with participants, families, school and health officials, and the general public. Doing so will minimize the confusion that might arise from communicating in terms of the percentiles used to derive the cut-points. For example, the CDC has used the 95th percentile from a previous decade to derive cut-points for obesity in children, yet more than 15 percent of youth currently exceed that 95th percentile. These are challenging issues to communicate. In cases where percentiles may allow for a clearer presentation of the results than cut-points, as with BMI, the year of data collection should be reported with the percentile. In this connection, researchers developing percentile data with which to derive cut-points should also report the time of data collection.

Establishing Cut-Points When the Relationship Between the Test of Interest and a Health Marker in Youth Is Known

In the ideal situation, when there is a concurrent relationship between a health outcome and a fitness test, the cut-points for the test are determined using a data mining procedure2 to establish statistical evidence for the relationship. While not common, this kind of concurrent relationship does exist and has been used for setting cut-points. For example, based on the concurrent relationship between body composition and a set of health outcome measures (total cholesterol, serum lipoprotein ratio, and blood pressure) (Williams et al., 1992), a set of cut-points was derived for evaluating body composition (Going et al., 2008). Similar applications have been reported for setting cut-points for cardiorespiratory endurance (Lobelo et al., 2009) and waist circumference in Chinese school-aged children (Liu et al., 2010) and for body composition and cardiorespiratory endurance tests (Going et al., 2011; Welk et al., 2011).

Establishing Cut-Points When a Concurrent Relationship Has Not Been Confirmed in Youth

Even if a concurrent relationship between a health outcome measure and a putative health-related fitness test has been well established in adults and cut-points exist for that population, such a relationship often has not been confirmed in youth. Because a negative health outcome (e.g., low-back pain, cardiovascular diseases) may take years to develop, children’s health


2Data mining involves varying cut-points, computing agreement-related statistics with the classification of health outcome measures each time, and determining the cut-points according to optimal statistical results (e.g., highest P- and kappa-coefficients, specificity index and relative risk statistics, findings and illustration of receiver operating characteristic [ROC] curve analysis). If the cut-points are set across a large range of groups, the data from these groups usually are smoothed before the data mining procedure is applied.

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement