Computer Model May 'Offer Better Picture' of Who Benefits from Bladder Cancer Therapy

February 13, 2019
Kristie L. Kahl
Kristie L. Kahl

Kristie L. Kahl is vice president of content at MJH Life Sciences, overseeing CURE®, CancerNetwork®, the journal ONCOLOGY, Targeted Oncology, and Urology Times®. She has been with the company since November 2017.

An integrative model helped to determine which patients with bladder cancer may benefit from checkpoint inhibitor therapy.

Integrative models of immune responses to checkpoint inhibitors may improve oncologists’ ability to anticipate clinical benefit among patients with bladder cancer, according to a recent study published in PLOS One.

“Checkpoint inhibitor immunotherapies have had major success in treating patients with late-stage cancers, yet the minority of patients benefit,” the researchers wrote. “Mutation load and PD-L1 staining are leading biomarkers associated with response, but each is an imperfect predictor. A key challenge to predicting response is modeling the interaction between the tumor and immune system.”

Therefore, researchers from University of Maryland (UMD), Microsoft Research and Memorial Sloan Kettering Cancer Center addressed this challenge with a new computer model approach trained to predict immune response in 21 patients after bladder cancer treatment based on 36 clinical, tumor and circulating features collected prior to therapy. The aim was to demonstrate this approach could identify a suite of features that accurately predicted a key immune system response to treatment while reducing over treatment by half, according to a UMD press release.

“If your goal is to treat everyone in that particular dataset who will respond, the type of multifactorial modeling we show in this paper will let you do that while treating many fewer people who won't respond," lead author Max Leiserson, assistant professor in the Department of Computer Science at UMD, said in the release, adding that the researchers were not just looking at the patient, but instead, at pecific marker of immune response, “which gave us a much better picture of what’s going on.”

The study showed that the computer model predictions could include as few as 38 percent of those who did not benefit from checkpoint inhibitor therapy, while still capturing 100 percent of the patients who did. In contrast, using current biomarkers, researchers would have to treat 77 percent of patients with non-durable clinical benefit to ensure that all who would derive benefit actually received treatment.

“People are realizing that predicting response is more and more appropriate and needed, and to be able to do this, the traditional kind of single biomarker approach isn’t always enough,” Leiserson said.

Moreover, the researchers discovered if they eliminated any one of the three categories of data from the model (tumor data, immune cell data or patient clinical data) the immune response was no longer predictable. For example, the model only predicted 23 percent of patients who would benefit from checkpoint inhibitor therapy.

“These features we identified may not be the only features that can be used to predict how a patient will respond,” Leiserson said. “There may be others that you could replace these with, but it’s about the method and the inclusion of all three categories of features.”

While this model is not ready in practice, the researchers plan to validate the model in a larger cohort of patients with bladder cancer, and to develop multifactorial models of durable clinical benefit from checkpoint inhibitor immunotherapy.

“When taken together, these data offer a tantalizing early look at how predictive models of the peripheral immune system, and predictive models of clinical response from peripheral immune system features, will play a role in helping to personalize immunotherapy treatment strategies and drug discovery,” the researchers concluded.