On difficulties to define prognostic factors for clinical practice in rheumatoid arthritis

In rheumatoid arthritis (RA), the identification of prognostic factors (PF) capable of predicting disease outcome, response to treatment or success of dose reduction is an important issue, as these factors are intended to serve as a basis for decision-making. The task is complex from the outset, as the definition of disease prognosis or therapeutic prognosis is not uniquevocal. The heterogeneity of the definitions used partly explains the failure to identify PF that can be applied at an individual level. But other factors also contribute. First, the scope of the disease studied is too broad, including nosologically different entities. Second, potential PF are only measured at a single point of time, whereas changes over a period of time should be taken into account to a greater extent, not forgetting the potential impact of the treatment received during this period. Beyond these limiting factors, one of the main obstacles to the identification of PF is probably the fact that the phase of the disease is not sufficiently taken into account. Predicting the disease outcome when it is well established is a more complex challenge than when it is just beginning, as many factors are likely to interfere. The same applies to therapeutic PF, which should be determined according to disease duration. Difficulties also arise from the approaches used, which are often restricted to a single field of interest whereas they should be much more integrative and call on new large-scale data analysis tools with a view to precision medicine. In RA, prognosis can be defined at two levels: disease outcome, including joint damage and risk of extra-articular manifestations and/or complications, and treatment outcome, including response to therapy, risk of adverse effects and drug-free remission.