Table 2.
Threshold analysis and diagnostic accuracy
| SUV cut-off | TP | FP | TN | FN | Sensitivity, % (95% CI) | Specificity, % (95% CI) | NPV, % (95% CI) | PPV, % (95% CI) | Benign call rate, % (95% CI) | |
|---|---|---|---|---|---|---|---|---|---|---|
| Visual assessment | ||||||||||
| All (n = 123) | 31 | 53 | 37 | 2 | 93.9 (79.8–99.3) | 41.1 (30.8–52.0) | 94.9 (82.7–99.4) | 36.9 (26.6–48.1) | 31.7 (23.6–40.7) | |
| Non-Hürthle cell nodules (n = 94) | 22 | 34 | 36 | 2 | 91.7 (73.0–99.0) | 51.4 (39.2–63.6) | 94.7 (82.3–99.4) | 39.3 (26.5–53.2) | 40.4 (30.4–51.0) | |
| Hürthle cell nodules (n = 29) | 9 | 19 | 1 | 0 | 100 (66.4–100) | 5.0 (0.1–24.9) | 100 (2.5–100) | 32.1 (15.9–52.4) | 3.4 (0.1–17.8) | |
| Quantitative analysis | ||||||||||
| All (n = 123) | ||||||||||
| SUVmax nodule, g/mL | 2.1 | 32 | 73 | 17 | 1 | 97.0 (84.2–99.9) | 18.9 (11.4–28.5) | 94.4 (72.7–99.9) | 30.5 (21.9–40.2) | 14.6 (8.9–22.1) |
| SUVpeak nodule, g/mL | 1.6 | 32 | 80 | 10 | 1 | 97.0 (84.2–99.9) | 11.1 (5.5–19.5) | 90.9 (58.7–99.8) | 28.6 (20.4–37.9) | 8.9 (4.5–15.4) |
| SUVmax-ratio | 1.2 | 32 | 56 | 34 | 1 | 97.0 (84.2–99.9) | 37.8 (27.8–48.6) | 97.1 (85.1–99.9) | 36.4 (26.4–47.3) | 28.5 (20.7–37.3) |
| SUVpeak-ratio | 0.9 | 32 | 75 | 15 | 1 | 97.0 (84.2–99.9) | 16.7 (9.6–26.0) | 93.8 (69.8–99.8) | 29.9 (21.4–39.5) | 13.0 (7.6–20.3) |
| Non-Hürthle cell nodules, AUS/FLUS + FN/SFNa (n = 94) | ||||||||||
| SUVmax nodule, g/mL | 2.1 | 23 | 54 | 16 | 1 | 95.8 (78.9–99.9) | 22.9 (13.7–34.4) | 94.1 (71.3–99.9) | 29.9 (20.0–41.4) | 18.1 (10.9–27.4) |
| SUVpeak nodule, g/mL | 1.6 | 23 | 61 | 9 | 1 | 95.8 (78.9–99.9) | 12.9 (6.1–23.0) | 90.0 (55.5–99.7) | 27.4 (18.2–38.2) | 10.6 (5.2–18.7) |
| SUVmax-ratio | 1.2 | 23 | 38 | 32 | 1 | 95.8 (78.9–99.9) | 45.7 (33.7–58.1) | 97.0 (84.2–99.9) | 37.7 (25.6–51.0) | 35.1 (25.5–45.6) |
| SUVpeak-ratio | 0.9 | 23 | 57 | 13 | 1 | 95.8 (78.9–99.9) | 18.6 (10.3–29.7) | 92.9 (66.1–99.8) | 28.8 (19.2–40.0) | 14.9 (8.4–23.7) |
| Hürthle cell nodules, HCN/SHCN (n = 29) | ||||||||||
| SUVmax nodule, g/mL | 5.2 | 9 | 15 | 5 | 0 | 100 (66.4–100) | 25.0 (8.7–49.1) | 100 (47.8–100) | 37.5 (18.8–59.4) | 17.2 (5.8–35.8) |
| SUVpeak nodule, g/mL | 4.7 | 9 | 13 | 7 | 0 | 100 (66.4–100) | 35.0 (15.4–59.2) | 100 (59.0–100) | 40.9 (20.7–63.6) | 24.1 (10.3–43.5) |
| SUVmax-ratio | 3.4 | 9 | 14 | 6 | 0 | 100 (66.4–100) | 30.0 (11.9–54.3) | 100 (54.1–100) | 39.1 (19.7–61.5) | 20.7 (8.0–39.7) |
| SUVpeak-ratio | 2.8 | 9 | 13 | 7 | 0 | 100 (66.4–100) | 35.0 (15.4–59.2) | 100 (59.0–100) | 40.9 (20.7–63.6) | 24.1 (10.3–43.5) |
AUS/FLUS, atypia of undetermined significance or follicular lesions of undetermined significance; CI, confidence interval; FN, false negative; FN/SFN, (suspicious for a) follicular neoplasm; FP, false positive; HCN/SHCN, (suspicious for a) Hürthle cell neoplasm; NPV, negative predictive value; PPV, positive predictive value; SUV, standardised uptake value; TN, true negative; TP, true positive
aSUV cut-offs were similar for AUS/FLUS (n = 55) and FN/SFN (n = 39) subgroups; results for these subgroups are presented in the Supplementary table 7