What was lost in translation in the DHA trial is whom you should intend to treat

The results of a randomized double-blind placebocontrolled trial with docosahexaenoic acid (DHA) supplementation in mild to moderate Alzheimer's disease (AD) published by Quinn and colleagues in JAMA argues against overall efficacy of DHA in slowing progression. However, certain caveats in the results caution against discarding DHA altogether, raising questions about oxidation, dosage, pharmacogenomics and stage of intervention. One potential misconception is that what works for prevention will slow progression in AD subjects. Preclinical studies with DHA supported the rationale for early stage intervention; and three epidemiological studies indicated DHA intake was associated with reduced risk in non-apolipoprotein E4 (ApoE4) carriers. Putative drugs are initially tested for impact on progression because prevention approaches are problematic. However, should a drug be discarded for prevention if it fails to modify progression? Consistent with epidemiology, DHA significantly benefited two measures of cognition in mild to moderate non-ApoE4 carriers. Although the results of this trial were overall negative, failing to modify other outcomes, this commentary discusses important questions raised by them. Should future trials pursue DHA in non-ApoE4 carriers for slowing progression? Since in vivo oxidation of DHA may have adverse effects, particularly in ApoE4 patients, should preclinical and clinical studies be performed to optimize dose and mitigate oxidation before pursuing intervention or prevention trials with DHA? And finally, should DHA be tested now for mild cognitive impairment or prevention?


Abstract
The results of a randomized double-blind placebocontrolled trial with docosahexaenoic acid (DHA) supplementation in mild to moderate Alzheimer's disease (AD) published by Quinn and colleagues in JAMA argues against overall effi cacy of DHA in slowing progression. However, certain caveats in the results caution against discarding DHA altogether, raising questions about oxidation, dosage, pharmacogenomics and stage of intervention. One potential misconception is that what works for prevention will slow progression in AD subjects. Preclinical studies with DHA supported the rationale for early stage intervention; and three epidemiological studies indicated DHA intake was associated with reduced risk in non-apolipoprotein E4 (ApoE4) carriers. Putative drugs are initially tested for impact on progression because prevention approaches are problematic. However, should a drug be discarded for prevention if it fails to modify progression? Consistent with epidemiology, DHA signifi cantly benefi ted two measures of cognition in mild to moderate non-ApoE4 carriers. Although the results of this trial were overall negative, failing to modify other outcomes, this commentary discusses important questions raised by them. Should future trials pursue DHA in non-ApoE4 carriers for slowing progression? Since in vivo oxidation of DHA may have adverse eff ects, particularly in ApoE4 patients, should preclinical and clinical studies be performed to optimize dose and mitigate oxidation before pursuing intervention or prevention trials with DHA? And fi nally, should DHA be tested now for mild cognitive impairment or prevention? impairment, one small trial [6] and the larger 485-subject MIDAS (Memory Improve ment with DHA Study) trial [7] found signifi cant cognitive benefi ts with DHA. While there have been two fi sh oil trials in unimpaired elderly in which no cognitive benefi ts were observed [8,9], subjects in both were cognitively normal at baseline, and the latter failed to show signifi cant cognitive decline in the placebo group. Th is study argues that fi sh oil is not a cognitive enhancer, but does not examine disease modifi cation in subjects with pathology-driven memory defi cits. Second, animal studies report DHA/fi sh oil act on two pathological endpoints that plateau by mild to moderate stages: Aβ accumulation [10] and loss of superior cortical drebrin, an excitatory synaptic marker [11] (reviewed in [1]). Dramatic medial and superior temporal drebrin loss plateaus early with mild cognitive impairment by MMSE 26 [12], so loss has already occurred in trial subjects. While DHA reduced both Aβ and tau pathology in 3xTg AD mice [13], that intervention was early (prepathology). In contrast, with late post-pathology intervention in human tau transgenic mice with signi fi cant neuron loss, we fi nd DHA treatment is insuffi cient to produce signifi cant cognitive and synaptic improve ments (GMC and SAF, Society for Neuroscience presen tations, 2010). Finally, epidemiological risk factors may be relevant to prevention, but not necessarily to treatment. Animal model data with DHA support early intervention for primary prevention or mild cognitive impairment and suggest a failure to impact tangle and neuron loss driven defi cits at later stages.
Although the data demonstrate that DHA has no general benefi t for AD, a concern remains as to whether the key negative eff ect may be driven by the failure of ApoE4 subjects to respond. Since non-ApoE4 carriers comprise a large segment of the US (approximately 75%) and AD (approximately 50%) populations, whether DHA may slow progression in non-ApoE4 carriers is important. Figure 3 in [2] indicates that 40% of non-ApoE4 carriers showed signifi cant (P = 0.03) stabilization of both ADAS-Cog and MMSE, but not with correction for multiple comparisons. Th e authors point out that three epidemio logical studies showed reduced risk with fi sh consump tion only in the ApoE4 non-carriers, but add that pharmacogenomic interaction was not seen with CDR, ADL or NPI. For example, NPI showed a trend indepen dent of genotype, worsening less (2.93 points) in the DHA group than in the placebo group (5.09 points, P = 0.11). Are pathogenic mechanisms impacting NPI, ADL, CDR and MMSE/ADAS-Cog the same? Th us, any pharmacogenomic potential of DHA requires clarifi cation.
For prevention or treatment, one might expect ApoE genotype-DHA interactions. Because ApoE4 accelerates pathogenesis, age-matched ApoE4 patients may have more intractable AD pathology. Further, one important target of DHA is insulin resistance [14], but drugs targeting insulin resistance (insulin or peroxisome proliferatoractivated receptor (PPAR)γ agonists) appears more eff ective at reducing cognitive defi cits in ApoE3 carriers than ApoE4 carriers [15]. ApoE is a major central nervous system lipid transport protein with isoform-dependent traffi cking likely to impact DHA compartmentalization in the brain. Finally, ApoE4 increases oxidative stress, and with six double bonds, DHA is readily oxidized.
Th is raises other critical issues that need to be addressed before pursuing a future trial: dose and oxidation. Th e authors discuss the need to investigate potential combinations of DHA with antioxidants in AD patients, given apparent benefi ts with combinations of fi sh oil and lutein or lipoate in small trials and with antioxidants in the Souvenaid trial. Oxidation of DHA to neuroprostanes is associated with synaptic loss. Further oxidation produces a toxic end-product, 4-hydroxyhexenal, that contributes to neuron death and defective uptake of glucose by neurons and glutamate by astrocytes. Clinical studies demonstrate that similar dosing with marine n-3 fatty acids (polyunsaturated fatty acids with a double bond at the third carbon), including DHA, can deplete vitamin E and increase some peripheral measures of oxidative damage, particularly with dosing up to 6 months [16]. Because DHA is enriched in the brain where oxidative damage is already increased in AD patients, antioxidant supplements optimized for AD brain appear crucial. Even though marine n-3 fatty acids can deplete vitamin E, high dose vitamin E (900 IU) did not reduce measures of lipid peroxidation in human plasma [17], so vitamin E supplementation is probably not suffi cient. In mice the lipophilic phenolic antioxidant food additive butylhydroxytoluene attenuated measures of lipid peroxidation in plasma after high intake of fi sh oil [18]. Th e preclinical studies with DHA in AD mouse models require encapsulation of DHA in the chow to minimize oxidation [10,11]. Also, using the US Food and Drug Administration's equation to estimate the human equivalent doses, the clinical trial dose was three-fold higher than the effi cacious preclinical dose in mice [10,11], and twice as high as in the MIDAS trial [7], raising questions about whether the dose may have been too high, potentially exacerbating oxidative damage.
Possible cognitive benefi ts in patient subgroups (pharma co genomic or otherwise) would be strengthened by evidence of a biomarker response, arguing for the need to validate neuroimaging, cerebrospinal fl uid or plasma biomarker responses in preclinical studies going forward. MRI was performed in a small subset of subjects, showing that volumentrics of the left hippocampus in the DHA group showed trends to be smaller than in the placebo group (P = 0.17), which may indicate brain shrinkage. In the AN1792 active Aβ vaccination, MRI shrinkage was attributed to plaque clearance. Since drugs may only work in a subset of patients, it would be helpful in large studies where neuroimaging or cerebrospinal fl uid biomarker analysis are less feasible to identify likely responders with plasma biomarkers. A diffi cult task at hand is to design future DHA or other trials with earlier intervention to include validated surrogate and/or diagnostic biomarkers that have shown DHA responses in animal models. For tracking adverse eff ects of DHA, it is important to measure blood vitamin E depletion and lipid peroxides (thiobarbituric acid reactive substances, malondialdehyde, or the specifi c byproduct of DHA oxidation, 4-hydroxyhexenal). Biomarker validation could track mechanisms and lower trial costs and facilitate choice of effi cacious doses before proceeding to longer term, more costly trials to evaluate conversion to AD.

Conclusion
Th e study by Quinn and colleagues provides additional rationale to test DHA for prevention, with focus on non-ApoE4 carriers, but problems with DHA dosing and oxidation need to be addressed (particularly if an antioxidant could correct a failed ApoE4 response to DHA). Additional preclinical studies of stage-dependent effi cacy and ApoE4-DHA interaction may help to clarify whether ApoE genotype aff ects outcomes and how this can be mitigated, possibly with antioxidants or nonsteroidal anti-infl ammatory drugs (NSAIDs). Beyond pharmacogenomic roadblocks emerging with DHA and other interventions, all of the epidemiology and most of the animal model data that have been generated are most relevant to early stage interventions, but have been translated in clinical trials in mild to moderate AD, potentially resulting in an intent-to-treat the wrong group. Th e pre-clinical conclusions may not be wrong, but simply still lost in this translation.