This Week in BMJ | Editor's Choice | Press releases | Advertisement details

Paper Saturday 22 February 1997

Lifetime socioeconomic position and mortality: prospective observational study

George Davey Smith, Carole Hart, David Blane, Charles Gillis, Victor Hawthorne

See editorial by Haines and Smith

Abstract

Objectives: To assess the influence of socioeconomic position over a lifetime on risk factors for cardiovascular disease, on morbidity, and on mortality from various causes.

Design: Prospective observational study with 21 years of follow up. Social class was determined as manual or non-manual at three stages of participants' lives: from the social class of their father's job, the social class of their first job, and the social class of their job at the time of screening. A cumulative social class indicator was constructed, ranging from non-manual social class at all three stages of life to manual social class at all three stages.

Setting: 27 workplaces in the west of Scotland.

Participants: 5,766 men aged 35-64 at the time of examination.

Main outcome measures: Prevalence and level of risk factors for cardiovascular disease; morbidity; and mortality from broad causes of death.

Results: From non-manual social class locations at all three life stages to manual at all stages there were strong positive trends for blood pressure, body mass index, current cigarette smoking, angina, and bronchitis. Inverse trends were seen for height, cholesterol concentration, lung function, and being an ex-smoker. 1,580 men died during follow up. Age adjusted relative death rates in comparison with the men of non-manual social class locations at all three stages of life were 1.29 (95% confidence interval 1.08 to 1.56) in men of two non-manual and one manual social class; 1.45 (1.21 to 1.73) in men of two manual and one non-manual social class; and 1.71 (1.46 to 2.01) in men of manual social class at all three stages. Mortality from cardiovascular disease showed a similar graded association with cumulative social class. Mortality from cancer was mainly raised among men of manual social class at all three stages. Adjustment for a wide range of risk factors caused little attenuation in the association of cumulative social class with mortality from all causes and from cardiovascular disease; greater attenuation was seen in the association with mortality from non-cardiovascular, non-cancer disease. Fathers having a non-manual occupation was strongly associated with mortality from cardiovascular disease: relative rate 1.41 (1.15 to 1.72). Participants' social class at the time of screening was more strongly associated than the other social class indicators with mortality from cancer and from non-cardiovascular, non-cancer causes.

Conclusions: Socioeconomic factors acting over the lifetime affect health and risk of premature death. The relative importance of influences at different stages varies for the cause of death. Studies with data on socioeconomic circumstances at only one stage of life are inadequate for fully elucidating the contribution of socioeconomic factors to health and mortality risk.

Introduction

The ubiquitous nature of the association between unfavourable socioeconomic circumstances in adulthood and premature death has been shown in many studies, with various socioeconomic indicators being related to morbidity and mortality.1-6 It has long been recognised that the effects of poor social circumstances in early life can have lasting influences,7-8 and studies relating childhood socioeconomic position to later risk  of illness and premature death have been performed.9-12 These studies have generally been inspired by suggestions that the early environment has specific influences which alter later susceptibility to disease. They have yielded somewhat equivocal findings, well illustrated by two reports from the same study coming to opposite conclusions about the relative importance of childhood and adulthood socioeconomic environment in relation to the risk of coronary heart disease.10 13

Recently the importance of considering the cumulative effect of socioenvironmental exposures over a life time has been recognised.14-16 The Department of Health's report Variations in Health concluded that "it is likely that cumulative differential lifetime exposure to health damaging or health promoting physical and social environments is the main explanation for observed variations in health and life expectancy."17 Few empirical data exist on such cumulative effects, however. In the national longitudinal survey of older men in the United States education, first occupation, occupation in middle age, and family assets contributed independently to the risk of premature death,18 but other data on risk factors were not available and patterns for different causes of death have not been examined. We investigated the contributions of indicators of socioeconomic position over the course of a lifetime to the risk of premature death in a large cohort of men in the west of Scotland who had detailed examinations in middle age and for whom 21 years of subsequent mortality follow up data were available.

Subjects and methods

This analysis is based on a cohort of men recruited from 27 workplaces in Glasgow, Grangemouth, and Clydebank between 1970 and 1973. The workplaces included engineering, manufacturing, and petrochemical plants; a publishing house; civil service departments; administrative and professional divisions from British Rail; architectural institutes; legal and dental offices; and banks. Response rates were available for the workplaces from which 87% of the sample was recruited. For these sites 70% of those invited completed the questionnaire and attended for examination. The achieved sample was 6,022 men and 1,006 women, from which the 5,766 men aged 35-64 at the time of examination are the basis for the present report. The mortality data in the women were not analysed as only 176 women had died.

The examinations used similar procedures to those used and previously described for the Renfrew-Paisley study population.19 An extensive questionnaire, completed by the subject, was checked at the screening examination. These examinations were conducted at clinics established for this purpose at or near the workplaces.

The information collected at baseline examination included:

Sociodemographic data - age, father's occupational social class, the social class of the participant's first regular job, excluding temporary work, and his occupational social class at the time of screening, and whether he drove a car regularly

Medical history - angina, previous heart attack and intermittent claudication from the Rose ques- tionnaire,20 respiratory symptoms from the Medical Research Council questionnaire21

Health related behaviours - detailed smoking history

Physical examination - height, weight, blood pressure while seated (taken with a London School of Hygiene and Tropical Medicine sphygmomanometer, diastolic pressure being recorded at the disappearance of the fifth Korotkoff sound), lung function measured with the Garthur Vitalograph, serum cholesterol concentration, and six lead electrocardiogram (leads I, II, III, aVR, aVL, and aVF).

Persistent phlegm was defined as usually bringing up phlegm from the chest first thing in the morning on most days for three months during winter each year. "Infective phlegm" was defined as usually bringing up phlegm from the chest first thing in the morning in winter and having had a period of increased cough and phlegm lasting for three weeks or more in the previous three years. Breathlessness was defined as a positive response to the question: "Do you get short of breath walking with people of your own age on level ground?" Bronchitis was defined as having persistent and "infective" phlegm and being breathless.20

Angina was considered present if chest pain or discomfort when walking uphill or hurrying was cited in the sternum or the left chest and arm; caused the subject to stop or slow down; went away when the subject stopped or slowed down; and went away in 10 minutes or less.20

Forced expiratory volume in one second (FEV1) was taken as the better of two expirations. To estimate impairment the expected value was obtained from linear regression equations of age and height, in which expected FEV1 = -278.06 + 4.33 x height - 3.06 x age. Age was measured in years and height in centimetres and the value was divided by 100 to give FEV1 in litres. The coefficients were derived from a healthy subset of the population who had never smoked and answered no to questions on phlegm and breathlessness and to questions asking whether they had wheezy or whistling chest and whether their breathing was affected by the weather. The adjusted FEV1 was calculated as (actual FEV1 divided by expected FEV1) x 100

The electrocardiograms were coded according to the Minnesota system.20 Any of the codes 1.1-1.3, 4.1-4.4, 5.1-5.3, and 7.1, which encompass diagnoses of definite myocardial infarction, myocardial ischaemia, and left bundle branch block, were considered to be evidence of ischaemia.

Data on father's, first, or current social class were missing in 199 men, who were excluded from the present analyses, which are therefore based on 5567 participants. The home address at the time of screening was retrospectively given a postcode, enabling an area based deprivation category at the time of the 1981 census as defined by Carstairs and Morris to be ascertained.22 Deprivation category varies from 1 (least deprived) to 7 (most deprived) and is based on four variables derived from census data at the level of the postcode sector: male unemployment, overcrowding, car ownership, and proportion of residents in social classes IV and V.

We identified men who had died over the 21 years of follow up by flagging at the NHS Central Registry in Edinburgh, which also provides death certificates coded according to the ninth revision of the International Classification of Diseases. We calculated mortality using a person years at risk lifetable approach. These rates have been standardised for age by the direct method, with the total study population as the standard. We tested for trends in age adjusted rates by proportional hazards regression using the proc phreg program in the SAS system,23 with age and the cumulative social class indicator (defined in the Results section) coded as a continuous variable as covariates. We standardised continuous variables for age using the proc glm program, with tests for trend for age adjusted means being obtained with the proc reg program. We standardised categorical variables for age by a direct method and tested for trends with the proc logist program.

Proportional hazards coefficients and their standard errors were calculated using Cox's model.24 Adjustment for age and other risk factors was performed by including terms for these in the proportional hazards models. Exponentiated hazards coefficients were taken as indicators of relative death rates.

Results

Table 1 shows the distribution of social class at different points during the lives of the study population. The largest category consisted of men whose fathers had had manual social class occupations, whose first job had been in a manual social class occupation, and who had a manual social class occupation at the time of screening. The only group that was particularly small consisted of men with fathers who had had a non-manual social class occupation, whose first job had been in a non-manual social class occupation, and who had a manual social class occupation at the time of screening.

Table 1 - Distribution of social class according to manual or non-manual occupation of study population
Social class at screening	First social class
	Non-manual	Manual
Manual
Father's social class:
Non-manual	34	177
Manual	247	2319
Non-manual
Father's social class:
Non-manual	956	165
Manual	923	746

A cumulative social class indicator was constructed by summing the number of occasions in which each participant's class location was manual or non-manual. In table 2 population characteristics are presented according to this cumulative social class indicator. The men in whom social class was manual on three occasions were on average one year older than the men in whom class was non-manual on all three occasions. After the remaining data on risk factors were adjusted for age we found that moving from the men who had consistently non-manual social class locations to those who had consistently manual social class locations showed trends of increasing systolic and diastolic blood pressure, body mass index, current cigarette smoking, and prevalence of bronchitis and angina and decreasing cholesterol concentration, height, lung function, and prevalence of never having smoked. The men consistently in manual social class locations were less likely than those consistently in non-manual social class locations to be regular drivers. Being a regular car driver is taken to be mainly a marker for ownership of a car, which has been used in various studies as an indicator of available income.3-4 The men consistently in manual social class locations were also more likely to live in a postcode sector with unfavourable socioeconomic characteristics at the time of the 1981 census.

Table 2 - Population characteristics according to cumulative social class. Values are age adjusted means or proportions unless stated otherwise
	Cumulative social class based on occupation
	All three non-manual (n=956)	Two non-manual, one manual (n=1122)	Two manual, one non-manual (n=1170)	All three manual (n=2319)	Total (n=5567)	P value for trend
Age at screening (years)*	47.5	47.7	48.6	48.6	48.2	>0.0001
Blood pressure (mm Hg):
Systolic	131.1	132.2	134.1	136.1	134.1	>0.0001
Diastolic	82.4	83.1	84.1	84.7	83.9	>0.0001
Cholesterol (mmol/l)	6.3	6.1	5.8	5.7	5.9	>0.0001
Height (cm)	176.0	174.4	172.6	170.8	172.8	>0.0001
Body mass index	24.7	25.1	25.4	25.3	25.2	>0.0001
FEV1 (%)	99.4	98.0	94.0	90.3	94.2	>0.0001
Percentage (No) of subjects:
With angina	4.4 (39)	5.4 (57)	6.5 (77)	7.5 (176)	6.3 (349)	0.0002
With electrocardiographic ischaemia	5.9 (55)	5.2 (57)	6.1 (74)	5.9 (143)	5.9 (329)	0.64
With bronchitis	0.6 (5)	0.5 (6)	2.9 (35)	3.4 (79)	2.2 (125)	>0.0001
Who had never smoked+	20.0 (194)	22.1 (254)	18.1 (209)	14.5 (330)	17.7 (987)	>0.0001
Who were current smokers+	45.4 (435)	46.8 (522)	55.4 (649)	63.5 (1471)	55.3 (3077)	>0.0001
Who were ex-smokers+	30.6 (289)	27.7 (306)	22.9 (270)	20.5 (482)	24.2 (1347)	>0.0001
Average No of cigarettes smoked per day‡	19.6	18.3	19.2	18.9	19.0	0.58
Percentage (No) of subjects:	69.3 (669)	57.5 (646)	54.3 (632)	40.8 (938)	51.8 (2885)	>0.0001
Who drove cars regularly
Living in area of residence of deprivation category 5-7	18.6 (173)	32.0 (354)	54.3 (638)	66.0 (1539)	48.6 (2704)	>0.0001
*Not age adjusted. +Excluding 154 men who smoked a pipe or cigars and two men with missing data. ‡For current cigarette smokers only.

Over the 21 years of follow up 1,580 members of the cohort died. Table 3 shows age adjusted mortality from all causes, cardiovascular disease, cancer, and other causes. Mortality increased with the extent to which the men had had manual social class locations at different times in their life. The trends were most consistent for mortality from all causes and cardiovascular disease; for mortality from cancer the main difference was between men who had had manual social class locations at all three stages of their life compared with the rest.

Table 3 - Age adjusted death rates (per 1000 person years) over 21 years of follow up according to cumulative social class*
	Cumulative social class based on occupation
	All three non-manual	Two non-manual, one manual	Two manual, one non-manual	All three manual
All causes
No of deaths	190	280	339	771
Mortality	100.1	123.4	135.1	153.9
Cardiovascular causes
No of deaths	84	145	197	388
Mortality	48.3	70.3	84.4	86.6
Cancer
No of deaths	73	86	95	247
Mortality	41.8	42.4	45.2	58.0
Non-cardiovascular, non-cancer causes
No of deaths	33	49	47	136
Mortality	19.1	25.7	22.6	34.0
*Significance tests for trend across age adjusted relative rates are given in table 6.

We fitted proportional hazards models with social class coded as manual or non-manual at the three different stages of life fitted individually and simultaneously. The social class of first occupation was not significantly associated with mortality from all causes or any subcategory of mortality when father's and current social class were taken into account (table 4). Father's social class and current social class were similarly related to mortality from all causes, both when treated singularly and when entered in the model with simultaneous adjustment for all social class measures. Within this general picture mortality from cancer and non-cardiovascular, non-cancer causes seemed to be more strongly associated with social class at the time of screening than with father's social class; the reverse was the case for mortality from cardiovascular disease.

Table 4 - Mortality by social class at three different stages of life. Values are age adjusted relative rates (95% confidence intervals), with individual and simultaneous adjustment for each social class indicator
	Father's social class	First social class	Current social class
All causes
Individual	1.44 (1.27 to 1.64)	1.29 (1.16 to 1.43)	1.40 (1.27 to 1.55)
Simultaneous	1.28 (1.11 to 1.47)	1.01 (0.89 to 1.16)	1.29 (1.14 to 1.47)
Cardiovascular causes
Individual	1.58 (1.32 to 1.89)	1.35 (1.16 to 1.56)	1.38 (1.20 to 1.59)
Simultaneous	1.41 (1.15 to 1.72)	1.08 (0.90 to 1.30)	1.20 (1.01 to 1.43)
Cancer
Individual	1.26 (1.02 to 1.56)	1.25 (1.04 to 1.50)	1.35 (1.13 to 1.61)
Simultaneous	1.11 (0.87 to 1.41)	1.04 (0.82 to 1.31)	1.28 (1.03 to 1.60)
Non-cardiovascular, non-cancer causes
Individual	1.45 (1.07 to 1.98)	1.18 (0.92 to 1.53)	1.59 (1.24 to 2.03)
Simultaneous	1.28 (0.91 to 1.80)	0.80 (0.58 to 1.10)	1.67 (1.22 to 2.28)

Two additional indicators of socioeconomic position in adulthood were available: car use and deprivation category of area of residence. Table 5 shows that these indicators are associated with mortality within the strata of cumulative social class. In analyses adjusted only for age not being a regular car driver was associated with a relative rate of mortality of 1.34 (95% confidence interval 1.21 to 1.48); inclusion of the cumulative social class indicator reduced this to 1.25 (1.13 to 1.39). Living in an area in deprivation categories 5-7 was associated with an age adjusted relative rate of mortality of 1.21 (1.10 to 1.34). After inclusion of the cumulative social class indicator this became 1.07 (0.96 to 1.19).

Table 5 - All cause mortality by cumulative social class, car driving, and deprivation category of area of residence. Values are age adjusted relative rates (95% confidence intervals)
	Cumulative social class
	All three non-manual	Two non-manual, one manual	Two manual, one non-manual	All three manual
Regular car driver:
Yes	1	1.28 (1.01 to 1.63)	1.36 (1.08 to 1.73)	1.57 (1.27 to 1.95)
No	1.22 (0.91 to 1.64)	1.52 (1.19 to 1.95)	1.76 (1.40 to 2.21)	2.00 (1.64 to 2.44)
Deprivation category of area of residence:
1-4	1	1.25 (1.01 to 1.56)	1.37 (1.09 to 1.72)	1.70 (1.39 to 2.09)
5-7	1.06 (0.74 to 1.52)	1.41 (1.10 to 1.82)	1.54 (1.25 to 1.90)	1.74 (1.45 to 2.09)

Table 6 shows relative rates of mortality adjusted for age and risk factors. For mortality from cardiovascular disease and all causes the relative rates were adjusted for age, smoking (number of cigarettes smoked, with an additional adjustment for being an ex-smoker), diastolic blood pressure, serum cholesterol concentration, body mass index, adjusted FEV1, angina, bronchitis, and electrocardiographic ischaemia. For mortality from cancer and from non-cardiovascular, non-cancer causes the relative rates were adjusted for age, smoking, body mass index, adjusted FEV1, and bronchitis. Such adjustment led to some attenuation in the increased risk of death in relation to social class, although for mortality from all causes, cardiovascular disease, and cancer this attenuation was not great. For mortality from non-cardiovascular, non-cancer causes more considerable attenuation of the trend of increasing risk of death with cumulative social class group was seen, with the association becoming non-significant at conventional levels.

Table 6 - Relative death rates (95% confidence intervals) by cumulative social class, adjusted for age and risk factors
	Cumulative social class				P value for trend
	All three non-manual	Two non-manual, one manual	Two manual, one non-manual	All three manual
All causes
Age	1	1.29 (1.08 to 1.56)	1.45 (1.21 to 1.73)	1.71 (1.46 to 2.01)	>0.0001
Age and risk factor*	1	1.30 (1.08 to 1.57)	1.33 (1.11 to 1.60)	1.57 (1.33 to 1.85)	>0.0001
Cardiovascular causes
Age	1	1.51 (1.16 to 1.98)	1.90 (1.47 to 2.45)	1.94 (1.53 to 2.45)	>0.0001
Age and risk factor*	1	1.57 (1.20 to 2.05)	1.78 (1.37 to 2.31)	1.92 (1.51 to 2.45)	>0.0001
Cancer
Age	1	1.04 (0.76 to 1.42)	1.06 (0.78 to 1.44)	1.44 (1.11 to 1.87)	0.001
Age and risk factor+	1	1.04 (0.76 to 1.42)	1.01 (0.74 to 1.37)	1.34 (1.03 to 1.75)	0.009
Non-cardiovascular, non-cancer causes
Age	1	1.31 (0.84 to 2.03)	1.16 (0.75 to 1.82)	1.75 (1.20 to 2.56)	0.002
Age and risk factor+	1	1.30 (0.83 to 2.02)	1.02 (0.65 to 1.60)	1.42 (0.97 to 2.09)	0.09
*Adjusted for age, smoking, diastolic blood pressure, cholesterol concentration, body mass index, adjusted FEV1, angina, bronchitis, electrocardiographic ischaemia. +Adjusted for age, smoking, body mass index, adjusted FEV1, and bronchitis.

Discussion

Cumulative socioeconomic disadvantage and mortality
Most of the numerous studies relating socioeconomic position to mortality have used measures of socioeconomic circumstances in adulthood.1-5 This in part reflects the view that exposures acting at this time have the most impact on the risk of premature death.8 However, the risk of premature death may reflect the accumulation of environmental insults or the cumulative effects of unfavourable behavioural or psychological factors, which progressively increase susceptibility to disease.25-26 The few studies that have assessed socioeconomic position throughout life show the strengths of this notion. Thus in a study based on record linkage of the 1960, 1970, and 1980 censuses in Norway particularly high risks of premature death were seen among men who had had limited education and then worked in manual occupations and lived in poor housing.27-28 Similar findings have come from the national longitudinal survey of older men in the United States.18

Our cohort was recruited from workplaces, but the study was initiated at a time of comparatively low unemployment and recruited men from across the social spectrum. The mortality differentials according to social class in our study are similar to those in men of the same age group in Scotland around 1981, the midpoint of the follow up period. For men aged 55 to 64 in the general Scottish population the death rates were 66% higher in social classes IV and V than in social classes I and II.29 In our study the age adjusted death rate was 61% higher in men of social classes IV and V than in men of social classes I and II. The general population data also show mortality differentials for the main causes of death that are similar in size to those seen in our study. Our workplace sample thus seems to be a reasonable model for studying factors underlying socioeconomic differentials in risk of premature death in the general population, and the non-response rate, which is comparable to that in other such studies, does not seem to have introduced any serious bias. Our findings (a) that socioeconomic position in early and later life contribute separately to the risk of premature death and (b) that the risk can be further differentiated by adding additional adult socioeconomic indicators (use of a car and area based deprivation category) to the cumulative social class indicator are likely to be generalisable to other populations.

The cumulative social class indicator shows graded associations with most of the risk factors and morbidity measures included in our study. Strong associations with mortality were also evident. Adjustment for risk factors measured in adulthood attenuated the association of social class with mortality from cancer and from non-cancer, non-cardiovascular causes more than it did that with mortality from cardiovascular disease. This may partly reflect the fact that a major risk factor for coronary heart disease - serum cholesterol concentration - was higher in the group with the more favourable socioeconomic experience.

The use of a cumulative social class indicator does not take into account directions of social class change. Thus a participant whose father had had a manual occupation, whose first occupation had been manual, and who had a non-manual occupation at screening would be in the same group as a participant whose father had had a non-manual occupation and who had manual occupations both at first and at screening. Other studies have suggested that social mobility is not an important contributor to overall mortality differentials,30 and detailed analyses of our data (to be reported elsewhere) show this too. We repeated the mortality analyses using all combinations of social class and found that the associations of mortality with cumulative social class were generally not dependent on the order in which different social classes came (basic tables available from us).

Socioeconomic position in childhood and mortality
There has been particular interest in the association between living conditions in childhood and risk of coronary heart disease in adult life. This follows from the work of Forsdahl, who suggested that early deprivation followed by later affluence increased the risk of coronary heart disease, an effect in part mediated by an increase in blood cholesterol concentrations.31-32 The attribution of increased risk of coronary heart disease to an interaction between poor socioeconomic circumstances in early life and later affluence has received little support from subsequent studies.9-10 33 Similarly, the suggestion that the effects of deprivation in early life are mediated through high blood cholesterol concentrations in adulthood has not been substantiated in later investigations.33 In our cohort men with fathers of a manual social class had lower, rather than higher, serum cholesterol concentrations than men with fathers of a non-manual social class.34 In contrast to these negative assessments of the Forsdahl hypothesis, the basic notion that unfavourable socioeconomic conditions in childhood predispose to increased risk of coronary heart disease in adulthood has received more support. Most9 11 13 33 35-37 but not all10 38 studies have found an association of childhood socioeconomic circumstances with risk of coronary heart disease, which was apparently not purely due to the adverse social conditions in adulthood of those born into poor circumstances. In a Swedish census follow up study men with fathers who had manual occupations had considerably higher risk of dying from coronary heart disease than had those whose fathers had non-manual occupations.35 For mortality from all causes this was much less evident, mortality being dependent on social class in adulthood much more than social class in childhood. This particular dependence of the risk of coronary heart disease on socioeconomic circumstances in childhood has also been observed in area based studies from Finland.39-40 Our study had similar findings, father's social class being particularly important for mortality from cardiovascular disease but not for mortality from cancer or non-cardiovascular, non-cancer causes. Analyses of the association between height and cause specific mortality, in which height is taken to be an indicator of childhood circumstances, show similar specificity.41 These findings are clearly relevant to the hypothesis that fetal development is associated with the risk of cardiovascular disease in later life42 as parental social class will influence early development. Socioeconomic factors in childhood will also influence growth, and recent evidence suggests that poor growth in childhood is also associated with higher mortality from cardiovascular disease in adulthood.43 Studies with data covering all stages of development are needed to determine which stages of development most affect the risk of cardiovascular disease in adulthood.

Conclusions
Our data show a clear cumulative effect of socioeconomic circumstances acting over a life time. Combining the cumulative social class index with additional indicators of socioeconomic position in adulthood led to further differentiation of the risk of premature death. This has important implications for studies that try to control for socioeconomic factors when analysing outcomes in relation to socially patterned exposures. Single measures of adult social class, traditionally used in such studies, will not adequately capture the full extent of socioeconomic differentials between groups with different exposures. Statistical adjustment for these single measures will therefore not control for socioeconomic differences, and apparently independent risk relations may remain confounded by factors related to socioeconomic environment.44

Specific patterns emerge within the general picture of higher death rates among people with less favourable socioeconomic trajectories during their lives.

Key messages

Health and risk of premature death are determined by socioeconomic factors acting throughout life Socioeconomic influences on particular causes of death may have different critical times The risk of premature death from cardiovascular disease is particularly sensitive to socioeconomic influences acting in early life Studies with data on socioeconomic circumstances at only one stage of life are inadequate for fully elucidating the contribution of socioeconomic factors to health

Firstly, mortality from cardiovascular disease seems to be more strongly related to cumulative social disadvantage than does that from cancer or non-cardiovascular, non-cancer causes.

Secondly, whereas social class in adulthood is the more important socioeconomic indicator over a life time for differentiating groups with differing risks of mortality from cancer and non-cardiovascular, non-cancer causes, the socioeconomic environment in childhood seems to be particularly important with respect to mortality from cardiovascular disease. These findings should help direct the attention of disease specific aetiological research to influences acting both in childhood and in adult life.

The Department of Health's report Variations in Health has directed attention to the accumulation of socially patterned adverse exposures over a life time.17 Our results add to the as yet limited data that show the necessity of such an approach. Any serious attempt to elucidate the contributions of socially distributed risk factors to the risk of disease in adulthood should aim to collect information covering the entire lifespan of study participants.14

Funding: The investigation of socioeconomic determinants of mortality in this cohort is supported by a grant from the NHS Management Executive, Cardiovascular Disease and Stroke Research and Development Initiative.

Conflict of interest: None.

References

1 Davey Smith G, Bartley M, Blane D. The Black report on socioeconomic inequalities in health 10 years on. BMJ 1990;301:373-7.

2 Moser K, Pugh H, Goldblatt P. Mortality and the social classification of women. In: Goldblatt P, ed. Longitudinal study: mortality and social organisation. London: HMSO, 1990:145-62.

3 Goldblatt P. Mortality and alternative social classifications. In: Goldblatt P, ed. Longitudinal study: mortality and social organisation. London: HMSO, 1990:163-92.

4 Davey Smith G, Shipley M J, Rose G. The magnitude and causes of socio-economic differentials in mortality: further evidence from the Whitehall study. J Epidemiol Community Health 1990;44:260-5.

5 Davey Smith G, Neaton J D, Wentworth D, Stamler R, Stamler J. Socioeconomic differentials in mortality risk among men screened for the multiple risk factor intervention trial. I. Results for 300,685 white men. Am J Public Health 1996;86:486-96.

6 Eachus J, Williams M, Chan P, Davey Smith G, Grainge M, Donovan J, et al. Deprivation and cause-specific morbidity: evidence from the Somerset and Avon survey of health. BMJ 1996;312:287-92.

7 Burnett J C. The prevention of hare-lip, cleft-palate, and other congenital defects: as also of hereditrary disease and constitutional taints by the medicinal and nutritional treatment of their mother during pregnancy. Homæpathic World 1880 Oct 1:437-51.

8 Kuh D, Davey Smith G. When is mortality risk determined? Historical insights into a current debate. Social History of Medicine 1993;6:101-23.

9 Burr M L, Sweetnam P M. Family size and paternal unemployment in relation to myocardial infarction. J Epidemiol Community Health 1980;34:93-95.

10 Lynch JW, Kaplan GA, Cohen RD, Kauhanen J, Wilson TW, Smith N L, et al. Childhood and adult socioeconomic status as predictors of mortality in Finland. Lancet 1994;343:524-7.

11 Gliksman M D, Kawachi I, Hunter D, Colditz G A, Manson J E, Stampfer M J, et al. Childhood socioeconomic status and risk of cardiovascular disease in middle aged US women: a prospective study. J Epidemiol Community Health 1995;49:10-5.

12 Elo I T, Preston S H. Effects of early-life conditions on adult mortality: a review. Population Index 1992;58:186-212.

13 Kaplan G A, Salonen J T. Socioeconomic conditions in childhood and ischaemic heart disease during middle age. BMJ 1990;301:1121-3.

14 Ben-Shlomo Y, Davey Smith G. Deprivation in infancy or adult life: which is more important for mortality risk? Lancet 1991;337:530-4.

15 Mann S L, Wadsworth M E J, Colley J R T. Accumulation of factors in influencing respiratory illness in members of a national birth cohort and their offspring. J Epidemiol Community Health 1992;46:286-90.

16 Davey Smith G, Bartley M, Blane D. Explanations for socio-economic differentials in mortality: evidence from Britain and elsewhere. Eur J Public Health 1994;4:131-44.

17 Department of Health. The health of the nation: variations in health: What can the Department of Health and the NHS do? London: DoH, 1995.

18 Mare R D. Socio-economic careers and differential mortality among older men in the United States. In: Vallin J, D'Souza S, Palloni A, eds. Measurement and analysis of mortality: new approaches. Oxford: Clarendon Press, 1990:362-87.

19 Hawthorne V M, Watt G M C, Hart C L, Hole D J, Davey Smith G, et al. Cardiorespiratory disease in men and women in urban Scotland: baseline characteristics of the Renfrew/Paisley (Midspan) study population. Scott Med J 1995;40:102-7.

20 Rose G A, Blackburn H, Gillum R F, Prineas R J. Cardiovascular survey methods. Geneva: World Health Organisation, 1982.

21 Medical Research Council. Definition and classification of chronic bronchitis for epidemiological purposes. Lancet 1965;i:775-9.

22 Carstairs V, Morris R. Deprivation and health in Scotland. Aberdeen: Aberdeen University Press, 1991.

23 SAS Institute. SAS language and procedures: usage 2, version 6. Cary, N C: SAS Institute, 1991.

24 Cox D R. Regression models and life tables. Journal of the Royal Statistical Society Series B 1972;34:187-220.

25 Jones HB. A special consideration of the ageing process, disease and life expectancy. Advances in Biology and Medical Physics 1956;4:281-337.

26 Alter R, Riley J. Frailty, sickness and death: models of morbidity and mortality in historical populations. Population Studies 1989;43:25-46.

27 Salhi M, Caselli G, Duchêne J, Egidi V, Santini A, Thiltgès E, Wunsch G. Assessing mortality differentials using life histories: a method and applications. In: Lopez A, Caselli G, Valkonen T, eds. Adult mortality in developed countries: from description to explanation. Oxford: Clarendon Press, 1995.

28 Wunsch G, Duchêne J, Thiltgès E, Salhi M. Socio-economic differences in mortality. A life course approach. European Journal of Population 1996;12:167-85.

29 Office of Population Censuses and Surveys. Occupational mortality 1979-1980, 1982-1983. London: HMSO, 1986.

30 Blane D, Davey Smith G, Bartley M. Social selection: what does it contribute to social class differentials in health? Sociology of Health and Illness 1993;15:1-15.

31 Forsdahl A. Are poor living conditions in childhood and adolescents an important risk factor for arteriosclerotic heart disease? Br J Prev Soc Med 1977;31:91-5.

32 Forsdahl A. Living conditions in childhood and subsequent development of risk factors for arteriorsclerotic heart disease. J Epidemiol Community Health 1978;32:34-7.

33 Notkola V, Punsar S, Karvonen M J, Haapakaski J. Socio-economic conditions in childhood and mortality and morbidity caused by coronary heart disease in adulthood in rural Finland. Soc Sci Med 1985;21:517-23.

34 Blane D, Hart C L, Davey Smith G, Gillis C R, Hole D J, Hawthorne WM. The association of cardiovascular disease risk factors with socioeconomic position during childhood and during adulthood. BMJ 1996;313:1434-8.

35 Vågerô D, Leon D. Effect of social class in childhood and adulthood on adult mortality. Lancet 1994;343:1224-5.

36 Gillum R F, Paffenbarger R S. Chronic disease in former college students. XVII. Socio-cultural mobility as a precursor of coronary heart disease and hypertension. Am J Epidemiol 1978;108:289-98.

37 Wannamethee S G, Whincup P H, Shaper G, Walker M. Influence of father's social class on cardiovascular disease in middle-aged men. Lancet 1996;348:1259-63.

38 Hasle H. Association between living conditions in childhood and myocardial infarction. BMJ 1990;300:512-3.

39 Valkonen T. Male mortality from ischaemic heart disease in Finland: relation to region of birth and region of residence. European Journal of Population 1987;3:61-83.

40 Koskinen S. Origins of regional differences in mortality from ischaemic heart disease in Finland. Helsinki: National Research and Development Centre for Welfare and Health, 1994. (NAWH search report 41.)

41 Leon D, Davey Smith G, Shipley M, Strachan D. Height and mortality in London: early life influences, socio-economic confounding or shrinkage? J Epidemiol Community Health 1995;49:5-9.

42 Barker D J P. Early nutrition and coronary heart disease. In: Davies DP, ed. Nutrition in child health. London: Royal College of Physicians of London, 1995:77-85.

43 Gunnell D, Davey Smith G, Frankel S, Nanchahal K, Braddon FEM, Peters TJ. Childhood leg length and adult mortality - follow up of the Carnegie survey of diet and growth in pre-war Britain. J Epidemiol Community Health 1996;50:580-1.

44 Davey Smith G, Phillips AN. Confounding in epidemiological studies: why "independent" effects may not be all they seem. BMJ 1992;305:757-9.

(Accepted 18 December 1996)

Department of Social Medicine,
University of Bristol,
Bristo BS8 2PR
George Davey Smith, professor of clinical epidemiology

Department of Public Health,
University of Glasgow,
Glasgow G12 8RZ
Carole Hart, statistician

Academic Department of Psychiatry,
Charing Cross and Westminster Medical School,
London W6 8RP
David Blane, senior lecturer in medical sociology

West of Scotland Cancer Surveillance Unit,
Ruchill Hospital,
Glasgow G20 9NB
Charles Gillis, director

University of Michigan,
School of Public Health,
Department of Epidemiology,
109 Observatory Road,
Ann Arbor,
MI 48109,
USA
Victor Hawthorne, professor of epidemiology

Correspondence to: Professor Davey Smith.