NEW YORK – A machine learning algorithm used to quantitatively score TROP2 expression in patients with advanced metastatic non-small cell lung cancer (NSCLC) appears to predict response to AstraZeneca and Daiichi Sankyo's datopotamab deruxtecan (Dato-DXd).
At the International Association for the Study of Lung Cancer's World Conference on Lung Cancer (WCLC) in San Diego on Sunday, data presented from the ongoing Phase III TROPION-Lung01 trial showed that NSCLC patients who had higher ratios of cytoplasmic to membrane-bound TROP2 experienced higher objective response rates and longer median progression-free survival on Dato-DXd, compared to those on docetaxel.
Dato-DXd is a TROP2-directed antibody-drug conjugate comprising a plasma-stable linker that when cleaved within the cell, releases a cytotoxic anti-TROP2 payload.
"Dato-DXd has already demonstrated statistically significant benefits in terms of progression-free survival in patients with advanced metastatic [NSCLC], but conventional immunohistochemistry scoring has not been able to predict the response to TROP2-directed antibody-drug conjugates," said Marina Chiara Garassino, a professor of medicine at the University of Chicago and the principal investigator of TROPION-Lung01.
A major hurdle in developing such predictive assays, according to Sanja Dacic, professor of pathology at Yale School of Medicine, stems from the fact that TROP2 is expressed in nearly all non-small cell carcinomas. "[This] tells you that if we use traditional scoring systems, they are simply going to fail as predictive biomarkers," said Dacic, who discussed Garassino's research at the conference.
However, with digital pathology's entrance into the clinic, she said novel methods are emerging for analyzing whole slide images, often using automated computational assays, which can quantitatively evaluate biomarkers such as TROP2.
Thinking along those lines, Garassino and her colleagues hypothesized that a more quantitative assessment of TROP2 expression on the cell membrane and in the cytoplasm might lead to providing better predictions. To do so, they applied a quantitative continuous scoring (QCS) algorithm developed by AstraZeneca.
This fully supervised computational pathology approach scans digitized immunohistochemically-stained pathology slides for patterns of optical density and quantifies the ratio of membrane-bound TROP2 to overall TROP2, which the investigators called the normalized membrane ratio (NMR). Patients below a certain cutoff point, indicating they have a higher proportion of cytoplasmic TROP2, are considered biomarker positive.
The TROPION-Lung01 trial included 604 participants, 352 of whom were evaluable for the TROP2 biomarker and 221 of whom had non-squamous NSCLC without actionable genomic alterations. TROP2 biomarker positivity was most common in the non-squamous group with actionable alterations (75 percent), followed by those without actionable mutations (63 percent), and patients with squamous cancers (43 percent).
The overall response rate among all evaluable TROP2 biomarker-positive patients on Dato-DXd was nearly 33 percent, compared to roughly 10 percent for those on docetaxel. Median progression-free survival in the biomarker-positive group was 6.9 months in the Dato-DXd arm, compared to 4.1 months in the docetaxel arm.
TROP2 biomarker-negative participants showed less of a difference in outcome measures between treatment arms. Nearly 17 percent on Dato-DXd responded, compared to approximately 15 percent on docetaxel, while median progression-free survival was about three months and four months, respectively.
Patients with non-squamous NSCLC without actionable genomic alterations had similar outcomes as the overall group. Data on patients with non-squamous NSCLC with actionable mutations were not presented.
Among the biomarker-positive patients, serious treatment-related adverse events affected approximately 29 percent of those given Dato-DXd, and 46 percent of docetaxel-treated patients. Within the biomarker-negative subset, approximately 22 percent of participants treated with Dato-DXd and 27 percent of those receiving docetaxel experienced grade 3 or higher events. Events included stomatitis, ocular surface events, and interstitial lung disease.
Based on the results from TROPION-Lung01, Garassino concluded that "TROP2 NMR has the potential to be the first TROP2 biomarker and the first computational pathology biomarker for predicting clinical response to Dato-DXd in [NSCLC]."
Dacic noted that while there is not a gold standard against which to compare Garassino's research, the study "nicely demonstrated that … using this assay, they can separate patients who are going to benefit from TROP2 treatments." Still, further validation is needed before this QCS-based approach can be clinically implemented.
The data presented at WCLC comes from a tightly controlled clinical trial, which Dacic said raises questions about how the assay will perform in less controlled settings. "The question is … whether in the real world, we are going to be able to reproduce what was demonstrated in a clinical trial," she said.
Dacic added that the field currently lacks consensus guidelines for implementing digital pathology algorithms into daily practice, which would also have to overcome financial and technical challenges. "These slides are not easy to scan, and the computational approaches are not easy to apply," she said, adding that broad implementation of these technologies constitutes a multimillion-dollar investment for each healthcare system involved.
Despite the challenges Dacic raised, she still lauded the research presented by Garassino and agreed that the assay "has the potential to be the first TROP2 biomarker and the first computational biology predictive biomarker."
In Garassino's view, scaling implementation of the TROP2 QCS assay should not be difficult. "The test is very easy," she said in an email, involving standard staining techniques, after which "the automatic system will test and calculate the ratio. You don't need supercomputers."
In addition to its use in the TROPION-Lung01 trial, Daiichi Sankyo is investigating the QCS method in the AVANZAR and TROPION-Lung10 trials. In these Phase III studies, AstraZeneca is evaluating Dato-DXd as a first-line therapy in advanced, metastatic NSCLC patients.