Obesity in middle age and future risk of dementia: a 27 year longitudinal population based study

Study population

We conducted a prospective analysis of 10 276 members of the Kaiser Permanente medical care programme of northern California who participated in voluntary periodic multiphasic health checks in San Francisco and Oakland, California, between 1964 and 1973. We identified participants aged 40-45 at the time of the multiphasic exam who were still members of Kaiser Permanente when outpatient diagnoses of dementia were available in 1994 (n = 25 290). After we excluded those who had died before 1994 (n = 2598), were no longer members (n = 10 407), and had missing information on sex (n = 9), 10 276 remained for analysis.

Kaiser Permanente is a non-profit, group practice integrated healthcare delivery system that includes hospitals and outpatient clinics that contract exclusively with a single group of physicians to provide all healthcare services to all members of the system. It covers more than a quarter of the population in the areas served, and members are representative of the sociodemographics of the local population in the service areas.¹¹

Data collection and mid-life adiposity

At the multiphasic exam, participants were interviewed, underwent a clinical examination, and gave a blood sample. Information was collected on demographics and medical history. Full details have been published elsewhere.^12-14 Height and weight were measured according to standardised procedures.¹² We categorised body mass index (weight/height²) as obese (≥ 30), overweight (25.0-29.9), normal (18.6-24.9), and underweight (≤ 18.5). Subscapular and triceps skinfold thickness were measured by trained technicians using Lange skinfold callipers (Cambridge Scientific Industries, Cambridge, MA) according to the criteria of the committee on nutritional anthropometry.¹⁵,¹⁶

Diagnosis of dementia and other illnesses in later life

We searched the databases of inpatient and outpatient medical records from the care programme (from January 1994 to April 2003) for diagnoses of dementia and other illnesses. All cause dementia diagnoses included: dementia, Alzheimer's disease, and vascular dementia (ICD-9 (international classification of diseases, ninth revision) codes of 2900.0, 7809.3, 3310.0, 2904.1, 2900.1), and the criteria for these diagnoses did not change during the ascertainment period. We determined incidence of ischaemic heart disease, hypertension, stroke, hyperlipidaemia, and diabetes. To gather information on mortality we used the California automated mortality linkage system, which has a sensitivity of 97% compared with the National Death Index,¹⁷ up to the end of 2000 and a matching linkage system, incorporating social security number, name, and address, from 2001 through the end of 2002. Mortality information was not available from January to April 2003.

Statistical analysis

We used SAS version 8.0 (SAS Institute, Cary, NC) for analyses. We used the log rank test to assess the association between time to diagnosis of dementia and characteristics measured at the multiphasic exam and Cox proportional hazard models to identify independent predictors of risk of late life dementia. Person years were calculated from onset of follow-up (1 January 1994) until onset of dementia or the earliest of death, end of Kaiser Permanente membership, or end of study (3 April 2003). We carried out χ² analyses to determine if there were any significant differences in the mid-life measures of adiposity and covariates by health plan membership status in 1994. Because measures of skinfold thickness varied significantly by sex (P < 0.0001) we divided the distribution into fifths for men and women and used these in the analyses. For the body mass index models, the reference group was participants with a normal body mass index, while for skinfolds, it was those in the lowest fifth.

We generated three models for each measure of adiposity (body mass index and subscapular and triceps skinfolds): firstly, a model adjusted for age in mid-life (age at time of multiphasic exam) and education; secondly, a model additionally adjusted for age at start of ascertainment of dementia (age in 1994), race, sex (with the exception of sex stratified models), smoking, alcohol use, and martial status; and, thirdly, a model additionally adjusted for mid-life comorbidity (high total cholesterol, diabetes, and hypertension) and late life comorbidity (diabetes, ischaemic heart disease, stroke, hypertension, and hyperlipidaemia).

Strengths and weaknesses of study

Strengths of the study include the longitudinal design with skinfold measures from two parts of the body as well as body mass index, the long follow-up, the breadth of information on other diseases in middle and old age, and the well characterised representative sample with equal access to medical care. The participants were middle aged when risk factors were assessed, thus subclinical dementia at baseline is highly unlikely.

This study also has limitations. We could assess dementia status only among participants who were still members of the health plan when dementia was diagnosed. However, there were no differences in adiposity or sociodemographic characteristics in those who didn't continue in the health plan. Information on weight cycling, dieting, nutrition, or mid-life measures of cognitive functioning were not collected, and other studies have shown that several different nutritional factors are associated with dementia.¹⁸,¹⁹ A measure of central obesity, such as waist circumference would have been informative, but this was not collected. As diagnoses of dementia were made during medical visits, some cases of dementia may have been missed in those who did not participate in visits. This, however, would bias results in an underestimation of the effects of obesity on dementia.

Mechanisms of effects

One plausible reason for an increased risk of dementia with adiposity is through cardiovascular disease and diabetes as both these conditions increase the risk of dementia.¹⁴,^20-22 Yet, adjustment for prevalence of diabetes and cardiovascular disease at mid-life and later did not attenuate the associations. Adiposity is also one component of the metabolic syndrome, which has also been shown to cause cognitive decline, particularly in those with high levels of inflammation.²³

Perhaps adiposity has a direct effect on neuronal degradation. Genetically obese, leptin receptor deficient rodents have impaired performance on spatial memory tasks²⁴ and long term potentiation of neurons in the hippocampus.²⁵ C reactive protein, an inflammatory marker, is increased in those with greater adiposity²⁶ and is associated with dementia²⁷ and cognitive decline.²³,²⁸ Recently, obesity in elderly women was shown to be associated with greater cerebral atrophy²⁹ and white matter hyperintensity.³⁰ Future studies of cytokines produced in fat cells and neuronal functioning should be useful.

This is the first study to assess skinfold thickness on risk of dementia. Another study found that high skinfold thickness in mid-life was associated with Parkinson's disease.³¹ Perhaps adiposity works together with other risk factors to increase neurodegenerative disease.

Body mass index predicted dementia more strongly among women. These findings, however, are consistent with the only published prospective study on body mass index and dementia.⁴ There could be several reasons for this. There were fewer obese and overweight men, thus the power to detect an effect was reduced. Current body mass index categories may be insensitive indicators of adiposity in men because of less variation in the distribution of visceral fat.⁵ Perhaps body mass index and risk of dementia in women is mediated through central obesity,³² which is highly associated with insulin resistance,³³ cardiovascular disease,³² increased adipocytokines,³⁴,³⁵ and inflammatory markers.³⁶ All of these conditions could contribute to dementia. In the present study diabetes and cardiovascular disease did not attenuate the effect, though the role of central obesity, inflammation, and adipocytokines on our findings is unknown.

Future directions

Future studies on obesity and dementia should consider distribution of adiposity and the role of adipocytokines on brain structure and function. If these results can be confirmed elsewhere, perhaps treatment of obesity might reduce the risk of dementia. Failure to contain the present epidemic of obesity may accentuate the expected age related increase in dementia.

Amendment

This is version 2 of the paper. In this version, in the results section of the abstract the comparison is between men and women in the highest fifth (not the higher fifth) and the risk of dementia in men and women in the highest fifth of tricep skinfold thickness is 59% greater (not 60%), in table 1 the footnote says that data on education were missing for 1185 (not 185), and the adjusted hazard ratio for obese men is 1.30 (95% confidence interval 0.84 to 1.87 (not 1.20) in table 3 and the third paragraph of results.