In this pharmacovigilance study, JAK inhibitors were most commonly associated with infectious adverse events, embolism and thrombosis, neoplasms and gastrointestinal perforation events. We also identified significant increase in adverse event reporting regarding musculoskeletal and connective tissue disorders. Finally, we found no association with major cardiovascular events.
In our study, infections were frequently reported for JAK inhibitors, as was expected according to safety data from clinical trials24,30. We found a significant increase in reporting compared with the full database for some microorganisms (viral [herpes and influenza]fungal, and mycobacterial infectious disorders) and two main organ locations (respiratory and urinary tract infections).
Herpes zoster has been identified as a complication of JAK inhibitors in clinical trials25,26,31,32 and in a pharmacovigilance study of adverse events reported from the United Statestwenty two. Of note, in our study, herpes viral infections (MedDRA HLT) also include herpes simplex virus. Herpes zoster induced most of the treatment discontinuation due to infections in some clinical trials25,26but few data are available for herpes simplex infections. We observed over-reported herpes viral infections, the highest level for baricitinib, then tofacitinib, then ruxolitinib. Associated risk factors that can affect herpes zoster and herpes simplex infections for patients receiving JAK inhibitors include age, glucocorticoid exposure25other combined therapy, and underlying immunologic dysregulation. For example, herpes zoster/simplex infections were more frequently reported in a pooled safety data analysis of baricitinib in atopic dermatitis than in rheumatoid arthritis33. Atopic dermatitis is known to be associated with herpes simplex infections, with a severe form called eczema herpeticum34. In our study, the increased risk of herpes viral infections with baricitinib versus the two other JAK inhibitors could be explained by the underlying disorder because the main indication (80%) was rheumatoid arthritis. The risk associated with ruxolitinib was difficult to assess because most patients with haematopoietic neoplasms could have received prophylactic valaciclovir.
Recent concerns about JAK inhibitors involved in embolism and thrombosis17,18,19,20,21,22. Although the initial beneficial effect of ruxolitinib for risk of thrombosis was assessed in patients with polycythemia vera and myelofibrosis35lack of evidence remains for this beneficial association. Regarding tofacitinib, in the meta-analyses including 12,410 tofacitinib-exposed patients from completed studies, the incidence rate of venous thromboembolism events was 0.25 (95% CI 0.19–0.33). In our study, We found significant disproportionality analysis results for embolism and thrombosis with the first three approved JAK inhibitors. Over-reported “embolism and thrombosis” adverse events were ranked the highest for baricitinib, then ruxolitinib, then tofacitinib.
These comparisons must be interpreted with caution. Indeed, we did not consider patient characteristics, risk factors for thromboembolism or dose and duration of treatments. In the meta-analysis of clinical trials of tofacitinib, patients with than without baseline cardiovascular risk factors were more likely to experience thromboembolic events20. Risk factors were age ≥ 50 years and with at least one criterion (current smoker, high-density lipoprotein level < 40 mg/dL, history of hypertension, diabetes, myocardial infarction or coronary heart disease). Incidence rates in patients without risk factors were very low and most patients who experienced thromboembolic events also had multiple cardiovascular risk factors at baseline. Similarly, all patients with thromboembolic events in a pooled analysis of clinical trials of baricitinib had multiple risk factors25. Therefore, the treatment must be adapted to the individual risk.
Regarding neoplasms, we found increased frequency of neoplasm reports and identified “skin neoplasms malignant and unspecified” as significant. The three JAK inhibitors were associated with increased frequency of “skin neoplasms malignant and unspecified”. This is an important finding because previous cohort studies of patients with rheumatoid arthritis did not find a difference between tofacitinib and biologic disease-modifying anti-rheumatic drugs in risk of non-melanoma skin cancer (adjusted hazard ratio 1.04 [95% CI 0.68–1.61])36. Rheumatoid arthritis is associated with increased risk of melanoma and non-melanoma skin cancer regardless of the exposure37,38. Thus, the increased frequency of “skin neoplasms malignant and unspecified” for ruxolitinib leads to a discussion of a class effect of the JAK inhibitor. “Respiratory and mediastinal neoplasms malignant” was frequently reported for all three JAK inhibitors. This finding confirmed the recent warning from Pfizer for tofacitinib39. Indeed, in this warning, the incidence rate of malignancies excluding non-melanoma skin cancer was 1.13 (95% CI 0.94–1.35), with lung cancer as the leading cancer. As for skin neoplasms, this signal concerned all three JAK inhibitors. Lastly, we found a significant increase in reporting for “leukaemias”, in particular for ruxolitinib, which is probably related to the underlying disease.
Some studies have concluded similar incidence rates of malignancies for patients receiving tofacitinib or baricitinib as for those receiving other drugs40 and for non-melanoma skin cancer41 or malignancies excluding non-melanoma cancer13,14,25. ‘Cancer immunoediting’, the process whereby the human immune system destroys cancer cells within the body, is thought to rely upon a variety of cytokines (for example, IFNγ) and cell types (such as NK cells) that could be affected by JAK inhibition42. Decrease NK cells could predispose to develop malignancies among patients treated by JAK inhibitors but this effect remains unclear30.
Exposure time within trials is relatively limited, and even if pharmacovigilance studies bring interesting data, longer follow-up is needed to further assess malignancy risk and to compare JAK inhibitors with each other.
In our study, we observed increased frequency of gastrointestinal perforations with the three JAK inhibitors. Few cases of gastrointestinal perforation have been reported for patients participating in clinical trials of baricitinib and tofacitinib or those covered by US Medicare/Marketscan10,23,24,25,26. These few cases were described only among patients with rheumatoid arthritis. To our knowledge, only one clinical trial of ruxolitinib for myelofibrosis reported such an event causing death in a patient in the placebo group11,43. In our study, gastrointestinal perforation was over-reported only with tofacitinib. However, cases were also reported for the other two JAK inhibitors. These adverse events would be more frequent for patients with inflammatory bowel diseases. Treatments other than JAK inhibitors such as non -steroidal anti-inflammatory drugs are associated with increased risk of gastrointestinal perforation, which is important to consider with JAK inhibitors.
Finally, the percentage of fatal cases resulted much higher for ruxolitinib than for other JAK inhibitors. We did not perform a detailed analysis and clinical review of the 7000 fatal cases. However, plausible explanation regarding the percentage for ruxolitinib relies on patient characteristics and indications.
Limitations of this study include under-reporting of events and few verifications of the clinical and laboratory tests or radiological findings leading to the diagnosis of the adverse events. Moreover, spontaneous reporting cannot be used to estimate prevalence or incidence of adverse events among patients exposed to drugs. A lack of case-causality constitutes also a main limit. Indeed, individual case safety reports are not fully reliable regarding causal association, due to lack of other potential causes described than suspected drug, and to missing data about time to onset of the adverse event. In this study, we first analyzed the more general level of the MedDRA hierarchy to retain groups of adverse events we further detailed. With this method, we missed potential signals which could be significant at deeper level of the MedDRA hierarchy. Finally, we did not perform subgroup analyses by duration or patient characteristics, which are not available for all ICSRs, but contribute to th e occurrence of adverse events. Despite these limitations, pharmacovigilance analyses enable the detection of safety signals. VigiBase relies on data provided by more than 130 countries and enhances the identification of adverse events. Disproportionality analysis is a suitable method to compare spontaneous notifications of groups of drugs with other drugs while avoiding the effect of the extent of use of the product and nature of the adverse events.