Table 5.

Performance of mesothelin and calretinin in women and the probability of a diagnosis of MPM, conditional on the observed biomarker concentrations

Probability %	Females
	Model 1				Model 2				Model 3
	Cut off Mesothelin [nmol/L]	TPR	FPR		Cut off Calretinin [ng/mL]	TPR	FPR		Cut off Mesothelin [nmol/L]	Cut off Calretinin [ng/mL]	TPR	FPR
90	3.24	0.42	0		-	0	0		-	-	0.33	0
80	-	0.42	0		-	0	0		2.13	3.09	0.50	0
70	-	0.42	0.02		4.98	0.08	0		-	-	0.50	0.02
60	1.87	0.58	0.02		3.09	0.17	0		-	-	0.58	0.02
50	-	0.58	0.02		-	0.17	0		-	-	0.58	0.02
40	1.48	0.58	0.03		1.44	0.25	0		1.35	1.23	0.75	0.02
30	1.35	0.75	0.03		0.97	0.50	0.06		1.21	1.12	0.75	0.05
20	1.14	0.92	0.12		0.49	0.58	0.24		1.05	0.72	0.83	0.09
10	0.90	0.92	0.20		0.23	0.92	0.49		0.90	0.23	0.92	0.20

Probability was used to estimate TPR and FPR: Probability = 1/(1+e^(-φ)). Because of the small number of female cases (12), not for all set probabilities corresponding marker concentrations were available. Logistic regression models: (1) with log-mesothelin as predictor, φ = exp [-1.77 + 3.35 * ln(mesothelin)]; (2) with log-calretinin as predictor, φ = exp [-0.72 + 0.99 * ln(calretinin)]; (3) with log-mesothelin and log-calretinin, φ = exp [-1.36 + 3.08 * ln(mesothelin) + 0.33 * ln(calretinin)]. TPR: true positive rate; FPR: false positive rate