Phase 2 Safety Trial Targeting Amyloid β Production With a γ-Secretase Inhibitor in Alzheimer Disease

Objective  To evaluate the safety, tolerability, and amyloid β (Aβ) response to the γ-secretase inhibitor LY450139 in Alzheimer disease.

Design  Multicenter, randomized, double-blind, dose-escalation, placebo-controlled trial.

Setting  Community-based clinical research centers.

Patients  Fifty-one individuals with mild to moderate Alzheimer disease were randomized to receive placebo (n=15) or LY450139 (100 mg [n=22] or 140 mg [n=14]), with 43 completing the treatment phase.

Intervention  The LY450139 groups received 60 mg/d for 2 weeks, then 100 mg/d for 6 weeks, and then either 100 or 140 mg/d for 6 additional weeks.

Main Outcome Measures  Primary outcome measures were adverse events, plasma and cerebrospinal fluid Aβ levels, vital signs, electrocardiographic data, and laboratory safety test results. Secondary outcome measures included the Alzheimer's Disease Assessment Scale cognitive subscale and the Alzheimer's Disease Cooperative Study Activities of Daily Living Scale.

Results  Group differences were seen in skin and subcutaneous tissue concerns (P=.05), including 3 possible drug rashes and 3 reports of hair color change in the treatment groups. There were 3 adverse event‐related discontinuations, including 1 transient bowel obstruction. The plasma Aβ₄₀concentration was reduced by 58.2% for the 100-mg group and 64.6% for the 140-mg group (P<.001). No significant reduction was seen in cerebrospinal fluid Aβ levels. No group differences were seen in cognitive or functional measures.

Conclusions  LY450139 was generally well tolerated at doses of up to 140 mg/d for 14 weeks, with several findings indicating the need for close clinical monitoring in future studies. Decreases in plasma Aβ concentrations were consistent with inhibition of γ-secretase.

Trial Registration  clinicaltrials.gov Identifier: NCT00244322

LY450139 is a functional γ-secretase inhibitor that is currently under development as a disease-modifying therapy for Alzheimer disease (AD). As reported previously, LY450139 rapidly reduces amyloid β (Aβ) concentrations in the brain, cerebrospinal fluid (CSF), and plasma of transgenic V717F human amyloid precursor protein mice (PDAPP mice)^1,2and in the plasma of humans.³In addition, the administration of LY450139 at doses of 30 mg/kg once daily for 5 months to PDAPP mice results in reduced accumulation of Aβ in the hippocampus and cortex, as measured by means of enzyme-linked immunosorbent assay.⁴Previous clinical studies using LY450139 with either healthy volunteers⁵or patients with AD⁶have shown short-term reductions in plasma Aβ₄₀levels up to approximately 40% using single daily doses up to 50 mg. A single-dose escalation study³using 60, 100, or 140 mg in healthy volunteers demonstrated a dose-proportional increase in drug levels in plasma and CSF and a dose-dependent reduction in plasma Aβ levels.

The tolerability of LY450139 treatment in clinical studies of up to 50 mg for 2 weeks⁵or 40 mg for 6 weeks⁶was generally good. However, concern exists about cumulative toxic effects that may not be revealed in short low-dose trials or higher single-dose biomarker studies. A potential cause of clinical toxic effects for γ-secretase inhibitors is the inhibition of Notch cleavage by these compounds.⁷Notch cleavage is integral to cell differentiation pathways in many organ systems, including the gastrointestinal tract and lymphoid cell lines. To our knowledge, the effect of multiple-dose administration of 100 and 140 mg of LY450139 in patients with AD has not been reported previously.

In this study, we aim to demonstrate the safety and tolerability of LY450139 over 14 weeks using single daily doses of 100 and 140 mg. We sought to determine whether extended exposure to these higher doses of LY450139 would be tolerated and would result in expected changes in plasma and CSF Aβ levels.

Fifty-one patients were enrolled at 6 academic research centers between October 1, 2005, and December 31, 2006. The protocol was reviewed and approved by the institutional review board at each participating site. All research participants and caregivers gave written informed consent. The Alzheimer's Disease Cooperative Study's Data Safety Monitoring Board, which is advisory to the National Institute of Aging, provided oversight on an ongoing basis.

Participants were 50 years or older and diagnosed as having probable AD, as defined by the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer Disease and Related Disorders Association criteria.⁸Individuals receiving stable doses of cholinesterase inhibitor drugs or memantine were included. Patients were excluded if they had a history of irritable bowel syndrome, chronic diarrhea, peptic ulcer, or gastroesophageal reflux disease; a history of cardiac disease; significant electrocardiographic (ECG) abnormalities; hematologic disorders; hepatic or renal disease; active malignancy within 5 years; or clinically important depressive, neuropsychiatric, cerebrovascular, or respiratory disease.

This is a multicenter, randomized, double-blind, placebo-controlled, dose-escalation trial (Figure 1). Participants were randomized to receive LY450139 or placebo using a 2:1 randomization scheme via a telephone-based interactive voice-response system. Patients randomized to the LY450139 groups received 60 mg/d for 2 weeks, then 100 mg/d for the next 6 weeks. At 8 weeks, the treatment arm was randomized again to receive 6 additional weeks of treatment at either 100 or 140 mg/d. Dose reductions were allowed for dose-limiting adverse events (AEs) during the treatment phase.

All AEs occurring during the study were characterized. Safety data, such as vital signs, ECG data, and laboratory test results, were collected throughout the trial. Vital signs were recorded at each study visit, with ECGs performed at baseline and at weeks 2, 4, 8, 10, 14, and 16.

Routine hematology and clinical chemistry laboratory tests were performed at screening; at weeks 2, 4, 8, 10, and 14 (treatment); and at a 26-week follow-up visit. Measurements of CD4, CD8, CD19, IgG, IgA, and IgM cells were obtained at baseline, treatment end, and the 26-week follow-up visit. T-cell function was assessed by means of flow cytometry.³Urinalysis was tested at screening and at weeks 8 and 14. Stool samples were tested for occult blood at each study visit after baseline. Optional DNA testing for apolipoprotein E subtype was performed.

Plasma samples for measurement of biomarkers and drug levels were collected at baseline and at weeks 14 and 16. At week 14, multiple samples were collected in 6 hours for pharmacodynamic modeling (30 minutes before and 0.5, 1, 2, 4, and 6 hours after receiving the drug). The CSF samples were collected into polypropylene tubes by means of a lumbar puncture at baseline and approximately 6 hours after the administration of the last dose of LY450139 at week 14.³Plasma and CSF concentrations of Aβ were measured from all the samples,³and concentrations of LY450139 were measured from all the samples after baseline.⁵

At baseline and at weeks 8 and 14, assessments were made of cognition and daily functioning using the Alzheimer's Disease Assessment Scale cognitive subscale (ADAS-Cog 11)⁹and the Alzheimer's Disease Cooperative Study Activities of Daily Living Scale (ADCS-ADL).¹⁰

Adverse event data are reported using intention-to-treat analyses (n=51) across 26 weeks, with a completers analysis for all comparisons of laboratory and clinical measures through week 14 (n=43). To maximize the sensitivity of the safety measures, no adjustments were made for multiple comparisons. All statistical tests were 2-sided. P<.05 was considered statistically significant for biomarker and clinical measures, and P<.10 was used for safety variables to indicate potential clinical significance. All analyses were performed using the R statistical package, version 2.3.1.¹¹

Data from all randomized participants for 26 weeks were included in the safety analyses. The Fisher exact test was performed to assess group differences in proportions of participants reporting specific types of AEs and serious AEs.

Analysis of covariance (ANCOVA) was used to assess 14-week change scores in all efficacy and biomarker outcomes of interest, with treatment group as a factor. Significance was based on the overall Fstatistic for interactions among all 3 treatment arms. Baseline values of outcome measures were entered as covariates into each model to account for any baseline group differences. In addition, any variable found to have significant between-group differences at baseline and to correlate with the outcome measure being tested was entered as a confounder into a post hoc ANCOVA model. Outcome variables included vital signs, laboratory measures, Fridericia-corrected QT intervals, log mean CSF Aβ₄₀and Aβ₄₂measures, and ADAS-Cog 11 and ADCS-ADL scores. A 2-way ANCOVA was used to investigate whether the treatment effect on the change in ADAS-Cog 11 scores was modified by donepezil hydrochloride therapy. Post hoc Bayesian estimates were used to determine pharmacokinetic variables.⁵At week 14, mean change in LY450139 and absolute mean plasma Aβ levels were compared for each time point over 6 hours using Kruskal-Wallis nonparametric tests of means.

Of 71 participants who were screened for the trial, 51 were eligible and randomized. Forty-three participants completed the treatment phase of the trial (Figure 2). No differences existed in dropout rates (84%) or medication compliance (96%-99%) among the 3 groups.

The 3 groups were balanced at baseline regarding all demographics, clinical measures (except Mini-Mental State Examination [MMSE] scores), apolipoprotein E ε4 distribution, and acetylcholine esterase inhibitor and memantine use (Table 1). Differences in baseline laboratory values are given in Table 1.

All AEs that occurred in 2 or more participants are given in Table 2. Only the category of skin and subcutaneous tissue disorders reached significance for differences between groups (P=.05). In this category, there were 3 possible “drug-related” rashes and 3 reports of hair color change (lightening) in the treatment groups, with none in the placebo group. Although not reaching statistical significance, the percentage of patients reporting nausea, vomiting, or diarrhea was 27% in the treatment groups and only 13% in the placebo group (P=.41). In addition, when combining reports of somnolence, fatigue, lethargy, and asthenia from different organ classes, we found that 40% of treatment subjects noted 1 or more of these symptoms, whereas only 13% in the placebo group had these symptoms (P=.18). There were 4 serious AEs (Table 3). A small-bowel obstruction was considered to be possibly related to the study drug; no medical intervention was required, the study drug was discontinued, and the obstruction resolved spontaneously.

Eight participants dropped out of the trial during the treatment phase because of AEs (n=3), personal conflict (n=3), physician decision due to difficulty in managing increasing agitation and paranoia (n=1), and a protocol violation consisting of a patient reporting a history of gastroesophageal reflux disease after randomization (n=1). The AEs included diarrhea, heme-positive stool (required dropout per protocol), and the previously described bowel obstruction. There were 3 site-initiated dose reductions for rash, abdominal discomfort, and nausea.

Between-group differences were seen in some safety laboratory value changes during the therapy phase (Table 4). None of these changes were considered clinically important based on the degree of change and the lack of associated adverse events. No statistically significant group differences existed in mean change in the ECG Fridericia-corrected QT interval over 14 weeks. The 140-mg group showed the greatest numeric change, with a prolongation of 19.3 milliseconds (4.80%) compared with an increase of 2.8 milliseconds (0.69%) in the placebo group (P=.18).

At week 14, a rapid increase in plasma drug levels was seen after drug administration (Figure 3). Peak plasma levels occurred at 1.26 hours (t_½=2.23 hours) for the 100-mg group and at 1.68 hours (t_½=2.59 hours) for the 140-mg group. A CSF sample could not be obtained for 1 participant in the 140-mg group. Six hours after drug administration at week 14, the 100- and 140-mg groups had mean (SD) CSF drug concentrations of 77.4 (34.3) ng/mL and 102.1 (30.3) ng/mL, respectively (Wilcoxon rank sum P=.02).

At week 14, just before drug administration, absolute mean plasma Aβ₄₀levels were increased by a mean (SD) of 32.0% (34.6%) in the 100-mg group and 35.2% (60.6%) in the 140-mg group compared with study baseline (Kruskal-Wallis P=.009) (Figure 4). No statistically significant elevations were seen in Aβ₄₂levels.

After drug administration at week 14, mean plasma Aβ₄₀levels declined rapidly in both treatment groups (Figure 4). Concentrations decreased significantly below baseline by hour 2 for both treatment groups (P=.02), and they continued to decline up to 6 hours after administration (P<.001). The maximum reduction in Aβ₄₀concentration was 58.2% for the 100-mg group and 64.6% for the 140-mg group (Figure 4). No statistically significant difference in Aβ₄₀reduction existed between the 100- and 140-mg groups. During this 6-hour period, 14 of 36 treated participants had 1 or more plasma Aβ₄₂measures decline below the lower limit of quantification measured by means of enzyme-linked immunosorbent assay (28.0 pg/mL), making these data statistically noninterpretable.

No significant reductions were seen in mean log CSF Aβ₄₀(F=2.025, P=.15) or Aβ₄₂(F=0.73, P=.49) levels in the full ANCOVA model. Percentage changes from baseline in absolute means are presented in Figure 5. Spearman rank correlations showed a trend association between CSF drug levels and percentage change in CSF Aβ concentration in the 140-mg group (Aβ₄₀: r=–0.51, P=.09; Aβ₄₂: r=–0.51, P=.09) but not in the 100-mg group (Aβ₄₀: r=–0.35, P=.17; Aβ₄₂: r=–0.20, P=.44). Change in absolute CSF Aβ levels showed trend associations with drug levels for the 140-mg group (Aβ₄₀: r=–0.43, P=.17; Aβ₄₂: r=–0.55, P=.07) and the 100-mg group (Aβ₄₀: r=–0.47, P=.06; Aβ₄₂: r=–0.21, P=.42).

No significant differences were seen in ADAS-Cog 11 (P=.36) and ADCS-ADL (P=.63) scores after 14 weeks of treatment among any of the 3 groups (Table 5). This lack of treatment effect was not modified by the inclusion of donepezil use in the ANCOVA model.

The MMSE score was the only variable found to be a confounding measure. The 3 treatment groups differed at baseline on MMSE scores, with a possible relationship found between baseline MMSE scores and change in CSF Aβ₄₀concentrations (Spearman ρ=0.30, P=.06). An ANCOVA model including MMSE score as a covariate revealed a trend toward significant reductions in CSF Aβ₄₀levels at week 14 compared with study baseline in the treatment groups compared with the placebo group (F=2.90, P=.07) (Figure 5).

This longitudinal study in elderly patients with AD was necessary for assessing safety and possible Notch-related toxic effects, which would not likely be revealed by single-dose studies.¹²Notch toxicity encompasses many of the drug-related tolerability concerns for γ-secretase inhibitors.¹³Notch is a transmembrane protein that plays a role in nuclear signaling, and, similar to the amyloid precursor protein, it seems to be cleaved by a presenilin-dependent γ-secretase complex.⁷It plays an important role in programmed cellular death. Organ systems with rapid cellular turnover have, therefore, been the primary concern for Notch-related toxic effects. Both gastrointestinal and immune cell functions have been altered in preclinical studies with LY450139 (data on file, Eli Lilly & Co). To our knowledge, no previous human study of LY450139 has demonstrated clinically significant toxic effects on the immune system.^3,5,6Diffuse macular rash on the extremities and torso and hair color changes in some participants were likely the result of treatment with LY450139. However, no evidence existed of other toxic effects associated with these. The rash and the hypopigmentation were reversible. Gastrointestinal symptoms, somnolence and asthenia, and ECG changes were not statistically significantly different between groups, yet they should still be considered potentially important drug-related AEs given the known mechanism of action of this drug and the limited statistical power of this study.

As shown in Figure 4, the plasma level of Aβ₄₀is higher than the study baseline level before drug administration at visit 14. In a previous single-dose study,³a biphasic response of plasma Aβ to LY450139 was observed at 60, 100, and 140 mg, with an initial reduction in plasma Aβ concentration followed by an elevation higher than baseline levels 8 to 10 hours after administration. Doses of 100 mg or greater prolonged plasma reductions and reduced plasma elevations over 24 hours compared with 60 mg.³Similar patterns of peripheral changes in Aβ levels have been seen in preclinical studies¹⁴of guinea pigs with LY450139 and in studies of other γ-secretase inhibitors.¹⁵However, similar increases in CSF and brain Aβ levels have not been demonstrated in previous preclinical studies of LY450139 at multiple intervals after drug administration up to 24 hours (data on file). In addition, studies using very low doses of LY450139 showed increased plasma Aβ levels without a period of reduction.^5,14Thus, 1 possible explanation for a transient increase in plasma Aβ concentration is that in peripheral, but not in central, tissue(s) γ-secretase inhibitors have a stimulatory effect on the enzyme at low concentrations that is overcome by the inhibitory effects at higher concentrations. Although steady-state Aβ reductions would be desirable, twice-daily dosing is not possible owing to observed Notch-related toxic effects in multiple organ systems in preclinical studies of Fischer 344 rats and dogs (data on file, Eli Lilly & Co). The clinical implications of this are unclear.

The expected clear changes in CSF Aβ levels were not demonstrated in this study despite robust reductions in plasma Aβ₄₀and dose-related LY450139 concentrations in the CSF. Correlation analyses between CSF drug levels and Aβ levels suggest a pharmacodynamic response but perhaps lack statistical power. In preclinical studies, CSF Aβ-lowering effects have been seen in PDAPP mice¹and dogs (data on file, Eli Lilly & Co). Lack of CSF changes in the setting of clear serum changes may reflect rapid transport of Aβ from CSF into plasma and a lengthened period for Aβ to reach equilibrium in the CSF. In a trial¹⁵of a similar γ-secretase inhibitor, changes in CSF Aβ concentration lagged behind changes in plasma Aβ concentration, with significant decreases seen at 12 hours. One study¹⁶demonstrated that carbon 13‐labeled Aβ levels did not reach equilibrium in lumbar CSF until approximately 13 hours after the beginning of the infusion. In the present study, Aβ levels were measured approximately 6 hours after morning drug administration. Longer periods may be required to identify CSF changes. Furthermore, transgenic PDAPP mice,¹wild-type mice,¹and guinea pigs¹⁴show a clear association between the reduction in plasma and brain Aβ levels after the administration of LY450139. Whether CSF changes in Aβ concentration can be detected in people with AD after oral administration of LY450139 requires additional studies.

The long-term efficacy of this drug is not known. Given the slow rate of clinical progression in AD, we did not expect to see drug effects on measures of cognition or ADLs in this 14-week trial. Without this, a full risk-benefit assessment cannot be made. This trial sufficiently demonstrates that LY450139 can be tolerated, although not without risk. Given the potential for disease-modifying effects of this Aβ-lowering agent, and the arguably acceptable tolerance and safety profile of LY450139 demonstrated in this study, further large-scale efficacy trials are justified. Based, in part, on the results of this phase 2 study, Eli Lilly & Co launched a multinational phase 3 trial in the second quarter of 2008, with an enrollment goal of 1500 patients with AD.

Correspondence:Adam S. Fleisher, MD, University of California, San Diego, 8950 Villa La Jolla Dr, Ste C227, La Jolla, CA 92037 (afleisher@ucsd.edu).

Accepted for Publication: April 11, 2008.

Author Contributions:All authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Fleisher, Siemers, Dean, and Thal. Acquisition of data: Fleisher, Siemers, Clark, Dean, Farlow, Galvin, Peskind, Quinn, Sherzai, and Thal. Analysis and interpretation of data: Fleisher, Raman, Siemers, Becerra, Dean, Farlow, Sowell, Aisen, and Thal. Drafting of the manuscript: Fleisher, Raman, Siemers, Dean, Sowell, and Thal. Critical revision of the manuscript for important intellectual content: Fleisher, Raman, Siemers, Becerra, Clark, Dean, Farlow, Galvin, Peskind, Quinn, Sherzai, Sowell, and Aisen. Statistical analysis: Raman, Siemers, Becerra, and Sowell. Obtained funding: Thal. Administrative, technical, and material support: Fleisher, Siemers, Dean, Farlow, Sherzai, Aisen, and Thal. Study supervision: Fleisher, Raman, Siemers, Clark, Farlow, P eskind, Aisen, and Thal.

Financial Disclosure:Drs Siemers and Dean are employees of Eli Lilly & Co, the manufacturer of the study drug. All the study sites were subcontracted through the ADCS at the University of California, San Diego, which held the primary contract with Eli Lilly & Co. Dr Farlow has received research grant funding support from Eli Lilly & Co that was independent of and unrelated to this trial.

Funding/Support:This study was fully funded by Eli Lilly & Co. The ADCS infrastructure is supported by cooperative grant UO1AG10483 from the National Institute on Aging.

Role of the Sponsor:The study design was developed by Eli Lilly & Co in consultation with the ADCS. The study was conducted by the ADCS, with data collection, management, analysis, and interpretation by the ADCS. Manuscript preparation was completed by the ADCS in consultation with Eli Lilly & Co.

Additional Information:The ADCS had access to all the data, with full, unrestricted publication rights.