Eligible studies

Seven RCTs and five LTE studies were eligible, including tofacitinib (two RCT and one LTE); upadacitinib (two RCT and two LTE); filgotinib (one RCT and one LTE), deucravacitinib (one RCT) and brepocitinib (one RCT and one LTE). Details of included studies [22,23,24,25,26, 29, 30, 40,41,42,43,44] and the flow chart of study selection are provided in Appendix 3–5, respectively. Out of seven eligible RCTs [22, 23, 25, 26, 29, 42, 44], seven were compared with placebo groups [22, 23, 25, 26, 29, 42, 44] and two were compared with TNFi group (adalimumab) [22, 25]. Among the five eligible LTE studies [24, 30, 40, 41, 43], five had placebo groups [24, 30, 40, 41, 43] and two had TNFi groups (adalimumab) [24, 40]. All studies were multicenter studies, with eight being global and four being European. Of 12 eligible RCTs and LTEs, ten (83.3%) were considered to have risk of bias across all domains and two (16.7%) were considered to have some bias (Appendix 6).

Incidence Rate of MACE and TE Events

For the combined RCT and LTE data, there was a total of 9097.7 person-years of exposure in the JAK inhibitor groups (mean: 758.1 person-years per study, SD for follow-up: 59.3 weeks, n = 7378 participants and mean follow-up: 64.7 weeks): tofacitinib with 4145.7 person-years (n = 2398); upadacitinib with 4314.5 person-years (n = 3563); filgotinib with 287 person-years (n = 130); deucravacitinib with 45.6 person-years (n = 137); and brepocitinib with 304.9 person-years (n = 370). For RCT data, there was a total of 1934.2 person-years of exposure in the JAK inhibitor groups (mean: 276.3 person-years per study, SD for follow-up: 15 weeks, n = 3493 participants and mean follow-up: 14.3 weeks): tofacitinib with 746.5 person-years (n = 1026); upadacitinib with 1066.5 person-years (n = 2113); filgotinib with 25 person-years (n = 65); deucravacitinib with 45.6 person-years (n = 137); and brepocitinib with 50.6 person-years (n = 152).

The total person-years of exposure in the placebo group amounted to 1365.7 person-years (mean: 124.2 person-years per study, SD 4.3 weeks, n = 3027 participants and mean follow-up: 20.3 weeks), and exposure in the adalimumab group was 2250.4 person-years (mean: 562.6 person-years per study, SD 48.5 weeks, n = 2136 and mean follow-up: 26.7 weeks).

For combined RCT and LTE data, there were 23 MACE events and 26 TE events, corresponding to an incidence rate (IR) of 0.25 MACE per 100 person-years and 0.29 TE per 100 person-years of exposure. In all study groups from eligible randomized controlled trials (RCTs), 12 major adverse cardiovascular events (MACE) and eight thromboembolism events (TE) reported, with an incidence rate of 0.62 MACE per 100 person-years and 0.41 TE per 100 person-years, respectively.

Network Meta-Analysis

Network plot geometry is shown in Fig. 1. Mixed comparisons among 11 interventions of the five JAKis through NMA are depicted in Figs. 2 and 3. Estimates of MACE and TE risk from direct comparisons between JAKis and placebo from (1) combined RCT and LTE data as shown in Fig. 4 and (2) all eligible RCTs are shown in Fig. 5. Using the SUCRA approach to rank the risk of MACE and TE among JAKis, placebo, and adalimumab, details are presented in Appendix 7.

Fig. 1

Network diagrams depicting treatment comparisons. On the network diagrams, the size of each node is weighted based on the number of participants receiving a certain treatment, and the thickness of the lines connecting two nodes is weighted according to the number of studies that applied the connected treatment. MACE major adverse cardiovascular events, TE thromboembolism events, RCT randomized controlled trials, LTE long-term extension

Fig. 2

Mixed network meta-analysis of MACE and TE compared to placebo. Comparisons should be read from left to right. The upper part is for MACE and the lower part is for TE. RCT + LTE data is in orange and RCT data is in blue. The risk estimate is located at the intersection of the column defining treatment and the row defining treatment. For MACE and TE risk, an RR > 1 in orange favors the row defining treatment, while an RR > 1 in blue favors the column defining treatment. MACE major adverse cardiovascular events, TE thromboembolism events, RCT randomized controlled trials, LTE long-term extension, RR risk ratio

Fig. 3

Mixed network meta-analysis of MACE and TE compared to adalimumab. Comparisons should be read from left to right. The upper part is for MACE and the lower part is for TE. RCT + LTE data is in orange and RCT data is in blue. The risk estimate is located at the intersection of the column defining treatment and the row defining treatment. For MACE and TE risk, an RR > 1 in orange favors the row defining treatment, while an RR > 1 in blue favors the column defining treatment. MACE major adverse cardiovascular events, TE thromboembolism events, RCT randomized controlled trials, LTE long-term extension, RR risk ratio

Fig. 4

Comparison of JAK inhibitors with placebo or adalimumab in RCT + LTE data. Derived from Figs. 2 and 3. MACE major adverse cardiovascular events, TE thromboembolism events, RCT randomized controlled trials, LTE long-term extension, RR risk ratio, JAK Janus kinase inhibitors, CI confidence interval

Fig. 5

Comparison of JAK inhibitors with placebo or adalimumab in RCT data. Derived from Figs. 2 and 3. MACE major adverse cardiovascular events, TE thromboembolism events, RCT randomized controlled trials, LTE long-term extension, RR risk ratio, JAK Janus kinase inhibitors, CI confidence interval

NMA Estimates of MACE and TE Risk Comparing JAKi with Placebo in NMA

For MACE risk, traditional pairwise meta-analysis of direct JAKi–placebo comparisons either in combined RCT and LTE data (n = 13) (Appendix 8.1.1) or in RCT data (n = 7) (Appendix 8.1.3) yielded similar estimates. Direct comparisons with placebo in combined RCT and LTE data (Figs. 2 and 4) showed filgotinib (200 mg), brepocitinib (10 mg, 30 mg, 60 mg) and tofacitinib (5 mg) had a nonsignificant higher risk for MACE with RR range from 8.23 to 1.29; while upadacitinib (15 mg, 30 mg), deucravacitinib (6 mg, 12 mg) and tofacitinib (10 mg) had a nonsignificant lower risk of MACE with RR range from 0.12 to 0.58. Direct comparisons with placebo in RCT data (Figs. 2 and 5) showed filgotinib (200 mg), brepocitinib (10 mg, 30 mg, 60 mg) and tofacitinib (5 mg, 10 mg) had a nonsignificant higher risk for MACE with RR range from 22.97 to 1.09; while upadacitinib (15 mg, 30 mg), deucravacitinib (6 mg, 12 mg) had a nonsignificant lower risk of MACE with RR range from 0.09 to 0.25.

Regarding TE risk, traditional pairwise meta-analysis of direct JAKi–placebo comparisons in combined RCT and LTE data (n = 6) (Appendix 8.2.1) or in RCT data (n = 6) (Appendix 8.2.3) yielded similar estimates. Direct comparisons with placebo in combined RCT and LTE data (Figs. 2 and 4) showed filgotinib (200 mg), brepocitinib (10 mg, 30 mg, 60 mg) had a nonsignificant higher risk for MACE with RR range from 51.08 to 1.29; while upadacitinib (15 mg, 30 mg), deucravacitinib (6 mg, 12 mg) and tofacitinib (5 mg, 10 mg) had a nonsignificant lower risk of MACE with RR range from 0.12 to 0.59. Direct comparisons with placebo in RCT data (Figs. 2 and 5) showed filgotinib (200 mg), brepocitinib (10 mg, 30 mg, 60 mg) had a nonsignificant higher risk for MACE with RR range from 51.08 to 7.86; while upadacitinib (15 mg, 30 mg), deucravacitinib (6 mg, 12 mg) and tofacitinib (5 mg, 10 mg) had a nonsignificant lower risk of MACE with RR range from 0.01 to 0.17.

Indirect comparisons in NMA demonstrated no significant differences in MACE and TE risk among different JAKis, except that filgotinib (200 mg) in RCT data had a significant lower risk of TE compared with tofacitinib (5 mg, 10 mg) and upadacitinib (30 mg) with RR range from 868.06 to 4510.37, and brepocitinib (30 mg, 60 mg) in RCT data showed a significant higher risk of TE compared with upadacitinib (30 mg) with RR range from 1883.34 to 2028.25. (Fig. 2).

NMA Estimates of MACE and TE Risk Comparing JAKi with Adalimumab in NMA

Traditional pairwise meta-analyses of direct JAKi–adalimumab comparisons in combined RCT + LTE data (n = 8) (Appendix 8.1.2) or in RCT data (n = 4) (Appendix 8.1.4) provided similar estimates. Direct comparisons with adalimumab in combined RCT + LTE data (Figs. 3 and 4) showed there is a significant protective effect of MACE risk for upadacitinib, either with upadacitinib (15 mg, RR: 0.08, 95% CI [0.01, 0.47]) or with upadacitinib (30 mg, RR: 0.11, 95% CI [0.02, 0.52]), while a nonsignificant lower risk of MACE for tofacitinib (5 mg, RR: 0.15, 95% CI [0.02, 1.38]) and tofacitinib (10 mg, RR: 0.16, 95% CI [0.02, 1.52]). Direct comparisons with adalimumab in combined RCT data (Figs. 3 and 5) showed there is a nonsignificant protective effect of MACE risk for upadacitinib (15 mg, 30 mg) and tofacitinib (5 mg, 10 mg) with RR range from 0.12 to 0.1.

Regarding TE risk, traditional pairwise meta-analyses of direct JAKi–adalimumab comparisons in combined RCT + LTE data (n = 6) (Appendix 8.2.2) or in RCT data (n = 2) (Appendix 8.2.4) provided similar estimates. Direct comparisons with adalimumab in combined RCT + LTE data (Figs. 3 and 4) showed there is a nonsignificant protective effect of TE risk for upadacitinib (30 mg, RR: 0.85, 95% CI [0.00, 157.86]) and tofacitinib (10 mg, RR: 0.78, 95% CI [0.00, 184.99]), while a nonsignificant higher risk of TE for upadacitinib (15 mg, RR: 2.88, 95% CI [0.01, 587.68]) and tofacitinib (5 mg, RR: 1.05, 95% CI [0.00, 248.51]). Direct comparisons with adalimumab in combined RCT data (Figs. 3 and 4) showed there is a nonsignificant protective effect of MACE risk for upadacitinib (15 mg, 30 mg) and tofacitinib (5 mg, 10 mg) with RR range from 0.06 to 0.08.

Indirect comparisons in NMA showed no significant differences regarding MACE and TE risk between different comparisons (Fig. 3).

Sensitivity Analysis

Sensitivity analyses, which omitted individual studies one at a time, showed no substantial differences in estimates (Appendix 9). Additionally, random-effects meta-regression demonstrated no significant differences in the impact of age and sex between JAKis and comparators groups on the relative incidence of MACE or TE in either eligible RCTs or in combined RCT and LTE data (Appendix 10). The comparison-adjusted funnel plots in Appendix 11 indicated an asymmetry regarding TE risk in RCT + LTE data. Traditional pair-wise funnel plots were not suggestive of significant publication bias or small study effects (Appendix 12; Egger’s test of funnel asymmetry: p = 0.1886 for MACE and p = 0.9307 for TE in combined RCT and LTE data; p = 0.2568 for MACE and p = 0.4261 for TE in RCTs). Study heterogeneity was low for both traditional pairwise meta-analysis (I2 = 0% in Appendix 8) and NMA analysis (Appendix 13).

Proposed Evidence with GRADE Approach in this NMA

Details of the GRADE assessment are shown in Appendix 14–16. Our NMA conclusions (Appendix 17) were compared with the leading recommendations for JAKis use in PsA [12,13,14] (Table 1), resulting in the following recommendations: (1) For patients with PsA with a 10-year CVD risk < 10% or in case bDMARDs are not an appropriate choice, JAKis (tofacitinib and upadacitinib at both doses) can be recommended at the same level of bDMARDs and csDMARDs, which coincides with the latest proposal for JAKi use in PsA [45]; (2) For patients with PsA with multiple MACE risk factors, upadacitinib at a lower dosage is recommended as first choice for long-term treatment; (3) For patients with PsA with TE risk factors, investigational JAKis are not currently recommended, if alternative licensed JAKis or bDMARDs or csDMARDs are available.

Table 1 Comparison between evidence in our NMA with GRADE approach and current recommendations of different meta-analyses