Nah, its not that, its that there are not enough controls to make the claim. Lets create a strawman and see if it works for you, lets say your "job" was fitness equipment tester, and the more you worked the more exercise time you got in. Historically there is a sort of bell shaped impact on health here where exercise increases your overall health up to a point at which point too much of a good thing starts impacting your health in a negative way. If you were at the global maximum on that curve, either working more or working less would cause you to be less healthy.

For people where they are more 'at risk' at work than they are not at work (say 'deployed soldier') working less is statistically safer, for folks who have the reverse (say submariners) the reverse is true.

The discussion usually comes around how do you narrow the claim to make it either provable or disprovable. Is it true for all programmers? how about for 'all java programmers' or 'all java programmers in the bay area'. See the challenge?