BMI, metabolic conditions and cognitive impairment

Healthy eating is generally associated with a normal or low BMI, whereas an unhealthy diet is more often associated with overweight or obesity. Nevertheless, the effect of dietary composition on cognition in old age should not be underestimated, since healthy eating habits may also associate with overweight at the individual level (Buckland et al. 2008). This may explain why some studies have found no consistent association between BMI and cognition, although the bulk of epidemiological evidence indicates that a higher BMI in midlife increases the risk of poorer cognition in old age (see 2.1.4.2).

In this study, midlife BMI was generally associated with a poorer cognitive test performance, but also with the categories of impaired cognition. The results suggest that the association between midlife BMI and cognitive performance in old age is partly mediated by metabolic conditions, since adjusting for these led to a loss of statistical significance. In addition, the effect of metabolic conditions on cognition was only significant when not adjusted for BMI, reflecting their strong association. However, it cannot be excluded whether a sample larger than ours (containing more lean subjects with metabolic conditions) would also demonstrate significant associations when adjusted for BMI. This possibility is supported by the fact that dietary habits are also known to associate with metabolic conditions, including T2DM (Martinez-Gonzalez et al. 2008, Nash and Nash 2008, Esposito et al. 2009) and cardiovascular diseases (Nash and Nash 2008), independently of BMI.

An association between a higher midlife BMI and cognitive impairment may reflect either the effect of harmful nutritional factors such as saturated fat, extra adipose tissue gathered in the body, metabolic conditions partly ensuing from a high BMI, or most probably their combination (Figure 12). In addition, several other variables, most notably education, are associated with these, and should therefore be taken into account as confounders in epidemiological studies (Silventoinen et al. 2005, McLaren 2007).

Figure 12. Three diet-related health variables associated with cognitive performance in old age and their causal expectations.

Associations between metabolic conditions and impaired cognitive performance in old age observed at the individual level in this study support the findings of previous studies, and the size of the study population and duration of follow-up were comparable with them. Among the metabolic conditions, diabetes was most strongly associated with cognitive performance.

Obviously, we cannot know whether and to what extent this association reflects the effect of hyperinsulinemia, often considered as the main mediating factor, glucose intolerance, and/or adiposity (Luchsinger and Mayeux 2007). Compared to T2DM, previous evidence on the harmful effects of midlife cardiovascular disease (Singh-Manoux et al. 2008), hypertension (Launer et al. 2000, Kivipelto et al. 2001, Qiu et al. 2005) and hypercholesterolemia (Kivipelto et al. 2001) on cognitive performance in old age is less. This highlights the importance of studying them in different populations, including Finns, who have a high national incidence of both potentially harmful metabolic conditions and memory disorders.

Significant associations were also found with respect to both cardiovascular disease and hypertension and impaired cognition in old age in this study (Figure 13).

Cognitive performance

in old age

Extra adipose

tissue

Metabolic conditions Nutritional

factors

Figure 13. Significant associations between the metabolic conditions under study and cognitive performance in old age at the individual level (unidirectional arrows) and within discordant twin pairs (bidirectional arrow). The latter supports a possible causal pathway between the two variables.

Since dementia-associated weight loss often starts when cognition can be considered only mildly impaired or even earlier (Grundman 2005), studying the effect of midlife weight change requires at least two measurements of BMI several years before the earliest symptoms of memory impairment. Although one study found that a large increase in BMI from early to late midlife was associated with a lower executive function (Sabia et al. 2009), our study demonstrated that a weight change of even half of that within an average of 5.6 years significantly reduced the cognitive performance in old age. The decreasing effect of weight loss on cognitive performance found in our study may be associated with subsequent weight gain and thus reflect the effect of weight fluctuation. Another study investigating the effect of weight change found that weight loss almost doubled the risk of AD compared to a comparable weight gain among women (Beydoun et al. 2008). Thus, the effect of weight fluctuation may be even more harmful than pure weight gain. On the other hand, the possibility that weight loss is a symptom of a process leading to cognitive impairment years later, and not necessarily part of the causal process, cannot be excluded. However, the earliest hospitalization for dementia in this study was only in 1993, and the low proportion of

Cognitive performance

in old age

Diabetes

Hypertension

Hyper-cholesterolemia

Cardiovascular disease

subjects with prescribed medications for dementia (3%) in 1996 suggest a low incidence of incipient memory disorders at the time when weight loss was assessed.

In this study, the ApoE ε4 allele that has been associated with an increased risk of impaired cognition was not significantly associated with cognitive performance when adjustments were made for age, sex and education. However, the results of linear regression analysis indicated a higher risk of poorer cognitive performance when having one ApoE ε4 allele, and an even higher risk when having two alleles, and are thus in concordance with those studies in which these associations have been statistically significant. Indeed, the risk increase in relation to having two ApoE ε4 alleles was also nearly significant in this study (p = 0.078), and significant associations could be demonstrable in a larger sample. Contributions of genetic and environmental factors to the association between BMI and cognitive performance had not previously been studied. We found that the additive genetic correlation was statistically significant, but only if education was not incorporated into the model, indicating that genetic factors associated with education correlate with those of BMI (Figure 14).

Indeed, a significant negative correlation between genetic factors (rA) affecting midlife BMI and education was demonstrated in this same study cohort. Among men and women born in 1915–1946, rA was -0.18 and -0.21 (Silventoinen et al. 2004), i.e. stronger than the genetic correlation between BMI and cognition in old age found in this study (rA=-0.12).

14a)

14b)

Figure 14. The upper figure (a) represents a situation in which education is not incorporated in the model and there is significant genetic correlation underlying the association between BMI and cognition. The incorporation of education in the model led to a loss of this significance (b).

In fact, the loss of a significant correlation found in this study supports the hypothesis of a causal relationship between BMI and poorer cognition. However, the existence of even a subtle common genetic pathway between BMI and cognition cannot be excluded in the basis of this study, since it could be demonstrable in a larger sample of twins.

6.2.2.3 Causal conclusions in epidemiological studies in relation to BMI,