. Retinal Optical Coherence Tomography Metrics Are Unchanged in Verubecestat Alzheimer's Disease Clinical Trial but Correlate with Baseline Regional Brain Atrophy. J Alzheimers Dis. 2021;79(1):275-287. PubMed.


Please login to recommend the paper.


  1. The BACE 1 inhibitors are a powerful and specific class of drugs. They were used at very high doses, aiming for 70-90 percent inhibition of APP processing (as were the γ-secretase inhibitors). With the benefit of hindsight, we should not be surprised that they ran into major side effects. It would be reasonable to assume that these side effects are the result of on-target inhibition of APP processing. After all, this was what they were designed to do. If the normal cleavage of APP is to release the extracellular and intracellular domains, and if the normal function of APP is for heterotypic interactions at the peri-synaptic zones which are used for synaptic plasticity (for learning and memory functions), why, then, should we not expect 80 percent inhibition to result in deficits in learning and memory?

    We have previously published data to suggest that the amount of hourly accumulation of Aβ is only 5 percent of total hourly production (Roberts et al., 2017). Similar data has been published by others, e.g. the Bateman lab. This suggests to us that a BACE 1 inhibitor used at a dose aimed to inhibit Aβ production by 5 to 10 percent would have been appropriate, at least in the mildest AD cases. It’s worth noting that the high-dose BACE 1 inhibitors did have an Aβ-lowering effect based on PET data.

    The atrophy seen with verubecestat in amyloid-rich regions (Sur et al., Brain) could be interpreted as an on-target effect on APP processing, as these amyloid-rich regions are possibly related to the amount of normal processing of APP in these topographic regions. Similarly, the cognitive deficits described by Wessels et al. (Alz Dem) could also be the effect of the drug on normal APP processing. Both of these statements are speculative, but could be tested experimentally.

    Hopefully we will see the rehabilitation of these drug classes (both β- and γ-secretase inhibitors) at some stage in the future. One scenario is a monoclonal antibody immunotherapy lowering the amyloid burden to baseline over one to two years, followed by maintenance therapy with an amyloid production inhibitor dosed at 5 to 10 percent inhibition, to keep the amyloid burden at baseline levels.


    . Biochemically-defined pools of amyloid-β in sporadic Alzheimer's disease: correlation with amyloid PET. Brain. 2017 May 1;140(5):1486-1498. PubMed.

    View all comments by Colin Masters
  2. I congratulate both companies on sharing more detailed data from the terminated trials and on their collaborative effort to compare two trials and inhibitors. This is very important to better understand the effects, but also the side effects, of BACE inhibitors. I am confident that the other companies will similarly report more data on their trials.

    The comparative paper demonstrates impressively that a detailed analysis of the diverse cognitive tests is important, as it shows previously unanticipated cognitive worsening of lanabecestat—not globally, but on a subset of tests, whereas improvement in other tests (verbal fluency) can now also be appreciated.

    Regarding the retina analysis, it is reassuring to see that the retina is not affected by BACE inhibition with verubecestat. This was initially a concern based on retina degeneration observed in one BACE1-deficient mouse line (Cai et al., 2012), but this phenotype was not seen in an independently generated BACE1-deficient mouse line (Rudan Njavro et al., 2020).

    Although the molecular basis of the cognitive side effects is not yet clear, the three studies together provide very helpful guidance for possible future BACE inhibitor trials. First, a detailed ophthalmological analysis of trial participants is not needed anymore. Second, specific cognitive tests, such as digit symbol coding, may help to evaluate in a short trial (e.g., three months) the right dose of BACE inhibition, ideally less than 50 percent, which precludes cognitive side effects. Third, the lack of an effect of BACE inhibition on NFL levels in individuals with diagnosed AD demonstrates that the future of BACE inhibition is for prevention and not treatment of AD.


    . β-Secretase (BACE1) inhibition causes retinal pathology by vascular dysregulation and accumulation of age pigment. EMBO Mol Med. 2012 Sep;4(9):980-91. PubMed.

    . Mouse brain proteomics establishes MDGA1 and CACHD1 as in vivo substrates of the Alzheimer protease BACE1. FASEB J. 2020 Feb;34(2):2465-2482. Epub 2019 Dec 27 PubMed.

    View all comments by Stefan Lichtenthaler
  3. The authors provide important findings that may help to answer some current uncertainties, as well as guide future consideration of this or a similar class of compounds:

    • Mechanism-based versus compound-based. Although both compounds share similarities, there are differences, and with all BACEi trials now having been discontinued, it is clear that this is related to this class of therapies. Although the mechanism remains uncertain, a related study published by the Merck team focusing on a detailed analyses of the neuroimaging data does provide important information on how measures of brain atrophy and amyloid plaque could help to clarify (see below).
    • The relatively robust finding of an early impairment on the Digit Symbol Coding test used in both trials suggests this test could be used in future trials with BACEi at lower doses as a sensitive, early marker of detrimental effects. 
    • Discrepancies between language and non-language tasks. Previous reports on the Janssen compound had suggested an improvement in tests of verbal fluency. Interestingly, both compounds reported here posted an improvement in both letter and category fluency. Although this effect was not reported in the Novartis BACEi trial, that trial was conducted in cognitively normal subjects, not all of whom had evidence of significant amyloid deposition. Thus, this effect may be real and stage-dependent. If so, it remains uncertain mechanistically but worth further study. 

    Unfortunately, we do not know whether these cognitive effects would be identified at lower levels of BACE inhibition.

    An important implication of the publication in the journal Alzheimer’s and Dementia is collaboration between two companies in combining data for further analyses and release to the public. This should serve as a model for future publications, particularly when there are compelling reasons (patient safety) and an obvious logic (significant overlap in the mechanism of the drug and the study outcomes).

    In the neuroimaging study performed by Merck, their academic colleagues have taken a closer look at the regional atrophy patterns, regional associations with amyloid plaque load, as well as the longitudinal CSF measures of “neurodegenerative” markers to better evaluate the adverse effects of BACE inhibition by verubecestat. They demonstrate that although the regional volumetric changes, which were worse in the verubecestat groups and similar for both doses, were not strongly associated with the adverse cognitive effects, they were similarly detected at the earliest time point of measurement and were non-progressive, just like the cognition effect. Although there was not a strong association between any imaging measure of cognitive performance, these updated analyses clearly indicate 1) a strong temporal relationship between greater brain atrophy and the initial cognitive worsening and 2) that an acceleration of the AD process or neurodegeneration are unlikely the mechanistic cause. Again, as is the case with cognition, whatever the mechanistic link is appears to occur at BACE inhibition greater than 50-60 percent.

    Previous studies of other amyloid-specific therapies have also identified worsening atrophy in the treated groups compared to placebo, hence the authors explored the modest decline in amyloid PET SUVR levels with the change in volumetric measures. They found no clear relationship, suggesting a clearing of amyloid plaques was unlikely to explain the change in brain atrophy. They did find a specific effect of worsening atrophy with BACE inhibition in brain regions with significant amyloid plaques and suggested that this was a potential link between brain atrophy from BACE inhibition. However, the areas of comparison (e.g., low-amyloid areas) were limited and unlikely to explain the differential volumetric losses compared to the placebo in the first place. Thus, it is remains uncertain whether BACE inhibition in areas with or without substantial amyloid plaque pathology might help to understand the detrimental effects of BACE inhibition. This last point is important, as it would be helpful in considering the potential future use of BACE inhibitors at lower doses in at-risk populations and whether amyloid-negative or -positive participants would be at greater risk of accelerated atrophy. The detailed analyses of the Novartis BACE inhibitor may provide some answers to this question.

    Similar to the adverse cognitive effects, it remains unknown whether the volumetric differences between verubecestat and placebo normalize once BACE inhibition stops. Nonetheless, these two studies complement each other in demonstrating that BACE inhibitors at APP inhibition of greater than 50 percent clearly result in an early, consistent worsening of cognition (with the exception of verbal fluency) and atrophy in multiple brain regions, but that in both cases a continued acceleration of the neurodegenerative process is unlikely to be the cause. Likewise, they both provide important safety guidance on measures that could be used in future trials for therapies targeting the secretase pathway as a prevention or symptom therapy in AD.

    View all comments by Eric McDade
  4. Fundamental questions around local and temporary cognitive side effects of BACE inhibitors, and also γ-secretase inhibitors, become resolved as soon as we allocate a physiological role to Aβ monomer homeostasis in synaptic neuronal function in normal individuals and spontaneous AD cases. BACE elevation in early amyloid pathology is then regarded not as a cause of the pathology, but rather, a compensatory response to lack of physiological β-amyloid concentration at the level of the synapse. This β-amyloid dysregulation is supposed to start as soon as folding control by chaperones is impaired in the physiological β-amyloid regulation cycle.

    Thus, damping the compensatory Aβ production via BACE inhibitors is the wrong approach to tackle Aβ pathophysiology in spontaneous AD. The right approach will need to be removal of early misfolded Aβ species to restore normal Aβ homeostasis at the synapse. BACE elevation is expected to normalize and take a role as biomarker upon successful Aβ oligomer treatment.

    A clinical role for BACE inhibitors at low concentrations might be reserved for co-therapy in certain cases of familial AD.


    . The Beta Amyloid Dysfunction (BAD) Hypothesis for Alzheimer's Disease. Front Neurosci. 2019;13:1154. Epub 2019 Nov 7 PubMed.

    View all comments by Heinz Hillen
  5. Our primary manuscript from the EARLY trial of the BACE inhibitor atabecestat is in press. In the paper, we’ve addressed the interesting findings noted in the papers by Merck and Lilly. The EARLY trial was truncated due to hepatic safety concerns but in the preclinical AD population for which we obtained data, we did not observe improvement in the semantic fluency subtest on RBANS, although this subtest did not decline as clearly as episodic memory measures. Whole brain volume on MRI decreased at six months and 12 months in a dose-dependent manner. Unfortunately, MRI data were very limited beyond 12 months.

    The reasons for the disparity between our results and those of Merck and Lilly are unclear and should be interpreted with caution as our results come from an incomplete trial. However, an obvious difference between the programs is the population studied. In EARLY, we enrolled preclinical AD participants with elevated brain amyloid and either no symptoms or subjective cognitive decline but considered cognitively unimpaired on the Clinical Dementia Rating (CDR global score 0). These participants likely started with more intact synaptic structure and function. How this would lead to the differences observed is a question requiring further study.

    Our trial design permitted us to follow participants in the study without treatment. After up to six months of follow-up off treatment, the cognitive worsening observed on treatment appeared to recover, with cognition returning to baseline levels. Consistent with the findings from other sponsors, we did not observe greater increases in neurofilament light chain (NfL) with atabecestat compared to placebo.

    Taken together, the EARLY trial data seem to corroborate the hypothesis that BACE inhibition does not cause irreversible neurodegeneration. It may be that with 18- or 24-month volumetric MRI data in the EARLY trial, the decrease in whole brain volume would have plateaued. However, the field should also consider the possibility that a preclinical AD population with more intact synapses may be more susceptible to effects of BACE inhibition.

    View all comments by Michael C. Donohue

Make a Comment

To make a comment you must login or register.

This paper appears in the following:


  1. New Data from Past BACE Inhibitor Trials Shed Light on Side Effects