If research on mice is any indication, minor genetic variations determine response to different diets. No one diet works well for all.
The researchers used four different groups of animal models to look at how five diets affect health over a six-month period. The genetic differences within each group were almost non-existent, while the genetics between any two of the groups would translate to roughly the same as those of two unrelated people. The researchers chose the test diets to mirror those eaten by humans -- an American-style diet (higher in fat and refined carbs, especially corn) and three that have gotten publicity as being 'healthier': Mediterranean (with wheat and red wine extract), Japanese (with rice and green tea extract) and ketogenic, or Atkins-like (high in fat and protein with very few carbs). The fifth diet was the control group who ate standard commercial chow.
Although some so-called healthy diets did work well for most individuals, one of the four genetic types did very poorly when eating the Japanese-like diet, for example. "The fourth strain, which performed just fine on all of the other diets, did terrible on this diet, with increased fat in the liver and markings of liver damage,"...
... with the Atkins-like diet: two genetic types did well, and two did very badly.
... the animal models tended not to do great on the American-style diet. A couple of the strains became very obese and had signs of metabolic syndrome. Other strains showed fewer negative effects, with one showing few changes except for having somewhat more fat in the liver.
With the Mediterranean diet, there was a mix of effects. Some groups were healthy, while others experienced weight gain, although it was less severe than in the American diet. [emphasis mine]
image source (text mine)