Sample size and sample composition for constructing growth reference centiles.

Area of Science:

• Pediatrics
• Biostatistics
• Growth Monitoring

Background:

• Growth reference centile charts are essential tools in child health assessment.
• Current study designs for collecting reference data lack clear guidance, leading to variable sample sizes.
• Methods like the LMS method and GAMLSS are used for centile construction.

Purpose of the Study:

• To establish a theoretical framework for optimal design of cross-sectional growth reference studies.
• To determine the ideal sample size and age distribution for precise centile estimation.
• To provide practical guidance for future growth study designs.

Main Methods:

• Utilized Generalised Additive Models for Location Scale and Shape (GAMLSS) to fit centiles for weight, height, BMI, and head circumference.
• Analyzed data from 6878 boys aged 0-21 years from the Fourth Dutch Growth Study.
• Employed simulation studies to explore the impact of sample size and age distribution (sample composition) on precision.

Main Results:

• Smoothing centiles increased effective sample size two- to threefold through 'borrowing strength'.
• Optimal sample composition varied by centile: λ=0.4 for the median (infant over-sampling) and λ=0.75 for the 2nd/98th centiles (less infant over-sampling).
• Precision was measured on the z-score scale, with standard errors of 0.041 for the median and 0.066 for extreme centiles.

Conclusions:

• Both sample size and sample composition are critical factors for optimal growth reference study design.
• Well-designed studies require 7,000-25,000 subjects per sex for accurate growth assessment.
• The findings offer practical recommendations for improving the methodology of growth reference studies.