Role of Multiparametric Magnetic Resonance Imaging in Predicting Pathologic Outcomes in Prostate Cancer.

Multiparametric magnetic resonance imaging (mpMRI) and the introduction of standardized protocols for its interpretation have had a significant impact on the field of prostate cancer (PC). Multiple randomized controlled trials have shown that the sensitivity for detection of clinically significant PC is increased when mpMRI results are the basis for indication of a prostate biopsy. The added value with regards to sensitivity has been strongest for patients with persistent suspicion for PC after a prior negative biopsy. Although enhanced sensitivity of mpMRI is convincing, studies that have compared mpMRI with prostatectomy specimens prepared by whole-mount section analysis have shown a significant number of lesions that were not detected by mpMRI. In this context, the importance of an additional systematic biopsy (SB) is still being debated. While SB in combination with targeted biopsies leads to an increased detection rate, most of the tumors detected by SB only are considered clinically insignificant. Currently, multiple risk calculation tools are being developed that include not only clinical parameters but mpMRI results in addition to clinical parameters in order to improve risk stratification for PC, such as the Partin tables. In summary, mpMRI of the prostate has become a standard procedure recommended by multiple important guidelines for the diagnostic work-up of patients with suspicion of PC.

INTRODUCTION

For a long time, the urological community has been on a quest to find the best means for early detection of prostate cancer (PC). An important milestone was the introduction of prostate-specific antigen (PSA)-testing. While initial results regarding a structured screening were controversial, with longer follow-up of the European Randomized Study of Screening for Prostate Cancer (ERSPC) trial [1] and more detailed assessment of the Prostate, Lung, Colorectal, and Ovarian Cancer Screening (PLCO) trial [23] the advantage is increasingly evident, with the number needed to screen dropping below the levels of the widely accepted breast cancer screening [4].

A variety of markers have since been introduced with improved “sensitivity” and specificity, but low availability and high cost have prevented their wide adoption.

The most significant progress in the diagnosis of localized and metastatic PC has been achieved in the field of imaging. In the past, the only relevant option for an imaging the prostate has been trans rectal ultrasound but due to its low sensitivity and specificity it has mainly been used for guidance in systematic biopsy (SB) and has not been considered as a major independent tool for diagnostic assessment of PC.

In the last decade magnetic resonance imaging (MRI) was introduced as a diagnostic tool in PC and a standardized protocol which defined image acquisition and reporting was developed and refined by the Prostate Imaging-Reporting and Data System (PI-RADS) versions 1, 2, and 2.1 [567]. In addition to detection of PC, MRI has been investigated as relevant tool for differentiating between indolent and aggressive PC [8].

In recent years more emphasis has been placed on defining clinically significant PC (csPC) to reduce over diagnosis and concomitant overtreatment. In most studies clinically insignificant PC is defined as Gleason Group (GG) 1, conterminous to Gleason score 6. In this context, improved imaging is urgently needed.

In this article we will discuss the effect that multiparametric magnetic resonance imaging (mpMRI) has had on PC diagnostics by examining recent literature and providing an outlook to possible future developments.

MAIN BODY

1. Material and methods

To identify current relevant literature on the role of mpMRI in predicting pathological outcomes in PC, a systematic literature search was conducted in January 2020 using PubMed (https://www.ncbi.nlm.nih.gov/pubmed/). Search terms used included ‘mpMRI’, ‘prostate cancer’, ‘multiparametric MRI’, ‘targeted biopsy’, ‘prospective’, ‘randomized controlled’, ‘detection rate’, ‘extracapsular extension’, ‘seminal vesical invasion’, ‘pathologic stage’, ‘PI-RADS’. Additionally, references of included articles were screened for further relevant publications. To be included, manuscripts had to be original articles written in English and manuscripts older than 10 years were excluded to ensure topicality.

2. Standardization of multiparametric magnetic resonance imaging of the prostate

With the introduction of MRI in the field of PC diagnostics there was a need for a structured approach to image acquisition and reporting. Based on multiple consensus meetings the European Society of Urogenital Radiology created a protocol called PI-RADS [5]. This guideline described the minimum requirements necessary to acquire an mpMRI as well as a standardized interpretation and reporting system. Regions of the prostate are scored from 1 (clinically significant disease is highly unlikely to be present) to 5 (clinically significant cancer is highly likely to be present) (Fig. 1).

Fig. 1

Examples of lesions (marked by the arrows) graded Prostate Imaging-Reporting and Data System (PI-RADS) 1 to 5 according to PI-RADS version 2.1. ADC: apparent diffusion coefficient, DWI: diffusion-weighted imaging.

This protocol has extensively been validated resulting in version 2 [6].

In this updated version, the required parameters are T2-weighted (T2W) imaging, diffusion-weighted imaging (DWI), and dynamic contrast enhanced (DCE) MRI. Data are assessed depending on the anatomical zone with T2W and DWI being the dominant techniques for the transitional zones and peripheral zones, respectively. However, the benefit provided by DCE images has recently been called in to question [9].

Since significant interreader variability was problematic [10], this led to a second update to version 2.1 with more distinct guidance for scoring DWI between 3 and 4/5 as well as scoring of DCE for non-focal lesions in the peripheral zone [11].

An alternative protocol has been published as a Likert score. Although most recent studies report results according to the PI-RADS classification, Likert has been shown to provide similar diagnostic accuracy [12].

3. The pre-biopsy setting

MpMRI has high sensitivity for PC ≥GG 3. This was shown by Ahmed et al [13] in the prostate MRI Study (PROMIS) trial, which compared the sensitivities of systematic transrectal prostate biopsy and mpMRI with a transperineal mapping biopsy as reference. For the primary definition of csPC (Gleason score ≥4+3 or cancer core length ≥6 mm) mpMRI and SB had a sensitivities of 93% and 48% (p<0.0001), respectively. For the more common definition of csPC (any Gleason score ≥3+4) mpMRI and SB reached respective sensitivities of 88% and 48% (p<0.0001). Although the impact of this study on daily clinical practice was quite significant with most relevant guidelines recommending mpMRI before any prostate biopsy [1415], it did receive some criticism. For example transperineal template biopsy was chosen as reference for presence of PC. However, in some patients SB was positive while template biopsy did not find PC leading to a specificity of histological analysis of SB of only 96%, thereby showing the weakness of this reference.

From other studies we now know, that small lesions, even if high grade, can be missed by mpMRI [16]. This shortcoming might have been missed due to the lack of whole-mount-sectioning. Following shortly thereafter was the publication of the Prospective Randomized Evaluation of Celeboxib Integrated Safety vs. Ibuprofen or Naproxen (PRECISION) trial. Here, MRI-targeted biopsy (MRI-TB) in cases of suspicious lesions was randomized to SB. Initially set up as a non-inferiority trial, the MRI pathway had a significantly higher detection rate for csPC of 38% compared to 26% for SB [17]. Many prospective studies have compared detection rates of MRI-TB and SB [17181920212223242526], and a summary can be found in Table 1. The results of some studies that compared mpMRI to whole-mount section analysis is shown in Table 2. While sensitivity differs substantially within the cohorts reported here (sensitivities ranging from 74.3% to 100%), an important finding is consistent in all of them, high sensitivity for csPC on a patient basis shows that mpMRI is a relatively reliable pretest before prostate biopsy [162728293031323334]. Equally important with the ongoing discussions on overdetection and overtreatment of PC is the limited sensitivity for low grade PC. With growing acceptance of active surveillance as the preferred approach for GG 1 PC the reduced detection rates of mpMRI is often seen as an additional advantage, and therefore omitting SB may be advantageous. On the other hand, Rouvière et al [26] found that SB and MRI-TB equally contribute to the detection rate of csPC. In their study csPC was detected by SB only, MRI-TB only, or by both approaches in 14%, 20%, and 66% of cases, respectively.

Table 1

Prospective studies comparing PC detection rates of TB and SB

First author	Year of publication	Subjects (n)	TB (n)	SB (n)	Detection rate				Techniques of biopsy guidance
					Any PC		csPC
					SB (%)	MRI-TB (%)	SB (%)	MRI-TB (%)
Park [21]	2011	85	44	41	9.8	29.5	-	-	Cognitive fusion
Panebianco [22]	2015	1,140	570	570	37.7	73.2	36.8	71.9	Cognitive fusion
Arsov [23]	2015	104	104	104	35	34	25	26	Software fusion
Baco [18]	2016	175	86	89	54	59	49	44	Software fusion
Tonttila [19]	2016	130	53	60	57	64	45	55	Cognitive fusion
Taverna [24]	2016	200	100	100	26	24	12	15	Cognitive fusion
Porpiglia [20]	2017	212	107	105	29.5	50.5	18.1	43.9	Software fusion
Kasivisvanathan [17]	2018	500	252	248	48	46.8	26	38	Cognitive or software fusion
van der Leest [25]	2019	626	626	626	45	39.5	23.3	25.4	In bore
Rouvière [26]	2019	251	251	251	52.2	41.4	29.9	32.3	Cognitive fusion

PC: prostate cancer, TB: MRI-targeted biopsy, SB: systematic biopsy, n: number of patients included in the study, MRI: magnetic resonance imaging, csPC: clinically significant prostate cancer.

Table 2

Sensitivity of mpMRI of the prostate with whole-mount-sectioning as reference

First author	Year of publication	Patients (n)	Histologic lesions (n)	Sensitivity
				Any prostate cancer		csPC
				Patients (%)	Lesions (%)	Patients (%)	Lesions (%)
Bratan [27]	2013	175	362	-	53–59	-	85–88
Le [34]	2015	122	283	80	47	-	72
Chung [29]	2018	455	-	46.8	-	-	-
Borofsky [30]	2018	100	162	-	-	99	84
Kim [32]	2018	730		73.2		74.3
Asvadi [33]	2018	425	425	76	-	-	-
Kido [28]	2019	95	136	-	-	83.2	72.1
Lee [31]	2019	107	237	100	46	100	75.5
Ito [16]	2020	136	274	89.7	39.4	95.6	56

mpMRI: multiparametric magnetic resonance imaging, n: number of patients included in the study, csPC: clinically significant prostate cancer.

4. Patients with negative systematic biopsy and persistent suspicion for prostate cancer

The diagnostic pathway for patients after initial negative SB has been a matter of intense discussion and controversy. Prior to the era of mpMRI, saturation biopsy in cases of persistent suspicion had been the preferred approach due to lack of alternatives. Since the introduction of mpMRI and MRI-TB, many studies have been performed in this clinical setting. The largest retrospective analysis was published by Hansen et al [35]. In 70.4% of all patients (343/487) where there was at least one lesion of PI-RADS ≥3, using 24 core SB as the reference, MRI-TB had a negative predictive value of 92% and 99% for csPC defined as GG ≥2 or GG ≥3, respectively [35]. These results have been confirmed by several studies [3637383940], ans a summary can be found in Table 3.

Table 3

Studies evaluating the diagnostic accuracy of mpMRI in patients with prior negative prostate biopsy

First author	Year of publication	Patients (n)	Suspicious MRI	Histological reference	Sensitivity of MRI-TB	PPV of MRI	NPV of MRI	Definition csPC
Abd-Alazeez [36]	2014	54	54 (100%)a	Transperineal template	76%		92%	≥GG 2
Hansen [38]	2016	295	204 (69%)a	24 core systematic	-	91%	91%	≥GG 2
Hansen [35]	2017	487	343 (70.4%)b	24 core systematic	-	40%	92%	≥GG 2
Tsivian [40]	2017	50	41 (82%)b	Transperineal template	-	51%	100%	≥GG 2
Mortezavi [39]	2018	86	50 (58.1%)b	Transperineal saturation	53.90%	42%	86.1%	≥GG 2
Dal Moro [37]	2019	123	101 (82.1%)a	Saturation biopsy	100%		100%	≥GG 2

Suspicious findings on mpMRI were compared to histological findings on biopsy to calculate sensitivity, PPV and NPV for mpMRI.

mpMRI: multiparametric magnetic resonance imaging, n: number of patients included in the study, MRI: magnetic resonance imaging, TB: MRI-targeted biopsy, PPV: positive predictive value, NPV: negative predictive value, csPC: clinically significant prostate cancer GG: Gleason Group.

aSuspicion defined as Prostate Imaging-Reporting and Data System 3–5. bSuspicion defined as Likert score 3–5.

Many clinical societies have since included a recommendation for performing mpMRI after an initial negative SB and persistent suspicion for PC [14]. Furthermore, while the role of additional SB during MRI-TB as an initial biopsy is still unclear, MRI-TB after a previous negative SB seems to be sufficient for the detection of csPC. Studies on this issue have shown that tumors found exclusively by SB were mostly GG 1 and not considered clinically significant, and SB added only 2.3% to the detection rate by SB according to a Cochrane meta-analysis [41].

This data resulted in a strong recommendation by most guidelines to perform an mpMRI after negative SB and persistent suspicion for PC [14].

5. Planning of focal therapy

The promising results of mpMRI in the pre-biopsy setting have led to a wide adoption of local staging. In the field of focal therapy which is still mostly considered experimental, several consensus statements have underlined the importance of mpMRI in planning focal therapy and during follow-up [4243]. This has led to the inclusion of mpMRI in many protocols of studies investigating focal therapy for PC [44]. While this adoption is supported by the high detection rate of MRI-TB and SB, sensitivity on a lesion basis is not comparably high. In recent studies, detection rates on a lesion basis for any PC and csPC vary between 39% to 59% and 56% to 88%, respectively [162728293031323334]. A summary of studies on this issue can be found in Table 2. The most important predictors of detection were size ≥0.5 mL and GG ≥3, while tumor location within the prostate did not show a significant difference [16]. The assessment of whether the treatment of mpMRI visible lesions is sufficient depends on which definition of csPC is applied. GG 2 is most commonly defined as intermediate risk and is therefore recommended for treatment. An mpMRI seems insufficient for single treatment guidance and should be supplemented by a template biopsy. Additionally, Le Nobin et al [45] reported, that mpMRI underestimates the size of lesions especially for higher cancer grades.

6. Prediction of locally advanced stages

To ensure satisfactory functional outcomes without limiting the oncological results a thorough assessment of the local disease stage is mandatory. The tools that are best evaluated to predict extracapsular extension (ECE) or seminal vesical invasion mostly rely on PSA-values and biopsy results. The Partin tables provide probabilities based on Gleason score, serum PSA levels and digital rectal examination (DRE) [46]. The reliability of these tools has repeatedly been assessed. For example Augustin et al [47] showed that the accuracy of the Partin tables for organ-confined disease depends upon the biopsy approach with an area under the curve (AUC), estimated by the receiver operating characteristic curve, in the initial biopsy setting of 0.73 and saturation biopsy (≥20 cores) of 0.585.

Yu et al [48] validated the Partin table tool for ECE in 11,185 patients resulting in an AUC of 0.62, while Bhojani et al [49] evaluated a cohort of 3,105 patients showing an AUC of 0.789. In summary a tool that relies solely on biopsy results, PSA and DRE does not seem to provide a satisfactory prediction of ECE.

These results have led to many studies which evaluated the use of mpMRI in this setting [5051525354], which are summarized in Table 4.

Table 4

Studies evaluation sensitivity, specificity and AUC for local staging with multiparametric MRI in prostate cancer

First author	Year of publication	Subjects (n)	Extracapsular extension			Seminal vesical invasion			Any pT3 stage
			Sensitivity	Specificity	AUC	Sensitivity	Specificity	AUC	Sensitivity	Specificity	AUC
de Rooij [50]	2016	5,681	0.57	0.91	-	-	-	-	-	-	-
de Rooij [50]	2016	5,677	-	-	-	0.58	0.96	-	-	-	-
de Rooij [50]	2016	4,001	-	-	-	-	-	-	0.61	0.88	-
Rayn [53]	2018	532	-	-	0.78	-	-	0.86	-	-	0.78
Martini [54]	2018	561	-	-	0.688	-	-	-	-	-	-
Gandaglia [51]	2019	614	0.25	0.93	-	0.21	0.98	-	-	-	-
Mehralivand [52]	2019	553	0.30	0.96	-	0.10	0.99	-	-	-	-

The assessment of experienced radiologist, blinded to the final pathology, were compared to prostatectomy specimen.

AUC: area under the curve, MRI: magnetic resonance imaging, n: number of specimens included in the study.

For ECE Martini et al [54] found an AUC of 0.688 using mpMRI in their cohort of 561 patients. Radiographic assessments always depend on high quality images and the experience of the interpreter. Zanelli et al [55] showed that this applies to local staging for PC as well with a difference in sensitivity between any two radiologists that differs between 0.583 to 0.667 for pathological stage T3 disease. Ma et al [56] attempted to circumvent this limitation with an automated analysis after contouring of the prostate. Each half of the prostate was investigated separately with an AUC of 0.821 in the validation cohort of 90 prostate lobes. An external validation of these promising results has yet to be performed.

Counterintuitively, Zanelli et al [55] found that adding clinical scoring to mpMRI did not improve results in most combinations. Her group evaluated three different readers of the same images alone and in combination with Partin tables and Cancer of the Prostate Risk Assessment (CAPRA) scores. With an AUC of 0.73–0.75 for mpMRI alone CAPRA scoring improved the AUCs of two readers to 0.76. The remaining combinations did not provide any benefit [55]. Feng et al [57] showed significant improvement of the AUC of clinical assessment tools by adding mpMRI leading to an AUC of 0.92 compared to 0.85 for Partin tables alone.

7. Molecular features of tumors not visible by multiparametric magnetic resonance imaging

Some PC lesions were found to be missed by mpMRI in the aforementioned studies. While some of these cases might be explained by low grade cancer or small tumor volume, some patients presenting with high grade PC do not have visible lesions on mpMRI.

A known reason for reduced visibility is a cribriform architecture found in some PC tumors. In a small cohort of 22 radical prostatectomy specimens, Truong et al [58] found that only 36% of all cribriform lesions (5/14) were visible on mpMRI. However, other studies have provided conflicting evidence. For example Tonttila et al [59] reported a visibility rate of lesions with any amount of cribriform or intraductal histology of 90.5% in a cohort of 95 radical prostatectomies. The sensitivity for lesions with this adverse histology is of utmost importance due to the recently reported worse outcome of these tumors, with higher rates of biochemical recurrence and metastasis after radical prostatectomy [60].

Houlahan et al [61] investigated the genome, transcriptome and histology of mpMRI-visible and invisible tumors. Detectable lesions showed a more aggressive molecular profile. In their study no single underlying factor for visibility could be identified, but rather a combination of several criteria including cribriform architecture or intraductal carcinoma, a higher amount of genomic alterations and overexpression of key non-coding transcripts, such as SCHLAP1 and snoRNAs [61]. A similar combination of different alterations has been described by Chua et al [62]. In their study, these alterations were also associated with higher risk of biochemical failure and metastasis.

8. Limitations of multiparametric magnetic resonance imaging

There are important limitations that must be discussed in the context of mpMRI. A DCE is required for a complete assessment according to PI-RADS, which may represent an issue for patients with decreased renal function. With an aging population more and more patients are receiving hip replacements or pacemakers which often lead to artifacts that limit the accuracy of reporting or represent a contraindication to perform an MRI altogether. Most importantly if an mpMRI becomes a requirement for every man with a suspicion of PC, then availability will become a growing problem.

For these challenges a combination of different solutions will be necessary.

Microultrasound is a new imaging modality that is currently being evaluated as a possible alternative. With a very high frequency and therefore high spatial resolution it has the potential to improve the sensitivity of ultrasound while maintaining high availability, low cost and few contraindications [636465].

Additionally, there is growing evidence that support protocols which rely less on mpMRI. Woo et al [9] performed a metanalysis showing that omitting DCE and by performing a biparametric MRI instead had similar sensitivity for csPC (0.87 vs. 0.86).

Finally, one should not ignore the power of diagnostic tools that have existed for many years: for a patient with a PSA of 100 ng/mL or cT4 by DRE the harm of postponed diagnosis will probably outweigh the benefit of a pre-biopsy mpMRI.

CONCLUSION

Within the last decade mpMRI of the prostate has become a standard of care method for detection of PC. Numerous studies have shown that the sensitivity for csPC on a patient level is high enough for it to be recommended in all guidelines.

The increasing importance of mpMRI as a diagnostic tool is also the result of the observation that mpMRI findings correlate with features of adverse outcome leading to negative results in many tumors that are considered as insignificant. However, a small proportion of high-grade tumors remain undetected by mpMRI. This is most relevant when considering focal therapy as a treatment option.