Relationship Between Amyloid-β Positivity and Progression to Mild Cognitive Impairment or Dementia over 8 Years in Cognitively Normal Older Adults

INTRODUCTION

Clinical-pathological models propose that Alzheimer’s disease (AD) begins with accumulation of amyloid-β (Aβ) followed by aggregation of tau, which result in cortical atrophy, cognitive decline and, ultimately, dementia [1, 2]. Biomarker studies using positron emission tomography (PET) neuroimaging or cerebrospinal fluid (CSF) sampling show that Aβ accumulation begins up to 20 years prior to the onset of dementia [3]. According to these models, cognitively normal (CN) older adults with elevated levels of Aβ (Aβ+) are in the preclinical stages of AD [2, 4]. Even though CN Aβ+ individuals are clinically asymptomatic, preclinical AD is characterized by subtle progressive cognitive decline, primarily in episodic memory [5], which may reflect insidious loss of cortical brain volume due to Aβ+ [6–8]. Prospective studies in both experimental and epidemiological cohorts have indicated that CN Aβ+ adults have higher risk of progression to clinical classification of mild cognitive impairment (MCI) or dementia compared to those without elevated Aβ (i.e., Aβ–) [9, 10]. Clinical trials of Aβ-lowering drugs have endeavored to recruit CN Aβ+ older adults based on their increased risk for incident MCI/dementia in an attempt to slow cognitive decline and prevent onset of symptomatic AD (e.g. Clinical Trials NCT02569398 and NCT02008357) [11]. Furthermore, recently proposed guidance from the US Food and Drug Administration (FDA) acknowledges the centrality of biomarkers such as Aβ+ for recruitment of participants for early AD clinical trials, where Stage 0 includes individuals who are asymptomatic but have evidence of AD pathophysiologic changes [12]. Understanding the risk of progression to MCI/dementia associated with Aβ+ over the long time periods characteristic of preclinical AD is therefore necessary to inform models of disease etiology and guide recruitment and outcome expectations in clinical trials or clinical studies of preclinical AD and AD.

Incidence of progression to MCI/dementia in large samples of Aβ+ CN adults ranges from 17.7% over an average of 3.7 years in the Mayo Clinic Study of Aging (MCSA, mean age 76) [10], 26.4% over 5 years in the Knight Alzheimer’s Disease Research Center study (Knight ADRC, mean age 72) [13], 19% over 6 years in the Australian Imaging, Biomarkers and Lifestyle study (AIBL, mean age 73) [14], to 32.2% over an average of 4 years in the Alzheimer’s Disease Neuroimaging Initiative (ADNI, mean age 74) [15]. In all studies, the incidence of progression was at least two times greater in Aβ+ compared to Aβ–. Together, these studies show that Aβ+ increases risk for clinical disease progression in CN adults, although the error associated with some risk estimates is increased by the small sample sizes at the longer follow-up intervals [16]. For example, in ADNI, data beyond 4 years were available for 16% of the initial Aβ+ sample [15]. Similarly, estimates of progression at 5 years were based on 25% of the baseline MCSA sample [10], and 35% of the Knight ADRC data were available beyond 5 years [17]. These small samples limit the use of complex analyses to examine the influence of other characteristics proposed to hasten progression to symptomatic disease (e.g., age, interactions between apolipoprotein E (APOE) ɛ4 carriage and Aβ status) [18]. APOE ɛ4 carriage is associated with increased risk of both Aβ+ and dementia in CN older adults [19, 20], and neuropsychological studies show that Aβ+ APOE ɛ4 carriers experience greater cognitive decline than Aβ+ APOE ɛ4 non-carriers [18, 22]; therefore, the prognostic value of Aβ and APOE ɛ4 may be increased by examining their effects combined. Finally, although there is evidence that vascular and metabolic conditions increase risk of cognitive impairment and dementia in CN adults [23, 24], few studies have sought to account for their influence on estimates of Aβ-associated clinical progression. Although the prevalence of vascular disease in the population-based MCSA is higher than that found in controlled experimental samples [25], it was reported that adjustment for vascular diseases did not change their progression risk estimates [10]. Whether this remains the case in a large experimental cohort is yet to be seen.

The first aim of this study was to examine the incidence of Aβ-associated progression to MCI/dementia among CN older adults. The second aim was to identify demographic and clinical characteristics that moderate the relationship between Aβ status and progression. The first hypothesis was that Aβ+ would be associated with greater risk of clinical disease progression over 8 years. The second hypothesis was that Aβ-associated progression to MCI/dementia would be increased further by APOE ɛ4 carriage. Post-hoc analyses explored how demographic and clinical risk factors influenced clinical disease progression in Aβ+ and Aβ–CN adults.

METHODS

Participants

Participants were enrolled in the AIBL flagship study of aging, details of which have been described elsewhere [26]. Briefly, volunteers were excluded from entry if they had any of the following: non-AD dementia, history of schizophrenia or bipolar disorder, current depression (Geriatric Depression Scale score >5), Parkinson’s disease, cancer (other than basal cell skin carcinoma) within the last 2 years, symptomatic stroke, uncontrolled diabetes, obstructive sleep apnea, past head injury with >1 hour of post-traumatic amnesia, or current regular alcohol intake above recommended limits [27]. Health status at each study visit was determined from clinical assessment comprising measurement of vital signs (height, weight, waist circumference, and blood pressure using an electric sphygmomanometer) and self-reported medical history. Blood pathology for all participants was assessed at baseline. All included participants were identified to have no, or medically well-controlled systemic illnesses at baseline. Ethics approval for the AIBL study was granted by St Vincent’s Health, Austin Health, and Edith Cowan University, and written informed consent was collected from all participants prior to clinical and neuropsychological assessment at every 18-month interval.

Currently, the AIBL study includes 787 CN adults aged above 60 years who have undergone Aβ PET neuroimaging. Participants were recruited in two waves: an inception cohort (n = 444) followed for up to 8 years, and an enrichment cohort (n = 343) followed for up to 4.5 years. Of the initial cohort, 70.5% have remained active in the study for all visits up to 8 years. This study sample was restricted to those who attended at least two visits over the 8-year period (n = 621). Data for 19 participants with inconsistent clinical classification (n = 16) or Aβ status (n = 3) during the study period were excluded from all analyses. Twenty participants had incomplete covariate data, which were imputed based on information contained in their clinical visit notes where possible. One participant who met exclusion criteria but was inadvertently included in the AIBL study was also excluded. Thus, the total sample for this study consisted of 599 older adults (Fig. 1).

OPEN IN VIEWER

Fig.1

Sample selection criteria. Selection criteria and final sample sizes for the present study.

RESULTS

Risk of progression to MCI or dementia

Aβ+ participants were significantly more likely to progress to MCI/dementia than Aβ– (17.7% vs 8.1%, OR: 2.43, 95% CI: 1.47–4.03). In all 5 survival models, every additional year of age at baseline conferred greater risk (7%) of progressing over the 8-year period for all participants (Table 2). In Model 2, Aβ+ status increased the risk of progression by 104% (Fig. 2a). In Model 3, APOE ɛ4 carriers had 114% greater risk than non-carriers (Fig. 2b). The risk conferred by Aβ+ was mediated by the addition of APOE ɛ4 into the model. Model 4 showed that the health factors had no influence on risk of progression, nor did they mediate risk due to Aβ+; no interactions were observed in post-hoc analyses. In Model 5, the Aβ and APOE ɛ4 interaction was significant: Aβ+ APOE ɛ4 carriers had 348% greater risk of progressing to MCI/dementia compared to all other participants. Further analysis showed that progression risk in Aβ+ APOE ɛ4 non-carriers (hazard ratio (HR): 1.08, 95% CI: 0.56–2.07) and Aβ–APOE ɛ4 carriers (HR: 0.72, 95% CI: 0.21–2.45) was not significantly greater than in Aβ–APOE ɛ4 non-carriers (Fig. 2c); therefore, it was appropriate to combine these three groups in the interaction analysis.

OPEN IN VIEWER

Table 2

Fine-Gray subdistribution models for the full study sample

MODEL 1	p	HR	95% CI
Age	<0.0005	1.07	1.04	1.10
Hypertension	0.70	0.91	0.56	1.47
BMI	0.30	1.03	0.97	1.09
MODEL 2	p	HR	95% CI
Age	<0.0005	1.06	1.03	1.09
Hypertension	0.61	0.88	0.55	1.42
BMI	0.20	1.04	0.98	1.09
Aβ status	0.004	2.04	1.25	3.33
MODEL 3	p	HR	95% CI
Age	<0.0005	1.07	1.04	1.10
Hypertension	0.78	0.94	0.58	1.50
BMI	0.19	1.04	0.98	1.09
Aβ status	0.04	1.65	1.02	2.65
APOEɛ4	0.001	2.14	1.36	3.36
MODEL 4	p	HR	95% CI
Age	<0.0005	1.07	1.03	1.10
Hypertension	0.77	0.93	0.58	1.49
BMI	0.17	1.04	0.98	1.10
Aβ status	0.04	1.65	1.03	2.66
APOEɛ4	0.001	2.10	1.34	3.28
Diabetes	0.62	0.82	0.37	1.80
Stroke/TIA	0.30	1.36	0.76	2.44
MODEL 5	p	HR	95% CI
Age	<0.0005	1.07	1.04	1.10
Hypertension	0.88	0.97	0.60	1.55
BMI	0.23	1.03	0.98	1.09
Aβ status	0.88	1.05	0.55	2.02
APOEɛ4	0.57	0.71	0.21	2.36
Diabetes	0.80	0.90	0.41	2.00
Stroke/TIA	0.50	1.23	0.68	2.24
Aβ+ and APOEɛ4+	0.03	4.48	1.14	17.60

HR, hazard ratio; CI, confidence interval; Aβ, amyloid-β; APOE ɛ4, apolipoprotein epsilon 4 allele carriage; BMI, body mass index; TIA, transient ischemic attack.

OPEN IN VIEWER

Fig.2

Cumulative hazard functions for APOE ɛ4 status, Aβ status, and combined APOE ɛ4 and Aβ status. Cumulative hazard functions shown for A) APOE ɛ4 status, B) Aβ status, and C) combined APOE ɛ4 and Aβ status, indicating that APOE ɛ4 carriage and Aβ+ are independently and cumulatively associated with increased risk of progression to MCI/dementia. Total sample size at each time point is displayed at the bottom.

DISCUSSION

Rates of Aβ+ progression from CN to MCI/dementia in the AIBL sample and the associated risk factors for both Aβ– and Aβ+ progression over 8 years are reported for the first time. The results supported the first hypothesis that Aβ+ would be associated with greater risk of progression to MCI/dementia. Eight-year risk of progression to MCI/dementia in CN Aβ+ adults from the AIBL study was increased by 65–104% compared to Aβ– (Table 2). This indicates that Aβ+ is an important prognostic marker for progression to MCI/dementia in CN older adults. The second hypothesis, that Aβ-associated risk for progression to MCI/dementia would be increased further by APOE ɛ4 carriage, was also supported. The large sample studied here allowed the risk of progression conferred by concurrent Aβ+ and APOE ɛ4 carriage to be estimated, taking into account health factors posited to influence progression to MCI/dementia as well as competing risks such as death or withdrawal due to illness. Risk of progression to MCI/dementia was 348% greater in Aβ+ APOE ɛ4 carriers compared to all other participants (Table 2). Similar findings were observed in the MCSA, although their risk estimate was reported relative to Aβ– APOE ɛ4 non-carriers (190%) [37]. In this study, no difference in risk was observed between Aβ– APOE ɛ4 non-carriers, Aβ– APOE ɛ4 carriers and Aβ+ APOE ɛ4 non-carriers, suggesting an additive effect between Aβ+ and APOE ɛ4 carriage that is greater than the sum of their individual contributions. This is consistent with results of neuropsychological studies showing that CN Aβ+ APOE ɛ4 carriers experience greater cognitive decline over time with earlier onset [18, 38]. In agreement with a recent report that episodic memory performance remains stable in Aβ– regardless of APOE ɛ4 status, APOE ɛ4 carriage did not increase the risk of clinical disease progression in Aβ– (Fig. 2c) [38]. Previous research shows that APOE ɛ4 reduces clearance of cerebral Aβ but does not affect rates of Aβ production [39]; therefore, the findings of this study indicate that the impaired clearance of Aβ due to APOE ɛ4 is most clinically significant in individuals who have high levels of Aβ.

As progression to MCI/dementia was observed in a small number of individuals with normal Aβ levels (8.1%), risk factors within the Aβ+ and Aβ– groups were investigated. Previous studies indicate that health conditions such as hypertension, diabetes, or stroke/TIA can contribute to cognitive decline and clinical disease progression in older adults [23, 40]. Although higher prevalence of hypertension and lower BMI was observed in the Aβ+ group, these differences were very small. Thus, both groups had similar cardiovascular risk profiles and these factors did not influence risk of progression within either Aβ group. For both Aβ groups, higher age at baseline was associated with increased risk of progression to MCI/dementia (Table 3). Although Aβ deposition, and therefore risk of disease progression, is known to increase with age [41], other neuropathological processes such as brain atrophy are also associated with age [42]. Higher relative Aβ deposition increased progression risk in the Aβ+ group by 150%; however, PET SUVR/BeCKeT for Aβ– progressors was not near the threshold for Aβ+ (median 1.16, range 1.02– 1.39) making it unlikely that progression in the Aβ– group was due to any unrecognized increase in Aβ. While it remains unknown whether abnormal levels of Aβ deposition play a causative role in the development of dementia due to AD, the recent NIA-AA Research Framework proposed that AD be defined by the presence of cortical Aβ and tau aggregates and neurodegeneration rather than by clinical symptoms due to the poor specificity of cognitive symptoms to detect AD neuropathological processes [4]. Taken together, these data suggest that Aβ-associated clinical progression is consistent with AD neuropathological changes, whereas progression in the absence of elevated Aβ deposition is the result of disease processes other than AD that accumulate with age [43]. This indicates that the prognostic value of Aβ+ is specific to dementia due to AD.

Despite the longer time interval and greater statistical control over demographic, health and clinical characteristics, the 17.7% incidence of progression due to Aβ+ was consistent with that observed previously over 6 years (19%) in the AIBL sample [14]. However, the current 8-year estimate of Aβ-associated progression remained lower than those reported in the ADNI (32.2%) and Knight ADRC (26.4%) cohorts over similar time periods [15, 44], and was equal to that reported by the MCSA (17.7%) over an average of 3.7 years [10]. The relatively lower incidence of progression in AIBL may reflect differences in the samples studied and the respective inclusion/exclusion criteria. For example, the ADNI and Knight ADRC cohorts were, on average, 2–4 years older than AIBL and the MSCA cohort was 6 years older than the AIBL cohort. The population-based MCSA cohort reports higher prevalence of risk factors other than Aβ+ for MCI/dementia: at baseline, 79.4% of the MCSA participants had hypertension, 18.7% had diabetes, and 14.3% had history of stroke [25], compared to 38.8%, 7.3% and 1.8%, respectively, in the AIBL CN cohort. Despite the increased presence of these factors and older age in the MCSA, the estimates of Aβ-associated progression were similar and may also reflect comparable methods for measuring Aβ deposition and defining Aβ+ using PET; however, the follow-up time for the MCSA was shorter than that for AIBL and may be expected to increase with similar follow-up. It is also possible that the higher incidence of disease progression in the other samples reflects some unreliability in their estimates due to small sample size or differences in method of Aβ+ classification. Both ADNI and Knight ADRC used different cut-off values and either PET neuroimaging or CSF amyloid sampling to classify Aβ status across participants, while both AIBL and the MCSA used only PET to measure Aβ levels in all participants; therefore, differences between the Aβ+ samples identified in these studies may also reflect different prospective estimates of incident MCI/dementia. Measuring Aβ levels with a common method for all participants rather than using different biomarkers to do so will increase the reliability of classification. Finally, the methods utilized to define clinical status vary across the different studies. The Knight ADRC relies on CDR score to classify participants and, while AIBL and the MCSA both use a consensus panel to determine clinical status, these panels define cognitive impairment differently (≤–1.5 SD on two tests versus ≤–1 SD on one domain score). ADNI also utilizes clinical consensus classifications; however, the public availability of this data has meant different researchers have also utilized CDR ratings and actuarial neuropsychological approaches to define clinical status, which itself has resulted in different estimates of Aβ-associated progression [15, 46]. These varying approaches to clinical classification and sample selection may explain the differences in estimates of Aβ+ progression between studies. Findings of consistency or inconsistency in outcomes between different samples is crucial because this provides information about the effects of potential sampling bias associated with the different studies on models of disease progression and the disease processes reported from these individual cohorts. Lower incidence of progression in the current study may reflect the larger sample sizes at longer follow-up times, the more stringent criteria for cognitive impairment, the use of consensus classification, and the more exclusive sample when compared to the other large studies of preclinical AD. Nonetheless, these data indicate that Aβ+ is an important factor for clinical disease progression in AD.

A recent consensus group reported the importance of established and putative risk factors for dementia among older adults, and stated that the predictive value of Aβ+ for progression to MCI/dementia was equivocal over 3 years [47]. While they agreed that age and APOE ɛ4 carriage were important risk factors of clinical progression to symptomatic disease, these factors are non-modifiable. The group, therefore, considered other potentially modifiable risk factors and concluded that hearing loss, hypertension and obesity in midlife, and smoking, depression, physical inactivity, social isolation, and diabetes in late-life held greater prognostic utility than did Aβ [47]. The present study examined risk factors for progression accounting for age and APOE ɛ4 carriage and showed that Aβ+ was a strong predictor of clinical disease progression in CN adults over 8 years, while health factors such as hypertension, diabetes and obesity were not. The current findings converge with that from other prospective studies of Aβ+ risk for MCI/dementia, suggesting that the consensus position be reconsidered with data from longer prospective studies, given that the preclinical stages of AD can extend for up to 20 years [3].

Some aspects of this study limit the generalizability of its findings to a wider population. First, the AIBL study utilized a convenience sample and recruited healthy and well-educated older adults. Participants who did not progress were more likely to have attended all study visits, suggesting the presence of a healthy survivor effect. This may have contributed to the lack of relationship between health factors and disease progression in this study; hence, conclusions drawn here about the influence of these aspects of health on late-life risk for MCI/dementia should be challenged in similar studies using population-based sampling methods, such as the MCSA, using midlife health risk factors where possible. Although the number of participants retained in AIBL at 8 years was larger than the sample sizes of the other studies at their longest intervals and AIBL had the longest average follow-up time, even longer follow-ups are necessary to elucidate the disease course and risk factors for MCI/dementia. For this reason, it is not known whether those who did not progress over the study period will go on to progress in the future. Furthermore, as Aβ-associated progression was the focus on this study, neurodegeneration measures were not examined. Despite these caveats, the current results indicate that Aβ+ has strong prognostic value for the development of clinical symptoms of dementia due to AD even when health factors and competing risks for progression are taken into account. APOE ɛ4 carriage provides further predictive value in the presence of elevated Aβ; therefore, Aβ+ APOE ɛ4 carriers are ideal candidates for early intervention trials of disease-modifying therapies.