Relationship Between PET-Assessed Amyloid Burden and Visual and Verbal Episodic Memory Performance in Elderly Subjects

Study sample

A total of 38 patients with mild dementia clinically compatible with AD, 43 aMCI subjects, and 34 cognitively unimpaired (CU) volunteers were enrolled in the study. The ethical committee of the Hospital das Clínicas, Faculdade de Medicina, Universidade de São Paulo (HC-FMUSP) approved the investigation (CAPPesq 368.037), and informed consent was obtained from all participants or their caregivers in the dementia group.

Patients in the mild dementia group fulfilled the 2011 criteria of the National Institute on Aging-Alzheimer’s Association (NIA-AA) [26], and the diagnostic criteria for dementia of the Alzheimer’s type according to the DSM-IV [27]. The presence of consistent cognitive deficits compatible with the diagnosis of probable AD according to the NIA-AA criteria was documented by extensive neuropsychological assessments including tests evaluating auditory and visual verbal episodic memory, visuoconstructive function, semantic memory, attention, processing speed, language, and executive functioning. The overall severity of cognitive decline was assessed using the Mini-Mental State Examination (MMSE) [28]. Patients in the mild dementia group were also evaluated with the Blessed Dementia scale [29] and the Informant Questionnaire on Cognitive Decline in the Elderly (IQ-CODE) [30] in order to assess cognitive decline, and the Hachinski Ischemic Score (HIS) to assess the possible contribution of cerebrovascular pathology to the development of dementia [31].

Patients with aMCI also underwent neuropsychological evaluations, and their diagnosis was based on the revised Petersen criteria [32, 33]. Patients should present the following features: a) a perception of cognitive decline in relation to a previous level (according to the participant and an informant); b) deficits in one or more cognitive domains below the expected norms for age and education, with at least one of the domains being related to memory (amnestic impairment as a single or multiple domain aMCI); c) independence for daily activities; d) absence of dementia; e) absence of significant vascular disease, trauma or general medical conditions as the attributable cause of the cognitive deficit. Memory impairment in the aMCI group was considered present if subjects had scores at least 1.5 standard deviation (SD) below the age and education stratified norms on one memory test, or 1.0 SD below norms on more than one memory test. Besides memory impairment, participants might have deficits in other cognitive domains.

Based on neurological examination and structured interviewing using the translated version of the Structured Clinical Interview for DSM-IV (SCID-IV) [34], no patients in the AD or aMCI groups had history of neurologic diseases (such as Parkinson’s disease, epilepsy, inflammatory diseases or stroke, with the exception of migraine) or major psychiatric disorders with the exception of previous history of depressive, anxiety or alcohol-abuse disorders. As detailed elsewhere [35], eighteen cognitively impaired subjects (8 in the dementia group and 10 in the aMCI group) presented history of mood and/or anxiety disorders (currently in remission), and fifteen of those were under psychopharmacological treatment when recruited for the study.

Subjects for the CU group were recruited via online and printed announcements in activity centers and groups for elderly individuals in the community, invitation of acquaintances of the relatives of dementia and aMCI subjects, and invitation of elderly subjects who were screened in the community during other research studies recently carried out by our institution. They also underwent neurological examination, structured interviewing using the SCID-IV and extensive neuropsychological assessments as described above. They all had a normal cognitive performance for their age range and no past or current history of the above major psychiatric disorders or neurological conditions. As detailed elsewhere [35], six control subjects presented history of mood and/or anxiety disorders (currently asymptomatic), two of which under current psychopharmacological treatment.

Additional exclusion criteria for both the cognitively impaired and control samples were: presence of severe medical conditions and any systemic disorders associated with cognitive impairment; family history of dementia with a dominant pattern of inheritance; uncorrected auditory or visual impairments that could interfere with the cognitive evaluation; and presence of gross vascular pathological findings or other lesions incidentally detected by brain structural magnetic resonance imaging (MRI) scanning.

No subjects in the three groups were under chronic use of any other drugs with action in the central nervous system. A total of 8 subjects (two in the dementia, two in the aMCI, and three in the control groups) had been exposed to benzodiazepines recently but not continuously. Other medication used regularly for stabilized general medical conditions across the three groups included: a) anti-hypertensive medications of the following classes: diuretics, angiotensin-converting enzyme (ACE) inhibitors, calcium channel blockers, beta blockers, vasodilators; b) oral hypoglycemic medications; c) statins; d) acetylsalicylic acid; e) calcium carbonate; f) and hormone replacement therapies with estrogen or levothyroxine.

The core set of neuropsychological tests that were applied to all subjects in the three groups within a mean of 6 weeks prior to PET imaging are provided in Supplementary Table 1. Episodic memory indices used for the investigation of associations with biomarker status in the present investigation included: the delayed recall score on the Rey Auditory-Verbal Learning Test (RAVLT) as a measure of verbal auditory episodic memory [36, 37], thereafter referred to as verbal memory; and the memory sub-score on the Short Cognitive Performance Test (Syndrom Kurztest, SKT) as a measure of verbal visual memory [38, 39], thereafter referred to as visual memory. The RAVLT requires the individual to encode and immediately recall a 15-item word list five times. Next, he/she is asked to encode and recall a new list (interference) and then to recall the original list. After 30 min, in the delayed recall, the individual needs to remember as many words as possible from the original word list. The test is based on listening and orally repeating words and there are no visual stimuli. In the present analyses, the variable of interest was the number of items recalled in the delayed recall. The SKT has an attention sub-domain (0–18 points) and a memory sub-domain (0–9 points) [38]. In the present study, the memory domain was included in the analyses. This domain is composed of three tasks, each one scoring 0–3 points: immediate recall of 12 colorful pictures of common objects after the individual names them (he/she is reminded of the not-recalled ones). After approximately 5 min, the individual needs to recall the pictures, and, finally, in the recognition task, he/she needs to identify the original 12 pictures among other 12 new pictures. The variable of interest in the study was the total score for the SKT memory sub-domain. Higher SKT scores represent worse memory performance.

PET data acquisition

Both [¹¹C]PIB and [¹⁸F]FDG were produced at the Centro de Medicina Nuclear of HC-FMUSP using an on-site cyclotron (PETtrace-880, GE-Healthcare) and PET/CT images were acquired using a Discovery-710 PET/CT-scanner (GE-Healthcare, Milwaukee, USA). Before data acquisition, a CT scan from the skull was acquired and used for attenuation correction, at: 120 kVp and 70 mA (140 mA and 0.5 s per rotation) for the [¹¹C]PIB-PET data acquisitions; and 120 kV, range 70–100 mA for [¹⁸F]FDG-PET data acquisitions.

Details of the [¹¹C]PIB radiochemical production were described elsewhere [23]. For the acquisition of [¹¹C]PIB -PET data, subjects were positioned with the brain in the center of the field of view of the PET equipment and [¹¹C]PIB was injected intravenously (296–740 MBq). The acquisition started immediately after the beginning of tracer injection. Images were acquired for 70 min using a 256×256 mm matrix. Images from 40 to 70 min after injection were condensed in time creating a static [¹¹C]PIB-PET showing extracellular Aβ deposits. Images were corrected for radioactivity decay, dead time, attenuation and scatter and reconstructed using the OSEM algorithm, 4 iterations and 16 subsets. For the acquisition of [¹⁸F]FDG-PET data, participants fasted for at least 6 h before imaging. Blood glucose levels were measured and recorded immediately before [¹⁸F]FDG injection to ensure levels below 200 mg/dL. Subjects remained at rest with eyes open for at least 30 min after intravenous [¹⁸F]FDG injection (177 to 344 MBq). Data were acquired in static mode, for 15 min, with time-of-flight information and a 256×256 mm matrix, and were reconstructed using OSEM with four iterations and sixteen subsets and a 3 mm Gaussian filter with TOF and point spread functions modelling (VUE Point FX with Sharp IR - GE Healthcare). Data were corrected for scattering, attenuation, dead time, and radioactive decay.

Protocol for defining Aβ positivity and measures of standardized uptake value ratio (SUVr)

The protocol used for the classification of individuals as A+ or A–was detailed previously [35]. In summary, two certified nuclear medicine physicians with more than five years of experience in the field, evaluated each [¹¹C]PIB-PET dataset independently and blinded to diagnoses. Images were rated as “positive” if there was an increase in [¹¹C]PIB uptake in cortical gray matter areas causing a loss of GM to white matter contrast in at least two of the six following areas: frontal, temporal, lateral parietal, precuneus, anterior cingulate and posterior cingulate cortices, and precuneus. They also rated the images as positive if only one single large cortical area had a strong diffuse [¹¹C]PIB uptake. A scan was rated as “negative” when there was strong white matter uptake and no significant gray matter uptake [40]. As a second step, each physician performed an analysis with the aid of the 3D-SSP semi-quantitative method designed for the clinical analysis of brain PET amyloid imaging (Cortex ID Suite software, GE healthcare), with the standard uptake values ratio (SUVr) of the cortical areas normalized to the cerebellar gray matter. A SUVr cutoff of 1.42 was used for A+ scoring for the composite of cortical areas [41], complementing the visual analysis. In cases of persisting discordance between the readers after the 3D-SSP analyses, consensus readings with both physicians were performed.

Additionally, in order to generate one single continuous variable of cortical Aβ load, the SUVr of [¹¹C]PIB uptake was extracted for 34 regions of interest placed on the prefrontal, orbitofrontal, parietal, temporal, parahippocampal, anterior cingulate, and posterior cingulate cortices and precuneus, normalized to whole cerebellar tracer uptake [41]. For the generation of SUVr for each region, [¹¹C]PIB-PET images were co-registered to a T1 weighted MRI of the same subject using the PNEURO tool of the PMOD 3.4 software (PMOD™ Technologies Ltd, Switzerland), and brain parcellation was carried out based on the HAMMERS atlas for generating different volumes of interest (VOI). Images were corrected for partial volume effect using the “VOI based” correction method. A meta-region of interest including all regions was used to generate a composite [¹¹C]PIB SUVr for each individual [42].

Visual classification of [¹⁸F]FDG PET data

The protocol used for the classification of individuals as N+ or N–was described previously [35]. In brief, evaluations of individual [¹⁸F]FDG-PET datasets were carried out by the same two certified nuclear medicine physicians cited above, who were blinded to clinical diagnoses. The two physicians scored each exam as “positive” or “negative” (first separately and then in consensus) based on visual interpretation assisted by the 3D-SSP semi-quantitative software (Cortex ID Suite, GE healthcare) [25]. They used three different methods of tracer uptake normalization (global uptake, pons or cerebellum as reference), based on the fact that normalization using different regions as reference may complement each another in terms of sensitivity and specificity [43–45]. Abnormal [¹⁸F]FDG-PET datasets were rated as typical of AD when there were unilateral or bilateral findings of decreased regional brain glucose metabolism affecting mainly the posterior temporo-parietal, posterior and inferior temporal cortices, the precuneus and the posterior cingulate cortex [25, 47].

Classification of subjects in four categories based on the presence of Aβ aggregation and neurodegeneration

Based on the above described individual classification of [¹¹C]PIB-PET and [¹⁸F]FDG-PET datasets as “positive” or “negative”, we classified each individual in one of the four following categories [35]: A+(N)+ (Alzheimer’s pathology with neurodegeneration); A+(N)–(Alzheimer’s pathology without neurodegeneration); A–(N)–(non-pathological findings); and A–(N)+(SNAP). We were not able to apply the complete AT^*(N) classification based on the 2018 NIA-AA RF since neither PET imaging nor CSF analyses evaluating tau pathology were performed in the present study [5].

Data management and statistical analysis

Demographic data, clinical details, and main neuropsychological characteristics for the three groups were managed using the REDCap (Research Electronic Data Capture) method [48, 49]. Statistical tests were run using the Statistical Package for Social Sciences (SPSS) version 17.0 (IBM Corp.).

Group comparisons of demographic data, clinical details, and A(N) frequencies were conducted using one-way ANOVA for continuous variables and chi-square tests for categorial variables.

Using the univariate generalized linear model option in SPSS, we then carried out comparisons between the four A(N) categories entered as fixed factors, with cognitive test scores (RAVLT delayed recall score, SKT memory sub-score, and MMSE score) as dependent variables and age, sex and years of education as covariates. Significant findings in such analyses were followed-up with post-hoc Schefffé’s tests (unequal sample sizes) for two-group comparisons.

Subsequently, we ran partial correlation analyses (again controlled for age, sex, and education) between cognitive test scores and continuous SUVr measurements. Such analyses were first carried out for the overall sample, then separately for the (N)+ and (N)–subsamples, and finally for the A(N) biomarker-based categories.

Finally, both the group comparisons and partial correlation analyses above were repeated with the overall sample divided according to clinically defined diagnoses rather than A(N) categories, including: the overall cognitively impaired sample (dementia plus aMCI subjects); and the separate dementia, aMCI and CU groups.

In all analyses above, a one-tailed p < 0.05 level of statistical significance was applied.

3.1 Demographic and clinical data, cognitive test scores, and biomarker profile for the sample

Demographic and clinical details for the overall sample (n = 115), and then separately for the clinically defined subgroups are provided in Table 1. There were no significant between-group differences in regard to age, sex, number of years of education, or socio-economic status (Table 1).

In regard to cognitive test results, the greatest discrimination between the two cognitively impaired groups and CU subjects was obtained when verbal memory scores (RAVLT) were assessed, with significantly lower scores both in the dementia and aMCI groups relative to CU subjects (Table 1). Worse performance indices on the visual memory (SKT) and the MMSE were detected in the dementia group relative to both aMCI and CU subjects (Table 1).

Regarding [¹¹C]PIB uptake, the dementia sample displayed higher levels of Aβ deposition relative to aMCI and CU subjects either when mean SUVr scores or the frequency of subjects classified as A+ were analyzed (Table 1).

There were highly significant differences between the clinically defined groups in regard to the A(N) categorization, with the proportion of A+(N)+ subjects being higher in the dementia group (65.8% of all dementia cases) in comparison to the proportion of individuals rated as A+(N)+ in the aMCI (16.3%) and CU (0%) groups. Conversely, the proportion of A+(N)–and A–(N)–subjects was higher in the aMCI and CU groups in comparison to dementia subjects (Table 1).

Demographic details for the four subgroups classified according to presence of Aβ aggregation and neurodegeneration are shown in Table 2. Mean age was higher in A+(N)+ subjects relative to A–(N)–subjects (p < 0.05), with no other significant differences between subgroups.

Aβ-cognition relationships in the overall sample and the four subgroups categorized according to PET biomarker profiles

In the overall sample, A+ subjects presented significantly lower verbal memory (F = 14.88, p < 0.001), visual memory (F = 24.50, p < 0.001), and MMSE scores (F = 12.68, p = 0.001) compared to those with Aβ negativity (A–) (ANCOVA model including age, sex, and years of education as covariates).

There were also significant differences between the four A(N) profiles for verbal memory, visual memory, and global cognition scores (Table 3). Post-hoc Scheffé’s tests indicated that scores on the three measures were lower in the A+(N)+ subgroup compared to the A+(N)–and A–(N)–subgroups. Mean values and levels of inter-subject variability in regard to cognitive test scores were highly similar between the A+(N)–and A–(N)–subgroups (Table 3). The A–(N)+ subgroup had lower values relative to the A–(N)–subgroup in regard to the RAVLT delayed recall and MMSE scores, but not for the SKT memory subscore (Table 3). There was also a significant difference between the A–(N)+ subgroup and the A+(N)–subgroup in regard to MMSE scores, (Table 3).

Using SUVr as a continuous measure of Aβ deposition, significant associations were found in the overall sample for: verbal memory scores (r = –0.37, p < 0.001); visual memory scores (r = 0.50, p < 0.001); and MMSE scores (r = –0.38, p < 0.001).

In regard to Aβ-memory relationships in separate (N) subgroups, correlation indices were not statistically significant in the (N)–subgroup either for verbal memory (r = 0.13, p = 0.154), visual memory (r = –0.15, p = 0.106), or MMSE (r = 0.04, p = 0.486) scores. Conversely, the (N)+ subgroup displayed significant correlation indices for both verbal (r = –0.31, p = 0.029) and visual (r = 0.51, p = 0.001) memory scores, but not for MMSE scores (r = –0.21, p = 0.100).

Finally, there was a significant association between higher Aβ load and lower cognitive test scores in the A+(N)+ subgroup for visual memory performance (r = 0.45, p = 0.008), but not for verbal memory (r = –0.20, p = 0.155) or MMSE scores (r = –0.22, p = 0.129). No significant correlations were detected between higher Aβ load and lower memory scores either in the separate A+(N)–subgroup (r≤0.34, p≥0.118), the separate A–(N)–subgroup (r≤0.11, p≥0.220), or the combined A+(N)–plus A–(N)–subgroups (r≤0.16 p≥0.100).

Scatterplots showing SUVr measurements and visual memory scores regarding to the above correlation analyses are provided in Fig. 1.

OPEN IN VIEWER

Fig. 1

(a) Scatter plot for the overall sample showing visual memory sub-scores of the Short Cognitive Performance Test (SKT) (on the x axis) and [¹¹C]PIB standardized uptake value ratios (SUVRs, on the y axis), obtained using tracer uptake measurements in a meta-region of interest encompassing the prefrontal, orbitofrontal, parietal, temporal, cingulate cortices, and the precuneus, normalized to the tracer uptake in the whole cerebellum. Worse SKT memory performance (as indicated by higher scores) was significantly related to higher values of [¹¹C]PIB SUVr (n = 113, r = 0.50, p < 0.001, partial correlation index controlled for age, sex and years of education). The other graphs show scatter plots for: (b) the subjects rated as presenting neurodegeneration as assessed by [¹⁸F]FDG-PET ((N)+), with significant findings in the partial correlation analysis (n = 41, r = 0.51, p = 0.001); (c) the subjects rated as presenting no neurodegeneration ((N)–), with non-significant partial correlation results (n = 72, r = –0.14, p = 0.122); and (d) the A+(N)+ subgroup (n = 31), with significant findings in the partial correlation analysis (n = 31, r = 0.45, p = 0.008).

Relationships between Aβ aggregation and memory test results in the clinically defined groups

In the cognitively impaired subsample (dementia plus aMCI subjects), A+ subjects presented significantly lower verbal memory and visual memory scores compared to A–subjects (Table 4). When the same comparison was conducted separately within the dementia group, there was a trend towards worse cognitive performance in A+ relative to A–subjects in regard to visual memory scores (p = 0.082) (Table 4). There were no significant memory score differences between Aβ-positive and Aβ-negative subjects either in the aMCI or CU groups (Table 4).

Using SUVr as a continuous measure of Aβ deposition, significant correlations were found in: the cognitively impaired sample (dementia plus aMCI) for both verbal (r = –0.32, p = 0.002) and visual (r = 0.46, p < 0.001) memory scores; and the dementia group for visual memory scores (r = 0.47, p = 0.003). Correlation indices were not statistically significant either in the aMCI group or the CU group for verbal (r < 0.26, p≥0.088) or visual (r < 0.18, p > 0.200) memory scores.