Skip to main content
. Author manuscript; available in PMC: 2012 Mar 1.
Published in final edited form as: Cancer Prev Res (Phila). 2011 Mar;4(3):375–383. doi: 10.1158/1940-6207.CAPR-10-0193

Table 4.

Results of models from Steps 1 to 3

Model Sensitivity*
≤ 12 mo
% (95% CI)
Sensitivity*
13–24 mo
% (95% CI)
Specificity*
% (95% CI)
ROC*
Area (95% CI)
Sensitivity at 98%
Specificity
% (95% CI)
Step 1 N=67 N=26 N=476 N=67
A1 34.3 (23–46) 7.7 (1–25) 96.8 (95.2–98.4) 0.721 (0.64–0.80) 32.8 (22–44)
B1 69.2 (58–80) 12.5 (3–31) 96.6 (94.9–98.3) 0.892 (0.84–0.95) 64.6 (53–76)
C1 34.3 (23–46) 11.5 (2–30) 95.1 (93.1–97.1) 0.712 (0.63–0.79) 25.4 (15–36)
D1 95.4 (90–99) 76.0 (59–93) 32.2 (27.4–36.5) 0.858 (0.80–0.92) 52.3 (40–64)
E1 37.9 (26–50) 3.9 (0–20) 89.8 (87.0–92.6) N/A N/A
CA125 63.1 (51–75) 0.0 (0–13) 98.5 (97.4–99.6) 0.890 (0.84–0.94) 64.6 (53–76)

Step 2 †† N=30 N=15 N=237 N=30
A2 53.3 (35–71) 6.7 (0–32) 96.6(94.3–98.8) 0.852 (0.77–0.94) 36.7 (20–54)
B2 80.0 (66–94) 21.4 (5–50) 92.2(88.7–95.7) N/A N/A
C2 70.0 (54–86) 6.7 (0–32) 91.9 (88.4–95.4) 0.848 (0.76–0.94) 46.7 (29–64)
D2 55.2 (37–73) 0.0 (0–22) 86.9 (82.5–91.3) 0.810 (0.72–0.90) 51.7 (34–69)
E2 30.0 (14–46) 13.3(2–40) 96.2 (93.7–98.7) 0.590 (0.46–0.72) 23.3 (8–38)
CA125 72.4 (56–89) 0.0 (0–22) 97.9 (96.0–99.8) 0.898 (0.82–0.98) 72.4 (56–89)

Step3 (Pan-site) N/A§ N/A§ N/A§ 0.911 (0.86–0.96) 68.2 (57–80)
*

Calculated based on cut-off specified by the model. Note that we included all general population controls in the calculation of specificity and ROC; these as a whole were comparable to the one year cases with respect to the matching variables of age and calendar year of blood draw.

Calculated based on the 67 cases diagnosed ≤12 months from blood draw.

This model did not produce a propensity score, thus the measure cannot be calculated.

Using data previously obtained in PLCO for the same subject and study year as the samples in the current study, and a cutoff of ≥35 U/ml.

††

For Step 2 model, data from the validation set is shown.

§

The pan-site model did not have a cutoff, therefore no sensitivity or specificity can be calculated.