Progressive interstitial lung disease in patients with systemic sclerosis-associated interstitial lung disease in the EUSTAR database

1A subset of patients with systemic sclerosis-associated interstitial lung disease (SSc-ILD) develop progressive ILD, which is associated with higher mortality, but the prevalence of progressive ILD and the overall disease course and patterns of SSc-ILD are unknown. Current clinical practice emphasises treatment initiation of SSc-ILD patients with progressive ILD.

2Around 30% of SSc-ILD patients experienced ILD progression during any 12-month period, and 67% of all SSc-ILD patients experienced progression at any time over the mean 5-year follow-up.

3ILD patterns in patients with SSc-ILD are very heterogeneous, with most patients showing both progressive and stable periods.

4Of all progressive SSc-ILD patients, only a minority showed a pattern of rapid, continuously declining forced vital capacity (FVC) with several consecutive episodes of FVC decline and no periods of FVC stability or improvement.

5These results highlight a pitfall in current clinical practice, where treatment is often initiated after FVC decline has happened, and thus when lung damage has already occurred. Novel treatment concepts are needed and should aim for prevention of progression to avoid irreversible organ damage. This study defines factors that can identify patients at risk for progression. The results also stress the heterogeneity and variability of the course of ILD in SSc, and highlight the need for close monitoring of all patients with SSc-ILD.

6Systemic sclerosis (SSc) is a rare autoimmune disease, frequently complicated by interstitial lung disease (ILD), which is associated with worse outcomes. Some patients with SSc-associated ILD (SSc-ILD) develop progressive ILD, showing decline in lung function and/or increasing extent of fibrosis by high-resolution CT (HRCT). The proportion of patients with SSc-ILD who develop progressive ILD and the pattern of disease course in these patients are incompletely understood. Prior analyses of disease course in SSc-ILD have found different disease patterns in different patient cohorts. However, these studies are limited by their small sample size, selected patient populations, significant referral biases and statistical instabilities of the trajectories. Randomised clinical trials provide valuable data, but the 12-month or 24-month duration often used is insufficient for assessment of long-term disease course. There also remains a high unmet need to specifically identify patients with SSc at risk of progressive ILD. Risk factors for SSc-ILD progression have been proposed by several studies; however, their clinical applicability and specific power to predict progression are limited. The optimal combination of risk factors to accurately predict progression has not been identified.

7Treatments are available for SSc-ILD, but to date, nintedanib is the only approved treatment shown to reduce lung function decline in patients with SSc-ILD. Current clinical practice emphasises treatment of patients with progressive ILD, and a recent study showed that nintedanib reduces decline of forced vital capacity (FVC) in progressive ILD associated with various underlying conditions, including connective tissue disease-associated ILD. However, waiting for FVC decline and/or extensive ILD involvement neglects the opportunity of early treatment intervention until after clinically meaningful lung damage has occurred. Novel treatment concepts are therefore aiming to prevent progression and avoid irreversible damage to organs. This requires an understanding of the course and patterns of ILD progression, and reliable prediction algorithms that allow the specific detection of patients at risk of progression at a very early stage. Unfortunately, in SSc-ILD, this knowledge is currently lacking, and treatment initiation is often delayed in clinical practice by waiting for lung function decline over the preceding year before initiation.

8The European Scleroderma Trials And Research (EUSTAR) group database is a large, real-world database representative of the general SSc population. It includes a wide range of patients with SSc-ILD, from those with mild and stable to advanced progressive disease.

9Thus, the aims of this study were: to assess the prevalence of progressive ILD over 12-month periods; to examine disease course and identify patterns of ILD progression in SSc over a 5-year period; and to identify risk factors predictive for progressive ILD in patients with SSc-ILD, using the EUSTAR database.

10Post hoc analyses of prospectively collected patient data from the EUSTAR database were conducted. The structure of the online database, the collected data set and definitions of clinical variables have been described in detail previously.

11Patients registered since 2010 in the EUSTAR database (start of the online version), aged ≥18 years, who fulfilled the 2013 American College of Rheumatology/European League Against Rheumatism SSc classification criteria; with presence of ILD by HRCT or X-ray; recorded disease duration; and with available measurements of FVC and diffusion capacity of the lungs for carbon monoxide (DLCO) at baseline and after 12±3 months were included.

12To reflect clinical practice with respect to patient follow-up, and the usual study duration in clinical SSc-ILD trials, absolute changes in FVC% predicted were first evaluated over a 1-year period (baseline to 12±3 months). FVC decline ≥10% predicted is frequently used to define significant ILD progression and was therefore selected in this study as the main outcome measure for progressive ILD. Furthermore, an FVC decline >5% predicted is greater than the estimated minimum clinically important difference at a group level and has previously been associated with increased mortality in SSc. Patients were therefore divided into four progressive ILD subgroups based on absolute change in FVC% predicted from baseline to 12±3 months: significant progression (decline of >10%); moderate progression (decline of 5% to 10%); stable ILD (decline or improvement of <5%); and improvement (improvement of ≥5%) (table 1). The prevalence of annual FVC changes was assessed prospectively in patients with available data over a mean follow-up of 5 years, using the definitions of progressive ILD described above.

13A decline in FVC of ≥10%, or a decline in FVC of 5% to 10% along with a decline in DLCO of 15%, is a proposed definition of progressive fibrosis. Therefore, we also assessed the prevalence of this combined endpoint.

14Because annual FVC patterns can change over time, we evaluated the overall lung function course in patients who had at least two 12-month periods with FVC measurements. These periods could be, but did not need to be, consecutive. For the overall FVC course, patients were divided into five subgroups based on the difference between the first and last available FVC measurement: major decline (FVC decline of >20%); significant decline (FVC decline of >10 and ≤20%); moderate decline (FVC decline of 5% to 10%); stable (FVC decline or improvement of <5%); and improvement (FVC improvement of ≥5%).

15The numbers of patients who experienced no 12-month period of decline, one period of decline (moderate or significant) or multiple periods of decline (moderate, significant or both) across all periods with data available over the 5-year follow-up were assessed. Patients with ILD progression were split into different progression patterns according to the number of FVC decline periods: rapid progression (no periods of FVC stability or improvement); progression (more periods of decline than stability/improvement); and slow progression (more periods of stability/improvement than decline).

16All-cause mortality was assessed in all patients with SSc-ILD, and in patients with progressive ILD, until last available follow-up.

17Candidate baseline variables to predict progressive ILD were selected based on reports from the published literature and expert opinion: sex, age, reflux/dysphagia symptoms, SSc subtype, antibody status (anti-topoisomerase antibody (ATA) anti-centromere antibody (ACA), anti-RNA polymerase III antibody (ARA)), baseline FVC, baseline DLCO, disease duration, skin involvement measured by modified Rodnan skin score (mRSS), erythrocyte sedimentation rate (ESR), C-reactive protein (CRP) level, dyspnoea class, treatment, synovitis and muscle weakness. Extent of lung fibrosis on HRCT was not included due to extensive missing data.

18All analyses were performed using IBM SPSS Statistics V.25 and Stata V.15. Pearson χ2 test, Fisher’s exact test or independent sample t-test was used, as appropriate. For correlation analyses, Pearson or Kendall’s tau-b coefficients were applied as appropriate. All multivariable analyses were preceded by estimation of correlation between risk factors. Univariable and multivariable logistic regression analyses with OR and 95% CI were applied to analyse the predictive ability of baseline variables for progressive ILD. In the multivariable analyses, 10 events per variable were needed, and the variables were selected by expert opinion.

19Univariable and multivariable linear mixed-effect models were performed to identify risk factors of longitudinal changes in FVC (% predicted) over the maximum 5-year follow-up period (baseline, 1, 2, 3, 4 and 5 years). Only patients with at least three serial FVC measurements were included in the analyses. Time and risk factors were fixed effects. Interaction effects between time and fixed factors were checked by including product terms in the models. Only significant interaction terms in the univariable analysis are presented, and they were further included in the multivariable model. Risk factors selected for multivariable analyses were based on expert opinion. All models included random intercept and slope, and an unstructured correlation matrix was used.

20EUSTAR is part of the World Scleroderma Foundation, which has patient representatives from the Federation of European Scleroderma Associations (FESCA) in its governing board.

21Within the EUSTAR database, 6004 patients included since 2010 aged ≥18 years fulfilled the SSc classification criteria and had lung imaging data available. Of these, 2259 (38%) had evidence of SSc-ILD on imaging, of which 826 had valid lung function data available after 12±3 months follow-up and were eligible for inclusion.

22Demographic and clinical characteristics of all eligible patients are shown in table 1. No significant difference was observed in the baseline characteristics of the 826 eligible patients and the 1433 ineligible patients (online supplementary table S1).

23When analysing the prevalence of progressive ILD within the initial 12-month period, we found that 100 patients (12%) had significant ILD progression, 123 (15%) had moderate progression, 396 (48%) were stable and 207 (25%) had improvement.

24In multivariable logistic regression analyses, FVC (OR 1.02; 95% CI 1.01 to 1.03; p<0.001), presence of reflux/dysphagia symptoms (OR 1.97; 95% CI 1.14 to 3.40; p=0.016) and mRSS (OR 1.06; 95% CI 1.00 to 1.12; p=0.036) at baseline were predictive for significant ILD progression at 12±3 months. No association was seen with age, sex, disease duration, antibody status, SSc subtype or immunosuppressant treatment.

25Prevalence and prediction of progressive ILD measured by the combined FVC and DLCO definition over the initial 12-month period were comparable to these data (online supplementary table S1 and figure S1).

26A total of 535 (65%) patients with SSc-ILD had ≥3 FVC measurements available during the mean 5-year (±2.2) follow-up period, allowing for assessment of long-term ILD course. Baseline characteristics did not differ between patients with ≥3 and patients with <3 FVC measurements (n=291 (35%)).

27To assess the overall disease course, we assessed FVC changes between baseline and last available FVC. We found that 49 (9%) showed major FVC decline (FVC decline of >20%); 75 (14%) had significant decline (FVC decline 10% to 20%); 76 (14%) had moderate decline (FVC decline 5% to 10%); 206 (39%) were stable (FVC changes <5%); and 129 (24%) experienced improvement in FVC (FVC improvement >5%) over the overall disease course (mean 5-year follow-up) (figure 1A). The prevalence of significant ILD progression was between 13% and 18% and the prevalence of moderate progression was between 9% and 10% in each 12-month period over this 5-year follow-up (figure 1B). These progressive periods rarely appeared in consecutive 12-month periods, and progressive periods were mostly followed by stable periods (figure 2A). Stable periods were followed by a progressive period in about 30% of cases (figure 2B). Irrespective of the severity of overall FVC decline (major, significant or moderate), most patients still experienced at least one 12-month period of stable or improving FVC (table 2). On the other hand, patients with stable or improved overall FVC could still experience 12-month periods of FVC decline; these declines were more frequently moderate (FVC decline 5% to 10%) than significant (FVC decline 10% to 20%). Only 178 (33%) patients experienced no period of FVC decline of ≥5% during any 12-month period (table 2).

28Most patients with SSc-ILD with an overall FVC decline over 5 years had a slow pattern of lung function decline, with more periods of stability/improvement than decline (58%); 34% showed a progressive pattern, with more periods of decline than stability/improvement and slow progression. Only 16 (8%) patients showed a rapidly declining FVC pattern, with several consecutive episodes of FVC decline and no periods of FVC stability or improvement. Patterns of progression in patients with moderate, significant and major overall decline in FVC%, stratified by the presence or absence of a decline in the first 12 months, are shown in figure 3.

29To identify SSc-ILD patients at risk of ILD progression, we assessed the predictive value of baseline clinical variables on FVC measurements over the 5-year follow-up period. In multivariable linear mixed-effect models, we identified male sex, presence of reflux/dysphagia symptoms and high baseline mRSS as the strongest predictors, with significant interaction effects between time and these variables. This indicates that FVC changed differently over time as a function of one of these predictors (ie, different slopes). Older age, higher DLCO, dyspnoea (New York Heart Association class 3 or 4) and increased ESR were also significantly predictive for FVC decline but without a time interaction effect, indicating that the FVC changed significantly over time but not differently between patients with or without these clinical features (table 3). Immunosuppressive treatment was not predictive for FVC decline over time.

30Of 826 patients with SSc-ILD, 85 (10%) died during follow-up. There were no significant differences in mortality rate between patients with significant ILD progression (11 (12%)), moderate progression (18 (15%)) or stable ILD (36 (9%)) over the initial 12±3-month period. In patients with overall FVC changes measured between baseline and last available FVC, death occurred in 9 of 49 (19%) patients with major decline; 7 of 75 (9%) patients with significant decline; 12 of 76 patients (16%) with moderate decline; 18 of 206 (9%) patients who were stable; and 9 of 129 (7%) patients with improvement, with differences not statistically significant. As there were only a small number of events, no regression analyses were performed.

31This is the largest study to prospectively analyse the prevalence of progressive ILD in patients with SSc-ILD, and the first to describe comprehensively the disease course and patterns of ILD progression in patients with SSc over the long term.

32The proportion of patients with SSc-ILD who experienced progressive ILD during the initial 12±3-month period was 27%, and in each 12-month period over the mean 5-year follow-up, 23% to 27% of patients experienced progression. These findings are in agreement with estimates of progressive ILD prevalence of 31% to 32% derived from serial lung function data in patients with SSc and an international physician survey.

33Here, we show that patterns of FVC are frequently inconsistent between consecutive 12-month periods. Most patients who experienced an overall decline in FVC had periods of FVC improvement and, conversely, some patients whose FVC improved overall had periods of FVC decline. Patients with overall major FVC decline (FVC decline >20% over the entire study period) usually had several 12-month periods with FVC decline >10% rather than FVC decline of 5% to 10%. Others experienced a slower, but cumulative course of declining FVC. Such patients with slower progression can easily be overlooked in clinical practice and in current treatment strategies that target patients who progress rapidly and with significant FVC changes. Smaller changes in FVC (5% to 10%) may in themselves be clinically significant, as seen in patients with idiopathic pulmonary fibrosis. In clinical practice, this means that FVC decline >5% should alert physicians, especially when multiple declines occur, even when not in consecutive periods. These results highlight a pitfall in current clinical practice, where treatment is often initiated after FVC decline has happened, and thus when lung damage has already occurred. Novel treatment concepts are needed and should aim for prevention of progression to avoid irreversible organ damage. These results also stress and highlight the need for close monitoring of all patients with SSc-ILD, as also recently suggested by the European expert consensus. Respiratory symptoms, changes in HRCT findings and desaturation on exercise tests should all be implemented in clinical practice to assess ILD progression and aid treatment decisions.

34Robust predictive risk factors are very important for the early identification of progressive patients. Our large, multicentre study demonstrates that skin fibrosis (higher mRSS), male sex and the presence of reflux/dysphagia symptoms are the strongest predictors for FVC decline over time. Other predictive parameters included the presence of inflammation (higher ESR) and shorter disease duration, which are already frequently used as enrichment strategies for clinical studies. These parameters may also be applied in daily clinical practice, helping to identify patients who should receive treatment early, even before the first FVC decline has occurred. However, if earlier treatment of patients at risk for FVC decline truly leads to better outcomes, it needs to be analysed in appropriate randomised controlled clinical trials in the future. Risk factors identified in this study are potential inclusion criteria for such a trial. Interestingly, contrary to our finding that higher FVC at baseline was predictive for ILD progression, previous studies suggested that lower FVC at baseline is a risk factor for progressive ILD. These studies included some SSc patients without ILD, and one study assessed patients within 3 years of SSc diagnosis. Furthermore, definitions of progression (FVC decline of ≥15%, FVC or DLCO decline of ≥15%, or FVC or DLCO falling below 55%) differed from those in our study. The strongest association with further FVC decline was seen in patients with baseline FVC <65% predicted, lower than the mean value in our study (86%). Our contrasting findings may reflect these differences and should be assessed in other unselected cohorts.

35Strengths of our study include the use of a large set of prospective, representative real-world data, which increases the applicability of our results to clinical practice and different definitions of ILD progression. Nonetheless, this study has several limitations. While the data were gathered prospectively, this was a post hoc analysis. No central lung function reading was conducted, increasing the variability of FVC and DLCO. Most patients with SSc-ILD in the database (1433/2259) did not have serial lung function data. Data on immunosuppressant use were only available for 244/826 eligible patients, and the exact date of initiation, treatment indication and cumulative doses are unknown. Several studies have suggested that the extent of lung fibrosis by HRCT is prognostic for disease progression and mortality in SSc-ILD. Although data regarding the extent of lung fibrosis were not available in the database for the present analysis, they may be a valuable addition in future studies. Recent analyses also suggest that mRSS progression is an important risk factor for later FVC progression, which was not analysed in this study. A lead time bias cannot be excluded, as this was not an incident cohort. Finally, ILD-specific mortality was not available in the EUSTAR database. Here, all-cause mortality was not influenced by ILD progression; as recently highlighted, it is likely that ILD-specific mortality differs between progressive and stable ILD patients.

36This study provides novel insights into the occurrence of progressive ILD in SSc-ILD. The results stress the heterogeneity and variability of the course of ILD in SSc. Close monitoring of patients with SSc-ILD and awareness of the variable course of progression is of high importance in considering when to initiate treatment.