Study: The link between physical fitness and reduced risk of mortality may be misleading

That fit people have a reduced risk of premature death from various diseases is a recurring result in many studies. New research from Uppsala University shows that people with high fitness levels in their late teens also have a reduced risk of dying from random accidents. This suggests that the associations seen in previous studies have probably been misleading.

Many observational studies have shown that people who exercise more and have good cardiorespiratory fitness early in life are at lower risk of premature death from causes such as cancer and cardiovascular disease. However, a new study published in the European Journal of Preventive Cardiology suggests that the association between physical fitness and a reduced risk of mortality may be misleading. 

We found that people with high fitness levels in late adolescence had a lower risk of dying prematurely, for example from cardiovascular disease, compared to those with low fitness levels. But when we looked at their risk of dying in random accidents, we found an almost similarly strong association. This suggests that people with high and low fitness levels may differ in other important ways, which is something that previous studies have not fully taken into account."

Marcel Ballin, associated researcher in epidemiology and lead author of the study 

Based on conscription data from over 1 million men

In the study, the researchers leveraged data from 1.1 million Swedish men who were conscripted for military service between the years 1972 and 1995. The men, who were on average 18 years old at the time of conscription, were divided into five groups based on their fitness level at the time. They were then followed until their 60s or until they died. With access to the National Cause of Death Register, the researchers were able to see their cause of death. They subsequently used different methods to study the association between fitness level in late adolescence and premature death.

The researchers started with a traditional analysis of mortality from cardiovascular disease, cancer and from all causes, as in previous observational studies. They adjusted their statistical models for factors such as BMI, age at conscription, year of conscription, and parents' income and education level. The results showed that the group with the highest fitness level had a 58 per cent lower risk of dying from cardiovascular disease, a 31 per cent lower risk of dying from cancer, and a 53 per cent lower risk of dying from all causes, compared with the group with the lowest fitness level.

Very similar risk of dying in random accidents

Next, the researchers examined how fitness was associated with the risk of dying in random accidents such as car accidents, drownings and homicides. They chose random accidents because they assumed that there ought to be no association between the men's fitness in late adolescence and the risk of dying in random accidents. This method is called negative control outcome analysis and involves testing the validity of your results for a primary outcome by comparing them with an outcome where no association ought to be found. If, however, an association is found, it may indicate that the groups studied are not actually comparable, and that the study suffers from what is typically referred to as confounding. The researchers found that men with the highest fitness levels had a 53 per cent lower risk of dying in random accidents. Yet, it is unlikely that the men's fitness would have such a big effect on their risk of dying in random accidents. 

These results were also confirmed when the researchers used the sibling comparison design. Using this method, the researchers compared the risk of premature death between siblings with different fitness levels to control for all the factors that the siblings share such as behaviours, environmental factors, and some genetic factors.

"It surprised us that the association with accidental mortality reflected the other associations, even after we controlled for all the factors that siblings share. This underlines how strong the assumptions are that you make in observational studies, since it appears to be very difficult to create comparable groups. The consequences may be that you overestimate the magnitudes of the effects you find," says Marcel Ballin.

Picture confirmed in other studies

The study is one of the largest of its kind in which researchers used negative control outcomes to investigate whether the associations between fitness and mortality are in fact valid. The results in this study are also supported by other research. 

"That the effects of good cardiorespiratory fitness may be overstated might sound controversial to some, but the fact is that if you look at the results from studies others than traditional observational studies, a more nuanced picture does emerge. A number of twin studies for example have found similar results. Some genetic studies also suggest that there are genes that affect both the propensity to be physically active or have a good fitness level, and the risk of developing diseases such as cardiovascular disease," Marcel says.

Important to base interventions on correct estimates

Marcel Ballin also argues that there are many different reasons for promoting physical activity. However, large-scale interventions or policy changes intended to apply to the entire population must be based on reliable estimates – otherwise there is a risk of expecting effects that have in fact been overestimated.

"Our results should not be interpreted as if physical activity and exercise are ineffective or that you should not try to promote it. But to create a more nuanced understanding of how big the effects of fitness actually are on different outcomes, we need to use several different methods. If we just ask the question in the same way, we will always get the same answer. It's only when we get the same answer to a question that we have asked in slightly different ways that we can be sure that the findings are accurate," says Marcel Ballin.