Presentation (CTI)

OBJECTIVES: In HER2-mutant NSCLC, much of the efficacy evidence is single-arm, so HTA submissions, including EU joint clinical assessments, increasingly rely on unanchored indirect treatment comparisons (ITCs) and external controls. Their validity rests on the strong assumption that all prognostic factors and effect modifiers have been adjusted for. We synthesized the evidence to define an evidence-graded minimum adjustment set for such comparisons.
METHODS: A targeted literature review searched Embase and MEDLINE for observational studies and clinical trials in advanced HER2-mutant NSCLC (January 2019-December 2025) reporting OS, PFS, ORR, or DoR by baseline subgroups. Candidate variables—age, sex, smoking, ECOG performance status, metastatic status, stage, histology, PD-L1, HER2-mutation subtype, and co-mutations—were graded as prognostic factors or effect modifiers by cross-study consistency and analytic rigor (multivariable adjustment; interaction testing where reported).
RESULTS: Thirty-three publications were included (19 observational; 14 from 11 trials). Demographic factors—age, sex, smoking—showed no consistent prognostic or effect-modifying role. ECOG performance status (notably for OS) and metastatic burden (brain, extra-thoracic) showed the most consistent prognostic associations. HER2-mutation subtype (especially YVMA/Y772_A775dup) and TP53 co-alterations showed prognostic signals and the strongest evidence of potential effect modification (TKI versus non-TKI), though this was heterogeneous and treatment-specific. PD-L1, histology, and stage were limited or inconsistent. The evidence derived predominantly from small, retrospective, single-arm, or exploratory subgroup analyses; most effect-modifier signals were hypothesis-generating.
CONCLUSIONS: Credible ITCs in HER2-mutant NSCLC should assess balance in, and where feasible adjust for, the factors most consistently associated with efficacy outcomes: ECOG performance status, metastatic burden, HER2-mutation subtype, and TP53 co-mutation status; demographic adjustment alone is insufficient. For unanchored comparisons, where adjusting for all factors is rarely feasible (NICE DSU TSD18), residual confounding should be acknowledged and quantified—reporting covariate balance, effective sample size, and the range of effects compatible with the data rather than a single estimate.