OBJECTIVE To compare the performance and discriminant properties of two instruments (a tree-structured regression model and a logistic regression-based nomogram), recently developed to predict lymph node invasion (LNI) at radical prostatectomy (RP), in a contemporary cohort of European patients. PATIENTS AND METHODS The cohort comprised 1525 consecutive men treated with RP and bilateral pelvic LN dissection (PLND) in two tertiary academic centres in Europe. Clinical stage, pretreatment prostate-specific antigen (PSA) level and biopsy Gleason sum were used to test the ability of the regression tree and the nomogram to predict LNI. Accuracy was quantified by the area under the receiver operating characteristic curve (AUC). All analyses were repeated for each participating institution. RESULTS The AUC for the nomogram was 81%, vs 77% for the regression tree (P = 0.007). When data were stratified according to institution, the nomogram invariably had a higher AUC than the regression tree (Hamburg cohort: nomogram 82.1% vs regression tree 77.0%, P = 0.002; Milan cohort: 82.4% vs 75.9%, respectively; P = 0.03). CONCLUSIONS Nomogram-based predictions of LNI were more accurate than those derived from a regression tree; therefore, we recommend the use of nomogram-derived predictions.
A nomogram is more accurate than a regression tree in predicting lymph node invasion in prostate cancer
Suardi N;
2008-01-01
Abstract
OBJECTIVE To compare the performance and discriminant properties of two instruments (a tree-structured regression model and a logistic regression-based nomogram), recently developed to predict lymph node invasion (LNI) at radical prostatectomy (RP), in a contemporary cohort of European patients. PATIENTS AND METHODS The cohort comprised 1525 consecutive men treated with RP and bilateral pelvic LN dissection (PLND) in two tertiary academic centres in Europe. Clinical stage, pretreatment prostate-specific antigen (PSA) level and biopsy Gleason sum were used to test the ability of the regression tree and the nomogram to predict LNI. Accuracy was quantified by the area under the receiver operating characteristic curve (AUC). All analyses were repeated for each participating institution. RESULTS The AUC for the nomogram was 81%, vs 77% for the regression tree (P = 0.007). When data were stratified according to institution, the nomogram invariably had a higher AUC than the regression tree (Hamburg cohort: nomogram 82.1% vs regression tree 77.0%, P = 0.002; Milan cohort: 82.4% vs 75.9%, respectively; P = 0.03). CONCLUSIONS Nomogram-based predictions of LNI were more accurate than those derived from a regression tree; therefore, we recommend the use of nomogram-derived predictions.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.