A widely-held theory is that the declining rate of dementia seen in many – but not all -- cohort studies is due to reductions in cardiovascular disease (CVD) risk factors such as mid-life hypertension, which one estimate from the United States holds responsible for 8% of dementias.

The Framingham Study suggests there has been a 20% decrease in dementia incidence each decade since 1975. And the Study provides ample evidence for the better control of blood pressure, and lipids and, of course, substantial reductions in the proportion of people smoking – though there have been increases in some risk factors, notably body mass index (BMI) and diabetes.

In looking at causality it is sometimes best to start with the diagnosis and then go backwards 

Overall, however, analyses of the Framingham data do not support the view that falling CVD risks underlie the reduced incidence of dementias, Carole Dufouil (Bordeaux University Hospital, France, and INSERM) told the symposium. Neither, she said, are the trends accounted for by an increase in the number of years of formal education – which is an alternative hypothesis.

But there was one nugget of evidence to show education may have a profound effect. A presentation on racial disparities (see below) contained the suggestion that a legally-enforced increase in the number of years African Americans attended school in the southern USA had narrowed the difference in dementia incidence when compared with other groups in the population.

We’re not looking at what we think we’re looking at

Most epidemiological research in dementia is actually identifying correlates of childhood cognitive development. It is not really about dementia at all, Richard Jones (Alpert Medical School of Brown University, Providence, Rhode Island, USA) argued. The strongest predictor of cognitive performance late in life is cognitive performance early in life. Data suggest, for example, that IQ at the age 80 is highly correlated with IQ at the age of 11. (Professor Jones believes the r value of 0.66 is remarkably high; and as high as many test/re-test correlations.)

The greater your cognitive ability early in life, the further you have to decline before you cross the clinical threshold for the diagnosis of dementia. The trajectory of disease is about the intercept as well as the slope.

The greater your cognitive ability, the longer you stay above the dementia threshold: IQ aged 11 strongly predicts IQ at 80

It has been hard to gain traction in epidemiological studies of prevention because what we have been studying is in fact childhood development. The best way round the problem is to look at change within individuals in longitudinal studies such as the birth cohorts that have been followed for decades in many European countries.

On the positive side, these ideas point to investment in healthy child development as a critical component in dementia prevention.

Jennifer Manly (Columbia University, New York, USA) said that there is mounting evidence of racial disparities in dementia determinants and outcome but warned of the many methodological challenges faced by such research. These include the unwillingness of some groups to participate in studies or present to memory clinics, due in part to the stigma associated with dementia. There is potential cultural bias in functional assessment and tests of cognition such as word fluency. And there is the possibility of survival bias since different cohorts experience different rates of attrition due to mortality.

We’ve probably still not identified the key drivers of dementia

The challenge of reverse causality

The complexities of interpreting epidemiological data were well illustrated by Mika Kivimäki (University College London, UK). In looking at causality it is often best to start with the diagnosis and then go backwards – something which is possible in longitudinal studies such as that of ten thousand UK civil servants. In the Whitehall study, participants have had risk factors assessed every five years since 1985.

This shows that many people who develop dementia have experienced rising levels of depression over the preceding six years or more. However, if we go back in those same people to the period ten to twenty years before – when we assume neurodegeneration was beginning – levels of depression are not higher. So what at first looks clearly like a causal factor turns out to seem unimportant.

BMI also tells a complicated story. At the time of diagnosis, BMI is lower in people with dementia than in those without the disease; and BMI tends to fall in the period before dementia is diagnosed. You have to go back 15-20 years to find a time when BMI was higher in those people who would go on to develop dementia many years later than in the wider group. Length of follow-up is therefore critical in understanding risk factors.

More generally, Mika Kivimäki was inclined to the same skepticism expressed above by Carole Dufouil in relation to the causes of the apparent fall in dementia incidence. There has been a 15-20% reduction in age-standardized CVD mortality in the period 2005-2015, but dementia mortality fell by less than 3%. So what is good for the heart is not necessarily good for the brain, he said. We’ve probably still not identified they key drivers of dementia.