Updated prevalence rates of overweight and obesity in 11- to 17-year-old adolescents in Germany. Results from the telephone-based KiGGS Wave 1 after correction for bias in self-reports

Study population

The analyses presented are based on the data from KiGGS Wave 1. The goals, concept
and design of KiGGS have been described elsewhere 1], 7], 17]. KiGGS Wave 1 (2009–2012), the follow-up of the KiGGS baseline study (2003–2006),
was carried out as a telephone-based survey. An essential aim of KiGGS is to regularly
provide population-based cross-sectional data on the health situation of children
and adolescents aged 0–17 years living in Germany. Amongst others, the study population
of KiGGS Wave 1 consists of re-invited participants from the baseline study (KiGGS
cohort). A total of 5258 re-invited children and adolescents aged 11 to 17 years participated
in KiGGS Wave 1 (response 73.9 %). The net sample was compared with the resident German
population regarding particular population characteristics and an analysis of the
relationship between the re-participation rate and certain characteristics collected
in the baseline study, suggesting a mostly unbiased sample after taking the sample
weights into account 7].

Trained study staff conducted standardised telephone interviews with adolescents aged
11 or older. For further information (e.g. socioeconomic status) and to collect information
about the younger participants, parents of children and adolescents aged 0–17 years
were interviewed. The software product Voxco Version (Voxco Inc., Montréal
QC, Canada) was used to manage the calls and collect the data. A written informed
consent from parents or caregivers was required prior to the interviews. The Federal
Office for Data Protection and the ethics committee of the Charité Medical University
Berlin approved the survey.

Additionally, the ‘Motorik Modul’ (MoMo), which collected data on motor fitness and
physical activity of children and adolescents aged 4 to 17 years in a physical examination,
was part of KiGGS Wave 1 16]. At the end of the telephone interview, approximately half of the KiGGS Wave 1 subjects
were asked to participate in MoMo. If they gave their consent, they received information
material and were contacted to make an appointment for the physical examination.

Self-reports and anthropometric measurements

In the telephone interview, adolescents were asked to report their height (without
shoes) and weight (without clothes) to an accuracy of 1 cm or 1 kg, respectively.

Trained staff, in the physical examination of the ‘Motorik Modul’?, took anthropometric
measurements. Body height was measured without shoes to an accuracy of 0.1 cm using
a portable stadiometer. Body weight, while the participant was wearing just underwear,
was measured to the nearest 0.1 kg using a calibrated electronic scale.

Body mass index (BMI) in kg/m
was calculated both from self-reported and from measured data. Weight status was determined
using age- and gender-specific cut-offs for strong underweight (3rd percentile),
underweight (?3rd percentile to 10th percentile), normal weight (?10th percentile
to ?90th percentile), overweight (90th percentile to 97th percentile) and obese
(?97th percentile), based on the national German reference 18].

Body perception

Each adolescent’s body perception (BP) was examined by asking the following questions
in the telephone interview: ‘Do you think you are …’ ‘much too thin’?, ‘a bit too
thin’?, ‘exactly the right weight’?, ‘a bit too fat’?, or ‘much too fat’? 19]. Responses were classified into the following categories: (1) ‘too thin’ (summarising
‘much too thin’ and ‘a bit too thin’), (2) ‘right weight’?, and (3) ‘too fat’ (summarising
‘a bit too fat’ and ‘much too fat’).

Socioeconomic status

The socioeconomic status of the participants was assessed by a multidimensional index
score. The parents were asked to report their education and occupational qualifications,
occupational status, and net income. This information was used to calculate the socioeconomic
sum score, which was categorised into the following groups: (1) low, (2) moderate,
and (3) high socioeconomic status 20].

Statistical analysis/correction procedure

The analyses focused on 2509 boys and 2446 girls, aged 11 to 17 years, who were interviewed
in KiGGS Wave 1. Cases with missing values for body perception (seven cases) were
excluded, which led to a total sample size of 4948 adolescents (2505 boys and 2443

Measured height and weight from MoMo were available from 899 boys and 872 girls aged
11 to 17 years at the time of the KiGGS Wave 1 telephone interview. Due to the time
lag between KiGGS Wave 1 and MoMo, there were 36 cases that were 17 years old at the
time of the KiGGS Wave 1 telephone interview, but turned 18 when they participated
in MoMo. These cases were included in the correction procedure. MoMo participants
who did not give information about their body perception in the telephone interview
were excluded from the correction procedure. Participants with a time lag between
the telephone interview and the examination part of MoMo of greater than 3 months
(90 days) were also excluded from the analyses. For the correction procedure of prevalence
rates derived from self-reports of KiGGS Wave 1, weight status derived from height
and weight measured in MoMo was thus available for 826 boys and 814 girls.

In previous analyses with data from KiGGS baseline, which provide both self-reported
and measured values, correction of prevalence rates for weight status derived from
self-reports with Kurth and Ellert’s 10] formula showed satisfying results 13]. Therefore, the correction procedure developed by Kurth and Ellert with data of KiGGS
baseline (formula 16) 10] (i.e. the statistical procedure of directly correcting the prevalence rates) was
replicated here in order to derive a new correction formula for KiGGS Wave 1. The
derivation of a new correction formula was necessary since the amount and patterns
of underreporting might vary (1) over time; and (2) with the change of survey mode
(face-to-face interview in KiGGS baseline vs. telephone interview in KiGGS Wave 1).
For the correction procedure, the prevalence based on measured height and weight from
MoMo was used, as well as the information regarding individual body perception.

Let R k
, k?=?1…3 denote the prevalence of body perception (BP) category k in KiGGS Wave 1, i.e. R k
=?P (BP?=?k), k?=?1…3 (‘too thin’?, ‘right weight’?, ‘too fat’). Q jk
is the uncorrected prevalence, based on self-reported height and weight in KiGGS Wave
1, of weight status category j in the group of adolescents with body perception k, i.e. it is the conditional probability

Q jk
 =?P (BMI reported
 ??I j
 |?BP?=?k) (j?= 1?…?6; k?= 1?…?3), with

I 1
=?Extremely underweight (P3)

I 2
=?Underweight (?P3 -??P10)

I 3
=?Normal Weight (?P10 -??P90)

I 4
=?Overweight (?P90 -??P97)

I 5
=?Obese (?P97)

I 6
=?Self-reported height and/or weight missing

Here, category I6
was newly introduced to represent adolescents with missing values for weight status
derived from self-reports, but with available information on weight status derived
from measurements and body perception.

The corrected prevalence rate for weight status category j in KiGGS Wave 1, P?(???
), is then given by:


with I1
defined as above, where the relationship between measured and self-reported values
is captured by the factors ?
, which are the conditional probabilities for measured weight status category i in the group of adolescents with body perception k and weight status category j based on self-reported height and weight, i.e. ??
 =?P?(BMI measured
 ??I i
?|?BMI reported
 ??I j
 , BP?=?k), with I1
defined as above.

The factors ?
were determined in the MoMo subsample. Weight status based on measured values was
taken from MoMo, whereas information about weight status based on self-reports and
body perception was taken from the telephone interview in KiGGS Wave 1. Then, 95 %
confidence intervals for the corrected prevalence rates were estimated via a bootstrap
procedure with 2000 replicates by resampling 166 sample points with replacement from
the 167 original sample points 21] and using the 2.5 and 97.5 % quantiles of the bootstrapped distribution as the confidence
limits. These confidence intervals include the effect of the additional uncertainty
introduced by the correction procedure.

All analyses, including the correction procedure, were performed with the survey procedures
of SAS release 9.4 (SAS Institute Inc., Cary, NC, USA, 2012), taking sample weights
and the clustering in sample points into account. The analyses were both conducted
separately for boys and girls and for the total sample. The sample weights were used
to correct for possible deviations of the sample from the population structure (as
of December 31, 2010) with regard to age, gender, region, parental education, and
nationality (whether or not they were German). Furthermore, they included a correction
for selective dropout based on a logistic regression modelling of the reparticipation
probability 7].

For a comparison of the corrected prevalence rates derived from KiGGS Wave 1 with
the prevalence rates from the KiGGS baseline study, the latter ones were recalculated
using sample weights analogous to KiGGS Wave 1 (i.e. using the population structure
as of December 31, 2010 and including parental education in the calculation of the
sample weights) 7].