This is a bit long for a Twitter thread, and I'd like to be able to reference it.
I intend it to be tentative, hypothetical, a bit of a question. However, it is written as an argument, a statement, since that is easier, I think, than having every sentence an awkward query or with too much of the necessary language of uncertainty. Please read it in that light.
I don't bristle at being corrected, but welcome it. Further, I take the time to write this out since I haven't seen it elsewhere. If the central point made exists in another form elsewhere I would appreciate being directed to it.
An additional caveat. This is through the eyes of the risk-cost/benefit to an individual patient — the fully informed rational patient, if you will. I realize there will always be charlatans selling promises and the "worried well" willing to buy those promises.
This all said, there was a recent New England Journal of Medicine randomized control trial of colon cancer screening in three European countries that caused a stir (1). This is not directly an analysis of that study, nor of colon cancer screening per se (which is a broader topic), but it prompted a few thoughts.
First, as to the state of the art, using colon cancer screening as the example. Below I link to this paper on colon cancer screening by Zauber(2) from 2010. It is a rich analysis, and she is evidently quite prominent in this field (3). One of the obvious limitations — see below, as well — to the cost effectiveness analysis she uses here is that the prices in dollars, while real, are somewhat arbitrarily set by CMS, etc. There is no "price mechanism." Further, these analyses run the risk of becoming a population utilitarian calculus only. One also runs the risk of attempting to quantify things difficult if not impossible to quantify.
Nonetheless, here you'll find talk of "discount rates," years of life saved (i.e. taking into account the age of finding the cancer), the "efficient frontier," etc. She attempts to account for the costs of complications, as well. It's not perfect, but much more sophisticated than simple talk of absolute risk reduction or effect on overall mortality. Obviously, it does not take into account last week's study. However, this type of analysis is the backdrop. (4)
Still — and this is my primary point here —there is no taking into account the costs/benefits from strictly the individual viewpoint, or that one cannot simply buy the average. An individual either dies from colon cancer prematurely or does not. The average benefit obscures this.
I). The age at time of death is not normally distributed. See the above graphical image. I think the intuition is that most of us would like to improve the chances of having good quality of life up to the modern mode – i.e. the most common age of death, which approaches something like a full potential life — rather than simply add on months/years at the end of life. So, part of that goal is to reduce the risk of being in the long left side of the curve.
II). Population based studies are obviously necessary to determine efficacy, but, on the other hand, one cannot buy the index. In which case, a screening tool (and we're only talking about a few common tumors here: breast, colon, lung, prostate, cervix) that can decrease the chances significantly of pre-mature death at reasonable total cost for that one relatively common cancer acts a bit like the risk management tool of "insurance" from the individual standpoint — since one has only one life.
III) There is still a cost/benefit, but it's more nuanced than a population average applied to the individual.
Here I’ll attempt to flesh out the argument a bit more, if not obvious, re "insurance."
Term life & homeowner’s are good products despite having a negative expected value. If one owned 10,000 residential homes across the country, however, as some sort of rental empire, homeowner's insurance is a bad deal. I own one, so it's wise. The owner of 10,000 homes saves on the insurer's profits and overhead minus claims paid. The individual homeowner pays the premia to share risk, hoping to never need the big payout.
Now, the average patient won’t get or die from, in this case, colon cancer, so why bother? However, no one gets the average benefit, the little benefit, one gets a big benefit or zero. Each individual is a bit like the individual homeowner who cannot diversity his risk.
In an effort to think through heterogeneous phenomena clearly, and put yet another way, one realizes the aim in the case of these relatively common cancers is not to bend the risk curve of something ultimately near universal if one lives long enough (atherosclerosis) but to significantly decrease the chances of that one particular cancer from ending one's life prematurely.
If one's goal is to reduce the odds of avoidable premature death, then as long as the disease is relatively common and occurs at young enough ages, the screening tool significantly effective at detection/disease-specific risk reduction, and not terribly onerous, then the overall mortality rate reduction (nice if you have it, certainly!) may very well be the wrong measure, if viewed through the above lens.
In this sense, it is a bit like insurance — not with a financial payout for a rare idiosyncratic catastrophic outcome (e.g. house fire), but in preventing premature death from that one thing. Again, like with insurance of individual homes, we’re all individuals unable to diversify our risk. No one gets the average outcome, and some leave decades of good life on the table.
(To be clear, unlike insurance, it is not a pooling of risk for a price, so the analogy is not precise. Perhaps one can also think of an analogy to maintenance/inspection of equipment to avoid rare, but critical, as opposed to just expected mechanical failures? However, this too has its limitations).
Further, and to reiterate, we can only effectively screen for a handful of relatively common tumors, this is not an argument to do frivolous, worthless things out of a fear of premature death, just an attempt to describe the goals in a way that captures the intuition a bit better.
There are, of course, significant quantification challenges. Ultimately, I suspect quantification is in a full sense impossible, or should be seen as an approximation, a shorthand at best.
In the Zauber paper referenced above, and which I highly recommend reading, in addition to the issues raised above, dollar costs are used for costs, whereas costs also include time, appointments, inconvenience, discomfort, false positives, distress, etc.
And, certainly, the rare costs of complications are not adequately denoted in dollar signs!
The advantage of using costs in dollars, however, is it allows for quantification, for calculating the discounted costs and benefits, for calculating the "efficient frontier," for calculating the cost per quality of life year saved. The disadvantage is it tends to simplify both those costs and benefits.
It also misses, I think, and as I note above, the "insurance" point. Yet, even for insurance type instruments and products the costs/benefits are not always clear. Term life, particularly for the household’s breadwinner, and homeowner's insurance are fairly straightforward. Others less so (5).
The challenges to quantification should not discourage us, however, from thinking as clearly as possible about the phenomenon in question, our goals, the potential costs and risks, about risk management, and about the potential benefits. Further, we should not settle for a flattened, overly simplistic analysis out of a desire for ease in calculation.
And, of course, one should not neglect the obvious point that costs and benefits in an innovative, dynamic system are themselves dynamic.
The ever important, if thorny, issue of who calculates the risk/benefit and who pays is another topic altogether.
Lastly, one brief statement regarding populations as collections of individual patients vs. populations per se. We are physicians with individual patients who come to us seeking guidance and care, not gardeners cultivating our fields and maximizing the yield. The former is Medicine, the latter something else entirely: the wrong path.
There's another thing to think about, which Atul Gawande explains really well:
ReplyDelete"we’ve assumed, he says, that cancers are all like rabbits that you want to catch before they escape the barnyard pen. But some are more like birds—the most aggressive cancers have already taken flight before you can discover them, which is why some people still die from cancer, despite early detection. And lots are more like turtles. They aren’t going anywhere. Removing them won’t make any difference."
https://www.newyorker.com/magazine/2015/05/11/overkill-atul-gawande
Thanks for reading! That's a fine analogy, but covered by disease specific mortality benefit. If I were to re-write this piece, I would emphasize that nearer the beginning -- this is essentially an argument about disease specific mortality data vs all cause mortality data, i.e. all cause mortality being a nice but not necessary study endpoint in cancer screening, given the above noted constraints: relatively common tumor, effects young enough age, risks/costs not prohibitive to the individual.
Delete