Our nationwide population-based longitudinal cohort study analysis shows that continued physical activity in patients with MCI is associated with a lower risk of DAT. It appears that the decision to start physical activity leads to a lower risk of DAT, while ceasing physical activity may cause the risk of DAT to increase again. In addition, a higher frequency of physical activity appears to prevent conversion from MCI to DAT (moderate intensity physical activity more than 5 days per week or vigorous intensity physical activity more than 3 days per week).
We observed that the Maintenance-PA group had 18% fewer dementia conversions than the Never-PA group, while the Initiation-PA group had 11% less dementia conversion than the Never-PA group. We interpret this to indicate that continuing physical activity occurring at both time points was more effective than initiating a new physical activity between the two time points. Evidence suggests that the longer the duration of physical activity, the greater the effect of physical activity on cognitive function. The findings for the Initiation-PA group may therefore reflect the shorter duration of physical activity compared to the Maintenance-PA group.
There are at least two major mechanisms by which continuous physical activity may prevent the conversion from mild cognitive impairment to dementia. Physical activity increases the expression of neurotrophic factors such as brain-derived neurotrophic factor (BDNF), insulin-like growth factor 1 (IGF-1) and vascular endothelial growth factor (VFGF). BDNF is important for maintaining neuronal development and for exercise-related improvements in cognitive function. IGF-1 and VFGF play important roles in neurogenesis and angiogenesis, and influence the induction of hippocampal BDNF. Animal and clinical studies have demonstrated that physical activity can increase neurotrophic factors such as BDNF and IGF-1. After physical activity, the release of BDNF from the brain is enhanced.[17, 18], with 6 months of resistance exercise sufficient to increase serum levels of IGF-1 in older adults.
Physical activity also increases cerebral blood flow (CBF). After 12 weeks of physical activity, CBF has been shown to increase in the anterior cingulate cortex and hippocampal CBF increased in elderly patients with subjective memory complaints after 16 weeks of physical activity. CBF is thought to maintain cerebral perfusion to help maintain brain volume. Previous studies have shown that physical activity is associated with increased regional gray and white matter volume including areas such as the hippocampus, and prefrontal and cingulate cortices.[22, 23]
Ceasing regular physical activity appears to halt the positive effects on the body that the physical activity was eliciting. One small study found that elderly subjects who undertook endurance training for more than 15 years had reduced CBF in specific brain regions including the hippocampus after stopping physical activity. Stopping physical activity also reduces cardiorespiratory fitness, muscle mass, and increases glucose intolerance which might adversely affect cognitive function.[25-27] These adverse effects associated with stopping physical activity arise within 10-20 days. Our findings also showed that the dementia conversion rate was higher in the Initiation-PA group than the Withdrawal-PA group, so the duration of the positive effects of physical activity does not appear to be very long.
We also investigated the association between regularity of physical activity and dementia risk. The Regular-PA group showed a 15% reduction in dementia conversion compared to the Irregular-PA group. However, not all studies have shown that physical activity improves cognitive function, and the effects of exercise on MCI have shown inconsistencies. In a review of 14 randomized controlled trials, the majority had insufficient evidence of the benefits of physical activity on cognitive function in patients with MCI, whereas another review showed that aerobic exercise elicited beneficial effects on cognitive function in MCI patients. A possible explanation for these discrepancies is that each study involved different conditions including the type, intensity, frequency, and duration of physical activity. There have also been three large multidomain trials, FINGER, MAPT, and PreDIVA, which sought to investigate whether lifestyle modifications prevent cognitive impairment. Among these studies, only the FINGER study showed positive results, which could be because it involved active physical training that was different to the PreDIVA and MAPT studies.
Effects of physical activity on cognitive function in demented patients were also unclear. A randomized control study reported that moderate to high intensity exercise training could not slow cognitive impairment in patients with mild to moderate dementia. Many of studies included in a meta-analysis of randomized control trials showed beneficial effects of physical activity on cognitive function in patients with dementia. However, because of unclear methodology, it was difficult for these studies to conclude that the effect of physical activity on dementia patients was proved. On the other hand, several relatively large studies included in the meta-analysis reported no effects on global cognition. 
We note several limitations to our study. First, the evaluation of physical activity was based on self-reporting by the participants and could be open to bias. Second, when we analyzed the association frequency of physical activity with dementia, we could not control for disease severity due to a lack of cognitive function information. However, since the subjects were diagnosed with mild cognitive impairment, the level of baseline cognition would presumably not be very different. Third, the participants’ physical activity characteristics, such as type, intensity, duration, frequency, was free to change during the study period, but we did not take into account the effect of such changes. Moreover, we only measured the presence or absence of physical activity at two distinct time points. Forth, the biomarker information could not be used for diagnosis. From the claim data used in this study, only clinical diagnosis based on ICD-10 could be obtained. Therefore, some of patients with dementia of the Alzheimer type in this study might be non-Alzheimer’s disease patients. Fifth, because of the data access policy, it was not possible to use data for people not included in the study, so it was impossible to perform additional analyzes on people who were not included in the inclusion criteria, such as those who had health examination only once or who did not be diagnosed MCI. In addition, we did not investigate the relation between physical activity and incident MCI. Sixth, although this study had a longitudinal design, it was difficult to have causal inference. Since this was an observational study, possible confounder might cause the effect. Lastly, as the cohort included participants of Korean ethnicity, caution should be taken if generalizing these findings to other populations. However, our study involved a very large sample size of over 247,000 adults. In addition, we focused on the continuity and regularity of physical activity, representing more specific factors that can reduce the risk of DAT.