Sarcoidosis incidence and prevalence: a nationwide register-based assessment in Sweden

Discussion

Prevalent sarcoidosis in 2013 ranged from 152 to 215 per 100 000 depending on the definition used. The annual incidence of sarcoidosis in Sweden during 2003–2012 was 11.5 per 100 000 and differed by –10% to +30% when varying the number of visits, requiring specialist care or requiring two visits within 1 year. One-third of the incident cases were treated with glucocorticoids, antimalarials or immunomodulating medications within 6 months of first sarcoidosis-coded visit in specialty care.

The incidence of 11.5 per 100 000 reported here is lower than a previous report from Sweden by Hillerdal et al. [3] (19 per 100 000 in 1966–1980), who identified sarcoidosis cases from radiographic screenings in a regional health survey. One-third of the cases in the report by Hillerdal et al. [3] were identified because of symptoms and the rest were diagnosed through the health survey or by chance. In contrast, we identified individuals with sarcoidosis who received a diagnosis in inpatient or outpatient care, which likely includes more symptomatic cases at the expense of overlooking cases that would have remained clinically undetected. The study population investigated by Hillerdal et al. [3] also had a different age distribution, capturing cases with a median age 10 years younger than in our study. We cannot rule out that the age at onset has increased over time, which has been suggested by previous studies [19, 20]. However, the older mean age at diagnosis observed in our study may be due to differences in age of onset of symptomatic versus asymptomatic disease or the inclusion of older age groups in our study. Hillerdal et al. [3] had low rates of screening for older residents, whereas we used a population-based design that captured all Swedish adult residents’ sarcoidosis diagnoses. A report from Denmark also using register-based nationwide data estimated an incidence of 7.2 per 100 000 [10]. Unlike our study, they based their estimates exclusively on hospitalisations, likely underestimating the incidence of sarcoidosis which can often be managed in outpatient care.

The incidence of sarcoidosis did not vary over time between 2003 and 2012. This was consistent with findings from previous studies from the UK, Japan and Australia [8, 15, 21]. Two studies from the USA observed an increase in hospitalisations with sarcoidosis between 1996 and 2000 [19] and 1998 and 2008 [22]. This could indicate that improved detection and diagnostic procedures over time led to an increased number of identified cases during hospitalisation in the USA. A major limitation to these hospitalisation studies is that they could not identify unique individuals and therefore the same individuals could be counted more than once [22]. In contrast, our study included outpatient specialist care and the use of individual-level data, which likely better identifies the majority of new cases of clinical disease.

Approximately 55% of the incident cases were male, indicating no large difference between the sexes, as has been observed in previous reports [6–11, 13]. Other studies have, however, observed a slight female predominance [4, 5]. This may be due to differences in data sources or due to true differences in the populations under study in terms of age distribution, genetics or sarcoidosis phenotype.

We found that the peak age at onset in males was 10 years earlier than in females, which is consistent with most previous studies [3, 5, 7, 10, 13, 23, 24], but not all [4, 14]. The earlier onset in men could be due to an environmental factor which is more common in men and experienced at a younger age, such as an occupational exposure. In the ACCESS (A Case Control Etiologic Study of Sarcoidosis) study, men were more likely to have pulmonary compared with systemic disease, which was associated with exposure to wood burning and agricultural organic dust [25]. Therefore, the two different peaks in men and women may represent two unique subtypes of sarcoidosis driven by sex-specific exposures. Alternatively, the onset could be delayed in women due to endogenous hormones which may protect against the occurrence of the disease. This is supported by the observation that pregnancy appears to have a favourable effect on sarcoidosis in some patients [26] and that later age at menopause is associated with a lower incidence of sarcoidosis [27].

The highest prevalence was observed in less densely populated areas of Sweden in the northwest. It has been hypothesised that sarcoidosis could be caused by an exposure more common in rural areas such as pine, soil, clay or occupation (e.g. metal industry, farming) [11, 13, 28]. Occupation is closely related to education and we observed the lowest incidence among the highest educated (12.9 per 100 000 per year). However, this did not differ greatly from those with high school nor less than high school education (14.8 and 14.2 per 100 000 per year, respectively). Education is an imperfect proxy for occupation because industrial exposures may affect individuals living in the vicinity regardless of whether they work in the industry. For example, a study from Switzerland reported the highest sarcoidosis prevalence in regions where agriculture and metal processing were common industries [11]. Further research investigating specific occupations as well as proximity to potential occupational exposures on the individual level are necessary.

The observed geographical variation could also be explained by the genetic composition of the individuals living in different regions in Sweden [29]. The southern regions of Sweden include Sweden's most populated cities, i.e. Stockholm, Gothenburg and Malmö, which are more urban and ethnically diverse than northern regions. Genetics play a role in sarcoidosis aetiology, as demonstrated by the familial aggregation observed in multiple studies and the identification of several susceptibility loci [30]. Whether familial risk explains the geographic variation observed here should be investigated further. Our results do not support the hypothesis that tick-borne bacteria such as Borrelia burgdorferi or Rickettsia helvetica cause sarcoidosis [31, 32] because ticks are more common in the southern and coastal regions where we see the lowest prevalence [33].

Due to our reliance on ICD codes to identify sarcoidosis and the difficulty in diagnosing some cases, some misclassification likely exists and we were unable to examine different subtypes such as Löfgren's syndrome. We were limited by the use of the first sarcoidosis-coded visit date as a proxy for date of incidence, and there is likely a lag between onset and diagnosis of a few months. Despite this potential misclassification, our estimates were quite robust and do not change greatly when varying the incidence and prevalence definitions. In a clinical cohort of well-characterised sarcoidosis cases from Karolinska University Hospital, 95% had two or more sarcoidosis-coded visits. This indicates that we capture the large majority of cases, but does not provide a positive predictive value of the ICD-10 diagnosis, which must be obtained from a validation study. Data on primary care visits for sarcoidosis were not included in this study, but most cases are referred to a specialist in Sweden, therefore these missing cases are likely few. Lastly, our prevalence estimate may include people who no longer suffer from sarcoidosis, therefore overestimating the burden of the disease on the population. However, this estimate represents a group that may have a reactivation of disease in the future and are therefore still of interest. The number of prevalent cases who were seen for care in 2013 may more realistically reflect those individuals living with sarcoidosis currently requiring clinical care; however, it misses individuals who still live with the disease but visit the doctor infrequently.

The generalisability of our study is excellent due to the population-based nationwide data used. However, it is uncertain whether our findings are generalisable to other populations with different genetic and racial compositions. The identification of over 16 000 prevalent cases and over 10 000 incident cases makes our investigation the largest study of individual-level data on sarcoidosis occurrence to date. Our study is further strengthened by the inclusion of cases diagnosed in both inpatient and outpatient care, which gives a clear picture of the occurrence of clinical disease in Sweden. Our register-based estimates were reasonably robust when using different combinations of ICD codes, care providers and treatments, demonstrating that these data are valuable for future investigations into the aetiology of sarcoidosis.

In conclusion, we observed a geographical variation in Sweden that has not been reported before and must be investigated further with regard to familial clustering of disease and environmental exposures. Data show that men and women are affected by the disease at different ages, with the age at diagnosis in men 10 years younger than in women. This indicates that sex plays a role in sarcoidosis occurrence through genetics, hormones, occupational exposures or another as-yet unidentified factor. Future studies should investigate risk factors for sarcoidosis in women and men separately to clarify the reasons for these differences.