. The antibody aducanumab reduces Aβ plaques in Alzheimer's disease. Nature. 2016 Aug 31;537(7618):50-6. PubMed.


Please login to recommend the paper.


  1. This straightforward, “rite-of-passage,” multiple ascending dose, safety and tolerability study demonstrated that an aducanumab dose range from 3 to 10 mg/kg reduced amyloid fibrils and was safe enough for use in the ongoing Phase 3 pivotal efficacy trials. The extent of ARIA and adverse events show that the 10 mg/kg dose, and possibly the 6 mg/kg dose, may be too high for ApoE ε4 carriers who have detectable levels of amyloid plaques by PET.

    The study was most remarkable for demonstrating that aducanumab did exactly what it was engineered to do: clear amyloid fibrils in a predictable, precise, dose-dependent manner (as indicated by the small SEs around the SUVR estimates of change in Figure 2a). This, alone, should cause a high level of enthusiasm.

    Unfortunately, and despite the authors’ disclaimer that “[t]he trial was not powered for the exploratory clinical endpoints, thus the clinical cognitive results should be interpreted with caution,” they nevertheless spend a lot of ink analyzing and discussing clinical efficacy claims from a study that was not designed to do so. There are several problems here:

    The sponsor treated the three separate and sequential dosing cohorts (placebo, 1, and 3 mg/kg; placebo and 10 mg/kg; placebo and 6 mg/kg cohorts) as though they were a parallel-group trial by comparing the four dosing groups to a placebo group pooled from the three cohorts. That is, they made the study seem as though patients were randomized contemporaneously to placebo or one of the four doses in a parallel-group, dose-ranging trial. In fact, each cohort was separately randomized: The first cohort was randomized earlier than the second, 10 mg/kg cohort; and the second cohort prior to the third, 6 mg/kg cohort. They called it a “staggered parallel-group design,” but really the cohorts were separate parallel-group studies occurring at different times and differing sites. So, for example, the last cohort, the 6 mg/kg dosing group, was compared to mainly placebo patients acquired nearly a year earlier and from different sites.

    The above and play of chance might explain why the pooled placebo group tended to score better on the CDR-sb and FCSRT at baseline than the treatment groups. At best, after covariate adjustments for baseline scores and ApoE, and not adjusting the p value for the multiple comparisons, there was a nominally significant effect for the CDR-sb and the MMSE for the 10 mg/kg dose; and this was accompanied by implausibly large mean differences from placebo, about 1.24 and 2.25 for the CDR-sb and MMSE, respectively. Only three of 16 CDR-sb and MMSE contrasts were nominally significant at the unadjusted 0.05 alpha error level, and the NTB and FCSRT (memory) tasks did not show significant effects (Extended Data Table 1). Outcomes from the Cognitive Drug Research (computerized) battery were not reported.

    If the individual dosing cohorts had been presented separately as they were in the very similarly designed bapineuzumab MAD study (Salloway et al., 2009), we probably would see substantial variation in the placebo change and clinical effects within each group. Without the display of the cognitive data by dosing cohort, one has to wonder whether the apparent CDR-sb and MMSE benefits at 10 mg/kg had to do with variable baselines between cohorts and sites, statistical corrections, or extreme comparisons.

    The enthusiasm for aducanumab should rest on its PK, PD, plaque-busting ability, preclinical biochemical characteristics, and clinical safety. One doesn’t need to invoke implausible clinical and cognitive effects to make a compelling case for the potential for aducanumab. 


    . A phase 2 multiple ascending dose trial of bapineuzumab in mild to moderate Alzheimer disease. Neurology. 2009 Dec 15;73(24):2061-70. PubMed.

    View all comments by Lon S. Schneider
  2. This is very exciting work. The authors included a short discussion on the biochemical characterization of this antibody. The main take-home message is that aducanumab does not recognize monomeric Aβ40 at concentrations approaching 1 micromolar. One of the puzzles about this antibody is how it appears selective for aggregated forms of Aβ and yet recognizes an N-terminal region of Aβ that normally cannot aggregate by itself and is a region that is likely to be highly flexible in aggregated and fibrillar forms of Aβ based on recent structural studies.

    View all comments by Michael Parker
  3. In this paper on the aducanumab Phase 1b study in prodromal and mild AD patients, the data on plaque removal are quite encouraging. As pointed out on AlzForum, the initial positive cognitive data need to be confirmed in the ongoing Phase 3 studies, and I agree with Lon Schneider’s comments on potential biases linked to the staggered design of the study.

    In my opinion, there is another source of potential bias in interpreting the cognitive results of the study; it refers to the clinical stage of dropouts. The clinical characteristics of the dropouts are not much detailed in the article, but from Table 1 and Extended Data Figure 2, it seems that 10 of 12 dropouts in the 10 mg/kg group were patients with mild AD at baseline, whereas there were only one or two dropouts among prodromal AD patients in the 10 mg/kg group (see table below). Conversely, in the placebo group the discontinuation rate was similar between prodromal and mild AD patients (21.1 percent versus 28.6 percent). The criteria for definition of prodromal AD were an MMSE of 24-30 (inclusive) and a CDR of 0.5, while for mild AD they were an MMSE of 20-26 and CDR of 0.5-1. Thus, the imbalance in the proportion of prodromal AD and mild AD dropouts in the 10 mg/kg may have contributed to the apparent slow decline observed in this treatment group.

    Why there was a much higher dropout rate in mild compared to prodromal AD patients in the 10 mg/kg group? One possible explanation could be a higher incidence of ARIA-E abnormalities in the mild AD group. Compared to prodromal AD, mild AD patients may have a higher Aβ burden in brain vessels and consequently a greater risk of edema or bleeding due to Aβ removal by aducanumab. Interestingly, in the ongoing ENGAGE and EMERGE Phase 3 studies, only patients with MMSE 24-30 and CDR=0.5 at baseline were included.

    I hope this may be useful in interpreting the cognitive results of the study.

    View all comments by Bruno Pietro Imbimbo
  4. Scientists who have experience isolating Aβ from human brains may find the data shown in Figure 1 puzzling. Our lab has found Aβ deposits in human brain to be as hard as rocks. Because Aβ deposits resist such detergents as SDS, we usually depend on formic acid, an extremely toxic acid, to dissolve them. In general, the concentrations of SDS used for washing human brain tissues and of formic acid for extracting Aβ are 20 percent and 99 percent, respectively. For those unfamiliar with the pathological biochemistry of AD, these concentrations are extremely high. In contrast, you can easily extract plaque Aβ from transgenic mice brain using guanidine hydrochloride. Therefore, one would predict that the effects of immunotherapy on humans and on mice will not resemble each other. Human Aβ deposits are much more difficult to remove by therapeutic antibodies because of their physicochemical properties.

    Furthermore, the images in Figure 1 suggest that the nonspecific binding of florbetapir to white matter, which would not be expected to harbor Aβ plaque, is higher before treatment than after. Did the immunotherapy decrease the amount of white matter in the brain?

    Scientists including Yasuo Ihara, Colin Masters, Konrad Beyreuther, Steven Younkin, Virginia Lee, and Denis Selkoe would know better than I do, as they have more experience, but I worry that the data in this paper seem a bit too good to be true.

    View all comments by Takaomi Saido

Make a Comment

To make a comment you must login or register.