Effects of Low Birth Weight on Insulin Resistance Syndrome in Caucasian and African-American Children

Study subjects were part of an ongoing longitudinal study of intra-abdominal fat and disease risk in children (19). The current study used data from 139 children (86 Caucasians, 53 African-Americans) aged 4–12 years (mean 8.1) at the start of the study and 8–14 years (mean 11.7) at the end of the study. Ethnicity was determined by self-report, based on both patents and both sets of grandparents being of the same ethnicity. This study used an accelerated longitudinal design by which limited longitudinal data on a specific age cohort can be linked to determine a common developmental trend over time (21). Children were followed annually for 3–5 years with an average of four repeated measures for each subject, which resulted in a total of 560 observations. The children were recruited from Birmingham, AL, and had been free of major illnesses since birth. The nature, purpose, and possible risks of the study were carefully explained to the parents before consent was obtained. This study was approved by the Institutional Review Board at the University of Alabama at Birmingham. All measurements were performed at the General Clinical Research Center and the Department of Nutritional Sciences at the University of Alabama at Birmingham during the school year (Fall and Spring) between 1994 and 1999.

At each annual visit, children were admitted to the General Clinical Research Center in the late afternoon for an overnight visit. Anthropometric measurements, including sexual maturation, were obtained. After 2000, only water and energy-free, noncaffeinated beverages were permitted until after the morning testing. In the morning after an overnight fast, blood was collected for hormone analysis and a tolbutamide-modified frequently sampled intravenous glucose tolerance test was performed. Two weeks later, the children arrived at the Energy Metabolism Research Unit at 0700 in the fasted state. Body composition was determined by dual-energy X-ray absorptiometry.

The original data on birth weight were obtained by parental recall. The subjects whose birth weight was <2,500 g (or 5.5 lb) were categorized as LBW, and those whose birth weight was ≥2,500 g were categorized as normal birth weight (NBW). Body weight during study visits was measured to the nearest 0.1 kg while subjects were wearing only light clothing. Height was measured to the nearest 0.5 cm with the subject standing without shoes, with heels together, and with the head in the horizontal plane.

Tanner’s criteria were used to estimate sexual maturation on the scale of 1–5, with stage 1 representing prepubertal and 5 representing adult (22,23). The same pediatrician assessed Tanner stage in all the children. Tanner stage was regrouped as 1 (stage 1), 2 (stage 2), and 3 (stage 3 or above) in current study.

Subcutaneous abdominal adipose tissue and intra-abdominal adipose tissue (or visceral fat) were measured by computed tomography with a HiLight/Advantage Scanner (General Electric, Milwaukee, WI) as described previously (24). Body composition (fat mass and fat-free mass) was measured by dual-energy X-ray absorptiometry using a Lunar DPX-L densitometer (Lunar Radiation, Madison, WI), as described previously (25).

Insulin sensitivity and the acute response to intravenous glucose were determined using the frequently sampled tolbutamide-modified intravenous glucose tolerance test, as described previously (19). In brief, at time 0, glucose (25% dextrose, 11.4% g/m²) was administered intravenously. Blood samples (2.0 ml) were then collected at the following times relative to glucose administration at 0 min: 2, 3, 4, 5, 6, 8, 10, 14, 19, 22, 25, 30, 40, 50, 70, 100, 140, and 180 min. Tolbutamide (125 mg/m²) was injected intravenously at 20 min. Sera were analyzed for glucose and insulin, as described previously (19), and values were entered into the MINMOD computer program (Version 3.0; Richard N. Bergman, Los Angeles, CA) for determination of insulin sensitivity (S_i). Acute insulin response (AIR), based on the area above baseline insulin concentration, was calculated by the trapezoidal method during 0–10 min. The disposition index was calculated from the product of S_i and the AIR (S_i × AIR) and used as an index of β-cell function (26).

HDL cholesterol and triglyceride were measured with the Ektachem DT II System (Johnson & Johnson Clinical Diagnostics, Rochester, NY). With this system, HDL cholesterol is measured after precipitation of LDL and VLDL with dextran sulfate and magnesium chloride. Control sera of low- and high-substrate concentration are analyzed with each group of samples, and values for these controls must fall within accepted ranges before samples were analyzed. The Ektachem DT II system is calibrated every 6 months with reagents supplied by the manufacturer.

A linear random-effects (or termed mixed-effects) modeling approach was conducted to capture the within- and between-subject variations in this longitudinal study. The major advantages of the linear random-effects model are its ability to account for the correlations appropriately between measurements inherent in a longitudinal study and its flexibility of modeling variations in variance and covariance of random-effect variables (27).

Fasting insulin, insulin sensitivity, AIR, disposition index, HDL cholesterol, and triglyceride were not normally distributed and, therefore, were log-transformed before data analyses. To obtain appropriate interpretations of intercepts (mean levels) and slopes (average trends) and to avoid possible colinearity in the random-effects models, age and other time-variant covariates (such as total fat mass) were further centered to their grand mean levels (i.e., original scores subtract from the grand means). The SAS PROC MIXED procedure (SAS Institute, Cary, NC) (28) was used to test the hypothesis proposed previously.

At the first visit, the total sample comprised 38.9% boys and 61.9% Caucasians; at this point, most of the children were in Tanner stage 1 (92.8%). The LBW group had a mean birth weight of 1.95 ± 0.2 kg, and the NBW group had a mean birth weight of 3.6 ± 0.1 kg. There were no significant differences in sex, ethnicity, average age, and average weight between LBW and NBW children; however, LBW children were lower in stature, had higher fasting insulin, and had a lower AIR, as summarized in Table 1.

Average fasting insulin level across full birth weight range was 12.7 μU/dl with a 95% CI of 11.9–13.5 μU/dl. Among African-American children, LBW was significantly associated with higher fasting insulin. Among children with LBW, the fasting insulin was significantly higher in African-Americans than Caucasians (Fig. 1A).

Average S_i level across full birth weight range was 4.01 × 10⁻⁵ min⁻¹ · μU⁻¹ · dl⁻¹ with a 95% CI of 3.70–4.44 × 10⁻⁵ · min⁻¹ · μU⁻¹ · dl⁻¹. S_i was significantly lower among African-American children than among Caucasian children (P < 0.01), regardless of birth weight (Fig. 1B).

The average AIR level across full birth weight range was 871.3 μU · dl⁻¹ · min⁻¹ with a 95% CI of 800.9–947.9 μU · dl⁻¹ · min⁻¹. Among African-American children, LBW was significantly related to a lower AIR (P < 0.01). Among children with NBW, African-Americans had higher AIR concentration than Caucasians (P < 0.001; Fig. 1C).

The disposition index was lower in both Caucasian and African-American children with LBW than those with NBW, although this difference did not reach statistical significance (P > 0.05). However, among children with NBW, African-American children showed a significantly higher disposition index than Caucasian children (P < 0.001; Fig. 1D).

Average visceral fat across full birth weight range was 31.2 cm² with a 95% CI of 26.7–36.7 cm². LBW was significantly associated with greater visceral fat among African-American children (P < 0.05), but not among Caucasian children. In children with NBW, African-Americans had a significantly lower visceral fat level than Caucasians (P < 0.01; Fig. 1E).

Average HDL cholesterol level across full birth weight range was 41.7 mg/dl with a 95% CI of 39.3–44.8 mg/dl. Among African-American children, LBW was significantly associated with lower HDL cholesterol than NBW (P < 0.05; Fig. 1F).

The significant interaction between LBW and ethnicity was detected for fasting insulin (P < 0.05) and visceral fat mass (P = 0.05), indicating that the effects of LBW on fasting insulin and visceral fat were more pronounced in African-American children (Fig. 1, Table 2).

The regression coefficients of the interaction term between LBW and age in Table 2 reflected the effects of LBW on the developmental trends of outcome variables. The results showed that LBW was associated with a significantly lower age-related decrement of AIR (P < 0.05) and the disposition index as a marker of β-cell-function (P < 0.01), but higher age-related increment of triglycerides (P < 0.01).

The regression coefficient of the interaction term between LBW and S_i in Table 2 indicates the effect of LBW on the slope of AIR regressed on S_i. The data showed that hyperbolic function between S_i and AIR was retarded among children with LBW (P = 0.04; Fig. 2).

More than 90% of both LBW and NBW children were in prepuberty at the start of the study (Table 1). LBW was not related to either the average Tanner stages or the transition across stages. However, faster transition of Tanner stages over age was examined in African-American as compared with Caucasian children (β = 0.50, P < 0.0001 vs. β = 0.25, P < 0.0001; γ = 0.13, P < 0.001 for testing the differences between the two slopes). Tanner stage was significantly associated with a higher level of fasting insulin, systolic blood pressure, and triglycerides and was marginally associated with lower insulin sensitivity. No interaction between Tanner stage and LBW on the components of IRS was identified (Table 2).

The present study tested the differential effects of LBW on the components of IRS in Caucasian and African-American children and examined the effects of LBW on the developmental trends of certain components of IRS in relation to age and total fat mass using a longitudinal random-effects modeling approach. The results showed that 1) LBW was significantly associated with increased fasting insulin concentration, visceral fat mass, triglycerides, decreased AIR, β-cell function, and HDL cholesterol among African-American children; 2) among children with LBW, there were significant differences in fasting insulin, S_i, AIR, and HDL cholesterol between Caucasians and African-Americans; 3) interactions between LBW and ethnicity on fasting insulin concentration and visceral fat were statistically significant; 4) LBW was associated with a more rapidly decreasing AIR and increasing triglyceride levels with regard to age; 5) the hyperbolic function between S_i and β-cell function was retarded among children with LBW; and 6) no significant associations between LBW and fasting glucose concentration, S_i, and systolic blood pressure were detected among both Caucasians and African-Americans in this study.

The finding in our longitudinal study that LBW was significantly related to fasting insulin in African-American children was consistent with findings in India (12,13), Britain (14,15), the U.S. (16), and South Africa (17). The lack of an association between LBW and fasting glucose in both Caucasian and African-American children was similar to those in Britain (14,15) but was in contrast to those in India (12,13) and South Africa (17). The U-shaped relation between the categorical scale of birth weight and 2-h glucose concentration, as observed in Pima Indians in the U.S. (16), may provide some information that the average glucose concentration would be quite similar when birth weight was dichotomized as LBW (<2,500 g) and NBW (≥2,500 g) as in our study. Despite the lack of association with fasting glucose, LBW was significantly related to elevated fasting insulin, a consistent marker of insulin resistance in normoglycemic individuals (1). Furthermore, this relation was quite stable over age after adjusted for total body fat mass (data not presented), which provided evidence that fetal malnutrition leads to permanent alterations in structure and function of endocrine pancreas such as elevated insulin concentration and reduced β-cell mass, as proposed in the “thrifty phenotype hypothesis” (29).

The significant association between LBW and decreased β-cell function determined in our study provided further evidence that retarded fetal growth may result in a reduced number of pancreatic β-cells and, therefore, a reduced capacity of producing insulin (29). The well-known hyperbolic relationship between S_i and AIR is based on an increase in AIR in response to a reduction in S_i (30). However, as also shown in Fig. 2, the increase in AIR in response to decreased S_i is repressed among children with LBW, suggesting impairment in β-cell response among children with LBW. Animal experiments have shown that the offspring of rats fed an isocaloric, low-protein diet during pregnancy have reduced birth weight, islet size, and vascularization (31) as well as impaired insulin secretion (32). Our study provided first evidence for the association between LBW, as a marker of poor fetal nutrition, and impaired β-cell function in human subjects.

No previous studies have examined the associations between LBW and abdominal obesity and dyslipidemia in children in the context of IRS. In contrast with one of the past studies in adults (3), we found that LBW was significantly related to increased visceral fat mass. In our past studies, visceral fat was independently associated with elevated triglycerides and insulin (19,20). In the present study, we found that it was unlikely that increased visceral fat caused elevated fasting insulin in children, because LBW was significantly associated with both elevated fasting insulin and visceral fat. Instead, our study provided some evidence that the effect of LBW on fasting insulin may be mediated through visceral fat. Consistent with previous studies in adults (2,6) and children (33), LBW was significantly associated with reduced HDL cholesterol level and marginally associated with elevated triglycerides in African-American children, even after adjusted for total body fat. Moreover, the progression of triglycerides over age was more rapid among children with LBW than those with NBW. Although the mechanism linking LBW and dyslipidemia is unknown, our findings seem to support the suggestion that impaired growth of the fetal liver in utero may lead to permanent changes in lipid metabolism (34).

The lack of an association between LBW and glucose intolerance, S_i, and systolic blood pressure was in contrast with recent studies in young adults (9,10,35) and, therefore, provided little evidence in supporting the suggestion that intrauterine programming of the hypothalamic-pituitary-adrenal axis may be a functional mechanism underlying the association between LBW and the IRS (9,36). However, there were significant differences in fasting insulin, S_i, AIR, and HDL cholesterol between Caucasian and African-American children with LBW, and the interaction between LBW and ethnicity on fasting insulin and visceral fat was significant as determined in our study, suggesting that the genetic and environmental influence on insulin resistance may be more manifest during childhood (8,10,19,37,38).

Puberty was not related to LBW in our study; however, it was significantly associated with higher levels of fasting insulin and triglycerides and marginally associated with lower S_i. This finding was consistent with recent studies in which precocious pubarche was found to be associated with hyperinsulinemia (39) and puberty was related to insulin resistance (40). In particular, more rapid transition between Tanner stages as identified in African-Americans may explain, at least in part, the more pronounced effects of LBW on the components of the IRS in this population.

A limitation of our study was the lack of information for gestation at birth; therefore, we were unable to distinguish children with normal gestation time from those with premature delivery. Fortunately, the interpretation of our findings might not be affected by this limitation, because previous studies have shown that the effect of LBW on insulin resistance was independent of duration of gestation (2,14). However, our study might underestimate the association between LBW and the components of IRS, because past studies showed that inclusion of preterm births could obscure the relation of birth weight with metabolic variables (3). Another weakness of this study was attributable to the dearth of validation of parental recall of birth weight. However, several previous studies have suggested that recalled birth weights are sufficiently accurate for use in children ≤16 years of age, regardless of socioeconomic status (41). Whincup et al. (15) found that the strength of the relation between insulin and birth weight from birth record data were very similar to those obtained from maternal recall.

Despite these limitations, there are at least three unique strengths in our study. First, a longitudinal design made it possible to observe the individual changes of physical and plasma measurements over age and to improve the study power (18,42). Second, this study used advanced measurement techniques for accurate clinical assessment of insulin action and secretion and body composition and fat distribution. Finally, associations between LBW and major components of the IRS were examined thoroughly, and to our knowledge, some of these associations were identified for the first time in African-American children.

In conclusion, the present study indicated that LBW was associated with an increased risk of IRS among children aged 4–14 years. African-American children with LBW were at the highest risk of developing IRS during childhood. The findings suggest that hyperinsulinemia, decreased β-cell function, dyslipidemia, and central obesity, which are associated with LBW, might be the first components of IRS during childhood. More follow-up is necessary to establish whether and when other symptoms develop in young adulthood and later life.

This study was supported by the U.S. Department of Agriculture (M.I.G.), the National Institute of Child Health and Development (M.I.G., Grants R29-HD-32668 and R01-HD/HL-33064), and by General Clinical Research Center Grant M01-RR-00032.

We thank Rey Figueroa for physician coverage, Barbara Gower for insulin and minimal model analysis, Tena Hilario for coordination of this project, the staff of the General Clinical Research Center, and the children and families who participated in this study.