The influence of aerobic fitness on obesity and its parent-offspring correlations in a cross-sectional study among German families

Study sample

The study was designed as a cross-sectional study, enrolling the families of index
children whose AF had been measured within the framework of the German part of the
IDEFICS (Identification and prevention of Dietary- and lifestyle-induced health Effects
In Children and infantS) study 25], 26]. To be eligible, these index children had to be older than five years and had to
have performed a shuttle-run test in the IDEFICS study showing either high (above
70th percentile) or low (below 30th percentile) AF. Participants were selected from
the extremes of the AF distribution to sharpen the contrast. Only first degree biological
relatives including father, mother and siblings aged 5–17 years were eligible for
this study. Participation of both parents and of at least one child, i.e. of a complete
nuclear family, was required for inclusion of the respective family in the analysis.
Subjects were not eligible if: (i) he/she suffered from serious heart disease or severe
asthma, (ii) he/she was not capable of taking part in a maximum stage exercise test
due to physical handicap, body height above 210 cm or body weight above 150 kg.

Overall, 278 families were eligible for participation in this study; 199 families
did not participate for the following reasons: no time (N?=?57), unavailability of one parent during the recruitment period (N?=?36), not convinced of the study purpose (N?=?3), other reasons (N?=?56) or because they were never reached personally (N?=?47). Eventually, 79 families with 158 adults and 137 children participated in the
study (participation proportion 28 %). Six participants did not successfully perform
the maximum stage test to assess their AF and were subsequently excluded: one father
because of an unexpected high resting pulse rate, one child because he/she did not
understand the task, two children because they belatedly refused to take part and
further two children because they did not meet the workout criteria described below.
Overall, a total of 289 subjects were enrolled in the study. The average family size
was 3.66 with 38 families participating with one child, 34 families participating
with two children and seven families participating with more than two children. This
final sample provided 80 % power to detect a minimum intraclass correlation coefficient
for AF of 0.14.

All examinations took place under supervision of a physician between March and August
2009. The study was approved by the ethics committee of the University of Bremen and
written informed consent was given by all parents for themselves and for their children.
In addition, oral assent was obtained from all participants.

Anthropometry

Measurement of body weight and assessment of body fat percentage based on bioelectrical
impedance was performed using an adapted version of electronic scale TANITA BC 420
SMA. Measurement was done non-fasted with bare feet and in underwear. Height was measured
using a telescopic stadiometer, waist and hip circumferences were measured with a
tape SECA 200, both following the International Standard for Anthropometric Assessment
27]. The calculated body mass index (BMI?=?weight[kg]/height
²
[m
²
]) was classified according to the international reference values available from the
WHO 28] for participants older than 18 years. The IOTF reference 29] was used to classify childhood BMI categories. Waist-to-hip ratio was calculated
dividing waist circumference [cm] by hip circumference [cm].

Aerobic fitness

AF expressed as the maximum oxygen consumption (VO
_2max
) was assessed by a maximum exercise test performed on a stationary bicycle (Ergoselect
200p). Prior to the test, each subject underwent a medical examination and an interview
by a physician including a medical history with special focus on cardiovascular diseases
and current medications. The measurement of AF followed the recommendations of the
American Heart Association 30]. Different protocols were applied for the exercise test depending on age and sex
and for children additionally depending on their weight: Fathers started with an initial
workload of 50 Watts (W) which was increased by 50 W every three minutes, whereas
mothers started with an initial workload of 40 W which was increased by 40 W every
three minutes. All children started with an initial workload of 20 W, which was increased
by 0.5 W per kg body weight every two minutes. All subjects performed the test until
voluntary exhaustion. During the test, expired gas was analyzed for concentrations
of O
₂
and CO
₂
(MetaMax 3B, Cortex Biophysik GmbH, Germany). Breath-by-breath values were processed
by a three-point median filter to exclude outliers and were averaged over 15 s intervals.
Finally, VO
_2max
was related to body weight and expressed as ml ? min
^-1
? kg
^-1
, a well-established approach to normalize VO
_2max
. The test result was considered acceptable if the parents met one of the following
two criteria: maximum heart rate (HR) in beats per minute of at least 200 – age in
years or a respiratory exchange ratio (RER) of at least 1.0. In children, maximum
voluntary exhaustion was assumed when a maximum HR of at least 185 beats per minute
was reached. HR and RER were continuously displayed and recorded during the test.
A 12-channel ECG was recorded during the test to control for pathological responses
of the heart during exercise. Maximum workload in W was assessed by linear approximation
if the last stage of the test was not maintained for the full allocated time. For
example, if a father aborted the test after 1:30 min while cycling with a loading
of 250 W, maximum workload was set to 200 W + (50 W ? 0.5)?=?225 W.

Physical activity

PA was measured using an uniaxial accelerometer (Actigraph or Actitrainer) which is
described in detail in Sirard et al. 31]. The participants were asked to wear the accelerometer on the right hip for seven
consecutive days during waking hours. The activity monitor recorded the number of
activity counts in 15 s epochs. Consecutive zero counts of over 20 min were classified
as non-wearing time. A day was considered as valid if the total minutes of wearing
time (minutes the accelerometer monitored any counts) was greater than ten hours.
A minimum of two valid days was required for PA variables to be included in the analysis.
Average minutes per day spent in moderate or vigorous physical activity (MVPA) were
calculated as average time per day spent above ?3000 counts per minute. With this
cut-off value, we chose a compromise between the many different recommended cut-points
for children, adolescents and adults. Average wearing time is used to adjust for the
proportion of MVPA since the time spent in MVPA is correlated with accelerometer wearing
time.

Smoking behavior

Parents were also asked to report their smoking behavior. Non-smokers were defined
as parents that had never smoked and ex-smokers were defined as parents that had stopped
smoking more than a year ago. Pack-years were calculated as the product of smoking
duration in years and dose in packs per day (20 cigarettes ? package
^-1
). A family was considered to be a smoking household if at least one parent stated
that someone smoked at home.

Statistical analysis

Referring to the first research question, familial aggregation of AF was assessed
in complete nuclear families using the intraclass correlation coefficient (ICC) obtained
from an analysis of variance (ANOVA) 32]. A significant F-value implies that members of a family are more alike in their AF than non-members
are. The ranges of raw VO
_2max
values differ between children and adults as well as between males and females. However,
comparability of AF values in children and adults as well as between sexes is required
to allow analyses of familial aggregation and heritability. To make the raw AF values
comparable across all study participants, a two-step design was used where in the
first step a linear regression was fitted with relative VO
_2max
as the dependent variable. The resulting residuals, that are the differences between
estimated and observed values of the relative VO
_2max
, were included as the dependent variable in an ANOVA to determine the ICC for AF
in the second step 13], 22]. More detailed, the basic linear regression model for relative VO
_2max
(step 1) included sex (X₁
), children’s age (X₂
, continuous, zero for parents), parents’ age (X₃
, continuous, zero for children), and two multiplicative interaction terms for age
and sex as independent variables

where the error term ? was assumed to be normally distributed with zero mean. Children’s age and parents’
age were modelled as two distinct continuous variables because age would appear as
a bimodal variable otherwise. Other covariates like PA (average MVPA, valid time,
both continuous), BMI (categorical), smoking household (binary) and a multiplicative
interaction term for BMI and average MVPA (continuous) were added to the basic model
as additional independent variables resulting in six different models (Table 2). Residuals from each of these models were used to calculate the ICC by an ANOVA
in a second step, where each family formed one cluster. Significance tests of the
ICC were adjusted for multiple testing using Bonferroni correction for six models,
i.e. the significance levels was chosen as ??=?0.05/6?=?0.00833. The sample size for
the ICC analysis was reduced by two participants due to the restriction to complete
nuclear families. Because a complete case analysis was performed, sample sizes varied
due to missing values for PA items and smoking information.

To explore the heritability of AF in complete nuclear families (second research question),
the mid-parent-offspring design as well as the single-parent-child design were used 33]. Heritability was estimated by regressions of offspring phenotype on, both, the average
phenotype of both parents (mid-parent-offspring design) and the phenotype of only
one parent (single-parent-child design). The single-parent-child design allows distinguishing
maternal from paternal effects. A linear mixed model of offspring AF on parental AF
(X) was performed for the mean of the AF of both parents (mid-parent-offspring design)

and adjusted for children’s individual PA (average MVPA, valid time) and children’s
BMI where ? and ? were assumed to be normally distributed with zero mean. Families (Z) were taken into account as clusters and were treated as a random variable in mixed
models. AF was included in the linear mixed model as the obtained residuals from the
basic model on the relative VO
_2max
(see step 1 above). The estimated regression coefficient ? represents the upper-limit estimate of the heritability (h
²
) for the mean parental AF and reflects the resemblance between first-degree relatives.
Using a single-parent design, the upper-limit estimate is equivalent to twice the
regression coefficient 33].

To answer the third research question, logistic mixed models were calculated to estimate
the association between AF (quintiles, W) and overweight/obesity (Y). Families or respective parents were taken into account as clusters (Z)

where ? was assumed to be normally distributed with zero mean. Residuals were classified
into quintiles mainly for two reasons: 1) there are no meaningful cut-offs to classify
residuals; 2) we observed a non-linearity of the odds ratios that would not have become
visible if we had modeled AF using a continuous covariate. As complete nuclear families
were not essential for this analysis, all 289 participants were used to estimate individual
odds ratios (OR) and their 95 % confidence intervals (CI) to measure the strength
and direction of the association. We performed a sub-group analysis for parents to
additionally adjust for smoking behaviors.