You might try reading the linked study, which answers the question of how RMR data were collected (hint: it's accurate). Actually, this entire site would benefit from actually reading the paper.
Just judging from the article (which leaves a lot to be desired), though, it sounds like a small sample size, with no control group, and hasn't been reproduced.
Edit: I'm not saying it's impossible; it's a known issue: http://ajcn.nutrition.org/content/69/6/1117.full
There are reasons that most doctors recommend lifestyle changes, slower maintainable weight loss (eg 1lb a week) and resistance training to combat muscle wasting. The Biggest Loser is already picking outliers, then putting them through programs that are definitely not sustainable long term. It's very possible that all this study shows is that losing weight TBL style is very bad for you, not that everyone gains the weight back. I still haven't after nine years.
https://www.reddit.com/r/Health/comments/4hhdsj/the_science_...
On the face of it it isn't obvious to me why using a least-square model trained on just the baseline of these 16 individuals (the way they predict RMR in the paper) is better than the standard equation, and if the standard equation is a much closer fit to measurement, this seems suspect. But I know next to nothing about metabolic measurements.