I love genetics. Among the many disciplines of biology, it is perhaps the most elegant – that is, it explains much with just a few simple principles. The genetics class I took at UCSC so entranced me that I gave up all ambition to study psychology or be a writer, and became a molecular biologist instead.
Although genetics says much about the nature of life, it has much less to say about death, at least in human populations. I’ve argued before (here and here and here) that genomics is doomed to fall short of expectations that it will lead to one medical breakthrough after another. It’s not because there is anything wrong with gene theory, it’s that environments play a much stronger role in determining lifespan.
We’ve known this for decades, largely from twin studies that compare disease incidence and mortality rates between identical twins and non-identical siblings. They typically find that heritability – a loaded word that we’ll come back to – is at most 15-30% for lifespan. But these studies usually are for restricted geographic areas that may be non-representative. And of course they rely on twins, who are not necessarily representative of total human populations.
Measuring heritability is not the same thing as measuring the influence of genes. Many important traits – cultural and economic – are also inherited and have an effect on lifespan. So a finding of 30% of heritability is an upper limit on the influence of genes. The actual contribution is less, and could be much, much less.
This paper in Genetics uses the Ancestry.com database to measure correlations of lifespan between related (eg, brothers) and non-related (eg, spouses and cousins-in-law) family members. The basic premise is that longevity correlation should decrease at the same rate as genetic relatedness – that is, to the extent that genetic relatedness determines lifespan.
Per the title, they find that previous estimates of genetic contribution to longevity are overestimated due to the effects of assortive mating – our propensity to choose mates who are like us.
They come up with an estimate of genetic contribution to lifespan determination of less than 10%. And the principal authors are from Calico, a company founded to use genomics (and other sciences) to develop life-extending technologies. Oops.
It’s always nice to have one’s prejudices confirmed, and so I found this a very satisfying result. But looking over the figures, I noticed a surprising pattern in the data. The correlation in longevity between unrelated family members was higher than that of relations, and decreased over time.
For instance, here’s spouses vs siblings:
And here’s other non-related family members
I think I understand the explanation for why spouses (whom you choose) correlate better than siblings (whom you are just stuck with) – it’s assortive mating. But it is less obvious to me why this correlation would decrease over time, and not just for spouses but for cousins-in-law.
I suspect the answer lies in the time frame. Given a median lifespan of 50 years, the birth cohort of 1800 would have mostly died off before a substantial fraction of the population benefitted from public health measures of infection control.
But they would have been the last generation to do so. Smallpox vaccination was becoming more common, and John Snow (this one, not this one) showed that cholera was spread by tainted drinking water in 1854, prompting cities to begin providing clean water to their citizens. Germ theory became established in the 1870s, and led to curative therapies and more vaccines by the 1880s. 
We all know this story. But how would it have led to greater (but declining) correlation of lifespan between unrelated family members?
Assortive mating driven by infant mortality. Infectious disease deaths are always much higher among infants than adults and took huge proportions of all children born in 19th century Europe and North America (where Ancestry’s data is concentrated). Two people who survived long enough to marry will inevitably have more-correlated lifespans than siblings, at least when a quarter to a half of all siblings were lost in infancy. But as infant mortality plummeted in the 19th century, so did this disease-driven correlation. QED.
So there we go. I got two of my pet prejudices confirmed – that genetics matters little and infectious diseases matter a lot whenever we dig into the history of human society and behavior.