An Example of a Pitfall in the Correlation of Excess Fat with Age-Related Disease
Terrible, slow moving age-related diseases that kill you also tend to make you lose weight along the way. Even the lengthy period of gradually increasing disability prior to full-blown disease can achieve that result. This point is very important to bear in mind when looking at association studies that map measures of weight versus disease risk, or life expectancy, or other health metrics. Are the studies using late snapshots of weight, or lifetime maximum weight, or some other measure and time, and does that choice of data succeed in avoiding entanglement with the loss of weight that serious age-related disease tends to produce? If it doesn't, then the result may be suspect.
The study I'll point out today examines a very large set of data, that of more than a million individuals. In the course of processing this data, the study authors well illustrate the point made above. For measures of weight taken decades prior to the development of age-related disease, excess weight correlates with raised risk of disease. But if measuring weight within a few years of the diagnosis of age-related disease, that correlation is reversed - in later life, the group of normal weight people includes some of the least healthy, who have lost weight since their earlier highs due to the early stages of disease and dysfunction. They developed age-related disease because they were overweight, but then their status becomes less visible to simple statistics as the weight is lost.
There is little doubt that carrying excess visceral fat tissue is very harmful to health. It is in the same ballpark as smoking, when measured in terms of lost years of life expectancy, increased lifetime medical expenditure, and risk of disease. The evidence for this is overwhelming, ranging from many human epidemiological studies of hundreds of thousands of individuals tracked over decades to demonstrations of extended life in mice achieved through surgical removal of visceral fat tissue. Still, while being one of the more straightforward associations to measure, it isn't so straightforward as to prevent a number of research groups falling into the trap of failing to account for disease-related weight loss in their data. This is why studies such as the one below exist.
Body mass index and risk of dementia: Analysis of individual-level data from 1.3 million individuals
The costs of dementia are enormous and increasing globally. Current clinical guidelines for dementia prevention view obesity as one of the modifiable risk factors, but the evidence is based on a relatively limited number of observational studies and the findings are mixed. The most recent meta-analysis, including 4 studies and 16,282 participants, suggested a 1.4-fold increased risk of dementia in the obese. The largest study in the field, published after the inclusion date for the meta-analysis, found no increase in dementia incidence among the obese. On the contrary, higher body mass index (BMI) was linked to lower dementia risk. The reasons for this discordance in findings are unclear.
One possibility is that the observed association between BMI and dementia is attributable to two processes: one is a direct association between higher BMI and increased dementia risk, and the other is an association confounded by weight loss during the preclinical dementia phase, which leads a harmful exposure to appear protective via reverse causation. This hypothesis is supported by the fact that clinical diagnosis of dementia is often preceded by a long (20-30 years) preclinical phase during which cardiometabolic changes, including weight loss, are common. Thus, lower BMI close to dementia onset might be a consequence of preclinical disease rather than a cause of dementia.
The purpose of the present analyses was to investigate the BMI-dementia association using raw unpublished data. We included 39 prospective cohort studies which comprised a total of 1,349,857 participants with no history of dementia; were population based with BMI assessed from all participants before the ascertainment of dementia; recorded hospital-treated dementia or dementia deaths; and had accrued a minimum of 3 years of follow-up. We found that higher BMI was associated with increased dementia risk when weight was measured more than 20 years before dementia diagnosis, but this association was reversed when BMI was assessed less than 10 years before dementia diagnosis. The findings of this study are consistent with the hypothesis that the BMI-dementia association is attributable to two processes: a direct (causal) effect and reverse causation as a result of weight loss during the preclinical dementia phase.
As the present meta-analysis is based on a series of studies in which investigators ascertained dementia in different ways, we had the possibility to undertake a validation exercise. Thus, we repeated the main analyses excluding dementia status drawn from death certificates. The same pattern of results was evident as in the main analyses: higher BMI was associated with greater risk of dementia when BMI was measured many years before dementia onset, whereas an inverse relationship was apparent when BMI was measured closer to dementia ascertainment.
In analyses exploring survival bias, we found that higher baseline BMI was associated with an increased risk of all-cause mortality before the age of 65 years but lower mortality risk after the age of 85 years (the median age of dementia diagnosis). These findings suggest that, compared with their normal weight counterparts, obese individuals were less likely to live long enough to develop dementia and more likely to die from conditions that are known to be related to increased dementia risk, such as diabetes and cardiovascular diseases. Given these findings, differences in survival may have contributed, if anything, to an underestimation of the strength of the association between BMI and dementia.