Paper plus: Physical activity and cardiovascular death

Domhnall MacAuley takes you through this months paper and explains what it means

This months paper is called "What level of physical activity protects against premature cardiovascular death? The Caerphilly study" (Yu et al). It comes from Heart, a cardiology journal published by the BMJ Publishing Group and the British Cardiology Society. To read the paper, click here.

Exercise is good for you. But, how much, what kind of exercise, and for how long? And, would you give the same advice to your granny? The general message about the benefits of exercise is not new; only when we look at the advice from many different sources do we appreciate the differences.

The first guidelines advised 20 minutes of vigorous exercise three times a week. More recent guidelines suggest that moderate exercise on a more regular basis, five days a week, is enough. In addition, these guidelines suggest that exercise can be cumulative and that we can achieve our exercise target with shorter bouts of exercise through the day. Guidelines are usually generated by an expert committee using the best research available. But, guidelines may take a pragmatic approach and, in a public health context, may suggest that moderate exercise will be more acceptable for most people.

Many studies of physical activity have been done. These include cross sectional studies comparing activity levels and risk factors in various population groups such as athletes and older people. The highest quality research study is the randomised controlled trial (RCT), and short term RCTs have looked at the effect of exercise on risk factors. A longer RCT of physical activity with more definite outcomes would be ideal but difficult. Cohort studies follow groups of people over time, measuring the outcome of different exposure to an agent-in this case, exercise. In practical terms, a cohort study probably gives us most of the information.

This cohort study case has the particular advantage of helping disentangle the benefits associated with vigorous activity and moderate activity.

The investigators took a group of people and followed them for an extended period of time to see how many died and what they died from; simple but not easy. The investigators wanted to recruit a group of healthy normal people who would be representative of the middle aged population-some active, some inactive; different occupations and background, etc, but similar to the local population. The study was based in Caerphilly, south Wales. The authors selected an initial sample of 2512 men aged 45-59 years, which represented 90% of the men of this age group in the local population, and followed them for 11 years.

Defining physical activity is not as easy as you might think. The best way to measure activity is to observe people and record every activity. This is not practical with a large number of people. In this study they used a questionnaire. Physical activity questionnaires vary from a few basic questions to a detailed recall diary looking back over weeks, months, or years. But measuring physical activity by questionnaire can be inaccurate, and even a subtle difference in the wording of a question can change the interpretation. The more detailed the diary or questionnaire, the better as it is more likely to represent true activity. Although questionnaires have shortcomings, they are often the best available method. In this study, the authors used the Minnesota leisure time physical activity questionnaire, which asks about the type and duration of physical activity over the previous 12 months. It is a well validated questionnaire that has been used in many studies, and it allowed the researchers to record activities and to apply the corresponding energy equivalents.

The cohort consisted of 2512 men. If you were to think of following up a small group of people-say your class in medical school-over a period of years, you would begin to appreciate the difficulties. People may have moved house, left the district, or even left the country. Circumstances change, and people may not be interested in taking part. It requires an expert and dedicated follow up team to keep a record of a population. In this study, the team followed this group of men for 11 years and examined them at five year intervals. They included some men who had moved into the district and, because they were interested in looking at a people who were healthy at the start, they excluded 393 men who may have had heart disease and 30 men who died within two years and were probably ill at the outset. Looking at any study, it is important that the researchers clearly define the study population to avoid possible bias, and the researchers have been quite explicit in this paper.

EYEWIRE

Carrying a bike up a hill must be rigorous exercise

The outcome measure was death. All the men who took part in this study were notified to the NHS central registry, and the death certificates were examined and coded. In theory this is a good measure, but death certificates are notoriously inaccurate. As a researcher, I know how important accurate death certification is, but, as a general practitioner, I also know that it can be difficult sometimes to know the cause of death and what to write on the certificate. The cause of death cannot be recorded by any means other than by postmortem examination, which is relatively uncommon. Statistical packages allow us to look at the relations between activity and the proportion of the population left at the end of a study. In this case they used a Cox proportional hazards model in survival analysis, but this is where we all need the expert advice of a statistician. Some of the data handling and log transformation in this study are quite complex, but the statistical tests were entirely appropriate.

In reporting any study, we first record the number of people who were included in the study, the numbers lost to follow up, and the number of people excluded. We then usually look at the basic findings for each variable and finally we look at the relation between the different findings. In this study, we are most interested in the relation between level of activity and death and would be tempted to look straight to the end of the results to find this. But, it is important to work our way through the initial results first.

Death and cause of death were the main outcomes used in this study. Those who were ill or had poor health were excluded, leaving an initial sample of 1975. Of those, 252 died; 111 from cardiovascular disease, 82 from coronary heart disease, and 98 from cancer. There were seven deaths from other causes unrelated to the study (congenital abnormalities, injury, or poisoning) and these were excluded from the analysis.

Physical activity was calibrated for each individual and coded as light, moderate, or heavy. Table 1 shows the energy expenditure each day for activities recorded as light, moderate, heavy, and the total of all three groups.

Total energy expenditure is the primary measure of activity and allows us to look at the number of deaths related to energy expenditure (grouped as bands of activity) in table 2. Looking at the death rates we see that there are more deaths from all causes, from cardiovascular, and from coronary heart disease, in the least active groups than in the more active groups. In addition, a trend was confirmed using a statistical test (P<0.05) across activity levels. Interestingly, more men died from cancer in the more active group, but we attach less importance to this as the test for trend was negative.

These findings confirm that physical activity is protective against death and particularly death from cardiovascular and coronary heart disease. But, other factors might affect this. So, the next stage in the analysis is to look at the affect of possible confounding factors. The authors include possible confounding factors such as age and other cardiovascular risk factors into their statistical model and show the risk of death (hazard ratio compared to those where were least active) adjusted for these possible confounding factors in table 3. This confirms that those who were more active had a lower risk than those who were moderately active or least active.

Table 4 is interesting; it shows that if you combine those who were moderately or least active, death rate differs little in relation to activity and is not significant. Table 5 is even more interesting; it looks at those who were classified into the heavy activity group and finds that, even within this group, men who were most active had the least risk. Tables 4 and 5 are remarkable because they suggest little benefit to be gained from moderate activity but that there is a reduced risk in those who are most active and risk is dose related-the more active the better. Table 6 considers work related physical activity. Many of the early studies looked at work activity, but there are now few occupations with a large manual component and this table showed no relation.

EYEWIRE

Games like squash and tennis are rigorous exercise

Yes. Vigorous activity is good for you, but moderate or light activity has no benefit. These results are interesting and important as they suggest that we may be giving the wrong health promotion message. Heavy leisure time physical activity and deaths from all causes, cardiovascular heart disease, and coronary heart disease were significantly and consistently inversely related. But moderate or light intensity leisure time activity had no benefits in terms of all-cause mortality or cardiovascular or coronary heart disease mortality. The UK guidelines are that adults should accumulate at least 30 minutes of moderately intensive activity on at least five days a week. The US surgeon general issues similar guidelines.

This study, and others quoted in this paper, suggest that these guidelines may be wrong. So, when you see guidelines recommending moderate intensity physical activity, you now know that this may be flawed.

We could identify some limitations but, to be fair, this is a good study. The authors list some of the potential weaknesses. One weakness is that the researchers only measured physical activity at baseline and that activity levels may have changed through the duration of the study. And, that lack of activity may be as a result of incapacity rather than the cause of illness. Measuring physical activity is difficult and we will always have doubts about the accuracy of physical activity measurement by questionnaire. But this is one of the best methods we have.

More recently, researchers have tried to validate physical activity measurements using motion sensors to aid activity measurement.

Using cause of death as recorded on death certificates may be inaccurate, but we have no other method. You may also have noticed that the study was of middle aged men only. Interestingly, many of the landmark studies of physical activity include only men, and we can only speculate that the findings apply to women also. Many of the population studies show that many more men than women are vigorously physically active during leisure time.

If moderate activity is of little value, then this has important implications for women. Studies have shown women do less vigorous exercise than men.

EYEWIRE

Researchers classed running as vigorous exercise; brisk walking is moderate

This cohort has been followed up carefully for a number of years and has produced a lot of useful data and spawned many research papers. Two other fascinating papers from this cohort suggest that, not only does vigorous physical activity offer health benefits but that sexual activity seems to have a protective effect on mens health.1 2 You may like to analyse these papers for yourself.

Domhnall MacAuley, associate editor, BMJ
Email: dmacauley@bmj.com

Competing interests: DM has published research on physical activity and have an informal attachment to the department that carried out this work.

studentBMJ 2003;11:175-218 June ISSN 0966-6494

1. Davey Smith G, Frankel S, Yarnell J. Sex and death: are they related? Findings from the Caerphilly Cohort Study. BMJ 1997;315:1641-4.
2. Ebrahim S, May M, Ben Shlomo Y, McCarron P, Frankel S, Yarnell J, et al. Sexual intercourse and risk of ischaemic stroke and coronary heart disease: the Caerphilly study. J Epidemiol Community Health 2002;56:99-102.