Reference Data for Attentional,Executive,Linguistic,and Visual Processing Tests Obtained from Cognitively Healthy Individuals with Normal Alzheimer’s Disease Cerebrospinal Fluid Biomarker Levels

INTRODUCTION

The biological presence of the Alzheimer’s disease (AD) is defined by amyloid-β (Aβ) and tau pathology. The early impact of Aβ and tau in cognition has been reported even in the cognitively healthy population [1, 2]. The effects of Aβ and tau have been mainly associated with a decline in memory function, but recent evidence suggested that beyond memory, other cognitive processes such as executive functioning are affected at early stages of preclinical AD [3]. A comprehensive assessment of cognitive function across cognitive domains, and not just of memory function, might help to delineate better the subtle cognitive decline associated with subjacent Aβ deposition, and aggregated tau at the preclinical stage of AD.

In 2018, the research criteria of the National Institute on Aging-Alzheimer’s Association (NIA-AA) defined subtle cognitive decline as the stage of transitional decline between healthy cognition and mild cognitive impairment (MCI) [4]. The characterization of subtle cognitive decline in preclinical AD aims to advance clinical diagnosis at earlier asymptomatic stages, and to improve sensitivity in clinical trials targeting cognitive function. The NIA-AA provided a classification system based on the status of the three AT(N) biomarker types that describe the biological profile along the AD continuum, Aβ (A), p-tau (T), and non-specific neurodegeneration ([N]) [4]. Thus, the AT(N) system allowed the classification of individuals with normal AD biomarkers, individuals within the AD continuum, or individuals with a non-AD pathologic change, providing a research framework for delineating the earliest relationship between biological changes and the decline of cognitive function at preclinical stages.

Normative data for clinical neuropsychological testing are conventionally obtained from cognitively unimpaired (CU) individuals. Thus, cognitive impairment could be identified by comparing individual performances with normal range while adjusting by sociodemographic factors, such as age, education, and sex. Nevertheless, this canonical procedure might yield some deflation in normative cognitive scores according to the presence of subtle cognitive decline in CU individuals due to incipient abnormal levels of AT(N) biomarkers indicating underlying AD pathology.

Since conventional neuropsychological norms might lack enough sensitivity for detecting subtle cognitive decline in preclinical stages, robust normative procedures typically account for the reliable performance of clinically stable populations (including only individuals who maintain a status of healthy cognitive performance on longitudinal follow-up and excluding those that latter on develop MCI/dementia). Robust normative approaches based on longitudinal assessment demonstrated an increase in the probability of labeling cognitive performance as below average in comparison to conventional norms [5–7], by establishing higher thresholds for what is considered normal performance or decline according to these robust procedures [8, 9].

An alternative approach is to rely on biomarker data for constructing robust references sensitive enough to identify subtle decline in cognitive performance. The presence of Aβ positivity in CU individuals ranging from 50 to 70 years has an estimated prevalence of between 10.4% and 23.1% [10], and the presence of either Aβ, tau pathology, or neurodegeneration in CU individuals older than 65 years has an estimated prevalence of up to 44% [11]. In fact, the effect of aging in the cognitive performance of older adults is attenuated when AD biomarker status is accounted for in the analysis [12]. Accordingly, excluding individuals with signs of altered Aβ, p-tau, neurodegeneration (neurofilament light) biomarker levels, cerebrovascular pathology, or uncontrolled systemic medical illness, results in more rigorous standards for detecting cognitive decline than those of conventional norms [13, 14]. Therefore, accounting for cerebrospinal fluid (CSF) AD biomarkers for the adjustment of normal performance in cognitive testing may enhance the sensitivity of normative neuropsychological data beyond standard published norms for detecting subtle cognitive decline in individuals not meeting yet MCI criteria [15, 16].

Previous research identified that the use of norms excluding individuals with positive Aβ biomarker levels, as criteria for a more robust normative procedure, allowed the identification of individuals at risk of dementia and increased the predictive accuracy of dementia progression using memory measures [16]. Thus, Bos et al. proposed a “normal or impaired” classification model based on three groups combining the classification of conventional norms and Aβ negative norms (i.e., Group 1: normal with both norms; Group 2: impaired only with robust Aβ negative norms, Group 3: impaired with both norms [16]). Such classification model provides a complementary interpretational framework for characterizing subtle cognitive decline. Although in Bos et al. [16] it was only necessary to adjust for AD biomarkers in memory measures, other evidence pointed out that removing preclinical AD participants from normative samples also results in higher means and less variability in visuospatial ability and executive function [17]. For this reason, we considered relevant to explore the usefulness of AD biomarker-based normative data in the assessment of other several cognitive domains apart from memory processes.

We have recently published norms for the verbal memory tests Free and Cued Selective Reminding Test and the WMS-IV Logical Memory Test in a sample of CU individuals with normal AD CSF biomarker levels. The use of such biomarker-based norms in combination to standard references may be useful for the identification of subtle impairments in memory function [15]. These biomarker-based normative data have demonstrated to capture cognitive decline in individuals with evidence of positive Aβ and tau status in a 3-year follow-up within the ALFA+study, while conventional norms failed to do so (in preparation).

In the present work, as a robust normative procedure we only used data of CU individuals with normal (non-pathological) levels of AD biomarkers. We excluded from our analyses individuals with positive status of underlying AD pathology, defined by CSF Aβ (A), p-tau (T), t-tau ([N]) biomarkers, with the aim to provide a more sensitive normative reference for cognitive evaluation. The AD biomarker-based normative procedure is a complementary interpretational framework for the neuropsychological assessment of subtle cognitive decline in preclinical AD.

METHODS

RESULTS

Demographic, presence of the APOE ɛ4 allele, inclusion cognitive outcomes, and biomarker data of the participants included in this study are presented in Table 1. Descriptive data of the neuropsychological assessment is presented in Table 2. Results of multiple linear regression analyses according to the estimated coefficient (beta) value for each variable and related p-value are presented in Table 3. Equations used to calculate z-scores accounting for relevant sociodemographic factors are presented in Table 4. Normative tables with the calculations developed and raw scores equivalence to percentiles are available as Supplementary Tables (1–20).

We assessed the proportion of impaired scores when using the biomarker-based in contrast to the standard references. Using biomarker-based references, the percentage of at least one or more lower scores among the 9 measures considered was 44.76%. Using the standard references, the percentage was 21.77%. The McNemar test for related proportions was statistically significant when comparing the distribution of impaired individuals between both references (McNemar X2 (df = 1, n = 248) = 52.155, p = <0.001). The Fleiss’ Kappa coefficient revealed low-intermediate agreement between datasets for the labeling (k = 0.470, z = 7.36, p = <0.001). To illustrate the different behavior of both norms we further provide a single test example, we assessed the proportion of impaired scores when using biomarker-based as compared to WMS-IV Spanish references in the Symbol Span. Using the current biomarker-based references, the percentage of lower scores for this single measure was 5.65%, while using the WMS-IV this percentage was 1.61%. The McNemar test for related proportions was statistically significant (McNemar X2 (df = 1, n = 248) = 6.75, p = 0.009). The Fleiss’ Kappa coefficient revealed low agreement between datasets for the labeling using this single measure (k = 0.308, z = 4.83, p = <0.001).

To facilitate the use of the reference data presented here and the previously published norms for the Free and Cued Selective Reminding Test and the WMS-IV Logical Memory Task developed with the same sample and approach [15], we provide online interactive tables with automatic calculators in the Supplementary Material.

DISCUSSION

In this study we provided regression-based normative data obtained from a sample of CU individuals (aged between 50 and 70 years) without underlying AD pathology, assessed by CSF determination of core AD biomarkers (AT[N]). Normative data is presented for the following tests: the Digit Span of the WAIS-IV, the Symbol Span of the WMS-IV, the TMT, the Coding of the WAIS-IV, the Matrix reasoning of the WAIS-IV, the Semantic Fluency Test, the Visual Puzzles of the WAIS-IV, and the Judgement of line orientation of the RBANS. The current biomarker-based cognitive references aim to provide an interpretational framework for the characterization of subtle cognitive decline in preclinical AD.

Given the potential limitations of conventional neuropsychological norms in identifying subtle cognitive decline, robust normative procedures are commonly employed to ensure the inclusion of clinically stable populations. These procedures involve selecting individuals who consistently demonstrate healthy cognitive performance over longitudinal follow-ups, while excluding those who later develop MCI or dementia. Such robust normative procedures aimed to increase the sensitivity of norms for the detection of MCI [5, 8]. However, robust normative procedures are not limited to clinical stability over longitudinal follow-up, nor to the detection of MCI. Recent biomarker-based normative approaches have expanded beyond these parameters by excluding data obtained from individuals with underlying AD pathology. Such norms render higher cut-offs and more sensitive standards than those provided by conventional norms aiming to detect the subtle cognitive decline that precedes MCI [12, 13]. By incorporating such innovative methodologies, researchers can enhance the accuracy and sensitivity of cognitive assessment tools, thereby facilitating the early identification of individuals at risk of dementia and potentially enabling interventions at preclinical stages.

The present work may help in the characterization of subtle cognitive decline. As defined by the NIA-AA in 2018, subtle cognitive decline is the stage of transitional decline before MCI [4]. This criterion proposed that this stage can be documented through subjective reports of cognitive decline or objectively by a longitudinal follow-up. Since most individuals with subjective reports of decline do not progress to dementia [25], objective cognitive measures may be more predictive of future dementia [16]. A direct approach is to use cross-sectional performance to predict clinical progression. The use of the common demographically adjusted cut-offs of< -1.5 SD below the mean, or percentile < 5, using a group of CU individuals without biological evidence of AD pathology as a reference has already been reported to improve the accuracy prediction of dementia progression [16]. By combining conventional and biomarker-based normative data, it is possible to identify those CU individuals who perform in the normal range according to standard normative data (without using negative-biomarker CU as reference) but fall in the impaired range according to normative data based on AD biomarkers (Group 2 classification according to Bos et al. [16]). We propose that cognitive scores showing these discrepancies could serve as indicators of subtle cognitive decline.

We assessed the sensitivity of the biomarker-based norms provided in this work in comparison to the conventional published references. We demonstrated that the proportion of lower scores (individuals labeled as impaired) significantly differed between the datasets, being higher for the biomarker-based norms. In addition, there was a low-intermediate agreement for impaired labeling between the norms. Considering the classification discrepancies between references and the fit of biomarker-based approaches for the characterization of preclinical AD, results suggest that the presented norms work as intended and indeed were more sensitive to subtle cognitive changes than the standard references.

As it has been previously reported, removing individuals with positive status of amyloid-β and/or tau pathology in normative data, yields higher reference scores than those of standard procedures [15–17]. As a result, the sensitivity was enhanced but the specificity of the norms decreased. To explore the characterization of subtle cognitive decline, the norms provided here do not aim to replace previously published ones. We suggest using these norms as a complementary tool (e.g., not in isolation but rather in combination with published conventional norms) for an interpretative framework in characterizing subtle cognitive decline. For reference, we provide hypothetical examples of the use of the normative data provided in this work in comparison with standard norms (see the Supplementary Material).

We found a relevant effect of age on the performance of all tasks, except for the Digit Span forward and backward parts. Education affected the performance for all tasks except for two variables within the attentional domain, the Digit Span forward, and the TMT-A. Education was positively associated with cognitive performance, except for the TMT-B, in which the association was negative, as expected (higher education associated with lower completion times). Sex affected the performance in all tasks except for two tasks within the executive domain, the TMT-B, and the WAIS-IV: Coding, and for the task in the linguistic domain, the Semantic Fluency Test. Males outperformed women on the TMT-A, WAIS-IV: Digit Span, WMS-IV: Symbol Span, WAIS-IV: Matrix reasoning, WAIS-IV: Visual Puzzles, RBANS: Judgment of line orientation, indicating that males tend to have higher scores than females. These associations between sex and cognitive performance showing better scores in men are consistent with previous studies in the studied cognitive domains [26–29], but are in contrast with our previously published normative data from the same sample in verbal episodic memory, in which women consistently outperformed men [15].

The influence of sociodemographic variables on cognitive function is well known [30]. However, available normative data typically presents adjustments for age, but adjustments for education are not always included, and including them for sex is even less frequent. It is of high relevance to adjust cognitive norms for sex, since not accounting for sex differences could lead to misdiagnoses [31]. Following this consideration, and in contrast with available normative data, the present study included adjustments for education for the WMS-IV: Symbol Span; the WAIS-IV: Coding; and the WAIS-IV: Matrix reasoning, and additional adjustments for sex for the WAIS-IV: Digit Span; the WMS-IV: Symbol Span; the TMT-A; the WAIS-IV: Matrix reasoning; and the WAIS-IV: Visual Puzzles.

The main limitation of our work resides in the partial applicability of the normative data provided. Individuals tested were aged from 50 to 70 years-old, thus constraining the age range of application. Nevertheless, the proposal of normative data based on AD biomarkers is especially useful in this age range, since the estimated prevalence of subjective cognitive decline is 25% for individuals older than 60 years [32]. The range of application is also constrained by the fact that our sample is mainly composed of individuals with at least elementary studies. Therefore, there is an important gap in the lower levels of education and the presented norms are not directly applicable to individuals with lower educational levels than those of the present investigation. Considering sample characteristics, the percentage of APOE ɛ4 carriers is higher in our sample than in the general population, and arguably, this may yield some bias in our norms. Nevertheless, although APOE ɛ4 might be associated with lower performance in this age range, we believe that we are controlling this effect by including AD biomarkers, since the effect of APOE ɛ4 on cognition is thought to be mediated by the presence of amyloid-β and tau pathology [33]. We also acknowledge that we used a highly sensitive cut-off for the CSF biomarker levels [19], therefore using more liberal cut-offs or different measurements such as PET imaging to define the reference group could result in different distributions.

In conclusion, we developed socio-demographically fully adjusted reference data obtained from individuals with evidence of normal (non-pathological) (AT[N]) biomarker levels in CSF. In combination with conventional norms, the biomarker-based cognitive references presented here may improve the objective characterization of subtle cognitive decline in preclinical AD.