Flutemetamol was negative for genotoxicity after in vivo exposure in rats to flutemetamol at the highest cumulative dose level tested, as measured in bone marrow micronucleus assays (157 and 27 mcg/kg/day for 2 days and 14 days, respectively) and an unscheduled DNA synthesis assay in rat hepatocytes (39 mcg/kg/day).
The effectiveness of VIZAMYL was evaluated in two single-arm clinical studies (i.e., Studies 1 and 2) in adult subjects with a range of cognitive function, including some terminally ill subjects who had agreed to participate in a postmortem brain donation program as well as healthy subjects. Subjects underwent a VIZAMYL injection and scan. The images were interpreted using a clinically applicable binary image interpretation method (negative or positive) by five independent readers blinded to all clinical information [see Dosage and Administration (2.3, 2.4, 2.5)]. PET images were reviewed first without, and subsequently with, brain CT or MRI images. Before image interpretation, all readers underwent in-person training or electronic media training.
To determine the agreement between the in vivo VIZAMYL image results and the postmortem amyloid beta neuritic plaque density, VIZAMYL results (negative or positive) were pre-specified to correspond with specific histopathology-derived plaque density scores, based upon a modification of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) criteria (Table 6), which use neuritic plaque counts as a necessary pathological feature of AD. Plaques were counted on microscope slides with modified Bielschowsky silver stained tissue sections. The global neuritic plaque density score for each subject was determined by averaging across the scores (0-3) for five gray matter fields per slide and then across six slides for each of eight regions; if any one region had a score of greater than 1.5, the subject was classified as positive for amyloid beta.
Study 1 evaluated performance characteristics (sensitivity and specificity) in terminally ill subjects by comparing the premortem VIZAMYL PET images to a postmortem truth standard of cortical amyloid beta neuritic plaque density. A total of 180 subjects were dosed with VIZAMYL and 176 were imaged. The mean age was 80 years (range 47 to 98 years) and 57% of the subjects were female. By medical history, 135 subjects had dementia, one subject had memory loss of unspecified nature, and 44 subjects had no cognitive impairment. Among the imaged subjects, 68 subjects who died during the study and had cerebral cortical amyloid status determined were included in the primary analysis. The time interval between the VIZAMYL scan and death ranged from 0 to 13 months, with a median of 2.6 months, and was less than one year for 66 subjects and between 12 to 13 months for 2 subjects. At autopsy, the global brain neuritic plaque density category (CERAD score as in Table 6) was available for 67/68 subjects: frequent (n = 19); moderate (n = 22); sparse (n = 14); and none (n = 12).
Study 2 evaluated inter-reader and intra-reader reproducibility of image interpretation using images from subjects with a truth standard (68 subjects who underwent an autopsy in Study 1 and 36 subjects who had known or suspected normal pressure hydrocephalus with in vivo brain biopsy) and subjects without a truth standard (80 subjects with amnestic mild cognitive impairment (aMCI), 33 subjects with probable AD (pAD), 28 cognitively normal subjects 55 years of age and older, and 31 young healthy subjects). Additionally, intra-reader reproducibility was assessed with images from 29 subjects (10%). Among the 276 subjects, the mean age was 68 years (range 20 to 95 years), 136 were females, and 251 were White. Readers were naïve to all forms of amyloid PET imaging and underwent electronic media training.
Among subjects who underwent autopsy (n=68; 43 positive and 25 negative based on histopathology), the sensitivity using the majority interpretation of the readers trained using electronic media was 93% (95% CI: 81%, 99%) and specificity was 84% (95% CI: 64%, 96%). The median (and range) of correct, false negative, and false positive reads were 59 (51, 61), 5 (3, 8), 3 (2, 14), respectively, for in-person training; and were 60 (55 to 61), 3 (3 to 6), 4 (2 to 10), respectively, for electronic media training.
Table 7 shows inter-reader reproducibility results among readers for various subject groups in Study 2. Inter-reader reproducibility analyses showed an overall Fleiss' kappa statistic of 0.83 (95% CI: 0.79, 0.86), which met the pre-specified success criterion (95% CI lower bound > 0.60). Intra-reader reproducibility analyses showed that between the two readings for each of the 29 duplicate scans, one of the five readers had complete agreement for all 29 scans, two readers had discordant reads for a single scan, and two readers had discordant reads for two scans. Intra-reader reproducibility for a subgroup of eight scans from aMCI subjects showed that all five readers had complete agreement for all duplicate scans.