Too many variables in the group members.
This is the major flaw of almost all 'lifestyle' studies. I say 'lifestyle' because we're essentially looking at people's lives. It's not a totally accurate word but it's close enough.
There's also the major issue of BMI being an unrealistic and unreliable statistic. How do you adjust for skeletal muscle? A highly vascularised tissue that reduces blood pressure, the development of which assists in reversing metabolic syndrome and type 2 diabetes. It's unquestionably a healthy addition to the usual untrained individual, but it increases the BMI of that person (we all know this).
BMI is the wrong statistic here. There isn't a catch-all one to use, so multiple should have been used to draw conclusions, and to create the study groups.
And again, meta-analysis should raise question marks.
As I said, it's not a worthless contribution and it makes the reader curious, question the dogma etc. That doesn't mean it's inherently reliable though.