Weight loss correlates with low carb, not low calorie

This is extracted from: “The Obesity Epidemic: What caused it? How can we stop it?”

Marjorie R. Freedman, Janet King, and Eileen Kennedy wrote an article called “Popular Diets: A Scientific Review”, where 17 studies of weight loss with different macro nutrient (carbohydrate, fat and protein) composition, were reported with the concomitant results.[i] The authors stated that “no published studies are excluded”, so I have assumed that these are the only studies between 1956 and 2001. The three authors concluded “Caloric balance (calories in vs. calories out), rather than macro nutrient composition is the major determinant of weight loss.”

Yet I analysed all the data for these 17 studies and found no relationship between low calorie intake and high weight loss (0.009 correlation coefficient), but a significant relationship between low carbohydrate intake and high weight loss (0.79 correlation coefficient) (Tables below). How can the authors have concluded as they did, unless they set out to prove an already held point of view?

“Popular Diets: A Scientific Review”[i]

Notes:

*   21 studies were included in tables 5a and 5b in the paper. Three had no results for weight change, so could not be included (Kekwick 1957, Bortz 1968 and Krehl 1967).

*   I excluded Larosa et al (1980), as the table said that the study was for 12 weeks when, in fact, the article and the original journal revealed that the subjects followed their normal diet for two weeks at the start and two weeks at the end of the 12 week period. We cannot determine what the weight loss per day was in relation to a measured carbohydrate and calorie intake period.

*   The 17 remaining studies are above. There was a material error in the article. This has been corrected in the table above, as follows: Study 8 (Golay et al) – the weight loss in table 5a is given as -8.0 kilograms over 6 weeks, which was commuted to -111 grams per day in table 5a. This is wrong and -8.0 kilograms over 6 weeks equates to -190 grams per day. However, the table appears to be wrong in more way than one and the original journal (and the text in the article) state that the weight change for the group on 37.5 grams CHO was -8.9 kilograms +/- 0.6 kilograms. I have taken the mid point of -8.9 kilograms and this equates to -212 grams per day. This is the number in the table for this appendix and the number I have used for the correlation.

*   A number of ranges were taken at mid point level, as follows:

–    Study 7 (Evans) – the range of weight loss was given as -76 to -119 grams per day, so I took the mid point of -97.5 grams per day;

–    Study 11 (Cedarquist) – the range of weight loss was given as -78 to -150 grams per day, so I took the mid point of -114 grams per day.

*   Five of the studies were directly comparable isocaloric studies i.e. the studies had parallel groups on a different carbohydrate level and the same calories per day (these are studies 6, 8, 10, 12 and 14). In every such isocaloric study in the article, the weight loss was substantially higher in the low carbohydrate group than in the higher carbohydrate group. At the highest calorie intake (Kasper et al, study 6, 1707 calories per day), the group consuming 56 grams of carbohydrate per day lost 300 grams per day vs. the 156 grams of carbohydrate per day group, which only lost 50 grams per day. This supports Kekwick and Pawan’s findings that large numbers of calories can be consumed and weight still be lost with a low carbohydrate diet composition.

Correlation between (low) calories & (high) weight loss

Where:

N = the number of pairs of scores (17)

∑xy = sum of the products of paired scores (-6,094,862)

∑x = sum of x scores (21,373)

∑y = sum of y scores (-4,854.5)

∑x2 = sum of x scores squared (28,593,253)

∑y2 = sum of y scores squared (1,865,057)

r = 0.009

Correlation between (low) carbs & (high) weight loss

Where:

N = the number of pairs of scores (17)

∑xy = sum of the products of paired scores (-147,786)

∑x = sum of x scores (711.5)

∑y = sum of y scores (-4,854.5)

∑x2 = sum of x scores squared (40,158.25)

∑y2 = sum of y scores squared (1,865,057.25)

r = 0.79

I reiterate, how can the authors have concluded as they did, unless they set out to prove an already held point of view?

References

[i] Marjorie R. Freedman, Janet King, and Eileen Kennedy, “Popular Diets: A Scientific Review”, Obesity Research, (March 2001).