What lessons can be learned from failed Alzheimer’s disease trials?

Abstract

Trials missing primary efficacy end points raise the question of whether the choice of drug or the limitations of disease biology were at fault. In some trials, drugs appear not to have achieved biochemical effect thresholds sufficient for clinical benefit. This suggests the need for improved drugs that are more active at tolerated doses. In other trials, it is unclear how the observed biomarker changes are related to potential efficacy. However, hints of efficacy from exploratory analyses support the idea that starting treatment earlier in the course of the disease might be more effective. A closer look at the failed trials will help de-risk future trials.

Keywords:

• aducanumab
• Alzheimer
• amyloid
• avagacestat
• bapineuzumab
• clinical
• secretase
• semagacestat
• solanezumab
• trial

Many Alzheimer’s disease (AD) trials have been unsuccessful in recent years, with no approvals of new medicines for this catastrophic and currently untreatable disease in over a decade. Marginal benefits can be obtained by treatment of symptoms, but there are no drugs proven to slow progression or prevent the disease. Multiple trials of small molecule and antibody drugs have been inspired by the amyloid hypothesis, so the lack of success makes it natural to ask if the amyloid hypothesis itself has failed. However, a closer look at the amyloid-based trials reveals technical limitations rather than a fundamentally misguided approach. Furthermore, many trials targeting non-amyloid mechanisms have also been unsuccessful. For meaningful evaluation of trials that did not benefit patients, a key question is whether or not appropriate biochemical effects of sufficient magnitude and duration were demonstrated.

Compelling evidence for the amyloid hypothesis stems from human genetics, which implicate a central role for the amyloid-β peptide (Aβ), particularly the form containing 42 amino acids, Aβ 42. Aβ 42 readily self-aggregates, thereby initiating the formation of amyloid plaques and a variety of neurotoxic Aβ oligomers. It is nevertheless unclear exactly what aspects of Aβ metabolism are critical, particularly because Aβ accumulation and plaque deposition start decades before the onset of dementia. Drugs, depending on their mechanisms of action, can affect Aβ metabolism in different ways, such as soluble Aβ lowering, Aβ sequestration, or amyloid plaque clearance. In principle, a rigorous trial would explore up to a maximum level of biochemical effect, although evidence suggests that 25% lowering of Aβ might be sufficient for mechanisms specifically affecting Aβ production [1].

Aβ is a peptide fragment derived from the amyloid precursor protein by two proteases; β-site amyloid precursor protein cleaving enzyme (BACE) and γ-secretase. Small molecule inhibitors and modulators of these enzymes directly affect Aβ biosynthesis, and so brain Aβ production can be monitored via CSF Aβ levels. Trials with γ-secretase inhibitors (GSI) and BACE inhibitors have demonstrated excellent translation of mechanism and Aβ lowering between preclinical species and humans [2,3]. Unfortunately, the highest tolerated doses for chronic treatment in patients have limited the extent of Aβ lowering that could be maintained. Thus, the GSIs semagacestat and avagacestat had minimal effects on CSF Aβ levels. For semagacestat, despite evidence of transient inhibition of γ-secretase [4], there was no significant decrease in steady state levels of CSF Aβ [5], and for avagacestat, it was ≤15% lowering at the tolerated dose [2]. The γ-secretase modulator, tarenflurbil, which selectively affects Aβ 42, likewise showed no significant CSF Aβ 42 lowering [6]. Thus, the scant biochemical effect on Aβ production is sufficient to explain the lack of benefit to patients in these trials. In ongoing trials, the BACE inhibitor MK-8931 may be on track to change this situation. It has so far escaped the off-target toxicities that led to withdrawal of prior BACE inhibitors, while chronically lowering CSF Aβ by nearly 80% at the top dose in patients [7].

Perhaps the most remarkable outcome of the GSI trials was the increased rate of cognitive decline, which is indeed the opposite of the intended benefit of therapy. While there is still no clear explanation for this finding, it has been proposed that the GSI mechanism itself could have contributed [8,9]. GSIs can increase Aβ production at low doses, and more so for Aβ 42 than for Aβ 40. Consistent with this, semagacestat increased the CSF Aβ 42/Aβ 40 ratio [5], although the same was not observed for avagacestat [2]. Ironically, such explanations tend to support the amyloid hypothesis. Additional explanations include intermittent inhibition of Notch or other γ-secretase substrates in the brain [9], or accumulation of the toxic processing intermediate upstream of Aβ, ‘β-CTF’ [10]. While the mechanism is unresolved, the GSI class is perceived to bear a cognitive liability in addition to dose-limiting peripheral side effects. Like γ-secretase, BACE also processes substrates known to have neuronal functions, although no evidence has yet emerged for cognitive worsening that might obscure the theoretical benefits of Aβ lowering. The known limitations make new AD trials with GSIs very unlikely, so future γ-secretase-targeted compounds would have to display a radically different mechanism. One possibility might be improved γ-secretase modulators, which selectively lower Aβ 42 while circumventing Notch inhibition and the undesirable effects of GSIs on Aβ metabolism. Such γ-secretase modulators, however, have not yet advanced beyond early phase trials in healthy subjects [11].

There have been trials of several compounds that interfere with Aβ metabolism by other mechanisms. PBT2, a Cu and Zn ionophore that affects Aβ metabolism, showed significant CSF Aβ 42 lowering. Remarkably, there was evidence of cognitive benefit after only 12 weeks of dosing, but it was not correlated with CSF Aβ 42 levels in individual patients [12], suggesting that CSF Aβ 42 lowering was not directly relevant to cognitive benefit in this case. A second small trial with PBT2 reported no significant effects on cognition or on brain imaging biomarkers [13]. Scyllo-inositol, a compound that binds to Aβ, showed significant CSF Aβ 42 lowering but missed primary efficacy end points in a Phase II trial. Exploratory analysis, however, suggested a potential benefit in a prespecified mild AD subgroup [14]. Tramiprosate, another Aβ binding compound, showed significant CSF Aβ 42 lowering in a small Phase II trial [15], and a Phase III trial reported cognitive benefits in a post hoc analysis [16], but missed its primary cognitive end points [17]. For PBT2, scyllo-inositol, and tramiprosate, CSF Aβ 42 lowering can be interpreted as evidence of pharmacological action in the brain, although it is unclear if this biomarker represents the mechanism of potential cognitive benefits for these drugs. It is therefore uncertain what contributed most to the inconclusive cognitive outcomes; whether it was an insufficient level of biochemical effect and duration, or an unsuitable drug mechanism to impact the disease. Perhaps improved formulations or higher doses of these types of drugs might be worth exploring if increased drug exposures can be tolerated, and if increased biochemical effect can be demonstrated.

Active and passive anti-Aβ immunization trials have demonstrated that anti-Aβ antibodies can enter and act in the brain. Despite not meeting their primary efficacy end points, these trials were informative in many ways, including potential efficacy, tolerability, biomarkers, and side effects, such as amyloid-related imaging abnormalities. The large efficacy trials of two antibodies, in particular, bapineuzumab [18] and solanezumab [19], have been thoroughly reviewed [20]. Another anti-Aβ antibody, gantenerumab, was recently withdrawn from a Phase III trial in prodromal AD, but evaluation will have to await the disclosure of results. Despite the negative primary outcomes for bapineuzumab and solanezumab, exploratory data analysis showed hints of cognitive benefit in mild AD patients, who represent the earliest stage of AD tested in these trials. For bapineuzumab, use of an alternative baseline threshold for the mini-mental state examination suggested a benefit, and for solanezumab a combined analysis of two Phase III trials indicated an improved Alzheimer’s disease assessment scale-cognitive subscale in the mild AD subgroup. Additional anti-Aβ antibodies in early phase studies showing evidence of efficacy at the earliest stages of AD are aducanumab and crenezumab. This supports the idea that treatment could potentially be more effective if started earlier in the course of the disease, where it might work to delay age-of-onset for cognitive symptoms in at-risk individuals. Most remarkable about the exploratory biomarkers was their divergence between antibodies. In ApoE4 carriers, bapineuzumab decreased the rate of amyloid plaque accumulation as measured by PET imaging, decreased the neurodegeneration biomarkers CSF tau and phospho-tau, but had no effect on CSF Aβ levels. In contrast, solanezumab did not affect plaque or phospho-tau levels, but had pronounced effects on CSF Aβ, in particular, causing significantly increased levels of antibody-bound Aβ. While the biomarker changes were interpreted as evidence of pharmacological action in the brain, it remains to be determined how they relate to the magnitude of biochemical effect or to potential cognitive benefits. In contrast, baseline biomarkers for inclusion criteria will clearly have value in future trials, because 36% of ApoE4 non-carriers in the bapineuzumab trial had negative PET scans for amyloid. Recently, an early phase study of aducanumab, using stringent inclusion criteria for early AD, was reported to significantly decrease both cognitive decline and plaque accumulation, which bodes well for Phase III trials.

In conclusion, the trials of small molecule inhibitors and modulators of Aβ biosynthesis have so far not demonstrated adequate biochemical effects at tolerated doses in patients. So a meaningful test of the amyloid hypothesis through effects on Aβ production will require improved drugs. The small molecules and antibodies that sequester or otherwise interfere with Aβ metabolism have shown some promise, pointing the way to improved or modified trials, but despite a wealth of information the quest for informative biomarkers remains a work in progress. In future, the most informative trials will continue to be those demonstrating relevant mechanism and adequate biochemical effects regardless of the therapeutic approach.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.