How Basket Trials Support Tumor-Agnostic Drug Approvals

A basket trial pools patients whose cancers arise in different organs but share one molecular alteration, then tests whether a drug aimed at that alteration works regardless of where the tumor started. When the shared target is rare, a single-arm design without a comparator is often the only practical way to gather evidence, and the drug is judged mainly by how many tumors shrink and for how long. The NTRK-fusion TRK inhibitors larotrectinib and entrectinib are the clearest worked example of this approach, and they also expose exactly where its evidence runs thin.

What a basket trial is built to do

Most oncology trials enroll one tumor type: breast, colon, lung. A basket trial inverts that logic. It enrolls by biomarker first and lets tumor histology vary, so a single protocol can hold a colon cancer, a salivary gland tumor, a sarcoma, and a thyroid cancer side by side because all of them carry the same driver. This design earns its place when a genetic alteration is shared across many cancers but is uncommon in each one individually.

NTRK gene fusions fit that profile. Fusions involving NTRK1, NTRK2, or NTRK3 produce a constitutively active TRK protein that drives tumor growth. They appear in a long list of cancers but are rare in most of them, and in only a few uncommon tumor types are they highly recurrent. No single-histology trial could enroll enough NTRK-fusion colon cancers, for instance, to say anything statistically meaningful. Pooling across histologies is what makes the study feasible at all.

The NTRK story as evidence

Larotrectinib received accelerated approval from the FDA on November 26, 2018 for adult and pediatric patients with solid tumors carrying an NTRK gene fusion, without a known acquired resistance mutation, that are metastatic or where surgery would cause severe morbidity, and who lack satisfactory alternatives. That approval rested on three multicenter, open-label, single-arm trials: LOXO-TRK-14001, SCOUT, and NAVIGATE. In the evaluable cohort that supported the initial approval, a large majority of patients had an objective response, responses were seen across a range of histologies, and they were durable. The FDA converted the drug to full approval on April 10, 2025 after a larger pooled population and longer follow-up, by which point the reported median duration of response was measured in years rather than months.

Entrectinib followed a parallel path. An integrated analysis of three phase 1-2 trials, ALKA-372-001, STARTRK-1, and STARTRK-2, was published in Lancet Oncology and later updated in Clinical Cancer Research. The updated analysis reported an objective response rate of about 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 intracranial responses in patients with central nervous system disease, which matters because brain metastases are common in several of the cancers involved. Two independent drugs, two independent basket programs, and convergent results across shared histologies together make a strong biological case that the driver, not the organ, governs the response.

Why regulators accept the design here

The FDA has described its thinking in a guidance for industry on tissue-agnostic drug development. The logic is conditional. When a biomarker is biologically central to a cancer's growth, when it recurs across many tumor types, and when each individual tumor-biomarker combination is too rare for a conventional randomized trial, a single-arm basket study measuring objective response rate and duration of response can support approval, usually through the accelerated pathway. Response rate serves as a surrogate: a tumor that measurably shrinks under a targeted drug is a reasonable, though not guaranteed, signal of clinical benefit. Accelerated approval on that basis is explicitly contingent on confirmatory evidence of clinical benefit over time.

This is the same framework that produced the first tumor-agnostic approval, pembrolizumab for microsatellite-instability-high tumors in 2017, and later approvals for other agnostic targets. The NTRK inhibitors are the cleanest case because the biology is unusually direct: a single fusion produces a single overactive kinase, and a drug that blocks that kinase acts on the mechanism itself.

The appraisal limits worth naming

A single-arm trial has no control group, so it cannot separate the drug's effect from the natural history of the disease or from patient selection. Response rate and duration of response are surrogates, and a high response rate does not automatically translate into longer survival, which is why confirmatory evidence is required rather than assumed. These are structural features of the design, not flaws in any one study.

Pooling across histologies carries its own caution. A biomarker-defined population is still biologically heterogeneous. The same NTRK fusion sits inside cancers with different genetic backgrounds, and the pooled response rate can be pulled up or down by whichever tumor types happen to be best represented. A treatment effect that is real on average may be weaker in a histology that contributed only a handful of patients, and the smallest subgroups leave the widest uncertainty. Resistance is a further limit: acquired mutations can restore kinase signaling and blunt the initial response, which is one reason durability data matter as much as the headline number.

None of this argues against the design. It argues for reading it precisely. A basket trial answers a specific question, whether targeting a shared driver produces responses across tumor types, and the NTRK inhibitors answer it convincingly. What such a trial cannot deliver on its own is a controlled estimate of survival benefit or a guarantee that every histology behaves like the pooled average. Holding both facts at once is what evidence-literate appraisal of tumor-agnostic approvals requires.

This article is educational and not medical advice.