Interestingly, there was an earlier 2017 study, from the 'proper' BMJ no less, which reached similar conclusions:

This study uses a different methodology -recruiting volunteers instead of combing through anonymous health records. I am guessing the latter would be the 'gold standard' but nonetheless this study also reaches very similar results.
40% reduction in cancer mortality, and 41% reduction in all-cause mortality. This finding is pretty well corroborated.