Background: Biomarkers are regularly utilized to select treatment within cancer clinical trials. However, there remains a lack of understanding regarding physician perspectives on what data is needed for physicians to comfortably use these markers to escalate or de-escalate chemotherapy. Methods: Semi-structured qualitative interviews with medical oncologists from different academic and community-based cancer centers were conducted to investigate perspectives on the utilization of biomarkers to de-escalate chemotherapy. Key topics explored included: (1) physician preference for biology-based (e.g. genomic profiles) vs. response-based (e.g. complete pathologic response) biomarkers, and (2) importance of personal familiarity with biomarkers. Interviews were audio-recorded and transcribed. Two independent coders analyzed transcripts using a constant comparative method in NVivo to identify major themes. Analysis was stratified by practice-type to elucidate differences between oncologists at academic and community practices. Results: Of the 39 participating physicians, 51% practiced in an academic setting and 49% practiced in a community setting. The majority of physicians (67% overall, 77% community, 59% academic) did not have a preference for biology-based vs. response-based biomarkers, if the data is equally strong and clinical use is appropriate for the clinical context (e.g. patient subtype). Many physicians were reassured by achieving a real-time therapeutic response, with 23% of physicians preferring response-based biomarkers. One physician stated, “I am still more comfortable with a real-time, well-validated biomarker, response marker, than I am with an overall predictive marker for a population”. In contrast, 10% (all academic) preferred biology-based biomarkers. One physician commented “I think the biology is probably more attractive because that potentially allows you to avoid treatment, whereas pathCR they've already had to get treatment to get there”. The majority of academic physicians (55%) felt that strong data was more important than personal familiarity with regards to implementation of novel biomarkers, as noted by one who stated, “As long as there's good data, I don't care.” 15% of community physicians shared a similar view. The majority of community physicians (54%) voiced familiarity to be more important in their comfort with biomarker use as noted by one physician who stated, “I think things I’m already familiar with, I'm more inclined to feel good about”. 18% of academic physicians held a similar perspective. Conclusions: Academic and community physicians’ perspectives regarding use of novel biomarkers overlap, with multiple factors playing a role in how these biomarkers are used in decision-making. Future research is needed to understand the impact of biomarker selection on clinical trial enrollment.