Sarcoidosis-related mortality in France: a multiple-cause-of-death analysis

Discussion

For this study, we took advantage of the exhaustive recording of all French death certificates by CépiDc since 2002 and performed the largest MCOD analysis in a European country. Mainly, we found that: 1) the mean age at death reported on death certificates that listed sarcoidosis was lower than for the general population over the same time period, by about 6 years, 2) the UCD was sarcoidosis in half of the cases, 3) before the age of 65 years deaths were more frequent in males, 4) there was a significant increase in standardised mortality rates over the study period and 5) sarcoidosis-related mortality was higher in northern regions.

Contrary to cohort studies that are often performed in restrained numbers of medical facilities and for limited periods of time, MCOD analyses provide comprehensive data at a country level with complete coverage of all deaths over selected periods of time, thus avoiding selection bias [17].

However, the MCOD analyses have some limitations: indeed, there is no way to assess the accuracy of diagnoses reported in the dataset (including histological confirmation). For example, sarcoidosis may not have been declared on the death certificates of some patients with a past clinically occult disease and/or spontaneous remission. More symptomatic, long-standing and severe phenotypes may be over-represented, leading to overestimation of sarcoidosis-related mortality. Thus, such findings may not be relevant to the whole spectrum of patients with sarcoidosis.

Moreover, in such studies, it is not possible to exclude that some patients with a alternative diagnosis (e.g. tuberculosis or sarcoid-like reactions) may have been over-reported to have sarcoidosis.

Some certifiers may also have appraised sarcoidosis complications, such as pulmonary hypertension or fibrosis, as the UCD, whereas others may have considered that they were NUCDs. Thus, mortality rates must not be considered as surrogates for the incidence of this disease. This means that MCOD analysis is not performed to provide information about fatality, but rather crude mortality rates [18–20].

Another limitation of our study is a lack of data on ethnicity, socioeconomic status and environment. We did not study the links between occupational exposure and sarcoidosis-related death. However, a recent paper has reported a greater odds ratio for cases where there was occupational exposure and a difference between genders regarding some specific exposures [21].

In accordance with our results, Swigris et al. [4] reported a mean±sd age of death of 57.2±1.8 years in decedents with sarcoidosis compared with 71.4±1.1 years for the nonsarcoidosis population in the USA over the period 1988–2007. Although it has been stated that sarcoidosis patients have lower survival rates than the general population, a confirmed increased risk of death in the whole spectrum of sarcoidosis patients can only be identified in prospective studies, which are very difficult to conduct on a large scale. Ungprasert et al. [8] recently reported on a prospective cohort study, and found no difference in mortality rates between sarcoidosis patients and the general population. The differences seen in the MCOD analyses (ours and Swigris et al.’s [4]) probably reflect the differences between a subgroup of patients with sarcoidosis severe enough to be notified on a death certificate and the general population.

Interestingly, as compared with the general population, the majority of decedents aged <65 years were males, whereas females were the majority in those who died when aged ≥65 years. These observations are in accordance with previous epidemiological reports of two incidence peaks in sarcoidosis: the first being observed in young adults (mainly males exhibiting systemic features) and the second observed in older patients (more likely to be Caucasian females with ocular sarcoidosis) [15, 16]. Our data agree with previous reports conducted in the USA and England, showing an increased female/male mortality ratio in older age strata [4, 5, 13]. Finally, we found an increased age at death over the study period that was significant only for females with sarcoidosis. Hypotheses include better management of sarcoidosis or related comorbidities through optimised monitoring, the use of novel therapeutics and better prevention of side-effects.

Over the study period, we observed a significant increase in age-standardised mortality rates. This finding has also been reported in previous studies [4, 5, 12, 13, 22]. Nevertheless, this increase seems to be restricted to Caucasian ethnicity [12]. Various hypotheses have been formulated, such as an increased incidence or increased severity of sarcoidosis, but the most likely reason may be improved recognition of this disease by physicians and the use of chest computed tomography in difficult cases.

Overall, sarcoidosis was mentioned as the UCD in 48% of death certificates. This rate remained stable throughout the study period, but tended to decline with increasing age, whereas sarcoidosis notification as a NUCD increased with age. We speculate that this could be caused by 1) a decreased severity of sarcoidosis in the elderly [23] or 2) the accumulation of comorbidities and/or complications in the elderly [16]. Such variations have been frequently reported in other MCOD analyses [24].

When sarcoidosis was the UCD, the most common NUCDs were chronic respiratory insufficiency, lung fibrosis and pulmonary hypertension. In previous reports, lung fibrosis was reported as a contributing factor in 9% of deaths, but may have been under-reported according to the authors [4]. Pulmonary hypertension was mentioned on 2.8–7.1% of death certificates of sarcoidosis decedents [4]. Considering that they are known complications of sarcoidosis (mainly stage 4 patients) [7], the number of patients dying from lung fibrosis or pulmonary hypertension should be interpreted carefully. Indeed, death certifiers may simply have written “sarcoidosis” on death certificates without mentioning lung fibrosis or pulmonary hypertension. Moreover, some certifiers may have considered these complications as the UCD, whereas other may have mentioned them as NUCDs [25]. Finally, this observation is likely to reflect a subgroup of patients with more cases of chronic or severe sarcoidosis and who have a greater risk of dying from sarcoidosis.

Cardiovascular aetiology was described as a NUCD in 17% of decedents with sarcoidosis as the UCD. This is consistent with the 25% rate reported in the USA [4]. Such a rate suggests that, at the time of death, sarcoidosis was severe enough to warrant coding by death certifiers. In line with our results, a recent publication has associated sarcoidosis with pulmonary embolism and congestive heart failure in patients admitted into UK hospitals over a period of 35 years [26]. In our study, it was not possible to distinguish between nonspecific and sarcoidosis-specific heart disease.

When sarcoidosis was mentioned as a NUCD, neoplasms were the main UCD followed by cardiovascular disease and infection. Neoplasms were reported more often in decedents aged >50 years; lung cancer being predominant in males, whereas females were more likely to have non-Hodgkin's lymphoma reported as the UCD. Importantly, observed/expected ratios, which quantify the increase or decrease in mortality due to sarcoidosis, as compared with the general population, were found to be >1 for only infectious diseases (including tuberculosis) and chronic respiratory disease. Although the observed/expected ratio was high for tuberculosis, the raw number of deaths was very low (29 cases). As stated above, there was also uncertainty about a diagnosis of tuberculosis due to the nature of the data. Thus, in some cases, sarcoidosis and tuberculosis may have been misdiagnosed.

A cardiovascular-associated increase in mortality was only significant for decedents aged 45–64 years. In contrast, haematologic malignancy was more likely to lead to death in decedents aged ≥85 years.

Strikingly, the overall observed/expected ratio was <1 among sarcoidosis decedents whose death certificates indicated a neoplasm. Further analyses by age and gender revealed that this was only true and significant for those aged between 45 and 65 years at death. Solid neoplasms other than lung cancer were significantly lower in sarcoidosis patients aged 45–64 years and lung cancer was only significantly lower in males aged 45–64 years. One possible explanation may be a lower number of smokers within sarcoidosis patients, as this was reported previously [27, 28]. Another explanation may be a lower probability of reporting sarcoidosis when a neoplasm was the UCD. In contrast, mortality caused by haematologic malignancy was higher in older individuals (aged ≥85 years) and mortality due to malignancy, either haematologic or lung cancer, was higher in females aged ≥85 years. In a previous study in the UK, Le Jeune et al. [29] reported an overall increased incidence of cancer in sarcoidosis, but this was mostly due to increased incidence of skin cancers. For pneumonia and cerebrovascular diseases, observed/expected ratios <1 were observed in females and in those aged ≥85 years, respectively. This may reflect better medical management (including smoking cessation counselling) and earlier diagnoses rather than protection due to sarcoidosis itself. In the USA, Misraeidi et al. [12] reported less deaths caused by pneumonia in Caucasian females with sarcoidosis than in a control group comprised of patients with rheumatoid arthritis. They also found that Caucasians in the sarcoidosis group were less likely to have diabetes or a neoplasm, but were more likely to have heart failure. Nevertheless, their study was designed to analyse racial differences and the comparison was made using an identified control group rather than the general population, thus precluding a formal comparison. In France, ethnic statistics are forbidden; thus, we were unable to compare patients from different ethnicities. Similarly, data on residency were not available.

Finally, we did find a north–south gradient of age-standardised mortality at the country level. Despite the lack of data on incidence and prevalence of sarcoidosis in France, this gradient is more likely to reflect the prevalence gradient constantly reported at the European continent level [15, 30–32]. Such a difference is hypothesised to be due to environmental or occupational exposures rather than disparities in healthcare management.