Neuropathology and measures of brain structure do not fully explain cognitive decline (1) nor age-related variation in cognitive function (2). This is evident in the finding of normal cognitive function in individuals who meet the diagnostic criteria for Alzheimer’s disease (AD) based on neuropathology (3,4). This well-established gap between brain and cognition may be explained by cognitive reserve (CR), wherein the effects of brain pathology or ageing on cognitive function are moderated by an individual’s ability to efficiently or flexibly use the brain’s resources to cope with task demands (5).
Accurate measurement of CR could improve the detection of, and risk assessments for, age-related cognitive decline and AD (6) and improve the measurement of intervention efficacy in clinical trials and intervention studies by enabling researchers to effectively statistically control for CR (7). Difficulties in measuring CR (8), however, limit this potential. The most direct measures of CR are likely to be obtained using functional neuroimaging (8). CR may be measured with functional MRI using resting-state and task-based functional connectivity. For example, a pattern of greater change in functional connectivity from resting state in response to task demands is associated with better cognitive performance, above and beyond the effects of cortical thickness (9). However, the considerable cost of MRI scanning (10) limits access to such measures, particularly in lower income countries (11). As such, socio-behavioural variables reflecting the degree of exposure to, or engagement in, various lifetime experiences are often used as proxies of CR (8).
The rationale for using proxies is that greater exposure to certain lifetime experiences increases the adaptability of cognitive and functional brain processes, thereby enabling a greater ability to cope with brain changes or damage (8). Considerable epidemiological evidence indicates a reduced risk and/or delayed onset of dementia and cognitive decline in individuals with greater educational attainment (12–14); occupational complexity/status (15–17); literacy and/or verbal intelligence (18–21); engagement in activities that were cognitively stimulating (22,23); leisure-related (24,25); physical (22,26–28); and social (22,23,29). Proxies also provide a single value with a simple interpretation – a higher degree of exposure reflects greater CR. Moreover, proxies are easy and inexpensive to obtain, and some, such as educational attainment, are routinely collected as part of most ageing studies. It is therefore not surprising that CR is most often measured using proxies (30).
Despite their advantages, the use of proxies to measure CR has been criticised. First, some proxies, such as educational attainment, are typically static measures (31) despite the fact that CR is considered to be a dynamic construct that can change over time (32). Second, some argue that a single proxy fails to reflect the full CR construct which is thought to be influenced by a range of experiences (33,34). Finally, proxies may also be associated with cognitive decline via mechanisms other than reserve (35). For instance, greater educational attainment is correlated with higher socioeconomic status (36) which is itself associated with slower cognitive decline (37) and reduced risk and prevalence of dementia (38,39). Low socioeconomic status is associated with various other factors, including stress and access to healthcare, which could exacerbate cognitive decline (38). As such, the protective effect of education on cognitive decline and dementia (but cf. 40 for an alternative perspective) may be via mechanisms related to socioeconomic status, rather than CR (41).
The limitations of individual proxies may be mitigated by averaging (cf. transformation methods such as principal component analysis) multiple proxies to create a composite proxy measure that still provides a single summary value with a simple interpretation (42–46). Composite proxies allow for a wider range of contributions to CR and enable the inclusion of dynamic proxies that can change over time, such as verbal intelligence or engagement in activities (31). Furthermore, composite proxies may attenuate the issue of non-CR mechanisms of individual proxies because alternative mechanisms (e.g., socioeconomic status) might only be associated with some proxies, such as educational attainment, but not others like social engagement. Some composite-type approaches, including factor analytic and latent variable models, measure CR using inappropriate reflective measurement models, where the observed CR proxies are effectively considered to be reflective (i.e., caused by) the latent CR construct (35). Composite proxies are a more appropriate formative measurement model, where the observed proxies are considered to form, or cause, CR. Moreover, this approach can reflect the unique additive contributions of individual proxies, whereas factor analytic models reflect only the shared variance across different proxies (8).
While the composite approach offers advantages over the use of single proxies, there is no agreed-upon gold-standard composite proxy (30) just as there is likewise no gold-standard individual proxy. Similarly, it is unclear which proxy should be used when assessing candidate neuroimaging measures of CR, as face validity is assessed via their association with CR proxies (47,48). The considerable variation (49,50) and lack of coherence in the use of proxies means that there is poor comparability across studies, as an effect observed for one proxy (e.g., educational attainment), may not be observed to the same degree for another (e.g., occupational complexity), even though both putatively reflect CR. It also provides researchers in the field of CR with an additional “researcher degrees of freedom” (51) such that several different proxies could be examined but only statistically significant results are reported.
To assess the validity of a potential measure of CR, a complete model of CR is required, which includes 3 components: a measure of CR (e.g., a proxy), a measure of brain structure/pathology, and a measure of cognitive function (8,52). This enables the assessment of the cognitive benefit criterion (48). This criterion can be satisfied via the observation of 1) an “independent effect” in which the candidate measure is positively associated with cognitive function, independent of brain structure, or 2) a “moderation effect” in which the candidate measure moderates the relationship between brain structure and cognitive function (8,47). The moderation effect is considered the ideal benchmark for CR, whereas the independent effect is considered a weaker level of evidence for a CR effect (8).
A systematic review of CR proxies from complete CR models reported inconclusive evidence for educational attainment, occupational complexity/status and leisure activity as proxies of CR in cognitively healthy cohorts (53). A single reviewed study provided evidence that greater engagement in cognitively stimulating activities in mid- and late-life provided CR effects (54). Other proxies were not assessed in this systematic review, although individual studies have reported positive evidence for CR effects in complete CR models. Verbal intelligence has been positively associated with cognition, controlling for global AD neuropathology or hippocampal atrophy in cognitively healthy (55,56) and cognitively impaired older adults (55). Physical activity was positively associated with cognition in the presence of neuropathology (57) but not hippocampal atrophy (56). Social engagement moderated the relationship between amyloid-beta deposition and cognitive decline (58). The composite of verbal intelligence and education moderated the relationship of subcortical grey matter (GM) volume and cortical thickness with fluid reasoning but not memory or processing speed and attention (46). This composite was also associated with memory controlling for GM volume (59) and global cognition controlling for a composite AD-biomarker (45). Although other composites have been associated with cognition (50), there is very little empirical evidence regarding their effects within complete CR models.
There is currently no conclusive evidence for the best individual or composite proxy for measuring or validating neuroimaging measures of CR, particularly with respect to cognitively healthy older adults. A methodology for solving this problem is the use of hierarchical linear moderated regressions to systematically assess standard CR proxies and their composites in complete models, an approach that enables the examination of both moderation and independent effects within the same analysis framework. This is important because, although moderation effects should ideally be observed to validate a CR proxy or measure (8), they are typically small in real-world data (60), explaining 1–3% of the variance in the outcome (61). Consequently, large sample sizes are required to detect typically small moderation effects (62). This issue is further exacerbated when measurement error is present in either variable in the interaction term (e.g., the CR proxy and measure of brain structure) used to assess the moderation effect (63) or when either variable in the interaction term is associated with the outcome variable (e.g., cognitive function; 65). Given the noted difficulties in identifying moderation effects, it is important to also consider the independent effect when assessing the validity of CR proxies.
Hierarchical linear regressions allow the robustness (i.e., frequency of effects using different measures of brain structure and cognitive function) and magnitude of both moderation and independent effects of different proxies to be compared. Here, in two separate community-dwelling older adult cohorts, we examined five common putative CR proxies – education, occupational complexity, verbal intelligence, leisure activities, and exercise – and all of their possible combinations. We included three brain structure variables, mean cortical thickness, hippocampal volume, and grey matter volume, in each model. Our primary aim was to identify the CR proxies with the most robust and largest effects across two datasets. More formally, we define effective CR proxies as those variables that have a significant independent or moderation effect on measures of cognitive function and brain structure.