Does eating less overall (but high protein), in conjunction with weight lifting (4 days a week), have the effect of inducing fat loss while growing muscle?

On the CBC site it is claimed that a regime of exercise, very low calorie dieting and high protein intake allowed test subjects to gain significant muscle mass while losing fat.

To our surprise the amount of muscle lost, even in the low protein group, was negligible and they lost about 3.5 kg of fat. The high protein group was able to gain muscle mass. They gained about two pounds of muscle and lost about 10 pounds of body fat.

Is this reporting an accurate representation of the study described? Is the study reputable?

  • 2
    So one group lost kilograms, and the other group lost pounds!
    – GEdgar
    Feb 10, 2016 at 14:47
  • This question is highly subjective. It depends on your age, current physical fitness level, body type, and so on. If I took a male that was 6'3" 150 pounds and overall pretty fit, and then put them on a low calorie diet you would get little to no gains. But the obvious is that you will gain muscle lifting 4 days a week compared to 0. It is like asking if your teeth are cleaner if you brush twice a day vs never. You probably need some kind of baseline in your question or it is too broad to cover in less than a 500 page book.
    – blankip
    Feb 10, 2016 at 18:12
  • 1
    @blankip it's not highly subjective if we post scientific results (which we do). Scientific papers take all those variables into account.
    – Sklivvz
    Feb 10, 2016 at 20:21
  • @Sklivvz - My point is the question isn't clear enough not that it is unanswerable. Is the title question for everyone, for people who normally don't workout, for people who work out all the time? Again if you take an olympic athlete and feed them a low calorie diet and make them workout... well they have been doing that for years and won't change much if at all. So yes its subjective because studies for the average non-working out person could be true and others false.
    – blankip
    Feb 11, 2016 at 4:35
  • 1
    @blankip it's a normal step for any serious research to estimate and address all confounding factors. For example, the sample of people subjected to the experiment must not have a systemic bias, as you suggest. That said, it's possible to have a good sample.
    – Sklivvz
    Feb 11, 2016 at 11:47

1 Answer 1


The best thing to do is to go to the original paper, which isn't too hard to find. There are also other studies cited in the FDA listing for this work which seem to corroborate the finding that exercise plus increased protein intake plus significant caloric reduction of diet will increase muscle mass while reducing fat. But all of these are small studies (6 to 38 participants total).

But when I read these papers I'm suspicious. Looking at the results commentary in the study referred to by CBC:

Results: As a result of the intervention, LBM increased (P < 0.05) in the PRO group (1.2 ± 1.0 kg) and to a greater extent (P < 0.05) compared with the CON group (0.1 ± 1.0 kg). The PRO group had a greater loss of fat mass than did the CON group (PRO: −4.8 ± 1.6 kg; CON: −3.5 ± 1.4kg; P < 0.05). All measures of exercise performance improved similarly in the PRO and CON groups as a result of the intervention with no effect of protein supplementation. Changes in serum cortisol during the intervention were associated with changes in body fat (r = 0.39, P = 0.01) and LBM (r = −0.34, P = 0.03).

Seems legit, but from a naive statistical point of view things don't seem to add up.

  • How can lean body mass be statistically different in the two study groups when their ranges overlap significantly (LBMPro = 0.2 to 2.2 kg, LBMCon = -0.9 to 1.1kg)?
  • Similar question around loss of fat mass (FatLossPro = -6.4 to -3.2kg, Fat LossCon = -4.9kg to -2.1kg)?

Maybe they used fancy stats that I wasn't trained in. Maybe the data isn't normally distributed (which then begs the question of why they included +- values as if the data was normally distributed...). Or maybe, like with a lot of scientific papers, the statistics are garbage and the peer-review process fails to address bad stats.

  • 1
    I suspect that the statements of statistical significance are with respect to the mean value of the statistics, in which case there can be an overlap in the statistical distributions of the two groups, while having a statistically significant difference in the means (e.g. some women are taller than some men, but with enough people in the study there is likely to be significant difference in their average heights).
    – user18604
    Feb 12, 2016 at 17:36
  • @Dikran - Reporting P<0.05 means that no more than 5% of the distributions overlap. Given that [FatLossPro plus one presumed standard deviation] > FatLossCon, saying P<0.05 is nonsensical as +- one SD corresponds to only two thirds (approximately) of the data. My bet is that the data is NOT normally distributed and the authors used a parametric test rather than a non-parametric one.... which from a stats perspective is just plain dumb.
    – Doug B
    Feb 12, 2016 at 18:57
  • 2
    no that is not the meaning of a p-value. Perhaps you should ask a question about this on the stats SE. Consider the example I gave earlier, I suspect the overlap between women's heights and mens is larger than 5%, but very obviously men are taller than women on average, and a sensible statistical test of that hypothesis would give a statistically significant result for sufficiently large sample sizes.
    – user18604
    Feb 12, 2016 at 19:39

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .