Alan Hollingsworth: Precision Medicine Plagued by Polysemy
Don’t fret. I had to look it up, too. Polysemy is the association of one word with two or more distinct meanings. Monosemy is a one-to-one match between a word and a meaning, and is found more often in specialized vocabularies dealing with science and technology. That said, we have plenty of polysemy to pass around in science, and it can overpower anything that so-called precision medicine has to offer.
So how did I stumble on this word-of-the-month – polysemy? In May 2017, I was referenced in the Journal of Volcanology and Geothermal Research. I’m not kidding. In my fiery young days as a trauma surgeon in the crater of Los Angeles, I might have been accused of having explosive moments, but Volcanology? Really? I thought it was a mistake when I received the online notice that I’d been quoted. But curiosity won out, and when I clicked on the beckoning prompt, I discovered that this geothermal Alan Hollingsworth and I were one and the same.
And there it was — polysemy — scattered throughout the article as if it rolled from our tongues on a daily basis.
It seems that volcanologists are struggling with the exact meaning of a particular word – overpressure – which, from my position of volcanic ignorance, seems fairly important to nail down. Yet, to quote from the article, “This (polysemy) is likely to be a particular problem in analysis of geothermal resources, where reservoir engineers, volcanologists and structural geologists may each confidently use overpressure but mean different things.”
So, of course these authors (from Norway, U.K. & Italy) went straight to PubMed and searched “breast cancer” to find support for their statement regarding the problem of precise definitions. Probably not. I have no idea how my editorial in the The Breast Journal caught their attention. Nonetheless, in 2015, my article dealt with – who knew? – polysemy. I was still 2 years away from hearing that word for the first time, but my title was, “The Beginning of Wisdom is the Definition of Terms,” a phrase I stole from Socrates who, by golly, was smart enough to avoid writing anything down. Socrates was openly opposed to literacy, believing that words short-circuited the learning process, forcing an illusion of knowledge while compressing big concepts into small containers (at least we think this is what he believed – after all, he wrote nothing).
My editorial intent was to discuss how clumsy definitions pollute even the most sophisticated statistical analyses. If we’re all talking about different things when we analyze “multicentricity,” for instance, we’re simply looking for trouble. Or, what about “local recurrence?” Do we mean breast parenchymal recurrence, chest wall recurrence after mastectomy (solitary vs. diffuse), or regional node recurrence? Precise definitions are lagging way behind precision medicine. Yet, as Socrates (allegedly) said, “The beginning of wisdom…” Well, you get the point.
Here’s another polysemic term – “breast cancer.” I’ve been on a campaign lately trying to undo the confusion about “breast cancer” in the historical autopsy studies. The overdiagnosis crowd is having a field day as breast cancer experts shoot themselves in their feet with misrepresentation of this very important data. These anti-screening epidemiologists and their ilk don’t have to lift a finger when clinical experts announce from the podium: “Breast cancer can be found in autopsy series 30% of the time, similar to prostate and thyroid cancer.”
If we define “breast cancer” to include DCIS then, yes, 30% can be found in the literature, although this is at the high end of a wide range (and appears to include some cases of ADH and borderline lesions). But let’s be sticklers for accuracy because we’re not talking about the usual DCIS:invasive ratio in autopsy series that is seen clinically or through screening. The autopsy findings are almost entirely DCIS. Most everyone agrees that there is an element of overdiagnosis in DCIS, the degree of which is under study.
But in contrast, the bitter controversy surrounds the degree of overdiagnosis with invasive breast cancer, and one of the most important bits of information to support or deny overdiagnosis is the disease reservoir in autopsy studies. When the term “breast cancer” is applied to invasive disease only, the autopsy numbers plummet to 1% (range 0 to 1.8%). This is the same as disease prevalence in the living, thus implying an overdiagnosis rate that hovers somewhere around zero. Under the true definition of an overdiagnosed cancer (one that never progresses) there are only two fates for these so-called “pseudo-cancerous” growths – quiescence or regression. The autopsy data, mostly from the pre-mammographic era, effectively rules out quiescence when it comes to invasive cancer. This forces the screening critics to focus on proving that tumor regression is real, in spite of the complete absence of a clinical correlate.
When anti-screening propaganda that draws on the autopsy data is tossed into the public arena, I submit that the author(s) use “breast cancer” in the broadest sense possible to stay within the bounds of legitimacy (and I’ll make the case next month that these same individuals would be the first to state that “DCIS is not really cancer” in another setting). How convenient it is to make DCIS “cancer” whenever it supports your case, then deny that that DCIS is cancer when it hurts your case. I can’t do anything about this contingency that uses sly semantics, given that their statements can be defended merely by capitalizing on polysemy. But what disturbs me is when pro-screening breast cancer experts unwittingly regurgitate the same misleading information.
This 1% rate of invasive breast cancer at autopsy is not even in the same ballpark as prostate cancer, the latter probability approximating age at the time of death. An 80 year-old male has an 80% chance of occult prostate cancer at death. A female, however, has only a 1% chance of invasive breast cancer at autopsy. How can we conceivably call these numbers “similar?” Yet, that claim seems to be the norm today thanks to polysemy.
When I was a surgery resident in the 1970s, it was common to describe nearly all breast cancer patients as having “infiltrating intraductal carcinoma.” And this was long before the word “oxymoron” spread through the country like a contagion. Our awareness of whether we were talking about in situ disease or invasive disease was basically zilch (same treatment for everyone), so the phrase flourished. Of course, mammography had not yet brought DCIS into the limelight, but it’s still a good example of semantics gone awry.
I’m going to save Part Two (with specifics) on this topic for next month, but the fact is that when we are discussing overdiagnosis, or disease reservoirs, it is absolutely mandatory to make clear if we’re talking about DCIS, invasive disease or a combination of the two. Otherwise, we might as well go back to using “infiltrating intraductal carcinoma.”
Original post: https://alanhollingsworth.wordpress.com/2017/07/ (to be continued August 2017)