Theoretical Considerations Related to Cutoff Points

Abstract

The cutoff points for creating anthropometric indicators of size and growth can be established by three distinct methods: statistical, risk-based, and prescriptive. The theoretical, philosophical, and technical bases for these are quite distinct, but the implications of each method for applications at population and individual levels can be explored by using a common conceptual model. This model posits that any observed anthropometric distribution is a mixed distribution of two (or more) subpopulations, representing some individuals who are or will remain healthy (the specificity distribution) and those who are or will become unhealthy (the sensitivity distribution). The performance and appropriateness of cutoff points based on statistical, risk-based, and prescriptive criteria depend upon the relative sizes of these two subpopulations in a given context, the distance between their means, and the strength and shape of the relationship between the anthropometric indicator and the health outcomes that define these two subpopulations. The risk-based and prescriptive methods both require substantial epidemiologic evidence if they are to fulfill their theoretical and public health expectations, and both face normative (ethical) trade-offs in establishing cutoff points. The prescriptive method faces even stronger normative challenges, especially in relation to overweight and obesity, because its explicit claim regarding the desirable size and growth of children and adolescents may understate the importance of individuality and overstate the strength of the relationship (and the evidence for the relationship) between size, growth, and future health. These concerns are most pronounced for applications at the individual level and for mild-to-moderate elevations of body-mass index and other indicators.