Quantitative results of in-depth analysis should be a good thing, one would think.
As a practicing physician I made many, many decisions in the care of patients, and a surprising number of those decisions did not have rigorous scientific studies backing them up. They may have been based on personal experience, accumulated opinions of experts or even some clinical data that suggested a certain procedure, or a specific drug, would provide the anticipated clinical outcome. That's the way a lot of medical decisions are made, although more and more effort is going into measuring and consolidating the clinical evidence in validated research settings to allow true evidence-based decisions.
As a scientist, well, valid analysis of experimental findings is everything. I recall as a graduate student way too many nights in the library poring over recently published scientific papers and dissecting the methods, the results, the analysis and the conclusions of the authors. It was a star in your cap at Journal Club if you found errors in either the selection or the execution of the statistical analysis. Their statistics approach could mean the difference between good research with a valid conclusion or a "filler" paper meant to pad the author's Curriculum Vitae.
In both the clinical and scientific research arenas one is striving for decision-making support and hoping that the authors of the studies were true adherents of scientific methods and principles. However… it's not always that way when statistics are involved. Although true attribution is vague, Mark Twain popularized the phrase "There are three types of falsehoods; lies, damed lies and statistics."
There were many instances in those research papers in which the authors were using their data to tell their story and therefore presented the numbers in a manner that supported their position. Similar numerical obfuscation abounds in both marketing and politics. Numbers presented supporting a particular point of view are to be suspect in proportionality to the vested interests of the presenters.
This brings me to my point - cancer patients should not read statistical analyses of outcome data for their cancer. This is an incredibly difficult thing to avoid. When one is first diagnosed with cancer it is natural to research every possible presentation on the subject. Many options abound - from true scientific literature to hucksters on the internet - and all are backed by numbers. Five-year survival rates, regression percentages, response fractions, dose-response curves, etc. etc. If you are a cancer patient, ignore them! They are not the truth you are looking for! They do not mean that if you do this, the results will, inevitably, be that. The scientific data are meant to nudge progress over time and support clinical decisions of the treatment providers. The marketing data of the hucksters are meant to separate you from your money. No one can tell you the odds of your responding to medication, or even of your own survival with accuracy any more than they can predict the March Madness outcomes.
This sentiment comes after reading a recent paper on the outcomes of the investigational drug I am (likely) being given. The paper hailed the fantastic outcomes of previous clinical studies with this drug and touted it as THE new therapeutic option for those with advanced melanoma. This all sounds good - until you get down to the final numbers. The statistical improvement in survival was about 20 months. Twenty month survival was not the conclusion I was looking for, but that's what I came away with. Now I could parse their conclusions and re-run all the numbers, but I suspect they did a good job and my quantitative analysis would come out the same as theirs. However, the study had a number of variables -- response rates, dosage differences, prior treatments, etc. that contributed to an aggregate 20 month enhancement of survival. I know, intellectually, that this doesn't mean that I only have 20 months to live but, emotionally, it is difficult to ignore that bottom line number.
But I can ignore it, and ignore their statistical analysis, and say that in my experience with this drug, I am 100% alive!
No comments:
Post a Comment