Clinical validity: Defining biomarker performance

Figures & data

Figure 1. A schematic representation of a diagnostic accuracy study (A: left) and an example of a group to be screened by a marker (KLK2) (B: right).

Table I. Diagnostic accuracy study results.

	Needle Biopsy
KLK2	Cancer	No Cancer
Positive	TP = 42	FP = 75
Negative	FN = 1	TN = 20	N =138

Results of a (hypothetical) diagnostic accuracy study to evaluate the accuracy of kallikrein-related peptidase 2 (KLK2) in patients with suspected prostate cancer. TP: True positives; FP: False positives; TN: True negatives; FN: False negatives.

Table II. Measures for reporting and interpreting results of diagnostic accuracy studies.

Measure	Definition	Estimate	95% CI
Error-based
Sensitivity (Se)	TP/(TP ± FN)	98%	90% to 100%
Specificity (Sp)	TN/(TN ± FP)	21%	17% to 22%
Overall Fraction Correct*	(TP ± TN)/N	45%	40% to 46%
Youden's J index*	Se + Sp − 1	19%	7% to 22%
Information-based
Positive Predictive Value	TP/(TP ± FP)	36%	33% to 37%
Negative Predictive Value	TN/(TN ± FN)	95%	79% to 99%
Positive Likelihood Ratio	[TP/(TP ± FN)]/[FP/(TN ± FP)]	1.24	1.09 to 1.28
Negative Likelihood Ratio	[FN/(TP + FN)]/[TN/(TN ± FP)]	0.11	0.02 to 0.59
Association-based
Diagnostic Odds Ratio	[TP × TN]/[FP × FN]	11.2	1.8 to 67.4
Error Odds Ratio*	[TP × FP]/[TN × FN]	157	41 to 855
Positive Predictive Ratio*	[TP/(TP + FP)/[FN/(TN × FN)]	7.4	1.6 to 43.1
Kappa*		0.13	0.05 to 0.15

Measures with an asterisk are given for completeness, but are not recommended.