Adult‐onset Still's disease in Western Australia: Epidemiology, comorbidity and long‐term outcome

Abstract Aim Adult‐onset Still’s disease (ASD) is a rare, potentially life‐threatening autoinflammatory condition. As reported prevalence shows regional variation and long‐term outcome data are scarce, we investigated epidemiology and long‐term health outcomes of ASD in Western Australia (WA). Methods Population‐based cohort study using longitudinally linked administrative health data from all WA hospitals between 1999 and 2013 for ASD patients (ICD‐10‐AM M06.1) and controls matched for age, gender, and index year. Rate ratios and odds ratios (RR/OR) with 95% confidence intervals (CI) compared ASD patients with controls. Results The average ASD incidence (n = 52) was 0.22/100 000 with 2.4/100 000 point‐prevalence as of December 31, 2013. ASD patients (median age 41.5 years, 59.6% female) had higher odds of previous liver disease (OR 2.67, 95% CI 1.31‐5.45), fever (OR 54.10, 95% CI 6.60‐433.0), rash (OR 15.70, 95% CI 4.08‐60.80), and serious infections (OR 4.36, 95% CI 2.11‐22.80) than controls. Despite biological disease‐modifying antirheumatic drugs in 27% of patients, ASD patients had higher odds for joint replacement (n = 7, 13.5%) (OR 45.5, 95% CI 4.57‐93.70), osteoporosis (OR 31.3, 95% CI 3.43‐97), and serious infections (RR 5.68; 95% CI 6.61‐8.74) during follow up. However, crude mortality (11.5% vs 7.5%; P = 0.34), survival at 1 and 5 years (P= 0.78), and last modified Charlson Comorbidity score (median 2 vs 2) were similar between groups. Conclusion The epidemiology and demographics of ASD in Western Australia fall within the internationally reported range. ASD patients present increased rates of liver disease, rash, and serious infections before disease onset. Mortality following ASD was not increased for 5 years despite high rates of chronic arthritis requiring joint replacement, serious infections, and osteoporosis.


| INTRODUC TI ON
Still's disease (SD) is a rare systemic inflammatory disease of unknown origin, first described in children by Sir George Still more than a century ago. 1 Bywaters et al described in 1971 a series of adult women with clinical features reminiscent of SD and proposed the term adult-onset Still's disease (ASD). 2 ASD has reported prevalence ranges from 1 to 34 cases per million [3][4][5] and typically presents with a mixture of symptoms where daily fever spikes, sore throat, evanescent rash, and polyarthritis in the presence of neutrophilia and elevated acute-phase reactants are the most frequent findings. 3 However, myalgia, lymphadenopathy, fulminant hepatitis, serositis, consumption coagulopathy, and myocarditis can also occur. [6][7][8] ASD is a diagnosis of exclusion and, as a result, is usually only considered when treatment of suspected infections has been unsuccessful. 3,6 ASD is classified as an autoinflammatory disease where a complex interplay of genetic, infectious, and other environmental factors trigger overproduction of proinflammatory cytokines, which drive the clinical manifestations and ultimately can lead to the life-threatening macrophage activation syndrome ("cytokine storm"). 7,8 The clinical course of ASD is unpredictable because it appears self-limiting in some patients but leads to recurrent exacerbations of systemic inflammation and/or the development of chronic deforming arthritis in many others. [9][10][11] With scarce data available from Australasia, we investigated the epidemiological characteristics, previous conditions, and long-term clinical outcomes in patients hospitalized for ASD in Western Australia (WA) over a 14-year period.

| Data sources
Data were derived from the WA Rheumatic Disease Epidemiological Registry (WARDER) that contains routinely collected longitudinal linked health data for patients with rheumatic diseases from hospitals for the entire state of WA as described elsewhere. 12 WARDER also contains a large age-and sex-matched "comparator" group of patients selected from the WA Electoral Roll on the basis of requiring hospital care in the study period but not having a registered diagnostic code for rheumatic disease in any data collection linked through the WA Data Linkage System (WADLS).

| Study cohort
For this population-level observational study we included persons We excluded patients with possible ASD before 1999 because no specific coding for ASD was available in the ICD-9-Clinical modification with code 714.2 "Other rheumatoid arthritis with visceral or systemic involvement" considered too ambiguous. In an earlier study, ICD-10 coding for ASD was found to have 83% positive predictive value for a clinical diagnosis of ASD and 78% sensitivity for fulfillment of the Yamaguchi criteria. 14 For this study, each ASD patient was matched with up to five controls matched for age and gender but also for the same year of requiring hospital admission as the incident ASD case. WARDER controls are not healthy controls because they required hospitalization for a wide range of indications other than inflammatory rheumatic disease (see Figure S1), but survival in WARDER controls has been shown to be similar to that in the general population. 15 Date and primary causes of death were extracted from the WA Death Registry.

| Outcome ascertainment
As the first hospital contact was not necessarily for ASD, we defined a time-zero (T 0 ), which for ASD patients was the date of ASD diagnosis and for each control the date that most closely mirrored T 0 for the matched ASD patient. We defined the lookback period as all observation time before T 0 and follow up as all observation time more than 30 days after T 0 . We defined the occurrence of serious infections as episodes leading to presentation at the Emergency Department and/or hospital admission resulting in an infectious disease code. 13 Study measures were the presence of specific ASD manifestations (Table S1) and the documented accrual of organsystem-specific, as well as overall weighted, comorbidity before and after T 0 according to the validated and prognostically important Charlson comorbidity index (CCI), in-hospital mortality, re-admission rate within 30 days, and survival at 1 and 5 years of T 0 .

| Statistical analyses
Descriptive statistics are presented as median plus interquartile range (IQR) for numeric variables and proportions for categorical variables, unless otherwise indicated. Historical population data for WA were obtained from the Australian Bureau of Statistics (https://www.abs.gov. au/stati stics/ peopl e/popul ation/ natio nal-state -and-terri tory-popul ation/ lates t-relea se#data-downl oads-data-cubes). Average annual incidence and point prevalence rates are given per 100 000 population with the total number of cases as numerator and a denominator based on the adult population in that year. A generalized log-linear regression model (Poisson) was used to analyze the trend in the number of cases per year. Differences for numeric results were compared by non-parametric methods (Kruskal-Wallis) and for proportions by χ 2 test with Yates correction where needed. All-cause hospitalization and Emergency Department visit rates (expressed as number per 100 person-years at risk) and odds for comorbidity/complications in ASD patients and controls were compared by conditional maximum likelihood estimates of odds ratios (OR) and rate ratios (RR) with 95% confidence intervals (CI). Kaplan-Meier estimates were used to compare survival between ASD and control with P values presented from the log-rank test. Analyses were performed using SPSS v27.0 software (IBM, Armonk, NY, USA) with two-sided P-values less than 0.05 considered to be statistically significant.

| Ethics
This project was approved by the Human Research Ethics Committee at the WA Department of Health (HREC 2016.24) with the condition to prevent potential identification by confidentializing small numbers (n < 5).

| RE SULTS
There were 52 incident cases of ASD in the study period, comprising 31 females (59.6%, median age 42 years) and 21 males (40.4%, median age 39 years). The average annual incidence for ASD was 0.22/100 000 (95% CI 0.14-0.32) and did not change significantly Demographic and previous medical details of the ASD patients (Table 1) showed few gender differences regarding age at onset, regional presentation, or insurance status. Median length of hospital stay for first ASD presentation was 9 days (IQR 4.5-21.5) with some patients (n < 5, 5.7%) requiring intensive care unit admission.
During a median lookback period of 203 months (IQR 40-280), ASD patients had higher rates/100 person-years for hospital admission and ED presentations before diagnosis than controls as well as higher odds of being diagnosed with liver disease, fever of unknown origin, rash, and serious infections ( Table 2). Forty-eight ASD patients (92.3%) had at least one hospital contact in the year before diagnosis. Most ASD patients (89.5%) and controls (96.9%) had recorded comorbidity before T 0 and although median CCI scores were similar between the groups, slightly more ASD patients had multimorbidity (m-CCI ≥2) ( Table 2).
Readmission within 1 month after discharge was more frequent in ASD patients than controls (19.2% vs 9.6%, P = 0.003) with five out of ten ASD patients readmitted with disease flare as the primary diagnosis. During a median follow up of 49 months (IQR 24-84) a total of 36 ASD flares occurred in 13 (25%) patients for a flare rate of 14.7/100 person-years (Table 3). Forty-eight ASD patients and 177 controls required hospital care beyond 30 days of follow up (94% vs 78%, respectively; OR 3.45, 95% CI 1.27-11.7) with higher rates for both admission and Emergency Department visits in ASD patients (Table 3). Among these ASD patients a total  (Table 3). There was however no significant difference between ASD patients and controls in crude mortality (6/52, 11.5% vs 18/228, 7.5%; P = 0.34) or survival rates over 5 years (P = 0.78) (Figure 2). Adjusting for the baseline presence of DM or CVD did not significantly influence survival rates between patients and controls (log-rank P = 0.74 and P = 0.82, respectively). The most frequent cause of death was malignancy in ASD patients (n = 6) (melanoma in two, metastasis of unknown primary in one with no registered cases of lymphoproliferative malignancies) and cardiovascular events in controls (n = 18) ( Figure S2).

| DISCUSS ION
In this population-based study stretching over 14 years, the aver-  Baseline comorbidity scores (without the rheumatic disease component, because this was excluded in controls) at ASD diagnosis were not higher than in non-ASD controls in our study, although the average CCI scores in the US survey, which included the score for rheumatic disease, were reportedly higher in ASD patients than controls (1.33 vs 0.81). 18 The specific comorbid conditions that could potentially predispose to ASD in this study were a higher rate of previous serious infections and existing liver disease.
Although infections have long been considered a trigger for ASD in (genetically) predisposed individuals, 23 Note: Small numbers are given as <5 because of Human Research Ethics Committee requirements to prevent identification. Data represent median values (interquartile range) or frequency (%) with rate ratio and odds ratios (95% confidence intervals).
TA B L E 2 Patient characteristics and previous medical conditions in ASD patients and controls during lookback period microbiological confirmation, it is less likely that ASD disease flares were misdiagnosed as serious infections, but it has been suggested that a particular combination of pathogenic microorganisms with genetic susceptibility (variations in, for example, human leukocyte antigen or cytokine genes) are likely essential determinants of ASD development and severity. 23,25 The prevalence of chronic lung disease over a nearly 20-year period before diagnosis was high although not different for ASD patients and controls. The relevant CCI codes include conditions such as bronchitis, smoking, asthma, and emphysema, which are not specific for ASD and are frequent diagnoses in hospitalized patients. Although abnormal liver enzymes are a recognized feature of ASD, it is unclear whether the higher rate of pre-existing liver disease in ASD patients versus controls (28.8% vs 13.2%) reflected impending ASD or classifies as a specific risk factor that will require further study.
There are limited data on long-term survival in ASD. 26 Crude mortality was 11.5% over a median follow up of 4.2 years in this study, which falls within the 25% mortality rate (n = 2 of 8) from  18 and while this may partly be due to immunomodulating treatment, the fact that a considerable proportion of ASD patients also had serious infections before diagnosis lends some support to the idea of an inherent susceptibility to infection in ASD patients. 36 The strength of this study is the reliance on a validated database with good diagnostic accuracy, reliable data linkage, long term follow up and inclusion of an age-and gender-matched control group to determine differences in the main outcomes, especially serious infection and mortality. The limitations of this study relate to the fact that our data are based on a physician-based discharge diagnosis of ASD and lack the detailed clinical and laboratory data to determine if patients fulfilled ASD classification criteria. However, a recent chart review found a 78% sensitivity for fulfilling Yamaguchi criteria in administrative hospital data, which reduces the likelihood of diagnostic error. 14 The population-wide capture of ASD through mandatory reporting of hospital discharge diagnoses also makes it unlikely that we have missed the rare ASD patients seen solely as outpatients. Nonetheless, the relatively small number of cases with wide confidence intervals makes it challenging to compare outcome measures and a larger (national) cohort with even longer follow up may provide additional information.

| CON CLUS IONS
ASD is as uncommon in WA as in other regions. Serious infection appears to be a major risk factor. Despite high rates of chronic arthritis and joint replacement, osteoporosis and serious infections, 5-year survival is not significantly negatively impacted. This may be because the most serious complication, hemophagocytic lymphohistiocytosis/MAS, is less common in population cohorts compared with specialist center studies.