In a Phase 2 trial of Bristol-Myers Squibb’s oral γ-secretase inhibitor avagacestat, AD patients who took high doses suffered more adverse effects and appeared to fare worse cognitively than those on placebo. Christopher van Dyck of Yale University School of Medicine, New Haven, Connecticut, showed this at the 2011 Alzheimer’s Association International Conference in Paris (see ARF conference story). Now, the detailed findings appear in print in the November Archives of Neurology, with company scientists Vladimir Coric and Robert Berman of BMS in Wallingford, Connecticut, as lead and senior authors, respectively. Experts find little promise in the trial results, calling them a red flag for γ-secretase inhibitors in general. If a shred of hope remains, they say, it may come from the company’s ongoing Phase 2 study in prodromal AD.

Avagacestat—also known as BMS-708163—initially looked good because it seemed to overcome a key problem of γ-secretase inhibitors: gastrointestinal and immune system troubles linked to other substrates, most notably Notch (Milano et al., 2004; Wong et al., 2004). In-vitro work reported by company scientists suggested the BMS inhibitor is ~190-fold more selective for blocking processing of amyloid-β precursor protein (APP) than of Notch (see ARF conference story). Phase 1 data seemed supportive, as single doses up to 800 mg and multiple doses up to 150 mg a day caused no adverse effects.

The six-month Phase 2 study tested four doses of the γ-secretase inhibitor against placebo in 209 people with mild to moderate AD at 41 sites in the U.S., Denmark, Finland, and Sweden. The lower doses (25 and 50 mg) were relatively well tolerated. Participants receiving 100 and 125 mg of the compound daily had gastrointestinal, skin, and other side effects that prompted study withdrawal more than twice as frequently as in the lower-dose or placebo groups.

At the two lower doses, cerebrospinal fluid (CSF) biomarker data were mixed. CSF levels of Aβ1-34, Aβ1-14, Aβ1-15, and Aβ1-16 rose in a dose-dependent manner. Produced by α-secretase processing of APP, these short Aβ isoforms provide indirect evidence of γ-secretase inhibition, suggesting target engagement (Portelius et al., 2011). However, Aβ1-40 and Aβ1-42 levels did not budge in treated patients. The authors attributed this to small sample size (only 45 of 209 participants agreed to lumbar puncture) and/or methodologic challenges associated with CSF measurements.

Perhaps most concerning was that the clinical data showed trends toward cognitive decline at the highest two doses. This is “eerily similar” to the data on Eli Lilly’s γ-secretase inhibitor semagacestat, suggested Sascha Weggen of Heinrich-Heine University in Düsseldorf, Germany. In a Phase 3 trial, the Lilly compound made AD patients worse, forcing the company to halt the study in 2010 (see ARF related news story). All told, Weggen found the Phase 2 avagacestat data “highly worrisome for γ-secretase inhibitors as a class.”

Michael Wolfe of Brigham and Women’s Hospital, Boston, Massachusetts, said the results indicate “that, first and foremost, we need to have agents that are much more selective for inhibiting the cleavage of APP over that of the Notch receptor.” In an e-mail to Alzforum, Wolfe suggested that the avagacestat data call into question the compound’s preference for APP. “If [the inhibitor] is indeed as selective as BMS says, then it needs to be more selective still,” he wrote.

Other research challenges the claimed APP vis-à-vis Notch selectivity (Chávez-Gutiérrez et al., 2012; Crump et al., 2012). These studies “did not see any selectivity with the BMS compound, Weggen noted, adding that his lab has “similar unpublished data on avagacestat, suggesting at best a 10-fold difference between inhibition of APP and Notch processing.”

Aside from Notch, some researchers believe that problems linked to γ-secretase inhibitors could stem from their effects on APP processing. Cell culture studies suggest that endogenous Aβ may be beneficial, promoting synaptic plasticity by regulating glutamate release (ARF related news story on Abramov et al., 2009). If true, then reducing Aβ levels greatly by way of γ-secretase inhibition could potentially harm cognition, noted Hugo Geerts of In Silico Biosciences in Berwyn, Pennsylvania. In another study, avagacestat treatment did not improve learning and memory in APP mice, and actually worsened cognition in normal mice (ARF related news story on Mitani et al., 2012), whereas Notch-controlled gene expression remained unchanged. The authors attributed the cognitive decline instead to accumulation of APP C-terminal fragments, suggesting these peptides could be harmful to people as well.

Meanwhile, scientists await results from the ongoing Phase 2 trial of avagacestat in prodromal AD, in which people at risk for AD receive 50 mg of the inhibitor daily for two to four years. “Interim data from this study and data from the completed Phase 2 study in mild to moderate Alzheimer’s will inform our plans for further development of this compound,” Sonia Choi of Bristol-Myers Squibb Public Affairs told Alzforum via e-mail. “We plan to complete an interim analysis this quarter.” She said the company is “not in a position to facilitate discussion” about the Phase 2 trial results at this time, meaning its scientists are muzzled for the time being. The prodromal AD trial began in May 2009. Enrollment was challenging, but is complete; the company screened 1,350 candidates to find about 270 who met criteria for amnestic mild cognitive impairment (aMCI) and had CSF evidence of AD pathology (see ARF conference story).—Esther Landhuis.

Coric V, van Dyck CH, Salloway S, Andreasen N, Brody M, Richter RW, Soininen H, Thein S, Shiovitz T, Pilcher G, Colby S, Rollin L, Dockens R, Pachai C, Portelius E, Andreasson U, Blennow K, Soares H, Albright C, Feldman HH, Berman RM. Safety and Tolerability of the γ-Secretase Inhibitor Avagacestat in a Phase 2 Study of Mild to Moderate Alzheimer Disease. Arch Neurol. Nov 2012;69(11):1430-1440. Abstract


Make a Comment

To make a comment you must login or register.

Comments on News and Primary Papers

  1. To understand why semagacestat and avagacestat failed, we need to understand the molecular properties of those compounds. Enzymatic mechanism studies offer insights into why those compounds can facilitate cognitive decline, and why current drug development strategies need to change, as we outlined in our recent paper (see Svedružić et al., 2013).


    . Modulators of γ-secretase activity can facilitate the toxic side-effects and pathogenesis of Alzheimer's disease. PLoS One. 2013;8(1):e50759. PubMed.

  2. My humble opinion is that this is a red flag for pure γ-secretase inhibitors (GSIs). The trend for worsening cognition was also seen with the Lilly GSI semagacestat. Even when taking into account the much more selective property of avagacestat to spare Notch signaling, this suggests an alarming and possibly fundamental issue with lowering soluble amyloid-β. We don’t know at this point what the situation might be with γ-secretase modulators.

    The large majority of the scientific community still believes that reducing amyloid-β will improve cognition (it does, at least, in animal studies), but there is an increasing realization that amyloid-β might have a beneficial role for synaptic function by regulating glutamate release (Abramov et al., 2009). Other studies suggest that some FAD point mutations in presenilins will lead to a loss of function, leading to lower Aβ production (Chavez-Guttierez et al., 2012). In the human brain, more irreversible processes might also be at play, as Eric Siemers from Lilly reported that the subjects were still performing worse six months after washout of semagacestat.

    Conversely, there are many arguments to be made for a toxic effect of β amyloid (oligomers). However, many of those are based on animal studies, which are often not that predictive of human outcomes, as they start from a different baseline in a different environment. I hope that with the data from all the current amyloid-modulating trials together, using quantitative systems pharmacology approaches, we could start piecing together a better picture of the shifts among the different forms of amyloid-β (soluble, oligomeric, or fibrillar) that are changed with the different therapeutic interventions and their effect on the human AD brain.

    BMS is pushing ahead with a predementia study; however, enrolling presymptomatic (MCI?) patients is not easy, as only about one in five is suited to enter an amyloid-focused trial. Unfortunately, I think the outcome will be very similar at doses that show a robust target engagement; i.e., cognition will likely worsen, not to mention the other severe adverse effects such as pneumonia, vasogenic edema, skin carcinoma, and renal dysfunction. Therefore, it will be ethically necessary to test the cognitive performance of the subjects every six months and have the review board halt the trial if there is the slightest evidence of a deterioration in the active group or futility in clinical outcomes.

    Finally, we should not forget that even in the best-case scenario, getting rid of the amyloid pathology might still leave us with a tauopathy.

    View all comments by Hugo Geerts
  3. These cognitive effects, along with the adverse events in the gastrointestinal tract and skin, are troubling, as they are similar to the outcomes Lilly had with semagacestat. It's not the end of γ-secretase as a target, but it says that, first and foremost, we need to have agents that are much more selective for inhibiting the cleavage of APP over that of the Notch receptor. Inhibiting Notch signaling has been the primary problem for γ-secretase inhibitors to date, and there is reason to question the evidence that avagacestat is very selective for APP vis-à-vis Notch. And if it is indeed as selective as BMS says, then it needs to be even more selective still.

    View all comments by Michael Wolfe
  4. This study describing Phase 2 results of the γ-secretase inhibitor avagacestat comes down to two facts: clear evidence of toxicity at the two higher doses—even with the relatively short treatment period of 24 weeks—and no evidence of target engagement at the two lower doses.

    The observed trend for cognitive decline with the two higher doses (significant for ADAS-cog at 100 mg) is eerily similar to the data from the semagacestat trial, and is highly worrisome for γ-secretase inhibitors as a class. The authors state in the discussion that the cognitive decline was not correlated with worsening MRI (more atrophy) and CSF tau results, and that this supports the conclusion that the cognitive decline was not due to an enhancement of the underlying neurodegenerative disease process. I would argue that this is irrelevant if the cognitive decline is mechanism based (due to γ-secretase inhibition). As a basic researcher, I would be most interested in what is causing this cognitive decline. I can see four (simplified) explanations.

    1. Notch inhibition: Avagacestat has been presented by BMS as a Notch-sparing GSI with a 240-fold window between APP and Notch inhibition in in-vitro studies. However, this has been questioned by two published studies, from the labs of Bart de Strooper (Chávez-Gutiérrez et al., 2012) and Yue-Ming Li (Crump et al., 2012). Neither of these studies saw any selectivity with the BMS compound. We also have similar unpublished data that suggest at best a 10-fold window between APP and Notch with avagacestat. Having said that, things could be different in vivo in humans. In the Archives paper, peripheral markers of Notch signaling did not change with avagacestat treatment. However, that does not exclude Notch inhibition in the brain, which BMS was not able to assess.

    2. Inhibition of APP processing: Recently, APP-transgenic mice were treated with avagacestat by Mitani et al. (2012), and the authors also found evidence for cognitive decline in sub-chronic dosing studies. Notch target gene expression appeared to be unchanged in the brains of the avagacestat treated mice, and the authors attributed the cognitive decline instead to the accumulation of APP C-terminal fragments. Neuronal toxicity of APP-CTFs has been proposed for many years.

    3. Inhibition of another, not yet identified substrate of γ-secretase.

    4. Non-mechanism-based toxicity of the compound. The first three explanations would all appear to exclude GSIs from further clinical development. The fourth explanation is possible, but does not seem very likely. Obviously, it will be extremely difficult to nail down the cause for the cognitive decline. For example, to assess the expression of Notch target genes in response to avagacestat, brain biopsy material would be required, which is impossible to get.

    Concerning target engagement, this study is, in my view, unequivocally negative for the two lower doses. There was no significant change in Aβ38/40/42 or tau/phospho-tau. The exploratory data, in which the authors looked at Aβ1-14/1-15/1-16/1-34 levels as a surrogate readout for γ-secretase activity, offer little. Even if these data could be reproduced in a larger study, and even if they indicated some reduction in γ-secretase activity (as the authors argue in the discussion), what does a decrease in Aβ1-34 mean if the "real" target Aβ42 is unchanged? A significant change in Aβ42 levels was only observed for the highest 125 mg dose. I do not see anything else in the results with the two lower doses that would make me hopeful for further development. Aside from the already known issues with avagacestat, more and more evidence suggests that very early (preventive) and long-term treatment with anti-amyloidogenic agents might be required to reach meaningful clinical outcomes. For such an intervention, avagacestat appears to be ill suited.

    View all comments by Sascha Weggen


News Citations

  1. Paris: Semagacestat Autopsy and Other News of Trial Tribulations
  2. DC: New γ Secretase Inhibitors Hit APP, Spare Notch
  3. Lilly Halts IDENTITY Trials as Patients Worsen on Secretase Inhibitor
  4. What’s My Line?—Fishing for the True Role of Aβ
  5. Paper Alert: γ-Secretase Modulators Trump Inhibitors
  6. CTAD: Turning the Ship Around Toward Early Trials

Therapeutics Citations

  1. Avagacestat

Paper Citations

  1. . Modulation of notch processing by gamma-secretase inhibitors causes intestinal goblet cell metaplasia and induction of genes known to specify gut secretory lineage differentiation. Toxicol Sci. 2004 Nov;82(1):341-58. PubMed.
  2. . Chronic treatment with the gamma-secretase inhibitor LY-411,575 inhibits beta-amyloid peptide production and alters lymphopoiesis and intestinal cell differentiation. J Biol Chem. 2004 Mar 26;279(13):12876-82. PubMed.
  3. . Novel aβ isoforms in Alzheimer's disease - their role in diagnosis and treatment. Curr Pharm Des. 2011;17(25):2594-602. PubMed.
  4. . The mechanism of γ-Secretase dysfunction in familial Alzheimer disease. EMBO J. 2012 May 16;31(10):2261-74. PubMed.
  5. . BMS-708,163 targets presenilin and lacks notch-sparing activity. Biochemistry. 2012 Sep 18;51(37):7209-11. PubMed.
  6. . Amyloid-beta as a positive endogenous regulator of release probability at hippocampal synapses. Nat Neurosci. 2009 Dec;12(12):1567-76. PubMed.
  7. . Differential effects between γ-secretase inhibitors and modulators on cognitive function in amyloid precursor protein-transgenic and nontransgenic mice. J Neurosci. 2012 Feb 8;32(6):2037-50. PubMed.
  8. . Safety and Tolerability of the γ-Secretase Inhibitor Avagacestat in a Phase 2 Study of Mild to Moderate Alzheimer Disease. Arch Neurol. 2012 Aug 13;:1-12. PubMed.

External Citations

  1. Phase 2 study in prodromal AD
  2. Phase 2 study

Further Reading


  1. . Safety and Tolerability of the γ-Secretase Inhibitor Avagacestat in a Phase 2 Study of Mild to Moderate Alzheimer Disease. Arch Neurol. 2012 Aug 13;:1-12. PubMed.

Primary Papers

  1. . Safety and Tolerability of the γ-Secretase Inhibitor Avagacestat in a Phase 2 Study of Mild to Moderate Alzheimer Disease. Arch Neurol. 2012 Aug 13;:1-12. PubMed.