Aim. To search for factors associated with the occurrence of a difficult-to-treat state, reflecting the characteristics of both the patient and his disease, as well as preceding treatment.

   Materials and methods. The analysis includes data of patients with rheumatoid arthritis (RA) from the Moscow Unified Arthritis Registry (MUAR). All included patients received вiological Disease-Modifying Antirheumatic Drugs (bDMARDs) and / or targeted synthetic DMARDs (tsDMARDs). Prior to this, all patients were treated with tsDMARDs. The group of difficult-to-treat patients included persons in whom attempts to treat bDMARDs / tsDMARDs with the first two classes of drugs were not successful, and attempts to use bDMARDs / tsDMARDs were continued. Patients in whom an attempt to treat bDMARDs / tsDMARDs of the first or second class proved effective and treatment with the appropriate drug lasted at least 2 years are not difficult-to-treat. All other cases were excluded from the analysis. The differences between the two groups were studied according to various indicators, including demographic, anthropometric, indicators describing the features of the disease, clinical manifestations of the onset and further course of the disease. From the indicators that demonstrated significant differences between the groups, significant independent predictors of difficult-to-treatstate were identified within the linear-logistic regression model by forward stepwise selection. Data describing preceding treatment were analyzed within the framework of a multifactorial linear logistic model with included selected independent predictors of achieving difficult-to-treat state (resistance to treatment).

   Results. In the study were included 1044 patients, including 112 patients (10.7 %) were classified as difficult-to-treat. As significant independent predictors of treatment resistance, the following were identified: a patient’s message about an increase in body temperature at the onset of the disease of more than 38°C (p < 0.001); a patient’s message about swelling of the hands during the disease (p = 0.033); the presence of rheumatoid nodules (p = 0.022); a relatively short time since the appearance of joint pain till contacting a doctor (p = 0.021). A significant relationship was found betweenbDMARD / tsDMARD used as first-line drug and the subsequent categorization of the patient as difficult-to-treat (p = 0.013). In patients receiving tofacitinib or rituximab in the first line of targeted therapy, the risk of subsequent treatment resistance was significantly lower than in patients receiving tumor necrosis factor inhibitors (iFNO): relative risk (RR) 0.017, 95 % confidence interval (CI) 0.04–0.72) and RR 0.47, 95 % CI 0.22–0.98), respectively. The analysis of the second line of targeted therapy revealed a significant advantage of etanercept over some other classes of bDMARDs / tsDMARDs in terms of the risk of subsequent treatment resistance. Thus, RR for other drugs of the iFNO class was 2.48 (95 % CI 1.13–5.44, p = 0.023); for abatacept – 3.15 (95 % CI 1.488–6.68, p = 0.003), and for tocilizumab – 2.57 (95 % CI 1.03–6.40; p = 0.043).

   Conclusion. The analysis of data from the registry of patients who were in real clinical practice made it possible to identify patients who accurately meet criteria of the European Alliance of Associations for Rheumatology for difficult-to-treat RA and patients who exactly do not meet them. A comparative analysis of the groups revealed predictors of following classifying the patient as difficult-to-treat. Based on the listed signs, high-risk groups of resistance to treatment with bDMARDs / tsDMARDs can be identified. The results obtained give grounds for the preferred use in such patients on the first line of treatment of tofacitinib or, possibly, other JAK kinase inhibitors, and on the second line of etanercept.