Table 3.

The ROC statistics comparing extraction methods.

		n	ROC area	Standard error	90% CI	p-value
Averaged concentrations	Extraction A	20	0.843	0.098	0.686–1.000	0.071
	Extraction B	20	0.956	0.048	0.876–1.000
Independent sample concentrations	Extraction A	59	0.828	0.057	0.734–0.921	0.001
	Extraction B	59	0.963	0.023	0.925–1.000

One of the replicates for a male in extraction A was lost during analysis reducing the total number of observations per extraction method to 59. A cut-off value of 1.9 pg/mg was associated with a sensitivity of 96% and a specificity of 91% when pulverised hair was used for analysis, the same cut-off value was associated with a sensitivity of 85% and a specificity of 73% when non-pulverised hair was used (n = 59). Calculating the ROC statistics using only mean hair testosterone-like immunoreactivity concentrations showed a trend towards better separation between the sexes favouring pulverised hair (n = 20, p = 0.071).