|Home | About | Journals | Submit | Contact Us | Français|
The article by Del-Rio-Navarro et al1 provides percentiles of height, weight and body mass index (BMI) for Mexican youth aged 10-18 years, calculated using a version of my LMS method. However their description of the method is incomplete and misleading.
They describe their approach as a least-mean-squares (LMS) procedure similar to that of Kuczmarski et al.2 They then say that “In the LMS technique, three parameters are estimated: the median (M), the generalized coefficient of variation (S) and the power in the Box-Cox transformation (L).” In fact the L-M-S they describe is the lambda-mu-sigma of my LMS method,3 which they don't cite. They go on to fit it in two stages rather than one, while analysing the data in one-year age groups rather than using continuous age, an approach that leads to bias. They may be interested to know that free lmsChartMaker software exists, designed for non-statisticians, to fit the LMS method as originally described.
The process of fitting the LMS method involves deciding on the smoothness of the three curves for L, M and S plotted against age. Judging from the fitted percentile curves in Figures 1-3, where the outer percentiles are appreciably rougher than the inner ones, the L curves were insufficiently smoothed.
Separately, the Discussion of the article states that “WHO has issued international definitions for overweight and obesity in children (Cole et al, 2000).” I should like to emphasise that WHO had no involvement in these cut-offs, which were sponsored by the International Obesity TaskForce (IOTF).4
The IOTF cut-offs were obtained by averaging across six nationally representative samples BMI percentile curves corresponding to BMI 25 and 30 kg/m2 at age 18 years.4 They have been highly cited since publication. Following the IOTF example the authors propose using as their national cut-offs similarly defined percentile curves, but based on their own data. It would have been interesting to see how closely their curves match the IOTF cut-offs– they are likely to be very similar. It is worth noting that the IOTF cut-offs would be just as effective as the Mexican percentiles for national obesity monitoring purposes, but the IOTF-based prevalence rates would have the advantage of being comparable to other published rates, which the nationally based cut-offs would not.