Abstract Background: Patients face complex treatment choices with a range of benefits/risks. Researchers seek new biomarkers to predict individualized risks/benefits. It is essential that biomarkers yield information useful to patients. Our prior research in metastatic breast cancer showed choice-based conjoint (CBC) allowed patients to express risk-benefit preferences. This study focused on patients with early stage breast cancer who received chemotherapy and quantified relative influence of benefit and specific toxicities on predicted treatment choice for common therapies. Methods: A CBC survey was developed and sent to patients. The survey showed 12 pairs of hypothetical treatment choices based on benefit/risk profiles similar to those of TC, AC, AC®T, and TAC to allow assessment of what drove selections. Choices included benefit (relative risk reduction range: 20%-50%) and likelihood of toxicities including: peripheral neuropathy (PN range: 0%-60% with degrees of severity/duration); congestive heart failure (CHF range: 0%-10%); and clinically relevant infection (CRI range: 5%-25%). Analysis (N = 417) allowed treatment choice prediction for any benefit/toxicity combination. Hypothetical biomarkers were modeled to determine shifts in toxicity/efficacy necessary to change treatment selection. Results: Severity of PN experience had significant influence in decisions. When simply asking patients for perceived impact of previous PN experience on a future decision, the majority who had mild/moderate PN felt a future decision would not be impacted by a similar experience. Conversely, a plurality (47%) who had prior severe PN would be less likely to take similar future treatment. With CBC, we examined the derived influence of risks/benefits on treatment decisions. Based on estimated benefit/risk profiles, 50% selected a regimen similar to TC > AC (22%) > AC→T (16%) > TAC (10%). Incremental benefit had substantial impact on selected regimen. The desire for a given regimen decreased with incremental increase for each toxicity; most dramatically for CHF. Patients with a perception of low recurrence risk had higher preference for a non-anthracycline based profile, while those with higher perceived recurrence risk had higher preference for a combined anthracycline/taxane based profile. Interestingly, PN-naive patients had higher preference for non-taxane based profile, whereas those who had prior PN had higher preference for some taxane-based profiles. With moderate/limited PN, substantially more PN-naive patients avoided a taxane profile (31%) vs. those who had experienced PN (18%). When focusing on likelihood of severe/irreversible PN, there was dramatic shift toward non-taxane based regimens: AC (53%) > TC (23%) > AC→T (20%) > TAC (3%). When modeling impact of biomarkers on therapy selection, as likelihood of PN in taxane profiles increased, the fraction that chose a non-taxane profile increased. As likelihood of CHF in anthracycline-based profile increased, the fraction that chose such a regimen decreased. Conclusion: Patients considered all information about benefit and risks when making treatment choices. The information effective hypothetical biomarkers could add impacted these choices. Personal experience with a given toxicity appeared to impact decision for future therapies. Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P6-08-05.