76 % greater weight loss is not statistically significant???

[Viking Dan]

Courtesy of Fighting the Fad: Low Carbohydrate/High Protein Diets (http://www.thedoctorwillseeyounow.com/articles/nutrition/lowcarb_19/)


More recently, several studies have looked at the effects of a very dramatic reduction in carbohydrate intake. Foster et al. 9 (http://www.thedoctorwillseeyounow.com/articles/nutrition/lowcarb_19/#ref9) educated moderately overweight subjects to follow either a loosely controlled low carbohydrate diet (subjects tried to reduce carbohydrate to an initial goal of 20 g/day of carbohydrate and were given a copy of Dr. Atkins New Diet Revolution) or a low fat diet with a calorie goal of 1200-1500 calories/day). The low carbohydrate diet initially resulted in significantly greater weight loss and at six months the weight loss in the groups was 7.0 vs. 3.2 kg. By the 12-month end of the study, however, the difference in weight between lower and higher carbohydrate diet groups was no longer significant (4.4 vs.2.5 kg).

Granted, I never took statistics, but that seems like a significant difference to me.

[LisaS]

in statistics, significant doesn't mean important (or anything like it) - it is more "how likely is the difference seen to be TRUE (due to whatever we are studying) rather than due to chance" . Sample size, error rates, etc come in to play.
here is a pretty good link that doesn't require you to take a stats class to understand:
http://www.statpac.com/surveys/statistical-significance.htm

[Viking Dan]

So...how big a sample does one need for nearly 2x the results to be significant?

[LisaS]

probably someone currently working in the sciences could give you the specifics in the context of research design and statistical analysis. From what you quoted, apparently for this sample differences of 7.0/3.2 is significant and 4.4/2.5 is not.

or take a statistics course - it is eye opening to learn about how statistics are used, what things really mean when they are cited, how polls work (or don't), what probability really is (and isn't)

statisticians words mean what they mean in statistics and not necessarily the common usage. what "not significant" means is that for that size sample and circumstance, they cannot say that the differences aren't due entirely to chance. not that the difference isn't big in terms of a pct.

[Viking Dan]

2.2x vs 1.7x doesn't seem like a big difference to me, but I do understand what you're saying.

[LisaS]

that's the thing about statistics, the absolute size (or even pct) of the difference doesn't especially matter.

it could be avg 1.3 lb vs avg 1.2 lbs - but if it was 95+% likely to be TRUE and not chance, it would be significant - in the statistical sense - even if you'd look at the raw numbers and say "looks like no difference at all - how could that be significant, they are almost the same"

[deirdra]

The numbers can even look quite different, but if the variability within group is great, then the differences in the averages would not be significant.

For instance, if the average loss was 4.4kg for one group and 2.5kg for the other, but the range of losses was 2-6kg for both groups, you cannot predict that you would lose more on Atkins. You might be the person who loses 2kg on Atkins and 3kg on the other.

[Gabriel Guzman]

Well, you can really see where the flaw is:




The energy that we take in when we eat must be either stored as body energy, or used during the conversion of food to chemical energy to power our cells, or released as heat, or excreted (as with uncontrolled diabetes). This is called nutrient partitioning. It would be wonderful if a particular diet could somehow influence the body to burn more of the calories it consumes and store fewer. In the ideal weight loss program, we would want the composition of the diet itself to cause us to lose body fat while not affecting muscle and other tissues.

Apart from the fact that the references supporting this part of the argument are now rather outdated, more recent research has shown that the partitioning of food does influence the body to use energy in a different way. In fact, evidence of this has been published many times but since it has always been counterintuitive, it has been either ignored or dismissed without appropriate rebuttal. For more information, a good review of this evidence, plus a sound argument on the thermodynamics of weight loss diets, I put an article for download in the Important Links and Resources (http://72.32.36.211/forum/showthread.php?t=68) forum. Tha article's name is "Thermodynamics of weight loss diets".




For example, on a 2000 calorie/day diet, shifting protein from 15% of dietary energy to 30%, with a proportional decrease in carbohydrate from 55% to 40%, would increase energy expenditure by only 23 calories a day, or about as many as you might burn by walking a few hundred yards.

The main flaw here is, in my opinion, the assumption that the body does not adjust to alternative sources for energy and doesn't become more efficient after that adjustment. In single-meal experiements, the above statement may be relevant, but after adaptation, particularly when carbohydrate is reduced even more and protein is adequately increased, the body becomes more efficient in using dietary fat or fat in stores for energy. The example is flawed because it assumes first of all that 2000 calorie/day as the necessary calorie requirement for an individual, which is no longer the case once the individual has switched to more fat utilization than carbohydrate. Another flaw is the assumption that increasing protein intake from 15% to 30% should produce the expected change as though the rest of the nutrients don't influence how that increment is used. Moreover, the apparent decrease in carbohydrate from 55% to 40% doesn't really consider the fact that some amino acids can be used to make glucose and in this example, the actual reduction in carbohdyrate is from 55 to 48% (this is adjusting for gluconeogenesis from amino acids). Thus, the real reduction is a mere 7%, instead of the apparent 15%. It's also important to consider that [if there is enough carbohydrate present in the diet/B], protein may be use in a pathway that leads to fat synthesis. An increase in protein without enough decrease in carbohydrate may not have a dramatic effect in energy utilization. Research has shown, however, that even changes of this type, which alter the ratio protein:carbohydrate still bring about positive changes, not in weight loss or energy utilization but in improvement of other parameters such as better glucose and insulin homeostasis as well as thermogenesis.

We could probably dissect this monument to colosal ignorance but much of it is just a repetition of what other with even less clue have written before. I guess, if we should take the following as an example then we know what to expect (or not to expect from the rest of the article):




Ketosis is the presence in the blood of high levels of acidic substances called ketones, which are produced when there is not enough glucose in the bloodstream, as in starvation, and the body turns to fat as a source of fuel. High levels of ketones make the blood abnormally acid and can be dangerous. Mild ketosis is thought to be a cause of excessive morning sickness in pregnancy; it can also be provoked by a crash diets.

First of all, and just to make sure that those who still don't know the difference, [B]ketosis is plain and simple the production of ketone bodies, nothing more, nothing less. What the paragraph above is trying to explain is ketoacidosis, which is the uncontrolled accumulatino of ketones and can be fatal. On a low carbohdyrate diet is virtually impossible to fall into ketoacidosis and repeated measurements of the amount of ketones have shown that they don't account for even half of the ketone conentration under ketoacidosis. Moreover, given the fact that people on a carbohydrate controlled diet also improve their insulin sensitivity (i.e. they become less insulin resistant), ketones can and actually are used more efficiently and don't accumulate in the blood. Uncontrolled diabetics with severe insulin resistant cannot use ketones efficiently and are, therefore, at high risk of ketoacidosis.