Thursday, January 23, 2014

White Noise: Trials of Pharmaceuticals for Alzheimer's Disease

"But we are not here concerned with hopes or fears, only with the truth as far as our reason allows us to discover it." - Charles Darwin

In yesterday's NEJM, the results of two trials of antiamyloid monoclonal antibodies (sonalezumab and bapeneuzumab)  for Alzheimer's Disease (AD) are published.  I became interested in the evidence for AD treatments after the recent trial of Vitamin E and Mematine for AD (the TEAM-AD VA Cooperative Trial) was published in JAMA earlier this month.  Regular readers know that I think that the prior probability that vitamins, minerals, and antioxidants are beneficial for any disease outside of deficiency states is very low.  The vitamin E trial was the impetus for some background investigation which I will summarize below.

Some diseases appear to be degenerative, multifactorial, progressive, and intractible in nature such that the hopes for a pharmacological cure or treatment are remote - the organs or tissues just "wear out", sometimes "prematurely".  Within my own specialty, Idiopathic Pulmonary Fibrosis (IPF) is one of those diseases.  No drug stalls or arrests the fibrotic process.  The only "treatment" is replacing the lung(s) via lung transplantation.  In orthopedics, osteoarthritis (OA) is emblematic.  The only way to treat it is to replace the entire joint - there is no drug we can administer that can arrest or reverse the degenerative and fibrotic changes caused by OA.  I worry that AD may similarly be refractory to pharmacological treatment.

Here are the data on the major, pivotal, and highly referenced trials of therapies for AD that I could find in prominent journals:
  • Sano et al, 1997, NEJM tested vitamin E, selegeline (a MAO inhibitor) and the combination in patients with moderate severity AD with the primary endpoint being the time to the occurrence of any of the following: death, institutionalization, loss of the ability to perform basic activities of daily living, or severe dementia (defined as a Clinical Dementia Rating of 3).  In unadjusted analyses, there were no significant differences between treatment and placebo.  In analyses adjusted for baseline differences in the Mini Mental Status Examination (MMSE), all three groups were superior to placebo.  This implies several things:  1.) randomization failed; 2.) MMSE was a predictor of outcome; 3.) we know all variables that correlate with outcome and can measure them - suppose that variable XYZ is also correlated with outcome in the opposite direction of MMSE and it is also imbalanced, but we can't measure it?  Needless to say, I am skeptical of adjusted results.  You can only adjust for variables you know about.  That's the point of randomization.
  • Peterson et al, 2005, NEJM tested, in patients with amnestic mild cognitive impairment, whether donepezil or vitamin E was superior to placebo in preventing the primary outcome of "clinically possible or probable AD".  Neither donepezil not vitamin E prevented the primary outcome.
  • Reisberg et al, 2003, NEJM tested memantine in patients with moderate to severe AD against placebo.  Here the primary outcome variables (there were two) were the Clinician's Interview-Based Impression of Change Plus Caregiver Input (CIBIC-Plus) and the Alzheimer's Disease Cooperative Study Activities of Daily Living Inventory modified for severe dementia (ADCS-ADLsev).  Marginal results hovering around the significance threshold improved when "observed cases" analyses were substituted for the last observation carried forward (LOCF) analyses, the latter being the pre-specified analysis plan.  In any case, we have two primary endpoints and two analysis methods and marginal results either way on the primary outcomes.
  • Tariot et al, JAMA, 2004 studied the addition of memantine to donepezil in patients with mod-sev AD.  Using [again] two primary outcomes, the SIB (Severe Impairment Battery) and the ADCS-ADL19 (Alzheimer's Disease Cooperative Study Activities of Daily Living 19 scale, modified to 19 from its original components for reasons that were not explicated), they found that memantine was superior to placebo in both outcomes.
  • Dysken et al, JAMA, 2013 compared vitamin E, memantine, and both to placebo in patients with mild to moderate AD.  They found that vitamin E, but not memantine nor the combination slowed the progression of AD as measured by the ADCS-ADL inventory (this time not modified to 19 items).
  • Interested readers can look for early trials of donepezil lasting 15 weeks and 24 weeks published in 1998.
If you are willing to dismiss hopes and fears, you may have the following concerns about these trials:
  • Nobody seems to agree on the best outcome for these trials
  • Almost every trial uses more than one primary outcome but does not adjust the significance level for this
  • Many trials use multiple analysis or adjustment techniques and the a priori plans are poorly specified
  • The outcome instruments are both subjective and opaque (but if you look at the ADCS-ADL, it does appear to have surface validity, and clinical relevance)
  • There is no discussion of the MCID (minimal clinically important difference) in the methods for any of the scales and instruments
  • There is no consistent predictable benefit of any drug or combination of drugs across these trials even allowing for differences in primary outcomes - it's a proverbial crap shoot
  • I won't even mention the potential market for these drugs that forms the basis for valuations of companies that can convince the FDA that theirs is beneficial
AD researchers need to get together and decide which darned scale is the best, most clinically relevant one and use that scale consistently as the solitary primary outcome.  This will facilitate the combination of trials into meta-analyses.  They also need to determine, and state, what is the MCID for the chosen scale.

Since my hopes and fears were long ago purged, I'm going to wager that AD is a degenerative disease and that we have no treatment that provides a consistent, reproducible, sustained, and clinically important benefit for patients with this disease, let alone a treatment that slows its progression.  The trials above are tantamount to white noise.  So it was no wonder that antiamyloid antibodies were of no benefit in the 2 newest NEJM trials, any more than it would be that chondroitin sulfate is of no benefit in OA - even the orthopedists know better.


  1. Great posting--thanks.

    The failed anti-amyloid results are sad but interesting. The obvious question is whether the amyloid inhibition was working or not. Would be nice to see effect size for amyloid inhibition, right? If that's small, then failure may not be completely discouraging for the mechanism.

    2. Your unstoppable degenerative process hypothesis is plausible in a dismal way, but not precise. So what does it mean, and why does it imply unstoppability? Some interpretations:

    (a) Sort of like entropy, when you wear something down it is structurally impossible to rebuild it. Could be, but that seems kind of unbiological, no? Wound healing rebuilds broken down tissue, after all.

    (b) it is because these syndromes amount to _premature aging_, and evolution has selected its own tradeoffs w.r.t. aging, and we aren't likely to find any way to turn any of that around.

    Maybe, but at least *some* age-correlated processes can be turned around nicely (eg pharmacological treatment of male pattern baldness works if done aggressively enough).

    I'm not saying that you're wrong, just noting that what you've said is a bit vague and doesn't necessarily point to any clear analysis that should make us get all hopeless quite yet. :-)

  2. In addition, to illuminate how much "noise" these instruments involve, the scales such as the ADCS-ADL be evaluated for test-retest reliability at study initiation, so that we can see how much any positive exceeds or does not exceed the rest-retest variability. There is no need to rely on past measures of test-retest variability when you're conducting a large(r) trial like these. This kind of feature in these trials will enhance their contribution to the evidence base.


Note: Only a member of this blog may post a comment.