This is Part 2 of a two-part series on AD/PD 2013 immunotherapy news. See also Part 1.

18 April 2013. At the 11th International Conference on Alzheimer’s and Parkinson’s Diseases, held 6-10 March in Florence, Italy, Thierry Bussiere and Jeff Sevigny of Biogen Idec in Weston, Massachusetts, showed new data on how a therapeutic Aβ antibody their company had licensed from the Swiss biotech Neurimmune was measuring up. The idea is that a fully human antibody derived from old people whose immune systems successfully hold Alzheimer’s disease at bay might be safer than antibodies engineered in the lab (see Part 1 of this series).

Alzforum has previously covered preclinical studies on BIIB037, an IgG1 against a conformational epitope found on β amyloid (see ARF related news story). In essence, BIIB037 binds to parenchymal amyloid and clears parenchymal plaques in transgenic mice without causing microbleeds. In Florence, Bussiere presented new data on a direct comparison of mouse versions of BIIB037 to three other clinical antibodies—bapineuzumab (3D6), solanezumab (m266), and gantenerumab (mGt)—in vitro and in mice.

Both BIIB037 and mGt bound fibrillar amyloid, but the latter bound vascular plaques as well as parenchymal ones, Thierry said. 3D6 bound soluble and fibrillar Aβ equally strongly; among fibrillar deposits, it showed a preference for vascular over parenchymal plaques, while m266 bound only soluble Aβ. Like m266, 3D6 caused a plasma spike due to its binding to soluble Aβ; BIIB037 did not. Nor did BIIB037 cause microhemorrhages in 13-week chronic dosing of 22-month-old mice with doses of up to 70 mg/kg. “We saw a trend for microhemorrhage only at 500 mg/kg—that is a very, very high dose,” Bussiere said. BIB037 starts clearing plaque in mice at 3 mg/kg.

Dosing and safety were the themes of Sevigny’s talk as well. Sevigny presented the first clinical data from an ongoing Phase 1 study comparing six doses of BIIB037 to placebo given as a single intravenous infusion in 48 people with mild to moderate Alzheimer’s. The doses ranged from 0.3 mg/kg to 30 mg/kg. In these studies, it is customary that once the data safety monitoring committee deems a given dose safe, the study proceeds to the next higher dose. Participants came in for assessments at 10 time points up to two years after dosing. To look for amyloid-related imaging abnormalities, aka ARIAs, every participant between baseline and study end underwent four MRI scans, which were read both locally and by a central reader at the imaging CRO Synarc based in Newark, California.

Thus far, 21 study participants had a total of 43 side effects, Sevigny reported. They included headache, diarrhea, and dizziness. Most were mild, some were moderate, none severe, and none became worse with higher doses, Sevigny said. Importantly, no new ARIA-E or ARIA-H came up during the study (some old people and people with AD have ARIA even without drug). Because this antibody seemed safe at the highest dose in this study, the scientists added a test of 60 mg/kg—an improbably high dose that amounts to 4 grams in a 150-pound person. “These are the highest of all therapeutic AD antibodies being tested these days,” Sevigny told Alzforum. The scientists plan to present results of the 60 mg/kg dose later this year.

For comparison, the multiple sclerosis antibody natalizumab (trade name Tysabri®) is generally given as a 300 mg infusion once a month, and the rheumatoid arthritis antibody infliximab (trade name Remicade®) at around 3 to 5 mg/kg.

The pharmacokinetics of BIIB037 was what researchers casually call “well behaved,” meaning it changed in a linear fashion across doses and showed little variability from person to person. Consistent with the preclinical work, the antibody generated no plasma spike, probably because it does not bind soluble Aβ, Sevigny said.

Last summer, Biogen Idec started a second Phase 1 trial to evaluate how people at an earlier stage of disease respond to multiple doses of the antibody. At up to 160 participants at up to 40 locations, this trial is much larger. Besides BIIB037 itself, it also tests whether diagnostic criteria for prodromal Alzheimer’s perform as robustly in the trenches, i.e., in clinics across the country, as they do in the hands of leading clinicians at specialty dementia clinics who articulated them (Dubois et al., 2010; Albert et al., 2011). “We know this diagnosis works in an ideal environment, but a multicenter trial is not an ideal environment,” Sevigny said.

Prospective participants go through a tiered process to gain admission into the trial. People who score between 20 to 30 on the MMSE, between 0.5 and 1 on the Clinical Dementia Rating, and 27 or lower on the Free and Cued Selective Reminding Test (FCSRT) are then referred for an amyloid PET scan with florbetapir and an MRI. The MRI serves to look for vascular lesions and to add anatomic precision to the amyloid scan, but only people whose amyloid PET scan is positive can enter the trial. Trialists across the field are intensely interested in how both the prodromal AD psychometric criteria as well as biomarker enrichment are doing in selecting people who truly are at the earliest symptomatic stage of Alzheimer’s (see ARF AD/PD 2013 story).

Several trials are testing these entry criteria now, and some are beginning to report their initial experience. At the last CTAD conference, researchers talked about a high screen failure rate at sites recruiting for prodromal AD trials (see ARF CTAD story). In Florence, Joyce Suhy of Synarc presented results for BIIB037.

Nineteen imaging centers, using 12 different scanner models from old to state-of-the-art, are participating in the BIIB037 prodromal trial. The sites send the scan data to Synarc. First author Jerome Barakos and colleagues processed them and showed the PET, concomitant MRI, and fused PET/MRI images to the trial’s two central neuroradiologists. These readers were trained on a binary classification method Avid had developed as part of florbetapir’s FDA approval package (see ARF related news story). The scientists also calculated each scan’s SUVR to compare the visual and quantitative methods of interpreting an amyloid scan.

Of the first 80 patients who came to get the scans, 44 had been clinically diagnosed as prodromal AD, with a mean MMSE of 28, and 36 as mild AD, with a mean MMSE of 23. In the visual read, 48 of these 80 people, or 60 percent, were amyloid positive. In the quantitative analysis, 55 were amyloid positive, i.e., concordant with their prior clinical diagnosis.

This means that the visual read had missed seven people who were positive by SUVR, in other words, generated some false negatives. At the same time, however, the comparison of the two methods showed that the visual read produced no false positives, Suhy said. On a technical note, Suhy noted that it’s important to correct the data when a patient moves in the scanner; otherwise, the motion-induced fuzziness in the images can generate a false-positive read. The false negatives were prodromal cases with a mean MMSE of 27 and a composite SUVR between 1.12 and 1.31 (the cutoff for positivity was 1.1). The κ value for how often the two readers agreed was 0.97.

Suhy interpreted the data to mean that enrichment by PET works. “The visual reading scheme allowed accurate detection of all quantitative amyloid-negative patients and their exclusion from this trial,” Suhy said. Around the borderline of amyloid positivity, the visual read missed some people. “We could have had seven more people enter the trial if we had used quantitative analysis, but that would have been at the cost of maybe including a false positive while adding time and process. Avoiding false positives is the main point,” Suhy said. Future multicenter trials, and also clinical practice, will tend to use visual reads because results come back faster and are not dependent on data-processing software.

Of the 32 people who were amyloid negative, a large majority had the prodromal diagnosis. This BIIB037 trial had a higher amyloid-negative rate after the clinical exam than did the bapineuzumab and solanezumab Phase 3 trials. In discussion, scientists said that this is not surprising, because the error rate of clinical diagnosis is generally higher at earlier stages.

Suhy’s talk generated intense discussion about how this enrollment strategy could be improved. Privately, some scientists said that the clinical testing for prodromal AD is far from optimized. In particular, at this point there is considerable uncertainty about how best to set cutoffs for what are, fundamentally, continuous variables.

For his part, Bruno Dubois, at Pitié-Salpêtière Hospital in Paris, suggested that a lower FCSRT cutoff would reduce the number of false negatives. The test’s ceiling is 48, but normative FCSRT data suggest that 27 is the mean for cognitively normal adults. “Maybe half the people in this room, with a little jetlag, would not achieve 27,” Dubois quipped. Dubois uses a cutoff of 17 to make a diagnosis of prodromal AD. “At that point, all our patients are biomarker positive,” Dubois said.

A cutoff of 17 includes only the most impaired end of the MCI spectrum, Dubois acknowledged. Site clinicians said that this number makes recruitment difficult. They spoke about patients whose condition looks like prodromal AD to them but who still score higher than 17 on the FCSRT and therefore never get to have a scan, much less join the trial. On the upside, people who do fall below 17 do not have to be turned away later after having undergone a biomarker test. The basic dilemma is that when the FCSRT cutoff is too high, some people get excluded after an invasive procedure (PET or lumbar puncture); when it is too low, the more mildly impaired people are blocked at the initial clinical step even though they may have a head full of amyloid. Dubois proposed that 20 might be the right compromise.

The BIIB037 prodromal trial cutoff of 27 was set in discussion with Ellen Grober at Albert Einstein College of Medicine in New York. Grober co-developed the FCSRT. The cutoff is higher in part because the U.S. version of the test is different from the French version, for example, using pictures, not words, said Sevigny. By comparison, Roche’s prodromal AD trial of the gantenerumab antibody uses a cutoff of 17. That trial does not permit memantine or cholinesterase inhibitors, whereas the BIIB037 trial does; hence, local factors also influence how a cut point is set.

“The exact clinical criteria for the prodromal trials may have to be tweaked based on first results of ongoing studies. We are all learning,” said Hansruedi Loetscher at Roche.—Gabrielle Strobel.

This is Part 2 of a two-part series on AD/PD 2013 immunotherapy news. See also Part 1.