Table 5.
A logistic regression model based on informative fusion markers [Fx (III, IV, ETS)], PCA3, and prebiopsy serum PSA to predict cancer risk on biopsy
| Univariable logistic regression models | Multivariable logistic regression model1 | |||||||
| Biopsy cohort (n) | Dependent variable | Diagnostic Variable | OR (95% CI) | P | Overall accuracy (%) | OR (95% CI) | P | Overall accuracy (%)2 |
| Fx (III, IV, ETS)3 | 10.11 (3.32–30.81) | <0.0001 | 73.9 | 7.10 (2.20–22.89) | 0.001 | |||
| 92 | Biopsy outcome | PCA34 | 1.34 (1.11–1.60) | 0.002 | 65.2 | 1.22 (1.00–1.49) | 0.045 | 77.25 |
| Serum PSA4 | 1.63 (0.96–2.76) | 0.07 | 58.7 | 1.58 (0.89–2.77) | 0.116 | |||
PSA, prostate-specific antigen.
1
Hosmer–Lemeshow Goodness-of-Fit of logistic regression model: P = 0.307.
2
Defined as (true positives + true negatives)/all.
3
TMP:ERG subtype III or IV, or TMP:ETS (ETV 1, 4, or 5) (binary categorical variable).
4
PCA3 and serum PSA were log-transformed continuous variables.
5
61.5% sensitivity and 88.7% specificity at 50% cut-off value.