Last week, I shared my prediction for the results of the POLARIX clinical trial. Today, the results of this massive trial were published in The New England Journal of Medicine. The POLARIX trial enrolled 879 patients with Diffuse Large B-Cell Lymphoma (DLBCL). Half of these patients received the current standard of care for DLBCL, the five-drug combination called R-CHOP. The other half of patients received a modified drug combination where vincristine, one of the five drugs in R-CHOP, was replaced with an antibody-drug conjugate called polatuzumab-vedotin (Polivy). This experimental combination is abbreviated at R-CHP+Polivy.
Based on a previous clinical trial of Polivy as salvage therapy in relapsed/refractory (r/r) patients, I modeled the increased clinical efficacy of Polivy compared to vincristine. I then simulated the progression-free survival (PFS) for the five drug combinations including either vincristine (R-CHOP) or Polivy (R-CHP + Polivy). Based on the clinically observed efficacy of Polivy in r/r patients, the predicted benefit of replacing vincristine with Polivy in first-line treatment was significant (Hazard ratio (HR): 0.812; 95% confidence interval (CI): 0.668 - 0.988). At the time this simulation was done, all that was known about the results of the POLARIX trial was that it was successful, and the PFS predicted by my model supported that outcome.
Today, the full results of the trial were reported including the progression-free survival curves for the control (R-CHOP) and experimental (R-CHP + Polivy) arms. The results are remarkable, making R-CHP + Polivy the first combination to improve on R-CHOP progression-free survival in two decades. The complete results now allow me to compare my predicted results with the clinically observed results (Figure 1).
My predicted trial result matches the clinical result in three key ways:
The magnitude of benefit from Polivy is similar with a predicted 7% improvement in PFS at 2 years matching a 6.5% improvement observed clinically.
The time at which a benefit from Polivy can be observed is similar with a separation between the R-CHOP and R-CHP + Polivy curves occurring at approximately 6 months.
The hazard ratios between trials are similar (predicted: 0.81; [0.668-0.988] vs. observed: 0.73; [0.57-0.95]).
This result demonstrates the ability of my mechanistic model of curative combination therapies to predict clinical trial outcomes before the trial is completed or even started. Since the average oncology clinical trial costs $60k/patient (source), the POLARIX trial, enrolling 879 patients likely cost at least $52 million to conduct. Fortunately, POLARIX was successful since the new treatment significantly improves progression-free survival compared to the standard of care. However, multiple previous clinical trials in DLBCL have not been successful (Ibrutinib, Lenalidomide, Bortezomib). Given the high financial cost and a large participant burden of conducting clinical trials, we want to be as sure as possible that a trial will succeed and have a reliable estimate of the magnitude of benefit a new treatment will confer compared to the current standard of care. This model provides such insights even before a trial is conducted, potentially reducing the number of failed clinical trials and providing guidance as to how the trial should be designed for the best possible success.
Disclaimer: This work is a work in progress. The methods and details of the model have been intentionally excluded since this work has not yet been published in a peer-reviewed journal.
Acknowledgments: I would like to thank my advisor, Dr. Adam Palmer, for his continued support of my research and the rest of the Palmer lab, Deb Plana, Sarah Cooper, Hannah Hwangbo, and Jacob Pantazis, for their feedback and moral support.
Comments