How Basket Trials Support Tumor-Agnostic Drug Approvals

For most of the history of cancer medicine, a drug was tested and approved for one place in the body: a breast drug, a colon drug, a lung drug. A basket trial breaks that habit. It enrolls patients by the molecular alteration inside the tumor and lets the organ of origin vary, then asks a single question, whether a drug aimed at that alteration shrinks tumors no matter where they arose. When the answer is a clear yes, regulators can approve the drug for a mutation rather than a body part, an outcome known as tumor-agnostic approval.

Key points#

• A basket trial groups patients by a shared genetic driver, not by tumor site, so one protocol can hold cancers from many different organs.
• The design suits alterations that turn up across many cancers but are rare in each, where a conventional single-organ trial could never enroll enough patients.
• NTRK-fusion drugs (larotrectinib and entrectinib) are the clearest worked example: high, durable response rates across a range of histologies.
• Because the target is rare, these trials are usually single-arm and judged by response rate, which supports accelerated approval but leaves survival benefit to be confirmed later.

Sorting by the lock, not the door#

Think of a tumor's driver mutation as the lock a targeted drug is cut to open. Traditional trials sort patients by which door they walked through, the organ where the cancer began. A basket trial sorts by the lock instead. If the same faulty lock sits inside a colon cancer, a salivary gland tumor, a sarcoma, and a thyroid cancer, one protocol can hold all of them together, because the drug is designed to act on the lock and should not care about the door.

This inversion earns its place under one specific condition: a genetic alteration that recurs across many cancer types but is uncommon within each of them. That combination is exactly what defeats the older approach. There are simply not enough patients of any single organ type to run a standard trial, yet added together across organs the numbers become workable.

Where the design proves itself: NTRK fusions#

NTRK gene fusions are the textbook case. When NTRK1, NTRK2, or NTRK3 fuses with another gene, the cell makes a TRK protein that is switched permanently on and pushes the tumor to grow. These fusions show up in a long list of cancers, are rare in most common tumors, and are highly recurrent in only a few unusual ones. No single-organ study could gather enough fusion-positive colon cancers to prove anything. Pooling across organs is what makes the evidence possible.

Two drugs carried this idea into regulatory approval. Larotrectinib received accelerated approval from the FDA in late November 2018 for adults and children with solid tumors carrying an NTRK gene fusion (without a known acquired resistance mutation) that are metastatic or where surgery would cause serious harm, and who have no satisfactory alternatives. The approval drew on three multicenter, open-label, single-arm trials. In the group of patients that supported it, most had an objective response, the responses appeared across several different histologies, and they lasted. In April 2025 the FDA converted the drug to full approval on a larger pooled population with longer follow-up, at which point the median duration of response was measured in years.

Entrectinib walked a parallel road. An integrated analysis of three early-phase trials was published in Lancet Oncology and later updated in Clinical Cancer Research. The updated analysis reported an objective response rate of roughly 61 percent across NTRK-fusion solid tumors, with a median duration of response near 20 months. Entrectinib also crosses the blood-brain barrier, and the pooled data showed responses inside the brain in patients with central nervous system disease, which matters because brain metastases are common in several of these cancers. Two separate drugs, two separate programs, and results that agree across the shared histologies build a strong biological case: here the driver, not the organ, governs the response.

The regulatory bargain#

The FDA has set out its reasoning in a guidance on tissue-agnostic drug development, and the logic is conditional rather than open-ended. When a biomarker is biologically central to a cancer's growth, when it recurs across many tumor types, and when each single tumor-plus-biomarker pairing is too rare for a randomized trial, a single-arm basket study can support approval, usually through the accelerated pathway. The measures that carry the weight are objective response rate (how many tumors shrink by a set amount) and duration of response (how long that lasts). A tumor that visibly shrinks under a targeted drug is a reasonable signal of benefit, though not a guarantee of it. That is why accelerated approval on this basis is explicitly tied to later confirmation of clinical benefit.

This is the same scaffold that produced the first tumor-agnostic approval, an immunotherapy for mismatch-repair-deficient tumors in 2017, and later approvals for other shared targets. The NTRK drugs are the cleanest version because the biology is so direct: one fusion makes one overactive kinase, and one drug blocks that kinase at its source.

What the design cannot tell you#

A single-arm trial has no comparison group. Without one, it cannot fully separate the drug's effect from the natural course of the disease or from the way patients were selected into the study. Response rate and duration of response are surrogates, and a high response rate does not automatically buy longer survival. These are built-in properties of the design, not defects in any particular study, and they are the reason confirmation is required rather than assumed.

Pooling across organs adds a second caution. A group defined by one biomarker is still biologically mixed. The same fusion sits inside tumors with very different genetic backgrounds, so the headline response rate you see can be nudged up or down by whichever organ types happened to enroll most patients. An effect that looks solid on average may be thinner in a histology that contributed only a handful of cases, and those small subgroups carry the widest uncertainty. Resistance is a further limit: acquired mutations can switch the kinase back on and erode an early response, which is one reason durability data deserve as much of your attention as the top-line number.

None of this argues against basket trials. It argues for reading them precisely. The design answers one question well, whether targeting a shared driver produces responses across tumor types, and the NTRK drugs answer it convincingly. What it cannot deliver on its own is a controlled estimate of survival or a promise that every organ behaves like the pooled average. Holding both of those truths at the same time is what a tumor-agnostic approval asks of you.