The comment from Eugeniu Soronov left me puzzled. So I looked at the references provided.
1) The meta-analysis of low-carb diets and heart health. Published as a Letter to the Editor of the British Journal of Nutrition, and not as a paper submitted through the peer review process, this analysis by Lukas Schwingshackl and Georg Hoffman leaves much to be desired. The criterion of “low-carb” diets as below 40% of calories as carbohydrates is laughable—most low-carb researchers regard 15% carbohydrate as the absolute maximum. In only one of the trials (Varady et al.) did the Forest plot not cross the zero mark, indicating that it was the sole trial to show a statistically significant difference in flow-mediated dilation. However, this trial consisted of a small number of subjects (17), lasted only 6 weeks, and had a very strange notion of appropriate protein content for a high-fat diet (35% of calories). The LF arm had a much more normal 20% protein content. Nevertheless, in its primary outcome, weight loss, the HF arm outperformed the LF arm (mean weight loss 6.6 and 4.7 kg respectively).
Consequently, the conclusion of the meta-analysis, that high-fat diets impair flow-mediated dilation, is not justified. Perhaps it might be more appropriate to make this claim about high-protein diets.
2) The study of patients with heart disease put on low-carb diets. By Merino et al, from the British Journal of Nutrition, this study at least used a more reasonable number of patients (247), but again had a very strange idea of “low-carb” at 29% of calories. The highest-fat-consumption quartile was still only marginally higher than typical dietary values, at 40% of calories. So this trial was neither low-carb nor high-fat, and can therefore tell us nothing about patients with heart disease on such a diet.
3) The study of all-cause mortality on low-carb diets. Noto et al. went to extreme lengths to pool large numbers of subjects on “low-carb” versus “high-protein” diets. However, the “low-carbohydrate score” they chose to use effectively destroyed any notion of determining the macronutrient percentages of calories for the different subgroups. Furthermore, all the studies the researchers adjudged worthy of inclusion were observational. So this meta-analysis is in fact quite worthless.
4) The study showing that obese patients eat less sugar but more fat than skinny people? First of all, while the number of participants is large, it is nevertheless an observational epidemiological study, and so cannot demonstrate any form of causality. Secondly, the differences found in energy intake were tiny: 34.3% versus 33.4% for fat, and 22% versus 23.4% for sugar. No mention is made of the starch variable in the abstract, and the full text is behind a paywall.
So, once again, no inference can be drawn about the effects of high- or low-carbohydrate intake on obesity.
Thank you for providing four wonderfully smelly red herrings, Eugeniu.