Risk-adapted first-line therapy selection in advanced EGFR-mutant NSCLC: a practical clinical algorithm integrating tumor biology and patient factors.

Advanced EGFR-mutant non-small cell lung cancer (NSCLC), specifically tumors harboring classical activating mutations (exon 19 deletions and L858R substitutions, which are the exclusive focus of this review), is no longer a single-treatment landscape. The MARIPOSA and FLAURA2 phase 3 trials have established that combination regimens of amivantamab plus lazertinib and osimertinib plus platinum-pemetrexed, respectively, extend both progression-free and overall survival compared with osimertinib monotherapy. Each carries substantially greater toxicity, treatment complexity, and cost. No validated, prospective framework exists to guide first-line selection between combination and single-agent strategies. Decision-making is currently anchored to trial eligibility criteria and institutional norms rather than individualized risk stratification. This review synthesizes evidence from landmark trials, molecular biomarker analyses, and real-world cohort studies to propose a practical risk-adapted algorithm integrating tumor biology, molecular co-alterations, patient performance status, comorbidities, logistical feasibility, and treatment preference. Patients with biologically high-risk features, including liver metastases, baseline central nervous system (CNS) involvement, circulating tumor DNA (ctDNA) elevation, or heavy disease burden, may derive the greatest absolute benefit from combination therapy. TP53 co-mutation warrants consideration but should be interpreted as a prognostic marker rather than a definitive treatment-selection criterion given conflicting predictive data across trials. The subcutaneous (SC) formulation of amivantamab, demonstrated in PALOMA-3 to reduce infusion-related reactions and administration time relative to the intravenous formulation, may improve the tolerability and logistical feasibility of MARIPOSA-based therapy, though prospective comparative evidence across formulations remains limited. Prospective biomarker-driven trials are needed to validate treatment selection; until that evidence matures, individualized shared decision-making guided by the proposed algorithm is the most defensible clinical approach.