Tarenflurbil, aka flurizan, a γ-secretase modulator that showed promise in early preclinical and clinical trials, failed in Phase 3, as reported in the December 16 JAMA. The results will come as no surprise to those following the flurizan saga. Alzforum previously covered the Phase 3 data when they were presented at ICAD last year (see ARF related news story). Writing in JAMA for the Tarenflurbil Phase 3 Study Group, Robert Green, Boston University School of Medicine, and colleagues confirm that the drug had no statistically significant effect in co-primary outcome measures of cognition and activities of daily living.

The 18-month, randomized, placebo-controlled, double-blind trial was carried out at 133 centers in the U.S. Patients were administered placebo, 400 mg, or 800 mg of the drug twice daily. After analysis of earlier Phase 2 trial data, the Phase 3 trial was modified, with the approval of the FDA, such that only the highest dose was administered and only to patients with mild AD; patients with moderate AD were dropped from the trial. In the final analysis there was no difference between treatment and placebo arms in either the co-primary outcomes (the ADAS-Cog 80-point version and the ADCS activities of daily living scale) or a range of secondary outcomes that included measures of function (the CDR sum-of-boxes), cognition (the MMSE), psychopathology (Neuropsychiatric Inventory), quality of life (QQL-AD), and caregiver burden.

This was the first Phase 3 trial of a γ-secretase modulator, a class of drugs designed to shift the proteolytic profile of the enzyme away from the longer Aβ42 and toward shorter, less-amyloidogenic species, such as Aβ37 or Aβ38, rather than blocking the enzyme completely. Blocking γ-secretase cleavage of other transmembrane proteins, including Notch, can lead to intolerable side effects.

It is not clear why the trial failed, but Thomas Montine, University of Washington, Seattle, and Eric Larson, Group Health Research Institute, also in Seattle, list some potential reasons in an accompanying JAMA editorial. They suggest that the drug may not have achieved the concentration in the brain necessary to modulate the secretase, a possibility that may never be proven one way or the other, since the trial did not measure CSF Aβ levels. An earlier, 21-day study lends credence to this idea, however, finding that up to 800 mg of the drug twice daily failed to reduce Aβ42 in plasma or CSF (see Galasko et al., 2007). Montine and Larson also expressed a slightly more disheartening explanation for the trial’s failure. “Commonly used experimental models of Alzheimer disease may inadequately reflect the complexity of cognitive impairment and dementia in older patients and thereby provide falsely promising leads.” If true, then that could spell trouble ahead for other AD trials.—Tom Fagan


  1. Tarenflurbil, a γ-secretase modulator, failed in Phase 3 trial. Recent clinical failures of amyloid treatment including tarenflurbil have brought confusion to the Alzheimer’s field, which should consider the pathogenic differences of amyloid between Alzheimer’s model mice and humans. Anti-amyloid treatment in model mice succeeded in recovering the cognitive abilities of mice, but not humans.

    Why? In their commentary, Montine and Larson explain the possibility that “Commonly used experimental models of Alzheimer disease may inadequately reflect the complexity of cognitive impairment and dementia in older patients and thereby provide falsely promising leads.”

    Our research team has found that homocysteic acid (HA), which is metabolized from homocysteine or methionine, is a pathogen for Alzheimer disease. This pathogen basically works under amyloid toxicity. We confirmed this pathogenic action of HA with a newly developed HA vaccine for 3xTg-AD mice (1). First, the mice showed higher HA levels prior to amyloid-induced pathological changes. Second, the HA vaccine recovered the cognitive disability of the mice independently from amyloid toxicity.

    Humans excrete HA into urine at 1,000 times higher concentration than mice, which implies that humans have more severe HA toxicity than do mice (2). When anti-amyloid treatment decreases the amyloid level in mouse brain, then HA toxicity appears in mice, but HA levels in mice are low and its toxicity is weak. In contrast, HA levels are much higher in humans, and HA itself shows strong toxicity. In conclusion, we should consider the pathogenic effect of HA in human Alzheimer disease.

    See also:

    Hasegawa, T. Differences in urinary excretion of homocysteic acid (HA) have been observed between humans and mice, suggesting that HA toxicity is more severe in humans than mice.


    . Treatment of Alzheimer's disease with anti-homocysteic acid antibody in 3xTg-AD male mice. PLoS One. 2010 Jan 20;5(1):e8593. PubMed.

  2. What has always been remarkable about preclinical work in mice is just how extraordinarily plastic amyloid deposits are in the mouse brain. I can recall numerous discussions with colleagues leading to the inevitable point that an inordinately high proportion of potential therapeutics significantly reduce Aβ in transgenic mice. This is a problem that echoes one in the cancer field (it has often been said that cancer was cured in mice a decade ago, but success at translating these findings into human patients has been poor). Not only does this situation cast doubt on our models of the disease itself, but makes one wonder if mice in general are a poor system in which to study Alzheimer disease.

  3. It's amazing to have to read the "blame it on the mouse models" excuse.

    Those working with mice know well: mice are not patients and definitely not AD patients. They "model" one or the other aspect of AD, in most cases (over)production of Aβ peptides and other APP metabolites, a fact often set aside in the pharma sector for temporary ease of mind.

    So is the fact that putting in not one but three or five mutations accelerates the process, but does not necessarily improve the preclinical relevance of the model. Adding the odd mutant PS1 is defensible for mechanistic revelations of its actions on "85+ substrates," but as an amyloid , this is hardly relevant for the aged Phase 3 trial participant.

    Models are what they are: living "test tubes" of higher-order complexity than a solution of recombinant enzyme, but never the "real thing." The only efficacy test tube of any compound or treatment is the clinic. It's too bad if they fail there…telling the model-makers that square one is still open for trying harder!

Make a Comment

To make a comment you must login or register.


News Citations

  1. Chicago: Flurizan Postmortem

Paper Citations

  1. . Safety, tolerability, pharmacokinetics, and Abeta levels after short-term administration of R-flurbiprofen in healthy elderly individuals. Alzheimer Dis Assoc Disord. 2007 Oct-Dec;21(4):292-9. PubMed.

Further Reading


  1. Chicago: Flurizan Postmortem

Primary Papers

  1. . Effect of tarenflurbil on cognitive decline and activities of daily living in patients with mild Alzheimer disease: a randomized controlled trial. JAMA. 2009 Dec 16;302(23):2557-64. PubMed.
  2. . Late-life dementias: does this unyielding global challenge require a broader view?. JAMA. 2009 Dec 16;302(23):2593-4. PubMed.