The role of AI technologies in precision oncology
In light of the recent understanding of the molecular complexity of cancer and the costs of corresponding targeted treatments such as immunotherapies and CAR-T, there are legitimate questions that arise regarding precision oncology. Are community cancer patients getting the best possible care? Are pharmaceutical companies optimising the market potential for these expensive therapies?
A recent study indicates that the answer to these questions is an unqualified “no”.1 The results, published in JCO Oncology, indicate that only 36% of eligible lung cancer patients received targeted therapies. Clearly, this is a disservice to the patient. At the same time, it also indicates that two thirds of this market is not being reached by the pharmaceutical companies that make these drugs. As with many issues in healthcare, better data, specifically clinical and market access data, can make a significant difference.
Barriers to delivering precision oncology
There are two underlying reasons for this situation. One is the circumstances of the typical community oncologist, and the other is lack of market insights available to drug manufacturers. The majority of cancer care in the US, about 80%, is provided in community clinics, not in academic medical centres or NCI accredited centres. The oncologists practicing in these centres see 20 to 25 patients a day and, unlike academic practitioners, don’t specialise in a specific type of cancer. As such, they deal with a variety of cancers, each of which are molecularly different, with different targeted therapies applicable to each. Further, the molecular tests needed for specific cancers are also different, complicating the task of knowing which test to order.
In the realm of drug manufacturers, the market information available to them is limited, in that they don’t actually know who is getting their drugs due to privacy regulations. During COVID, the access to physician offices by pharmaceutical representatives was eliminated, and many physicians have maintained these policies. This has led to lack of direct contact with oncologists, resulting in lack of insights on the oncologist’s knowledge about new drugs, molecular testing protocols, and the doctor’s rationale for choosing one drug over another. Essentially, the pharma industry doesn’t have the information needed to drive greater adoption of its products.
Emerging solutions: AI and data at the point of care
The use of AI in oncology has tremendous potential to improve many facets of cancer care, including diagnosis, drug discovery, clinical trials, and treatment decision making. Companies such as OncAI and Ataraxis are focused on diagnosis for specific cancers, while Ryght AI is improving the clinical trial process.
In the treatment decision area, AI agents can collect and organise all of the relevant information that could inform an oncologist’s treatment plan. These agents act to identify the relevant molecular diagnostic tests for the patient’s cancer, and integrate that data with the formulary and coverage information from the patient’s payer. At the point when the test results are available, additional agents create summaries of the molecular and pathology test results. Combined with an AI generated summary of the patient history from the EMR, AI agents can provide a comprehensive, accurate, and user-friendly view of the patient’s situation.
Once a treatment plan has been formulated, the oncologist’s rationale for the treatment can be discerned from the physician note and a brief set of directed queries to the oncologist. The resulting insights lead to the answers to important questions: Was a molecular test ordered, and was it the right test? Were the results received in a timely manner? Did the test results influence the treatment decision? Was patient cost a factor in the decision? Did the payer influence the decision in an appropriate or inappropriate manner?
This information is aggregated across physicians and practices to provide drug manufacturers with specific physician, practice, and payer insights, which enable them to address market access issues. These issues could be as simple as not being listed on a payer formulary or not being a “preferred” drug for a specific oncology practice. The insights are also valuable for practice management to understand the practice’s uniformity of care and appropriate use of treatment guidelines.
Oncology focused AI technologies have the potential to remedy the current situation where only 36% of eligible patients are receiving targeted therapies. By focusing on assistance to the front-line oncologist, AI can enable improved diagnosis, clinical trials, and timely, accurate, and efficient decision making on the part of the oncologist. All of these aspects directly benefit patients with treatment plans that address their cancer with the most effective possible treatment, leading to improved patient outcomes. Market access insights directly from the front-line clinics and improved decision making in the clinic will not only benefit patients, but will also lead to greater market penetration for drug manufacturers.
References
- JCO Precis Oncol 6, e2200246(2022), Volume 6, DOI: 10.1200/PO.22.00246
About the author
Dave Parkhill is the CTO and co-founder of OncoRx Insights. He was formerly a member of the start-up team at Altais, a Blue Shield of California spin-off, and was previously with Global Healthcare Exchange, Hitachi Consulting, IBM Global Services, and Andersen Business Consulting. He has extensive healthcare experience with both provider and payer organisations and has founded two previous start-up companies.
