Abstract
“Treasure your exceptions.”
A
One of the GRG's primary areas of interest, over many years, has been supercentenarians (those living in excess of 110 years), and that is the link to the memorable event. As I write this I am somewhat tempting providence, because the event in question has not actually occurred quite yet, but we are close to having a living 117-year-old for the first time this century. (As is our tradition for the first issue of the year, we report elsewhere in this issue a list of living and recently deceased supercentenarians.) Misao Okawa, if still alive on March 5th, will become the fifth validated survivor to 117, and, the next in line, Gertrude Weaver, is only 4 months younger.
What of the more general, demographic situation? Roughly speaking, on the basis of the well over 1000 individuals recorded by the GRG over its existence, the mortality rate (risk of death per year) of supercentenarians is quite close to a constant 50% across the whole range of ages between 110 and 117—in other words, roughly half of those who have reached 110 made it to 111, half of those survived to 112, and so on. But here's the thing: When one looks a little more closely, it is by now un-arguable that the mortality rate through age 115 is not in fact constant, but steadily rises. Fewer than one-third (11/35, to be precise) of GRG's recorded 115-year-olds have made it to 116, and only three more still can (i.e., are still alive at age 115). The most interesting thing about this finding, in fact, is how uninteresting it is: Many commentators, myself included, have confidently predicted that there would be a substantial departure from the ultra-familiar Gompertz trajectory, resulting from a selection effect whereby those with intrinsically slower aging come to dominate the population, but there seems not to be. (Strictly speaking, I am again jumping the gun here, because the rate of increase in the 110–113 range is ostensibly slower than the 10% per year seen in the age range when most people die; but it's about the same as for those dying a decade younger, and a recent study looking at ages up to 106 has shown that there are various biases in the way that such rates are normally calculated, whose elimination restores the 10%/year rate. 1 ) The seemingly inescapable conclusion is that there is very, very little heterogeneity within the human population in the “individual Gompertz slope”—the rate at which a given individual's risk of death increases with age. You might be wondering whether this picture is blurred by improvements in longevity that would make populations born in different years akin to apples and oranges, but it seems that the advances that have allowed so many more people than in prior decades to reach 80, or 90, or even 100 have not yet significantly affected the supercentenarian elite—indeed, it has affected centenarians much less than was anticipated not long ago 2 —so, luckily, we can safely aggregate all cohorts for which there is reliable data.
So far so good. This is an intriguing finding, but what does it mean? When heterogeneity of this or that trait is small, generally there are no dramatic outliers whatsoever, but not always. And in this case, we have a clear exception to that rule—the super-duper centenarians. A hint is already apparent at 117: The proportion of 116-year-olds who made it to 117 is currently 4/9 if we censor Okawa and Weaver, higher than 11/35 (and the hint becomes stronger when we note that all four of them made it to 117½). But the real action happens after that. Of the four (so far, as noted above) individuals who reached 117, two reached 118—and they both reached 119. Moreover, of those two, one reached 120—and she reached not just 121 but 122 before finally succumbing.
So what? It's obvious that one cannot derive any confident conclusions by doing standard statistical analysis with so few data points. But can one do so with more sophisticated statistics? I suspect that one now can, precisely because of the heterogeneity finding I've belabored above. The fact of so slight a deviation from the Gompertz pattern up to age 116, with its implication in terms of an upper bound on individual Gompertz slope heterogeneity, places correspondingly tight limits on how far out outliers (even just one or two of them) can lie before the null hypothesis described by a fitted curve becomes challenged.
My suspicion above clearly awaits rigorous scrutiny before it can be accepted, but for now, for sake of argument, let's suppose that the few longest-lived people of all time must indeed be in some way so special that we might need to reconsider the true universality of the heterogeneity result. What practical, biomedical significance (you knew this was coming—why else would I write this editorial?) would that have? Evidence of a completely different kind is beginning to emerge indicating that such outliers may not be so paradoxical as they seem, and moreover that they have the potential to teach us things of great biomedical import.
The evidence in question is genetic. It's been more than 20 years since the discovery that one's age-specific mortality risk, arising from risk of both atherosclerosis and Alzheimer's disease, was substantially affected by one's genotype for the apolipoprotein E gene. 3 It was a huge shock to pretty much everyone that no other such genes were discovered for many years, this situation only changing when really huge studies were undertaken that could detect extremely small contributions to longevity. But the thing is, all such work has been restricted to genes in which multiple alleles are present in a substantial proportion of the population. Really rare alleles are basically not explored. And this matters hugely, because the only way that a polymorphism can persist in the population for long enough to become common is if there is evolutionary value in the variability—either a heterozygous individual is at an advantage, as in sickle cell anemia, or the population as a whole benefits from diversity—and that's just not going to happen very often. Rare alleles, on the other hand, can exist even if they have a bona fide selective disadvantage in the long term (so long as it's mild), just because random chance sometimes takes time to eliminate them. And such alleles may be good for us in old age even if they are mildly bad for us in youth (which is what evolution cares about). Sure enough. they are starting to turn up—in apolipoprotein A-1, 4 in transthyretin, 5 and even in the amyloid precursor protein 6 —and at least sometimes their effects on longevity seem to be huge.
Might alleles like these be responsible for the super-duper centenarian phenomenon? If they are rare, just having one or two of them would be enough to make someone an outlier. They wouldn't even need to be all that rare to fly under the radar of the sort of demographic analysis that I allude to above: 0.1% would certainly do that. Actually finding them systematically may be really hard, but hey, that's what big data is for, right? And applying such discoveries medically is a long way off, for sure, but the sooner we know what to do, the sooner we'll do it.