The 8th Clinical Trials Conference on Alzheimer’s Disease, held November 5-7 in Barcelona, Spain, displayed the potential pitfalls of each of the many decisions that go into designing trials in early, pre-dementia AD. Amid a broad sense of optimism that anti-amyloid antibodies might actually be working, there was also a sobering humility about just how hard it may be to prove it. Curiously, two opposing trends demonstrated the challenges of trial design jointly faced by two of the most closely watched investigational immunotherapies. Gantenerumab, an N-terminal Aβ42 antibody by Roche, re-emerged to look rather seaworthy after appearing to sink under the flat-out negative result of an interim analysis in December 2014. Aducanumab, a similar antibody being developed by Biogen, looked to observers to be taking on just a little water at CTAD following its breezy Phase 1b results presented at AD/PD this past March. Of course, both antibodies are still the same, both programs are moving forward, and most observers at CTAD believed that the efficacy signals in the overall datasets for both of them are probably real. The respective antibodies' ups and downs as they chart a path through the clinical trials process reflect the challenges of conducting innovative trials when multiple elements—the tools to select participants, the management of side effects, and the ways of measuring progression—are evolving along the way. Read on for an update on both antibodies’ journeys.
First, gantenerumab, which was presented at four talks with various kinds of subgroup analysis at CTAD. Roche’s Robert Lasser prefaced the analyses by saying, “Scarlet RoAD is still a negative trial. We wanted to know what happened, and now we believe there was a signal.” To recap, the trial’s independent data safety monitoring board ordered a halt to dosing last December, and in July 2015 at AAIC, Lasser subsequently showed that the futility analysis prompting this decision had shown no benefit for gantenerumab treatment. That day, Lasser also showed an initial subgroup analysis to suggest that the trial population had comprised about equal parts people who did not progress over the trial’s two-year time frame, people who progressed slowly, and people who progressed quickly (see Dec 2014 news; Aug 2015 conference news). Importantly, the scientists now say, the interim analysis, done on 312 people who had completed the two-year trial, excluded the 190 people who by then had dropped out. It therefore did not factor in, as Roche scientists claimed at CTAD, that the placebo group had lost a disproportionate number of fast progressors. Their departure, in essence, rendered it impossible to detect a treatment benefit in their fellow fast progressors on gantenerumab.
The subgroup analyses presented at CTAD painted a picture of a trial that was masking a response in fast progressors because these people were leaving the placebo group to start symptomatic drugs such as acetylcholinesterase inhibitors or memantine, which the Scarlet RoAD protocol had forbidden. The trial targeted prodromal AD, i.e., it enrolled still-functional people who worry intensely about their memory, but did not allow them to take approved AD medications while they were helping test this investigational drug. As the trial went on, the fast progressors in the placebo group noticed their memory worsening, concluded correctly they must be on placebo or the drug was not working, and left the trial to go on symptomatic medications instead, Roche scientists said. In the treatment group, it was different. The fast progressors responded to gantenerumab, so they perceived little decline and stayed in the trial. But because the treatment group also contained slow progressors and non-progressors, the treatment benefit of the fast progressors could not be discerned.
There was intense buzz in the hallways about how AD trials remain full of surprises as the field learns collectively. Roche apparently had not anticipated that participants who were barely progressing, combined with a selective de-riching of fast progressors from the placebo group, could obscure a treatment effect in the interim analysis. “By losing untreated fast progressors, and turning treated fast progressors into slow progressors, an analysis that looks only at completers would cancel out a treatment effect,” one scientist summed up. Alzheimer clinicians at Scarlet RoAD sites in both Europe and the United States spoke only on condition of anonymity, but they agreed with this assessment. In fact, some had expressed displeasure in private all along about having to withhold AD medications. Clinicians recommended that stable use of approved Alzheimer’s medications be allowed in the next gantenerumab trial to pre-empt this incentive to leave. Acetylcholinesterase inhibitors are not indicated as standard of care for prodromal AD, particularly in Europe; however, in practice patients who participate in clinical trials know these drugs exist and can help a little, and they expect the best possible care, including access to these drugs, clinicians say.
Some trialists, notably Rachelle Doody of Baylor College of Medicine in Houston, have urged sponsors to track different progression rates in their trials (see Dec 2014 CTAD story). In 2010, when Scarlet RoAD started, trials were only just beginning to target prodromal, biomarker-confirmed Alzheimer’s, and requiring biomarker evidence of amyloid positivity was the big new thing. Researchers had no handle on how to predict progression rates at baseline. Five years later, that is changing. At CTAD, numerous groups, armed with data correlating progression with baseline characteristics in various observational cohorts, discussed the performance of experimental prediction algorithms—combining elements from ApoE genotype, baseline amyloid load, or performance on episodic memory tests or online tests. The emphasis now has moved to getting better at finding pre-dementia patients who are likely to decline measurably during the lifetime of a treatment trial.
At CTAD, Roche’s Lasser noted that of Scarlet RoAD’s 797 enrolled patients, a fifth progressed to dementia in the two years of the trial. At baseline, they had in common greater functional impairment on the CDR-sb and the FAQ. Beyond that, however, nothing jumped out of this dataset to lend itself as an easy inclusion criterion to find fast progressors, Lasser said. Also at CTAD, Paul Maruff of Cogstate, Melbourne, Australia, reported that in the AIBL natural history dataset, ApoE4 strongly predicted future memory decline in cognitively normal people; however, Lasser said that in Scarlet RoAD’s population of already mildly symptomatic people, ApoE4 offered no further predictive value.
To explore progression in the Scarlet RoAD cohort, Sylvie Retout of Roche took a modeling approach with data from ADNI. Showing spaghetti plots of stark differences in individual people’s trajectories over time, Retout said that small hippocampal volume and high scores on the FAQ and CDR-sb at baseline marked people who then declined rapidly. Retout used these features from ADNI to classify the Scarlet RoAD population at baseline to predict slow, medium, and fast progressors. Indeed, two years later, the predicted slow progressors had worsened two points on ADAS-cog, the “medium” group by five points, and the predicted “fast” group by six points. Other outcome measures used in this trial, such as the CANTAB and even the MMSE, showed similar trends; however, the CDR-sb did not, Retout said. About 250 patients per arm were available for these post hoc clinical analyses.
Separately, Retout sliced the data not by baseline predictors but by plasma gantenerumab concentration, i.e., exposure to study drug. She split the high-dose arm into people with low, medium, and high blood gantenerumab levels, and found that efficacy correlated with exposure. The largest treatment benefit, six points on the ADAScog, accrued to people with the highest blood antibody concentration, Retout said. This finding, as well, showed up on CANTAB and MMSE, but not CDR-sb. “The efficacy signal increases with gantenerumab plasma concentration,” Retout said.
In an attempt to understand the effect of the drug in the overall study population, including dropouts, Juergen Dukart of Roche computed slopes of progression for all patients in the trial. He then compared the slopes between the treatment arms. At CTAD, Dukart reported that this analysis revealed dose-dependent slowing of disease progression on all clinical scales used in the trial. He said that in a subsequent sub-analysis looking separately at dropouts and patients who completed the trial, the dropouts in the placebo group turned out to have had the fastest rate of progression when compared to patients in both treatment arms in all clinical scales used. Among the placebo group, people who left the trial were progressing up to four times faster than the people who stayed, Dukart said. When comparing the dropouts' reasons for leaving, those who left the placebo group more often cited lack of efficacy, whereas treatment group dropouts were more often due to cited side effects or other reasons, Dukart said. From the patient’s point of view, this makes sense, clinicians later commented. If you progress slowly or not at all, you cannot tell if you are on placebo or drug and see no reason to change. If you decline rapidly, you conclude the trial is not helping and may consider alternatives.
Biomarker findings appear to line up with the company’s claim that the higher dose of gantenerumab may have done a little bit of good in some patients. Roche’s Tania Nikolcheva presented the trial’s florbetapir PET substudy. One hundred and fourteen people initially enrolled, but the stop to dosing left only 27 placebo, 21 low-dose, and 19 high-dose patients available for analysis of repeat scans at one year. Beyond the overall finding, previously reported at AAIC, of a modest 5 percent dent in global brain amyloid load on the higher dose, at CTAD Nikolcheva showed new data relating amyloid removal to gantenerumab exposure. Splitting participants into three bins of low, medium, or high blood gantenerumab levels, she saw that the high-exposure group had the biggest difference, up to 10 percent less brain amyloid than placebo. Dukart presented a separate post-hoc, voxel-wise analysis of the PET data, claiming that when analyzed by brain region, amyloid reduction on the higher dose reached up to 18 percent relative to placebo in areas where amyloid is known to build up strongly in AD, such as the precuneus. These differences were dose-dependent and statistically significant, Dukart said. He also correlated gantenerumab exposure in the CSF to amyloid PET SUVR and, like Nikolcheva’s plasma-based exposure analysis, connected higher exposure to more amyloid removal. “We are confident that the drug is doing something in a dose-dependent manner,” Dukart said.
On CSF biomarkers, a 10 percent reduction of phospho-tau on the higher gantenerumab dose had been reported at AAIC. At CTAD, Nikolcheva added that more reduction in CSF p-tau tracked with more reduction in amyloid PET. This correlation was dose- and time-dependent, Nikolcheva said. She reported a p value of 0.039 and a correlation coefficient of 0.43 for this association; 23 patients were available for this exploratory post-hoc analysis. “This indicates that targeting amyloid can influence downstream processes reflected by CSF tau,” Nikolcheva said, adding “Clearly we need to reproduce this data in a larger set and with higher doses.”
Even when taken with heaping grains of salt, as befits interpretation of post hoc analyses, the data suggest, if nothing else, that the gantenerumab doses in this trial were too low, researchers at CTAD widely agreed. The current open-label extension of Scarlet RoAD aims to offer participants higher doses of gantenerumab. The next Phase 3 trial will attempt to dose up to four times higher. In conversation, Roche scientists make a comparison to Biogen’s aducanumab. This antibody started showing treatment effects at 3 mg/kg, which roughly corresponds to the 225 mg at which Roche now believes to have seen a signal in fast progressors. The challenge for both programs then becomes one of managing the ARIA side effects, enabling amyloid clearance without hurting the patients.
Aducanumab has entered Phase 3 trials following the Phase 1b PRIME trial of 165 enrolled patients. This antibody has similarities (though also differences, see May 2015 news) in its Aβ42 binding characteristics and ARIA liability to gantenerumab. At CTAD, Biogen researchers presented bits of new data in three presentations. Jeff Sevigny reported an analysis on regional reduction in brain amyloid as measured by florbetapir PET. He noted that amyloid deposition was not driven by a single region but occurred in a set of regions known from natural history studies to accumulate amyloid deposition early on in AD. They include the anterior cingulate cortex, the frontal cortex, the posterior cingulate and parietal cortex, and the striatum, among others. In addition, Sevigny showed a correlation at one year between change in PET SUVR and two clinical measures, the CDR-sb and the MMSE. Aducanumab-treated participants who had one standard deviation or less of reduction on amyloid PET, as defined by the placebo group at week 54, continued to decline on these clinical measures, whereas clinical progression was slower in treated people who had more than a one standard deviation amyloid reduction. Sevigny cautioned that this correlation was drawn with exploratory clinical endpoints in this trial, which was primarily a safety and target engagement study.
More puzzling for observers was Sevigny’s report that aducanumab showed no benefit on two other exploratory outcome measures at one year. They are the free and cued selective reminding test (FCSRT) and a neuropsychological test battery (NTB). “We saw no treatment effect, but we did see floor effects, suggesting these tests are unsuitable for this amyloid-positive population with an MMSE of 20 or higher,” Sevigny said. In the minds of clinicians, this finding somewhat weakened the trial’s efficacy conclusion, which is generally considered strongest when a treatment benefit occurs consistently across all assessments. It also revived speculation heard earlier that the PRIME study’s exploratory clinical benefit on the CDR-sb may have been subtly influenced by a potential rater bias that can be inherent in an escalating-dose trial.
Vissia Viglietta of Biogen presented a post hoc analysis of those people among randomized PRIME population who most closely matched the type of patient Biogen is currently enrolling into its Phase 3 trials of aducanumab. The Phase 3 patients are similar to the PRIME population spanning the prodromal and mild parts of the AD continuum. One difference is that prospective patients will get picked for a screening PET scan based on a score at or below 85 on the RBANS cognitive scale, as opposed to the FCSRT test that was used in PRIME for prodromal subjects. Viglietta said that the RBANS better predicts amyloid positivity and might reduce screen failures due to amyloid negativity, which have bedeviled both PRIME and Scarlet RoAD, as well as the ongoing A4 trial. Of the 165 patients in PRIME, 92 met the Phase 3 criteria, Viglietta said, and their baseline characteristic made them seem a little less clinically affected, with two-thirds being prodromal and one-third mild AD, whereas two-third of the patients in PRIME overall had mild AD and one-third were prodromal. “This is overall a slightly milder population,” Viglietta said. The results in this subgroup were similar to PRIME overall in that the antibody dose-dependently removed brain amyloid while being associated with ARIA-E in a dose- and ApoE-dependent pattern. On the CDR-sb and MMSE, results still favored aducanumab.
Viglietta presented details about the aducanumab Phase 3 program. Two identical Phase 3 trials, called ENGAGE and EMERGE, will each combine ApoE 4 carriers and non-carriers, but dose them differently in an attempt to mitigate the risk of ARIA. In ApoE 4 carriers, the dose will be slowly titrated up to 3 or 6 mg/kg. Likewise, ApoE 4 non-carriers will slowly ramp up the dose and stop at 10 mg/kg. “We are trying to treat every patient at their optimal dose,” Viglietta said.
Biogen estimates that it will enroll about 3,000 patients across both studies, and randomize people 2:1 into treatment versus placebo. These trials allow symptomatic AD medications but require that participants have been on a stable regimen for eight weeks prior to starting aducanumab or placebo.
The studies will run for 18 months; people who complete it and have at that time an MMSE above 15 can then join a 24-month long-term extension study. In the extension study, they will know they are getting the antibody, Viglietta said, but will be assigned to a dose and participants will remain blinded to which arm they were in during the first 18 months. This will enable Biogen to conduct a delayed-start analysis to assess both lasting efficacy and safety similar to what Lilly attempted with solanezumab (Jul 2015 conference news). Enrollment is underway in the United States and soon to begin in other countries, for a total of 300 sites around the world, Viglietta said.
The aducanumab and gantenerumab results fueled discussion throughout CTAD about the CDR-sb as a primary outcome for pivotal Phase 3 trials in very mildly affected patients. This informant- and patient-based rating began as a tool to stage dementia, and is widely used for that purpose in the NIA-funded system of Alzheimer Disease Research Centers and in many clinics beyond. Clinicians broadly agree on its face value and ability to pick up inter-individual change, and sites throughout the world are able to administer it. When the ADAS-cog fell out of favor as being insufficiently sensitive in the increasingly early AD populations that treatment trials were targeting, the CDR-sb drew notice for its performance across the disease continuum in ADNI and for its mention in an FDA guidance on conducting trials in early AD (Kozauer and Katz, 2013). At CTAD 2015, however, debate intensified about the CDR-sb as a single primary outcome measure. Aducanumab, as well as the Merck BACE inhibitor verubecestat, have essentially hinged their success on it. The design of the next Phase 3 gantenerumab trial in prodromal AD is not yet known (for one perspective on this question, see also Q&A with Steve Ferris of New York University).—Gabrielle Strobel
- End of the RoAD for Gantenerumab? Roche Declares Prodromal Alzheimer’s Trial Futile
- Aducanumab, Solanezumab, Gantenerumab Data Lift Crenezumab, As Well
- CTAD Shows Alzheimer’s Field Trying to Reinvent Itself
- Shape of a Hug: How the Embrace of a Therapeutic Aβ Antibody Really Matters
- Outcomes, Outcomes: Cognition is Crux of New Alzheimer’s Trials
- Kozauer N, Katz R. Regulatory innovation and drug development for early-stage Alzheimer's disease. N Engl J Med. 2013 Mar 28;368(13):1169-71. Epub 2013 Mar 13 PubMed.
No Available Further Reading