Predicting Response to PD-1 Checkpoint Blockade Using Deep Learning Analysis of Imaging and Clinical Data

Recent Phase III studies have demonstrated the effectiveness of atezolizumab (PD-L1) in metastatic triple-negative breast cancer [3] and small cell lung cancer, while the standard of care for Stage III non-small cell lung cancer has changed with positive results of the PACIFIC Phase III study, where durvalumab (PD-L1) administered after chemoradiation showed a significant increase in overall survival.

Low response rates, generally in the 15% to 20% range in most diseases when used as a single agent, high therapy cost globally ($150,000 or more per year in the U.S) and serious immune-mediated adverse events, particularly when PD-1/PD-L1 inhibitors are combined with the CTLA-4 inhibitors (ipilimumab). Unpredictable and low patient response rates coupled with high drugs costs and serious toxicities can significantly burden healthcare systems, third-party payers and patients. Clearly, diagnostic tools to stratify patients according to response likelihood are necessary as PD-[L]1 checkpoint inhibitors continue to gain adoption.

The standard-of-care biomarker is an immunohistochemistry (IHC) test that measures levels of the PD-L1 protein expressed in tumor samples. Tumor mutational burden, presence of Tumor-Infiltrating Lymphocytes and inflammatory cytokines are being explored in multiple clinical trials involving PD-(L)1 often in combination with additional immuno-oncology (IO) therapies In such an approach, a non-invasive imaging scan can provide insight and information on the patient's entire tumor burden rather than a sample of a subset of lesions (as provided by biopsy or serum-based assays). When diagnostic images that depict all treatable lesions are further analyzed with computational techniques such as machine-learning and artificial intelligence, resulting in the identification of relevant imaging biomarkers, an accurate overall assessment of patient response to PD-[L]1 therapy may be attainable.

Inclusion criteria

Patients between 18 and 100 years of age -

Exclusion criteria

Patient under 18 years of age

-