Steady decrease in systemic sclerosis mortality rates at younger ages over the past five decades

Abstract

Objective

We aimed to assess SSc mortality by age in the general population over the past five decades.

Methods

This is a population-based study using a national mortality database and the census data for all US residents. We calculated the proportions of deaths for SSc and for all other causes (non-SSc) by age, and calculated age-standardized mortality rates (ASMRs) for SSc and non-SSc, and the ratio of SSc-ASMR to non-SSc-ASMR by age groups for each year from 1968 through 2015. We performed joinpoint regression to estimate the average annual percent change (AAPC) for each of these parameters.

Results

SSc was recorded as the underlying cause of death in 5457 decedents aged ≤44 years, 18 395 aged 45–64, and 22 946 aged ≥65 from 1968 through 2015. At ages ≤44, the proportion of annual deaths decreased more for SSc than for non-SSc: AAPC, –2.2% (95% CI, –2.4% to –2.0%) for SSc vs –1.5% (–1.9% to –1.1%) for non-SSc. Consistently, SSc-ASMR decreased from 1.0 (95% CI, 0.8–1.2) in 1968 to 0.4 (0.3–0.5) per million persons in 2015, a cumulative decrease of 60% at an AAPC of –1.9% (95% CI, –2.5% to –1.2%) at ages ≤44. The SSc-ASMR:non-SSc-ASMR ratio also decreased [cumulative –20%; AAPC –0.3% (95% CI, –1.15% to 0.55%)] in the ≤44-years group. In contrast, those aged ≥65 experienced a steep increase in the SSc-ASMR [cumulative 187.0%; AAPC 2.0% (95% CI, 1.8–2.2)] and the SSc-ASMR:non-SSc-ASMR ratio [cumulative 395.4%; AAPC 3.3% (95% CI, 2.9–3.7)].

Conclusion

Mortality for SSc has steadily decreased at younger ages over the past five decades.

Open in new tabDownload slide

Introduction

Mortality for SSc has begun to decline in the 2000s, but still remains high relative to mortality from all other (non-SSc) causes [1]. Most patients with SSc are young: in a large US cohort, 91% of patients had SSc onset prior to age 65 and 46% before age 45 [2]. Young patients with SSc had a delayed diagnosis and more cumulative organ damage, including oesophageal disease and myositis [3], and they can die prematurely of these complications. There have been conflicting reports about SSc mortality by age. While some studies found no significant difference between the standardized mortality ratios in SSc patients younger than 65 and those older than 65 at diagnosis [4], others found the standardized mortality ratio to be higher in younger patients [3]. In addition, some studies found that older age at disease onset was an independent predictor of SSc mortality risk [5, 6]. Most previous studies describing SSc mortality by age were based on patient cohorts at referral centres [3–8], which do not capture changes in SSc incidence over time and do not reflect the true burden of SSc mortality in the general population. These limitations might have contributed to inconsistent findings across previous studies.

We used the Centers for Disease Control and Prevention (CDC)’s national mortality database (which encompasses >99% of deaths across the USA) to calculate population-based estimates of SSc deaths and age-standardized mortality rates (ASMRs) by age over a period of 48 years from 1968 to 2015. To evaluate improvement in SSc mortality in the context of improvement in overall mortality for all causes in the general population over time [9], we computed the ratio of SSc mortality rates to non-SSc mortality rates by age groups.

Methods

Data source

We used the CDC Wide-ranging Online Data for Epidemiologic Research (WONDER) web application to gather data on SSc deaths from 1968 (the earliest year for which the CDC published county-level mortality data) through 2015. Deaths were attributed to SSc if an international classification of diseases (ICD) code for SSc was listed as the underlying cause of death on the death certificates: ICD-8, 734.0; ICD-9, 710.1; and ICD-10, M34 [1]. Information on age was ascertained using standard methods [10].

We obtained annual death counts in the entire US population and separately by age. We then calculated the percentages of total SSc deaths and non-SSc deaths in various age groups for each year over the 48-year period.

The work is exempt from regulations involving human subjects and is compliant with the Declaration of Helsinki.

Mortality rates

For the calculation of mortality rates, the size of the population (total and each group) was obtained from the United States Census Bureau files for each year. We quantified age-specific crude mortality rates for SSc and non-SSc for each year from 1968 through 2015 as the number of deaths in each year divided by the number of persons in the US general population in that same year [11]. This was done within age strata for the total US population for the three age groups, and repeated for non-SSc deaths.

To calculate the ASMR for the population for each year from 1968 through 2015, we combined the yearly age-specific crude mortality rates with the age distribution of the US population in 2000, as described previously [11]. This was done separately for each age group and for both SSc deaths and non-SSc deaths. We then computed the ratio of SSc-ASMR to non-SSc-ASMR for each year, by age groups.

Statistical analysis

We used joinpoint regression to fit piecewise-linear (or broken-line) trends to the annual percent SSc and non-SSc deaths, the annual SSc-ASMR, non-SSc-ASMR, and SSc-ASMR:non-SSc-ASMR ratio by age groups over the 48-year period [1]. Joinpoint regression identifies a set of joinpoints—the calendar years for which the change in the slope (of proportions of death or ASMR) is statistically significant—and computes the slope (year-to-year percentage change in annual ASMR or in percent of deaths) and the 95% CI over each linear trend segment between adjacent joinpoints [11, 12]. This approach identifies the year when the trend (i.e. slope) changes significantly and determines the magnitude of the change. Joinpoint regression analyses were conducted using the National Cancer Institute Joinpoint Regression Program, which uses a grid search method to find the best locations for the joinpoints based on a least-squares fit of the data and uses a permutation test with Monte Carlo simulation to determine the optimal number of joinpoints (see the Supplementary Methods, available at Rheumatology online).

We calculated the annual percent change (APC) for each trend, and then calculated the average annual percent changes (AAPCs) with 95% CIs for the entire study period. AAPCs represent changes in trends and offer the advantage of reporting a weighted average of the APCs for each segment [12].

To investigate whether the differences in the proportions of SSc and non-SSc deaths in the various age groups is more than expected by chance, we performed χ² statistics with Yates’s correction (GraphPad Prism 6.07) and quantified the odds ratio with its 95% CI.

Results

SSc was recorded as the underlying cause of death in 5457 decedents aged ≤44 years, 18 395 aged 45–64, and 22 946 aged ≥65 in the USA from 1968 through 2015. The proportions of deaths in those aged ≤44 and those aged 45–64 were higher for SSc than for non-SSc deaths (Table 1).

Trends in annual percent deaths

The proportion of annual SSc deaths in the ≤44-year age group decreased from 23.4% in 1968 to 5.7% in 2015, at an AAPC of –2.2% (95% CI, –2.4% to –2.0%) (Fig. 1). The percent of SSc deaths also decreased in the 45–54-year age group over the 48-year period. No statistical change was noted in the 55–64-year age group, but in the ≥65-year age group the proportion of SSc deaths increased from 23.8% in 1968 to 59.3% in 2015 at an AAPC of 2.0% (95% CI, 1.3–2.7).

Noticeably, the proportions of annual deaths decreased at a higher pace for SSc than for non-SSc causes in the ≤44 and 45–54 years age groups, but increased more for SSc than for non-SSc causes in the ≥65 age group (non-overlapping CI for AAPCs).

In 1968, the proportions of total deaths in ages ≤44 years was higher for SSc than for non-SSc (23.4% vs 13.5%, P < 0.0001), but in 2015 the proportion of total deaths was slightly lower for SSc than for non-SSc (5.7% vs 6.9%, P = 0.1) (Supplementary Fig. S1, available at Rheumatology online). The 4.1-fold decrease in the proportion of SSc deaths (23.4% in 1968 to 5.7% in 2015) was due to both decreased absolute SSc death counts at ages ≤44 years (from 109 deaths in 1968 to 68 deaths in 2015) and increased absolute SSc death counts at ages ≥65 years (from 111 deaths in 1968 to 709 deaths in 2015) (Table 2).

Further analysis of SSc deaths at younger ages revealed that in those aged ≤19 years, the proportions of total SSc deaths decreased from 2.4% in 1968 to 0% in 2015, in those aged ≤34 years, it reduced from 10.5% in 1968 to 1.9% in 2015.

Trends in SSc mortality rate according to age groups

The ≤44-years age group did not experience a period of statistically significant increase in SSc-ASMR at any time over the 48-year period (Fig. 2, left panel). In contrast, those aged ≥65 had a continuous increase in SSc-ASMR for 32 years before decreasing in the last 16 years of the study period. The 45–64-years age group exhibited two rise-and-fall trends over the 48-year period.

In contrast to the increasing and decreasing trends in SSc-ASMRs, non-SSc-ASMRs decreased or stayed stable throughout the 48-year period in all age groups (Fig. 2, middle panel). Consequently, the SSc to non-SSc ASMR ratio initially increased at greater APCs in ≥65 and 45–64-year age groups, but not in ages ≤44 (Fig. 2, right panel). Starting in the 2000s, the ratio decreased in all age groups, although the differences were not statistically significant in those aged ≥65 years.

The AAPC showed significant decreases in non-SSc-ASMR in all age groups (Fig. 3). For SSc-ASMR, only persons aged ≤44 years had a significant annual decline (AAPC, –1.9%); those aged ≥65 years had the highest annual increases (AAPC, 2.0%).

Similar trends were seen in the cumulative percent change in ASMRs between 1968 and 2015 (Table 2). Persons aged ≤44 years experienced a 60% cumulative decrease in SSc-ASMR in the 48-year period, whereas those aged ≥65 had the largest cumulative increase (187.0%). Of all age groups, only persons aged ≤44 years had a cumulative decrease in the ratio of SSc-ASMR to non-SSc-ASMR in 48 years (–20.0%); those aged ≥65 years had a dramatic cumulative increase (395.4%).

Discussion

Analyses of all deaths recorded across the USA over a 48-year period revealed that the proportions of deaths in younger age groups have decreased at a higher rate for SSc than for non-SSc causes. The mortality attributed to SSc has steadily and significantly improved in the younger age group (≤44 years) over the past 48 years, but those ≥65 years of age had a dramatically increased SSc mortality relative to mortality from all other causes. The decreases in absolute SSc death counts, proportionate SSc death burden, and in the ratio of SSc mortality rates to non-SSc mortality rates at ages ≤44 years suggest that the premature mortality at the population level has reduced more for SSc than for non-SSc (the general population) over the 48-year study period, which is remarkable considering SSc has had no U.S. Food & Drug Administration–approved treatment until recently.

Several factors may underlie the age-related pattern of changes in SSc mortality that we observed. The significantly reduced mortality at ages ≤44 years may reflect improved survival in younger patients owing to increasing use of disease-modifying drugs over time [1, 13]. Cumulative toxicities of these medications with time could account for increasing SSc mortality in older age [1]. Greater risks of pulmonary hypertension, renal impairment and cardiac disease among those with late-age (>65 years) onset SSc [2] could also contribute to higher mortality in old age. Nevertheless, the ratio of SSc to non-SSc mortality rates has begun to decrease from 2000 onwards for those aged ≥65 years. Many advances in SSc evaluation and management in the late 1990s to the early 2000s, such as the introduction of prostanoids, endothelin receptor antagonists, and phosphodiesterase-5 inhibitors to treat pulmonary arterial hypertension, early screening for alveolitis and selective use of CYC and MMF for SSc lung disease, early recognition and treatment of SSc renal crisis with angiotensin-converting enzyme inhibitors, establishment of classification criteria for early SSc, and identification of the relationships between autoantibodies and prognosis could have contributed to decreasing SSc mortality in all age groups in the 2000s.

Improving detection of SSc disease over time, identification of SSc-associated autoantibodies, and the introduction of classification criteria for early SSc in 2001 might have led to increasing recognition of SSc, with earlier interventions at younger ages and more older individuals with SSc. Increasing detection of ACA-positive individuals with age could also result in more people with a SSc diagnosis at older ages [2]. The mean age at SSc diagnosis in the Danish National Registry increased from 53.7 ± 15.6 during 1995–2001 to 56.1 ± 15.5 during 2009–2015 [14], which could partly account for the increasing SSc prevalence and mortality rates at older ages. Finally, changes in physicians’ practice of recognizing SSc and recording it on the death certificates (including electronic reporting and usage of ICD-10 codes) over decades could have partly contributed to the SSc mortality trends that we observed.

The age-related SSc mortality trends could also be related to differences in SSc mortality by race/ethnicity. Indeed, black persons have a 13-years lower median age of death from SSc compared with white persons, and a 5.1-times the odds of dying from SSc before 65 years of age [15]. Disconcertingly, the gap in the proportions of premature SSc deaths between black and white persons has widened over the last five decades [15]. Efforts are underway to evaluate the relationship between race/ethnicity, age, and other demographic factors in influencing SSc mortality.

The limitations of studies using mortality databases were recently elaborated [1, 16], such as the difficulty in ascertaining the accuracy of the physicians’ coding on death certificates. However, misclassification of SSc (i.e. recording SSc as the cause of death on death certificates for decedents that did not have SSc) is unlikely to have influenced our inferences in a substantial way [1]. Unlike administrative health-care databases, in which possible and probable diagnoses are often recorded, death certificates usually have the most proximate illness recorded as the cause of death. Consequently, SSc is unlikely to be recorded as a cause of death for decedents that did not have SSc. However, SSc may be under-recorded as the cause of death in a proportion of patients whose proximate causes of death were cardiovascular disease, cancer, or infection [8, 17], which are common in older individuals [16]. However, such underreporting (i.e. SSc not recorded as the cause of death on death certificates when in fact SSc predisposed the patient to the listed proximate cause of death) in older patients would underestimate SSc mortality at older ages, where we found a cumulative increase of 187% in the SSc mortality rate and of 395% in the ratio of SSc-ASMR to non-SSC-ASMR. Thus, the profound age-related trends in SSc mortality we detected cannot be explained by a low specificity of the cause of death on death certificates, as random misclassification increases the similarity between study groups. Other limitations of our study include lack of data on disease severity, treatment history, organ involvement, and social variables. Finally, the inferences made using aggregate mortality data in this study must be verified by collecting individual-level data to account for population heterogeneity.

Under-recording of SSc on death certificates may occur, because at the time of death many SSc patients may be under the care of physicians who may have a limited awareness of SSc as the cause of death. SSc mortality studies like ours may help improve this knowledge gap in health-care workers, with increased recognition of SSc and its varied complications as the cause of death and improved physicians’ coding on death certificates.

Nevertheless, our study provides an unbiased assessment, based on a large sample size, of the population-based burden of SSc mortality by age. Such mortality statistics from national databases serve as important indicators of the burden of specific diseases at the population level, with implications for health policy planning and resource allocation. Thus, until we have large-scale, population-based, prospective studies on SSc outcomes, mortality data from studies like ours may influence research and health-care policy, such as the allocation of resources to the elderly at risk of high SSc mortality. Use of direct age standardization to calculate annual mortality rates, use of joinpoint regression as a computational approach to identify mortality trends, and calculation of the ratios to assess SSc mortality trends relative to changes in mortality from all other causes in the general population are other strengths of our approach.

In conclusion, the reduced mortality burden from SSc relative to mortality from all other causes in the general population prior to age 45 years is encouraging. However, the relative mortality burden for SSc is still high in ages 45–64 years, and has dramatically increased in those aged ≥65 years over the past five decades. These observations emphasize the need for prospective studies to collect individual-level data, with a goal to identify SSc mortality risk factors that could be modified to improve outcomes in older individuals with SSc.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

Data used in this study are from a national mortality database maintained by the Centers for Disease Control and Prevention (CDC)’s National Vital Statistics System. The exact data used in this manuscript can also be obtained by contacting the corresponding author.

Contribution statement

E.Y.Y. and R.R.S. had full access to the data and take full responsibility for the integrity of the data and the accuracy of the analysis. E.Y.Y. performed data collection and data analysis. All three authors contributed to the literature search, preparation of the figures and tables, the study design, data analysis and interpretation, and the writing of the manuscript. R.R.S. supervised all aspects of this study.

Funding

This work was supported in part by the National Institutes of Health (R01-AI080778 and R01-AR056465). E.Y.Y. was supported by NIH T32-DK-07789, NIH T32-HD-007512, and the UCLA Children’s Discovery and Innovation Institute.

Disclosure statement: The authors have declared no conflicts of interest.

References