E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6

ava i lable at www.sciencedirect .com

journal homepage: www.europeanurology.com

Platinum Priority – Prostate CancerEditorial by Andrew J. Vickers and Hans Lilja on pp. 467–468 of this issue

Preoperative Prostate-Specific Antigen Isoform p2PSA and Its

Derivatives, %p2PSA and Prostate Health Index, Predict Pathologic

Outcomes in Patients Undergoing Radical Prostatectomy for

Prostate Cancer

Giorgio Guazzoni a, Massimo Lazzeri a,*, Luciano Nava b, Giovanni Lughezzani a,Alessandro Larcher a, Vincenzo Scattoni c, Giulio Maria Gadda a, Vittorio Bini d, Andrea Cestari a,Nicolo Maria Buffi a, Massimo Freschi e, Patrizio Rigatti c, Francesco Montorsi c

a Department of Urology, San Raffaele Turro, Vita-Salute San Raffaele University, Milan, Italy; b Department of Urology, Fondazione ‘‘Opera S. Camillo’’, Casa

di Cura S. Pio X, Milan, Italy; c Department of Urology, Vita-Salute San Raffaele University, Milan, Italy; d Department of Internal Medicine, University of

Perugia, Italy; e Department of Pathology, Vita-Salute San Raffaele University, Milan, Italy

Article info

Article history:Accepted October 26, 2011Published online ahead ofprint on November 4, 2011

Keywords:

Prostate-specific antigen

[�2]proPSA

Prostate health index

Prostate cancer

Radical prostatectomy

Pathologic examination

Predictive models

Abstract

Background: Currently available predictive models fail to assist clinical decision makingin prostate cancer (PCa) patients who are possible candidates for radical prostatectomy(RP). New biomarkers would be welcome.Objective: Test the hypothesis that prostate-specific antigen (PSA) isoform p2PSA andits derivates, percentage of p2PSA to free PSA (%p2PSA) and the Prostate Health Index(PHI), predict PCa characteristics at final pathology after RP.Design, setting, and participants: An observational prospective study was performed in350 consecutive men diagnosed with clinically localised PCa who underwent RP.Measurements: We determined the predictive accuracy of serum total PSA (tPSA), freePSA (fPSA), fPSA-to-tPSA ratio (%fPSA), p2PSA, %p2PSA, and PHI. The primary end pointwas to determine the accuracy of these biomarkers in predicting the presence of pT3disease, pathologic Gleason sum�7, Gleason sum upgrading, and tumour volume<0.5 ml.Intervention: Open retropubic and robot-assisted laparoscopic RP was performed.Pelvic lymphadenectomy was performed according to baseline oncologic parametersand the surgeon’s judgement.Results and limitations: The %p2PSA and PHI levels were significantly higher in patientswith pT3 disease, pathologic Gleason sum �7, and Gleason sum upgrading (all p values<0.001). Conversely, %p2PSA and PHI levels were significantly lower in patients withtumour volume <0.5 ml ( p < 0.001). By univariate analysis, both %p2PSA and PHI wereaccurate predictors of pT3 disease, pathologic Gleason sum �7, Gleason sum upgrading,and tumour volume<0.5 ml. By multivariate analyses, the inclusion of both %p2PSA andPHI significantly increased the predictive accuracy of a base multivariate model(excluding the tumour volume prediction for both variables, and Gleason sum upgradingfor the model including %p2PSA) that included patient age, tPSA, fPSA, f/tPSA, clinicalstage, and biopsy Gleason sum.Conclusions: We found that p2PSA and its derivatives are predictors of PCa character-istics at final pathology after RP and are more accurate than currently available markers.

soc

Department of Urology, San Raffaele Turro, Vita-Salute San Raffaele University,, 20127 Milan, Italy. Tel. +39 022643.3357; Fax: +39 0226433442.

# 2011 European As

* Corresponding author.Via Stamira D’Ancona 20

E-mail address: lazzeri.max

0302-2838/$ – see back matter # 2011 European Association of Urology. Publis

iation of Urology. Published by Elsevier B.V. All rights reserved.

imus@gmail.com (M. Lazzeri).

hed by Elsevier B.V. All rights reserved. doi:10.1016/j.eururo.2011.10.038

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6456

1. Introduction

In the setting of prostate cancer (PCa) screening with

prostate-specific antigen (PSA), an increasing number of

men are identified with low-stage and low-grade disease.

Currently available treatments for men with localised PCa

include active surveillance, radical prostatectomy (RP),

radiation therapy, or focal therapy [1]. Ideally, the best

treatment should provide the patient with perfect oncologic

and functional outcomes. To date, established clinical

parameters used in the PCa setting, such as PSA, digital

rectal examination (DRE), or biopsy Gleason sum, fail to

accurately predict PCa aggressiveness. Consequently, sev-

eral predictive and prognostic tools (eg, lookup tables,

classification and regression-tree analyses, artificial neural

networks, and nomograms) have been developed to assist

physicians in the clinical decision-making process [2].

However, currently available models remain imperfect in

their predictive ability, and the use of new molecular

biomarkers is required to reduce their error margin.

Although several biomarkers recently have been investi-

gated with promising results, their actual impact on clinical

practice appears to be negligible [3].

Recent studies have shown that PSA isoform [�2]proPSA

(p2PSA) and its derivatives, percentage of p2PSA to free PSA

(%p2PSA; [(p2PSA pg/ml) / (free PSA ng/ml � 1000)] � 100)

and the Prostate Health Index (PHI; [p2PSA / free PSA] �HtPSA) (Beckman Coulter; Brea, CA, USA), improve the

accuracy of total PSA (tPSA) and percentage of free PSA

(%fPSA) in predicting the presence of PCa at prostate biopsy,

and they are also related to PCa aggressiveness at biopsy

[4–9]. However, to our knowledge, no study has yet

investigated the role of p2PSA and its derivatives in the

prediction of PCa aggressiveness at final pathology after RP.

The aim of this study was to test the hypothesis that these

biomarkers predict the pathologic PCa characteristics

within a prospectively collected contemporary cohort of

patients who underwent RP for clinically localised PCa at a

single high-volume institution.

2. Materials and methods

2.1. Patient population and clinical evaluation

Our observational descriptive study cohort consisted of 350 consecutive

patients with biopsy-proven, clinically localised PCa, who were

prospectively recruited between June 2010 and April 2011 and who

underwent either open or robot-assisted laparoscopic RP at our

institution. Extended lymph node dissection was performed in

290 patients following baseline oncologic parameters and the decision

of the surgeon. None of the patients included in the current study

received neoadjuvant androgen-deprivation therapy or drugs that could

alter the PSA values, such as dutasteride and finasteride. Patients with

marked blood protein alterations (plasma normal range: 6–8 g/100 ml),

who were suffering from haemophilia, or who had been previously

polytransfused were excluded from the study because these conditions

may alter the concentration of p2PSA. The hospital ethics committee

approved the study, and all patients signed written informed consent.

The primary end point of the study was to determine the accuracy of

%p2PSA and PHI in predicting the presence of extracapsular disease (pT3),

the presence of pathologic Gleason sum �7 PCa, the presence of Gleason

sum upgrading (defined as an upgrading of Gleason sum from �6 to �7),

and the presence of tumour volume<0.5 ml. P2PSA, %p2PSA, and PHI were

considered the index tests and were compared with the established

biomarker reference standard tests: tPSA, free PSA (fPSA), and %fPSA.

A blood sample was drawn on the day before surgery and prior to any

manipulations (eg, DRE, enema) that might cause a transient increase of

biomarkers. Blood samples were processed with the UniCel DxI800

Immunoassay System analyser (Beckman Coulter Inc., Brea, CA, USA) and

were managed according to the criteria described by Semjonow et al.

[10]. Blood sample analysis was performed using the Hybritech

calibrated Access tPSA (Beckman Coulter; Brea, CA, USA) and fPSA

assay. RP specimens were processed and evaluated according to the

Stanford protocol [11] by a single, experienced, genitourinary patholo-

gist (MF) blinded to index-tests results. PCa was identified and graded

according to the definitions of the 2005 consensus conference of the

International Society of Urological Pathology [12].

2.2. Statistical analysis

The Kolmogorov-Smirnov test was used to assess the normal distribution

of variables. The Mann-Whitney U test was used for comparisons of ordinal

and non-normally distributed continuous variables. Spearman’s rho

coefficient analysis was used to test correlations between continuous

and ordinal variables. Bivariate and multivariate logistic regression

models were fitted for the prediction of the presence of pT3, the presence

of pathologic Gleason sum�7 PCa, the presence of Gleason sum upgrading,

and the presence of tumour volume <0.5 ml. Logistic regression model

goodness-of-fit was checked using the Hosmer and Lemeshow test [13].

Odds ratios (ORs) with 95% confidence intervals were also calculated.

Multivariate logistic regression models were complemented by

predictive accuracy tests. Predictive accuracy was quantified as the area

under the receiver operating characteristics curve (AUC), where a value

of 100% indicated perfect prediction and 50% was equivalent to a toss of a

coin. To test the ability of %p2PSA and PHI in determining the four

outcomes of interest, these variables were added to the base multivariate

model. The gain in predictive accuracy was quantified and AUCs were

compared using the DeLong method [13]. To evaluate whether

incorporating %p2PSA and PHI levels into the statistical models

improved the accuracy of the prognostic model and the consequent

clinical management of patient, we used decision-curve analysis (DCA)

[14]. Decision curves are constructed by plotting net benefit against

threshold probability. In our study, this analysis estimates the

magnitude of benefit resulting from altering clinical management

in patients with different threshold probabilities of PCa. The threshold

probability is the minimum probability of outcome of interest at which

a patient (or clinician) would opt for intervention. As the threshold

probabilities can vary from patient to patient, the net benefit is

calculated across a range of probabilities.

All calculations were carried out with Predictive Analytic Software

v.17.0.2 (IBM Corp., Armonk, NY, USA). A two-sided p value of <0.05 was

considered significant. Decision curves were plotted by using a macro

developed in Excel (Microsoft, Redmond, WA, USA) by one of the authors

(VB).

3. Results

The descriptive characteristics of the study population are

shown in Table 1. Mean patient age was 64.6 yr (range:

42–80). Preoperative median tPSA, fPSA, and fPSA-to-tPSA

ratio (f/tPSA) values were, respectively, 5.89 ng/ml (range:

0.64–19.79), 0.82 ng/ml (range: 0.09–5.63), and 13% (range:

0.2–44). Median p2PSA, %p2PSA, and PHI values were,

Table 1 – Patient characteristics and descriptive statistics

Preoperative variables n = 350 patients Pathology variables n = 350 patients

Age, yr

Mean (median)

Range

BMI, absolute value

Mean (median)

Range

tPSA, ng/ml

Mean (median)

Range

fPSA, ng/ml

Mean (median)

Range

f/tPSA, ng/ml

Mean (median)

Range

p2PSA, pg/ml

Mean (median)

Range

%p2PSA, ratio

Mean (median)

Range

PHI, absolute value

Mean (median)

Range

Biopsy Gleason sum, no. (%)

�6

=7

�8

Clinical stage, no. (%)

cT1c

cT2

64.6 (65)

42–80

25.9 (25.4)

19.8–40.1

6.95 (5.89)

0.64–19.79

0.95 (0.82)

0.09–5.63

0.143 (0.134)

0.002–0.441

20.06 (14.73)

1.62–212.76

2.12 (1.96)

0.60–6.40

54.42 (47.44)

10.95–195.42

187 (53.4)

130 (37.2)

33 (9.4)

302 (86.3)

48 (13.7)

Gleason sum, no. (%)

�6

=7

�8

Tumor stage, no. (%)

pT2a,b

pT2c

pT3a

pT3b

Nodal stage, no. (%)

pNx

pN0

pN1

Surgical margins, no. (%)

R0

R1

Tumor, %

Mean (median)

Range

Tumor volume, ml

Mean (median)

Range

Tumor volume, ml (%)

<0.5

no. (GS)

>0.5

Lobe involvement, no. (%)

Monolateral

Bilateral

Prostate weight, g

Mean (median)

Range

89 (25.4)

220 (62.9)

41 (11.7)

22 (6.3)

220 (62.9)

69 (19.7)

39 (11.1)

60 (17.1)

250 (71.4)

40 (11.4)

275 (78.6)

75 (21.4)

10.1 (7.0)

1–80.0

4.40 (2.59)

0.20–69.00

26 (7.4)

4 (3+4)

18 (3+3)

4 (3+2)

324 (92.6)

22 (6.3)

328 (93.7)

53.9 (50.0)

19.0–130.0

BMI = body mass index; PSA = prostate-specific antigen; tPSA = total PSA; fPSA = free PSA; f/tPSA = ratio of fPSA to tPSA; %p2PSA = percentage of p2PSA to fPSA;

PHI = Prostate Health Index; GS = Gleason score.

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6 457

respectively, 14.73 pg/ml (range: 1.62–212.76), 1.96%

(range: 0.6–6.4), and 47.44 (range: 10.95–195.42).

Overall, 261 patients (74.6%) were diagnosed with

pathologic Gleason sum �7 PCa. One hundred and eight

(30.8%) and 40 (11.4%) patients were diagnosed with pT3

and pN1 disease, respectively.

Tumour volume <0.5 ml was observed in 26 patients

(7.4%). Finally, of 172 patients with biopsy Gleason sum�6,

99 (57.5%) showed an upgrading of the Gleason sum,

defined as Gleason sum �7 at final pathology.

As shown in Figure 1, both %p2PSA and PHI levels were

significantly higher in patients with pT3 disease, pathologic

Gleason sum �7, and Gleason sum upgrading (all p values

<0.001). Conversely, %p2PSA and PHI levels were signifi-

cantly lower in patients with tumour volume <0.5 ml

( p < 0.001).

Spearman’s rho (r) analysis showed that both %p2PSA

( p = 0.231) and PHI ( p = 0.487) were not significantly

related to patient age. In addition, PHI showed no

correlation with prostate volume ( p = 0.203).

On bivariate analyses, %p2PSA and PHI were accurate

predictors of the presence of pT3 disease, pathologic

Gleason sum �7, Gleason sum upgrading, and tumour

volume <0.5 ml.

Predictive accuracy was quantified as the AUC for each

outcome of interest and different cut-offs at different levels

of sensitivity and specificity were reported (Table 2a–2d).

The multivariate logistic regression models predicting

pT3 disease (Table 3), pathologic Gleason sum �7 (Table 4),

Gleason sum upgrading (Table 5), and tumour volume

<0.5 ml (Table 6) showed that both %p2PSA and PHI

achieved independent predictor status (all p values�0.001).

[(Fig._1)TD$FIG]

Fig. 1 – Percentage of p2PSA to free PSA (%p2PSA) and Prostate Health Index (PHI), each relative to tumour stage, Gleason sum, Gleason upgrading, andtumour volume.

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6458

On multivariate analyses, the inclusion of both %p2PSA and

PHI significantly increased the predictive accuracy of a base

multivariate model (excluding the tumour volume predic-

tion for both variables and Gleason sum upgrading for

model including %p2PSA) that included patient age, tPSA,

fPSA, f/tPSA, clinical stage (cT1c vs cT2), and biopsy Gleason

sum (in prediction models of pT3 and tumour volume).

The predictive accuracy increase varied from 2.4% to 6.0%

when considering the four different outcomes. For example,

when predicting Gleason sum�7 at final pathology, including

%p2PSA or PHI in the multivariable model significantly

increased its accuracy from 72.3% to 78.3% for both predictors.

Figure 2–5 present the decision curve analysis for the

models shown in Table 3–6. Models including %p2PSA and

PHI clearly result in greater net benefit in pathologic

outcome probability (except in pT3 prediction) when it is

plotted against various threshold probabilities.

4. Discussion

In the current study, we investigated the relationship

between p2PSA and its derivatives, namely %p2PSA and PHI,

and PCa characteristics at final pathology in a contemporary

population of patients treated with RP for clinically

localised PCa. The results of the study supported the

hypothesis that p2PSA, and in particular its derivatives,

%p2PSA and PHI, may predict the final pathologic outcomes.

On univariate analyses, %p2PSA and PHI emerged as the

most accurate predictors of several pathologic disease

characteristics, namely, the pT3 disease, pathologic Gleason

sum �7, Gleason sum upgrading, and tumour volume

<0.5 ml. In addition, both %p2PSA and PHI achieved

independent predictor status when considering all four

pathologic end points (all p values <0.001). Multivariate

analyses revealed that the inclusion of %p2PSA or PHI in

multivariate models increased their accuracy in predicting

the four pathologic outcomes from 2.4% to 6%, although it

did not reach the level of statistical significance for tumour

volume (both %p2PSA and PHI) and for Gleason sum

upgrading (only %p2PSA). Our current pathology-based

results suggest that %p2PSA and PHI significantly discrimi-

nate men with pathologic Gleason sum �7 and predict

Gleason sum upgrading. %p2PSA and PHI might be used to

stratify the risk of harbouring clinically insignificant or

Table 2a – Sensitivity and specificity at three levels of predictive variables (high sensitivity, best combination, high specificity) for predictionof PT3 disease at pathologic staging

Criterion Sensitivity, % 95% CI Specificity, % 95% CI

Total PSA, ng/ml

�3.3 90.7 83.6–95.5 14.0 9.9–19.1

�6.2 64.8 55.0–73.8 62.4 56.0–68.5

�11.5 22.2 14.8–31.2 91.7 87.5–94.9

Free PSA, ng/ml

�0.37 92.6 85.9–96.7 11.6 7.8–16.3

�0.83 52.8 42.9–62.5 55 48.5–61.3

�1.48 18.5 11.7–27.1 90.5 86.1–93.9

%fPSA

�0.20 90.7 83.6–95.5 19.4 14.6–25.0

�0.13 60.2 50.3–69.5 58.3 51.8–64.5

�0.08 18.5 11.7–27.1 90.1 85.6–93.5

p2PSA, pg/ml

�8.3 89.8 82.5–94.8 18.6 13.9–24.1

�15.9 62.0 52.2–71.2 60.7 54.3–66.9

�29.8 29.6 21.2–39.2 90.5 86.1–93.9

%p2PSA

�1.5 90.7 83.6–95.5 24.8 19.5–30.7

�2.0 64.8 55.0–73.8 62.4 56.0–68.5

�2.9 25.0 17.2–34.3 90.5 86.1–93.9

PHI

�34.6 90.7 83.6–95.5 28.5 22.9–34.6

�50.7 68.5 58.9–77.1 64.1 57.7–70.1

�71.8 34.3 25.4–44.0 90.1 85.6–93.5

CI = confidence interval; PSA = prostate-specific antigen; %fPSA = percent free PSA; %p2PSA = percentage of p2PSA to free PSA; PHI = Prostate Health Index.

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6 459

more aggressive PCa at final pathology and therefore might

be adopted in preoperative counselling for patients with

clinically localised PCa.

To the best of our knowledge, the current study is the

first showing the relationship between isoform p2PSA and

Table 2b – Sensitivity and specificity at three levels of predictive variablof Gleason score I7

Criterion Sensitivity, %

Total PSA, ng/ml

�3.3 90.8

�5.3 64.8

�10.6 19.5

Free PSA, ng/ml

�0.37 90.0

�0.80 52.9

�1.70 8.8

%fPSA

�0.21 90.0

�0.14 62.8

�0.09 26.4

p2PSA, pg/ml

�7.6 89.7

�13.7 61.3

�29.0 18.4

%p2PSA

�1.3 91.9

�1.85 64.8

�2.4 38.7

PHI

�30.9 90.8

�44.0 66.3

�56.2 44.8

CI = confidence interval; PSA = prostate-specific antigen; %fPSA = percent free PSA

final pathology in patients who underwent RP for clinically

localised PCa. In recent years, several efforts have been

made to find biomarkers or more complex tools that could

help clinicians preoperatively determine PCa pathologic

characteristics and prognosis. An extensive review was

es (high sensitivity, best combination, high specificity) for prediction

95% CI Specificity, % 95% CI

86.6–94.0 22.5 14.3–32.6

58.6–70.5 62.9 52.0–72.9

14.9–24.9 91.0 83.0–96.0

85.7–93.4 11.2 5.5–19.7

46.6–59.1 51.7 40.8–62.4

5.7–12.9 91.0 83.0–96.0

85.7–93.4 30.0 18.1–37.4

56.7–68.7 62.9 52.0–72.9

21.2–32.2 89.9 81.7–95.3

85.3–93.1 21.3 13.4–31.3

55.1–67.2 60.7 49.7–70.9

13.9–23.6 91.0 83.0–96.0

88.0–94.9 24.7 16.2–35.0

58.6–70.5 66.3 55.5–76.0

32.8–44.9 91.0 83.0–96.0

86.6–94.0 34.8 25.0–45.7

60.2–72.0 66.3 55.5–76.0

38.7–51.1 89.9 85.6–93.5

; %p2PSA = percentage of p2PSA to free PSA; PHI = Prostate Health Index.

Table 2c – Sensitivity and specificity at three levels of predictive variables (high sensitivity, best combination, high specificity) for predictionof Gleason sum upgrading from 6 to 7

Criterion Sensitivity, % 95% CI Specificity, % 95% CI

Total PSA, ng/ml

�3.6 90.9 83.4–95.7 21.9 13.1–33.1

�5.3 60.6 50.3–70.3 61.6 49.5–72.8

�10.6 15.2 8.7–23.8 90.4 81.2–96.0

Free PSA, ng/ml

�0.38 89.9 82.2–95.0 9.6 4.0–18.8

�0.80 53.5 43.2–63.6 50.7 38.7–62.6

�1.70 3.0 0.7–8.6 90.4 81.2–96.0

%fPSA

�0.22 89.9 82.2–95.0 23.3 14.2–34.6

�0.145 60.6 50.3–70.3 60.3 48.1–71.5

�0.10 26.3 18.8–37.1 90.4 81.2–96.0

p2PSA, pg/ml

�7.4 92.9 86.0–97.1 20.6 12.0–31.6

�13.5 57.6 47.2–67.5 56.2 44.1–67.8

�30.5 12.1 6.4–20.2 91.8 83.0–96.9

%p2PSA

�1.2 91.9 84.7–96.4 19.2 10.9–30.1

�1.8 61.6 51.3–71.2 64.4 52.3–75.2

�2.3 36.4 26.9–46.6 90.4 81.2–96.0

PHI

�28.0 89.9 82.2–95.0 26.0 16.5–37.6

�42.2 60.6 50.3–70.3 61.6 49.5–72.8

�55.7 37.4 27.9–47.7 90.4 81.2–96.0

CI = confidence interval; PSA = prostate-specific antigen; %fPSA = percent free PSA; %p2PSA = percentage of p2PSA to free PSA; PHI = Prostate Health Index.

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6460

recently reported by Lughezzani et al. [2]. They concluded

that although predictive and prognostic tools represent

valuable aids, more accurate, flexible, and easily accessible

tools were needed to simplify the practical task of

Table 2d – Sensitivity and specificity at three levels of predictive variablof tumour volume <0.5 ml

Criterion Sensitivity, %

Total PSA, ng/ml

�11.2 92.3

�4.9 73.1

�3.1 15.4

Free PSA, ng/ml

�1.97 92.3

�0.77 57.7

�0.37 11.5

%fPSA

�0.09 92.3

�0.148 65.4

�0.23 30.8

p2PSA, pg/ml

�24.3 88.5

�12.1 65.4

�7.2 23.1

%p2PSA

�2.1 92.3

�1.7 73.1

�1.25 34.6

PHI

�49.3 92.3

�41.3 73.1

�28.5 53.8

CI = confidence interval; PSA = prostate-specific antigen; %fPSA = percent free PSA

prediction and choose the best treatment option based

on these parameters. Auprich et al. and Nakanishi et al.

explored the relationship of the new biomarker PCa antigen

3 (PCA3), and PCa characteristics at final pathology.

es (high sensitivity, best combination, high specificity) for prediction

95% CI Specificity, % 95% CI

74.8–98.8 13.9 10.3–18.1

52.2–88.4 66.1 60.6–71.2

4.5–34.9 89.5 85.6–92.6

74.8–98.8 6.2 3.8–9.4

36.9–76.6 53.7 48.1–59.2

2.6–30.2 89.8 86.0–92.9

74.8–98.8 23.1 18.7–28.1

44.3–82.8 63.6 58.1–68.8

14.4–51.8 91.7 88.1–94.4

69.8–97.4 25.0 20.4–30.1

44.3–82.8 66.1 60.6–71.2

9.0–43.7 90.7 87.0–93.7

74.8–98.8 45.4 39.9–51.0

52.2–88.4 61.7 56.2–67.0

17.2–55.7 92.3 88.8–94.9

74.8–98.8 50.9 45.3–56.5

52.2–88.4 67.0 61.6–72.1

33.4–73.4 90.7 87.0–93.7

; %p2PSA = percentage of p2PSA to free PSA; PHI = Prostate Health Index.

Table 3 – Univariable and multivariable analyses predicting the probability of pT3 at pathologic staging

Predictors Univariable analysis Multivariable analysis

OR (95% CI);p value

AUC of individualpredictor variables, %

Base model Base modelwith %p2PSA

Base modelwith PHI

OR (95% CI);p valuey

OR (95% CI);p value

OR (95% CI);p value

Age 1.035 (1.001–1.071); 0.041 57.2 1.027 (0.989–1.067); 0.166 1.035 (0.996–1.076); 0.078 1.036 (0.996–1.076); 0.078

tPSA 1.152 (1.085–1.224); <0.001 65.6 1.007 (0.868–1.169); 0.925 1.016 (0.875–1.180); 0.830 0.929 (0.795–1.087); 0.359

fPSA 1.503 (1.073–2.106); 0.018 56.4 2.378 (0.846–6.679); 0.100 2.145 (0.764–6.022); 0.170 2.208 (0.784–6.213); 0.138

f/tPSA 0.363 (0.183–0.718); 0.004 60.7 0.452 (0.177–1.157); 0.098 0.519 (0.203–1.324); 0.170 0.503 (0.198–1.282); 0.150

Biopsy Gleason sum 3.706 (2.559–5.368); <0.001 71.3 3.308 (2.272–4.816); <0.001 2.929 (2.004–4.281); <0.001 3.017 (2.064–4.410); <0.001

Clinical stage 1.510 (0.927–2.458); 0.098 54.4 1.525 (0.860–2.704); 0.149 1.471 (0.821–2.635); 0.194 1.498 (0.836–2.684); 0.175

p2PSA 1.040 (1.022–1.059); <0.001 64.9 – – –

%p2PSA 2.220 (1.626–3.032); <0.001 68.0 – 1.802 (1.264–2.569); 0.001 –

PHI 1.030 (1.019–1.040); <0.001 71.9 – – 1.021 (1.009–1.034) 0.001

AUC of multivariable models, % 78.2 80.6 80.7

Gain in predictive accuracy, % – 2.4y 2.5y

OR = odds ratio; AUC = area under the curve; PHI = Prostate Health Index; CI = confidence interval; PSA = prostate-specific antigen; tPSA = total PSA; fPSA = free PSA; f/tPSA = ratio of free to total PSA; %p2PSA = percentage

of p2PSA to free PSA.*The AUC reflects the predictive values of individual variables (columns) and of the multivariable models in predicting the probability of having pT3.y p � 0.05 versus base model.

Table 4 – Univariable and multivariable analyses predicting the probability of Gleason sum I7 at pathologic grading

Predictors Univariable analysis Multivariable analysis

OR (95% CI);p value

AUC* of individualpredictor variables, %

Base model Base modelwith %p2PSA

Base modelwith PHI

OR (95% CI);p value

OR (95% CI);p value

OR (95% CI);p value

Age 1.029 (0.995–1.064); 0.096 56.8 1.043 (1.005–1.082); 0.025 1.057 (1.016–1.100); 0.006 1.053 (1.013–1.095); 0.009

tPSA 1.176 (1.084–1.276); <0.001 65.7 1.077 (0.895–1.295); 0.434 1.147 (0.919–1.432); 0.224 0.996 (0.803–1.235); 0.996

fPSA 1.092 (0.749–1.592); 0.647 50.6 1.579 (0.546–4.572); 0.399 1.214 (0.343–4.295); 0.763 1.072 (0.308–3.732); 0.913

f/tPSA 0.796 (0.512–1.236); 0.309 67.4 0.328 (0.143–0.751); 0.008 0.409 (0.164–1.023); 0.056 0.446 (0.183–1.087); 0.075

Clinical stage 1.718 (0.972–3.038); 0.063 55.2 1.799 (0.974–3.325); 0.061 1.578 (0.833–2.987); 0.161 1.661 (0.882–3.129); 0.116

p2PSA 1.035 (1.010–1.061); 0.006 61.9 – – –

%p2PSA 3.154 (2.048–4.856); <0.001 70.4 – 3.485 (2.140–5.675); <0.001 –

PHI 1.052 (1.034–1.070); <0.001 74.0 – – 1.047 (1.027–1.067); <0.001

AUC of multivariable models, % 72.3 78.3 78.3

Gain in predictive accuracy, % – 6.0y 6.0y

OR = odds ratio; CI = confidence interval; AUC = area under the curve; PSA = prostate-specific antigen; PHI = Prostate Health Index; tPSA = total PSA; fPSA = free PSA; f/tPSA = ratio of free to total PSA; %p2PSA = percentage

of p2PSA to free PSA.* The AUC reflects the predictive value of individual variables (columns), as well as of the multivariable models in predicting the probability of Gleason sum �7.y p = 0.004 versus base model.

EU

RO

PE

AN

UR

OL

OG

Y6

1(

20

12

)4

55

–4

66

46

1

Table 6 – Univariable and multivariable analyses predicting the probability of tumour volume <5 ml at pathologic examination

Predictors Univariable analysis Multivariable analysis

OR (95% CI);p value

AUC* of individualpredictor variables, %

Base model Base modelwith %p2PSA

Base modelwith PHI

OR (95% CI);p value

OR (95% CI);p value

OR (95% CI);p value

Age 0.979 (0.926–1.034); 0.446 56.4 0.980 (0.918–1.046); 0.540 0.956 (0.890–1.027); 0.216 0.958 (0.893–1.028); 0.230

tPSA 0.866 (0.753–0.995); 0.042 67.4 0.823 (0.570–1.188); 0.297 0.693 (0.424–1.131); 0.142 0.789 (0.477–1.304); 0.355

fPSA 1.012 (0.559–1.830); 0.969 52.3 1.291 (0.226–7.383); 0.774 2.745 (0.235–32.12); 0.421 3.915 (0.267–57.36); 0.319

%fPSA 2.136 (1.253–3.640); 0.005 67.9 1.561 (0.454–5.366); 0.480 1.056 (0.232–4.797); 0.944 0.866 (0.186–4.021); 0.854

Biopsy Gleason sum 0.113 (0.039 – 0.330); <0.001 75.0 0.101 (0.030–0.343); p<0.001 0.115 (0.029–0.452); 0.002 0.124 (0.032–0.476); 0.002

Clinical stage 0.888 (0.361–2.182); 0.796 51.2 1.046 (0.386–2.832); 0.929 1.283 (0.456–3.613); 0.637 1.227 (0.436–3.453); 0.699

p2PSA 0.938 (0.888–0.991); 0.022 68.5 – – –

%p2PSA 0.201 (0.091–0.960); <0.001 75.9 – 0.233 (0.097–0.560); 0.001 –

PHI 0.927 (0.895–0.960); <0.001 79.9 – – 0.937 (0.901–0.975); 0.001

AUC of multivariable models, % 83.3 87.1 87.5

Gain in predictive accuracy, % – 3.8 4.2

OR = odds ratio; CI = confidence interval; AUC = area under the curve; PSA = prostate-specific antigen; %p2PSA = percent 2PSA; PHI = Prostate Health Index; tPSA = total PSA; fPSA = free PSA; %p2PSA = percentage of p2PSA

to free PSA.* The AUC reflects the predictive values of individual variables (columns) as well as of the multivariable models in predicting the probability of tumour volume <0.5 ml.

Table 5 – Univariable and multivariable analyses predicting the probability of Gleason sum upgrading from 6 to 7 at pathologic grading

Predictors Univariable analysis Multivariable analysis

OR (95% CI);p value

AUC* of individualpredictor variables, %

Base model Base modelwith %p2PSA

Base modelwith PHI

OR (95% CI);p value

OR (95% CI);p value

OR (95% CI);p value

Age 1.004 (0.960–1.049); 0.873 52.2 1.027 (0.979–1.078); 0.271 1.049 (0.995–1.106); 0.074 1.045 (0.992–1.100); 0.099

tPSA 1.142 (1.027–1.271); 0.015 62.0 1.1450 (0.886–1.479); 0.301 1.273 (0.950–1.707); 0.106 1.161 (0.872–1.545); 0.306

fPSA 0.953 (0.589 –1.542); 0.845 48.4 0.951 (0.224–4.040); 0.946 0.580 (0.124–2.711); 0.489 0.400 (0.079–2.038); 0.270

f/tPSA 0.469 (0.282–0.780); 0.004 64.9 0.532 (0.180–1.567); 0.252 0.675 (0.219–2.083); 0.495 0.846 (0.269–2.660); 0.775

Clinical stage 2.115 (1.048–4.265); 0.036 57.4 2.186 (1.039–4.598); 0.039 1.923 (0.887–4.169); 0.098 2.005 (0.928–4.332); 0.077

p2PSA 1.018 (0.991–1.046); 0.192 57.0 – – –

%p2PSA 2.514 (1.492–4.238); 0.001 66.6 – 3.036 (1.657–5.561); <0.001 –

PHI 1.042 (1.020–1.064); <0.001 68.6 – – 1.045 (1.019–1.071); <0.001

AUC of multivariable models, % 69.6 74.7 75.3

Gain in predictive accuracy, % – 5.1 5.7y

OR = odds ratio; CI = confidence interval; AUC = area under the curve; PSA = prostate-specific antigen; PHI = Prostate Health Index; tPSA = total PSA; fPSA = free PSA; f/tPSA = ratio of free to total PSA; %p2PSA = percentage of

p2PSA to free PSA.* The AUC reflects the predictive value of individual variables (columns), as well as of the multivariable models in predicting the probability of Gleason sum upgrading from 6 to 7.y p < 0.05 versus base model.

EU

RO

PE

AN

UR

OL

OG

Y6

1(

20

12

)4

55

–4

66

46

2

[(Fig._2)TD$FIG]

-0.100

0.000

0.100

0.200

0.300

1009080706050403020100

Net

ben

efit

Threshold probability, %

Model 1

Model 2

Model 3

Treat all

Treat none

Fig. 2 – Decision curve analysis of the effect of prediction models on the detection of pT3 at radical prostatectomy. Net benefit is plotted against variousthreshold probabilities. Model 1 is a basic model including age, total prostate-specific antigen, free PSA (fPSA), percent fPSA (%fPSA), biopsy Gleason sum,and clinical stage. Model 2 is a basic model including all the factors in model 1 plus percentage of p2PSA to fPSA (%p2PSA). Model 3 is a basic modelincluding all the factors in model 1 plus the Prostate Health Index.

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6 463

Although these studies confirmed the reliability of PCA3

sum in determining tumour volume, they were unable to

show a significant relationship between this biomarker and

several other important PCa characteristics, such as the

presence of extracapsular extension or pathologic Gleason

sum�7 disease [15,16]. Conversely, the results of our study

showed that %p2PSA and PHI are related to both PCa volume

and aggressiveness. Auprich et al. showed that PCA3 was an

independent risk factor of low-volume disease, defined as a

computer-assisted planimetrically measured tumour vol-

ume <0.5 ml [15]. In the PSA era, mean tumour volume is

drastically decreased, and evidence suggests that smaller

tumours are less aggressive and less frequently associated

with progression [17]. Although the volume threshold for

[(Fig._3)TD$FIG]

0.800

0.700

0.600

0.500

0.400

0.300

0.200

0.100

0.000

-0.1000 10 20 30 40

Threshold

Net

ben

efit

Fig. 3 – Decision curve analysis of the effect of prediction models on the detectplotted against various threshold probabilities. Model 1 is a basic model includ(%fPSA), and clinical stage. Model 2 is a basic model including all the factors inmodel including all the factors in model 1 plus the Prostate Health Index.

clinically significant PCa is still controversial, data from

autopsy series, cystoprostatectomy, and RP series support

the concept that the 0.5-ml threshold volume with no

Gleason 4 or 5 can be used for the definition of clinically

insignificant tumours [17–19]. In our study we found that

PHI, but not %p2PSA, significantly increases the predictive

accuracy of a basic model including patient age, tPSA, fPSA,

f/tPSA, clinical stage and biopsy Gleason sum for tumour

volume <0.5 ml. In our series, 15% of patients with a PCa

volume <0.5 ml had Gleason sum 7 disease, indicating that

this parameter could be a subject of interest and discussion

in further studies.

In the current era of PCa screening with PSA, an

increasing number of men are diagnosed with lower stage

50 60 70 80 90 100

probability, %

Model 1

Model 2

Model 3

Treat all

Treat none

ion of pathologic Gleason sum I7 at radical prostatectomy. Net benefit ising age, total prostate-specific antigen, free PSA (fPSA), percent fPSAmodel 1 plus percentage of p2PSA to fPSA (%p2PSA). Model 3 is a basic

[(Fig._4)TD$FIG]

-0.100

0.000

0.100

0.200

0.300

0.400

0.500

0.600

1009080706050403020100

Ne

t b

en

efi

t

Threshold probability, %

Model 1

Model 2

Model 3

Treat all

Treat none

Fig. 4 – Decision curve analysis of the effect of prediction models on the detection of the presence of Gleason sum upgrading at radical prostatectomy. Netbenefit is plotted against various threshold probabilities. Model 1 is a basic model including age, total prostate-specific antigen, free PSA (fPSA), percentfPSA (%fPSA), and clinical stage. Model 2 is a basic model including all the factors in model 1 plus percentage of p2PSA to fPSA (%p2PSA). Model 3 is a basicmodel including all the factors in model 1 plus the Prostate Health Index.

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6464

and grade of PCa. These men are offered numerous options,

including active surveillance, RP, radiotherapy, and focal

therapy [1]. The use of an accurate and robust biomarker,

alone or in combination with other variables, that enables

prediction of a low-volume low-grade PCa could be

extremely useful in clinical decision making.

Our study has other strengths. It consisted of a

prospective observational study in a homogeneous con-

temporary cohort of men with clinically localised PCa who

were candidates for RP. Second, we adopted a standardised

and centralised pathologic evaluation: a highly experienced

genitourinary pathologist reviewed all the specimens.

Finally, all blood samples were managed according to the

guidelines of Semjonow et al. and no archived serum was

[(Fig._5)TD$FIG]

-0.010

0.000

0.010

0.020

0.030

0.040

0.050

0.060

403020100

Ne

t b

en

efi

t

Threshold

Fig. 5 – Decision curve analysis of the effect of prediction models on the detectiobenefit is plotted against various threshold probabilities. Model 1 is a basic mofPSA (%fPSA), biopsy Gleason sum, and clinical stage. Model 2 is a basic model(%p2PSA). Model 3 is a basic model including all the factors in model 1 plus th

used [10]. Finally, although not all readers may be familiar

with the DCA, it should be considered a novel method for

evaluating new diagnostic tests (ie, biomarkers) and

prediction models (ie, logistic regression models). It was

developed to address the limitations of traditional biosta-

tistical methods, which focus on accuracy, calibration, and

discrimination using metrics such as sensitivity, specifici-

ty, or AUC [14]. Such methods are mathematically simple,

can be used irrespective of whether the predictor is binary

or continuous, and generally have an intuitive interpreta-

tion, but these methods also could have little clinical

relevance.

Despite its strengths, our study is not devoid of

limitations. We did not compare p2PSA and its derivates

1009080706050 probability, %

Model 1

Model 2

Model 3

Treat all

Treat none

n of the presence of tumour volume <0.5 ml at radical prostatectomy. Netdel including age, total prostate-specific antigen, free PSA (fPSA), percentincluding all the factors in model 1 plus percentage of p2PSA to fPSAe Prostate Health Index.

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6 465

with other commonly used PSA measurements such as PSA

density (PSAD) and PSA velocity (PSAV). Giannarini et al.

assessed whether PSAD and PSAD of the transition zone

(PSADTZ) were more accurate than PSA alone in predicting

the pathologic stage of PCa [20]. They found that PSAD and

PSADTZ failed to outperform PSA in preoperative stage

prediction of PCa. Similar results were reported by O’Brien

et al, who assessed whether pretreatment PSAV or PSA

doubling time predicts outcome in men undergoing RP [21].

These authors concluded that there was no clear evidence

that PSAV substantially enhanced the predictive accuracy of

a single pretreatment tPSA value. Most of the predictive

tools for PCa candidates for RP included biopsy Gleason as a

variable [2]. In our analysis, biopsy Gleason was included in

the multivariate logistic regression analysis for prediction

of pT3 and tumour volume <0.5 ml but not for pathologic

Gleason and Gleason upgrading. Such strategy is due to the

fact that the measured biopsy Gleason was estimated with

a certain margin of error: The correlation with the

pathologic Gleason was poor: r = 0.503. Consequently, if

we considered the pathologic Gleason an outcome and if

the preoperative Gleason had been very precise, there

would not have been the need to value the pathologic one.

There are other limitations, such as the relatively small

number of patients, the absence of stratification according

to risk categories (ie, D’Amico), and the lack of a nomogram

including PHI. Finally, although the pathologic examina-

tion was performed by an experienced pathologist, we do

not have a second reference pathologist to confirm our

findings.

5. Conclusions

The current study showed that p2PSA and its derivatives are

predictors of several PCa characteristics at final pathology

and are more accurate than currently available markers in

determining PCa aggressiveness. Further studies are re-

quired to externally validate our findings.

Author contributions: Massimo Lazzeri had full access to all the data in

the study and takes responsibility for the integrity of the data and the

accuracy of the data analysis.

Study concept and design: Guazzoni, Lazzeri.

Acquisition of data: Scattoni, Freschi, Cestari, Buffi, Larcher, Gadda.

Analysis and interpretation of data: Guazzoni, Nava, Lazzeri.

Drafting of the manuscript: Lazzeri, Montorsi, Lughezzani.

Critical revision of the manuscript for important intellectual content:

Montorsi.

Statistical analysis: Bini, Lughezzani.

Obtaining funding: None.

Administrative, technical, or material support: None.

Supervision: Rigatti, Guazzoni.

Other (specify): None.

Financial disclosures: I certify that all conflicts of interest, including

specific financial interests and relationships and affiliations relevant to

the subject matter or materials discussed in the manuscript (eg,

employment/ affiliation, grants or funding, consultancies, honoraria,

stock ownership or options, expert testimony, royalties, or patents filed,

received, or pending), are the following: None.

Funding/Support and role of the sponsor: Beckman Coulter Inc. and

Beckman Coulter Italy were involved in collection of data for this study.

Access Hybritech p2PSA ([-2]proPSA) reagents were provided by

Beckman Coulter Inc. and Beckman Coulter Italy; UniCel DxI800

Immunoassay System analyzer (Beckman Coulter Inc., Brea, CA, USA)

was provided by Beckman Coulter Italy.

References

[1] Lazzeri M, Guazzoni G. Focal therapy meets prostate cancer. Lancet

2010;376:1036–7.

[2] Lughezzani G, Briganti A, Karakiewicz PI, et al. Predictive and

prognostic models in radical prostatectomy candidates: a critical

analysis of the literature. Eur Urol 2010;58:687–700.

[3] Quinn D, Gross M. Show us a sign: the search for ‘‘game changing’’

prostate cancer biomarkers. Lancet Oncol 2011;12:204–6.

[4] Stephan C, Kahrs AM, Cammann H, et al. A [-2]proPSA-based

artificial neural network significantly improves differentiation be-

tween prostate cancer and benign prostatic diseases. Prostate 2009;

69:198–207.

[5] Sokoll LJ, Sanda MG, Feng Z, et al. A prospective, multicenter,

National Cancer Institute Early Detection Research Network study

of [�2]proPSA: improving prostate cancer detection and correlating

with cancer aggressiveness. Cancer Epidemiol Biomarkers Prev

2010;19:1193–200.

[6] Le BV, Griffin CR, Loeb S, et al. [-2]Proenzyme prostate specific antigen

is more accurate than total and free prostate specific antigen in

differentiating prostate cancer from benign disease in a prospective

prostate cancer screening study. J Urol 2010;183: 1355–9.

[7] Jansen FH, Van Schaik RHN, Kurstjens J, et al. Prostate-specific antigen

(PSA) isoform p2PSA in combination with total PSA and free PSA

improves the diagnostic accuracy in prostate cancer detection. Eur

Urol 2010;57:921–7.

[8] Guazzoni G, Nava L, Lazzeri M, et al. Prostate-specific antigen (PSA)

isoform p2PSA significantly improves the prediction of prostate

cancer at initial extended prostate biopsies in patients with total

PSA between 2.0 and 10 ng/ml: results of a prospective study in a

clinical setting. Eur Urol 2011;60:214–22.

[9] Catalona WJ, Partin AW, Sanda MG, et al. A multicenter study of

[-2]pro-prostate specific antigen combined with prostate specific

antigen and free prostate specific antigen for prostate cancer de-

tection in the 2.0 to 10.0 ng/ml prostate specific antigen range.

J Urol 2011;185:1650–5.

[10] Semjonow A, Kopke T, Eltze E, et al. Pre-analytical in-vitro stability

of [-2]proPSA in blood and serum. Clin Biochem 2010;47:926–8.

[11] Van der Kwast TH, Amin MB, Billis A, et al. International Society of

Urological Pathology (ISUP) Consensus Conference on Handling and

Staging of Radical Prostatectomy Specimens. Working group 2: T2

substaging and prostate cancer volume. Mod Pathol 2011;24:

16–25.

[12] Epstein JI, Allsbrook Jr WC, Amin MB, et al. The 2005 International

Society of Urological Pathology (ISUP) Consensus Conference on

Gleason Grading of Prostatic Carcinoma. Am J Surg Pathol 2005;

29:1228–42.

[13] DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas

under two or more correlated receiver operating characteristic

curves: a nonparametric approach. Biometrics 1988;44:837–45.

[14] Vickers AJ, Elkin EB. Decision curve analysis: a novel method for

evaluating prediction models. Med Decis Making 2006;26:565–74.

[15] Auprich M, Chun FK-H, Ward JF, et al. Critical assessment of

preoperative urinary prostate cancer antigen 3 on the accuracy

of prostate cancer staging. Eur Urol 2011;59:96–105.

[16] Nakanishi H, Groskopf J, Fritsche HA, et al. PCA3 molecular urine

assay correlates with prostate cancer tumour volume: implication

E U R O P E A N U R O L O G Y 6 1 ( 2 0 1 2 ) 4 5 5 – 4 6 6466

in selecting candidates for active surveillance. J Urol 2008;179:

1804–9.

[17] Mouraviev V, Villers A, Bostwick DG, Wheeler TM, Montironi R,

Polascik TJ. Understanding the pathological features of focality, grade

and tumour volume of early-stage prostate cancer as a foundation for

parenchyma-sparing prostate cancer therapies: active surveillance

and focal targeted therapy. BJU Int 2011;108:1074–85.

[18] Autorino R, Di Lorenzo G, Damiano R, et al. Pathology of the prostate

in radical cystectomy specimens: a critical review. Surg Oncol 2009;

18:73–84.

[19] Anast JW, Andriole GL, Bismar TA, Yan Y, Humphrey PA. Relating

biopsy and clinical variables to radical prostatectomy findings: can

insignificant and advanced prostate cancer be predicted in a

screening population? Urology 2004;64:544–50.

[20] Giannarini G, Scott CA, Moro U, Pertoldi B, Beltrami CA, Selli C. Are

PSA density and PSA density of the transition zone more accurate

than PSA in predicting the pathological stage of clinically localized

prostate cancer? Urol Oncol 2008;26:353–60.

[21] O’Brien MF, Cronin AM, Fearn PA, et al. Pretreatment prostate-

specific antigen (PSA) velocity and doubling time are associated

with outcome but neither improves prediction of outcome beyond

pretreatment PSA alone in patients treated with radical prostatec-

tomy. J Clin Oncol 2009;27:3591–7.