The picture is taken from this study: https://www.tandfonline.com/doi/abs/10.3810/psm.2011.09.1933 (use Sci-Hub if you do not belong to an institution paying for access)
(quote) "A cross-section of 40 high-level recreational athletes (“masters athletes”) who were aged 40 to 81 years and trained 4 to 5 times per week underwent tests of health/activity, body composition, quadriceps peak torque (PT), and magnetic resonance imaging of bilateral quadriceps."
The 81 yr old that trains 4-5x (!) a week is a lot more special than the 40 yr old (that's my hypothesis at least). Thus any regression is spurious, since we are sampling from different quantiles! So you don't know anything about the correlation with age.
I'm not disputing that training 4-5x a week keeps muscles strong and the body lean, but it's hardly prescriptive. It's like winning the lottery will make you richer when you retire.
I've tried to reach the authors about this when the article came out, but alas. A letter to the editor in a to me unknown journal seemed not the best use of my time. I've even got code demonstrating the fallacy with quantile regression ;)
In addition, studying masters athletes partially limits the applicability of these data to the general population of aging adults who do not aggressively exercise 4 to 5 times per week.
It's unfortunate that such a statement won't make it into popular treatments, but the paper doesn't get very far out over its skis.
Since I don't know for sure which direction causality points, I'm going to take the path of maximal self-determinism and try to stay fit as long as possible.
I get up every weekday morning at 5 to 6am to do a spin or group fitness class and lift 1 to 2 times per week Arnold superset style. I'm loving this lifestyle. (I guess the gym being Equinox helps.)
That may not be true for a large proportion of people in their 80s.