377 Using artificial intelligence (AI) to compare patient perspective of PD-1 and BRAF inhibitors for melanoma treatment

Abstract

Due to side effects and adverse psychosocial factors, there can often be a disconnect between clinical impression and the patient perspective of treatment. Melanoma patients frequently use social media to discuss their disease sentiments and outcomes. We analyzed social media on melanoma treatments, PD-1 inhibitors (pembrolizumab/Keytruda, nivolumab) and BRAF inhibitors (dabrafenib, vemurafenib), associated with Patient Global Impression of Change (PGIC) terms to compare and identify patient burden. 12,599,313 publicly available online social media text data were extracted and run through Brandwatch Artificial Intelligence-powered database to categorize treatment-specific posts with PGIC terms associated with sentiment. Out of 52,962 posts related to a select list of melanoma treatments, we identified Keytruda (6,080), nivolumab (1,614), dabrafenib (529), and vemurafenib (329) posts. The top ten types of posts by volume for each treatment were predominantly positive for patient impression of change of treatment (improving, well) in contrast with associated negative emotions (fear and sadness). Patient-perceived better treatments were associated with decrease fear. Keytruda at a higher positive PGIC (92.3%) had markedly less fear (56.9%) compared to nivolumab positive PGIC (78.0%) and fear (81.6%). Similarly, dabrafenib positive PGIC (86.4%) had less fear (71.9%) compared to vemurafenib positive PGIC (78.0%) and fear (81.6%). Our initial results provide an indication for greater understanding of patient perspective and translation into more effective clinical and pharmaceutical response.