Are JAKis more effective among elderly patients with RA, smokers and those with higher cardiovascular risk? A comparative effectiveness study of b/tsDMARDs in Sweden

Objectives To investigate whether the relative effectiveness of janus kinase inhibitors (JAKis) versus tumour necrosis factor inhibitors (TNFi) or other biological disease-modifying antirheumatic drugs in rheumatoid arthritis differ by the presence or absence of risk factors for cardiovascular (CV) disease, age, sex and smoking. Methods Through Swedish registers, we identified 13 493 individuals with 3166 JAKi, 5575 non-TNFi and 11 286 TNFi treatment initiations 2016–2022. All lines of therapy were included, with the majority in second line or higher. Treatment response was defined as the proportion reaching European Alliance of Associations for Rheumatology (EULAR) good response and Clinical Disease Activity Index (CDAI) remission, respectively, within 6 months. Crude percentage point differences in these proportions (JAKis, and non-TNFis, vs TNFis) overall and by risk factors were observed, and adjusted for confounders using linear regression models. Predicted probabilities of response and remission were estimated from adjusted Poisson models, and presented across CV risk and age. Results Overall, adjusted percentage point differences indicated higher response (+5.0%, 95% CI 2.2% to 7.9%) and remission (+5.8%, 95% CI 3.2% to 8.5%) with JAKis versus TNFis. The adjusted percentage point differences for response in those above 65, at elevated CV risk, and smokers were +5.9% (95% CI 2.7% to 9.0%), +8.3% (95% CI 5.3% to 11.4%) and +6.0% (95% CI 3.3% to 8.7%), respectively. The corresponding estimates for remission were +8.0% (95% CI 5.3% to 10.8%), +5.6% (95% CI 3.0% to 8.2%) and +7.6% (95% CI 5.5% to 9.7%). Conclusions As used in clinical practice, response and remission at 6 months with JAKis are higher than with TNFi. Among patients with risk factors of concern, effectiveness is similar or numerically further increased. For individualised benefit-to-risk ratios to guide treatment choice, safety and effectiveness in specific patient segments should be considered.


Steroid use
Dispensation of ATC code H02AB06 recorded in year prior to treatment initiation.Indicator variable (Y/N).

Comorbidities
History of joint surgery History of joint surgery recorded in the 10 years prior to cohort entry.

History of diabetes
History of diabetes recorded in the 10 years recorded prior to cohort entry.Defined as a record in the National Patient Register (inpatient and outpatient components, ICD10: E10-E14, O24).Indicator variable (Y/N).

History of hyperlipidemia
History of hyperlipidemia recorded in the 10 years recorded prior to cohort entry.Defined as a record in the National Patient Register (inpatient and outpatient components, ICD10: E78).Indicator variable (Y/N).History of hypertension History of hypertension recorded in the 10 years recorded prior to cohort entry.Defined as a record in the National Patient Register (inpatient and outpatient components, ICD10: I10-I15 Indicator variable for CV risk factor (Y/N) based on emulating the oral surveillance inclusion criteria.
To have the CV risk factor, the following conditions must be met: At least one CV risk factor: • Family history of CVD ever (For female first-degree relatives: I20-I25 in NPR at age 65 or younger.For male first-degree relatives: I20-I25 at age 55 or younger) Relatives were identified using the Swedish Multigeneration Register and diagnoses identified from the National Patient Register (NPR).

Socioeconomics Education
Highest education achieved as recorded in the year prior to cohort entry.Data obtained from the Longitudinal integrated database for health  *Fully-adjusted risk ratios estimated from Poisson regression models adjusting for sex, age, CV risk (Y/N), plus line of therapy, baseline DAS28CRP (for response outcome) and baseline CDAI (for remission outcome) seropositive RA, HAQ, VAS pain, RA duration, history of joint surgery, concomitant csDMARD use, concomitant methotrexate use, prednisolone use in previous year, year of treatment initiation, origin and education.Estimates presented by sex, age, CV risk and smoking include a cohort-sex, cohort-age, cohort-CV risk and cohort-smoking interaction in the model, respectively.Separate models fitted for JAKi versus TNFi, and non-TNFi versus TNFi.
BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) Hypertension in the past 5 years (any diagnosis of I10-I15 NPR), or dispensation of an antihypertensive drug in past year (C02A, C03, C07A, C07F, C08, C09) • Dispensation of a lipid-lowering drug in the past 183 days (ATC: C10A, C10B, PDR) • Diabetes in the past 5 years (any diagnosis of E10-E14, NPR) or dispensation of an antidiabetic drug in past year (ATC codes A10A, A10B, A10XA01) • CVD in the past 5 years (main diagnosis of I20-I25, NPR)

Table S2 :
Number of treatment initiations imputed based on discontinuation within 6 months after treatment initiation Non-responders/remission imputed for those discontinuing treatment due to any reason except pregnancy and inactive disease **Responder/remission imputation performed where individuals discontinued due to inactive disease BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance Supplemental material placed on this supplemental material which has been supplied by the author(s) *

Table S3 :
Baseline characteristics of RA patients initiating JAKis, non-TNFis and TNFis in Sweden, non-

Table S4 :
Baseline characteristics of RA patients initiating JAKis, non-TNFis and TNFis in Sweden, imputed dataset, imputed variables only

Table S5 :
P-values for the inclusion of effect-modifying interaction terms from linear regression models.

Table S6 :
Covariates values used for predicting probability of outcome

Table S7 :
Adjusted risk ratios comparing EULAR good response and CDAI remission in JAKi and non-TNFi (verus TNFi)