Skip navigation

Exercise and mortality

Benefits of cohort studies

Is Bandolier obsessed with systematic reviews of randomised controlled trials? Surely not! Systematic reviews and meta-analysis may be the best way of assessing the effectiveness of interventions, but other study architectures can help understanding, especially when more than one study points in the same direction for the same effect.

Cohors, cohortis

For the classically minded we have previously given the definition of a cohort ( Bandolier 24 ). This might conjure up a picture of Roman legionaries tramping to the limits of the Empire (and if you want to see just how good their shoes were, check out some of the museums along Hadrian's Wall).

The point, though, is that they did walk everywhere, and up until quite recently daily walks to work or school of several miles were not remarkable. They are now, which is why a study of the effects of walking in non-smoking retired men in Hawaii [1] is important.

What a difference a mile makes

The study examined 707 non-smoking retired men of Japanese ancestry aged 61 to 81 years who were enrolled in the Honolulu Heart Program, which has been going since 1965. When they enrolled between 1980 and 1982 (then aged 45 to 68 years) they had a physical examination. They were asked about the average distance walked each day.


The average distance walked each day was 1.8 miles (2.9 km). In the 12 years of follow up there were 208 deaths - 33 from heart disease, 19 from stroke, 68 from cancer and 88 from other causes. The death rate was examined according to whether men walked less than 1 mile, 1 to 2 miles, or more than 2 miles a day.

There were no differences between these three groups in terms of cholesterol, HDL, weight, hypertension, diabetes, diet or alcohol consumption. After 10 years, walking patterns were checked, and confirmed that men, by and large, maintained their walking patterns.

The less men walked, the more likely they were to die (Figure).

The reduction in mortality came not just from reduced heart disease or stroke, but was also from cancer and other causes of death. The risk of death in men who walked less than 1 mile a day was 1.8 times that of men who walked more than 2 miles a day.

Put another way, the 21% chance of dying over 12 years in men who walked more than 2 miles was reached in just seven years in men who walked less than 1 mile, and in 10 years in those who walked 1-2 miles a day.

It is even possible to put a crude NNT on it. If 43% of men who walked less than 1 mile a day died over 12 years, compared with 22% of those who walked more than 2 miles, the NNT is 1/(0.43-0.22) = 4.8. That is for every five men who walk at least two miles a day, one fewer will die over 12 years compared with those who walked less than one mile a day.

Nature or nurture

Does the inverse relationship between physical activity or walking and mortality hold just for men in Hawaii? A comprehensive study of twins in Finland tells us that activity has effects beyond genetics, and for women as well as men.

The study [2] examined the physical activity of just under 8,000 male and 8,000 female twins in 1975 (all born before 1958). Exclusions were chronic disease and deaths from injuries, suicide and homicide. After a comprehensive questionnaire subjects were classified into three broad bands of physical leisure activity:

  • Sedentary subjects - not participating in leisure time physical activity.
  • Occasional exercisers - participated in physical activity less than six times a month.
  • Conditioning exercisers - exercising at least six times a month for a mean duration of 30 minutes with an intensity corresponding to vigorous walking or jogging.


Overall the death rate between 1977 and 1994 was 12% for those who were sedentary, 7.4% for occasional exercisers, and 4.9% for conditioning exercisers. Twin pairs discordant for death were examined to determine whether mortality of physically active subjects differed from that of their sex- and age-matched sedentary siblings. The result (Figure) showed that physical activity conferred a lower risk of death (odds ratio below 1) when compared with those who were sedentary in 1975 and adjusted for baseline smoking occupational group and alcohol use.


These studies demonstrate the benefit of cohort studies in identifying risk factors. The direction and magnitude of the effect of moderate exercise is similar in both studies: over 12 years walking two miles a day reduced the risk of death by half in the Hawaiian study, while in Finland over 17 years moderate exercise reduced the risk by 60%. Two different studies, in different populations, using different cohorts, but with much the same answer.

Encouraging people to walk or take moderate exercise is likely to benefit their health - and the magnitude of that benefit is large. Bandolier calculated an NNT of about 5 for the Hawaiian study. Compare that to NNTs of 10 at best for statins in secondary prevention.

This might be useful information in the new Health Service. With more involvement of Local Authorities, perhaps more can be done to develop safe, well-lit, and pleasant places to walk. If anyone wants to know how this might be done, take a day trip and see some of the delightful small Dutch towns with their traffic-free streets and interesting town-centres.

Finally it is good to report that people are already acting on this sort of information. There is one general practice in the UK which has already received a grant to study the benefits of walking when prescribed by general practitioners.


  1. AA Hakim, H Petrovitch, CM Burchfield et al. Effects of Walking on mortality among nonsmoking retired men. New England Journal of Medicine 1998 338: 94-9.
  2. UM Kujala, J Kaprio, S Sarna, M Koskenvuo. Relationship between leisure-time physical activity and mortality. The Finnish twin cohort. JAMA 1998 279: 440-4.

previous or next story in this issue