Back

Platinum Priority – Prostate Cancer
Editorial by Michael R. Abern and Stephen J. Freedland on pp. 210–211 of this issue

Initial Prostate Biopsy: Development and Internal Validation of a Biopsy-specific Nomogram Based on the Prostate Cancer Antigen 3 Assay

By: Jens Hansena b , Marco Aupricha b , Sascha A. Ahyaia, Alexandre de la Taillec, Hendrik van Poppeld, Michael Marbergere, Arnulf Stenzlf, Peter F.A. Muldersg, Hartwig Hulandb, Margit Fischa, Clement-Claude Abbouc, Jack A. Schalkeng, Yves Fradeth, Leonard S. Marksi, William Ellisj, Alan W. Partink, Karl Pummerl, Markus Graefenb, Alexander Haeseb, Jochen Walzm, Alberto Brigantin, Shahrokh F. Shariato and Felix K. Chuna lowast

European Urology, Volume 63 Issue 2, February 2013, Pages 201-209

Published online: 01 February 2013

Keywords: Biomarker, Decision curve analysis, Nomogram, Initial prostate biopsy, Internal validation, Prostate cancer, Prostate cancer antigen 3

Abstract Full Text Full Text PDF (340 KB)

Abstract

Background

Urinary prostate cancer antigen 3 (PCA3) assay in combination with established clinical risk factors improves the identification of men at risk of harboring prostate cancer (PCa) at initial biopsy (IBX).

Objective

To develop and validate internally the first IBX-specific PCA3-based nomogram that allows an individual assessment of a man's risk of harboring any PCa and high-grade PCa (HGPCa).

Design, setting, and participants

Clinical and biopsy data including urinary PCA3 score of 692 referred IBX men at risk of PCa were collected within two prospective multi-institutional studies.

Intervention

IBX (≥10 biopsy cores) with standard risk factor assessment including prebiopsy urinary PCA3 measurement.

Outcome measurements and statistical analysis

PCA3 assay cut-off thresholds were investigated. Regression coefficients of logistic risk factor analyses were used to construct specific sets of PCA3-based nomograms to predict any PCa and HGPCa at IBX. Accuracy estimates for the presence of any PCa and HGPCa were quantified using area under the curve of the receiver operator characteristic analysis and compared with a clinical model. Bootstrap resamples were used for internal validation. Decision curve analyses quantified the clinical net benefit related to the novel PCA3-based IBX nomogram versus the clinical model.

Results and limitations

Any PCa and HGPCa were diagnosed in 46% (n=318) and 20% (n=137), respectively. Age, prostate-specific antigen, digital rectal examination, prostate volume, and PCA3 were independent predictors of PCa at IBX (all p<0.001). The PCA3-based IBX nomograms significantly outperformed the clinical models without PCA3 (all p<0.001). Accuracy was increased by 4.5–7.1% related to PCA3 inclusion. When applying nomogram-derived PCa probability thresholds ≤30%, only a few patients with HGPCa (≤2%) will be missed while avoiding up to 55% of unnecessary biopsies. External validation of the PCA3-based IBX-specific nomogram is warranted.

Conclusions

The internally validated PCA3-based IBX-specific nomogram outperforms a clinical prediction model without PCA3 for the prediction of any PCa, leading to the avoidance of unnecessary biopsies while missing only a few cases of HGPCa. Our findings support the concepts of a combination of novel markers with established clinical risk factors and the superiority of decision tools that are specific to a clinical scenario.

Take Home Message

This novel, internally validated, initial biopsy–specific nomogram based on the prostate cancer antigen 3 (PCA3) assay outperforms a clinical prediction model without PCA3 for the prediction of any prostate cancer, thus avoiding unnecessary biopsies while missing only a few cases of high-grade prostate cancer.

Keywords: Biomarker, Decision curve analysis, Nomogram, Initial prostate biopsy, Internal validation, Prostate cancer, Prostate cancer antigen 3.

1. Introduction

In current clinical practice, only about 4 of 10 men who are suspected of harboring prostate cancer (PCa) due to an elevated prostate-specific antigen (PSA) level >4 ng/ml or a suspicious digital rectal examination (DRE) show a positive initial prostate biopsy (IBX) [1] and [2]. In other words, one could argue that about 60% of these IBX-negative men could have been saved from further prostatic evaluations. Instead, most of these IBX-negative men most likely remain suspicious and undergo even further subsequent prostatic evaluations consisting of serial PSA tests, DREs, and potentially multiple repeated prostate biopsies. These are associated with potential adverse biopsy-related events such as infections, bleeding, or urinary obstruction [3] as well as the emotional confrontation with a potential cancer diagnosis [4]. Thus it is of great clinical importance to assess PCa risk correctly prior to an IBX to prevent a significant proportion of men from having to undergo such a vicious cycle.

Previously reported prostate biopsy nomograms for IBX settings, which are based mainly on clinical variables, demonstrated satisfactory predictive accuracy (PA) estimates between 67% and 77%, respectively [1], [2], [5], [6], [7], [8], and [9]. Novel diagnostic markers, which carry the potential to overcome PSA's low specificity in the diagnostic arena, may help improve PA estimates of such nomograms. Several studies have shown that a commercially available urinary prostate cancer antigen 3 (PCA3) assay in combination with established clinical risk factors may support clinicians to better identify men at risk of PCa [9], [10], [11], [12], [13], [14], [15], [16], and [17]. Its principle resides in PCA3's messenger mRNA overexpression in malignant prostate tissue, detected in the patient's urine after DRE [18]. Consequently, one externally validated PCA3-based mixed-biopsy nomogram to assess an individual's PCa risk became available [9]. Unfortunately, this robust nomogram is not biopsy-scenario specific because it is based on IBX and repeat biopsy men. Therefore, clinicians may be reluctant to use it because the odds of harboring PCa significantly change with the number of biopsy sessions [19].

To address this void, we developed and validated the first PCA3-based IBX-specific nomogram that allows a more accurate identification of any PCa and high-grade prostate cancer (HGPCa), as well as avoidance of unnecessary IBX.

2. Material and methods

2.1. Patient population

Our study cohort consisted of 772 men subjected to ≥10 cores at IBX due to suspicious DRE and/or elevated total PSA levels (2.5–10 ng/ml) from two prospective multi-institutional studies from Europe and North America [12] and [13]. Patients with symptoms of urinary tract infections, any medical therapy affecting PSA levels, and/or a history of PCa or invasive treatment for benign prostatic hyperplasia were not recruited for the studies, as previously described [9]. Patients with suspicious DREs were included, even if their PSA levels were >10 ng/ml. However, due to clinical as well as statistical considerations, men with a PSA ≥20 ng/ml (n=5) were not considered for the IBX nomogram development, and 75 men were excluded due to missing information on prostate volume. This resulted in 692 men for our analyses. Independent ethics committees approved the study protocol, and informed consent was obtained from all patients.

2.2. Clinical evaluation

PSA levels were measured before DRE and transrectal ultrasound (TRUS). DRE findings were classified as unsuspicious versus suspicious. TRUS-derived total prostate volume was calculated in all patients using the prolate ellipse formula (0.52×length×width×height) [20]. Urinary PCA3 scores were assessed according to the manufacturer's instructions [21]. In all individuals, ≥10-core systematic laterally directed TRUS-guided biopsies were performed. All biopsy specimens were evaluated by an experienced uropathologist at each participating center.

2.3. Statistical analyses

The PCA3 assay score was explored with respect to a possible cut-off value that could be more informative than the unaltered continuous variable and previously used cut-off formats [9] and [15]. The PCA3 cut-off value (≤21 vs >21) was identified using the minimum p value approach [22]. Previously reported PCA3 cut-offs were used in subsequent univariable and multivariable logistic regression models (LRMs). Multicolinearity analyses between PCA3 and other clinical variables, which were included in the multivariate model, were also performed. Invariably, PCA3 score was not correlated to any other variable.

LRMs addressed the presence of any PCa at IBX. Risk factors were age, DRE, PSA, prostate volume, and PCA3. The number of biopsy cores was not considered because we aimed to develop a model that helps to better identify men suspicious of harboring PCa prior to the first prostate biopsy. Multivariable LRM coefficients were used to construct various nomogram sets with and without PCA3. Area under the curve (AUC) of the receiver operator characteristic analysis was used to quantify the PA of each multivariable model predicting any PCa at IBX. The AUC of high-grade PCa prediction (HGPCa, defined as Gleason sum ≥7) was derived from the predicted probabilities of any PCa according to a method previously described by Vickers et al. [23]. In this analysis, men with IBX Gleason sum ≤6 were classified the same way as men with negative IBX. For internal validation and to reduce overfit bias, 200 bootstrap resamples were used [24]. Model PA differences were tested for statistical significance using the Mantel-Haenszel test. The extent of over- or underestimation of the observed versus predicted PCa rate at biopsy was explored graphically. Various nomogram-derived probability cut-offs were tested to assess the ability to predict PCa and HGPCa.

Finally, to determine whether PCA3 inclusion in a clinical model improves its predictive net benefit, we relied on decision curve analyses (DCAs) [25]. DCAs examine the theoretical relationship between the threshold probability of IBX outcome (PCa) and the relative value of false-positive and false-negative results to determine the value (net benefit) of a predictive model [25].

All tests were two sided with a statistical significance set at p<0.05. Analyses were conducted using the R statistical package (R Foundation for Statistical Computing, v.2.1.13).

3. Results

Overall, 46% of the men (n=318) were diagnosed with PCa and HGPCa was detected in 20% (n=137) of men at IBX. Men with PCa at biopsy had significantly higher PCA3 scores (median: 46; interquartile range [IQR]: 23–82 vs median: 17; IQR: 34–60), higher PSA levels (median: 5.9 ng/ml; IQR: 4.5–7.9 ng/ml vs 5 ng/ml; IQR: 4.1–6.6 ng/ml), higher rates of suspicious DRE (38%, n=122 vs 20%, n=75), and smaller prostates (median: 36cm3; IQR: 28–47cm3 vs median: 45cm3; IQR: 34–60cm3) relative to men with negative biopsies, respectively (all p<0.001). Patient characteristics are displayed in Table 1 after stratification according to absence of PCa, presence of low-grade PCa (Gleason ≤6), and presence of HGPCa (Gleason ≥7), respectively.

Table 1 Prostate cancer risk factors at initial biopsy (n=692)

Entire initial biopsy cohort No cancer at initial biopsy Low-grade PCa at initial biopsy HGPCa at initial biopsy p value
No. of patients (%) 692 (100) 374 (54) 181 (26) 137 (20)
Age, yr <0.001
Median 64 62 65 67
IQR 58–69 57–68 59–70 60–72
PSA, ng/mL <0.001
Median 5.2 5 5.4 6.3
IQR 4.3–7.2 4.1–6.6 4.2–7.5 4.8–8.3
PSA ≥10.1 ng/ml 28 (4.0) 8 (2.1) 13 (7.2) 7 (5.1) 0.01
Total prostate volume, cm3 <0.001
Median 40 45 38 34
IQR 30–55 34–60 29–50 27–45
DRE <0.001
Unsuspicious, no. (%) 495 (71.5) 299 (79.9) 138 (76.2) 15 (10.9)
Suspicious, no. (%) 197 (28.5) 75 (20.1) 43 (23.8) 122 (89.1)
PCA3 assay score <0.001
Median 27 17 42 55
IQR 12–59 10–34 19–74 29–91
≤17, no. (%) 239 (34.5) 189 (50.5) 37 (20.4) 13 (9.5) <0.001
>17, no. (%) 453 (65.5) 185 (49.5) 144 (79.6) 124 (90.5)
≤21, no. (%) 287 (41.5) 220 (58.8) 52 (28.7) 15 (10.9) <0.001
>21, no. (%) 405 (58.5) 154 (41.2) 129 (71.3) 122 (89.1)
≤24, no. (%) 319 (46.1) 235 (62.8) 62 (34.3) 22 (16.1) <0.001
>24, no. (%) 373 (53.9) 139 (37.2) 119 (65.7) 115 (83.9)
≤35, no. (%) 412 (59.5) 284 (75.9) 85 (47.0) 43 (31.4) <0.001
>35, no. (%) 280 (40.5) 90 (24.1) 96 (53.0) 94 (68.6)
Biopsy cores, no. 0.01
Median 10 10 10 10
IQR 10–12 10–12 10–12 10–12
Positive cores, no. <0.001
Median 2 5
IQR 1–3 3–7
Biopsy Gleason sum, no. (%)
≤6 181 (100)
7 114 (82.6)
≥8 24 (17.4)

PCa=prostate cancer; HGPCa=high-grade PCa (defined as biopsy Gleason sum ≥7); IQR=interquartile range; PSA=prostate-specific antigen; DRE=digital rectal examination; PCA3=prostate cancer antigen 3.

In univariable and multivariable LRMs, PCA3 (regardless of its coding), PSA, DRE, and prostate volume achieved independent predictor status of PCa at IBX (all p<0.05; Table 2 and Table 3). In univariable LRMs, continuously coded PCA3 represented the most informative parameter in the prediction of any PCa (AUC: 0.739) and HGPCa (AUC: 0.729), respectively, followed by the PCA3 cut-off of 21 (AUC: 0.689 for any PCa and 0.690 for HGPCa, respectively; Table 2). Each model's bias-corrected accuracy is displayed in Table 3. The PCA3-based nomogram (PCA3 cut-off: 21) demonstrated the highest bias-corrected AUC of 80.7% in the prediction of any PCa and resulted in a significant accuracy increment of 7.1% (p<0.001), compared with the base model without PCA3 (Table 3). Applying the models for the prediction of HGPCa, the PCA3-based model with a cut-off of 21 showed the highest AUC with 82.9%, leading to a significant gain of 5.4% accuracy (p<0.001). No significant differences in terms of accuracy increment were recorded between the five different PCA3-based models (all p>0.05). Figure 1 displays the PCA3-based nomogram (PCA3 cut-off: 21) with its associated calibration plot.

Table 2 Univariable logistic regression models predicting any prostate cancer and high-grade prostate cancer at initial biopsy

Variables Any PCa HGPCa*
OR (95% CI) p value AUC (95% CI) AUC (95% CI)
Age, yr 1.04 (1.03–1.06) <0.001 0.606 (0.569–0.642) 0.615 (0.579–0.652)
Serum PSA, ng/mL 1.16 (1.09–1.24) <0.001 0.600 (0.564–0.637) 0.634 (0.599–0.670)
Prostate volume, cm3 0.98 (0.97–0.98) <0.001 0.651 (0.616–0.687) 0.639 (0.604–0.675)
DRE (suspicious vs unsuspicious) 2.48 (1.77–3.48) <0.001 0.591 (0.554–0.628) 0.682 (0.647–0.717)
PCA3 assay score, continuously coded 1.01 (1.01–1.02) <0.001 0.739 (0.706–0.771) 0.729 (0.696–0.762)
PCA3 assay score >17 vs ≤17 5.48 (3.81–7.88) <0.001 0.674 (0.639–0.709) 0.656 (0.621–0.691)
PCA3 assay score >21 vs ≤21 5.35 (3.81–7.51) <0.001 0.689 (0.654–0.723) 0.690 (0.656–0.725)
PCA3 assay score >24 vs ≤24 4.71 (3.4–6.52) <0.001 0.682 (0.647–0.717) 0.687 (0.653–0.722)
PCA3 assay score >35 vs ≤35 4.68 (3.38–6.49) <0.001 0.678 (0.644–0.713) 0.675 (0.641–0.710)

* The AUC for the prediction of HGPCa was calculated from the predicted probabilities of any PCa according to a method previously described by Vickers et al. [23]. In this analysis, men with initial biopsy (IBX) Gleason sum ≤6 were classified the same as men with negative IBX.

PCa=prostate cancer; HGPCa=high-grade prostate cancer (defined as biopsy Gleason sum ≥7); OR=odds ratio; CI=confidence interval; AUC=area under the curve; PSA=prostate-specific antigen; DRE=digital rectal examination; PCA3=prostate cancer antigen 3.

Table 3 (a) Multivariable logistic regression models predicting any prostate cancer at initial biopsy; (b) areas under the curve for prediction of high-grade prostate cancer at initial biopsy*

Variables Multivariable analyses predicting any PCa at IBX
Base model Base model plus PCA3 (cc) Base model plus PCA3-17 Base model plus PCA3-21 Base model plus PCA3-24 Base model plus

PCA3-35
OR (95% CI); p OR (95% CI); p OR (95% CI); p OR (95% CI); p OR (95% CI); p OR (95% CI); p
(a)
Age, yr 1.05 (1.03–1.07); <0.001 1.03 (1.01–1.06); 0.005 1.02 (1–1.05); 0.04 1.02 (1–1.05); 0.06 1.02 (1–1.05); 0.04 1.03 (1–1.05); 0.02
Serum PSA, ng/ml 1.24 (1.15–1.33); <0.001 1.24 (1.14–1.33); <0.001 1.24 (1.15–1.34); <0.001 1.23 (1.14–1.33); <0.001 1.23 (1.14–1.33); <0.001 1.23 (1.14–1.33); <0.001
Prostate volume, cm3 0.97 (0.96–0.98); <0.001 0.97 (0.96–0.98); <0.001 0.97 (0.96–0.98); <0.001 0.97 (0.96–0.98); <0.001 0.97 (0.96–0.98); <0.001 0.97 (0.96–0.98); <0.001
DRE (suspicious vs unsuspicious) 1.79 (1.23–2.59); 0.002 1.76 (1.20–2.58); 0.004 1.88 (1.27–2.79); 0.002 1.76 (1.19–2.61); 0.005 1.78 (1.21–2.64); 0.004 1.81 (1.22–2.78); 0.003
PCA3 assay score 1.01 (1.01–1.02); <0.001 5.14 (3.43–7.69); <0.001 4.93 (3.38–7.20); <0.001 4.19 (2.92–6.02); <0.001 3.94 (2.75–5.65); <0.001
Bias-corrected AUC for prediction of any PCa (%)

(95% CI)
73.6

(70.6–77.2)
78.1

(75.1–81.3)
80.4

(76.5–82.5)
80.7

(76.8–82.8)
78.8

(76.1–82.1)
78.6

(75.7–81.8)
Increment AUC (%);

Mantel-Haenszel test (p value)


4.5;

<0.001
6.8;

<0.001
7.1;

<0.001
5.2;

<0.001
5.0;

<0.001
(b)
Bias-corrected AUC for prediction of HGPCa (%)§

(95% CI)
77.5

(74.4–80.6)
80.9

(77.9–83.8)
81.7

(78.8–84.6)
82.9

(80.1–85.7)
82.3

(79.5–85.2)
81.1

(78.1–84.0)
Increment AUC (%);

Mantel-Haenszel test (p value)


3.4;

0.008
4.2;

<0.001
5.4;

<0.001
4.8;

<0.001
3.6;

<0.001

* Head-to-head comparison of predictive accuracy estimates of the clinical base nomogram versus different PCA3-based nomograms for prediction of any prostate cancer and high-grade prostate cancer.

Comparison of AUC increment was made for each PCA3-based nomogram to the clinical base model.

The clinical base model consisted of age, PSA, DRE, and prostate volume.

§ The AUC for prediction of HGPCa was calculated from the predicted probabilities of any PCa according to a method previously described by Vickers et al. [23]. In this analysis, men with IBX Gleason sum ≤6 were classified the same as men with negative IBX.

PCa=prostate cancer; IBX=initial biopsy; OR=odds ratio; CI=confidence interval; PCA3 (cc)=prostate cancer antigen 3 assay score, continuously coded; PCA3-17=PCA3 assay score threshold 17; PCA3-21=PCA3 assay score threshold 21; PCA3-24=PCA3 assay score threshold 24; PCA3-35=PCA3 assay score threshold 35; PSA=prostate-specific antigen; DRE=digital rectal examination; AUC=area under the curve; HGPCa=high-grade prostate cancer (defined as biopsy Gleason sum ≥7).

gr1

Fig. 1 (a) Prostate cancer antigen 3 (PCA3) assay score nomogram predicting prostate cancer on initial biopsy (IBX); (b) local regression nonparametric smoothing plot showing the calibration of the PCA3-based IBX nomogram. Instructions for physicians: To obtain nomogram-predicted probability of prostate cancer (PCa), locate patient values at each axis. Draw a vertical line to the “Point” axis to determine how many points are attributed for each variable value. Sum the points for all variables. Locate the sum on the “Total points” line to be able to assess the individual probability of cancer on IBX on the “PCa probability” line. Instructions for readers: Perfect predictions correspond to the 45° line. Points estimated below the 45° line correspond to nomogram overprediction, whereas points situated above the 45° line correspond to nomogram underprediction. A nonparametric smoothed curve indicates the relationship between predicted probability and observed frequency of PCa at IBX. Vertical lines indicate the frequency distribution of predicted probabilities. PSA=prostate-specific antigen.

In DCAs, the PCA3-based IBX nomogram (PCA3 cut-off: 21) was clearly superior to the base model without PCA3 with a higher net benefit for all threshold probabilities >18% (Fig. 2). For example, applying a nomogram-derived probability threshold of 30% above which a man would be considered to harbor PCa, use of the PCA3-based IBX nomogram would result in a net reduction of 16 unnecessary prostate biopsies per 100 patients without a decrease in the number of men with PCa who duly have an IBX, compared with the scenario where all men undergo IBX [25]. The net reduction of unnecessary biopsies associated with the use of the PCA3-based IBX nomogram relative to a base model without PCA3 was 9 per 100 biopsies. Specifically, IBX could be avoided in 207 men without PCa (55%), and 41 men with PCa (13%) would not undergo IBX. More importantly, of these 41 men only 3 (2%) harbor HGPCa (all Gleason score 3+4; Table 4). In comparison, using the base model without PCA3, 10 men (7%) with HGPCa would have been incorrectly advised against IBX.

gr2

Fig. 2 Decision curve analysis of predicting prostate cancer (PCa) on initial biopsy (IBX) using prostate cancer antigen 3 (PCA3) assay score (cut-off: 21), age, prostate-specific antigen (PSA) level, digital rectal examination (DRE) result, and prostate volume compared with age, PSA level, DRE result, and prostate volume in 692 patients with either suspicious DRE or elevated PSA levels. The dashed line indicates the prediction model that included PCA3, age, PSA level, DRE result, and prostate volume; the solid black line shows the prediction model that included only age, PSA level, DRE result, and prostate volume. The horizontal line along the x-axis assumes that no patient will have PCa (ie, no patient should undergo a prostate biopsy), whereas the solid gray line assumes that all patients will have PCa (ie, all patients will need to undergo prostate biopsy). For example, above a threshold probability of 18%, the PCA3-based IBX nomogram is superior to the base model and to a scenario where all patients would be treated.

Table 4 Numbers of biopsies performed and detection rates of any prostate cancer and high-grade prostate cancer (Gleason sum ≥7) according to different nomogram-derived probability cut-offs of the prostate cancer antigen 3 assay score (PCA3)-based initial biopsy nomogram (PCA3 cut-off: 21) and the base model without PCA3

Model Probability cut-off, % Biopsies performed, n (%) Biopsies not performed, n (%) Biopsies not performed in men without PCaa, n (%) Any PCa detectedb,n (%) Any PCa missed, n (%) NPV for any PCa, % HGPCa detectedb,n (%) HGPCa missed, n (%) NPV for HGPCa, %
All biopsies 692 (100) NA NA 318 (46) NA 100 137 (100) NA 100
Base model§ 15 657 (95) 35 (5) 32 (9) 315 (99) 3 (1) 91 137 (100) 0 (0) 100
20 613 (89) 79 (11) 68 (18) 307 (97) 11 (3) 86 135 (98.5) 2 (1.5)* 97
25 568 (82) 124 (18) 98 (26) 292 (92) 26 (8) 79 135 (96) 6 (4)* 95
30 525 (76) 167 (24) 131 (35) 282 (89) 36 (11) 78 134 (93) 10 (7)* 94
35 469 (68) 223 (32) 170 (45) 265 (83) 53 (17) 76 129 (91) 13 (9)** 94
40 410 (59) 282 (41) 208 (56) 244 (77) 74 (23) 74 126 (85) 20 (15)*** 93
Basel model§ plus PCA3 15 597 (86) 95 (14) 84 (22) 307 (97) 11 (3) 88 136 (99.3) 1 (0.7) 99
20 536 (77) 156 (23) 136 (36) 298 (94) 20 (6) 87 135 (98.5) 2 (1.5) 99
25 492 (71) 200 (29) 172 (46) 290 (91) 28 (9) 86 135 (98.5) 2 (1.5) 99
30 444 (64) 248 (36) 207 (55) 277 (87) 41 (13) 83 134 (98) 3 (2) 99
35 405 (59) 287 (41) 232 (62) 263 (83) 55 (17) 81 129 (94) 8 (6)‡‡ 97
40 384 (55) 308 (45) 245 (66) 255 (80) 63 (20) 80 126 (92) 11 (8)‡‡ 96

§ The clinical base model consisted of age, PSA, DRE, and prostate volume.

a Percentage is indicative of specificity.

b Percentage is indicative of sensitivity.

* One man had Gleason score 4+4; the other men had Gleason score 3+4.

** One man had Gleason score 4+4; one man had Gleason score 4+3; the other men had Gleason score 3+4.

*** One man had Gleason score 4+4; three men had Gleason score 4+3; the other men had Gleason score 3+4.

All men had Gleason score 3+4.

‡‡ One man had Gleason score 4+4; the other men had Gleason score 3+4.

PCa=prostate cancer; NPV=negative predictive value; HGPCa=high-grade prostate cancer (prostate cancer with biopsy Gleason sum ≥7); PCA3=prostate cancer antigen 3 assay score.

To further substantiate our findings, an additional PCA3-based nomogram to specifically predict HGPCa at IBX (data not shown) was built. However, in direct comparison, it neither showed superior performance nor did it allow detection of more HGPCa relative to the present PCA3-based nomogram to predict any PCa (Table 5).

Table 5 Numbers of biopsies performed and detection rates of high-grade prostate cancer (Gleason sum ≥7)*

Model Probability cut-off, % Biopsies performed, n (%) Biopsies not performed, n (%) HGPCa detected, n (%) HGPCa missed,n (%) NPV, %
PCA3-based IBX nomogram to predict any PCa 15 597 (86) 95 (14) 136 (99.3) 1 (0.7) 99
20 536 (77) 156 (23) 135 (98.5) 2 (1.5) 99
25 492 (71) 200 (29) 135 (98.5) 2 (1.5) 99
30 444 (64) 248 (36) 134 (98) 3 (2) 99
35 405 (59) 287 (41) 129 (94) 8 (6) 97
PCA3-based IBX nomogram specifically to predict HGPCa 15 329 (48) 363 (52) 120 (88) 17 (12) 95
20 256 (37) 436 (63) 104 (76) 33 (24) 92
25 194 (28) 498 (72) 91 (66) 46 (34) 91
30 158 (23) 534 (77) 78 (57) 59 (43) 89
35 141 (20) 551 (80) 71 (52) 66 (48) 88

* According to different nomogram-derived probability cut-offs of the PCA3 assay–based IBX nomogram (PCA3 cut-off: 21) to predict any PCa relative to a PCA3-based IBX nomogram to predict specifically HGPCa.

Percentage indicates sensitivity.

PCA3=prostate cancer antigen 3 (assay score); PCa=prostate cancer; HGPCa=high-grade prostate cancer (biopsy Gleason sum ≥7 was considered high grade); IBX=initial prostate biopsy; NPV=negative predictive value.

4. Discussion

We used data of a large multi-institutional PCA3 cohort of European and North American patients who underwent IBX to successfully develop the first IBX-specific PCA3-based nomogram. After bias correction, accuracies of this novel clinical decision tool are 80.7% for prediction of any PCa and 82.9% for prediction of HGPCa, respectively, which is remarkably high (Table 5). Of note, these accuracies were only achieved by combining PCA3 with established clinical risk factors. Application of the PCA3-based IBX nomogram may avoid unnecessary IBX in a considerable number of men at the expense of missing only a few men harboring HGPCa (Table 4), and inclusion of PCA3 led to a more distinct reduction of unnecessary biopsies relative to a clinical base model (age, PSA, DRE, and prostate volume) without PCA3 (Table 4). An additional PCA3-based nomogram to specifically predict HGPCa was not superior to the present PCA3-based nomogram to predict any PCa. Consequently, only the PCA3-based IBX nomogram to predict any PCa is presented and discussed.

In total, any PCa was detected in 46% of patients, whereas 20% had HGPCa (Gleason sum ≥7), which appears comparable with the 42% reported by Nam et al. [1], albeit rates reported from screening trials are lower (24.5%) [26]. This may primarily be explained by differences between early detection and screening cohorts. And in contrast to the current study, the vast majority of patients in the screening trials underwent a sextant biopsy [26] and [27].

Regarding overall accuracy, the novel PCA3-based IBX nomogram compares favorably with previously published models [2], [3], [5], [6], [7], [8], and [9] as displayed in Table 6. However, comparison of different decision tools can be difficult due to inherent differences of development populations that cannot all be accounted for. Due to missing clinical information, we were unable to perform a true head-to-head comparison. For example, Zaytoun et al. [2] or Nam et al. [1] recently developed IBX prediction tools relying on age, PSA, percentage of free PSA (%fPSA), DRE, race, and family history. Unfortunately, the latter two variables were unavailable in our patient cohort (Table 1), and %fPSA was not considered in our study because not all of the assessed patients had available %fPSA information. Thus the size of the study cohort would have been too small.

Table 6 Overview of current available initial prostate biopsy nomograms

Study Patients, n Biopsy setting Variables Predictive accuracy, % Validation type
Karakiewicz et al. [6] 1762 IBX Age, PSA, DRE, %fPSA 77 External
Nam et al. [1] 3108 IBX Age, race, family history of PCa, symptom score, PSA, %fPSA, DRE 74 Split sample
Chun et al. [5] 2900 IBX Age, DRE, PSA, %fPSA, sampling density 77 External
Kawamura et al. [7] 1037 IBX; ethnic specific (Japanese) Age, DRE, PSA, %fPSA, PV 73 Split sample
Sooriakumaran et al. [8] 599 IBX Age, PSA, DRE, %fPSA 72 External
Chun et al. [9] 809 Mixed biopsy Age, PSA, DRE, PV, biopsy history, PCA3 75 External
Zaytoun et al. [2] 1551 IBX Age, PSA, %fPSA, family history of PCa, abnormal DRE, race 73 Split sample
Present study 692 IBX Age, PSA, DRE, PV, PCA3 (cut-off: 21) 81 Internal

IBX=initial biopsy; PSA=prostate-specific antigen; DRE=digital rectal examination; PV=prostate volume; %fPSA=percent free prostate-specific antigen; PCa=prostate cancer; PCA3=prostate cancer antigen 3 (assay score).

In contradistinction to those nomograms [1] and [2], we demonstrate within the present PCA3-based IBX nomogram (Fig. 1) that consideration of prostate volume within biopsy nomograms is crucial, as previously shown [7], [9], [26], and [28]. Different biologic and statistical explanations may be advanced to explain prostate volume's effect within multivariable diagnostic models [29], but a more practical consideration may be even more important. Specifically, location and direction of biopsy cores need to be adjusted for volume already in the IBX setting and not only within repeat biopsy decision tools.

A disadvantage of our novel PCA3 IBX nomogram resides in the nature of its validation [1] and [2]. Clearly, an external validation represents the gold standard. However, first calibration of the novel PCA3 IBX nomogram (Fig. 1b) virtually overlaps the line of perfect predictions. Second, although not the gold standard, an internal validation using 200 bootstrap resamples were used indicating the robustness of our novel tool. Finally, the transatlantic, multi-institutional nature of our study supports our hypothesis that after validation this novel tool will be applicable to both North American and European patients. Taken together, we demonstrate the first IBX-specific, internally validated PCA3-based nomogram based on most contemporary international patients with the so far highest reported overall discriminative ability and performance.

Another recurrent question refers to the clinical meaningfulness of the 4.5–7.1% overall accuracy increment related to PCA3 inclusion in the current study. De la Taille et al. reported a comparable adjusted overall accuracy increment of 5.5% at an IBX setting [12]; Chun et al. demonstrated accuracy increments related to PCA3 inclusion ranging between 2.3% and 4.6% in a mixed-biopsy cohort [9]. Thus it is important to emphasize that consideration of one novel marker enables us to correctly predict IBX outcome in 9 additional men at risk out of 100 (net reduction) relative to the clinical model without PCA3 [25]. This diagnostic PCA3 impact may appear small, but if once extrapolated to the total number of men undergoing an IBX in Europe and North America, respectively, the diagnostic PCA3 effect may then appear substantial. Of note, applying a nomogram-derived probability threshold of ≥30% advising an IBX, only 2% of men (n=3) with HGPCa would have been missed (Table 4). Exclusion of PCA3 would result in 7% of men (n=19) with HGPCa to be missed. However, when the proportion of 2% of missed HGPCa is considered to be too high, lower probability thresholds should be applied, at the expense of fewer potentially avoided biopsies. Recently, Perdona et al. performed a head-to-head comparison between the mixed-biopsy PCA3-based nomogram by Chun et al. and the updated Prostate Cancer Prevention Trial (PCPT) risk calculator including PCA3 [16]. At a probability threshold of 30%, the nomogram of Chun et al. would have saved 28% of biopsies while missing 6.8% of any PCa and 3.7% of HGPCa. The PCPT risk calculator would have saved 17.9% of biopsies while missing 1.4% of any PCa and no HGPCa [16]. Comparable with our finding, these data demonstrate that the trade-off between avoiding a potentially unnecessary initial biopsy and therefore potentially missing a significant PCa has to be carefully balanced by patients and physicians.

Avoidance of unnecessary biopsies is crucial because prostate biopsies may cause adverse biopsy-related events such as infections, bleeding, or urinary obstruction [3], as well as significant psychological distress [4] and [30]. The novel PCA3-based IBX nomogram has the potential to better identify men with PCa and HGPCa and equally important, those in whom an IBX may be spared (Table 4). Taken together, the urinary PCA3 score adds significant information to combinations with clinical risk factors and enables us to better risk stratify men prior to IBX.

Despite these encouraging findings, limitations certainly apply to our study. First, the lack of a central pathology review may have introduced outcome variability. However, all participating centers represent tertiary referral centers with experienced uropathologists. Second, even though this study relies on a relatively large PCA3 patient cohort, the absolute number of included patients is limited (Table 6). It is noteworthy that predictive models such as nomograms depend strongly on their development data [31]. Consequently, PA discrepancies between different PCA3 codings between univariable and multivariable analyses may be biased due to the number of investigated patients. For example, inclusion of continuously coded PCA3s in the nomogram resulted in the lowest increment of the nomogram's PA, whereas a PCA3 cut-off of 21 showed the highest PA increment. Conversely, continuously coded PCA3s had the highest PA in univariable analyses. The lack of a formal head-to-head comparison between our novel nomogram and other existing predictive tools for initial biopsy may represent a further limitation of the current study. Lastly, due to the lack of external validation, validation studies are clearly needed to confirm our findings prior to recommending a safe use of this novel PCA3-based IBX nomogram.

5. Conclusions

We clearly demonstrate that the construction of a PCA3-based IBX nomogram is more accurate than clinical models or a previously published mixed-biopsy cohort-based PCA3 nomogram. When applying nomogram-derived PCa probability thresholds ≤30%, only a few cases of HGPCa (≤2%) would be missed while avoiding up to 55% of biopsies. External validation of this novel PCA3-based IBX-specific nomogram is recommended prior to its clinical routine use.

Author contributions: Felix K. Chun 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: Hansen, Auprich, Ahyai, Chun.

Acquisition of data: Auprich, de la Taille, van Poppel, Marberger, Stenzl, Mulders, Abbou, Schalken, Fradet, Marks, Ellis, Partin, Pummer, Haese, Walz, Chun.

Analysis and interpretation of data: Hansen, Auprich, Briganti, Chun.

Drafting of the manuscript: Hansen, Auprich, Chun.

Critical revision of the manuscript for important intellectual content: Auprich, Ahyai, de la Taille, van Poppel, Marberger, Stenzl, Mulders, Huland, Fisch, Abbou, Schalken, Fradet, Marks, Ellis, Partin, Pummer, Graefen, Haese, Walz, Briganti, Shariat, Chun.

Statistical analysis: Hansen, Auprich, Chun.

Obtaining funding: None.

Administrative, technical, or material support: None.

Supervision: Graefen, Auprich, Chun.

Other (specify): None.

Financial disclosures: Felix K. Chun certifies 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: Yves Fradet is president, CMO, and shareholder of DiagnoCure, the company that developed and licensed the test to GenProbe. Alan Partin is a study investigator for GenProbe. Hendrik van Poppel and Jack A. Schalken are editorial board members of the pca3.org Web site. Jack A. Schalken is an inventor on PCA3-related PCT applications owned by Radboud University Nijmegen MC. The remaining authors have nothing to disclose.

Funding/Support and role of the sponsor: Gen-Probe, Inc helped design and conduct the study, collect the data, and review and approve the manuscript.

References

  • [1] R.K. Nam, A. Toi, L.H. Klotz, et al. Assessing individual risk for prostate cancer. J Clin Oncol. 2007;25:3582-3588 Crossref.
  • [2] O.M. Zaytoun, M.W. Kattan, A.S. Moussa, J. Li, C. Yu, J.S. Jones. Development of improved nomogram for prediction of outcome of initial prostate biopsy using readily available clinical information. Urology. 2011;78:392-398 Crossref.
  • [3] R.K. Nam, R. Saskin, Y. Lee, et al. Increasing hospital admission rates for urological complications after transrectal ultrasound guided prostate biopsy. J Urol. 2010;183:963-968
  • [4] D.A. Katz, D.F. Jarrard, C.A. McHorney, S.L. Hillis, D.A. Wiebe, D.G. Fryback. Health perceptions in patients who undergo screening and workup for prostate cancer. Urology. 2007;69:215-220 Crossref.
  • [5] F.K.-H. Chun, A. Briganti, M. Graefen, et al. Development and external validation of an extended 10-core biopsy nomogram. Eur Urol. 2007;52:436-445 Abstract, Full-text, PDF, Crossref.
  • [6] P.I. Karakiewicz, S. Benayoun, M.W. Kattan, et al. Development and validation of a nomogram predicting the outcome of prostate biopsy based on patient age, digital rectal examination and serum prostate specific antigen. J Urol. 2005;173:1930-1934 Crossref.
  • [7] K. Kawamura, H. Suzuki, N. Kamiya, et al. Development of a new nomogram for predicting the probability of a positive initial prostate biopsy in Japanese patients with serum PSA levels less than 10 ng/ml. Int J Urol. 2008;15:598-603 Crossref.
  • [8] P. Sooriakumaran, M. John, P. Christos, et al. Models to predict positive prostate biopsies using the Tyrol screening study. Urology. 2011;78:924-929 Crossref.
  • [9] F.K. Chun, A. de la Taille, H. van Poppel, et al. Prostate cancer gene 3 (PCA3): development and internal validation of a novel biopsy nomogram. Eur Urol. 2009;56:659-668 Abstract, Full-text, PDF, Crossref.
  • [10] D.P. Ankerst, J. Groskopf, J.R. Day, et al. Predicting prostate cancer risk through incorporation of prostate cancer gene 3. J Urol. 2008;180:1303-1308 discussion 1308 Crossref.
  • [11] M. Auprich, A. Haese, J. Walz, et al. External validation of urinary PCA3-based nomograms to individually predict prostate biopsy outcome. Eur Urol. 2010;58:727-732 Abstract, Full-text, PDF, Crossref.
  • [12] A. de la Taille, J. Irani, M. Graefen, et al. Clinical evaluation of the PCA3 assay in guiding initial biopsy decisions. J Urol. 2011;185:2119-2125 Crossref.
  • [13] I.L. Deras, S.M. Aubin, A. Blase, et al. PCA3: a molecular urine assay for predicting prostate biopsy outcome. J Urol. 2008;179:1587-1592 Crossref.
  • [14] A. Haese, A. de la Taille, H. van Poppel, et al. Clinical utility of the PCA3 urine assay in European men scheduled for repeat biopsy. Eur Urol. 2008;54:1081-1088 Abstract, Full-text, PDF, Crossref.
  • [15] L.S. Marks, Y. Fradet, I.L. Deras, et al. PCA3 molecular urine assay for prostate cancer in men undergoing repeat biopsy. Urology. 2007;69:532-535 Crossref.
  • [16] S. Perdonà, V. Cavadas, G. Di Lorenzo, et al. Prostate cancer detection in the “grey area” of prostate-specific antigen below 10 ng/ml: head-to-head comparison of the updated PCPT calculator and Chun's nomogram, two risk estimators incorporating prostate cancer antigen 3. Eur Urol. 2011;59:81-87
  • [17] M. Auprich, A. Bjartell, F.K.-H. Chun, et al. Contemporary role of prostate cancer antigen 3 in the management of prostate cancer. Eur Urol. 2011;60:1045-1054 Abstract, Full-text, PDF, Crossref.
  • [18] D. Hessels, J.M.T. Klein Gunnewiek, I. van Oort, et al. DD3PCA3-based molecular urine analysis for the diagnosis of prostate cancer. Eur Urol. 2003;44:8-16 discussion 16 Crossref.
  • [19] F.K.-H. Chun, P.I. Karakiewicz, A. Briganti, et al. Prostate cancer nomograms: an update. Eur Urol. 2006;50:914-926 discussion 926 Abstract, Full-text, PDF, Crossref.
  • [20] L.A. Eskew, R.L. Bare, D.L. McCullough. Systematic 5 region prostate biopsy is superior to sextant method for diagnosing carcinoma of the prostate. J Urol. 1997;157:199-202 discussion 203 Crossref.
  • [21] J. Groskopf, S.M. Aubin, I.L. Deras, et al. APTIMA PCA3 molecular urine test: development of a method to aid in the diagnosis of prostate cancer. Clin Chem. 2006;52:1089-1095 Crossref.
  • [22] M. Mazumdar, J.R. Glassman. Categorizing a prognostic variable: review of methods, code for easy implementation and applications to decision-making about cancer treatments. Stat Med. 2000;19:113-132 Crossref.
  • [23] A. Vickers, A. Cronin, M. Roobol, et al. Reducing unnecessary biopsy during prostate cancer screening using a four-kallikrein panel: an independent replication. J Clin Oncol. 2010;28:2493-2498 Crossref.
  • [24] E.W. Steyerberg, F.E. Harrell Jr., G.J. Borsboom, M.J. Eijkemans, Y. Vergouwe, J.D. Habbema. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol. 2001;54:774-781 Crossref.
  • [25] A.J. Vickers, E.B. Elkin. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making. 2006;26:565-574 Crossref.
  • [26] M.J. Roobol, H.A. van Vugt, S. Loeb, et al. Prediction of prostate cancer risk: the role of prostate volume and digital rectal examination in the ERSPC risk calculators. Eur Urol. 2012;61:577-583 Abstract, Full-text, PDF, Crossref.
  • [27] F.H. Schroder, J. Hugosson, M.J. Roobol, et al. Screening and prostate-cancer mortality in a randomized European study. N Engl J Med. 2009;360:1320-1328 Crossref.
  • [28] R. Al-Azab, A. Toi, G. Lockwood, G.S. Kulkarni, N. Fleshner. Prostate volume is strongest predictor of cancer diagnosis at transrectal ultrasound-guided prostate biopsy with prostate-specific antigen values between 2.0 and 9.0 ng/ml. Urology. 2007;69:103-107 Crossref.
  • [29] A. Briganti, F.K. Chun, N. Suardi, et al. Prostate volume and adverse prostate cancer features: fact not artifact. Eur J Cancer. 2007;43:2669-2677 Crossref.
  • [30] N.S. Awsare, J.S. Green, B. Aldwinckle, D.C. Hanbury, G.B. Boustead, T.A. McNicholas. The measurement of psychological distress in men being investigated for the presence of prostate cancer. Prostate Cancer Prostatic Dis. 2008;11:384-389 Crossref.
  • [31] F.K. Chun, P.I. Karakiewicz, H. Huland, M. Graefen. Role of nomograms for prostate cancer in 2007. World J Urol. 2007;25:131-142 Crossref.

Footnotes

a Department of Urology, University Hospital Hamburg-Eppendorf, Hamburg, Germany

b Martini Clinic Prostate Cancer Center, University Hospital Eppendorf, Hamburg, Germany

c Hôpital Henri Mondor, Créteil, France

d Universitair Ziekenhuis Gasthuisberg, Leuven, Belgium

e University of Vienna, Vienna, Austria

f Uniklinikum Tübingen, Tübingen, Germany

g Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands

h Department of Urology, Université Laval, Quebec City, Quebec, Canada

i Urological Sciences Research Foundation, Culver City, CA, USA

j University of Washington Medical Center, Seattle, WA, USA

k Department of Urology, Johns Hopkins University Medical Institution, Baltimore, MD, USA

l Department of Urology, Medical University of Graz, Graz, Austria

m Department of Urology, Institut Paoli-Calmettes Cancer Centre, Marseille, France

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

o Department of Urology, Weill Medical College of Cornell University, New York, NY, USA

lowast Corresponding author. University Hospital Hamburg-Eppendorf, Department of Urology, Martinistr. 52, 20246 Hamburg, Germany. Tel. +49 40 428 03 3443; Fax: +49 40 428 03 6837.

Both authors contributed equally.

Place a comment

Your comment *

max length: 5000