To evaluate the efficacy and safety of extracorporeal hemoadsorption (HA) bridging to sequential Janus kinase inhibitor (JAKi) maintenance therapy in patients with difficult-to-treat rheumatoid arthritis (D2T RA).

This retrospective, single-center, observational cohort study enrolled 60 D2T RA patients treated at the Department of Rheumatology, Xi'an Fifth Hospital, from January 2020 to December 2024. All patients met the 2021 EULAR definition of D2T RA, including failure of ≥2 b/tsDMARDs with different mechanisms of action after inadequate csDMARD response, persistent moderate-to-high disease activity (DAS28-ESR > 3.2), and management perceived as problematic by both physician and patient. The majority of patients (85.0%) had failed exactly two b/tsDMARDs, while 10.0% had failed three and 5.0% had failed four or more. All patients were RF-positive and 98.3% were anti-CCP positive, representing a seropositive, inflammatory-driven D2T phenotype.

The treatment protocol comprised three phases: Phase 1 (Week 0–1), three sessions of extracorporeal hemoadsorption using HA280 neutral macroporous resin columns (Jafron Biomedical, Zhuhai, China) on days 1, 3, and 5, each processing 1.5 times the calculated plasma volume; Phase 2 (Week 1–2), JAKi initiation within 24–48 hours after the last hemoadsorption session, with agent selection based on prior JAKi exposure — patients with prior tofacitinib failure received baricitinib 4 mg once daily, those with prior baricitinib failure received tofacitinib 5 mg twice daily, and JAKi-naïve patients received either agent per physician discretion (36 patients received tofacitinib, 24 received baricitinib); Phase 3 (Week 2–24), JAKi plus methotrexate (10–15 mg/week) maintenance therapy with stable glucocorticoid dose.

The primary endpoint was drug retention rate at week 24. Secondary endpoints included DAS28-ESR change, ACR20/50/70 response rates, EULAR response, proportion achieving low disease activity (LDA) and remission, inflammatory marker dynamics (ESR, CRP, TNF-α, IL-6), IgG kinetics, HAQ-DI change, and safety outcomes. Efficacy was analyzed using non-responder imputation (NRI) as the primary analysis and completer analysis (CA) as a sensitivity analysis.

The drug retention rate at week 24 was 78.3% (47/60, 95% CI: 67.3–89.3%). All discontinuations (n = 13) occurred between weeks 12 and 24, attributed to lack of efficacy (n = 4), adverse events (n = 3), loss to follow-up (n = 3), patient withdrawal (n = 2), and relocation (n = 1). Retention rates were comparable between tofacitinib (80.6%) and baricitinib (75.0%) groups (log-rank P = 0.89).

Mean DAS28-ESR decreased progressively from 5.59 ± 0.74 at baseline to 4.31 ± 0.85 at week 12 and 4.24 ± 0.96 at week 24 (Δ = −1.38, P < 0.001). Individual disease activity components showed consistent improvements: TJC28 decreased from 15.8 to 10.4, SJC28 from 12.2 to 7.2, and patient global VAS from 69.0 to 43.2 mm. HAQ-DI improved from 1.76 to 1.38 (Δ = −0.38, P < 0.001). By NRI analysis, ACR20, ACR50, and ACR70 response rates at week 24 were 51.7%, 25.0%, and 10.0%, respectively (CA: 66.0%, 31.9%, 12.8%). LDA was achieved in 13.3% and EULAR good/moderate response in 50.0% by NRI.

Circulating inflammatory markers demonstrated sustained reductions: TNF-α decreased by 35.4% (45.8 to 29.6 pg/mL), IL-6 by 38.1% (27.8 to 17.2 pg/mL), ESR by 48.5%, and CRP by 64.2% at week 24. IgG levels showed a transient 16.5% reduction at week 2, with rapid recovery to near-baseline by week 4, indicating that the non-selective adsorptive effect of HA280 is temporary without sustained immunoglobulin depletion. This dissociation between rapid IgG recovery and sustained cytokine reduction strongly suggests that the therapeutic benefit derives from early cytokine debulking rather than prolonged immunoglobulin reduction. Exploratory subgroup analysis suggested that patients with higher baseline DAS28-ESR (≥5.5) showed a trend toward greater ACR50 response (35.5% vs 13.8%, P = 0.075), supporting a preferential benefit in the persistent inflammatory phenotype.

Adverse events occurred in 35.0% of patients (21/60) and were predominantly mild (95.5%). Hemoadsorption-related events included transient hypotension (6.7%), dizziness (5.0%), and fatigue (3.3%), all self-limiting. JAKi-related events included upper respiratory infection (5.0%), gastrointestinal discomfort (5.0%), and herpes zoster (1.7%, incidence 3.6/100 patient-years), consistent with the reported background rate for JAKi monotherapy in RA. No serious adverse events, major adverse cardiovascular events, venous thromboembolism, or malignancies were observed over 27.7 patient-years of follow-up.

Sequential hemoadsorption induction followed by JAKi maintenance therapy demonstrated acceptable efficacy and safety in D2T RA patients. The NRI-based ACR20 of 51.7% and drug retention rate of 78.3% compared favorably with published data on JAKi monotherapy in similar populations (ACR20 35–45%) and real-world sequential biologic switching (retention 60–70%). The dissociation between rapid IgG recovery and sustained cytokine reduction supports an "inflammatory debulking and therapeutic window" model, wherein rapid extracorporeal clearance of the pro-inflammatory milieu creates a reduced-inflammation environment facilitating more effective JAKi engagement with downstream JAK-STAT signaling pathways. This strategy may offer a novel therapeutic approach for D2T RA patients with a persistent inflammatory phenotype who have exhausted conventional treatment options and warrants further validation in prospective controlled trials with baseline phenotyping to identify optimal patient selection criteria.