Childhood Poverty and Abdominal Obesity in Adulthood: a Systematic Review

Adverse socioeconomic conditions in childhood can have lasting effects on health, but evidence is lacking from prospective studies concerning the effects of early poverty on abdominal obesity in adulthood. Cross-sectional studies in adults from middle and high-income countries show that current socioeconomic status is inversely related to obesity in women, but the pattern in men is not consistent. A systematic review was undertaken to assess the influence of early socioeconomic status on waist circumference , hip circumference, and waist-hip ratio in adulthood. Thirteen relevant articles were located (five cross-sectional and eight cohort), including only one from a middle-income country and the remainder from high-income settings. In all the studies, childhood poverty was associated with higher levels of abdominal obesity in women. In men, the associations were weaker, and no clear pattern emerged.


Introduction
Obesity is an important risk factor for incidence of cardiovascular diseases, diabetes, hyperten sion, and hypercholesterolemia and related mortality 1 .Prevalence of obesity has increased rapidly in high, medium, and low-income coun tries 2,3,4,5 .Socioeconomic conditions have an important influence on prevalence of obesity, as well as on other risk factors associated with car diovascular diseases 4,6,7,8 .
Obesity in adults has been estimated mainly through body mass index (BMI) 9 .Recent stud ies show that abdominal obesity, evaluated ac cording to waist-hip ratio (WHR), is capable of predicting incidence of cardiovascular diseases better than other anthropometric measures, in cluding BMI 10,11 .
WHR is used as an indicator of the amount of fat tissue deposited in the waist area in relation to the hip structure 10,11 .However, increased WHR can result from high waist circumference and/ or reduced hip circumference 12,13 .These two components of WHR can be affected by distinct factors and in different stages of growth 14,15 .In addition, waist circumference is directly associ ated with risk of cardiovascular diseases, while hip circumference shows an inverse relation ship 12,14,16 , thereby supporting the recommen dation that the two components be evaluated separately 13 .
Cad. Saúde Pública, Rio de Janeiro, 25 Sup 3:S427-S440, 2009 Since obesity has multifactorial origins, the emerging hypothesis of "early onset of adult hood diseases" has led to increasing interest in understanding the mechanisms by which early exposures might affect nutritional and health status over the course of life.Adverse conditions experienced in the intrauterine period and/or in the early months of life can alter organ functions and predispose to the development of diseases in adulthood due to a process of biological pro gramming 17,18,19,20,21 .
The effects of childhood poverty on health and nutritional status in the early years of life have been widely studied 2,22 , but few longitudi nal studies have evaluated the consequences for obesity in adulthood 23 .Nearly all studies on the association between socioeconomic status and obesity in adults have a cross-sectional design, in which both variables are measured simultane ously 4,6,8,23 .By reviewing the literature, we found 40 studies evaluating the relationship between childhood socioeconomic status and obesity in adulthood, of which only three were conducted in medium or low-income countries: two in Bra zil 24,25 and one in China 26 .There are even fewer articles specifically evaluating the association between socioeconomic status and abdominal obesity.
Given the possibility that abdominal obesity is determined early, the current systematic review investigates the association between childhood socioeconomic status and waist circumference, hip circumference, and WHR in adulthood.

Methodology
The systematic literature review included the MEDLINE, Springerlink, Ovid, LILACS, and SciELO databases.Various combinations of Med ical Subject Headings (MeSH) descriptors were used for the exposure ("socioeconomic status", "socioeconomic factors, income", "education", "social class", "employment", "occupational sta tus", "poverty") and outcomes ("waist-hip ratio", "hip", "obesity", "abdominal tissue", "adiposity").Articles published until July 2008 were consid ered.The search was conducted independently by the two first authors of the present article.The reference lists were compared and the relevant articles selected by title and abstract.All the apparently relevant articles were obtained and reviewed by the first author.The bibliographic references cited by the selected articles were also reviewed, and if relevant they were included in the review.
The review selected all studies with a longi tudinal or cross-sectional design that evaluated childhood socioeconomic status as the exposure and waist circumference, hip circumference, and/or WHR when the individuals were at least 18 years of age as the outcomes.The review did not include studies limited to carriers of a spe cific disease.
The results of different articles were com pared, whenever the data allowed, based on the slope index of inequality (SII).The SII represents the linear regression coefficient (beta) with the health indicator as the outcome and an ordinal variable representing socioeconomic status as exposure 27 .This index allows comparing stud ies that have used different categories of socio economic status, since the beta parameter rep resents the difference between the extremes in social distribution, regardless of the classification used.In the analyses presented below, a value of "0" was assigned to poorer individuals and "1" to wealthier ones.
The results were interpreted using the con ceptual model described in Figure 1.The arrows in the figure indicate the possible causal mech anisms.The most distal determinant is the in heritance of genetic characteristics, which sup posedly affects the model's other components, including intra-gestational maternal adiposity, as well as the child's intrauterine and post-natal life 3,23,28 .Genetic load can also influence the parents' socioeconomic status, as in the case of discrimination against black 29 or obese indi viduals 30 .In the second hierarchical level, par ents' socioeconomic status can also affect life time adiposity 4,6,7,8,23 .This effect is supposedly mediated by the family's nutritional habits and physical activity, the child's disease history (par ticularly that of infectious diseases), and the indi vidual's own socioeconomic status in adulthood 3,23,28 .This model suggests that controlling for genetic factors, if possible, would be justified by the possibility that such factors are influencing both socioeconomic level and nutritional status.Meanwhile, the results adjusted for behavioral factors (diet, smoking, etc.), childhood nutrition al status, and/or current socioeconomic status should be interpreted with caution, due to the possibility that these factors are at least partially influenced by the family's socioeconomic status, thereby constituting mediating factors in the tar get relationship.

Figure 1
Conceptual model for possible associations between childhood socioeconomic status and obesity in adulthood.

Results
Figure 2 summarizes the selection process for the reviewed articles, based on the QUOROM guide lines 31 .A total of 1,387 possibly relevant articles were located, of which 134 were selected to be read in full.Ninety-four of these were excluded because they did not evaluate family socioeco nomic status when the individual was five years of age or older.Of the remaining articles, only 13 evaluated waist circumference, hip circumfer ence, and/or WHR in adulthood as the outcome 26,32,33,34,35,36,37,38,39,40,41,42,43 .
All the included articles were published in the last ten years.Twelve were performed in highincome countries (11 in Europe 32,33,34,35,36,37,38,39,40,42,43 and one in New Zealand 41 ) and one in Chi na 26 .Three British articles 33,36,37 used data from the same cohort; two Finnish articles 34,35 like wise analyzed the same cohort; and two Spanish articles 38,39 used data from the same cross-sec tional study, evaluating the outcome differently (continuous or categorical variable) or in distinct moments of life.Thus, the 13 articles refer to nine different studies.
The independent variable in all the articles was defined on the basis of the parents' occu pation, except in the Chinese study, which used household assets during the individual's child hood.Waist circumference was analyzed as the outcome by nine articles 26,32,33,35,36,37,38,39,42 , five of which also evaluated WHR 32,33,35,36,42 .In the four remaining articles, WHR was analyzed alone 34,40,41,43 .Only the Finnish article 35 and two Brit ish articles 36,42 evaluated hip circumference as outcome.All the articles presented their results separately for men and women, except the New Zealand study 41 .
Nine articles reported losses ranging from 40% to 69% of the initial sample 32,33,34,35,36,37,40,41,43 , while in three others the losses were approxi mately 25% 38,39,42 .Only the article from China reported smaller losses (10% in men and 1% in women) 26 .Six articles provided a detailed de scription of the losses 32,35,37,38,39,40 , and in two Selection process for studies evaluating the association between childhood socioeconomic status and abdominal obesity in adulthood (according to QUOROM guidelines).articles they were more common among the poor 35,37 .Table 1 summarizes the articles included in the review.The principal results will be discussed according to the target outcome.

Waist circumference
In women, the nine articles that evaluated the relationship between early socioeconomic status and waist circumference in adulthood found sig nificant inverse associations 26,32,33,35,36,37,38,39,42 .This was observed both in articles where waist circumference was treated as a continuous vari able 26,32,33,35,39,42 , as well as in those where it was treated as greater than or equal to 88cm 37,38 .It was possible to calculate SII for six articles, all generating negative values.For five articles in which the outcome was continuous 26,32,35,36,39 , women from wealthy families in their childhood presented waist circumferences 2.2cm (China) and 6.8cm (Great Britain) smaller than in women from poor families.In the only study that used a categorical outcome, the index was 11.1 percent-Cad.Saúde Pública, Rio de Janeiro, 25 Sup 3:S427-S440, 2009 Table 1 Studies that evaluated associations between childhood socioeconomic status and abdominal obesity in adulthood.age points, once again with a lower prevalence among the wealthier 38 .

Reference
Current socioeconomic status was included as a co-variable in six articles: three showed a re duction in the coefficients with persistence of the association 32,36,37 , in two the relationship disap peared completely, 26,39 and one showed a slight increase in the effect measure 38 .In Kuh et al. 33 , in the 1946 British cohort, current socioeconom ic status was included with weight at seven years and current BMI, which led to complete loss of the original association.
In the nine articles with data on men 26,32,33,35,36,37,38,39,42 , the pattern in the associa tions was not as clear, nor was the magnitude of effects as pronounced as in women.Three arti cles on the 1946 British cohort 33,36,37 , one British survey 42 , and a Finnish article 32 showed inverse associations between childhood socioeconomic status and waist circumference.Two of these ar ticles calculated the SII 32,36 , presenting values of -2.0cm (Great Britain) and -2.6cm (Finland).The two Spanish articles showed a non-linear re lationship, and waist circumference was larger among elderly individuals belonging to the ex treme categories of childhood socioeconomic status, especially the poorer 38,39 .Meanwhile there was a direct relationship in Chinese elderly (SII equal to 2.8cm) 26 .Only one article, with data from Finland 35 , found no association.
When current socioeconomic status was in cluded in the analyses, two articles showed no change in the associations 26,39 , in two there was a reduction in the coefficients 32,36 , and in two others the association disappeared 37,38 .In Kuh et al. 33 , with the inclusion of anthropometric variables (in addition to current socioeconomic status) there was a reduction in the coefficients, but the association remained significant.

Hip circumference
Only three articles evaluated the association be tween childhood socioeconomic status and hip circumference in adulthood.In Finland, there was no difference between the various catego ries of socioeconomic status for either men or women 35 .Meanwhile, two British articles found an inverse relationship, particularly in women 36,42 .In two of these articles it was possible to cal culate the SII 35,36 , ranging from -0.6 to -4.30cm in women and 0.0 to 0.5cm in men.None of the articles included adjusted analyses.

WHR
In women, the three Finnish articles 32,34,35 and three other British articles 33,36,42 found inverse associations between childhood socioeconomic status and WHR.In four articles it was possible to calculate the SII 32,34,35,36 , all of which showed negative values.In three, WHR was analyzed as a continuous variable 32,35,36 , with the SII ranging from -0.018 to -0.027.In the article that treated WHR as a categorical variable, the SII was -3.2 p.p. 34 .Two other British studies 40,43 showed no relationship between childhood socioeconomic status and WHR in adulthood.
In men, of the seven articles that evaluated WHR as the outcome, four reported significant inverse associations 32,33,35,36 , two found no re lationship 34,42 , and one found a direct associa tion 40 .In three articles it was possible to calcu late SII, two of which showed negative values (-0.010 and -0.025) 35,36 .In the study by Kivimaki et al. 32 , the authors reported a significant inverse association after adjusting for age and current so cioeconomic status, but the SII calculated from the crude data was not significant (SII = 0.007).
In the study from New Zealand, in which the analyses were performed without stratifying by gender, there was an inverse relationship be tween early socioeconomic status and WHR in adulthood 41 .
In three articles the analyses were adjusted for current socioeconomic status 32,36,41 , and in both men and women the associations remained significant despite the drop in the coefficients.The same occurred in Kuh et al. 33 when adult anthropometric variables and current socioeco nomic status were included in the analyses.

Other analyses
Three studies assessed the association between change in socioeconomic status between child hood and adulthood and abdominal obesity.A study from the United Kingdom 36 and another from New Zealand 41 showed a higher risk of ab dominal obesity in men and women that had al ways been poor as compared to the non-poor.Meanwhile, a British survey of elderly women showed no association 43 .

Discussion
Except for a study in China 26 , the articles re viewed here are from high-income countries.Most of the analyses show that among women, childhood poverty is associated with increases in waist circumference, hip circumference, and/ Cad.Saúde Pública, Rio de Janeiro, 25 Sup 3:S427-S440, 2009 or WHR in adulthood.For men, the associations were weaker and did not show such a clear pat tern as in women.Most of the authors found in creases in waist circumference and WHR in men that were poor in childhood, but some studies found no association, and in still others the mean values were higher among the wealthy.This sug gests that the influence of socioeconomic condi tions on obesity is more evident and occurs ear lier in women than in men.
The current findings are similar to those re ported in the review by Parsons et al. 23 , with a predominance of inverse associations between childhood socioeconomic status and BMI in adults of both sexes, although no direct associa tion was reported.Part of the difference in the results in men according to the current review probably lies in the origin of individuals that were evaluated.In the review by Parsons et al. 23 , only Northern Europe and the United States were rep resented, but in the current study other countries were included (New Zealand, Spain, and China), among which different associations were report ed in men.
These differences in obesity levels between countries have been documented previously.Two systematic reviews of cross-sectional stud ies conducted in high-income countries showed that obesity was more frequent among poor women 4,8 .This pattern had already been shown in the first European and North American stud ies published in the 1930s 8 .Meanwhile, among men, the two reviews detected both inverse and direct associations and even lack of association between current socioeconomic status and obe sity; importantly, direct associations are less fre quent in studies performed in the last 20 years.Meanwhile, in medium and low-income coun tries, obesity was previously a disease of the so cioeconomic elite in both sexes 8 , but since 1980 this association has become predominantly in verse in women, while in men the majority of the studies reported either direct associations or a lack of association 4,5,6 .This pattern of nutritional transition is particularly evident in countries that have experienced greater economic growth, af fecting women earlier than men 4,6 .
In the current review, the only study held out side of high-income countries found an inverse association with abdominal obesity in women and a direct association in men 26 , thus reflect ing a transition pattern.Brazilian studies found a direct association between childhood socioeco nomic status and BMI in late adolescence and early adulthood in men 24,25 , and an inverse as sociation in women 44 .Thus, although the results of cross-sectional studies are subject to reverse causality -whereby obesity can determine cur rent socioeconomic status rather than vice-versa 4,6,8,28,30 -, the transition process described above is consistent with our finding of clearer associa tions in women than in men for the effects of ear ly poverty.This long-term effect on abdominal obesity appears not to result only from persisting socioeconomic adversities in adulthood, since based on the results of the present review, cur rent socioeconomic status was a partial mediator in the associations.
To understand the associations, it is neces sary to contextualize the current results.When the individuals studied in the current review were born, childhood obesity was more frequent among boys and girls from wealthy families, even in high-income countries 8 .This suggests that especially for women, from childhood to adulthood there is a reversal in this association.A similar reversal for girls -but not for boys -was recently reported in a Brazilian cohort 44 .Vari ous studies point to adolescence as the period in which this "socioeconomic reversal of fatness" occurs 7,8,28,45,46,47 , apparently as a consequence of biological and cultural factors, although the mechanisms involved are not very clear.
Biologically, women appear to be better adapted than men to survive adverse condi tions in early childhood, which is confirmed by higher morbidity and mortality in boys 48,49 .Fe males have more subcutaneous fat tissue since birth, which favors surviving early adverse con ditions 49,50,51 .Periods of nutritional restriction in early childhood lead to permanent deficits in the amount of lean mass in adulthood 52 , but central adipose tissue appears to be selectively preserved 53 .Even if nutrient supply improves over life, central fat distribution is maintained in adulthood, particularly in women, as demon strated by Ravelli et al. 54 in their study on hunger in the Netherlands during World War II.In ado lescence, sex hormones lead to increased fat de posits in women, as an energy reserve to guaran tee procreation and breastfeeding 49,50,51 , thereby increasing the differences between the sexes.The same hormonal action probably contributes to highlighting the effects of early poverty on cen tral adiposity in women.
Behavioral characteristics of poor women in adolescence may also contribute to the socio economic reversal of fatness: higher parity 55 and less concern with body image 23,56,57,58 or weight reduction 23,28,57 .Meanwhile, for men, there are similar proportions of individuals from distinct socioeconomic strata who wish to lose or gain weight during adolescence.However, the control measures acquired by adolescents are mainly targeted to gaining body mass (both fat and lean) 23,56,57,58 .
Cad. Saúde Pública, Rio de Janeiro, 25 Sup 3:S427-S440, 2009 Other conditions related to childhood socio economic status and that can affect obesity pat terns in adulthood need to be considered.Vari ous studies point to parental obesity as an im portant risk factor for the children's obesity, both in childhood and adulthood 3,23,28,59 .This could suggest that the observed associations with so cioeconomic status are the consequence of pa rental obesity.None of the studies in the current review included the parents' nutritional status as an adjustment variable.However, the review by Shrewsbury et al. 7 based on cross-sectional studies in children found that in six of the nine articles that evaluated the association between socioeconomic status and childhood obesity and that included parental adiposity as a co-variable, the direction of the associations did not change.This suggests that socioeconomic status is more than a simple marker of parental obesity.At any rate, it is unlikely that genetic factors can explain the differences between men and women accord ing to early childhood socioeconomic status.
Race, which reflects both the inheritance of genetic characteristics and can represent a proxy for socioeconomic conditions, is also associated with obesity 3,23,28,60 .The populations evaluated in the current review consisted of quite homo geneous ethnic groups, so they did not allow evaluating this variable's impact on abdominal obesity.
Childhood nutritional habits and patterns of physical activity have also been mentioned as possible explanatory factors for the observed differences.However, these factors have shown conflicting results in different studies 23,28 .These and other factors need to be explored better to define their importance in these associations.
Various limitations should be considered when interpreting the results of the current review.All the studies used the parents' occupation to as sess socioeconomic status, with the exception of the Chinese study, which used household assets.However, the categories used to characterize the parents' occupation varied, thus hindering com parison.Even so, in the eight studies in which it was possible to calculate SII 26,32,34,35,36,38,39,43 , the resulting trends were similar to the original esti mates.Other important exposure variables were not evaluated, like maternal schooling, which has been identified as a more stable and relevant risk factor than parents' occupation for this type of association 7,8,23,28 .Secondly, various studies only present data for WHR, and not separately for waist and hip circumference.These two cir cumferences contribute in opposite directions to WHR and risk of cardiovascular diseases 12 , and can also be influenced differently by early socio economic status.The study from New Zealand did not stratify the results by gender 41 and was thus unable to determine whether the observed inverse association was similar for men and women.Finally, although various studies have reported considerable loss-to-follow-up rates, they do not explain the differences observed be tween the genders or the consistency of the find ings with the results of cross-sectional studies.
Early poverty increases the risk of abdominal obesity in women, but among men the pattern was not as clear.The negative effect of poverty on abdominal obesity in men may become more evident as the nutritional transition progresses.The effects of early poverty on adults' health are shown by studies approaching different outcomes 2,29,52 .According to the current review, abdomi nal obesity -with its important consequences for morbidity and mortality from degenerative diseases -constitutes an additional risk factor that could be prevented through social policies to reduce the inequalities that exert a permanent influence on growth and development in the ear ly stages of life.