When Reiko Horst was diagnosed with early-stage breast cancer, she knew next to nothing about the disease. To help determine the best treatment plan, her medical oncologist at Fred Hutchinson Cancer Center recommended doing a test called Oncotype DX to help predict Horst’s risk of the cancer returning and the benefit of chemotherapy.
A score under 25 indicates a low risk of recurrence for a patient who opts for hormone treatment and minimal benefit from chemotherapy. A score over 25 suggests a higher risk that the cancer may come back. Horst scored a 20. And that’s where her confusion began.
The test, which genetically analyzes a small portion of a patient’s tumor, is designed to give patients and their physicians more clarity about how to proceed with treatment. But Horst, as many patients do, felt overwhelmed and uncertain about how to interpret her score and its associated risk percentages.
We are living in an information age where seemingly everything is quantified: sports scores and their accompanying statistics that drill down into the finest points of the game, the keywords that entice readers to click on an article, the percentage chance of rain every hour. But just because we can assign numbers to everything doesn’t mean that the data generated always provide clarity for patients.
“When I first got my Oncotype, that was the first time I felt stuck in terms of what do I do,” Horst said.
Meeting with her medical oncologist, Jennifer Specht, MD, helped Horst make sense of the data. Specht explained that Horst’s results indicated that opting for chemotherapy would provide a 1% benefit over deciding against it. That helped Horst make the choice to move forward with a lumpectomy and radiation without chemotherapy because she deemed the potential side effects and health implications of not feeling good from chemo were not worth the tiny upside.
Next, she had to decide whether to go with a more effective treatment regimen that would put her into menopause or one that would be less effective but wouldn’t plunge her into chemically-induced menopause. Again, Specht shared percentages, but it was the personal conversations between doctor and patient that helped Horst the most.
At age 48, she figured she wasn’t that far from natural menopause anyway so she chose the more effective treatment, enrolling in a clinical trial of a shot similar to leuprolide (Lupron) administered every three months, plus a daily aromatase inhibitor pill.
"If I looked at all the risk statistics and information available alone, I would be buried in numbers," said Horst. "I reviewed the data and weighed my options. Dr. Specht took great care in getting to know me and what was important to me, so I put my trust in her. She supported me and made it more personal by saying, 'If you were my sister or good friend, this is the path I would recommend,'" Horst said. "I value her experience."
Horst’s decision-making process illustrates that despite an increasing amount of data available to people with cancer, it’s common for patients to lean heavily on their providers’ guidance.
"I appreciated the numbers to help me make a decision,” said Horst. “But I really do feel the relationship that Dr. Specht established with me throughout my breast cancer journey was the biggest factor in my trusting her opinions and expertise.”
Specht, who holds the Jill D. Bennett Endowed Professorship in Breast Cancer at UW Medicine, recognizes that many patients feel like Horst do. In her Fred Hutch profile, Specht notes that options for treating cancer have become increasingly complex, stating: “My job is to make you an expert in this field, sharing as much information as you would like so that you feel confident and equipped to manage your journey.”
The field of breast cancer has evolved to support a wide range of data-driven tools. There are prognostic markers that indicate how likely it is for breast cancer to recur; there are predictive markers that indicate how likely it is that a specific treatment will work. Explaining them to patients can be challenging. “I have hard conversations with patients who’ve completed treatment,” Specht said. “They are cancer-free but know cancer can recur years later and want to know: how can I estimate that risk for them?”
“People handle access to more data and more information very differently,” she said. “Patients at our center tend to want more information, and that may be part of why they choose a specialty cancer center like Fred Hutch.”
Hannah Linden, MD, FACP, clinical director of the Breast Cancer Program at Fred Hutch, has a joke she likes to tell patients: “The lottery is widely successful in this country because people don’t understand probability.”
Many cancer patients tend to think they’ve lost the lottery, Linden said, so they’re willing to try anything. But Linden works to help them understand the data surrounding their particular situation by using graphs or dot plots.
“A patient hears cancer and that’s it — they are just terrified," she said. "Whatever we say we can do to help them they are going to do. That gives us way too much power and authority and we need to tone it down. It’s important to give them a graphical picture and talk it through.”
Linden uses data in combination with an assessment of a patient’s current quality of life and what their goals are to recommend a course of treatment. “If a patient is stage 1, chances are the surgeon cured them. For a patient with advanced disease, I don’t want to extinguish hope, but I do want to be realistic.”
Many of Linden’s patients don’t ask about numbers.
“The whole numeracy thing is hard,” said Linden, who holds the Athena Distinguished Professorship of Breast Cancer Research at UW Medicine. “I always offer to talk about it because I want them to understand. I offer to pull up a graph. I try to encourage people to look at it. But I don’t discuss it if they don’t want to, or I offer to talk in generalizations. In science, we have to figure out a way to explain things better so patients understand what the numbers mean.”
As chief data officer at Fred Hutch, Jeff Leek, PhD, is working on that. It’s his job to make sure physicians have the data they need to make recommendations.
“Increasingly, cancer care is about data, odds and risk,” said Leek, whose team works on determining the odds associated with genetic profiles and imaging data, among other factors. “Our job as data professionals is to make sure they have the best data possible so they can leverage their knowledge to make the data understandable and translatable.”
Still, interpreting data on an individual level can be difficult, even for a data expert like Leek.
“There’s tons of research that shows that humans are bad at making decisions under uncertainty,” said Leek, who holds the J. Orin Edson Foundation Endowed Chair. “Even if you perfectly understand risk, you also need to understand your particular value judgments. How I understand what it means to have a 2% risk of death from cancer may be different from how you understand it, depending upon if you have a history of cancer or if you have kids. Each person has their own personal decision-making system.”
Further complicating the situation is the relatively meager training in data and statistics that medical students typically receive. Yet doctors need to have a thorough understanding of data science considering that 50% of sentences in abstracts published in major medical journals between 2010 and 2019 contain terms that require statistical training to comprehend, according to a Center for Open Science paper that Leek co-authored in 2020.
But before doctors need to make sense of the data, they need to be able to get their hands on it. And that’s not always easy. Cancer registries don’t track recurrence endpoints, says biostatistician Ruth Etzioni, PhD, who holds the Rosalie and Harold Rea Brown Endowed Chair at Fred Hutch.
“We have a hard time predicting recurrence or metastasis because there’s no population data,” said Etzioni, who is working with data to model recurrence and metastasis in prostate cancer. “That’s one of our big projects.”
Actual data aside, Etzioni believes that patients can misunderstand what data reveals. For example, patients may ask what’s the chance that they’ll die from their cancer.
But as Leek points out, the answer may depend upon other factors. “If I’m predicting am I going to die of cancer in 10 years, am I considering this in terms of if I don’t die of anything else? If I’m very sick with other conditions, the chances I get 10 years are low, so the chance of dying from cancer is also low. Those are two different predictions.”
Eric Holland, MD, PhD, envisions using big data to generate and fine-tune recommendations for individual patients. A neurosurgeon and senior vice president and director of the Human Biology Division at Fred Hutch, he helped develop Oncoscape, a tool that allows researchers to build a “map” that aggregates patient data and generates a picture of how other patients with similar cancers have responded. In theory, patients could pinpoint themselves on the map and find their nearest “neighbors,” allowing them to explore treatments that other patients have chosen. Eventually, Holland, who holds the Endowed Chair in Cancer Biology, hopes Oncoscape will transition to clinical use.
Finding the raw data has required cooperation from collaborators. More data yields more predictive value, said Sonali Arora, a computational biologist working with Holland.
“The more patients we can get onto the map, the more we can better define sub-clusters,” she said.
Holland is casting a wide net. “For meningioma, we’re going all over the planet to get it,” Holland said. “Once you get enough samples from around the world, it ensures that your data is representative of humanity. You realize that a lot of cancer subtypes happen to humans everywhere.”
Are you interested in reprinting or republishing this story? Be our guest! We want to help connect people with the information they need. We just ask that you link back to the original article, preserve the author’s byline and refrain from making edits that alter the original context. Questions? Email us at email@example.com
Bonnie Rochman is a staff writer at Fred Hutchinson Cancer Center. A former health and parenting writer for Time, she has written a popular science book about genetics, "The Gene Machine: How Genetic Technologies Are Changing the Way We Have Kids—and the Kids We Have." Reach her at firstname.lastname@example.org.