BACKGROUND: No standard exists for the adjustment of lung function for height and age in children. Multiple regression should not be used on untransformed data because, for example, forced expiratory volume (FEV1), though normally distributed for height, age, and sex, has increasing standard deviation. A solution to the conflict is proposed. METHODS: Spirometry on representative samples of children aged 6.5 to 11.99 years in primary schools in England. After exclusion of children who did not provide two repeatable blows 910 white English boys and 722 girls had data on FEV1 and height. Means and standard deviations of FEV1 divided by height were plotted to determine whether logarithmic transformation of FEV1 was appropriate. Multiple regression was used to give predicted FEV1 for height and age on the transformed scale; back transformation gave predicted values in litres. Other lung function measures were analysed, and data on inner city children, children from ethnic minority groups, and Scottish children were described. RESULTS: After logarithmic (ln) transformation of FEV1 standard deviation was constant. The ratios of actual and predicted values of FEV1 were normally distributed in boys and girls. From the means and standard deviations of these distributions, and the predicted values, centiles and standard deviation scores can be calculated. CONCLUSION: The method described is valid because the assumption of stable variance for multiple regression was satisfied on the log scale and the variation of ratios of actual to predicted values on the original scale was well described by a normal distribution. The adoption of the method will lead to uniformity and greater ease of comparison of research findings.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.