Design: Phase 3, international, randomised, open-label
Population: 338 patients with EGFR-mutated NSCLC progressing on osimertinib, MET overexpressed or amplified
Primary outcome: Progression-free survival by blinded central review
Effect: PFS HR 0.55 (0.42–0.72); OS HR 0.68 (0.50–0.91); 12-month OS 70% vs 53%

MET amplification or overexpression is one of the commonest mechanisms by which EGFR-mutated NSCLC escapes osimertinib. SAFFRON, an international open-label phase 3 trial published in NEJM on 25 October, screened 1,746 patients progressing on osimertinib and randomised 338 with MET overexpression (IHC 3+ in at least 90% of cells), MET amplification (10 or more copies on FISH), or both. They received osimertinib plus the MET inhibitor savolitinib, or platinum–pemetrexed.

Progression-free survival improved (HR 0.55, 95% CI 0.42–0.72), with 36% against 10% progression-free at 12 months. Overall survival also improved at final analysis (HR 0.68, 0.50–0.91), with 70% against 53% alive at 12 months. Grade 3 or worse adverse events were similar (64% vs 63%). Peripheral oedema affected about half, with anaemia and diarrhoea in about a quarter.

The practical message is upstream: the benefit depends on finding MET-driven resistance, which requires a tissue biopsy at progression with IHC and FISH. Only about one in five screened patients qualified. The trial was industry funded, and savolitinib is not approved in India; its regulatory status elsewhere should be checked before use.

At progression on osimertinib, obtain a tissue rebiopsy where feasible and test for MET by IHC and FISH.
Plasma testing can miss MET overexpression, which is defined on tissue.
Where MET-targeted combinations are not accessible, platinum–pemetrexed remains the standard after osimertinib.
Peripheral oedema is common with savolitinib; monitor weight and limbs and manage early.
Re-test for small-cell transformation and other resistance mechanisms in the same biopsy.

Why it matters

It turns MET testing at osimertinib progression from optional to decision-making, because the result now changes survival.

The statistics, in plain English

The overall survival hazard ratio of 0.68 means about a one-third lower death rate over follow-up; the interval (0.50–0.91) excludes 1.0. Open-label design affects symptom-based outcomes more than survival, which is objective.