The Project on Preterm and Small-for gestational-age infants (POPS) study is a nationwide multicentre prospective follow-up study, which comprises 94% of all liveborn very preterm (<32 weeks of gestation) and/or very low birthweight (<1,500 g) infants born in the Netherlands in 1983, and has documented birth, growth and a number of other characteristics from birth onwards [12, 13]. At follow-up visits at age 3 months and 1 year post-term, weight was recorded. At age 19 years, all 637 alive subjects born with a gestational age below 32 weeks who were free from congenital skeletal deformations, Down’s syndrome, chromosomal abnormalities, multiple congenital deformations or inborn errors of metabolism, and who were not born to mothers with gestational diabetes, were approached by mail to participate in the POPS-19 study. Subjects with diabetes mellitus, or on thyroid hormone or systemic corticosteroids, as well as pregnant women, were excluded. The approval of the medical ethical committees of all participating centres was obtained for the POPS-19 study.
Subjects, who gave written informed consent to participate, were seen after an overnight fast between 0830 and 1000 hours between April 2002 and May 2003 at one of the outpatient clinics of the ten participating centres. Assessors were blinded with respect to the perinatal characteristics of the subjects.
Venous blood was obtained after 30 min in a supine position. Thereafter, anthropometry was performed, for which assessors had received extensive training prior to the study, and re-training during the entire study period at 2-month intervals. Subjects were measured barefoot while wearing underclothing only. Weight was measured to the nearest 0.1 kg on a balance scale, and height to the nearest 0.1 cm with a fixed stadiometer. Waist and hip circumferences were measured at 0.1-cm accuracy using standard methods . Four skinfold thickness measurements were taken in duplicate with a calibrated skinfold calliper on the left side of the body: at triceps, biceps, subscapular and iliacal regions. From these measurements, fat mass and the corresponding fat-free mass were calculated using the equations of Durnin and Rahaman . A more detailed description of skinfold thickness measurements obtained in the POPS-19 study has been published elsewhere .
Blood samples were stored at −80°C, and thawed only once immediately before analysis. Glucose was measured in a fully automated computerised laboratory system with an Hitachi 747 (Hitachi, Tokyo, Japan) chemistry analyser, and insulin and C-peptide were measured with highly sensitive RIAs (Linco, St Charles, MO, USA; detection levels 0.1 mU/l and 0.03 mmol/l, respectively; interassay CV 4.7–12.2% and 3.2–9.3% at different levels, respectively). A homeostatic model assessment for insulin resistance index (HOMA-IR) was calculated . Insulin and C-peptide levels, and HOMA-IR were considered as parameters of insulin resistance. Insulin level and HOMA-IR correlate strongly with S
i assessed by the frequently sampled IVGTT in young persons [18, 19].
Auxological data at birth and on subsequent occasions were converted to standard deviation scores (SDSs), to correct for (gestational) age and sex, using Swedish references for preterm infants , and recently collected Dutch references [14, 21, 22], respectively.
Results in Tables 1 and 2 are presented as means±SD, or medians (interquartile range) if variables were not normally distributed (insulin and HOMA-IR). These variables were log-transformed before statistical comparison.
Birthweight is a strong predictor of postnatal size. Therefore, the multivariate linear regression model developed by Li et al. was used to distinguish between the separate effects of birthweight SDS, and of postnatal size (at the age of 3 months, 1 year and 19 years) on the parameters of insulin resistance . First, the effect of birthweight SDS on the parameters of insulin resistance was studied. Subsequently, residual (observed-expected) postnatal size was entered into the model. Expected postnatal size was based upon birthweight SDS only. Hence, residual postnatal size can be interpreted as growing more (or less) than would be expected from a given birthweight SDS. Thereafter, the interaction term (birthweight SDS×residual, with subtraction of means) was entered. Recently , the algebraic concept of this model was explained, showing that it can be rewritten to the model by Lucas et al.  In the applied model, postnatal size is made statistically unrelated to birthweight SDS.
Analyses with birthweight SDS and infancy weight gain were repeated with adjustment for the possible confounders sex, race (white or non-white), socio-economic status (≤ or >2), multiple pregnancy (singleton or non-singleton), gestational age (≤ or >30 weeks), parity (0 or >0) and hypertension during pregnancy (yes or no). Analyses with adult size and body composition were repeated with adjustment for sex, race and socio-economic status.
Statistical significance was defined as a p value ≤0.05. Non-linear associations were tested by first producing quarters of birthweight SDS and infancy weight gain. These quarters were compared with respect to the parameters of insulin resistance.