Estimation of Age-specific Reference Intervals for Skewed Data

Summary

Objectives: To compare Cole’s LMS method with Wright and Royston’s Exponential-Normal (EN) method for estimating reference intervals and generating smooth centile curves for the body mass index (weight in kg/height in meters squared) measurements of children aged 6 to 13 years.

Methods: In the LMS method, the parameters L (the power needed to normalize the data), M (median) and S (coefficient of variation) are modeled as smoothed fits of maximum likelihood estimates. In the Exponential-Normal method, the three parameters mean, standard deviation and skewness are estimated separately using multiple regression techniques.

Results: The centiles generated by the LMS and EN methods are close in most of the age groups. The 2.5^th and 97.5^th quantiles of the interval of the differences between the loss function scores of the LMS and EN methods calculated by bootstrap was found to include zero, indicating that the difference in loss function scores of the two methods is random and not systematic.

Conclusions: The two methods are simple to use and generate comparable centile curves.

Keywords

Normal distribution - body mass index - regression - growth

• References

• 1 Van’t Hof M, Wit JM, Roede MJ. A method to construct age references for skewed skinfold data, using Box-Cox transformations to normality. Hum Biol 1985; 57: 131-9.
• 2 Box GEP, Cox DR. An analysis of transformations. J R Statist Soc B 1964; 26: 211-52.
• 3 Cole TJ. Fitting smoothed centile curves to reference data (with discussion). J R Statist Soc A 1988; 151: 385-418.
• 4 Cole TJ. The LMS method for constructing normalized growth standards. Eur J Clin Nutr 1990; 44: 45-60.
• 5 Cole TJ, Green P. Smoothing reference centile curve: the LMS method and penalized likelihood. Statist Med 1992; 11: 1305-19.
• 6 Royston P, Wright E. A method for estimating age-specific reference intervals (‘normal ranges’) based on fractional polynomials and exponential transformation. J R Statist Soc A 1998; 161: 79-101.
• 7 Royston P, Altman DG. Regression using fractional polynomials of continuous covariates: parsimonious parametric modelling. Appl Statist 1994; 43: 429-67.
• 8 Wright E, Royston P. Simplified estimation of age-specific reference intervals for skewed data. Statist Med 1997; 16: 2785-803.
• 9 Moussa MAA, Shaltout AA, Nkansa-Dwamena D. et al. Factors associated with obesity in Kuwaiti children. Eur J Epidemiol 1999; 15: 41-9.
• 10 Stata Corp.. Stata Statistical Software: Release 5.0 College Station. TX: Stata Corporation; 1997
• 11 Manly BFG. Exponential data transformations. Statistician 1976; 25: 37-42.
• 12 Altman DG. Construction of age-related reference centiles using absolute residuals. Statist Med 1993; 12: 917-24.
• 13 D’Agostino RB, Balanger A, D’Agostino Jr RB. A suggestion for using powerful and informative tests of normality. Am Statistician 1990; 44: 316-21.
• 14 Bonellie SR, Raab GM. A comparison of different approaches for fitting centile curves to birthweight data. Statist Med 1996; 15: 2657-67.
• 15 Moussa MAA, Shaltout AA, Nkansa-Dwamena D, Mourad M. et al. Association of fasting insulin with serum lipids and blood pressure in Kuwaiti Children. Metabolism 1998; 47: 420-4.