Cardiovascular risk factors and memory decline in middle-aged and older adults: the English Longitudinal Study of Ageing

Cardiovascular risk factors (CVRFs) (tobacco smoking, hypertension, obesity, physical inactivity and diabetes) are highly prevalent among midlife and older adults [1], and constitute leading causes of mortality. Of the 2.5 million Americans who died in 2005, tobacco smoking killed 1 in 5, elevated blood pressure was responsible for 1 in 6 deaths, and obesity and physical inactivity took 1 in 10 lives [2]. Besides this augmented risk for death, population-based studies have also identified CVRFs as strongly related to higher risk for accelerated cognitive decline [3, 4] and dementia [5]. For example, current smokers are between 50 and 80% more likely to develop Alzheimer’s Disease (AD) in the future, compared with those who never smoked [6, 7], and older adults who engage in high leisure-time physical activity have approximately 45% less risk of AD than people with the lowest level [8]. The effect of CVRFs on cognitive deterioration is mostly, but not exclusively, driven by cardiovascular diseases (CVD) [9]. Other pathophysiological mechanisms through which CVRFs might impact on cognitive decline or risk for dementia include inflammation and oxidative stress, cerebral small vessel diseases, cerebral hypoxia and hypoperfusion, or neurodegeneration in the brain, which in turn might also increase the risk for heart disease [10]. Several epidemiological studies have shown that the effect of several CVRFs on cognitive deterioration follows a dose-response pattern [10]. Thus, composite scores of CVRFs are often used as potential tools to detect at-risk older adults for cognitive decline and dementia.

Despite the association between CVRFs in older adults and risk for dementia being well-established [5, 9, 11], several authors postulate that this relationship might be age-dependent [12]. High blood pressure during midlife has been shown to be associated with an increased risk of AD in late life [13, 14], whereas research on the effect of high blood pressure in late life and dementia has yielded mixed results [9]. Similarly, the association of obesity and AD seems to be also age-dependent, with obesity in midlife reported as risk factor for AD [15], and underweight in late-life being more important in predicting AD [16]. Other studies have also demonstrated that aggregated CVRFs at midlife are associated with cognitive decline in the middle age [17, 18], suggesting that this effect can be observed at early stages. The associations between CVRFs during late life and dementia are less clear, with several studies finding very small or null effects on subsequent cognitive deterioration [13, 19].

Comparison across these studies focusing on middle-age versus late-life might be challenging because of distinct study designs, measures and sampling characteristics. To the authors’ knowledge, no studies have compared the impact of a summary score of CVRFs on subsequent cognitive deterioration in population-based middle aged versus older cohorts. The impact of a summary score of CVRFs at baseline on subsequent trajectories of cognition (verbal episodic memory) over a period of 10 years was investigated in a longitudinal, nationally-representative cohort of adults aged 50 to 64 and 65 to 79 years old. We aimed to study whether this association followed a dose-response pattern by exploring the interaction between time and CVRFs. Models were adjusted for several confounders to determine whether this association was above and beyond the effect of potential explanatory variables. The stratification of models by two age cohorts (middle-aged and older adults) allowed us explore age differences in the association between CVRFs and cognitive decline.

Table 1 presents a descriptive analysis of the total sample (N = 7377) and by age cohorts.

Compared with the younger cohort, older participants presented higher rates of low education level (p < 0.001), and higher proportions of two or more non-CVDs and CVDs (p < 0.001). The prevalence of number of CVRFs was also higher in the older cohort (p < 0.0001). Mean score of episodic memory at baseline was lower than middle-aged people (p < 0.001). Number of depressive symptoms was also higher (p < 0.01). Some 22.5% of older participants died by the end of the study, and 45.6% completed all the assessments.

Overall, our findings suggest a greater cardiovascular risk burden on contemporaneous memory decline in midlife, but not in late-life, supporting the hypothesis that the deleterious effect of cumulative CVRFs is age-dependent [10]. They also suggest that the effect of CVRFs follows a dose-response association with verbal episodic memory decline which is independent of other potential confounders.

As has been previously suggested, the presence of CVRFs in midlife is related to increased risk for late-onset cognitive impairment and dementia [27, 28]. Anstey et al. [17] speculated that, if these risk factors for late-life dementia occur in midlife, some effects on cognitive functioning, despite small, might be detectable in middle age. These authors found that greater CVRF burden, measured with a composite score of several risk factors, was associated with faster decline in reaction time. Similarly, other previous longitudinal population-based studies found worse cognitive performance associated with CVRFs in midlife [4, 18, 29]. In our study, middle-aged participants with no CVRFs show an improvement in their memory performance over 10 years. This lack of cognitive deterioration over time, or even some improvement due to practice effect, has been documented in previous research based on general populations, especially among younger and middle-aged adults [17, 30]. However, our results show that participants aged 50 to 64 years who had one, two or three or more CVRFs did not experience this improvement and did significantly worse compared with those free of CVRFs. Albeit subtle, the observed effect of CVRFs on verbal episodic memory is potentially important and meaningful, since one would expect memory not to deteriorate to an observable extent in middle-aged adults [31, 32].

Our findings also indicate that prospective association between cumulative CVRFs and memory in participants aged 50–64 years is dose-response with the memory score getting lower for every additional CVRF. The existing evidence concurs with that and suggests that composites of CVRFs have been shown to have a dose-dependent effect on the risk for dementia [33]. Since these risk factors frequently co-exist, intervention focusing on the combined effect of multiple CVRFs is preferable rather than focusing only on individual risk factors [3].

As for older adults (i.e., aged 65 to 79), we found that there were a significant decline of memory over time, but this was not associated with the number of baseline CVRFs. Previous literature has yielded mixed results when analysing the effect of summary risk scores of CVRFs in later life on risk of dementia, with some population-based studies reporting a positive effect [34, 35], and others failing to report significant effects of CVRFs in late-life on subsequent deterioration of cognitive function [5]. It has been also suggested that CVRFs no longer act as risk factors for dementia [10]. For example, obesity during midlife has been demonstrated to be a risk factor for late-life dementia, whereas for older ages, being underweighted is related to increased risk for cognitive impairment [12]. This U-shaped relationship has been similarly found for hypertension [36]. Higher levels of blood pressure at midlife which decrease over time has been shown to be associated with increased white matter lesions [37]. In a previous study conducted on the ELSA survey [38], we similarly found that the number of CVDs was only significantly related to subsequent latent trajectories of verbal episodic memory in midlife but not in late-life. A recent investigation with the ELSA study also reported that the co-occurrence of diabetes type II and elevated depressive symptoms was associated with accelerated cognitive decline, especially among those aged 50–64 years [39]. Therefore, it is possible that cognitive deterioration in older adults is being affected by different pathways or risk factors. Overall, middle-aged participants present with better levels of health at baseline than older participants. The number of non-CVDs and CVDs were significantly higher in the older cohort, and this might have greatly impact on their memory function. For example, musculoskeletal disease, lung disease or arthritis have been previously linked to cognitive decline [40]. Despite the fact that we controlled for the presence of multimorbidity (including both CVDs and non-CVDs), it is possible that the effect of chronic conditions on participants’ memory function might be more salient on the older cohort than in midlife, thus attenuating the detrimental effect of CVRFs in older participants. There is also the hypothesis that the duration of the exposure to CVRFs might impact on the rate of cognitive decline. Some previous research has supported this when studying diabetes in middle-aged adults [41, 42], hypertension [43], or smoking [44]. Future studies to investigate how age of onset of these CVRFs is impacting on cognitive functioning are warranted.

Survival bias might also partially explain this lack of differences. Presence of CVRFs in younger ages or midlife are related to higher risk for death [45], and thus older participants in our study might be healthier than expected. Cohort effects can also explain this lack of significance. Older and younger cohorts might be exposed to different risk factors or historical events during their lifespan [46].

Some limitations should be considered when interpreting our findings. First, some CVRFs at baseline were self-reported and they could have been affected by recall bias or low accuracy, especially for the older cohort [47]. Second, the age of onset for CVRFs was not considered in this study despite the fact that the duration of these conditions might play an important role on subsequent memory decline. Third, we focused on verbal episodic memory. CVRFs might affect differently the trajectories of other cognitive domains (e.g., attention, verbal fluency). Moreover, one could not discard floor effect, especially in the older age group. Fourth, despite the fact that we excluded those individuals who self-reported being diagnosed with dementia or other brain disorders, it is possible that some others might have presented with milder forms of neurodegenerative disorders, such as mild cognitive impairment (MCI). This would be particularly important if the presence of MCI is contributing to lower baseline cognitive scores and muting the test-retest effect at follow-up, especially in the older group. Finally, non-response might have also introduced some bias. However, we controlled for the confounding effect of completion.

Our findings support the deleterious effect of aggregate CVRFs in midlife (50–64 years) on subsequent decline over a period of 10 years, whereas this association was not found when CVRFs were measured in older adults (over 65 years). Despite being subtle, the decline observer among middle-aged adults with a high cardiovascular burden was significantly different from that observed for people with no CVRFs, where an improvement was observed. These differences in cognitive decline might increase as people aged, leading to greater risk for future dementia, such as AD. Interventions over these modifiable conditions at midlife could help stop this cognitive deterioration.