A noninterventional, hybrid study (retrospective and prospective study) with two arms comparing tofacitinib to bDMARDs treatments in patients with RA after the failure of csDMARDs was performed from March 2017 through September 2019. The retrospective period corresponds to the time from when the physician prescribed the studied drug until the patient decided to participate in the study at maximum of three weeks, while the prospective period is the period from recruitment until six months of follow-up. This study was conducted in 13 sites from Colombia (10 sites) and Peru (3 sites). Patients were followed up for 6 months and were in primary care. Changes to the treatment and use of concomitant medications during the follow-up were within current practice guidelines and were decided upon by their rheumatologist under real-world conditions. The protocol was approved by the Independent Ethics Committee at each center. All the patients provided written informed consent. Clinical trial number: NCT03073109.

The patients were ≥18 years of age, had been diagnosed with moderate to severe RA for more than 6 months before enrollment, had a Disease Activity Score 28-joint based on erythrocyte sedimentation rate (DAS28-ESR) score ≥3.2, with inadequate response to the continuous use of methotrexate or combination of csDMARDs for at least 12 weeks before the study entry, with no previous bDMARDs use, and had been prescribed tofacitinib or bDMARDs in the last three weeks at doses established by the American College of Rheumatology (ACR) guidelines published in 2015 and following medical criteria. The definition of RA diagnosis was based on the criteria used in medical practice for each rheumatologist.

The exclusion criteria were patients who were not able to answer the questionnaires; patients diagnosed with autoimmune rheumatic diseases other than RA and Sjogren's syndrome; patients treated with bDMARDs as monotherapy; patients who had participated in other studies; pregnant or breastfeeding women; and patients with any current malignancy or history of malignancy with the exception of adequately treated or excised nonmetastatic basal cell or squamous cell cancer of the skin or cervical carcinoma in situ, lymphoproliferative disorders, a history of lymphoma, leukemia, or signs and symptoms suggestive of current lymphatic disease.

Sample size was estimated based on the primary objective to compare the disease activity between both groups as measured by the Routine Assessment of Patient Index Data 3 (RAPID3). Taking into account the results obtained in prior studies with bDMARDs, a difference of 2 points in the RAPID3, with a standard deviation (SD) of 6 in both arms, a confidence level of 95% and a statistical power of 80%, approximately 142 patients were required in each arm. With a correction factor of 15% considering the proportion of patients who could drop out during the study, the minimum number of patients required was approximately 160 per arm. Sampling was conducted at the main sites using tofacitinib in the participating countries.

Convenience sampling was conducted due to the limited access to patients with tofacitinib. For patients using a biological DMARD, data was collected randomly with replacement; for this purpose, at the beginning of the week patients from each site that had started treatment with biological DMARD the previous week was listed consecutively following the order of visit and was selected through random numbers generated by computer. The candidate patients were contacted by phone and those who do not want to participate was replaced by the next patient until completing the required number of patients.

All the outcomes defined in the study were measured through PROs at month 0, it was the time when the patient signed the consent inform where the treatment was already prescribed, and at month 6 through questionnaires completed by the patients. The scales used had been previously validated in the Spanish language and registered the information reported in the medical records. The primary assessed outcome was disease activity, which was measured through the RAPID3. The other involved outcomes were functional status, which was measured through the adapted Health Assessment Questionnaire – Disability Index (adapted HAQ-DI) and quality of life, which was measured with the EuroQoL 5-Dimension 3-Level (EQ-5D-3L) Latin-American language. The only outcome measure that was extracted from the medical records was the Disease Activity Score 28-joint count assessment-erythrocyte sedimentation rate (DAS28-ESR) because it is routinely used in clinical practice.

High severity disease activity using DAS28-ESR was defined as score higher that 5.2, 5.1–3.2 as moderate severity, 3.1–2.6 as low severity, and lower than 2.6 as in remission. With RAPID3, patients who scores between 10 and 4.3 was defined high severity, 2.3 and 4.0 as moderate, 2.0 and 1.3 as low severity, and 0 and 1.0 near remission.

Adverse event (AE) frequency and severity occurring due to the prescription of therapies under evaluation were assessed after 6 months of follow-up. AE frequency and severity were measured from the information on the medical record or were spontaneously reported by the patients during the medical visit.

Other data related to the patient's demographic and clinical characteristics, treatment details such as previous (glucocorticoid, methotrexate or leflunomide) and concomitant treatments, and barriers in the health care system were abstracted from the medical records to describe the population and other possible potential confounders.

Health insurance corresponded to the type of insurance in which the patient has access to the treatment. It can be through public or private insurance, complementary which is additional insurance to the previously mentioned, or bought by the patient. Access barrier was defined as any constraints caused by administrative issues with the health care insurance or supplier reported by the patient during the follow up. Interruption was considered if the patient loses one day or more with treatment. The time to supply was measured only for the first prescription dosage delivery from the Health Maintenance Organization (HMO) or supplier to the patient, and it was expressed as the number of days required for the delivery of treatment from the time of prescription.

Matched analysis was planned through propensity score matching; however, the achieved number of recruited patients did not allow propensity score matching. Descriptive statistics were produced for all variables. These included estimates of the mean, standard deviation (SD), 95% confidence intervals of the mean, median, interquartile ranges and frequency distributions for continuous-scale variables and frequency distributions for categorical-scale variables.

DAS28-ESR was used to measure disease activity estimated by the rheumatologist. Several PROs were used to measure functional status (adapted HAQ-DI), quality of life (EQ-5D-3L) and disease activity (RAPID3) were analyzed by directly estimating the difference in means using least square means (LSM) between patients treated with tofacitinib and patients treated with bDMARDs and between periods. A bivariate and multivariable analysis was performed to identify the association between the baseline variables and the disease activity. Linear regression was conducted for the multivariable analysis of all PROs. The adjusted full model was composed by all potential confounding variables such as demographic and clinical characteristics, concomitant treatment, previous treatment and access variables (e.g. age, gender, country of origin, previous treatments, neutrophils, access barriers, insurance, baseline clinical data such as DAS28). The reduced model was developed from the results of multivariable analysis selecting the variables with p value less than 0.05. The detail results of multivariable analysis for mean changes of PROs comparing treatments were reported in the Supplementary material Tables S1–S4. The comparison of score for each group of treatments at the baseline and month 6 was conducted with a t paired in bivariate analysis and mixed-effect regression for multivariable analysis. The statistical criterion to select the final model was looking for the minor variance comparing with simple model. Multiple imputation was used to manage the missing data from the different variables using the package MICE in R software (version 4.0.5).

During the study period, 170 patients were enrolled in the study, including 100 patients treated with tofacitinib and 70 treated with bDMARDs (etanercept (39.7%), rituximab (16.2%), tocilizumab (16.2%), infliximab (13.2%), golimumab (8.8%), abatacept (2.9%), adalimumab (1.1%), and certolizumab (1.1%)). Eighty-six patients were from Colombia, and 83 patients were from Peru. At the 6-month follow-up, 92.9% (158) of enrolled patients completed the study, including 90 in the tofacitinib group and 68 in the bDMARDs group. Nine patients withdrew due to a loss to follow-up (tofacitinib (n=8) and bDMARDs (n=1)) for administrative reasons, however it was not possible to stablish if patients continued with the treatment; 2 patients treated with tofacitinib discontinued for other reasons. Only 1 patient from the bDMARDs group discontinued the study due to AEs. Although the mean time to supply at the start of the treatment was similar in both groups, more dispersion was observed in patients who received tofacitinib.

The mean age was 53.53 years (SD 13.77); 88% were women, and 97% were living in urban areas. The mean disease duration since diagnosis was 6.31 (SD 7.02) years. Methotrexate (58.82%), leflunomide (19.41%), or any chloroquine (11.76%) were the most frequent csDMARDs previously used. Corticosteroids were used previously in 82.94% of patients.

When comparing baseline data, the main differences were found in insurance characteristics and previous treatments. Patients had mainly private health insurance (n=91), followed by public health insurance (n=53). More patients with tofacitinib had coverage through private health insurance (67%), while the majority of the bDMARDs group had coverage through public health insurance (59%). The group that received tofacitinib reported more access barriers than patients with bDMARDs (29% vs 14%). Another observed difference in both studied groups was in the proportions of patients with concomitant treatment with leflunomide, previous use of methotrexate, and previous use of prednisolone. Although the mean time to supply to start the treatment was similar in both groups, more dispersion was reported in patients who received tofacitinib (Table 1).

The patients had severe disease activity according to the DAS28-ESR reported at baseline, 5.9 (SD 4.5) for bDMARDs and 5.2 (SD 1) for tofacitinib. Seven percent of patients prescribed tofacitinib had any comorbidity, whereas this value was 1.4% for the bDMARDs group. Lymphocyte counts, neutrophil counts, and the numbers of swollen joints and tender joints were similar between the study groups.

The disease activity was evaluated with DAS28-ESR. DAS28-ESR data were available for all patients at baseline; however, at the second visit, DAS28-ESR data were reported in only 86 patients treated with tofacitinib and 61 treated with bDMARDs.

The reduction of DAS28-ESR for both groups was statistically significant comparing the score at baseline and month 6 (p value<0.001) (Fig. 1). During the 6 months, the difference of score in DAS28-ESR in both visits were similar between treatments both bivariable and multivariate analyses (p value 0.099 and 0.728, respectively) (Supplementary material Table 5S). For low disease activity, 11% more bDMARDs patients than tofacitinib patients reported low disease activity (Fig. 2). Comparing the score of DAS28-ESR between baseline and month 6 visit, a reduction of the activity was found in both treatment groups with consistency of the results with bivariate and multivariable analysis (Supplementary material Table 6S).

Fig. 1.

Results of the PROs at baseline and 6 months for each treatment. *p value <0.001 for the difference between 6-month outcome value and baseline outcome value. bDMARDs: biological disease-modifying anti-rheumatic drugs; DAS28-ESR: Disease Activity Score 28-joint count assessment-erythrocyte sedimentation rate; EQ-5D-3L: EuroQoL 5-Dimension 3-Level; HAQ-DI: Health Assessment Questionnaire – Disability Index; PRO: patient-reported outcomes; RAPID3: Routine Assessment of Patient Index Data 3; SE: standard error.

(0,22MB).

Fig. 2.

Proportion of rheumatoid arthritis patients according to severity at baseline and 6 months. (A) Proportion of patients based on DAS28-ESR classification. (B) Proportion of patients based on RAPID3 severity score.

(0,27MB).

An exploratory analysis, the difference between treatments in the change in the baseline and month 6 was statistically no significant (p value 0.695).

The measurement of outcomes (RAPID3, adapted HAQ-DI, EQ5D) was conducted at baseline in 100 patients in the tofacitinib group and in 70 patients in the bDMARDs group. Ninety and sixty-eighty patients in the tofacitinib and bDMARDs groups, respectively, completed the study and had RAPID3, adapted HAQ-DI and EQ5D scores recorded at the last visit.

Reduction of RAPID3 scores from baseline to month 6 was observed in both studied treatments with statistical difference between visits (p value>0.001) (Fig. 1 and Table 2). At the end of follow-up, a similar proportion of patients treated with tofacitinib and bDMARDs achieved remission, while 18% more patients from the bDMARDs group than the tofacitinib group achieved low disease activity (Fig. 2). The RAPID score obtained at month 6 compared to baseline for both groups (bDMARD and tofacitinib) were minor with difference statistically significant being steady results in the bivariate and multivariable analysis (Supplementary material Table 7S).

The reduction in RAPID3 score during the follow-up between the bDMARDs and tofacitinib groups was not statistically significant different (p value 0.154). The multivariable analysis composed by significant variables has the same tendency remained.

Pain and health status are other dimensions included in the RAPID3. Patients who received tofacitinib reported a mean reduction in pain of 3.94±0.89, while bDMARDs patients reported a mean reduction of 3.28±0.77. In the health status dimension, the tofacitinib group presented an improvement of 3.79±0.47, whereas the bDMARDs group presented an improvement of 2.78±0.47. Differences in both groups for these outcomes were not found (p value 0.577 and 0.104).

Regarding other secondary PROs, patients who received tofacitinib or bDMARDs reported an improvement in their functional status and quality of life (Table 2). There were not statistically significant difference between groups in the mean change during 6 months (p value: 0.256 and 0.986, respectively).

The results of the bivariate analysis, multivariable analysis with all variables, and reduced model comparing the score for all PROs by treatment and by visits were consistent (Supplementary material Tables 7S–12S).

Twenty patients treated with tofacitinib and sixteen treated with bDMARDs reported any AEs. The most frequent events (over 2%) in patients treated with bDMARDs were diarrhea, pharyngitis, falls, headache, urinary tract infection, and nasopharyngitis, while those for tofacitinib group were headache, influenza, and pharyngotonsillitis. Herpes simplex and herpes zoster were reported in 2 patients treated with tofacitinib (Table 3). Cardiovascular, malignancies or thromboembolism AEs were not reported. Two serious AEs were reported by one patient treated with tofacitinib (appendicitis and peritonitis that were reported as not related to the treatment).