Cancer development in patients hospitalized with systemic lupus erythematosus: A population‐level data linkage study

To explore the association between systemic lupus erythematosus (SLE) with the risk of cancer development and subsequent 5‐year mortality in Western Australia (WA).


| INTRODUC TI ON
Systemic lupus erythematosus (SLE) is a chronic autoimmune disease underwritten by the production of autoantibodies, immune complex deposition, and complement activation, which produces a variety of clinical manifestations, organ damage, and premature mortality. 1 Management of SLE patients often requires intensive and long-term use of immunosuppressive medications, the side effects of which, increase the patient's burden of disease. Consequently, SLE patients have poorer survival, especially after the onset of infection, cancer, renal failure, myocardial infarction, and central nervous system disorders.
Meta-analyses suggest that SLE is associated with increased risk of more than 16 types of cancer, and a decreased risk for prostate cancer and cutaneous melanoma 2,3 ; however, those findings did not account for the impact of environmental triggers, ethnicity, alcohol use, and smoking status. 2 Other studies support an increased risk of cervical, 4 hematological, particularly non-Hodgkin's lymphoma, and kidney cancers 5 ; with some showing a decreased risk of prostate cancer and bladder cancer, 6 or no association with breast, uterine, ovarian, colorectal, or brain cancers. 7 Australian data were not available for aggregation in the meta-analyses, 2,3 including data on cancer development in Aboriginal Australians with SLE; an ethnic group with higher demand for hospitalizations, including SLE-related hospitalizations. 1 Furthermore, Australia has unique environmental exposures, including very high levels of background ultraviolet radiation and high levels of silica dust particles, which makes data from this region important to include in future studies. 8 Therefore, this study explores the association between SLE with overall and typespecific cancer development in a cohort of SLE patients hospitalized in Western Australia (WA) between 1980 and 2014, and subsequent 5-year mortality.

| MATERIAL S AND ME THODS
This whole-of-population cohort study used data linked by the Western Australian Data Linkage System (WADLS), which uses probabilistic matching (99.7% accuracy) to identify individuals across administrative health data sets. 9 The Western Australian Administrative data have been validated for the identification of rare diseases, and suitable for producing epidemiological estimates 14 ; and, having at least one ICD-9-CM code of SLE (ICD-9: 694.4 and 710) having a positive predictive value of 70%-96% in administrative or claims data, 15 which is even higher (≥97%) when patients are seen by a specialist. 16 Given that, 89%-93% of SLE patients meeting classification criteria are hospitalized during their disease course 17 ; and that the HMDC also captures same-day drug infusions (although not the specific drug types) and procedures, our cohort will represent the majority of SLE patients in WA over the study period.
Systemic lupus erythematosus patients (n = 2111) were propensity score matched (1:10) to 22 110 control individuals free of rheumatic disease conditions. Time zero (study entry) was defined as the index hospitalization for SLE and a random hospitalization in the non-exposed comparators. SLE patients and controls were matched for the year of the index (first ever) hospital contact, year of time zero (study entry), age at time zero, sex, and Aboriginality. 1 Racial and ethnic demographic data are not comprehensively captured in the WADLS. However, self-reported Aboriginality is recorded during hospital admissions, with sensitivity over 90%. 18

| Statistical analysis
Results are summarized as a mean with standard deviation, median with interquartile range or frequency and proportion. Betweengroup differences were assessed with the t-test, Mann-Whitney U test, χ 2 test, and odds ratios (OR) with 95% confidence intervals (CI) as appropriate. We determined the association between a hospitalization for SLE and cancer development with univariate and multivariate Cox proportional hazards regression models.
Sub-group and period specific (overall, 1980-1999, and 2000-2014) analysis looked at the association between a hospitalization for SLE and cancer development across the sexes, Aboriginality, age groups (<40, 40-49, 50-59, 60-69, and ≥70 years of age), and cancer site. The multivariate model (Model 1) for the association between SLE and cancer development risk was adjusted for prevalent cancer at time zero, smoking history (ever smoked), and the number of modified Charlson Comorbidity Items (0 = no CCI items, 1 = 1-2 CCI items, and 2 = 3 or more CCI items). 20 As SLE patients are susceptible to hematological diseases and infections throughout the disease course and viruses can increase the risk of cancer development. We performed further sub-group analyses (sex, age groups, Aboriginality, and year of index SLE hospitalization) to determine the association between risk of hospitalization for SLE with hematological, non-hematological and virus-induced cancers, e.g. human papillomavirus (HPV) -related cancers. 22 We

| Association between a hospitalization for SLE with the risk of cancer development
A total of 426 (20.2%) SLE patients developed cancer compared with 3664 (17.4%) in the general population comparators ( Figure 1 and   Table S5). The risk of developing oropharyngeal, vulvo-vaginal, softtissue sarcoma, non-melanoma, nervous system (excluding brain), and hematological cancer, especially non-Hodgkin's lymphoma, was significantly higher in SLE patients.  (Table S6). However, Aboriginal Australians had higher risk of incident cancer development of the hepatobiliary, respiratory (lung), musculoskeletal, and female reproductive systems (Table S6). Within Aboriginal Australians, a SLE-related hospitalization conferred no additional (age-and sex-adjusted) risk of cancer development from 1980-2014 (Table 2); although, we had insufficient end points to explore this association within specific cancertypes (data not shown).

| Association between a hospitalization for SLE with mortality after cancer development
After cancer development, hospitalized-SLE patients had increased multivariate-adjusted risk of mortality during follow-up (aHR 1.31, 95% CI 1.12-1.52); this association was reached by 3-years and established at 5-years (Table 3). mortality. Five-year mortality after cancer development was higher in hospitalized-SLE patients, who were <50 years-old (highest), female, non-Aboriginal Australian, or hospitalized from 1980-1999 (Table 4). Five-year mortality after cancer development was lower in hospitalised-SLE patients who were >70 years-old (  2 and an international multi-centered cohort study. 23 Additionally, our unadjusted findings of an increased risk of cancer in female, but not male, SLE patients, aligned with others. 2,23 However, after further adjustment for confounding risk factors the increased risk of cancer was attenuated, which aligned with other studies. 23,24 Our lower than expected SLE prevalence in female patients (85.1% vs. >90%) has been seen in other hospitalized SLE cohorts from Argentina (83.3%) 25 and Western Europe (87%), 26

and in another
Western Australian population-level study with a stricter definition of SLE (≥2 ICD-10 codes at least 30 days apart), 27 and might reflect male patients having higher healthcare resource utilization for SLEspecific or comorbidity reasons, rather than a selection bias. 28   Although comparable studies are limited with respect to mortality after cancer development, our findings aligned with an SLE-specific study, 43 and a broader sample of patients with rheumatic diseases. 44 Our data suggest that virus-induced cancers, 45    We showed that SLE patients had twice the risk of developing cancer of the meninges and spine, but not the brain (cerebrum/ cerebellum). The lack of association between SLE and brain cancer confirms existing data, 2 but there are some data to suggest that SLE patients are susceptible to developing meningioma. 68 We did not find an increased odds of viral infections of the central nervous system at time zero. Although, it is unclear what underlies the potential mechanism between SLE and these rare cancers, future studies could investigate the role of neuropsychiatric disease activity and cancer development of the nervous system.
Hospitalized SLE patients had a 78% increased risk of developing hematological cancer, similar to a meta-analysis, 2 an international multi-centered cohort study, 23 and others. 60  Systemic lupus erythematosus patients were at increased risk of some but not all cancer types compared with the general population. Clinicians should be particularly aware of the increased risk of virus-induced cancers, particularly of the oropharyngeal and vulvovaginal areas; as well as, the increased risk of soft-tissue sarcoma, non-melanoma skin cancer, and non-Hodgkin's lymphoma over time.

| Strengths and limitations
SLE patients had increased risk of 5-year mortality after cancer development, with the highest risk in those under 50 years old, and those with skin or prostate cancer.

AUTH O R CO NTR I B UTI O N S
The authors are responsible for the content and writing of the paper.
Authors (WDR, DBP, HK, CAI, JCN) had access to the data in this study and WDR takes complete responsibility for the integrity and accuracy of the data analysis.

ACK N O WLE D G E M ENTS
The authors wish to thank the staff at the Western Australian  Research & Training Scholarship.

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare that they hold no conflict(s) of interest in relation to this study or its findings.