European Radiology

, Volume 22, Issue 2, pp 468–475

Prostate MRI: diffusion-weighted imaging at 1.5T correlates better with prostatectomy Gleason grades than TRUS-guided biopsies in peripheral zone tumours

Authors

    • CDPI Clinics—Abdominal and Pelvic ImagingRio de Janeiro Federal University
  • Jelle O. Barentsz
    • Department of RadiologyRadboud University Nijmegen Medical Centre
  • Luiz Carlos Duarte de Miranda
    • Department of UrologyRio de Janeiro Federal University
  • Emerson Leandro Gasparetto
    • Department of Radiology, CDPI and Multi-Imagem ClinicsRio de Janeiro Federal University
Oncology

DOI: 10.1007/s00330-011-2269-1

Cite this article as:
Bittencourt, L.K., Barentsz, J.O., de Miranda, L.C.D. et al. Eur Radiol (2012) 22: 468. doi:10.1007/s00330-011-2269-1
  • 534 Views

Abstract

Objectives

To investigate the usefulness of Apparent Diffusion Coefficients (ADC) in predicting prostatectomy Gleason Grades (pGG) and Scores (GS), compared with ultrasound-guided biopsy Gleason Grades (bGG).

Methods

Twenty-four patients with biopsy-proven prostate cancer were included in the study. Diffusion-weighted images were obtained using 1.5-T MR with a pelvic phased-array coil. Median ADC values (b0,500,1000 s/mm²) were measured at the most suspicious areas in the peripheral zone. The relationship between ADC values and pGG or GS was assessed using Pearson’s coefficient. The relationship between bGG and pGG or GS was also evaluated. Receiver operating characteristic (ROC) curve analysis was performed to assess the performance of each method on a qualitative level.

Results

A significant negative correlation was found between mean ADCs of suspicious lesions and their pGG (r = −0.55; p < 0.01) and GS (r = −0.63; p < 0.01). No significant correlation was found between bGG and pGG (r = 0.042; p > 0.05) or GS (r = 0.048; p > 0.05). ROC analysis revealed a discriminatory performance of AUC = 0.82 for ADC and AUC = 0.46 for bGG in discerning low-grade from intermediate/high-grade lesions.

Conclusions

The ADC values of suspicious areas in the peripheral zone perform better than bGG in the correlation with prostate cancer aggressiveness, although with considerable intra-subject heterogeneity.

Key Points

Prostate cancer aggressiveness is probably underestimated and undersampled by routine ultrasound-guided biopsies.

Diffusion-weighted MR images show good linear correlation with prostate cancer aggressiveness.

DWI information may be used to improve risk-assessment in prostate cancer.

Keywords

Magnetic resonance imagingProstate cancerDiffusion magnetic resonance imagingBiopsybiological markers

Introduction

Prostate Cancer (PCa) is the most common solid organ malignancy among adult men worldwide and the second leading cause of cancer deaths in American men [1]. The disease varies widely in clinical behaviour among individuals, ranging from aggressive and rapidly progressive tumours that are commonly treated with radical prostatectomy and/or radiotherapy, to non-aggressive indolent tumours that may be followed through active surveillance protocols [2].

In order to achieve the ideal treatment choice and risk stratification in these patients, a series of parameters and methods have been devised, including demographics, histopathological data, prostate-specific antigen (PSA) levels, and Digital Rectal Examination (DRE), which can be combined in different kinds of nomograms [36]. Among the histopathological information, the Gleason Grading System [7, 8] is the most commonly accepted and widely used parameter for evaluating the aggressiveness of PCa. It not only reflects the neoplastic tissue characteristics, but also provides information on long-term prognosis and outcomes [9, 10]. However, it is well-known that the accuracy of routine transrectal ultrasound (TRUS)-guided biopsy Gleason Score (GS) for the prediction of the entire prostatectomy specimen GS is suboptimal, ranging between 28% and 74% in most of the series [1114]. In addition, the result is most likely even worse when considering the subdivision into primary and secondary Gleason grades. Therefore, other parameters may be needed to further refine this information in order to achieve better correlation with the prostatectomy data.

Diffusion-Weighted Imaging (DWI) is a functional technique that assesses the random movement of water molecules in different physical media through the use of Magnetic Resonance (MR) imaging. In biological tissues, the impedance of water molecule diffusion partially reflects tissue cellularity [1517] and the presence of intact cellular membranes [17]. Tissues and structures associated with restricted diffusion, which are quantitatively represented by a low Apparent Diffusion Coefficient (ADC), include abscesses, fibrosis, haematomas, cytotoxic oedema, and particularly highly cellular tumours. Moreover, ADC measurements have already been correlated with cancer aggressiveness in other body regions, such as the brain or the breast [1820].

Several preliminary studies have demonstrated that neoplastic lesions of the peripheral zone display significantly lower ADC values than normal prostate tissue [2127], suggesting that DWI may play an important role in the determination of the “cellularity component” in functional prostate imaging. In support of this hypothesis, it has been recognised that DWI can be used as a potential imaging biomarker [28] for a wide variety of solid tumours.

Unsurprisingly, the attempt to correlate cellularity indexes, such as the Gleason score, with the quantification of water diffusion by means of ADC values seems to be a natural progression of knowledge. Consequently, a few retrospective studies that used different methodologies have consistently shown that there is a significant negative correlation between the ADC values of PCa and Gleason scores [2934] in either biopsy or prostatectomy specimens. However, to our knowledge, no such study has ever compared the performance of ADC measurements with that of routine TRUS-guided biopsies in the correlation with prostatectomy Gleason scores on a case-by-case basis.

Therefore, this study aimed to prospectively investigate the usefulness of the ADC values of suspicious areas in the peripheral zone on the prediction of prostatectomy specimen Gleason scores compared with the performance of routine TRUS-guided biopsy samples. For the aforementioned reasons, we have hypothesised that DWI should provide a better correlation with the actual tumour aggressiveness than routine TRUS-guided biopsy samples.

Materials and methods

This prospective study was approved by the local institutional review board, and all patients were required to sign informed consent.

Between March 2009 and September 2010, 35 consecutive patients (mean age, 63 years; range, 52–78 years) with biopsy-proven PCa were referred for a clinically routine MR imaging examination of the prostate, with the plan for radical prostatectomy. The PSA levels ranged from 3.45 to 21.1 ng/mL (median 7.6 ng/mL).

All patients were submitted to routine TRUS-guided sextant prostate biopsies between 1 to 6 months before the MR examination. Each examination had at least 6 samples and all of them including the Gleason scores for each positive sample. For statistical analysis, only the sample with the highest Gleason score per patient was considered.

Magnetic resonance images were acquired at 1.5-T (Magnetom Avanto, Siemens Medical, Erlangen, Germany) using a 6-channel pelvic phased-array surface. Before MRI, the patients were given 10 mg of n-methyl-scopolamine bromide (Buscopan®, Boehringer Ingelheim, Brazil) intravenously in order to suppress peristalsis. The study protocol consisted of high-resolution T2-weighted (T2w) sequences in the axial (TR 4750 ms, TE 101 ms, FOV 160 × 160 mm2, matrix 256 × 230, slice thickness 3 mm, no gap, 3 averages), coronal (TR 3000 ms, TE 101 ms, FOV 160 × 160 mm2, matrix 256 × 230, slice thickness 3.5 mm, 20% gap, 2 averages) and sagittal (TR 3800 ms, TE 100 ms, FOV 170 × 170 mm2, matrix 320 × 240, slice thickness 3 mm, 10% gap, 2 averages) planes, diffusion-weighted images in the axial plane (TR 3000 ms, TE 88 ms, b-values 0, 500, 1000 mm/s2, 3-image trace, FOV 200 × 200 mm2, matrix 150 × 150, slice thickness 3.5 mm, no gap, 8 averages), an axial T1-weighted sequence (TIRM; TR 2100 ms, TE 20 ms, TI 829.7 ms, FOV 200 × 180 mm2, matrix 256 × 200, slice thickness 3 mm, 10% gap, 2 averages), and dynamic contrast enhancement (DCE) evaluation (T1 3D GRE; TR 4.08 ms, TE 1.43 ms, no fat saturation, FOV 280 × 280 mm2, matrix 192 × 192, slice thickness 3 mm, 1 average, 40 measurements, 6.8 s per measurement) after intravenous infusion of gadolinium-chelate (DOTAREM, Guerbet, Aulnay-sous-Bois, France). The MRI system software (Siemens Medical, Erlangen, Germany), using all three b-values in a monoexponential model, was used to calculate ADC maps. DCE post-processing was also done on the MRI system software, generating the following semi-quantitative colour maps: washin (phases 3–10), washout (phases 10–20), Maximum Intensity Projection over time (MIPt) and Positive Enhancement Integral (PEI). Quantitative K-trans maps were as well obtained, using commercially available software (syngo Tissue 4D, Siemens Medical, Erlangen, Germany).

The MR images were routinely read by one radiologist who had 5 years of experience in prostate imaging (L.K.B), and in every case the most suspicious lesion in the peripheral zone was delineated based on at least 3 of the following criteria: low signal intensity on T2w images, nodular lesion on T2w images, focal/sextant restricted diffusion on the ADC maps (ADC < 1000 × 10-6 mm/s2), or qualitatively high wash-in and/or wash-out rate on DCE colour maps. A finding was only considered suspicious if the area exceeded 0.25 cm2 (i.e. 0.5 × 0.5 cm) on the axial plane (Fig. 1). The radiologist then drew a region of interest (ROI) to match the shape and extent of each suspicious lesion on the ADC maps in the axial plane at the slice with its best depiction. The ROI surface areas ranged between 0.51 (21 pixels) to 3.44 cm2 (160 pixels), with an average of 1.18 cm2 (55.2 pixels). Only the median ADC values in each lesion were considered for the statistical analysis.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-011-2269-1/MediaObjects/330_2011_2269_Fig1_HTML.gif
Fig. 1

A 72-year-old man with prostate cancer, presurgical PSA level of 6.5 ng/mL, biopsy Gleason scores (grades listed) of 7 (4 + 3) and prostatectomy Gleason scores (grades listed) of 9 (5 + 4). a T2-weighted image on the axial plane (TR/TE 4750/101), showing a focal low-signal intensity nodular lesion in the posterolateral midgland (arrowheads). b k-Trans map derived from DCE evaluation post-gadolinium, depicting mild focal hyperpermeability on the suspicious lesion (arrowheads) compared with the contralateral peripheral zone. c ADC map on the same level, showing focal restricted diffusion at the suspicious area, with the yellow outline representing the ROI (ADC = 690 × 10-6 mm/s2). d Macroscopic histopathology slice at the midgland level, revealing the index lesion outlined in red at the left lobe, and showing good topographic correlation with the MR findings

The same surgical team performed all the radical prostatectomies no later than 2 months post-MR examination. The specimens were submitted for routine histopathological evaluation, which followed the current recommendations for reporting of prostate carcinoma [35] and Gleason grading system [7, 8]. Radical prostatectomy specimens were coated with standard marking ink and fixed in buffered formaldehyde. To prepare standard blocks, axial slices through the prostate were obtained at 4- to 5-mm intervals in a plane perpendicular to the posterior surface of the prostate. Depending on the size, each slice was then divided into halves (right or left) or quadrants (right anterior, right posterior, left anterior, and left posterior), that were inserted into standard numbered blocks. Haematoxylin and eosin–stained histological sections were produced of each half or quadrant. Each significant focus of adenocarcinoma was recorded for the report, regarding its sextant location and Gleason score. No whole-mount section maps were produced for correlation with MR images. The histopathological report included the sextant of the most significant (or ‘index’) prostatic lesion [36], the 2002 pTNM staging, and the Gleason grading with its primary and secondary components. For disambiguation matters, the term “prostatectomy Gleason score with the two grades listed” (pGG) was adopted to describe the primary and secondary components (e.g. pGG 7 = 3 + 4), whereas simply “Gleason Score” (GS) refers only to the sum of the components (e.g. GS 7). Additionally, a qualitative grading scale was also proposed, with ‘low-grade’ lesions consisting of GS 6 or lower, and ‘intermediate/high-grade’ lesions consisting of GS 7 or higher.

In order to determine the correlation between ADC values of suspicious lesions and the specimen’s pGG, Pearson’s correlation coefficients were calculated. The same was performed to correlate each patient’s “biopsy Gleason Scores with the two grades listed” (bGG) to the pGG. A significant result was considered when p < 0.05. Statistical analyses were performed with SPSS version 17.0.01 (IBM, Chicago, IL, USA).

Results

Of the 35 consecutive patients, 28 contained clinically significant peripheral zone tumours (> 0.5 cm3). Additionally, four of these 28 patients were excluded from the study because of the following: extensive extra capsular extension (1), clinical contraindications to surgery, (aortic aneurysm and haemophilia) (2), and prostatectomy at another site (1), leaving 24 patients for statistical analysis. The patient demographics and tumour characteristics are summarised in Table 1. In all of the 24 prostatectomy specimens, histological analysis confirmed the main tumour foci in the same sextants as the suspicious areas indicated by MR images.
Table 1

Patient demographics

Parameter

Number

Number of patients

24

Clinical Demographics

Median PSA ng/ml (range)

7,36 (3,45 - 21,1)

Mean Age yrs (range)

63 (52–78)

Pathology Demographics

Biopsy Gleason Scores (grades listed)

5 (2 + 3)

1

5 (3 + 2)

1

6 (3 + 3)

12

7 (3 + 4)

6

7 (4 + 3)

4

Prostatectomy Gleason Scores (grades listed)

6 (3 + 3)

7

7 (3 + 4)

9

7 (4 + 3)

4

8 (4 + 4)

1

9 (4 + 5)

1

9 (5 + 4)

2

Stage

pT2a

3

pT2b

1

pT2c

14

pT3a

4

pT3b

2

The mean ADC value of suspicious areas on preoperative MR imaging was 763.7 × 10-6 mm/s2 (range 307–1219 × 10-6 mm/s2), and the mean ADC values among the different Gleason scores (with and without the grades listed) are specified in Table 2. Statistical analysis with Pearson’s coefficient showed a significant negative correlation between ADC values and pGG (r = −0.55; p < 0.01), and an even stronger negative correlation with GS (r = −0.63; p < 0.01) (Fig. 2a, c). On the other hand, there was poor correlation between bGG and pGG (r = 0.042; p > 0.05) or GS (r = 0.048; p > 0.05) (Fig. 2b, d). There was also no significant correlation between bGG and the ADC values (r = −0.095; p > 0.05).
Table 2

Apparent diffusion coefficient value (ADC) by tumour aggressiveness

Gleason Score (grades listed)

ADC (x10−6 mm/s2)

6 (3 + 3) (n = 7)

914 ± 195

7 (3 + 4) (n = 9)

733 ± 128

7 (4 + 3) (n = 4)

775 ± 91

8 (4 + 4) (n = 1)

609

9 (4 + 5) (n = 1)

307

9 (5 + 4) (n = 2)

654 ± 50

Gleason Score

6 (n = 7)

914 ± 195

7 (n = 13)

746 ± 115

8 (n = 1)

609

9 (n = 3)

538 ± 203

https://static-content.springer.com/image/art%3A10.1007%2Fs00330-011-2269-1/MediaObjects/330_2011_2269_Fig2_HTML.gif
Fig. 2

Correlation between Apparent Diffusion Coefficient (ADC) and biopsy Gleason grades (bGG) with prostatectomy Gleason grades (pGG) or Gleason scores (GS). A,B. Scatter plot graphs, with each point representing the measured ADC value of a suspicious area at the peripheral zone, in correspondence with either the pGG (graph A) or GS (graph B). Straight line represents the linear relationship between the ADC value and pGG (A) or GS (B), and was estimated at −0.55 and −0.63, respectively, from Pearson’s correlation analysis. p < 0.01 indicates that the negative relationship is statistically significant for both. C,D. Scatter plot graphs, with each point representing reported bGG of tumours at the peripheral zone, in correspondence with either the pGG (graph C) or GS (graph D). Straight line represents the linear relationship between bGG and pGG (C) or GS (D), and was estimated to be 0.042 and 0.048, respectively, from Pearson’s correlation analysis. p > 0.05 indicates that the relationship is statistically insignificant for both

The overall accuracy of biopsy samples for estimating Gleason grade was 37.5%, and slightly better for the estimation of GS (45.8%). Up to 45.8% of the patients (n = 11) had their Gleason grades upgraded from biopsy to prostatectomy, while 16.7% (n = 4) were downgraded.

On the qualitative evaluation, ‘low-grade’ tumours showed a mean ADC value of 914.0 × 10-6 mm/s2 (720–1219 × 10-6 mm/s2), and ‘intermediate/high-grade’ tumours had a mean ADC value of 701.8 × 10-6 mm/s2 (307–917 × 10-6 mm/s2). A Receiver Operating Characteristic (ROC) curve analysis was then performed (Fig. 3) and revealed a discriminatory performance of AUC = 0.82 and AUC = 0.46 for ADC and bGG, respectively, in discerning low-grade from intermediate/high-grade lesions in the peripheral zone.
https://static-content.springer.com/image/art%3A10.1007%2Fs00330-011-2269-1/MediaObjects/330_2011_2269_Fig3_HTML.gif
Fig. 3

ROC curve on the discriminating performance of ADC to differentiate between low-grade and intermediate/high-grade lesions (Green line), with AUC = 0.82, compared with the routine TRUS biopsy Gleason grade groups (bGG, Blue line), with AUC = 0.46

Discussion

In this study, we have prospectively evaluated the performance of diffusion-weighted MR imaging in the correlation with final prostatectomy Gleason scores in peripheral zone prostate cancers. This performance was also compared with that of the TRUS-guided biopsy Gleason scores on a case-by-case basis.

Our study has shown that the ADC values of suspicious areas in the peripheral zone on functional prostate MR imaging are strongly correlated with cancer aggressiveness at the final specimen histopathology (r = −0.55 and r = −0.63 for pGG and GS, respectively). This is most likely attributed to increased cellular density, structural changes of the stroma, textural disorganisation, and fibrosis, which are known causes of restricted water motion, and consequently low ADC values [37]. These results are in agreement with those of previous publications [2834] and support the concept of DWI as an imaging biomarker.

Furthermore, the most innovative result of the present study was the finding that ADC values on preoperative prostate MR correlate significantly better with the final prostatectomy Gleason grades than TRUS biopsy Gleason grades, with an approximately 13-fold difference (r = −0.55 vs. r = 0.042, respectively). Although the number of patients included is rather small, the results were significantly different, so as to justify such a comparison. The discrepancy between biopsy and specimen Gleason grades is already a notorious issue for pathologists and urologists, and is probably attributed to sampling error and multifocality/heterogeneity of cancer foci [1114]. Additionally, previously published data have shown that the diagnostic accuracy of clinically significant PCa improves with extended biopsies, usually involving 10 or more cores [38]. Considering that routine, unenhanced TRUS bears low accuracy for cancer localisation [39], and that as many as 32% of tumoral foci are isoechoic [40], most biopsy procedures are based on randomly distributed sextant biopsy samples that may not include the most aggressive tumour component or even the index lesion itself. In this context, the authors hypothesise that the best performance of ADC values for the correlation with the Gleason scores may be due to the actual visualisation of probable cancer foci by this technique, enabling the assessment of the most suspicious and probably most significant areas. However, as our results and previously published results have shown, the intra-subject heterogeneity and overlap of ADC values among Gleason scores is considerable, so that no single cut-off value has been established to allow confident differentiation between the different categories. Therefore, DWI alone may not yet be the best tool for estimating cancer aggressiveness.

Based on this assumption, investigators have tried to develop biopsy techniques that benefit from DWI findings, either through TRUS/MR fusion [41] or even through direct MR-guided biopsy [42]. A recent study by Turkbey et al. [32] retrospectively compared the Gleason scores of TRUS/MR fusion-directed biopsies with the ADC values of suspicious lesions on endorectal 3-T prostate MRI, and found a significant correlation between those parameters (Spearman’s ρ = −0.60), but made no mention of prostatectomy findings. In addition, Hambrock et al. [43] showed that the detection rate of MR-guided biopsies may be as high as 59% in patients with repeated negative TRUS biopsies, with up to 93% of those patients harbouring clinically significant cancers. Nevertheless, no publication to date has specifically assessed the correlation between MR-guided biopsies and final prostatectomy Gleason scores.

Another possible application for the results of our study in clinical practice is the development of nomograms that include DWI information in their calculations, as a way of preventing the understratification of patients with low biopsy Gleason scores and concomitant low ADC values, or vice versa. An example for this statement is a study by Wang et al. that demonstrated the incremental value of endorectal prostate MR imaging data to the Kattan nomogram in the prediction of seminal vesicle invasion [44]. In our study, the qualitative analysis by the ROC curve has demonstrated the potential of ADC measurements to discriminate between ‘low-grade’ and ‘intermediate/high-grade’ lesions. The good performance of ADC (AUC = 0.82) in comparison to that of TRUS biopsies (AUC = 0.46) may be further explored as a refinement to the current nomograms in larger prospective studies.

This study was designed with the intention to closely simulate the steps taken at routine clinical/surgical management of PCa. Therefore, despite its direct applicability in the clinical setting, a number of intrinsic limitations are to be considered. First, because histopathology was not based on whole-mount specimens, there is no direct correspondence of the suspicious areas on MR imaging with the actual tumour lesions. Nevertheless, all of the drawn ROIs were located at the most significant sextants appointed by histopathology. Second, only patients who were candidates for a prostatectomy were included in this study, which may represent a selection bias either through the exclusion of active-surveillance (and probably low GS) patients, or through the exclusion of advanced (and probably high GS) cases. Third, because images were interpreted by a single reader, the inter-observer agreement was not assessed in this study.

In conclusion, our results suggest that the ADC values of suspicious areas on prostate MR imaging are strongly correlated with prostatectomy specimen Gleason scores of peripheral zone cancers. Moreover, ADC values perform significantly better than TRUS biopsy Gleason scores in the correlation with the prostatectomy Gleason scores. Those results were achieved with widely available MR imaging technology (1.5-T systems), without the use of an endorectal coil, therefore being ready for clinical practice. Although there are still no definite ADC cut-off values recommended for the differentiation among Gleason scores, the use of DWI on the estimation of PCa aggressiveness may already prove valuable for the guidance on biopsy samples and for the integration into risk-assessment nomograms.

Copyright information

© European Society of Radiology 2011