New Models for RA May Predict Infection, Stroke, and Myocardial Infarction

Safety, tolerability through 24 months was consistent with 12-month tofacitinib analysis.
Researchers sought to develop and validate prediction models for a variety of adverse outcomes in patients with rheumatoid arthritis who initiated methotrexate therapy.

New prediction models for serious infection, stroke, and myocardial infarction (MI) were shown to perform well among patients with rheumatoid arthritis (RA) starting first-line methotrexate (MTX) monotherapy, according to study results published in Seminars in Arthritis and Rheumatism.

Researchers conducted a study of 15,311 internally validated individuals with RA and a larger cohort of individuals with RA for external validation. Data were collected from 15 claims databases across 9 countries. They used L1-regularized logistic regression to estimate adverse health outcome risk for different time intervals after treatment initiation: 3 months (infection, pancytopenia, and leukopenia), 2 years (stroke and MI), and 5 years (colorectal, uterine, and breast cancers). Predictors included medical history and demographic variables, and 14 other databases were used for external validation. The area under the receiver operator characteristic curve (AUC) was used to assess predictor performance.

Models predicting serious infection (AUC: internal 0.74 [95% CI, 0.68-0.80], external 0.62 to 0.83), stroke (AUC: internal 0.77 [95% CI, 0.73-0.81], external 0.63 to 0.95), and MI (AUC: internal 0.76 [95% CI, 0.72-0.81], external 0.56 to 0.82) performed favorably in terms of both calibration and discrimination. Internal discrimination was modest for other outcomes (AUC <0.65), and these models did not undergo external validation.

Limitations to this study included potential misclassification, potential lack of generalizability, and a lack of sensitivity analyses on the inclusion criteria.

The study authors concluded, “We do not believe we have sufficient evidence to recommend the use of the models in clinical practice, but research could be conducted to prove their value as implemented in administrative or EHR software to generate automatic reminders for individuals at high risk [for] these outcomes. The models may enable clinicians to provide better personalized care to RA patients initiating first-line MTX monotherapy.”

Disclosure: Several study authors declared affiliations with the pharmaceutical industry. Please see the original reference for a full list of authors’ disclosures.


Yang C, Williams RD, Swerdel JN, et al. Development and external validation of prediction models for adverse health outcomes in rheumatoid arthritis: a multinational real-world cohort analysis. Published online June 15, 2022. Semin Arthritis Rheum. doi:10.1016/j.semarthrit.2022.152050