P-386 Novel data visualization and unsupervised machine learning techniques to support optimal management of toxicity profiles of encorafenib plus cetuximab in patients with BRAFV600E-mutant metastatic colorectal cancer
Annals of Oncology(2023)
摘要
The BRAF inhibitor encorafenib, in combination with cetuximab (EC), has recently been approved for patients with BRAFV600E-mutated metastatic colorectal cancer (BRAFV600E-mCRC) after prior systemic therapy. As the combination is the first BRAF inhibitor-based regimen approved for BRAFV600E-mCRC the goal of this analysis was to explore novel and distinctive data visualization outputs to enable a potentially more practical interpretation and prediction of the safety profile of encorafenib + cetuximab and thus better management of adverse events. Adverse Events (AE) of interest (AEI’s), including dermatological AEs, arthralgia/myalgia, nausea/vomiting, diarrhea, abdominal pain, fatigue/asthenia, and nephrotoxicity, were previously defined, examined and published from the EC group of the multicenter, randomized, phase 3 BEACON study. The presence of AE, time to AE onset and time to AE resolution were investigated and their association with clinical characteristics (age, gender) were explored using data visualization including stacked bars, multistate models’ visualization, bubble timeline analysis, boxplots, heatmap graphs and infographics based unsupervised machine learning results. Representation with a bubble timeline analysis provides visual advantages to ease interpretation, within the same graph, the frequency of AEI occurrence, the grades of AEI, time to onset and/or time to resolution, depending on each AEI category. This aids a global overview and linkage patterns of safety and timelines. This analysis was conducted on 216 patients in the BEACON study treated with EC. The most commonly occurring AEI were dermatological toxicity (75.5%), arthralgia/myalgia (56.0%) and fatigue/asthenia (56.0%) with most AEI being Grade 1 or 2. The probability of experiencing AEI was higher in women than in men, especially for arthralgia/myalgia (W:65,4%; M: 47,3%), abdominal pain (W: 43,3%; M: 26,8%), and nausea/vomiting (W: 55,8%; M: 38,4%). However, the observed median onset time was 2 weeks earlier for men for abdominal pain and arthralgia/myalgia. Bubble timeline analysis highlight dermatological toxicities, nausea/vomiting and fatigue occur sooner than other AEI types (respective median onset time (in weeks): 2.4, 2.3, 2.0), and that nausea/vomiting, diarrhea and abdominal pain resolve faster (respective median recovery time (in weeks): 0.8, 0.7 and 1,8 weeks. Multistate models were tested to provide a visualization of the evolution of probability of onset and probability of recovery for each AEI groups since the inclusion. For dermatological toxicity, the highest probability of onset was between 2 and 3 months (38% at 3 months). Patients who experienced a first dermatological AE had less chance to develop a second one (at 3 months, probability of 15% vs 54%). Additional analyses will be presented at the meeting. This novel approach allows a more nuanced perspective and approach to understanding AEI patterns, including linking disparate AE’s. This type of visualization can help heath care givers to better predict EC side effects which is generally well tolerated in most patients, with most AEIs being mild to moderate in severity. Analyses providing novel visual aids to support the interpretation of safety profiles of EC should help oncologists and heath care givers in daily clinical practice to improve overall outcomes for patients.
更多查看译文
关键词
colorectal cancer,toxicity profiles,encorafenib,e-mutant
AI 理解论文
溯源树
样例
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要