Article Text
Abstract
Objective: To update a systematic review on the association between childhood socioeconomic circumstances and cause-specific mortality. Studies published since 2003 include a far greater number of deaths than was previously available justifying an update of the previous systematic review.
Methods: Individual-level studies examining childhood socioeconomic circumstances and adult overall and cause-specific mortality published between 2003 and April 2007.
Results and conclusions: The new studies confirmed that mortality risk for all causes was higher among those who experienced poorer socioeconomic circumstances during childhood. As already suggested in the original systematic review, not all causes of death were equally related to childhood socioeconomic circumstances. A greater proportion of new studies included women and showed that a similar pattern is valid for both genders. In addition, the new studies show that this association persists among younger birth cohorts, despite temporal general improvements in childhood conditions across successive birth cohorts. The difficulties of establishing a particular life-course model were highlighted.
Statistics from Altmetric.com
Four years ago we conducted a systematic review evaluating the association between childhood socioeconomic circumstances and adult mortality.1 Up to that point a small number of studies had included women. It was also possible that this association would diminish among younger cohorts who have not been exposed to the same sort of childhood hardships as previous cohorts, many of whom had been born in the early decades of the 20th century. There have now been a number of additional studies that shed some light on these issues. Moreover, some of these new studies are very large. In total there are almost 100 000 new deaths available (total number of deaths in newly published studies = 99 165; 66 558 deaths in a Swedish study2 and 20 887 deaths in a study from Norway3), which is far greater than the number of deaths previously available (total number of deaths in the original systematic review was 26 796; numbers of deaths were not available in three studies).1 They provide greater power to establish whether childhood socioeconomic position is associated with specific causes of death and justify an update of the previous systematic review. Finally, some newer studies have specifically attempted to investigate the mechanisms and pathways that could explain this association.
METHODS
The same search criteria used in the original systematic review were applied to the update.1 We retrieved individual-level studies examining childhood socioeconomic circumstances and adult overall and cause-specific mortality published between 2003 and April 2007 (full details on the search strategy are available from the authors). In addition, all articles citing the original systematic review and the reference list of all new articles were searched for new publications. The criteria for inclusion were the same as for the original systematic review. Two studies were excluded from this update. A re-analysis of the collaborative study and an analysis of the British Regional Heart Study analysed fatal and non-fatal coronary heart disease (CHD) jointly and were therefore not included here.4 5 A study conducted within the framework of the Polish branch of MONICA grouped father’s and own educational level as a measure of social class and was also excluded.6 The list of all articles, those included in the initial review and those included in this update, is provided in the appendix on the JECH website. In this article we will refer to the new articles included in this update.
As for the original review we did not perform formal meta-analysis, as the necessary conditions of comparability of exposures and outcomes, together with homogeneity of association direction and strength, were not met.7
RESULTS
We located 11 studies and five updates of previously published studies that included longer follow-up, more detailed causes of death or performed additional analysis mainly aimed at evaluating mechanisms explaining the association between childhood socioeconomic circumstances and adult mortality. All were prospective studies. Four different publications reported2 8–10 the association in similar birth cohorts from Sweden, although there were slight differences in the actual birth cohorts included. Similarly, there were two different reports results from a Norwegian population-wide sample.3 11 The sample differences between these studies were small and we considered them as reporting the same population. They appear in the tables as one new study and the results of the most complete study are summarised in the tables on the JECH website.
Most studies evaluated northern European populations, although there were for the first time reports from France, Belgium and South Korea. Finally, three additional studies reported results from the United Kingdom and two from the United States. Seven of the 11 new studies included men and women, one included women only and three only men. All re-analyses of previous studies were based on male populations.
All studies, except the French cohort, reported an independent inverse association between childhood socioeconomic position and adult mortality—that is, those who experienced the worse socioeconomic circumstances during childhood had a higher overall mortality risk in adulthood. Not all causes of death were associated with early life circumstances. These included some cancers, such as breast cancer, and external causes of death.
Dying from stroke was equally related to early and adult socioeconomic position (SEP) independently of the type of stroke. Higher suicide mortality was observed among Norwegian women whose families had higher income during their childhood.3 11 12 The Swedish study found an association in the same direction although the confidence intervals around the estimate could not rule out a chance finding.2
Interestingly, an adoptees study showed that the biological and not the adoptive parents’ SEP was associated with higher mortality.13 The only exception was suicide risk for which there was weak evidence for a higher risk among adoptees whose adoptive parents had non-manual occupations.
Adjusting for adult SEP accounted, in most studies, for some or all the association. The two biggest studies reported that the person’s education was the main intermediate explaining the association between early life socioeconomic circumstances and later mortality.2 3 Similarly, in the Oslo Mortality Study, own education explained, to a large extent, the effect of the sanitary conditions and the economic deprivation indices in childhood.14 IQ level did partly explain the socioeconomic gradient in the Scotland twenty-07 study.
DISCUSSION
All studies published since the original systematic review confirmed that mortality risk for all causes of death was higher among those who experienced poorer socioeconomic circumstances during childhood, except for the Gazel study from France. A greater proportion of these new studies included women and showed that this general pattern is valid for both genders. As already suggested in the original systematic review, these new studies confirmed that not all causes of death were equally related to childhood socioeconomic circumstances.
In the large Swedish study, men and women of lower SEP in childhood had higher risk of dying from stomach, liver and lung cancer, diabetes, CHD, stroke, respiratory diseases and diseases of the digestive system. In addition, men with poor socioeconomic background had higher mortality as a result of nervous system conditions, alcoholic cirrhosis, unintentional injuries and homicide.2 Consistent with previous reports,15 the new studies found childhood SEP was more strongly associated with stomach cancer mortality than adult SEP,2 3 16 17 adding support to the hypothesis that exposure to Helicobacter pylori early in life results in stomach cancer in adulthood.18 On the other hand, the studies from Sweden and Norway, and the mothers of the 1958 cohort found both types of stroke, ischaemic and haemorrhagic, had a similar social patterning, thus not supporting earlier reports where worse childhood SEP was a stronger predictor for haemorrhagic stroke.2 3 14 16
These new studies have highlighted that the health effects of poor socioeconomic conditions in childhood persist among younger birth cohorts despite not generally having experienced the level of socioeconomic strain previous birth cohorts had. The magnitude of the relative risk associated with different causes of death, comparing younger with older cohorts, tended to be similar, despite the limitations of comparing the magnitudes of hazard ratios (HRs) from different studies adjusted for a variety of different confounders. Among the young cohorts the age-adjusted HR and 95% confidence interval (CI) for all-cause mortality, comparing either the HR of manual versus non-manual father’s occupation2 19 or using the relative index inequality,3 was 1.27 (1.24 to 1.30) in the Swedish cohort2 and 1.41 (1.34 to 1.48) in the Norwegian cohort.3 The equivalent HR for an older cohort, such as the collaborative study, was 1.44 (1.27 to 1.63).19 Adjustment for adult SEP and adult risk factors diminished this association similarly in all studies. CHD mortality was 1.54 (1.45 to 1.64) in the Swedish study,2 2.17 (1.89 to 2.49) in the Norwegian cohort3 and was 1.52 (1.24 to 1.87) in the older cohort.19 Again, similar reductions occurred in all studies after multivariable adjustments. Although the magnitude of the age-adjusted HR associated with stomach cancer mortality was smaller in the Swedish cohort 1.32 (1.10 to 1.59),2 it was similar and even slightly higher in the Norwegian cohort 2.34 (1.54 to 3.54)3 compared to that of the older cohort of the collaborative study 2.06 (0.93 to 4.57).19 As already mentioned it is extremely difficult to extract conclusions when comparing the actual magnitude of the association between younger and older birth cohorts. Not only did different studies measure SEP differently and adjust for different confounders, but also the relative contribution of CHD risk factors and childhood SEP may be different by age (age effect), by cohort (birth cohort effect) or by period (period effect). In addition, these are likely to be related to each specific context (for example, country). The most appropriate conclusion should probably be limited to the fact that the association is found in both younger and older birth cohorts.
Osler et al evaluated childhood circumstances in two consecutive generations and found poorer socioeconomic circumstances were associated with higher mortality among those born in 1953,20 but also among their parents.21 This study allowed them to investigate whether the effect of socioeconomic conditions was transmitted across two generations. They found a linear increase in mortality risk in the cohort born in 1953 with increasing number of parents and grandfathers of working class.20 This association remained after adjusting for own adult occupational class. These results show that the consequences of poor socioeconomic circumstances are cumulative across generations. The fact that the influence of socioeconomic position can be traced through more than one generation may partly explain why this association persists in younger cohorts who have experienced generally better childhood socioeconomic conditions than their parents.
Warner and Hayward using data from the National Longitudinal Survey of Older Men, reported that father’s occupation in addition to family structure, explained in part the race gap in overall mortality between black people and white people.22 Black men’s higher rates of death were associated with worse socioeconomic circumstances during early life, including not living with the biological parents.22 This effect was mediated through educational achievement and adult SEP.
Data from areas or countries where this association had not been previously reported found similar results. Data from South Korea showed that lower father’s education predicted a higher mortality in adulthood, whereas there was some evidence of a higher mortality risk for those with fathers in manual occupations although it did not reach conventional statistical significance.23 The authors point out the limitation of using a manual/non-manual dichotomy for a cohort who were most probably working as agricultural labourers and which may not accurately measure the social hierarchy in South Korea.23 The effect of paternal education was partly accounted for by a comprehensive array of adult measures of SEP. In a historical study from Ardennes (Belgium) wealth during childhood continued to be an important determinant of mortality at age 50.24 The first data available from France, from the Gazel cohort study, showed that sustained socioeconomic disadvantage throughout life was associated with higher mortality before 65 years of age, but father’s occupation was not independently associated with premature mortality.25 In this study worse childhood SEP resulted in higher mortality as a result of breast cancer among women, which has not been found in other studies. The authors suggest that worse survival rather than a higher incidence is the most likely explanation.25 The association between childhood SEP and overall mortality is likely to differ between countries that have different patterns of mortality. Mortality due to cancer is the main cause of death in France, which experiences the lowest CVD mortality in Europe.26 CVD mortality is the main cause of death in most other countries included in this review.27 This review has shown that childhood SEP is not related to adult cancer mortality, therefore it is plausible that currently childhood SEP is not independently associated with overall mortality in France.
As suggested in the original review, recalled socioeconomic measures of childhood in adulthood tended to underestimate the true association. An updated analysis of the Kuopio study found that objective measures of childhood SEP collected during childhood were more accurate than those recalled from adulthood, with the objective childhood SEP indicator conferring a higher adult mortality, particularly for CVD and CHD mortality and acute coronary events.28 Conversely, a previous report that used reported childhood socioeconomic data recalled in adulthood did not find such an association.29
Different indicators of childhood SEP were used in different studies. As we had pointed out in other publications in this journal: “There is no single best indicator of SEP suitable for all study aims and applicable at all time points in all settings. Each indicator measures different, often related aspects of socioeconomic stratification and may be more or less relevant to different health outcomes and at different stages in the life course. The choice of SEP measure(s) should ideally be informed by consideration of the specific research question and the proposed mechanisms linking SEP to the outcome.”30 31 This point is relevant for both adult and childhood SEP indicators.
Several studies investigated potential mechanisms to explain how early life socioeconomic circumstances related to adult mortality. The difficulty of teasing out a particular life course model (cumulative, social mobility or critical period) that could explain the association between childhood SEP and adult mortality has been highlighted in several publications32 33 and was also stressed in the analysis of the Sweden population living in Scania.10 The authors reported that all models fitted equally the observed risks and highlighted the inherent impossibility in distinguishing, for example, “the effect of social mobility out of the workforce per se from the effect of being outside the workforce”.10 Moreover, not all theoretical combinations of social classes and trajectories throughout the life course do in fact occur, thus limiting the ability to actually empirically differentiate particular life course models.
Osler et al evaluated whether early life SEP related to adult mortality through genetic, prenatal or post-natal family environment by studying the effect that paternal social class from the adoptive and the biological father had on the adopted person’s mortality.13 They found that adoptees with biological fathers with high social class had a lower rate of late mortality (after the fifth decade), mainly because of lower CVD, infectious and respiratory disease mortality, although only the latter outcome reached conventional levels of statistical significance. Adoptive father’s SEP was not related to the person’s future mortality. The only exception was a higher suicide risk, particularly among adoptive families of high income. Own social class in adulthood did not modify this pattern and made only slight modifications to the magnitude of the associations. Thus, the authors suggest that genetic or environmental factors occurring prenatally are more likely to explain the association between childhood socioeconomic circumstances and adult mortality than post-natal family socioeconomic environment.13 However, some limitations make the interpretation of this study difficult. Social class among the biological fathers did not predict their own mortality and among the adoptive fathers it showed only a “tendency for an inverse relation”, despite this association being widely described in all industrialised populations including Denmark.34 The effects of adoptive and biological fathers’ social class were similar irrespective of the age the adoption took place, which would be expected if indeed the effects were due to genetic characteristics or in-utero exposures. Adjusting for own adult social class did not modify these results. These findings contrast with what most other studies report in terms of cumulative effects of socioeconomic conditions throughout the life course and that part of this association is mediated by adult social class. Thus, the possibility that adoptive families and adoptees are different in important aspects compared to the general population remains a potential explanation for these results. It could also be, as the authors state, a chance finding. This study needs to be replicated in other populations to clarify some of these points.
The study based on the Norwegian population found that women of higher childhood social class had higher suicide mortality in adulthood.12 This association was explained in part by family situation in adulthood, but not by adult SEP. Among men, similar but weaker association after adjustment for adult SEP, was observed. Two other studies report an association in the same direction (higher childhood social class related to higher suicide among women) but in neither study did the results reach conventional statistical significance,2 35 suggesting that evidence supporting this finding is weak. The Norwegian study did not include lone parents. Children from lone parents experience poorer socioeconomic circumstances during childhood and if they are at increased risk of suicide excluding them from the analysis could bias the results.
In summary, the newer evidence available from this update allows us to more strongly affirm that poor socioeconomic circumstances during childhood are associated with higher mortality among men and women and that this association persists among younger cohorts. The new evidence highlighted the difficulty in establishing a particular life course model to explain this association, but several studies established the importance of education as mediator between early life SEP and adult mortality.
What this paper adds
The association between childhood socioeconomic circumstances and cause-specific mortality is present in men and women.
This association persists in younger cohorts despite them not having been exposed to the same sort of childhood hardships as previous cohorts
Education is an important mediator between early life socioeconomic position and adult mortality
Policy implications
This systematic review provides strong evidence that poor socioeconomic circumstances during childhood are associated with higher mortality among men and women and that this association persists among younger cohorts. Tackling health inequalities from the start of life needs to be a policy priority if we are to reduce adult health inequalities.
REFERENCES
Supplementary materials
web only appendices 62/5/387
Files in this Data Supplement:
Footnotes
▸ Additional tables and an appendix are published online only at http://jech.bmj.com/content/vol62/issue5
Competing interests: None.
Linked Articles
- In this issue