When Being Thin Is Not a Virtue

Obesity is strongly associated with type 2 diabetes mellitus, hypertension, altered cholesterol metabolism, and survival in later life.¹ Not surprisingly, it is also a central poor player in the story of vascular risk factors and late-life dementia. Investigations of associations between obesity at earlier life time points and dementia^2–5 have found that persons with obesity are more likely to develop dementia by the time of follow-up. In contrast, a number of studies that have examined weight or body mass index (BMI) in the decade or so before the onset of dementia have found that low body weight or low BMI is associated with incident dementia.^6–12 The study by Dahl et al. in this issue of the Journal reports similar observations but goes one step farther. Dahl et al. find that obesity appearing within 10 years of dementia onset is not a risk factor but is, in fact, protective. This finding held for women, and there was a trend for the same in men. It seems likely that the difference between men and women may simply be because the smaller number of men with incident dementia yielded less power to detect an association. The main result of Dahl et al. probably holds for men and women and is entirely consistent with the story relating low body weight in later life and dementia, although it is in conflict with the studies that considered obesity in midlife to be a risk factor.

How can a condition go from being a risk factor to a protective factor? If obesity is the result of poor dietary habits, sedentary lifestyle, and other unhealthy lifestyle choices, it should be a risk factor regardless of age. The fact that obesity appears to switch direction of risk may seem puzzling, but there probably are some good explanations.

One hypothesis is that obesity itself is not the true causal carrier of risk but rather is associated with something else that is, such as unhealthy lifestyle, other dysfunctional health behaviors, or diseases (such as diabetes mellitus or altered glucose-insulin regulation). Because obesity occurs with higher frequency in people who also have diabetes mellitus, hypertension, and poor lifestyle choices, it is difficult to isolate obesity as a risk factor. Even when diabetes mellitus and hypertension are controlled in analytical models, the definitions of diabetes mellitus and hypertension may include benign and malignant forms of those disorders and therefore not account correctly for the variance associated with the true causal metabolic or vascular disorders. Recent studies have documented a metabolically benign form of obesity,¹³ and there are variations in associations between obesity and diabetes. In the Dahl et al. study, only 11.9% of the obese subjects had diabetes mellitus (A. Dahl, personal communication). In contrast, in the Kaiser study, in which midlife obesity was a risk factor, 59% of the obese subjects also had diabetes mellitus.⁴ The lower rates of diabetes mellitus in the Dahl et al. study than in the Kaiser study could contribute to the difference in associations between obesity and dementia. Furthermore, there could be an interaction with age. Obesity in late life might not have the same associations with vasculopathic processes as it does in middle-aged persons. Another aspect of obesity's relationship to risk is that simple measurements of weight or BMI may be inferior measures of adiposity. Abdominal girth may be better as a genuine indicator of pathology.⁴ For all of these reasons, obesity may have a variable relationship to dementia.

A second possibility to account for obesity as protective in late life is that obesity is associated with excess mortality in midlife, culling out those at risk for heart disease and stroke. Selective mortality may operate mainly through obesity's associations with more-malignant conditions such as poorly controlled diabetes mellitus and hypertension, consistent with the first hypothesis. Through succumbing to related diseases (e.g., heart disease) or producing dementia earlier, obesity in late life identifies survivors who possess other traits that are especially protective, the so-called hardy survivor effect. Nevertheless, in the Dahl et al. study, it was underweight, not obesity, that characterized the 419 (35%) study subjects who died between the first and second study visits. So if obesity was a risk for mortality earlier in life, it had ceased to be observable in this elderly sample (mean age 70.8 at baseline).

A third perspective is the empirical fact that obesity is one end of a bidirectional spectrum of adiposity. The constructs of weight and BMI have a U-shaped risk profile that includes risk at the underweight and overweight ends of its distribution. The reasons for the risks differ between the two ends of the weight spectrum, and conditions that affect weight vary with age. Weight status may be a risk factor and a manifestation of disease. Obesity may be a true risk factor for dementia and mortality when appearing in midlife. In later life, underweight may be a manifestation of an underlying brain disease or a systemic disease. In younger persons in whom chronic diseases are less common, the morbidity associated with being underweight may be underrepresented compared with an elderly cohort. Therefore, the balance between risks for being underweight versus overweight favor overweight as a risk factor in younger persons, whereas in elderly people, such as in the Dahl et al. study and others, conditions—neurological and systemic—associated with underweight predominate as risk factors for incipient dementia.

This is not the first study that showed how a putative risk factor was found to be protective in very elderly people. A study of metabolic syndrome in persons aged 85 and older found that it was associated with lower rates of dementia.¹⁴ Those authors raised similar concerns about the importance of comorbidity and selective survival. Although midlife hypertension is considered to be a risk factor for cognitive decline,^5,15 in late life, hypotension may also be a risk factor.^16,17 Cholesterol may also reverse its associations with dementia from midlife to later life.¹⁸

One thing is clear. Obesity by itself is not a simple risk factor for dementia. Most importantly, the construct of weight status is bidirectional. Obesity is more than just pounds on a scale; metrics of adiposity should reflect association with risk or protection. The company that obesity keeps—namely diabetes mellitus and hypertension, as well as healthfulness of lifestyle—also should be considered. Finally, the association between obesity and lower risk for dementia should not be construed as a recommendation for obesity in late life. The behaviors and biology that promote loss of weight are the pathological processes of importance in late life, and overeating will not reverse those.