Analysis of the association between Janus kinase inhibitors and malignant skin tumors using the Food and Drug Administration Adverse Event Reporting System

Malignant skin tumors are adverse events of concern regarding Janus kinase (JAK) inhibitors. This study aimed to evaluate the association between JAK inhibitors and adverse events of malignant skin tumors, and to characterize the main features. Data (2012–2021) were collected using the US Food and Drug Administration Adverse Event Reporting System (FAERS). Adverse event cases of JAK inhibitors as the primary suspected drug were extracted for further analysis. Disproportionality analysis evaluated the association between JAK inhibitors and malignant skin tumor events by estimating the reporting odds ratio (ROR) and the information component (IC) with 95% confidence intervals (95% CI). A total of 142,673 cases with JAK inhibitors as a primary suspected drug were collected, including 1400 malignant skin tumor events. Ruxolitinib, upadacitinib, tofacitinib, and baricitinib were included in the disproportionality analysis. Three JAK inhibitors were associated with malignant skin tumor events, namely ruxolitinib (ROR 5.40, 95% CI 5.03–5.81; IC 2.39, 95% CI 2.14–2.62), upadacitinib (ROR 4.79, 95% CI 4.03–5.71; IC 2.24, 95% CI 1.62–2.77), and tofacitinib (ROR 1.67, 95% CI 1.53–1.83; IC 0.73, 95% CI 0.43–1.02). The median time to onset time was 378.5 days. We found association between malignant skin tumors and ruxolitinib, upadacitinib, and tofacitinib. More attention should be paid to these events when prescribing JAK inhibitors in clinical practice.


Introduction
Janus kinase (JAK) inhibitors are a new class of drugs for treating various conditions including rheumatoid arthritis and myeloproliferative neoplasms.They inhibit the Janus kinase/signal transducers and are activators of the transcription pathway (JAK/STAT), which are coupled to receptors for ligands that regulate immune responses and the bone marrow [1][2][3].The pathway is critical in orchestrating multiple biological functions [4].Several JAK inhibitors have been approved and are widely used in clinical practice.More JAK inhibitors are being developed.However, inhibition of JAK can induce unexpected adverse effects, such as malignant skin tumors [5][6][7].
Malignant skin tumors are characterized by abnormal proliferation and differentiation of skin tissues.The Global Burden of Disease 2019 reported that the incidence and mortality of malignant skin tumors were 0.12% and 0.21%, respectively.In addition to ultraviolet exposure, immunosuppressive drugs have been identified as a significant risk factor [8]. JAK inhibitors can alter the functions of the JAK/STAT pathway, preventing the human immune system from destroying cancer cells [9].The US Food and Drug Administration (FDA) has issued a black box warning for malignant adverse events (AEs) of upadacitinib, tofacitinib, and baricitinib.From March 2014 to July 2020, the FDA conducted a prospective head-to-head safety trial comparing tofacitinib with tumor necrosis factor inhibitors [10].The results showed a higher incidence of skin malignancies with both doses of tofacitinib (5 mg or 10 mg) [11].Ruxolitinib could increase the risk of basal and squamous cell carcinoma, and the high rate might be related to the long duration of follow-up [12].Several case studies have shown invasive events of malignant skin tumors in patients treated with JAK inhibitors, and long-term studies are expected to confirm the association [13][14][15][16][17].
When considering the safety of a new treatment, it is necessary to put clinical trial data into context.The incidence of absolute AEs is generally lower in clinical trials than in the real-world cohort.The FDA Adverse Event Reporting System (FAERS) is widely used due to its large amount of data and free public access.Data mining based on FAERS is effective in the early detection of safety issues in newly marketed drugs and the continuous monitoring of safety issues in older medications, even for non-serious AEs that typically receive little attention.Long-term pharmacovigilance studies of JAK inhibitors are still scarce.

Aim
This study aimed to evaluate the association between JAK inhibitors and adverse events of malignant skin tumors and to characterize the main features of drug-related malignant skin tumors.

Ethics approval
The FAERS database contains anonymized patient information hence no ethical approval was required.

Data source
This was a retrospective observational pharmacovigilance case non-case study.We downloaded 40 quarters of FAERS data (2012-2021) into a local database.FAERS is a post-marketing database containing AEs, medication error reports, and product quality complaints.FAERS collects global AEs spontaneously reported by pharmaceutical companies, healthcare professionals, and consumers.Data are uploaded quarterly and are available for free download.FAERS data are stored in seven tables: DEMO (patient demographic and administrative information), DRUG (drug information), REAC (adverse event information), OUTC (patient outcomes information), THER (drug therapy date information), INDI (drug indication), and PRSR (report source).

Target drugs and identification of malignant skin tumor event
Figure 1 shows the flow chart for identifying AE cases and malignant skin tumor events.Duplicate reports were removed from the DEMO table according to the FAERS instruction file [37].If the CASEIDs (a number used to identify a FAERS case) were the same, the latest FDA_DT (the date the FDA received the case) was selected.If the CASEID and FDA_DT were the same, the higher PRIMA-RYID (a unique number for identifying a FAERS report) was selected.
The remaining tables were PRIMAYID mapped with the deduplicated DEMO table.Drug names in the DRUG table were standardized using MedEx1.3.8 and assigned codes and common drug names [38].The AE data in FEARS are coded using the Preferred Terms (PTs) of the Medical Dictionary for Regulatory Activities (MedDRA) [39].A narrow search of the Standardized MedDRA Query (SMQ) was used to identify malignant skin tumor events in the REAC table (SMQ code: 20, 000, 204), including 102 PTs (Supplemental Table S1).
JAK inhibitors (ruxolitinib, upadacitinib, tofacitinib, baricitinib, decernotinib, filgotinib, momelotinib, and peficitinib) were screened in the standardized DRUG list as target drugs.JAK inhibitors reported as the primary suspected drug were included, and the corresponding reporting code (PRIMARYID) and the medical record code (CASEID) were obtained.Relevant adverse events were screened, and malignant skin tumor events were identified.

Serious adverse events
Serious adverse events were death, hospitalization, disability, life-threatening or other serious medical events.These outcomes are coded in FAERS.The numbers and proportions of specific serious events were calculated.

Data mining
The local FAERS data sets were managed using Microsoft SQL Server (2017).Characteristics of AE cases were collected, including age, sex, reporter, reporting country, reporting year, and case outcome.Disproportionality analysis of reporting odds ratio (ROR) and the information component (IC) was used to detect the association (signal) between malignant skin tumor events and JAK inhibitors [40].The ROR is one of the measures used in disproportionality analysis, while IC is a procedure that employs Bayesian theory to transform the raw values of the relative reporting ratio.For ROR, a signal is detected if the case number ≥ 3 and the lower limit of the two-sided 95% confidence interval (95% CI) > 1.The higher the ROR value, the stronger the association.For IC, the signal is detected if the lower limit of 95% CI > 0. This study determined the drug-AE association when both the frequency and Bayesian methods detected signals simultaneously.Supplemental Table S2 shows the specific algorithms for ROR and IC.
We excluded cases reporting the drugs with known risks of malignant skin tumors, including hydroxyurea, methotrexate, cyclosporine, etanercept, infliximab, and adalimumab, to verify the robustness of the analysis.As an indication for JAK inhibitors, patients with rheumatoid arthritis with an increased risk of malignancy were also excluded.To verify the feasibility of the data and data mining methods, we selected voriconazole as a positive control drug, as the association between voriconazole and the increased risk of malignant skin tumors has been well described in research and the drug label [41].Furthermore, a stratified sensitivity disproportionality analysis was used to assess the potential role of age.

Statistical analysis
Descriptive analysis was applied to summarize and present patient clinical characteristics.Different JAK inhibitors' onset times and prognosis of malignant skin tumor events were compared using Pearson's chi-square test.p < 0.01 was considered statistically significant.SPSS 23.0 and Microsoft Excel 2016 were used for statistical analysis.

General characteristics
The following four JAK inhibitors were analyzed: ruxolitinib, upadacitinib, tofacitinib, and baricitinib.Four JAK inhibitors (decernotinib, filgotinib, momelotinib, and peficitinib) had limited AE cases and were not included.As shown in Fig. 2, malignant skin tumor events were further classified based on PTs as skin cancer (40.8%), basal-cell carcinoma (23.4%), malignant melanoma (11.5%), skin squamous-cell carcinoma (10.4%), and skin neuroendocrine Table 1 shows the characteristics of the cases.More male than female patients (1.0% vs. 0.7%) were reported, and the difference was statistically significant.The highest proportion of malignant skin tumor events was in patients over 65 (25.3%).The percentages of AEs reported by healthcare professionals and non-healthcare professionals were similar (49.6% vs. 49.8%).USA (87.0%) reported the highest number of cases.Reports of malignant skin tumor reports due to JAK inhibitors increased, with more than 85% of cases reported within the last five years.
Table 3 shows the association between JAK inhibitors and malignant skin tumor events at the PT level.S7 shows the association between voriconazole and malignant skin tumor events at the SMQ and PT level.The remaining sensitivity analysis is shown in Supplemental Table S8.

Time to the onset of malignant skin tumor events
The onset time of malignant skin tumor events after JAK inhibitor treatment was compared.After removing unreported, repeated, and incorrect time interval reports, 242 cases were obtained.The median time to the onset of malignant skin tumor events was 378.5 days, with an interquartile range (IQR) of 59.8 to 3924.0 days.The median onset time was 405.0 days (IQR 60.5-845.0)for ruxolitinib, 315.5 days (IQR 167.3-463.8)for baricitinib, 525.5 days (IQR 250.8-997.5)for tofacitinib, and 64.0 days (IQR 29.5-190.5)for upadacitinib.Upadacitinib-treated patients had a significantly shorter onset time than those receiving ruxolitinib (P = 0.010) and tofacitinib (P < 0.01).However, no significant differences were observed in the onset time between patients receiving ruxolitinib and tofacitinib (P = 0.082).
To further explore the safety of JAK inhibitors, the onset time of all AEs of JAK inhibitors was calculated.More than 60% of the AE cases occurred within the first ten weeks of treatment (Fig. 3).Additionally, more than

Prognosis analysis
The fatality outcome of malignant skin tumor cases associated with JAK inhibitors was analyzed.It is important to note that the results reflect the proportion of deaths relative to the number of patients with AEs, not the entire population of JAK inhibitor users.The fatality proportion among all patients who experienced malignant skin tumor events was 4.4%.Ruxolitinib had the highest fatality proportion at 6.8% (51/755), followed by tofacitinib at 1.9% (9/470).
Regarding serious adverse events, 20.1% (275/1,366) of patients treated with JAK inhibitors experienced them.The proportion of serious adverse events associated with baricitinib was lower than that of ruxolitinib (13.3% vs. 22.8%) but higher than that of upadacitinib (13.3% vs. 7.9%).These differences were statistically significant, as confirmed by pairwise comparisons using the chi-square test.

Discussion
This study presented a comprehensive and systematic pharmacovigilance analysis of the association between JAK inhibitors and malignant skin tumor events based on real-world FAERS data.JAK inhibitors were selected as the primary suspect drug, and both ROR and IC were used to analyze the association of the drug with malignant skin tumor events.The residual confounders were considered.A sensitivity analysis was performed to rule out possible interference caused by combined drug use and indications.Furthermore, the investigation delved beyond the package insert labeled instructions to thoroughly analyze prescriptions.Differences in onset time and prognosis of malignant skin tumors among JAK inhibitors were also assessed.Drug-related malignant skin tumor events primarily affected males (1.0% vs. 0.7%, P < 0.001), although women were more likely to take JAK inhibitors.The result was comparable to the epidemiological characteristics of skin cancers [8], which may be explained by the increased exposure of men to ultraviolet radiation [42].Furthermore, malignant skin tumor events were reported mainly in the ≥ 65-year-old group.The results were consistent with the findings of age-related drug toxicity studies and malignant skin tumor characteristics [42][43][44].Furthermore, the number of malignant skin tumor events related to JAK inhibitors increased significantly between 2017 and 2021.The following three factors could explain this increase: 1) malignant skin tumors are adverse effects that occur only after a more extended period, while baricitinib was not approved until 2018 and upadacitinib was not approved until 2019, 2) the presence of notorious bias: an increase in malignant skin tumor reporting following the FDA warnings in 2021.However, a study showed no over-reporting of AEs due to the notorious bias in FAERS [45], and 3) JAK inhibitors are sometimes used off-label.
Three JAK inhibitors, ruxolitinib, upadacitinib, and tofacitinib, were significantly associated with malignant skin tumor events.Ruxolitinib had the highest number of reports and the strongest signal.Previous studies suggested that ruxolitinib could alter the immune microenvironment and potentially facilitate skin carcinogenesis [12].However, the stronger association of ruxolitinib could also be explained by other inherent mechanisms that could lead to skin tumors in patients with malignancies since ruxolitinib has been used primarily in myelodysplastic neoplasms.Furthermore, ruxolitinib was strongly associated with skin neuroendocrine carcinoma.The study also found that upadacitinib was associated with malignant melanoma, which had a high mortality rate and should be taken seriously.Baricitinib did not detect a signal, consistent with the findings of a comprehensive database analysis [46].However, due to the small number of reports of baricitinib, signal underestimate was possible.Concomitant drugs did not impact the results; only ruxolitinib detected a signal after excluding patients with rheumatoid arthritis, which may have resulted from the fact that patients with rheumatoid arthritis had an increased risk of malignancy [47].
The time interval between JAK inhibitor treatment and the malignant skin tumor events varied depending on the drugs.Upadacitinib had a significantly shorter time to onset than other inhibitors.Because upadacitinib is an approved drug for atopic dermatitis, which presents as similar to skin cancer, it can sometimes be difficult to tell if skin cancer is drug-related or a misjudgment.This should be considered in the reported cases.Tofacitinib and ruxolitinib had the longest time to onset.In the collected case studies of JAK inhibitors and malignant skin tumor events [13][14][15][16][17], the onset time ranged from one month to five years, with a median of 365.0 days, slightly shorter than in our study.Our study had a longer observation period and allowed complete data analysis of the safety of JAK inhibitors.The results showed that some AEs, such as malignancy, take a long time to develop.Most AEs occurred within the first ten weeks of treatment, but the proportion increased again after two years of treatment.
As the most commonly used immune inhibitors, JAK inhibitors have shown great potential in clinical practice.For example, ruxolitinib may reduce atopic dermatitis, psoriasis, and vitiligo [48].Of the 142,673 reports we collected, 5,835 (4.09%) reported off-label uses.The off-label use of JAK inhibitors did not increase the proportion of malignant skin tumor events (P = 0.634) in our analysis.Despite the support of clinical trials, the benefit/risk balance should be thoroughly evaluated before making such clinical decisions.
The mechanism by which JAK inhibitors may increase malignant skin tumor risk is not clear.JAK inhibitors are drugs that inhibit the JAK-STAT pathway, which is an important mediator of interleukin and interferon signaling [4].While proper functioning of the JAK/STAT pathway results in pro-inflammatory responses and cell growth and proliferation, improper functioning can result in diseases of the immune system, as well as cancers [49,50].Studies have suggested that JAK inhibitors can affect various cytokines (for example, IFNγ) and cell types (such as NK cells), thus disrupting the "cancer immunoediting" process by which the immune system destroys cancer cells within the body to induce malignant skin tumors [9].
Our study has several limitations.First, FAERS is a spontaneous reporting system with unavoidable biases, such as under-reporting, over-reporting, or missing data.Second, we could not assess the causal relationships because the study Fig. 3 Constituent ratios of time to adverse events for the four JAK inhibitors design is not suitable for assessing the temporal relationships between exposure (JAK inhibitors) and adverse events (malignant skin tumors).Assessing causality requires a prospective study design.Third, the analysis of deaths was limited because ruxolitinib is approved for myeloproliferative neoplasms with high mortality.Finally, we could not assess the causal relationship between malignant skin tumors and JAK inhibitors based on spontaneous reporting data.

Conclusion
This study suggested an association between JAK inhibitors and malignant skin tumors.Ruxolitinib had the strongest association, particularly neuroendocrine skin carcinoma, followed by upadacitinib.Tofacitinib had the lowest association.More clinical studies should verify the causal relationship.Patients at risk for skin cancer should have regular skin examinations while taking JAK inhibitors.

Fig. 1
Fig. 1 Flow chart for identifying malignant skin tumor cases related to JAK inhibitors from the FAERS database.MST: malignant skin tumors, FAERS: FDA Adverse Event Reporting System

Table 1
Characteristics of cases reported with JAK inhibitors as the primary suspect drug from the FAERS database

Table 2
Signal detection between JAK inhibitors and malignant skin tumor events at the SMQ level MST malignant skin tumors; ROR reporting odds ratio;95% CI 95% confidence interval; IC information component

Table 3
Signal detection between JAK inhibitors and malignant skin tumor events at the PTs level MST malignant skin tumors; ROR reporting odds ratio; 95% CI 95% confidence interval; IC information component