Subtle Cognitive Decline and Biomarker Staging in Preclinical Alzheimer’s Disease

INTRODUCTION

The pathogenesis of Alzheimer’s disease (AD) is believed to begin many years prior to clinical diagnosis, and there is a great need to better detect and characterize the earliest phases of the disease process. The National Institute on Aging and the Alzheimer’s Association (NIA-AA) published research criteria for “preclinical” AD [1], an asymptomatic phase in which individuals are classified as cognitively normal but have biomarkers associated with AD (i.e., evidence of amyloid-β or tau protein alterations) and possible evidence of “subtle cognitive decline,” which has yet to be definitively operationalized. The criteria propose a staging method based on the notion that biomarkers of AD follow an invariable temporal sequence in accordance with the amyloid cascade hypothesis [2, 3]. In Stage 1, individuals show biomarker evidence of cerebral amyloid accumulation. Stage 2 is characterized by amyloid positivity plus evidence of neuronal injury or neurodegeneration. Stage 3 involves these two criteria plus evidence of subtle cognitive decline. However, this temporal sequence has been called into question [4, 5], as studies have shown that biomarkers of neurodegeneration can occur before amyloidosis and many individuals who progress to AD demonstrate a pattern of biomarker abnormalities that fail to conform to the amyloid cascade hypothesis [6–13]. In addition, several studies have shown that cognitive decline, reflected by sensitive episodic memory measures, is at least as sensitive as, and sometimes superior to, cerebrospinal fluid (CSF) and imaging biomarkers in predicting the development of AD [14–17]. Our studies in mild cognitive impairment (MCI) have shown significant heterogeneity in cognitive profiles associated with early stages of AD beyond the typical “amnestic” profile [18–22]. Thus, it is plausible that there are different neurobiological pathways to AD [23].

Previous studies have demonstrated the imprecision of the conventional MCI diagnosis, including its instability over time (e.g., high rates of reversion to cognitive normalcy) [24–26] and a high rate of false positive diagnostic errors [20–22]. Such imprecision in diagnosis will likely become even more pronounced as studies attempt to identify “subtle cognitive decline,” which according to the NIA-AA criteria [1] “may include slightly abnormal performance on sensitive cognitive measures, self-complaint of memory decline, or subtle neurobehavioral changes.” We aimed to operationalize subtle cognitive decline within the same conceptual framework we have previously used to define MCI and to examine associated CSF biomarker abnormalities within the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort.

METHODS

Data were obtained from the ADNI database (http://adni.loni.usc.edu). ADNI was launched in 2003 by the National Institute on Aging, National Institute of Biomedical Imaging and Bioengineering, Food and Drug Administration, private pharmaceutical companies, and non-profit organizations. The primary goal of ADNI is to test whether neuroimaging, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. ADNI is the result of efforts of many co-investigators from a range of academic institutions and private corporations. This study was approved by an ethical standards committee on human experimentation at each institution. Written informed consent was obtained from all participants or authorized representatives participating in the study. For more information, including criteria for eligibility, see http://www.adni-info.org.

RESULTS

At baseline, 48.9% of the sample (n = 279) was positive for amyloidosis, 63.3% (n = 361) for neurodegeneration, and 13.3% (n = 76) for subtle cognitive decline. There was no difference in age, education, or gender between individuals who demonstrated subtle cognitive decline and those who did not (all p-values >0.05). Of the 76 participants who demonstrated subtle cognitive decline, 65 had impaired scores on two of the six neuropsychological measures in different cognitive domains (criterion 1 of our operational definition), and 11 had an FAQ score between 6–8 but no impairment on neuropsychological testing (indicating functional decline; criterion 2 of our operational definition). Comparison of those who met criteria for subtle cognitive decline based on cognitive performance versus functional performance showed no significant differences in age, education, gender, amyloid positivity, neurodegeneration, APOE status, or rates of progression to AD or MCI (all p-values >0.05).

Forty-four of the 570 participants (7.7%) progressed to dementia at a subsequent follow-up visit (mean time of diagnosis = 31.4 months, SD = 23.1). Of these,43 met the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS/ADRDA) criteria for AD. The remaining participant was diagnosed with progressive supranuclear palsy and was therefore excluded from the progression analyses. One-hundred-thirty four additional participants (23.5%) progressed to a diagnosis of MCI (mean time of diagnosis = 18.5 months, SD = 21.3) based on the actuarial neuropsychological criteria.

DISCUSSION

Using a novel operational definition of “subtle cognitive decline” that examined the performance of individual neuropsychological test scores and captured both cognitive and functional decline, we classified cognitively normal participants based on the NIA-AA criteria [1] and separately based on the number of abnormal biomarkers or cognitive markers that each individual possessed. Using the NIA-AA criteria [1], we found that Stage 2 (amyloidosis plus neurodegeneration) and SNAP (neurodegeneration with normal amyloid levels) were the most common stages in individuals who showed evidence of preclinical AD. When we used the number of biomarkers or cognitive markers without regard for their temporal order, we found that neurodegeneration alone was 2.5 times more common than amyloidosis alone at baseline. For those who demonstrated only one abnormal marker at baseline and later progressed to MCI/AD (n = 37), it was most common to have neurodegeneration alone (n = 24), and nearly equally common to have amyloidosis alone or subtle cognitive decline alone (n = 8 versus 5) as the first sign of prodromal AD. These findings suggest that most individuals did not follow the temporal order proposed by the amyloid cascade hypothesis, which is that amyloid biomarkers become abnormal first, followed by biomarkers of neurodegeneration, and then clinical symptoms [2, 3]. Our results are consistent with research showing that biomarkers of neurodegeneration can occur before amyloidosis [10], as well as findings that individuals with subtle cognitive decline but normal neurodegeneration biomarkers (i.e., the Unclassified group) have a relatively high rate of progression to AD [36]. By presuming that all patients follow a singular invariant expression of the disease, we may be failing to identify individuals with differing initial presentations who nevertheless will progress to AD. It should be noted that the NIA-AA diagnostic criteria for AD [37] and MCI [38] are agnostic with respect to the temporal order of biomarker changes. Evidence of amyloidosis or neural injury increases diagnostic confidence to an “intermediate” likelihood, and having both increases it to a “high” likelihood that the diagnosis of dementia or MCI is due to AD. Our method of staging preclinical AD based on the number of abnormal markers, irrespective of their order of appearance, similarly demonstrated increasing likelihood of progression.

A comparison of our findings to other studies [34–36] aimed at examining the NIA-AA criteria [1] for preclinical AD are shown in Fig. 4. Our study found relatively high rates of amyloid positivity (48.9%) and neurodegeneration (63.3%), which is reflected in the higher percentage of participants in Stage 2 and Stage 3 compared to previous studies. This discrepancy is purely the result of the way in which we defined “cognitively normal” versus “MCI.” Rather than using the diagnoses assigned by ADNI, which has been shown to produce a high rate of false positive diagnostic errors (i.e., roughly one-third of those diagnosed with MCI by ADNI actually perform normally on more extensive cognitive testing and have normal CSF biomarker levels) [21, 22], we used actuarial neuropsychological criteria [22, 27] to determine cognitively normal versus MCI classifications. Of the 570 deemed cognitively normal by the actuarial criteria and included in the current study, only 250 were also classified as normal by ADNI while the remaining 320 were diagnosed with MCI by ADNI (based on subjective memory complaints, a Clinical Dementia Rating Scale score of 0.5, and low performance on a single memory test) [39]. If our analyses are conducted only on the 250 individuals classified as cognitively normal by ADNI, the percentage of participants in each preclinical AD classification stage based on NIA-AA criteria [1] is comparable to previous studies. In particular, our results are quite similar to the Toledo et al. [36] study, which also used an ADNI sample (Stage 0 : 29% versus 32% ; Stage 1 : 13% versus 15% ; Stage 2 : 25% versus 22% ; Stage 3 : 2% versus 3% ; SNAP: 30% versus 23% ; Unclassified: 2% versus 5% in the current study versus the Toledo et al. [36] study, respectively). However, the main findings of our study remain the same in this smaller sample of 250. Specifically, of the 2% (n = 5) who progressed to dementia at follow-up, three of these five individuals had one abnormal biomarker at baseline: one with amyloidosis and two with neurodegeneration. In addition, two of the five individuals were in the SNAP category at baseline based on NIA-AA criteria [1]. Thus, the majority did not show amyloid biomarker abnormalities as the firstsign of AD.

Although our findings do not abide by the amyloid cascade model [2, 3], they do concur with the theoretical notions of Braak and colleagues [4], who have proposed an alternative model to the amyloid cascade hypothesis. They propose that detectable AD pathologic markers (CSF Aβ and tau pathology, neuronal loss, cognitive symptoms) all representlate-appearing and relatively simultaneous changes in the AD pathophysiologic cascade, which appear in a very narrow time sequence. The perceived differences in the sequence of biomarker abnormalities may be more a reflection of the varying sensitivity of the biomarkers or of our ability to detect change than to the timing of these neurobiological changes. The fact that subtle cognitive decline has traditionally been viewed as the last marker to be affected in preclinical AD may be due to the routine use of crude, insensitive measures of cognitive abilities (i.e., rating scales or screening measures) or variations in the way subtle cognitive decline is defined. By contrast, sensitive memory measures (e.g., Rey Auditory Verbal Learning Test) have been shown to be the earliest markers to become abnormal in the progression to AD relative to screening measures, rating scales, and biomarkers [14], and are more robust predictors of conversion from MCI to AD than most biomarkers [15, 16].

Our novel method of staging preclinical AD is at least as predictive of progression to MCI/AD as the stages proposed by the NIA-AA criteria [1]. This approach that simply tallies the number of abnormal markers provides a more parsimonious method of characterizing preclinical AD, since the staging system described by others [9, 35] results in ambiguous “SNAP” and “Unclassified” categories. In the current study, the SNAP group was similar to Stages 0 or 1 with regard to APOE-ɛ4 frequency, FAQ total scores, and progression rates, while the Unclassified group was similar to Stages 1 or 2 with regard to APOE-ɛ4 frequency and progression rates and similar to Stage 3 with regard to FAQ total scores. It is not clear if there is anything to be gained by creating these separate SNAP and Unclassified groups, particularly given that SNAP (i.e., suspected non-AD pathophysiology) itself has been shown to be a misnomer by research demonstrating that both amyloid-first and neurodegeneration-first biomarker profile pathways to preclinical AD exist [10].

One might argue that biological markers offer a more specific marker of AD pathophysiology than cognitive decline, which could be due to a number of other non-AD pathologies, and that elevating the notion of subtle cognitive decline to be on par with the highly specific amyloid pathology may ultimately decrease specificity of the AD diagnosis. In the current study, nearly all (43/44) of the participants who progressed to dementia met NINCDS/ADRDA criteria for a diagnosis of AD. Nonetheless, an individual who demonstrates the biomarkers and cognitive markers examined in this study may also show non-AD pathology and could go on to develop a mixed dementia syndrome. Indeed, the notion of “pure” AD pathology is increasingly thought to be far more rare than multiple underlying neuropathologies (e.g., AD, cerebrovascular disease, Lewy bodies, hippocampal sclerosis, TDP-43) [40–42] for the vast majority of late-onset AD. Thus, a combination of biomarkers and sensitive neuropsychological tests that detect these various underlying pathologies is an improvement on any individual biomarker and can substantially improve prediction of dementia risk [17].

A strength of the present study was the use of individual test scores and actuarial neuropsychological criteria to operationalize subtle cognitive decline within the same conceptual framework we have previously used to define MCI [27]. This method of defining subtle cognitive decline may improve diagnostic consistency and reliability compared to the use of “self-complaint of memory decline or subtle neurobehavioral changes,” [1] which has been shown to cloud rather than clarify diagnosis [43], or the use of composite scores [9, 35] in which sensitive and insensitive test performances are averaged together, with the result that the composite score’s sensitivity to subtle cognitive decline is likely reduced. A direction for future research will be to examine individual neuropsychological markers of subtle cognitive decline against different methods of constructing composite scores (e.g., principle components analysis, item response theory) [44] to determine which method most reliably predicts progression. In addition, future studies are needed to explore the impact of using different normative reference methods (e.g., education-based norms) to define subtle cognitive decline.

One limitation of our criteria for subtle cognitive decline is that they may be too strict to capture all individuals with very early cognitive changes (i.e., those who have declined cognitively but are still performing in the normal range on neuropsychological testing or those who show only one impaired score). On the other hand, it is possible that false positive errors could result from applying our actuarial criteria for subtle cognitive decline to populations where individuals deviate from the demographic characteristics of ADNI (e.g., low education or premorbid abilities). Despite this potential risk, our previous work has demonstrated that using our actuarial criteria for MCI actually substantially reduces false positive diagnostic errors compared to conventional criteria that rely on subjective complaints, rating scales, and a single memory test [20, 22]. Another limitation of the current study is that ADNI participants undergo only a relatively brief battery of neuropsychological tests. A more extensive neuropsychological evaluation would enhance our ability to detect and more fully characterize subtle cognitive decline, provided that we are mindful of the possible increase in base rates of low scores with increasing numbers of tests [45, 46] and the resulting potential for misclassification [31]. By clearly defining “subtle cognitive decline” via sensitive and reliable neuropsychological test scores, and by re-exploring timelines of biomarker and cognitive changes, we hope to improve diagnostic precision of preclinical AD and sharpen our ability to identify early signs associated with different neurobiological pathways to the diagnosis of AD.