Changes in Brain Volume with Bapineuzumab in Mild to Moderate Alzheimer’s Disease

Study design

The phase 3 studies were multicenter, randomized, double-blind, placebo-controlled, 18-month trials of intravenous bapineuzumab in subjects aged 50 to 89 with mild to moderate AD dementia [11] who had baseline scores on the Mini-Mental Status Exam (MMSE) between 16 and 26, inclusive, and a Rosen-Modified Hachinski Ischemic Score ≤ 4 [12]. All procedures performed on the human subjects in this clinical trial were conducted in accordance with the Helsinki Declaration of 1975 and the ethical standards of the participating institutions, and were approved by local and/or central Institutional Review Boards or Ethics Committees. Patients were excluded for clinically significant neurological disease other than AD; screening visit brain MRI scan indicative of other significant abnormality, including ≥ 2 microhemorrhages, prior hemorrhage >  1 cm³, ≥ 2 lacunar infarcts, prior infarct >  1 cm³, or space occupying lesions; a major psychiatric disorder; history of stroke or seizures; current anticonvulsant, antiparkinsonian, anticoagulant, or narcotic medications; immunosuppressive or cancer chemotherapy; or cognitive enhancers other than acetylcholinesterase inhibitors or memantine at a stable dose for at least 120 days.

The primary clinical endpoints for both studies included the Alzheimer’s Disease Assessment Scale–Cognitive [13] and the Disability Assessment for Dementia [14] administered at baseline, at every treatment visit, and at Week 78 (or early termination visit).

Each study included three key biomarker secondary endpoints: Brain fibrillar amyloid burden as estimated from carbon-11-labelled Pittsburgh compound B (¹¹C-PiB) positron emission tomography (PET), cerebrospinal fluid (CSF) concentrations of tau phosphorylated at threonine 181 (CSF p-tau), and vMRI. PiB-PET and CSF p-tau results will be presented in separate publications.

Study 302 (APOE^*ɛ4 carriers) was to randomize 1,000 subjects to placebo or bapineuzumab 0.5 mg/kg, given every 13 weeks for six doses. 480 subjects were to be included in the vMRI substudy to achieve 90% power to detect a difference (placebo-bapineuzumab) in BBSI of 4.15 ( ± 13.7) mL at Week 71.

Study 301 (APOE^*ɛ4 noncarriers) was to randomize 1,300 subjects to placebo, bapineuzumab 0.5 mg/kg, or 1.0 mg/kg, given every 13 weeks for six doses. 610 subjects were to be included in the vMRI substudy to achieve 90% power to detect a difference (placebo-bapineuzumab) in BBSI of 5.05 ( ± 13.7) mL at Week 71. The study included a 2.0-mg/kg arm, which was discontinued due to increased occurrence of amyloid-related imaging abnormalities—edema or effusion (ARIA-E) [15, 16]. These subjects continued in the study, but further dosing was reduced to 1.0 mg/kg, and they were not included in efficacy or biomarker analyses.

Acquisition and processing of vMRIs

vMRIs were performed at a baseline assessment 2 to 6 weeks prior to randomization, and at Weeks 19, 45, and 71 following randomization. Image acquisition and quality control were designed to be consistent with the Alzheimer’s Disease Neuroimaging study (ADNI-1) [17]. Volumetric measurements were based on 3DT1-weighted sequences obtained in the coronal or sagittal plane on 1.5T or 3.0T scanners. Images were collected and quality controlled by an imaging core laboratory (Synarc, Newark, CA, USA). Quantitative image analysis was performed by the Core Analysis Laboratory at the Dementia Research Centre at University College London Institute of Neurology, using the Medical Image Display and Analysis System, blind to patient details or treatment allocation. All scans were corrected for intensity non-uniformity [18]. Each individual’s follow-up scan was assessed for consistency of acquisition, positionally matched (co-registered) to baseline, and corrected for non-uniform intensity fields using differential bias correction [19].

Outcome measures

The principal vMRI outcome was rate of whole brain atrophy, indexed by annualized BBSI [20]. Annualized BBSI was derived at each time point as change in brain volume from initial MRI measured by BSI method, divided by time elapsed (days) from initial MRI, multiplied by 365.25. Additional outcomes included rate of ventricular expansion, indexed by annualized ventricular BSI (VBSI), and rate of mean (right and left) hippocampal atrophy, indexed by annualized mean hippocampal BSI (HBSI). In addition, volume change was also presented separately for right and left hippocampi. By convention, positive values for BBSI and HBSI measurements indicate volume reduction in these structures, and positive values for VBSI indicate ventricular volume increase.

Data analysis

Treatment-related differences in annualized BBSI (mL/y) at Weeks 19, 45, and 71 were estimated statistically in both studies, as well as for the pooled population of both studies, using a mixed model for repeated measures (MMRM). Observed value of annualized BBSI was the response variable, and the fixed-effect model terms included treatment (bapineuzumab 0.5 or 1.0 mg/kg [Study 301 only] or placebo), visit (scheduled week), treatment-by-visit interaction, baseline whole brain volume (WBV), baseline WBV-by-visit interaction, baseline MMSE total score stratum, use of acetylcholinesterase inhibitor or memantine stratum, APOE^*ɛ4 allele copy number stratum (Study 302 only), study (pooled studies only), and age at baseline. An unstructured variance-covariance matrix was used. Treatment differences (bapineuzumab minus placebo) were estimated using least squares means with factor levels weighted according to overall baseline sample proportions. A similar MMRM analysis was performed for mean, right, and left HBSI. For VBSI, MMRM did not include baseline ventricular volume. Primary analysis was based on Week 71 treatment difference.

Additional MMRM analyses of BBSI, VBSI, and mean HBSI were carried out for separate subgroups based on severity of dementia at baseline (mild: MMSE = 21–26; moderate: MMSE = 16–20). Compared with MMRM used for overall estimation of treatment-related difference, MMRM used for subgroup analyses excluded baseline MMSE total score stratum. Additionally, a direct estimation of the effect of dementia severity on BSI measures was carried out, comparing mild versus moderate placebo subjects within each trial and the pooled studies. Compared with MMRM used for overall estimation of treatment-related difference, MMRM used for between-severity comparison excluded the fixed-effect terms treatment and treatment-by-visit interactions, but included additional terms of dementia severity and dementia severity-by-visit interaction.

All p-values reported are nominal. No adjustments for multiplicity were performed.

Brain volume changes with treatment

Brain volume was decreased at an annual rate of 16.2 to 19.1 mL/y across all treatment groups in both studies, equivalent to 1.6% to 1.9% /y of baseline WBVs (Fig. 1, Table 2). Estimated differences between bapineuzumab-treated and placebo subjects for annualized BBSI were not significant for the 0.5-mg/kg dose in the carrier study or the 0.5- or 1.0-mg/kg doses in the noncarrier study. However, in the pooled carrier and noncarrier analyses, subjects receiving bapineuzumab 1.0 mg/kg (all noncarriers) had a significantly greater rate of brain atrophy compared with pooled placebo subjects (approximately equal numbers of carriers and noncarriers). Cumulative increase in BBSI at each time point relative to baseline is presented in Fig. 2; the rate of brain volume loss was approximately linear over 71 weeks.

A significantly greater increase in ventricular volume was observed for bapineuzumab-treated carriers relative to placebo in the 0.5-mg/kg dose, and bapineuzumab-treated noncarriers relative to placebo in the 1.0-mg/kg dose, but not the 0.5-mg/kg dose. In the pooled carrier and noncarrier analyses, there were significantly greater annualized VBSI increases for both bapineuzumab doses relative to placebo.

There was a trend for a greater rate of mean hippocampal atrophy for the 1.0-mg/kg dose relative to placebo in noncarriers; this trend became statistically significant for this dose in the pooled carrier and noncarrier analyses, although all subjects treated with the 1.0-mg/kg dose came from the noncarrier study. When left and right annualized HBSI were assessed separately, bapineuzumab-treated noncarriers were found to have significantly greater annualized left HBSI relative to placebo at Week 71 (1.0 mg/kg, Δ=  18.5 ± 7.1 mm³/y, p = 0.010), though annualized right HBSI differences were not significant (Fig. 3). In carriers, neither right nor left annualized HBSI showed significant differences between bapineuzumab and placebo.

Effect of disease severity

In order to assess whether rate of brain atrophy was influenced by disease severity, and whether this might have interacted with treatment, separate analyses of volume change were performed in subjects with mild (baseline MMSE 21–26) and moderate (baseline MMSE 16–20) AD. In placebo subjects, rates of volume change over the 18-month study ranged from 30% to 80% greater in moderate compared with mild AD; for all comparisons, p <  0.001 (Fig. 4, Table 3).

Although the magnitude of treatment effect differed across severity subgroups, there did not appear to be a consistent pattern for the different BSI measures (Fig. 4). When analyses were restricted to subjects with mild AD, there was a significantly greater rate of brain atrophy in APOE^*ɛ4 carriers treated with bapineuzumab 0.5 mg/kg relative to placebo but not in APOE^*ɛ4 noncarriers for either dose. In the pooled carrier and noncarrier analyses, rate of brain atrophy was greater for both doses in subjects with mild AD relative to placebo. In contrast, there were no statistically significant treatment differences for rate of brain atrophy in subjects with moderate AD.

Significant treatment-related increases in ventricular volume were seen for carriers with mild AD at 0.5 mg/kg and noncarriers with mild AD at 1.0 mg/kg relative to placebo; these differences were significant for noncarriers with moderate AD. For annualized mean HBSI, none of the differences achieved statistical significance.

Effects of AD and disease severity on vMRI

Rate of brain volume reduction was consistent with rates reported in observational studies of mild to moderate AD [21–24]. Reduction of brain parenchymal volume is accompanied by dilatation of both the ventricular and subarachnoid spaces. In placebo subjects, annualized VBSI, presumed to reflect a portion of this ex vacuo change, accounted for about 30% of annualized BBSI. The hippocampus, a site of early neurofibrillary pathology, consistently has a high rate of atrophy relative to its size in AD. In placebosubjects, annualized mean HBSI was slightly greater for carriers than noncarriers (97.8 versus 87.5 mm³/y, respectively, equivalent to –4.4% versus –3.6% /y of baseline hippocampal volumes), again consistent with rates of atrophy in observational studies [24, 25]. Additionally, subjects with moderate dementia had a greater rate of brain atrophy compared with those with mild dementia, consistent with previously reported longitudinal studies in which within-subject acceleration has been detected over time intervals longer than present bapineuzumab studies [26–28].

Treatment-related differences in vMRI

Annualized BSI measurements were, for nearly all comparisons, greater for bapineuzumab-treated than placebo subjects, though differences were relatively small and often not statistically significant. In the case of annualized BBSI, observed between-treatment differences of 1 to 1.5 mL/y did not achieve statistical significance, and were equivalent to about 6% to 9% of the difference that can be attributed to disease progression over the assessment period (i.e., baseline to Week 71 differences for placebo subjects). Although BBSI was greater for the bapineuzumab 1.0-mg/kg group than placebo in the pooled analysis, this comparison is confounded by the fact that subjects in the active dose group all came from Study 301 (APOE^*ɛ4 noncarriers), while placebo subjects came from Studies 301 and 302, and therefore include carriers and noncarriers.

In both studies, annualized VBSI was nominally greater for all bapineuzumab treatment groups relative to placebo; this was statistically significant for 0.5 mg/kg in carriers, 1.0 mg/kg in noncarriers, and both doses in the pooled analysis. It is of interest that the magnitudes of between-treatment differences for VBSI and BBSI are approximately the same, suggesting that ex vacuo effects of brain parenchymal volume reduction associated with treatment are relatively greater for ventricular than subarachnoid spaces. As noted above, in the placebo subjects, VBSI accounted for about 30% of atrophy attributable to disease progression relative to BBSI.

Treatment-related differences based on disease severity

It was hypothesized that there might be differences in treatment effect based on disease severity. This was not apparent for the clinical outcomes [10]. Although between-treatment differences in BBSI were seen only in subjects with mild AD, between-treatment differences in VBSI were significant regardless of severity. One intriguing possibility suggested by these results is that different pathological processes drive brain volume reduction and ventricular expansion, manifest as a differential sensitivity to the effects of immunotherapy at different stages of disease.

Consistency with prior reports of immunotherapies

Given the evolving nature of research in vMRI as a biomarker and the exploratory intent of these analyses, there was no correction of alpha for multiplicity, hence no rigorous control of Type 1 error. It is possible that some of the above treatment-related differences may have arisen by chance. However, this seems unlikely for two reasons: The pattern of results was consistent as all differences that achieved nominal significance indicated a greater volume loss for bapineuzumab compared with placebo; and the pattern of results are similar to those observed in other clinical trials of amyloid immunotherapy.

In the previously mentioned phase 2 clinical trial of bapineuzumab [9], there were no significant treatment effects in the modified intention-to-treat sample. When the data were analyzed by APOE^*ɛ4 carrier status, two treatment differences were apparent: in APOE^*ɛ4 carriers, VBSI was greater in bapineuzumab-treated than placebo subjects; in noncarriers, BBSI was greater for placebo than bapineuzumab-treated subjects. This was interpreted as supporting a possible favorable treatment effect of bapineuzumab. However, when one compares across treatment and APOE^*ɛ4 groups, it may be seen that BBSI for bapineuzumab-treated noncarriers was more nearly similar to both placebo and bapineuzumab-treated carriers; placebo noncarriers, numerically the smallest group (n = 22), had a much greater BBSI. Failure to replicate this treatment difference, and similarity in the magnitude of BBSI in the phase 3 studies when placebo noncarriers in the phase 2 study are excluded, would suggest that this latter finding may have arisen by chance due to a larger BBSI in the smallest group of subjects.

In a trial of immunization with aggregated human Aβ_1-42 (AN1792) in AD, treatment was stopped after reports of meningoencephalitis, which occurred in about 6% of subjects [29]. Of the 300 AN1792-treated patients who received one to three of five planned injections, 59 (19.7%) developed the predetermined antibody response. Double-blind assessments were maintained for 12 months. No significantdifferences were found between antibody responder and placebo groups for prespecified primary efficacy outcomes, but improvement was noted in some secondary cognitive outcomes [30]. Notably, antibody responders also had reduction in brain volume relative to placebo subjects as indexed by a greater BBSI and VBSI [31]. This apparent dissociation of brain volume loss and cognitive function remains unexplained, but possible reasons for volume loss include removal of amyloid, associated plaque components and related inflammatory cells, and associated cerebral fluid shifts. Interestingly, in a follow-up study 4.5 years later [32], in the small number of patients who were evaluated for brain volumetry, no differences between treated responders and placebo wereobserved.

Possible explanations for effect of bapineuzumab on BBSI, VBSI, and HBSI

If it is assumed that treatment with bapineuzumab over periods of 71 weeks may have resulted in additional brain volume loss up to 2.9 mL/y (as observed for BBSI in carriers with mild AD; Fig. 4), what are the possible explanations for this effect?

The first consideration is an effect of accelerating the progression of volume loss due to disease. This seems unlikely, as there were no treatment-related statistically significant differences for multiple cognitive or functional endpoints for any dosage of bapineuzumab, and the overall pattern of results as reported by Salloway et al. [10] does not show even a nominal difference in favor of placebo that is consistent across the data set. It must be noted, however, that there is a greater variability in clinical endpoints than for the BSI measures and a relatively small increase in the rate of clinical change, proportional to the 6–9% increase in BSI relative to total change from baseline, might be difficult to detect. Since the progression of atrophy in AD has a characteristic topographical pattern, it is possible that regional volumetric assessment might differentiate disease progression from other pharmacodynamic effects on brain volume.

Another potential explanation is loss of cortical volume related to clearance of amyloid plaques, which may account for up to 9% of brain volume in some cerebral regions in AD [31]. This has been suggested as a possible explanation for volume loss seen in subjects in the AN1792 study [31]. In fact, previous pathologic studies in patients who came to autopsy after being immunized with AN1792 have shown not only a loss of plaques, but also a reduction in p-tau and associated inflammation [33, 34]. A recent study found that multiple inflammatory markers were reduced in AN1792-immunized patients, leading the authors to conclude that immunization may result in reduced activated microglia and inflammation over the long term [34]. Since the total volume of cerebral plaques (and importantly the associated inflammatory and neuritic peri-plaque pathology) that exist pre- or post-treatment in any AD subject cannot be quantitated by noninvasive means, it would be difficult to estimate the magnitude of this effect in any one subject or group of subjects. Newer imaging agents that measure tau and inflammatory markers may be promising modalities in the future to further evaluate the response of these pathologiesin vivo in response to treatment.

It has also been suggested that observed changes in brain volume could result from fluid shifts. Fluid shifts resulting in brain volume reduction must involve reduction of intracellular, extracellular, or intravascular compartments, or some combination of these. Imaging methods used in this study do not permit a selective estimate of volume change in any one of these compartments, nor do the known pharmacological effects of bapineuzumab suggest a mechanism by which bapineuzumab would drive water outside of cells, shrink interstitial fluid volume, or reduce cerebral blood volume. In fact, the observation of ARIA-E, which occurred in 9.5% of APOE^*ɛ4 noncarriers treated with bapineuzumab 0.5 to 1.0 mg/kg, and 21.2% of carriers treated with 0.5 mg/kg, suggests that treatment with bapineuzumab produces a transient increase in extracellular volumes in some subjects, which would be expected to increase brain volume globally, resulting in a reduction in rate of BBSI. This was not seen in the data, perhaps because the number of subjects with ARIA-E was relatively small, and the process was likely to have resolved by the final vMRI assessment.

As noted above, although differences in BBSI were not statistically significant in either trial, VBSI differences were significant in both trials. Mechanisms that might explain the apparent selective increase in rate of ventricular volume expansion with bapineuzumab immunotherapy are not clear. Preclinical studies have suggested there may be a transient increase in amyloid deposition in blood vessels as amyloid is cleared from the parenchyma [35]. This may lead to compromised drainage of interstitial fluid with a resulting increase in interstitial volume, which may manifest as ARIA-E [36]. Related to this hypothesis is the possibility that ongoing clearance and shunting of parenchymal Aβ into perivascular spaces leads not only to an increase in parenchymal interstitial fluid but may result in elimination of this excess fluid into the ventricles. It has been suggested that in AD the ventricles may act as a sink for interstitial fluid not readily cleared through amyloid-laden perivascular spaces, resulting in ventricular enlargement [37].