Prediction in rheumatoid arthritis to enable individualised therapy

Rheumatoid arthritis (RA) is a chronic inflammatory disorder of autoimmune nature, primarily manifesting as polyarthritis. It is a complex disorder, of combined genetic and environmental etiology, that can lead to substantial loss of mobility and irreversible joint damage, if left untreated. Today, modern breakthroughs have allowed tremendous strides to be made in treating RA, with remission now being an attainable goal following appropriate and efficient treatment. Nevertheless, interpatient-variability remains high within treatment response, where most patients attempt multiple disease-modifying antirheumatic drugs (DMARDs) throughout the course of their disease. In this thesis, we investigated opportunities for tailoring treatment and care at an individualized level, primarily through the identification of factors that would allow stratification of patients, and more rapid advancement towards relevant therapy. We concentrated our attention on common genetic variants and their contribution to aspects of RA patient prognosis, with a particular focus on cardiovascular comorbidity and methotrexate (MTX) treatment response.

In Study I, we explored the increased risk of cardiovascular disease experienced by patients with RA, by studying the genetic overlap between RA and myocardial infarction (MI). Employing a sample consisting of 26,637 Swedish RA patients and RA-free controls, we performed a genome-wide association study (GWAS), paired the resulting GWAS summary statistics with publicly available GWAS summary statistic data on MI, and assessed the genetic overlap between the two traits through estimation of genetic correlation. Ultimately, we found that genome-wide genetic correlation between RA and MI was minor overall (rg=0.13, 95%CI -0.03-0.29) with no robust evidence of strong local genetic correlation, concluding that overall genetic overlap between the two phenotypes was minor.

The remaining three studies explored the genetic underpinnings of treatment response in patients treated with MTX, focusing on persistence to treatment, i.e. remaining on MTX with no additional DMARDs prescribed. In Study II we investigated whether persistence to treatment with MTX aggregated within families, and quantified the underlying family-based heritability of the phenotype. Using 357 pairs of first-degree relatives concordant for early RA and treatment with MTX as first-line therapy, we found no evidence of familial aggregation of persistence at one year per the estimated risk ratio (RR=1.02, 95%CI 0.87-1.20), with a modest effect for persistence at three years (RR=1.41, 95%CI 1.14-1.71). Heritability estimates supported this, with a minor heritability for persistence at one year (h2=0.08, 95%CI 0-0.43) and a modest heritability for persistence at three years (h2=0.58, 95%CI 0.27-0.89), from which we concluded that a familial component was present for persistence at three years, but not for persistence at one year.

Developing on the previous study, in Study III we used data on common genetic variants to study the genetic component of persistence at a molecular level. Here, we used fully imputed genotype data on 3902 early RA patients treated with MTX as first-line therapy to estimate heritability and perform a GWAS on the two phenotypes of persistence at one and three years, respectively. Heritability estimates suggested a modest genetic component for persistence at three years (h2=0.45, 95%CI 0.15-0.75), with minor heritability observed for persistence at one year (h2=0.14, 95%CI 0-0.40). Nevertheless, we were unable to identify any genome-wide significant associations (p < 5 x 10-8) between common genetic variants and either of the two phenotypes, although several hundred variants exhibited suggestive associations (p < 5 x 10-5), from which we concluded that any genetic influence on persistence to treatment with MTX is likely to be strongly polygenic.

Lastly, in Study IV, we used the totality of available data to assess our ability to predict the outcome of persistence to treatment with MTX, using machine learning. We used data from our extensive register linkage, combining data on demographics, clinical presentation as well as medical- and prescribed drug history, with fully imputed genotype data at common genetic variants, for model training. Using a cohort of 2432 early RA patients initiating MTX as firstline therapy, we achieved minor improvements over random chance, per the area under the curve, in predicting persistence at one year (AUC=0.62, 95%CI 0.57-0.68) and persistence at three years (AUC=0.63, 95%CI 0.58-0.68), with negligible improvement in overall prediction quality upon including genotype data. We thereby concluded that despite the extensive and granular training data – including genotype data on a genome-wide set of common genetic variants – predictive quality was generally weak.

List of scientific papers

I. Sysojev, A.O., et al., Minor Genetic Overlap Among Rheumatoid Arthritis, Myocardial Infarction, and Myocardial Infarction Risk Determinants. Arthritis Rheumatol. 2024. 76(9): p.1344-1352.
https://doi.org/10.1002/art.42918

II. Sysojev, A.O., et al., Does persistence to methotrexate treatment in early rheumatoid arthritis have a familial component? Arthritis Res Ther. 2022. 24(1): p. 185.
https://doi.org/10.1186/s13075-022-02873-z

III. Sysojev, A.O., et al., Genome-wide investigation of persistence with methotrexate treatment in early rheumatoid arthritis. Rheumatology (Oxford). 2024. 63(5): p. 1221-1229.
https://doi.org/10.1093/rheumatology/kead301

IV. Sysojev, A.O., et al., The Impact of Genetics on Predicting Methotrexate Treatment Outcomes in early Rheumatoid Arthritis. [Manuscript]