Peripheral blood methylation may become useful biomarker in knee osteoarthritis
Presenter: Matlock Jeffries, MD, Oklahoma Medical Research Foundation, Oklahoma City, OK, United States
Peripheral blood DNA methylation-based machine learning models for prediction of knee osteoarthritis progression: biospecimens and data from the Osteoarthritis Initiative and Johnston County Osteoarthritis Project. Abstract 1116.
An analysis of peripheral blood epigenetic patterns may predict which patients with knee osteoarthritis are likely to progress.
Knee osteoarthritis (OA) is a heterogenous disease characterized by a variety of clinical and molecular phenotypes, which makes it difficult to predict the course of the disease. None of the widely available biomarkers can accurately assess the likely progression of knee OA.
“One problem is that existing models require multiple values over multiple timepoints, which is not practical for clinical trials or for practice,” said presenter Matlock Jeffries, MD, Associate Member, Oklahoma Medical Research Foundation in Oklahoma City, OK. “So we tried epigenetic markers, peripheral blood DNA methylation patterns, which tend to be more stable over time than biologic markers. A pilot study gave us an area under the curve of .81, which prompted us to expand to a larger cohort.”
Researchers obtained buffy coat DNA from 554 participants in the Foundation for the National Institutes of Health (FNIH) Osteoarthritis Biomarkers Consortium. They analyzed for DNA methylation using 850,000 methylation sites, including progressors with pain, radiographic evidence, or both pain and radiographic evidence compared to nonprogressors. This method was also applied to 2 validation cohorts of 128 patients from the Johnston County Osteoarthritis Project and an independent cohort of 55 patients from the prior pilot study.
“We found we could get similar results looking at just the most predictive methylation markers, which turned out to be 13 sites, and developed a parsimonious model using those 13 methylation sites,” Jeffries said.
The full model predicted radiographic progression with 87% accuracy (Area Under the Curve [AUC] = .94), pain progression with 89% accuracy (AUC = .97), pain plus radiographic progression with 72% accuracy (AUC = .79), and any progression with 78% accuracy (AUC = .86). The parsimonious model had similar results with 89% accuracy for radiographic progression (AUC = .94), 90% accuracy for pain (AUC = .95), 76% accuracy for pain plus radiographic progression (AUC = .85), and 82% for any progression (AUC = .89).
Matlock pointed out that the full models were not able to discriminate radiographic-only versus pain-only in patients, with 58% accuracy (AUC = .62).
The performance of the parsimonious cohorts stood up nicely in the independent cohorts, with 80% accuracy (AUC = .88) to predict radiographic progression, he said. “Our models performed quite well on three different cohorts,” Jeffries said.
In conclusion, Jeffries said, “Peripheral blood DNA methylation models can accurately predict radiographic and/or pain progression 2 to 5 years in advance based on a single-timepoint baseline blood draw. Models cannot discriminate pain-only versus radiographic-only progression.”
In the future, the researchers plan to develop and validate an inexpensive, high-throughput version of these models to make them clinically relevant for widespread use. “This prediction of progression is best suited to enrich clinical trials in patients likely to progress. We are also very interested in answering the clinical question, in a patient with early intermittent knee pain, of whether we can predict those who will develop clinically significant OA in the next 5 years, which is part of our ongoing studies,” Jeffries said. “Perhaps if we get expensive disease-modifying drugs, we might want to intervene in patients who are likely to progress rapidly.”
References
Dunn CM, Nevitt MC, Lynch JA, et al. A pilot study of peripheral blood DNA methylation models as predictors of knee osteoarthritis radiographic progression: data from the Osteoarthritis Initiative (OAI). Sci Rep 2019; 9(1):16880. doi.10.1038/s41598-019-53298-9
Dunn C, Sturdy C, Velasco C, et al. Peripheral blood DNA methylation-based machine learning models for prediction of knee osteoarthritis progression: biospecimens and data from the Osteoarthritis Initiative and Johnston County Osteoarthritis Project [abstract 1116]. Arthritis Rheumatol 2022; 74(suppl 9). https://acrabstracts.org/abstract/peripheral-blood-dna-methylation-based-machine-learning-models-for-prediction-of-knee-osteoarthritis-progression-biospecimens-and-data-from-the-osteoarthritis-initiative-and-johnston-county-osteoarth
Disclosures
Matlock Jeffries: No disclosures