“Failures, repeated failures, are finger posts on the road to achievement. One fails forward toward success.” C.S. Lewis.
Flashed onto the projection screen by Roche’s Robert Lasser, the Irish novelist’s quote symbolized the “chin-up!” attitude that prevails among developers of anti-amyloid immunotherapy these days. The most anticipated news at this year’s Alzheimer’s Association International Conference were clinical trial updates on four antibodies that tackle various versions of the Aβ peptide—monomer, aggregated forms, or both. As it turned out, none of the data was momentous in its own right. By itself, each antibody’s performance was mixed, “meh,” or negative, respectively. Still, taken together, they fed a confidence across AAIC that Aβ immunotherapy may become a treatment, albeit one at only the very early stages.
Chief among the concerns is the side effect called amyloid-related imaging abnormality, ARIA. Clinicians are gaining more experience with it across the different antibodies currently in clinical trials. As they do, their attitude is evolving from tossing out an otherwise promising investigational therapeutic because of ARIA toward wanting to study and minimize it. “The problem is ARIA,” commented David Holtzman of Washington University, St. Louis, “There are challenges ahead managing this side effect, but I am optimistic that it looks like a solvable problem in the long run.” Philip Scheltens of VU Medical Center in Amsterdam said, “We need to be less risk-averse in Alzheimer’s disease. We should carefully dose up until side effects tell us to hold off. In cancer, we tell the patient: 'You will get nauseous from these drugs.'”
Taken on their own, none of the antibodies wowed the AAIC audience with new data. Aducanumab’s 6-milligram dose did not fit perfectly on all endpoints into the empty space waiting for it between the 3 mg and the 10 mg dose the way the audience—especially investors—had expected. Gantenerumab was flat-out negative on all endpoints in the SCarlet RoAD trial in prodromal Alzheimer’s disease, save for some blips in biomarker and subgroup analysis. Solanezumab came across as straining to prove disease modification for a small effect with what some called a geeky new statistical approach. Yet somehow, in the aggregate, the signs in the tea leaves, paired with consensus that the field’s measurement techniques are slowly improving, still added up to a positive vibe. By the end of AAIC, Roche had announced that it was moving into Phase 3 with gantenerumab and also crenezumab, the AC Immune antibody it is developing with Genentech (see Jul 2014 conference news). Aducanumab has begun enrolling for Phase 3, and solanezumab will finish dosing in its third Phase 3 trial next October.
Aducanumab: Just Shy of Sky-High Expectations
Jeff Sevigny of Biogen showed new results from the Phase 1b trial of aducanumab, aka BIIB037, in people with prodromal AD. This presentation wrapped up the double-blinded portion of this study, and follows Sevigny’s presentation at the AD/PD meeting in Nice, France (see March 2015 conference news), of a 6 mg/kg dose and its placebo arm. Biogen had added this arm to the study when it became apparent that the 10 mg/kg dose was causing ARIA in a large proportion of patients. At AD/PD, Sevigny had only been able to show six-month data for this dose.
By one year, this dose had achieved a statistically significant reduction of brain amyloid as per florbetapir PET. This was widely seen at AAIC as the trial’s strongest, uncontested success. In exploratory analyses, the antibody also appeared to slow clinical progression as measured by the CDR-SB. The point value of this effect tracked with the dose-response relationship reported at AD/PD for the 1, 3 and 10 mg/kg doses. Specifically, Sevigny showed that the placebo group declined by 1.87 points, vis-à-vis 1.72 in the 1 mg/kg group, 1.37 in the 3 mg/kg group, 1.11 in the 6 mg/kg group, and 0.63 in the 10 mg/kg group. Except for the highest dose, each individual dose effect was not statistically significant on its own, but the overall dose-dependence of aducanumab’s effect on CDR-SB was significant in a test called linear trend of dose response, Sevigny said.
The blemish was the MMSE. There, the absolute value of the 6 mg dose (worsening by 1.96) did not fall between the 3 and 10 mg/kg doses but hovered nearer the 1 mg/kg dose. The point values were 2.81 reduction for placebo, 2.18 in the 1 mg/kg group, 0.7 in the 3 mg/kg group, and 0.56 in the 10 mg/kg group. The 3 and 10 mg/kg doses were close together and statistically significant, the 6 and 1 mg/kg doses were not significant. As with the CDR-SB, the linear test of dose response for overall dose dependence was significant, Sevigny reported.
The 6 mg/kg dose results were closely anticipated because of the expectations raised after Biogen’s AD/PD presentation. In defense, Sevigny told Alzforum, “We are talking about point estimates on scales that are being deployed in a Phase 1b study powered for PET imaging. As a drug developer and clinician, I find these results fantastic. Keep in mind what we can expect to see in this type of study with 30 people per active arm.”
The immediate stories by analysts and in the financial press were contradictory, and on July 24, when Biogen adjusted 2015 sales projections for its multiple sclerosis drug Tecfidera downward, the stock dropped. Alzheimer’s scientists largely shrugged off the market for now and focused on the new results. “The aducanumab data look good,” said Colin Masters of the University of Melbourne in Australia. “The new data look consistent with what Biogen have already reported. The fact that they got a signal across the measurements is encouraging for the field,” Holtzman concurred.
ARIA-E, a poorly understood kind of edema, was common in this trial. It occurred more often at higher doses and in people who carry the ApoE4 allele. In carriers, the incidence of ARIA-E climbed from 5 percent in the 1 and 3 mg/kg groups to 43 percent in the 6 and 55 percent in the 10 mg/kg group. In non-carriers, it still was 9 percent in the 3 mg/kg, 22 percent in the 6 mg/kg, and 17 percent in the 10 mg/kg group. A fraction of those with ARIA-E discontinued treatment; 56 percent continued, some at a lower dose. Patients had no further ARIA-Es after the first instance, Sevigny told Alzforum.
Of the ARIA-Es, 89 percent developed early in the course of treatment. “Since the AD/PD meeting, we have only had one additional case. Something happens biologically in a significant fraction of patients; if it does not happen early then it tends not to happen later, ” Sevigny said. A third of ARIA-E cases were symptomatic, and resolved, the remainder were only detected on the monitoring MRIs.
Other scientists were cautiously optimistic about what this might mean for the future. Several clinicians noted that it will be important for aducanumab’s future to learn how these patients fare in the long term. “The hope is we can learn to manage it. It depends on how severe the ARIA is clinically,” said David Knopman of the Mayo Clinic in Rochester, Minnesota.
Besides ARIA, headaches were common. Sevigny said they were unremarkable and resolved quickly. A few people developed antibodies against aducanumab, but these antibodies neither caused clinical symptoms nor reduced aducanumab blood levels, he told Alzforum.
Biogen has started two Phase 3 studies in a total of 2,700 mildly symptomatic patients in the United States and abroad. Both enroll ApoE4 non-carriers and carriers, the latter on a lower dose. Sevigny would not disclose the doses in Phase 3. About ARIA, he said that his company is focusing research on this phenomenon. “We are committed to being world experts in ARIA. It’s not like we say, ‘There is ARIA and we have to live with it.’ We are working on improving it,” Sevigny said.
Prior trials of prodromal AD used the free and cued selective reminding test, but struggled with high screen failure rates. Sevigny said patients would have to have measureable evidence of a cognitive deficit, but declined to say which memory test or exact screening procedure Biogen is using to speed up enrollment.
On a poster at AAIC, Yaming Hang of Biogen showed some pharmacokinetic characteristics of aducanumab. The higher a person’s blood aducanumab level, the stronger their amyloid removal in different brain regions of interest, regardless of ApoE genotype, Hang said. The distribution of the antibody in the body and its clearance depends on a person’s weight, and cumulative exposure in people who received multiple injections is linear, according to the poster. In pharmacology-speak, this implies to scientists that the antibody is “well-behaved,” Hang said.
Gantenerumab: Flatline Trial Shows Little Blips of Life
At AAIC, scientists picked up the pieces from gantenerumab’s setback last December, when a futility analysis halted the prodromal AD trial called SCarlet RoAD because the data thus far made it seem unlikely that the trial would show a treatment effect (see Dec 2014 news). Dosing stopped, but patients kept coming for assessments, and last month at AAIC, Lasser and Scheltens presented the trial’s clinical and biomarker results.
Lasser’s disappointment came through when he noted that a trial of this size engages thousands of people in dozens of countries—physicians, site staff, and patients and their partners. For SCarlet RoAD, 3,089 prospective participants were screened to enroll 799 people with a defined deficit on the FCSRT memory test and CSF evidence of Aβ deposition. They got monthly shots of either 105 or 225 mg of gantenerumab under the skin; ApoE4 carriers received the lower dose.
Lasser reported that, as with aducanumab, the main side effect of gantenerumab was ARIA. In most instances it caused no clinical symptoms, i.e., was evident only on the mandated safety MRIs. The incidence of ARIA-E and H ranged from 0.4 to 14 percent, rising both with gantenerumab dose and with the number of ApoE 4 alleles a patient carried.
Placebo and both gantenerumab dose groups were no different on the CDR-SB, MMSE , ADAS-Cog13, and FAQ outcome measures. “There was no efficacy,” Lasser and Scheltens both said explicitly. The researchers dug into the data to analyze subgroups in order to learn from the data, not to finesse a positive result from a negative trial, Scheltens emphasized to reporters in a media briefing. That analysis hinted that the patients whose disease progressed the fastest may have responded, though that made no difference in the overall outcome. Different progression rates from person to person, and the field’s inability to predict with any precision how quickly a given person will progress, are longstanding problems in Alzheimer’s disease trials. In this instance, the fast progressors—i.e., those whose hippocampal volume and CDR-SB performance declined the most over the duration of the trial—appeared to benefit, especially those whose serum levels of gantenerumab were high. “In fast progressors, we detected a concentration-dependent treatment effect on ADASCog and MMSE,” Lasser said. “This is a post hoc analysis, however.”
Looking at biomarkers, Scheltens spotted signs of biological activity in the trial’s wreckage. At AAIC, he reported that a florbetapir PET substudy showed that brain amyloid load dipped by about 5 percent in the high-dose group. This dose further reduced CSF total tau by 2.94 percent compared to an increase of 3.11 percent in those on placebo, and reduced CSF phospho-tau by 7.52 percent compared to a 2.62 percent increase in placebo. The point values for the mild dose fell in between. “This is the first antibody to show consistent improvement in target and downstream biomarkers of neurodegeneration. It seems to change something in the cascade,” Scheltens said.
This amounts to a negative clinical result in the face of an effect on amyloid and tau pathology. The audience question shot out immediately: “Does this mean the amyloid hypothesis is disproven? You have target engagement but no clinical benefit.” True, Scheltens conceded, but he suggested instead that the dose of gantenerumab in this trial was simply too low to make a clinical difference.
In preclinical studies, gantenerumab has performed similarly to aducanumab. The high dose in the SCarlet RoAD trial roughly compares to the 3 mg/kg dose in the aducanumab trial. Those SCarlet RoAD participants with the highest serum gantenerumab levels showed the most amyloid removal, and an overall comparison of the dose and pharmacokinetic relationship of both antibodies suggests that gantenerumab exposure in this trial was at the bottom of the range where aducanumab started showing efficacy. “I think it is right to continue evaluating gantenerumab,” Scheltens told Alzforum. Knopman and Holtzman agreed that the dose may have been too low.
Matthew Frosch of Massachusetts General Hospital, Boston, recommended digging deeper into the data of the PET substudy to learn if brain areas with ARIA showed more amyloid clearance than others.
Upping the gantenerumab dose would likely mean more side effects. “ARIA thus far has been self-limiting, without any long-term consequences,” Lasser said. For now, participants in the SCarlet RoAD trial can opt to take the antibody in open label at a higher dose, and future high-dose trials are being discussed with the EMA and FDA. This news was not lost on attentive participants in the DIAN-TU trial of gantenerumab (Aug 2015 news), who wondered in their social media group whether their dose could be increased, as well.
Solanezumab: What Does a Smidgen of Disease Modification Mean?
With this most advanced antibody against monomeric Aβ, the news at AAIC was different. Solanezumab, an antibody against Aβ’s midregion, is currently in its third Phase 3 trial. Rather than sharing data from it, Paul Aisen, University of Southern California, San Diego, presented a new statistical analysis of 3½-year treatment data from an open-label extension of the first two Phase 3 trials of solanezumab, called Expedition 1 and 2. “This is the first time the delayed-start method has been formally implemented in an AD trial,” Aisen said.
This far out, he reported, it is clear that people who chose open-label solanezumab after having been on placebo for the first 18 months benefitted, but they never caught up to fellow participants who had been on solanezumab from the get-go, i.e., 18 months longer. Hence, the claim goes, the antibody lastingly modifies the underlying disease process rather than just patching up its symptoms for a while, and for this reason should be started as early as possible. In a searching discussion session, however, regulatory scientists gave this delayed-start approach a cold shoulder. While complimenting the statistics, they mostly emphasized that achieving a large treatment effect is far more important than trying to prove that a small effect is disease-modifying.
The Expedition 1 and 2 trials were negative. Even so, a secondary analysis of the pooled mild subgroups, done by Lilly and also by the Alzheimer Disease Cooperative Study, both spotted an efficacy signal the scientists calculated to reflect 34 percent slower progression (Siemers et al., 2015; Oct 2012 news). In the present study, Lilly scientists implemented a trial design for showing disease modification that was proposed by former FDA scientist Paul Leber, who directed the agency’s Division of Neuro-Pharmacological Drug Products from 1981 to 1999 (Leber, 1996).
Called “delayed start,” it makes use of the commonly practiced open-label extensions, whereby everyone in the trial is offered drug after the placebo-controlled phase is over. A delayed-start study then analyzes whether the decline trajectories of the two groups merge over time (indicating symptomatic treatment) or stay parallel (indicating disease modification). The idea is that because people on placebo lose more neurons during the first 18 months than the people on drug, switching to drug at 18 months will start to slow down their neuronal loss at that point, but they will never catch up to the people who started taking the disease-modifying from the beginning.
The trouble with this design has always been that it is statistically complex, especially because more and more people drop out when trials drag on for years and comorbidities and advancing AD overwhelm the aging participant’s ability to come to the clinic for monthly shots and assessments. It has been tried once before, in the ADAGIO trial of rasagiline for Parkinson’s. There it was only a partial success, either because this transmitter-based drug truly has mixed effects, or because of the statistical challenges inherent in this design (Rascol et al. 2011). Conceptually, antibody therapeutics should give an either-or answer, Lilly’s Eric Siemers told Alzforum.
Of the solanezumab Expedition 1 and 2 participants, 95 percent chose open-label drug. Both patients and study staff continued to stay blinded to what patients had been on previously. Lilly’s statistician Hong Liu-Seifert developed a new way of analyzing the long-term data. It tests whether the treatment difference at the end of the placebo phase persists going forward or, in statistics lingo, stays “non-inferior.” At AAIC, Aisen presented data on up to two years of delayed-start treatment. In essence, the difference persisted on the ADAS-cog and the ADCS-ADL scales for a year. Over the course of the second year, as more people dropped out, statistical significance was gradually lost, but to the naked eye the lines of decline of the two groups stayed parallel. In other words, the people who had been on drug since day one continued to do somewhat better. By 3½ years, about half the people had dropped out, about equally in both the early and delayed-start groups.
“I was surprised that after two years of open-label we still see those hypothetical curves,” Siemers told Alzforum. He added that this additional data makes it more likely that solanezumab’s effect is real, not due to chance.
During an AAIC panel discussion of the delayed-start results, however, government regulators politely poured cold water on the effort to show disease modification for the Expedition 1 and 2 trials. Aisen moderated the discussion. Lisa LaVange directs biostatistics at the Food and Drug Administration. She said that while it would be wonderful to get disease modification, the treatment difference at the beginning of the delayed-start phase needs to be large enough to withstand the statistical complexity and the missing data problem of this type of trial. In essence, sponsors should show first that the difference is meaningful, and only then focus on showing that it is durable, LaVange said. Aisen noted that with disease modification, the clinical effect tends to grow over time, and asked if that lowers the bar for what is clinically meaningful at 18 months. “I don’t know. It makes no sense to me to show you are not inferior to something that is not meaningful to begin with,” Lavange said.
Rusty Katz, formerly of the FDA, acknowledged that Seifert’s approach addressed some of the statistical challenges that had bedeviled the ADAGIO trial, but he reminded the audience that Expedition 1 and 2 were negative trials. “We live and die by primary analysis. All subsequent analyses are suspect, and in this situation it is hard to know what the p values mean,” Katz said. Nick Kozauer of the FDA agreed that while disease modification is conceptually important, it should not be conflated with a big effect, “What’s really important is a meaningful effect,” Kozauer said.
Aisen next asked whether disease modification is important for drug approval. Billy Dunn of the FDA gave a clear answer. “Absolutely not. We work hard to dissuade sponsors from thinking that demonstrating disease modification in a modest fashion assumes some intrinsic importance over the pursuit of a dramatically large effect.” Dunn went on to explain that disease modification is intrinsically linked to the persistence of effect in the absence of drug. He agreed that the construct of disease modification is laudable. “But we should not take our eyes off the prize: a large treatment effect. I won’t comment on this particular data, but we are not dealing with a monster effect,” Dunn said.
In conversation, other clinicians agreed that they would prefer to see clinical meaningfulness on the primary outcome, get the drug to patients, and then figure out exactly how it works. But some biostatisticians greeted the delayed-start data more warmly as a sign that clinical trials methodology is improving. Take Suzanne Hendrix of Pentara Corporation, Salt Lake City. “I am excited that we can analyze this now. The data is clean, there is so little noise that we can see this small effect for 36 months,” Hendrix said. Being able to detect small effects means the field is not missing something that could be made to be meaningful with more work. Sometimes it is possible to build on a small effect by improving formulation, dosing, time of intervention, and especially by combining drugs (see Part 1).
“In many prior AD studies, we have been overwhelmed with variability and noise. Just look at spaghetti plots—they are all over the map. Even means plots often zigzag. It’s good to see smooth curves with tight error bars. It means that we now know better what we are seeing,” Hendrix said.
The ongoing Expedition 3 trial of solanezumab will add the same delayed-start analysis once its randomized phase concludes in October 2016. So far, dropout is low, at 8 percent, Siemers said. That may be because the participants have mild disease, but also because they know they have brain amyloid and that solanezumab looks to be quite safe.
The antibody crenezumab had no data presented at AAIC, but on the last day, word went around that its sponsor, Genentech, had decided to evaluate it in prodromal to mild AD in Phase 3 (see press release). The news came after a long period of deliberation following crenezumab’s Phase 2 results presented at AAIC 2014 (Jul 2014 conference news). Despite differences in IgG subclass and binding characteristics, crenezumab is widely considered to be clinically similar to solanezumab (see May 2015 news). With that, all of these four anti-Aβ antibodies remain standing in the long slog toward a new treatment for Alzheimer’s disease. —Gabrielle Strobel
- Crenezumab Disappoints in Phase 2, Researchers Remain Hopeful
- End of the RoAD for Gantenerumab? Roche Declares Prodromal Alzheimer’s Trial Futile
- Families Challenge DIAN Scientists to Do Even More, Faster
- The Solanezumab Benefit: Oh, So Small, But Probably Real
- Massive Mouse Study Bolsters Rationale for Combination Therapy
- Shape of a Hug: How the Embrace of a Therapeutic Aβ Antibody Really Matters
- Siemers ER, Sundell KL, Carlson C, Case M, Sethuraman G, Liu-Seifert H, Dowsett SA, Pontecorvo MJ, Dean RA, Demattos R. Phase 3 solanezumab trials: Secondary outcomes in mild Alzheimer's disease patients. Alzheimers Dement. 2015 Aug 1; PubMed.
- Leber P. Observations and suggestions on antidementia drug development. Alzheimer Dis Assoc Disord. 1996;10 Suppl 1:31-5. PubMed.
- Rascol O, Fitzer-Attas CJ, Hauser R, Jankovic J, Lang A, Langston JW, Melamed E, Poewe W, Stocchi F, Tolosa E, Eyal E, Weiss YM, Olanow CW. A double-blind, delayed-start trial of rasagiline in Parkinson's disease (the ADAGIO study): prespecified and post-hoc analyses of the need for additional therapies, changes in UPDRS scores, and non-motor outcomes. Lancet Neurol. 2011 May;10(5):415-23. Epub 2011 Apr 7 PubMed.
No Available Further Reading