Aetiology of coronary heart diseaseBMJ 2001; 323 doi: https://doi.org/10.1136/bmj.323.7324.1261 (Published 01 December 2001) Cite this as: BMJ 2001;323:1261
Fetal and infant growth and socioeconomic factors in adult life may act together
- Michael Marmot (), professor
- International Centre for Health and Society, Department of Epidemiology and Public Health, University College London, London WC1E 6BT
Papers p 1273
It is a common tendency in science to argue that in order for my pet hypothesis to be correct yours has to be wrong. Indeed, it is somewhat unsatisfying for both yours and mine to be right at the same time. Barker's work on the fetal and infant origins of coronary heart disease has stimulated such reactions.1 The two positions are that coronary heart disease is determined by factors associated with growth in utero or in the first year of life or that it is caused by social conditions and lifestyle acting later in life. Among the arguments for the former is the body of work that Barker and colleagues have produced over 15 years.2 Arguing for the importance of contemporaneous influences is that coronary heart disease rates can be reduced within five years by reducing cholesterol levels,3 giving up smoking,4 or reducing blood pressure levels.5 Rates of coronary heart disease are strongly related to social class6 and changed rapidly in response to the social and economic crises affecting the former Soviet Union and other former socialist countries of central and eastern Europe.7
So which is it? Are rates of coronary heart disease determined by factors related to fetal and infant growth or by influences acting on adults as shown by differences in coronary heart disease according to measures of socioeconomic position? In order to assess independent effects we need to hold one variable constant while assessing the effect of the other.
Even as a thought experiment this is not quite realistic. People are not randomly assigned, in adult life, to better or worse social conditions. Where they started in life affects where they end up. Under real life conditions, it has been difficult to say which is more important, early life or later life. The evidence has been mixed. 8 9
The latest paper by Barker et al in this issue (p 1273) suggests that independence of early and later life effects is the wrong way to look at the problem.10 The two types of influence may interact. Barker and colleagues followed a cohort of Finnish men, born in a Helsinki hospital in 1934-44, to determine the predictors of incident coronary heart disease. Among the variables associated with incident coronary heart disease were low ponderal index or thinness at birth, low weight at 1 year of age, low social class of father, low level of education achieved, low adult social class, and low income in adult life. These variables are, however, correlated with each other. In a multivariate analysis, they showed that low weight at 1 year of age, low education, and low adult social class were independent predictors of incident coronary heart disease. The other variables dropped out.
This type of multivariate analysis is useful, but its results need to be interpreted in the light of a causal model. Barker and colleagues make clear in their paper the importance of causal thinking. They report that only 12% of boys whose fathers were labourers reached the highest educational level compared with 37% of other boys. Father's social class may drop out of the multivariate model when education is included because of the strong link between them. Father's social class may be causally important because of its strong association with boys' educational achievement.
The same thinking does not apply to income. In a two way analysis they show that income is unrelated to the incidence of coronary heart disease once education is taken into account. One cannot therefore argue that education affects risk of coronary heart disease because boys with high education become men with high income. Income does not predict at all within educational strata. Adult social class, based on occupation, does predict independent of education and father's social class. Barker and colleagues enter the debate on material versus psychosocial explanations for inequalities in health 11 12 by suggesting that this set of findings fits a little better with a psychosocial explanation than a strictly material one.
The most interesting finding of the paper, however, is not the independence of childhood and adult conditions but their interaction. This can be looked at in two ways. Thinness at birth (low ponderal index) was related to coronary heart disease incidence only in the presence of low adult social class. Alternatively, low adult social class was related to coronary heart disease more strongly in men who had been thin at birth. The relation of adult social class to coronary heart disease was particularly strong in the subset of men, thin at birth, whose body mass index increased relative to the average during the first 12 years of life.
The thrust of the paper is not to explain away the link between social position and coronary heart disease, but to try to understand how it operates. In showing the importance of both early life conditions, adult conditions, and their interaction, this paper rises above the debate. It will reinforce the call to look at the influence of conditions throughout life in determining social inequalities in disease in adulthood.13