Can Artificial Intelligence (AI) improve the detection and treatment of Interstitial Lung disease in Scleroderma?

By Associate Professor Jo Sahhar

How common is Interstitial Lung Disease (ILD) in scleroderma?

Systemic sclerosis or Scleroderma (SSc) is a relatively rare disease with a prevalence estimated at 6000 cases across Australia. or 8/100,000. Although SSc is uncommon, Interstitial Lung Disease (ILD), also known as lung or pulmonary fibrosis, occurs frequently in SSc, with 25- 80 % of SSc patients being affected, depending on the ILD definition used. 

More than 2000 patients across Australia are enrolled in the Australian Scleroderma Interest Group (ASCS) database and undergo baseline and annual screening for ILD. A recent examination of this database demonstrated that about 25% of patients have ILD on HRCT, although this may underestimate the prevalence, as not all patients have undergone HRCT chest to screen for ILD. Of those identified with ILD on HRCT, 50% had evidence of progressive SSc-ILD with about 15% progressing in the first 12 months, other patients have mild, stable or non-progressive ILD

What is the impact of ILD in patients with Scleroderma?

The impact of SSc-ILD on patients and the community is substantial.  Due to thickening of the tissue in the lung, oxygen exchange is impaired, leading to dry cough, shortness of breath and fatigue. Data from the EUSTAR scleroderma database in Europe showed ILD was the leading cause of death in patients with diffuse cutaneous SSc (responsible for 53% of deaths) and the third most common cause of death in those with limited cutaneous SSc, accounting for 35% of deaths. In a cohort of patients with early disease from Australia, Canada and Spain, ILD was responsible for 21% of deaths. SSc-ILD is also responsible for reduced quality of life in patients with ILD. 

What treatments are available for ILD in patients with Scleroderma?

Are these treatments safe, effective and available?

Until recently, there were limited treatment options for ILD.  Trials of Cyclophosphamide treatment showed modest improvement in lung function but improvement was not maintained after stopping treatment, and long-term follow-up showed high rates of side effects such as infections, cancers and bladder problems. Mycophenolate mofetil was shown to be as effective as cyclophosphamide with fewer side effects and is often used as the first treatment in patients with ILD, but some patients still have progression of ILD despite these therapies. 

 Fortunately, over the last few years, there have been trials of newer therapies for ILD including nintedanib, an effective antifibrotic therapy for patients with progressive lung fibrosis and rituximab, an immunosuppressive therapy previously used in rheumatoid arthritis, which have been shown to stabilise lung function in ILD. Autologous stem cell transplantation has also been shown to improve lung function and reduce ILD-associated mortality. However, while these therapies are more effective at stabilising lung function, they do have associated side effects and risks such as gut symptoms, infections and heart issues so it is very important for us to be able to identify which patients are most likely to have progressive lung disease and require these new therapies and to differentiate these patients from those with mild, stable disease in whom the risk of these therapies may outweigh the benefit. Currently, PBS funding of expensive antifibrotic therapies requires demonstration of disease progression based on symptoms, breathing tests and HRCT chest.  Some studies have shown that being able to detect and treat lung fibrosis earlier may be associated with better outcomes 

Why and how do we screen for ILD in patients with Scleroderma?

We know that by the time patients present with the symptoms of ILD, such as cough and shortness of breath, they have already lost significant lung function. Therefore, regular screening for ILD in patients with SSc before they develop symptoms is needed to detect early disease and monitor for progression of fibrosis. 

Consensus screening guidelines for SSc-ILD suggest that patients with scleroderma should have pulmonary function tests (PFTs) when they are first diagnosed and then at least annual testing with more frequent tests, every 3-6 months in high-risk patients or where ILD has already been diagnosed and is being monitored or treated.   Currently high. Resolution CT (HRCT) of the chest is used to confirm or exclude ILD and is recommended in all patients at presentation so we have a baseline scan. Scans are repeated if lung function or symptoms decline. Where previously scans were performed less commonly due to concerns about radiation, now lower and safer doses of radiation are used in performing HRCT chest, allowing us to monitor patients. with serial scans over time to help guide treatment.  However, baseline and serial scans are not routinely performed in all centres, with access to scans being limited, particularly in rural centres

Are current screening tests effective and reliable at detecting early disease and monitoring or progression of ILD?

While breathing tests are helpful in screening, they can be unreliable at times as they can be affected by other factors such as tight chest skin, variable or reduced patient effort due to muscle disease, fatigue or infection and by technical variation between machines.

We know that HRCT chest is more sensitive at detecting early ILD than lung function tests.  Moreover, studies performed utilising data from the Australian scleroderma cohort study (ASCS) have confirmed that a greater extent of fibrosis on CT at baseline (> 20%) has been shown to predict poor outcomes, as has a decline in lung function. However, how HRCTs are performed and reported in the community is variable. Many scans are performed outside of larger teaching hospitals, with radiology reports lacking the necessary detail that is required to identify patients with early or progressive disease who will benefit from therapy. Moreover, recent studies indicate that even more than just 10% of lung involvement may predict future mortality. But currently the ability of radiologists to detect this early disease is limited. There is also significant variability in the reporting of disease extent, which is problematic when trying to demonstrate progression of fibrosis on serial CT scans to enable access to expensive therapies for patients.  Access to high-quality services is particularly limited in rural settings where access to scanning facilities and reporting by expert lung radiologists is very limited, resulting in potential delays in diagnosis and treatment. 

 Can the use of Artificial Intelligence (AI) tools improve ILD screening?

Recently published studies by Australian lung specialist and researcher, Prof Tamera Corte and post -doctoral fellow Dr. Caitlin Fermoyle, who work at the Royal Prince Alfred (RPA) Hospital in Sydney, in Australian patients with other forms of lung fibrosis, not related to scleroderma, showed that the use of AI tool SOFIA was effective in identifying patterns of progressive lung disease and predicting outcomes in patients with lung fibrosis. The performance of SOFIA, a tool which can read scans without human input, was superior to that of expert lung radiologists in predicting outcomes. This study was performed in collaboration with Simon Walsh, an expert in AI from the Imperial College in London, who developed SOFIA and who has now developed another AI tool, Qureight which has been designed to detect 4 different parameters of fibrotic lung disease which may be useful in assessing disease extent and other important findings in scans of patients. with Scleroderma ILD.

ASCS HRCT Chest repository -applying SOFIA and QUREIGHT to assess the utility of AI in the detection and management of Scleroderma ILD (SScILD)

The Australian Scleroderma Interest Group (ASIG) is a collaborative group of Australians. Researchers who care for patients with scleroderma and recruit patients to be involved in the Australian Scleroderma Cohort Study (ASCS). The data. collected annually from patients in this study enables us to perform research to understand more about SSc and its treatment to improve patient outcomes and quality of life. ASIG is interested in determining whether the use of these new AI tools, SOFIA and Qureight may enhance screening and management of ILD in patients with SSc.

 With support from Boehringer Ingelheim, ASIG has established a digital HRCT repository on the XNAT platform, on which we have collected and stored over 2000 de-identified HRCT chest scans from patients enrolled in the ASCS.  We will be. working collaboratively with Professor Tamera. Corte and Simon Walsh in a project which will test whether SOFIA and Qureight can detect early ILD, detect disease progression on serial scans and be used to monitor response to therapy. We will also investigate whether analysis of the baseline scans with these tools allows us to better predict progression so that we can identify patients who require early or more aggressive treatment. If shown to be effective, these tools could be incorporated into everyday practice, helping to improve access to accurate, timely and reliable CT scan assessment in patients with or at risk of scleroderma-associated ILD.

 Availability of more accurate and reproducible HRCT assessments could also improve the ability to recruit patients and assess outcomes in clinical trials of new therapies being developed and trialled for the treatment of SScILD.

 Moreover, we hope in the future to utilise the scans in the digital repository to learn more about other aspects of scleroderma, including Pulmonary arterial hypertension, lung nodules, oesophageal disease and scleroderma associated with occupational silica and dust exposure. 

This work is being performed by Dr Emily Lin, rheumatology fellow and PhD student at Monash Health and Monash University and Dr Caitlin Fermoyle. Emily is being supervised by A/Prof Joanne Sahhar and Professor Eric Morand at Monash Health and by Prof Mandana Nikpour and Prof Tamera Corte at RPA. The project will draw upon the 2000 scans collected in our digital repository as well as the clinical data, including lung function data stored on ASCS participants from ASIG centres around Australia.   

 This project builds upon work that ASIG has performed and published previously in collaboration with experts A/Profs Tamera Corte and Nicole Goh examining the role of HRCT chest and serial lung function in predicting outcomes in SScILD and on work performed with Prof Corte, Dr Adelle Jee and Dr Matthew Parker examining the use of blood “biomarkers” in detecting and predicting ILD.

 None of this research would be possible without the contribution of more than 2000 patients who have participated in the Australian scleroderma Cohort Study (ASCS) and consented to the use of their de-identified clinical data, scans and blood samples.

 We hope that this study will improve our ability to accurately detect, treat and monitor ILD in patients with SSc, enhance assessment of new therapies and improve outcomes and quality of life for patients with SSc.