OUP user menu

Effect of amyloid on memory and non-memory decline from preclinical to clinical Alzheimer’s disease

Yen Ying Lim, Paul Maruff, Robert H. Pietrzak, David Ames, Kathryn A. Ellis, Karra Harrington, Nicola T. Lautenschlager, Cassandra Szoeke, Ralph N. Martins, Colin L. Masters, Victor L. Villemagne, Christopher C. Rowe
DOI: http://dx.doi.org/10.1093/brain/awt286 221-231 First published online: 30 October 2013

Summary

High amyloid has been associated with substantial episodic memory decline over 18 and 36 months in healthy older adults and individuals with mild cognitive impairment. However, the nature and magnitude of amyloid-related memory and non-memory change from the preclinical to the clinical stages of Alzheimer’s disease has not been evaluated over the same time interval. Healthy older adults (n = 320), individuals with mild cognitive impairment (n = 57) and individuals with Alzheimer’s disease (n = 36) enrolled in the Australian Imaging, Biomarkers and Lifestyle study underwent at least one positron emission tomography neuroimaging scan for amyloid. Cognitive assessments were conducted at baseline, and 18- and 36-month follow-up assessments. Compared with amyloid-negative healthy older adults, amyloid-positive healthy older adults, and amyloid-positive individuals with mild cognitive impairment and Alzheimer’s disease showed moderate and equivalent decline in verbal and visual episodic memory over 36 months (d’s = 0.47–0.51). Relative to amyloid-negative healthy older adults, amyloid-positive healthy older adults showed no decline in non-memory functions, but amyloid-positive individuals with mild cognitive impairment showed additional moderate decline in language, attention and visuospatial function (d’s = 0.47–1.12), and amyloid-positive individuals with Alzheimer’s disease showed large decline in all aspects of memory and non-memory function (d’s = 0.73–2.28). Amyloid negative individuals with mild cognitive impairment did not show any cognitive decline over 36 months. When non-demented individuals (i.e. healthy older adults and adults with mild cognitive impairment) were further dichotomized, high amyloid-positive non-demented individuals showed a greater rate of decline in episodic memory and language when compared with low amyloid positive non-demented individuals. Memory decline does not plateau with increasing disease severity, and decline in non-memory functions increases in amyloid-positive individuals with mild cognitive impairment and Alzheimer’s disease. The combined detection of amyloid positivity and objectively-defined decline in memory are reliable indicators of early Alzheimer’s disease, and the detection of decline in non-memory functions in amyloid-positive individuals with mild cognitive impairment may assist in determining the level of disease severity in these individuals. Further, these results suggest that grouping amyloid data into at least two categories of abnormality may be useful in determining the disease risk level in non-demented individuals.

  • beta-amyloid
  • neuropsychology
  • mild cognitive impairment
  • Alzheimer’s disease
  • PET imaging

Introduction

Recent prospective studies show that amyloid-β positivity, as detected by PET neuroimaging, is associated with substantial decline in episodic memory decline over 18 and 36 months in individuals who meet clinical criteria for mild cognitive impairment (MCI) as well as in otherwise healthy older adults, even in the absence of any change in clinical disease status (Villemagne et al., 2011; Doraiswamy et al., 2012; Lim et al., 2012, 2013a, b, c; Small et al., 2012; Ellis et al., 2013). In contrast, healthy older adult and MCI groups with low amyloid-β show no deterioration in episodic memory over the same time intervals (Villemagne et al., 2011; Lim et al., 2012, 2013a; Ellis et al., 2013). In MCI, the amyloid-β-related memory decline confirms that the clinical abnormalities observed are indicative of incipient Alzheimer’s disease (Dubois and Albert, 2004; Albert et al., 2011), whereas in healthy older adults, it shows that Alzheimer’s disease-related neurodegeneration can be detected years before individuals meet any clinical staging criteria for early Alzheimer’s disease (Mintun et al., 2006; Aizenstein et al., 2008; Rowe et al., 2010).

Recent pathophysiological models of Alzheimer’s disease show that amyloid-β accumulation slows once individuals meet clinical criteria for Alzheimer’s disease (Jack et al., 2010, 2013; Bateman et al., 2012; Villemagne et al., 2013). Further, although cross-sectional studies report robust negative associations between amyloid-β and memory in MCI, no such associations are observed in Alzheimer’s disease (Pike et al., 2007; Mormino et al., 2008). It is possible, therefore, that when compared to the rate of amyloid-β accumulation observed in non-demented individuals, the rate of cognitive decline may also decrease once clinical criteria for Alzheimer’s disease are met (Jack et al., 2010; Bateman et al., 2012). However, as yet, the nature and magnitude of amyloid-β-related cognitive change in the preclinical and clinical stages of Alzheimer’s disease have not been compared directly. Further, in the majority of studies from the Australian Imaging, Biomarkers and Lifestyle (AIBL) study cohort, amyloid-β-related memory decline has been characterized using a single neuropsychological test of episodic memory (e.g. verbal list learning; Ellis et al., 2013), an episodic memory composite score (Villemagne et al., 2011, 2013), or with a brief computerized cognitive test battery (i.e. CogState brief battery; Lim et al., 2012, 2013a, b). Although these data provide converging evidence that in non-demented individuals, amyloid-β positivity is associated with episodic memory decline, there has been no thorough investigation of the extent to which any amyloid-β-related decline extends to other aspects of cognition (e.g. attention, language and executive function). Further, as decline in cognition is the hallmark of clinically diagnosed Alzheimer’s disease (Almkvist, 1996; Mohs et al., 2000; Minati et al., 2009), it provides a strong basis for comparison of the nature and magnitude of cognitive decline in the preclinical and prodromal stages of the disease.

The aim of this study was to investigate the nature and magnitude of change in a comprehensive range of neuropsychological outcomes over 36 months in individuals with Alzheimer’s disease, patients with amnestic MCI and in healthy older adults with a positive and negative amyloid-β scan. The first hypothesis was that all aspects of cognition would remain stable over 36 months in healthy older adults and in individuals with amnestic MCI who are amyloid-β−. The second hypothesis was that amyloid-β positivity in healthy older adults and in individuals with MCI or Alzheimer’s disease would be associated with a significant decline in memory over 36 months. We then explored the extent to which changes in other aspects of cognition were associated with amyloid-β positivity and whether these changes differed according to clinical group. Further, as previous studies have reported an effect of apolipoprotein (APOE) ɛ4 carriage on cognitive decline (Caselli et al., 2009; Lim et al., 2012; Ellis et al., 2013), we explored whether APOE ɛ4 moderated any cognitive decline associated with amyloid-β positivity. Finally, as we have shown previously that non-demented individuals who have high amyloid-β positivity [e.g. standardized uptake value (SUV) ratio ≥ 1.9] progress to MCI or Alzheimer’s disease at a faster rate than non-demented individuals who have low amyloid-β positivity (e.g. SUV ratio 1.5–1.9) (Rowe et al., 2013a), we explored the extent to which the level of amyloid-β positivity affects rates of cognitive decline over 36 months.

Materials and methods

Participants

All participants in the current study were recruited from the AIBL Study of Ageing (Ellis et al., 2009; Rowe et al., 2010). The process of recruitment and diagnosis classification has been described in detail previously (Ellis et al., 2009). In this study, only healthy older adults and individuals with MCI or Alzheimer’s disease who had undergone amyloid-β imaging with PET and who had completed the AIBL neuropsychological battery at baseline and at 18-month and 36-month follow-up were included. Demographic characteristics of each participant group are shown in Table 1.

View this table:
Table 1

Demographic and clinical characteristics of each participant group

HA-Aβ− (n = 244)HA-Aβ+ (n = 76)MCI-Aβ− (n = 16)MCI-Aβ+ (n = 41)AD-Aβ+ (n = 36)P-value
Female, n (%)126 (51.6)41 (53.9)9 (56.3)19 (46.3)21 (58.3)0.856
APOE ɛ4, n (%)59 (24.2)45 (59.2)1 (6.3)28 (68.3)28 (77.8)<0.001
Age, mean (SD)68.62 (6.10)73.87 (7.32)77.38 (8.69)80.41 (6.73)74.81 (8.46)<0.001
Premorbid IQ, mean (SD)108.15 (7.13)110.00 (6.37)105.00 (11.70)109.05 (7.03)103.47 (8.82)<0.001
HADS depression, mean (SD)2.61 (2.21)2.54 (2.37)3.47 (1.92)3.43 (2.38)3.25 (2.89)0.096
HADS anxiety, mean (SD)4.15 (2.75)4.36 (2.98)5.13 (2.59)4.58 (2.45)4.83 (3.84)0.473
MMSE, mean (SD)28.91 (1.19)28.74 (1.19)27.63 (2.42)27.02 (2.12)21.22 (4.56)<0.001
CDR-SB, mean (SD)0.04 (0.17)0.04 (0.14)0.94 (0.57)1.04 (0.74)4.20 (1.40)<0.001
Verbal Episodic Memory, mean (SD)0.09 (0.96)−0.09 (1.06)−1.67 (1.01)−2.55 (0.74)−3.46 (0.72)<0.001
Visual Episodic Memory, mean (SD)0.00 (1.00)−0.27 (1.08)−1.06 (1.20)−1.79 (1.20)−3.12 (0.98)<0.001
Executive Function, mean (SD)−0.03 (1.02)−0.08 (1.00)−1.36 (0.86)−0.85 (1.19)−1.73 (2.28)<0.001
Language, mean (SD)0.11 (0.98)0.11 (0.89)−1.54 (1.85)−1.16 (1.67)−2.82 (2.09)<0.001
Attention, mean (SD)0.00 (1.00)−0.22 (0.80)−1.09 (1.57)−0.86 (1.32)−2.50 (1.61)<0.001
Visuospatial, mean (SD)0.21 (0.73)0.10 (0.80)−0.60 (1.39)−0.70 (1.47)−3.09 (3.54)<0.001
  • MMSE = Mini-Mental State Examination; CDR-SB = Clinical Dementia Rating Scale, Sum of Boxes; HADS = Hospital Anxiety and Depression Scale.

  • Of the 320 healthy adults who underwent PET neuroimaging, 178 were scanned using 11C-Pittsburgh compound B, 73 using 18F florbetapir, and 69 using 18F flutemetamol. Of the 57 adults with MCI who underwent PET neuroimaging, 49 were scanned using 11C Pittsburgh compound B, four using 18F florbetapir, and four using 18F flutemetamol. Of the 36 adults with Alzheimer’s disease who underwent PET neuroimaging, 35 were scanned using 11C Pittsburgh compound B, and one using 18F florbetapir. Participants who underwent PET neuroimaging using Pittsburgh compound B were scanned at baseline and at each 18 month follow-up time point. Participants who underwent PET neuroimaging using florbetapir or flutemetamol were scanned at the 36 month follow-up time point.

All participants with Alzheimer’s disease met NINCDS-ADRDA criteria for Alzheimer’s disease (McKhann et al., 1984), and in all cases, the clinical review panel (chaired by D.A.) reviewed all available data to ensure that the diagnosis was consistent with these agreed criteria. Similarly, all available data for participants with MCI were reviewed to ensure that their classification was consistent with internationally agreed criteria (Petersen et al., 1999; Winblad et al., 2004). For participants with Alzheimer’s disease, an additional inclusion criterion was a score of 18 to 26 on the Mini-Mental State Examination (Folstein et al., 1975). All participants with Alzheimer’s disease and MCI received a Clinical Dementia Rating scale sum of boxes score and a total score (Morris, 1983). Exclusion criteria at baseline were: schizophrenia; depression (Geriatric Depression Score of ≥6); Parkinson’s disease; cancer (other than basal cell skin carcinoma) within the past 2 years; symptomatic stroke; uncontrolled diabetes; or current regular alcohol use exceeding two standard drinks per day for women or four per day for men. Clinical classification was blinded to amyloid-β imaging data.

The study was approved by and complied with the regulations of the institutional research and ethics committees of Austin Health, St. Vincent’s Health, Hollywood Private Hospital and Edith Cowan University (Ellis et al., 2009). All participants provided written informed consent before participating in the study.

Measures

Positron emission tomography neuroimaging and APOE ɛ4 genotyping

Amyloid-β imaging with PET was conducted using either 11C-Pittsburgh compound B (Pittsburgh compound B), 18F-florbetapir or 18F-flutemetamol. PET methodology has been described in detail previously (Rowe et al., 2010; Vandenberghe et al., 2010; Wong et al., 2010). Given the different pharmacokinetic characteristics, a different acquisition protocol was adopted for each tracer. Thirty minute acquisitions were started 40 min after injection of Pittsburgh compound B, 20-min acquisitions were performed 50 min after injection of florbetapir and 90 min after injection of flutemetamol. For Pittsburgh compound B and flutemetamol, PET SUV data were summed and normalized to the cerebellar cortex SUV, resulting in a region-to-cerebellar ratio termed SUV ratio. For florbetapir, the SUV ratio was generated using the whole cerebellum as the reference region (Clark et al., 2011). In line with previous studies, the SUV ratio was classified dichotomously as either negative or positive. For Pittsburgh compound B and flutemetamol, a SUV ratio threshold ≥1.5 was used (Rowe et al., 2010; Vandenberghe et al., 2010). In the case of florbetapir, based on the results of a phase III study (Clark et al., 2011), an SUV ratio threshold of ≥1.1 was used to discriminate between amyloid-β− and amyloid-β+. As participants were scanned at different follow-up time points (Table 1), we classified participants as amyloid-β− or amyloid-β+ based on the SUV ratio that was obtained closest to their 36 month follow-up time point.

An 80 ml blood sample was also taken from each participant, 0.5 ml of which was forwarded for APOE genotyping at a clinical pathology laboratory.

Cognitive and clinical assessments

All participants were assessed with the clinical rating scales and neuropsychological battery from the AIBL study which have been described in detail elsewhere (Ellis et al., 2009; Rowe et al., 2010). The clinical status of participants was determined by data which included the Mini-Mental State Examination (Folstein et al., 1975), and the Clinical Dementia Rating scale (Morris, 1983). Premorbid intelligence was estimated using the Wechsler Test of Adult Reading (Wechsler, 2001), and levels of depressive and anxiety symptoms were assessed using the Hospital Anxiety and Depression (Snaith and Zigmond, 1986).

Procedure

Participants underwent an extensive medical, psychiatric, and neuropsychological assessment upon enrolment into the AIBL study. The same assessments were repeated 18 and 36 months after baseline. In this study, we report PET neuroimaging and APOE ɛ4 genotyping data obtained at baseline, and neuropsychological data obtained at baseline, 18 months and 36 months to examine the rate of cognitive change in relation to baseline levels of amyloid-β.

Data analysis

Individual outcome measures on individual tests were standardized against the baseline mean and standard deviation for the healthy older adult group, and then averaged to compute a cognitive composite score for verbal episodic memory [Logical Memory delayed recall, California Verbal Learning Test, Second Edition (CVLT-II) long delay recall, and CVLT-II d’]; visual episodic memory [Rey Complex Figure Test 30 minute delayed recall, CogState One Card Learning task, and CogState One Back task); executive function [Stroop Colours/Dots, Letter Fluency, and Category Fluency Switching (Fruit/Furniture)]; language [Category Fluency (Animals/Boys’ Names) and Boston Naming Test]; attention (Digit Symbol, CogState Detection task, and CogState Identification task); and visuospatial function (Rey Complex Figure Test Copy, and Clock Drawing). The process of selecting cognitive tasks for the formation of each composite score, and the validation of each cognitive composite score has been described in detail previously (Harrington et al., 2013).

A series of repeated measures linear mixed model analyses (using maximum likelihood estimation and an unstructured covariance matrix) were conducted to examine the relation between group [healthy older adult amyloid-β− (HA-Aβ−), healthy older adult amyloid-β+ (HA-Aβ+), MCI amyloid-β− (MCI-Aβ−), MCI amyloid-β+ (MCI-Aβ+) and Alzheimer’s disease amyloid-β+ (AD-Aβ+)] and time (baseline, 18 month, and 36 month) on cognitive change. Linear mixed modelling was used because of its ability to model both fixed and random effects, which accounts for multiple sources of variability, and because it provides improved estimates of within-subject coefficients (i.e. random effects) in longitudinal studies. In these analyses, group, time, APOE status (ɛ4 carrier, ɛ4 non-carrier), and the group × time interaction were entered as fixed factors; participant as a random factor; age and premorbid intelligence as covariates; and cognitive composite score as the dependent variable. For each cognitive composite score, mean slope estimates were computed for each group. The magnitude of difference in the rates of change (i.e. slopes) of the HA-Aβ+, MCI-Aβ−, MCI-Aβ+, and AD-Aβ+ groups in relation to the HA-Aβ− group was expressed using Cohen’s d and 95% confidence intervals (CIs) (Cohen, 1988).

To investigate whether there was an effect of APOE ɛ4 on amyloid-β-related change in cognition, and to maximize the power to detect any subtle relationships, participants were classified into amyloid-β− (HA-Aβ−, MCI-Aβ−) and amyloid-β+ (HA-Aβ+, MCI-Aβ+) groups. The AD-Aβ+ group was excluded from this analysis, as there were no individuals with Alzheimer’s disease that were amyloid-β−.

Linear mixed model analyses were then conducted to examine the relationship between group (amyloid-β−, amyloid-β+), APOE (ɛ4 carrier, ɛ4 non-carrier), and time (baseline, 18 months, and 36 months) on change in each cognitive composite score. In these analyses, only main effects or interactions involving APOE were interpreted.

To determine whether level of amyloid-β positivity (i.e. high amyloid-β+ or low amyloid-β+) was associated with increased rates of cognitive decline in non-demented individuals, we combined the HA-Aβ+ and MCI-Aβ+ groups. Receiver operating characteristic analysis of the AIBL Alzheimer’s disease and healthy older adult cohorts were also performed (Rowe et al., 2013a). This indicated that for Pittsburgh compound B, an SUV ratio of 1.9 was the optimal cut-off point for distinction of individuals with Alzheimer’s disease from age-matched healthy older adult. Thus, this higher SUV ratio cut-off point was used to define high amyloid-β+ and low amyloid-β+ (SUV ratio 1.5-1.9). Similarly, for flutemetamol, an SUV ratio threshold ≥2.19 was used to discriminate between high amyloid-β+ and low amyloid-β+ (SUV ratio 1.5–2.19) non-demented individuals, and for florbetapir, the SUV ratio threshold used was ≥1.29. Linear mixed model analyses were conducted to examine the relationship between amyloid-β positivity (low amyloid-β+ versus high amyloid-β+) and time (baseline, 18 months and 36 months) on change in each cognitive composite score.

Results

Demographic differences between amyloid-β− and amyloid-β+ subgroups in healthy older adults, mild cognitive impairment and Alzheimer’s disease

There were statistically significant differences between the five groups in age and premorbid intelligence. In particular, the MCI-Aβ− and AD-Aβ+ groups had significantly lower premorbid intelligence when compared with the HA-Aβ−, HA-Aβ+ and MCI-Aβ+ groups, P < 0.001. The HA-Aβ− and HA-Aβ+ groups were also significantly younger than the MCI-Aβ− and MCI-Aβ+ groups, P < 0.05. Groups did not differ on symptoms of depression or anxiety (Table 1). There were significantly more APOE ɛ4 carriers in both the HA-Aβ+ and MCI-Aβ+ groups.

Comparison of rates of cognitive change in HA-Aβ+, MCI-Aβ−, MCI-Aβ+ and AD-Aβ+ groups relative to the HA-Aβ− group

Group × time interactions were statistically significant for all cognitive composite scores (Supplementary Table 1). The mean slope for each clinical group for each composite is given in Table 2. The magnitude of the difference in slopes for each group from that of the HA-Aβ− group is presented for each composite score in Fig. 2. In Fig. 2, the 95% CIs presented for each effect size also allows interpretation of significant differences (i.e. where 95% CIs do not overlap).

View this table:
Table 2

Mean slopes and standard deviation of the slopes for each participant group, and the magnitude of difference in slopes of HA-Aβ+, MCI-Aβ−, MCI-Aβ+ and AD-Aβ+groups from the HA-Aβ−group

Mean slope (SD)Cohen’s d (95% CIs)
CompositeHA-Aβ− (n = 244)HA-Aβ+ (n = 76)MCI-Aβ− (n = 16)MCI-Aβ+ (n = 41)AD-Aβ+ (n = 36)HA-Aβ− versus HA-Aβ+HA-Aβ− versus MCI-Aβ−HA-Aβ− versus MCI-Aβ+HA-Aβ− versus AD-Aβ+
Verbal Episodic Memory0.019 (0.415)−0.185 (0.369)0.274 (0.278)−0.253 (0.323)−0.278 (0.358)0.50−0.620.670.73
(0.24–0.76)(−1.13–−0.11)(0.34–1.01)(0.37–1.08)
Visual Episodic Memory0.077 (0.559)−0.191 (0.494)0.019 (0.374)−0.259 (0.439)−0.334 (0.539)0.490.110.620.74
(0.23–0.75)(−0.40–0.61)(0.28–0.95)(0.38–1.09)
Executive Function−0.124 (0.627)−0.011 (0.553)0.141 (0.418)−0.218 (0.482)−0.768 (0.513)−0.19−0.430.151.05
(−0.44–0.07)(−0.94–0.08)(−0.18–0.49)(0.69–1.41)
Language−0.03 (0.490)−0.150 (0.435)−0.172 (0.332)−0.568 (0.379)−0.739 (0.408)0.250.291.131.48
(−0.01–0.51)(−0.21–0.80)(0.78–1.47)(1.10–1.84)
Attention−0.123 (0.473)−0.152 (0.420)0.010 (0.318)−0.341 (0.376)−0.543 (0.461)0.06−0.290.470.89
(−0.19–0.32)(−0.79–0.22)(0.14–0.81)(0.53–1.25)
Visuospatial−0.311 (0.734)−0.465 (0.654)−0.372 (0.489)−0.946 (0.566)−2.022 (0.621)0.220.080.892.37
(−0.04–0.47)(−0.42–0.59)(0.55–1.23)(1.97–2.77)
  • Values in bold are significant at the P < 0.05 or P < 0.001 level.

Post hoc comparison of group mean slopes over 36 months indicated that, relative to the HA-Aβ− group, the HA-Aβ+ group showed a significantly greater rate of decline over 36 months on the verbal episodic memory, and visual episodic memory composites, with the magnitude of these differences, by convention, moderate (Cohen, 1988) (Table 3 and Fig. 1A and B). No differences in group mean slopes were observed between the HA-Aβ− and HA-Aβ+ groups on any of the other cognitive composite scores. Similarly, relative to the HA-Aβ− group, the MCI-Aβ+ group showed a greater rate of decline over 36 months on the verbal episodic memory, and visual episodic memory composites, and also decline on the language, attention, and visuospatial function composites (Table 2 and Fig. 2). The difference in group mean slopes was by convention, moderate-to-large in magnitude. Relative to the =HA-Aβ− group, the AD-Aβ+ group showed greater rates of decline over 36 months on all cognitive composite measures, and the magnitude of difference in group mean slopes was large (Table 2 and Fig. 2). Of note, the MCI-Aβ− group showed no decline on any of the cognitive composite scores relative to the HA-Aβ− group, rather, they showed improvement on the verbal episodic memory composite with this improvement, relative to the HA-Aβ− group, moderate-to-large in magnitude (Table 2).

Figure 1

Linear trend of performance on the verbal memory composite (A) and the visual memory composite (B) for HA-Aβ−, HA-Aβ+, MCI-Aβ−, MCI-Aβ+, and AD-Aβ+ groups, from baseline to 36 months.

Figure 2

Magnitude of difference (Cohen’s d) in the rate of change in each cognitive composite score between the HA-Aβ+, MCI-Aβ+, and AD-Aβ+ groups relative to the HA-Aβ− group (represented by ‘0’ line). Error bars represent the 95% CIs of the difference in the rate of cognitive change.

View this table:
Table 3

Effect of APOE on the relationship between amyloid-β and cognitive change over time across all participant groups

AgeP- valueIQP- valueAPOEP- valueAmyloid-βP- valueTimeP- valueAmyloid -β × APOEP- valueAmyloid-β × TimeP- valueAPOE × TimeP- valueAmyloid-β × APOE × TimeP-value
Verbal Episodic Memory(1,376) 82.44<0.001(1,376) 24.58<0.001(1,378) 0.020.884(1,407) 2.650.104(1,366) 9.220.003(1,376) 0.510.474(1,366) 16.94<0.001(1,366) 4.880.028(1,366) 3.190.075
Visual Episodic Memory(1,377) 91.91<0.001(1,373) 19.94<0.001(1,380) 4.400.037(1,415) 0.220.642(1,370) 3.920.048(1,379) 1.810.179(1,370) 18.19<0.001(1,370) 0.000.968(1,370) 0.200.656
Executive Function(1,373) 44.65<0.001(1,371) 47.35<0.001(1,370) 1.970.161(1,406) 0.370.541(1,354) 7.980.005(1,369) 0.370.544(1,354) 0.040.848(1,354) 0.000.954(1,354) 0.770.381
Language(1,377) 78.28<0.001(1,377) 64.62<0.001(1,374) 0.520.470(1,393) 5.060.025(1,357) 38.57<0.001(1,373) 0.070.790(1,357) 26.61<0.001(1,357) 0.010.945(1,357) 0.220.639
Attention(1,378) 83.30<0.001(1,375) 46.43<0.001(1,377) 0.140.707(1,412) 0.420.515(1,353) 30.55<0.001(1,375) 0.870.351(1,354) 4.040.045(1,354) 0.410.520(1,354) 0.000.974
Visuospatial(1,373) 20.40<0.001(1,373) 22.17<0.001(1,336) 0.420.516(1,371) 1.850.175(1,360) 152.01<0.001(1,335) 0.160.690(1,360) 22.15<0.001(1,360) 0.730.394(1,360) 0.040.838
  • Values in bold are significant at the P < 0.05 or P < 0.001 level. The values in brackets are degrees of freedom, and the values NOT in brackets are F values, denoted as (df) F.

The effect of APOE ɛ4 on the relationship between amyloid-β and cognitive change

Analysis of the effect of APOE ɛ4 carriage on the relationship between amyloid-β and cognitive change over 36 months showed no significant interactions for any cognitive composite score (Table 3). The only statistically significant effect identified from this analysis was the main effect of APOE on cognitive change over 36 months for the verbal memory composite score (Table 3).

The effect of low or high amyloid-β positivity on cognitive change in non-demented individuals

Amyloid-β positivity × time interactions were statistically significant for verbal and visual episodic memory and language composites (Table 4). Relative to low amyloid-β+ non-demented individuals, high amyloid-β+ non-demented individuals showed a significantly greater rate of decline over 36 months that was moderate in magnitude, for the verbal episodic memory, visual episodic memory, and language composites only (Table 4). This analysis was then repeated in only the HA-Aβ+ group. When compared with the low amyloid-β+ healthy older adult group, the high amyloid-β+ healthy older adult showed significantly greater rate of decline over 36 months that was moderate in magnitude for the verbal episodic memory [d (95% CI) = 0.44 (0.00–0.89)] and visual episodic memory [d (95% CI) = 0.73 (0.27–1.18)] composites.

View this table:
Table 4

Results of linear mixed model analyses examining change in cognitive performance over 36 months in high and low amyloid-β+ non-demented individuals, mean slopes and standard deviation of the slopes for high and low amyloid-β+ groups, and Cohen’s d of the difference between slopes

CovariatesFixed factorsLow amyloid-β+ (n = 46)High amyloid-β+ (n = 73)
CompositeAgeP-valueIQP-valueAmyloid-β positivityP-valueTimeP-valueAmyloid-β positivity × TimeP-valueMean (SD)Mean (SD)Cohen’s d (95%CI)
Verbal Episodic Memory(1,117) 43.92<0.001(1,117) 6.380.013(1,114) 0.710.401(1,102) 16.51<0.001(1,102) 6.820.010−0.062 (0.458)−0.287 (0.656)0.38
(0.01–0.75)
Visual Episodic Memory(1,114) 29.09<0.001(1,116) 4.880.029(1,114) 0.060.803(1,100) 9.370.003(1,100) 7.940.006−0.015 (0.651)−0.360 (0.932)0.41
(0.04–0.78)
Executive Function(1,116) 12.660.001(1,117) 14.34<0.001(1,102) 0.390.533(1,100) 1.710.194(1,100) 0.000.967−0.075 (0.634)−0.080 (0.907)0.01
(−0.36–0.38)
Language(1,115) 30.84<0.001(1,116) 11.320.001(1,115) 2.510.116(1,101) 25.00<0.001(1,101) 5.470.021−0.143 (0.573)−0.394 (0.819)0.34
(0.00–0.68)
Attention(1,116) 42.51<0.001(1,116) 26.37<0.001(1,105) 0.010.934(1,88) 20.38<0.001(1,88) 0.160.691−0.193 (0.499)−0.231 (0.715)0.06
(−0.31–0.43)
Visuospatial(1,113) 10.330.002(1,114) 5.240.024(1,95) 1.260.264(1,107) 49.09<0.001(1,107) 2.630.108−0.476 (0.947)−0.762 (1.351)0.24
(−0.14–0.60)
  • Values in bold are significant at the P < 0.05 or P < 0.001 level; amyloid-β positivity indicates main effect of membership in the low amyloid-β+ (i.e. SUV ratio 1.5–1.9) or high amyloid-β+ (SUV ratio >1.9) groups. The values in brackets are degrees of freedom, and the values NOT in brackets are F values, denoted as (df) F.

Discussion

The first hypothesis that all aspects of cognitive function would remain stable in healthy older adults and individuals with MCI who were amyloid-β− was supported partially. Specifically, amyloid-β− healthy older adults showed no change in verbal or visual episodic memory (Fig. 1) or any other aspect of cognitive function over 36 months (Table 2). However, amyloid-β− individuals with MCI showed improvement in verbal memory over 36 months that was, by convention (Cohen, 1988), moderate-to-large in magnitude (Table 2 and Fig. 1A). The absence of decline in memory or any other cognitive function over 36 months in amyloid-β− healthy older adults replicates and extends previous observations of cognitive stability over 6, 18 and 36 months in negative amyloid-β healthy older adults from AIBL (Villemagne et al., 2011; Lim et al., 2012, 2013b; Ellis et al., 2013) and other cohorts (Doraiswamy et al., 2012; Small et al., 2012; Snitz et al., 2013). In individuals with amyloid-β− MCI, the finding of no cognitive decline is also consistent with the results of previous 18 and 36 month studies (Doraiswamy et al., 2012; Lim et al., 2013a, b), which converge to suggest that even when characterized clinically within a highly specialized Alzheimer’s disease memory clinic, MCI without a positive amyloid-β biomarker most likely reflects neurological or psychiatric conditions other than Alzheimer’s disease (Dubois and Albert, 2004). The moderate-to-large improvements in verbal episodic memory over 36 months observed in the amyloid-β− MCI group suggests that the abnormality in memory observed at the baseline assessment may have reflected, at least in part, the effects of situationally-raised levels of depression or anxiety that can occur when individuals with memory difficulties are confronted with formal cognitive tests (Beaudreau and O'Hara, 2008), the effects of subclinical mental health, sleep, or physical health. These factors may have then resolved with experience in the AIBL study, and consequently the magnitude of memory impairment was reduced. However by themselves, these factors cannot account fully for the baseline memory impairment in the amyloid-β− MCI group as all individuals in this group were still classified as meeting clinical criteria for MCI by a consensus panel after 36 months in the study.

The second hypothesis that in healthy older adults, individuals with MCI and individuals with Alzheimer’s disease, amyloid-β positivity would be associated with decline in memory over 36 months was supported. For amyloid-β+ healthy older adults, the rate of decline in verbal and visual episodic memory over 36 months was moderate in magnitude when compared with memory changes in amyloid-β− healthy older adults (Fig. 1). No difference between amyloid-β− or positive healthy older adults was observed in the rate of change for any cognitive function other than memory. These data confirm previous examinations of the current sample conducted over shorter time intervals or with other measures of memory (Lim et al., 2012, 2013a, b). They are also consistent with observations made in other prospective studies of older individuals who have elevated amyloid-β on PET neuroimaging to suggest that specific memory decline is the hallmark of preclinical Alzheimer’s disease (Darby et al., 2011; Villemagne et al., 2011; Doraiswamy et al., 2012; Small et al., 2012; Snitz et al., 2013). When compared with amyloid-β− healthy older adults, individuals with amyloid-β+ MCI also showed greater rates of decline in verbal and visual episodic memory, with the magnitude of the decline moderate to large by convention (Fig. 2). However, in amyloid-β+ MCI, decline was also observed for language, attention and visuospatial function, with the magnitude of decline in these other domains of cognition equivalent to that observed for the decline in episodic memory (Fig. 2). Finally, when compared with amyloid-β− healthy older adults, amyloid-β+ patients with Alzheimer’s disease showed large decline across all areas of cognitive function (Fig. 2) with the largest decline observed for language and visuospatial function.

The observation that there was substantial overlap between the confidence intervals for the magnitude of decline for amyloid-β+ healthy older adult, MCI and Alzheimer’s disease groups (Fig. 2) suggest that the rate of decline in verbal and visual episodic memory was equivalent between these groups over 36 months. Among amyloid-β+ individuals who met clinical criteria for MCI, decline in cognition extended from memory to language, attention and visuospatial functions, with the magnitude of this decline large but equivalent across these cognitive domains. In some of our own previous studies of non-demented individuals with elevated amyloid-β (Villemagne et al., 2011, 2013), we have categorized cognitive functions as either memory or non-memory and observed that in general, amyloid-β-related cognitive change in non-demented older individuals was specific to episodic memory. However, results of the current study indicate that when examined more carefully, clinically significant decline occurs in cognitive domains other than memory, most notably in language and visuospatial function, among individuals with amyloid-β+ MCI. One interesting aspect of these data is that no decline was observed for executive function in individuals with amyloid-β+ MCI. This result contrasts with prior suggestions that impairment in executive function is an important characteristic of early Alzheimer’s disease (Baudic et al., 2006; Traykov et al., 2007). One reason for this may be that we classified individuals with early Alzheimer’s disease based on the clinical classification of MCI with additional biomarker confirmation (i.e. amyloid-β positivity), whereas previous studies that have found impairment in executive function were conducted only in clinically classified MCI. Further, several authors have suggested that the presence of impairment in executive function in individuals with MCI may be indicative of a non-Alzheimer’s disease aetiology (Dubois and Albert, 2004; Petersen, 2004).

Recent pathophysiological models of Alzheimer’s disease suggest that the rate of amyloid-β accumulation usually slows after individuals meet clinical criteria for Alzheimer’s disease (Jack et al., 2010, 2013; Villemagne et al., 2013). Thus, when the low amyloid-β levels of younger adults are considered, the temporal relationship between amyloid-β accumulation and age across the lifespan tends to become sigmoidal in nature (Jack et al., 2010; Bateman et al., 2012; Villemagne et al., 2013). These same models suggest that like amyloid-β accumulation, the rate of cognitive decline may also reach a plateau once individuals meet clinical criteria for Alzheimer’s disease, especially given that once individuals meet clinical criteria for Alzheimer’s disease, while the association between amyloid-β burden and cognitive performance is weak or non-existent, there is a high association between the rates of amyloid-β accumulation and the rates of cognitive decline, probably reflecting the simultaneous or parallel slowing of both processes (Jack et al., 2013; Villemagne et al., 2013). Results of the current study show that despite meeting clinical criteria for Alzheimer’s disease, decline in memory continues at the same rate as that observed for non-demented amyloid-β+ older individuals. However, in these same individuals who are amyloid-β+ and have Alzheimer’s disease, rates of decline in all non-memory domains are substantially increased when compared to those observed for amyloid-β+ healthy older adults. Taken together, as decline in memory is characteristic of Alzheimer’s disease even in the earliest preclinical stage (Braak and Braak, 1991; Sperling et al., 2011), memory decline in Alzheimer’s disease also persists at the same rate. However, as the disease progresses, other cognitive domains become affected and therefore, the rate of decline in the non-memory domains is more pronounced when compared to that observed in the earlier stages of the illness (i.e. preclinical stage) where there is little or no decline in these same non-memory functions. This suggests that the assessment of cognitive domains in addition to memory may provide greater insight into the disease progression of individuals once they meet clinical criteria for Alzheimer’s disease.

The APOE ɛ4 allele did not moderate the relationship between amyloid-β and decline for any cognitive domain in individuals who are amyloid-β− (amyloid-β− healthy older adults and individuals with MCI) or individuals who are amyloid-β+ (amyloid-β+ healthy older adults and individuals with MCI) (Table 4). We have reported previously that the APOE ɛ4 allele did not moderate amyloid-β-related decline in memory in studies conducted over shorter time intervals (i.e. 6 and 18 months) (Lim et al., 2012, 2013c; Ellis et al., 2013). In these previous studies, inclusion of APOE ɛ4 status (ɛ4 carrier or non-carrier) and amyloid-β status (negative or positive) in our statistical models has resulted in decline being observed for both APOE ɛ4 and amyloid-β status (Lim et al., 2012; Ellis et al., 2013); however, the strength of APOE ɛ4-related cognitive decline has always been substantially less than that observed for amyloid-β. In the current study, we observed this effect again in the combined sample of healthy older adults and individuals with MCI, assessed over a longer period of time than has been done previously. The only effect of APOE ɛ4 status observed was that for decline in verbal episodic memory. Taken together, these data support the hypothesis that while APOE ɛ4 carriage is a risk factor for amyloid-β positivity, once this occurs, APOE ɛ4 does not moderate disease progression, at least at a level which can be detected by cognitive assessments.

In healthy older adults, SUV ratio has been consistently shown to have a skewed distribution, as opposed to the normal distribution typically observed in clinical and cognitive variables (Jack et al., 2008; Rowe et al., 2010). As such, in order to parametrically determine associations between amyloid-β levels and cognitive markers, we and others have separated amyloid-β into two categories (positive and negative). However, as sample sizes of individuals who had undergone neuroimaging have grown, there is now some evidence to suggest that additional prognostic information can be derived from the degree of amyloid-β positivity. Recently, we showed that when a receiver operating characteristic analysis of healthy older adult and Alzheimer’s disease groups was conducted, an SUV ratio of ≥1.90 was the optimal cut-off for a diagnosis of Alzheimer’s disease (Rowe et al., 2013a). Further, a high amyloid-β+ scan (e.g. SUV ratio>1.90) in non-demented individuals (i.e. healthy older adult and individuals with MCI) was associated with significantly higher positive predictive rates of progression to the next disease stage than a low amyloid-β+ scan (e.g. SUV ratio 1.50–1.90) in non-demented individuals (Rowe et al., 2013a). These data suggest that there is some dose-dependent effect of the level of amyloid-β positivity on memory decline in non-demented individuals. For example, non-demented individuals with higher levels of amyloid-β positivity may be closer in time to a clinical diagnosis of Alzheimer’s disease (Rowe et al., 2013a; Villemagne et al., 2013). In accord with this previous observation, the results of this study show that high amyloid-β+ non-demented individuals show greater rates of decline in episodic memory and language when compared to low amyloid-β+ individuals. Taken together, these results suggest that grouping data into at least two categories of abnormality may be useful in determining the disease risk level in non-demented individuals.

In this study, we have reported estimates of the mean and standard deviation of the slope for the relationship between amyloid-β and each cognitive composite score over 36 months for all stages of the disease. The finding that amyloid-β positivity is associated with cognitive decline across all stages of Alzheimer’s disease accords with drug development strategies that have aimed at reducing the effect of amyloid-β on the brain. Therefore, the slope estimates presented here may serve as a guide to inform the rate of amyloid-β-related decline in cognitive function that would be expected to occur in placebo groups at each stage of the illness. Further, they could be used to estimate effect sizes and sample sizes that would be required to provide sufficient statistical power to detect the extent to which amyloid-β- modifying drugs can halt or delay amyloid-β-related cognitive decline.

An important caveat when interpreting the results of this study is that the AIBL study is not an epidemiological but a convenience sample. The selection of MCI groups was biased towards the inclusion of individuals with amnestic MCI. Further, in the recruitment of healthy older adults, participants in AIBL were highly educated, and few had existing or untreated medical, neurological, or psychiatric illnesses. As such, it would be important for these findings to be replicated in amyloid-β+ individuals in population-based samples, such as the Mayo Clinic Study of Aging (Roberts et al., 2008), where it is possible that amyloid-β-related decline in cognition may be greater than that observed here. A second caveat is that participants who underwent 18F-florbetapir or 18F-flutemetamol PET imaging were scanned only at their 36-month follow-up assessment, and the number of individuals who underwent neuroimaging using the three different compounds was uneven (Table 1). Importantly though, 11C-PiB, 18F-florbetapir and 18F-flutemetamol are commonly used to measure use Aβ levels (Clark et al., 2011; Rowe et al., 2013b; Vandenberghe et al., 2010), and we have shown previously that accumulation of Aβ in HA and MCI groups occur at a rate of less than 0.05 per year (Villemagne et al., 2013). Finally, we did not consider whether healthy older adult and MCI amyloid-β+ groups progressed to the next disease classification at a faster rate than amyloid-β− individuals. This was because the primary aim of this study was to consider the disease as a continuous process, and to determine the rate of cognitive decline associated with amyloid-β positivity in each clinical group. Further, previous studies of the same group have shown that after even after removal of individuals for whom the disease had progressed to the next stage (i.e. from healthy to MCI, or from MCI to Alzheimer’s disease), decline in memory remained of equivalent magnitude (Lim et al., 2012, 2013b; Ellis et al., 2013), suggesting that this memory decline is a characteristic of individuals with amyloid-β+ groups, and not merely a consequence of amyloid-β+ groups containing more individuals for whom the disease had progressed.

These limitations notwithstanding, results of this study suggest that the combined detection of a positive biomarker for Alzheimer’s disease and objectively-defined decline in memory are reliable indicators of the earliest stages of Alzheimer’s disease, and that amyloid-β positivity in individuals with MCI provides additional confirmation that the underlying aetiology is due to Alzheimer’s disease. Importantly, these findings suggest that amyloid-β+ healthy older adults may be promising candidates for clinical trials aiming to modify or halt the progression of Alzheimer’s disease in its very early stages.

Funding

Funding for the study was provided in part by the study partners [Australian Commonwealth Scientific Industrial and Research Organization (CSIRO), Edith Cowan University (ECU), Mental Health Research Institute (MHRI), Alzheimer’s Australia (AA), National Ageing Research Institute (NARI), Austin Health, CogState Ltd., Hollywood Private Hospital, Sir Charles Gardner Hospital]. The study also received support from the National Health and Medical Research Council (NHMRC) and the Dementia Collaborative Research Centres program (DCRC2), as well as ongoing funding from the Science and Industry Endowment Fund (SIEF).

Conflict of interest

YYL, KE, KH, and CLM report no conflicts of interests. PM is a full-time employee of CogState Ltd., the company that provided the CogState tests used in this study. RHP is a scientific consultant to CogState Ltd. DA has served on scientific advisory boards for Novartis, Eli Lilly, Janssen, Prana and Pfizer Inc.; has received funding for travel from Janssen and Pfizer Inc., has served as Editor-in-Chief for International Psychogeriatrics; has received speaker honoraria from Pfizer Inc. and Lundbeck Inc.; and has received research support from Eli Lilly and Company, GlaxoSmithKline, Forest Laboratories Inc., Novartis, and CSIRO. NTL is the current Editor-in-Chief for International Psychogeriatrics, receives research support from NHMRC, has served on scientific advisory boards for Lundbeck and Novartis, has served as a consultant for Schwabe and has received speaker honoraria and travel support from Pfizer, Novartis and Lundbeck. RM is a consultant to a small biotech company, Alzhyme, although currently receives no financial benefits or fees; also holds stock for Alzhyme which is not listed. VV serves as a consultant for Bayer Schering Pharma; and receives research support from NHMRC and NEDO, Japan. CR previously served on scientific advisory boards for Bayer Schering Pharma, Elan Corporation, and AstraZeneca; has received speaker honoraria from GE Healthcare and Bayer Schering Pharma; and receives research support from GE Healthcare, Bayer, Piramal, Avid Radiopharmaceuticals, SIEF, NHMRC, DCRC, an anonymous Foundation and the Alzheimer’s Association.

Supplementary material

Supplementary material is available at Brain online.

Acknowledgements

The AIBL investigators thank Richard Head of CSIRO for initiating and facilitating the AIBL collaboration. Alzheimer’s Australia (Victoria and Western Australia) assisted with promotion of the study and the screening of telephone calls from volunteers. The AIBL team wishes to thank the clinicians who referred patients with Alzheimer’s disease to the study: Associate Professor Brian Chambers, Professor Edmond Chiu, Dr Roger Clarnette, Dr Mary Davison, Dr John Drago, Dr Peter Drysdale, Dr Jacqueline Gilbert, Dr Kwang Lim, Dr Dina LoGiudice, Dr Peter McCardle, Dr Steve McFarlane, Dr Alastair Mander, Dr John Merory, Professor Daniel O’Connor, Dr Ron Scholes, Dr Mathew Samuel, Dr Darshan Trivedi, and Associate Professor Michael Woodward. We thank all those who participated in the study for their commitment and dedication to helping advance research into the early detection and causation of Alzheimer’s disease.

Abbreviations
AIBL
Australian Imaging, Biomarkers and Lifestyle
AD-Aβ+
Alzheimer’s disease amyloid-β+
HA-Aβ+/−
healthy older adult amyloid-β+/negative
MCI-Aβ+/−
mild cognitive impairment amyloid-β+/negative
SUV
standardized uptake value

References

View Abstract