. Shalev T, Haviv Y, Segal E, Ehrenfeld M, Pauzner R, Levy Y, the literature. Lupus. 14:315 320 Langevitz P, Shoenfeld Y (2006) Outcome of patients with sclero- 24. Frausova D, Brejnikova M, Hruskova Z, Rihova Z, Tesar V (2008) derma admitted to intensive care unit. A report of nine cases. Clin Outcome of thirty patients with ANCA-associated renal vasculitis – Exp Rheumatol 24:380–386 admitted to the intensive care unit. Ren Fail 30:890 895 10. Haviv Y, Shovman O, Bragazzi NL, Sharif K, Yavne Y, Shoenfeld 25. Khan SA, Subla MR, Behl D, Specks U, Afessa B (2007) Outcome Y, Amital H, Watad A (2017) Patients with vasculitides admitted to of patients with small-vessel vasculitis admitted to a medical ICU. the intensive care unit: implications from a single-center retrospec- Chest. 131(4):972–976 tive study. J Intensive Care Med:088506661771722. https://doi.org/ 10.1177/0885066617717223 Publisher’snoteSpringer Nature remains neutral with regard to 11. Knaus WA, Draper EA, Wagner DP, Zimmerman JE (1985) jurisdictional claims in published maps and institutional affiliations. APACHE II: a severity of disease classification system. Crit Care Med 13(10):818–829