Oncologists Will Use AI To Deliver Personalized Cancer Care

There is so much data now available that clinicians do not have time to interpret all of it. 

AI will be able to analyze the information from disparate sources - testing, genomics, pharmaceuticals, patient history - distill it and recommend solutions to doctors. JL 

Andrew Ip comments in Scientific American:

Genomic testing makes it possible to genetically profile cancers and identify the specific mutations underlying individual tumors. Clinical studies are assessing new drug candidates or exploring new uses for existing drugs. In practice, oncologists are now confronted with a firehose of information that is too much to take in. As a result, cancer care is often not as personalized as it could be. The next step is to harness AI to sift through the genomics, drug trials, patient demographics and outcomes, and scientific research—and then deliver a clinical recommendation. This is known as a clinical decision support system (CDS).

We’re at the dawn of a new era of personalized medicine, in which treatment of each patient’s cancer is customized to their genetic makeup. Genomic testing makes it possible to genetically profile cancers and identify the specific mutations underlying individual tumors. Clinical studies are assessing new drug candidates or exploring new uses for existing drugs.

In theory, these findings should empower oncologists to find treatments that precisely fit each cancer and achieve better outcomes for patients. But in practice, oncologists are now confronted with a firehose of information that is too much to fully take in. As a result, cancer care is often not as personalized as it could be.

THE ONCOLOGIST’S CHALLENGE

As a hematologist/oncologist specializing in lymphoma, I know that it’s possible to obtain genetic and molecular details on each patient’s cancer, and that this information can offer potentially lifesaving clues. But we don’t always know which molecular diagnostic test is appropriate for each patient.

When we do obtain genetic sequence data, it’s often unclear what it means for a patient’s prognosis. It’s hard to select the most promising medicine when different studies point to different applications. For example, several inhibitors of the enzyme anaplastic lymphoma kinase (ALK) have proven effective in treating lymphoma, non-small cell lung cancer, and neuroblastoma, while other findings suggest one of the inhibitors can treat pediatric oncology patients.

Furthermore, some of us face challenges gaining access to genomic sequencing and getting precision medicines to our patients. Some health insurers do not reimburse for genomic testing or pay far less than the actual cost. Others are reluctant to pay for off-label drug use.

According to one recent study from Hackensack Meridian Health, oncologists in community settings who were treating non-small cell lung cancer patients had difficulty handling tumor samples, faced long turnaround times for laboratory tests, and had limited access to targeted therapies. To make it more difficult, next-generation sequencing results are often provided as a pdf, which cannot be digitally integrated with a patient’s electronic health records.

Meanwhile, as the science or precision medicine steams forward, oncologists struggle to keep up. Updates are coming too quickly for oncologists to read each one, or to learn of them all at an annual conference. It’s simply too much for an oncologist to analyze, interpret and apply, given that a cancer doctor’s working day is filled with patient visits and managing associated clinical documentation, administrative communications, and insurance paperwork. 
 

LEVERAGING GENOMICS

Facing these myriad challenges, oncologists must have strong support to deliver optimal cancer care and now, they’re beginning to get it. This support can help them access new data, understand it, and apply it to improve care for their cancer patients.

A recent case of a severely ill cancer patient at the Hackensack Meridian Health John Theurer Cancer Center, where I practice oncology, illustrates this progress. The patient had two separate biopsies taken. The analyses pointed to different conclusions, and one identified an incurable form of cancer. But the patient’s genomic sequencing results revealed a mutation revealing a very curable form of cancer. We treated the patient with a scientifically proven chemotherapy, and the patient quickly showed marked improvement.

To ensure our patients have access to treatment informed by genomic profiling, Hackensack Meridian Health and a partner organization, the Genomic Testing Cooperative, recently established an in-house genomic profiling center. The service analyzes 434 genes for solid tumors, searching for DNA and RNA mutations and chromosomal structural abnormalities. For blood cancers, the service generates a 177 gene panel hematology profile.

At HMH, we are also building a database that correlates genomic results, cancer types, cancer medicines and patient outcomes, and we expect it to help oncologists quickly analyze a patient’s cancer and use that information to improve care. Other major cancer centers are working on similar projects.

AUTOMATED EXPERTISE

The next step is to harness artificial intelligence to sift through all the data—genomics, drug trials, patient demographics and outcomes, and scientific research—and then deliver a clinical recommendation. This is known as a clinical decision support system (CDS). IBM Watson was designed to do this, but it was not up to the task. Other leading technology companies are building cancer treatment guidance software. Within a decade we should have a CDS system that could revolutionize cancer care.

Eventually, a clinical decision support system will integrate clinical data and National Comprehensive Cancer Network guidelines — a set of evidence-based recommendations meant to help health care professionals prevent, diagnose, and treat cancer. It will also integrate individual patient information and automatically submit it to insurance companies for approval. This would streamline one of the greatest administrative burdens confronting physicians.

For now, though, oncologists can get their best guidance from roundtable discussions with multi-disciplinary specialists—radiologists, pathologists, medical oncologists, radiation oncologists, and surgeons—known as molecular tumor boards. They review cases, assess all the available data, and provide treatment recommendations. During the COVID-19 pandemic, we have been conducting our conferences virtually, as have other health systems. Having a group of specialized experts analyze an individual patient’s molecular testing data and clinical charts and compare it with research data is hugely helpful for oncologists, especially those working in small community settings.

For all the power of technology-enabled treatments, dedicated health care professionals will always be needed. Compassionate doctors, nurses, nutritionists, and social workers can consider each patient as an individual, with not only physical needs, but also mental, spiritual, and social needs. Cancer patients do best when they maintain good nutrition, regular physical activity, a positive mental attitude, and have adequate social support.

This holistic approach to care, combined with targeted cancer therapies, is the true essence of personalized medicine, which will deliver increasingly effective cancer treatment.