CANTORAL included 504 patients; of these, 272 patients were CV risk-enriched (CV+ cohort; patients meeting CV risk criteria), and 232 patients were not (CV− cohort; patients not meeting CV risk criteria). The mean ± SD (median [interquartile range (IQR)]) tofacitinib exposure was 17.0 ± 10.5 months (17.0 [6.6, 24.5] months) for the CV+ cohort and 15.6 ± 10.2 months (14.3 [6.7, 23.6] months) for the CV− cohort. As of the July 2021 data-cut, 140 (51.5%) and 112 (48.3%) patients in the CV+ and CV− cohorts, respectively, had completed the study; reasons for discontinuation were: withdrawal of consent (30 [22.7%] and 33 [27.5%] patients), AE (34 [25.8%] and 17 [14.2%] patients), lack of efficacy (20 [15.2%] and 27 [22.5%] patients), other reason/reason unknown (24 [18.2%] and 20 [16.7%] patients), loss of efficacy (11 [8.3%] and 16 [13.3%] patients), patients lost to follow-up (10 [7.6%] and 7 [5.8%] patients) and death (3 [2.3%] and 0 [0%] patients).

Baseline characteristics for the overall population in CANTORAL have been reported previously [20]. As reported in Table 1, patients in the CV+ versus CV− cohort were older, had a numerically longer disease duration and were more likely to be male. Moreover, a higher proportion of current smokers were included in the CV+ versus CV− cohort. At tofacitinib initiation, patients in the CV+ cohort had higher HZ vaccination rates, were numerically less likely to be bDMARD-naïve, had numerically lower concomitant conventional synthetic DMARD (csDMARD) use and were more likely to use statins than patients in the CV− cohort.

The prevalence of comorbidities is reported in Table 1. Notably, a history of prior or current chronic pulmonary disease/asthma and malignancy was more commonly reported in the CV+ versus CV− cohort. The prevalence of CV comorbidities: diabetes mellitus, hypertension and myocardial infarction was higher in the CV+ than CV− cohort, and coronary artery disease was numerically higher. In the CV+ cohort, 24 (8.8%) patients had a history of ASCVD versus 1 (0.4%) patient in the CV− cohort.

Within the CV+ cohort, patients aged 65 years or older (N = 149) were less likely to be current smokers, and numerically more likely to be former smokers and receive statins than patients aged 50–64 years (N = 123) (Supplemental Table S1). The prevalence of comorbidities, including CV comorbidities, was mostly numerically higher in patients aged 65 years or older than patients aged 50–64 years within the CV+ cohort.

Within the CV− cohort, patients aged 50 years or older (N = 125) were less likely to be current smokers and numerically less likely to be bDMARD-naïve, more likely to be former smokers and had higher rates of HZ vaccination than patients aged 18–49 years (N = 107) (Supplemental Table S2).

In general, disease characteristics were similar between the CV+ and CV− cohorts (Table 1). In the CV+ cohort, disease activity levels were generally similar in patients aged 65 years or older versus 50–64 years (Supplemental Table S1). Within the CV− cohort, disease activity levels were generally comparable between patients aged 50 years or older versus 18 to 49 years (Supplemental Table S2).

The CV+ cohort experienced a higher incidence of TEAEs (138.5 [95% CI 127.7, 150.0] vs. 112.5 [101.8, 124.1] events/100 PY) and SAEs (17.0 [13.3, 21.3] vs. 5.6 [3.4, 8.7] events/100 PY) and a numerically lower incidence of discontinuations due to AEs (12.4 [9.3, 16.2] vs. 17.2 [13.2, 22.1] events/100 PY) than the CV− cohort (Table 2). Within the CV+ cohort, TEAEs were similar in patients aged 50–64 years versus 65 years or older; however, the incidence of SAEs was higher in patients aged 65 years or older versus 50–64 years (Supplemental Table S3). In the CV− cohort, the incidence of TEAEs and SAEs was numerically higher in patients aged 50 years or older than those aged 18–49 years (Supplemental Table S4). The most common SAEs were infections and infestations regardless of age in both cohorts. Deaths occurred in three patients (IR 1.2 [0.4, 2.8] events/100 PY) in the CV+ cohort (renal failure and sepsis; salmonella sepsis; myocardial infarction; all patients aged 65 years or older) and one patient (aged 78 years) in the CV− cohort (IR 0.3 [0.0, 1.6] events/100 PY) (acute myeloid leukaemia and pneumonia).

The most frequent AEs in CV+ and CV− cohorts included drug ineffective (9.7 [7.0, 13.1] and 14.3 [10.6, 18.8] events/100 PY), upper respiratory tract infection (4.4 [2.7, 6.9] and 3.1 [1.5, 5.5] events/100 PY), urinary tract infection (4.4 [2.7, 6.9] and 2.8 [1.3, 5.2] events/100 PY), hypertension (3.2 [1.7, 5.4] and 1.4 [0.5, 3.3] events/100 PY), headache (2.8 [1.5, 4.9] and 2.8 [1.3, 5.2] events/100 PY), diarrhoea (2.3 [1.1, 4.2] and 2.3 [1.0, 4.5] events/100 PY), pneumonia (3.0 [1.6, 5.1] and 0.8 [0.2, 2.4] events/100 PY), bronchitis (2.1 [1.0, 4.0] and 1.1 [0.3, 2.8] events/100 PY) and nausea (1.4 [0.5, 3.0] and 2.8 [1.3, 5.2] events/100 PY) (Table 2).

For the CV+ and CV− cohorts, IRs (events/100 PY) for AEs of special interest and specific MACE, malignancies (excluding NMSC), NMSC, thromboembolism events and hepatic events are reported in Table 2 and stratified by age in Supplemental Table S3 and Table S4. Serious infection IRs were 5.5 [3.5, 8.2] and 1.7 [0.6, 3.7] in the CV+ and CV− cohorts, respectively. The HZ vaccination rates were 51.8% (141/272 patients) and 35.8% (83/232 patients) in the CV+ and CV− cohorts, respectively. Six cases of HZ (serious/non-serious) occurred in the CV+ cohort (IR 1.4 [0.5, 3.0]); two cases occurred in two patients who were vaccinated (IR 0.5 [0.1, 1.7]), and four cases occurred in four patients who were unvaccinated (IR 0.9 [0.2, 2.3]). Four cases of HZ (serious/non-serious) occurred in the CV− cohort (IR 1.1 [0.3, 2.8]); one case occurred in a patient who was vaccinated (IR 0.3 [0.0, 1.6]), one case occurred in a patient who was unvaccinated (IR 0.3 [0.0, 1.6]) and two cases occurred in two patients with unknown vaccination status. In the CV+ cohort, MACE IR was 1.6 [0.6, 3.3] (all patients aged 65 years or older; two events of cardiac failure in one patient, both non-fatal; one fatal myocardial infarction and one non-fatal myocardial infarction, both not treatment related; three events of cerebrovascular accident in two patients; one patient had one non-fatal small stroke [not treatment-related], and one patient had one non-fatal stroke [not treatment-related], and one unknown event [unknown if treatment related]). Two of five patients with MACE had ASCVD; the remaining three patients with MACE had ASCVD risk factors. No MACE occurred in the CV− cohort. Both venous thromboembolism events occurred in the CV+ cohort (IR 0.5 [0.1, 1.7]).

Malignancies (excluding NMSC) IRs were 2.1 [1.0, 4.0] in the CV+ cohort (time to malignancy, 86–615 days) versus 0.3 [0.0, 1.6] in the CV− cohort (time to malignancy [prostate cancer], 652 days in one patient aged 63 years at baseline). NMSC occurred in two patients in the CV+ cohort (IR 0.7 [0.1, 2.0]) and two patients in the CV− cohort (IR 0.6 [0.1, 2.1]). Hepatic events IRs were 0.7 [0.1, 2.0] and 1.7 [0.6, 3.7] in the CV+ and CV− cohort, respectively. No gastrointestinal perforations were reported, as previously described [20].

For AEs of special interest, in the CV+ cohort, IRs were numerically higher among patients aged 65 years or older than aged 50–64 years for serious infections (7.7 [4.6, 12.2] vs. 3.0 [1.1, 6.5]), MACE (3.0 [1.2, 6.2] vs. 0.0 [0.0, 1.8]) and malignancies (excluding NMSC) (2.6 [1.0, 5.6] vs. 1.5 [0.3, 4.4]). These trends were less evident among patients in the CV− cohort when stratified by age (50 years or older vs. 18–49 years).

Mean time to all-cause discontinuation (CV+ 24.5 months [95% CI 22.4, 26.6] vs. CV− 22.4 months [20.2, 24.5]) and discontinuations due to AEs (including intolerance; CV+ 35.7 months [33.9, 37.4] vs. CV− 37.0 months [35.6, 38.4]) was similar between cohorts. Time to discontinuation due to lack/loss of efficacy was longer in the CV+ cohort (36.3 months [34.7, 38.0]) than in the CV− cohort (32.4 months [30.3, 34.4]) (Fig. 2).

Kaplan–Meier time to study discontinuation: a all-cause; b due to AE (including intolerance); c due to lack or loss of efficacy in patients who were (CV+) or were not (CV−) CV risk-enriched. aDefined as time from treatment initiation until premature study discontinuation or last available assessment date. AE adverse event, CI confidence interval, CV cardiovascular, n number of patients with event, N number of evaluable patients, NE non-estimable

At month 6, 51.5% (95% CI 44.6, 58.3) and 54.6% (46.9, 62.1) of patients had achieved CDAI LDA in the CV+ and CV− cohorts, respectively, with rates generally attained by month 3 and maintained to month 18. In the CV+ cohort, a numerical decrease in the proportion of patients achieving CDAI LDA was observed from months 12–18 (Fig. 3a).

Proportion of patients with CV risk enrichment (CV+) and those who were not CV risk-enriched (CV−) achieving a CDAI-defined LDA and remission; b DAS28-4(ESR)-defined LDA and remission; c DAS28-4(CRP) < 3.2 or DAS28-4(CRP) < 2.6a and d estimated change from baseline in HAQ-DI. Denominators for the response rates were the number of patients available at that visit. aLDA and remission values for DAS28-4(CRP) (< 3.2 and < 2.6, respectively) have not been validated [21], but are commonly used in rheumatology [22]. The dotted line at month 6 indicates the time point for assessment of the co-primary endpoints (achievement of CDAI LDA and remission). Black arrow (panel d) indicates the direction of improvement [23]. Δ change from baseline, CDAI Clinical Disease Activity Index, CI confidence interval, DAS28-4(CRP) Disease Activity Score in 28 joints, C-reactive protein, DAS28-4(ESR) Disease Activity Score in 28 joints, erythrocyte sedimentation rate, HAQ-DI Health Assessment Questionnaire-Disability Index, LDA low disease activity, LS least squares, n number of patients meeting response criteria, N number of patients with available data at each visit

The proportion of patients achieving CDAI remission in the CV+ cohort at month 6 was 12.0% (8.2, 17.2), with rates achieved by month 3 and maintained to month 18; in the CV− cohort, 19.6% (14.3, 26.4) of patients achieved CDAI remission at month 6, with an increase to 25.3% (17.6, 34.8) by month 18 (Fig. 3a).

CDAI LDA or remission rates (months 3 and 6) by prior bDMARD experience are shown in Supplemental Fig. S1. At month 6, across both cohorts, the proportion of patients who were bDMARD-naïve achieving CDAI LDA and remission was numerically higher than those who were bDMARD-experienced, with similar trends noted at month 3.

The proportions of patients achieving DAS28-4(ESR) LDA and remission were comparable between the CV+ and CV− cohorts with 31.0% (25.0, 37.7) and 28.8% (22.4, 36.2) achieving LDA, respectively, and 16.5% (12.0, 22.3) and 19.0% (13.7, 25.7), respectively, achieving remission at month 6 (Fig. 3b). Proportions of patients achieving DAS28-4(CRP) < 3.2 were 44.0% (37.3, 50.9) and 39.3% (32.1, 46.9) in the CV+ and CV− cohorts, respectively, at month 6. DAS28-4(CRP) < 2.6 was achieved by 31.5% (25.5, 38.2) and 28.8% (22.4, 36.2), respectively (Fig. 3c). Similar results were observed from months 3 to 18 in both cohorts, although a numerical reduction was observed from months 12 to 18 in the CV+ cohort (Fig. 3b, c).

Estimated mean changes to month 18 in CDAI, DAS28-4(ESR), DAS28-4(CRP), TJC28, SJC28, PtGA, MDGA and patient-reported pain were similar between CV+ and CV− cohorts (Supplemental Tables S5 and S6).

Estimated mean change in HAQ-DI was − 0.3 (− 0.4, − 0.2) in the CV+ cohort and − 0.3 (− 0.4, − 0.3) in the CV− cohort at month 6, which was achieved by month 3 and maintained to month 12. A further numerical decrease from baseline to − 0.4 (− 0.5, − 0.3) was observed in the CV− cohort from months 12 to 18 (Fig. 3d and Supplemental Table S5).