The worldwide prevalence of overweight and obesity in the adult population is currently estimated to be 35 %, rising to over 70 % in some population groups, with rates of obesity nearly doubling between 1980 and 2008^{( 1 )}. This is a global phenomenon, with the prevalence of overweight and obesity in low- and middle-income countries on the rise, mirroring their transition to the obesogenic environments more commonly associated with ‘Western’ living. The relationship between obesity and risk of non-communicable disease in adulthood has long been established, with a wealth of evidence demonstrating the considerable impact of excess adiposity on the health and wellbeing of populations⁽ Reference Visscher and Seidell ^{2 )}. The trends in adult overweight and obesity are reflected in childhood obesity prevalence, and the potential impact of early-onset obesity is of great concern. It has been estimated that 200 million school-aged children were overweight or obese in 2010 worldwide, with a prevalence of >20 % in European states and >30 % in regions of North America^{( 3 )}. In the UK, the 2007 Foresight report showed that about 8–10 % of girls and boys were obese in 2004, and this was projected to rise to an estimated 25 % by 2050^{( 4 )}. The present paper considers the long-term implications of childhood obesity for metabolic and cardiovascular health, focusing on the independent contribution of childhood obesity to adult disease risk, as distinct from associations mediated by tracking of obesity and dietary-related factors across the lifespan.

Impact of childhood obesity on long-term metabolic and cardiovascular health

A large and consistent body of evidence demonstrates that overweight and obesity during childhood and adolescence have adverse consequences in terms of chronic disease risk in adulthood. This evidence has been systematically reviewed a number of times recently. In 2009, Owen et al.⁽ Reference Owen, Whincup and Orfei ^{5 )} reviewed the evidence of an association between BMI in childhood or early adulthood and CHD outcomes in later life, with almost all studies showing a positive association. The authors concluded that BMI from about age 7 years shows a consistently positive relation to risk of CHD in adulthood, which did not appear to be confounded by adult smoking or social class. The authors noted that the independent contribution of early obesity to CHD risk could not be established, as analyses adjusting for the effect of adult BMI were limited and inconclusive. A second systematic review⁽ Reference Reilly and Kelly ^{6 )} set out with similar objectives, but widened the inclusion criteria to include outcomes related to any adult physical morbidity or premature mortality. Again, the review demonstrated a highly consistent body of evidence that overweight and obesity in childhood and adolescence were associated with increased risk of adult morbidity and mortality, particularly related to cardiometabolic dysfunction. The largest review of this nature⁽ Reference Park, Falconer and Viner ^{7 )} brought together thirty-nine studies examining the association between childhood BMI and a variety of adult outcomes, including type-2 diabetes, hypertension, CHD, stroke, asthma, cancer and all-cause mortality. The review found that childhood overweight was consistently associated with increased risk of type-2 diabetes, hypertension, CHD and mortality, whereas evidence for a relationship with stroke outcomes, adult-onset asthma or cancer were mixed and inconclusive.

It is important to note that all three systematic reviews commented on the unknown independent contribution of childhood obesity⁽ Reference Owen, Whincup and Orfei ^{5 –} Reference Park, Falconer and Viner ^{7 )}, as distinct from the contribution mediated by tracking of BMI from childhood into adult life. Of the studies reported, only a few had adjusted for adult BMI and, where adjustments had been made, the results were found to be inconclusive. Discussion and commentary in this area often stray towards asserting that childhood obesity is an independent risk factor for adult disease, giving the impression of a permanency of effect and weighting the argument for prioritising childhood for weight management interventions. This may at first seem like a subtlety in the reporting of the data, but is in fact very important when considering identification of at-risk groups and the relative impact of the timing of interventions on their long-term effectiveness. The present review will first consider potential mediators of the relationship between childhood obesity and adult disease risk, focusing on the tracking of BMI and related behavioural factors across the lifespan, before reviewing the evidence of an independent association between childhood BMI and adult disease risk.

Tracking of BMI from childhood to adulthood

The associations observed between childhood obesity and adult disease risk are likely to be partially mediated by tracking of BMI from childhood to adulthood, with the increased likelihood that overweight or obese children become overweight or obese adults explaining their increased risk of disease. The extent to which this explains the associations is an important question in terms of the relative contribution of exposures at different stages of the lifespan to risk of disease in adulthood. In 2008, Singh et al.⁽ Reference Singh, Mulder and Twisk ^{8 )} published a systematic review of data relating to the persistence of childhood overweight and obesity. Studies were selected if they had a prospective or retrospective longitudinal design, and had taken at least one anthropometric measurement of BMI, skin fold or waist circumference during youth (age ≤18 years) and during adulthood (age ≥19 years). Such data had been collected as part of eighteen different studies and the data published in twenty-five articles. From these, Singh et al. extracted estimates of the risk or proportions of overweight youth who went on to become overweight adults and compared these with non-overweight youth. As expected, all studies showed an increased risk for overweight or obese youth to become overweight adults. However, there was considerable variability in the estimated risks, reflecting homogeneity between studies in terms of the anthropometric methods used, the interval between childhood and adulthood measurements, and the criteria used to classify BMI. Based on the high-quality studies, the risk of overweight or obese children or adolescents becoming overweight or obese adults was at least twice as high as their normal-weight counterparts, with the highest relative risk reported being 22·3. The persistence of overweight was generally higher with increasing BMI (i.e. the risk of adult overweight was greater for obese compared with overweight children) and with increasing age (i.e. the risk of adult overweight was greater for overweight and obese adolescents compared with children). The percentage of overweight adolescents becoming overweight adults varied between 22 and 58 %. Overall, the percentage of obese adolescents becoming overweight/obese adults varied between 24 and 90 %. The authors termed the persistence of overweight into adulthood as ‘moderate’, but also noted that the majority of overweight adults were not overweight during childhood.

It is an inevitable limitation of using longitudinal life-course data that the populations generally are not contemporary. The persistence of overweight in current generations, which have a greater prevalence of obesity^{( 1 )} and a more obesogenic environment⁽ Reference Lobstein ^{9 )}, may therefore be greater than that reported to date. However, at present, these data must be relied on to give us the best current estimate of BMI tracking. The moderate level of tracking of BMI from childhood to adulthood⁽ Reference Singh, Mulder and Twisk ^{8 )} may be explained by a combination of genetic and environmental factors⁽ Reference Silventoinen, Bartels and Posthuma ^{10 –} Reference Silventoinen and Kaprio ^{13 )}, and is likely to partially mediate the increased risk of adult disease in those who were overweight in childhood. The genetic influences on adiposity and tracking of BMI across the lifespan are beyond the scope of this review, but have been reviewed elsewhere⁽ Reference Silventoinen and Kaprio ^{13 )}. The tracking of two key modifiable behaviours related to adiposity will be reviewed briefly, before considering the independent effects of childhood obesity on adult disease risk.

Tracking of obesity-related behaviours from childhood to adulthood

Dietary intake and physical activity are two key modifiable behaviours related to risk of obesity. It is important to consider how these factors track between childhood and adulthood, and therefore contribute to the persistence of overweight and obesity. A recent systematic review⁽ Reference Craigie, Lake and Kelly ^{14 )} identified thirty-eight longitudinal studies, which assessed at least one measure of diet, physical activity or inactivity at baseline (aged <18 years) and at follow-up (at least 5 years after baseline and aged >18 years). The majority of studies investigated tracking of physical activity (twenty-seven papers describing sixteen cohorts) rather than dietary intake (eleven papers describing five cohorts), and these outcomes were generally reported separately. Despite the considerable challenges associated with measuring complex behaviours over time, the review demonstrated clear evidence of tracking of both physical activity and diet between childhood and adulthood, with the strength of tracking of a similar order for both behaviours. The emphasis of the studies included was primarily on cardiovascular risk, with dietary assessment focusing on dietary patterns or specific nutrients rather than the total energy intake per se. However, the dietary pattern is predictive of obesity risk, with energy dense, high-fat, low-fibre patterns prospectively associated with excess adiposity during childhood and adolescence⁽ Reference Ambrosini, Emmett and Northstone ^{15 )}. In addition, the dietary pattern data are useful for identifying at-risk groups or overall dietary behaviours to target with intervention⁽ Reference van Dam ^{16 ,} Reference Slattery ^{17 )}.

Craigie et al.⁽ Reference Craigie, Lake and Kelly ^{14 )} identified a clear need for further investigation of tracking of diet and physical activity within the context of obesity risk. Limitations surrounding population dietary assessment and long-term follow-up mean that the evidence base is relatively weak, with several studies having a short follow-up period and inconclusive results. Data from the Young Finns Study enabled the assessment of dietary patterns over a longer-term period of 21 years⁽ Reference Mikkila, Rasanen and Raitakari ^{18 )}. The study identified two substantially different dietary patterns, termed ‘traditional’ and ‘health conscious’, which were clearly identifiable across the three study points. Food choices expressed as dietary pattern scores showed tracking across the 21-year period (1980–2001), with Spearman's correlation coefficients of about 0·35, reasonable considering the time interval, transition from childhood to adulthood and the short term measurement of food consumption used. Approximately 30–40 % of subjects originally belonging to the extreme quintiles of the energy-adjusted pattern scores persisted in the same quintile 6 and 21 years later. Within this same study population, significant tracking of physical activity and inactivity were also observed from adolescence to young adulthood⁽ Reference Raitakari, Porkka and Taimela ^{19 )}. Similarly to the tracking of BMI, the strength of tracking of both diet and physical activity was generally greater with increasing age at baseline, suggesting greater stability of obesity-related behaviours from late adolescence v. earlier childhood and with shorter interval between baseline and follow-up⁽ Reference Craigie, Lake and Kelly ^{14 )}. Similar findings were observed when investigating the tracking of blood pressure, a key cardiovascular risk factor associated with obesity⁽ Reference Chen and Wang ^{20 )}.

Independent effects of childhood obesity on adult disease risk

The data reviewed demonstrate tracking of overweight and obesity, as well as two primary modifiable behaviours related to risk of obesity, from childhood to adulthood⁽ Reference Singh, Mulder and Twisk ^{8 ,} Reference Craigie, Lake and Kelly ^{14 ,} Reference Raitakari, Porkka and Taimela ^{19 )}, providing a pathway by which the risk of adult disease could originate from childhood. However, it is important to consider whether a component of the association between childhood obesity and adult disease risk could be independent of adult BMI. For example, if overweight or obese children shift to a healthy weight during the late adolescence or the early adulthood period, does their risk reduce to that of healthy-weight adults who have no history of childhood obesity? Similarly, are obese adults who have been obese since childhood at greater risk than those who were a healthy weight during childhood? While not a primary focus of the research reviewed so far, this subtlety in the data is an important point when considering the permanency of effects of childhood dietary behaviours and the relative impact of interventions at different stages of the life-course.

The general concept that the nutritional status during the developmental period can have permanent lifelong consequences is well supported. There is a wealth of information from human epidemiology and animal modelling studies to demonstrate that nutrition during early life can impact on developmental processes in a way that permanently affects the organ structure, physiological function and gene regulation, thereby affecting the risk of developing a range of disease risk factors and outcomes in adult life⁽ Reference Langley-Evans and McMullen ^{21 )}. However, it remains unclear whether childhood overweight could have similar remodelling effects such that, even if a cohort of overweight children achieved a healthy BMI in adulthood, the increased risk associated with their childhood obesity remained. Previous systematic reviews have clearly demonstrated an association between childhood BMI and adult disease risk, but have also highlighted a gap in the evidence base surrounding the unknown independent contribution of childhood BMI⁽ Reference Owen, Whincup and Orfei ^{5 –} Reference Park, Falconer and Viner ^{7 )}. We therefore systematically reviewed the literature with the specific objective of assessing whether the association between childhood BMI and adult disease risk were fully dependent on adult BMI⁽ Reference Lloyd, Langley-Evans and McMullen ^{22 ,} Reference Lloyd, Langley-Evans and McMullen ^{23 )}.

Studies were selected for inclusion in the systematic reviews if they had a measure of BMI in childhood (aged ≤18 years, either classified according to the center for disease control (CDC)⁽ Reference Kuczmarski, Ogden and Grummer-Strawn ^{24 )} or international obesity taskforce cutoffs (IOTF) guidelines⁽ Reference Cole, Bellizzi and Flegal ^{25 )} or analysed as a continuous variable) and a marker of metabolic or CVD risk or outcome in adult life (age >18 years). Regression coefficients or estimates of relative risk were extracted from each study, and the reviews extended those previously published by explicitly searching for analyses which had adjusted for adult BMI. Reasonable quality was established through the inclusion and exclusion criteria; for example, studies were excluded if they used arbitrary BMI thresholds for overweight and obesity, or if the data were self-reported. However, the quality assessment using the Newcastle–Ottawa Scale⁽ Reference Wells, O'Connell and Peterson ^{26 )} identified some reasonably consistent weaknesses. The majority of studies failed to demonstrate that the outcome was not already present during childhood and assessed disease risk markers in the cohort at an adult age which was considered too young for outcomes to be shown adequately. In contrast, those studies which assessed risk markers at an older age or focused on actual disease outcomes or mortality inevitably used populations which were less contemporary. Regression coefficients or estimates of the relative risk could be extracted from all studies, but there was considerable heterogeneity in the data presented. Most notably, the ages at which exposures and outcomes were measured differed substantially between studies, ranging from 2 to 18 years for childhood BMI and 18–71 for adult outcome.

Despite methodological heterogeneity, the findings were relatively consistent. Similarly to other systematic reviews⁽ Reference Owen, Whincup and Orfei ^{5 –} Reference Park, Falconer and Viner ^{7 )}, the studies consistently showed positive associations between childhood BMI and adult disease risk, across a range of metabolic and cardiovascular measures. However, there was very limited and weak evidence that childhood obesity is an independent risk factor for adult disease. The majority of analyses that adjusted for adult BMI showed attenuation of the associations between childhood BMI and adult disease risk, indicating that the associations were mediated by tracking of BMI from childhood to adulthood.

Peripheral blood pressure was one of the most common risk factors assessed as an outcome in the studies selected. Of the eight studies considering the relationship between childhood BMI and adult blood pressure⁽ Reference Lauer and Clarke ^{27 –} Reference Li, Law and Power ^{34 )}, six showed positive associations⁽ Reference Lauer and Clarke ^{27 –} Reference Freedman, Khan and Dietz ^{29 ,} Reference Hardy, Wadsworth and Langenberg ^{32 –} Reference Li, Law and Power ^{34 )}. However, when adjusted for adult BMI, only two associations remained positive⁽ Reference Hardy, Wadsworth and Langenberg ^{32 ,} Reference Field, Cook and Gillman ^{33 )} and four studies demonstrated significant negative associations (i.e. lower BMI in childhood was associated with greater adult blood pressure)⁽ Reference Freedman, Khan and Dietz ^{29 ,} Reference Wright, Parker and Lamont ^{30 ,} Reference Hardy, Wadsworth and Langenberg ^{32 ,} Reference Li, Law and Power ^{34 )}. These included the two studies with the oldest adult cohorts⁽ Reference Wright, Parker and Lamont ^{30 ,} Reference Li, Law and Power ^{34 )}, which scored well in terms of quality of assessment and would be expected to give a better representation of lifetime risk of developing hypertension. Of the two studies that showed a positive association in the adjusted data set, one had the fewest number of participants and the adult age was young⁽ Reference Field, Cook and Gillman ^{33 )}. There were therefore very few hypertensive cases to consider (16/130 men). Six studies considered the association between childhood BMI and carotid intima media thickness (CIMT)⁽ Reference Wright, Parker and Lamont ^{30 ,} Reference Oren, Vos and Uiterwaal ^{35 –} Reference Freedman, Patel and Srinivasan ^{39 )}, with five of these studies showing a positive association⁽ Reference Oren, Vos and Uiterwaal ^{35 –} Reference Freedman, Patel and Srinivasan ^{39 )}. However, after adjusting for adult BMI only one positive association remained, and this was for the cumulative change in BMI from childhood to adulthood⁽ Reference Freedman, Patel and Srinivasan ^{39 )}. It must be noted that these studies assessed CIMT at a relatively young age in adulthood, so may be limited in their ability to determine the long-term risk associated with childhood obesity. The one exception was the study by Wright et al.⁽ Reference Wright, Parker and Lamont ^{30 )} which measured CIMT at age 50 years and showed no association with childhood BMI. Using data from the Bogalusa Heart Study, Freedman et al.⁽ Reference Freedman, Dietz and Tang ^{37 )} showed no relationship between childhood BMI status and adult CIMT among adults who were of normal BMI. Interestingly, those in the 90th centile for CIMT actually had lower BMI up to age 7 years.

Analyses of the association between childhood BMI and circulating biomarkers of metabolic function showed a similar pattern. The relationship with total cholesterol was variable between studies⁽ Reference Sinaiko, Donahue and Jacobs ^{28 –} Reference Wright, Parker and Lamont ^{30 ,} Reference Lauer, Lee and Clarke ^{40 )}, but both of the analyses which adjusted for adult BMI showed a negative association between the total cholesterol and childhood BMI⁽ Reference Freedman, Khan and Dietz ^{29 ,} Reference Wright, Parker and Lamont ^{30 )}. A positive association between plasma LDL cholesterol and BMI⁽ Reference Freedman, Khan and Dietz ^{29 )}, and negative associations between HDL cholesterol and BMI⁽ Reference Sinaiko, Donahue and Jacobs ^{28 ,} Reference Freedman, Khan and Dietz ^{29 )} were observed, but these were attenuated or reversed in the adjusted data sets. Similarly the data were mixed for TAG in the unadjusted data sets⁽ Reference Sinaiko, Donahue and Jacobs ^{28 –} Reference Wright, Parker and Lamont ^{30 ,} Reference Salonen, Kajantie and Osmond ^{41 )}, but only negative associations were reported for data sets adjusted for BMI⁽ Reference Freedman, Khan and Dietz ^{29 ,} Reference Wright, Parker and Lamont ^{30 )}. Three out of six⁽ Reference Sinaiko, Donahue and Jacobs ^{28 –} Reference Wright, Parker and Lamont ^{30 ,} Reference Freedman, Khan and Serdula ^{42 –} Reference Thearle, Knowler and Krakoff ^{44 )} studies looking at insulin concentrations or resistance showed a positive association with childhood BMI⁽ Reference Sinaiko, Donahue and Jacobs ^{28 ,} Reference Freedman, Khan and Dietz ^{29 ,} Reference Freedman, Khan and Serdula ^{42 )}, but these were again attenuated or reversed with adjustment for BMI. Similarly to the cardiovascular outcomes, the two studies using older adult cohorts showed negative relationships with childhood BMI⁽ Reference Wright, Parker and Lamont ^{30 ,} Reference Martin, Holly and Smith ^{43 )}.

The studies which assessed metabolic and CVD outcomes rather than risk factors were of better quality in terms of length of follow-up⁽ Reference Eriksson, Forsen and Tuomilehto ^{45 –} Reference Lawlor, Martin and Gunnell ^{47 )}. However, the nature of these studies meant that no adjustments for adult BMI were made. Only one paper used type-2 diabetes mortality as an outcome measure and found no association with childhood BMI status⁽ Reference Bjorge, Engeland and Tverdal ^{48 )}. Mixed results were found for the risk of metabolic syndrome, with the only study to adjust for adult BMI showing a negative association⁽ Reference Salonen, Kajantie and Osmond ^{41 )}.

To summarise, there was very limited and weak evidence that childhood obesity is an independent risk factor for adult disease. Instead, the associations appeared dependent on tracking of BMI from childhood to adulthood, with the majority of studies which adjusted for adult BMI showing attenuation or reversal of the associations observed. This conclusion has been supported by a more recent review⁽ Reference Park, Falconer and Viner ^{7 )}, which set out with similar objectives and included a wider spectrum of adult disease (e.g. asthma and cancers). Interestingly, the analyses adjusted for adult BMI were more suggestive of an inverse association between childhood BMI and adult disease risk factors⁽ Reference Lloyd, Langley-Evans and McMullen ^{22 ,} Reference Lloyd, Langley-Evans and McMullen ^{23 )}. Children at the lower end of the BMI range appeared most susceptible to the metabolic risks associated with adult obesity, including dyslipidaemia, elevated blood pressure (and, less convincingly, insulin resistance). Additionally obese adults appeared more likely to exhibit healthier metabolic profiles if they were also obese in childhood, with childhood obesity conferring a protective effect if BMI was reduced in adulthood.

While these patterns in the data became clear after systematically extracting unadjusted and adjusted data sets for the purpose of review, there is very little comment on the attenuated or negative associations in the individual papers or narrative reviews on the subject. Interpretation or reporting bias seems implicit in driving the focus on the more intuitive positive associations between childhood obesity and adult disease risk in the unadjusted data sets.

Statistical considerations: the reversal paradox

In highlighting the predominance of negative associations in the adjusted data sets, it must be noted that there are potential analytical issues associated with the use of statistical adjustment for an exposure measure in this way. Anthropometric measures are often considered to be confounders for health-related outcomes, with regression analyses adjusted accordingly. However, in many instances, body weight or BMI are not true confounders but part of the causal pathway between exposure and outcome⁽ Reference Kirkwood and Sterne ^{49 )}. In the data sets being discussed in this review, adjusted and unadjusted effect estimates are being compared which distinguish the indirect effect of childhood BMI (i.e. acting through adult BMI as the specified intermediate) from its direct effects (i.e. acting via pathways independent of adult BMI). Using this method of decomposition of effects⁽ Reference Kaufman, Maclehose and Kaufman ^{50 )}, if adjustment for the proposed intermediate exposure greatly attenuates the association between exposure and outcome, it is inferred that the exposure's effect is mediated predominantly through a pathway involving that intermediate. If minimal attenuation is observed, it is assumed that the exposure is acting through pathways which are independent of the proposed intermediate, in this case adult BMI. However, there are potential issues with the use of statistical adjustment in this way, as demonstrated in a noteworthy statistical simulation experiment conducted by Tu et al.⁽ Reference Tu, West and Ellison ^{51 )}. Computer simulations of three hypothetical relations between birth-weight and adult blood pressure were used to investigate the effect of statistically adjusting for adult weight, as a potential mediator of the relationship. An association between birth weight and blood pressure could be induced, exaggerated or reversed by adjustment for adult body weight, depending on the simulated direction of the unadjusted association and the strength of the correlations between adult body weight and exposure or outcome. Of particular relevance to the data sets systematically reviewed by ourselves, the simulated modest positive relationship between birth-weight and blood pressure (Scenario 3) could be reversed by strengthening the correlations between these variables and the proposed intermediate adult weight. Given the strength of association between adult BMI and CVD risk factors, it is plausible that the associations between childhood BMI and CVD risk factors could be at risk of the same reversal paradox when adjustments for adult BMI are made. Therefore, while the data are clear in terms of demonstrating an attenuation of the association between childhood BMI and adult CVD risk when adjusting for adult BMI, the reverse associations observed must be interpreted with caution.

Could childhood obesity be protective at a given level of adult BMI?

Although the inverse associations between childhood BMI and adult disease risk that are observed after adjustment for adult BMI are somewhat counterintuitive, it is important not to perpetuate any interpretation bias by only attempting to explain them as a possible statistical artefact. It is known that there are considerable variations between individuals in the level of adiposity at which insulin resistance is exhibited; some relatively lean individuals are insulin resistant, whereas some very obese individuals are not⁽ Reference Bonora, Kiechl and Willeit ^{52 )}. Obese individuals who present as ‘metabolically normal’ are of considerable interest, with 16·6 % of obese adults categorised as metabolically healthy using relatively stringent criteria⁽ Reference Wildman, Muntner and Reynolds ^{53 )}. Uncertainty surrounds the important question of what determines the point at which an otherwise ‘healthy’ storage tissue begins to promote the development of components of the metabolic syndrome. Importantly for the research questions being examined in this review, metabolically normal obesity has been associated with an earlier onset of obesity⁽ Reference Brochu, Tchernof and Dionne ^{54 )} and insulin resistance has been found to be progressively lower with longer duration of obesity⁽ Reference Muscelli, Camastra and Gastaldelli ^{55 )}. These studies were based on small samples of the population, but the data are consistent with the hypothesis that childhood obesity may offer protection against the metabolic effects of obesity in adulthood. Adipocyte size is inversely related to insulin sensitivity⁽ Reference Lundgren, Svensson and Lindmark ^{56 )} and risk of type II diabetes⁽ Reference Weyer, Foley and Bogardus ^{57 )}. Obese individuals with relatively few large adipocytes exhibit lower insulin sensitivity than those with the same degree of obesity but many small fat cells⁽ Reference Lundgren, Svensson and Lindmark ^{56 ,} Reference Weyer, Foley and Bogardus ^{57 )}. Of particular relevance to this project is the evidence that obesity with relatively small adipocytes has been associated with onset during childhood⁽ Reference Salans, Cushman and Weismann ^{58 )}. Adipocyte number, size and relative distribution between different adipose tissue depots are all therefore implicated in determining the metabolic consequences of a positive energy imbalance. The mechanisms responsible for the development of different patterns of adipose morphology are unknown. It is of particular interest that these are related to early life factors, but further research is required to understand the influence of excess adiposity during childhood on adipocyte number, size and relative distribution between depots.

Methodological considerations: sampling and assessment

Given the long-term nature of the associations being examined in this literature, it is inevitable that the childhood assessments were made in cohorts studied before the current obesity epidemic. It has been shown that the persistence of overweight between childhood and adulthood is generally higher with increasing childhood BMI⁽ Reference Singh, Mulder and Twisk ^{8 )}. Widespread and extreme obesity are more common now^{( 1 )}, combined with an obesogenic environment⁽ Reference Lobstein ^{9 )}; so current estimates may underestimate the strength of tracking of obesity between childhood and adulthood in contemporary populations. Additionally, research has demonstrated that many obese children already manifest some metabolic complications, including impaired glucose tolerance, dyslipidaemia, hypertension, fatty liver disease and systemic low-grade inflammation⁽ Reference Ford ^{59 –} Reference Bell, Byrne and Thompson ^{61 )}. Cross-sectional analysis of obese children showed that the prevalence of metabolic syndrome increased directly with the degree of obesity⁽ Reference Weiss, Dziura and Burgert ^{60 )}, with levels of insulin resistance exacerbated by increased deposition of lipid in the visceral and intramyocellular compartments at any given BMI⁽ Reference Weiss, Taksali and Dufour ^{62 )}. While the reviewed data sets only provide weak evidence that an association between childhood and adult disease risk exists independently of adult BMI, it remains a possibility that the levels of childhood obesity in contemporary society may have a greater permanency of effect. Increased understanding of the particular phenotypic characteristics, which are associated with the presence of obesity-related cardiometabolic dysfunction during childhood and adolescence itself remains an important area of research.

Another limitation of the majority of longitudinal cohort studies published to date is that they rely on BMI data as a measure of adiposity, when it is known to reflect both fat and fat-free mass. This is a particularly important point when considering the association between BMI and metabolic disease, given the opposing influences which abdominal adipose tissue and skeletal muscle have on insulin sensitivity and glucose disposal⁽ Reference Galgani, Moro and Ravussin ^{63 )}. BMI is also highly age-dependent in children, with a progressive increase in lean (as well as fat) mass throughout childhood and adolescence⁽ Reference McCarthy, Samani-Radia and Jebb ^{64 )}. Interestingly, Wright et al.⁽ Reference Wright, Parker and Lamont ^{30 )} showed a relatively weak correlation between childhood and adulthood BMI, likely to reflect the increasing contribution of other factors to adult body composition as the follow-up period lengthens. This study also found that percentage body fat during adulthood was not associated with childhood BMI, despite childhood and adult BMI being correlated. Overrepresentation of children with a lower lean body mass at the lower end of the BMI scale may complicate interpretation of the associations observed, given the known tracking of build⁽ Reference McCarthy, Samani-Radia and Jebb ^{64 )} and involvement of lean mass in glucose metabolism⁽ Reference Galgani, Moro and Ravussin ^{63 )}.

Conclusion

The data reviewed provide little evidence to suggest that childhood overweight and obesity is an independent risk factor for metabolic and cardiovascular risk during adulthood, and that which does exist should be considered weak. Instead, the data demonstrate that the relationships observed are dependent on tracking of BMI between childhood and adulthood, alongside persistence of dietary and physical activity behaviours. Adjustment for adult BMI uncovers unexpected negative associations between childhood BMI and adult risk factors, suggesting a protective effect of childhood obesity at any given level of adult BMI. Further investigation is required to explain these findings, both in terms of potential biological pathways and statistical artefacts. It is important to stress at this point that it is not being suggested that childhood obesity is without negative consequence. Childhood obesity is clearly associated with a range of adverse physical, psychological and social outcomes⁽ Reference Park, Falconer and Viner ^{7 ,} Reference Ford ^{59 –} Reference Bell, Byrne and Thompson ^{61 ,} Reference Puhl and Latner ^{65 )}. In addition, the tracking of adiposity and related behaviours from childhood to adulthood increases the risk of morbidity and mortality related to chronic disease in adulthood. However, the data are important in supporting the positive message that the long-term consequences of childhood obesity are avoidable; and that there remains opportunity for intervention across the lifespan. This is in contrast to the notion that childhood obesity has permanent irreversible effects on cardiovascular health, and supports the well-established evidence that weight loss in adulthood can lead to important positive health outcomes. This nuance in understanding long-term risk is particularly important when considering the effectiveness of interventions at different stages of the lifespan. Using an established risk factor focused modelling tool Dynamic Modelling for Health Impact Assessment, it has been demonstrated that reducing obesity prevalence in adulthood has much greater outcomes in terms of chronic disease and life expectancy than the same reduction in obesity prevalence during childhood⁽ Reference Lhachimi, Nusselder and Lobstein ^{66 )}. In addition, the health gains associated with a large reduction in childhood obesity were offset by a relatively small increase in risk of becoming overweight or obese in adulthood. It is important to target interventions to the stages of the life-course at which they are most likely to be effective and which offer the best long-term benefits. This has to be balanced against the potential for negative consequences if the timing of intervention coincides with critical stages of neurological, behavioural and physical developments. Understanding the pathways by which childhood obesity influences long-term disease risk will help inform the design and focus the targets of health promotion programmes during these sensitive stages.