When physicians are overwhelmed with more clinical trial data than they can digest, the natural default is to rely on conventional chemotherapy regimens. Like his fellow oncologists, Jeremy Warner, M.D., associate professor of medicine and biomedical informatics at Vanderbilt University Medical Center, understands that this default position is uncomfortable and untenable.
“There are over 100 common cancers and even more uncommon ones. Each has its own distinct treatment approaches, vulnerabilities to treatment, resistances, and treatment toxicity risks,” he said. “The question became, with all the effective new therapies emerging from the clinical trials pipeline, is there a technological means of helping clinicians capitalize on these?”
To provide just that, ten years ago Warner spearheaded an effort, now joined by a team of nearly 40 practicing physicians across the U.S. and Canada, to painstakingly assemble an online knowledge base, HemOnc.org. They categorize chemotherapy regimens along parameters such as type and stage of the cancer, and whether treatments are curative or palliative and indicated as a first, second or third line of therapy.
“What we are providing goes way beyond what you might see in conventional guidelines or an article you might read about how to treat a certain cancer,” Warner said. “We’ve now been able to take what is available on the web and create a computable ontology that can be used to power novel software tools.”
Importantly, one of these tools ranks treatment options in order of efficacy. “Some of these have never been directly compared. With this, clinicians can make optimally informed choices,” he said.
“With all the effective new therapies emerging from the clinical trials pipeline, is there a technological means of helping clinicians capitalize on these?”
A Computable Ontology
The user of this ontology can describe the specifics of a cancer and then access suggested treatment regimens based on the latest research and clinical trial data. “Our goal is to keep up with the literature and maintain this as an up-to-date knowledge base that practicing oncologists can rely on,” Warner said. This includes access to one of the largest catalogues – with more than 4,000 entries – of brand name chemotherapy drugs currently available.
Warner says the program has important distinctions from commercial software products on the market. “There are groups that do a broad scraping of data, using PubMed or ClinicalTrials.gov, for example,” he said. “But our team are all practicing clinicians who comb the literature and put our heads together to examine and curate these trials. There is no parallel yet for an expert human taking a look.”
Clinicians can utilize the regimen browser – a graphical user interface – to explore a network of thousands of treatment options and assess how individual regimens compare against others. “This gives them the all-important comparison that they need in decision-making. For example, the very simple regimen of single agent tamoxifen, used in breast cancer, has been compared to at least 52 other regimens in published clinical trials,” Warner said.
“Our team are all practicing clinicians who comb the literature and put our heads together to examine and curate these trials.”
An important component for HemOnc.org is the focus on toxicities. “Historically, toxicity is not nearly as consistently reported as efficacy. Yet, many patients stop treatment because of side effects, not because it’s not working,” Warner said. “With so many new drugs, new information on cardiotoxicity and other problems are cropping up at a rapid pace. We want to capture this.”
In the future, the ontology might also be integrated with EHRs to populate the chemotherapy prescribing component. While dosing guidelines for drugs such as antibiotics are routinely pre-populated in a new EHR system, Warner says chemotherapy dosing is not. “Using this provides a way of accessing valuable information that is traditionally ‘locked up’ in the free text of electronic health records,” Warner said.
The regimen browser and ranking algorithms are currently built out for breast cancer and multiple myeloma, with more in the near-term pipeline. Next steps include incorporating genomically informed cancers and chemotherapy cost factors into the mix. “About 30 percent of our website users are from outside the U.S. – 181 countries and territories, many of which are under-resourced. We don’t necessarily want to only include the newest and most expensive treatments,” Warner said.
There are currently more than 1,300 cancer therapies in development in the U.S. Clinicians will need to work hand in hand with the best augmented intelligence tools in order to sort through the options. Initiatives like HemOnc.org are here in the nick of time.