To the Editor:

The recent paper by Zhao et al. (1) claims to identify detrimental effects of the ketogenic diet (KD) on cognitive function in rats. Despite their comprehensive approach to testing cognitive performance, this paper has a serious flaw making it premature to conclude that the KD may pose a potential risk for long-term brain development. Our concern is based on the fact that the KD used by Zhao et al. (1) had a fat-to-protein plus carbohydrate ratio of 8.6:1, which is a ratio more than 2-fold higher than found in any version of the KD used in children. This extreme ratio appears to have reduced food intake thereby causing much lower weight gain than seen in children given a KD for control of intractable seizures.

The impaired food intake created two related problems. First, ketones cannot meet more than 20–30% of brain energy requirement. By extreme limitation imposed by the 8.6:1 KD on dietary intakes of both carbohydrate and protein (the main gluconeogenic substrate), there was a real risk that the brain was starved of energy substrates. Second, users of the classical KD aim explicitly to meet about 75% of the child's calculated energy demands. This inevitably leads to some restriction in weight gain but most children on the KD remain close to their ideal body weight (2). In contrast, weight gain in the KD-treated animals reported by Zhao et al. (1) was 47% and 35% of their two different controls, i.e. about half the average aimed for clinically. Such brain energy substrate restriction and such severely compromised weight gain both totally confound interpreting outcomes related to cognition. Impaired weight gain, although commonly observed in young rats on the KD, can be avoided if a KD with a clinically relevant fat to protein plus carbohydrate ratio of about 4:1 is used, and if the eventual high fat content of the KD is introduced gradually (3).

The authors acknowledge that weight gain was poor and that protein content of the diet was low (41% of control). However, combining the low protein concentration with lower food intake would make the relative protein intake in the KD group about 40% of 41%, or about 20% of that in controls. Without pair-fed non-KD controls that have the same restricted protein-energy intake and weight gain, altered cognitive outcomes cannot be attributed more to the KD than to the protein-energy restriction. Experiments utilizing a paired approach require a KD and a corresponding non-ketogenic control diet that provide matched protein intakes. One possible formulation for such diets has been reported (4).

Zhao et al. (1) demonstrated that absolute brain size was smaller in their two KD groups. However, extrapolating from their figures and compared to the relevant controls, the brain to body weight ratio was 1.53-fold higher in the status epilepticus KD group and 2.1-fold higher in the non-status epilepticus KD group. These results demonstrate the well-known observation that brain growth is less affected than body growth during severe protein-energy restriction and provide no useful information about the possible risk of the KD per se for brain development. Given these serious confounders, the title and abstract are misleading.

Several papers suggesting that “high fat diets” may impair cognitive performance were cited (5–7), but these papers did not use fat intakes anywhere close to that used by Zhao et al. (1) or used clinically in children on the KD so the comparisons were inappropriate. Other studies not cited have reported improved long-term neurological prognosis in children on the same formulation of KD used for control of intractable epilepsy (8–10).

Almost all children on the KD for control of intractable seizures have already been unsuccessfully treated with at least three anti-epileptic medications over a period of many months to years. Hence, the delay in starting the KD plus the behavioral and cognitive side effects of several anti-epileptic drugs seriously confound assessing the effects of the KD itself on cognitive development in children.

We have also worked with these experimental challenges in children with epilepsy and in animal models of the KD. Living with seizures is both physically and emotionally debilitating, so verifying whether the KD (a treatment of last resort) affects cognitive development is important but also requires the “most rigorous evidence-based” approach possible (11). There are two relevant issues – is cognitive development at risk on the classical KD used in children and is the reported model appropriately designed? We believe that methodological problems prevent Zhao et al. (1) from making an adequate case against the KD on either of these issues.