Grace Lu-Yao, MD
Not enough is known about older populations of patients with cancer. They tend to be frailer, they have comorbidities, and they may be on multiple medications. Yet when it comes to treating them for cancer, it’s often the case that the only available data are based on trials of younger patients in far healthier condition.
All of that is about to change, says Grace Lu-Yao, MD, a pioneer in the use of big data to gain clearer understanding of how patients with cancer are affected by treatments. Lu-Yao, for example, worked with the Prostate Patient Outcomes Research Team on a large-scale study to document the treatment patterns of prostate cancer and to understand the value of conservative management for low-risk prostate cancer. The finding, first published in the Journal of the American Medical Association
was so revolutionary at the time that it was initially resisted by the cancer community, which had difficulty accepting that quality-adjusted life years saved following conservative management of prostate cancer could be roughly equivalent to surgery or radiation therapy.
Clinical data-sharing partnerships established this summer by CancerLinQ, the FDA, and the National Cancer Institute (NCI) may bring more discoveries like this. Under these agreements, vast amounts of patient data—essentially real-world evidence (RWE)—will be shared among these institutions and the many cancer treatment centers operating across the United States. What happens in the community clinic, for example, will be relayed to the FDA to help inform drug policy decisions; and what is gathered and stored in the NCI’s Surveillance, Epidemiology, and End Results program will be available to far-flung clinics that need to know how a certain drug will affect geriatric populations—something that only recorded data gathered from multiple cancer centers may show.
“The difficulty we have is that most evidence, the gold standard, comes from clinical trials,” said Lu-Yao, associate director for Population Science at the Sidney Kimmel Cancer Center at Jefferson University in Philadelphia. “Yet clinical trials a lot of the time really are quite stringent. Their criteria are such that about half of elderly patients will not be eligible for clinical trials. Therefore, the result from clinical trials may not be applicable to these so-called vulnerable or frail populations.”
Lu-Yao, a widely respected epidemiologist, said that population science is made for this new wave of RWE. Many of the studies she has done have been based on analyses of large data pools, and her insights into prostate cancer treatment exemplify how investigation of existing information can be as useful as active trials that involve tests of drug efficacy. In other words, much information already exists. It just needs to be crunched and interpreted and put to good use.
More data are exactly what is needed to improve geriatric oncology, agrees Hyman Muss, MD, who directs the Geriatric Oncology Program at the University of North Carolina (UNC) Lineberger Comprehensive Cancer Center. Checkpoint inhibitors are so new that not enough is known about how they should be used in older populations. “Very few older patients have been on these studies, so we really don’t know how they fair with these treatments, especially when it comes to fatigue and functional change. I suspect there are not going to be dramatic differences in nausea and vomiting and other adverse events, but to take an older patient who is barely making it in the community [because of existing health problems] and add profound fatigue or some muscle weakness is going to really change that patient's life, and we know very little about it,” Muss said.
The shortcomings of clinical trials can be remedied in various ways by the broad sweep inherent in big data analysis, Lu-Yao explained. Rare adverse events that would not be spotted in smaller patient samples will be recognized and cataloged in the larger analysis. This can remove more of the uncertainty from the treatment process.
Lu-Yao has a special interest in developing risk-assessment tools that incorporate aggregated patient data and clarify the potential problems from a given medication. At Kimmel, she and others are building the first race-specific prediction model for prostate cancer. This would incorporate biomarkers, comorbidity status, and all clinical prognostic factors so that physicians can tell patients how likely they are to have disease progression or die from competing causes. The tool would enable doctors to judge whether aggressive treatment is a good option for a patient.