Investigators reported that conventional pathologist scoring of TROP2 by immunohistochemistry did not reliably predict response to TROP2-directed antibody–drug conjugates in non-small-cell lung cancer (NSCLC). To address that gap, the study team developed a fully supervised computational pathology pipeline named TROP2 quantitative continuous scoring (QCS). QCS quantifies TROP2 signal in individual tumor cells and separates subcellular compartments—membrane and cytoplasm—on routine IHC-stained slides. From those compartmental measurements the authors derived a normalized membrane ratio (NMR), defined as the ratio of membrane signal relative to the sum of membrane plus cytoplasmic signal. The derivation dataset came from the biomarker-evaluable population of the TROPION-PanTumor01 study (n = 115), where the algorithm was trained and the NMR measure established.
The QCS method directly addresses subcellular localization rather than relying on a single visual score. This allowed the authors to create a continuous metric that reflects the relative abundance of membrane-localized antigen—an important conceptual distinction because ADCs such as datopotamab deruxtecan (Dato-DXd) target cell-surface proteins. The source did not provide detailed algorithmic hyperparameters or every step of staining normalization in the article summary; those technical specifics are not reported here.
The study team optimized an NMR cut-point in the PanTumor01 set and then tested that cut-point in the primary intended-use population of the randomized TROPION-Lung01 trial. The primary intended-use group comprised patients with advanced nonsquamous NSCLC without actionable genomic alterations (n = 221), and the trial’s biomarker-evaluable population included 352 patients overall. Authors noted the cut-point was validated in that cohort to enable an assessment of association between NMR status and clinical outcomes across treatment arms.
This validation step separated the biomarker derivation data from the clinical-efficacy dataset. That design element strengthens claim testing, because the cut-point was not both derived and tested in the same small sample. The article summary did not include detailed numbers of concordance, false positives or false negatives for the cut-point beyond the outcome associations reported below.
Data indicated a treatment-dependent association between NMR status and progression-free survival (PFS). In patients who received datopotamab deruxtecan, median PFS was longer for NMR-positive versus NMR-negative samples: 7.2 months versus 4.0 months. The hazard ratio comparing NMR-positive to NMR-negative in the Dato-DXd arm was 0.40 (95% confidence interval 0.25–0.64). By contrast, in patients treated with docetaxel, median PFS was 4.1 months for NMR-positive and 4.4 months for NMR-negative samples, with a hazard ratio of 0.94 (95% confidence interval 0.59–1.49).
The study team reported a statistical interaction between treatment assignment and biomarker status (treatment by biomarker interaction P = 0.0098). Authors interpreted that finding as evidence that NMR may function as a predictive biomarker for benefit from Dato-DXd versus standard chemotherapy in this population. The analysis focused on progression-free survival; the summary did not enumerate other endpoints such as objective response rate or overall survival in this text, and these were not stated in the abstract excerpt provided here.
The authors placed their findings within the broader context of ADC therapy and the high prevalence of TROP2 expression in NSCLC tumor samples. They highlighted that a quantitative, subcellular measure—membrane localization relative to cytoplasmic signal—differed in its association with outcome depending on treatment, which supports a biomarker role tied to mechanism of action for a membrane-directed ADC.
Remaining questions documented or implied by the source include whether the NMR cut-point will generalize across independent laboratories and routine clinical workflows, and how NMR correlates with other clinical endpoints beyond progression-free survival. The article summary did not report algorithmic implementation details, harmonization steps for IHC across sites, nor specific Dato-DXd dosing regimens within the presented text; those items were not provided in the source material summarized here. The summary also did not list regulatory milestones or approvals tied to the biomarker; any such developments are not reported in this article excerpt.
Limitations that the authors acknowledged, as reflected in the reported data, include that the biomarker was developed in a trial-derived biomarker-evaluable set and tested within a single randomized study cohort. External validation in independent prospective datasets and reporting of additional clinical endpoints remain unanswered clinical questions arising from the source. The article ends with the validated association and the statistical interaction signal, and with the clear implication that further work—reported outside this summary—will be needed to establish implementation, reproducibility and broader clinical utility.