4. Database development and reference d ata
4.4.2.2. Height velocity
Growth standards based only on cross sectional data are different in form from those
that represent individual longitudinal curves of growth (Cole 1994; Tanner &
Whitehouse 1976), and this is particularly relevant over the period of the adolescent
growth spurt. One solution in clinical practice is to use longitudinal or 'distance' charts
for growth monitoring. An alternative measure, which can be quantified, is to use
height velocity standards, either on charts or as SDS. Height velocity standards are
centred around the peak height velocity, with more extreme centiles representing the
growth rate for early and late maturers.
Height velocity reference data on which these standards are based have not been
updated in the UK since 1966. The Harpenden growth study, from which the standards
are derived, are based on measures from 49 boys and 41 girls followed throughout
their adolescence (Tanner, Whitehouse, & Takaishi1966). Measurements were taken
at 3 monthly intervals during the growth spurt and at 6 monthly intervals before and
after this time. Each individual's growth data was smoothed to remove measuring error
and to take into account seasonal and other variation. The resulting curves were then
read off at 3 monthly intervals and velocity curves derived. SDS for peak height
velocity were then obtained by arranging the curves over one another, so that the peak
coincide, thus giving a spread of growth velocity around the central peak. The average
age at which the peak was achieved was obtained from averaging the growth velocity
curves. From these, centiles for height velocity were obtained (Tanner &
Whitehouse1966), and SDS calculated as in appendix 4.2: section 11.2.
4.4.2.3. Skinfoid thickness
Skinfold thickness reference data were taken from the Tanner-Whitehouse standards
(1975) from birth to 19 years, converted to SDS using the LMS method (Davies, Day,
takes into account both the skewness of the normative skinfold data, and the fact that
the degree of skewness varies with age. Tables for L, M, and S values for triceps and
subscapular measurements are given in Davies et al. (Davies, Day, & Colei 993), and
SDS calculated as described in appendix 4.2; 11.1.
Skinfold thickness, as with other indices of nutritional status, are undergoing secular
change (Chinn & Rona 1994), particularly for girls. However, as yet there is no rival to
the Tanner-Whitehouse standard with reference to skinfold thickness, last revised 30
years ago, although revision has been called for (Paul et al. 1998). In addition,
discrepancies within ethnic groups of skinfold thickness have been noted, but as yet
no racially diverse reference standards exist (Rona & Chinn 1987).
4.4.2.4. Arm circumference
The most widely quoted reference for mid arm circumference is from Jelliffe (Jelliffe &
Jelliffe 1980), based on a study of healthy Polish children (Velzeboer et al. 1983).
However, the more recent study of 2,555 Dutch children from birth to 19 years of age
in the town of Osterwolde between 1979 and 1980, provide detailed percentiles of mid
upper arm circumference (Voorhoeve 1990). The values from the Dutch data are
approximately 0.5 cm above the figures for Jelliffe's national standard. Although not
ideal, in view of the increasing trend towards overweight (Reilly, Dorosty, & Emmett
1999), they are likely to be an improvement on 1964 data from Polish children. In
addition, what bias there is will tend to under-represent the degree of relative
malnutrition in underweight patients to today's child and adolescent population.
The second source of reference data available for the relevant age group for this study
is from a Spanish cohort (Hernandez et al. 1998). Although more up to date, the
numbers in the study are smaller. Comparisons between the Dutch and Spanish data
(males and females) are shown in figures 4.1. Overall similarity between the data sets
can be seen. For the purposes of the present study the Dutch reference data were
fit of data over the age distribution. SDS were calculated from the 50^ centile figures
and SD entered into a reference table, using the SD function described in appendix
4.2; section 11.2.
4.4.2.5. Puberty
Standards for pubertal development and derived puberty centiles (Tanner &
Whitehouse 1976) are based on the data from the original 1966 Harpenden cohort, and
were assessed using the method described by Tanner (1962). These standards are
undoubtedly out of date, although as yet no better reference dataset has been
developed. Tanner et al. advise that pubic hair and penis size or breast size be rated
separately, since the two often develop with different timings.
Unfortunately for the purposes of the present study, in which delay of puberty would be
an anticipated finding, 'puberty SDS' are harder to derive. Therefore, whilst a cross
over between stage 1 and stage 2 of puberty can be defined, there is no median or
mean for stage 1 puberty (prepubertal) and for stage 5 (adult maturity). Pubertal stage
has therefore been used in the analyses as a grouping variable where appropriate.
4.4.2.6.
Bone Density
Reference data for DXA scans in children are inadequate. The main obstacle is the
ethics of exposing large numbers of normal children to radiation, albeit low quantities.
This was particularly true for the older, pencil beam scanners, and efforts to collect
good quality reference data for children are now being made. It was beyond the scope
of this study to produce reference data on normal populations. In addition, the number
of different scanners makes comparison across machines problematic, and reference
data obtained on one machine is not directly transferable to another machine. The two
Figure 4.1: Comparison of mid-arm circumference reference data from two sources by
gender
MIDARM CIRCUMFERENCE OIRLS
Series 1 - Series 2
AGE
M I D A R M C I R C U M F E R E N C E B O Y S
Series 1: Spanish Mid-upper arm circumference (Hernandez,M., Catallet.J.,
Narvai.J.L., Rincon,J.M., Ruiz,!., Sanchez,E., Sobradillo,B., and Zurimendi,A.: Curvas y Tablas de Crecimiento. Fundacion F Obregozo, Bilbao. Ed. Garsi, Madrid 1998)
Series 2: Dutch Mid-upper Arm Circumference (Voorhoeve, H.W.A. Journal of Tropical
Paediatrics Vol 36 October 1990)
Table 4.1 : Sources of normative data for children by DXA machine and age range Machine Age range Age Interval No. of children Source Limitations Hologic Q D R 4500
8-17 yrs 3 years 666 Hologic (adapted
from Faulkner et al. (1996) Calcif. Tissue Int. 59:5:344)
No difference between male and female values. Age bands too wide. Hologic QDR1000/W 4-14 yrs 1 year 83F 95 M Cambridge schoolchildren 1994 -1997 Personal communication Only up to 14 years. No. of children in higher age groups limited. Lunar DPX software version 3.6 4-25 yrs 0.25 Years 138 F 139 M Lu et al. (1994) J Bone Miner. Res. 9: 1451
Best available data, except toward end of puberty Hologic QDR- 1000/W 180 children and young adults Nysom et al. (1998) Acta Radiologica 39 632-636)
Lumbar data from whole body scan lower than from a dedicated lumbar scan
Figure 4.2: Comparison of bone density data using different DXA scanners.
Between scanner
variation - GirlsQ S CD ■ H OI4 500 -a- HoIIOOO - ^ H o M O O O ^ L u n D P X 3 . 6 H olTotB A g e 10 13 16 B e t w e e n s c a n n e r variation - Boys o s m 0.6 0.4 0.2 7 10 Age 13 16 HOI4500 HollOOO HollOOO LunD PX HolTotBod 149
Q4500 models; and Lunar. The reference data summarised in table 4.1 were
available at the time of the study: all have their limitations for the present study cohort.
The scan data collected during the study included z scores calculated from the
reference data used by the Hologic QDR4500 software. In order to improve the
interpretation of the results obtained for subjects, the options were:
■ recalculate z scores from better data but collected on different machines
■ examine the relationship between reference data from various machines,
looking for sources of systematic error.
The latter was performed, on the basis that if there was a significant non-linear
relationship between the data sources, that z scores could be recalculated. Figure 4.2
show the relationship between the four sets of reference data at stepped time points
from age 7 to 17 (the age of range of the study cohort) for girls. Findings were
comparable for boys. Hologic 4500 readings were comparable to those from other
samples and machines in a consistent way. The slight bias towards lower readings
was not thought significant enough to merit reanalysis of the data.
4.5.
Summary
The database was designed to collate and present data in a flexible format that
allowed it to be transformed to compare with reference data. Using queries, results
were drawn from subsets of the data. Microsoft Access is not a statistical package,
and data needed to be analysed in appropriate statistical packages. Any alteration or
transfer of data from one source to another introduced the risk of error, with the added
effect that once outside the main database, delete and alteration functions need to be
systematically backdated to include the entire dataset. Methods used by the database
to minimise such errors are described.