Intended for healthcare professionals

CCBYNC Open access
Research

Effects of a low carbohydrate diet on energy expenditure during weight loss maintenance: randomized trial

BMJ 2018; 363 doi: https://doi.org/10.1136/bmj.k4583 (Published 14 November 2018) Cite this as: BMJ 2018;363:k4583

Re: Effects of a low carbohydrate diet on energy expenditure during weight loss maintenance: randomized trial

I have read the responses of Hall et al. and the replies of Ludwig et al. Admittedly I have no special expertise in these subjects, but I believe I have identified a serious flaw in Dr. Hall’s arguments.

He says that his analysis is correct, while Dr. Ludwig’s is incorrect. But Ludwig’s analysis gives a highly significant result. Dr. Hall, if he is right, should be able to explain why Ludwig’s analysis gives a highly significant result. The only explanation implied by Hall’s arguments is that it is random chance. We know “p-hacking” can occur if there are many possible analyses. Then one can choose the analysis that gives a significant result by random chance. This is clearly not the situation here, since this is the primary endpoint, and there only a small number of possible analyses.

Hence, since random chance is very unlikely, there should be a reason that Dr. Ludwig’s analysis gives a significant result. Dr. Ludwig has a compelling explanation for this. Dr. Hall, if he is to be convincing, should give an explanation for the significant result. The non-significant result of Hall is explained plausibly by both Hall and Ludwig, so that result does not distinguish who is correct. The significant result is explained much more plausibly by Dr. Ludwig, so it is compelling evidence that Ludwig is correct and Hall is wrong.

Andrew Moulthrop

Competing interests: No competing interests

06 December 2018
Andrew A Moulthrop
Engineer
None
Los Angeles, CA