Mark Tremblay and Douglas Willms have analyzed data from 3 Canadian surveys and drawn conclusions about secular trends in the relation between body mass index and age.1 Unfortunately, the samples they used are not random. In order for results to be generalized to the population at large, analyses must take sampling weights into account. The variances estimated from unweighted regression analyses underestimate the variance in the population, and more reliable variances are generally now computed using bootstrap methodology.2
These methodologic issues have important implications for the authors' findings. Although their results might provide some information about body mass index in Canadian children, there is no guarantee that they are representative of results for the country as a whole.