We propose a deep generative approach using latent temporal processes for modeling and holistically analyzing complex disease trajectories, with a particular focus on Systemic Sclerosis (SSc). We aim to learn temporal latent representations of the underlying generative process that explain the observed patient disease trajectories in an interpretable and comprehensive way. To enhance the interpretability of these latent temporal processes, we develop a semi-supervised approach for disentangling the latent space using established medical knowledge. By combining the generative approach with medical definitions of different characteristics of SSc, we facilitate the discovery of new aspects of the disease. We show that the learned temporal latent processes can be utilized for further data analysis and clinical hypothesis testing, including finding similar patients and clustering SSc patient trajectories into novel sub-types. Moreover, our method enables personalized online monitoring and prediction of multivariate time series with uncertainty quantification.

Background: Systemic sclerosis (SSc) is an autoimmune disease with high mortality with lung involvement being the primary cause of death. Progressive interstitial lung disease (ILD) leads to a decline in lung function (forced vital capacity, FVC% predicted) with risk of respiratory failure. These patients could benefit from an early and tailored pharmacological intervention. However, up to date, tools for prediction of individual FVC changes are lacking. In this paper, we aimed at developing a trustworthy machine learning system that is able to guide SSc management by providing not only robust FVC predictions, but also uncertainty quantification (i.e. the degree of certainty of model prediction) as well as similarity-based explainability for any patient P (i.e. a list of past SSc patients with similar FVC trajectories like P). We further aimed to identify the key clinical factors influencing the model's predictions and to use model-guided data representation to identify SSc patients with similar sequential FVC measurements. Methods: We trained and evaluated machine learning (ML) models to predict SSc-ILD trajectory as measured by FVC% predicted values using the international SSc database managed by the European Scleroderma Trials and Research group (EUSTAR), which comprises clinical, laboratory and functional parameters. EUSTAR records patients' data in annual assessment visits, and, given any visit, we aimed at predicting the FVC value of a patient's subsequent visit, taking into account all available patient data (i.e. baseline and follow-up visit data up to the time point where we make the prediction). For the training of our ML models, we included 2220 SSc patients that had at least 3 recorded visits in the EUSTAR database, were at least 18 years old, had confirmed ILD and sufficient clinical documentation. We developed sequential ML models implementing the attentive neural process formalism with either a recurrent (ANP RNN) or transformer encoder (ANP transformer) architecture. We compared these architectures with baseline sequential models including gated recurrent neural networks (RNNs) and multi-head self-attention transformer-based networks. Baseline non-sequential models included tree-based models such as gradient boosting trees, and regression-based models with varying regularization schemes. Our experiments used stratified 5-fold cross-validation to train and test the models using the average root mean squared error (RMSE), weighted RMSE, and mean absolute error (MAE) as performance metrics. We computed the coverage and Winkler score for uncertainty quantification, SHAP values for grading the input features importance and used the data embeddings of the ANP architectures for both similarity-based explainability and the identification of similar SSc patient journeys. Results: Patients' baseline FVC scores ranged from 22 to 150% predicted with a mean (SD) of 90.53% predicted (21.52). Our deep learning models showed better performance for FVC forecasting, compared to tree- and regression-based models. The top performing ANP RNN architecture was able to closely model future FVC values with average (SD) performance of 8.240 (0.168) weighted RMSE and 6.94 (0.190) MAE that was further used as feature generator for a logistic regression trained to predict a FVC% decline of at least 10% points achieving 0.704 AUC score. In comparison, a naive baseline using the mean FVC value as a predictor achieved much lower FVC forecasting capabilities, with 18.718 (0.317) weighted RMSE, and 17.619 (0.599) MAE. SHAP value analysis indicated that prior FVC measurements, diffusion of carbon monoxide (DLCO) values, skin involvement, age, anti-centromere positivity, dyspnea and CRP-elevation contributed most to deep-learning-based FVC predictions. Regarding uncertainty quantification, ANP RNN achieved 79% coverage (i.e. the model would provide uncertainty estimates that included the true future FVC value in 79 out of 100 predictions) out of the box, and 90% using an additional conformal prediction module with a corresponding Winkler score of 892 (indicating the width of the uncertainty estimate plus penalty for mistakes), smaller than any other model at the same coverage level. We further demonstrate how the data abstraction provided by the ANP RNN model (embeddings) allows for deriving similar patient trajectories (for similarity-based explanation). Conclusions: Our study demonstrates the feasibility of FVC forecasting and thus the ability to predict ILD trajectories in individual SSc patients using deep learning. We show that model predictions can be paired with uncertainty quantification and similarity-based model explainability, which are crucial elements for deploying trustworthy ML algorithms. Our study is thus an important first step towards reliable automated ILD trajectory (i.e. FVC%) prediction system with potential clinical utility. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement This work is funded by the Swiss National Science Foundation (project number 201184) ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: European Scleroderma Trials and Research group (EUSTAR) gave ethical approval for this work. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes The raw dataset is owned by the EUSTAR group, and may be obtained by request after the approval and permission from EUSTAR board.

In this paper, we propose a deep generative time series approach using latent temporal processes for modeling and holistically analyzing complex disease trajectories. We aim to find meaningful temporal latent representations of an underlying generative process that explain the observed disease trajectories in an interpretable and comprehensive way. To enhance the interpretability of these latent temporal processes, we develop a semi-supervised approach for disentangling the latent space using established medical concepts. By combining the generative approach with medical knowledge, we leverage the ability to discover novel aspects of the disease while integrating medical concepts into the model. We show that the learned temporal latent processes can be utilized for further data analysis and clinical hypothesis testing, including finding similar patients and clustering the disease into new sub-types. Moreover, our method enables personalized online monitoring and prediction of multivariate time series including uncertainty quantification. We demonstrate the effectiveness of our approach in modeling systemic sclerosis, showcasing the potential of our machine learning model to capture complex disease trajectories and acquire new medical knowledge.

Background/Purpose There is a need for better treatments that modify the disease course and improve symptoms of idiopathic pulmonary fibrosis (IPF) and other progressive fibrosing interstitial lung diseases (PF-ILD). BI 1015550, a preferential phosphodiesterase 4B inhibitor, prevented lung function decline in a Phase II study in IPF. Based on its anti-inflammatory and antifibrotic properties it may provide an additional treatment option, alone or with nintedanib, in patients with IPF or PF-ILD.

BackgroundHigh resolution computed tomography (HRCT) is the gold standard for the diagnosis of systemic sclerosis associated interstitial lung disease (SSc-ILD). Although there is agreement in performing HRCT as a screening test at time of SSc diagnosis, some physicians do not regularly perform baseline HRCTs. In addition, it is unclear according to which criteria HRCTs should be repeated during the follow-up of baseline ILD negative patients.ObjectivesTo develop a risk score for the presence of SSc-ILD (the ILD-RISC), to guide physicians in ordering both baseline and follow-up HRCTs.MethodsThe steering board included six SSc-ILD experts from referral centers, two fellows and a patient research partner. Items for regression analysis were selected according to face validity, feasibility, scientific background, and personal experience using the nominal group technique (NGT). The prediction model for the presence of ILD was developed from baseline visits of SSc patients from the six centers using multivariable logistic regression with backward selection. Patients were randomly divided into a derivation and validation cohort consisting of 66% and 34% of patients respectively. Patients with missing data in the selected covariates and in the ILD status (outcome) were excluded. After identifying a cut-off favoring sensitivity >85% from the ROC curve analysis, the derived ILD-RISC score was applied first in the validation cohort and then longitudinally in a cohort of SSc patients with negative baseline HRCT.ResultsThe steering board selected 13 variables deemed important in the identification of SSc-ILD: sex, age, disease duration from first non-Raynaud’s phenomenon symptom, skin subset (diffuse/limited), presence of esophageal symptoms, digital ulcers (DU) ever, arthritis ever, smoking ever, increased inflammatory markers, NYHA functional class, SSc autoantibody status (SSc_Atb), FVC% and DLCO%. Among 780/3240 patients fulfilling the inclusion criteria, 533 (43% ILD) and 247 (48% ILD) respectively constituted the derivation and the validation cohort. In the derivation cohort, a model including FVC%, DLCO%, DU ever, age and SSc_Atb (Table 1A) showed an OR of 133.9 (95% CI 53.4-335.9) and an AUC of 79.1% (95% CI 75.3-83.0%) for the presence of ILD on HRCT (Figure 1). An ILD-RISC score ≥0.3 showed sensitivity of 85.6% and specificity of 53.6%, NPV of 83.2% and PPV of 58.2%, which were replicated in the validation cohort (Table 1B). Among 819 patients with negative baseline HRCT, 170 (20.8%) developed ILD during a 3.8±3.0 years follow up (1988 visits). Longitudinally, the ILD-RISC score showed comparable sensitivity and specificity (Table 1B).Table 1.A)ILD-RISC MODEL VARIABLESOR95% CIp valueDigital ulcers, ever2.0581.347-3.145<0.001Age1.0261.010-1.0420.001SSc_ATBAnti-centromere0.3340.198-0.563<0.001Anti-topoisomerase I2.3791.326-4.2670.004Anti-RNA-polymerase III1.4070.636-3.1130.399Anti-Pm/Scl2.5560.916-7.1350.073None of the aboveComparatorFVC%0.9900.979-1.0020.091DLCO%0.9710.960-0.982<0.001B)ILD-RISC SCORE PERFORMANCEDerivation cohortValidation cohortLongitudinal cohortAll Patients/ILD patients533/229247/119819/170AUC %, 95% CI79.1 (75.3 – 83.0)76.4 (71.0 – 82.7)72.6 (68.9 – 76.2)Sensitivity %, 95% CI85.6 (80.4 – 89.9)85.7 (78.1 – 91.5)80.4 (73.9 – 86.0)Specificity %, 95% CI53.6 (47.8 – 59.3)49.2 (40.3 – 58.2)50.5 (48.2 – 52.9)Negative Predictive Value %, 95% CI83.2 (77.2 – 88.1)78.8 (68.2 – 87.1)96.3 (94.9 – 97.4)Positive Predictive Value %, 95% CI58.2 (52.7 – 63.5)61.1 (53.2 – 68.5)13.9 (11.8 – 16.1)ConclusionWe developed and validated the ILD-RISC score to predict the presence of ILD at time of diagnosis and evaluated its performance during follow-up. The ILD-RISC may be useful in routine practice when resources for HRCTs might be limited. In particular, it may also help to decide when to order HRCTs at follow up, thus limiting unnecessary HRCTs and reducing the burden for patients and institutions.Disclosure of InterestsCosimo Bruni Speakers bureau: Actelion, Consultant of: Boehringer-Ingelheim, Eli-Lilly, Grant/research support from: New Horizon fellowship, FOREUM, EUSTAR, GILS, Lorenzo Tofani: None declared, Håvard Fretheim: None declared, Sophie Liem: None declared, Arthiha Velauthapillai: None declared, Hilde Jenssen Bjørkekjær: None declared, Imon Barua: None declared, Ilaria Galetti: None declared, Alexandru Garaiman: None declared, Mike O. Becker Speakers bureau: Mepha, MSD, Novartis, GSK, Bayer and Vifor, Consultant of: Mepha, MSD, Novartis, GSK, Bayer and Vifor, Grant/research support from: Mepha, MSD, Novartis, GSK, Bayer and Vifor, Anna-Maria Hoffmann-Vold Consultant of: Actelion, ARXX therapeutics, Bayer, Janssen, MSD, Lilly, Roche, Boehringer-Ingelheim, Medscape., Jeska de Vries-Bouwstra Speakers bureau: Payment for presentations and educational events by Boehringer Ingelheim and Janssen, Consultant of: Janssen and Boehringer Ingelheim, Abbvie, Grant/research support from: Roche, Galapagos and Janssen, ZonMW and ReumaNederland, Madelon Vonk: None declared, Jörg H.W. Distler Shareholder of: J.H.W.D. is stock owner of 4D Science and Scientific head of FibroCure., Grant/research support from: J.H.W.D. has received research funding from Anamar, Active Biotech, Array Biopharma, aTyr, BMS, Bayer Pharma, Boehringer Ingelheim, Celgene, Galapagos, GSK, Inventiva, Novartis, Sanofi-Aventis, RedX, UCB., Marco Matucci-Cerinic Speakers bureau: Biogen, Bayer, Boehringer-Ingelheim, CSL Behring, Eli-Lilly., Consultant of: Actelion, Biogen, Bayer, Boehringer-Ingelheim, CSL Behring, Eli-Lilly., Grant/research support from: Actelion, Oliver Distler Speakers bureau: Bayer, Boehringer Ingelheim, Janssen, Medscape, Consultant of: Abbvie, Acceleron, Alcimed, Amgen, AnaMar, Arxx, AstraZeneca, Baecon, Blade, Bayer, Boehringer Ingelheim, Corbus, CSL Behring, 4P Science, Galapagos, Glenmark, Horizon, Inventiva, Kymera, Lupin, Miltenyi Biotec, Mitsubishi Tanabe, MSD, Novartis, Prometheus, Roivant, Sanofi and Topadur, Grant/research support from: Kymera, Mitsubishi Tanabe, Boehringer Ingelheim

OBJECTIVE:To estimate the extent of and the reasons for ineligibility in randomized controlled trials (RCTs) of SSc patients included in the EUSTAR database, and to determine the association between patient's features and generalizability of study results.METHODS:We searched Clinicaltrials.gov for all records on interventional SSc-RCTs registered from January 2013 to January 2018. Two reviewers selected studies, and information on the main trial features were retrieved. Data from 8046 patients having a visit in the EUSTAR database since 2013 were used to check patient's eligibility. The proportion of potentially eligible patients per trial, and the risk factors for ineligibility were analysed. Complete-, worst- and best-case analyses were performed.RESULTS:Of the 37 RCTs included, 43% were conducted in Europe, 35% were industry-funded, and 87% investigated pharmacological treatments. Ninety-one percent of 8046 patients included could have participated in at least one RCT. In complete-case analysis, the median [range] proportion of eligible patients having the main organ complication targeted by each study was 60% [10-100] in the overall sample of trials, ranging from 50% [32-79] for trials on skin fibrosis to 90% [34-77] for those targeting RP. Among the criteria checked, treatment- and safety-related but not demographic were the main barriers to patient's recruitment. Older age, absence of RP, and lower mRSS were independently associated with the failure to fulfill criteria for any of the included studies.CONCLUSIONS:Patient's representativeness in SSc-RCTs is highly variable and is driven more by treatment- and safety-related rather than demographic criteria.

Objectives Treatments for SSc-associated interstitial lung disease (SSc-ILD) differ in attributes, i.e. mode of administration, adverse events (AEs) and efficacy. As physicians and patients may perceive treatments differently, shared decision-making can be essential for optimal treatment provision. We therefore aimed to quantify patient preferences for different treatment attributes. Methods Seven SSc-ILD attributes were identified from mixed-methods research and clinician input: mode of administration, shortness of breath, skin tightness, cough, tiredness, risk of gastrointestinal AEs (GI-AEs) and risk of serious and non-serious infections. Patients with SSc-ILD completed an online discrete choice experiment (DCE) in which they were asked to repeatedly choose between two alternatives characterized by varying severity levels of the included attributes. The data were analysed using a multinomial logit model; relative attribute importance and maximum acceptable risk measures were calculated. Results Overall, 231 patients with SSc-ILD completed the DCE. Patients preferred twice-daily oral treatments and 6-12 monthly infusions. Patients' choices were mostly influenced by the risk of GI-AEs or infections. Improvement was more important in respiratory symptoms than in skin tightness. Concerning trade-offs, patients accepted different levels of increase in GI-AE risk: +21% if it reduced the infusions' frequency; +15% if changing to an oral treatment; up to +37% if it improved breathlessness; and up to +36% if it reduced the risk of infections. Conclusions This is the first study to quantitatively elicit patients' preferences for treatment attributes in SSc-ILD. Patients showed willingness to make trade-offs, providing a firm basis for shared decision-making in clinical practice.

Background: Systemic sclerosis (SSc) is a heterogeneous autoimmune disease frequently leading to interstitial lung disease, which highly impacts on mortality. Objective: To compare the prognostic importance of baseline and persistent inflammatory SSc phenotypes. Methods: SSc patients from the EUSTAR cohort were stratified by baseline inflammatory/non-inflammatory (CRP ≥/<5mg/l at first visit) and persistent inflammatory/non-inflammatory phenotypes (CRP ≥5mg/l at ≥80%/<20% of visits). Cox regression was used to compare mortality risks. Results: Of 5619 patients (>1 visit) 33% were stratified as baseline inflammatory; of 2883 patients (>2 visits) 14% were stratified as persistent inflammatory. Inflammatory patients had more frequently diffuse-cutaneous disease, anti-Scl-70 autoantibodies, interstitial lung disease, pulmonary hypertension, lower forced vital capacity and lower diffusing capacity for carbon monoxide. Patients with persistent inflammation had a strongly increased risk of all-cause mortality (HR 7.1 [95%CI 3.7-13.5], p<0.001, A) compared to non-inflammatory patients, whereas this association was weaker when based on a single CRP measurement (HR 2.6 [95%CI 2.1-3.2], p<0.001, B). Conclusions: The persistent inflammatory phenotype is strongly associated with mortality risk, and repetitive CRP measurements might facilitate therapeutic decisions and estimation of prognosis.

BackgroundIn the SENSCIS trial conducted in a population of subjects with SSc-ILD with a mean time since first non-Raynaud symptom of 3.5 years and 52% with diffuse cutaneous SSc (dcSSc), nintedanib reduced the rate of decline in FVC (mL/year) over 52 weeks by 44% versus placebo. Risk factors for a rapid decline in FVC in patients with SSc include early SSc, elevated inflammatory markers, significant skin involvement, and dcSSc. Patients with SSc with these risk factors for rapid progression of ILD are typically given immunosuppressants but not nintedanib.ObjectivesTo analyse the rate of decline in FVC and the effect of nintedanib on FVC decline in subjects with risk factors for a rapid decline in FVC in the SENSCIS trial.MethodsIn post-hoc analyses of data from the SENSCIS trial, we analysed the rate of decline in FVC (mL/year) over 52 weeks in all subjects and in those with early SSc (<18 months since first non-Raynaud symptom), elevated inflammatory markers (C-reactive protein ≥6 mg/L and/or platelets ≥330 x 109/L), or significant skin fibrosis using two approaches (modified Rodnan skin score [mRSS] 15-40 or mRSS >18) at baseline. We also analysed the rate of decline in FVC over 52 weeks in subjects with one of these risk factors and dcSSc.ResultsOf 575 subjects analysed, 79 (13.7%) had <18 months since first non-Raynaud symptom, 210 (36.5%) had elevated inflammatory markers, 172 (29.9%) had mRSS 15-40 and 118 (20.5%) had mRSS >18. Of 299 subjects with dcSSc, 29 (9.7%) had <18 months since onset of first non-Raynaud symptom, 129 (43.1%) had elevated inflammatory markers, 162 (54.2%) had mRSS 15-40 and 118 (39.5%) had mRSS >18. In the placebo group, the rate of decline in FVC over 52 weeks was numerically greater in subjects with these risk factors for rapid decline in FVC compared with all subjects (Figure 1). Across the subgroups, the rate of decline in FVC was numerically lower in subjects treated with nintedanib than placebo (Figure 1).Figure 1.Rate of decline in FVC (mL/year) over 52 weeks in (A) all patients and in patients with risk factors for rapid decline in FVC at baseline and (B) all patients and in patients with dcSSc and risk factors for rapid decline in FVC at baseline in the SENSCIS trial.ConclusionThe SENSCIS trial included a broad range of subjects with a fibrotic ILD complicating SSc, including those with risk factors for a rapid decline in FVC. In the placebo group, subjects with these risk factors had a more rapid decline in FVC over 52 weeks compared with the overall trial population. By targeting fibrosis with nintedanib, the rate of decline in FVC in patients with risk factors for FVC decline was reduced in patients treated with nintedanib compared with placebo.AcknowledgementsThe SENSCIS trial was funded by Boehringer Ingelheim. Toby M Maher and Masataka Kuwana were members of the SENSCIS trial Steering Committee.Disclosure of InterestsDinesh Khanna Shareholder of: Stocks - Eicos Sciences, Inc, Consultant of: AbbVie, Acceleron, Actelion, Amgen, AstraZeneca, Bayer, Boehringer Ingelheim, Corbus, CSL Behring, Galapagos NV, Genentech/Roche, Gilead, GlaxoSmithKline, Horizon Therapeutics, Merck Sharp & Dohme, Mitsubishi Tanabe Pharma, Prometheus, Sanofi-Aventis, Theraly, United Therapeutics, Grant/research support from: Bayer, Bristol-Myers Squibb, Horizon Therapeutics, Immune Tolerance Network, National Institutes of Health, Pfizer, Employee of: Leadership/Equity position – Chief Medical Officer - CiviBioPharma/Eicos Sciences, Inc, Toby Maher Speakers bureau: Boehringer Ingelheim, Galapagos, Genentech, Consultant of: AstraZeneca, Bayer, Blade Therapeutics, Boehringer Ingelheim, Bristol-Myers Squibb, Galapagos, Galecto, GlaxoSmithKline R&D, IQVIA, Pliant, Respivant, Roche, Theravance and Veracyte, Grant/research support from: AstraZeneca, GlaxoSmithKline, Elizabeth Volkmann Speakers bureau: Boehringer Ingelheim, Consultant of: Boehringer Ingelheim, Grant/research support from: Boehringer Ingelheim, Corbus, Forbius, Horizon, Kadmon, Yannick Allanore Consultant of: AbbVie, AstraZeneca, Bayer, Boehringer, Mylan, Janssen, Medsenic, Prometheus, Sanofi, Roche, Grant/research support from: Alpine Immunosciences, Medsenic, OSE Immunotherapeutics, Vanessa Smith Speakers bureau: Actelion Pharmaceuticals, Boehringer-Ingelheim Pharma GmbH&Co, Janssen-Cilag NV, UCB Biopharma Sprl, Consultant of: Boehringer-Ingelheim Pharma GmbH&Co, Janssen-Cilag NV, Grant/research support from: Belgian Fund for Scientific Research in Rheumatic diseases (FWRO), Boehringer-Ingelheim Pharma GmbH&Co, Janssen-Cilag NV, Research Foundation - Flanders (FWO), Shervin Assassi Speakers bureau: On speaker bureau for Integrity Continuing Education, Consultant of: Abbvie, AstraZeneca, Boehringer Ingelheim, CSL Behring, Novartis, Grant/research support from: Boehringer Ingelheim, Janssen, Michael Kreuter Speakers bureau: Boehringer Ingelheim and Roche, Consultant of: Boehringer Ingelheim and Roche, Grant/research support from: Boehringer Ingelheim and Roche, Anna-Maria Hoffmann-Vold Speakers bureau: Actelion, Boehringer Ingelheim, Lilly, Medscape, Merck Sharp & Dohme, Roche, Paid instructor for: Boehringer Ingelheim, Consultant of: Actelion, ARXX, Bayer, Boehringer Ingelheim, Lilly, Medscape, Merck Sharp & Dohme, Roche, Grant/research support from: Boehringer Ingelheim, Masataka Kuwana Speakers bureau: AbbVie, Asahi Kasei Pharma, Astellas, Boehringer Ingelheim, Chugai, Eisai, GlaxoSmithKline, Janssen, Nippon Shinyaku, Ono Pharmaceuticals, Tanabe-Mitsubishi, Consultant of: AstraZeneca, Boehringer Ingelheim, Corbus, Kissei, Mochida, Grant/research support from: Boehringer Ingelheim, MBL, Ono Pharmaceuticals, Christian Stock Employee of: Christian Stock is an employee of Boehringer Ingelheim, Margarida Alves Employee of: Margarida Alves is an employee of Boehringer Ingelheim, Steven Sambevski Employee of: Steven Sambevski is an employee of Boehringer Ingelheim, Christopher P Denton Speakers bureau: Boehringer Ingelheim, Janssen, Consultant of: Abbvie, Acceleron, Boehringer Ingelheim, Corbus, CSL Behring, GlaxoSmithKline, Roche, Grant/research support from: ARXX Therapeutics, GlaxoSmithKline, Horizon Therapeutics, Servier

In 1897–98, Paul Gauguin created the monumental painting Where Do We Come From? What Are We? Where Are We Going?—from the French D'où venons-nous? Que sommes-nous? Où allons-nous?—during a time of a great personal grief involving the death of his beloved daughter due to pneumonia. Studying this masterpiece on the riddle of human existence, and its depressing background, we were reminded of the current situation for interstitial lung diseases (ILDs), in terms of both our understanding of the diseases themselves and the path that clinicians, researchers, and patients have taken in responding to and influencing the course of disease and the patient experience.

Background: Some patients with SSc-ILD develop dyspnoea secondary to parenchymal lung disease, while others do not report dyspnoea even when their lung function is impaired. It is unclear whether the presence of dyspnoea is associated with a worse course of SSc-ILD or with response to therapy. Objectives: To investigate the rate of decline in FVC in patients with SSc-ILD in the SENSCIS trial in subgroups by patient-reported dyspnoea at baseline. Methods: The SENSCIS trial enrolled patients with SSc-ILD with first non-Raynaud symptom within ≤7 years before screening, extent of fibrotic ILD ≥10% on HRCT and FVC ≥40% predicted. Patients were randomised to receive nintedanib or placebo until the last patient reached week 52. In post-hoc analyses, we analysed the rate of decline in FVC (mL/year) over 52 weeks in patients with and without dyspnoea at baseline based on the question about dyspnoea in the St. George’s Respiratory Questionnaire (SGRQ). Patients who reported having shortness of breath “most days a week”, “several days a week” or “a few days a month” (rather than “only with chest infection” or “not at all”) over the last month were considered to have dyspnoea at baseline. A random slope and intercept model was used to assess the rate of decline in FVC (mL/year) and an interaction test was applied to assess potential heterogeneity in the treatment effect of nintedanib between the subgroups. Results: Of 576 patients, 69.8% had dyspnoea at baseline. At baseline, in patients with and without dyspnoea, respectively, mean (SD) extent of fibrotic ILD on HRCT was 37.7 (21.7)% and 31.6 (19.4)%; mean (SD) FVC was 71.0 (16.3) and 76.5 (16.8) % predicted; 50.7% and 44.8% were taking mycophenolate; 53.5% and 41.9% were taking corticosteroids. In the placebo group, the rate of decline in FVC (mL/year) was similar in patients with and without dyspnoea at baseline (Figure). The effect of nintedanib versus placebo on reducing the rate of decline in FVC (mL/year) was numerically more pronounced in patients without dyspnoea (difference: 79.8 [95% CI: 9.8, 149.7]) than with dyspnoea (difference: 25.7 [-19.9, 71.3]), but the exploratory interaction p-value did not indicate heterogeneity in the treatment effect between subgroups (p=0.20). Conclusion: In the SENSCIS trial, patients with SSc-ILD who had dyspnoea at baseline had a numerically greater extent of fibrotic ILD on HRCT and numerically lower FVC % predicted at baseline. The rate of decline in FVC in the placebo group was similar in patients with and without dyspnoea. Nintedanib had a numerically greater treatment effect in patients without dyspnoea. These data suggest that the presence of dyspnoea should not be used as a criterion for starting nintedanib in patients with SSc-ILD. Acknowledgements: The SENSCIS trial was funded by Boehringer Ingelheim. Medical writing support was provided by Fleishman Hillard Fishburn, London, UK. The authors meet criteria for authorship as recommended by the International Committee of Medical Journal Editors (ICMJE). Disclosure of Interests: Elizabeth Volkmann Consultant of: Boehringer Ingelheim, Grant/research support from: Corbus and Forbius, Michael Kreuter Speakers bureau: Boehringer Ingelheim, Consultant of: Boehringer Ingelheim, Grant/research support from: Boehringer Ingelheim and Roche, Anna-Maria Hoffmann-Vold Speakers bureau: Actelion, Boehringer Ingelheim, Lilly, Merck Sharp & Dohme and Roche, Consultant of: Actelion, Arxx Therapeutics, Bayer, Boehringer Ingelheim, Lilly, Medscape, Merck Sharp & Dohme and Roche, Grant/research support from: Boehringer Ingelheim, Marlies Wijsenbeek Speakers bureau: Boehringer Ingelheim (fees paid to institution) and Hoffmann-La Roche (fees paid to institution), Consultant of: Boehringer Ingelheim (fees paid to institution), Bristol-Myers Squibb (fees paid to institution), Galapagos NV (fees paid to institution), Hoffmann-La Roche (fees paid to institution), NeRRe Therapeutics (fees paid to institution), OncoArendi Therapeutics (fees paid to institution), Respivant Sciences (fees paid to institution) and Savara (fees paid to institution), Grant/research support from: Boehringer Ingelheim (fees paid to institution) and Hoffmann-La Roche (fees paid to institution), Vanessa Smith Speakers bureau: Boehringer Ingelheim and Janssen-Cilag NV, Consultant of: Boehringer Ingelheim, Grant/research support from: Belgian Fund for Scientific Research in Rheumatic diseases (FWRO), Boehringer Ingelheim, Janssen-Cilag NV and Research Foundation - Flanders (FWO), Dinesh Khanna Shareholder of: Eicos Sciences, Inc. (less than 5%), Consultant of: Acceleron Pharma, Actelion, AbbVie, Amgen, Bayer, Boehringer Ingelheim, CSL Behring, Corbus, Gilead Sciences, Galapagos NV, Genentech/Roche, GlaxoSmithKline, Horizon Therapeutics, Merck Sharp & Dohme, Mitsubishi Tanabe Pharma, Sanofi-Aventis and United Therapeutics, Grant/research support from: Bayer, Bristol-Myers Squibb, Horizon Therapeutics, Immune Tolerance Network, National Institutes of Health and Pfizer, Employee of: Chief Medical Officer- CiviBioPharma/Eicos Sciences, Inc., Christopher Denton Speakers bureau: Boehringer Ingelheim, Corbus, Janssen, and Mallinckrodt Pharmaceuticals, Consultant of: Acceleron Pharma, Arxx Therapeutics, Bayer, Boehringer Ingelheim, Corbus, CSL Behring, Galapagos NV, GlaxoSmithKline, Horizon Therapeutics, Janssen, Mallinckrodt Pharmaceuticals, Roche, Sanofi and UCB, Grant/research support from: Arxx Therapeutics, GlaxoSmithKline and Servier, Wim Wuyts: None declared, Corinna Miede Employee of: Currently an employee of mainanalytics GmbH, contracted by Boehringer Ingelheim, Margarida Alves Employee of: Currently an employee of Boehringer Ingelheim, Steven Sambevski Employee of: Currently an employee of Boehringer Ingelheim, Yannick Allanore Consultant of: Boehringer Ingelheim, Medsenic, Menarini and Sanofi, Grant/research support from: Alpine Pharmaceuticals

Background: Interstitial lung disease (ILD) has shown to worsen outcome in patients with Coronavirus disease-19 (COVID-19). Systemic sclerosis (SSc) is a severe, multi-organ disease associated with ILD. Objectives: To assess factors associated with severe outcome in SSc-ILD from European Scleroderma Trial and Research (EUSTAR). Methods: SSc patients from EUSTAR with COVID-19 were prospectively collected between 15.03.-31.12.2020. Severe outcome was defined as need of ventilation/ECMO or death. Risk factors evaluated were sex, age, comorbidities, immunosuppressive treatment, SSc subtype, autoantibodies, and other SSc associated organ manifestations. Descriptive statistics and logistic regression models were applied. Results: Among 178 SSc patients with COVID-19, 90 (51%) had ILD. Mean age of SSc-ILD patients was 59 years, 24 (27%) were male and 33 (37%) had >1 non-SSc associated comorbidity. At COVID-19 infection, 2 (8%) SSc-ILD patients used Prednisone>10mg; 4 (15%) MTX, 3 (12%) Azathioprine, 10 (39%) Mycophenolate and 5 (19%) Rituximab. Over median 5.5 weeks, 26/90 (29%) developed a severe outcome, including 17 (20%) deaths. Older age, male sex, non-SSc comorbidities and SSc associated renal or cardiac disease were associated with severe outcome in univariable logistic regression (figure). Conclusion: SSc-ILD patients with COVID-19 have a high risk of mortality. Male patients, older age, non-SSc comorbidities and other SSc organ manifestations risk a more severe outcome.

Interstitial lung disease (ILD) is a relatively frequent manifestation of systemic autoimmune rheumatic disorders (SARDs), including systemic sclerosis (SSc), rheumatoid arthritis (RA), idiopathic inflammatory myopathies (IIM), systemic lupus erythematosus (SLE), primary Sjögren’s syndrome (pSS), and anti-neutrophil cytoplasmic antibody (ANCA) associated vasculitis. Interstitial pneumonia with autoimmune features (IPAF) has been proposed to describe patients with ILD who have clinical or serological findings compatible with SARDs but they are not sufficient for a definite diagnosis. ILD may present with different patterns among patients with SARDs, but most commonly as nonspecific interstitial pneumonia (NSIP), with the exception of RA and ANCA vasculitis that more often present with usual interstitial pneumonia (UIP). The natural history of ILD is quite variable, even among patients with the same SARD. It may present with subclinical features following a slow progressively course or with acute manifestations and clinically significant rapid progression leading to severe deterioration of pulmonary function and respiratory failure. The radiographic pattern of ILD, the extent of the disease, the baseline pulmonary function, the pulmonary function deterioration rate over time and clinical variables related to the primary SARD, such as age, sex and the clinical phenotype, are considered prognostic factors for SARDs-ILD associated with adverse outcomes and increased mortality. Different modalities can be employed for ILD detection including clinical evaluation, pulmonary function tests, high resolution computed tomography and novel techniques such as lung ultrasound and serum biomarkers. ILD may determine the clinical outcome of SARDs, since it is associated with significant morbidity and mortality and therefore screening of patients with SARDs for ILD is of great clinical importance.