Reflections on null findings—the case of physical activity and risk of dementia

Mika Kivimäki and Archana Singh-Manoux

Guidelines for dementia prevention highlight the role of lifestyle; it is estimated that one in three dementia cases worldwide could be prevented by changes in lifestyle habits. Physical activity is a recommended target for a simple, low cost strategy to reduce the burden of dementia. According to observational studies, individuals who are physically active have a 20% to 40% lower risk of dementia compared with those who are physically inactive.

Yet it is widely acknowledged that the evidence supporting this recommendation is uncertain, as no interventional study to date has been large enough or long enough to demonstrate the benefits of physical activity in the prevention of dementia. In the DAPA trial of dementia patients, cognitive decline was actually slightly more rapid in the exercise group than in the usual care arm. However, few researchers have challenged the role of physical activity for dementia prevention.

In observational studies, a single measure of physical activity cannot separate the people for whom this reflects typical behaviour from those whose behaviours have been modified during the long preclinical phase of dementia. Our analysis of repeat measures of physical activity, starting in midlife, in 10 000 men and women showed declining physical activity in the last 10 years leading up to dementia diagnosis. In these analyses there was no association between physical activity and dementia when the measurements of exposure and disease onset were separated by more than 10 years.

We undertook confirmatory analyses using individual level data from 19 long term cohort studies and 400 000 participants. A 1.4-fold association between physical inactivity at baseline and subsequent dementia was evident when the measurement of physical activity was less than 10 years before dementia diagnosis. Yet, again, no association was seen when we measured physical activity more than 10 years before dementia onset. In the same data, physical inactivity was associated with an increased risk of diabetes, coronary heart disease, and stroke irrespective of the length of the follow-up.

Null findings are rarely popular. Experts weigh in on the inadequate measurement of the exposure or outcome to explain away the findings. Although physical activity in our study was self-reported, it did show the expected associations with incident diabetes, coronary heart disease, and stroke. Finding out whether a patient developed dementia by checking electronic health records may also be seen to be problematic, as it detects only diagnosed cases, not dementia biomarkers. However, this is unlikely to be a source of major bias in our findings as diabetes, coronary heart disease, and stroke did, as expected, increase the risk of dementia. In addition, the short term associations between physical inactivity and dementia in our study were similar to those in previous studies with more comprehensive dementia ascertainment.

An important issue, seldom addressed in studies of dementia, is that an in depth assessment of dementia status based on biomarkers (amyloid, tau, atrophy) or an in depth clinical evaluation requires continued long term participation. Results based on such a selected set of participants may not be generalisable. Record linkage studies have the advantage of analyses on every participant in the study.

We are intrigued by studies with a long follow-up that report a strong association between physical inactivity and dementia. For example, a highly cited study of 1450 older adults from Finland found that physical activity at least twice a week halved dementia risk in a 20 year follow-up. But in the same study, sedentariness was not associated with BMI, lipids, or blood pressure and, in addition, physically inactive participants were less likely to develop diabetes. More recently, a well known Swedish cohort study with a 40 year follow-up concluded that low fitness levels explain 80% of the dementia burden. This estimate of the population attributable fraction for a single risk factor seems implausible given that the estimate in other studies for all lifestyle factors together is only 35%.

We are not arguing that physical activity is not important for health. Our results support a physically active lifestyle as a way to reduce the risk of cardiometabolic disease. But in light of current evidence, promises that physical activity could significantly reduce an individual’s risk of dementia are, in our view, unfounded. Null findings are important as they highlight the need to identify new prevention targets and inform interventions to test causality.

Professor Mika Kivimäki is director of the Whitehall II study at UCL and primary investigator of the IPD-Work consortium. Twitter @MikaKivimaki

Archana Singh-Manoux is research professor at Inserm, France, and leads the cognitive ageing component of Whitehall II.

Competing interests: We declare no competing interests