EarlyBird Diabetes Study


Recruitment : EarlyBird is a random sample of the 1995-6 Plymouth birth cohort. All Plymouth primary schools were identified and their head teachers asked for agreement to participate in the study. The 54 schools that consented were stratified into quartiles according to their proportion of free school meals as a socio-economic proxy, and a random selection from each made accordingly. Registration into the study as family units comprising the child and his/her parents was invited during parent induction meetings for school entry and parents expressing interest were given a full written explanation. The parents were included to make up trios for three reasons. They provide the closest approximation available to the child’s metabolic status a generation hence, and they permit the separate analysis of father/offspring and mother/offspring analysis of inherited features. Phenotypic characteristics passing between father and offspring are more likely to prove genetic, whereas the same characteristic shared by mother and offspring might equally be a gestational (intergenerational) effect. The inclusion of parents is also critical to retention. Exclusion criteria in the children were existing diabetes, pathological states likely to affect growth or body composition, moderate or severe physical disability, long term use of oral steroids, and in the parents diabetes, chronic illness requiring medication, long-term use of corticosteroids, pregnancy, lactation and use of the oral contraceptive. Following ethical approval and a parent’s written consent, a total of 307 children (137 girls, 170 boys, mean age 4.9 years) who started school between January 2000 to January 2001 became the EarlyBird cohort. 

Procedure : The children are reviewed fasting from 8 am by a research nurse within a hospital department of child health. Anthropometric measures and blood pressure are repeated at six-monthly intervals and other tests every 12 months. A limited number of anthropometric measurements were made on the parents at base-line, and a single blood sample taken for DNA, insulin resistance and markers of the metabolic syndrome. All data are anonymised and archived on optical disc. 

1. Candidate factors affecting insulin resistance 

  • Anthropometry : Height is measured to the nearest 1mm (Leicester Height Measure), weight to the nearest 0.2 kg, skinfold thickness at five sites (triceps, biceps, subscapular, suprailiac and para-umbilical) by Holtain skinfold calipers, mid-arm circumference, waist and hip by tape measure. A minimum of three ‘blind’ repeats are made of each anthropometric measure at each visit. Ultrasound was used at baseline to measure the intra-abdominal sagittal diameter between aorta and anterior abdominal wall as a proxy for visceral fat mass, but found to be insufficiently precise.
  • Body composition : Regional fat mass, %fat, fat-free mass, water and bone are deduced from dual energy X-ray absorptiometry (DPX, Lunar Corporation, Maddison, USA ), and the total body measures also by bioimpedance (Tanita Corporation, Japan). Body composition gives a more precise definition of true fat mass and, importantly, of its distribution in the body.
  • Physical activity : Daytime physical activity at school, after school and at weekends is measured continuously over seven-day periods by means of a CSA piezo-electric accelerometer (MTI, Fort Walton Beach, FL). The monitor is fitted before the child leaves the department and retrieved from school one week later. The data are downloaded onto a PC for storage and analysis.
  • Resting Energy Expenditure : Resting metabolic rate is measured in the lying position by indirect calorimetry using a GEM (Nutrem, Manchester) over a period of 15 minutes, during which time the child can choose to listen to a story or watch a video.
  • Food frequency : Standardised food frequency questionnaires, widely used by dieticians, are answered by the parent.
  • Genetics and demography : Two ml aliquots from each serum sample are archived at –80 o C after each visit, and blood cells retained from parents and children for DNA analysis of candidate genes. Parents completed a baseline questionnaire detailing medical history and socio-economic circumstances (income band, educational achievement, occupation, post-code and eligibility for free school meals) which is updated every six months. 
  • Heart rate variation : Beat-to beat variation in heart rate is increasingly recognized as a measure of autonomic tone, itself a determinant of insulin resistance. A two-lead rhythm-strip ECG is attached to the children while undergoing REE measurement, and the beat variation pattern analysed by specialist software.

Accelerometers provide extraordinarily precise information, and are re-writing the literature on the physical activity of children. They sample movement 600 times a minute, register the clock time, duration and intensity of each movement, and can generate a detailed ‘day in the life of x’ graph for any child.

2. Behaviour of Insulin resistance and insulin secretory capacity:

Insulin resistance and insulin secretory capacity are derived from fasting measures of insulin and glucose in venous blood samples, using Homeostasis Model Assessment – HOMA. The earlier of two reports describing HOMA provides formulae for the calculation of insulin resistance and beta cell function, the later update supplies a software programme. We have compared the log insulin resistance from the EarlyBird cohort calculated according to formula with the log insulin sensitivity according to the programme separately in the children, mothers and fathers, and found them to correlate in each case (r= -0.993) There was major deviation, however, in the case of beta cell function and we have used the programme when analysing insulin secretory capacity. 

3. Metabolic Impact

Blood pressure (considered as a metabolic response to insulin resistance) is taken by semi-automated sphygmomanometer (Welch-Allyn, Beaverton, OR) and the mean of the second and third of three recordings used in the analysis. Topical analgesic cream (Emla, Astra-Zeneca) is applied one hour before venepuncture. Insulin and sex hormone-binding globulin (SHBG) are measured by immunometric assay on a DPC Immulite analyser, using kits manufactured by Diagnostic Products Corporation (Los Angeles). Insulin cross-reactivity with proinsulin is less than 1%. Glucose, cholesterol, HDL cholesterol, triglycerides and uric acid are measured on a Cobas Integra 700 analyser (Roche Diagnostics, Lewes, East Sussex, UK). A full haematological profile is recorded. Glycated haemoglobin is measured by automated high performance liquid chromatography using a Menarini Biomen HA 8140 analyser. Follicle-stimulating hormone (FSH) and luteinising hormone (LH) are measured by automated chemiluminescent sandwich immunoassay on an Advia Centaur analyser (Bayer Diagnostics, Newbury, Berkshire, UK). Serum IGF1 (microELISA, Univ Glasgow), Tanner staging and peak height velocity will be used to monitor the progress of puberty.

Adiponectin (microELISA, Univ Glasgow) is a recently discovered mediator of insulin action and regulator of fat mass. Leptin (microELISA, Univ Glasgow) is produced by fat cells, an indicator of fat mass and a satiety signal. C-reactive peptide (hcCRP, microELISA, Univ Glasgow) is a marker of inflammation. Insulin resistance is fundamentally a state of low grade inflammation.

Quality Control, Validation and Compliance : We have formally assessed the precision of all anthropometric measures. The performance of the accelerometers has been rigorously investigated by ourselves and others. We have also established 12-month test-retest correlations on the full EarlyBird cohort for the accelerometers measuring physical activity (r= 0.49), GEM measuring REE (r= 0.49) and HOMA measuring insulin resistance (r= 0.44). HOMA is now well established in the estimation of insulin resistance in children. The euglycaemic clamp, gold standard for the measurement of insulin resistance, is both impractical and unethical in healthy young children. HOMA compares favourably, with correlation coefficients exceeding 0.8.  The special concerns for clinical investigation in small children are recognised. No investigation is considered unduly invasive, and none is performed without the child’s assent. To date, the Study can claim over 95% compliance on all tests. Every effort is made to minimise attrition by close family support, group activities, newsletters, parties and so on. Parents are assured that all data will be coded and anonymised, and that any abnormal results will always be reported to their general practitioner.

Main Outcome Measures

  • Children :  1. Candidate factors : Birth weight, height, weight, BMI, skinfolds at five sites, waist, mid-arm and hip circumference, body composition, blood pressure, resting energy expenditure, physical activity and diet.

    2. Insulin resistance and secretory capacity : HOMA-IR and HOMA-ISC. 

    3. Metabolic impact : Blood pressure, full blood count, haemoglobin and haematocrit, HbA1C, total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, uric acid, IGF-1, gonadotrophins and SHBG. DNA is prepared and serum aliquots from each visit archived at -80°C. 

  • Parents
    1. candidate factors: Height, weight, BMI, waist circumference. 
    2. insulin resistance and insulin secretory capacity : HOMA-IR and HOMA-ISC. 
    3. metabolic impact: Full blood count, haemoglobin and haematocrit, HbA1C, total cholesterol, HDL cholesterol, LDL cholesterol, triglycerides, uric acid, gonadotrophins, and SHBG. DNA is prepared and serum aliquots archived at -80°C.

Statistical Analysis : All statistical analyses are performed using SPSS for windows version 10.1. Standard deviation scores (SDS) are calculated for birth weight, current weight, height and BMI in order to standardise for age and sex (based on 1990 UK reference curves) using the conversion programme issued from the Child Growth Foundation (London, W4 1PW, UK). T-tests or one-way analysis of variance (ANOVA) compare means, and correlations are established by Pearson’s coefficient. Chi-Square is used for the analysis of categorical variables. A sample of size 300 will detect a correlation of 0.23 or more, assuming a two-tailed test at the 5% significance level. When considering males or females only, a sample size of 150 will detect a correlation coefficient of 0.27.

With accumulation of data (EarlyBird now has five annual time-points to analyse), and increasing use is being made of trend analysis.