Career Development Award
Julie Deutsch, MD
Johns Hopkins School of Medicine
Baltimore
MD
Dr. Deutsch’s proposal centers around finding better pathologic predictors of response to neoadjuvant IO in early stage NSCLC. She will utilize machine learning/artificial intelligence to test an algorithm that she and her team have developed that assesses percent residual viable tumor (%RVT), which is the amount of tumor left at the time of surgery. Dr. Deutsch will also characterize tissue specimens using a novel immunofluorescence platform to identify cell types and spatial relationships that are associated with patient benefit to immunotherapy+chemotherapy. This approach can help inform which patients should receive a given therapy, how they will respond, and additional possible targets for the development of new therapies.
Research Summary
Immunotherapy revolutionized the treatment of lung cancer, and is now being extended so patients can receive therapy before surgery. This was supported by a large clinical trial, CheckMate 816 (CM816), where patients with lung cancer showed improved survival when treated with immunotherapy+chemotherapy before surgery, compared to chemotherapy alone followed by surgery. However, there is an unmet need to identify who is most likely to benefit from such an approach. To address this gap, we will apply novel, next-generation pathology biomarkers utilizing machine learning/artificial intelligence and multispectral imaging. Specifically, we have shown that the amount of tumor left at the time of surgery, termed percent residual viable tumor (%RVT), predicts survival. To date, %RVT assessment is primarily performed visually on glass slides using a light microscope. We developed a machine learning-based algorithm for assessing %RVT on digitized glass slides using a small cohort of patients at Johns Hopkins to improve standardization and throughput in preparation for broad usage. Here, we will test the algorithm’s performance in a larger cohort of patients (the CM816 patients). Additionally, we will characterize tissue specimens using the novel multiplex immunofluorescence AstroPath platform, which uses algorithms first developed in astronomy, to identify cell types and spatial relationships that are associated with patient benefit to immunotherapy+chemotherapy. Our goal is to use cutting-edge technologies to improve the care of lung cancer patients by informing which patients should receive a given therapy, how well patients will do after receiving therapy, and possible additional targets for the development of new therapies.
Technical Abstract
As seen in the phase III trial CheckMate 816 (CM816), neoadjuvant anti-PD-1+chemotherapy improves survival for patients with resectable non-small cell lung cancer (NSCLC), with pathologic response as a major trial endpoint. Our team led the Central Pathology Review for CM816, and we showed the first prospective evidence that the full spectrum of % residual viable tumor (%RVT) associates with event free survival. Given the data supporting pathologic response as a survival surrogate, %RVT will likely be incorporated into the next generation of clinical trials and may ultimately guide clinical decision-making. %RVT is primarily evaluated using visual assessment of routine hematoxylin and eosin-stained slides. We developed a machine learning-based approach to score %RVT, which allows for a standardized approach that can be completed rapidly for a large volume of patients, and we propose to test this algorithm in resection specimens from CM816. Additionally, we will use multiplex immunofluorescence (mIF) to quantify individual features of pathologic response, locate them within the larger tumor bed, and determine the relative contribution in predicting patient outcomes. Furthermore, we will use the novel AstroPath platform, a mIF whole-slide imaging platform that uses algorithms first developed in astronomy to generate tumor-immune maps, to identify additional pre- and on-treatment biomarkers of response. Our goal is to leverage emerging technologies (i.e, machine learning and mIF) to develop the next generation of pathology biomarkers, including pathologic response assessment, and to identify additional features that can potentially be targeted in combination with anti-PD-(L)1+chemotherapy to improve clinical benefit in patients with NSCLC.