EURURO Vol. 72 No. 6

Trial Risk Calculator (PCPTRC) 2.0 risk calculator

[17]

, providing risk

estimates for low-grade versus high-grade cancer with PSA, age, DRE,

race, and prior biopsy as parameters. Performance parameters including

specificity, sensitivity, negative predictive value (NPV), and positive

predictive value (PPV) were examined, and their clinical consequences

for each of the two indications were recorded in terms of biopsies

avoided as defined according to selected cutoff values. Statistical

analysis was conducted using either Analyse-It for Microsoft Excel

v.4.65.2 (Analyse-it Software, Leeds, UK) or Stata/MP 13.1 for Windows

(StataCorp, College Station, TX, USA).

Results

The overall prevalence of PCa in the entire cohort was 53.3%

(139/261), while the prevalence of high-grade PCa was 33.7%

(88/261). As expected, there was no significant correlation

between IsoPSA

and serum PSA levels

( Fig. 1 ;

Pearson

correlation coefficient 0.2). As measured by ROC, IsoPSA

outperformed standard PSA for both study endpoints

( Table 2

and

Fig. 2

). For the cancer versus no cancer endpoint, the AUC

was 0.79 (95% CI 0.73–0.84) for IsoPSA versus 0.61 (95% CI

0.54–0.67) for total PSA (

0.001). For high-grade cancer

versus low-grade cancer/benign histology, the AUC was 0.81

(95% CI 0.74–0.86) for IsoPSA versus 0.69 (95% CI 0.61–0.75)

for total PSA (

0.005). Calibration curves for both end-

points demonstrated very good correspondence between

predicted and observed results, with a Hosmer-Lemeshow

statistic of 8.21 (

= 0.41,

0.05) for cancer versus no cancer

and 9.86 (

= 0.28,

0.05) for high-grade cancer versus low-

grade cancer/benign histology

( Fig. 2

Table 3

shows the

sensitivity, specificity, PPV, and NPV for IsoPSA versus total

PSA at selected cutoff values for risk probability for the

endpoints cancer versus no cancer (KR-CNC) and high-grade

PCa versus low-grade PCa/benign histology (KR-HG). Recog-

nizing that the decision to perform or forgo a biopsy depends

on clinical suspicion and both patient and physician risk

tolerance,

Table 3

suggests some clinically relevant scenarios

illustrating the superior predictive power of IsoPSA. For

example, as an exclusion test for separating PCa from benign

disease, a risk probability cutoff value of KR-CNC = 35% for

IsoPSA provides a good balance of high sensitivity (90%) and

specificity (48%), whereas at similar sensitivity (87%)

standard PSA (at 4 ng/ml) has significantly inferior specificity

(15%). As an exclusion test to identify patients at high risk of

high-grade PCa, a KR-HG cutoff of 17% yields NPV of 96%,

while KR-HG of 70% yields PPV of 76%.

DCA also demonstrated superior performance for IsoPSA

compared to the modified PCPTRC 2.0 risk calculator. For

expression as net benefit,

Fig. 3

illustrates DCA results for a

relevant risk probability range for the two study endpoints.

If the goal is to identify all patients with cancer of any grade,

KR-CNC = 35% results in a 48% reduction in unneeded

biopsies (from 122 to 63). For the objective of identifying

Table 3 – Performance metrics versus PSA at selected cutoff values for risk probability

Cancer vs no cancer

High-grade PCa vs low-grade PCa/BH

Total PSA

KR-CNC

Total PSA

IsoPSA KR-HG

Low RP

High RP

Prevalence (%)

Cutoff

4 ng/ml

35%

4 ng/ml

17%

70%

Sensitivity (%)

Specificity (%)

NPV (%)

PPV (%)

AUC

0.61

0.79

0.69

0.81

PSA = prostate-specific antigen; PCa = prostate cancer; BH = benign histology; KR-CNC = IsoPSA

result for cancer versus no cancer; KR-HG = IsoPSA

result for

high-grade cancer (Gleason 7) versus benign/low-grade cancer (Gleason 6); RP = risk probability; NPV = negative predictive value; PPV = positive predictive

value; AUC = area under the receiver operating characteristic curve.

[(Fig._1)TD$FIG]

Fig. 1 – Serum prostate-specific antigen (PSA) concentration versus

IsoPSA

value for all patients.

Table 2 – Area under the ROC curve for IsoPSA versus total PSA for

high-grade cancer versus low-grade cancer/benign histology and

for cancer versus no cancer

Model

Area under the ROC curve (95% CI)

valu

e a

IsoPSA K

Total PSA

KR-HG

0.81 (0.74–0.86)

0.69 (0.61–0.75)

0.005

KR-CNC

0.79 (0.73–0.84)

0.61 (0.54–0.67)

0.001

KR-CNC = IsoPSA

result for cancer versus no cancer; KR-HG = IsoPSA

result for high-grade cancer (Gleason 7) versus benign/low-grade cancer;

PSA = prostate-specific antigen; ROC = receiver operating characteristic.

Delong and Delong test.

E U R O P E A N U R O L O G Y 7 2 ( 2 0 1 7 ) 9 4 2 – 9 4 9

945