13 August 2008. As new experimental drugs are getting ready for clinical testing against Alzheimer disease, it’s instructive to learn how pharmaceutical companies are navigating the transition from Phase 2 to 3. At this juncture, they sometimes base the decision on whether to start a long and costly Phase 3 program on less than rock-solid efficacy data (see ARF ICAD story). At the International Conference on Alzheimer’s Disease, held 26-31 July in Chicago, the pharmaceutical company Eli Lilly and Company took the opposite approach to Elan/Wyeth. Both competitors are developing a humanized therapeutic antibody, which would be the first biologic treatment for AD. Elan/Wyeth uses a humanized version of an N-terminal Aβ antibody that recognizes forms of Aβ peptide in amyloid plaques; Lilly uses a mid-region antibody (m266) that captures soluble Aβ and tests the peripheral sink hypothesis of Aβ clearance (DeMattos et al., 2002). But rather than running a sizable and long Phase 2 trial (230 patients/19 months in the case of Elan/Wyeth), and scrutinizing its results for hints for efficacy (see ARF ICAD story), Lilly decided to sidestep clinical efficacy entirely. It instead ran a small and short (52 patients/3 months) Phase 2 trial that was heavily geared toward fluid biomarkers, and used those results as the basis to move into Phase 3 by 2009. Safety and biomarkers, but not efficacy, are also what drove Lilly's Phase 2 testing of its γ-secretase inhibitor, which is currently enrolling for Phase 3 (see Fleisher et al., 2008). Here are the details:
At ICAD, Eric Siemers and Ron DeMattos of Eli Lilly and Company in Indianapolis separately presented results of the trial and its attendant biomarker research, respectively. The Phase 2 trial tested four different doses of LY2062430, a therapeutic antibody, in people with mild to moderate AD in order to determine which doses and which interval between doses to use in Phase 3. The patients received 12 weekly infusions of placebo or antibody arranged to contain either 100 mg every four weeks, 100 mg every week, or 400 mg every four weeks, or 400 mg every week. There were 10 or 11 patients per group. The injections and most assessments ended at week 12, but patients will be followed for a year. At ICAD, Siemers showed interim data from day 112/week 16, i.e., partway into the follow-up period. The study was recently locked for definitive analysis, Siemers said.
The trial assessed its primary outcome of safety by looking for host immune responses against the Lilly antibody and for fluid or inflammatory signals by MRI and CSF, in addition to the usual adverse event monitoring. It also measured ADAS-Cog at the beginning and at 12 weeks, did plasma and CSF research, and a little neuroimaging.
On safety, Siemers said that no side effect cropped up that could be attributed to the antibody. None of the 42 treated patients showed evidence of edema, microhemorrhage, inflammation, nor were there laboratory abnormalities or troubling signs of immune response. Five of the patients, four of them dosed weekly, showed antibody titers to LY2062430, but these titers had no detectable effects. “The safety data could not be better,” Siemers told this reporter.
If the antibody is so safe, is it doing anything at all? In plasma, it caused both Aβ40 and 42 concentration to shoot up several orders of magnitude, as was also seen in the Phase 1 trial of this compound (see ARF Keystone story). The scientists showed a graph with different-sized saw tooth zigzag lines of plasma Aβ levels curving up in response to each injection, and they calculated the pharmacodynamics to conclude that monthly injections would be acceptable for Phase 3.
A tiny fraction, some 0.1 percent, of the antibody crosses into the CSF. There, too, total Aβ40 and 42 concentrations increase up to four orders of magnitude in a dose-dependent way, Siemers reported. “We expected this data,” Siemers said. But there was a surprise, as well. The Lilly scientists developed assays to distinguish in CSF between free Aβ and Aβ bound to antibody, doing so separately for Aβ40 and 42. At ICAD, they reported that the unbound, i.e., free, CSF Aβ40 decreased with greater doses of antibody, whereas the free CSF Aβ42 increased with rising doses of antibody. According to the poster, the Aβ40 drop happened in the 3,000 picogram/ml range, the Aβ42 rise in the 200 picogram/ml range. (CSF Aβ42 concentrations are low in people with AD.)
What does this mean? “We had to think about this,” Siemers said. “To us it suggests that the plaques are beginning to dissolve and Aβ42 slowly leaves the brain.” Siemers said further that at the time of spinal tap, all injected antibody is saturated with Aβ, so that additional Aβ42 coming off the plaques would stay free. Siemers briefly mentioned brain imaging with IMPY in this trial. However, experts in amyloid imaging later noted that IMPY is too poor a marker of amyloid to support firm conclusions, adding that Lilly will hopefully be able to use PIB or one of the 18F amyloid imaging agents that are being tested clinically.
For his part, DeMattos described in more detail fluid biomarker research aimed at characterizing how the humanized m266 antibody acts on soluble Aβ. In essence, DeMattos was looking for markers that would be able to track whether equilibria between different pools of Aβ in the body’s periphery and CNS shift after treatment. DeMattos developed an acid urea gel technique that allowed him to resolve full-length Aβ from within a mixture of modified or truncated species. He then used the 3D6 and the 21F12 antibodies to label the peptides on Western blots, on which, he noted, Aβ species run to their true size. This approach visualized a pattern of different truncated Aβ species extracted from human AD brain. It confirmed published work by others. DeMattos estimated that about 75 percent of Aβ extracted from human AD brain and ending at amino acid 42 is truncated. In DeMattos’s hands, a small fraction of that appeared to be 3-42 pyroglutamate derivatives of Aβ (see ARF related Keystone story), which is thought to be an early aggregating form of Aβ.
Soluble Aβ in brain interstitial fluid is generally thought to be in equilibrium with plaque Aβ, which draws Aβ out of solution. The hope is that m266 will do the opposite, i.e., draw Aβ out of plaques and toward clearance. It’s unproven at present that increasing free CSF Aβ42 will do that, but the new biomarker data hint that this may happen, DeMattos said. When applied to the plasma samples taken during the Phase 2 trial, the acid urea technique detected an accumulation of not only full-length Aβ but also a range of N-truncated fragments that probably originate in the brain. This includes, for example, the pyroglutamate forms. They were never before seen in biological fluids and are considered specific to plaques, DeMattos said. It also includes what DeMattos called “fragment 2,” an as yet uncharacterized Aβ snippet that might be plaque-specific as well. “We see a robust dose-dependent and time-dependent accumulation of this fragment 2 series. We think it is being liberated from plaque. We think we are shifting the equilibrium,” DeMattos said.
Curiously, DeMattos also reported seeing a higher concentration of pyroglutamate Aβ fragments in plasma of some of 16 healthy controls, who had received a single injection of m266. They were people who were also positive for IMPY imaging, suggesting to DeMattos that they had amyloid in their brain and that the bolus of m266 might have drawn some of it out. Overall, the increase in free Aβ42 to DeMattos means that m266 is dissolving various different forms of Aβ out of plaques, from where they make their way to the CSF and then to the blood. The ICAD presentations included no data on whether all that Aβ bound up in plasma is then being degraded and excreted at pace. “The importance to us of this biomarker data is that we know the drug hits its target,” DeMattos said. Added Siemers “We have made a decision to go to Phase 3 with this antibody, and will be starting in 2009.”—Gabrielle Strobel.