Table 3.

Discriminatory accuracy of extrusion cut-offs

		OA structural feature
		Radiographic OA
Cut-off	3 mm		4 mm
Sensitivity	81 (73–87)		63 (60–65)
Specificity	49 (45–52)		82 (81–83)
Positive predictive value	24 (20–27)		41 (40–43)
Negative predictive value	93 (90–95)		92 (91–92)
% correctly classified	54 (51–57)		79 (78–80)
		BML
Cut-off	3 mm		4 mm
Sensitivity	74 (68–79)		46 (43–48)
Specificity	51 (47–55)		79 (77–81)
Positive predictive value	37 (33–41)		48 (46–50)
Negative predictive value	83 (79–87)		79 (78–79)
% correctly classified	57 (54–61)		70 (68–71)
		Cartilage damage
Cut-off	3 mm		4 mm
Sensitivity	68 (63–73)		39 (37–41)
Specificity	54 (49–58)		84 (82–85)
Positive predictive value	54 (50–59)		67 (65–69)
Negative predictive value	67 (63–72)		63 (62–64)
% correctly classified	60 (57–63)		64 (63–65)

Numbers are estimated as % with 95% confidence intervals in parentheses. Radiographic osteoarthritis (OA) = Kellgren and Lawrence (KL) grade ≥ 2, bone marrow lesions (BML) = WORMS grade ≥ 1, and cartilage damage = WORMS grade ≥ 2