participants suggested. Researchers could use animal models to suggest which combinations of agents would be most effective at countering the specific molecular defects in patients’ tumors. If clinical testing of such combinations cause toxicity or resistance, the same combinations could be retested in animal models to assess better dosing, scheduling, or the molecular resistance mechanisms and what other agents might counter them. Studies of biopsies collected from patients with tumor progression could also point toward more effective combinations of agents.
Several participants pointed out the need for higher standards for both preclinical and clinical effectiveness. Tumor shrinkage is likely to be a better endpoint in laboratory studies than blocking tumors from forming or from growing, whereas overall survival is likely to be a better endpoint than time to progression in clinical studies, they argued.
Given the numerous possible combinations and limited number of patients and other resources, there has to be some prioritization of what combinations should be tested clinically. Suggestions for prioritizing included testing only those combinations that:
• Perform well and consistently in several xenograft models;
• Have a biological mechanism for which there is an assay;
• Have demonstrated adequate pharmacokinetics and some evidence of activity or target engagement at clinically relevant doses and exposures;
• Are composed of the best in each class of agents that are pharmacologically compatible; and
• Have validated biomarkers for patient selection and pharmacodynamics.
Some participants cautioned that effective combinations should not be judged on the basis of the single-agent activity of their components, as many combinations have been found to be effective even though studies of the single agents did not show significant effects. There also was concern about the additive, synergistic, or unexpected toxicities that can result from combinations, particularly those that target the same pathways. Researchers need to explore more creative innovations in the approach to dosing and scheduling to avoid toxicity and improve efficacy, several participants suggested. Agents could be used intermittently or sequenced in a manner that makes sense from a biological, mechanistic perspective.
To aid both preclinical and clinical investigations, more basic information could be gathered on genetic expression, and feedback and network responses to signaling perturbations and DNA damage. Some participants stressed that information is also needed on the non-genetic effects that influence treatment, including the microenvironment of the tumor, the