I usually start from the premise that insulin makes you fat. The most simplistic prediction from this is that eating carbohydrate raises insulin and this insulin is what makes you fat. Over the years I have looked at data from all sorts of places, particularly the extremes such as the Kempner Rice Diet and/or the Potato Diet, which clearly work and which appear to do so (to me!) via lowering the level of systemic insulin acting on adipocytes. In particular I consider that first pass hepatic insulin extraction has a huge effect on the systemic insulin level and the subsequent exposure of adipocytes to that insulin. The carbohydrate-insulin-model of obesity as set up by detractors is simplistic in the extreme.
This paper was published recently:
The carbohydrate-insulin model does not explain the impact of varying dietary macronutrients on body weight and adiposity of mice
Of all of the combinations tested the one which we are interested in is where protein was held constant at 10% of calories and carbohydrate was varied from 10% of calories up to 80% of calories. The remaining calories were made up of fat giving a pure carb vs fat comparison. At the end of the study period we have this graph where I have added in the percentages of calories from fat in red along the x axis:
If you draw a straight line through the data points your weight line slopes downwards as fat percentage drops, pretty well. Fat makes you fat:
Now, clearly, there is a missing data point. That is the bodyweight from a diet group with zero carbohydrate, 10% protein and 90% fat. This combination was not included in the study.
So all else from here forward in this post is now pure speculation.
Help is at hand for the missing data point in this paper
A high-fat, ketogenic diet induces a unique metabolic state in mice
Here we also have C57Bl/6 mice and in this case they were fed F3666 diet which looks like this:
"The proportion of calories deriving from different nutrients was as follows: ... KD:95% fat, 0% carbohydrate (0% sucrose), 5% protein"
This is not a perfect fit as the protein is about half that used by Speakman and there is (obviously) no sucrose, but it's the best anyone can do in the absence of the omitted group essential to complete the graph. The mice which were eventually put on to F3666 were initially made obese with a sucrose/fat combination before being put on to their ketogenic diet. Their weight dropped from approx 37g on the sucrose/fat diet to approx 27g on F3666, ie they ended up about 3g lower in total bodyweight than the control mice fed approx 10% of calories from fat in a carbohydrate based diet throughout. Here's the graph we all know from years ago. Open triangles show the drop in weight when F3666 was introduced at around day 80:
So what might a zero carb, fairly low protein group of C57Bl/6 mice look like? They might well end up slightly below the weight of Speakman's 80% carb group mice, (ie those eating closest to mouse chow), or they might end up slightly heavier due to the higher protein content limiting the total fat percentage which could be provided. I feel a compromise might be to use 35g, the same as Speakman's 80% carb group, the closest chow equivalent.
I've added the zero carb speculative data point to the blue line on the graph at 35g bodyweight which now looks like this:
and now we can curve fit the bodyweights like this:
Ah, that's better. Ketosis at the left, "carbosis" at the right. Nice.
I love rodent studies. You just have to understand that setting them up correctly is essential to obtaining the result you want. You also have to know what you want.