I assumed it was a spurious correlation. Maybe rapidly progressive Alzheimer's has a very early effect on the enjoyment of exercise, for example.
The studies keep coming though. This one, for example:
BBC NEWS | Health | Exercise 'slows down Alzheimer's'Hmmphh. Still not a randomized study, so it's not very persuasive. These associative studies are more wrong than right.
... While there was no relationship between brain size and exercise in people tested who did not have Alzheimer's, Dr Burns said the four-fold difference in those who did was evidence that exercise might help.
He said: 'People with early Alzheimer's disease may be able to preserve their brain function for a longer period of time by exercising regularly and potentially reducing the amount of brain volume lost.
'Evidence shows decreasing brain volume is tied to poorer cognitive performance, so preserving more brain volume may translate into better cognitive performance.'...
What gives me more hope, are the animal model studies (emphasis mine):
Still voluntary exercise. I'd have preferred they forced the mice to exercise, say a treadmill that dumps the non-runners into water.
... To directly test the possibility that exercise (in the form of voluntary running) may reduce the cognitive decline and brain pathology that characterizes AD, the study utilized a transgenic mouse model of AD rather than normal mice. The transgenic mice begin to develop AD-like amyloid plaques at around 3 months of age. Initially, young mice (6 weeks or 1 month of age) were placed in cages with or without running wheels for periods of either 1 month or 5 months, respectively. Mice with access to running wheels had the opportunity to exercise any time, while those without the wheels were classified as “sedentary.”
On 6 consecutive days after the exercise phase, the researchers placed each mouse in a Morris water maze to examine how fast it could learn the location of a hidden platform and how long it retained this information ... the mice that used the running wheels for 5 months took less time than the sedentary animals to find the escape platform. The exercised mice acquired maximal performance after only 2 days on the task, while it took more than 4 days for the sedentary mice to reach that same level of performance...
So mark me down as cautiously optimistic, though very puzzled about mechanism.
At present I'll grant a 40% probability that exercise will really slow the inevitable  onset of Alzheimer's -- presuming the exercise isn't associated with head injury risk. (So my inline skating hobby is not a preventive measure.)
Even at a 40% 'might help' that's a much higher protective probability than anything else I've heard of other than avoiding head injury .
Since we really do need to work until 70, we can't be cognitively impaired before then. That means we need to delay the normal dementing process by at least ten years. It's time to dust off that old Nordic Track ...
 Live long enough, you get Alzheimer's -- along with vascular dementia. Genetics, head injury and (perhaps) exercise only determine the speed of decline.
 I classify the modern enthusiasm for live combat right up there with the 1970s enthusiasm for snorting coke.