Majallah-i Dānishgāh-i ̒Ulūm-i Pizishkī-i Qum (Jun 2011)
Using LMS Method in Smoothing Reference Centile Curves for Lipid Profile of Iranian Children and Adolescents: A CASPIAN Study
Abstract
Background and Objectives: LMS is a general monitoring method for fitting smooth reference centile curves in medical sciences. They provide the distribution of a measurement as it changes according to some covariates like age or time. This method describes the distribution of changes by three parameters; Mean, Coefficient of variation and Cox-Box power (skewness). Applying maximum penalized likelihood and spline function, the three curves are estimated and fitted and optimum smoothness is expressed by three curves. This study was conducted to provide the percentiles of lipid profile of Iranian children and adolescents by LMS. Methods: Smoothed reference centile curves of four groups of lipids (triglycerides, total-LDL- and HDL-cholesterol) were developed from the data of 4824 Iranian school students, aged 6-18 years, living in six cities (Tabriz, Rasht, Gorgan, Mashad, Yazd and Tehran-Firouzkouh) in Iran. Demographic and laboratory data were taken from the national study of the surveillance and prevention of non-communicable diseases from childhood (CASPIAN Study). After data management, data of 4824 students were included in the statistical analysis, which was conducted by the modified LMS method proposed by Cole. The curves were developed with a degree of freedom of four to ten with some tools such as deviance, Q tests, and detrended Q-Q plot were used for monitoring goodness of fit models. Results: All tools confirmed the model, and the LMS method was used as an appropriate method in smoothing reference centile. This method revealed the distributing features of variables serving as an objective tool to determine their relative importance. Conclusion: This study showed that the triglycerides level is higher and HDL-C level is lower in Iranian children and adolescents than their counterparts in Western countries. Future studies with larger sample size and with higher density at the end points and equal distribution of measurements in changing limits of covariates would hopefully reach more precise findings.
