Articles

Platinum Priority – Prostate Cancer
Editorials by Monique J. Roobol on pp. 223–225 and by Carvell T. Nguyen and Michael W. Kattan on pp. 226–228 of this issue

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

By: Giorgio Guazzonia, Luciano Navaa, Massimo Lazzeria lowast , Vincenzo Scattonib, Giovanni Lughezzania, Carmen Maccagnanob, Fernanda Dorigattic, Ferruccio Ceriottic, Marina Pontilloc, Vittorio Binid, Massimo Freschie, Francesco Montorsib and Patrizio Rigattib

European Urology, Volume 60 Issue 1, August 2011, Pages 214-222

Published online: 01 August 2011

Keywords: Prostate-specific antigen, PSA, proPSA, Diagnostic accuracy, Prostate cancer, Prostate biopsy, PSA isoforms

Abstract Full Text Full Text PDF (473 KB)

Abstract

Background

Total prostate-specific antigen (tPSA), ratio of free PSA (fPSA) to tPSA (%fPSA), and PSA density (PSAD) testing have a very low accuracy in the detection of prostate cancer (PCa). There is an urgent need for more accurate biomarkers.

Objective

To compare the diagnostic accuracy of PSA isoform p2PSA and its derivatives in determining the presence of PCa at initial biopsy with the accuracy of other predictors in patients with tPSA 2.0–10 ng/ml.

Design, setting, and participants

We conducted an observational prospective study in a real clinical setting of consecutive men with tPSA 2.0–10 ng/ml and negative digital rectal examination who were scheduled for prostate biopsy at a tertiary academic center.

Intervention

Outpatient transrectal ultrasound-guided prostate biopsies were performed according to a standardized institutional saturation scheme (18–22 cores).

Measurements

We determined the diagnostic accuracy of serum tPSA, %fPSA, PSAD, p2PSA, %p2PSA [(p2PSA/fPSA)×100] and the Beckman Coulter Prostate Health Index (phi; [p2PSA/fPSA×√tPSA]).

Results and limitations

Overall, 107 of 268 patients (39.9%) were diagnosed with PCa at extended prostate biopsies. Statistically significant differences between patients with and without PCa were observed for age, prostate and transition zone volume, PSAD, %p2PSA, and phi (all p values<0.05). In univariate accuracy analysis, phi and %p2PSA were the most accurate predictors of PCa (area under the curve: 75.6% and 75.7%, respectively), followed by transition zone volume (66%), prostate volume (65%), patient age (63%), PSAD (61%), %fPSA (58%), and tPSA (53%). In multivariate accuracy analyses, both phi (+11%) and %p2PSA (+10%) significantly improved the accuracy of established predictors in determining the presence of PCa at biopsy (p<0.001). Although %p2PSA and phi were significantly associated with Gleason score (Spearman ρ: 0.303 and 0.387, respectively; p≤ 0.002), they did not improve the prediction of Gleason score ≥7 PCa in multivariable accuracy analyses (p > 0.05).

Conclusions

In patients with a tPSA between 2.0 and 10 ng/ml, %p2PSA and phi are the strongest predictors of PCa at initial extended biopsies and are significantly more accurate than the currently used tests (tPSA, %fPSA, and PSAD) in determining the presence of PCa at biopsy.

Take Home Message

The ratio of prostate-specific antigen (PSA) isoform p2PSA to free PSA and the Prostate Health Index were shown to be the strongest predictors of prostate cancer (PCa) at initial extended biopsies in men with a total PSA (tPSA) between 2.0 and 10 ng/ml, and they are significantly more accurate in determining the presence of PCa at biopsy than the currently used tests: tPSA, ratio of free PSA to total PSA, and PSA density.

Keywords: Prostate-specific antigen, PSA, proPSA, Diagnostic accuracy, Prostate cancer, Prostate biopsy, PSA isoforms.

1. Introduction

Prostate-specific antigen (PSA) has a limited specificity and sensitivity in determining the presence of prostate cancer (PCa), especially in the total PSA (tPSA) range between 2 and 10 ng/ml [1]. Despite the introduction of several PSA derivatives, such as free to total PSA ratio (%fPSA; [(fPSA/tPSA)×100]), PSA density (PSAD), and PSA velocity, our ability to determine the presence of PCa at initial prostate biopsy remains limited. Consequently, approximately 75% of patients who do not harbor PCa may receive unnecessary prostate biopsies [2]. On the contrary, several patients may harbor PCa despite low levels of tPSA or high levels of %fPSA [3].

A meta-analysis showed that the use of %fPSA is able to improve the diagnostic performance among men with a tPSA 2–10 ng/ml compared with tPSA alone [4]. However, at a sensitivity of 95%, the specificity of %fPSA was only 18% in the 4–10 ng/ml tPSA range and 6% in the 2–4 ng/ml tPSA range [4]. On the contrary, %fPSA showed a significantly lower area under the curve (AUC) (66.7%) relative to PSAD (73.9%) for detecting PCa in men with a tPSA level <4 ng/ml [5]. Thus, in the tPSA range between 2 and 10 ng/ml, no single biomarker can accurately predict the result of an initial prostatic biopsy, although in the last few years, several prediction tools have been developed to help clinicians in the biopsy decision path [6].

According to preliminary studies, PSA isoform p2PSA and its derivatives, namely %p2PSA and the Beckman Coulter Prostate Health Index (phi), a mathematical combination of tPSA, fPSA, and p2PSA, may significantly improve the accuracy of tPSA and %fPSA in predicting the presence of PCa at prostate biopsy [7], [8], [9], [10], [11], [12], and [13]. However, the results of these studies were limited by their retrospective nature, use of stored blood samples, limited sample size, as well as by the lack of standardized biopsy protocols.

Based on the limitations of the previously published studies and because only scant information about the role of p2PSA and its derivatives in the tPSA range between 2.0 and 10 ng/ml is available, we investigated the accuracy of these biomarkers discriminating between patients with or without PCa within a prospectively collected contemporary cohort of prostate biopsy candidates in a clinical setting.

2. Materials and methods

2.1. Study population

The study population consisted of patients with tPSA between 2.0 and 10 ng/ml and with a negative digital rectal examination (DRE) who were prospectively referred to our department for a first set of prostate biopsies by their referenced urologist for suspected PCa. Patients with bacterial acute or chronic prostatitis, patients subjected to previous endoscopic surgery of the prostate for benign prostatic hyperplasia, and patients treated with drugs that may alter serum PSA levels were excluded from the study. The study was approved by the hospital ethics committee and reported according to the Standards for the Reporting of Diagnostic Accuracy guidelines.

2.2. Study design

The current study was an observational prospective study in a contemporary cohort of consecutive patients subjected to outpatient prostate biopsy between March and August 2010. The primary end point of the study was to determine the diagnostic accuracy of p2PSA, %p2PSA ([(p2PSA pg/ml)/(fPSA ng/ml×1000)]×100), and Beckman Coulter phi ([p2PSA/fPSA]×√tPSA; index tests) and to compare it with the accuracy of established PCa predictors (tPSA, fPSA, %fPSA, and PSAD; reference standard tests). The secondary end point was to evaluate the relationship between p2PSA, %p2PSA, phi, and Gleason score at biopsy.

2.3. Methods

A blood sample was drawn before any prostatic manipulations that might cause a transient increase of biomarkers. The blood samples were processed by Access2 Immunoassay System analyzer (Beckman Coulter, Brea, CA, USA) and managed according to Semjonow et al [14]. The analysis of the serum samples was performed using Hybritech calibrated Access tPSA and fPSA assays. Transrectal ultrasonography (TRUS) was used to determine prostate and transition zone volume. Patients underwent ambulatory TRUS-guided prostate biopsies according to a standardized institutional saturation scheme, which consisted of at least 18–22 biopsy cores taken from the prostate gland to obtain the highest detection rate [15]. Specimens were processed and evaluated by a single experienced genitourinary pathologist blinded to the results of index tests. PCa was identified and graded according to the 2005 consensus conference of the International Society of Urological Pathology definitions [16]. Patients diagnosed with high-grade prostatic intraepithelial neoplasia or atypical small acinar proliferation of prostate were not considered as having developed the outcome of interest (PCa).

2.4. Statistical analysis

A study sample of 150 subjects with a PCa rate at biopsy of 50% was needed to detect receiver-operating characteristic (ROC) curve area differences of 15% at a hypothesized rank correlation coefficient between predictive variables of 0.400, in both PCa-positive and PCa-negative groups. With this sample size, the study power (1-β) at an α error of 5% is 80%.

The Kolmogorov-Smirnov test was used to assess the normal distribution of variables. The student t test and the Mann-Whitney U test were used for comparisons of normally and not normally distributed continuous variables, respectively. The student t test and the Mann-Whitney U test were used for comparisons of parametric and nonparametric continuous variables, respectively. The chi-square test was used for comparisons of categorical variables. Patients were stratified according to the presence or absence of PCa at biopsy.

Multivariate logistic regression models were fitted for the prediction of the presence of PCa at biopsy and for the prediction of Gleason score ≥7 in the PCa group, incorporating as explanatory variables all the variables that showed a corrected p value (pc) ≤0.25 in comparisons between groups [17]. To avoid multicollinearity problems, predictors in strong correlation with other explanatory variables were dropped from the models. Logistic regression models goodness of fit was checked using the Hosmer-Lemeshow test. Multivariate logistic regression models were complemented by predictive accuracy tests. Predictive accuracy was quantified as the area under the ROC curve (AUC), where a value of 100% indicates perfect prediction and 50% is equivalent to a toss of a coin. To test the ability of %p2PSA and phi in determining the presence of PCa at biopsy, these variables were added to the base multivariate model. The gain in predictive accuracy was quantified and AUCs were compared using the DeLong method [18]. To reduce overfit bias, multivariate predictive accuracy tests were subjected to 200 bootstrap resamples. Multivariate logistic regression models for the prediction of PCa at biopsy were also fitted in the tPSA 4–10 ng/ml subgroup. Finally, the relationship between p2PSA (and its derivatives) and Gleason score at biopsy was tested using the Spearman ρ coefficient analysis.

All statistical analyses were performed using SPSS v.16.0 or S-Plus professional. A two-sided p value <0.05 was considered significant. To plot ROC curves, Prism 4.0 (GraphPad Software, La Jolla, CA, USA) was used. In comparisons between groups, the probability (p) values were corrected (pc) for the number of tests made, using the Bonferroni formula.

3. Results

Table 1 lists the demographic and clinical characteristics of the overall study population. PCa at initial biopsy was diagnosed in 107 of 268 patients (39.9%). When comparing patients with and without PCa at biopsy, patients with PCa were significantly older (mean age: 65.5 vs 61.7 yr; pc=0.002). Median tPSA (5.2 vs 5.6 ng/ml; pc=0.989), fPSA (0.8 vs 0.9 ng/ml; pc=0.217), %fPSA (0.15 vs 0.17; pc=0.251), and p2PSA (15.0 vs 13.0 pg/ml; pc=0.104) did not differ significantly between men with and without PCa. Conversely, PSAD was significantly higher (0.12 vs 0.09; pc=0.043) in patients with PCa relative to their counterparts without PCa. Additionally, in the PCa group, %p2PSA (2.0 vs 1.4; pc < 0.001) and phi (44.3 vs 33.1; pc < 0.001) values were significantly higher than in patients without PCa.

Table 1 Descriptive characteristics of the study population

Overall Absence of PCa Presence of PCa p value pc value*
No. of patients (%) 268 161 (60.1) 107 (39.9)
Age, yr 63.3±8.2 61.7±7.8 65.5±8.3 0.00017** p=0.0022
Gleason score NA NA 6 (5–9)
Gleason score categories (%)
<7 NA NA 55 (51.4)
≥7 NA NA 52 (48.6)
Prostate volume, ml 54 (9–190) 58 (20–190) 46 (9–163) 0.00007 p=0.0007
Transition zone volume, ml 27 (1–150) 31 (7–150) 22 (1–95) 0.00004 p=0.0004
Total PSA, ng/ml 5.7 (2.0–9.9) 5.6 (2.0–9.9) 5.2 (2.0–9.9) 0.340 p=0.989
Total PSA categories, No. (%)
2–4 ng/ml 50 (18.7) 29 (18.0) 21 (19.6) 0.751 1.000
4–10 ng/ml 218 (81.3) 132 (82.0) 86 (80.4)
Free PSA, ng/ml 0.8 (0.1–3.3) 0.9 (0.2–3.3) 0.8 (0.1–3.1) 0.022 0.217
%fPSA 0.17 (0.04–0.46) 0.17 (0.06–0.46) 0.15 (0.04–0.31) 0.026 0.251
PSAD 0.10 (0.03–0.43) 0.09 (0.03–0.29) 0.12 (0.03–0.43) 0.004 0.043
p2PSA, pg/ml 14.0 (1.5–98.6) 13.0 (1.5–64.1) 15.0 (2.5–98.6) 0.010 0.104
%p2PSA 1.6 (0.4–4.6) 1.4 (0.4–4.6) 2.0 (0.9–4.1) <0.00001 <0.001
phi 37.1 (6.5–109.6) 33.1 (6.5–109.6) 44.3 (21.7–106.5) <0.00001 <0.001

* Corrected p value.

** Student t test

Mann-Whitney test

Chi-square test

PCa=prostate cancer; NA=not available; PSA=prostate-specific antigen; %fPSA=free to total PSA ratio; PSAD=PSA density; phi=Prostate Health Index.

Parametric data are expressed as mean plus or minus standard deviation.

Nonparametric data are expressed as median, minimum, and maximum.

In bivariate logistic regression models, patient age (p < 0.001), PSAD (p=0.001), fPSA (p=0.024), %fPSA (p=0.008), p2PSA (p=0.041), %p2PSA (p<0.001), and phi (p<0.001) were significantly associated with the presence of PCa at biopsy (Table 2a). In addition, prostate volume (p=0.003) and transition zone volume (p=0.002) were statistically significant predictors of PCa at biopsy. On the contrary, tPSA was not significantly associated with the presence of PCa (p=0.395). In univariate accuracy analysis, %p2PSA (AUC: 75.7%) and phi (AUC: 75.6%) were the most accurate predictors and significantly outperformed %fPSA (AUC: 57.9%) and PSAD (AUC: 60.8%) in the prediction of PCa at biopsy (p≤ 0.001) (Fig. 1). At 90% specificity, the sensitivities of phi (42.9%) and %p2PSA (38.8%) were significantly higher than those of %fPSA (20.0%) and PSAD (26.5%) (p≤ 0.014) (Table 3).

Table 2a Logistic regression analyses predicting the probability of prostate cancer (PCa) at biopsy in the overall population with total prostate-specific antigen (tPSA) 2–10 ng/ml (n=268; the area under the curve reflects the predictive value of individual variables [columns] as well as of the multivariable models in predicting the probability of having PCa)

Predictors AUC of individual predictor variables (95% CI) Bivariate analysis Multivariate analysis
Base model Base model with %p2PSA Base model with phi
OR (95% CI); p value OR (95% CI); p value OR (95% CI); p value OR (95% CI); p value
Age 0.63 (0.57–0.68) 1.062 (1.028–1.096); p < 0.001 1.097 (1.055–1.140); p<0.001 1.093 (1.049–1.139); p<0.001 1.095 (1.050–1.141); p<0.001
Prostate volume 0.65 (0.59–0.71) 0.986 (0.976–0.995); p=0.003 0.988 (0.976–0.999); p=0.037 0.994 (0.982–1.007); p=0.367 0.994 (0.981–1.007); p=0.363
Transition zone volume1 0.66 (0.56–0.72) 0.979 (0.967–0.993); p=0.002
tPSA 0.53 (0.47–0.59) 0.948 (0.838–1.072); p=0.395 1.078 (0.902–1.290); p=0.408 1.145 (0.944–1.389); p=0.168 0.889 (0.724–1.091); p=0.260
fPSA 0.58 (0.52–0.64) 0.561 (0.339–0.928); p=0.024 0.538 (0.279–1.037); p=0.064 0.471 (0.191–1.160); p=0.471 0.494 (0.198–1.235); p=0.131
%fPSA2 0.58 (0.52–0.64) 0.007 (0.000–0.276); p=0.008
PSA density3,4 0.61 (0.54–0.67) 1.008 (1.004–1.013); p=0.001
p2PSA5 0.59 (0.53–0.65) 1.026 (1.001–1.052); p=0.041
%p2PSA 0.76 (0.71–0.81) 4.284 (2.629–6.982); p<0.001 4.576 (2.540–8.243); p<0.001
phi 0.76 (0.70–0.81) 1.062 (1.040–1.084); p<0.001 1.072 (1.045–1.100); p<0.001
AUC of multivariable models (95% CI) 0.72 (0.65–0.78) 0.82 (0.76–0.87) 0.83 (0.77–0.88)
Gain in predictive accuracy (95% CI) 0.10 (0.05–0.16)* 0.11 (0.06–0.17)*

* p<0.001 (relative to the base multivariate model; DeLong method).1Not included in base multivariate model because of the strong correlation with prostate volume (ρ=0.946).2Not included in base multivariate model because of the strong correlation with fPSA (ρ=0.687).3Not included in base multivariate model because of the strong correlation with prostate volume (ρ=0.700).4Expressed as tPSA (pg/ml)/prostate volume (ml) to scaling OR in a more intelligible range.5Not included in base multivariate model because of the strong correlation with fPSA (ρ=0.734).

gr1

Fig. 1 Receiver operating characteristic curves depicting the accuracy of individual predictors of prostate cancer at initial extended biopsies.PSA=prostate-specific antigen; tPSA=total PSA; %fPSA=free to total PSA ratio.

Table 3 Sensitivities at 90% specificity in predicting the presence of prostate cancer at initial biopsy

Variables Cut-off value Sensitivity at 90% specificity, % 95% CI, % p value*
tPSA ≥8.9 5.1 0.8–10.1 <0.001* <0.001
%fPSA ≤0.29 20.0 8.5–29.0 <0.001* 0.001
PSAD ≥0.16 26.5 12.4–34.1 0.003* 0.014
%p2PSA ≥2.18 38.8 22.4–43.2 0.814*
phi ≥48.5 42.9 32.1–54.1 0.814

* The phi compared with other variables (DeLong method).

%p2PSA compared with other variables (DeLong method).

CI=confidence interval; tPSA=total prostate-specific antigen; %fPSA=free to total PSA ratio; phi=prostate health index; PSAD=PSA density; 2PSA=[-2]proPSA.

In multivariate logistic regression models testing the predictors of PCa at biopsy, patient age (p<0.001) and prostate volume (p=0.037) achieved independent predictor status (Table 2a). Both %p2PSA and phi significantly increased the accuracy of the base multivariate model by 10% and 11%, respectively (p<0.001). Similarly, in the group of patients with tPSA 4–10 ng/ml (n=218), the inclusion of phi and %p2PSA significantly increased multivariate predictive accuracy from 72% to 83% (+11%) in both models (p<0.001) (Table 2b).

Table 2b Logistic regression analyses predicting the probability of PCa at biopsy in patients with tPSA between 4 and 10 ng/ml (n=218)

Predictors AUC of individual predictor variables (95% CI) Bivariate analysis Multivariate analysis
Base model Base model with %p2PSA Base model with phi
OR (95% CI); p value OR (95% CI); p value OR (95% CI); p value OR (95% CI); p value
Age 0.64 (0.57–0.70) 1.066 (1.029–1.106); p<0.001 1.102 (1.055–1.151); p<0.001 1.100 (1.049–1.153); p<0.001 1.100 (1.049–1.153); p<0.001
Prostate volume 0.65 (0.58–0.71) 0.987 (0.977–0.997); p=0.010 0.988 (0.975–0.999); p=0.047 0.993 (0.980–1.007); p=0.334 0.994 (0.980–1.007); p=0.348
Transition zone volume1 0.67 (0.59–0.73) 0.979 (0.965–0.993); p=0.003
tPSA 0.55 (0.48–0.62) 0.896 (0.760–1.057); p=0.193 0.981 (0.783–1.228); p=0.867 1.051 (0.825–1.337); p=0.689 0.846 (0.658–1.087)
fPSA 0.60 (0.53–0.66) 0.539 (0.310–0.937); p=0.029 0.491 (0.209–1.154); p=0.103 0.560 (0.221–1.417); p=0.221 0.567 (0.222–1.451); p=0.237
%fPSA2 0.57 (0.50–0.63) 0.017 (0.000–1.137); p=0.057
PSA density3,4 0.60 (0.52–0.67) 1.008 (1.003–1.013); p=0.003
p2PSA5 0.60 (0.54–0.67) 1.028 (1.001–1.056); p=0.040
%p2PSA 0.78 (0.72–0.83) 4.533 (2.624–7.830); p<0.001 4.889 (2.507–9.534); p<0.001
phi 0.76 (0.70–0.81) 1.061 (1.038–1.086); p<0.001 1.069 (1.040–1.099); p<0.001
AUC of multivariable models (95% CI) 0.72 (0.65–79) 0.83 (0.78–0.89) 0.83 (0.78–0.89)
Gain in predictive accuracy (95% CI) 0.11 (0.06–0.17)* 0.11 (0.05–0.16)*

* p<0.001 (relative to the base multivariable model; DeLong method).1Not included in multivariate base model because of the strong correlation with prostate volume (ρ=0.950).2Not included in multivariate base model because of the strong correlation with fPSA (ρ=0.840).3Not included in multivariate base model because the strong correlation with prostate volume (ρ=0.818).4Expressed as tPSA (pg/ml)/prostate volume (ml) to scaling OR in a more intelligible range.5Not included in multivariate base model because of the strong correlation with fPSA (ρ=0.704).

AUC=area under the curve; CI=confidence interval; OR=odds ratio; phi=prostate health index; %fPSA=free to total prostate-specific antigen ratio; tPSA=total PSA; fPSA=free PSA.

In Table 4, we report the results of univariate and multivariate logistic regression analyses testing the predictors of PCa with Gleason score ≥7. In multivariate analysis, phi and %p2PSA were able to increase the accuracy of the base model by 7%. However, these results did not achieve statistical significance. Finally, the Spearman ρ coefficient analysis demonstrated a significant relationship between Gleason score, %p2PSA (ρ: 0.303; p=0.002), and phi levels (ρ: 0.387; p<0.001).

Table 4 Logistic regression models testing the predictors of Gleason score ≥7 in prostate cancer group

Predictors AUC of individual predictor variables (95% CI) Bivariate analysis Multivariate analysis
Base model Base model with %p2PSA Base model with phi
OR (95% CI); p value OR (95% CI); p value OR (95% CI); p value OR (95% CI); p value
Age 0.65 (0.55–0.74) 1.068 (1.016–1.123); p=0.010 1.108 (1.042–1.179); p=0.001 1.128 (1.051–1.210); p < 0.001 1.129 (1.053–1.212); p < 0.001
Prostate volume1 0.60 (0.50–0.70) 0.991 (0.977–0.1.005); p=0.186
Transition zone volume1 0.61 (0.51–0.71) 0.983 (0.963–1.003); p=0.100
tPSA 0.61 (0.52–0.71) 1.286 (1.034–1.600); p=0.024 1.191 (0.909–1.561); p=0.205 1.342 (0.994–1.812); p=0.055 0.989 (0.730–1.339); p=0.941
fPSA1 0.54 (0.44–0.64) 0.892 (0.448–2.152); p=0.965
%fPSA 0.63 (0.54–0.73) 0.001 (0.000–0.744); p=0.041 0.000 (0.000–0.073); p=0.011 0.000 (0.000–0.372); p=0.031 0.000 (0.000–0.297); p=0.028
PSAD2 0.64 (0.54–0.73) 1.006 (1.000–1.011); p=0.054 1.001 (0.994–1.008); p=0.736 0.998 (0.990–1.006); p=0.585 0.998 (0.990–1.006); p=0.622
p2PSA1 0.54 (0.45–0.64) 1.021 (0.986–1.059); p=0.244
%p2PSA 0.68 (0.58–0.76) 3.149 (1.478–6.710); p=0.003 5.859 (1.974–17.389); p=0.001
phi 0.72 (0.63–0.81) 1.056 (1.023–1.090); p=0.001 1.072 (1.025–1.122); p=0.003
AUC of multivariable models (95% CI) 0.74 (0.64–0.84) 0.81 (0.73–0.90) 0.81 (0.72–0.89)
Gain in predictive accuracy (95% CI) 0.07 (0.00–0.14)* 0.07 (0.00–0.14)*

* p>0.05 (relative to the base multivariate model; DeLong method).1Not included in base multivariate model because pc>0.25 in comparison between groups (data not shown).2Expressed as tPSA (pg/ml)/prostate volume (ml) to scaling OR in a more intelligible range.

AUC=area under the curve; phi=prostate health index; OR=odds ratio; CI=confidence interval; tPSA=total prostate-specific antigen; fPSA=free PSA; %fPSA=free to total PSA ratio; PSAD=PSA density; p2PSA=[-2]ProPSA; phi=prostate health index.

4. Discussion

It has already been established that both %fPSA and PSAD are significantly better than tPSA in predicting the presence of PCa at biopsy. However, although PSAD was more accurate than %fPSA in the prediction of PCa in the tPSA range of 2–4 ng/ml, it showed the same accuracy of %fPSA in the tPSA range of 4–10 ng/ml [5] and [18].

In the current study, both %p2PSA and phi showed significantly higher accuracy in predicting the presence of PCa at biopsy relative to tPSA, %fPSA, and PSAD in a group of patients with negative DRE where tPSA levels were not statistically significantly different between individuals with and without PCa. Specifically, in accuracy analyses, %p2PSA and phi were 23% more accurate than tPSA in detecting patients with PCa. Similarly, at 90% specificity, the sensitivities of phi (42.9%) and of %p2PSA (38.8%) were significantly higher than those of tPSA (5.1%), %fPSA (20.0%), and PSAD (26.5%). The inclusion of %p2PSA or phi in a multivariate logistic regression model that included the most informative clinical predictors of PCa in our population resulted in a 10% and an 11% increase of its predictive accuracy, respectively.

Although the accuracy of the base multivariable model (71%) was similar to the accuracy of several currently available tools predicting the presence of PCa at initial extended biopsies [6], the inclusion of %p2PSA and phi along with other PCa predictors resulted in a significantly higher predictive ability (81% and 82%, respectively). Unfortunately, we were not able to test the accuracy of these models directly because they included several variables that were not available in our patient population, such as DRE (all patients had negative DREs) and sampling density (the number of cores was standardized).

We supported the higher accuracy of %p2PSA and phi relative to %fPSA and PSAD in the subgroup of patients with tPSA 4–10 ng/ml. Conversely, we were not able to test the accuracy of these variables in the 2–4 ng/ml subgroup, due to the low number of individuals in this patient subgroup (n=50).

We also confirmed a direct relationship between p2PSA derivatives and PCa aggressiveness. Specifically, %p2PSA and phi also represented the most accurate predictors of PCa with a Gleason score ≥7, outperforming patient age and %fPSA. However, although %p2PSA and phi appeared to increased the accuracy of a multivariate model by 7%, this result did not achieve statistical significance due to the relatively small number of patients included in the analyses (n=107). Additionally, we demonstrated a significant relationship between Gleason score, %p2PSA (p=0.002), and phi levels (p<0.001). This information may be helpful to identify those men at an increased risk of harboring clinically relevant PCa in the range of 2–10 ng/ml. Similar results were shown by Sokoll et al, who observed a direct relationship between p2PSA levels and Gleason score in men with tPSA levels between 2 and 10 ng/ml [11]. Conversely, Jansen et al were not able to demonstrate unequivocally a relationship between Gleason score and p2PSA levels [13]. It is noteworthy that, in both studies, a central pathologic review was not performed.

Although supporting the finding of a superior accuracy of p2PSA and its derivatives in PCa detection and characterization relative to the reference standard tests (tPSA, %fPSA, and PSAD) [7], [8], [9], [10], [11], [12], and [13], our study also overcomes several limitations of the previously published studies [7], [8], [9], [10], [11], [12], and [13]. First, it consisted of a prospective observational study in a contemporary cohort of prostate biopsy candidates. Conversely, most of the previously published studies were based on a retrospective analysis, and consequently they were more susceptible to selection bias [9], [10], and [13]. Furthermore, no information was provided on how serum was obtained, managed, stored, and analyzed, whereas in our study, blood samples were managed according to Semjonow et al guidelines [14]. Our study appears to be the first where p2PSA and its derivatives were prospectively evaluated in a setting different from a screening program.

Second, to optimize the PCa detection rate, we used a standardized saturation biopsy scheme consisting of 18–22 cores [15]. Conversely, all previous studies did not adopt a standardized biopsy protocol [7], [8], [9], [10], [12], and [13]. Only in the study by Sokoll et al were patients subjected to at least 10 biopsy cores, but no further information was provided [11]. Third, to increase the reliability of our findings, all the specimens were analyzed by a single blinded, experienced genitourinary pathologist.

Despite its strengths, our study is not devoid of limitations. p2PSA is Hybritech calibrated and can be used for phi calculation only in combination with Hybritech tPSA and fPSA, both Hybritech and World Health Organization calibrated. In addition, a selection bias may limit the results of the current study because both cases and controls were selected on the basis of their tPSA values, thus resulting in a low ability of tPSA to determine the presence of PCa. The relatively small sample size of our study and the lack of race and family history information represent additional limitations that underscore the need for larger multi-institutional validation studies. Finally, due to the relatively small number of patients in the study cohort, we did not determine phi or %p2PSA cut-off values that may be used in clinical practice. Multi-institutional efforts aimed at determining the best cut-off values according to the patient characteristics are warranted.

5. Conclusions

In patients with a tPSA between 2.0 and 10 ng/ml, %p2PSA and phi are the strongest predictors of PCa at initial extended biopsies and are significantly more accurate than the currently used tests (tPSA, %fPSA, and PSAD) in determining the presence of PCa at biopsy. The implementation of %p2PSA and phi in clinical practice may significantly increase our ability to detect PCa, lowering the rate of unnecessary biopsies and reducing the rate of overtreatment in patients with clinically insignificant PCa. Further multicentric confirmatory studies are mandatory.

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, Nava, Lazzeri.

Acquisition of data: Maccagnano, Scattoni, Freschi, Dorigatti, Ceriotti, Pontillo.

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

Drafting of the manuscript: Lazzeri, Lughezzani, Scattoni.

Critical revision of the manuscript for important intellectual content: Montorsi.

Statistical analysis: Bini, Lughezzani, Scattoni.

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: Access Hybritech p2PSA ([-2]proPSA) reagents were in part provided by Beckman Coulter Inc. and Beckman Coulter Italy; Access 2 Immunoassay system was provided by Beckman Coulter Italy.

References

  • [1] F.H. Schroder, M.J. Roobol. Defining the optimal prostate-specific antigen threshold for the diagnosis of prostate cancer. Curr Opin Urol. 2009;19:227-231 Crossref.
  • [2] E.A. Heijnsdijk, A. der Kinderen, E.M. Wever, et al. Overdetection, overtreatment and costs in prostate-specific antigen screening for prostate cancer. Br J Cancer. 2009;101:1833-1838 Crossref.
  • [3] I.M. Thompson, D.K. Pauler, P.J. Goodman, et al. Prevalence of prostate cancer among men with a prostate-specific antigen level < or=4.0ng per milliliter. N Engl J Med. 2004;350:2239-2246 Crossref.
  • [4] A.W. Roddam, M.J. Duffy, F.C. Hamdy, et al. Use of prostate-specific antigen (PSA) isoforms for the detection of prostate cancer in men with a PSA level of 2–10 ng/ml: systematic review and meta-analysis. Eur Urol. 2005;48:386-399 Abstract, Full-text, PDF, Crossref.
  • [5] C. Stephan, G. Stroebel, M. Heinau, et al. The ratio of prostate-specific antigen (PSA) to prostate volume (PSA density) as a parameter to improve the detection of prostate carcinoma in the range of < 4 ng/mL. Cancer. 2005;104:993-1003 Crossref.
  • [6] F.K.-H. Chun, J.I. Epstein, V. Ficarra, et al. Optimizing performance and interpretation of prostate biopsy: a critical analysis of the literature. Eur Urol. 2010;58:851-864 Abstract, Full-text, PDF, Crossref.
  • [7] S.D. Mikolajczyk, L.S. Millar, T.J. Wang, et al. A precursor form of prostate-specific antigen is more highly elevated in prostate cancer compared with benign transition zone prostate tissue. Cancer Res. 2000;60:756-759
  • [8] W.J. Catalona, G. Bartsch, H.G. Rittenhouse, et al. Serum pro prostate specific antigen improves cancer detection compared to free and complexed prostate specific antigen in men with prostate specific antigen 2 to 4 ng/ml. J Urol. 2003;170:2181-2185 Crossref.
  • [9] L.J. Sokoll, Y. Wang, Z. Feng, et al. [-2]Proenzyme prostate specific antigen for prostate cancer detection: a National Cancer Intitute early detection research network validation study. J Urol. 2008;180:539-543 Crossref.
  • [10] C. Stephan, A.M. Kahrs, H. Cammann, et al. A [-2]proPSA-based artificial neural network significantly improves differentiation between protate cancer and benign prostatic diseases. Prostate. 2009;69:198-207 Crossref.
  • [11] L.J. Sokoll, M.G. Sanda, Z. Feng, 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;19:1193-1200 Crossref.
  • [12] B.V. Le, C.R. Griffin, S. Loeb, 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;183:1355-1359 Crossref.
  • [13] F.H. Jansen, R.H.N. Van Schaik, J. Kurstjens, et al. Prostate-specific antigen (PSA) isoform p2PSA in combination with total PSA and free PSA improves diagnostic accuracy in prostate cancer detection. Eur Urol. 2010;57:921-927 Abstract, Full-text, PDF, Crossref.
  • [14] A. Semjonow, T. Köpke, E. Eltze, et al. Pre-analytical in-vitro stability of [-2]proPSA in blood and serum. Clin Biochem. 2010;47:926-928 Crossref.
  • [15] V. Scattoni, M. Raber, F. Abdollah, et al. Biopsy schemes with the fewest cores for detecting 95% of the prostate cancers detected by a 24-core biopsy. Eur Urol. 2010;57:1-8 Abstract, Full-text, PDF, Crossref.
  • [16] J.I. Epstein, W.C. Allsbrook Jr., M.B. Amin, et al. The 2005 International Society of Urological Pathology (ISUP) Consensus Conference on Gleason Grading of Prostatic Carcinoma. Am J Surg Pathol. 2005;29:1228-1242 Crossref.
  • [17] D.W. Hosmer, S. Lemeshow. Applied logistic regression. (Wiley, New York, NY, 2000)
  • [18] E.R. DeLong, D.M. DeLong, D.L. Clarke-Pearson. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics. 1988;44:837-845 Crossref.

Footnotes

a Department of Urology, San Raffaele Turro, Vita-Salute San Raffaele University, Milan, Italy

b Department of Urology, Vita-Salute San Raffaele University, Milan, Italy

c Diagnostica e Ricerca San Raffaele S.P.A., San Raffaele Scientific Institute, Milan, Italy

d Department of Internal Medicine, University of Perugia, Italy

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

lowast Corresponding author. Department of Urology, San Raffaele Turro, Vita-Salute San Raffaele University, Via Stamira D’Ancona 20, 20127 Milan, Italy. Tel. +39 022643.3357; Fax: +39 0226433442.

Place a comment

Your comment *

max length: 5000