Well it's 55,000 people, so it seems like one of those cohort studies where there are many correlations, but then you get down to the 4,000 deaths and it's no longer that big a study after all.
Although they say this:
COMPETENCY IN MEDICAL KNOWLEDGE: Leisure-time running, even at low intensity or pace, reduces all-cause and cardiovascular mortality independently of sex, age, body mass index, health behavior, and medical conditions. Reduction in mortality is related to continued running activity over time, and running is as important as such other prognostic variables like smoking, obesity, or hypertension.
But really to address your point this should be done with nonrunners who start running compared to nonrunners in a case/control study where the benefits can actually be measured using end points other than death. And then you can do a Students t test rather than a correlation coefficient (am I right?).