Comparison of predictive accuracy for pathologically organ confined clinical stage T1c prostate cancer using human glandular kallikrein 2 and prostate specific antigen combined with clinical stage and Gleason grade.

PURPOSE

Previously human glandular kallikrein 2 (hK2) has been implicated to predict pathologically organ confined prostate cancer (PCa) in patients with stage T2 disease. Now we evaluated the usefulness of hK2, as measured by 2 entirely different immunoassay designs, to enhance the discrimination of pathologically organ from nonorgan confined clinical stage T1c PCa.

MATERIALS AND METHODS

A consecutive series of pretreatment serum from 148 men with clinical stage T1c PCa was used in 2 equally sensitive and specific methods to measure total hK2 with independent reagents and entirely different assay designs. Total prostate specific antigen (tPSA) and free PSA (fPSA) were measured and percent fPSA was calculated. We determined the algorithm, hK2*tPSA/fPSA, from data generated by each hK2 assay, calculated means, medians and ranges for each analyte and algorithm, and calculated the significance of differences on univariate analysis. Using pretreatment PSA, clinical stage and biopsy Gleason grade we then developed a multivariate logistic regression base model to predict organ confined cancer and we compared predictions of the base model supplemented by the different hK2 measurements.

RESULTS

hK2 and hK2 based algorithms obtained by each hK2 assay were significantly different for pT2a/b vs pT3a or greater PCa (p = 0.034 to 0.0001) compared to tPSA (p = 0.06), fPSA (p = 0.90) or percent fPSA (p = 0.059). However, AUC (0.67 to 0.70) calculated by ROC analysis of the 4 models containing hK2 derived information was not significantly larger than that of the base model (AUC = 0.64, p = 0.52).

CONCLUSIONS

The current data confirm that hK2 alone or hK2*tPSA/fPSA measured by 2 immunoassays is significantly lower in men with pT2a/b vs pT3a or greater PCa compared to tPSA, fPSA or percent fPSA on univariate analysis of a validation set of clinical stage T1c prostate cancer treated at an American center of excellence for prostate cancer surgery. However, the incorporation of preoperative hK2 into multiparameter predictive models for pT2 cancers did not increase predictive accuracy in this cohort of men.