Standardized processes should be used in the identification and development
of intermediate endpoint biomarkers (IEB) for the prediction of patient-sp
ecific disease outcomes, Using our own experiences, we outline some of our
standardized processes, Using computer-assisted image analysis, we develope
d a new biomarker of genetic instability, termed quantitative nuclear grade
(QNG). The QNG biomarker is derived using nuclear images analyzed from the
tumor areas of Feulgen-stained 5-mum biopsy or radical prostatectomy tissu
e sections. From the variances of 41 to 60 different nuclear size, shape, a
nd chromatin organization features, a QNG solution is computed using either
logistic regression or artificial neural networks. QNG can then be used as
an input for models that solve for a patient-specific probability to accur
ately predict disease outcomes. Preoperatively, QNG predicted both the path
ologic stage and progression of prostate cancer using biopsies (P <0.0001).
Postoperatively, QNG proved extremely valuable in the prediction of bioche
mical progression using radical prostatectomy specimens with more than 10 y
ears of follow-up (P <0.0001). We also demonstrate the identification of no
vel, differentially expressed, prostate cancer genes using RNA fingerprinti
ng methods and the clinical utility of testing for these genes in both bloo
d and tissue samples. Also illustrated is the improvement of serum biomarke
r performance by combining molecular forms of PSA with new biomarkers. In c
onclusion, the development of new IEBs requires planning based upon an unde
rstanding of the molecular pathogenesis of disease. IEB selection and clini
cal evaluation should employ standardized methods of testing and validation
, followed by publication. QNG is 1 example of a new, highly predictive, IE
B for prostate cancer that has been developed using these processes.