Can AI Tool Guide CDK4/6 Inhibitor Use in Breast Cancer?

SAN ANTONIO — A machine learning model, which incorporates both clinical and genomic factors, helps predict which patients with hormone receptor–positive (HR+)/human epidermal growth factor receptor 2–negative (HER2–) metastatic breast cancer will benefit from adding a CDK4/6 inhibitor to frontline endocrine therapy, according to new research. 

There is a huge need to identify patients who may or may not benefit from adding CDK4/6 inhibitors at the time of metastatic diagnosis to guide escalation and de-escalation strategies in advance, senior investigator Pedram Razavi, MD, PhD, said during a presentation at the San Antonio Breast Cancer Symposium 2024.

While the combination of a CDK4/6 inhibitor and endocrine therapy has improved outcomes in this patient population, responses to CDK4/6 inhibitors vary widely. 

Being able to better predict outcomes could also help some patients avoid unnecessary side effects and financial toxicity from escalated upfront approaches, added Razavi, scientific director of the global research program at Memorial Sloan Kettering Cancer Center in New York City. 

Razavi and his colleagues used a machine learning tool developed at Memorial Sloan Kettering (OncoCast-MPM) to compare three different models (clinicopathological, genomic, and combined) to predict progression-free survival outcomes when adding a CDK4/6 inhibitor to endocrine therapy. 

The models used clinical and genomic factors known to be associated with outcomes or resistance to either a CDK4/6 inhibitor or endocrine therapy. “All of these variables are potentially available when the patients are diagnosed with metastatic disease, making such machine learning models broadly applicable,” Razavi noted in a press release.

The clinicopathological features model included tumor grade, location of metastases, and prior treatments. The genomic features model included tumor mutational burden and gene mutations such TP53, RB1, PTEN, among others. And the combination model included both clinical and genomic features. 

The researchers developed the models using a training cohort of 761 patients with HR+/HER2– metastatic breast cancer who received first-line endocrine therapy with CDK4/6 inhibitor combinations and had tumor genomic analysis before or within 3 months of starting treatment. The models were then tested in a cohort of 326 patients.

During the training phase, the clinicopathological features model identified three risk groups (high, intermediate, and low), with median progression-free survival ranging from 6.3 months in the high-risk group to 24.5 months in the low-risk group (hazard ratio [HR], 3.95).

The genomic features model achieved stratification in three groups with progression-free survival ranging from 9.9 months in the high-risk group to 23.1 months in the low-risk group (HR, 2.86).

The combined clinical and genomic model delivered results with four risk categories (including two intermediate groups as well as the low and high-risk groups), with median progression-free survival ranging from 5.3 months in the high-risk group to 29 months in the low-risk group (HR, 6.45), with the two intermediate groups demonstrating progression-free survival durations of 10.7 months and 19.8 months, respectively. 

During the testing phase, the models yielded nearly identical progression-free survival and HR results, indicating the models are robust.

All three models performed well, surpassing the conventional clinical risk models based on a single or a few clinical features. “But the power of the analysis shone when we started combining the clinical and genomic features together,” Razavi said in the press release. 

Given the limitations of the study, including single-institution design, retrospective data analysis, and potential referral bias associated with specialized cancer centers, the researchers plan to validate the clinico-genomic model using external cohorts to ensure “robustness and generalizability,” Razavi said at a press briefing. 

Commenting on the data, briefing moderator Carlos Arteaga, MD, noted that the response to CDK4/6 inhibitors is not black and white. 

“Some patients respond really well and some patients don’t respond at all, and there are all the in-betweens,” and this study suggests that we’re going to have predictive models soon that can help patients and physicians better understand how helpful a CDK4/6 inhibitor will be for them personally, said Arteaga, with Simmons Comprehensive Cancer Center at UT Southwestern Medical Center, Dallas.

The study was partially and indirectly supported by the National Institutes of Health, Department of Defense, Susan G. Komen for the Cure, Breast Cancer Research Foundation, AstraZeneca, Sophia Genetics, Novartis, and Tempus. 

Razavi has disclosed relationships with Novartis, AstraZeneca, Pfizer, Lilly/Loxo, Tempus, Prelude Therapeutics, NeoGenomics, Natera, SAGA Diagnostics, Paige.ai, Guardant, Myriad and Foresight. Arteaga reported relationships with Pfizer and Lilly.