JAK Safety Debate Continues: Stronger VTE signal, less clear cancer signal Save

Few topics in rheumatology have reshaped prescribing behaviour as dramatically as the safety concerns surrounding JAK inhibitors following ORAL Surveillance and the subsequent FDA and EMA warnings. Even now, years later, the field remains divided over how much of the observed risk reflects real-world practice.

At EULAR 2026, two major registry studies revisit perhaps the two most controversial aspects of the JAK story: venous thromboembolism (VTE) and malignancy. Raffray and colleagues present updated nationwide comparative safety data examining VTE risk across biologic and targeted therapies in RA (OP085). Using the Swedish Rheumatology Quality Register and linked national health datasets, the investigators have analysed more than 26,000 treatment initiations between 2017 and 2024. Patients treated with JAK inhibitors experienced nearly double the adjusted risk of VTE compared with those receiving TNF inhibitors, with an adjusted hazard ratio of 1.89. Absolute rates remain relatively modest, with 8.3 events per 1000 person-years with JAK inhibitors versus 4.3 with TNF inhibitors. However, the consistency of the signal is increasingly notable.

Perhaps the most important aspect of the study was that the elevated VTE risk persisted even when analysis was restricted to treatment starts from 2021 onward, after regulatory warnings had already influenced prescribing behaviour. One common criticism of observational studies has been that clinicians may preferentially avoid JAK inhibitors in high-risk patients after ORAL Surveillance, thereby distorting comparisons through channelling bias. Yet despite this likely shift in prescribing patterns, the VTE signal remains.

Interestingly, the Swedish group also observed a possible increase in VTE risk with IL-6 inhibitors compared with TNF inhibitors, a finding that will likely generate discussion given the relative lack of attention this class has received in thrombosis debates. Whether this reflects a true biological signal or residual confounding remains uncertain.

Turning to malignancy, in the multinational JAK-pot collaboration, Ciribè and colleagues pooled data from 14 European and Canadian RA registries encompassing more than 60,000 treatment courses to examine subtype-specific cancer risk with JAK inhibitors, TNF inhibitors, and biologics with other mechanisms of action (OP086).

Numerically, overall cancer rates excluding non-melanoma skin cancer were somewhat higher with JAK inhibitors than TNF inhibitors. However, after adjustment for confounding factors, the increase was not statistically significant. More importantly, no significant increase emerged for any individual cancer subtype, including lung cancer, breast cancer, colorectal cancer, prostate cancer, genital cancers, haematologic malignancies, or non-melanoma skin cancer.

This matters because ORAL Surveillance generated particular concern around lung cancer and lymphoma, especially in older smokers. Yet in this large real-world dataset, there was no clear replication of this result.

The study did, however, identify higher adjusted cancer rates among patients receiving non-TNF biologics such as rituximab, abatacept, and IL-6 inhibitors. Interpreting this finding is challenging. These medications are often preferentially used in patients with prior malignancy, smoking history, greater frailty, or more complex comorbidity profiles, making confounding by indication difficult to fully eliminate even with sophisticated statistical adjustment.

Taken together, these studies reinforce the idea that the JAK safety conversation is evolving into something more sophisticated than a simple question of whether these drugs are “safe” or “unsafe.”

The VTE signal now appears increasingly credible across multiple healthcare systems and datasets. At the same time, the absolute event rates remain relatively low, and we still lack clarity around whether risks differ substantially between individual JAK inhibitors, dosing strategies, or patient subgroups.

Cancer risk remains much harder to define. Large observational datasets continue to produce mixed or neutral findings, in contrast to the stronger signal seen in ORAL Surveillance. It remains possible that malignancy risk is concentrated within specific high-risk populations rather than representing a broad class effect across all patients with RA.

For clinicians, the practical message is increasingly one of individualised risk assessment. Age, smoking history, prior VTE, cardiovascular disease, prior malignancy, and overall comorbidity burden likely matter more than any single drug label. Equally important is remembering that RA itself confers increased baseline risks for thrombosis and malignancy, complicating attempts to disentangle treatment effect from disease biology.

The debate surrounding JAK inhibitors is far from settled. These EULAR data add to our understanding of comparative risk, and another reminder of the need for careful patient selection and informed shared decision-making.