Ketogenic diets have been used for over 50 years in a variety of CNS conditions. For example, in a 3-month randomized controlled trial, a ketogenic diet resulted in an almost 75% reduction of seizure frequency in childhood epilepsy [33
]. The long-term tolerability and effects of ketogenic diets have also been previously reported [17
]. In addition, there are preclinical data on the beneficial effects of intermittent fasting, caloric restriction and ketogenic diets on brain amyloid deposition and toxicity (for review see [22
Ketogenic diets result in many changes, other than simply elevating circulating ketone body levels, which may confer neuroprotection. For example, ketogenic diets have been found to increase levels and activity of uncoupling proteins [34
]. To address whether administration of ketone bodies alone are neuroprotective, several studies have examined if infusion of BHB would protect cells from a variety of cytotoxic agents and conditions. Infusion of BHB was found to protect rodents from 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) [35
], hypoxia [20
], traumatic brain injury [37
] and glutamate toxicity [38
]. Such studies suggest that ketone bodies alone offer a possible therapeutic intervention under several conditions (for review see [21
In this study, we examined if chronic induction of mild ketosis would be beneficial to patients with mild to moderate AD. Chronic induction of ketosis in AD patients could be achieved by compliance to a ketogenic diet. However, ketogenic diets require strict adherence to low carbohydrate intake. Compliance to low carbohydrate intake may be difficult for Alzheimer's patients due to the well documented shift in food preference toward sweet, carbohydrate-rich foods [25
]. We took advantage of the unique properties of MCTs to induce ketosis without the need for dietary change. AC-1202 successfully induced mild ketosis in AD subjects. Administration of AC-1202 at 10 grams (1/2 recommended amount) at Baseline significantly elevated average serum BHB levels by 157%, 2 hours post-dose. Administration of 20 grams of AC-1202 (full recommended amount) significantly elevated average serum BHB levels by 330% on Day 45 and by 401% on Day 90. In contrast, in the Placebo group, average serum BHB levels declined between pre- and post-dose sampling at each study visit, possibly due to suppression of endogenous BHB production after eating a carbohydrate-rich breakfast. Notably, AC-1202 was able to elevate serum BHB even when the subjects ate breakfast. These findings are consistent with the ketogenic properties of MCTs [30
AC-1202 is an MCT composed almost entirely of C8:0 fatty acids. It is possible that some beneficial action could be attributed to the C8 fatty acids rather than ketone bodies. Several reports have implicated MCTs in increasing fatty acid oxidation with possible roles in weight loss [40
]. However, in general, very little C8 reaches the blood stream after ingestion of an MCT. Medium chain triglycerides containing C8 fatty acids undergo complete hydrolysis in the gut lumen. The released C8 are poor substrates for esterification and instead are transported by the portal vein directly to the liver. Within the liver, C8 fatty acids undergo obligate oxidation. Thus, very little C8 makes it in to circulation (for overview see [29
]). In studies in humans dosed with C8 containing MCTs, very little free C8 is found in aterial blood [43
]. While we cannot entirely rule out the role of C8 fatty acids in mediating some of the cognitive effects seen in the study, the correlation of cognitive performance with circulating BHB suggests that ketosis plays a prominent role in MCT therapy.
The levels of ketosis obtained with AC-1202 were mild and similar to those seen in the early phases of very low carbohydrate diets, and much lower than levels found during starvation or diabetic ketoacidosis. Very low carbohydrate diets have been examined mainly for weight loss and management of type II diabetes (for review see [44
]). In studies of low carbohydrate diets, average levels of BHB after 2 weeks range from 0.4 mM to 0.65 mM, and these levels frequently decrease over time and may return to Baseline after 10 to 12 weeks [45
]. Ketogenic diets differ from a low carbohydrate diets by imposing stricter limitations on protein intake, and higher, more sustained levels of BHB (above 1 mM) have been reported [49
]. As a comparison, much higher levels of serum BHB are found during 5–6 weeks of starvation (4–8 mM) [51
] and in cases of diabetic ketoacidosis (9–10 mM) [52
] (Table ). AC-1202 induced transient increases of ketosis that reached average levels of 0.3 to 0.4 mM in the 2 hour post-dose sample on Days 45 and 90. The positive cognitive effects noted in E4(-) subjects suggests that higher levels of ketosis may not be required in AD patients to produce beneficial outcomes, and that safe, mild elevations in BHB can be effective.
Levels of BHB associated with AC-1202, fasting, and dietary regimens
While the direct effects of ketone bodies on AD metabolism have not been definitively shown, preclinical studies provide clues as to the possible mechanism of action. The induction of mild ketosis by AC-1202 in an aged dog model was found to improve respiration rates in the parietal regions of the brain, by improving mitochondrial function and reducing mitochondrial oxidative damage [53
]. This is consistent with studies demonstrating improved mitochondrial function in rat hearts perfused with BHB [15
]. In addition, Kashiwaya et. al. tested the ability of BHB to protect cultured hippocampal neurons from the toxic effects of Aβ42. Four millimolar BHB was found to significantly protect the neurons from Aβ42 [23
]. The proposed mechanism of protection was improved mitochondrial efficiency. In cell culture studies, incubation with ketone bodies reduced need for glycolysis, increased metabolites in the first third of the TCA cycle, and increased the redox potential of the NAD/NADH couple [15
]. In addition, exposure to BHB has been reported to increase autophagy [54
] and activate HIF-1(alpha) [55
]. Further experimentation will be required to understand the precise mechanism whereby AC-1202 improves cognitive performance in AD patients. Yet, the rapid response seen in an earlier study [31
], suggests that improvement in cellular metabolism plays a prominent role. For a discussion of the role of ketone bodies in Alzheimer's disease see Henderson [14
While the cognitive effects were not significant in the overall sample, a pre-defined examination of cognitive effects stratified by genotype yielded significant effects in E4(-) participants. The ADAS-Cog difference between AC-1202 and Placebo of 4.77 points at Day 45 and 3.36 points at Day 90 is notable given that many of the subjects in both groups were already receiving cholinesterase inhibitors and/or memantine. After the two week Washout (Day 104), there was no difference between groups. E4(+) subjects did not differ between groups at any time point.
Near the end of the study, newly enrolled participants were intentionally assigned to AC-1202 or Placebo groups by an independent monitor to balance the number of subjects who completed the study in each group. Since this intentional assignment may have introduced bias into the study, we conducted an analysis of "randomized only" subjects. The analysis of the randomized only subjects mirrors both the ITT w/LOCF population, and the per protocol population. In each analysis, there is no significant effect in the primary outcomes. Yet, in each analysis, significant effects in change from Baseline in ADAS-Cog scores compared to Placebo were found in APOE4(-) subjects, at both Day 45 and Day 90. In addition, in each case, the significant effects on APOE4(-) subjects is lost after the two week washout (Day 104).
Intention-to-treat w/LOCF analysis is beneficial in that it captures data from participants who may not have tolerated the complete study regimen. Yet, since AD is a progressive disease, a reliance on the ITT w/LOCF analysis may have biased the results toward AC-1202, given that the AC-1202 group experienced a higher dropout rate compared to the Placebo group. To address this concern, we conducted an analysis of "per protocol" subjects. Per protocol subjects were defined as subjects who completed all study visits and efficacy measures and had no values carried forward. Despite the relatively small numbers, the results of the per protocol analyses mirror the ITT w/LOCF results, with better results seen in the per protocol population. Among all per protocol subjects, a 2.53 point difference between groups in change from Baseline in ADAS-Cog was found at Day 45 compared to a 1.91 point difference in the ITT population. As with the ITT w/LOCF analysis, better efficacy was found in the per protocol E4(-) sub-population. On Day 45, a 5.73 point difference between groups in change from Baseline in ADAS-Cog was found in the E4(-) per protocol population compared to a 3.05 point difference in the E4(-) ITT population. On Day 90, a 4.39 point difference between groups in change from Baseline in ADAS-Cog was found in the E4(-) per protocol population compared to a 3.36 point difference in the E4(-) ITT population. The significant differences found between AC-1202 and Placebo groups in the per protocol analysis suggest that the positive efficacy results are not simply due to the use of imputed data. The consistent finding of an effect in E4(-) subjects in the total population, per protocol, and randomized only subgroups, suggests that this result is not due to the introduction of bias.
If AC-1202 is producing a positive effect in E4(-) subjects, it is reasonable to hypothesize that such an effect would be most notable in subjects who complied to the dosing schedule and actually took a substantial percentage of the investigational product. This hypothesis is supported by the analysis of dosage compliant subjects and the correlation between total dosage and improvement in ADAS-Cog at Day 90. The subjects that demonstrated the most response to therapy at Day 90 were E4(-) subjects who were dosage compliant (Figure ). Among E4(-) subjects, mean change from Baseline in ADAS-Cog score at Day 90 (independent of Placebo) was -1.7 points in the ITT w/LOCF population, -2.4 points in the per protocol population, and -3.9 points in the dosage compliant population. Dosage compliant subjects also performed better at Day 45 (-3.1 points) than either the per protocol population (-2.3 points) or the ITT w/LOCF population (-1.7 points). In addition, among E4(-) subjects, there was a significant correlation between total dose administered and change in ADAS-Cog score from Baseline at Day 90.
Figure 8 Summary graph of mean change from Baseline at Day 90 for ITT w/LOCF, per protocol and dosage compliant groups stratified by APOE4 carriage status. Red columns represent subjects receiving AC-1202. Blue columns represent subjects receiving Placebo. Error (more ...)
The positive effects of AC-1202 in E4(-) subjects is further supported by analysis of serum BHB levels and cognitive performance. AC-1202 resulted in significant elevation of serum BHB relative to Placebo at all study visits when investigational product was administered. In addition, a correlation between circulating BHB levels at the two-hour time point and improvement in ADAS-Cog score was noted in E4(-) subjects at Day 90. No significant correlation was found in E4(+) participants. Hence, higher levels of ketosis appear to confer greater benefit in the E4(-) group.
A common metric for clinically significant changes in AD trials is a 4 point change in ADAS-Cog after 6 months. This is frequently represented as the percent of subjects in each group who achieved this benchmark [56
]. In the present study, participants were only on therapy for 3 months, yet many of the subjects reached this level of improvement. Among E4(-) subjects receiving AC-1202, 31% (9/29) experienced a -4 point or greater improvement compared to 7.7% (2/26) in the Placebo group at Day 90. Among E4(-) subjects receiving AC-1202 who were also dosage compliant, 50% (8/16)) experienced a -4 point or greater improvement compared to 10% (2/20) in the Placebo group at Day 90.
We can only speculate on the mechanisms underlying the genotype-specific effects seen in this study, but there are reasons to suggest such an effect is not spurious. For example, E4(-) AD subjects seem to have greater relative benefits associated with some other therapies, such as infusion with glucose and insulin [58
], nasal insulin, [59
] or the insulin sensitizing agent rosiglitazone [60
]. One hypothesis is that there may be lower mitochondrial enzyme function in E4(+) versus E4(-) as noted in AD brain tissue samples [61
]. Reduced mitochondrial function may inhibit the ability of E4(+) participants to utilize ketone bodies and this may explain the apparent unresponsiveness to AC-1202 reported here.
An alternative explanation may be a differential insulin sensitivity of AD subjects based on APOE genotype [10
]. Ketone bodies are transported into the brain by monocarboxylate transporters [64
]. Levels of monocarboxylate transporters in the microvasculature are known to be low in adult mammals, yet elevated in diabetes and in other conditions where insulin resistance occurs [65
]. The milder insulin resistance in E4(-) AD subjects may allow them to more efficiently import ketone bodies into the brain and hence respond to AC-1202. Such a model is consistent with the observation that serum concentrations of BHB correlated with improvement in cognitive performance in E4(-) subjects, but not in E4(+) subjects. However, these models remain speculative and further experimentation will be required to confirm these differences and underlying mechanisms.
The rapid effects of induced ketosis in an earlier study [31
] and the loss of statistically significant differences in ADAS-Cog scores after the two week washout, suggests that AC-1202 may function to improve neuronal or glial metabolism in the presence of AD pathology, but may not slow the disease process (at least in a 3 month study). These results suggest that improvement in cognition may be feasible by mechanisms that do not address more typical targets, such as amyloid or tau. Improvement in mitochondrial efficiency, or activation of protective pathways, may provide a viable means to address AD. In support of this view, in animal models of AD, positive cognitive outcomes have been reported by interventions that do not lower amyloid or tau levels. For example, Halagappa et. al. demonstrated in a triple transgenic mouse model of AD that an intermittent fasting regime (which may elevate ketone bodies) produced cognitive benefits without affecting levels of Aβ or Tau [66
As this is a new area of AD research, our findings must be interpreted in that context. Participants administered Placebo demonstrated an unusual degree of worsening in ADAS-Cog scores at Day 45. The magnitude of this change is somewhat unexpected, although other studies have reported similar rates of decline [67
]. The lack of strong efficacy in MMSE and CGIC may be due to relative insensitivity of these tests, the small number of subjects, and/or the short duration of the trial. In addition, the reported deviations in the CGIC assessment scales cautions against over-interpretation of these results. The relatively high drop outs due to GI effects with our initial dosing regimen led to an intentional assignment by an independent monitor of a small number of subjects to the Placebo and AC-1202 groups, yet this did not appear to influence the overall outcome. It is possible that GI side effects might have biased outcomes in some subtle manner, although it must be noted that high rates of GI side effects are also noted in cholinesterase inhibitor trials. Group randomization was not stratified by APOE genotype, yet distribution of APOE4 alleles was similar between groups. Lastly, since this was a 3-month trial, this study cannot address the safety and efficacy of longer periods of administration.