The study of women, infant feeding and type 2 diabetes after GDM pregnancy and growth of their offspring (SWIFT Offspring study): prospective design, methodology and baseline characteristics

Study participants and setting

The SWIFT Offspring Study cohort is an observational study of 466 mother-infant pairs
recruited from among women with recent GDM who were participants in the SWIFT Study
77]. The parent SWIFT Study enrolled 1,035 women diagnosed with GDM by the Carpenter
and Coustan criteria during pregnancy and who delivered a singleton, pregnancy of
?35 weeks gestation at a Kaiser Permanente Northern California (KPNC) hospital from
August 2008 through December 2011. The study design and recruitment protocol for the
SWIFT Study have been described in detail elsewhere 77].

A description of the coordinated in-person exams (E) for the parent SWIFT Study and
the in-person infant exams (IE) and additional maternal exams (ME) for the SWIFT Offspring
Study are shown in Fig. 1.

Fig. 1. Diagram of the Parent SWIFT Study and SWIFT Offspring Studies Coordinated Data Collectin
Timelines for Women and their Offspring

From August 2009 to December 2011, the SWIFT Offspring Study enrolled 466 mother-infant
pairs after mothers provided written informed consent for their own and the infants’
participation in three in-person exams conducted at 6–9 weeks (baseline), 6 months
and 12 months of age (Fig. 1). The study employed standardized protocols and research quality calibrated instruments
to measure infant head circumference, weight, supine length, abdominal circumference
and three skinfold thicknesses. Mothers completed questionnaires that gathered data
on breastfeeding and formula feeding, and infant health, development, supplemental
dietary intake, sleep habits, temperament and other behaviors, as well as reported
paternal weight and height, reproductive and health history, inter-current pregnancies,
perinatal outcomes, postpartum depression and lifestyle behaviors (i.e., smoking,
sleep habits, physical activity and dietary intake) during the postpartum period.
Saliva specimens were also collected from infants.

Recruitment and eligibility criteria – the parent SWIFT study

SWIFT participants were recruited from 12 Kaiser Permanente Northern California (KPNC)
medical centers and medical office facilities in 2008–2011 throughout the 5,000 square
mile KPNC region. The SWIFT Offspring study recruitment and follow-up began in August
2009 and follow up continued through December 2013. Eligibility criteria were as follows:
a healthy, live born singleton infant ?35 week’s gestation and birth weight of ?2,500
g born to mothers enrolled in the SWIFT study. Infants with one or more serious medical
condition(s) (e.g., failure to thrive, physical impairment affecting feeding ability,
chronic infectious disease, severe jaundice, or other metabolic disorder) were excluded.
Mothers provided written informed consent for their infants after being fully informed
of all aspects of the study, including potential risks and benefits, and were given
an additional incentive for their child’s participation at each study exam. Characteristics
of the SWIFT Offspring study of mother-infant pairs (n?=?466) are displayed in Table 1.

Table 1. Maternal and offspring (n?=?466 mother-infant Pairs) characteristics (enrolled 2009–2011)

Sample size and power analysis

Minimum detectable differences in means (either at exam or change between exams) of
a continuous variable (e.g., change in weight-for-age z-score, skinfold thickness,
waist girth) are expressed in standard deviation units, in analyses of offspring and
mothers at 6–9 weeks (offspring only, Aim 1), at 6 months and at 12 months (Table 2). These estimates are based on the standard t-test for differences in means, in a
comparison of breastfeeding (66.7 %) vs. none/minimal (33.3 %). In Aim 1, for analyses
of infants, we have sufficient power (0.80) to detect a difference between lactation
groups in mean change in weight for age z-scores of at least 0.28 standard deviation
(s.d.) units at 6–9 weeks and 0.31 s.d. units at 6 months. Given an expected standard
deviation of z-score of 1.1 49], the minimum detectable difference in mean z-scores translates to 0.30 and 0.34 at
6–9 weeks and at 6 months, respectively. In Aim 2, for analyses of infant regional
adiposity variables at 12 months, the minimal detectable difference in means across
lactation categories is 0.29 s.d. units (e.g., 0.20 weight-for-length z-score, 0.86 cm
in abdominal circumference, 0.70 mm subscapular, and 2.9 mm sum skinfolds, based on
infant data in GDM offspring 29], 30]). In Aim 3, analyses of adiposity in mothers at 6 months and 12 months postpartum,
we have sufficient power (0.80) to detect a difference in means of 0.31 s.d. units
at 6 months, and 0.29 s.d. units at 12 months (e.g., differences of 2.0 kg/m 2] for BMI, 4.2 cm for waist girth at 6 months, and 1.85 kg/m 2] for BMI, 3.9 cm for waist girth at 12 months). In summary, detectable effect sizes
are relatively small, and clinically important.

Table 2. Minimum detectable differences in means (standard deviation units, s.d.) for comparison
of Intensive (exclusive or mostly) breastfeeding vs. Intensive (exclusive or mixed)
formula infant feeding (two-sided test, significance level?=?0.05, power?=?0.80).
[Total number of mother-infant pairs enrolled with one or more study exams; n?=?466]

Data collection and methodology

Table 3 displays a summary of data collection for the SWIFT Offspring study, including variables
collected by the parent SWIFT study and the Offspring study. Supplemental infant food
intake, physical activity, sleep habits, temperament, infant health and other infant
behaviors and characteristics were also assessed via interviews of mothers at study
exams.

Table 3. SWIFT and SWIFT Offspring studies: description of data elements and timeline

The SWIFT study data elements, (women with GDM)

SWIFT participants (mothers delivered of GDM pregnancies) attended three in-person
visits (6–9 weeks, 12 months and 24 months) at which time trained research staff obtained
anthropometric and body composition measurements, administered surveys to collect
behaviors, socio-demographics, and reproductive history, and conducted the 2-h 75
g oral glucose tolerance test (OGTT). Mothers kept diaries and were queried in detail
about infant feeding practices (breastfeeding intensity and duration, formula feeding),
and infant dietary intake (Table 3). Data collection occurred during telephone interviews in late pregnancy, and at
4–5 weeks and at 6 months post-delivery, as well as monthly mailed surveys from 3–11
months post-delivery and from health plan electronic medical records (EMR) (Fig. 1).

The SWIFT offspring study (mother-infant pairs), data elements

Study protocol at each exam included anthropometric measurements, and both self-and-interviewer-administered
questionnaires. Electronic medical records systems provided key information on perinatal
characteristics and pediatric clinical health outcomes. We obtained clinical data
on newborn birth weight, length, gestational age, Apgar scores, neonatal intensive
care unit (NICU) admissions and health conditions, as well as maternal prenatal 3-h
oral glucose tolerance test (OGTT) results, weight at delivery, type of GDM treatment,
severity of GDM, and pre-pregnancy weight from health plan EMR. Bilingual English-Spanish
research assistants administered the Spanish versions of the consent form and each
questionnaire to participants whose preferred language was Spanish. Additionally,
validated parent-report measures of infant temperament were collected, as growing
evidence supports the putative association between a ‘difficult’ infant temperament
and later child obesity 78], 79]. In fact, parents may bottle feed or television to assuage a difficult temperament
or negative emotionality 80]–83], although this issue has never been studied in the context of GDM offspring.

Study variables

We present the methodology for measurements of the study variables below including
primary outcome measure (i.e., change in infant weight-for-length z-score,WLZ); primary
independent variable (i.e., breastfeeding intensity and duration); and intrauterine
exposures, sociodemographics and lifestyle covariates. Our selection of covariates
was based in part on early life and behavioral risk factors examined in prior studies
6], 74], 84].

Infant growth – weight-for-length Z-scores

At in-person exams, trained research staff obtained anthropometric measurements (weight,
length, health circumference and abdominal circumference) at 6–9 weeks, 6 months and
12 months of age using the WHO Multi-center Growth Reference Study standardized procedures
85]. Weight and length were used to calculate the z-scores to evaluate growth compared
to the WHO standard referent population which is based on the growth of healthy breastfed
infants and young children raised in environments that do not constrain growth 86]. For children under 24 months of age, use of the 2006 WHO international growth charts
is recommended by the Centers for Disease Control and Prevention (CDC) 87], 88]. The primary measure of infant growth is defined as change in infant weight-for-length
z-score (WLZ) from birth to 6–9 weeks, 2 to 6 months, and 6 to 12 months of age based
on the World Health Organization (WHO) growth charts 87]. We also calculated weight-for-age (WAZ) and length-for-age (LAZ) z-scores to evaluate
infant growth at each age.

Infant feeding assessments

Breast milk and formula feeding intensity and duration

Infant feeding practices, including breastfeeding intensity and duration, were assessed
prospectively from birth by asking mothers to record amount of formula fed using a
weekly diary, and during a telephone interview conducted at 4–5 weeks of age to determine
study eligibility. Research staff administered questionnaires to assess infant feeding
practices since birth throughout the first year of life during the three in-person
study exams, and mothers returned monthly questionnaires by mail. Mothers were asked
to report whether they had ever breastfed their child, and if they were currently
breastfeeding. If they were not currently breastfeeding, women were asked to specify
the date and/or child’s age as well as the reasons for discontinuing breastfeeding.
They also reported the frequency of breastfeeding and formula feeding including the
number of expressed breast milk feedings by bottle, and formula supplementation (quantity)
per 24 h within the past 7 days during each month. Details included the frequency
of day and night feedings, frequency of breast milk expression and feeding by bottle,
provision of other liquids (water, tea, etc.), use of formula and quantity, type and
brand used, use of cereal, sweeteners, other liquids (i.e., Pedialyte), juices, sugar-sweetened
water and beverages, and timing of the introduction of solid foods and liquids and
the quantities.

At study baseline (6–9 weeks of age), five breastfeeding intensity groups were devised
based on cumulative intake since birth: 1) exclusive breastfeeding (0 ounces of formula,
and no other liquids); 2) mostly breastfeeding (formula ?6 ounces of per 24 h; 3)
Mixed breast milk and formula (?7–17 ounces of formula per 24 h) or inconsistent pattern
of feeding; 4) mostly formula feeding and some breast milk, (17 ounces per 24 h);
and 5) exclusive formula feeding (no breast milk and less than 3 weeks) 89]. We combined the mixed/inconsistent group with the mostly formula feeding group to
configure four infant feeding groups at study baseline.

A combined measure of breastfeeding intensity and duration from birth through one
year of age was configured based on the prospectively collected data from the in-person
exams, telephone interviews, and monthly questionnaires collected for the parent SWIFT
Study. We used the methodology of Piper et al. 90] to calculate a lactation intensity ratio summary score from birth to 12 months. For
each month (months 1–12), we calculated a lactation intensity ratio (LIR) based on
the number of breastfeedings and the amount and number of milk feedings during an
average 24-h period for the previous 7 days (1 week) reported by women. The 24-h recall
period is the best method to assess breastfeeding practices because it has greater
validity 91] and it minimizes recall error as compared to longer recall periods.

We also collected information on infant dietary intake of supplemental foods/liquids
(type, amount, frequency). The LIR was developed for the 1988 National Maternal-Infant
Health Survey 90] as an intensity ratio, calculated from the number of breast milk feeds (on average
in 24 h) divided by the total number of all milk feeds (on average in 24 h), within
a range from 0 to 1.0. Exclusive breastfeeding for a given month received the highest
score of 1.0. Exclusive formula feeding received a score of 0 for all time periods.
Partial breast-feeding for any month received an LIR of less than 1.0. The scores
for lactation were summed across all months until weaning, and an “overall” lactation
score was calculated based on the sum of the LIRs for each month over their entire
duration of breastfeeding.

The LIR for each month and overall LIR summary score were calculated as follows:

Supplemental dietary intake

Monthly mailed questionnaires asked mothers to report the types of foods and liquids
fed to the infant each month from 3 to 11 months of age. At the three in-person exams
at 6–9 weeks, 6 months and 12 months of age, research assistants asked mothers to
recall the average daily frequency that infants were fed fruit juices, sweetened juices
and drinks, sugar added to water or tea, and any cereal in the bottle or other foods
fed to the child in the past 7 days. At 6 months of age, a comprehensive dietary history
was administered to characterize the dietary intake during the past two weeks, and
to assess the transition to supplemental foods and other liquids for each month from
birth to 6 months. Mothers reported the types and amounts of foods and liquids consumed,
the ages when these were first introduced, and the average dietary history for each
month through age 6 months. When the infant reached 12 months of age, mothers also
completed a survey about the types of milk, including artificial formula, cow’s milk,
soy milk or other milk sources and quantity of milk consumed by the child per 24 h.

Anthropometric assessments

Infant measurements

Anthropometric measurements for infants were obtained at 6–9 weeks, 6 months and 12 months
of age using the procedures developed for the WHO Multicenter Growth Reference Study
86]. Prior to data collection, the research staff were properly trained and certified
to follow the WHO standardized procedures for anthropometric measurements 92]. At each in-person exam, trained research assistants measured the infant’s head circumference,
abdominal circumference, skinfold thicknesses (at the triceps, suprailiac and subscapular
sites), weight and supine length via standardized protocols. Two measurements were
obtained during each procedure, with a third measurement obtained if the difference
between the first two measurements was greater than 1 cm for length, 0.1 kg for weight,
1 cm for abdominal circumference, and 0.1 mm for skinfold thickness. Measurements
of infant weight and length were used to calculate the weight-for-length, weight-for-age,
and length-for-age z-scores, and change in z-scores using the WHO Growth Standards
(http://www.who.int/childgrowth/en/) as the referent population 92].

Weight was measured on a digital scale (Tanita, Model BD590 infant scale) that calibrates
to zero and is accurate to the nearest 5 g. Length was measured to the nearest 0.1 cm
using an infantometer (Seca, Model 417 infantometer). Infant head circumference was
measured using the Abbot Nutrition of Abbot Laboratories measuring tape. Abdominal
circumference was measured to the nearest 1 mm using a tape measure (Gulick Model
67020, ¼ inch measuring tape) made of material that does not stretch and the time
of the last feeding episode was recorded. Time of last feeding was recorded for each
infant before obtaining the abdominal circumference measurement. Skinfold thicknesses
were measured to the nearest 0.2 mm using a Holtain Tanner/Whitehouse Skinfold Calipers
which are calibrated at 0 mm at each exam to assess regional adiposity. The sum of
the three skinfold measurements was used to evaluate overall body adiposity.

Maternal measurements

Maternal body weight and waist circumference were measured at 6–9 weeks, 6 months
and 12 months postpartum. Women were weighed on a research quality, calibrated digital
scale (Tanita, Model WB110A, 100A) to the nearest 0.1 lb in light clothing and were
asked to empty pockets, and remove any heavy jewelry and shoes. Height was measured
in bare or stocking feet to the nearest centimeter using a stadiometer (Seca, Model
69072) to the nearest 0.1 in.. Body mass index (BMI) was calculated as weight (kg)
divided by height (m) squared and used to evaluate overall adiposity. Waist circumference
(waist girth) was measured on a bare abdomen in triplicate to the nearest centimeter
at the level of the right ischium using a Gulick II Plus 300 cm anthropometric tape (Model 67019). The Gulick II Plus tape has a tensioning device attached to the measuring tape that provides a standard
amount of tension (4 ounces) while a measurement is being taken.

Infant health and behavioral assessments

Infant health status

The survey includes questions about the child’s health conditions, current medication
use, allergic reactions, hospitalizations since birth, number of teeth the baby currently
has, and any serious long-term medical conditions.

Infant sleep habits

Infant sleep habits were assessed by a questionnaire that gathered details on where
the child sleeps, the position of the child when sleeping, duration of nighttime and
daytime sleep (including naps), number of night awakenings per night, longest period
of sleep without waking, duration of wakefulness during the night (between 10 p.m.
and 6 a.m.), length of time needed to put child to sleep, how the child falls asleep
at night (e.g., while feeding, being rocked, etc.), usual sleep and waking time each
day, and whether the child’s sleep is considered a problem. The questionnaire was
developed in collaboration with Dr. Kathryn Lee 93], 94].

Infant development and sedentary activity

A brief questionnaire asked mothers to report the ages when the child could first
roll over without assistance, was able to sit up without support, began to crawl and
walk without support, and how many teeth developed. Mothers also reported how much
time their child spent watching television or videos on a weekly basis.

Infant behaviors/temperament questionnaires

Rothbart scales

The Rothbart Infant Behavior Questionnaire (IBQ-R) contains 184 items that measures
14 dimensions of infant temperament, including ‘soothability’ and ‘distress to limitations.’
Additionally, a 15-item subscale assesses the child’s gross motor activity, including
movement of arms and legs and squirming and locomotor activity. Objective measures
of physical activity in infants have not been validated, but qualitative proxy measures
of physical activity, such as temperament scales, have been linked with obesity in
young children. The IBQ-R was administered at 6–9 weeks and at 6 months of age 95].

Goldsmith scale

We utilized the scale to assess child behavior at one year of age. The Toddler Behavior
Assessment Questionnaire (TBAQ) contains 108 items that measure six scales of parent-reported
temperament-related behavior in 16–36 month old children. A 26-item abbreviated version
of the TBAQ was administered at 12 months of age in this study 96]. The scales include activity level, anger, fear, pleasure, and interest. The scale
has been validated and internal consistency reliability exceeds .80 for each scale
97].

Other risk factors assessments

Using self-and interviewer-administered questionnaires at in-person exams, telephone
contacts and monthly mailed brief surveys, we collected data on numerous risk factors,
including intrauterine exposures. Women reported family history of diabetes, previous
GDM diagnosis, treatment for GDM, and other perinatal complications, newborn outcomes,
pre-pregnancy weight, current medical conditions, medication use, pregnancies after
enrollment (inter-current pregnancies), and contraception methods, including hormonal
contraceptive use.

These questionnaires assessed maternal socio-demographics, medical history, alcohol
consumption, smoking, and postpartum depression. Clinical risk factors, including
the severity of glucose intolerance during pregnancy utilizing the 3-h 100 g OGTT,
were also collected through a variety of methods including the health plan electronic
medical records, study phone interviews, in-person surveys and monthly study mailings.

Other data collection procedures

To streamline the in-person exams, a subset of questionnaires were mailed to mothers
at least a week prior to their 6- and 12-month exams and participants submitted the
completed questionnaires to research assistants at exams. If the mailed questionnaires
were incomplete or not received, they were completed during the in-person exam. For
quality control purposes, the questionnaires received by mail were reviewed by research
staff for completion and accuracy.

Quality control procedures

Detailed study operations manuals were developed to standardize the data collection
procedures across the study sites. Research staff completed trainings led by the Project
Manager which included a series of shadowed study activities for each data collector
that were evaluated before the person could begin performing any study activity. The
training phase included observation of the staff while conducting their first few
measurements with actual study participants. Throughout the study period, refresher
trainings involving all data collectors were conducted twice per year.

Biospecimen collection procedures

Saliva samples were collected from the infants at 6 months of age and older using
Oragene DNA kits which were stored at room temperature at the Division of Research
for future genetics studies. Saliva samples have been validated as an appropriate
method for DNA collection of sufficient quantity and quality for large-scale genetic
epidemiological studies 98], 99]. Procedures for the saliva collection from infants involved obtaining written informed
consent from the mother. Research assistants checked for consent before performing
collection and prepared the Oragene DNA kits. The kits included 2 sponges, a cap,
and a collection tube with funnel top. Prior to saliva collection, the mother was
asked about the infant’s last feeding time and saliva was collected at least 30 min
since last feeding. The research assistants recorded whether a sweetener was used
during saliva collection. During collection, infants were kept in an upright position
and a sponge was placed in cheek pouch along the gums to soak up as much saliva as
possible and gently moved along this area for at least 30 s. The saturated sponge
was placed in the V-notch of the funnel and saliva wrung out using a twisting and
pushing motion against the inner wall of the V-notch. The procedure was then repeated
in the other cheek until the amount collected reached the fill line of the collection
tube. Staff received training from the Oragene DNA specialist to ensure proper collection
of infant saliva via standardized procedures.