Measuring KEYNOTE-522 response from TNBC patient biopsies ex vivo using the E-slice assay.

Abstract

1132 Background: The KEYNOTE-522 regimen demonstrated a significant survival advantage of adding neoadjuvant pembrolizumab (pembro) to standard-of-care (SOC) chemotherapy (chemo) in TNBC patients. However, most patients will not benefit from this costly and potentially toxic therapy. A diagnostic test that can stratify responders and non-responders in advance will reduce overtreatment in this population. The E-slice assay is a proprietary 3D human tumor tissue culture platform that enables rapid, personalized drug sensitivity testing. E-slices preserve immune cells in their native state ex vivo, enabling immunotherapy response measurement from tissue resident immune cells. Methods: In this prospective clinical study, newly diagnosed triple-negative breast cancer (TNBC) patients scheduled to receive KEYNOTE-522 treatment at MD Anderson Cancer Center were enrolled. E-slices were generated from 14-gauge core biopsies and treated with IgG control, pembrolizumab, or a combination of pembrolizumab with paclitaxel + carboplatin followed by doxorubicin + cyclophosphamide. Diagnostic performance of the E-slice assay was evaluated using sensitivity, specificity, and ROC analysis, with the optimal cutoff determined by the Liu method. Concordance between assay-based predictions and clinical treatment response was assessed using Cohen’s kappa, and paired classification differences were tested with McNemar’s test. Patient tumor responders were defined solely by pathological complete response (pCR) or not, at the time of surgery, after standard-of-care neoadjuvant chemoimmunotherapy treatment. Results: A blinded interim analysis was scheduled after the first 10 patients evaluated in the clinic for pCR. The mean and SD of age were 52.10±12.52. The mean and SD for E-slice combo response, relative to DMSO @ day12 were 0.68±0.18. The optimal cutoff point for E-slice combo response was 0.78. The proportion of responders in patients was 0.8 (8/10). The proportion of responders from E-slice combo response was 0.7 (7/10). There was no significant difference based on McNemar’s test (p=1.0). The concordance rate is 75% (95% CI: 42.81%-94.51%). The sensitivity, specificity, positive predictive value, negative predictive value, and area under ROC curve at cutoff point for E-slice were: 87.5% (95% CI: 47.3%-99.7%), 100% (15.8%-100%), 100% (59%-100%), 66.7% (9.43%-99.2%), and 0.94 (0.82-1.0), respectively. Conclusions: E-slice assay predicts TNBC patient responses to pembrolizumab alone or pembrolizumab + chemo by providing drug sensitivity data in a clinically actionable time frame. Confusion matrices E-slice combo response. E slice Response Non-Responders Responders Total Patient Response Non-responders 2 0 2 Responders 1 7 8 Total 3 7 10 E-slice Responder was defined as E-slice combo response relative to DMSO @ day12 <0.7816. Non-responder was defined as ≥0.7816.