Safety of BRAF+MEK Inhibitor Combinations: Severe Adverse Event Evaluation

Aim: The selective BRAF and MEK inhibitors (BRAFi+MEKi) have substantially improved the survival of melanoma patients with BRAF V600 mutations. However, BRAFi+MEKi can also cause severe or fatal outcomes. We aimed to identify and compare serious adverse events (sAEs) that are significantly associated with BRAFi+MEKi. Methods: In this pharmacovigilance study, we reviewed FDA Adverse Event Reporting System (FAERS) data in order to detect sAE reporting in patients treated with the combination therapies vemurafenib+cobimetinib (V+C), dabrafenib+trametinib (D+T) and encorafenib+binimetinib (E+B). We evaluated the disproportionate reporting of BRAFi+MEKi-associated sAEs. Significant associations were further analyzed to identify combination-specific safety signals among BRAFi+MEKi. Results: From January 2018 through June 2019, we identified 11,721 sAE reports in patients receiving BRAFi+MEKi. Comparison of BRAFi+MEKi combinations demonstrates that skin toxicities, including Stevens–Johnson syndrome, were disproportionally reported using V+C, with an age-adjusted reporting odds ratio (adj. ROR) of 3.4 (95%CI, 2.9–4.0), whereas fever was most significantly associated with D+T treatment with an adj. ROR of 1.9 (95%CI, 1.5–2.4). Significant associations using E+B treatment include peripheral neuropathies (adj. ROR 2.7; 95%CI, 1.2–6.1) and renal disorders (adj. ROR 4.1; 95%CI, 1.3–12.5). Notably, we found an increase in the proportion of Guillain–Barré syndrome reports (adj. ROR 8.5; 95%CI, 2.1–35.0) in patients administered E+B. Conclusion: BRAFi+MEKi combinations share a similar safety profile attributed to class effects, yet concomitantly, these combinations display distinctive effects that can dramatically impact patients’ health. Owing to the limitations of pharmacovigilance studies, some findings warrant further validation. However, the possibility of an increased risk for these events should be considered in patient care.


Study Design and Data Sources
The observational, retrospective, pharmacovigilance study is based on reports linked to suspect drugs in the FDA Adverse Events Reporting System (FAERS) [35,36]. The FAERS serves as the FDA's repository of safety reports submitted by patients, health care professionals and pharmaceutical companies, and currently contains over 18 million reports [37]. Since the FAERS is a publicly available and anonymized database, institutional review board approval and informed consent was waived.

Procedures
This study included serious events leading to death, life-threatening conditions, disability, or hospitalization from January 2018 to December 2019. Despite available reports before 2018, this period was used to compare the reports of three BRAFi+MEKi combinations during the same timeframe. Of note, reports of AEs associated with E+B started in 2018. Adverse events were classified according to the Medical Dictionary for Regular Activities (MedDRA), which organizes terms into a hierarchy [38]. Duplicate cases were removed and general AEs such as 'off-label use' were excluded. Adverse events specific to combination therapy in the analysis were those notified as suspected to be caused by both BRAFi and MEKi as primary or secondary suspects. Additional administrative information was extracted for each report, including patient characteristics (sex and age); country of origin; drugs (dosage regimen, start and end dates of administration); indication for the drug; reactions (onset date, outcome, and response to rechallenge and dechallenge).

Statistical Analysis
Disproportionality analysis using the 'case-control' approach allows assessing whether the reporting rate of AEs is differentially elevated relative to expected rates of reporting in a pharmacovigilance database [39,40]. Disproportionate reporting rates of AEs could be indicative of potential risks associated with a particular agent or drug combination. To compare treatment-induced toxicities among different BRAFi+MEKi regimens, we also performed a disproportionate analysis between the combinations as a case-noncase study (e.g., V+C versus D+T and E+B). Standard measures of disproportionality include the information component (IC) when comparing to all the records in the dataset and the reporting odds ratio (ROR) when comparing between different treatment subgroups [41,42]. Thus, the latter analysis (ROR) involves only patients with BRAFi+MEKi and the rest of the database is excluded. The IC, given by a Bayesian confidence propagation neural network [41], and ROR, a point estimate calculated as the observed odds ratio [43], are widely used and are currently employed by various reporting agencies, as well as by the World Health Organization (WHO) [41,44,45].
The decision rule for a potential signal generation was based on a positive IC value (>0), a minimum number of reactions of 3, and a threshold of 0.05 for the Bayesian version of the false discovery rate (FDR) [46]. This analysis was performed on high-level group terms of drug-specific adverse events versus the full database (Supplementary Table S1). To compare reactions between BRAFi+MEKi combinations that are also disproportionally reported in the full database, we used only AEs linked to the identified signals above. In the ROR analysis, the signal-generation method was based on mid-p-values, as proposed by Ahmed et al. [47], and a signal was considered significant if ROR > 1 and the corrected p-value for multiple testing using the Benjamini-Hochberg (BH) correction was <0.05 [48]. Only combinations with a minimum amount of 0.5% reports were selected, corresponding to five reports in the smallest combination regimen (E+B). Given the considerable overlap between symptoms for various causes as well as redundancies in reporting AEs, we considered events for further evaluation if they were also significant in associated MedDRA terms. Since relevant medical information such as cancer staging was absent and therefore could not be accounted for, death and neoplasm occurrences as AEs were excluded from subsequent analysis. The disproportionality analysis and the age-adjusted ROR analysis were performed using the PhViD [49] and epiR [50] packages, respectively, in the R statistical computing environment (v3.5.2) [51].

Patient Characteristics
The FAERS database included 11,721 reports of sAEs among patients receiving V+C (2345), D+T (8411) and E+B (965), corresponding to the time period (January 2018 to December 2019) ( Table 1). The total number of sAEs reported in the FAERS was 3,285,265. Patient characteristics, drug indications, and country of origin are shown in Table 1.

Toxicity Profile of BRAFi and MEKi Combinations
The reported frequencies of sAEs for the three BRAFi+MEKi combinations can be compared directly using ROR analysis. The detailed ROR estimates of V+C, D+T, and E+B for sAEs detected as signals are listed in Supplementary Table S2. To facilitate the analysis and interpretation of the data, the ROR for each identified sAE was compared: the respective shifts in BRAFi+MEKi ROR, expressed as an increase or decrease in a logarithmic scale, are displayed in Figure 1. ROR of dermatological side effects including rash, generalized rash, maculo-papular rash, erythema multiforme and rash with eosinophilia and systemic symptoms were considerably higher in V+C compared to D+T and E+B, whereas pyrexia and elevated C-reactive protein (CRP) were disproportionally higher in D+T compared to V+C and E+B. Furthermore, gastrointestinal events including colitis, diarrhea, nausea and upper abdominal pain and renal side effects including renal disorder/impairment/failure and increased creatinine were disproportionally higher in E+B compared to other BRAFi+MEKi combinations. Acute kidney injury (AKI) was higher in V+C, compared to E+B; however, both were considerably higher than D+T. Notably, Guillain-Barré syndrome (GBS) and seizures were significantly higher in E+B compared to V+C and D+T.

Discussion
Novel therapeutic strategies relying on the selective inhibition of the MAPK pathway have proven to significantly prolong the survival of patients with advanced BRAF-mutated melanoma [1][2][3][4][5][6][7]. However, almost all patients treated with BRAFi+MEKi combination therapy develop AEs, with grade 3-4 AEs occurring in most patients [19]. Since the combination therapy with D+T has been approved for adjuvant use in melanoma patients [7] and the indications for BRAFi+MEKi have been extended to other cancer types [52][53][54], the pool of patients is expanding, and thus more patients are expected to develop life-threatening events. As no specific BRAFi+MEKi combination has shown superiority over the rest, there are no evidence-based standard-of-care recommendations of which therapy to choose in the metastatic setting. Therefore, knowledge of the safety profile can be used to tailor the therapy to patients and diagnose adverse reactions early. To our knowledge, we report the most extensive characterization of sAEs associated with BRAFi+MEKi combinations so far, through analysis of the FAERS pharmacovigilance database.
The results show a significant incidence of skin disorders, fever, and GI conditions associated with BRAFi+MEKi combinations, corresponding to the recognized class effects of the treatment [1,13,19].
A comparison between the combinations shows that skin toxicities and fever are overrepresented in V+C and D+T, respectively, while GI conditions were reported significantly less using D+T. These findings are consistent with safety data from phase III clinical studies [1][2][3][4][5][6]19], and corroborate our disproportionality analysis.
Confirmatory studies are necessary to determine therapeutic activity but are limited in providing an accurate depiction of the safety profile, as the conditions under which medicines are used post-marketing may not be reflected in clinical trials, and they also lack sufficient power to detect infrequent or rare events [55,56]. Pharmacovigilance surveillance thus remains the cornerstone for identifying drug-related complications in spite of recognized drawbacks [57]. Herein, we report renal toxicities, neurological disorders and hypotension as additional possible safety considerations for clinicians involved in the care of patients treated with BRAFi+MEKi therapy, and specifically with E+B combination. Data on renal toxicities from V+C and D+T are relatively sparse [2,4,26,58,59] and the association with E+B is even less clear [19,60,61]. We show that nephrotoxicity associated with E+B is significantly more disproportionate among BRAFi+MEKi and portends poor outcomes. Serious renal disorders can occur as early as 17.5 days (IQR 5-54) after initial exposure to E+B. Although the incidence of these renal toxicities cannot be established with the FAERS, some data are available. In the COLUMBUS study, 3% of grade 3-4 AEs encompass renal toxicities [5], which is similar to the proportion reflected in our results of 3.5% (n = 34). Of note, despite having longer follow-up durations, only 1% (grade 3-4 AE) and 0% (any grade AE) of the events were reported in phase III studies for V+C [4] and D+T [62], respectively. This pattern agrees with our findings and provides further support for our results. Furthermore, our report includes the identification of numerous cases with serious peripheral neuropathies and seizures associated with BRAFi+MEKi, which is, to our knowledge, the largest collection of such reports to date. Compared to the full database and BRAFi+MEKi combinations, E+B was associated with a notable reporting odds ratio of peripheral neuropathies and specifically GBS, representing 0.5% of the reports. Nonetheless, previous treatment with immune checkpoint inhibitors (ICIs) cannot be ruled out as a possible contributor or the source for GBS. Several lines of evidence support a major role of MAPK inhibitors as well as encorafenib in the pathophysiology of neurological disorders and GBS. First, the MAPK pathway mediates the dedifferentiation of Schwann cells in the presence of normal axonal signaling and required during the repair process following nerve injury [63]. In addition, the MAPK signaling pathway was found to be significantly enriched in overexpressed genes among patients with GBS [64]. Second, increasing number of case reports and case series were published in the literature, implicating BRAFi or MEKi in neuropathy [65][66][67][68][69][70][71] as well as GBS [69,72,73]. Third, a relatively frequent drug-related AE occurring with encorafenib monotherapy was transient Bell's palsy [5,18], a mononeuritic variant of GBS, in most cases [74]. This reaction appeared in 8% of the patients and 11% in the dose expansion phase using encorafenib, but has rarely been reported in association with other BRAFi [18]. Facial paresis has also been reported using the combination E+B, but was considerably decreased [5], which can be explained by the suppression of the paradoxical activation of the MAPK pathway [24].
Few limitations of disproportionality analysis need to be recognized. Pharmacovigilance studies have complementary strengths in signal detection that require further validation, optimally in prospective trials. Replication of safety issues in other continental reporting agencies such as the WHO pharmacovigilance database and European Medicine Agency is biased by overlapping safety reports between databases [57]. Further, not all sAEs are submitted to the FAERS, and thus, some reports could be underreported or misrepresented. However, the FAERS provides a key advantage in signal detection, especially of less frequent events, such as GBS, due to the extensive collection of reports from 190 countries, according to the dataset utilized in this study.
One of the major limitations of the analysis involves the inability to quantify the incidence of identified risks, as the total number of patients exposed to a given product is unknown [75]. Further, the background reporting rate could be inflated by an event associated with other drugs, thereby masking the detection of a suspected drug [76]. This phenomenon tends to reduce the sensitivity of some events [77]. Inherent weakness involving the FAERS as well as other spontaneous reporting systems is the inability to verify the clinical findings to justify the reported diagnosis and access relevant data, such as co-morbid illness. Further, a lack of accurate information regarding the line of treatment limits the interpretation of the results. Previous treatment with ICIs could have contributed to the observed toxicities, such as hyperthyroidism which is a common immune-related AE [78]. Furthermore, inconsistent reporting results in missing information regarding demographics, time to onset of AE, drug dosing, dechallenge, rechallenge and comedications. Yet, many successful efforts exist, showing the value of disproportionality to detect safety signals after approval, including examples as early as in the 1980s, where a significant association was found between valproic acid use during pregnancy and spina bifida aperta [79]. Nonetheless, a risk still exists that disproportionality analysis of post-marking data could be misleading. An example of another validated association involves the identification of cefaclor to induce serum sickness-like syndrome [80] and the association between rofecoxib and thrombotic reactions [81]. This relationship, among others, which could have been predicted soon after marketing and long before withdrawal [81], demonstrates the ability to find signals quickly after drug approval [82,83]. Although causality cannot be established conclusively based on post-marketing data alone, it is often served as a basis for adjudicating relationships with sufficient certainty in pharmacovigilance settings [84].
In conclusion, our study shows that beyond skin disorders and fever associated with V+C and D+T, respectively, BRAFi+MEKi therapy can cause other combination-specific toxicities that include gastrointestinal, renal and neurological disorders associated with E+B. Most of the serious adverse reactions commence soon after administration and resolve after dechallenge. These complications should be considered in therapeutic decision making, and careful monitoring is recommended.
Author Contributions: N.A. and T.M. conceived the idea. All authors contributed to the design of the study. T.M. acquired the data for the study, performed the analyses, and drafted the report. G.M. supervised the study. All authors interpreted the results and contributed to the writing of the final report. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by the Ella Lemelbaum Institute Funds, by the Samueli Foundation Grant for Integrative Immuno-Oncology and by the Israel Science Foundation Personalized Medicine Grant 3495/19. Tomer Meirson was supported by the Foulkes Foundation fellowship for MD/PhD students. The funding sources had no role in the study design, collection, analysis and interpretation of data.

Conflicts of Interest:
The authors declare no conflict of interest.