There are 3 basic patterns of disease progression in patients with RA. In the first pattern, patients may have a single episode of joint inflammation that remits spontaneously with or without treatment and then never recurs. A second pattern is that of chronic joint inflammation that may be controlled with medication but would be active if the patient was not treated. The third pattern is intermediate to the other patterns, in which there are remitting and recurring attacks of joint inflammation that may be controlled with medication. In general, the pattern of RA disease activity dictates the approach to treating the patient. If the patient were to have only a single episode of joint inflammation that subsequently would go into permanent remission, one could treat that episode but not need to maintain the patient on long-term therapy. However, if the patient were to have persistent disease activity, then chronic treatment is necessary.

These different patterns are especially relevant for patients who question how long they will need to be taking medications to treat their RA. For most individuals who receive methotrexate or other medications to treat their RA, we are not inclined to stop treatment because of the potential for disease recurrence. Based on shared decision making with patients, the duration of treatment with antirheumatic drugs is often indefinite because we do not want to risk a recurrence of disease activity. Thus, it would be valuable to have ways by which to identify a patient’s disease course, so as to allow treatment to be planned accordingly. This might allow treatment to be de-escalated in those patients who do not need as intensive treatment for their RA and continued in those who do.

Marks and colleagues recently reported findings from a study investigating the self-reporting of distinct RA disease courses. In an anonymous Web-based questionnaire, patients were asked to choose descriptors for their RA, which served to classify the disease course as constant, flaring and remitting, progressive, or nonprogressive. The investigators then analyzed the range of disease courses to learn how the different disease activity patterns might relate to characteristics of RA treatment, including degree of improvement with disease-modifying antirheumatic drugs. This was an interesting approach, but it should be developed further with validation of the various disease patterns by a rheumatologist. A machine learning approach, using computers to analyze massive amounts of data, might be used to identify distinct subsets of patients with unique patterns of disease activity. Guan et al recently used machine learning to predict RA patients’ responses to treatment, using both clinical and genetic data. This approach identified subpopulations of patients with RA who did not respond well to tumor necrosis factor inhibition.