Gauging the future of clinical pathways in oncology
November 4, 2024Amidst advances in therapies and technology, the core need for clinical oncology pathways remains.
By Dr. Andrew Hertler
About a decade ago, I left community oncology after 30 years in practice and joined a company that curated high-value clinical pathways. The need for pathways was clear. The pipeline of new anticancer therapies, combined with the relentless pace of new data about their efficacy and toxicity, meant that our field must continuously reassess which options work best for various treatment scenarios. The task was more than an individual provider or practice could take on.
Oncology has only grown in complexity since then. The cancer armamentarium keeps expanding, addressing new indications with more targeted and personalized therapies. But we are also in the midst of a sea change in information technology, notably the rapid maturation of artificial intelligence. It’s natural to wonder how we might harness AI’s power to sift through massive troves of data to help identify preferred regimens.
So, I was interested to hear how my peers in the field were thinking about these trends when I recently attended the Association of Value-Based Cancer Care’s annual meeting in New York. I was a panelist for two discussions, including one that asked, "Have clinical pathways outlived their usefulness in the world of AI?" I left with a few takeaways.
Oncologists need clinical pathways as much as ever, if not more.
So far this year, the National Comprehensive Cancer Network has issued more than 200 updates across its 90 guidelines, which inform pathways. And those guidelines are not easily digested. As one fellow panelist told the audience, they do not have the time or resources to keep up with it. Some community oncologists try to keep up by listening to podcasts or reaching out directly to their peers in academic medical centers, according to a recent Wall Street Journal article. I left with little doubt that pathways will continue to play a critical role, if not a larger one, as genetic testing and precision medicine helps us tailor treatments. Which brings us to the next takeaway…
Artificial intelligence has the potential to make pathways even more granular.
Today’s high-quality oncology pathways should already support the delivery of precision medicine. For example, they should encourage the use of genetic testing to guide regimen selection. Yet, AI might help us make treatment recommendations that are even more tailored to the patient’s specific case. For example, we know that some patients respond well to immune checkpoint inhibitors, while others do not. Yet oncologists struggle to predict with any confidence which of those two groups a given patient belongs. However, one machine learning model has proven adept at identifying those patients who are highly likely to benefit, based on a combination of clinical data points. By spotting patterns in data that typically escape clinicians, such tools may point the way towards more effective pathways and better drug selection.
Still, the design and curation of pathways remains a human endeavor.
Recently I wrote an article arguing that AI will not replace human judgment in the final decisions about which regimens belong on pathway. I was pleased to find broad agreement among fellow panelists on this topic. While AI could dramatically reduce the amount of time spent gathering and organizing clinical data, relying on data alone often results in no clear winners. Pathway decisions often come down to clinical judgment calls — for instance, whether a minor increase in overall survival is worth a more significant risk of serious side effects. And often the decisions are not that clear cut. Questions about the quality of evidence, lack of comparative effectiveness data, and other issues muddy the picture. Panels of experts — oncologists, pharmacists, and others — are best equipped to debate the issues and develop consensus (though unanimous agreement often remains elusive).
In summary, AI is already facilitating the rapid collection and analysis of the clinical data needed to build oncology pathways. It reveals patterns in data that ofttimes escape human analysis. However, AI should not replace humans in a pathway design process that has an element of values and goals that will differ even amongst the best of clinicians.
About the Author