Table 4.
Meta-analysis of HPV16 methylation pooled sensitivity, specificity and area under the curve (by study defined cut-off) in detecting disease ≥CIN2/HSIL and disease ≥CIN3/HSIL.
| Gene | Studies (N)a | Threshold (%) | Pooled sensitivity % (95% CI) | Pooled specificity % (95% CI) | Pooled estimated AUC (95% CI) | I2 (95% CI) | tau2 (95% CI) |
|---|---|---|---|---|---|---|---|
| BIVARIATE MODEL | |||||||
| ≥CIN2/HSIL | |||||||
| All studies (L1/L2/E2//E5/E6/E7/LCR) | 11 | 0.4–15.9/3rd tertile | 75 (59–86) | 65 (57–73) | 0.73 (0.69–0.76) | 95.1 (93–97) | 0.76 (0.00–1.89) |
| L1/L2/E5 | 11 | 0.4–15.9/3rd tertile | 76 (63–86) | 64 (56–72) | 0.73 (0.69–0.77) | 95.0 (93–97) | 0.43 (0.00–1.10) |
| L1/L2 | 10 | 0.4–15.9/3rd tertile | 77 (63–87) | 64 (55–71) | 0.73 (0.69–0.76) | 95.5 (94–97) | 0.00 (0.00–1.34) |
| L1 | 10 | 0.4–15.9/3rd tertile | 82 (73–88) | 57 (49–65) | 0.73 (0.69–0.77) | 85.9 (78–93) | 0.00 (0.00–2.99) |
| L2 | 5 | 0.4–15.9/3rd tertile | 75 (63–84) | 66 (42–84) | 0.77 (0.73–0.81) | 32.9 (0–98) | 0.00 (0.00–100) |
| L1 6367 | 5 | 3.1–10/3rd tertile | 74 (64–82) | 53 (41–66) | 0.70 (0.66–0.74) | 73.0 (48–98) | 0.35(0.00–0.98) |
| L1 6389 | 5 | 3.1–10/3rd tertile | 77 (67–84) | 52 (38–65) | 0.71 (0.67–0.75) | 67.9 (38–98) | 0.00 (0.00–1.64) |
| UNIVARIATE MODEL | |||||||
| ≥CIN2/HSIL | |||||||
| All studies (L1/L2/E2//E5/E6/E7/LCR) | 14 | 0.4–15.9/3rd tertile | 74 (60–84) | 65 (54–75) | 0.782 (0.669–0.864) | 39.5 (0–67) | 0.50 (0.00–1.58) |
| L1/L2/E5 | 14 | 0.4–15.9/3rd tertile | 75 (62–84) | 64 (54–74) | 0.810 (0.712–0.880) | 37.4 (0–61.7) | 0.46(0.00–1.24) |
| L1/L2 | 13 | 0.4–15.9/3rd tertile | 75 (61–85) | 64 (53–74) | 0.808 (0.685–0.891) | 39.7 (0.00–67.14) | 0.49 (0.00–1.49) |
| L1 | 13 | 0.4–15.9/3rd tertile | 81 (73–87) | 55 (44–66) | 0.808 (0.685–0.891) | 0.00 (0.00–70.7) | 0.00 (0.00–2.62) |
| L2 | 5 | 0.4–15.9/3rd tertile | 81 (69–89) | 69 (46–85) | – | 0.00 (0.00–100) | 0.00 (0.00–100) |
| LCR | 3 | 0/3rd tertile | 45.3 (19–73.4) | 75.2(47.6–91.0) | – | 0.00 (0.00–100) | 0.00 (0.00–100) |
| LCR/L1/L2 | 3 | 5.3/3rd tertile | 54.6 (26.1–80.4) | 74.2 (45.8–90.7 | – | 0.00 (0.00–100) | 0.00 (0.00–100) |
| L1 6367 | 5 | 3.1–10/3rd tertile | 75 (62–85) | 52 (37–67) | 0.720 (0.665–0.769) | 89.9 (68–99) | 0.00 (0.00–1.45) |
| L1 6389 | 5 | 3.1–10/3rd tertile | 82 (68–91) | 48 (33–63) | 0.834 (0.691–0.918) | 98.9 (97–100) | 0.18 90.00–6.58) |
| L1 5611 | 3 | 10.8/3rd tertile | – | – | 0.830 (0.739–0.894) | 96.4 (87–100) | 0.22(0.05–9.24) |
| L1 6650 | 3 | 10.8/3rd tertile | – | – | 0.779 (0.653–0.868) | 97.8 (92–100) | 0.30 (0.08–12.06) |
| L1 7145 | 5 | 10.8/3rd tertile | – | – | 0.729 (0.649–0.796) | 95.6 (88–100) | 0.17 (0.06—1.55) |
| ≥CIN3/HSIL | |||||||
| L1 | 7 | 3.1–15.9 | 76 (66–84) | 75 (57–87) | 0.903 (0.804–0.955) | 0.00 (0.00–58.6) | 0.00 (0.00–0.77) |
CIN: cervical intraepithelial neoplasia; WGA: Whole Genome Average.
Bivariate models applied wherever possible in preference. When not possible, univariate model were applied.
a Number of studies and heterogeneity refers to pooled sensitivity meta-analyses, on occasion the number of studies for AUC meta-analyses was less due to less available data.