Defining lupus and patient impact
Similar to other autoimmune conditions, lupus is a chronic disease that can cause pain and inflammation anywhere within the body and induce a wide range of other symptoms, including major fatigue, hair loss, cognitive issues, heart disease, strokes, disfiguring rashes and painful joints. With so many varying symptoms and an unknown cause for what triggers the symptoms, diagnosing and treating lupus can be difficult, leaving patients with little relief.
Despite the mysterious nature of lupus, it impacts more than 1.5M Americans with 16,000 new cases diagnosed each year in the U.S. We do know that certain populations are at higher risk for developing lupus, including women (who represent 90% of patients), African Americans, and individuals aged 15 to 44. While it’s not considered a hereditary disease, people who have a family member with lupus or another autoimmune disease are at a higher risk of developing lupus.
The most common form of lupus is systemic lupus erythematosus (SLE), affecting 70% of the lupus patient population. Two-thirds of people with lupus experience manifestation in the skin, known as cutaneous lupus erythematosus (CLE). Lupus Nephritis, a frequent complication of SLE, impacts 1 in 2 adults with lupus, and can lead to chronic kidney disease (CKD). The two overarching challenges with all manifestations of lupus are early and accurate diagnosis and optimizing treatment strategies for each patient, which can take months or even years.
There is currently no cure; standard of care is disease management. According to recent research conducted by the Mayo Clinic, lupus rates may be increasing, while risks of severity and death remain unchanged.
The value of real-world data for lupus
When it comes to understanding these patients and their unmet clinical needs, real-world evidence is key, particularly as inclusion and exclusion criteria for clinical studies are often highly restrictive and typically not representative of the overall lupus population. OM1 maintains a dataset of nearly 55,000 SLE patients in the United States, with data collected continuously from 2013. The clinical data within the OM1 PremiOM™ SLE Dataset includes medical encounters, laboratory tests, prescriptions, and clinical observations. The dataset also includes extracted and estimated measures of disease activity and severity, which can be used to describe the clinical course of lupus.
Disease activity scores are key to assessing patients’ disease burden and response to treatments. In the case of SLE, the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI), is pivotal for outcome measurement. The SLEDAI has been used extensively in clinical trials and to some extent, in clinical practice. It is simple to score and can be used to identify patients in remission both on and off- therapy. Yet, despite its relative advantages, use of SLEDAI in clinical practice is inconsistent and availability of clinician-recorded SLEDAI scores in real-world datasets is limited.
Machine learning amplifies disease activity scores
Increasing the availability of SLEDAI scores in real-world datasets would yield new opportunities for comparisons to trial data and for research on treatment patterns and patient outcomes in a real-world setting. Even for patients with disease activity scores available, estimation of additional scores at different timepoints creates a more comprehensive, longitudinal view of the patients’ journeys over time. To date, statistical methods of imputation have been used to address missing disease activity scores in research studies, but patients in real-world datasets may have no recorded disease activity scores, limiting the viability of this approach.
A case-study view
In what appears to be the first of its kind approach, OM1 developed a model to estimate disease activity scores in rheumatological conditions, such as SLE and rheumatoid arthritis (RA). The machine learning model was developed to estimate an individual patient’s SLEDAI score category (no activity, mild activity, moderate activity, or high/very high activity) at the time of a medical encounter, using unstructured data within clinical notes. Both the training cohort of encounters and a separate validation cohort were created from the OM1 PremiOM™ SLE Dataset. Model performance was assessed using the area under the receiver operating characteristic curve (AUC). The published results demonstrated that:
- A machine learning model can perform well when estimating SLEDAI score categories for patients with SLE using unstructured clinical notes
- The model could be used to estimate SLEDAI score categories from real-world data sources, making the model a valuable tool to generate real-world evidence
- The approach used to develop this model could be applied to disease activity measures for rheumatological conditions beyond lupus
A Hopeful Future
Although there is currently no cure, there is great promise in the fight against lupus. This year the Lupus Research Alliance and the FDA formed a novel private-public partnership called Lupus Accelerating Breakthroughs Consortium (Lupus ABC) aimed to advance therapeutic development for lupus. This partnership follows the recent entrance of anifrolumab to the market, the first treatment advancement for lupus in more than a decade.
Real-world data and artificial intelligence are both critical to the future of clinical research. Leveraging these tools can help the healthcare community get the right treatment to the right patient at the right time.
To learn more about how OM1 is applying real-world data and AI to advance lupus research and highlight clinical outcomes, contact us at [email protected]