Clinical effects of Aβ immunization (AN1792) in patients with AD in an interrupted trial

Patients.

Eligible patients were 50 to 85 years of age, met the criteria for a diagnosis of probable AD as defined by the National Institute of Neurologic and Communicative Disorders and Stroke–AD and Related Disorders Association,⁹ and had an MRI brain scan supporting the clinical diagnosis of AD. Additional inclusion criteria were a score of 15 to 26 on the Mini-Mental State Examination (MMSE¹⁰), a Rosen-Modified Hachinski Ischemic score of ≤4, and written, informed consent from the patient and the patient’s caregiver for the original protocol and subsequent amendments. Patients were excluded if they had clinically significant neurologic disease, other than AD, that might affect cognition; a major psychiatric disorder, systemic illness, or symptoms that could affect the patient’s ability to complete the study; a Hamilton Psychiatric Rating Scale for Depression score of >12; used anticonvulsant, antiparkinsonian, anticoagulant, narcotic, or immunosuppressive medications within 3 months prior to baseline; used medication with the potential to affect cognition (unless maintained on a stable low to moderate dose regimen for at least 3 months prior to baseline); or used medications for cognitive enhancement other than a stable dosing regimen of an acetylcholinesterase inhibitor (≥6 months).

Study design and treatment.

This randomized, placebo-controlled, double-blind, phase IIa clinical trial was conducted at 28 centers in the United States and Europe between September 2001 and December 2002. A total of 372 patients with mild to moderate AD were randomly assigned in a double-blind manner to receive treatment with a suspension of AN1792 225 μg (Elan Pharmaceuticals, Inc., South San Francisco, CA) and QS-21 50 μg (Antigenics, Framingham, MA) containing 0.4% polysorbate-80, or normal saline (placebo) in a 4:1 ratio. Randomization was performed by an independent statistician using a computerized, random-number generator and treatment was assigned by a central computer. Randomization was stratified by acetylcholinesterase inhibitor use (yes or no) and baseline MMSE score (15 to 20 or 21 to 26).

In each group, treatment was administered as a single 0.5 mL IM injection into the deltoid muscle and, according to the original protocol, dosing was planned to occur on day 0 and at months 1, 3, 6, 9, and 12 during the 15-month study. Patients in the active treatment group who had not developed a predefined serum anti-AN1792 antibody titer (IgG total tier > 1:2,200) were to be discontinued at month 8, after administration of four injections. Due to the premature discontinuation of the immunization, patients received only one to three injections. All patients enrolled in the study (including those who had previously discontinued early) were invited to participate in the safety follow-up period, which continued blinded and lasted for at least 9 months after their last dose of study treatment. Assessments were performed at weeks 1, 2, and 4, and thereafter at monthly intervals (with additional visits at months 3.5 and 6.5) until month 12 or early termination.

Site-specific local independent ethics committees approved the original protocol, amendments, and related informed consent forms prior to implementation. The study was conducted in accordance with the International Conference on Harmonisation Tripartite Guideline on Good Clinical Practice and in compliance with the Declaration of Helsinki 1964 as modified in October 2000.¹¹ An eight-member independent Data Safety Monitoring Committee assessed the safety of the study drug throughout the trial.

Outcome measures.

The protocol was amended following the occurrence of meningoencephalitis such that cognitive outcomes were reduced to secondary measures, as safety and tolerability became the primary outcome measures of the study. The primary outcome measures were monitored throughout the trial by adverse event (AE) reporting, physical and neurologic examinations, injection site reactions, vital signs, and laboratory evaluations (biochemical and hematologic tests and urinalyses). Physical and neurologic assessments, an ECG, and an MRI brain scan were performed at screening and the final visit. Vital signs were evaluated at each visit.

An ELISA was used to determine anti-AN1792 immunoglobulin (Ig) G (total) and IgM levels in serum and CSF samples. The ELISA had a lower limit of detection of 1:25 for CSF IgG, CSF IgM, and serum IgM, and a lower limit of detection of 1:50 for serum IgG. Serum was collected at approximately monthly intervals, and CSF was collected in a subset of patients at baseline and month 12. Antibody responders were predefined as serum anti-AN1792 IgG (total) titer ≥1:2,200 at any time after injection 1. This titer was selected as the minimum anti-Aβ titer predicted to be of clinical benefit, based on preclinical data (Elan Pharmaceuticals, Inc., data on file).

Predefined secondary evaluations included the following cognitive and functional tests at baseline, month 6, and month 12: MMSE, score range 0 to 30;¹⁰ ADAS–Cog, score range 0 to 70;¹² AD Cooperative Study–Clinical Global Impression of Change (ADCS–CGIC), seven-point scale;¹³ Disability Assessment for Dementia (DAD), score range 0 to 100%;¹⁴ Clinical Dementia Rating (CDR) scale,¹⁵ score range 0 to 3; and a Neuropsychological Test Battery (NTB). The NTB consisted of the following nine components: Wechsler Memory Visual–Immediate (WMVis-I, score range 0 to 18); Wechsler Memory Verbal–Immediate (WMVer-I, 0 to 24); Rey Auditory Verbal Learning–Immediate (RAVL-I, 0 to 105); Wechsler Memory–Digit Span (WMDS, 0 to 24); Controlled Word Association Test (COWAT); Category Naming Test (CNT); Wechsler Memory Visual–Delayed (WMVis-D, 0 to 6); Wechsler Memory Verbal–Delayed (WMVer-D, 0 to 8); and Rey Auditory Verbal Learning–Delayed (RAVL-D, 0 to 30). The cognitive scales were administered by two blinded raters: the first administered the ADAS–Cog, DAD, and NTB; the second administered the ADCS–CGIC and CDR. In addition to assessments of individual component scores, raw scores on each of the nine NTB tests were converted to z-scores using the sample baseline mean and SD for each test. The resultant z-scores were averaged to obtain a composite z-score including all nine NTB tests. The NTB was further explored by subgrouping the overall composite NTB z-score into immediate memory (WMVis-I, WMVer-I, RAVL-I), delayed memory (WMVis-D, WMVer-D, RAVL-D), executive function (WMDS, COWAT, CNT), and all memory (all six memory tests) composite z-scores. Change from baseline was calculated as the post-baseline composite z-score minus the baseline score, such that a positive change indicates an improvement from baseline. Each of the five composite NTB measures was analyzed using analysis of covariance with change from baseline z-score as the response and baseline value as the covariate, with the same adjustments as in the principal analysis. In addition to the cognitive tests, ELISA determinations of CSF tau and Aβ42 protein,¹⁶ and brain, hippocampal, and ventricular volume, as determined by MRI (reported elsewhere [Fox et al. 2004, submitted]), were analyzed.

Statistical analysis.

Sample sizes and power calculations were generated according to the original primary endpoints of the study (whole brain volume determined by MRI and cognitive change as determined by ADAS–Cog) using a two-sided test at the 5% level of significance. Based on MRI data,¹⁷ a sample size of 75 patients per group would provide 78% power to detect a 30% change in rate of reduction of whole brain volume (placebo vs active). Additionally, a sample size of 75 patients per group would provide 63% power based on an ADAS–Cog three-point difference (placebo vs active) in annualized mean change and an SD of eight points. Both estimates assume that analysis would compare the placebo group to the “antibody responder” patients from the active group. Based on data from a previous phase I study in elderly AD patients treated with AN1792(QS-21),⁷ it was estimated that approximately 25% of patients would develop serum anti-AN1792 titers of ≥1:2,200 (defined as antibody responders). Accordingly, in order to achieve approximately 75 antibody responders, 300 patients were planned for enrollment into the active group and 75 patients for the placebo group.

AEs were evaluated in all patients who were randomized and received at least one injection of study drug (safety population). Geometric mean serum anti-AN1792 titers (and 95% CI) were calculated for all visits at which titers were assessed for the antibody responder and placebo groups.

As it was anticipated that only 25% of active patients would become antibody responders, the prespecified efficacy analyses were conducted on the efficacy-evaluable population, rather than on an intent-to-treat (ITT) population. The efficacy-evaluable population consisted of antibody responders and placebo patients, i.e., 1) all patients injected with AN1792(QS-21) who had a baseline ADAS–Cog evaluation, an anti-AN1792 IgG (total) serum titer ≥1:2,200 at any post-injection visit, and at least one post-injection ADAS–Cog evaluation; and 2) all patients injected with placebo who had a baseline ADAS–Cog efficacy evaluation and at least one post-injection ADAS–Cog evaluation.

The ADAS–Cog, DAD, CDR, and NTB subscales and CSF levels of tau and Aβ42 were analyzed using analysis of covariance with change from baseline as the response, baseline value as the covariate, and treatment and stratum (acetylcholinesterase inhibitor use [yes, no], MMSE score at screening [15–20, 21–26], and geographic location [Europe, United States]) as independent variables. The MMSE was analyzed using the same type model except that the MMSE score at screening was not included as a covariate. The Cochran–Mantel–Haenszel mean score test, stratified by acetylcholinesterase inhibitor use, MMSE score at screening, and geographic location, with equally spaced scores for the ordered levels of the response variable, was used to compare the ADCS–CGIC distributions for each treatment. In addition, all analyses were performed excluding encephalitis patients, and similar results were observed unless stated otherwise.

Patients.

A total of 372 patients were randomized to receive study treatment. One patient who had been randomly assigned to the placebo group received AN1792(QS-21) due to a dosing error (figure 1); this patient has been included in the AN1792(QS-21) group for all data summaries. Accordingly, 300 patients were treated with AN1792(QS-21) and 72 patients received placebo. Baseline characteristics and patient demographics were generally similar between the two study groups in both the safety and efficacy-evaluable populations (table 1). Patients in both the active and placebo groups had low mean Rosen-Modified Hachinski Ischemic total scores (suggesting a non-ischemic etiology of dementia); the scores were slightly lower (p = 0.004), but not clinically different, in the AN1792(QS-21) group in the safety population, and there was no difference between antibody responders and nonresponders. The majority of patients (>90%) received two of the planned six injections before the sponsors halted the study (see table 1). Of the 300 subjects who received AN1792(QS-21), 59 (19.7%) were antibody responders (i.e., developed a serum anti-AN1792 IgG (total) titer ≥1:2,200 at any time after injection 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Patient disposition. ^aOne patient randomized to the placebo group received AN1792(QS-21) due to a dosing error; ^bcompleters were defined as those patients who completed the study in accordance with the revised protocol.

Safety and tolerability.

AEs were reported in 266/300 (88.7%) patients who received AN1792(QS-21) and 59/72 (81.9%) placebo-treated patients, and occurred with similar frequency in antibody responders and nonresponders (table 2). Treatment-related AEs, serious AEs (SAEs), and AEs leading to treatment discontinuation were more frequent in AN1792(QS-21) recipients than in placebo recipients, and occurred more frequently in antibody responders than in nonresponders. Deaths and clinically important abnormal laboratory values were no more common in AN1792(QS-21)-treated patients than in placebo patients (see table 2).

Adverse events.

Table 3 summarizes the most frequently reported AEs. Infection (19%), headache (17.3%), diarrhea (9.7%), and encephalitis (6%) were reported more frequently (by ≥5%) among AN1792(QS-21) patients than in placebo recipients. Treatment-related AEs were reported in 77/300 (25.7%) patients in the AN1792(QS-21) group (22 responders and 55 nonresponders), of which 24/300 (8%) were reported as severe (see table 2) and the majority of these were associated with encephalitis.

The most common treatment-related AEs were encephalitis, headache, and confusion. Encephalitis occurred exclusively (p < 0.0001) in the AN1792(QS-21) group (18/300, 6.0%) as described in detail elsewhere.⁸ Thirteen of the 18 patients with encephalitis were classified as antibody responders, although the magnitude of antibody response was variable (see table E-1 on the Neurology Web site at www.neurology.org). Injection site reactions were reported in both groups; however, they tended to be more significant and of longer duration among patients treated with AN1792(QS21) vs placebo. Mild tenderness was the most common symptom.

Other SAEs.

Non-fatal SAEs were reported in 74 patients (see table 2); however, separate SAEs that resulted in death (cerebral infarct, accidental injury, neoplasm) were also reported in 3 of these patients. Twenty-two patients developed 27 treatment-related SAEs (encephalitis [n = 18], encephalopathy [n = 2, one of which also had encephalitis], confusion [n = 2], grand mal convulsion [n = 1], retinal vein thrombosis [n = 1], cerebral hemorrhage [n = 1], convulsion [n = 1], and hemiplegia [n = 1]), all of whom received AN1792(QS-21). Of these events, 13 cases of encephalitis, 1 case of encephalopathy, and 1 case of grand mal convulsion were reported in anti-AN1792 responders. The overall incidence of cerebral hemorrhage reported during the study was low and similar between AN1792(QS-21) (2/300 [0.7%]) and placebo groups (1/72 [1.4%]).

Seven patients died during the study follow-up period, and the incidence of death was similar in the AN1792(QS-21) and placebo groups (1.7% [5/300] vs 2.8% [2/72]; see table 2). Deaths in the placebo group were caused by neoplasm or cerebral hemorrhage, and those in the AN1792(QS-21) group were due to myocardial infarction (n = 2), broken neck, progression of AD, or non-hemorrhagic cerebral infarct. With the exception of one event of myocardial infarction, all deaths in the AN1792(QS-21) group occurred in antibody nonresponders. Only the cerebral infarction (which occurred 205 days after the second dose of AN1792(QS-21) in a nonresponder patient without encephalitis) was considered by the investigator as related to study treatment. Two additional deaths occurred after the end of the study follow-up period. One antibody nonresponder patient died secondary to aspiration pneumonia 15 months after the second injection of AN1792(QS-21), and one antibody responder patient died from progression of AD 13 months after the third injection. Consent was given for a brain autopsy examination in both patients.^18,19

Additional safety variables.

There were no notable trends in biochemistry, hematology, urinalysis, blood pressure, or heart rate over time or between groups observed during the study. Although clinically significant EKG abnormalities were reported in eight patients treated with AN1792(QS-21), none were considered related to treatment.

Immunologic findings.

Fifty-nine of the 300 patients (19.7%) treated with AN1792(QS-21) were antibody responders. The majority of these patients (47/59) received two injections of study drug before developing a positive response (serum anti-AN1792 IgG [total] titer ≥1:2,200 at any time after injection 1); three patients developed positive responses after one injection and the remaining nine patients after three injections. In the safety population (figure 2), peak geometric mean serum anti-AN1792 titers in treated patients were 1:174 (month 4) for IgG and 1:469 (month 2) for IgM. In the efficacy-evaluable population (see figure 2), peak geometric mean serum anti-AN1792 titers in antibody responders occurred at month 4, and were 1:3,848 for IgG and 1:4,534 for IgM.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Geometric mean serum titers in patients immunized with AN1792(QS-21) in the safety (all treated, n = 300) and efficacy-evaluable (antibody responder, n = 59) populations. (A) Anti-AN1792 IgG (total); (B) anti-AN1792 IgM. Patients were considered antibody responders if they had a serum anti-AN1792 IgG (total) titer ≥1:2,200 at any time after injection 1. Data presented are the reciprocal of the geometric mean titer.

No significant differences in geometric mean serum anti-AN1792 IgG or IgM titers were observed between patients who developed encephalitis and those who did not (see table E-1). A higher percentage of the patients with encephalitis were antibody responders compared with AN1792(QS-21)-treated patients without encephalitis (13/18 [72.2%] vs 46/282 [16.3%]; p < 0.0001). Of the five antibody nonresponders who developed encephalitis, four showed peak anti-AN1792 IgG titers ≥1:100 (range 1:641 to 1:1,114) and peak anti-AN1792 IgM titers ≥1:100 (range 1:297 to 1:12,157); the fifth patient had peak IgM titer of 1:1,201, but no elevation of IgG titer.

Of the 57 AN1792(QS-21)-treated patients who consented to post-baseline CSF sampling, anti-AN1792 antibody (IgG or IgM) was detectable in the CSF of nine patients, five of whom were antibody responders (serum IgG ≥1:2,200). Of the nine patients with detectable CSF antibodies, four developed encephalitis (three serum antibody responders and one serum antibody nonresponder).

Cognitive function.

There were no differences between treatment groups in cognitive, disability, and global change scores, as measured using the ADAS–Cog, DAD, CDR, MMSE, and ADCS–CGIC scales. These measures declined from baseline in antibody responders and placebo-treated patients during the course of the study follow-up period (table 4).

The results of the NTB revealed that antibody responders had an improvement compared with placebo in the WMVer-D scale (p = 0.047), and a small improvement from baseline in the WMDS scale (p = 0.094) (table 5). Mean adjusted change scores in the WMVis-I, WMVis-D, WMVer-I, RAVL-I, RAVL-D, COWAT, and CNT scales among antibody responders were not statistically different from those in placebo recipients.

Analysis of the nine-component composite NTB z-score indicated less worsening (p = 0.020) in the antibody responder group compared with the placebo group at month 12 (see table 5). This treatment difference was also apparent at month 12 in the all-memory (p = 0.033) composite z-score.

The NTB composite z-scores were regressed on the geometric mean antibody titers from four groups (placebo, titers 0 to 1:99, 1:100 to 1:2,199, and ≥1:2,200). Concentration-response analysis indicated relationships between geometric mean titer and the nine-component (p = 0.006), all memory (p = 0.009), immediate memory (p = 0.044), and delayed memory (p = 0.016) composite z-scores; that is, greater improvements from baseline in these NTB z-scores were associated with higher IgG antibody titers. The concentration-response relationship between geometric mean titer and the executive function composite z-score was not significant (p = 0.107), although the direction of the trend favored higher IgG antibody titers.

CSF tau and Aβ42.

In the subset of antibody responders (n = 11) and placebo recipients (n = 10) who had baseline and post-baseline CSF samples, mean (± SD) baseline CSF levels of microtubule-associated tau protein were similar (740 ± 243 for antibody responders vs 811 ± 368 pg/mL for placebo; p = 0.605). A marginal difference in CSF Aβ42 was observed between antibody responders and placebo recipients at baseline (588 ± 119 vs 469 ± 152 pg/mL; p = 0.059). Baseline CSF tau and Aβ42 were within the expected ranges for patients diagnosed with probable AD.²⁰ CSF tau in antibody responders was reduced compared with baseline (−204 ± 57 pg/mL), and the change was greater than in the placebo group (42 ± 52 pg/mL; p < 0.001) (figure 3). Treatment with AN1792(QS-21) had no effect on CSF levels of Aβ42.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Change from baseline in CSF tau and Aβ42 in the efficacy-evaluable population of patients with baseline and post-baseline lumbar punctures. (A) CSF tau; (B) CSF Aβ42. Baseline levels of CSF tau and Aβ42 did not differ significantly between antibody responder (n = 11) and placebo (n = 10) groups. Data presented are raw mean (horizontal bar) and individual subject (symbol) changes from baseline.