Despite an adaptive design to grease its wheels, a Phase 2 clinical study of the anti-Aβ antibody BAN2401 demonstrated no clinical benefit after 12 months in people with early AD, according to a December 21 announcement by trial sponsors Eisai and Biogen. Sponsors now await the 18-month time point to draw conclusions about the efficacy of the treatment and decide whether to move into Phase 3. While the findings come as a disappointment to those hoping for speedy results, they leave open the possibility of success for this proof-of-concept trial.
- A Phase 2 trial of the Aβ protofibril antibody BAN2041 showed no cognitive benefit after 12 months.
- This despite an adaptive enrollment scheme to tease out early treatment benefit.
- As per the trial design, participants will continue to 18 months.
“The BAN2401 trial results so far (to the extent that we know them) indicate that the therapy is neither extremely effective nor clearly ineffective,” commented Don Berry of University of Texas MD Anderson Cancer Center in Houston, who helped design the trial. “On balance and in context of the sorry state of drug development in Alzheimer’s disease, that’s good news,” he added.
Phase 2 clinical trials aim to confirm the safety and tolerability of a drug, find doses that engage their targets, and assess whether a glimmer of hope exists for clinical efficacy in much larger Phase 3 trials. Proving efficacy is a tall order in AD trials, in part because cognition declines slowly and no surrogate biomarkers are available. To run an adequately powered Phase 2 trial of this kind, hundreds of participants must be tested for 18 months or longer. The costly nature of this endeavor has led to some drugs moving into Phase 3 based on smaller Phase 2 studies that were insufficiently powered to detect cognitive changes.
Reeling from a string of Phase 3 failures, researchers have been looking for ways to reverse the trend. Adaptive Phase 2 trials, which have been employed in oncology for years, emerged as a potential savior (Mullard, 2014). Clinicians adapt the trial protocol to data that emerges at successive interim analyses. One adaptation adjusts randomization schemes to lend more statistical power to dose groups with nascent positive signals. Ideally, this allows for an earlier determination of efficacy, or lack thereof.
Researchers employed Bayesian statistics to design such a trial for BAN2401, a monoclonal antibody specific for Aβ protofibrils. Eisai and Biogen licensed the antibody from BioArctic, a company started by Lars Lannfelt at Uppsala University, Sweden, after he discovered the Arctic mutation in APP (Apr 2011 conference news). Researchers announced their plans for the innovative trial back in 2012, and published a detailed description last year (Nov 2012 conference news; Satlin et al., 2016).
In a nutshell, the study started by enrolling 196 participants with MCI due to AD, or mild AD. Fifty-six were randomized to the placebo group. The remaining 140 were randomized equally to five different dose groups: 5 and 10 mg/kg monthly, and 2.5, 5, and 10 mg/kg biweekly. Each was assessed for cognition at baseline and every three months thereafter. The researchers conducted the first interim analysis after these initial 196 participants were enrolled, and then another every time 50 more joined the study. At each analysis, they assessed performance on ADCOMS, a cognitive composite test developed by Eisai (Wang et al., 2016).
For the first few interim analyses, the researchers only assessed whether the cognitive effects crossed a futility threshold, below which the trial would be halted. Once 350 people were enrolled, the options also included a success threshold, above which recruitment for the trial would stop but enrolled participants would continue the 18-month trial. Neither futility nor success thresholds were crossed, so the trial continued to full enrollment of 856 participants at 12 months.
Importantly, Bayesian statistics were used after each interim analysis to reshuffle the randomization deck in favor of dose groups showing signs of benefit. In all, 17 interim analyses were conducted in the first 12 months of the trial, but exactly how randomization adapted in response to each analysis will not be revealed until the trial data are unblinded.
A change in decline on the ADCOMS score, relative to placebo, between baseline and 12 months served as the trial’s primary endpoint, and success was judged as an 80 percent or greater probability of achieving at least a 25 percent drop in rate of decline. The 12-month results indicate that while the trial failed to meet this endpoint, it also did not reach the threshold for futility. The double-blind trial will now continue until both cognitive and biomarker/brain imaging measures can be assessed as secondary outcomes.
“The 12-month outcome was the trial’s primary endpoint. But the mature 18-month results will be critical in deciding whether BAN2401 is worth pursuing as a single agent,” Berry wrote to Alzforum. Importantly, Berry added that the futility analysis ruled out the possibility of clear negative effects or a total lack of response.
Lannfelt agreed. “After 17 interim analyses, the study has not been stopped for futility,” he told Alzforum. He added that given the stringent criteria for early success (a 25 percent drop in rate of decline), he would have been surprised if the trial had met this primary endpoint. Lannfelt noted that once the 18-month results are in and the study is unblinded, researchers will be able to critically assess whether the adaptive design helped tease out a signal of efficacy, if any emerge.
What do these results mean for the field? Not everyone is convinced that adaptive trials are the wave of the future. Paul Aisen of the University of Southern California in San Diego commented that while adaptive designs have proven useful in the oncology field where endpoints are more objective, they may be problematic in AD trials. “The greatest concern about Bayesian adaptive approaches in AD trials relates to our cognitive/clinical outcome measures,” he wrote to Alzforum. “These measures are highly variable and noisy. Adapting on interim, incomplete data risks chasing noise and making erroneous decisions regarding efficacy or futility.”
Still, other studies are moving forward with the approach. For example, IMI-EPAD (Innovative Medicines Initiative–European Prevention of Alzheimer’s Dementia) will use Bayesian statistics to run its adaptive platform trial, which will compare multiple therapies head-to-head in Phase 2 trials, then select the winners for Phase 3 (Aug 2016 conference news).—Jessica Shugart
- Barcelona: Antibody to Sweep Up Aβ Protofibrils in Human Brain
- CTAD: Adaptive Antibody Trial to Try Bayesian Statistics
- Coming to a Center Near You: GAP and EPAD to Revamp Alzheimer’s Trials
- Mullard A. Multicompany trials adapt to disciplines beyond cancer. Nat Med. 2014 Jan;20(1):3. PubMed.
- Satlin A, Wang J, Logovinsky V, Berry S, Swanson C, Dhadda S, Berry DA. Design of a Bayesian adaptive phase 2 proof-of-concept trial for BAN2401, a putative disease-modifying monoclonal antibody for the treatment of Alzheimer's disease. Alzheimers Dement (N Y). 2016 Jan;2(1):1-12. Epub 2016 Feb 4 PubMed.
- Wang J, Logovinsky V, Hendrix SB, Stanworth SH, Perdomo C, Xu L, Dhadda S, Do I, Rabe M, Luthman J, Cummings J, Satlin A. ADCOMS: a composite clinical outcome for prodromal Alzheimer's disease trials. J Neurol Neurosurg Psychiatry. 2016 Sep;87(9):993-9. Epub 2016 Mar 23 PubMed.