Mortality and causes of death in patients with Sjögren disease in Western Australia: a population-based study

Abstract

Objectives

Sjögren disease (SjD) is a systemic autoimmune disorder with emerging evidence showing increased mortality rates compared with the general population. We compared mortality risk and causes of death in Western Australian (WA) SjD patients with age/sex-matched controls and general population mortality rates.

Methods

The WA Rheumatic Disease Epidemiological Registry (WARDER) was accessed in this data linkage study to identify patients with primary SjD, SjD associated with rheumatoid arthritis (SjD + RA) and SjD associated with other connective tissue disease (SjD + CTD) using ICD codes and age/sex matched controls (1:5) between 1980 and 2013. Mortality of SjD cohorts was compared with both age/sex-matched controls from WA and population mortality rates. Multiple cause of death and competing risk analysis was performed for SjD cohorts compared with controls.

Results

A total of 3716 controls (1345 deaths) were compared with 457 primary SjD (204 deaths), 218 SjD + RA (149 deaths) and 77 SjD + CTD (38 deaths). Risk of mortality was increased in primary SjD [HR 1.8 (95% CI 1.5, 2.0)] and in associated SjD {SjD + RA: HR 2.1 [95% CI (1.8, 2.5)]; SjD + CTD: HR 3.1 (95% CI 2.2, 4.3)} compared with controls. Survival in primary SjD was 77% (95% CI 0.63, 0.94) at the end of follow-up (25 years). Mortality risk in primary and associated SjD was increased due to infection, ischaemic heart disease, lymphoid malignancy and interstitial lung disease.

Conclusion

All-cause and specific-cause mortality was increased in primary and associated SjD compared with age/sex-matched controls in an Australian population.

Graphical abstract

Graphical Abstract.

Rheumatology key messages.

• All-cause mortality was increased in Western Australian patients with primary and associated SjD.

• Primary and associated SjD patients had increased mortality from infection, lymphoid malignancy, cardiovascular and respiratory disease when compared with controls.

• The increased mortality in SjD highlights the critical need for disease-modifying therapies.

Introduction

Sjögren disease (SjD) is the second most common autoimmune rheumatic disease (AIRD), with international prevalence conservatively estimated between 0.01% and 0.05% (noting significant heterogeneity in this analysis) [1] and limited epidemiological data in the Australian context [2]. SjD affects exocrine glands causing dry eyes and mouth, in addition to serious systemic manifestations such as interstitial lung disease (ILD), lymphoma and vasculitis. SjD can be associated with other connective tissue diseases (CTD) (known as ‘associated SjD’) [3], occurring predominantly with rheumatoid arthritis (RA) (53%), systemic lupus erythematosus (SLE) (28%) and systemic sclerosis (13%) [4]. Since there are no approved treatments that prevent disease progression, SjD is managed empirically [5]. While increased lymphoma risk is well-characterized in SjD [6], emerging evidence suggests infection, cardiovascular and respiratory disease are also significant causes of mortality [7]. Knowledge regarding SjD mortality in Australia is lacking, including for Aboriginal and Torres Strait Islander peoples [8].

People with primary SjD have increased mortality rates compared with the general population. A large international cohort study by Brito-Zerón et al. (N = 11 372) demonstrated two-fold higher risk of all-cause and SjD-related death in patients with systemically active disease, with cardiovascular disease and infection causing almost 60% of deaths and malignancy causing 18% [9]. Similarly, a systematic review reported increased all-cause mortality in primary SjD {standardized mortality ratio (SMR) 1.46 [95% confidence interval (CI) 1.10, 1.93)]} compared with the general population [10]. However, data on mortality in associated SjD are limited and mortality is thought to be driven by the underlying AIRD [11, 12].

Between 5% and 10% of SjD patients develop non-Hodgkin lymphoma (NHL) [6] (up to 44-fold increased risk vs the general population [1, 13, 14]) which accounts for 30% of SjD-related deaths [7]. Evidence is conflicting regarding solid organ tumour risk in SjD, with a recent systematic review reporting only increased risk of thyroid cancer [relative risk (RR) 2.58; 95% CI 1.14–4.03], but not other solid organ malignancies [15]. Subsequently, a longitudinal cohort study reported elevated solid tumour risk in SjD patients [SIR 1.54 (95% CI 1.53, 1.55)] [16] and a prospective cohort study found that solid organ malignancies contributed to 14.4% of all deaths in SjD patients, primarily due to breast cancer, cholangiocarcinoma and hepatocarcinoma [7].

Given international evidence about increased mortality in SjD, the lack of Australian data regarding SjD mortality and limited, conflicting evidence surrounding causes of mortality in SjD patients compared with controls, we conducted this data-linkage study to investigate mortality risk and cause of death in SjD patients in Western Australia (WA).

Methods

Data sources

The WA Rheumatic Disease Epidemiological Registry (WARDER) contains population-level health data for patients requiring public or private hospital-level care for rheumatic diseases in WA (1980–2015). WARDER data is derived from the Hospital Morbidity Data Collection (HMDC; 1980–2015), Emergency Department Data Collection (EDDC; 2002–2015), WA Cancer Registry (1982–2015) and WA Births, Deaths and Marriages Registry (1980–2015) via the WA Data Linkage System (WADLS). International Classification of Diseases (ICD) Ninth Revision Clinical Modification (CM) (ICD-9-CM) was used until 1998, when it was replaced by the Tenth Revision Australian Modification (ICD-10-AM). The high linkage accuracy (99.7%) of WADLS enables longitudinal follow-up of individuals across multiple health datasets [17]. Aboriginal and/or Torres Strait Islander status was defined via self-report (in accordance with Australian Institute of Health and Welfare recommendations [18]).

Cohort definitions

Patients with SjD were identified from hospital admission and emergency presentation data by ICD diagnosis codes [710.2 (ICD-9-CM) or M35.0 (ICD-10-AM)] and classified as either primary SjD (if no codes were identified for other rheumatological conditions) or associated SjD [if codes were present for RA (SjD + RA) or other CTD (SjD + CTD)] (see Supplementary Table S1). The date of the first ICD code for SjD was the patient index date, reflecting first capture rather than SjD onset in this prevalent cohort. Controls for each SjD patient (5:1 ratio), matched for birth year, sex and Aboriginal and/or Torres Strait Islander status, were selected from WADLS patients requiring hospital-level care (but with no relevant rheumatological diagnosis) and assigned the same index date as their matched SjD patient, then controls were subsequently pooled for analysis.

Outcomes

The primary outcome was death and the secondary outcome was ICD-coded cause of death (COD) derived from the Medical Certificate of Cause of Death. COD consisted of both underlying (or direct) COD (which was only unambiguously assigned for the SjD cohort) and multiple causes of death data (MCoD). The underlying COD is ‘the disease or injury which initiated the train of morbid events leading directly to death’ [19]. The contributory COD are conditions that gave rise to the underlying COD, even with long intervals between the condition and death [19]. MCoD data incorporates underlying and contributory conditions, allowing more complete examination of all contributing COD (including chronic conditions and multimorbidity) on health outcomes [20].

Death linkage data was available between July 1980 and December 2014 for SjD patients and between November 1988 and December 2014 for controls. MCoD data were derived from record axis data, in which all information recorded on the death certificate is processed and captured as ICD codes. These data were available for the subset of deaths occurring between December 1996 and December 2013 in both SjD patients and controls, captured in up to 20 variables within the death linkage dataset.

A supplementary outcome was the number of comorbidities derived from the Elixhauser comorbidity index.

Statistical analysis

Statistical analyses and visualization were performed using Stata software (v19.0, StataCorp LLC, College Station, TX, USA). Analysis primarily involved time-to-event (survival analysis) with both left censoring at the maximum of 15th birthday/linkage start date, and right censoring at the minimum of date of death/99th birthday/linkage end date. The analysis primarily compared primary SjD and all SjD cohorts combined with controls due to insufficient numbers of patients in associated SjD cohorts for reliable inference.

Demographic data at diagnosis for patients and controls were summarized using frequencies (percentages) or median and interquartile range (IQR).

Mortality and underlying COD in SjD compared with population mortality rates: SMR and relative survival

Expected values for relative survival and SMR calculations utilized age, sex and calendar year-matched WA population mortality rates (available from the Australian Bureau of Statistics Data Explorer Datasets [21]) and for underlying COD utilized Australian age, sex and calendar year-matched cause-specific mortality rates from the General Record of Incidence of Mortality (GRIM) books [22].

Marginal relative survival (the ratio of observed to expected survival) was estimated using the Stata ado ‘stpp’ [23] which estimates the relative survival function at each death time. SMR is the ratio of the cumulative observed deaths to cumulative expected deaths. Expected deaths were estimated by a life-table approach (1-year intervals) using the Stata ado ‘strs’ [24] adopting the Ederer II method [25] and SMR calculated by Poisson regression (with a robust variance-covariance matrix). Results were age-standardized by weighting to the Australian 2001 standard population (age groups 15–44, 45–54, 55–64, 65–74, 75+ years) ensuring comparability between results for the different cohorts.

SMR were calculated for both overall mortality and underlying COD. The ICD codes used for analysis of underlying COD are reported in Supplementary Table S2.

Mortality and MCoD analysis in SjD compared with controls

Survival models for mortality in SjD compared with controls utilized age as the analysis time scale, with delayed entry to accommodate age differences at time of entry into the analysis. Overall mortality in SjD cohorts was compared with control mortality rates by Cox regression, with hazard ratios (HR) estimated under the proportional hazards assumption. Time-varying effects (violation of the proportional hazards assumption) were investigated using a Royston-Parmar model, estimated using the Stata ado ‘stpm3’ [26] utilizing splines to model the time scale. The best fitting model, as judged by the Akaike information criteria, was equivalent to a Weibull model, df [1], with time-varying effects, also df [1], for each participant subgroup.

Classification of specific COD listed in MCoD data was based around the hierarchical structure of major ICD chapters (with the addition of dementia, diabetes, renal disease and liver disease). The ICD codes utilized are reported in Supplementary Table S2.

Analysis utilized a competing risk approach (Fine-Gray Model) with individual deaths classified as either including the COD of interest or not (a competing event). This enables estimation of the cumulative incidence function (CIF), the probability that an individual COD contributed to death. Results are reported as subHazard Ratios (sHR), which measure the effect of covariates on the CIF, for both primary SjD and all SjD compared with controls.

Comorbidity burden

The Elixhauser Comorbidity Index is a validated method for identifying medical conditions (such as malignancy, heart failure, diabetes, alcohol use and obesity) using ICD codes to predict key patient outcomes (e.g. in-hospital mortality) [27]. Comorbidities were assessed at each hospital admission utilizing hospital discharge diagnosis information, which included up to 22 ICD-coded diagnosis variables: a principal diagnosis, co-diagnosis and 20 additional diagnosis variables.

Participants were included if they had at least one hospital event after their index date. Individual comorbidities were assessed cumulatively during follow-up, using available hospital admission data, with the Stata ado elixhauser [28] which utilizes the Quan et al. [29] ICD coding for 31 comorbidities. The total number of comorbidities for each individual by the last hospital admission (excluding the ‘Rheumatoid Arthritis, Collagen/Vascular Disease’ component and collapsing over diabetes and hypertension components) was analysed between patient subgroups by negative binomial regression, adjusted for sex and age (at last hospital admission) classified into five groups (20–44, 45–54, 55–64, 65–74, 75+ years old) with reporting of marginal population-averaged means.

Mortality in SjD patient cohorts vs matched controls was compared using Cox proportional hazards regression. In a sensitivity analysis, mortality HRs were estimated across three models: unadjusted (Model 1) and adjusted for covariates sex (Model 2) and sex plus Elixhauser comorbidities (Model 3).

Ethics approval

The project was approved by the Human Research Ethics Committee at the WA Department of Health (HREC # 2016.24).

Results

Cohort demographic characteristics

The study population characteristics are shown in Table 1. The analysis included 3716 controls (1345 deaths) over a maximum of 25 years of follow-up and 457 primary SjD (204 deaths), 218 SjD + RA (149 deaths) and 77 SjD + CTD (38 deaths) patients, with a total of 752 SjD patients (391 deaths) over a maximum of 33 years of follow-up. The cohorts were predominantly female [controls: 87% (N = 3246); primary SjD: 86% (N = 391); SjD + RA: 89% (N = 195); SjD + CTD 92% (N = 71), respectively]. The SjD + CTD patients were approximately a decade younger at their index date {median (IQR) age: [54 (44, 70) years] compared with pooled controls [63 (51, 73) years], primary SjD [64 (50, 75) years] and SjD + RA [65 (56, 72) years]}. Aboriginal and/or Torres Strait Islander peoples comprised 2% (N = 14) of the primary and associated SjD cohort and 2% (N = 69) of controls. The mean number of cumulative Elixhauser comorbidities (age and sex adjusted) in 3190 controls and 746 SjD patients was higher in primary SjD (3.8, 95% CI 3.4, 4.1) than controls (1.9, 95% CI 1.8, 2.0).

Table 1.

Sjögren disease and control cohort characteristics.

	Controls (N = 3716)	Primary SjD (N = 457)	SjD + RA (N = 218)	SjD + CTD (N = 77)
Deaths, n	1345	204	149	38
Follow-up (person-years)	46736.4	4527.5	2541.1	947.0
Females, n (%)	3246 (87%)	391 (86%)	195 (89%)	71 (92%)
Index age (years), median (IQR)	63 (51, 73)	64 (50, 75)	65 (56, 72)	54 (44, 70)
Aboriginal and/or Torres Strait Islander, n (%)	69 (2%)	7 (2%)	≤5	≤5
Elixhauser comorbidities (95% CI)a	1.9 (1.8, 2.0)	3.8 (3.4, 4.1)	4.7 (4.2, 5.3)	6.0 (4.8, 7.3)

^aThe mean cumulative number of Elixhauser comorbidities by the final hospital admission, adjusted for age and sex.

CI: confidence interval; CTD: connective tissue disease; IQR: interquartile range; RA: rheumatoid arthritis; SjD: Sjögren disease.

Mortality risk in SjD

We first evaluated all-cause mortality in SjD compared with WA population mortality rates, using relative survival (Fig. 1) and SMR (Table 2). Relative survival was decreased in all SjD cohorts [primary SjD: 0.77 (95% CI 0.63, 0.94), SjD + RA: 0.68 (95% CI 0.55, 0.85), SjD + CTD: 0.57 (95% CI 0.36, 0.91)] compared with WA population mortality rates (Supplementary Table S3), which was reflected in SMR significantly >1. Of note, all-cause mortality in our age/sex-matched controls was comparable to WA population rates [relative survival at 25 years 0.99 (95% CI 0.95, 1.04)].

Figure 1.

Marginal Relative Survival for Sjögren disease cohorts and controls compared with all-cause population mortality rates. Shaded areas represent 95% confidence intervals. (A) Marginal relative survival in age- and sex-matched controls compared to population mortality rates. (B) Marginal relative survival in primary SjD compared to population mortality rates. (C) Marginal relative survival in SjD + RA compared to population mortality rates. (D) Marginal relative survival in SjD + CTD compared to population mortality rates. CTD: connective tissue disease; RA: rheumatoid arthritis; SjD: Sjögren disease

Table 2.

Standardized mortality ratios and mortality hazards ratios for Sjögren disease cohorts.

Cohorts	SMR (95% CI)a	P-value	HR (95% CI)b	P-value
Controls	0.9 (0.9, 1.01)	0.10	1	–
Primary SjD	2.2 (1.8, 2.6)	<0.001	1.8 (1.5, 2.0)	<0.001
SjD + RA	2.9 (2.3, 3.6)	<0.001	2.1 (1.8, 2.5)	<0.001
SjD + CTD	3.3 (2.2, 4.9)	<0.001	3.1 (2.2, 4.3)	<0.001
All SjD combined	2.8 (1.7, 4.4)	<0.001	2.0 (1.8, 2.2)	<0.001

^aAll-cause standardized mortality ratios for Sjögren disease cohorts compared with Western Australian population mortality rates at 25 years of follow-up.

^bAll-cause mortality hazards ratios for Sjögren disease cohorts compared with age- and sex-matched controls.

CI: confidence interval; CTD: connective tissue disease; RA: rheumatoid arthritis; SjD: Sjögren disease; SMR: standardized mortality ratio. Bold text indicates significant P-values.

All-cause mortality HR were increased in each SjD cohort compared with controls (Table 2), reflecting the same trends as SMR. The increased mortality for all SjD compared with controls was HR 2.0 (95% CI 1.8, 2.2). However, there was evidence of violation of the proportional hazards assumption (i.e. HR were not constant over the period of follow-up) for each SjD group (P < 0.001), suggesting that mortality risk varied over time. Adjustment for sex had minimal effect on the mortality HR, as SjD cohorts and controls were well matched for sex; however, additional adjustment for the covariate Elixhauser comorbidity burden resulted in modest attenuation of the mortality HR in the SjD cohorts, which remained significant (Supplementary Table S4).

The hazard rates for each participant subgroup, allowing for non-proportional hazards, is shown in Supplementary Fig. S1A. The time-varying mortality hazard rates increased substantially from 60 years of age for SjD cohorts compared with controls. Supplementary Fig. S1B demonstrates that the time-varying effect is a modest increase in HR with increasing age.

Underlying COD in SjD

We next compared underlying COD in primary and combined SjD compared with Australian population cause-specific mortality rates. The most frequent underlying COD were circulatory diseases, neoplasms, respiratory disease and musculoskeletal diseases, collectively accounting for ∼75% (294/391) of deaths, all increased in SjD compared with Australian population cause-specific mortality rates (Table 3). Although less frequent (N = 10), death due to infection was also increased in the combined SjD cohort.

Table 3.

Underlying cause of death: standardized mortality ratios in Sjögren disease patients compared with Australian population cause-specific mortality rates.

GRIM category	Primary SjD			All SjD combined
	Deaths, n	SMR (95% CI)	P-value	Deaths, n	SMR (95% CI)a	P-value
Circulatory diseases	73	1.4 (1.2 1.8)	0.001	131	1.7 (1.3, 2.1)	< 0.001
Coronary heart disease	40	1.6 (1.1, 2.2)	0.007	69	1.9 (1.4, 2.5)	<0.001
Cerebrovascular disease	10	0.8 (0.3, 2.0)	0.56	16	0.9 (0.5, 1.5)	0.61
Neoplasms	43	2.3 (1.2, 4.6)	0.02	66	1.9 (1.1, 3.4)	0.03
Lung cancer	9	2.9 (0.9, 9.1)	0.06	15	3.0 (1.3, 7.0)	0.01
Respiratory diseases	22	3.5 (1.7, 7.2)	<0.001	47	4.2 (2.7, 6.5)	<0.001
Infections	6	–	–	10	7.1 (2.5, 19.9)	<0.001

^aAge standardized to the Australian 2001 Standard Population.

CI: confidence interval; GRIM: General Record of Incidence of Mortality books; SjD: Sjogren disease; SMR: standardized mortality ratio. Bold text indicates significant P-values.

MCoD in SjD

We then examined MCoD for primary SjD and combined SjD compared with age- and sex-matched controls. All-cause mortality sHR for this data subset were substantively the same as those reported in Table 2 for the full dataset (data not shown). Findings from the MCoD analysis (Table 4) were broadly similar to the underlying COD analysis for the full cohort (Table 3), but enabled a more nuanced analysis of specific causes in the presence of controls.

Table 4.

Multiple cause of death analysis in patients with Sjögren disease compared with controls.

Cause of death	Group	Deaths, n	sHR (95% CI)	P-value
Chapter I: Certain Infectious and Parasitic Diseases	Controls	57	1	–
	Primary SjD	17	2.8 (1.6, 4.8)	<0.001
	All SjD	38	3.7 (2.5, 5.6)	<0.001
Chapter II: Neoplasms	Controls	269	1	–
	Primary SjD	41	1.4 (1.0, 2.0)	0.04
	All SjD	62	1.3 (0.9, 1.7)	0.13
Solid organ malignant neoplasms	Controls	254	1	–
	Primary SjD	35	1.3 (0.9, 1.8)	0.20
	All SjD	53	1.1 (0.8, 1.5)	0.48
Lung cancer	Controls	45	1	–
	Primary SjD	7	–	–
	All SjD	10	1.2 (0.6, 2.4)	0.63
Malignant neoplasms of lymphoid, hematopoietic and related tissue	Controls	19	1	–
	Primary SjD	10	4.7 (2.2, 10.0)	< 0.001
	All SjD	12	3.4 (1.6, 6.9)	0.001
Chapter IX: Diseases of the Circulatory System	Controls	564	1	–
	Primary SjD	89	1.6 (1.3, 2.0)	<0.001
	All SjD	147	1.6 (1.3, 1.9)	<0.001
Coronary heart disease	Controls	246	1	–
	Primary SjD	47	1.9 (1.4, 2.6)	<0.001
	All SjD	69	1.6 (1.2, 2.1)	0.001
Other (cardiomyopathies, dysrhythmias, etc)	Controls	153	1	–
	Primary SjD	21	1.3 (0.8, 1.98)	0.35
	All SjD	35	1.2 (0.9, 1.78)	0.28
Hypertensive disease	Controls	131	1	–
	Primary SjD	17	1.2 (0.7, 1.98)	0.76
	All SjD	31	1.2 (0.9, 1.90)	0.23
Cerebrovascular disease	Controls	146	1	–
	Primary SjD	18	1.1 (0.7, 1.85)	0.66
	All SjD	27	1.0 (0.7, 1.50)	0.96
Heart failure	Controls	149	1	–
	Primary SjD	10	0.6 (0.3, 1.10)	0.09
	All SjD	22	0.8 (0.5, 1.2)	0.26
Pulmonary circulation disorders	Controls	13	1	–
	Primary SjD	≤5	–	–
	All SjD	13	5.4 (2.5, 11.6)	<0.001
Chapter X: Respiratory Diseases	Controls	303	1	–
	Primary SjD	57	1.8 (1.4, 2.5)	<0.001
	All SjD	108	2.2 (1.7, 2.7)	<0.001
Respiratory infections	Controls	955	1	–
	Primary SjD	158	1.9 (0.6, 5.4)	0.26
	All SjD	266	2.2 (1.0, 5.1)	0.054
Obstructive pulmonary disease	Controls	89	1	–
	Primary SjD	17	1.8 (1.0, 3.0)	0.03
	All SjD	31	1.9 (1.3, 2.9)	0.002
Bronchiectasis	Controls	≤5	–	–
	Primary SjD	6	–	–
	All SjD	13	14.1 (5.1, 39.2)	<0.001
Interstitial lung disease	Controls	≤5	≤5	–
	Primary SjD	9	20.4 (6.7, 61.8)	<0.001
	All SjD	15	20.9 (6.4, 61.8)	<0.001
Chapter III: Diseases of the blood and blood-forming organs & certain disorders involving the immune mechanism	Controls	23	–	–
	Primary SjD	8	3.2 (1.4, 7.1)	0.006
	All SjD	15	3.5 (1.8, 6.7)	<0.001
Other: Dementia	Controls	200	1	–
	Primary SjD	20	0.9 (0.6, 1.4)	0.55
	All SjD	31	0.8 (0.6, 1.2)	0.26
Other: Diabetes	Controls	76	1	–
	Primary SjD	16	2.0 (1.1, 3.4)	0.015
	All SjD	21	1.5 (0.9, 2.4)	0.11
Other: Renal disease	Controls	69	1	–
	Primary SjD	14	1.8 (1.0, 3.3)	0.038
	All SjD	18	1.4 (0.8, 2.3)	0.21
Other: Liver disease	Controls	10	1	–
	Primary SjD	8	–	–
	All SjD	10	5.3 (2.2, 12.6)	<0.001
Rheumatological disease a	Controls	≤5	–	–
	Primary SjD	22	–	–
	All SjD	66	–	–
Sjögren disease b	Controls	≤5	–	–
	Primary SjD	15	–	–
	All SjD	19	–	–

^aNo comparative analysis performed as insufficient control numbers. Rheumatological disease incorporates the ICD codes listed in Supplementary Table S2, including connective tissue diseases, rheumatoid arthritis, polymyalgia rheumatica and vasculitis.

^bNo comparative analysis performed as insufficient control numbers.

CI: confidence interval; subHR: subhazard ratio; SjD: Sjögren disease. Bold text indicates significant P-values.

Analysis was restricted to deaths which occurred between December 1996 and December 2013 when MCoD data was available. Analysis numbers consisted of 3427 controls (955 deaths), 415 primary SjD (158 deaths), 156 SjD + RA (81 deaths) and 64 SjD + CTD (27 deaths). Results are reported as subHazard Ratios (sHR) from a competing risks analysis with death due to other causes the competing event.

Combined SjD cohorts showed significantly increased mortality from infections [sHR 3.7 (95% CI 2.5, 5.6)] and circulatory disease [sHR 1.6 (95% CI 1.3, 1.9)] including ischaemic heart disease [sHR 1.6 (95% CI 1.2, 2.1)] and pulmonary circulation disorders [sHR 5.4 (95% CI 2.5, 11.6)] compared with age/sex-matched controls. Increased respiratory deaths [sHR 2.2 (95% CI 1.7, 2.7)] in combined SjD cohorts were due predominantly to ILD [sHR 20.9 (95% CI 6.4, 61.8)] and obstructive pulmonary disease [sHR 1.9 (95% CI 1.3, 2.9)] [including bronchiectasis (sHR 14.1 (95% CI 5.1, 39.2)] compared with controls, but there was no significant increase in mortality due to respiratory infections.

While mortality was increased in combined SjD cohorts from lymphoid neoplasms [sHR 3.4 (95% CI 1.6, 6.9)] compared with age- and sex-matched controls, death due to neoplasms overall [sHR 1.3 (95% CI 0.9, 1.7), P = 0.13] and solid organ malignancies [sHR 1.1 (95% CI 0.8, 1.5), P = 0.48] were not significantly elevated. Cross-referencing with International Classification of Diseases for Oncology (ICD-O) codes in cancer linkage data showed that the majority of haematological cancers in SjD were NHL/B cell lymphomas (ICD-O 9590–9729, N = 8). MCoD analysis did not confirm the association with lung cancer deaths observed in the underlying COD analysis (Table 3). The CIF for the most frequently identified major ICD chapter COD in SjD (cardiovascular diseases, respiratory diseases, neoplasms and infections) compared with controls is shown in Fig. 2, with increased cumulative incidence of these COD throughout the lifespan in SjD cohorts compared with controls, consistent with time-varying mortality patterns described earlier. In addition, the CIF represents the probability that a death with a specific cause has occurred by a given age, with the CIF over a lifespan providing an estimate of the average proportion of SjD patients who have died with these specific causes.

Figure 2.

Cumulative Incidence Function (by age) for the four major International Classification of Diseases Codes chapter classifications most frequently contributing to mortality in Sjögren disease patients and controls (multiple cause of death analysis). (A) Cumulative incidence function (CIF) of circulatory diseases in SjD cohorts by age. (B) CIF of respiratory diseases in SjD cohorts by age. (C) CIF of neoplasms in SjD cohorts by age. (D) CIF of infections in SjD cohorts by age. CTD: connective tissue disease; CIF: cumulative incidence function; RA: rheumatoid arthritis; SjD: Sjögren disease

Discussion

Significant knowledge gaps exist regarding risk and causes of mortality in primary and associated SjD in Australia. We found all-cause mortality risk was increased in SjD patients compared with WA population mortality rates. The increased all-cause mortality risk that we observed in primary SjD patients [SMR 2.2 (95% CI 1.8, 2.6)] compared with WA population mortality rates is comparable to findings of larger international cohorts [SMR 1.46 (95% CI 1.10–1.93)] [10]. Higher all-cause risk of mortality was also observed in SjD + CTD [SMR 3.3 (95% CI 2.2, 4.9)] and SjD + RA [SMR 2.9 (95% CI 2.3, 3.6)], consistent with international reports of higher all-cause mortality risk in the associated AIRD {SLE [SMR 2.66 (95% CI 2.09, 3.39)] [30], systemic sclerosis [SMR 2.72 (95% CI 1.93, 3.83)] [31] and RA [SMR 1.59 (95% CI 1.34, 1.70)] [32]} and previous WA data [33–35]. Our findings are generally consistent with previously published mortality risk for primary SjD and the associated CTD. All-cause mortality HR were also increased in each SjD cohort compared with age- and sex-matched controls, reflecting the same trends as population SMR.

Marginal relative survival was decreased in all SjD cohorts compared with WA population mortality rates, most significantly for the SjD + CTD cohort. Importantly, the marginal relative survival of the age- and sex-matched control population was comparable to WA population mortality rates at 25 years, demonstrating that survival rates of age- and sex-matched controls were representative of the WA population.

Our MCoD analysis indicated that circulatory disease was an important contributor to mortality in SjD patients [sHR 1.6 (95% CI 1.3, 1.9)], with significantly increased mortality risk due to ischaemic heart disease and pulmonary circulation disorders (including pulmonary embolism, primary and secondary pulmonary hypertension) compared with matched controls. A recent systematic review found no significant association between SjD and cardiovascular mortality [RR 1.48 (95% CI 0.77, 2.85), P = 0.24], despite SjD patients having higher risk for cardiovascular morbidity such as coronary artery disease [RR 1.34 (95% CI 1.06, 1.38), P = 0.01] vs control populations [36]. We did not have information surrounding other cardiovascular risk factors in our cohort (e.g. smoking rates), which may have contributed to our finding of elevated cardiovascular mortality.

Respiratory disease contributed significantly to mortality in our SjD cohorts, especially ILD [sHR 20.9 (95% CI 6.4, 61.8)], bronchiectasis [sHR 14.1 (95% CI 5.1, 39.2)] and obstructive pulmonary disease [sHR 1.9 (95% CI 1.3, 2.9)], underlining the importance of identification and management of respiratory manifestations in SjD [37]. Patients with SjD-related ILD by definition have systemic disease activity [37], known to be associated with increased mortality [9].

Elevated mortality risk due to lymphoid neoplasms in our cohort compared with matched controls [sHR 3.4 (95% CI 1.6, 6.9)] is supported by previous literature [7] and also reflects systemically active SjD. We found conflicting evidence of mortality from solid organ malignancy in SjD patients, with increased risk of mortality from lung cancer in SjD patients compared with Australian population cause-specific mortality rates [SMR 3.0 (95% CI 1.7, 5.2)], but no significantly increased risk of mortality from solid organ malignancies overall in SjD patients [sHR 1.1 (95% CI 0.8, 1.5), P = 0.48] compared with age- and sex-matched controls. Multiple studies have shown elevated solid tumour risk in SjD [7, 16, 38] with malignancy contributing to a significant proportion of overall deaths [7, 9]. Our conflicting findings about solid organ malignancy may be due to small numbers of deaths from malignancy (N = 46), a relatively short follow-up window of 25 years and publicly funded access to malignancy screening and treatment in Australia for citizens and permanent residents.

Mortality risk due to infection was significantly increased in the combined SjD cohort compared with age- and sex-matched controls. The importance of infection as a COD in SjD has been shown consistently in multiple studies [7, 9, 39]. In addition to underlying immune dysfunction in AIRD, serious infection risk is elevated with chronic glucocorticoid treatment used in severe AIRD and SjD complications such as ILD [40].

The predominance of female patients in our study is consistent with the known higher ratio of female vs male SjD patient demographic (9:1) [4]. The higher burden of Elixhauser comorbidities in SjD cohorts may contribute to increased mortality, with further research needed to determine whether these comorbidities are manifestations of SjD, reflect a shared predisposition or result from adverse effects of treatments. To our knowledge, this is the first study of SjD mortality that identified self-reported Aboriginal and Torres Strait Islander ethnicity in primary SjD patients [8], although subgroup analysis of COD was not possible due to small numbers (N = 14). Reports indicate that in Aboriginal and Torres Strait Islander peoples, SLE mortality rates are six times higher [41] and RA mortality rates are three times higher than the general Australian population [42], revealing the critical need for improved research and models of care.

This study comprises the most comprehensive analysis of mortality risk and causes in SjD using both curated ICD-coded COD information and a matched control population, in addition to population mortality rates, which alone are used in most studies of mortality in SjD [7, 10, 36, 43]. Limitations include absence of information regarding the SjD diagnostic criteria used, diagnosis date, disease duration (with inclusion of patients at different stages of disease), measures of disease activity, ethnicity, smoking status and socio-economic status, which may have influenced mortality outcomes [44] and generalizability of results. Patients were drawn from a hospital-based cohort with potentially more severe SjD and/or comorbidities than patients not needing hospitalization. Furthermore, while controls were drawn from a hospitalized cohort and may differ from the general population (especially with respect to comorbidity burdens), we demonstrated mortality outcomes (e.g. relative survival) in controls that are commensurate with WA population data. Additional limitations include the absence of baseline comparison between SjD and control cohorts for disease composition, which may have led to unbalanced confounding factors and potential bias in mortality comparison. To partially address this, a sensitivity analysis was performed of the mortality HRs adjusted for cumulative comorbidities, with modest reduction of HRs, which remained significant. As data linkage of the cohort was only until 2013, outcomes may not reflect contemporary therapies for the AIRD associated with SjD (noting there are no proven disease-modifying treatments for SjD) and improvements in therapies and multidisciplinary care for SjD complications such as ILD. Finally, COD classification based on death certificates may be incomplete, due to reliance on medical professionals’ subjective judgment.

In conclusion, primary and associated SjD were associated with significantly increased all-cause mortality and mortality due to infection, lymphoid malignancy, cardiovascular and respiratory disease compared with age- and sex-matched controls in an Australian population. As systemic disease activity predicts mortality in SjD [9], this study confirming elevated mortality risk of Australian SjD patients highlights the importance of assiduous clinical monitoring and management of systemic disease activity, and the critical need for accelerated development of disease-modifying therapies for SjD. The significant contribution of cardiovascular disease and infection to mortality necessitates early cardiovascular risk factor management and vigilance for preventing and treating infections to improve overall survival. Further research is warranted to explore underlying mechanisms leading to increased cardiovascular morbidity and mortality in SjD, as well as impacts of socio-economic status, geographic location, comorbidities, and ethnicity on SjD mortality in Australia.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

The data underlying this article were provided by the Western Australia Data Linkage System. The data cannot be shared publicly to minimize reidentification of individuals. Data will be shared on request to the corresponding author with permission of the Western Australia Data Linkage System.

Funding

C.P. would like to acknowledge the support of the Australian government via the Australian Government Research Training Program Scholarship.

Disclosure statement: M.R. discloses clinical trials research support from AbbVie, Amgen, Argenx, AstraZeneca, BMS, GSK, Horizon, J&J, LG Chem, and Novartis; and consultancy fees/speaker honoraria from AbbVie, Amgen, Boehringer Ingelheim, J&J, Novartis, Pfizer, Recordati, Sandoz, UCB. C.L.H. discloses clinical trials research support from GSK, Servier, Axsome, Merck and AbbVie.; and speaker honoraria from Novartis. Research funding support from Arthritis Australia, NHMRC, MRFF and The Hospital Research Foundation. F.B.V. received funding from National Health and Medical Research Council (Emerging Leadership 1, grant 1196112) and is supported by funding from Arthritis Australia & Arthritis and Osteoporosis Western Australia (Grant 26-AA/AWA-F). F.B.V. has received support from Janssen-Cilag, AstraZeneca and CSL for projects, in-kind support from Olink and Sengenics, and support from Pfizer, Lupus Research Alliance, and SomaLogic for conference or meeting sponsorship. The other authors do not report any relevant conflict of interest.