The INTERBIO-21 ^st Study was conducted between February 2012 and June 2018 at seven sites: Pelotas (Brazil), Kilifi (Kenya), Nairobi (Kenya), Karachi (Pakistan), Soweto (South Africa), Mae Sot (Thailand) and Oxford (UK), of which two (Kilifi and Mae Sot) were rural and the others urban. The sites in Pelotas, Nairobi and Oxford also participated in Phase I of the INTERGROWTH-21 ^st Project ^{40– 42} .

Pelotas (Brazil): The middle-income city of Pelotas, in the southernmost region of the country where the richer Brazilian states are located, was also the Latin American site for Phase I ⁴⁰ . Pelotas is the third most populous city in the state of Rio Grande do Sul, with 350,000 inhabitants (92% living in urban areas) and 4,000 births per year. More than 99% of these births take place in the city’s four maternity hospitals. In 2007, Pelotas had a per capita gross domestic product (GDP) of R$8248 (US$4933). A total of 47% of women in Pelotas receive >9 years of formal education, with 21% receiving more than 12 years. Data from the Pelotas 2004 birth cohort study indicate that the LBW and FGR rates are 10% and 12%, respectively, and that the mean birth weight of new-borns is 3150 g ⁴³ . The city has an Epidemiology Research Centre based at the Federal University of Pelotas, which has been conducting epidemiological research on maternal and child health nutrition for more than 30 years and is also a WHO Collaborating Centre in the field of nutrition. The same research team also participated in the WHO Multicentre Growth Reference Study (MGRS), which generated the WHO Child Growth Standards ⁴⁴ .

Kilifi (Kenya): Kilifi County Hospital (KCH) is located in a rural, malaria endemic, coastal area, 55 km north of Mombasa, which is the second poorest county in Kenya. The hospital, has a catchment area of approximately 280,000 people with 3,000 births per year. The antenatal HIV prevalence is 7.9% ⁴⁵ . KCH hosts the Kenyan Medical Research Institute (KEMRI)/Wellcome Trust Research Programme, a partnership with the University of Oxford that, since 1989, has pioneered work on laboratory-based, epidemiological and clinical research. In the late 1990s, an antenatal ultrasound service was established facilitating pregnancy-related research. The research program has a computerized, Health and Demographic Surveillance System (KHDSS) that catalogues a sub-population of people living in KCH’s predominant catchment area ⁴⁶ . Each individual is given a personal identification number and births, deaths, in- and out-migration are recorded at 4-monthly household visits. The KHDSS provides a means to encourage attendance at antenatal clinics, particularly for booking early in pregnancy.

Nairobi (Kenya): The relatively wealthy Parklands suburb of Nairobi, Kenya, was the sub-Saharan site for Phase I ⁴¹ . Almost all births (>4,000) in this geographically delimited urban area, which mostly houses affluent Kenyan families, take place in three hospitals, the largest of which, The Aga Khan University Hospital (AKUH), participated in the study. AKUH is a private, not-for-profit, tertiary care institution, which is accessed predominantly by the middle and high socio‐economic sectors. These women are able to access the hospital services either through medical insurance cover or direct payment; thus, the pregnant population served is at relatively low risk of FGR. In 2008, the mean birth weight was 3101 g; the low birth weight and perinatal mortality rates were 11% and 17%, respectively. Nairobi is a non-endemic malaria area but HIV remains a significant problem in the city with a prevalence of 10% in the female population although much lower at 1% among women attending this hospital for antenatal care. Thanks to wide access to drugs and other aspects of care, during the period of this study almost all women living with HIV or newly diagnosed during antenatal screening had commenced antiretroviral therapy.

Karachi (Pakistan): The Aga Khan University Hospital (AKUH) is a philanthropic, not-for-profit teaching institution, which caters to a range of socio-economic groups through an effective patient welfare program. The hospital is the most advanced private-sector tertiary care institution in Karachi, the largest city in Pakistan with an estimated population of about 20 million people. The AKUH is also affiliated with four secondary care hospitals (combined total approximately 15,000 births per year) from which high-risk cases are referred as needed: three in Karachi and one about 100 miles away. The main AKUH has 5,800 births per year, a third of which are high-risk. The per capita income in Karachi is approximately $1,427. At AKUH, the first antenatal visit is usually in the first trimester, at which time a dating scan is performed ⁴⁷ . Nuchal translucency screening is offered to high-risk women and a fetal anomaly scan is offered between 19–23 weeks’ gestation.

Soweto (South Africa): Chris Hani Baragwanath Academic Hospital (CHBAH), which is attached to the University of the Witwatersrand, Johannesburg, South Africa, is the only government hospital serving two sub-districts in southwestern Johannesburg, which includes the Greater Soweto area, with a population of approximately 1.3 million. CHBAH provides district hospital referral services for seven community health centers in Greater Soweto, in which midwives conduct low-risk births; approximately 5% of births are tertiary referrals from other districts and provinces. The hospital has approximately 17,000 births per year, 75% of which are medium-to-high risk; this represents 53% of all births in Greater Soweto ⁴⁸ . The HIV prevalence is 29% ⁴⁹ . The population is urbanized, indigenous African, and working-class. The average annual income is substantially less than the national average, at approximately US$4,000. CHBAH is linked to the Soweto First 1000 Days Study (S1000) at the MRC/Wits Departmental Pathways for Health Research Unit (DPHRU) ⁵⁰ .

Mae Sot (Thailand): The Shoklo Malaria Research Unit (SMRU), located on the Thailand-Myanmar border, is a field station of the Mahidol-Oxford Research Unit (MORU) of Mahidol University, Bangkok. SMRU was established in 1986 as a research centre for studying the epidemiology, prevention and treatment of resistant malaria (including in pregnancy) among the 130,000 people living in refugee camps along the border. As there is no safe and effective drug for preventing Plasmodium falciparum and Plasmodium vivax in pregnancy, women are encouraged to attend the antenatal clinic at SMRU every 2 weeks from as early as possible in the first trimester, and they are systematically screened for malaria at each antenatal consultation and treated if positive ⁵¹ . In addition, since 2008, local health workers with limited education have been trained at SMRU to take fetal growth measurements with great accuracy ⁵² . Three of its field sites, with a total of approximately 2,100 births per year, participated in the study. In 2008, in newborns >28 weeks’ gestation, the mean birth weight was 2908g; the LBW and perinatal mortality rates were 15.6% and 3.4%, respectively.

Of the sites, two (Wang Pha and Mawker Thai) are clinics for migrants; the third, the Maela camp, is the largest refugee camp along the border. Both populations remain marginalized: the educational level is low and most of available income is obtained through intermittent work paid below the minimum wage and some women in the camp receive a supplement of refugee food rations. In 2012–13, HIV prevalence was low in refugees and migrants: 0.27% and 0.61%, respectively ⁵³ .

Oxford (UK): The John Radcliffe Hospital, Oxford was one of the two sites in Europe that participated in Phase I ⁴² . Oxfordshire has a population of more than 650,000 people, which includes a large proportion of young, middle‐class, well‐educated, professional families. A total of 37% of the Oxfordshire population hold a university degree, 16% higher than the national average. The hospital covers approximately 75% of more than 8,000 pregnancies that occur annually in this county. The general pregnant population served is at low risk of FGR. In 2008, the mean birth weight was 3334 g; the LBW and perinatal mortality rates were 6% and 0.5%, respectively. In addition, 99% of mothers delivering in the unit have completed secondary school or university level education. The hospital also houses the University of Oxford’s Nuffield Department of Women’s & Reproductive Health, which is where the INTERGROWTH‐21 ^st Project Coordinating Unit is located.

These sites contributed cases and controls to the INTERBIO-21 ^st Newborn Case-Control Study, selected using the definitions described below.

The INTERBIO-21 ^st Newborn Case-Control Study consisted of two components evaluating pregnancy characteristics, birth outcomes, neurodevelopment and biological markers associated with the preterm birth and SGA syndromes.

The first component aimed to compare preterm phenotypic cases to term newborn controls.

Preterm cases were singleton, naturally conceived babies, live-born at 23 ⁺⁰ to 37 ⁺⁶ weeks’ gestation ^{2, 3} , whose mothers were ≥18 years of age and resided in the hospital’s catchment area (to avoid recruiting women referred for tertiary care from another geographical region), and whose gestational age was estimated by ultrasound measurement of either crown-rump length <14 ⁺⁰ weeks’ gestation or head circumference <24 ⁺⁰ weeks’ gestation ⁵⁴ . For the planned analyses, we will stratify the cases by gestational age (defined a priori) into those born <37 ⁺⁰ weeks, and those born ≥37 ⁺⁰ but <38 ⁺⁰ weeks’ gestation, and according to the previously described phenotypes to explore interaction effects ¹¹ . Cases include groups A and C ( Table 1).

For the case/non-case analyses, controls for preterm cases were singleton, naturally conceived babies, live-born at 38 ⁺⁰ to 41 ⁺⁶ weeks’ gestation, and appropriately grown for gestational age (AGA), i.e. with a birth weight for gestational age/sex ≥10 ^th centile of the INTERGROWTH-21 ^st Newborn Size Standards ²² , whose mothers were ≥18 years of age and resided in the hospital’s catchment area, and whose gestational age was estimated by ultrasound measurement of either crown-rump length <14 ⁺⁰ weeks’ gestation or head circumference <24 ⁺⁰ weeks’ gestation ⁵⁴ . These controls include groups D and B. For group B, the sample will be down-weighted to represent their actual occurrence in the population.

A cut-off of 37 ⁺⁶ weeks instead of 36 ⁺⁶ weeks’ gestation was used to define a preterm case because of the evidence of a small but nevertheless increased risk of respiratory and other adverse neonatal outcomes (including mechanical ventilation, sepsis, hypoglycemia, NICU admission, and hospitalization for 5 days or more) in those ‘term’ babies born between 37 ⁺⁰ and 37 ⁺⁶ weeks’ gestation ⁵⁵ .

The second component aimed to compare SGA phenotypic cases to term newborn controls.

SGA cases were singleton, naturally conceived babies, live-born at 23 ⁺⁰ to 41 ⁺⁶ weeks’ gestation, with a birth weight for gestational age/sex <10 ^th centile of the INTERGROWTH-21 ^st Newborn Size Standards ²² , whose mothers were ≥18 years of age and resided in the hospital’s catchment area, and whose gestational age was estimated by ultrasound measurement of either crown-rump length <14 ⁺⁰ weeks’ gestation or head circumference <24 ⁺⁰ weeks’ gestation ⁵⁴ . Cases include groups B and C ( Table 1).

For the case/non-case analyses, controls for SGA cases were singleton, naturally conceived babies, live-born at 38 ⁺⁰ to 41 ⁺⁶ weeks’ gestation, and AGA, i.e. with a birth weight for gestational age/sex ≥10 ^th centile of the INTERGROWTH-21 ^st Newborn Size Standards ²² , whose mothers were ≥18 years of age and resided in the hospital’s catchment area, and whose gestational age was estimated by ultrasound measurement of either crown-rump length <14 ⁺⁰ weeks’ gestation or head circumference <24 ⁺⁰ weeks’ gestation ⁵⁴ . These controls include groups D and A. For group A, the sample will be down-weighted to represent their actual occurrence in the population.

For the second analytical approach—the case-base analyses—controls will include newborns from all four groups. For groups A, B and C, the samples will be down-weighted to represent their actual occurrence in the population.

Cases were recruited consecutively and one newborn control was recruited immediately after each preterm case was recruited; similarly, another newborn control was recruited immediately after each SGA case was recruited. Both sets of newborn controls will be pooled to create a control group for use in the comparative analyses with both preterm and SGA cases separately (as well as those cases born preterm and SGA), resulting in two newborn controls per case and a considerable increase in statistical power.

At all sites, trained, dedicated research staff screened all women presenting for delivery on a daily basis using a tablet (iPad, Apple, USA)-based interface with the data management system ( https://doi.org/10.5281/zenodo.1442668 ⁵⁶ ). The software (available on request), that was specially written for the study, selected the correct proportion of preterm and SGA cases and corresponding controls according to birth weight and gestational age. Thus, each newborn recruited fell into one of the four groups shown in Table 1.

The software selected a higher proportion of newborns with earlier gestational ages (for preterm cases) and lower birth weights, i.e. <3 ^rd centile (for SGA cases), using the sampling fractions shown in Table 2 so as to avoid recruiting excessive numbers just below the cut-offs that represent the majority of SGA and preterm newborns, i.e. moderate SGA and late preterm newborns. Oversampling cases at the lower end of the gestational age and birth weight distributions was important to have a large enough sample size to study the highest risk sub-groups; it was also expected to increase the statistical power of the study by producing a higher proportion of exposures and adverse neonatal outcomes.

Slight changes in the sampling fractions (see arrows in Table 2) were recommended by the study’s epidemiological advisors and introduced for preterm cases in November 2012 and for SGA cases in November 2013 to reach the recruitment rates initially planned. These changes were anticipated because the actual recruitment rate of cases was difficult to predict. The adjustments were facilitated by the tablet software.

We aimed to recruit at least 2,000 cases and 2,000 controls in total from the study sites. However, we recognised then that power calculations are a great challenge in any field-study of this magnitude and even more difficult when exploring risk factors with relatively unknown degrees of association and prevalence in such populations. The key issue is to reach a balance between logistical demands, including the need to maintain data quality in these populations, and power calculations especially for the planned genetic and epigenetic studies. In addition, when the study was designed in 2012, it was extremely difficult to provide reliable power calculations for epigenetic studies: the field was too new and very few relevant studies had been conducted. The compromise was to use experience gained from genome-wide association studies to facilitate sample size estimations. Thus, 1,500 cases and 1,500 controls (ratio 1:1) would be required, assuming a methylation proportion of 0.3 and 0.2 in cases and controls, respectively, to detect an odds ratio of 1.7 (population attributable fraction of 0.12) with a significance threshold alpha of 5.0 x 10 ^-7 and 90% power. These calculations included a continuity correction allowing for normal approximation of the binomial distribution.

The methods used to estimate gestational age, as well as the training, standardization and quality control processes are described elsewhere ^{57– 59} . In brief, crown-rump length measurements were taken <14 ⁺⁰ weeks’ gestation in a mid-sagittal view of the horizontal fetus in a neutral position, with an angle of insonation as close as possible to 90°. The image could not fill less than 30% of the monitor screen. The callipers were placed on the outer borders of the head and rump, and gestational age was estimated using the INTERGROWTH-21 ^st standards for pregnancy dating ¹⁶ .

Head measurements were taken <24 ⁺⁰ weeks’ gestation in an axial view at the level of the thalami, with an angle of insonation as close as possible to 90° using the same ultrasound machine at each site (Philips HD-9, Philips Ultrasound, USA with curvilinear abdominal transducers C5-2, C6-3, V7-3). The head had to be oval in shape, symmetrical, centrally positioned, filling at least 30% of the monitor. The midline echo (representing the falx cerebri) had to be broken anteriorly, at one-third of its length, by the cavum septum pellucidum. The thalami had to be located symmetrically on either side of the midline. The head circumference was measured using the ellipse facility on the outer border of the skull, and gestational age was estimated using the INTERGROWTH-21 ^st standards for late pregnancy dating ¹⁷ .

Femur length was measured using a longitudinal view of the fetal thigh closest to the probe and with the femur as close as possible to the horizontal plane. The angle of insonation of the ultrasound beam was approximately 90° with the full length of the bone visualised, unobscured by shadowing from adjacent bony parts and the femur had to fill at least 30% of the monitor screen. The intersection of the callipers was placed on the outer borders of the edges of the femoral diaphysis (outer to outer) ensuring clear femoral edges; ultrasound artefacts of the femoral edges such as the proximal “trochanter” or pointed femoral spurs were not included in the measurement (the detailed methodology and a graphical display of how the bone structures are localised are available at www.intergrowth21.org.uk).

The ultrasonographers at each site were selected on the basis of their technical expertise, motivation, reliability and ability to speak the local language(s). Through rigorous training they gained theoretical knowledge and familiarity with the study protocol, operations manual, data collection and quality control measures. Centralized hands-on training and initial standardization were also conducted ⁶⁰ , and the Oxford-based Ultrasound Quality Control regularly carried out site-specific standardization to ensure proper use of the ultrasound equipment, calibration and adherence to the protocol. Quality control was maintained throughout the study by taking a random 10% sample of all ultrasound images and assessing their quality using a validated scoring system ⁵⁸ .

The anthropometric measurement protocols and quality control procedures were identical to those used in Phase I ^{61, 62} . A team of anthropometrists was specially recruited, trained and standardized for the study; all training materials were based on the original WHO MGRS protocols ⁴⁴ .

In brief, newborn anthropometric measures were ideally obtained in all neonates within 12 h of delivery (and no later than 24 h), using identical equipment at all sites: electronic scale for birth weight (Seca, Hamburg, Germany) and a specially designed Harpenden infantometer (Chasmors Ltd, London, UK) for recumbent length. The equipment was selected for accuracy, precision and robustness, as reported in previous studies ⁴⁴ , and calibrated twice weekly. Head circumference was measured using a metallic non-extendable tape (Chasmors Ltd, London, UK). All lead anthropometrists were trained centrally and, in turn, trained the local anthropometrists to measure newborns according to the study protocol. The Anthropometric Standardization Unit based in Oxford regularly monitored the performance of all the anthropometrists.

The quality control measures required anthropometrists at each study site to take and record all measurements independently and compare their values with the maximum allowable differences. They also checked the forms visually after each session to ensure appropriate remeasurements were performed when necessary ⁶² .

We will use an un-weighted composite outcome including at least one of the following conditions: neonatal death until hospital discharge of the newborn, stay in NICU for ≥7 days or other severe neonatal complications, such as intraventricular hemorrhage and necrotizing enterocolitis. We have used such a composite outcome (that, when appropriate, also included stillbirths) previously ^{63, 64} ; it requires limited standardization of clinical diagnoses across hospitals and is well accepted as a marker in large, international, population-based studies of newborns that are severely ill ^{65, 66} . We believe this is a good proxy for adverse neonatal outcomes across countries.

Biological samples were collected from participants to establish a biorepository for a series of planned nutritional, biochemical, omic and histological studies. The protocols for processing and storing the samples are described briefly below. For each sample type, kits containing all the necessary supplies, including tubes pre-labelled with a unique aliquot identifier, were prepared by GAPPS (Global Alliance to Prevent Prematurity and Stillbirth, Seattle, USA) and supplied to each participating site.

Prior to sample collection beginning, the lead laboratory technician from each site was brought to Oxford for a centralized training session. Everyone was trained in collecting, processing and storing samples, as well as recording the associated data. Further training sessions were conducted on-site for the full laboratory teams every 6-12 months by the global laboratory lead to ensure adherence to the protocols and to retrain any staff if necessary.

Maternal blood: Maternal blood was collected at delivery and routinely processed within 12 h of collection. Plasma was divided into up to six 1-ml aliquots; the buffy coat was aspirated slowly using a circular motion and stored in a pre-labelled tube. After gentle inversion to ensure thorough mixing, the whole blood in the EDTA tubes was divided into 1.5 ml aliquots. After processing, all relevant information on the collection and processing of the plasma, buffy coat and whole blood specimens was recorded on the specially designed e-forms and the aliquots were stored at −80°C.

Umbilical cord blood: Cord blood was collected within 30 mins of delivery of the placenta (or whilst the placenta remained in utero). After gentle inversion to ensure thorough mixing, the whole blood in the EDTA tubes was divided into 1.5 ml aliquots before being stored at -80°C. After collection, the blood in the trace element tube was inverted to ensure thorough mixing and then left for 30 mins before processing to prevent the formation of a gel disc. It was then centrifuged for 10 mins at 1200 g to separate the plasma (top layer) and buffy coat (middle white layer), both of which were retained, from the bottom layer of red blood cells that was discarded. The plasma was stored in up to three 1 ml aliquots in Nalgene polypropylene tubes (Fisher Scientific, Leicestershire, UK); the buffy coat was aspirated slowly using a circular motion and also stored in a Nalgene polypropylene tube. These tubes were selected as they allow subsequent trace element analysis of the plasma sample. After processing, all relevant information on the collection and processing of the whole blood, plasma and buffy coat was recorded on e-forms and the aliquots were stored at -80°C.

Placenta: If processing within 1 h of delivery was possible, two placental tissue punches (~8 mm diameter x full placenta thickness) from the placental disc, avoiding the site of umbilical cord insertion and at least 3 cm from the edge of the placenta, were collected and placed in a tube preloaded with 3 ml RNAlater (Sarstedt, Nümbrecht, Germany) for future estimation of RNA. A 0.5 cm membrane strip, cut using scissors or a scalpel from the rupture site to the edge of the placental disc, was also placed in a tube preloaded with 3 ml RNAlater. These samples were stored at 4°C for a minimum of 24 h and a maximum of 4 weeks before freezing at −80°C. In an area of the placenta adjacent to the RNAlater sampling points, two tissue punches of similar size (1 cm diameter x full placenta thickness) and a small sample of membrane (1 cm wide) were frozen in liquid nitrogen or dry ice. A second membrane sample (1 cm wide) and sample of the placental disc were placed in formalin for future histology. These samples were stored at room temperature for 48–72 h after which the formalin was removed, the tissue washed with 70% ethanol and then transferred to a tube containing 4 ml 70% ethanol for long-term storage.

If the samples could not be processed within 1 h, the placenta was stored at 2–8°C for later processing. Within 12 h of delivery, samples were collected for freezing and histology (but not RNA estimation); if sample processing was possible only at >12 but <24 h of delivery, samples were collected for histology only.

Two photographs were taken of the placenta showing the whole placenta and umbilical cord with the ‘fetal’ and ‘maternal’ sides uppermost. A metric ruler was placed at right angles next to the tissue to indicate the size of the placenta. The weight of the placenta, trimmed of the cord and all its membranes, was recorded.

Maternal feces: A sample of maternal feces (approximately 5 g), if passed at delivery, was collected and stored at -80°C.

Buccal swabs: DNA was collected from the infants at 1 and 2 years of age using a buccal swab collection kit (MAWI DNA Technologies, CA 94545, USA). Up to four swabs were gently rubbed against the inside of the infant’s cheek and placed into a single vial of buffer. After collection, the swabs were discarded and the cloudy buffer containing the DNA was stored and transported at ambient temperature.

All clinical data were managed in a system very similar to the one used in Phase I of the INTERGROWTH-21 ^st Project ⁶⁷ . In brief, the data were initially collected on paper forms capturing information relating to ultrasound estimation of gestational age, pregnancy & delivery, and any fetal/neonatal abnormalities; these forms were securely stored. The data were then entered at the local level into an on-line data management system, based on the one developed specifically for the INTERGROWTH-21 ^st Project (MedSciNet, London, UK). This on-line system, which resides on a secure MedSciNet server, facilitated quality control, correction of errors or missing values, and the initiation of data analysis soon after completion of data collection. A review process within the system, which involved weekly queries to each site via Skype if necessary, ensured that all key data were complete. Blinded data from the ultrasound machines were transferred directly to the database in Oxford.

All sample-related data were collected on an electronic form (e-form) and the data were uploaded onto a separate data management system (Sapphire, Labvantage Solutions Ltd, High Wycombe, UK) that was specifically modified for the study. This system, which resides on a secure University server, allows samples to be tracked from the time of collection through processing, storage at the study sites, and transport to the central storage facility in Oxford. Each participant was given a unique identifier number, which was used to link the clinical and sample databases. Individual aliquots of each sample type were also given a unique number.

All the electronically stored data were stripped of personal identifiers, which are held separately and securely on site. The anonymised databases are only accessible to designated personnel, including the Bill & Melinda Gates Foundation as part of a data sharing agreement. Users from each study site can only view their own data at present and a limited number of global administrators can see all the data on a secure server.

These systems provided the Data Management Unit in Oxford with a detailed daily record of patient enrolment and data entry, at both individual and institutional levels, to monitor progress. Corresponding actions, such as telephone calls, web conferences and site visits took place within a week of detecting a problem at a study site to ensure that appropriate corrective measures were taken.

Body composition was estimated within 96 h of birth at two study sites, Mae Sot and Oxford, using air displacement plethysmography (ADP) (PEA POD®, COSMED, Rome, Italy), which derives the proportion of fat mass (FM) and fat-free mass (FFM) by measuring body volume (air displacement) and weight. ADP has considerably improved our understanding of newborn body composition and the implications of feeding regimens especially for preterm and growth restricted newborns.

This non-invasive technique is rapid, reliable, robust to moderate levels of infant activity, and acceptable to parents ⁶⁸ . Using the same PEA POD methodology, we have previously reported normative data at term following the prescriptive approach, plus differential FM, body fat percentage and FFM patterns for babies born preterm (34 ⁺⁰ to 36 ⁺⁶ weeks’ gestation) and with impaired fetal growth ²⁴ .

The PEA POD, which is designed for use in infants up to 6 months of age or 8 kg in weight, was routinely calibrated and used according to the manufacturer’s instructions in a temperature-controlled room. The newborn baby was evaluated undressed in the test chamber for 2 min and, if necessary, duplicates of irremovable items (clamps, tubes or tags) were measured before the examination.

This ancillary study tests the hypothesis that pregnancy-related etiological factors associated with preterm birth and SGA produce differential body composition proportions that can be evaluated in the neonatal period. It is expected to contribute greatly to the improved phenotypic characterization of these complex syndromes.

At 1 and 2 years of age, anthropometric measurements (weight, length, head circumference, mid-upper arm circumference, triceps skinfold thickness and subscapular skinfold thickness) were taken, and data on the infant’s general health, diet and motor development skills collected.

In addition, neurodevelopment is being assessed at 2 years of age using the INTERGROWTH-21 ^st Neurodevelopment Package ²⁷ . This is a multi-dimensional instrument for early child development ideally suited for both research and screening purposes in field studies and large populations. It was designed to be implemented by non-specialist assessors at individual and small group levels across varied socio-economic and multi-cultural settings. In brief, the Package assesses:

These tools were used to assess the eligible 2-year old children in Phase I of the INTERGROWTH-21 ^st Project ²⁹ . Similarities across study sites were measured using variance components analysis and standardised site differences (SSD). In 14 of the 16 domains, the percentage of the total variance explained by between-site differences ranged from 1.3% (cognitive score) to 9.2% (behavior score); 8% or less for the visual and motor items, and <9% for WHO milestones. Of the 80 SSDs comparisons, only six were > ±0.50 units of the pooled SD for the corresponding item. These data demonstrate that the children of healthy, adequately nourished, well-educated pregnant women, recruited from five diverse geographical and cultural study sites, who receive recommended antenatal care, display consistent similarities at 2 years of age across a comprehensive set of neurodevelopmental outcomes. The corresponding normative values have been produced for the evaluation of the INTERBIO-21 ^st populations and we would like to continue following these children until at least the age of 5 years.