Life-Course Socioeconomic and Behavioral Influences on Cardiovascular Disease Mortality: The Collaborative Study

Abstract

Objectives. This study sought to demonstrate life-course influences on cardiovascular disease (CVD).

Methods. Data were derived from a prospective observational study in which the main outcome measure was death resulting from CVD.

Results. Combining 4 socioeconomic and behavioral risk indicators into a measure of life-course exposure produced 5 groups whose relative risks of CVD mortality ranged from 1.00 (the group with the most favorable life-course exposures) to 4.55 (the group with the least favorable life-course exposures). If the entire study population had had the CVD mortality risk of the subsample with the most favorable risk factor profile, approximately two thirds of cardiovascular deaths would not have occurred.

Conclusions. CVD risk is influenced in a cumulative fashion by socioeconomic and behavioral factors acting throughout the life course.

In 1941, Antonio Ciocco and colleagues concluded that the findings of their follow-up study of Maryland schoolchildren support the common view that “disease in adulthood is often brought about by the cumulative effects over a long period of time of many pathological conditions, many incidents, some of which take place and are even perceived in infancy.”1 This life-course approach to chronic disease lost favor over the subsequent half-century but has recently been revived.2 According to the life-course view, factors acting in early life accumulate and interact with factors acting in later life in the production of adulthood disease.

Cardiovascular disease (CVD) is in many ways the paradigmatic adulthood health problem for the life-course perspective. Genomic and nongenomic intergenerational factors2,4; intrauterine environment5; growth, nutrition, health, and social circumstances in childhood6–9; and a variety of behavioral and socioeconomic factors in adulthood may all contribute to the development of CVD. Most studies to date have focused on only 1 stage of the life course, on only 1 risk factor category, or on cross-sectional measures of CVD risk. Here we demonstrate the importance of life-course influences on CVD in a large cohort of men from western Scotland.

METHODS

This analysis was based on a cohort of 5766 men enrolled in a prospective observational study (the Collaborative Study). Recruited from workplaces in the west of Scotland between 1970 and 1973, respondents were aged 35 to 64 years at the time of entry into the study. Full details regarding the present study population and the methods of data collection used have been provided elsewhere.9,10

Data on sociodemographic characteristics (age, father’s occupation, age at leaving full-time education, occupation at time of screening) and health-related behaviors (tobacco and alcohol use) were collected during baseline examinations, and these data were included in the present study. Alcohol consumption was dichotomized as 15 units or more per week and fewer than 15 units per week. According to previous reports involving this cohort, increases in CVD mortality rates have been associated with alcohol consumption levels above 15 weekly units.11

The home addresses of the men at the time of screening were retrospectively postcoded, enabling us to calculate the Carstairs index—an area-based deprivation measure12—from 1981 Census of Scotland data). Deprivation categories varied from 1 (least deprived) to 7 (most deprived); these values were dichotomized as 1 to 5 and 6 to 7. Participants with missing data on any of the variables just described were excluded from the present analyses, which were based on 5628 participants.

Deaths among study participants are flagged via the National Health Service central registry in Edinburgh; we ascertained mortality rates over a 25-year follow-up period. Here we analyze CVD mortality (International Classification of Diseases, Ninth Revision, codes 390–459).13

Logistic regression was used in calculating age-adjusted odds ratios for associations between variables. Proportional hazards coefficients and their standard errors were calculated after adjustments for age. Exponentiated coefficients were taken as indicators of relative rates of death. Populationattributable-risk fractions were calculated via standard methods.14

RESULTS

The social and behavioral exposures we examined were generally strongly interrelated (Table 1 ▶). People who were disadvantaged with respect to childhood social circumstances (as indexed by father’s social class) were more likely to quit school early, to have manual jobs in later adulthood, to live in deprived areas as adults, to be cigarette smokers, and to have high levels of alcohol consumption.

TABLE 1.

—Associations Between Risk Factors

	Age-Adjusted Odds Ratio (95% Confidence Interval)
	Screening Social Class	Current Smoking	Heavy Alcohol Consumption	Deprivation Category	Education
Father’s social class	8.26 (7.05, 9.69)	1.45 (1.28, 1.64)	1.82 (1.57, 2.10)	3.36 (2.83, 3.98)	8.44 (7.21, 9.88)
Screening social class		1.89 (1.70, 2.11)	2.78 (2.47, 3.14)	4.15 (3.66, 4.72)	9.27 (8.18, 10.51)
Current smoking			1.95 (1.74, 2.20)	1.64 (1.46, 1.85)	1.60 (1.44, 1.79)
Heavy alcohol consumption				1.99 (1.76, 2.25)	2.17 (1.92, 2.45)
Deprivation category					3.35 (2.95, 3.81)

Over the 25-year period examined, 1187 men died of CVD. Men whose fathers were employed in manual occupations had a relative CVD mortality risk of 1.61 (95% confidence interval [CI] = 1.39, 1.88) in comparison with men whose fathers were employed in nonmanual occupations.

Table 2 ▶ presents relative rates of CVD mortality according to father’s social class and later-life risk factors. Behavioral factors (smoking and heavy drinking), age at leaving full-time education, and later-life socioeconomic factors (social class, deprivation category) all provided additional discrimination of CVD mortality risk when they were combined with father’s social class. In all cases, the separate contributions of father’s social class and the other risk indicator were significant at conventional (P < .05) levels. In no cases were there substantial or conventionally significant interactions between father’s social class and the later-life risk indicator.

TABLE 2.

—Age-Adjusted Relative Rates of CVD Mortality, by Father’s Social Class and Later-Life Risk Factors

	Smoking
Father’s social class	Other	Current
Nonmanual
No. (CVD deaths)	703 (75)	657 (133)
Relative rate (95% CI)	1	2.20 (1.66, 2.93)
Manual
No. (CVD deaths)	1811 (339)	2457 (640)
Relative rate (95% CI)	1.80 (1.40, 2.31)	3.11 (2.45, 3.95)

	Alcohol
	< 15 units/wk	≥ 15 units/wk
Nonmanual
No. (CVD deaths)	1065 (158)	295 (50)
Relative rate (95% CI)	1	1.28 (0.93, 1.76)
Manual
No. (CVD deaths)	2847 (621)	1421 (358)
Relative rate (95% CI)	1.53 (1.28, 1.82)	2.13 (1.77, 2.57)

	Screening social class
	Nonmanual	Manual
Nonmanual
No. (CVD deaths)	1148 (161)	212 (47)
Relative rate (95% CI)	1	1.43 (1.03, 1.97)
Manual
No. (CVD deaths)	1691 (354)	2577 (625)
Relative rate (95% CI)	1.56 (1.30, 1.88)	1.85 (1.55, 2.19)

		Deprivation category
	1–5	6–7
Nonmanual
No. (CVD deaths)	1182 (175)	178 (33)
Relative rate (95% CI)	1	1.17 (0.80, 1.69)
Manual
No. (CVD deaths)	2831 (614)	1437 (365)
Relative rate (95% CI)	1.58 (1.34, 1.87)	1.78 (1.49, 2.14)

	Age at leaving full-time education, y
	> 14	≤ 14
Note. CVD – cardiovascular disease; CI = confidence interval.
Nonmanual
No. (CVD deaths)	1126 (165)	234 (43)
Relative rate (95% CI)	1	1.18 (0.84, 1.65)
Manual
No. (CVD deaths)	1630 (299)	2638 (680)
Relative rate (95% CI)	1.48 (1.22, 1.79)	1.77 (1.49, 2.10)

In line with a strategy previously applied to analysis of this cohort,9 we constructed a combined index of situations in which people were disadvantaged with respect to a health risk indicator. Combining 4 risk indicators—father’s social class, participant’s own social class, smoking, and alcohol use—produced a strongly graded risk association (Table 3 ▶). At the extreme ends of the index were participants with no unfavorable indicators (i.e., those whose fathers were employed in nonmanual occupations, who themselves had nonmanual jobs in adulthood, and who neither smoked nor were heavy alcohol drinkers at the time of screening) and those with 4 unfavorable indicators (i.e., those whose fathers were employed in manual occupations, who themselves had manual jobs in adulthood, and who were smokers and heavy drinkers at the time of screening). The intermediate categories simply summed the number of unfavorable risk indicators.

TABLE 3.

—Cardiovascular Mortality According to Cumulative Risk Indicator (Father’s Social Class, Screening Social Class, Smoking, Alcohol Use)

Indicator Profile	Sample, No.	No. of Cardiovascular Deaths	Relative Risk (95% CI)
4 favorable, 0 unfavorable	517	47	1
3 favorable, 1 unfavorable	1299	227	1.99 (1.45, 2.73)
2 favorable, 2 unfavorable	1606	354	2.60 (1.92, 3.52)
1 favorable, 3 unfavorable	1448	339	2.98 (2.20, 4.05)
0 favorable, 4 unfavorable	758	220	4.55 (3.32, 6.24)

Note. CI – confidence interval.

Finally, all 6 socioeconomic and behavioral risk indicators included in these analyses were combined, and a similar gradient of increasing risk was observed. Relative risks, extending from the most favorable exposure group to the least favorable exposure group, were as follows: 1.00, 1.98 (95% CI = 1.39, 2.82), 2.57 (95% CI = 1.82, 3.64), 2.67 (95% CI = 1.89, 3.77), 2.83 (95% CI = 2.01, 3.98), 4.00 (95% CI = 2.84, 5.63), and 4.48 (95% CI = 3.06, 6.55). Population-attributable-risk fraction calculations indicated that if the entire cohort had had the risk indicator profile of the most favorable group, 63.4% of CVD deaths would have been averted. This fraction was slightly different from that produced by combining the 4 risk indicators shown in Table 3 ▶ (62.6%).

DISCUSSION

We have demonstrated that substantial differences in CVD mortality risk exist between groups defined by a small set of socioeconomic and behavioral risk factors. Previous reports have linked cumulative socioeconomic disadvantage to higher levels of CVD mortality,9 morbidity,9,15 and risk factors.15–17 Similar cumulative socioeconomic influences on all-cause mortality,9,18–20 self-rated health,21 and physical, psychological, and cognitive functioning22 have been demonstrated.

In our previous studies of the association between cumulative social disadvantage and all-cause and CVD mortality risks,9 health-related behaviors were used as control variables rather than additional risk indicators. In such presentations, the influence of smoking on socioeconomic gradients can be detected, but the additional contribution of smoking to health outcomes cannot be determined. A few previous studies have described the joint effects on mortality risk of socioeconomic position in adulthood and smoking,23,24 and one study addressed the joint effects of these indicators on carotid intima-media thickness.25

An important issue relates to whether adulthood disease risk results from interactions between early-life and later-life risk indicators or is more straightforwardly influenced by the accumulation of risk across the life course. We previously demonstrated that, in the present cohort, accumulation rather than interaction characterized the effects of childhood and adulthood social circumstances on CVD mortality risk.7 In the present study, we have shown that there are no important interactions between early-life socioeconomic circumstances and behavioral risk factors in adulthood, in keeping with evidence that there are no important interactions between smoking and adulthood social position.23,24 The influence on CVD risk of socioeconomic factors acting across the life course seems to accumulate along with that of 2 important healthrelated behaviors—smoking and heavy drinking—in adulthood.

In previous investigations involving this cohort, we demonstrated that heavy drinking is associated with increased CVD risk (in particular, increased risk of stroke).11 Here we have shown that this association is not dependent on socioeconomic confounding, either by childhood or adulthood social circumstances. In our cohort, alcohol consumption patterns probably took the form of binge drinking (rather than sustained low-level consumption), and such patterns have been associated with increased CVD risk in other studies.26

Favorable exposures and adverse, health-damaging exposures are not randomly distributed across individuals; they generally cluster within particular groups. People who are disadvantaged with respect to a given exposure tend to be disadvantaged with respect to others, as can be seen in Table 1 ▶. There are clear causal chains acting in this regard. Unfavorable childhood social circumstances increase the risk of finishing education with few credentials, which in turn leads to an unfavorable occupational trajectory in adulthood and to membership in social groups that encourage the development and maintenance of certain patterns of health-damaging behaviors.27–28 Interventions that simply select 1 item from this chain—almost invariably 1 of the health-related behaviors—and fail to recognize the societal basis for distribution of risk are unlikely to be successful, as illustrated by the frequent failure of behavioral programs targeted to individuals to reduce CVD risk.30

As mentioned earlier, if our entire study population had exhibited the CVD mortality risk of the subsample with favorable life-course socioeconomic and behavioral factor profiles, nearly two thirds of cardiovascular deaths would not have occurred during the follow-up period. This was the case despite the loss of information entailed by dichotomizing our exposure variables; if more categories had been used, an even greater proportion of cardiovascular deaths could have been attributed to these simple indicators.

We have previously shown that healthrelated selection between childhood and adulthood social positions does not account for CVD mortality differentials31; thus, these risk indicators appear to be exogenous causes (at the distal level, in the case of the socioeconomic measures) of increased cardiovascular mortality. Improving social circumstances and reducing harmful health behaviors could lead to substantial decreases in the population burden of CVD.