This article is more than 5 years old. Information may no longer be current.
The denominator problem in clinical oncology
I have a fairly typical academic oncology practice.
As a bone marrow transplant physician, I primarily see patients who have acute or chronic leukemia, lymphoma or myeloma, and I assess their candidacy for autologous or allogeneic transplant.
I provide care throughout the transplant continuum, from initial consultation through the peritransplant period and into long-term follow-up.
Many of my clinical and research efforts focus on improving care and outcomes for patients like the ones my colleagues and I see each day. Recently, though, an important question has come up: Who don’t I see in clinic, and what does that mean about the care and outcomes for those I do?
In my opinion, the “denominator problem” encompasses and crystallizes many of the most pressing issues in contemporary oncology.
In academic practice, and more generally in the current era of rapid advances in biotechnology, our universe is constrained in ways we may not appreciate — in other words, by the old psychological maxim, “What you see is what you get.”
As we think about what big data truly can offer the oncologic community, it becomes clear that insights into the denominator problem may become one of the most important contributions of the big-data movement.
Here are a few examples of the challenges to be solved as part of the denominator problem:
(Im)precision medicine
In a national analysis of Medicare data, Reeder-Hayes and colleagues found that only 50% of white women and 40% of black women with HER-2–positive breast cancer received trastuzumab (Herceptin, Genentech) therapy, an effective and evidence-based “precision” treatment for this disease.
Additionally, racial disparities existed: After adjustment for other relevant factors, black women were 25% less likely than white women to receive trastuzumab within a year of diagnosis.
If we can’t deliver known “precision” therapies to half of the women we think would benefit from them, we should be very cautious proceeding into the territory of universal genomic testing, where the evidence is much less clear, and where — as the evidence is developed — disparities in delivery of precision therapies are likely to be amplified.
Surveillance analyses of real-world big data will be needed to ensure that dissemination — and equitable dissemination at that — keeps pace with scientific progress.
Skyrocketing costs of care
The issue of rising cancer costs has been discussed in depth over the last several years, with the term “financial toxicity” entering the lexicon. Even against this background, recent findings from Dusetzina and colleagues were striking.
The researchers compared new oncology products launched between 2000 and 2010 with those launched after 2010. Mean monthly spending during the first year on the market increased by 63% across all products, reaching $9,013.
When costs rise this high, who is prescribing these drugs and who is receiving them?
We have evidence and anecdotes about coinsurance, financial assistance and other interventions to soften the out-of-pocket impact of these costs upon patients. We also have data about personal bankruptcies and other evidence of financial toxicity.
But what does access look like from a microgeographic perspective, and what factors influence this? We are at risk for a major equity issue in access to new cancer therapeutics, and big-data analyses will be needed to protect the social good.
Accurate appraisals of intensive technologies
We no longer consider age to be a primary limitation for autologous or allogeneic hematopoietic cell transplantation. In a systematic review and meta-analysis of older patients with acute myeloid leukemia, Rashidi and colleagues reported a 3-year RFS of 35% in individuals aged older than 60 years.
Though these patients are representative of who we see in clinic each week as transplant physicians, an accompanying editorialist pointed out that these transplants account for a tiny fraction of individuals in this age category with AML during this time period.
Given this limitation, how can one draw generalizable conclusions from this small slice of the universe?
In academic practice, what we see is what we get. Similar issues exist in autologous transplantation for myeloma, where a modest proportion of transplant-eligible patients actually are evaluated for transplant. Who refers patients for transplant, and why?
Large-scale data and emerging techniques, such as social network analysis, may start to provide answers to these questions. To get a handle on the true denominator, how can we ensure that patients potentially eligible for transplant actually are evaluated so we can have informed discussions with eligible patients and better refine our methods for selection of appropriate candidates?
Solving this problem likely will require a multistakeholder approach involving patients, providers and payors. In the emerging era of T-cell–based immunotherapies, issues of access and feasibility will become amplified.
Trial extrapolation and performance status
The issues that fall under the denominator problem include long-known health services research challenges. An example: Are the results of cancer clinical trials generalizable in the real world?
There also are emerging challenges. For example, are clinician-assessed Karnofsky and ECOG scales the right way to assess performance status in real-world oncology, and to help with all of the treatment decisions that are affected by performance status determination? I’ll address these and other topics in future columns.
Conclusion
We are reaching a critical inflection point in oncology as the pace of technological innovation and development increases.
We should not forget that the original moonshot — putting a man on the moon — was so remarkable because it affected the collective national consciousness and inspired Americans with an idea that determination and scientific progress could make all things possible.
For the national cancer moonshot to do this, we must ensure that the benefits — and risks — of advancements in cancer research and care are understood as they affect all Americans, wherever they may live.
This is the challenge that big data must address.
References:
Dusetzina SB. JAMA Oncol. 2016;doi:10.1001/jamaoncol.2016.0648.
Gale RP. Biol Blood Marrow Transplant. 2016;doi:10.1016/j.bbmt.2015.12.029.
Rashidi A, et al. Biol Blood Marrow Transplant. 2016;doi:10.1016/j.bbmt.2015.10.019.
Reeder-Hayes K, et al. J Clin Oncol. 2016;doi:10.1200/JCO.2015.65/8716.
Subbiah V, et al. JAMA Oncol. 2016;doi:10.1001/jamaoncol.2016.0078.
West HJ. JAMA Oncol. 2016;doi:10.1001/jamaoncol.2016.0075.
For more information:
William Wood, MD, is an assistant professor of medicine in the division of hematology/oncology at University of North Carolina in Chapel Hill. He also is a HemOnc Today Editorial Board member. He can be reached at UNC Health Care System, Division of Hematology and Oncology, 101 Manning Drive, Chapel Hill, NC 27514; email: wawood@med.unc.edu. You also may follow him on Twitter (@WoodBD).
Disclosure: Wood reports no relevant financial disclosures.