The future of systemic sclerosis classification

A recent review published in Nature Rheumatology focused on research into the classification of scleroderma, which suggests that the current classification does not capture the heterogeneity of the condition.

A recent review published in Nature Rheumatology1, entitled 'Towards a new classification of systemic sclerosis' focused on the work conducted to assess whether the dichotomous approach to grouping those with scleroderma in either the localized cutaneous systemic sclerosis (lcSSc) category or the diffuse cutaneous systemic sclerosis (dcSSc) category was the most appropriate and effective. Authored by Dr Monique Hinchcliff and Dr J. Matthew Mahoney of Yale School of Medicine, the paper recognises that the current classification does not allow for the variability of the clinical course of the disease, meaning that the utility of this classification is limited.

SSc is described as having heterogeneity, whereby the condition presents itself in different manifestations in different people. This is due to a range of biological factors, and means that it can be viewed as being on a spectrum. The European Scleroderma Trials and Research (EUSTAR) cohort captured data for 24 clinical features from a group of people with SSc, and applied clinical informatics techniques to develop a SSc classification scheme that has a data-driven basis. Two dominant groups were defined, both mainly associated with the extent of skin involvement, however 39% of lcSSc patients and 19% of dcSSc patients did not fall into either group. This emphasises the clinical need for a new classification strategy that goes beyond the current grouping system.

Leading from this, the investigators from EUSTAR sought to evaluate if 6 categories that are more specific would capture further variation in patient survival and would prove to be more comprehensive. EUSTAR has broad community support, with over 11,000 patients at 137 referral centres in Europe, the USA and Asia, which acts as a huge resource for gathering data for clinical informatics. A major challenge however is missing data, a hurdle that often occurs with patient registries. 52% of patients in the EUSTAR cohort were not followed up on, which is a significant proportion. It is evident that the SSc community needs to find novel ways to ensure that patients with SSc are included and maintained in registries. The investigators referred to the potential of partnering with patients' groups, industry and philanthropic organisations, as this could dramatically improve registry retention.

The investigation also demonstrated the importance and ever-growing role of machine learning within clinical data analysis. The traditionally statistical approaches have been preferred in the past during this form of data analysis, but an increasing number of investigations are now involving new statistical approaches or new machine learning techniques. In this EUSTAR study, a combination of these approaches was utilised. The authors of the paper stated that future large-scale studies should account for the availability and requirements of analytical methods for clinical data. This would require additional input from data scientists during all study phases, from the initial feasibility studies through to post study analyses.

The paper made reference to the ongoing debate regarding high-throughput molecular data, with some support for rheumatic disease classifications that are based only on data available to clinicians, versus those who believe that high-throughput data should be included. High-throughput data refers to the results gather Interestingly, the EUSTAR study does not make use of high-throughput molecular data, instead choosing to use organ damage data to inform the risk of more organ damage and death. The authors of the paper argue that a major drawback of this is that indicators are missed that can actually be detected at a molecular level; for example, skin gene expression can predict the extent of skin involvement in SSc.

The conclusive comments from the paper are that optimum SSc classification will combine all the characteristics of SSc with machine learning strategies. The 6 clusters observed by the EUSTAR cohort indicate that more detailed classification is required during the diagnosis of scleroderma, further allowing for better risk stratification. This process is reliant on further detail being gathered on the molecular and cellular mechanisms of the condition, as well as a shared commitment by funding agencies, health insurance providers and rheumatologists, and communication between the patient community and all professionals. The ideal outcome from a new classification system would be the identification of meaningful subsets during the early stages of SSc before the development of any internal damage, which would enable doctors to adjust and monitor treatment methods to limit the extent of illness and help maintain quality of life.

If you are interested in helping SRUK to fund work like this, then please donate today. We rely on the generosity of our community to continue to support groundbreaking research in both scleroderma and Raynaud's.

If you would like information on scleroderma in different parts of your body, please visit: Scleroderma and your body

If you would like to attend our regional conference in Glasgow in October, then please register here: Glasgow Conference 2019

[1] Hinchcliff, M., & Mahoney, J. M. (2019).Towards a new classification of systemic sclerosis. Nature Reviews Rheumatology.