Article Text

Download PDFPDF

Prediction of progression of interstitial lung disease in patients with systemic sclerosis: the SPAR model
  1. Wanlong Wu1,2,
  2. Suzana Jordan1,
  3. Mike Oliver Becker1,
  4. Rucsandra Dobrota1,
  5. Britta Maurer1,
  6. Håvard Fretheim3,
  7. Shuang Ye2,
  8. Elise Siegert4,
  9. Yannick Allanore5,
  10. Anna-Maria Hoffmann-Vold3,
  11. Oliver Distler1
  1. 1 Department of Rheumatology, University Hospital Zurich, Zurich, Switzerland
  2. 2 Department of Rheumatology, South Campus, Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University, Shanghai, China
  3. 3 Department of Rheumatology, Rikshospitalet, Oslo University Hospital, Oslo, Norway
  4. 4 Department of Rheumatology and Clinical Immunology, Charité University Hospital Berlin, Berlin, Germany
  5. 5 Rheumatology A Department, Cochin Hospital, Paris Descartes University, Paris, France
  1. Correspondence to Dr Oliver Distler, Department of Rheumatology, University Hospital Zurich, Zürich 8091, Switzerland; oliver.distler{at}usz.ch

Abstract

Objectives To identify the predictive clinical characteristics and establish a prediction model for the progression of mild interstitial lung disease (ILD) in patients with systemic sclerosis (SSc).

Methods Patients with SSc from two independent prospective cohorts were included in this observational study. All patients fulfilled the 2013 American College of Rheumatology/European League Against Rheumatism criteria, had mild ILD at baseline diagnosed by High-Resolution Computed Tomography (HRCT), available baseline and ≥1 annual follow-up pulmonary function tests and no concomitant pulmonary hypertension or airflow obstruction. ILD progression was defined as a relative decrease in forced vital capacity (FVC)%≥15%, or FVC%≥10% combined with diffusing capacity for carbon monoxide %≥15% at 1-year follow-up. Candidate predictors for multivariate logistic regression were selected by expert opinion based on clinical significance. A prediction model for ILD progression was established in the derivation cohort and validated in the multinational validation cohort.

Results A total of 25/98 and 25/117 patients with SSc showed ILD progression in the derivation cohort and the validation cohort, respectively. Lower SpO2 after 6 min walk test (6MWT) and arthritis ever were identified as independent predictors for ILD progression in both cohorts. The optimal cut-off value of SpO2 after 6MWT for predicting ILD progression was determined as 94% by receiver operating characteristic curve analysis. The derived SPAR model combining both predictors (SPO2 and ARthritis) increased the prediction rate from 25.5% to 91.7% with an area under the curve (95% CI) of 0.83 (0.73 to 0.93).

Conclusions The evidence-based SPAR prediction model developed in our study might be helpful for the risk stratification of patients with mild SSc-ILD in clinical practice and cohort enrichment for future clinical trial design.

  • systemic sclerosis
  • pulmonary fibrosis
  • arthritis
View Full Text

Statistics from Altmetric.com

Introduction

Systemic sclerosis (SSc) is a heterogeneous autoimmune disease, characterised by vascular damage, inflammation and fibrosis of skin and various visceral organs.1 Interstitial lung disease associated with SSc (SSc-ILD) is a common complication and leading cause of death in SSc.2 3 Nowadays, treatment options are still limited and challenging. The European League Against Rheumatism (EULAR) recommended that cyclophosphamide (CYC) should be considered for the immunosuppressive treatment of SSc-ILD. Mycophenolate mofetil (MMF) has shown similar effects to CYC.4–6 However, due to their known toxicity, overall mild to moderate and short-lasting effects, CYC and MMF are generally administered only to selected cases with risk for ILD deterioration.4

Previous studies explored baseline predictors for lung progression and mortality in SSc-ILD. Diffuse cutaneous subset, presence of antitopoisomerase-I antibodies, decreased baseline forced vital capacity (FVC) and diffusing capacity for carbon monoxide (DLCO) were reported to be significantly associated with lung progression in patients with SSc, while positive anticentromere antibody (ACA) was protective.7–13 However, different definitions of lung deterioration were applied. When patients with SSc-ILD were preselected as the study population, only extensive lung disease was predictive for both lung progression and mortality in previous studies.14–16 Older age, decreased baseline FVC, short-term pulmonary function trends and exercise peripheral oxygen saturation (SpO2) were shown to be predictive for mortality in patients with SSc-ILD.14 15 17–22 However, these results were limited by small sample size, lack of validation and insufficient adjustment for potential confounders.23 Moreover, few extrapulmonary factors were assessed for prognostic value in SSc-ILD.

The natural disease course of SSc-ILD is highly heterogeneous. Some patients with minimal ILD will remain stable while others could deteriorate rapidly. Clinicians would like to start active treatment earlier for those who are at risk of lung progression. However, currently no data are available to distinguish between progressive and stable patients when mild ILD is diagnosed. These patients might be easily overlooked and undertreated due to better pulmonary function tests (PFTs) often within normal values.24 This defines an urgent clinical need for the identification of baseline predictors of lung progression in patients with mild SSc-ILD. A clinically applicable model to identify patients at risk for progression of mild SSc-ILD should become even more meaningful in view of the multiple ongoing phase II/III prospective randomised placebo-controlled clinical trials with targeted therapies in SSc, which might open new treatment options for SSc and SSc-ILD in the near future.25

The objectives of the current study were (1) to identify baseline clinical characteristics that can predict the progression of mild SSc-ILD at 1-year follow-up in a derivation cohort; (2) to establish a simple-to-use and practical prediction model by combining independent predictors for lung progression in mild SSc-ILD; (3) to validate the derived prediction model in a multinational validation cohort.

Methods

Study cohorts

Patients were included from the longitudinal cohort with prospectively collected data from the University Hospital Zurich (derivation cohort). Data from three other SSc expert centres (Oslo, Paris, Berlin) combined as a validation cohort were also included.

All included patients fulfilled the following criteria: diagnosis of SSc according to the 2013 American College of Rheumatology/EULAR (ACR/EULAR) classification criteria,26 diagnosis of mild ILD by HRCT at baseline visit, available PFTs at baseline and annual follow-up visits (annual defined as 12±3 months), no concomitant pulmonary hypertension according to right heart catheterisation or echocardiography as judged by the local investigators, no evidence of substantial airflow obstruction defined as forced expiratory volume in 1 s/FVC<70%.

Clinical data

Demographic and clinical parameters including age, gender, disease duration, SSc antibodies, PFTs, 6 min walk test (6MWT), arthritis status and modified Rodnan Skin Score (mRSS) at baseline were obtained from all included centres. The parameters in each local database were collected following the European Scleroderma Trials and Research recommendations (for details about oximetry assessment see online supplementary material).3

Supplementary file 1

Assessment of ILD

PFTs were obtained at baseline and annual follow-up visits in all patients. HRCT lung images were available at baseline in all included patients. Reticular pattern abnormalities and superimposed ground-glass opacities defined as equivalent to fibrosis were assessed semi-quantitatively as previously described.12 15 27 Pulmonary fibrosis at baseline was expressed as percentage of total lung volumes, an extent of <20% pulmonary fibrosis was considered as mild ILD and an extent of pulmonary fibrosis ≥20% was defined as extensive ILD (see online supplementary material).

Study end point

The end point of this study, progression of mild SSc-ILD, was defined present if either of the following parameters was fulfilled at 1-year follow-up: a relative decrease in FVC% predicted ≥15%, or relative decrease in FVC% predicted ≥10% combined with DLCO% predicted ≥15%. Patients fulfilling these criteria were classified as progressors while the others were non-progressors. If more than one follow-up visit was available, the visit fulfilling the above criteria for progression was chosen as the follow-up visit and the visit 1 year before as the baseline visit.

We defined the thresholds of PFTs’ decline for the end point based on the criteria recommended for idiopathic pulmonary fibrosis trials by the American Thoracic Society/European Respiratory Society and previous clinical trials in SSc-ILD.5 6 28 The evolution period of 1 year has been chosen since it is the routine follow-up interval for patients with SSc in clinical practice and also frequently used in clinical trials for SSc-ILD.

Statistical analysis

Candidate predictors were selected using the nominal group technique by SSc experts (OD, A-MH-V, YA, ES, SY, BM, RD, MOB, WW), who were asked to suggest clinically meaningful variables with face validity based on previous studies and feasibility (see online supplementary material). The following hypothesis was tested in this study: one or more of six candidate predictors including ILD-associated variables (FVC, DLCO, SpO2 after 6MWT) and extrapulmonary variables (mRSS, arthritis ever, disease duration) can independently predict the progression of mild SSc-ILD. Potential confounding demographic and SSc-associated variables (age, gender, anti-Scl-70 positivity, ACA positivity) were adjusted in the multivariate analysis.

Baseline characteristics were described and compared between progressors and non-progressors by univariate analyses followed by Bonferroni correction. The independent sample t-test, Mann-Whitney U test, Χ2 test and Fisher’s exact test were conducted, as appropriate.

Relationships between the candidate predictors and the occurrence of ILD progression were investigated using multivariate logistic regression. Each significant continuous parameter was then converted to categorical variable by determining the optimal cut-off value using receiver operating characteristic (ROC) curve analysis. Multiple imputation was used to address missing data in the validation cohort and the pooled cohort. The logistic regression models from both the original dataset and the multiply imputed dataset were presented.

The risk prediction model was established by combining independent predictors. Discriminatory performance by applying ROC curve analysis (eg, area under the curve (AUC), sensitivity and specificity) and predictive performance (eg, positive and negative predictive values) of the prediction model were examined. The prediction model was also separately tested in patients stratified by disease duration and immunosuppressive treatment at baseline.

All above-mentioned statistical analyses were conducted in both the derivation cohort and validation cohort. Significance was defined as p<0.05. The statistical analysis was performed using SPSS V.23.

Results

Study populations

A total of 98 patients with SSc with mild ILD at baseline who met the inclusion criteria were analysed in the derivation cohort. Of those, 25 (25.5%) patients showed ILD progression at 1-year follow-up (classified as progressors). Patients were predominantly female (79.6%) with a mean age of 57.3 years. The median disease duration was 4.7 years (IQR 1.9–7.4). Thirty patients (30.6%) had diffuse cutaneous involvement. The mean FVC (% predicted) was 101.9% and the mean DLCO (% predicted) was 76.5%. The detailed demographic and clinical characteristics at baseline are provided in table 1.

Table 1

Patients’ demographic and clinical characteristics at baseline in the derivation cohort

In the validation cohort, 25 of 117 (21.4%) patients with SSc with mild ILD at baseline were ILD progressors at 1-year follow-up. The description of baseline characteristics and univariate comparisons are summarised in table 2.

Table 2

Patients’ demographic and clinical characteristics at baseline in the validation cohort

Follow-up of lung function

The available follow-up (time between baseline and latest follow-up visit) was similar for progressors and non-progressors (derivation cohort: 3.3±1.4 vs 3.0±1.1 years, p=0.330; validation cohort: 3.4±2.1 vs 4.7±3.4 years, p=0.168). However, progressors had significantly lower FVC values at latest follow-up compared with non-progressors (derivation cohort: 82.1±14.7 vs 102.2±20.0, p<0.001; validation cohort: 81.4±20.2 vs 98.2±18.5, p=0.001) and more FVC decline since baseline (online supplementary table S1).

Univariate and multivariate analysis

Results of univariate comparison of baseline parameters between progressors and non-progressors are summarised in tables 1 and 2 for the derivation and validation cohort, respectively. In the derivation cohort, progressors had higher mRSS, incidence of arthritis ever, joint contractures and dyspnoea New York  Heart   Association grade ≥2. Worse results for 6MWT including 6MWD, SpO2 before 6MWT, SpO2 after 6MWT and oxygen desaturation were also observed in progressors. By using Bonferroni correction, the modified critical p value (α) was determined as 0.002. Significantly higher incidence of arthritis ever and lower SpO2 after 6MWT were found in progressors compared with non-progressors. Arthritis was first reported 2.7 years in average before baseline among the 27 patients who ever had arthritis; oxygen desaturation after 6MWT was first observed 0.6 year in average before baseline among the 28 patients who had oxygen desaturation. There was no significant difference in FVC or DLCO. Similar results were found in the validation cohort.

The logistic regression model including all six candidate predictors identified lower SpO2 after 6MWT (OR=0.77, 95% CI 0.64 to 0.94, p=0.009) and arthritis ever (OR=6.84, 95% CI 2.19 to 21.33, p=0.001) as independent predictors for progression of mild SSc-ILD at 1 year (figure 1A, for details see online supplementary table S2).

Figure 1

Multivariate logistic regression models in the derivation cohort, the validation cohort and the pooled cohort (both original and multiply imputed datasets in the validation cohort and the pooled cohort, adjusted for age, gender, anti-Scl-70 positive, ACA positive in the pooled cohort). 6MWT, 6 min walk test; anti-Scl-70, antitopoisomerase-I antibodies; ACA, anticentromere antibody; DLCO, diffusing capacity for carbon monoxide; FVC, forced vital capacity; mRSS, modified Rodnan Skin Score; SpO2, peripheral capillary oxygen saturation.

The same multivariate logistic regression model was tested in the validation cohort. Lower SpO2 after 6MWT (OR=0.65, 95% CI 0.47 to 0.90, p=0.009) and arthritis ever (OR=5.27, 95% CI 1.29 to 21.44, p=0.020) were confirmed as independent predictors for ILD progression. Similar results were obtained using multiply imputed datasets (figure 1B,C for details see online supplementary table S2).

We then combined the derivation cohort and the validation cohort into a pooled cohort. In order to adjust for possible confounding effects of age, gender and profile of antibodies on ILD progression, a new multivariate logistic regression model was conducted in the pooled cohort, which confirmed lower SpO2 after 6MWT (OR=0.75, 95% CI 0.64 to 0.89, p=0.001) and arthritis ever (OR=7.91, 95% CI 3.11 to 20.13, p<0.001) as significant predictors. Similar results were also obtained using multiply imputed datasets (figure 1D,E for details see online supplementary table S3).

Identification of the optimal cut-off value for SpO2 after 6MWT

A ROC curve analysis was then performed that identified 94% as the best cut-off value for SpO2 after 6MWT to distinguish progressors from non-progressors. For predicting ILD progression, SpO2 after 6MWT≤94% had a sensitivity of 68.0% and a specificity of 84.7%, together with an AUC of 0.76 (95% CI 0.64 to 0.88) (p<0.001) in the derivation cohort (figure 2). The subsequent ROC curve analysis also identified 94% as the best cut-off value for SpO2 after 6MWT (AUC (95% CI)=0.81 (0.70 to 0.92)) in the validation cohort, with a sensitivity of 55.6% and a specificity of 93.5% (p<0.001).

Figure 2

(A) Percentage of progressors and non-progressors per SpO2 after 6MWT in the derivation cohort. (B) Sensitivity analysis for ILD progression depending on different cut-off values for SpO2 after 6MWT in the derivation cohort. 6MWT, 6 min walk test; AUC, area under the curve; ILD, interstitial lung disease; SpO2, peripheral capillary oxygen saturation.

Prediction models of progression of mild SSc-ILD

We then established prediction models by combining the two categorical predictors. Altogether, four models were tested and compared with discriminatory and predictive performance (table 3). Model 3 increased the prediction success rate from 25.5% in the whole unselected cohort to 91.7% in the optimised enrichment cohort. for example, among the 12 patients who fulfilled both the SpO2 after 6MWT≤94% and arthritis ever criterion, 11 had ILD progression.

Table 3

Prediction models of progression of mild SSc-ILD in the derivation cohort

In a simplified scoring system, presence of both SPO2 after 6MWT≤94% and ARthritis ever were set to 1, while both SPO2 after 6MWT>94% and ARthritis never were set to 0, giving a SPAR score ranging from 0 to 2. We tested the discriminatory and predictive performances of the SPAR prediction model in both derivation and validation cohorts. The AUCs for the SPAR score were 0.83 (95% CI 0.73 to 0.93) (p<0.001) in the derivation cohort and 0.82 (95% CI 0.70 to 0.94) (p<0.001) in the validation cohort, respectively. The successful prediction rate for ILD progression increased with an increase in SPAR score in both cohorts. The prediction rate with a SPAR score of 0 was 7.4%/6.3%, with a score of 1 it was 32.3%/36.0%, and with a score of 2 it was 91.7%/85.7% in the derivation/validation cohorts, respectively (table 4).

Table 4

Sensitivities, specificities, PPVs and NPVs of the SPAR score in the derivation and validation cohorts

Subgroup analyses of disease duration and immunosuppressive treatment

In order to investigate whether the predictive value of the SPAR model on ILD progression could be influenced by disease duration and immunosuppressive treatment at baseline, we tested the discriminatory and predictive performances of the SPAR model in the subgroups of patients with short (≤5 years) and long (>5 years) disease duration, treated and untreated patients in the derivation and validation cohort, respectively. The SPAR model could still significantly distinguish ILD progressors and non-progressors in any subgroup (for details see online supplementary table S4–S5).

Discussion

Among patients with mild SSc-ILD, those prone to faster deterioration are more likely to benefit from active therapeutic interventions than those prone to stabilisation. Here, we present predictors of rapid lung worsening in patients with SSc with mild ILD and identify a simple prediction model (SPAR model) combining SpO2 and arthritis, both readily accessible clinical characteristics, to predict ILD progression at the 1-year follow-up from real-life data in two independent SSc cohorts from multiple centres.

We chose to focus on mild ILD as the study population, since this subgroup frequently has a relatively normal PFTs value and will be recognised as progressors only when lung volumes drop dramatically 1 year later. Therefore, such progressors could have a better chance to be monitored more carefully in daily practice if a prediction model is available.

Although a moderate definition of ILD deterioration (either FVC decline ≥10% or DLCO decline ≥15%) was commonly used in previous observational studies and clinical trials,9 15 29–31 we set a stricter definition. That is because we assumed patients with mild ILD would have better lung function than overall patients with SSc-ILD. Thus, a greater decline should be more clinically meaningful when related to worse outcome and survival. The baseline PFT values reported in our study were better than previous studies and clinical trials, which confirmed that our cohorts indeed had mild functional SSc-ILD as assessed by HRCT.5 6 19 20 Although there is possibility that some patients with smaller changes in PFTs than in our definition ultimately deteriorate to clinically severe cases, our data showed that progressors had significantly lower FVC values and more FVC decline than non-progressors during a similar follow-up period, indicating that we successfully captured the real long-term ‘ILD progressor’ population in our study.

It should be noted that exercise SpO2 and presence of arthritis were identified as significant predictors rather than some known risk factors for progression such as PFTs. A possible explanation could be that these known risk factors have been derived from cohorts with established, more extensive ILD. While they might be important predictors in these advanced cases, they might be less important in earlier and mild cases. Indeed, FVC was within the normal range in our cohort, and thus might not be sensitive enough for the prediction of progression in mild SSc-ILD.

Exercise-induced peripheral oxygen saturation was significantly lower in progressors and proved to be predictive for ILD progression, consistent with previous results for overall mortality.22 This finding indicates that the insufficient pulmonary physiological reserve might be ahead of lung function volumes’ decline in mild SSc-ILD, which suggests that measurement of oxygen saturation during 6MWT might be a more sensitive parameter than PFTs for predicting the progression of SSc-ILD.

The current study also identified arthritis ever as an independent predictor of mild SSc-ILD progression. This association was not found in a previous study,7 which might be explained by two major differences in our study design. Only patients with ILD extent <20% lung involvement on HRCT were included in our cohort, which was different from the cohort selection of overall SSc-ILD or patients with SSc in previous studies. In addition, we have chosen ‘ever happened’ rather than ‘current’ arthritis. The presence of anti-CCP was low in our cohorts (3.5%~6.5%), and showed no significant difference between progressors and non-progressors, indicating that overlap syndrome with rheumatoid arthritis could hardly influence our findings.

There are several strengths of our study. First, we validated our results in a multinational cohort including patients from three centres. Second, we collected serial lung function tests and extensive clinical data in both derivation and validation cohorts. Third, the results were derived from the real-life data, which could probably guide the daily practice. Finally, we excluded the patients with pulmonary hypertension and substantial airflow obstruction, which could also impair the lung function, to make our cohort more consistent.

There are also limitations of our study. First, our data were derived from several local databases. Despite extensive quality control, there were still missing data as common for real-life registries. Those parameters with >50% missing values (eg, history of smoking) could not be considered for multivariate analysis even by imputation methods. Hence, we might have missed some potential predictors. Second, peripheral oxygen saturation was obtained by finger oximetry in some centres. In order to avoid biases by Raynaud phenomenon and peripheral vasculopathy, forehead oximetry would have been more reliable.32 However, measuring exercise SpO2 by finger oximetry in SSc-ILD reflected arterial oxygen saturation (SaO2) and provided a meaningful prognostic value in previous studies.22 Third, one should not misinterpret the relatively high percentage of lung progression in our study, because we had a specific cumulative rather than cross-sectional selection of progressors. Finally, although patients from three international centres had been included as a validation cohort in our study, external validation with qualified data in other international cohorts should be performed to confirm our findings.

In conclusion, our study defined lower SpO2 after 6MWT and arthritis ever as independent baseline predictors for progression of mild SSc-ILD at 1-year follow-up. The derived evidence-based SPAR model might help physicians to identify patients at risk for progressive lung fibrosis, which might have implications for therapeutic strategies in clinical practice.

Acknowledgments

The authors would like to thank Nicole Schneider for excellent administration and data entry to the Zurich SSc cohort.

References

View Abstract

Footnotes

  • Handling editor Josef S Smolen

  • Contributors Study conception and design: OD, WW, SJ, A-MH-V, YA, ES, SY, BM, RD, MOB. Acquisition of data: WW, OD, SJ, A-MH-V, HF, YA, ES. Analysis and interpretation of data: WW, OD, SJ. Drafting the article: WW, OD, SJ. Revising the article: A-MH-V, YA, SY, ES, HF, MOB, BM, RD. All authors have finally approved the submitted version to be published.

  • Funding This study was supported by a grant from Boehringer Ingelheim.

  • Competing interests OD has consultancy relationship and/or has received research funding from Actelion, Bayer, Boehringer Ingelheim, ChemomAb, espeRare foundation, Genentech/Roche, GSK, Inventiva, Italfarmaco, Lilly, Medac, MedImmune, Mitsubishi Tanabe Pharma, Novartis, Pfizer, Sanofi, Sinoxa and UCB in the area of potential treatments of scleroderma and its complications, and a patent mir-29 for the treatment of systemic sclerosis licensed. YA has received grants from BMS, Genentech-Roche, Inventiva, Sanofi and consulting fees from Actelion, Bayer, Biogen, Boehringer, Genentech-Roche, Galapagos, Inventiva, Medac, Pfizer, Sanofi and Servier, about the treatment of systemic sclerosis. BM has received research funding in the area of systemic sclerosis and related conditions from AbbVie, Protagen, EMDO, Novartis, German SSc Society, Pfizer, Roche, Actelion, MSD, OPO Foundation and a patent mir-29 for the treatment of systemic sclerosis licensed. WW, SJ, MOB, RD, HF, SY, ES and A-MH-V have nothing to disclose.

  • Patient consent Obtained.

  • Ethics approval Each participating centre has obtained approval from the local ethics committee for including a patient’s data in the SSc cohort after the patient has given written informed consent.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.