October's FDA approvals, read as a feasibility problem

Several of October's FDA oncology actions are worth reading less for the efficacy headlines and more for how narrowly each one defines its patient population. For anyone who plans, activates, or recruits for the trials that follow, the population definition is the feasibility problem.

Three approvals, three ways of narrowing the population

By biomarker. The October 8 approval of atezolizumab with chemotherapy covers adjuvant treatment of stage III colon cancer that is mismatch repair deficient (dMMR). In the ATOMIC trial (711 adults), disease-free survival favored the atezolizumab arm, with a hazard ratio of 0.50. Whether a patient belongs to this population depends on a laboratory result, not on how the disease looks clinically.

By treatment timing. The October 7 approval of tucatinib with trastuzumab and pertuzumab applies to maintenance treatment of HER2-positive breast cancer after induction therapy. In HER2CLIMB-05 (654 adults), median progression-free survival was 24.9 months versus 16.3 months with placebo. The eligible window opens at a specific point in a patient's treatment course, and closes if that point passes.

By disease subtype, with extrapolation. The October 14 expansion of nogapendekin alfa inbakicept with BCG extends to high-grade papillary-only BCG-unresponsive non-muscle invasive bladder cancer. The efficacy basis was cohort A of QUILT-3.032 (77 patients; 62% complete response), and the FDA noted it extended those findings from the carcinoma in situ population rather than relying on a separate trial in papillary-only disease.

What this means for feasibility

None of this is new in principle. What stands out is how many dimensions now stack at once — biomarker status, line and timing of therapy, and precise disease definition — and how often the data needed to apply them sits in different places.

A few practical consequences:

• Prevalence estimates get fragile. A feasibility assessment that counts "stage III colon cancer patients" overstates the pool if dMMR status isn't captured in the same dataset. The same goes for any criterion tied to a test that may not have been ordered.
• Timing windows punish slow activation. If eligibility opens only after induction ends, a site that activates late can miss a meaningful share of its own patients.
• Protocol language matters. When a population is defined by subtle distinctions, free-text eligibility criteria invite inconsistent interpretation across sites.

The patient-matching side

This is also the underlying reason patient-facing trial search is hard. A patient may reasonably know their cancer type and stage but not their biomarker result or exactly where they sit in a treatment sequence. The information needed to match them to a study exists — it's just rarely structured in a way either a patient or a search tool can use directly.

That gap, more than any single approval, is the argument for treating eligibility criteria as structured data from the start of protocol development rather than as text to be interpreted later.

This post reflects the author's commentary on publicly posted FDA approval notifications. Refer to the sources below for the full approval details.