Evaluating the evidence for models of life course socioeconomic factors and cardiovascular outcomes: a systematic review

The different extant hypotheses on the influence of life course SES on CVD can be grouped into four broad conceptual models: the latent effects, pathway, social mobility, and cumulative life course models. Most studies reviewed tested the influence of life course SES on CVD outcomes via the operation of one or more of these models.

Studies were categorized by their life course study design(s) and study question(s) into four groups: early SES → outcome, early SES → risk factor, social trajectory, and cumulative SES studies. Several studies appear in multiple categories, as they employed more than one of these designs. Table 1 describes the research questions and hypotheses typically considered by each design.

These studies examined the effect of childhood and/or adolescent SES on risk of adult CVD (e.g., incident myocardial infarction (MI), CVD death, incident/fatal stroke) [37‐56] and typically tested the latent effects hypothesis (i.e., examining the "direct" effect of early-life SES on adult CVD risk). They commonly measured the independent effect of childhood SES by statistically adjusting for later-life SES and CVD risk factors in regression models.

Table 2 lists these 20 studies according to CVD outcome utilized (see Additional file 1 for greater detail on these studies). All 19 studies conducting unadjusted or age-adjusted analyses reported a point estimate consistent with an inverse association between early-life SES and risk of one or more of the adult cardiovascular outcomes. Authors of 14 studies concluded that their findings supported some or all of their study hypotheses [39, 41‐45], 47[49‐55]. Fourteen of 16 studies adjusting for adult SES and/or CVD risk factors reported some indication of an inverse association; however, in only five did the authors suggest that their results support the hypothesis of an inverse association between early life SES and CVD.

For studies with CHD and stroke outcomes limited to early-life SES (i.e., not controlling for adult SES or other risk factors) almost all included a point estimate consistent with an inverse association. However, in adjusted analyses, authors of only one study with a CHD-related outcome and fewer than half the studies of CVD mortality reported inverse adjusted associations. Thus, the existence of a "direct effect" of early-life SES after adjusting for adult risk factors was not strongly supported.

To prevent statistical confounding by adult conditions, most studies employed statistical models that adjusted the association between childhood SES and adult CVD risk for adult SES and behavioral/ physiological CVD risk factors. This may be an over-adjustment, as certain risk factors (e.g., BMI, smoking) are part of the pathways through which low childhood SES may influence CVD risk. Little information is usually available on how participants' early-life SES may have influenced levels of adult CVD risk factors. Additionally, if unknown variables influence both the distribution of a CVD risk factor adjusted for and the distribution of the outcome measure, then adjustment will result in an incorrect partitioning of the total effect of early-life SES into "direct" and "indirect" components [57‐59]. Thus, early SES → outcome studies which seek to determine the "direct," adjusted effect of early-life SES on CVD risk may incorrectly estimate this effect [60], leading to questions about the accuracy of such estimates.

Some studies examined and qualitatively compared the effects of adjustment for different classes of CVD risk factors. Gliksman et al. (1995), for example, analyzed the association between father's occupational class and adult CVD in a series of models adjusting for different classes of potential mediators or confounders. This allowed for a more complete description of the impact of adult environment, behavior, and physiologic risk factors versus the latent impact of early life socioeconomic factors [39]. Studies utilizing more than one life course study design ("mixed-design studies") take a related approach and are considered in the Discussion section.

Life course studies in the early SES → risk factor group evaluated the influence of early- and/or mid-life SES or living conditions on later-life behavioral or physiologic CVD risk factors [14, 40, 44, 52, 61‐73]. These designs are often similar to those used in the early SES → outcome studies group except that CVD risk factors are the outcomes. The life course hypotheses typically examined, however, include both the latent effects and pathway conceptual models. The typical study design involves a statistical model where early-life SES measures predict the levels of CVD risk factors in adulthood.

Table 3 outlines the 17 studies reviewed (see Additional file 2 for detailed summaries). Five of six studies reported associations between low early-life SES and little or no adult leisure-time physical activity [14, 61, 63, 64, 66], five of five found associations with high adult alcohol intake [52, 61, 63, 68, 73], and eight of twelve studies reported associations with higher smoking rates as study findings [14, 44, 52, 63, 65‐68]. Nine of 11 studies reported (unadjusted) associations between lower early-life SES and elevated BMI or WHR [14, 44, 63, 64, 67, 68, 70, 71, 73]. The impact of a statistical adjustment for other CVD risk factors varied, including no impact [59], attenuation with a persistence of the significant effect [46, 55], and strong attenuation with associations no longer apparent [54, 78].

As with the early SES → outcome studies, concerns that lack of adequate adjustment or over-adjustment may lead to biased estimates are germane to this group of studies. The issue of covariate adjustment is complex in these studies, given that many CVD risk factors are considered. Certain physiologic risk factors (e.g., insulin resistance, dyslipidemia) may be linked to early-life SES through latent physiologic effects of negative childhood exposures, continuous exposure to negative physiologic stimuli, or other physiologic or life course pathway mechanisms [14, 63, 70, 74]. In contrast, the links between early-life SES and levels of behavioral risk factors (e.g., smoking, diet), and certain physiologic risk factors (e.g., BMI, hypertension), may operate principally through learned behaviors, behavioral responses to negative psychosocial stimuli, or other psychosocial life course pathway effects. The associations may develop through the operation of multiple physiologic and psychosocial pathways, making "correct" covariate adjustment challenging within a single model. Thus, statistical adjustment should take into account the specific life course pathways hypothesized to be operating [39]. Previous findings should be used to generate a priori hypotheses; nonetheless, there is considerable opportunity for inappropriate adjustments to be made.

Studies evaluating the social mobility life course model typically considered the impact of inter-generational or intra-generational social mobility on CVD and CVD risk factors [38, 44, 49, 64, 72, 73, 75‐79]. Inter-generational mobility was usually determined by contrasting the participant's father's occupational SES to the participants'. Intra-generational SES was typically defined as a change in occupational SES from early adulthood to later adulthood. Six of 11 studies evaluated adult CVD risk factors as outcomes [49, 64, 72, 73, 75, 77]; seven evaluated CVD mortality or CHD [38, 44, 75‐79]. Four adjusted only for age [38, 49, 76, 80].

Table 4 outlines these studies (see Additional file 3 for greater detail). Of 10 studies carrying out statistical analyses [38, 44, 49, 72, 73, 75‐79], six did not report associations between upward or downward mobility and either elevated levels of CVD risk factors or increased CVD morbidity or mortality when compared to stable low-SES or high-SES trajectories [38, 44, 49, 75, 78, 79]. Nine studies, however, reported the suggestion of inverse, although not always statistically significant, relationships between social mobility and a CVD-related outcome [38, 44, 72, 73, 75‐79]. Two studies found marked differences in CVD mortality risk between upwardly mobile individuals and individuals maintaining the same SES across time. However, one study reported increased CVD risk among the upwardly mobile [77] and one reported decreased mortality risk [79].

Four studies examined differences in CVD risk between stable low-SES trajectory and stable high-SES trajectory individuals [38, 44, 75, 79]. Three reported that individuals with stable low-SES trajectories had a greater CVD risk than stable high-SES trajectory individuals [44, 75, 79]; the fourth reported a marginally significantly greater risk [38]. These results are similar to those of most cumulative SES studies (described below) in that greater exposure to low SES was associated with increased CVD risk.

In these studies the unit of analysis is a trajectory, permitting the impact of change over time to be examined. However, the socioeconomic trajectories in most reviewed studies were limited to two time points, and groups compared tended to share the same SES at one of these time points. These similarities may partly explain why seven of ten studies did not report an association between social mobility and risk of CVD risk factors or events. Social trajectory studies incorporating SES at three or more time points allow for the analysis of more informative trajectories than studies evaluating SES at only two points [75]. Yet, even with large studies, analysis becomes cumbersome if more than two or three SES levels at three time points are measured. Additionally, the impact of uncommon (e.g. downward) trajectories are difficult to study due to the small numbers of individuals who typically comprise them.

These studies tested the operation of the cumulative life course conceptual model, typically by summing the number of times participants experienced unfavorable SES situations during early, middle or later life, and creating SES indices representing the accumulation of these experiences [45, 52, 53, 68, 79, 81, 82]. For example, three of these studies summed the number of times a participant (or their parents during the participant's childhood) had been in a manual occupation to create an index of accumulated low-SES exposure [68, 81, 82]. Cumulative SES was also measured using indices of occupational class, socioeconomic categories, exposures to negative socioeconomic experiences/conditions, income, and housing conditions. [45, 52, 53, 68, 81, 82].

Table 5 summarizes the seven cumulative SES studies (see Additional file 4 for greater detail). All authors reported that participants' cumulative life course exposure to low SES conditions was associated with increases in CVD outcome, supporting the cumulative life course SES hypothesis. Several studies indicated that cumulative SES was a more powerful predictor of CVD morbidity and/or mortality than adult or early-life SES alone [45, 53, 79, 81]. In studies that adjusted for CVD risk factors, graded associations were attenuated but remained strong in two studies [81, 82] and were greatly attenuated in another [45].

Davey-Smith et al. (2002) employed a unique cumulative SES measure, combining early and later-life occupational class experience with the CVD risk factors of smoking and heavy alcohol consumption [52]. They reported a marked, graded relationship between the number of negative SES exposures and risk of CVD mortality.

Only a large cohort study of Norwegians reported a statistically significant supra-multiplicative interaction between early- and later-life SES on risk of CVD [53]. Another study reported marginally statistically significant supra-multiplicative interaction effects between early and later life [45].

The reported associations between cumulative SES and CVD risk are more consistent than those reported in the other life course study designs. However, three of the seven studies evaluated were based on the same cohort [52, 81, 82]. Cumulative life course SES variables include current SES, and therefore may be conflating the effect of current SES with that of SES over the life course. For example, in Pensola et al. (2003), social mobility analyses suggested that the observed gradient in CVD mortality associated with cumulative manual occupational class was driven primarily by the impact of current occupational status. Additionally, cumulative indices implicitly assume that a specific negative life experience or situation has the same impact regardless of when it occurs in an individual's lifetime, with no distinction made between the impact of childhood versus adult events on risk of disease [45]. Furthermore, some cumulative SES studies combined disparate measures into a single, lifetime index variable, or summed the number of times different negative events or exposures occur [45, 52]. This approach assumes that different types of exposures have an equivalent impact on the risk of CVD. Evidence supporting these assumptions is not provided in these studies.

Certain general limitations and assumptions of life course SES studies should be considered. First of all, SES is a theoretical construct operationalized using various measures (e.g., income, occupation, education) that tap into different components of this construct. Despite several proposed SES evaluation schemes [83‐85], there is no overarching theory in the biomedical literature providing a rationale for the use of specific SES measures. Occupational status, the SES measure most commonly used in the studies reviewed, is assumed to be an adequate proxy of SES, but in fact represents only one component of SES. As associations of a given SES measure with CVD risk factors are not always consistent with those of other SES measures [47, 60], it is important to consider that results may depend upon the proxy SES measure employed [64, 81].

Most life course studies used retrospective cohorts or a case-control design, relying on participants' recall of early life SES. There has been little systematic evaluation of the validity of recalled early life circumstances or of the potential for such recall errors to bias associations. In studies directly comparing the impact of childhood and adult SES on adult CVD risk, greater error in childhood (vs. adult) SES measures may underestimate the true impact of child SES [86, 87]. Additionally, selection bias due to either loss to follow-up or selective survival may distort findings. Studies using cohorts followed from birth or early life probably do not have these limitations [14, 28, 47, 62, 71, 72, 77, 88‐93] and may avoid the problem of changing occupational status and workplace functions across time if they establish SES scales at the time of measurement and update them as appropriate. As more studies analyze data from prospective cohorts starting in early life, concerns about unequal measurement error and the changing status of specific occupations should decrease [14, 54, 73, 89, 93, 94].

Most studies were limited to white males. While the number of studies including women has increased in recent years, minority groups remain underrepresented in most life course studies. These groups may interact differently with the economic and/or social systems of the majority group; there are also established associations between minority status, low SES, and increased CVD risk [5, 84, 85]. The lack of information on the manner in which life course SES relates to adult CVD risk in minorities should be addressed.

The majority of the early SES → outcome group studies' results describe an inverse association between early life SES and risk of adult CHD, non-fatal MI, and CVD mortality. Similarly most early SES → risk factor studies suggest that low early-life SES negatively impacts levels of adult CVD risk factors. Most studies reported relationships between lower early-life SES and elevated alcohol intake, BMI/WHR, and insulin resistance, as well as decreased leisure physical activity. Together, the early SES → outcome and early SES → risk factor studies support the hypothesis that low early-life SES is inversely associated with adult risk of CVD. However, support for the hypothesis of a "direct effect" of early-life SES on risk of CVD mortality is equivocal as findings were less consistent after risk factor adjustment. Additionally, reported associations between early-life SES and adult health behaviors detract from the evidence supporting a latent effects life course model, as they suggest that CVD risk may be driven by learned behaviors.

The operation of social chains of risk or life trajectories described by a pathway life course model are difficult to observe or test, since almost all early SES → risk factor studies consider only two points during the life course. However, significant associations between early-life SES and the adult risk factors suggest behavioral, psychosocial or environmental links that may be best explained as a pathway effect. As described below, the pathway model will need further examination through the use of series of linked studies.

While nine of 10 social trajectory studies reported some relationship between inter- or intra-generational social mobility and CVD-related outcomes, only four were identified as providing evidence in favor of a social mobility hypothesis. Conversely, most studies comparing stable low- and high-SES trajectory groups reported markedly increased CVD mortality among the low-SES group, with many socially mobile groups having a CVD risk between the stable high and low groups. Determining whether these associations are a function of social mobility, or whether they are due to exposure to low SES at some point in the life course and not to mobility, is problematic. Future studies should focus on differentiating between the influences of cumulative SES and social mobility.

Of the conceptual models discussed, the cumulative life course model was the most consistently supported. However, the weight of these findings is limited as much of the data came from one cohort [57, 60, 61]. Also, specific methodological concerns (e.g., equal weighting of life course periods, conflation of current and life-course SES) need to be considered when interpreting these findings. Future studies conducted in other cohorts that address these concerns will help clarify the viability of cumulative life course SES hypotheses.

Researchers have noted the limitations of life course SES study designs examining only one potential pathway linking SES and CVD. Accordingly, recent studies tend to use a combination of SES measures, CVD risk factor and/or outcomes and life course study designs [40, 44, 45, 52, 53, 68]. These mixed-design studies allow for a comparison of how well different life course SES conceptual models fit the patterns observed in the same data. Such studies are also less likely to overlook patterns of association in the data than studies which only evaluate one possible relationship between SES and CVD.

As modern life course cohorts are followed from childhood, longer prospective studies or series of cross-sectional studies that evaluate associations between early-life experiences and risk factor levels at several time points may provide the opportunity to observe the operation of pathway life course effects. For example, two studies in the early SES → risk factor group [14, 90] compared levels of CVD risk factors at two points in young adulthood (22 and 33 years of age). The authors observed an association between childhood SES and adult obesity, which was smaller for participants at 33 than at 23 years of age [14], suggesting a decreasing impact of early life SES on adult health with increasing age. Such information is only obtained through longer ongoing studies or series of interrelated studies.

Individuals' socioeconomic environments may impact their CVD risk factor levels and CVD risk. Community socioeconomic conditions may influence individuals' ability to engage in leisure physical activity or eat a healthy diet [95‐97]. Some life course studies have evaluated the community socioeconomic environment, primarily through the use of indicators of neighborhood deprivation [44, 48, 54, 81, 82]. Future studies should likewise consider the effect of physical and psychosocial environmental factors.

Life course SES studies have generally failed to consider the length of exposure to the various socioeconomic conditions measured. As this may influence the impact of negative SES experiences on adult health, future cumulative life course studies may benefit from evaluating the effect of length of exposure into their indices.