BACKGROUND: The extent to which patients with cancer are willing to accept uncertainty about the clinical benefit of new cancer drugs in exchange for faster access is not known. This study aims to examine preferences for access versus certainty, and to understand factors that influence these preferences. METHODS: A US nationally representative sample of older adults were recruited via Cint, an online platform for survey research, to take part in an online discrete choice experiment. To be eligible, respondents had to self-report some experience with cancer-ie, they themselves, a close friend or a family member, previously or currently diagnosed with cancer. In the experiment, respondents chose between two cancer drugs, considering five attributes: functional status, life expectancy, certainty of the survival benefit of a new drug, effect of the drug on a surrogate endpoint, and delay in US Food and Drug Administration (FDA) approval time. The first primary outcome was the relative importance of certainty of survival benefit and wait time to respondents. The second primary outcome was willingness to wait for greater certainty of survival benefit, including subgroup analysis by cancer experience, age, education status, race or ethnicity and income. Secondary outcomes were changes in sensitivity to certainty and wait time, depending on the drug's effect on a surrogate endpoint, respondents' functional status, and life expectancy. The study plan was registered with ClinicalTrials.gov, NCT05936632. FINDINGS: Between July 7 and July 20, 2023, 998 eligible respondents completed the survey. 870 respondents (461 [53%] male, 406 [47%] female, and three [<1%] other) were included in the final analysis. Respondents showed strong preferences for high certainty of survival benefit (coefficient 2·61, 95% CI 2·23 to 2·99), and strong preferences against a 1-year delay in FDA approval time (coefficient -1·04, 95% CI -1·31 to -0·77). Given very low certainty a drug would provide survival benefit (no evidence linking a surrogate endpoint to overall survival), respondents were willing to wait up to 21·68 months (95% CI 17·61 to 25·74) for high certainty (strong evidence) of survival benefit. A drug's effect on a surrogate endpoint had no significant impact on drug choices (coefficient 0·02, 95% CI -0·21 to 0·25). Older respondents (aged ≥55 years), non-White, lower-income (<$40 000 per year) individuals, and those with the lowest life expectancy, were most sensitive to wait time. INTERPRETATION: Many cancer drugs approved through the FDA's accelerated approval pathway do not offer any survival benefit to patients. In this study, individuals expressed strong preferences for certainty that a cancer drug would offer survival benefit. Some individuals also expressed a higher willingness to wait for greater certainty than would be necessary to assess the survival benefit (over progression-free survival benefit) of most cancer drugs used in the metastatic setting. FUNDING: The London School of Economics and Political Science Phelan United States Centre.