Patient selection: men with serum PSA <= 10 ng/mL
Men with prostate carcinoma represented almost 25% of the study population.
Parameters:
(1) serum PSA level in ng/mL
(2) PSA density (serum PSA divided by prostate volume in mL or cubic cm)
(3) transrectal ultrasound (TRUS)
(4) age in years
(5) prostate volume in cubic cm
prostate volume in cubic cm =
= 0.52 * (length in cm) * (width in cm) * (height in cm)
PSA density =
= (serum PSA in ng/mL) / (prostate volume in mL)
Criteria for a male at high risk:
(1) PSA > 1.55 ng/mL, PSA density > 0.165
(2) PSA > 1.55 ng/mL, PSA density <= 0.165, hypoechoic areas on TRUS
(3) PSA > 1.55 ng/mL, PSA density <= 0.165, no hypoechoic areas on TRUS, age > 55 years, volume <= 44 mL
(4) PSA > 1.55 ng/mL, PSA density <= 0.165, no hypoechoic areas on TRUS, age > 55 years, volume 50.26 to 80.8 mL
where:
• The results of using prostate volume to determine risk causes me some concern. A volume of <= 44 mL is associated with high risk, a volume of 44.01 to 50.25 mL with low risk, 50.26 to 80.8 high risk, and > 80.8 as low risk.
• There probably are some interesting interactions between the volume and the PSA density. On a preliminary look st the combinations some of the patients in Groups 3 or 4 have PSA > 5.
|
Group
|
Percent of Patients in Group
|
Percent of Patients with Prostate Cancer
|
|
1
|
49%
|
58%
|
|
2
|
26%
|
22%
|
|
3
|
22%
|
9.2%
|
|
4
|
17%
|
6.5%
|
Performance:
• In the validation set the sensitivity was 93% with a specificity of 31%.
• The area under the ROC curve was 0.72, which is nothing to brag about.
