The Expert Consensus on the Diagnosis and Treatment of Advanced Non-Small-Cell Lung Cancer with HER2 Alterations (2025 Edition) places HER2 mutation, gene amplification, and protein overexpression under the same category of “HER2 alterations”.[1] Yet these abnormalities occupy different biological layers: mutation changes the DNA sequence, amplification changes gene dosage, and overexpression describes a protein phenotype. They therefore do not inherently constitute a single disease entity; the more important distinction is which abnormality defines oncogene dependence and which mainly indicates drug accessibility. Classic HER2-positive breast cancer is predominantly defined by ERBB2 amplification and consequent protein overexpression, whereas activating ERBB2 mutations occur in only a low-single-digit proportion—approximately 1.6% in an early large sequencing analysis—and usually represent a separate, non-amplified disease.[2] The HER2 landscape in lung cancer is different. The best-established HER2 driver state is an activating mutation in the tyrosine kinase domain, particularly an exon 20 insertion. Such mutations occur in approximately 2%–4% of lung adenocarcinomas and account for roughly 5% of identified oncogenic driver events in large molecular series; they are usually mutually exclusive with other canonical drivers and do not require HER2 amplification or high IHC expression.[3–5] The breast-cancer model of “amplification–overexpression–HER2 dependence” therefore cannot be transplanted directly into lung cancer merely because the same receptor is involved.

The modest activity of earlier pan-HER TKIs—including afatinib, dacomitinib, and pyrotinib—left a central uncertainty unresolved: was HER2 mutation simply not a strong tumour-maintaining driver, or had these drugs failed to shut down mutant HER2 adequately? Their limited selectivity and imperfect structural fit for HER2 exon 20 insertions, together with dose-limiting diarrhoea and rash caused by wild-type EGFR inhibition, prevented a clean answer.[1,6] Zongertinib has substantially narrowed this uncertainty. In previously treated NSCLC with HER2 TKD—tyrosine kinase domain—mutations, the confirmed objective response rate was 71%, with a median progression-free survival of 12.4 months; in patients who had received no systemic therapy for advanced disease, the objective response rate reached 76% and median progression-free survival was 14.4 months.[7,8] When a TKI designed to inhibit mutant HER2 while largely sparing wild-type EGFR produces sustained tumour regression in most patients, the result functions as a clinical test of HER2 dependence. HER2 TKD-mutant lung cancer is therefore approaching the EGFR-like model of “genotype–kinase addiction–selective TKI”.

The success of HER2-directed antibody–drug conjugates demonstrates something different. In DESTINY-Lung02, trastuzumab deruxtecan achieved an objective response rate of 49% at the recommended dose in previously treated HER2-mutant NSCLC.[9] However, an ADC acts mainly through binding to membrane HER2, receptor internalisation, and release of a cytotoxic payload. Translational work further suggests that, in HER2-mutant lung cancer, ubiquitination—the attachment of ubiquitin tags that promotes recognition, internalisation, and trafficking of HER2 towards lysosomal processing—and the subsequent movement of the receptor into the cell may explain ADC uptake better than protein overexpression alone.[10] ADC activity therefore establishes HER2 as an exploitable delivery portal, but does not by itself prove addiction to HER2 signalling. More importantly, lung cancer has not reproduced the evidentiary structure of HER2-low breast cancer. DESTINY-Breast04 used a defined IHC category and demonstrated improvements in both progression-free and overall survival in a randomised phase III trial, whereas expression-selected NSCLC studies remain predominantly single-arm and have shown weaker response signals than mutation-selected cohorts.[11,12] The most compelling ADC activity in lung cancer has emerged first in mutation-selected, rather than expression-selected, populations. This is not a simple replay of the HER2-low breast-cancer paradigm.

HER2 TKIs and HER2 ADCs should therefore not be regarded merely as two interchangeable forms of “HER2-targeted therapy”. A TKI tests and suppresses oncogene dependence and may be considered cause-directed therapy; an ADC exploits receptor accessibility and internalisation and is better understood as phenotype-directed therapy. This distinction also suggests a cautious sequencing hypothesis. In a tumour driven by a clonal activating HER2 TKD mutation, beginning with a selective TKI to suppress the central oncogenic signal, followed after progression by an ADC if surface HER2 expression and internalisation remain preserved, may be more biologically coherent than treating the two approaches as freely interchangeable. Current evidence is insufficient to establish an optimal sequence, so this should remain a mechanism-based hypothesis rather than a clinical rule. Regardless of the eventual sequence, one distinction is already clear: mutation defines kinase dependence, whereas expression defines drug accessibility; both concern HER2, but they do not convey the same information. HER2 in lung cancer is not a second HER1, because only the disease defined by activating HER2 mutations is beginning to resemble the HER1/EGFR model.