Which Prostate Cancers are Undetected by Multiparametric Magnetic Resonance Imaging in Men with Previous Prostate Biopsy? An Analysis from the PICTURE Study

Background Multiparametric magnetic resonance imaging (mpMRI) has improved risk stratification for suspected prostate cancer in patients following prior biopsy. However, not all significant cancers are detected by mpMRI. The PICTURE study provides the ideal opportunity to investigate cancer undetected by mpMRI owing to the use of 5 mm transperineal template mapping (TTPM) biopsy. Objective To summarise attributes of cancers systematically undetected by mpMRI in patients with prior biopsy. Design, setting, and participants PICTURE was a paired-cohort confirmatory study in which men requiring repeat biopsy underwent mpMRI followed by TTPM biopsy. Outcome measurements and statistical analysis Attributes were compared between cancers detected and undetected by mpMRI at the patient level. Four predefined histopathological thresholds were used as the target condition for TTPM biopsy. Application of prostate-specific antigen density (PSAD) was explored. Results and limitations When nonsuspicious mpMRI was defined as Likert score 1–2, 2.9% of patients (3/103; 95% confidence interval [CI] 0.6–8.3%) with definition 1 disease (Gleason ≥ 4 + 3 of any length or maximum cancer core length [MCCL] ≥ 6 mm of any grade) had their cancer not detected by mpMRI. This proportion was 6.5% (11/168; 95% CI 3.3–11%) for definition 2 disease (Gleason ≥ 3 + 4 of any length or MCCL ≥ 4 mm of any grade), 4.8% (7/146; 95% CI 2.0–9.6%) for any amount of Gleason ≥ 3 + 4 cancer, and 9.3% (20/215; 95% CI 5.8–14%) for any cancer. Definition 1 cancers undetected by mpMRI had lower overall Gleason score (p = 0.02) and maximum Gleason score (p = 0.01) compared to cancers detected by mpMRI. Prostate cancers undetected by mpMRI had shorter MCCL than cancers detected by mpMRI for every cancer threshold: definition 1, 6 versus 8 mm (p = 0.02); definition 2, 5 versus 6 mm (p = 0.04); any Gleason ≥ 3 + 4, 5 versus 6 mm (p = 0.03); and any cancer, 3 versus 5 mm (p = 0.0009). A theoretical PSAD threshold of 0.15 ng/ml/ml reduced the proportion of patients with undetected disease on nonsuspicious mpMRI to 0% (0/105; 95% CI 0–3.5%) for definition 1, 0.58% (1/171; 95% CI 0.01–3.2%) for definition 2, and 0% (0/146) for any Gleason ≥ 3 + 4. Conclusions Few significant cancers are undetected by mpMRI in patients requiring repeat prostate biopsy. Undetected tumours are of lower overall and maximum Gleason grade and shorter cancer length compared to cancers detected by mpMRI. Patient summary In patients with a previous prostate biopsy, magnetic resonance imaging (MRI) overlooks few prostate cancers, and these tend to be smaller and less aggressive than cancer that is detected.


Introduction
Prebiopsy multiparametric magnetic resonance imaging (mpMRI) has excellent test accuracy, validity, and reliability for detection of clinically significant prostate cancer [1][2][3][4][5] resulting in its incorporation into national and international guidelines [6,7]. However, as with all cancer risk-stratification strategies, not every prostate cancer is detected by mpMRI [1]. Understanding the nature of disease that is undetected by mpMRI is important, particularly given the increasing preference for omission of prostate biopsy in cases of nonsuspicious prebiopsy imaging [6]. We have recently shown that in biopsy-naïve patients, so-called mpMRI-invisible cancer is significantly smaller in tumour size and has lower maximum and overall Gleason scores compared to mpMRI-visible disease [8].
Recent investigation into mpMRI performance in patients with prior biopsy has shown favourable features of undetected disease [9], consistent with a body of evidence identifying reassuring genetic, molecular, histopathological, and clinical characteristics for mpMRI-undetected cancer in biopsy-naïve patients [8,[10][11][12]. Nonetheless, concern remains regarding the potential for significant prostate cancer going undetected on mpMRI [13]. Existing evidence for men with prior biopsy is limited by imperfect reference standards, retrospective study designs, lower mpMRI magnetic strength, or poor image quality due to close timing between prior biopsy and imaging [14][15][16][17].
The Prostate Imaging Compared to Transperineal Ultrasound-guided biopsy for significant prostate cancer Risk Evaluation (PICTURE) study was a prospective pairedcohort confirmatory study that compared the diagnostic performance of mpMRI against a strict reference standard in 249 patients with prior prostate biopsy who required further risk stratification [3,[18][19][20]. Patients underwent prebiopsy mpMRI at 3T, followed by transperineal template prostate mapping (TTPM) biopsy (the reference test) in which biopsies were taken at 5 mm intervals throughout the prostate. Here we present a comparison of cancer attributes (at the patient level) between patients with mpMRI-detected and mpMRI-undetected disease in the PICTURE study.

2.
Patients and methods

Study population
In brief, PICTURE was a prospective single-centre trial in which patients with prior systematic transrectal ultrasound (TRUS)-guided biopsy and ongoing clinical suspicion underwent prebiopsy 3T mpMRI, followed by TTPM biopsy under general anaesthesia. The mpMRI parameters used are reported in full in the main PICTURE report [3].

Post hoc analysis
Once stratified by cancer threshold, patients were divided into mpMRIdetected (Likert score 3-5) and mpMRI-undetected (Likert score [1][2] groups. An additional threshold for tumour visibility was also evaluated (mpMRI-detected group, Likert score 4-5; mpMRI-undetected group, Likert score 1-3). Prostate-specific antigen density (PSAD) was calculated by dividing serum PSA by mpMRI-derived prostate volume (using the prolate ellipsoid method). Overall Gleason score was defined as the predominant Gleason pattern across the entire prostate and constituted the final pathological score. Maximum Gleason score was defined as the highest Gleason pattern found per patient.

PSAD
Overall, median PSAD was 0.18 ng/ml/ml (IQR 0.12-0.28) across the entire cohort. For men with prostate cancer, PSAD did not significantly differ between those with mpMRIdetected disease (Likert score 3-5) and those with mpMRIundetected disease (Likert score 1-2). However, application of theoretical PSAD thresholds above which a biopsy would be indicated altered the rates of undetected significant prostate cancer. Multiple hypothetical PSAD thresholds were evaluated for all cancer definitions and mpMRI detection thresholds (Table 4).

Discussion
In summary, our post hoc analysis of the PICTURE cohort showed that for patients with previous TRUS-guided biopsy, the proportion of the most aggressive prostate tumours undetected by 3T mpMRI is very low (2.9%). Overall, our findings in this patient subgroup support results from other investigators who found that prostate cancers undetected by mpMRI are significantly smaller and have lower pathological grade than those that are detected [9,26]. The results presented here also closely mirror our recent interrogation of the PROMIS data set in which undetected cancer had favourable characteristics at histopathology [8], highlighting parallels in mpMRI performance between patients with and without prior biopsy. Collectively, these findings support avoidance of biopsy in men requiring repeat risk stratification with nonsuspicious mpMRI, especially when PSAD is low (eg, <0.15 ng/ ml/ml). Furthermore, while not the primary focus of this analysis, the restratification performed in PICTURE also demonstrates the utility of mpMRI in predicting pathological upgrading (Supplementary Table 1), with 92% (120/131) of men with upgraded disease (compared to their preenrolment status) having positive or suspicious mpMRI findings (Supplementary Table 2).
Using PICTURE, our study provides a robust description of prostate cancers that mpMRI does not detect by using 5 mm TTPM biopsy as the reference standard. While this exhaustive approach may not represent the modern clinical approach (and thus may detect cancers with inherently different risk profiles) and is associated with higher risk of urinary retention and impairment of genitourinary function [27], it does overcome several methodological challenges intrinsic to whole-mount radical prostatectomy, especially selection bias. In addition to providing a unique insight into patients requiring further risk stratification, the PICTURE  data set also offers an advantage over PROMIS by providing histopathological-radiological correlation at a higher MRI magnet strength (PROMIS exclusively examined 1.5T mpMRI, while PICTURE exclusively examined 3T mpMRI) [1,3]. It is interesting to note that application of numerous different PSAD thresholds resulted in a more pronounced reduction in nondetected cancer than was noted in our previous analysis, and this is potentially attributable to higher overall PSAD in PICTURE. Our analysis has some limitations. PICTURE was a singlecentre study conducted at an experienced academic centre [3] and thus importantly lacks the generalisability provided by multicentre trials such as PROMIS [1]. Another limitation of this analysis is the per-patient strategy, in which single overall mpMRI scores were assigned (Likert score [1][2][3][4][5]. This approach mirrors real-life diagnostic settings; however, it may limit detailed tumour conspicuity investigation because of the inherent possibility of concurrent visible and invisible tumours, risking the possibility of ignoring invisible tumours owing to the overall positive mpMRI scores generated by visible lesions. However, our original PICTURE report, which included targeted biopsy (not included here), demonstrated that such scenarios are uncommon [19]; nevertheless, there are still situations, particularly as target-only biopsy becomes more common, in which nonvisible tumours may be overlooked in real-life clinical settings when only visible lesions are targeted. Furthermore, the benefits that we have demonstrated with the use of PSAD cutoffs for men with nonsuspicious mpMRI may be limited in reality, as they require full 5 mm TTPM in order to detect the same levels of significant disease that we have shown (in reality, a simple 12-core systematic TRUSguided biopsy is more likely to be offered, which would have much lower detection rates). Lastly, whilst the cancer yield was high in this cohort (probably because of our chosen population, ie, men with prior risk stratification), the most aggressive cancers (eg, grade group 4-5) were uncommon, and thus analyses regarding detection and nondetection of this disease generated wide CIs, suggesting limited study power for this particular question.
As with our previous reports in this field [8], we have shown that mpMRI detects nearly all high-grade prostate   cancers [1,3,8]. This is particularly important following the recent 29-yr update of the SPCG-4 trial, which demonstrated that these cancers are most strongly associated with prostate cancer-related death [28]. Combining these data suggests that mpMRI might deliver useful prognostic information and requires prospective evaluation. This is supported at multiple levels.
First, it appears that the genomic features of disease progression are enriched in mpMRI-detected tumours. Furthermore, this phenomenon goes beyond tumour volume and grade, which are (as we have demonstrated here) more favourable in undetected cancers. Indeed, mpMRI-detected tumours ostensibly harbour a greater proportion of molecular features of progression, including PTEN loss, biochemical recurrence (BCR)-associated genes (eg, CENPF), and elevated genomic scores (eg, Oncotype DX, Decipher, and Prolaris) compared to undetected disease [10,29], thus reinforcing the potential prognostic utility of mpMRI conspicuity. To validate this, future research should focus on exploring the molecular basis of cancer conspicuity on mpMRI in larger patient cohorts, and this, in part, is the focus of the ReIMAGINE trial (NCT04063566) investigating the role of genetic biomarkers in conjunction with mpMRI for diagnosis of prostate cancer.
Second, additional histopathological features of mpMRIundetected disease beyond tumour grade and size are also reassuring. For example, contrary to early accounts, aggressive prostate cancer subtypes (eg, cribriform pattern disease) now in fact appear to be predominantly detected by mpMRI according to pooling of data from multiple studies [30][31][32]. This is important, as these pathological entities are more strongly associated with BCR after radical prostatectomy.
Finally, it appears that undetected tumours on mpMRI behave favourably in the long-term setting, as demonstrated by retrospective clinical data [33] and through prediction of biochemical failure following radical prostatectomy [34]. Likewise, in the active surveillance context, tumour detection status on mpMRI may potentially provide greater utility than pathological grade alone. Recent findings from a contemporary mpMRI-directed active surveillance cohort suggest that mpMRI-detected moderate-risk prostate cancer behaves like low-risk prostate cancer, and conversely that mpMRI-undetected low-risk cancer behaves more like moderate-risk prostate cancer [35]. To expand further on existing evidence in this field, additional analysis of mpMRI-undetected prostate cancer at biological, histopathological, and clinical levels is currently under way.

Conclusions
In patients with prior prostate biopsy, mpMRI is highly unlikely to overlook clinically significant prostate cancer. Tumours undetected by mpMRI have significantly lower overall and maximum Gleason grade and are smaller in size. These results further support the utility of mpMRI not only for biopsy-naïve patients but also for those who have been advised to under further biopsies for accurate risk stratification. Ongoing work investigating longitudinal long-term mpMRI-correlated clinical outcomes will be instrumental in revealing the implications of various baseline mpMRI phenotypes over time.
Author contributions: Joseph M. Norris 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: All authors.
Analysis and interpretation of data: Ahmed, Emberton, Norris.
Drafting of the manuscript: Norris.
Critical revision of the manuscript for important intellectual content: All authors.