To date, the strongest evidence for therapeutic ketosis in neurodegenerative disorders supports a role for cognitive improvement in AD and MCI. Henderson et al. (24) conducted a controversial clinical trial in 2009, finding that an oral ketogenic compound, AC-1202 (a caprylic acid [C8] triglyceride), was superior to placebo in improving cognitive outcomes among subjects with mild-to-moderate AD who were APOε4- (24). These phase 2 trial results appeared promising and were used to justify the marketing of AC-1202 (commercially known as “Axona”) as a medical food. This designation led the FDA to issue a warning letter in 2013, asserting that Axona was incorrectly labeled as a medical food and should be held to a more stringent burden of evidence as an investigational new drug. An updated formulation, AC-1204, subsequently failed to achieve its primary endpoint in a phase 3 clinical trial, which could have been related to sub-optimal levels of blood ketone bodies secondary to limited bioavailability (31). Whereas, the AC-1202 formulation was associated with significant elevations in blood ketone bodies (with a mean post-dose serum β-HB concentration of 0.36–0.39 mmol/L), the AC-1204 formulation achieved a lower mean post-dose serum β-HB concentration of 0.109–0.272 mmol/L.

Despite the media backlash against the controversial marketing of Axona, these clinical trials represented advances in this field of research. To date, the Henderson et al. (24) trial is among the most highly powered studies of ketogenic interventions in neurodegenerative disorders with 152 enrolled subjects. The trial employed a randomized, double-blinded design, comparing a ketogenic intervention to placebo via objective outcome assessment and intention-to-treat analysis. Under strict AAN criteria, however, the trial experienced dropout rates that preclude classification as Class I evidence (at least 80% of enrolled subjects must complete a study for consideration as Class I evidence). A total of 52 subjects were lost to follow-up or discontinued the protocol (a completion rate of 65.8%), resulting in a Class II classification.

An important takeaway from the Henderson et al. (24) dataset was the conclusion that the benefits derived from ketogenic intervention may be significantly modulated by APOε4 genotype. Henderson et al. (24) observed no significant cognitive benefit associated with AC-1202 supplementation among participants who were APOε4+, while participants who were APOε4- demonstrated significant cognitive improvements. This general trend was supported by limited evidence in subsequent studies. A small, randomized crossover study comparing a ketogenic diet to a control diet in 26 individuals with AD suggested that ketone metabolism may be less beneficial for APOε4 carriers, though the authors acknowledged potential confounding factors affecting their dataset (32). In 2018, Torosyan et al. (27) performed a small, double-blinded, randomized controlled trial comparing caprylidene (a caprylic acid [C8] triglyceride) to placebo in 16 subjects with mild-to-moderate AD, with regional cerebral blood flow assessed as the primary outcome. Increased left superior lateral temporal cortical, left inferior temporal cortical, anterior cerebellar, and hypothalamic blood flow were observed only among subjects who were APOε4-, with no significant effect observed among subjects who were APOε4+ (27).

In 2020, Xu et al. (30) conducted a double-blinded, randomized controlled trial comparing 17.3 g/d of medium-chain triglyceride (MCT) oil to placebo in 53 subjects with mild-to-moderate AD (30). MCT oil supplementation was associated with statistically significant cognitive improvements compared to placebo, but these results were observed only among subjects who were APOε4-. A close look at this dataset reveals a subtle signal suggesting the possibility that MCT oil may preserve cognitive function among APOε4+ patients, as APOε4+ participants receiving MCT oil demonstrated marginal cognitive improvement (i.e., stabilization) while APOε4+ participants receiving placebo demonstrated cognitive decline, though this effect was not statistically significant. Only three of the 49 subjects in this dataset were APOε4+, so any analysis of APOε4+ participants in this sample lacks statistical power, precluding the validity of any interpretations beyond hypothetical reasoning. Albeit with methodological limitations, these cumulative findings suggest the possibility that individuals who are APOε4+ may derive relatively less benefit (if any at all) from ketogenic interventions, at least when dosed equivalently to individuals who are APOε4-. APOε4 expression has been associated with inhibition of the PPAR-γ/PGC-1α signaling pathway, deficits in glycolysis and mitochondrial respiration, enhanced neuroinflammation, and dysregulated lipid metabolism (23, 44–47). Although more data is needed to better characterize the potential modulating role of APOε4 status in therapeutic ketosis, some authors have hypothesized that altered metabolism in the setting of ε4 allele expression may limit the body's capacity to metabolize ketones (in turn, modulating the potential neurological benefits derived from ketogenic interventions) (48, 49).

In contrast, a small, recent clinical trial (n = 20) found that subjects with mild-to-moderate APOε4+ AD responded to an MCT oil intervention (33). The trial design included two phases, the first of which was an eight-month, randomized, double-blinded, placebo-controlled crossover study, in which each arm received MCT oil vs. olive oil (placebo) for 4 months prior to crossover. During this phase, cognitive outcomes for the experimental arm did not reach statistical significance. The second phase of the trial involved an open-label extension of MCT oil among all participants for seven months, after which the cognitive test battery was repeated. Overall, at the conclusion of this extended protocol (after completing the crossover trial and the seven-month open-label MCT extension) 80% of the subjects demonstrated either cognitive improvement or slowed rates of cognitive decline, as calculated via Mini-Mental State Exam (MMSE) score trends. 19 participants were APOε4+. When analyzing the subgroup that received MCT oil in the second four-month crossover period then continued MCT supplementation during the seven-month open-label extension, the authors reported that uninterrupted use of MCT oil for a total of 11 months (i.e., the placebo-MCT arm in the crossover trial) was associated with significant improvements in Cognigram^® 1 cognitive scores. A closer look at baseline values, however, reveals that the placebo-MCT arm actually had significantly higher Cognigram^® 1 scores at baseline, possibly influencing results. When baseline values are included in the comparison, the placebo-MCT arm experienced a 1.47% increase in Cognigram^® 1 scores from baseline to study conclusion, while the MCT-placebo arm experienced a 19.97% increase in Cognigram^® 1 scores during the same timeframe. While these results are encouraging for future research, they would be categorized as Class III evidence given the presence of non-equivalent baseline characteristics.

Other studies provided further evidence supporting the efficacy of ketogenic interventions in AD, though potential modulating effects of APOε genotype were frequently not assessed (25, 26, 28, 29). Collectively, ketogenic interventions in AD are supported by one Class I study and three Class II studies. This justifies a “B” (probably effective) recommendation under AAN criteria, with the caveat that this would only apply to patients who are APOε4-. Among patients who are APOε4+, the current data justifies a “U” (data inadequate or conflicting) recommendation under AAN criteria. This may change in the near future as one additional Class III study would justify a “C” (possibly effective) recommendation, at least regarding evidence of cognitive stabilization (rather than improvement per se).

The multi-phase BENEFIC trial provided some of the strongest evidence to date supporting the use of ketogenic interventions in patients with MCI (35, 36). In this trial, 122 subjects with MCI were randomized to receive either a ketogenic MCT formula or placebo daily for 6 months. Supplementation with the ketogenic MCT formula was associated with significant improvements in cognitive tests compared to placebo. No significant effect of APOε4 status on ketosis or cognitive outcomes was observed, though the study was not adequately powered to assess APOε subgroups. This trial is categorized as Class II evidence, as the completion rate for the protocol was 68%.

The BENEFIC trial was preceded by a small study of 23 older adults with MCI conducted by Krikorian et al. (34). Subjects were randomized to either a ketogenic diet or a high-carbohydrate diet for 6 weeks, with the ketogenic diet being associated with significantly improved verbal memory (34). Another small trial conducted by Neth et al. (19) involved a randomized crossover design comparing a modified Mediterranean-Ketogenic diet to a low-fat American Heart Association (control) diet among 20 subjects with subjective memory complaints or MCI (19). Memory improved with both diets, though this may be attributable to practice effects. Compared to the control diet, a modified Mediterranean-Ketogenic diet was associated with significant improvements in the Free and Cued Selective Reminding Test, whereas similar effects were not observed for total story recall and ADAS-Cog12 outcomes. The modified Mediterranean-Ketogenic diet was associated with significant improvements in peripheral metabolic profile, cerebral perfusion, and cerebral ketone body uptake as compared to the control diet. Collectively, more data is needed to assess ketogenic interventions in MCI (particularly to assess the potential modulating role of APOε4 status) but the existing literature justifies a “B” (probably effective) recommendation per AAN criteria in light of multiple consistent Class II studies.

A small (n = 7) proof-of-concept study assessing the efficacy of a ketogenic diet for patients with PD was conducted by VanItallie et al. (2). Five out of seven participants completed the protocol, with all five completers showing improved total UPDRS scores and improved motor subscores. This study achieved remarkably high levels of ketonemia among the three most adherent subjects (mean serum β-HB concentration of 6.6 mmol/L). In 2018, Phillips et al. (41) conducted a larger trial using a less stringent ketogenic diet as compared to the diet employed by Vanittalie et al. (2). 47 subjects with PD were randomized to consume either the modified ketogenic diet or a low-fat (control diet) for 8 weeks. Mean serum β-HB concentration increased to 1.15 mmol/L in the ketogenic group. The ketogenic group demonstrated a 41% improvement in UPDRS I scores (non-motor daily living experiences), as compared to an 11% improvement in the control group, with the largest between-group differences observed for fatigue, daytime sleepiness, pain and other sensations, urinary problems, and cognitive impairment. A randomized controlled pilot trial conducted by Krikorian et al. (42) compared a ketogenic diet to a control diet among 14 individuals with MCI in the setting of PD (42). Mean serum β-HB concentration increased to 0.31 mmol/L in the ketogenic group. Relative to the control group, the ketogenic group demonstrated significant improvements in memory. No significant improvements in motor symptoms were observed.

More recently, Norwitz et al. (40) performed a randomized, placebo-controlled, crossover study in 14 subjects with PD, comparing the effects of acute administration of a ketone ester drink and a taste-matched, carbohydrate-based control drink on endurance exercise performance (40). The ketone ester drink raised participants' β-HB level to a mean of 3.5 mmol/L within 30 min of consumption. Consumption of the ketone ester drink was associated with a 24% increase in exercise endurance capacity as compared to the isocaloric placebo drink, suggesting that ketone ester supplementation has the potential to modulate motor performance in PD.

More research is needed to elucidate the potential for ketogenic interventions to alleviate motor features of PD, particularly given the varying degrees of ketonemia achieved with differing protocols (as detailed above). Based on limited preliminary data, it is possible that more potent formulations (and correspondingly higher circulating β-HB levels) may be needed to significantly influence motor features. It is also possible that multi-pronged interventional approaches (of which therapeutic ketosis may represent a single dimension) may be required to address the complexity of PD pathophysiology (50). This line of research currently supports a “C” (possibly effective) recommendation for non-motor symptoms and a “U” recommendation (data inadequate or conflicting) for motor symptoms in PD based on AAN criteria. With promising preliminary findings and multiple clinical trials in the pipeline, however, this may soon evolve in the coming years.

Although most of the reviewed studies demonstrating cognitive and/or motor symptomatic improvement may (at least in part) support the “alternative fuel” hypothesis, there is limited evidence suggesting that therapeutic ketosis may favorably affect AD-specific biomarkers in patients at risk for AD. Although not all biomarker outcomes featured in the reviewed studies reached significance (38), Neth et al. (19) found that therapeutic ketosis favorably affected CSF Aβ42, CSF tau, and CSF neurogranin in patients at risk for AD (19). These presumed disease-modifying (i.e., preventive) effects suggest (at least in part) support for the signaling hypothesis of therapeutic ketosis in neurodegeneration. However, more research is needed to validate this hypothesis and identify optimal biomarkers for clinical translation.

In contrast to the more immediate symptomatic effects associated with alternative fuel substrate provision in patients with advanced brain glucose metabolic deficits, the effects of such “signaling” mechanisms may follow a delayed time course over weeks to months or longer, as such effects will be mediated through signaling pathways and epigenetic changes as opposed to more immediate substrate-level ‘fuel' availability (14, 51, 52). Although the reviewed studies demonstrated the possibility of cognitive improvement in MCI (36), it is also possible that mitigation, slowing, or even prevention of disease progression may be seen rather than more immediate cognitive improvement. This may necessitate future study designs with primary outcome measures (e.g., clinical severity ratings vs. use of proxy biomarkers) that differ between late vs. early stages of neurodegenerative disease. If this line of reasoning regarding therapeutic ketosis as a means of preventive medicine proves to be true, modification of disease trajectory may be proportionally greater when intervention is initiated in prodromal stages of neurodegeneration. Given the pleiotropic effects of therapeutic ketosis, various mechanisms will undoubtedly be at play in both early and late stages of neurodegeneration, though the relative rates and clinical importance of differing mechanisms may vary.

Several important themes emerged in our analysis of the existing literature. First, it is important to note the limitations of the AAN Classification of Evidence approach. While it provides an objective framework for assessing the strength of evidence for an intervention, it should be noted that this framework is optimally applied in the assessment of pharmaceutical interventions. In particular, a primary criterion for Class I evidence under AAN criteria is concealed allocation, which is often not achievable for a dietary intervention. As such, several high-quality studies that we reviewed were precluded from being classified as Class I evidence, with relatively stronger classifications applied to dietary supplements that can be compared to taste-matched placebo. In addition, the majority of clinical trials we identified were small studies limited in statistical power and treatment duration. Larger-scale, pivotal trials are justified in these populations to validate preliminary findings.

Another important point of consideration is the variety of ketogenic interventions utilized in differing study protocols. In the interest of providing objective criteria for comparison, we included the mean serum β-HB concentration measured among participants in Table 1 whenever possible, often noticing striking differences in the relative degrees of ketonemia achieved by various protocols. The possibility of a dose-response relationship for therapeutic ketosis depending on the clinical context offers an interesting hypothesis to guide future research.

In particular, the mixed preliminary results observed among subjects who are APOε4+ suggest the possibility of a dose-response curve applying to these patients. The strongest evidence supporting non-response to a ketogenic intervention among patients who are APOε4+ was provided by the Henderson et al. (24) trial, which reported a mean post-dose β-HB concentration in the range of 0.36–0.39 mmol/L for the experimental arm (24). For reference, we have included a table outlining clinical interpretations for respective degrees of ketonemia (see Table 3).

A case report published by Newport et al. (53) found that a patient with severe manifestations of AD who was APOε4+ responded clinically to a ketone ester intervention (53). The patient had initially improved during treatment with coconut oil and medium-chain triglyceride (caprylic and capric) fatty acids, as his MMSE score had improved from 12 to 20; his Activities of Daily Living (ADL) score had risen 14 points; and he had exhibited gradual improvement in gait, social participation, and memory. However, his clinical condition deteriorated while participating in a clinical trial assessing the γ-secretase inhibitor, semagacestat. Increasingly desperate for a remedy, his wife (a physician who had assumed his caregiving responsibilities) suggested a trial of a ketone ester intervention (a more potent formulation as compared to coconut oil and medium-chain triglycerides). Subsequently, the patient began augmenting the coconut oil and medium-chain triglyceride regimen with high doses of exogenous ketones (starting at 21.5 g of [R]-3-hydroxybutyl [R]-3-hydroxybutyrate monoester three times daily [TID], then titrating to 28.7 g TID), which was associated with marked clinical improvement. Of note, post-dose serum β-HB measurements obtained during the ketone ester intervention period ranged from 3 mmol/L to 7 mmol/L. This case study suggests evidence of a possible dose-response function that may be dependent on the severity of cognitive impairment, the latter being also a function of more severe glucose metabolic deficits in the brain. Figure 1 illustrates the progressive glucose metabolic deficits that occur during the disease trajectory from normal cognition (few cortical deficits) to mild cognitive impairment (mild-to-moderate deficits) to dementia (severe and extensive deficits) with serial FDG PET scans obtained over 5 years in a different case study of a patient with PD. Similarly, Figure 2 demonstrates more severe glucose metabolic deficits associated with progression from normal cognition to AD, while ketone metabolism pathways are preserved or even up-regulated in the presence of glycolytic dysfunction.

It is thus possible that patients who are APOε4+ may require relatively higher degrees of ketosis (i.e., more potent interventions) to achieve clinically significant outcomes. Table 4 provides a conceptual outline for the stepwise progression of ketogenic intervention potency, with the crucial caveat that more may not always be better. Greater degrees of ketonemia necessitate greater safety precautions, as higher levels may bear greater potential for adverse effects. Attention should also be paid to the half-life of more potent interventions, as pulsatile vs. more sustained levels of ketosis may have differing effects on human physiology.

Conceptualizing therapeutic potency solely as a function of serum β-HB level, however, fails to capture the complexity of intersecting factors influencing metabolic outcomes. Perhaps most pertinently, it is widely known that ketone metabolism is altered in the setting of high glycemic intake with resultant elevations in insulin signaling. Post-mortem brain studies of young adults who are APOε4+ have shown evidence of not only glycolytic enzyme and mitochondrial dysfunction, but also ketone metabolism pathway changes with evidence of mixed compensatory and failing elements (56). Therefore, it is possible that a combined low-glycemic intake and ketone body supplementation approach may be required in these patients as, otherwise, continuation of high glycemic intake may further compromise mitochondrial function, thereby limiting mitochondrial capacity for optimal ketone body oxidation. There is evidence that APOε4+ status may be a risk factor for not only dysglycemia but also for dysfunctional PPAR-γ/PGC-1α signaling pathway functions, resulting in altered mitochondrial biogenesis (23). This observation bears clinical relevance as the potential effectiveness of a ketogenic intervention in APOε4+ carriers to bypass a block in the glycolytic pathway may be reduced if a block in mitochondrial oxidation capacity is also present.

Lastly, an important point of consideration involves long-term risks and safety, as most of the studies identified were of short duration (with intervention periods generally ranging from 3 weeks to 3 months). Notable exceptions include the 15-month Juby et al. (33) crossover study and the six-month BENEFIC trial, both of which assessed MCT supplementation. Further studies are thus needed to assess the long-term safety of ketogenic interventions, particularly for more potent exogenous ketone formulations and strict ketogenic diet protocols.