Our Blog

A New Look at Data in Veterinary Oncology

March 18, 2020

Background

Ethos Discovery seeks to deliver curative outcomes (aka survivorship) for pet animals with cancer. This is somewhat of a revolution since the term “cure” has been avoided in the field. It has been avoided so much that we do not document or recognize curative outcomes occur when they occur. As a first step towards delivering curative outcomes and raising survivorship a change in perspective and our assessment of existing data is needed. So we understand survivorship rates for common cancers and then are liberated to develop strategies to raise these rates of survivorship.

Clinical outcomes for veterinary cancers are often determined from Kaplan-Meier survival data. Rather than processing the full complexity of these data, many veterinary clinical fields, most note-worthy veterinary oncology, have focused on a single data point in the description of these complex data. This single data point is most often the median. This narrow approach falsely presumes that the median to be a more important point of data within these longitudinal studies than any other single point. Furthermore, this approach conveys the false perception that survivorship does not occur, and then perpetuates the problem by misdirecting clinical trial designs to prioritize improvements in the median rather than the distinctive and more important goal of raising survivorship. These shortcomings in the assessment of clinical data were eloquently summarized by the Evolutionary Biologist, Stephen Jay Gould in his essay describing his own cancer diagnosis, entitled “The Median isn’t the Message”. It is important to note that this perspective on the median is not only an impediment to progress and innovation, it is necessarily misaligned with the interests and needs of the client whose pet has a cancer diagnosis. A broader view of these complex Kaplan Meirer data and specifically looking beyond the median refers to the long-term outcomes seen at the right side plateau of a survival curve.

Ethos Discovery has adopted this shift in perspective towards curative outcomes in veterinary oncology and have begun to look beyond the median, and deliver a therapeutic strategy to improve curative outcomes and survivorship. Indeed, we have initiated comprehensive analyses of survivorship in canine osteosarcoma and canine hemangiosarcoma (through met-analyses) that will be submitted for publication in late 2020. We hope this approach has traction and will incentivize individuals and veterinary cancer-related organizations and journals to begin to report survivorship and the develop programs to improve survivorship.

The shift in Perspective is not simple: There is a long history of training that has avoided the curative data in front of our eyes for decades

The median is a simple number to report, publish, and remember. It has therefore been the focus of training programs in veterinary oncology, curiously even more so than similar human fields, i.e. pediatric oncology. It is very likely that every veterinary oncologist can cite a Median Disease Free Survival in osteosarcoma following surgery and chemotherapy. Such a median in pediatric osteosarcoma is not something known or of any interest to a Pediatric Oncologist; however, the 5-year survival and need to improve it is well known in this community.

Beyond our training, there are other reasons for veterinary oncologists to fear the use of the word “cure” in a client discussion. There are weaknesses to looking beyond the median, and these weaknesses should be recognized and understood rather than ignoring these valuable data at the right-side of the survival curve. A recognized weakness is the difficulty with long term follow up i.e., 1-2 years) in prospective studies and the continued over representation and reliance on retrospective studies. Accordingly, survivorship data from any single study is likely to be unreliable and would benefit from meta-analyses.

Anatomy of the Kaplan-Meier Survival Curve

The Kaplan-Meier survival curve is meant to convey a complex set of longitudinal data in a population of individuals. The left portion of the survival curve includes the early deaths (or other measured events, i.e. relapse) of individuals in a population. In most cancer populations this left portion of the curve has its slope reduce to eventually reach a plateau. When this plateau is stable, few further events occur and a definition of survivorship can be suggested and statistically be defined. Indeed, for most human cancers this plateau occurs 5 years after diagnosis, hence the use of 5 year survivorship as a common curative benchmark. In veterinary oncology the kinetics of time, disease, species, and cancer biology are not likely to allow a single benchmark and will likely require disease and species specific definitions of this survivorship plateau. As discussed above and by Stephen Jay Gould, a common assessment of these data ignores the above described complexity and focusses on a single data point, that being the median. The median is emphasized in the discussion of these complex data and often discretely reported in both oral and written summaries of the population outcome described by the curve.

This focus on the median makes the false presumption that the median is a more important or more likely outcome for an individual, and more important than any other discrete data point on the curve. This focus on the median has an additional consequence which is to convey the absence of survivorship for individuals in the population. When conveyed to families whose pet has been diagnosed with cancer, this perspective emphasizes a palliative outcome for patients.

An additional consequence of this clinical focus on the median has been a drug development focus for novel therapeutics to shift the median to the right. The problem with this consequence is that the median shift is not often a clinically valuable improvement for pet owners, and often requires large clinical studies to demonstrate statistically significant although clinically unimportant changes in outcome. A distinct approach would be much smaller studies, require much coarser intervals for observation, that focus on larger clinically meaningful and statistically significant improvements in survivorship (i.e. curative outcomes).

Curative Outcomes Exist Today.

Within the Kaplan-Meier survival curves there is a reproducible group of data described as the plateau at the right side of these longitudinal data. This plateau represents a time after which ongoing attrition (death or end of disease free interval) becomes less frequent and statistically less likely to occur. After such a point in time it becomes increasingly unlikely for attrition or events to occur. At some point this plateau then represents the population of individuals described as survivors or cures. Interestingly enough, such plateaus are remarkably consistent over decades of Kaplan-Meier survival analysis for most veterinary cancers, but remain largely unreported. A focus on this plateau would allow clinicians to discuss survivorship and curative outcomes with clients as they consider options for therapy for their pets. For example, in the setting of canine osteosarcoma, a median-centric interpretation of decades of clinical data, would include the following clients discussion: “If we pursue amputation of your dog’s leg and then follow this with chemotherapy, there is approximately 50% chance that your dog will remain free of disease in 12 months.”

A plateau-centric interpretation of the same data would include the following client discussion:

“if we pursue amputation of your dog’s leg and then follow this with chemotherapy, there is approximately a one in three chance that your dog will not have this disease affect your dog’s further longevity.”

It is reasonable and most fair for both assessments of the data to be shared with families as they consider options for therapy.

Weaknesses and Fears of the Plateau and Cure

There are reasonable concerns and fears surrounding the plateau and the inference of curative outcomes.

First, in veterinary oncology we recognize the limitations of our dose intensity and are aware of the complexity of the cancer phenotype. Collectively unresolvable concerns make discussion of cure as uncomfortable and a less likely outcomes than the median.

Tied to these realities is our understanding that any mention of the term “cure” will be the only part of a one-hour discussion that a client hears, and that we will invariably fail to meet expectations if and when an outcome that is less than curative is delivered.