Interpreting skinfold sums. Use of absolute or relative typical error?

The main aims of this study were to: 1) quantify how divergent the sum of seven skinfolds (Sigma7(skinfolds)) of athletes have to be before a general index of measurement imprecision (typical error, TE) is no longer appropriate to very lean or somewhat overweight athletes, and 2) discuss the application of absolute or relative (TE%) typical errors. The Sigma7(skinfolds) was measured in duplicate on 101 athletes by one level 3 anthropometrist with one calibrated skinfold caliper. The TE and TE% for E7(skinfolds) were calculated for all data (TEALL, TE%(ALL)), as well as for results less than 50 mm (TE<50, TE%(<50)), inclusive of 50-74.9 min (TE50-74.9, TE%(50-74.9)), inclusive of 75-99.9 min (TE75-99.9, TE%(75-99.9)) and 100 min and greater (TE,100, TE%,100). At least 20 samples were taken for each measurement range. Limits of agreement (LoA) at the 68% and 95% confidence level were also calculated for all data and the four skinfold ranges. The TE, TE 68, and 95% LoA increased as a direct function of the total E7(skinfolds) and the general TEALL was inappropriate for TE<50 and TEgreater than or equal to100. In contrast, the %TE was very similar within each confidence level (ranging from 1.6-2.1% and 3.1-4.2% for the 68% and 95% LoA, respectively) regardless of the skinfold total. To minimise inaccurate feedback to individuals, anthropometrists dealing with skinfolds of elite athletes should establish an absolute TE on a homogeneous sample of athletes with a E7(skinfolds) in a narrow band, for instance <50 mm. We also urge prudence in interpreting change in skinfold totals and suggest that the 95% level of confidence is appropriate in most instances. (C) 2003 Wiley-Liss, Inc.