Medicine Meet Math | Commission on Cancer Oncology Lecture | Clinical Congress 2020
As a young surgeon I belonged to a generation of surgeons who were numerator surgeons. We knew and were influenced greatly by the satisfying successes, and the dramatic failures, but we had no idea of the denominator of either. We did not know the denominator for all those in-between, which were neither memorable successes, nor even more memorable failures. To change that we had to develop prospective databases. To do that we had to know what we needed to collect prospectively. That spawned many retrospective reviews, which reinforced the inadequacy of randomly collected information, and helped define what we would need to collect prospectively.
Prospective cancer databases are a rich resource for natural history. They enable us to define risk factors for recurrence and survival. Survival not just as an absolute (Overall) event but for disease specific survival, the latter critical to define efficacy of an intervention that is disease directed. Such databases allow for the identification of where a cancer is most likely to occur. By such identification, screening and follow up are more efficiently defined. They are a deep resource for defining outcomes based on specific histopathology. Nowhere is that more clear than soft tissue sarcoma, something that I have been interested in for the last four decades. Careful attention to comprehensive databases tells us about groups of patients but not about individual patients. When asked now, how many different types of cancer are there, I respond with the observation that the current and next generation of translational scientists will hopefully come to where the number of cancers is equivalent to the number of patients with cancer. This will incorporate not just the malignancy but the environment in which a cancer grows. We will define the impact not just on the cancer but on the cancer’s host environment, the patient.
Once we have a database and are aware of the natural history, we learn a great deal. We quickly learn that five-year survival is an arbitrary event. For some cancers, one- or two-year survival was all that mattered, in others like sarcoma, we see recurrences at 10, 15 and 20 years after the initial diagnosis and treatment, which at five years had been presumed curative. This then was the basis to appreciate that math was meeting medicine. We now have the tools to take these enriched databases and learn far more than natural history. We can now not just describe group or stage outcomes but outcomes for an individual patient. These initial attempts at predictive nomograms were crude but they have constantly improved as the data on which they are based improves.
Even more important for what has taken place since this lecture was delivered, is how such databases can be integrated into the societal impact of poverty and diversity. We are far more cognizant that cancer care is not just about treatment, but as much about the patient in whom the tumor arises or progresses. The differing outcomes based on race, ethnicity and lifestyle behaviors give us great insight into what we must address as we examine the role of cancer care within society.
We have long neglected the societal impact of unnecessary and ineffective treatment. We can now look to how much is too much. How personally rewarding to patient and surgeon is it to be able to say that the prognosis is sufficiently good that any attempt to improve such outcome, must be critically balanced against potential harm.
The application of math to medicine can tell us much more. We can predict the value or lack thereof of multiple diagnostic or imaging applications. I provide a classic example of our over diagnosis and over investigation of some patients with breast cancer. This too has a societal impact as we struggle with the unsustainable cost of current cancer care. It is not just new and progressive approaches but new and progressive approaches that do not involve over treatment as no treatment is ever without risk. We need to examine how many patients are we justified in treating with significant side effects for a very small numbers to benefit. An unpleasant but necessary exercise.
I describe some of the approaches that we have taken and some that are to be expected and not yet foretold. We must select patients at the time of diagnosis who are most likely to benefit from our intervention, we can no longer wait and waste time on long term outcome studies. We have no choice; medicine must meet and embrace math to determine what is best for the individual patient and society as a whole. Difficult choices are better made when the math supports perceived experience.