Magnetic Resonance Imaging Follow-up of Targeted Biopsy–negative Prostate Lesions

Background: The optimal radiological follow-up of prostate lesions negative on magnetic resonance imaging (MRI)-targeted biopsy (MRI-TB) is yet to be optimised. Objective: To present medium-term radiological and clinical follow-up of biopsy-negative lesions. Design, setting, and participants: The records for men who underwent multiparametric MRI at


Introduction
Although we and others have shown that the decision to sample Prostate Imaging-Reporting and Data System (PI-RADS) or Likert 3 prostate magnetic resonance imaging (MRI) phenotypes can be informed by factors such as prostate-specific antigen density (PSAD) and lesion characteristics, targeting of clearly suspicious lesions (PI-RADS or Likert 4-5) is widely considered uncontroversial (at least for the time being) [1,2]. Nonetheless, one in five highly suspicious lesions are negative on MRI-targeted biopsy (MRI-TB): in PRECISION, 31% of PI-RADS 4/5 lesions and 6% of 5/5 lesions did not yield cancer on targeting [3]. The challenges in negative MRI-TB scenarios are (1) to distinguish false-positive MRI from false-negative biopsy findings and (2) to determine the level of follow-up stringency required to ensure no new or misclassified disease is missed. As mentioned, we previously addressed the former question by showing that MRI false positivity is associated with specific clinicoradiological attributes such as lower PSAD and T2 or contrast-enhanced sequence signals without substantial diffusion restriction [2]. In this paper we address the latter question by providing medium-term radiological outcomes for biopsy-negative lesions. We specifically demonstrate that such lesions largely resolve over time and their resolution goes hand in hand with a reduction in PSAD.

Clinic setting
The University College London Hospitals (UCLH) prostate cancer onestop clinic is a specialist service specifically designed for men with an

Baseline characteristics
The baseline characteristics are presented in by far the most prevalent noncancerous pathology on MRI-TB (Table 2 and Fig. 2A), which was present in 64 men (74%). This was followed by acute inflammation in 42 men (46%), PIN in 33 (36%), chronic inflammation in 18 (20%), atypia in 13 (14%), and granulomatous inflammation in three (3%). Lesion-level analyses generally echoed these findings (Fig. 2B). Multiple correspondence analysis of prostate-level pathology demonstrated that acute inflammation was often observed in specimens with atrophy, while the presence of PIN was more often noted in specimens with atypia (Fig. 2C). Regardless of their hierarchy, lesions were mainly distributed in the peripheral zone, and most spanned the middle and apical prostate segments ( Supplementary Fig. 1).

Clinical, radiological, and MRI-TB follow up
The median follow-up for the entire cohort was approximately 1.8 yr (656 d, IQR 359-1008). Follow-up information is given in Table 1. Of 91 men, 58 had a second MRI (median interval from baseline 376 d, IQR 361-412). The median prostate volume at the second MRI was 55 ml (IQR 45-69), median PSA was 7 ng/ml (IQR 4.8-9.2), and median PSAD decreased to 0.11 ng/ml 2 (IQR 0.08-0.18). These changes were accompanied by substantial shifts in MRI scores (Fig. 3): only five out of 58 men (9%) had a persistently high overall Likert score of 4, while 40 (69%) were scored as Likert 3 and the remaining 13 (22%) as Likert 2 (no visible lesion). In addition, of the 45 men with Likert 3 scores, most now had only one lesion (28 men; 62%). Eleven men had a third MRI (with a persisting Likert 4 score in one patient), and the proportion of men without lesions increased further in comparison to the second scan (four men; 36%). Of the six men who had repeat MRI-TB during the study period, two were subsequently diagnosed with prostate cancer and both had persistent Likert 4 scores (at baseline and at least one follow-up MRI). The first patient (age 57 yr; no abnormal pathology) was diagnosed 3.7 yr (1357 d) after baseline and underwent radical prostatectomy (RP) for a 7-ml Gleason 3 + 4 cancer. The second patient (age 68 yr; PIN, atypia, and atrophy on pathology) was diagnosed 3.3 yr (1222 d) after baseline with Gleason 4 + 4 disease at RP. The remaining four patients with repeat MRI-TB were negative for cancer; their pathology is given in Table 2. Four patients in total had benign prostatic hyperplasia treatment (transurethral resection, laser enucleation, or embolization), all within 2.3 yr from baseline.

Discussion
This work further contributes to the evidence base on the fate of TB-negative prostate MRI lesions. We showed that most of these highly scoring phenotypes are associated with atrophy or other noncancerous pathologies and short-term radiological resolution (both in terms of score severity and lesion number) is their most likely outcome, usually accompanied by a decrease in PSAD. In a similar series of 88 TBnegative cases described by Meng and colleagues [4], the rate of persistently high-scoring lesions was higher at 27% (12/45 men with a second MRI). This could be attributed to multiple factors, including different targeting, MRI acquisition, and reporting protocols (PI-RADS v2 was used instead of Likert). The rate of cancer on subsequent MRI-TB was five out of eight men, in comparison to two out of six in our series. Other authors more recently reported that the rate of subsequent clinically significant cancer on follow-up targeting ranges from 4.9% to 23% [5,6]. An equally interesting parallel finding in the Meng study is the low rate of subsequent cancer on concurrent systematic biopsy in those with decreasing radiological scores (2/13 men; 15%) [4]. These results could imply that nonresolution of targeted lesions foreshadows prostate cancer diagnosis, while subsequent cancer risk outside these areas is low. This view is corroborated by recent work in larger cohorts: Kinnaird and colleagues [5] reported no clinically significant cancer in men with negative follow-up MRI. In a carefully designed study by Barletta and colleagues [6] with median follow-up of 31 mo for 68 men who underwent a followup TB, the 3-yr survival free from a clinically significant cancer diagnosis survival was 92% for those with positive and 65% for those with negative follow-up MRI.
Similarly, Kornienko and colleagues [7] reported that among 84 men with PI-RADS 4 or 5 lesions and repeat MRI, more than half of the lesions were downgraded to PI-RADS 3 at a median of 28 mo, while clinically significant disease was detected in 10/24 of these men who also underwent repeat biopsy, all with persistent MRI lesions.
Pepe and Pennisi [8] estimated that the rate of some form of inflammation for PI-RADS 5 lesions was approximately 37%. Meng and colleagues [4] also found that inflammation, atypia, and other pathologies were prevalent at baseline-just as in our population-but these were not predictive of subsequent cancer. This should not be necessarily taken as an indication against a causal relationship between inflammation and cancer, for reasons discussed elsewhere [9]. However, determining the histopathology of TBnegative lesions is important because, despite past assertions, formal studies on the topic have only started to appear recently: glandular-stromal alterations, acute or chronic inflammation, and vascular changes have now been found in most false-positive MRI lesions and, interestingly, are more prevalent and more synchronous in MRI-TB tissue in comparison to systematic biopsy cores from the same patients [10][11][12]. Although our population did not undergo standard systematic sampling, our multiple correspondence analysis showed that atrophy on MRI-TB is closely associated with acute inflammation, whereas PIN often cooccurs with atypia. Whether this implies the existence of two distinct MRI false-positive types with a different microstructural basis or a reporting artefact remains to be indicating that their prevalence is similar across targetable areas, regardless of whether the lesion was primary (index) or secondary. (C) Multiple correspondence analysis (MCA) of prostate-level pathology data confirmed that almost half of the data variance observed can be explained by two principal components, one related to atrophy and acute inflammation (horizontal axis) and the second one related more to the presence of PIN and atypia (vertical axis). In the MCA plot, names ending in _NO indicate the absence of a particular pathology, and those ending in _YES indicate presence. Gr = granulomatous inflammation; Atyp = atypia.
S X X X ( X X X X ) X X X -X X X seen. We also note that biopsy-negative lesions in our population mostly spanned the middle and apical peripheral zone.
The main limitations of our study include the almost complete lack of repeat biopsy data and the nonblinding of the reporting uroradiologists to the initial biopsy result, which could influence scoring of follow-up MRIs. The lack of systematic sampling also comes with a small risk of missing significant disease, although this risk has been deemed to be small and inherently acceptable in the TB era [13]. Even though the Likert scheme was used instead of PI-RADS v2, there is substantial experience with this score in our institution; it is also endorsed by UK policy makers and its performance against PI-RADS v2 has been favourably assessed by other authors [14][15][16]. In addition to the main limitations, we should also note that the assessment of noncancerous pathologies is not as standardised as the reporting for prostate cancer pathology. Here, reporting of noncancerous entities was binary (simple presence or absence of inflammation or PIN) even though a quantitative approach could be necessary for more accurate histology-MRI correlations. Finally, although our study is informative, longer follow-up is an important prerequisite for advocating a particular monitoring strategy, which we cannot propose on the basis of our findings. For the moment, we note that most TB-negative lesions improve over time and the short-term risk of subsequent significant cancer in the small number of men who undergo repeat targeting is associated with lesion persistence rather than resolution. Equally, the combination of negative follow-up MRI with low PSAD can be reassuring: Barletta and colleagues [6] reported a 3-yr rate of clinically significant disease of 6% among such cases. Collectively, these observations align with systematic review evidence supporting the view that repeat MRI with follow-up targeting in cases of persistently suspicious imaging findings is a reasonable management approach, although larger studies are required [17].

Conclusions
In conclusion, negative MRI-TB of highly-scoring prostate lesions is associated with noncancerous entities such as PIN, inflammation, and atrophy. Such lesions tend to regress to less suspicious phenotypes, commonly accompanied by a reduction in PSAD after the first year. Although repeat biopsy should always be considered if there is suspicion of missed significant disease (especially when lesions persist over time), repeat PSA imaging in selected cases could be a clinical alternative that spares men from early repeat biopsies.