Reduced final height and indications for insulin resistance in 20 year olds born small for gestational age: regional cohort studyBMJ 1997; 315 doi: https://doi.org/10.1136/bmj.315.7104.341 (Published 09 August 1997) Cite this as: BMJ 1997;315:341
- Juliane Leger, paediatriciana,
- Claire Levy-Marchal, paediatriciana,
- Juliette Bloch, statisticiana,
- Agnes Pinet, paediatriciana,
- Didier Chevenne, biochemistb,
- Dominique Porquet, biochemistb,
- Dominique Collin, obstetricianc,
- Paul Czernichow, paediatriciana
- a Paediatric Endocrinology and Diabetes Unit and INSERM CJF 93-13, Hôpital Robert Debré, 75019 Paris, France
- b Department of Biochemistry, Hôpital Robert Debré
- c Maternity Unit, Centre Hospitalier de Haguenau, Haguenau, France
- Correspondence to: Dr Leger
- Accepted 19 May 1997
Objective: To investigate whether the association between low birth weight and increased risk of developing impaired glucose tolerance, insulin resistance, hypertriglyceridaemia, and hypertension in middle age is apparent by the age of 20 in people born small for gestational age.
Design: Regional cohort study.
Setting: Maternity registry, Haguenau, France.
Subjects: 236 full term singleton babies born small for gestational age (birth weight or length, or both, below third centile) during 1971-8 and 281 with normal birth weight (between 25th and 75th centile). All subjects were contacted and evaluated at a mean (SD) age of 20.6 (2.1) years.
Main outcome measures: Adult height; concentrations of glucose, insulin, and proinsulin during an oral glucose tolerance test; lipid and fibrinogen concentrations; and blood pressure.
Results: After sex and target height were adjusted for, subjects who had been born small for gestational age were significantly shorter at age 20 than those with a normal birth weight (men 4.5 cm shorter (95% confidence interval 6.0 to 3.0 cm); women 3.94 cm shorter (5.2 to 2.7 cm). After sex and body mass index were adjusted for, mean plasma glucose concentration 30 minutes after a glucose load, fasting insulin concentration (in women), and insulin and proinsulin concentrations 30 and 120 minutes after a glucose load were significantly higher in subjects who had been born small for gestational age than in those with a normal birth weight. Mean lipid and fibrinogen concentrations and blood pressure were not different between the two groups.
Conclusions: Intrauterine growth retardation has long term consequences such as reduced final height. Raised insulin and proinsulin concentrations are present in young adults born small for gestational age and could be markers of early changes in insulin sensitivity.
An inverse relation between birth weight and increased risk of developing impaired glucose tolerance, insulin resistance, hypertriglyceridaemia, and hypertension (syndrome X) in middle age has been reported
In this study reduced final height and raised serum insulin and proinsulin concentrations during oral glucose tolerance testing were found in young adults born small for gestational age compared with young adults with an appropriate birth weight for gestational age
Reduced fetal growth was not associated with impaired glucose tolerance, higher blood pressure, or abnormalities in lipid and fibrinogen concentrations
Further longitudinal studies are required to examine the risk of developing the other elements of syndrome X later in life
Intrauterine growth retardation has long term consequences for postnatal growth and development. Affected infants are known to be at a higher than average risk of illness and death from several neonatal disorders.1 Although most (85-90%) will catch up in height during the first two years of life,2 3 some children will remain short in childhood and adulthood.3 4 5 6 A recent population based study following babies born small for gestational age to adulthood showed that they had a sevenfold higher risk of being short than subjects who were not small for gestational age.3
Recent findings suggest that the associated conditions of hypertension, non-insulin dependent diabetes or insulin resistance, and dyslipidaemia are more common in adults who had an abnormally low birth weight.7 8 9 10 Increased death rates from cardiovascular disease have also been reported.11 12 These studies indicate that these metabolic disorders may originate from impaired growth and development during fetal life.13 14 15 16 None of the studies, however, specifically investigated subjects born small for gestational age since length of gestation was not precisely known. We conducted a large population based study to examine the effect of intrauterine growth retardation on final height and metabolic variables in a cohort of young people.
Subjects and methods
We identified subjects for study from a population based registry of the metropolitan area of the city of Haguenau in France. This registry records information on all pregnancies, deliveries, and perinatal events in the area. Haguenau Maternity Hospital is the main maternity centre for this region of north east France, and the population is mainly white. The degree of ascertainment of the registry is greater than 80%.17 We selected subjects born between 1971 and 1978. During this period 10 830 singleton babies were born in the maternity hospital and included in the registry.
The growth standards of the population of Haguenau are different from those of the population of France in general because of the large number of Germanic people, so our standards were local values for birth weight, height, and head circumference by gestational age and sex as derived from live births registered during 1971-85. Gestational age was determined from the date of the mother's last menstrual period and by physical examination during pregnancy, and it was confirmed by ultrasound measurements when available. We included all 452 singleton subjects who had been small for gestational age and born at term (≥37 weeks of gestation) during 1971-8. Being small for gestational age was defined as having a birth weight or length, or both, below the third centile of the local standard values. The control subjects were 451 singletons who had had a normal birth weight for gestational age (between 25th and 75th centile); they were selected from the register as the first person with an appropriate birth weight who was born immediately after a subject who was small for gestational age without any attempt to match for sex and gestational age.
Altogether we identified 903 subjects for the study; 16 subjects with aberrant measurements (normal birth weight of greater than 3000 g but extremely short birth length less than 35 cm) due to recording errors at birth were excluded. Of the remaining 887 subjects, 517 (58%) agreed to take part in the study, 236 of whom had been small for gestational age and 281 controls. Table 1 shows the reasons for non-participation. As expected, the number of deaths was significantly higher in the group that was small for gestational age (32 (15%) v 9 (5%); P=0.002). The other causes for non-participation were equally distributed between the two groups. The proportion of women was greater among participants than non-participants (55% (285) v 47% (1881); P=0.02).
Among the control subjects we found no difference between participants and non-participants in gestational age at birth, placental weight, birth weight, and length at birth. Among the subjects who had been small for gestational age, the only variable that was significantly different between participants and non-participants was mean birth length, which was significantly higher for the participants (47.0 (SD 2.1) cm v 46.2 (3.3) cm; P=0.005). This was partly explained by the greater severity of intrauterine growth retardation among those who had died, their mean length at birth being 45.7 (3.9) cm and mean birth weight 2260 (380) g.
Degree of intrauterine growth retardation was evaluated by expressing birth weight and length as a standard deviation score and correcting for gestational age and sex according to the local growth standards. To evaluate intrauterine nutritional state the ponderal index was calculated as the ratio of birth weight (g) to the cube of length at birth ((cm)x100) corrected for gestational age according to the standard of Miller and Hassanein.18 We divided subjects into two groups according to whether the ponderal index was less than or equal to the third centile or greater than the third centile.
We asked each subject's parents for their current height and weight, and 97% of them were measured in the same standardised way as their child. Target height was calculated from mid-parental heights adjusted for the sex of the child.19
Subjects who agreed to participate in the study attended the hospital for a morning, and we took blood samples for measurement of fasting lipid concentrations (total cholesterol, high density lipoprotein cholesterol, triglycerides, apolipoproteins A1 and B) and concentrations of fibrinogen, glucose, insulin, and proinsulin. Subjects also had an oral glucose tolerance test with a glucose load of 75 g, and further blood samples were taken after 30 and 120 minutes for measurement of plasma glucose and serum insulin and proinsulin concentrations. We obtained a medical history using a questionnaire. The prevalence of diabetes, hyperlipidaemia, and hypertension was assessed in subjects and their parents and did not differ between the two groups.
The height of all subjects was measured twice to the nearest 0.1 cm by one paediatrician using the same wall mounted stadiometer; the average value was used in the analysis. Weight was measured on portable scales to the nearest 0.1 kg, and weight for height was assessed as body mass index (weight (kg)/(height)2 (m)). Waist circumference was measured at the level of the umbilicus and hip circumference was measured at the level of the greater trochanter.
We measured blood pressure in the right arm of seated subjects after five minutes' rest using an automated device (Dinamap, Critikon, Neuilly Plaisance, France) and a cuff of the recommended size for the mid-upper arm circumference. Three measurements were made at an interval of one minute, and the average of the last two measurements was used in the analysis.
All blood samples except those for measurement of glucose, fibrinogen, and lipid concentrations were centrifuged and serum was stored at -20°C until assayed. Glucose (mmol/l), cholesterol (mmol/l), high density lipoprotein cholesterol (after precipitation by phosphotungstic acid and magnesium chloride), and triglyceride (mmol/l) concentrations were measured by enzymatic methods. Fibrinogen (g/l) was measured by an automated method based on the time of fibrin formation after addition of thromboplastin-calcium reagent. Apolipoprotein A1 (g/l) and apolipoprotein B (g/l) were measured by kinetic nephelometry using specific antibodies to apolipoprotein A1 and B (Beckman, France).
Insulin concentration (pmol/l) was measured by a specific immunoradiometric assay using two monoclonal anti-insulin antibodies (Bi-insulin immunoradiometric assay, ERIA Diagnostics Pasteur, France). Intact proinsulin, split (32,33) proinsulin, and des (31,32) proinsulins did not cross react.20
Proinsulin concentration (pmol/l) was determined by an immunoradiometric assay based on a previously described method21; briefly, proinsulin was measured by using two monoclonal antibodies, one recognising the insulin part of the proinsulin molecule (3B1, Biochem Immunosystems, United Kingdom) and the other recognising the C peptide part (PEP-001, Novo Nordisk, France) of the proinsulin molecule. Molar cross reactivities of proinsulin intermediates were 80-100% compared with those of intact proinsulin. The detection limit was 0.3 pmol/l with an intra-assay coefficient of variation of less than 6% and an interassay variation of less than 9%.
Statistical analysis was performed with SAS software (SAS Institute, Cary, NC). Results are expressed as means (SD). Insulin, proinsulin, triglyceride, and apolipoprotein A1 concentrations had skewed distributions, which were normalised by logarithmic transformation.
One subject who was being treated for cystic fibrosis and diabetes, three who were being treated for hypertension, and four who were being treated for hyperlipidaemia were excluded from the analyses of glucose tolerance, blood pressure, and lipid concentrations respectively.
The significance of differences between groups was assessed by the χ2test, Fisher's exact test, and Student's t test as appropriate. A P value of 0.05 was considered to be significant.
Multiple linear regression models (GLM procedure) were fitted with biological parameters and blood pressure as dependent variables and group (small or appropriate for gestational age), sex, and body mass index as explanatory variables. First order interaction between groups and sex was tested. If significant, separate analyses were carried out for men and women. The relations between biological parameters and gestational age, adjusted for sex and body mass index, were investigated in separate linear models according to whether subjects had been small or of an appropriate size for gestational age, and a significant interaction was found between group and gestation. A stepwise multilinear regression was used to assess the relations between systolic blood pressure and biological parameters such as glycaemia and insulin, proinsulin, and lipid concentrations after adjustment for sex, body mass index, and family history of high blood pressure.
Multiple linear regression was also used to assess the relation between final height and group (small or appropriate for gestational age), adjusted for sex and target height.
The study protocol was reviewed and approved by the faculty ethics committee and all subjects and parents gave signed written consent.
Table 2 shows the clinical characteristics at birth of the two study groups. No significant differences were found in age, sex distribution, and gestational age between the two groups. Among the subjects small for gestational age, birth weight was below the third centile in 128 (54%), birth length was below the third centile in 50 (21%), and both birth weight and length were below the third centile for 58 (25%) subjects. The birth weight was from 2 SDS (standard deviation scores) to 3 SDS below the mean and more than 3 SDS below the mean in 101 and 14 subjects, respectively. The birth length was from 2 SDS to 3 SDS below the mean and more than 3 SD below the mean in 81 and 30 subjects, respectively. Fifty six (24%) had had a ponderal index below the third centile at birth. Subjects who had been small for gestational age had the following risk factors for intrauterine growth retardation: pregnancy induced hypertension (63 subjects, 27%), maternal smoking during pregnancy (72, 32%), congenital abnormalities (28, 12%), and maternal short stature (22, 9%); some subjects had more than one factor.
The mean (SD) age of the population at the time of the study was 20.6 (2.1) years (range 16.6-24.5 years). As shown in table 3, the mean height, weight, and head circumference at the time of study were significantly lower in those who had been small for gestational age than in those born at normal birth weight. The mean body mass index and ratio of waist to hip measurements were similar in the two groups. As the mean height of men was significantly lower from 16.6 to 18 years than from 18 to 24.5 years, we assumed that some of the younger men had not achieved their final height and excluded those younger than 18 from analysis of final height (19 who had been small for gestational age and 21 controls). However, we included all the women as they had completed puberty, with menarche having occurred at least 18 months before the study. No significant difference was found in mean (SD) age at menarche between the two groups (12.6 (1.6) v 12.9 (1.7) years). The mean target height was lower in subjects who were small for gestational age but was significantly different only for women (P=0.0001). Parents of subjects who had been small for gestational age were shorter than those of control subjects (P=0.02 for paternal height, 172.3 (6.7) v 173.8 (6.4) cm, P=0.0001 for maternal height 160.8 (6.2) v 163.1 (5.6) cm).
After adjustment for target height, a significant deficit in final height was found in those who were small for gestational age (men: -4.50 (95% confidence interval -6.0 to -3.0) cm, women: -3.94 (-5.2 to -2.7) cm). Twenty nine (13.4%) subjects who had been small for gestational age were short (>2 SD below mean height of controls) compared with seven (2.6%) of those who had had a normal birth weight.
Oral glucose tolerance test
Table 4 shows the results of oral glucose tolerance testing in the two groups. After sex and body mass index were adjusted for, fasting plasma glucose concentration, fasting serum insulin (for men), and proinsulin concentrations did not differ between the two groups. Among women, fasting serum insulin concentrations were significantly higher in those who had been small for gestational age than in controls.
Glucose concentrations 30 minutes after a glucose load and insulin and proinsulin concentrations 30 and 120 minutes afterwards were significantly higher in subjects who had been small for gestational age. After sex and body mass index were adjusted for, insulin and proinsulin concentrations at all studied times were negatively correlated with gestational age in the group that was small for gestational age but not in the control group (regression coefficients were -0.078 for fasting insulin (P=0.009), -0.093 for insulin at 30 minutes (P=0.006), -0.093 for insulin at 120 minutes (P=0.03), -0.063 for fasting proinsulin (P=0.03), -0.084 for proinsulin at 30 minutes (P=0.007), and -0.064 for proinsulin at 120 minutes (P=0.05)). Consequently, the difference between the groups in serum insulin and proinsulin increased with decreasing length of gestation. This finding was independent of the degree of intrauterine growth retardation as measured by numbers of standard deviations below the mean birth weight or birth length or of the ponderal index at birth.
Neither the ratio of proinsulin to insulin concentrations nor the ratio of the 30 minute increment in insulin concentration to the 30 minute increment in glucose concentration after the oral glucose load, taken as indicators of insulin secretion,22 23 were different between the two groups.
No association was found between the significant results of the glucose tolerance test and the final height, the degree of intrauterine growth retardation at birth (2-3 SD below the mean and >3 SD below the mean), the ponderal index at birth (low or normal), or risk factors associated with intrauterine growth retardation during pregnancy in subjects born small for gestational age.
The mean systolic and diastolic blood pressure values in subjects who had been small for gestational age were not significantly different from those in control subjects after adjustment for sex, body mass index, and height. The mean (SD) systolic and diastolic blood pressure values were respectively 124.4 (10.5) v 125.3 (10.3) mm Hg and 62.7 (8.3) v 64.8 (8.8) mmHg in men and 117.0 (9.9) v 115.3 (10.3) mm Hg and 63.4 (8.8) v 62.6 (8.1) mm Hg in women. No relation was found between blood pressure and either placental weight or gestational age.
Stepwise multiple regression analysis showed that factors positively associated with systolic blood pressure were body mass index (P=0.0001), sex (P=0.0001), age (P=0.01), height (P=0.01), tobacco consumption (P=0.02), fasting plasma insulin concentration (P=0.002), and total cholesterol concentration (P=0.03).
Serum lipid and fibrinogen concentrations
After sex and body mass index were adjusted for no significant difference was found in serum lipid or fibrinogen concentrations (adjusted also for tobacco consumption) between the two groups (table 5). Triglyceride concentrations were positively related to fasting insulin and proinsulin concentrations and to insulin and proinsulin concentrations at 30 and 120 minutes (P<0.001). No relation was found between high density lipoprotein cholesterol, insulin, and proinsulin concentrations.
Our study demonstrates that being born small for gestational age is associated with reduced final height and increased serum insulin and proinsulin concentrations with normal glucose tolerance in young adults. In common with other studies3 4 we also found a significant reduction in parental height in subjects who had been small for gestational age; maternal height was more reduced than was paternal height, as short women are known to have a higher risk of babies with intrauterine growth retardation.24 Mean final height in the group born small for gestational age remained significantly lower after adjustment for the target height in both sexes. Reduced mean final height (without correction for target height) has been observed in adolescents aged 16-18 who had been born at term but small for gestational age.3 4 5 6 Subjects with a normal birth weight in our study were 4 cm taller than the target height in both sexes, documenting the upward secular drift in height.
Being small for gestational age has been defined according to birth weight or birth length, and different cut off points have been used such as two standard deviations below the mean or the third, fifth, or 10th centile.24 When Karlberg and Albertsson-Wikland used two standard deviations below the mean birth length as the cut off point 7.9% of subjects who had been small for gestational age were more than two standard deviations below the mean in height at 18 years of age, and 6.4% of subjects were in this category when small for gestational age was defined in terms of birth weight.3
Several studies have reported the efficacy of growth hormone in promoting growth of short children who were born small for gestational age.25 26 However, as treatment with growth hormone may be a risk factor for insulin resistance,27 it could increase any insulin resistance present before puberty and induce glucose intolerance. Although no increase in insulin resistance was found in recent studies of glucose metabolism in short children treated with growth hormone, regardless of the cause of shortness,25 26 28 children should be monitored during treatment.
Earlier studies on increased risks of insulin resistance, hypertension, and dyslipidaemia did not have good information on gestational age and therefore could not determine the effects of premature delivery or malnutrition on the results.8 10 29 Nevertheless, these studies did show that low birth weight was associated with a higher risk of non-insulin dependent diabetes mellitus or impaired glucose tolerance.7 9 In a biethnic population in San Antonio, Texas, low birth weight was an independent risk factor for insulin resistance.10 More recently a study of 1333 men aged 50-60 who were born and living in Uppsala confirmed that reduced fetal growth is associated with insulin resistance and non-insulin dependent diabetes mellitus and suggested a specific association with thinness at birth.29 This relation seems to be mediated through insulin resistance rather than through ß cell dysfunction.29 30 31
We investigated whether these abnormalities could be detected in young adults born small for gestational age. We found that, independent of current body mass index and sex, plasma glucose concentration 30 minutes after a glucose load, serum insulin concentrations 30 and 120 minutes after a load, and proinsulin concentrations 30 and 120 minutes after a load were significantly higher in subjects who had been small for gestational age compared with the controls; in women serum fasting insulin concentrations were also significantly higher. A weak association between birth weight and plasma glucose concentration 30 minutes after a load has also been reported in children and young adults.32 33 34 Our study shows that metabolic abnormalities described in middle age, which seem to indicate insulin resistance as a responsible factor, are already present in 20 year olds born small for gestational age. Gestational age at birth influences insulin and proinsulin concentrations in subjects born small for gestational age, suggesting that a shorter length of gestation increases the association of reduced fetal growth with insulin resistance later in life.
An increased ratio of proinsulin to insulin concentrations and a decreased ratio of the 30 minute increment in insulin concentration to the 30 minute increment in glucose concentration after oral glucose load may reflect ß cell malfunction.22 23 However, no significant changes in these ratios were observed between subjects who had been small for gestational age and control subjects. People who are small for gestational age may therefore have only insulin resistance, with no defects in insulin secretion. This hypothesis is supported by the results of the insulin response to the intravenous glucose tolerance test, which did not correlate with birth weight or ponderal index at birth.29 30 31
Previous studies have shown that cardiovascular risk factors associated with insulin resistance such as raised blood pressure, high serum triglyceride concentrations, and low concentrations of high density lipoprotein cholesterol were more prevalent at 1 year old in babies born with a low birth weight or whose weight was low at 1 year.8 10 15 16 35 36 Plasma fibrinogen concentration has also been shown to be independently associated with an increased rate of ischaemic heart disease and is also linked with growth in infancy.37 However, we found no increase in risk factors for coronary heart disease in subjects who had been small for gestational age, and lipid abnormalities characteristic of the insulin resistance syndrome were not associated with low birth weight in a study carried out in Sweden.29
We found no association between being small for gestational age and blood pressure at age 20, although other studies have found such associations.38 A large population study found that systolic and diastolic blood pressure in adults and children were related to birth weight36 37 38 39 40 41 42 43 44 and that low birth weight was associated with increased blood pressure.36 37 38 39 40 41 42 43 The association was particularly strong when low birth weight occurred in conjunction with increased placental weight.41 45 The association between birth weight and blood pressure was attenuated by standardisation for gestational age both in children46 and in adults,36 and low birth weight (<2500 g) was not associated with increased systolic or diastolic blood pressure in adolescence.44 47 48 49 Adjustment for insulin concentrations 60 minutes after an intravenous glucose tolerance test reduced the significance of the associations between blood pressure and birth weight in Swedish men aged 50, suggesting that abnormalities in insulin secretion may be one of the factors mediating the association.36
Hyperinsulinaemia is an independent risk factor for ischaemic heart disease,50 and insulin resistance may occur in the absence of obesity or high blood pressure.51 Populations such as ours offer a unique opportunity for describing the natural course of hyperinsulinaemia and its outcome, and follow up should help resolve the controversy over its causes.
We found that young adults born small for gestational age had hyperinsulinaemia but not hypertension or dyslipidaemia. Whether other elements of syndrome X were present or will develop later requires further longitudinal study. Finally, insulin resistance should be documented by other more direct techniques. If it is present at a young age steps could be taken to prevent the development of non-insulin dependent diabetes—for example, by preventing obesity.
We thank all the people in Haguenau who gave us their time, and E Mairot and M C Walter for help with fieldwork. We thank P Schneegantz, J C Ongagna, and the laboratory staff at the Haguenau and Robert Debré hospitals; K Benali for entering the data on a computer; C Limoni for help in analysing the anthropometric data; and J Bouyer, P Lazar, and E Papiernick for initiating the study, preserving the records, and allowing us to use them. We also thank ERIA Diagnostics Pasteur Laboratories for help in determining insulin concentrations.
Funding: This work was supported by Pharmacia Upjohn Laboratories.
Conflict of interest: None.