Xeljanz Edges Humira Just a Bit for Real-World RA Patients, Study Suggests

Tofacitinib (Xeljanz) appeared ever-so-slightly more effective than adalimumab (Humira) for rheumatoid arthritis (RA) in a study that used real-world patient data to simulate a randomized head-to-head trial.

Leaders of Australia’s OPAL network of rheumatologists reported that, among 699 RA patients starting their first biologic or targeted therapy for the disease, the mean 28-joint Disease Activity Score including C-reactive protein levels (DAS28-CRP) was 0.2 points lower for those on tofacitinib after 3 months (P=0.02).

However, by month 9 the difference was a nonsignificant 0.03 points, though still numerically in tofacitinib’s favor, according to Claire T. Deakin, PhD, of OPAL Rheumatology Ltd. in Sydney, and colleagues writing in JAMA Network Open.

The researchers emphasized that, even at the 3-month time point, both agents “led to substantive reductions in mean DAS28-CRP, consistent with remission.”

Overall, the results generally supported those from manufacturer-sponsored clinical trials and published treatment guidelines, Deakin’s group indicated.

Tofacitinib is an oral drug that inhibits Janus-associated kinase enzymes; adalimumab, an injectable that inhibits tumor necrosis factor, is the long-time leading biologic drug for RA. Although a previous head-to-head, randomized trial did compare the two agents, clinicians tend to be somewhat skeptical of such studies: their strict inclusion and exclusion criteria leave it unclear how the drugs may perform in the broader run of patients who don’t meet those criteria because of comorbidities or other factors.

OPAL Rheumatology is a nonprofit organization in which about one-third of all rheumatologists in Australia participate, sharing electronic medical record data from their rheumatic disease patients. All told, the OPAL database covers more than 200,000 patients. Patients consent to this use of their data.

RA patients make up about one-quarter of the OPAL database. Out of these approximately 50,000, Deakin and colleagues identified 699 who started adalimumab (n=464) or tofacitinib (n=235) as their first biologic/targeted drug during April 2015 through December 2020 and who had clinical data including spanning the period from 6 months before initiating these medications to at least 9 months afterward. The baseline and follow-up data had to include at least one component of the DAS28-CRP scoring system. All of the adalimumab patients were using original branded Humira, not a biosimilar version.

Patients’ general characteristics did not differ markedly between the treatment groups and were about as expected for those initiating biologic/targeted therapy for the first time. About 70% were women, mean age 57, with median time since diagnosis of about 1 year. Median DAS28-CRP score was 5.7 in the adalimumab group and 5.4 for those starting tofacitinib.

One of the researchers’ missions was to try not to exclude patients because of missing data. Consequently, they employed “multiple imputation by chained equations using the random forest algorithm under the missing-at-random assumption … to impute missing data for the components of the DAS28-CRP at baseline and follow-up, generating 10 imputed data sets.” Only patients with zero observed DAS28 components at follow-up were left out of the final analysis. Results were also weighted to account for the modest differences between treatment groups at baseline; thus, baseline DAS28-CRP was 5.3 for both treatment groups after this adjustment.

By month 3, mean weighted scores were 2.6 with adalimumab and 2.4 with tofacitinib. At month 9, both groups had scores of 2.3 when rounded to one decimal place. Safety was not examined in the study, which was as much intended to demonstrate a “target trial emulation” research approach as to provide clinical guidance.

Deakin and colleagues acknowledged that, despite their efforts to mimic a randomized trial, there could have been undetected reasons underlying the original decisions to start one drug versus the other, which then might have influenced the outcomes.