22 July 2004. Perhaps…yes? At first blush, this is the tentative conclusion one must draw if the results of two presentations yesterday on the formal analysis of Elan’s AN-1792 vaccine are both right. The presentations were part of the 9th International Conference on Alzheimer’s Disease and Related Disorders, which ended earlier today here in Philadelphia. One group reported no overall cognitive improvement but upward trends in visual memory. Another jolted the audience by showing dramatic pictures of how patients who made antibodies in response to the vaccine clearly lost more brain volume than did patients who did not respond or were on placebo. Yes, you read it right: Responders had more brain shrinkage than non-responders. In the packed lecture hall, this counter-intuitive result led to a palpable sense of “I-told-you-so” and some snickering among those who doubt the wisdom of this approach. It left the presenters looking, for the moment, a bit sheepish as they tried to explain the paradox of how it could be that a treatment can stabilize thinking while apparently reducing gray matter. Last time most people checked, serial brain volumetry was considered a measure of continuing deterioration in AD.
The field has eagerly awaited results of this 27-center trial for two years, ever since dosing ended early because of encephalitis in about six percent of patients (see ARF related news story and ongoing Alzforum coverage since then). The surprising imaging result presented yesterday stirred up the audience enough to make this session arguably the liveliest in what has otherwise become a large and scripted conference that leaves little room for public discussion at the end of oral presentations.
Here is a summary of the data. Sid Gilman of the University of Michigan, Ann Arbor, first presented overall results of this phase2a trial of the AN-1792 vaccine containing pre-aggregated Aβ plus the adjuvant WS-21. The primary objective of the trial was to establish safety and tolerability, as well as first-pass efficacy data. “Responders” were pre-defined as patients who mounted an antibody response of a given size; the assessments included, among others, ADAS-cog, MMSE, clinical dementia rating (CDR), and brain volumetry. A secondary objective was to study the immune response generated by the vaccine, and changes in CSF Aβ and tau levels. The trial took place in the U.S. and Europe. Three hundred outpatients with mild-to-moderate AD who had had the disease for an average of four years received vaccine or placebo injections. The original protocol called for six shots, but after the encephalitis turned up, the company ended dosing so that the patients ended up receiving between one and three doses total (274 patients received two shots.) The investigators kept the study blinded for 12 months to exclude bias from the analysis.
Of the 300 patients, 59 mounted an antibody response. In the cognitive analysis, the investigators compared placebo, responders, and non-responders at baseline versus 12 months. Overall, they did not find significant differences between the placebo and responder groups, Gilman said. When investigators looked at performance in individual tests, they found that there appeared to be a trend. In four different Wechsler visual memory tests, those on placebo deteriorated as expected, while the responders showed small improvements, Gilman said. Only the result of one such test, the Wechsler verbal-delayed test, was statistically significant. In all other memory tests, the placebo group worsened more quickly than did responders, but none of these differences were statistically significant. Tests of executive function showed no differences.
When the investigators broke up the 59 responders into a group that made low antibody titers and one that made high titers, they saw no difference between patients on placebo and the non-responders, less decline in the low-responder group, and small improvement in the high-responder group. Subgroup analysis lowers power and tends to prompt heated discussion about what the numbers mean.
The CSF measurements showed no difference in Aβ levels but a reduction of tau in the responders. Most of the neuropathology findings were reported in this space already (see James Nicoll’s presentation in the ARF St. Moritz conference summary). The news since then is that two additional autopsy cases show broadly similar results. One, analyzed by Eliezer Masliah’s group at University of San Diego, California, showed focal depletion of plaques but persistent tangles, some lymphocyte infiltration and Aβ immunoreactivity. The other, presented here in Philadelphia in a separate talk by Nicoll, of University of Southampton in England, is of a person who died from a cardiac aneurysm only four months after getting the vaccine. That case shows considerable amyloid still in place but signs of active phagocytosis, i.e., abundant microglia surrounding the plaques with gobs of Aβ inside them, as well as some astroglia. In all, neuropathology on two cases with encephalitis and two without is available to date. The Masliah case is in press; for the second Nicoll case the manuscript is in preparation.
Overall, Gilman concluded that despite its difficulties, the trial had established a proof of concept that amyloid immunotherapy can elicit an immune response that leads to clearance of amyloid.
It was Nick Fox, of the Institute of Neurology, London, however, who delivered the big surprise when he presented results of the MRI component of the trial. The original protocol called for volume measurements at baseline, 12, and 15 months. After the encephalitis turned up, it was revised to drop the third scan. The investigators took three MRI measures. Whole brain measurement has relatively low resolution but the scan-rescan error is low and matching scan-rescan images has been automated. A patient with AD typically loses 20 to 25 milliliters per year, Fox said. Hippocampal measurement affords higher resolution but also has a higher error rate and requires manual outlining of the structure. AD patients typically lose three to six percent of hippocampal volume per year. Finally, ventricular enlargement complements the first two measures. It is less sensitive to artifact and changes by about 5 ml per year in AD patients, according to Fox.
The 288 patients who had paired scans included 57 on placebo and 45 responders. Fox said he reviewed all scans blinded to subject and time point. When the blind came off, the results fell out this way: Antibody responders had greater decrease in brain volume than did people on placebo. Correlation to titer showed that the more antibodies a patient had, the more volume they lost. Brain ventricles enlarged more in high responders than low responders. All three measures were highly statistically significant. High responders also showed more hippocampal loss, but this result was not statistically significant. Some people lost up to 60 ml from their brains. In relating brain volume changes to cognitive performance, Fox said that high responders had a better outcome even though they lost more brain volume and their ventricles enlarged more. They also had the largest drop in tau. The groups were balanced at baseline, and patients on placebo had volume changes exactly like what would be expected from the literature; non-responders fell in with those on placebo.
How to explain this surprising result? The brain volume loss could represent accelerated neuronal loss, perhaps as a result of sub-clinical encephalitis. Fox considers this unlikely because of the fall in CSF tau and the stable cognitive performance. Another possibility is that the brain collapses around areas of cleared amyloid, and the measured volume loss reflects the volume of the removed amyloid itself. Fox did not argue this point strongly, and questions from the audience and comments later on expressed doubt that the removed amyloid could account for this much volume change. There is disagreement about how much amyloid there is in brain, and ways of measuring it vary, but several researchers said that it is in the range of at most a gram. The brain loss could reflect fluids, or perhaps something else entirely is going on, some underlying qualitative change in the brain that the imaging did not visualize. See more on this puzzle below.
There has been intense research on various imaging approaches to diagnose and track the progression of AD. Indeed, brain volumetry has become a contender in the imaging field’s current efforts to agree on, standardize, and validate one or a couple of broadly accepted, robust measures as a consensus biomarker. To most casual observers, at least, shrinkage in cortex and hippocampus coupled with enlargement of the ventricles implies that AD is getting worse. But up until now, brain imaging research has been conducted separately from treatment trials, as trials are just starting to include a brain imaging component. Consequently, the apparent contradiction between cognition and imaging came as a surprise to many. Some researchers noted that perhaps animal studies of immunotherapy could help clarify this issue more quickly than future trials.
The session ended with this mischievous question from Peter Davies to Nick Fox: “If I treat a group of patients with an anti-tangle drug, will you cut me the same amount of slack?” Fox: “Yes!”
If you think the story ends here, hang on for three more paragraphs. Today, in the last session of the conference, Roger Nitsch of the University of Zurich presented the two-year follow-up in the analysis of the Swiss cohort of 30 patients, which his group conducts independently of the trial as a whole. This space has covered much of Nitsch’s presentation before (see ARF St. Moritz conference summary; ARF New Orleans summary). Here is what’s new since then: Nitsch’s team imaged their patients at the Zurich site, and they have completed a third scan on them. Like Fox, the Swiss researchers also see brain shrinkage, and the patients with the most robust cognitive stabilization had the most antibodies and also the most brain volume loss.
However, this volume loss ended after about a year, and the two-year scan showed a slight increase again. Then the team’s neuropsychologist conducted a Wechsler test of visual memory that draws on hippocampal function. “We have patients that are better now on it than they were at the beginning of the trial. I ask the audience, could this be a neuroregenerative effect?” Nitsch asked provocatively. It is important to note that at 2.8 percent (p=0.05), the reported uptick in brain volume appears to fall barely outside of the confidence interval. This group of high-responders numbers 11 patients, and all have had a second-year scan. It appeared to this writer as if the high responders and their comparison group who had not had an antibody response (n=3) did not start out with the same volume at baseline, so this data must be considered preliminary. Nitsch's group is continuing to follow their patients closely and expect to scan their brains again at future time points.
For now, Nitsch offered this speculative proposition: As the amyloid burden is cleared out, the brain also might shed some of the astrocytosis that typically accompanies AD. (In an earlier lecture at this conference, Stephen Paul of Eli Lilly and Co. had shown evidence suggesting that astrocytes clear amyloid; prior work on astrocytes and amyloid phagocytosis includes Wyss-Coray et al., 2003). The volume taken up by amyloid plus astrocytosis, plus other factors, would come closer to explaining the volume loss, Nitsch added. Privately, other neurologists agreed this possibility is worth testing.
Elan is continuing to develop immunotherapy approaches, as are academic laboratories and other companies. Elan has begun a phase 1 safety trial of a passive vaccine. As always, we invite comments and corrections.—Gabrielle Strobel.