Last Sunday, Neurochem, Inc., the Montreal drug company developing Alzhemed as a new amyloid-reducing therapy for Alzheimer disease, announced that the results of its recently completed North American phase 3 trial to test tolerability, efficacy, and safety of Alzhemed in 1,052 AD patients were inconclusive. The Food and Drug Administration (FDA) ruled that the statistical models used to analyze both the cognitive efficacy data as well as the brain volume data were problematic, and that the results obtained could not support a claim for clinical efficacy. The decision comes after Neurochem announced problems with the trial at a conference this June (see ARF related news story).

Multiple factors contributed to the inconclusive results. Overall, variability among the 67 clinical sites in the trial overwhelmed the observed treatment effects, said Paul Aisen of Georgetown University in Washington, DC. Aisen served as scientific advisor to Neurochem and was the principal investigator of the trial of Alzhemed (aka tramiprosate). In particular, changes in people’s concomitant treatment with cognitive-enhancing drugs such as cholinesterase inhibitors, memantine, and antidepressants affected the results for the primary cognitive endpoints based on neuropsychological testing. Unexpected problems also arose in the control group and confounded the interpretation of Alzhemed efficacy. These, too, were most likely due to changes in concomitant medications, Aisen said. Thirty percent of the control group did not decline in cognition over the 18-month trial period; indeed, a portion of the control group unexpectedly demonstrated a significant improvement in cognition.

The company did note a trend in hippocampal volume change as measured by MRI. Patients treated with Alzhemed tended to show an increased atrophy rate as compared to the control group, suggesting a drug effect perhaps similar to what was seen in the halted AN-1792 trial (see ARF related news story). The data on hippocampal volume change was to be compared to secondary measures of whole-brain and entorhinal cortex volume change, also measured by MRI.

The next steps for Neurochem and the FDA are complicated, Aisen said. They depend on the recommendations of a newly appointed scientific advisory board led by Rachelle Doody of Baylor College of Medicine in Houston, Texas. The board’s report is expected by December 2007. The FDA is encouraging Neurochem to continue working on the U.S trial data set post-hoc. Aisen added that the company has committed to work with the FDA toward obtaining more conclusive results from the ongoing phase 3 trial of Alzhemed in European Union countries. The FDA has told Neurochem that the agency is open to considering modifications of study design, such as enrolling more people and changing treatment duration, as well as changes to the statistical analysis measures used in the EU trial.

One positive aspect of the European Alzhemed trial is that memantine treatment, a confounding factor in the U.S. trial, will not be allowed in that trial. A special challenge of the European trial is that it introduces another set of variables in that multiple sites in the member countries conduct cognitive testing in multiple languages.

Alzhemed’s fast-track status has not changed, although any approval date now will depend on the completion of the EU trial, which currently is an 18-month trial that enrolled the last patients earlier this summer. This would mean that data collection would be complete by end of 2008, with results expected by mid-2009.

On a general note, Aisen said that the problems that arose from this trial are not easily avoided. Indeed, he believes that this troubled trial will provide valuable lessons for the field of clinical AD research at large. The trial highlights the need to better manage the confounding variables and to minimize their effects on the results, and the new board is charged with providing advice on that. Patients who participated in the U.S. trial will have access to Alzhemed in an open-label extension for 1 year following conclusion of the 18-month treatment period. Neurochem expects that all data analysis of the North American trial will be ready to report by next summer’s ICAD meeting in July 2008 in Chicago.—Gwendolyn T. Wong.


  1. Since Alzhemed is a disease modifyer, the use of biomarkers by quantification of Abeta in CSF and/or plasma to assess its efficacy in this trial may be a real asset.

  2. The Alzhemed study was disappointing for the actual treatment arm. But it does suggest – unexpectedly - that a substantial fraction of patients with the current Alzheimer medications (i.e. anticholinesterases and memantine) alone (the control arm) performs quite well over a period of 18 months, at least in cognitive performance. Except for a small number of open-label studies (some going out to 4 years), this hypothesis hasn’t been tested to such an extent. In addition, the freedom to change medications allowed for many patients to settle on the most optimal treatment in terms of benefit and side-effect.
    Unfortunately we don’t know how much brain amyloid load was changed in the treatment arm, as the study was started well before the amyloid imaging probes became available. Such information would have given us some idea of how much of the cognitive benefits could be linked to a reduction of brain amyloid for patients in this stage of the disease.

    The major lesson from this trial is that the currently marketed medications provide an unexpected high overall cognitive benefit, which raises the hurdle for any future ‘disease modifying’ trial.

  3. This news story is a bit too optimistic and accepting of the company line. In truth, Neurochem managers knew in early April that their trial was negative and gave away as much in their April 19 press release by discussing the need for post hoc statistical models. Since then, they have issued carefully parsed public statements, press conferences, investor teleconferences, and a meeting presentation. In my opinion, close observers and certainly statisticians recognize the spin. The company’s statements sound like a distraction to avoid the outright admission that tramiprosate showed no effect compared to placebo. Even the company’s most recent news release does not quite get to this fact, nor does the ARF news story. Wall Street analysts and some investors finally got it, as the stock price tanked.

    The company spent over 4 months dredging, sifting, slicing, dicing, and remodeling their database, positing one, another, or several potential “confounding” variables that might have influenced outcomes. All the while, they issued periodic press releases stating that their “external team of statisticians” was hard at work making adjustments to the statistical models “as set out in the statistical plan”; making this sound about as formidable and intellectually challenging an endeavor as decoding the human genome.

    At the Alzheimer’s Association’s prevention meeting in June, Neurochem held a press conference and scientific presentation and issued a news release where it spoke about “potential confounding factors,” “numerical differences in favor of tramiprosate,” and “differences between groups on the primary disease modification endpoint as measured by magnetic resonance imaging,” all without providing any data. At the meeting, they presented a slide listing about 19 such factors from which they included from four to six (the slide is ambiguous) in their “adjusted statistical model(s).” Each factor involves an aspect of antidepressant, vitamin E, memantine, or cholinesterase inhibitor use, disuse, and/or dose change. They stated again, “no statistical comparisons can be conducted yet on the preliminary data as the primary statistical model is not reliable…. No conclusions can be drawn presently,” yet they obviously did such statistics.

    This level of vagueness has no place in scientific presentations, and arguably no place in drug company communiqués, especially when the presentations can be interpreted as federally required disclosures of information on material changes in the company. It strains the company’s credibility and is virtually impossible to interpret. By withholding data and analysis results while talking around the results, there is really nothing of substance to learn.

    The company presentations speak as though some sites may be at fault for site-to-site variability. The significance of this is not apparent, since we expect site variability; it is one reason why we do multicenter trials. They suggest that there was something about the discontinuation of cholinesterase inhibitors (after an average of nearly 2 years’ use) or memantine (average of nearly 1 year’s use), or that antidepressant and vitamin E use suppressed favorable outcomes. The latter is odd because antidepressant use is frequent in AD trials, and vitamin E is not cognitively effective. In the absence of data from the company, one must wonder how strong the potential effect of tramiprosate can be if drugs that have very modest to no cognitive effects are confounders.

    Moreover, Neurochem further obfuscates their outcomes by seeming to blame the placebo group for not worsening as expected. ARF quotes that “30 percent of the control group did not decline in cognition,” but this is probably close to what is expected. For example, the ADASc worsens by about 6 to 7 points in other 18-month trials with a standard deviation of about 8 or more. Therefore, about a third of a placebo sample would score the same or better than at baseline.

    If any of these “potential confounding factors” really influenced outcomes or suppressed an otherwise significant tramiprosate outcome, then Neurochem would have said as much. No, they would have shouted it, as would their sell-side financial analysts.

    The difficulty, of course, is that the company makes assertions without showing supporting data. They imply variously that the trial may have significant outcome elements to it, would have been significant if there had not been “confounders,” or that the trial failed because there was something wrong with the placebo group. Trust us, they seem to say. Better yet, believe our external statisticians, our advisors, or what we say the FDA said in a confidential meeting: “FDA deems…results inconclusive,” or “FDA is encouraging Neurochem to continue working on the U.S. trial data set post hoc.” In effect, they cloak themselves in perceived external authority such that it appears that Neurochem, FDA, and consultants make these statements together.

    I doubt that a commission of academics—however capable and well-intentioned—can provide anything additional beyond more breathing space and cover for the company. What can they do with the trial that the company hasn’t already attempted for more than 4 months? Even so, the company has stated that they will hold back outcomes of this trial until July 2008 at ICAD, the next Alzheimer’s Association science meeting.

    The company knows well their limited options with regard to both their trials and their business. Neurochem’s development program (and possibly its survival) rests on the European trial that finished enrolling in August and whose last patient would complete about February 2009. The company can increase sample size, tweak the statistical plan, shorten the treatment period, or end the trial and move on. Yet it has only $80 million cash, $45 million in debt, a burn rate of $3 million per month, two other iffy drug programs, and probably some angry investors with leverage. In this context, using an advisory board that won’t produce a report until Christmas strikes me as a play for time while the company does other things and hopes for good luck. Maybe they will get positive opinions from FDA or EMEA about their other, similar drug, eprodisate (branded Kiacta) for amyloid A amyloidosis, or maybe they will reveal phase 2 results from their trial of tramiprosate (here brand named Cerebril) to prevent hemorrhagic stroke in patients with cerebral amyloid angiopathy. I hope so, because good news there would increase the likelihood that the European AD trial is completed.

    Neurochem could have made this easier on everyone if they had presented their results in a straightforward manner when they knew them. Then they could have shown any unusual site-to-site variability, confounds, and discussed threats to the trial’s validity. Such forthrightness would have most directly benefited patients in the trials, physicians, other consumers, and stockholders. A straightforward presentation of their results might even gain them new investors. It is not too late for this.

    The evolving Neurochem story and its relationship with the Alzheimer’s Association is as much an example of the big business of how little companies finesse research results that have a material impact on their futures, as it is a story of scientific reporting when there are no facts to report. These may be the lessons yet to learn from this trial.

    Relevant disclosures: During the last year, I have consulted with the manufacturers of one marketed drug for AD, Forest, Lundbeck, and Merz; consulted with companies developing drugs for AD but without marketed drugs for AD, including AstraZeneca, Elan, GSK, Roche, Sanofi-Aventis, Takeda, Wyeth, Voyager, Myriad, and others. I serve as co-PI of Myriad North American AD trial. I have received a grant from the Alzheimer’s Association and am a member of the NIA ADCS steering committee.

    Neurochem news releases and financial statements.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Washington: Site-to-Site Differences Delay Result of Alzhemed Trial
  2. Philadelphia: Can a Shrinking Brain Be Good for You?

Further Reading