Table 3. Estimated odds ratios for prediction of prostate cancer status from logistic regression models evaluating the effect of risk groups and cutoffs for PSA, fcDNA and GADD45a methylation obtained from RPART analysis.
|
Risk groups defined by univariate analysis of cutoffs for PSA, fcDNA and GADD45a methylation (sum of four CpGs) |
|||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Risk group | PSA | fcDNA | GADD45a methylation | Benign (N) | Cancer (N) | Total (N) | Prediction |
Models |
|||
| OR (95% CI) | P | AUC | |||||||||
| 1 | <16.9 | — | — | 48 | 21 | 69 | Benign | Reference | 0.691 | ||
| 2 | ⩾16.9 | — | — | 0 | 13 | 13 | Case | 60.9 (3.11, >999) | 0.007 | ||
| 1 | — | <188 | — | 43 | 17 | 60 | Benign | Reference | 0.698 | ||
| 2 | — | ⩾188 | — | 5 | 17 | 22 | Case | 8.6 (2.7, 27.0) | 0.0002 | ||
| 1 | — | — | <89.6 | 47 | 21 | 68 | Benign | Reference | 0.681 | ||
| 2 | — | — | ⩾89.6 | 1 | 13 | 14 | Case | 29.1 (3.57, 237.1) | 0.002 | ||
|
Risk groups defined by PSA, fcDNA and GADD45a methylation (sum of four CpGs) | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Risk group | PSA | fcDNA | GADD45a methylation | Benign (N) | Cancer (N) | Total (N) | Prediction |
Model 1 (AUC=0.937) |
Model 2 (AUC=0.952) |
||
| OR (95% CI) | P | OR (95% CI) | P | ||||||||
| 1 | <16.9 | <188 | <89.6 | 42 | 2 | 44 | Benign | Reference | Reference | ||
| 2 | <16.9 | <188 | ⩾89.6 | 1 | 6 | 7 | Case | 73.7 (7.4, 731) | <0.001 | 77.8 (6.7, 902) | <0.001 |
| 3 | <16.9 | ⩾188 | — | 5 | 13 | 18 | Case | 41.7 (8.1, 215.9) | <0.0001 | 36.4 (8.1, 188) | <0.0001 |
| 4 | ⩾16.9 | — | — | 0 | 13 | 13 | Case | 458.9 (18.7, >999) | <0.001 | 348.7 (18.7,>999) | <0.001 |
Abbreviations: OR=odds ratio; CI=confidence interval.
Model 1: effect of the composite variable risk group. Model 2: effect of the composite variable risk group adjusted for race (Black vs White; OR=0.57, P=0.636), ethnicity (non-Hispanic vs Hispanic; OR=4.25, P=0.262) and age (1-year increase; OR=0.97, P=0.607).