So having healthy grandparents gives a survival advantage.

At that age, the effect should be reversed as people often need care themselves rather than caring for grandchildren (who may be adults by that time anyway)

I suppose the "grandmother theory" originally only explained why women live beyond menopause, not into their 70s, 80s or 90s.

Maybe even more importantly, though, 100 years ago other, unrelated diseases of old age were likely to get you well before Alzheimer's really got bad.

When resources get scarce, there are a lot of ugly realities.

http://newsroom.ucla.edu/stories/jared-diamond-on-aging-1505...

"Mostafavi and his colleagues tested more than 8 million common mutations, and found two that seemed to become less prevalent with age"

Old people are less likely to have this gene than young people. This can have two possible explanations: old people with this gene are more likely to are already dead (and thus not show up in the study), or the gene is becoming more prevalent in the younger generations. The statement "People without these mutations have a survival edge and are more likely to live longer" favors the first explanation.

This is not the effect of natural selection on gene prevalence. (And it would be weird for Alzheimer's, which usually occurs at an age where reproduction is often finished, and maybe even raising the kids is done already).

Note that they are not suggesting these genes have been the subject of selection; it is the hypothesized number of such late-acting genes that were not found that are assumed to have been selected against.

At least with smoking, we've been experiencing exactly that decrease for generations. Smoking has become vastly unpopular compared to how prevalent and socially accepted it used to be. Social factors had been one of the main driving forces, but that doesn't mean it has to be the only driving force.