Phase II (NAVIGATE-AD study) Results of LY3202626 Effects on Patients with Mild Alzheimer’s Disease Dementia

INTRODUCTION

Alzheimer’s disease (AD) is a progressive degenerative neurological disorder that results in the slow decline of cognition and function with a characteristic symptom of memory loss [1]. There is an unmet need for disease-modifying treatments in AD, as currently available therapies are symptomatic and do not affect the underlying disease pathology.

Patients with AD display severe brain atrophy with neurofibrillary tangles and amyloid plaques at autopsy [2]. The definitive etiology and cause of AD are still poorly understood; however, there is evidence supporting the ‘amyloid hypothesis’ that amyloid-β (Aβ) peptides aggregate to form amyloid plaques which act as an initial trigger of AD [3]. Aβ plaques have demonstrated neuronal toxicity and are hypothesized to lead to synapse loss, neurofibrillary tangle formation, and eventual neuronal cell death. The inhibition of Aβ formation is therefore a logical strategy towards developing a therapy for AD.

Aβ is part of the amyloid-β protein precursor (AβPP), which is a transmembrane protein widely expressed on the cell surface, particularly in neurons. AβPP has been found to be cleaved through two cleavage pathways involving three secretase enzymes: α-secretase, γ-secretase, and β-secretase (now called β-site APP-cleaving enzyme [BACE]1). Cleavage of AβPP by α-secretase precludes the formation of Aβ as the site is located within the Aβ sequence. In the second pathway, β-secretase cleaves the AβPP molecule, generating membrane-associated C99 and releasing a larger secreted fragment called secreted AβPP. Next, γ-secretase cleaves C99 in a heterogeneous fashion within the membrane releasing a variety of species that aggregate in protofibrils, and then fibrils, which seem to comprise the mass of Aβ plaques in AD brain tissue [4]. While both γ- and β-secretase inhibition represent effective means of precluding the formation of Aβ, BACE inhibition may provide improved safety and tolerability.

The accumulation of aggregated tau protein in the brains of patients with AD is also a characteristic pathology associated with the disease. The density and neuroanatomical localization of tau neurofibrillary tangles correlate strongly with neurologic symptoms and AD progression [5]. The recent development of the [¹⁸F]AV-1451 (flortaucipir) positron emission topography (PET) tracer allows for the ability to detect and measure tau protein in the brains of patients with suspected diagnosis of AD [6]. Use of this tracer shows increasing tau accumulation signal in healthy controls compared to mild cognitive impairment with progressive increases in patients with mild and moderate AD. The anatomical distribution seen on PET imaging corresponds well to the histopathological staging of Braak and Braak [7, 8]. Markers of tau pathology have also been shown to correlate more closely with changes in patient cognition compared to Aβ markers [9–11].

LY3202626 is a synthetic small molecule potent oral BACE1 inhibitor developed for the treatment of AD dementia. LY3202626 has been shown to reduce plasma and cerebrospinal fluid (CSF) Aβ_1–40 and Aβ_1–42 in mice, dogs, and humans. A Phase I study investigated the safety, tolerability, pharmacokinetics (PK), and pharmacodynamics (PD) of single- and multiple-ascending doses of LY3202626 given orally, in healthy subjects and patients with AD. In this study, single and multiple doses of LY3202626 were well tolerated and demonstrated a robust, prolonged reduction in plasma Aβ concentrations [12].

In the setting of a number of BACE inhibitors undergoing clinical development at the same time, a Phase II proof-of-concept clinical development approach was taken to estimate the extent to which LY3202626 impacted disease progression, and to better understand the mechanism of action of BACE inhibition on neurodegeneration biomarkers prior to initiating a Phase III program. The Phase II study (NAVIGATE-AD) aimed to assess whether suppression of Aβ production in the brain by LY3202626 inhibition of the BACE1 enzyme could slow the progression of AD tau progression as assessed by PET imaging and AD progression as assessed by clinical outcome measures. This Phase II study prioritized high levels of enzyme inhibition (expected 70–90% inhibition). Flortaucipir PET scans were chosen as the primary outcome endpoint for efficacy as a means to assess for cerebral tau neurofibrillary tangle load, a pathology known to correlate highly with cognition.

MATERIALS AND METHODS

RESULTS

The trial was terminated early following an interim data analysis concluding that there was a low likelihood of identifying a statistically significant effect of LY3202626 in slowing functional or cognitive decline. At the point of early termination, 1,149 patients had been screened, 316 patients randomized, and 47 patients (14.9%) had completed the study. The planned randomization ratio of 1:1:1 to the 3 mg/day, 12 mg/day, and placebo arms was also altered, with randomization to the 3 mg arm stopped after enrollment of 55 subjects in order to prioritize investigation of the higher LY3202626 dose in response to negative clinical efficacy data regarding other BACE inhibitors [28–30]. Patient disposition is summarized in Fig. 1.

Fig. 1

Summary of patient disposition.

At baseline, patient demographics were similar across the treatment arms (Table 1). The mean patient age was 72.9 years, with the majority of the treatment population being female (60.8%), white (82.9%), and from the United States (74.1%). One patient was randomized but did not receive study drug. A total of 269 patients (85.1%) discontinued from the study, with the most common reason for early discontinuation being the termination of the study by the sponsor (Fig. 1).

Analysis of efficacy endpoints

Flortaucipir PET measurements

As the study terminated early, the number of evaluable patients in each arm for the primary analysis was limited to 11, 15, and 15 patients in the placebo, 3 mg, and 12 mg arms, respectively.

Neither of the LY3202626 arms demonstrated a statistically significant separation from the placebo arm in change from baseline to 52 weeks for flortaucipir PET measurement. The least-squares (LS) mean changes were 0.02 for the 3 mg and 0.03 for the 12 mg arm in comparison to 0.00 for placebo. A similar lack of significant change was observed following the calculation of annualized change from baseline to completion or early termination of the study (Fig. 2).

Fig. 2

LS mean (SE) change from baseline and annualized change from baseline in flortaucipir SUVr. LS, least-squares; N, number of subjects in the analysis population; n, number of subjects who contributed data both at baseline and at specified visit; SE, standard error; SUVr, standardized uptake value ratio. Note: Annualized change was calculated by extrapolating observed change based on time between scans.

Measures of cognition, function, quality of life, and neuropsychiatric symptoms

No clinically meaningful separation was observed between the LY3202626 arms and placebo in the analysis of secondary efficacy measures of cognition, function, and composite measures of cognition and function (ADAS-Cog₁₃, ADCS-iADL, and iADRS, respectively) following 52 weeks of treatment (Fig. 3). Some spurious statistically significant differences were observed in iADL and iARDS scores following 52 weeks of treatment; however, these findings should be interpreted with caution due to the limited sample size at 52 weeks following early termination of the study. These secondary analyses were not adjusted for multiplicity. There was no clinically meaningful separation between the LY3202626 arms and placebo in the exploratory efficacy measures of cognition, function, subjective quality of life, and neuropsychiatric symptoms (MoCA and MMSE; FAQ and ECog; BASQID; and NPI, respectively) (Fig. 4) after 52 weeks of treatment. At the significance level of 0.2, there were few seemingly spurious statistically significant changes for both LY3202626 arms in ECog global score and NPI total score at Week 52; however, these results should be interpreted with caution due to the limited sample sizes.

Fig. 3

LS mean (SE) change from baseline on ADAS-Cog13, ADCS-iADL, and iADRS. ADAS-Cog13, Alzheimer Disease Assessment Scale–cognitive subscale; ADCS-iADL, Alzheimer’s Disease Cooperative Study–Instrumental Activities of Daily Living Inventory; iADRS, Integrated Alzheimer’s Disease Rating Scale; LS, least-squares; SE, standard error.

Fig. 4

Exploratory efficacy endpoint analyses. Results presented are change from baseline. MoCA, FAQ, MMSE, and ECog assessment were analyzed using a mixed models repeated measures model. Baseline is defined as last observation before or at randomization. BASQID, Bath Assessment of Subjective Quality of Life in Dementia; ECog, Measurement of Everyday Cognition; FAQ, Functional Activities Questionnaire; LS, least-squares; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; NPI; Neuropsychiatric Inventory; PBO, placebo; SE, standard error.

Exploratory outcomes: vMRI, cerebral perfusion, florbetapir, and NfL

Effect on brain volume. vMRI treatment comparisons of regional brain volume changes were conducted using the following three statistical approaches within the ANCOVA model (the first two were defined a priori in the protocol and the third was added as a consequence of early termination): 1) change from baseline to 52 weeks as outcome, with treatment, age, and baseline as covariates; 2) annualized change as outcome, with treatment, age, and baseline as covariates; and 3) change from baseline to last observation as outcome, with treatment, age, baseline, and between-time scan as covariates. The first analysis, which compared measurements at baseline and 52 weeks, had the fewest number of patients (placebo [8/58]; LY3202626 3 mg [15/38]; LY3202626 12 mg [12/60]). The third analysis uses time between baseline to last scan as a covariate in the model to account for a last scan that may not have occurred at 52 weeks due to early discontinuation.

Volumes of twelve brain regions were analyzed. In summary, using any of the three statistical approaches, treatment differences in regional brain volume change (bilateral sum) with greater reduction seen in any dose of LY3202626 to placebo were seen in the following regions (overall p for annualized change provided): entorhinal cortex (p = 0.044), hippocampus (p = 0.019), medial temporal lobe (p = 0.021), lateral (p = 0.034) and inferior parietal lobes (p = 0.032), and isthmus of the cingulate gyrus (p = 0.152). Using the annualized change analyses, no statistically significant volume difference for either LY3202626 dose to placebo in the following six brain regions was observed: whole brain, ventricles, whole temporal lobe, superior temporal lobe, prefrontal lobe, and precuneus (Fig. 5). Supplementary Table 1 presents all vMRI treatment comparisons of regional brain volume changes for each of the three analytical approaches described above.

Fig. 5

Annualized LS mean change from baseline as measured by vMRI. LS, least-squares; PBO, placebo; SE, standard error; vMRI, volumetric magnetic resonance imaging.

Cerebral perfusion. Due to early study termination, the number of evaluable patients for the florbetapir PET cerebral perfusion analysis who reached 52 weeks was limited to 10, 16, and 9 patients in the placebo, 3 mg, and 12 mg arms, respectively. No significant overall treatment effect (at 0.2 significance level) was observed using the perfusion SUVr methodologies applied. The LS mean changes were: –0.05 for the 3 mg arm, –0.03 for the 12 mg arm, in comparison to –0.03 for placebo. In the post-hoc annualized change analysis, there was reduced perfusion in the 12 mg group compared to placebo (p = 0.009).

Florbetapir. The number of evaluable patients for the florbetapir PET analysis was limited to 10, 16, and 10 patients in the placebo, 3 mg, and 12 mg arms, respectively. Neither of the LY3202626 doses had a statistically significant separation (at 0.2 significance level) from the placebo arm in change from baseline to 52 weeks for florbetapir PET measurements. The post-hoc annualized change analyses comparing each of the LY3202626 doses and placebo were not statistically significant.

Neurofilament light chain. NfL is a promising biomarker for measuring neurodegeneration. Plasma NfL levels were measured in 310 samples at baseline and 150 samples at post-baseline (111 samples at Week 36, and 39 samples at Week 52). Annualized changes were calculated by extrapolating last observed change based on time between data points. There was no statistically significant difference in the change of plasma NfL levels between placebo and the LY3202626 arms (Supplementary Table 2).

DISCUSSION

This Phase II study prioritized the testing of high BACE1 inhibition (70–90% inhibition of CSF Aβ, 3 mg and 12 mg of LY3202626 daily, respectively) over 52 weeks for the reduction disease progression in patients with mild AD dementia and confirmed amyloid pathology. This proof-of-concept study included a number of biomarkers aimed at understanding the effect of BACE inhibition on downstream neurodegenerative pathology and changes (e.g., flortaucipir, NfL, vMRI) and their relation to clinical outcomes of efficacy and safety. The study was stopped early, after an interim analysis was added, due to potential safety concerns emerging from the clinical trial results of other BACE inhibitors. The interim analysis was added to assess potential worsening of clinical outcomes as a consequence of treatment with a BACE inhibitor (as reported in other studies of BACE inhibitor compounds) and to evaluate futility. As a result of early termination, there were a limited number of patients who fully completed the study or even reached a later assessment visit. In examination of enrolled patients using prespecified and additional statistical analyses, treatment with BACE1 inhibitor LY3202626 did not slow disease progression (as assessed by flortaucipir PET scan) or reduce the clinically significant decline in cognition or function, as compared with placebo.

Another consideration in interpreting the negative results of this study is the appropriateness of the administered dose. As discussed previously, the study randomization was altered to prioritize investigation of the 12 mg daily dose following reports of negative clinical efficacy outcomes regarding another BACE inhibitor [29]. Treatment with the 3 mg dose of LY3202626 reduced the concentrations of Aβ_1–40 and Aβ_1–42 by 85.8% and 68.1% from baseline, respectively, which confirms that the drug had the intended PD effect of reducing the production of Aβ. Finally, the mild AD population enrolled may have been too far along in their disease process to respond to a BACE inhibitor treatment. A BACE inhibitor trial was terminated in the preclinical AD population due to findings of dose-related cognitive worsening and neuropsychiatric adverse events [31], though it has been hypothesized that a viable low dose BACE inhibition regimen could be identified in the future [32]. A number of other trials, such as the A4 study [33] or the AHEAD 3–45 Study (NCT04468659) are attempting to target the amyloid pathway with other mechanisms of action in preclinical AD.

In this study, administration of LY3202626 3 mg or 12 mg once daily for 52 weeks to patients with mild AD dementia and evidence of amyloid pathology was generally well tolerated. Despite substantial reductions in the plasma levels of circulating Aβ following the last treatment visit, no significant difference in clinical efficacy for cognition and function between LY3202626 and placebo were observed at either dose, which were seen in other Phase III studies testing BACE inhibitors [28–30, 34]. Moreover, no significant changes in amyloid deposition (as measured by florbetapir SUVr) or in cerebral tau neurofibrillary tangle load (as measured by flortaucipir SUVr) were observed between either treatment arm and placebo. Other markers for neurodegeneration showed mixed results, with no significant change in NfL between LY3202626 and placebo, but increased hippocampal volume loss for LY3202626 doses compared to placebo, findings which have also been observed in clinical trials with other BACE inhibitors [28, 30, 34]. In the vMRI analysis, a number of brain regions demonstrated statistically lower volumes in LY3202626-treated patients compared to placebo-treated patients for the respective regions of interest including the hippocampus. Most concerning would be the interpretation that volumetric changes on MRI reflected actual brain atrophy. However, MRI is a noninvasive imaging modality and, while volume analysis does include neuronal parenchyma, it also includes non-neuronal constituents (i.e., non-neuronal cells) and volume components (i.e., fluids), thus making these volume changes difficult to interpret. In this study, changes in cognition were not different between LY3202626-treated patients and those administered placebo and, therefore, it appears that MRI volume reductions were not correlated with cognitive worsening.

NfL is a biomarker for neurodegeneration and, although it was numerically higher in LY3202626 groups compared to placebo, the difference was not statistically significant. The post-hoc assessment of florbetapir PET cerebral perfusion showed a statistically significant reduction in perfusion for the 12 mg dose compared to placebo in the extrapolated annualized analysis, but no statistical significance for either dose in the LS mean change analysis for completers.

This study was limited by sample size due to early termination. In addition, without serial longitudinal follow-up MRIs or NfL measurements, we cannot definitively resolve a safety concern regarding volume changes. However, a recent study (including longitudinal data) has shown hippocampal volume reduction in umibecestat treated patients [35]. This study tested doses that resulted in mean CSF Aβ inhibition of 70–87%, similar to the LY3202626 doses tested in our study. The longitudinal analysis showed no progression of hippocampal volume loss between Week 26 and Week 52 scans. Furthermore, hippocampal volume reductions were not correlated with cognitive worsening, and volume reductions reversed in two months following discontinuation of umibecestat. Although this volume correction is perhaps reassuring, the mechanisms driving this reversal are not clear. Investigators theorized that contributors to the volume changes could include amyloid removal or fluid shifts. More careful monitoring is warranted in future BACE inhibitor studies.

Adverse events were more common following treatment with LY3202626 than with placebo, but no notable differences were observed between the 3 mg and 12 mg arms. Specifically, there were no significant differences between LY3202626 and placebo for weight loss or hair hypopigmentation as seen with other BACE inhibitors [28, 34]. With regard to non-clinical retinal concerns, there were no statistical differences for TEAEs of eye disorders or for retinal evaluations using fundoscopic and OCT assessments. A significant increase in AEs related to the psychiatric disorders system organ class were reported in both the 3 mg and 12 mg arms compared to placebo. Similar increases have been reported previously in trials of BACE inhibitors, such as verubecestat [28]. Furthermore, a recent study of atabecestat noted a greater number of AEs related to cognition, depression, sleep, dreams, and anxiety for patients receiving the BACE inhibitor compared to placebo [29]. Despite the fact that treatment with these compounds resulted in a greater magnitude and severity of psychiatric events (including suicidal ideation) and showed a clearer dose response relationship, these findings suggest that psychiatric reactions are adverse drug reactions associated with the BACE inhibitors class of molecules when administered at high levels of target engagement, as has been the case in several BACE inhibitors to date [28–30, 36].

The PK of LY3202626 were consistent with those described in healthy subjects, and a PK model derived from the observable plasma concentrations adequately described the overall shape of the plasma concentration-time profile and the overall dispersion of the data. High inter-subject and inter-visit variability in PK parameters were observed; however, a similar level of variability was observed in previous studies of healthy subjects, and does not seem to be unique to patients with AD. The range of LY3202626 inhibition was focused on a high degree of BACE inhibition, which was achieved using relatively low doses. The lower spectrum of BACE inhibition, such as in the range of amyloidogenic peptide reduction, 30–40%, similar to the Icelandic mutation [37], was not explored. While a lower degree of BACE inhibition might result in an improved safety profile, the benefit-safety profile of lower doses would have to be confirmed in a prospective study. For clinical trials that observed cognitive worsening, speculated mechanisms have focused on axonal organization defects caused by BACE1 over-inhibition and on the consequences of BACE inhibition in non-clinical studies, indicating negative effects on synaptic function and loss of dendritic spines in mice treated with BACE inhibitor MK-8931 [38, 39]. This biological effect on synapses and dendrites is reversible after cessation of BACE1 inhibitor treatment [38] and, while it has not been confirmed clinically, the effect appears to be dose-dependent in non-clinical studies, with lower doses of BACE inhibition having no effect on dendritic structure or spine plasticity [40]. Concordantly, recent studies have suggested that the negative effect of BACE inhibitors on cognition might be independent of the presence of brain amyloid and appears reversible after compound washout [41]. Some authors have suggested that lower BACE inhibition may have potential in future trials [39, 42, 43].

The dual-phase florbetapir PET may hold promise as a dual-biomarker imaging approach in the measurement of amyloid pathology (“late phases” started 50 minutes post-injection and regional cerebral perfusion; “early phases” started at the time of tracer administration). This approach could reduce patient radiation exposure and imaging sessions. To our knowledge, NAVIGATE-AD is the first multi-center interventional trial in patients with AD where dual-phase florbetapir acquisition was implemented. Further, longitudinal characteristics of perfusion imaging remain largely unknown. Our study demonstrated a significant (LS mean change of –0.03) decrease of perfusion for placebo participants. Among cortical Automated Anatomical Labeling atlas regions of interest, the largest relative decrease in perfusion was observed in temporal, lateral parietal, and cingulate regions. No significant association between baseline perfusion SUVr and amyloid burden was observed, suggesting an independence of these two measurements obtained using a dual-phase florbetapir scan. At the same time, a significant (p < 0.001) negative association between perfusion and composite neocortical tau SUVr was found. Significant associations between perfusion SUVr and ADAS-Cog₁₃ and iADL were also observed at baseline. These analyses suggest the potential utility of regional perfusion estimates derived from early-phase florbetapir scans as a tractable alternative to ¹⁸F-fluorodeoxyglucose measurements in AD trials [44]. Further exploration and validation would be required in clinical studies with more definitive positive or negative clinical outcomes to strengthen the method’s utility.

In conclusion, as a BACE1 inhibitor targeting 70–90% inhibition, LY3202626 was generally well-tolerated. A significant increase in AEs related to psychiatric disorders was observed following treatment, consistent with findings in studies with other BACE inhibitors. Despite high target engagement, no clinically significant change in the primary endpoint of tau burden or in the rate of functional or cognitive decline was observed following 52 weeks of dosing (or with annualized change analysis). An efficacy interim analysis led to the early termination of the study.