BJU International

Multiparametric magnetic resonance imaging of the prostate can improve the predictive value of the urinary prostate cancer antigen 3 test in patients with elevated prostate-specific antigen levels and a previous negative biopsy

Authors


Alessandro Sciarra, Prostate Unit – Department Urology, University Sapienza, Viale Policlinico 155, 00161 Rome, Italy. e-mail: sciarra.md@libero.it

Abstract

Study Type – Clinical (prospective trial)

Level of Evidence 2b

What's known on the subject? and What does the study add?

In clinical practice, we know that it is necessary to identify new biomarkers that can better detect prostate cancer (PC), at the same time as reducing the number of unnecessary biopsies. Recently, studies have suggested that the most relevant clinical scenario in which the prostate cancer antigen 3 (PCA3) score could be used comprises patients with a previous negative prostate biopsy and persistently elevated PSA levels. At the same time, although multiparametric MRI is not currently used as a first approach for diagnosing PC, it can be useful for directing targeted biopsies, especially in those patients with elevated PSA levels and a previous negative TRUS-guided biopsy. Considering all of these aspects, the present study aimed to evaluate the role of multiparametric MRI as an additional diagnostic tool for improving the accuracy of the urinary PCA3 test in patients with increased PSA levels and a previous negative prostate biopsy.

Our hypothesis is that the potential value of the PCA3 test as a biomarker for PC diagnosis could be improved by the use of multiparametric MRI in directing prostate biopsy. In the present study, we show that, in cases with a previous negative biopsy and persistently elevated PSA levels submitted to multiparametric MRI to direct biopsies, the sensitivity of the PCA3 test significantly improved (79% vs 68%). However, further larger randomized studies on this combination using a new biomarker and a new imaging modality for PC diagnosis are expected.

OBJECTIVE

  • • To evaluate the role of multiparametric magnetic resonance imaging (MRI) as an additional diagnostic tool for improving the accuracy of the urinary prostate cancer antigen 3 (PCA3) test in patients with an increase in prostate-specific antigen (PSA) levels and a previous negative prostate biopsy.

PATIENTS AND METHODS

  • • The present study comprised a prospective randomized study on patients with a previous negative transrectal ultrasonography (TRUS)-guided prostate biopsy and elevated PSA levels.
  • • In total, 180 cases were analyzed, and all were submitted to PCA3 assay.
  • • Patients in group A were submitted to a second random TRUS-guided prostate biopsy, whereas patients in group B were submitted to a multiparametric MRI examination and then to a second TRUS-guided prostate biopsy.

RESULTS

  • • At the second biopsy, a histological diagnosis of prostate cancer was found in 26 of 84 cases (30.9%) in group A and in 29 of 84 cases (34.5%) in group B.
  • • In group A, the sensitivity and specificity of the PCA3 score were 68.0% and 74.5% respectively (positive predictive value of 53.1%, negative predictive value of 84.6% and accuracy of 72.6%).
  • • In group B, the sensitivity and specificity of the PCA3 score were 79.3% and 72.7%, respectively (positive predictive value of 60.5%, negative predictive value of 86.9% and accuracy of 75.0%).
  • • For the PCA3 score, the area under the receiver-operator characteristic curve was 0.825 (95% confidence interval, 0.726–0.899) in group A and 0.857 (95% confidence interval, 0.763–0.924) in group B (P < 0.001).

CONCLUSION

  • • In patients with a previous negative biopsy and persistently elevated PSA levels, the use of multiparametric MRI for indicating sites suitable for rebiopsy can significantly improve the sensitivity of the PCA3 test in the diagnosis of prostate cancer.
Abbreviations
DCEI

dynamic contrast-enhanced imaging

DWI

diffusion-weighted imaging

MRSI

magnetic resonance spectroscopic imaging

NPV

negative predictive value

PCA3

prostate cancer antigen 3

PPV

positive predictive value.

INTRODUCTION

The diagnosis of prostate cancer (PC) is mainly based on three tests: DRE, PSA levels and TRUS-guided biopsy [1]. In particular, a diagnosis based on PSA levels shows a significant trade-off between sensitivity and specificity, and all attempts to increase the sensitivity have been balanced by a reduction of specificity [2–4]. Therefore, in clinical practice, it is necessary to identify new biomarkers that can better detect PC, at the same time as reducing the number of unnecessary biopsies. Recently, the PC antigen 3 (PCA3) gene has been considered as the most promising new biomarker for the diagnosis of PC [5–8]. Data from a recent systematic review and meta-analysis showed that the sensitivity of the PCA3 test is in the range 46.9–82.3%, specificity is in the range 56.3–89%, positive predictive value (PPV) is in the range 59.4–97.4% and negative predictive value (NPV) is in the range 87.7–98% [9]. Moreover, some studies [10–13] suggest that the most relevant clinical scenario in which the PCA3 score could be used is in patients with a previous negative prostate biopsy and persistently elevated PSA levels. In 233 men with negative biopsy findings but persistently elevated serum PSA levels (>2.5 ng/mL), Marks et al. [10] reported a PCA3 (theshold >35) score sensitivity of 58% and a specificity of 72%. The European Urological Association guidelines [1] confirm that, although PCA3 may have potential value for identifying PC in men with initially negative biopsies despite elevated PSA levels, such a determination remains experimental.

After the correct use of biomarkers, the second step in the early diagnosis of PC is histological confirmation at prostate biopsy. Although random TRUS-guided biopsy remains the gold standard for detecting PC, it have been reported to miss up to 30% of cancers [1,14]. The European Urological Association guidelines [1] also suggested that biopsy strategies with an increased number of cores (‘saturation biopsy’) reduced false-negative rates, although the incidence of PC detected with this technique is between 30% and 43% (dependent on the number of cores sampled) and may be associated with increased patient morbidity. On the basis of these outcomes, several studies suggested the need for a sensitive and accurate imaging modality to detect PC and to direct biopsy. Recently, there has been great interest in multiparametric MRI, which combines anatomic imaging with magnetic resonance spectroscopic imaging (MRSI), diffusion-weighted imaging (DWI) and dynamic contrast-enhanced (DCEI). Multiparametric MRI aims to be a successful tool for improving many aspects of PC management [15]. Although multiparametric MRI is not currently used as a first approach in the diagnosis of PC, it can be useful for directing targeted biopsies, especially for those patients with elevated PSA levels and a previous negative TRUS-guided biopsy [15–18]. In particular, Cirillo et al. [16] found that the use of MRSI combined with MRI had a sensitivity of 100%, a specificity of 51.4%, a PPV of 48.6%, a NPV of 100% and an accuracy of 66.3% with respect to indicating PC sites suitable for rebiopsy. Using similar patients, Sciarra et al. [17] found that the use of combined MRSI and DCEI had a sensitivity of 93%, a specificity of 89%, a PPV of 89%, a NPV of 93% and an accuracy of 91% with respect to predicting the detection of PC at rebiopsy. Considering all of these points, our hypothesis is that the potential value of PCA3 as a biomarker for PC diagnosis could be improved by the use of multiparametric MRI for directing prostate biopsy. In particular, it is possible that some of the cases detected as being PCA3 false positive at random biopsy could become true positive cases when using multiparametric MRI to better direct prostate biopsy. Therefore, the present study aimed to evaluate the role of multiparametric MRI as an additional diagnostic tool for improving the accuracy of the urinary PCA3 test in patients with increased PSA levels and a previous negative prostate biopsy.

MATERIALS AND METHODS

STUDY DESIGN AND POPULATION

The present study comprised a prospective randomized single-centre study on patients with a previous negative TRUS-guided prostate biopsy and persistently elevated PSA levels. The study was conducted after receiving approval of the protocol from our institutional board committee and informed consent was obtained from all patients. In total, 180 consecutive patients who were referred to our Department of Urology from March 2008 to March 2011 were recruited into the present study. All cases had a first random TRUS-guided prostate biopsy negative for PC or high-grade prostate intraepithelial neoplasia, persistently elevated PSA levels (total PSA >4 ng/mL) and negative DRE. Exclusion criteria for the present study were: previous hormonal, surgical or radiation therapies for prostate diseases; inadequate sample for PCA3 analysis; and cases in which anatomic imaging with complete MRSI and DCEI was not possible. Inclusion criteria for the present study were: first negative prostate biopsy, persistent total PSA >4 ng/mL and negative DRE. All first biopsies were conducted homogeneously as part of the patient's urological evaluation (10 cores laterally directed, random TRUS-guided prostate biopsy). At baseline, all patients were submitted to a PCA3 assay; 168/180 (93.3%) cases yielded sufficient RNA for analysis and were included in the present study. Patient characteristics at inclusion are provided in Table 1. Patients included in the present study were randomly (1 : 1) assigned to two groups (Fig. 1). In group A, patients (n= 84) were submitted to a second random TRUS-guided prostate biopsy no later than 90 days after the first biopsy. In group B, patients (n= 84) were submitted to a multiparametric MRI examination with MRSI, DWI and DCEI and then to a second TRUS-guided prostate biopsy no later than 2 weeks after MRI and no later than 90 days after the first biopsy. In cases of prostate areas indicated by MRI as being suspicious for cancer, samples targeted on these areas were associated with the random biopsy.

Table 1.  Characteristics of the population randomized into groups A and B
CharacteristicGroup AGroup B P
  1. PCA3, prostate cancer antigen 3.

Cases (n)8484
Age (years), mean (sd); median, range63.22 (7.14); 64, 46–7564.09 (7.36); 64, 46–760.820
Total PSA level (ng/mL), mean (sd); median, range6.89 (2.09); 6.4, 4.1–13.17.07 (3.48); 6.2, 4.2–15.50.448
PCA3 score, mean (sd); median, range39.25 (35.02); 26.5, 3.0–183.048.21 (47.73); 31.0, 3.0–225.00.544
Suspicious at multiparametric MRI, n (%)39 (46.4)
Prostate cancer at second biopsy, n (%)26 (30.9)29 (34.5) 
Prostate cancer Gleason score ≤7 (3+4), n (%)19 (22.6)24 (28.5) 
Prostate cancer Gleason score ≥7 (4+3), n (%)7 (8.3)5 (5.9) 
Figure 1.

Study design. PCA3, prostate cancer antigen 3.

PCA3 URINE ASSAY

The PCA3 clinical test requires urine to be collected after an attentive DRE to increase the number of prostate cells shed into the urine [7]. An attentive DRE included firm pressure on the prostate from the base to the apex and from the lateral to median lobe, with three strokes per lobe and sufficient pressure to slightly depress the prostate surface [10,11]. After the DRE, patients collected their initial void of 20–30 mL of urine and then PCA3 and PSA mRNAs were isolated from a total of 2.5 mL of urine for transcription-mediated amplification (ProgensaTM PCA3 assay; Gen-Probe Inc., San Diego, CA, USA). PCA3 scores were reported as the quantitative PCA3/PSA mRNA ratio × 1000 to normalize PCA3 to the amount of prostate RNA present in the urine sample. Cases with insufficient PSA mRNA were considered inconclusive and excluded (12 of 180; 6.7%). Based on previous experience as reported in the literature [4,10,19], a threshold value of 35 was considered for the prediction of PC.

MULTIPARAMETRIC MRI

All examinations were performed using a 3T scanner (Magnetom Vario, Siemens Medical Solutions, Erlangen, Germany; gradient strength, 45 mT/m; slew rate, 346 T/m/s; rise time, 400 µ/s), equipped with surface phased array (Body Matrix, Siemens Medical Solutions) and an endorectal coil (e-Coil, Medrad, combined with an Endoan-Interface; Siemens Medical Solutions). Morphologic imaging of the prostatic gland was carried out by acquiring turbo spin echo T2-weighted sequences in the axial, sagittal and coronal planes. The technique used for imaging with MRSI, DCEI and DWI of the prostate has been described previously [17,20]. Images were analyzed for consensus by two radiologists (V.P. and R.P.), with both having long experience in urogenital MRI. As described previously [17,21–23], for a comparison of the imaging findings with the pathological data, the peripheral zone of the prostate was divided into sextant according to fixed criteria. The location of the MRSI voxels and the DCEI and DWI areas used for the analysis was correlated with the sextants defined by MRI. Assessment of the spatial correspondence between the MRI findings and the pathological evaluation was performed independently in all cases by two investigators (V.P. and R.P.) and then revised by a third investigator (L.D.) who was not a reader of the images. This correspondence was achieved on the basis of the x- and z-coordinates derived from the T2-weighted MRI and on the basis of the sextant division of the peripheral zone of the prostate, as described previously [17].

PROSTATIC BIOPSY

All biopsies were perfomed in our department by a single physician (M.C.) with a long experience of this procedure. All TRUS and biopsies were peformed using an end-fire ultrasonographic transducer and biopsy gun with an 18-gauge needle (Esaote Technos MP with a C10-5 transducer). As is common practice in our institution, standard random, laterally directed 10-core biopsies were performed for each patient. In cases with areas described by MRI as being suspicious for cancer, two additional cores were taken from each area that was labelled abnormal. Biopsy targeting was carried out in zones corresponding to those analyzed with multiparametric MRI, on the basis of the x- and z-coordinates derived from the T2-weighted MRI, as described previously [17]. Histological assessments were carried out blind to the results of the MRI.

CORRELATION OF PCA3 WITH PATHOLOGICAL FINDINGS AND STATISTICAL ANALYSIS

Correlation analysis was performed aiming to identify a relationship between the PCA3 score and pathological results at random prostate biopsy in group A and between the PCA3 score and pathological results at random + targeted biopsy after MRI examination in group B. Statistical data analysis was carried out using the statistical software MedCalc Software for Windows, version 9.3 (). P < 0.05 was considered statistically significant. Differences between group means were analyzed by the Mann–Whitney U-test. Sperman coefficients and logistic univariate analysis were used to determine the association between the PCA3 scores and pathological results in the two groups. All variables were also included in logistic multivariate models. Receiver-operator characteristic (ROC) curve comparison for each analysis phase was also estimated.

RESULTS

At baseline, no statistically significant difference in mean age, PSA levels and PCA3 scores was found between the two randomized groups (Table 1). At the second biopsy, a histological diagnosis of prostate adenocarcinoma was found in 26 of 84 cases (30.9%) in group A and in 29 of 84 cases (34.5%) in group B (Table 1). In group B, the multiparametric MRI resulted in a suspicion of PC in 39 of 84 cases (46.4%) (Table 1) and a mean (sd, range) of 12.17 (1.29, 10–14) of biopsy samples was performed in this group. In particular, in group B, a prostate adenocarcinoma was found in four cases showing a normal MRI and in 25 cases with a suspicious MRI. In all suspicious MRI cases found to be positive for PC at the second biopsy, the localization of cancer on histological examination also corresponded to the site indicated by the multiparametric MRI.

The performance of the PCA3 test for predicting biopsy outcomes was evaluated in terms of ROC curve analysis, sensitivity, specificity, PPV, NPV and accuracy by comparing the PCA3 score with the histological results of the biopsy used as reference method in the two groups (group A = second random biopsy; group B = second random + MRI-targeted biopsy). For the PCA3 score, the area under the ROC curve was 0.825 (95% CI, 0.726–0.899) in group A and 0.857 (95% CI, 0.763–0.924) in group B (P < 0.001) (Fig. 2). In group A, the sensitivity and specificity of the PCA3 score (theshold >35) for predicting a positive biopsy were 68.0% and 74.5%, respectively, with a PPV of 53.1%, a NPV of 84.6% and an accuracy of 72.6%. In group B, the sensitivity and specificity of the PCA3 score were 79.3% and 72.7%, respectively, with a PPV of 60.5%, a NPV of 86.9% and an accuracy of 75.0% (Table 2). All of these data show that, in group B, the PCA3 score could represent a better predictor of prostate biopsy outcomes, in particular in terms of sensitivity, compared to group A (P < 0.001). No significant difference (P= 0.089) in the predictive value of the PCA3 score was found with respect to distinguishing PC cases on the basis of the Gleason score: ≤7 (3+4) vs ≥7 (4+3).

Figure 2.

Receiver-operator characteristic (ROC) curve analysis in group A vs group B by comparing the prostate cancer antigen 3 score with the histological results of the biopsy as a reference method (group A, area under the ROC curve = 0.825 vs group B, area under the ROC curve = 0.857; P < 0.001).

Table 2.  Prostate cancer antigen 3 (PCA3) score (>35) performance as a predictor of prostate biopsy outcomes in groups A and B
PCA3 scoreGroup AGroup B
  1. AUC, area under the ROC curve.

Sensitivity0.680.793
Specificity0.7450.727
PPV0.5310.605
NPV0.8460.869
Accuracy0.7260.750
AUC0.8250.857

DISCUSSION

The value of urinary PCA3 has been evaluated primarily in patients with a previous negative biopsy and persistently elevated PSA levels. This category of patients represents a dilemma to physicians because of the costs, risk of morbidity and emotional disadvantages related to repeated prostate biopsies. In this setting, the PCA3 test showed a sensitivity of 47–58%, a specificity of 71–72%, a PPV of 39–43% and a NPV of 78–83% [9–11]. All of the studies used pathological results from random TRUS-guided biopsies as a reference to define the predictive value of the PCA3 test for PC. Although random TRUS-guided biopsy is the reference method for the histological diagnosis of PC, several studies highlight the high (30%) false negative rates associated with this method [1,14]. The hypothesis in the present study is that the potential value of the PCA3 test as a biomarker for PC diagnosis could be improved by the use of multiparametric MRI for directing prostate biopsy. In particular, it is possible that some cases detected as being PCA3 false positive at a random biopsy could become true positive cases using multiparametric MRI to better direct prostate biopsy. Particularly in cases with a previous negative random biopsy and persistently elevated PSA levels, the combination of anatomic, metabolic and functional dynamic informations offered by multiparametric MRI promises to be useful for directing targeted biopsies [16,17]. In particular, the use of strict criteria [17], as obtained combining MRSI and DCEI analysis on the entire prostate gland, can be used to direct targeted samples in association with a random rebiopsy.

In the present study, we show that, in cases with a previous negative biopsy and persistently elevated PSA levels, the predictive value of the PCA3 test could be improved using multiparametric MRI to direct prostate biopsy. In particular, in the group in which multiparametric MRI was used to associate targeted samples to the classic random biopsy (group B), the sensitivity of the PCA3 test significantly improved (79% vs 68%). In addition, the area under the ROC curve value was significantly higher (0.85 vs 0.82) (P < 0.001) in group B, although the difference was limited. However, the specificity of PCA3 did not increase when adding MRI (74.5% vs 72.7%), showing that specificity remains the key problem in prostate cancer testing.

The strength of the present study is mainly a result of the randomized prospective design, the blinded assessment of the spatial correspondence between multiparametric MRI findings and the pathological evaluation performed by three investigators with long experience in MRI studies, and the homogeneous criteria used to select and evaluate the population included in the study. The limitations of the present study mainly relate to difficulties in ensuring the correspondence of TRUS-guided biopsy spatial accuracies with suspicious areas on MRI, as reported previously [17]. However, the correspondence between the localization of cancer on histological examination and the site indicated by multiparametric MRI in those cases with a suspicious MRI who were positive for PC at biopsy supports our methodology. It is true that group B patients (on the basis of results of the MRI) had more biopsy samples (mean 12.17) than patients in group A; otherwise, we consider that the better results obtained in group B are related more to the MRI methodology than to the increased number of biopsy samples. Relative to the PCA3 test, and considering the primary aim of the present study, we did not analyse different PCA3 score theshold values but, instead, used the most recommended value of 35 [4,10] to define a test as positive.

In conclusion, in patients with a previous negative biopsy and persistently elevated PSA levels, the use of multiparametric MRI to guide prostate biopsy can increase the accuracy of this procedure with respect to confirming the diagnosis of PC. Regarding increasing the accuracy of prostate biopsy, multiparametric MRI can also increase the sensitivity (but not the specificity) of a marker such as PCA3, so that cases detected as being PCA3 false positive can become PCA3 true positive. Further larger randomized studies on this combination using a new biomarker and a new imaging modality for PC diagnosis are expected.

CONFLICT OF INTEREST

None declared.

Ancillary