Background. Childhood obesity is related to adult levels of lipids, lipoproteins, blood pressure, and insulin and to morbidity from coronary heart disease (CHD). However, the importance of the age at which obesity develops in these associations remains uncertain.
Objective and Design. We assessed the longitudinal relationship of childhood body mass index (BMI, kg/m2) to adult levels of lipids, insulin, and blood pressure among 2617 participants. All participants were initially examined at ages 2 to 17 years and were reexamined at ages 18 to 37 years; the mean follow-up was 17 years.
Results. Of the overweight children (BMI ≥95th percentile), 77% remained obese (≥30 kg/m2) as adults. Childhood overweight was related to adverse risk factor levels among adults, but associations were weak (r ∼ 0.1–0.3) and were attributable to the strong persistence of weight status between childhood and adulthood. Although obese adults had adverse levels of lipids, insulin, and blood pressure, levels of these risk factors did not vary with childhood weight status or with the age (≤8 years, 12–17 years, or ≥18 years) of obesity onset.
Conclusions. Additional data are needed to assess the independent relationship of childhood weight status to CHD morbidity. Because normal-weight children who become obese adults have adverse risk factor levels and probably will be at increased risk for adult morbidity, our results emphasize the need for both primary and secondary prevention.
- CHD =
- coronary heart disease •
- BMI =
- body mass index •
- HDL =
- high-density lipoprotein
Childhood obesity is associated cross-sectionally with various risk factors for coronary heart disease (CHD).1 ,2Several longitudinal associations have also been documented, with overweight children and adolescents having an increased risk of adult obesity,2–5 adverse levels of several CHD risk factors in adulthood,6–8 and various adult comorbidities.9–15 It has been suggested16that 3 stages of growth may be critical for the development of persistent obesity that influences comorbidities in adulthood: the prenatal period,17 the period of adiposity rebound (ages 4–8 years),18 and adolescence.12
The contribution of the age of obesity onset to adult disease remains uncertain because few studies have obtained weight status in both childhood and adulthood and information on obesity-related morbidities. In addition, the studies that have found childhood obesity to be predictive of adult morbidity9–15 have focused on obesity that was already present rather than on incident obesity. If the age of obesity onset were found to influence the subsequent severity of obesity, risk factor levels, or adult co-morbidities, preventive efforts aimed at these ages could be given a higher priority.
The Bogalusa Heart Study19 has examined children, adolescents, and adults for cardiovascular disease risk factors in a biracial (one-third black) community in Louisiana. Because the panel design has resulted in repeated examinations among many participants, this study is well suited to examine the relationship of the age of obesity onset to the severity and complications of adult obesity.
Bogalusa, a community of approximately 40 000 people, is 70 miles northeast of New Orleans and is fairly typical of semirural towns in the South. Between 1973 and 1994, 7 cross-sectional studies of schoolchildren, with participation rates >80%, were conducted in Ward 4 of Washington Parish.20 With the exception of the first (1973–1974) study, which had an upper age limit of 14 years, all 5- to 17-year-old schoolchildren were eligible for examination. A person could have participated in up to 5 of these cross-sectional studies.
Participants who had been examined as children were also eligible to be reexamined in 4 studies of adults,21 which were conducted in 1982–1983, 1985–1986, 1988–1991, and 1995–1996. The upper age range of these adult studies was extended to include all participants examined as children, with the 1995–1996 study having a upper age of 37 years. For example, an 8-year-old examined in 1974 could have been reexamined in 1985 (age 19 years), 1988 (22 years), and 1995 (29 years).
A total of 9597 children participated in the Bogalusa Heart Study, and these analyses are based on 2617 (27%) participants who were reexamined as adults; people in this cohort were examined an average of 2 times before age 18 and 2 times during adulthood. As compared with children lost to follow-up, children who were reexamined as adults were more likely to be white (67% vs 65%), female (58% vs 45%), and older (10.0 years vs 8.8 years). After adjustment for these characteristics, however, mean (baseline) body mass index (BMI, kg/m2) and triceps skinfold thickness were similar among participants and nonparticipants (P > .25).
The current analyses use information on weight and height obtained at the initial childhood examination (mean age 10 years) and anthropometric and risk factor information obtained at the final adult examination (27 years). Analyses that examine the age of obesity onset are limited to obese adults who had been examined at least twice (≤8 years and 12–17 years) as children.
The examination procedures used in the Bogalusa Heart Study have been described in detail.19 The triceps skinfold thickness was measured 3 times in succession with Lange skinfold calipers, and the mean value is included in the analyses. Approximately 74% of the participants in the current study were examined as children in 1973–1974, and the triceps skinfold was the only skinfold measured in that cross-sectional examination.
Weight was measured to the nearest 0.1 kg using a balance beam scale, and height was measured to the nearest 0.1 cm with a manual height board. The BMI was used as an index of relative weight, and sex- and age-specific BMI percentiles for children were calculated from national data (1963–1994) using a modification of the LMS technique.22 This method summarizes age trends in 1) the power (the λ in LMS) of the Box-Cox transformation needed to approximate normality, 2) the mean (M), and 3) coefficient of variation (S).23 Any BMI percentile can be calculated from these 3 parameters.
Adults in the current study with a BMI ≥30 kg/m2were considered obese, and children and adolescents with a BMI ≥95th percentile were considered overweight.24 Adults with a BMI <25 kg/m2 were classified as normal weight, as were children with a BMI <50th percentile; we considered children with a BMI between the 85th and 95th percentile to be at risk of overweight. Although there are several classifications of obesity, these cutoff points follow the recent recommendations of several US government agencies.24
Lipids, Insulin, and Blood Pressure
All analyses were performed on fresh blood samples in the Bogalusa Heart Study Core Laboratory. Participants were instructed to fast for 12 hours before the screening, and compliance was assessed by interview; lipid and insulin levels in the current study are based on these fasting determinations.
Serum concentrations of cholesterol and triglycerides were measured before 1986 using chemical procedures (Auto Analyzer II, Technicon, Tarrytown, NY) and enzymatic procedures (Abbott VP, North Chicago, IL) have been used since 1986. In the current analyses, 87% of the lipid determinations were performed using enzymatic procedures. Both procedures met the performance requirements of the Centers for Disease Control Lipid Standardization Program.
The concentration of low-density lipoprotein cholesterol was determined from the densitometric (electrophoretic) ratio and cholesterol contents following the heparin-calcium precipitation of β- and preβ-lipoproteins.25 Plasma insulin determinations were performed using a radioimmunoassay procedure (Phadebas Pharmacia, Piscataway, NJ). Right arm, sitting, systolic and diastolic blood pressures were measured by trained observers with a mercury sphygmomanometer.19
We focused on the relationship of childhood (mean age 10 years) obesity to adult (mean 27 years) risk factor levels, and the effects of race, sex, and age (at both baseline and follow-up) were controlled in all analyses. Because the distributions of triglyceride and insulin levels were skewed, these variables were log-transformed in several analyses, and geometric means were obtained after exponentiation. BMI categories in childhood and adulthood were cross-classified to assess the predictive value and sensitivity of childhood BMI, and this association was also examined graphically using a locally weighted scatterplot smoother (lowess).26
Associations between childhood BMI levels and adult risk factor levels were examined using Spearman (rank) correlations and by contrasting adult risk factor levels among participants whose childhood BMI was above the 95th percentile (overweight) or below the 50th percentile (normal weight). Stratification (by adult weight status), partial correlations, and regression models were used to determine whether the relationship of childhood BMI to adult risk factor levels was independent of weight status in adulthood. The importance of change in weight status was also examined, and this variable was constructed by subtracting the childhood BMI percentile from the adult BMI percentile; the latter was calculated from sex-specific rankings of BMI among all adults in the current study.
To examine the importance of the age at obesity onset to risk factor levels in adulthood, we focused on a subsample of the obese adults who had been previously examined 2 times: before 9 years of age and between ages 12 and 17 years. Risk factor levels were then contrasted between 3 groups of obese adults who differed by age of obesity onset: those with a BMI >95th percentile in both childhood and adolescence, those with a BMI >95th percentile in adolescence but not in childhood, and those with an adult BMI >30 kg/m2 but with BMI levels <95th percentile in both childhood and adolescence. These 3 groups were considered to have obesity onset in childhood, adolescence, or adulthood, respectively.
Mean levels of various characteristics at baseline and follow-up are shown in Table 1. The mean follow-up period was 17 years, with adult ages ranging from 18 to 37 years; about one-third of the cohort was black, and slightly more than half were women. As indicated by the baseline BMI percentiles, the weight status of children in the current study was very similar to national data. The 8.4-kg/m2 increase in mean BMI levels during follow-up, corresponding to a yearly average increase of 0.5 kg/m2, was accompanied by adverse changes in risk factor levels. Triglyceride levels showed the largest increase over the time period, with (geometric) mean levels increasing from 63 to 92 mg/dL, whereas mean levels of high-density lipoprotein (HDL) cholesterol decreased from 65 to 50 mg/dL. Based on self-reported information, approximately 1% of the adults had been diagnosed with diabetes mellitus or were using medication for diabetes.
Overweight children were very likely to become obese adults (Table 2): of the 186 children with a BMI ≥95th percentile, 144 (77%) had an adult BMI of ≥30 kg/m2. In contrast, only 7% of the 1317 normal-weight children became obese adults. Of the 581 adults who were obese (final column), 144 (25%) had been overweight as children and another 129 (22%) had a childhood BMI between the 85th and 94th percentile. The correlation between childhood BMI and adult BMI wasr = 0.58, and additional analyses indicated that this association did not vary substantially by age, race, or sex. For example, 52 (87%) of the 60 overweight children under 8 years of age became obese adults.
The relationship of childhood BMI percentiles to adult BMI is examined in the upper panel of Fig 1. The estimated adult BMI increased rapidly with childhood BMI percentiles above the median: Whereas a child at the 50th percentile had a predicted adult BMI of 24 kg/m2, a child at the 90th percentile was predicted to have an adult BMI of 31 kg/m2. Although the shape of the relationship did not vary substantially by race or sex, at comparable childhood BMI percentiles, black girls had an ∼3 kg/m2 higher adult BMI level than did white girls, reflecting racial differences in adult BMI levels (data not shown).
The relationship of BMI levels and triceps skinfold thickness to adult risk factors is shown in Table 3. Although many of the differences were small, associations with BMI levels were consistently stronger than were the corresponding associations with triceps skinfold thickness; this difference was seen among both adults and children. As expected, adult BMI levels were moderately related to the examined risk factors, with levels of insulin (r = 0.59) and triglycerides (r = 0.36) showing the strongest associations. Adult risk factors were also associated with childhood BMI (middle rows), but the correlations were only about 50% as large as those with adult BMI levels. For example, correlations with HDL cholesterol levels were −.31 (adult BMI) and −.14 (childhood BMI), and correlations with systolic blood pressure were .21 and .09. (Associations with childhood BMI varied only slightly by race and sex; data not shown.) Controlling for adult BMI (bottom row) eliminated the positive (inverse for HDL cholesterol) associations with childhood BMI, indicating that the effect of childhood weight status was mediated through its association with the adult BMI level. An increase in the relative BMI percentile during follow-up was also associated with adverse risk factor levels, with adult levels of insulin showing the strongest (r = 0.34) association with BMI change.
The greater importance of adult (vs childhood) BMI levels can also be seen in the bottom panel of Fig 1, in which regression-predicted adult insulin levels are shown. (Insulin levels showed the strongest association with levels of both childhood and adult BMI.) As indicated by the solid line, childhood BMI (x axis) was strongly related to adult levels of insulin, with a child at the 85th BMI percentile estimated to have a 5 mU/L higher insulin level in adulthood than a child at the 15th percentile. However, adjustment for adult BMI (dashed lines) indicated that at a specific adult BMI (18.8 kg/m2 or 30.7 kg/m2, the 15th and 85th percentiles), childhood BMI was only weakly (and inversely) associated with adult insulin levels. In contrast, the large difference in insulin levels according to adult BMI was independent of childhood weight status.
A stratified analysis is shown in Table 4, in which mean risk factor levels are shown within categories of both childhood and adult BMI. As compared with adults who were normal-weight children, those who had been overweight in childhood had adverse levels of all risk factor levels (first 2 columns), with the largest differences seen for triglyceride and insulin levels and in the prevalence of diabetes mellitus. The large (34.9- to 22.5-kg/m2) difference in adult BMI levels between the 2 groups reflected the strong tracking of weight status, and subsequent columns contrast mean risk factor levels by categories of both adult and childhood BMI. Within each adult BMI category, risk factor levels differed only slightly between participants who had been normal-weight or overweight as children. Among adults with a BMI of <25 kg/m2, for example, those who had been normal-weight children had a mean HDL cholesterol level of 54 mg/dL whereas those who had been overweight children had a mean level of 55 mg/dL; the mean HDL cholesterol level in both groups of obese adults (final 2 columns) was 42 mg/dL. The only statistically significant difference in these stratified analyses was for triglycerides: Among normal-weight adults, those who had been overweight during childhood had a 23 mg/L lower (53 vs 76 mg/dL) mean level than did those who been normal-weight children. Interestingly, obese adults who had been overweight children had a 4.9 kg/m2 higher mean adult BMI than did those who had been normal-weight children.
We then examined the relationship of the age at obesity onset to adult risk factor levels (Table 5). (These analyses were limited to 195 obese [BMI ≥30 kg/m2] adults who had been examined in both early childhood [<8 years] and adolescence [12–17 years].) Although the 38 obese adults who had been overweight before 8 years of age were much heavier (mean BMI 41.7 kg/m2) in adulthood than were those whose became overweight in adolescence or adulthood, a childhood onset of obesity was not associated with adverse risk factors in adulthood. In general, mean risk factor levels varied only slightly across the 3 groups, and none of the differences approached statistical significance.
In addition to being associated with adverse levels of CHD risk factors,1 ,2 childhood obesity is predictive of adult risk factor levels6–8 and CHD morbidity.9–15However, the relative contributions of childhood and adult obesity in these associations are uncertain, and the current findings provide additional insight into possible mechanisms. We found that the associations between adverse risk factor levels in adulthood and childhood obesity, which were weaker than those with adult BMI, resulted entirely from the strong tracking (persistence) of weight status from childhood to adulthood. Furthermore, risk factor levels among obese adults did not differ between those who had been normal weight or overweight in childhood. Although participants who became overweight early in life were more obese as adults than were those who became obese after childhood, the age of obesity onset did not show consistent (or statistically significant) associations with adult risk factor levels.
The tracking (persistence) of obesity has been well documented,2–5 ,15 ,27 and several investigators have reported that ∼50% of overweight children become obese adults.2 ,28 However, this estimate would be expected to vary substantially according to the prevalence of obesity, the BMI cutoff points used, and the time period examined. For example, the recent secular trends in obesity among Americans,20 ,29particularly for marked obesity, might be expected to increase the positive predictive value of childhood overweight. Whereas 40% of 7-year-olds with a BMI >95th percentile in the 1958 British birth cohort were obese at 33 years old,4 we observed a predictive value of 87% among children under 8 years of age. Additional studies are needed to document these trends, but the longitudinal relationship of childhood obesity to fibrous plaques30 and coronary artery calcification31indicates that the development of CHD may depend on the cumulative lifetime effects of obesity. The very high BMI level (mean 41.7 kg/m2) we observed among obese adults who had also been overweight in childhood may portend a marked increase in obesity-related comorbidities in the near future.
Obesity is not homogenous, and it has been suggested32that adults who have always been obese, and who are more likely to have a peripheral distribution of body fat differ from obese adults who were normal-weight children. It has also been hypothesized that various stages of growth may be critical for the development of persistent obesity and co-morbidities16: the prenatal period,17 the period of adiposity rebound,18and adolescence.12 For example, adolescence is a time of the deleterious accumulation of visceral fat,33 ,34 and an early adiposity rebound may be caused by hyperplastic obesity.18 Additional research is needed into these growth stages, and the complexity of the interrelationships is emphasized by the observation that a low birth weight in combination with childhood obesity is associated with an increased CHD risk.14
Overweight children are at increased risk for adult morbidity and mortality (particularly CHD).9–15 However, the strong tracking of obesity between childhood and adulthood raises the possibility that the increased risk associated with childhood overweight is mediated by adult weight status. Unfortunately, the few studies that have information on BMI levels in both childhood and adulthood have yielded conflicting results. The Harvard Growth Study, for example, found that obese adolescent boys (but not girls) were at increased risk for CHD mortality and that this association was independent of adult BMI.12 In contrast, others9 ,35 have concluded that the effects of childhood obesity were mediated by adult weight status, and some findings9 suggest that the risk may be highest among normal-weight children who became obese adults. In agreement with our findings (Table 4), others12 ,36 have also found that diabetes mellitus in adulthood is more closely linked to adult obesity than to childhood weight status.
Obese children have been found to have adverse risk factor levels in adulthood.6–8 ,37 However, the possible importance of adult BMI levels in these associations has not been previously assessed, and we found that the longitudinal associations are mediated by adult weight status and the strong tracking of BMI from childhood to adulthood. Adult BMI levels also reflect changes in weight status after childhood, and other studies have emphasized the effect of weight gain (or loss) on risk factor levels.8 36–38 In addition, both adolescent weight status and subsequent weight gain have been found to be independent predictors of risk factor levels among adults.8 Associations with weight change may explain, in part, the slightly protective effect of childhood BMI after adjustment for adult weight status (Table 3 and Fig 1). For a given adult BMI, a heavy adolescent would have experienced a smaller weight gain during follow-up than would a thin adolescent.
Several limitations of the current study should be considered. We examined risk factor levels in early adulthood, and our findings may not apply to levels among middle-aged and older adults. It is also possible that chronic conditions, such as atherosclerosis and CHD morbidity, are more strongly influenced by the cumulative, long-term effects of obesity than are risk factor levels. In addition, BMI does not measure body composition, and correlations between BMI and body fatness among children have ranged from approximately .6 to .8.39 However, we did find similar results using the triceps skinfold thickness (Table 3), and some investigators have found that BMI predicts various complications as well as skinfold thicknesses do.40 Although the substantial loss to follow-up in the current study was not associated with childhood BMI levels or skinfold thicknesses, the small number of children in the analyses of obesity onset age resulted in limited power.
Our results emphasize the importance of both primary and secondary prevention. Although we found that childhood weight status is not independently related to adult risk factor levels, it is likely that the cumulative lifetime risk of CHD is greatest among those who are persistently overweight throughout their years at risk. However, because the risks for adult morbidity are elevated among normal-weight children who become obese in adulthood, a population-based approach is also needed. Additional data are needed to assess whether the relationship of childhood weight status to CHD is independent of adult BMI.
This work was supported by Grants HL 15103 and HL 32194 from the National Heart, Lung, and Blood Institute, National Institutes of Health; and by funds from the Centers for Disease Control and Prevention and the Robert W. Woodruff Foundations.
- Received December 27, 2000.
- Accepted April 5, 2001.
Reprint requests to (D.S.F.) CDC Mailstop K-26, 4770 Buford Highway, Atlanta, GA 30341-3717. E-mail:
- Power C,
- Lake JK,
- Cole TJ
- Guo SS,
- Chumlea WC
- Lauer RM,
- Lee J,
- Clarke WR
- Lauer RM,
- Clarke WR
- Nieto FJ,
- Szklo M,
- Comstock GW
- Gunnell DJ,
- Frankel SJ,
- Nanchahal K,
- Peters TJ,
- Davey Smith G
- Eriksson JG,
- Forsen T,
- Tuomilehto J,
- Winter PD,
- Osmond C,
- Barker DJ
- Dietz WH
- ↵Berenson GS, ed. Cardiovascular Risk Factors in Children.New York, NY: Oxford University Press; 1980
- Freedman DS,
- Srinivasan SR,
- Valdez RA,
- Williamson DF,
- Berenson GS
- ↵Kuczmarski RJ, Ogden CL, Gummer-Strawn LM, et al. CDC growth charts: United States. Hyattsville, MD: National Center for Health Statistics; 2000 (Advance data from Vital Health Stat 314). Available at:http://www.cdc.gov/nchs/data/ad314.pdf
- Kuczmarski RJ,
- Flegal KM
- Srinivasan SR,
- Frerichs RR,
- Webber LS,
- Berenson GS
- ↵Cleveland WS. The Elements of Graphing Data. Monterey, CA: Wadsworth Advanced Books and Software; 1985:170–178
- Albrink MJ,
- Meigs JW
- Daniels SR,
- Morrison JA,
- Sprecher DL,
- Khoury P,
- Kimball TR
- Colditz GA,
- Willett WC,
- Stampfer MJ,
- et al.
- Sinaiko AR,
- Donahue RP,
- Jacobs DR Jr,
- Prineas RJ
- Spiegelman D,
- Israel RG,
- Bouchard C,
- Willett WC
- Copyright © 2001 American Academy of Pediatrics