Zidesamtinib

ROS1 fusions drive roughly 1 to 2 percent of non-small cell lung cancers, and that subgroup skews young, skews never-smoker, and carries brain metastases at a high rate ^[3]. The first generation of ROS1 inhibitors, crizotinib and entrectinib, produced response rates above 60 percent, and then ran into two walls: resistance mutations at the solvent front, of which ROS1 G2032R is the most common, and poor penetration into the central nervous system ^[3].

Zidesamtinib was built against both walls at once, and against a third problem the later inhibitors introduced. Lorlatinib, repotrectinib and taletrectinib reach the brain, but they also hit the tropomyosin receptor kinase family, and TRK inhibition is what produces the dizziness, ataxia, cognitive change and mood change that force dose reductions ^[3]. In an accelerated mutagenesis screen zidesamtinib suppressed more than 1,500 pooled ROS1 mutants with essentially no resistance emerging, outperforming crizotinib, entrectinib and repotrectinib, and a 2.2 angstrom cocrystal structure with ROS1 G2032R shows it accommodating the mutated residue while clashing with the TRK kinases, which is the structural basis of the selectivity ^[1].

The approval rests on ARROS-1, a single-arm, open-label, multi-cohort phase 1/2 trial. In the 117 patients who had already had at least one ROS1 inhibitor, the overall response rate was 44 percent; 82 percent of responders were still responding at six months and 69 percent at twelve. Among the 50 patients with measurable disease in the brain, 48 percent responded intracranially. That is a single-arm accelerated-approval dataset with no control group, and a confirmatory trial is the condition attached to it.

Zidesamtinib is the first medicine in lung cancer for GSK, which acquired Nuvalent, the company that discovered it.

The label carries warnings for central nervous system adverse reactions, QTc interval prolongation, interstitial lung disease and pneumonitis, skeletal fractures, myalgia with creatine phosphokinase elevation, pancreatic toxicity and embryo-fetal toxicity. The most common adverse reactions at 15 percent or above were oedema in 38 percent, peripheral neuropathy in 25 percent, constipation in 17 percent, fatigue in 16 percent and shortness of breath in 15 percent. The grade 3 or 4 laboratory abnormalities seen in at least 2 percent were raised lipase in 8 percent, raised creatine phosphokinase in 4 percent, raised triglycerides in 3.5 percent, low lymphocytes in 2.7 percent and low haemoglobin in 2.3 percent. Cognitive disorders, psychiatric disorders, stomatitis, ataxia, pneumonitis, QTc prolongation, pancreatitis and ankle fracture each occurred in under 10 percent. Strong and moderate CYP3A inhibitors and inducers are to be avoided.

The TRK-sparing design is a claim about a smaller neurological burden than lorlatinib or repotrectinib, not a claim of none; cognitive and psychiatric events still appear in the label, and no head-to-head trial against another ROS1 inhibitor has been run.

Zidesamtinib began as NVL-520 at Nuvalent, a company founded on the idea that the neurological side effects of the existing ROS1 and ALK inhibitors were an engineering problem rather than a fixed cost of the class. The preclinical case was published in Cancer Discovery in 2023: selective for ROS1, sparing of TRK, active against the secondary resistance mutations, and able to reach the brain ^[2]. The ARROS-1 trial opened in January 2022. Results in patients already treated with crizotinib and with repotrectinib were presented at the 2024 ESMO congress ^[5]. GSK acquired Nuvalent, and the FDA approved the drug on 22 July 2026, ahead of its 18 September target date, under Breakthrough Therapy and Orphan Drug designations.

Within thoracic oncology zidesamtinib is treated as the current answer to a specific and well defined problem: what to give a ROS1-positive patient whose disease has come back on a first or second inhibitor, particularly in the brain. Reviews written before the approval already named it as the agent expected to address the TRK-driven neurological toxicity of the earlier drugs while covering G2032R ^{[3] [4]}. The reservations are the ones that attach to any accelerated approval on a single arm: a 44 percent response rate has no control group beside it, overall survival is unmeasured, optimal sequencing against taletrectinib and repotrectinib is undefined, and resistance will eventually appear here too.