Merck dropped a Valentine’s Day downer announcing they will stop the EPOCH trial of verubecestat in mild to moderate Alzheimer’s disease. A prespecified interim analysis by an external data-monitoring committee gave the trial “virtually no chance of finding a positive effect,” according to a Merck press release. The trial had been slated to run until July of this year. Merck’s APECS trial, which uses the same doses of verubecestat to treat people with prodromal AD, will continue. APECS results are expected by 2019.

Researchers were disappointed but not that surprised by the news. Scientists had questioned the rationale for using a BACE inhibitor in people who had accumulated enough Aβ deposition in the brain to have frank dementia (see Alzforum’s 2012 interview with Merck’s Johan Luthman). “The field needs some success, especially after the solanezumab results, but I was not holding my breath for this particular trial,” noted Robert Vassar, Northwestern University, Chicago, who co-discovered BACE. “It was conducted in people with mild to moderate AD, and since many cross-sectional studies have shown that amyloid has plateaued and there is substantial synaptic and neuron loss [in these patients], it seems highly unlikely that stopping or slowing production of Aβ at that stage would have a clinical benefit,” he told Alzforum. Verubecestat blocks BACE cleavage of the amyloid precursor protein, which is the first step in production of Aβ. Vassar also noted that because EPOCH, like early solanezumab and bapinezumab trials, did not require patients have a positive amyloid test, 25 to 30 percent of them likely had dementia other than Alzheimer’s. “That could kill any positive signal,” said Vassar.

Others agreed that trial may have started too late. “This is sad news for Alzheimer patients and their families,” wrote Stefan Lichtenthaler, German Center for Neurodegenerative Diseases in Munich, adding “the lack of a positive effect strengthens the need to test BACE inhibitors as early as possible” (see full comment below). Paul Aisen, University of Southern California, San Diego, agreed. “I remain optimistic that very early treatment with these drugs may by quite effective,” he wrote (see full comment below). Bart DeStrooper, KU Leuven, Belgium, noted that it may be tricky to determine when to start treating people with a BACE inhibitor because scientists still don’t know exactly how Aβ causes AD. It may be a trigger only very early in disease, becoming irrelevant later on, he wrote (see full comment below).

New trials may tell. In addition to Merck’s prodromal trial of verubecestat, other BACE inhibitors are being put to the test at various stages of the disease. The EARLY trial being run by Reisa Sperling, Brigham and Women’s Hospital, Boston, will test Johnson & Johnson’s BACE inhibitor JNJ-54861911 in people who test positive for amyloid but have no signs of dementia, while the DIAN trials unit will test this inhibitor in asymptomatic people who carry familial Alzheimer’s disease mutations. The Generation trial, run by Novartis, Amgen, and The Banner Alzheimer’s Institute in Phoenix, will test Novartis’ CNP520 in people who have two copies of the ApoE4 variant, a major genetic risk factor for AD (see Aug 2016 news). 

“The bottom line is we need prevention trials in asymptomatic individuals,” said Vassar. The paradigm for BACE inhibitors is statins, he noted. These drugs are given in early middle age, often for life, and they are successful for secondary prevention of heart disease by lowering cholesterol. Likewise, researchers hope that by lowering Aβ production, BACE inhibitors might prevent AD, he said.  “Giving a statin to someone in heart failure will not keep them alive, and in AD at the time of diagnosis we basically have [brain] organ failure,” Vassar said.      

Matt Kennedy from Merck told Alzforum that the company hopes to at least learn enough from this trial to inform other trials going forward. Kennedy said that a large number of the patients enrolled in the trial, which started in 2012, already have completed the 18-month dosing regimen. In previous trials Kennedy and colleagues had extensively tested target engagement, showing that verubecestat lowered production of Aβ by as much as 85 percent at the doses used in the EPOCH trial—12 and 40 mg daily (see Nov 2016 news). “We don’t have the data yet, but we’ll comb through [it] carefully and hope we can use it to inform about the prodromal trial and what other new ways we can use the molecule in different patients,” he told Alzforum. Kennedy noted that without seeing the full data, it is too soon to draw any formal conclusions about BACE inhibition in patients with mild to moderate AD. “But for BACE inhibition, we still feel that this molecule is appropriate to interrogate that mechanism,” he said.

On the plus side, the data monitoring committee noted no safety signals that warranted stopping EPOCH. Kennedy confirmed that with Alzforum. “It is heartening to know that the safety signal was not an issue,” he said. That also bodes well for other BACE inhibitors under evaluation.—Tom Fagan


  1. This is sad news for Alzheimer’s patients and their families. Scientifically, it is not surprising that verubecestat did not have a positive effect, given that the target patient population was already at an advanced disease stage. The lack of a positive effect strengthens the need to test BACE inhibitors as early as possible. It is great that Merck continues the APECS study, where prodromal AD is targeted, but I am convinced that we need to go even earlier in order to see a reasonably large effect size. This requires preventive trials and Merck should engage in such trials, still using verubecestat.

    The really good news, though, is that chronic BACE inhibitor dosing did not appear to induce major adverse events in patients. I hope that Merck will publish the apparently “minor” side effects, as this will be an important further validation of BACE1 as a drug target.

  2. This failure was indeed predicted as a possible outcome of Aβ therapy (see Figure 4 of Karran et al., 2011). If Aβ is only a trigger then it will be really difficult to get the therapy right, because we do not understand when and how it triggers disease. Most likely Aβ is a combination of trigger and threshold in the early phase of the disease, becoming irrelevant in the later phase.

    It is clear with the repetitive failure of the trials that we need urgently to step away from the amyloid cascade and the Cliff Jack progression curves in our thinking. These are extremely simplistic reductions and far too linear in their assumptions. Dementia is a complex phenotype involving not only synapse or neuron loss, but a strong disruption of cellular homeostasis. Aβ toxicity is only one part of a series of triggers as discussed in our “cellular phase” paper (see De Strooper and Karran 2016 and Alzforum webinar). 

    We need to think about Alzheimer’s disease as a multifactorial disorder and a multi-hit disorder just like all other complex disorders. Once astroglia and microglia, are involved it is no longer a biochemical process anymore. Taking away the biochemical problem will only be helpful before the cellular phase starts.

  3. This is very disappointing. But for a BACE inhibitor in particular, earlier intervention is likely critical. If AD is an amyloid brain disease (an oversimplification), then BACE inhibition would be expected to prevent the disease. But reducing Aβ production at the dementia stage, when amyloidosis has been present for many years and neurodegeneration is extensive, might have little effect on the clinical course. I remain optimistic that very early treatment with these drugs may by quite effective.

  4. One should not forget that certain side effects of drugs that interfere with APP cleavage, like BACE inhibitors, affect synaptic plasticity and cognition in mice (Filser et al., 2015; Zhu et al., 2016). Of course we do not know if this holds true for humans, especially in old age. Although verubecestat showed no obvious side effect on cognition in healthy individuals, it is not clear whether the drug could influence synaptic plasticity at this dose of treatment. But if it does, it may not be become obvious in clinical AD trials that have cognition as outcome measure, because it might mask the therapeutic benefit of reducing Aβ. The observation that compounds do not affect cognition in healthy controls or not yet demented people could be due to the "reserve capacity" of the brain to cope with a certain extent of synapse loss. If Merck deems this concern valid, it would be interesting to test whether individuals within EPOCH who do not have a significant plaque load (Amyloid PET) show a more severe decline in cognition after Verubecestat treatment.


    . Pharmacological inhibition of BACE1 impairs synaptic plasticity and cognitive functions. Biol Psychiatry. 2015 Apr 15;77(8):729-39. Epub 2014 Oct 29 PubMed.

    . Beta-Site Amyloid Precursor Protein Cleaving Enzyme 1 Inhibition Impairs Synaptic Plasticity via Seizure Protein 6. Biol Psychiatry. 2016 Dec 26; PubMed.

  5. Merck pushing forward on clinical trials for a BACE1 inhibitor in such late stages of AD is certainly puzzling, as others have already noted. It is to be expected that, once Aβ has plateaued and sufficient Aβ-related damage has been done, there will be little symptomatic effect from inhibiting BACE1. If the solution to ineffectiveness of modifying BACE1 activity is to administer an inhibitor before any cognitive, biomarker, or any other signs or symptoms at all appear, this raises a different problem. Herms has already mentioned the risk of altering synaptic plasticity from long-term BACE1 inhibition. To this we would like to further add that significant BACE1 activity is not restricted to the central nervous system. BACE1 also functions in the liver, where it is the primary enzyme to process α2,6-syalyltransferase (ST6Gal1) from membrane-bound to secreted form (Kitazume et al., 2005). ST6Gal1 is vital in response to radiation, oxidative, and mechanical damage of the liver and other tissues (Kitazume et al., 2009; Lee et al., 2012). Upregulation of ST6Gal1 accompanied noninvasive human bladder tumors while downregulation accompanied muscle-invasive bladder cancer (Antony et al., 2014). Silencing of ST6Gal1 inhibited human influenza virus infection of respiratory epithelial cell cultures (Wu et al., 2014), but humoral response to influenza virus infection is deficient in ST6Gal1 KO mice (Zeng et al., 2009). In terms of direct hepatotoxicity, an ST6Gal1 SNP (rs10937275) is associated with drug-induced liver toxicity (Aithal et al., 2015), and hepatotoxic extracts from comfrey and the pyrrolizidine alkaloid riddelliine downregulated expression of ST6Gal1 when fed to rats (Guo et al., 2007). In addition, levels of circulating (cleaved) ST6Gal1 did not correlate well with levels of typically measured liver enzymes, such as ALP, ALT, or AST (Kitazume et al., 2009), suggesting that conventional liver panels may not reveal earlier stages of all forms of liver damage.

    We are all already aware of an earlier drug trial that failed due to what was described as hepatotoxic effects (Jun 2013 news). While the presumption is that any hepatotoxicity is an "off-target" result, given BACE1's proven activity in the liver, it is possible that any apparent hepatotoxicity could be due to fully on-target activity. BACE1 could act on a "correct" substrate, but in a site other than what would be desired (Lahiri et al., 2014). Such "on-target, off-site" activity (OnTOS) could be confounding drug discovery. Taking OnTOS into account may reduce such risks. It may be fruitful to investigate organ-specific BACE1 downregulation, such as via organ-specific profiles of BACE1-regulating miRNA levels. Alternatively it may be fruitful to explore targeting of BACE1 inhibitors to specific subcellular compartments, so as to avoid inhibiting BACE1 processing of neuregulin-1, for example.

    In addition, drugs such as BACE1 inhibitors presumably would have to be administered for years or decades, especially if one must start them before the slightest sign of Alzheimer's and continue them indefinitely. Such a process could produce a lifetime “polypharmic” cascade, where a person begins taking a "preventative" drug with no known acute side effects, but chronic administration results in long-term unacceptable side effects, which would be treated by another drug, which would have to be administered as long as the first drug is taken. This combination could result in further slow-emerging side effects due to OnTOS, leading to another drug prescription, and so on in an increasing spiral.

    Prevention is better than cure, and a treatment that invades least is an ideal. A large amount of evidence has amassed that environmental factors play significant roles in inducing Alzheimer's disease and other dementias (Maloney and Lahiri, 2016). Thus, appropriate use of lifestyle measures such as exercise, appropriate diet, and sociocultural enrichment may end up playing as large or larger roles than pharmaceutical interventions in the future of medical response to Alzheimer's. Such methods would avoid a polypharmic cascade, altogether. In the meantime, it could be useful to more explicitly consider OnTOS—on-target molecular effects that occur in an unwanted (or even unforeseen) site.


    . In vivo cleavage of alpha2,6-sialyltransferase by Alzheimer beta-secretase. J Biol Chem. 2005 Mar 4;280(9):8589-95. PubMed.

    . Molecular insights into beta-galactoside alpha2,6-sialyltransferase secretion in vivo. Glycobiology. 2009 May;19(5):479-87. PubMed.

    . Cleavage of ST6Gal I by radiation-induced BACE1 inhibits golgi-anchored ST6Gal I-mediated sialylation of integrin β1 and migration in colon cancer cells. Radiat Oncol. 2012 Mar 27;7:47. PubMed.

    . Epigenetic inactivation of ST6GAL1 in human bladder cancer. BMC Cancer. 2014 Dec 2;14:901. PubMed.

    . Gene silencing of β-galactosamide α-2,6-sialyltransferase 1 inhibits human influenza virus infection of airway epithelial cells. BMC Microbiol. 2014 Mar 27;14:78. PubMed.

    . The generation of influenza-specific humoral responses is impaired in ST6Gal I-deficient mice. J Immunol. 2009 Apr 15;182(8):4721-7. PubMed.

    . Genome-Wide Association Studies in Drug-Induced Liver Injury: Step Change in Understanding the Pathogenesis. Semin Liver Dis. 2015 Nov;35(4):421-31. Epub 2015 Dec 16 PubMed.

    . Comparison of gene expression profiles altered by comfrey and riddelliine in rat liver. BMC Bioinformatics. 2007 Nov 1;8 Suppl 7:S22. PubMed.

    . Lessons from a BACE1 inhibitor trial: off-site but not off base. Alzheimers Dement. 2014 Oct;10(5 Suppl):S411-9. Epub 2014 Feb 12 PubMed.

    . Epigenetics of dementia: understanding the disease as a transformation rather than a state. Lancet Neurol. 2016 Jun;15(7):760-74. Epub 2016 May 9 PubMed.

  6. Let us hope that prodromal AD will benefit from verubecestat treatment. A problem with even this relatively early stage of disease is that Aβ may have already accumulated, which will make a negative study difficult to interpret. This confound is in addition to variation caused by inclusion of other dementias, as mentioned by Bob Vassar. It has been reported that BACE1 transgenic mice performed poorly in a Morris water maze in the absence of Aβ deposits, regardless of the presence of transgenic human APP (Rockenstein et al., 2005). It will be useful to see if verubecestat can correct behavior deficits in this BACE1 model, to demonstrate that high BACE1 activity, rather than accumulation of BACE1 protein, is the cause of memory failure in these mice. The question may also be relevant to AD, because BACE1 accumulates in the AD brain (Ahmed et al., 2010; Li et al., 2004).


    . High beta-secretase activity elicits neurodegeneration in transgenic mice despite reductions in amyloid-beta levels: implications for the treatment of Alzheimer disease. J Biol Chem. 2005 Sep 23;280(38):32957-67. PubMed.

    . BACE1 and BACE2 enzymatic activities in Alzheimer's disease. J Neurochem. 2010 Feb;112(4):1045-53. PubMed.

    . Amyloid beta peptide load is correlated with increased beta-secretase activity in sporadic Alzheimer's disease patients. Proc Natl Acad Sci U S A. 2004 Mar 9;101(10):3632-7. PubMed.

  7. The reported premature halt of the BACE inhibitor is a substantial setback, as the field desperately needs some success. Obviously, we need to scrutinize the clinical data in more detail before we can identify possible reasons. However, together with the (modest) clinical improvement of the passive vaccination trials, this probably suggests a deeper issue with all amyloid-modulating interventions. One way to interpret these data consistently is to assume that Aβ is much less neurotoxic than previously assumed and exerts its effect on cognition by interacting with neurophysiological processes such as glutamate (Wang 2013) and nicotinic neurotransmission, to name a few. In addition, there is experimental evidence that some Aβ peptides are actually improving synaptic function.

    In the spirit of thinking outside of the box, we have developed a Quantitative Systems Pharmacology computer model of the interaction between Aβ and glutamate and α7 nAChR, based on electrophysiological data on paired-pulse stimulation that can recapitulate a number of unexpected clinical outcomes (Geerts 2014). Simulation of BACE inhibition with pharmacodynamic effects on biomarkers from clinical studies (Van Maanen 2016) suggests that the clinical effect ranges from worsening at low levels of Aβ to a modest improvement at higher levels of Aβ. In addition, high doses of BACE inhibition are less effective, likely because of the beneficial effect of short Aβ isoforms on glutamate neurotransmission. All these considerations make it likely that, on average, clinical trial results with a BACE inhibitor will be very modest. 

    In contrast, the effects of passive vaccination aimed at monomers is much less sensitive to Aβ load but are equally modest. Clinical data from both the solanezumab and verubecestat trials can confirm or refute these predictions, and it would be great to have access to these clinical data as they might have important implications for the ongoing trials in prodromal or healthy volunteers.

    These model outcomes might also explain the fact that many of the amyloid interventions work well in transgenic animals, as they often have (artificially) high amyloid levels.

    As for patient selection, cholinesterase inhibitors are able to improve clinical readouts in mild to moderate AD patients, so there is certainly still room for improvement.


    . Using in silico mechanistic disease modeling to address the effect of amyloid beta manipulation on cognitive clinical readouts. Alzheimers Dement. 2011 Jul; 7(4 Suppl),S774–5.

    . Systems Pharmacology Analysis of the Amyloid Cascade after β-Secretase Inhibition Enables the Identification of an Aβ42 Oligomer Pool. J Pharmacol Exp Ther. 2016 Apr;357(1):205-16. Epub 2016 Jan 29 PubMed.

    . Multiple effects of β-amyloid on single excitatory synaptic connections in the PFC. Front Cell Neurosci. 2013;7:129. Epub 2013 Sep 3 PubMed.

Make a Comment

To make a comment you must login or register.


Therapeutics Citations

  1. Verubecestat
  2. Solanezumab
  3. Atabecestat
  4. Umibecestat

News Citations

  1. Q&A With Merck’s Johan Luthman
  2. New Ways to Target Aβ and BACE Show Promising Phase 1 Data
  3. Paper Alert: Verubecestat Preclinical and Phase 1 Data Published

External Citations

  1. Merck press release
  2. trial
  3. APECS trial

Further Reading