Book review: Super Agers: An Evidence-Based Approach to Longevity, by Eric Topol.
I was somewhat disappointed, partly because the title misled me.
The book is broad and shallow, as if he’s trying to show off how many topics he’s familiar with. Too much of the book consists of long lists of research that Topol finds interesting, but for which I see little connection with aging. He usually doesn’t say enough about the research for me to figure out why I should consider it promising.
He mostly seems to be saying that the number of new research ideas ought to impress us. I care more about the quality of the most promising research than about the quantity of research.
The book is mostly correct and up-to-date, but I’m unclear what kind of reader would get much out of it.
Cancer
My favorite chapter was about cancer.
The main strategy that he endorses involves detecting and treating cancer much earlier than is currently done.
Is he aware of the downside that this would risk more harm, from the side effects of treating cancers that would have remained benign? Probably, but he’s not as explicit about that as I would like.
He foresees mitigating that harm by better prediction about what treatment, if any, to use, and about whether screening for cancer is worthwhile.
Currently doctors often decide whether to screen a patient for cancer based purely on the patient’s age. It’s possible to make more sophisticated decisions using an AI that considers hundreds of variables.
It seems likely that many experts are underestimating the benefits of AI-enhanced predictions, but I’m left feeling uncertain about whether those benefits will be large compared to the risks of unnecessarily treating harmless early-stage cancers.
I’m glad that Topol rejects the standard practice of classifying cancer solely by the organ that it affects. He reports that it’s feasible, and at least sometimes beneficial, to classify cancers by their mutations or other molecular characteristics. It sounds like status quo bias is deterring much of the medical establishment from adopting that advice.
Topol is optimistic about medical research, but puzzled about deploying the results:
Why aren’t we routinely assessing body-wide inflammation and building on the success of anti-inflammatory drugs … that significantly reduced fatal cancers … in a large randomized trial? Why aren’t we aggressively tackling air pollution, which is unquestionably promoting cancer?
What I Didn’t Like
Topol’s average level of optimism about medical research is roughly appropriate. But it’s an overly uniform optimism that neglects some serious problems such as Eroom’s Law (named for working the opposite way of Moore’s Law). Here’s an example of comments, marginally related to aging, that overstate progress:
a forecast was made that we might have a COVID vaccine … within eighteen months. Ridiculous! No vaccine program had ever come close to such a short time frame!
False. In the 1957 flu pandemic, it took about 5 months to deliver 40 million doses.
Did COVID vaccines require more innovation? Not that I can tell. COVID vaccines such as Covaxin and Sinopharm used relatively old technology. There were multiple vaccines for other coronaviruses to learn from (for dogs, cats, pigs, poultry, and cattle).
A better statement of what was unusually fast about the COVID vaccine programs was that they achieved that speed while following modern standards. The effects of those standards sound more in line with Eroom’s Law than with the progress that Topol depicts.
Topol quotes Fauci as implying that it might take 15 years for a COVID vaccine. Why were the “expert” forecasts so far off? Topol wants to credit unusually good vaccine programs. I blame the forecasters, and the people who decided which forecasters deserved publicity.
Conclusion
I liked parts of the book, but was annoyed at how much of it strayed from topics that interest me.
> It sounds like status quo bias is deterring much of the medical establishment from adopting that advice.
My understanding from anecdotes is that an individual doctor who favors anything nonstandard, whether it is a test, diagnosis, treatment, or prevention, can expect to have a lot of trouble getting the patient’s insurance to pay for it. I imagine that an individual employee of the insurance company would find it equally hard to go against their policy. And the executives are almost certainly tasked with increasing profit at acceptable legal risk, rather than improving patient health.
So I guess the main bias that needs addressing is systemic, maybe structural, rather than an “average belief of individuals about what might help the patient”, though that may matter as well.
Pingback: The Death of Cancer | Bayesian Investor Blog