The control here is the weight loss itself. In most diet comparisons, whatever the eating pattern does to metabolism arrives tangled up with however many pounds came off, and the pounds do plenty on their own. The Washington University group cut that knot by measuring each participant before and after they had all arrived at roughly the same 10 percent reduction in starting weight [3]. Samuel Klein, the senior author, says the weight question was settled anyway, with only small differences between these three patterns in earlier work [9]. What is left in the comparison is composition.
Then the denominator. Fifty-five people with obesity, prediabetes and fatty liver disease [1], split across three diets, leaves roughly 18 per arm [16]. If about half the keto arm reached remission of prediabetes [7], that is on the order of nine people [17]. Nine is a real signal inside a matched-weight-loss design and a thin base for a remission rate anyone should quote back at a patient.
The Scientific American account does not report how long participants ate this way, how they were allocated to arms, or the actual size of each arm [18]. It also says nothing about people who buy their own groceries: every arm received pre-cooked frozen meals and met a dietician weekly [2]. That is how you get adherence clean enough to interpret, and it is also why these effect sizes read as ceilings rather than field estimates.
There is a live question about the active ingredient. Klein reads the result as an argument about sugar, saying that cutting sugar intake would be the first thing he would tell anyone with obesity to do, independent of weight loss [10]. The keto arm restricted carbohydrate hard enough to force the body to burn fat for energy [13], so carbohydrate load and ketosis moved together. Jeff Volek of Ohio State, who was not involved, calls the design sound and notes that work in his field has already tied ketosis to liver insulin sensitivity and triglyceride production [11]. Sai Das of Tufts, also uninvolved, calls keto highly promising [12]. A three-arm trial with no low-sugar, non-ketogenic comparator cannot adjudicate between those readings.
My view, with its conditions attached: for a patient who already has fatty liver and prediabetes, this is a reason to try carbohydrate restriction first, because muscle insulin sensitivity improved on all three patterns [4] and the liver measures separated on one [6]. Outside that phenotype the trial ranks nothing. And liver fat is a marker standing in for the inflammation and cancer risk Klein cites as the reason to care [15], which is not something 55 people over one weight-loss episode can show you.
