In almost all countries of the world people of South Asian origin are at increased risk of cardiovascular disease (CVD) when compared with people of European origin.1^–3 These excess rates are particularly marked in the UK, where mortality rates are 30% and 60% higher in Indian and Pakistani male migrants, respectively, than in the general population for coronary heart disease (CHD) and 16% and 41% higher, respectively, for stroke.3 The reasons for this excess mortality have not been explained by increased insulin resistance and its sequelae and other explanations need to be sought.4 The excess mortality in British South Asians runs counter to the prevailing wisdom that a “healthy migrant” effect in first-generation migrants reflects self-selection of the healthiest individuals.5

In England and Wales, this increased risk applies across the adulthood socioeconomic spectrum, whilst the expected gradient in association with lower socioeconomic position (SEP) is now similar to that experienced in the general population of England and Wales. Between 1991 and 1993 Indian men in non-manual occupations had standardised mortality rates (SMRs) for CHD of 143 and those in manual occupations had SMRs of 205 compared with men born in England and Wales.6 A similar pattern was seen for cerebrovascular disease mortality.6 Atherosclerotic CVD is considered to have prenatal origins7 and many factors may contribute cumulatively to its development across the life-course including socioeconomic circumstances and behaviours in both childhood and adulthood.8 9 To our knowledge, no studies have examined the effects of childhood (premigration) SEP in association with adult CVD risk and mortality in British South Asians.

We investigated the effects of markers of childhood and adulthood SEP in a cohort of British South Asian men.

METHOD

Study design and participants

We used data on 1420 South Asian men recruited to the Southall population-based study conducted in 1988–1989 in West London. The men have since been followed for deaths for an average of 17 years. The study was previously described in detail10 11 and approved by Ealing research ethics committee. Main recruitment was from age, ethnicity and sex-stratified random samples from general practice lists. An additional sample (23%) was recruited from four West London factories, chosen on the basis of the ethnic mix of their workforce. Men with serious morbidities were excluded. The overall response rate was 62%. South Asian ethnicity was identified on the basis of name, country of birth and appearance, supplemented by direct enquiry to determine ethnic subgroup.

Baseline measurements

Following an overnight fast, measurements included anthropometry, blood pressure, resting electrocardiogram, blood samples and oral glucose tolerance tests. A self-administered questionnaire included items on lifestyle, medical history, childhood and adulthood SEP. The questionnaire was written in English and appropriate interpreters were available to aid completion where necessary. Childhood SEP markers were based on recall of circumstances at age 12 years of father’s occupation and educational status, presence of bath and/or running water at home, number of people per room, home tenure and land ownership. Since length of education may be considered as a marker of childhood circumstances as well as an independent influence on later risk trajectories, participants were asked to recall their age on starting and leaving full-time education. We used length of education in addition to paternal occupational status, as primary markers of childhood SEP. Markers of adult SEP were own current or most recent occupation at baseline, qualifications/training, supervisory status and home tenure. Occupational status (own and father’s) was coded according to the Registrar General’s Classification of Occupations, 1980.12

Mortality follow-up

All participants were flagged for death notification by the Office for National Statistics between baseline and 15 October 2006, with cause of death coded according to the International Classification of Diseases (ICD 9th or 10th revisions). We defined cardiovascular disease deaths with mention anywhere on the death certificate of ICD-9 codes 390.0–460.0 or ICD-10 codes I100–I990. These codes include coronary heart disease (CHD) (ICD-9 codes 410.0–414.9 or ICD-10 codes I200–I250) and cerebrovascular disease (ICD9 codes 430–438.9 or ICD10 codes I600–I699). Participants may have more than one listed cause of death. A recent tracing exercise (2005), using NHS numbers, indicated that 4% of the South Asian study population were recorded as neither dead nor traceable to addresses in the UK at that time and that six people were recorded as having left the UK permanently (unpublished data).

Analyses

Analyses were conducted in STATA version 9 (Statacorp, College Station, Texas 77845, USA).

Diabetes was defined by World Health Organisation 1999 criteria,13 hypertension was defined as systolic blood pressure ⩾ 140 or diastolic blood pressure ⩾85 mm Hg or treated hypertension. Systolic blood pressure as a continuous variable was ranked and participants with treated hypertension were ranked at the top. Insulin resistance was described by the homeostasis model assessment method (HOMA).14 Baseline characteristics relating to childhood and adulthood SEP, migration patterns and conventional cardiovascular risk factors were examined in the whole group and in each ethnic subgroup. Skewed data were log transformed.

Paternal occupational status was coded as non-manual (class I–IIInm) and included skilled manual (IIIm) and semi/unskilled manual (IV–V). Adult occupational status was coded as for paternal, with an added division of non-manual into professional (I–II) and clerical (IIInm). Due to small numbers of participants and events in some categories, some analyses were conducted on amalgamated categories. We summarised, from least to most favourable, the combined effects of childhood and adulthood SEP by categorising length of full-time education (<11 years and 11+ years) and father’s occupation (manual or non-manual) within three strata of adult occupational status (semi/unskilled manual, skilled manual and non-manual). Eleven years of education represented the upper limit of the lowest tertile. In sensitivity analyses we examined the effects of changing the cut-point to the median (12+ years of education).

Univariate Cox proportional hazards models were used to assess survival in relation to SEP measures in childhood and adulthood. Multivariate analyses explored potential explanatory and confounding risk factors. Variables were entered into models in a forward stepwise progression and retained if the p value was ⩽0.10. Survival analyses were applied to the whole group, as numbers of deaths were small in the ethnic subgroups and there were no between-group differences in mortality rates. However, interactions between SEP measures and being of Sikh or non-Sikh origin were explored (this decision was based on differences in smoking habits and size of groups). The proportional hazards assumption was formally tested in each model using the STATA stphtest and log cumulative hazard curves. Interactions between childhood and adulthood SEP markers were assessed. Linearity of effects of combined childhood and adulthood markers of SEP as a continuous representation of life-course SEP from least to most favourable was assessed using the likelihood ratio test.

RESULTS

Baseline characteristics

All men were migrants to the UK. Their mean age at baseline was 51.2 (SD 7.0) and they were born between 1921 and 1949. Fifty-two percent were Punjabi Sikhs, 15% were Muslim (50% born in Pakistan and 37% born in India), and 9% and 11%, respectively, were Punjabi and Gujarati Hindus. Baseline characteristics of each ethnic subgroup are shown in table 1.

Fewer years of education (<11) were associated with a greater prevalence of adult diabetes (27% vs 20%, p = 0.030), higher systolic blood pressure (age-adjusted median 129 vs 127 mm Hg, p = 0.005) and larger waist circumference (94 vs 93 cm, p = 0.014); for other conventional risk factors associations between years of education and risk factors were trivial. Men in manual occupations were more likely to be current or ex-smokers (non-Sikhs: 50% vs 38%, p = 0.006); we excluded Sikhs from this comparison as their religion prohibits smoking and very few of them had ever smoked. Men in manual occupations had lower fasting triglycerides (1.66 vs 1.79 mmol/l, p = 0.039) and higher systolic blood pressures (125 vs 128 mm Hg, p = 0.075) than those who were in non-manual occupations. There were no other significant differences in baseline risk factors by adult occupational status. (See supplementary table A.)

Mortality

Cardiovascular disease was recorded as a cause of death for 143 of the 217 men who died during the median follow-up period of 17.1 years. Age-adjusted mortality rates for cardiovascular disease did not differ by ethnic subgroup (table 2).

In age-adjusted analyses, each year of residence in England was associated with a 2% increase in risk of cardiovascular disease death (p = 0.062) (table 2), although this association was attenuated on multivariate adjustment.

Men with 11 or more years of education had lower cardiovascular disease mortality rates than those with fewer than 11 years of education (age-adjusted HR 0.66 (95% CI 0.47 to 0.94), p = 0.02). This association was partially attenuated on adjustment for adult occupational status (HR 0.74 (95% CI 0.52 to 1.06), p = 0.10) (table 2), but was not attenuated on adjustment for any other measures of childhood SEP, age on migration, ethnic subgroup or any measured conventional cardiovascular risk factors including smoking. There were no significant interactions between markers of childhood and adulthood SEP.

Men in non-manual occupations in adulthood had lower CVD mortality rates than those in semi-skilled or unskilled manual occupations. There was a clear dichotomy in mortality rates between those in all non-manual occupations (I–IIIn) and those in all manual occupations (IIIm–V) (HR 0.55 (95%CI 0.35 to 0.88), p = 0.01), whilst there was little differential between subcategories within the manual and non-manual groups (table 2). These differences between manual and non-manual groups were unaffected by adjustments for other measures of adulthood SEP, childhood SEP, age on migration, ethnic subgroup or any measured conventional cardiovascular risk factors including smoking.

When childhood and adulthood markers of SEP were combined, cardiovascular disease mortality rates were substantially reduced in those men who were most advantaged in both childhood and adulthood when compared with those who were least advantaged. Hazard ratios for the most advantaged vs least advantaged were similar whether paternal occupational status (HR 0.30, 95% CI 0.11 to 0.85, p = 0.022) or years of education (HR 0.39, 95% CI 0.21 to 0.73, p = 0.003) were used to define childhood SEP (fig 1). These findings were not explained by adjustment for other measured cardiovascular risk factors and did not vary by ethnic subgroup. Men who had 11+ years of education appeared to gain the greatest reduction in CVD mortality in association with adult advantage (fig 1). Similarly, men whose fathers were in non-manual occupations gained a greater reduction in CVD mortality in association with adult non-manual occupations than those whose paternal occupations were manual (paternal occupation  =  non-manual, HR 0.30 (0.11 to 0.85) and paternal occupation  =  manual, HR 0.63 (0.36 to 1.10) (see supplementary table B). The effects of changing socioeconomic circumstances could not be examined effectively due to the small numbers of participants and events in these groups, since the majority of men remained socioeconomically stable between childhood and adulthood.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Age-adjusted hazard ratios for circulatory disease death by years of education and adult occupational status

Exclusion of participants with baseline CHD did not alter univariate or multivariate associations with cardiovascular disease mortality.

DISCUSSION

Socioeconomic disadvantage estimated in childhood and middle age appeared to confer a marked increase in risk of cardiovascular disease death in South Asian men. Conventional risk factors were unable to explain the mechanisms through which these effects operate.

There was an indication of a cumulative effect of childhood and adulthood SEP in predicting cardiovascular disease mortality, and, whilst the effects of adult occupational status appear to partially outweigh those of childhood SEP measures, we observed a marked reduction in CVD deaths in those who were advantaged in both childhood and adulthood, although it should be emphasised that numbers of events were small in these subgroups. Paternal occupational status and length of education were most strongly associated with CVD mortality in men whose adult occupations were non-manual. These findings in South Asian men are consistent with those of the British Regional Heart Study, where the influence of paternal occupation on non-fatal myocardial infarction and CHD was observed only in those men whose adult occupation was non-manual.15 Likewise, a study of Glasgow University students found that paternal occupation on admission to university was a strong predictor of later CVD mortality16 and the Scottish workplace-based Collaborative Study found that CVD mortality risk was influenced by accumulation of risk across the life-course.17

Stronger associations with CVD mortality were observed when duration of education, rather than paternal occupation, was used as the marker of childhood circumstances. We might conclude that in this migrant population length of education is either a better marker of childhood SEP or a more potent determinant of adulthood circumstances, and therefore CVD mortality, than other measured childhood markers including paternal occupation. This latter finding is unsurprising given the difficulty of meaningful categorisation of paternal occupational status in markedly different cultural, economic, geographical and historical circumstances and is consistent with the findings of Davey Smith et al, who found that age of leaving education was a strong predictor of CVD mortality in Scottish men, although outweighed by adult occupational status.18

Most men who were less advantaged in childhood had manual occupations in adulthood. However, it appeared that the majority of men whose childhood circumstances were favourable in terms of years of education and paternal occupation were also in manual occupations in adulthood in England – perhaps a manifestation of the “ethnic penalty” paid by migrants, who are disadvantaged in comparison with equally qualified members of the host population of the same age.19

Previous studies of SEP and health in South Asian migrants to the UK are few and there are no studies which relate childhood SEP to mortality in South Asians. Marmot et al in 1984 found no socioeconomic gradient in mortality between 1970 and 1978.5 As the migrant population has aged, later studies,6 20^–22 like ours, have reported the expected mortality gradient in South Asian migrants in relation to adulthood SEP. A mechanism for this change in the socioeconomic gradient since 1970 has been proposed by Williams et al,23 following a study in 159 young South Asian Glaswegians. Williams hypothesised that the class structure of British South Asians would be broken down on migration and reconstructed with redistribution of class chances over time. Thus, class gradients in migrants’ health would not be visible immediately, but would occur slowly, firstly being manifest in health behaviours and then later in chronic illness and mortality. The Newcastle Health Project conducted between 1993 and 1997 (325 men) lends support to this hypothesis in finding a mixed picture with regard to socioeconomic gradient in anthropometric, biochemical measures and blood pressure in South Asians, although the gradient was becoming established with regard to risk behaviours (physical inactivity and smoking).24 We could also extend the Williams hypothesis to explain the limited and mixed associations between SEP and risk factors and biochemical markers measured at the time of our baseline study. Adult manual occupational status, although associated with smoking (in non-Sikhs) and increased blood pressure, was not associated with other risk factors with the exception of unexpectedly lower triglyceride levels, suggesting that the class gradient in risk factors and biochemical markers was at that time in an interim phase following changes in health behaviours, whilst smoking, as a risk behaviour, had a well-established class gradient in the non-Sikhs.

LIMITATIONS

SEP was measured at two time points only, one of which is based on recall of childhood circumstances from the perspective of middle or late middle age. Applying British interpretations of markers of SEP in South Asian migrants may be inappropriate and may result in underestimation of the effects of childhood SEP. Our markers are crude indicators of lifetime circumstances which undoubtedly involved rapidly changing cultures, language, economics, climate and discrimination. Accuracy of recall of childhood circumstance indicates 91% concordance on paternal educational level and 81% concordance on childhood household social class, with no differences in recall by adult SEP, ethnicity or age.25 However, stronger associations have been observed with death and cardiovascular disease later in life when historical records rather than recall were used to define childhood disadvantage.26 Thus, it is possible that the associations between childhood SEP on CVD mortality are underestimated in our study.

Possible sources of bias arise from the baseline response rate of 62%. It is possible that non-responders may have differed in some way which could affect associations between SEP and mortality, for example they may have been less healthy or less educated than responders, and this is likely to lead to an underestimate of any associations between SEP and mortality. In addition we cannot be sure that we have identified deaths which occurred overseas; however, a recent tracing exercise indicated that only 4% of the study group could not be traced to a UK address, suggesting that very few deaths have been missed.

It should also be noted that numbers of participants and events are very small in some categories and caution is needed in drawing firm conclusions.

A further limitation is the absence of morbidity follow-up; further study is needed to confirm that these findings are consistent for non-fatal cardiovascular disease and such study may elucidate the mechanisms by which life-course SEP affects cardiovascular health in this and other migrant populations.

CONCLUSIONS

This study brings new evidence of the importance of adult occupational status in determining risk of cardiovascular disease death in South Asian migrants to the UK. Years of education, and to a lesser extent paternal occupation, as markers of childhood SEP, had cumulative effects with adulthood socioeconomic circumstances on risk of CVD death; these cumulative effects were strongest in men whose own occupation was non-manual and were unexplained by conventional risk factors measured in middle age. Further studies of morbidity and mortality are needed in order to confirm and explain these findings.