AI Has Significantly Advanced, Changed the Management of Colorectal Cancer

Over the past decade, there has been a paradigm shift in how we manage different malignancies, including colon cancer, explained Arturo Loaiza-Bonilla, MD, MSEd, FACP, during his keynote presentation at the 2023 NCODA International Spring Forum.

Loaiza-Bonilla, medical director of oncology research at Capital Health, advisory committee member - Government Relations Committee at the American Society of Clinical Oncology (ASCO), co-founder and chief medical officer at Massive Bio, president of the Pennsylvania Society of Oncology and Hematology, and teaching faculty in the Department of Medicine at Drexel University College of Medicine, explained further that many of the changes have occurred because of a greater level of knowledge about biomarkers for each cancer type.

Today, the traditional morphological and pathological anatomical classification of cancer is no longer seen as being able to accurately predict biological and clinical behavior, prognosis, or response to treatment, according to Loaiza-Bonilla. Further, he explained that it has also become important to not only identify clinical stage, but also use biological or molecular markers that determine outcome (i.e., prognosis) and therapy (i.e., prediction).

Arturo Loaiza-Bonilla, MD, MSEd, FACP, discusses the potential of AI to transform cancer care during a presentation at the 2023 NCODA International Spring Forum.

“The paradigm that we used to use, which was based on location, for example colon cancer, as we're [sic] discussing today, now has become a little bit less relevant because we're looking at, for example, side effects, and we’re looking at different biomarkers themselves,” Loaiza-Bonilla said during the presentation.

This shift in the management of malignancies came about because of advances in biomedical technologies that have come to market in recent years, according to Loaiza-Bonilla. These advancements first made waves with the advent of ImmunoHistoChemistry (IHC) for breast cancer, which quickly became relevant to practice in the field. This biomarker technique was used to detect the level of protein expression.

“But now we're starting to look at HER2, and that evolved into [fluorescence in situ hybridization (FISH)]. Now we have next-generation sequencing, and that is becoming more and more prominent, and actually cheaper by the day,” Loaiza-Bonilla said. “The Human Genome Project was like, ‘For how many years [this was] millions of dollars, now we can do [sic] genomic testing for 100 bucks.’ So that's really changed a lot of things in this space.”

In the future, Loaiza-Bonilla explained that we are moving toward biomarker multi-omics analysis, which is a large-scale biomarker analysis currently underway using plasma and tumor tissue samples collected pre- and post-treatments (NCT02394843). Loaiza-Bonilla noted that potential biomarkers on outcomes will be reported in upcoming meetings.

“Biomarker multi-omics analysis will require us to look for secluding tumor DNA for these patients, and also new RNA signatures in analyses are going to be key for this,” Loaiza-Bonilla said.

However, Loaiza-Bonilla explained that, although he’s a “passionate enthusiast” for using AI in cancer care, there are many questions remaining about the “how” of its use in health care. For example, there are a lot of language models coming up that are now being trained to answer things as if they are humans themselves, but with a greater capacity to apply a breadth of data from the web beyond human capacity.

“GPT-4 now has access to the web. So you can tell them, ‘Can you tell me who Arturo Loaiza-Bonilla is,’ and it actually—they may make up a few things sometimes, such as where I graduated—but they're pretty much right on the whole thing,” Loaiza-Bonilla said. “Basically, you can ask them, ‘Okay, I have this patient with this disease [on this drug]. What are the expected side effects?’ Or ‘How do I handle this patient?’ It’s just become like Dr. Google, but this is Dr. Microsoft Open AI on steroids.”

Yet, Loaiza-Bonilla noted that questions remain regarding how this software could be used in a more focused way for cancer care. According to Loaiza-Bonilla, oncology has experienced 2 major technological evolutions: the molecular evolution and the evolution of big data. However, a limitation of the latter is the sheer quantity of data available to be analyzed from numerous sources, which have yet to be well organized.

“Now we're [asking], ‘Can we use AI for this?’” Loaiza-Bonilla said. “The benefits of using AI technology in colorectal cancer and beyond is becoming relevant.”

Loaiza-Bonilla noted that data from MIT showed that cancer diagnosis can be significantly improved using AI as well, with real-time data updates, personalized attention, and ultimately better results at lower costs. Additionally, Loaiza-Bonilla explained that it will be important to look into using AI to find the better coverage for patients, identify things missing from biomarker testing, or identify whether a cancer center is missing a certain drug for its patients.

Now, using optical character recognition, Loaiza-Bonilla explained that it has even become possible to use a machine learning approach when looking at drug sequencing for a patient with colorectal cancer.

“You put the data in on the patient, and depending on what you use first or second, it is going to tell you the potential better option for the patient in terms of sequencing. So, for example, it’s telling us that the overall survival when you use FOLFOXai can be optimized depending on what genomic alterations the tumor has,” Loaiza-Bonilla said.

Additionally, Loaiza-Bonilla noted that deep learning has become possible for a tumor board. This is the work he started doing at Massive Bio, the company he co-founded.

“We started with genomics,” Loaiza-Bonilla said. “Basically looking at all the genomic alterations and aligning them up with patients’ data and providing some input in terms of what are the next steps and start from treatment—targeted or clinical trials. We use it through an app, which is free for every patient to use. So this is not a pitch; this is free. Anyone can use it at any time. Any patient with cancer can get access to it, the same as any provider, and it tells them what clinical trials are in proximity within a 50-mile radius, as well as the potential treatment options for them.”

Reference

Loaiza-Bonilla A. Attacking Prostate Cancer: Latest Updates for Treating Patients. Presented at 2023 NCODA International Spring Forum; March 16, 2023.