1. Introduction
The Prostate Imaging—Reporting and Data System (PI-RADS) score is important for standardized prostate magnetic resonance imaging (MRI) acquisition and reporting [
1,
2,
3,
4]. Depending on the nature of the cohort and by following the PI-RADS guidelines, a not negligible number of lesions will be scored as PI-RADS 3, which is termed equivocal [
5]. The prevalence of clinically significant prostate cancer (csPCa, often defined as International Society of Urological Pathology [ISUP] grade ≥2 PCa) in biopsied PI-RADS 3 cases varies from 3% to 50% in the literature [
6,
7].
PI-RADS 3 lesions are challenging because their characteristics in MRI have a great overlap with benign conditions [
8,
9]. On the other hand, tumors that are less visible using T2-weighted (T2W) and apparent diffusion coefficient (ADC)-based tissue contrast might be classified as PI-RADS 3, despite the presence of Gleason ≥4 patterns [
10]. The PI-RADS guidelines propose recommendations for MRI-directed biopsy strategies, but do not clearly state how to deal with these category 3 imaging findings. In men with an overall PI-RADS score 3 in prostate MRI, an MRI-directed biopsy should be considered; however, biopsy can be avoided or deferred in carefully chosen patients if they are not at high risk of csPCa [
2]. Thus, while a targeted biopsy (TBx) may appear to be the logical next step in PI-RADS 3 cases, monitoring lesion characteristics with follow-up MRI and thereby postponing biopsies also seems to be an acceptable option [
11].
The risk stratification of suspicious MRI lesions could help to avoid unnecessary biopsies. Studies on sub-classifying PI-RADS 3 lesions are limited and include most often small cohorts. The available data indicate that prostate-specific antigen density (PSA-D) might be useful in predicting the presence of csPCa [
7,
11,
12,
13,
14,
15,
16,
17]. From the perspective that size matters, subcategorizations of PI-RADS 3 lesions based on size have been proposed [
18,
19,
20,
21]. However, sufficient evidence to develop clear directions on lesion characteristic-specific management of PI-RADS 3 lesions in csPCa diagnosis is still lacking, especially in men scheduled for MRI and TBx for the first time.
Using one of the largest series of men with an overall PI-RADS score of 3 on prostate MRI undergoing a first TBx session currently available, we aimed to investigate whether stratifying PI-RADS 3 lesions based on the largest (index) lesion diameter could aid in avoiding TBx sessions and low-risk PCa diagnoses without missing the diagnosis of csPCa. In addition, acknowledging earlier publications of its potential usefulness we also studied PSA-D as stratification tool and combined this information with the largest index lesion diameter and age in a multivariable prediction model.
4. Discussion
Patients with PI-RADS 3 index lesions scheduled for first-time TBx represent a diagnostic problem, as there is controversy regarding whether these men should be biopsied or could safely be monitored with follow-up MRI. In this study, using a large international dataset, we found that csPCa with a detection percentage of 3–13% (based on different definitions) on first-time TBx was uncommon in PI-RADS 3 cases. This could imply that PI-RADS score 3 cases represent a category of men in whom initially monitoring with follow-up MRI could be a realistic option. The csPCa (I) detection at the patient level was 13% in our total population, with almost half of the cancers being not higher than ISUP grade 2 PCa without CR and/or IDC. Our csPCa (I) detection percentage is in line with the recent meta-analysis of Maggi et al. which included 25 studies, showing a csPCa detection percentage in PI-RADS 3 cases of 18.5% (95% CI 16.6–20.3; range 3.4–47) [
7]. The reasons for the wide range of csPCa detection in PI-RADS 3 index lesions in the literature include, among others, the considerable interobserver variability in the characterization of equivocal lesions caused by reader experience and differences in technical performance, the prevalence of csPCa in different populations, and TBx-related factors. These factors could potentially also explain the difference in csPCa (I) found between the Düsseldorf and Rotterdam cohorts. The low event rate in the Düsseldorf cohort limited us to performing logistic regression analyses separately per cohort. For these analyses, the two cohorts were regarded as one, in line with analyses on heterogeneous cohorts [
24].
Instead of a monitoring all or a biopsy-all PI-RADS 3 patient strategy, a more realistic approach to avoid TBx sessions and low-risk PCa diagnoses would be to apply a risk stratification strategy. Risk stratification for TBx decision solely based on the largest index lesion diameter did not aid in avoiding TBx sessions while assuring csPCa (I) detection. On the contrary, risk stratification based on PSA-D only or a multivariable approach including the next to largest index lesion diameter, PSA-D, and age could result in avoiding a substantial number of TBx sessions and low-risk PCa diagnoses at the cost of missing only limited numbers of csPCa (I) diagnoses. These results suggest that when TBx is considered in men with PI-RADS 3 index lesions scheduled for first-time TBx, risk stratification based on PSA-D or preferably a multivariable model-based risk stratification approach is advisable.
To the best of our knowledge, this is the first study to investigate largest (index) lesion diameter as a stratification tool in a large daily clinical cohort of men with PI-RADS 3 index lesions. According to our results, the largest index lesion diameter is not a significant predictor of csPCa in PI-RADS 3 cases. To lower the risk of statistical overfitting, largest index lesion diameter is included in the multivariable prediction model [
25]. In our cohort, slightly more csPCa (I) was detected in PI-RADS 3 index lesions with a diameter ≥10 mm (14%, 31/217), compared to index lesions with a diameter <10 mm (11%, 8/75). This finding could suggest that csPCa is relatively rare in smaller PI-RADS 3 index lesions. Rais-Bahrami et al. suggest that small MRI index lesions (≤7 mm) may correspond to benign lesions or indolent cancers [
18]. Furthermore, Rosenkrantz et al. proposed to upgrade a PI-RADS 3 to a PI-RADS 4 lesion on the basis of larger size [
20,
26]. These assumptions do, however, not take into account the scenario that in the studied series small PI-RADS 3 index lesions harboring csPCa could have been mis-sampled by TBx. The absence of csPCa in the TBx specimens would then mean that csPCa was missed and not that there was no csPCa present [
6,
27]. However, if this would really be the case follow-up of the lesions with MRI could overcome the problem of missing a timely csPCa diagnosis.
PSA-D showed to be a significant clinical predictor of csPCa in our cohort. Applying solely PSA-D as risk stratification tool could result in 25% less TBx sessions and 11% less low-risk PCa diagnoses, missing only 5% csPCa (I). This high predictive value of PSA-D in PI-RADS 3 cases is in line with previous studies, and also with studies reporting on TBx and SBx histopathology outcomes [
17]. Venderink et al. showed that offering a biopsy to only PI-RADS 3 men with a PSA-D of ≥0.15 ng/mL
2 resulted in 42% of biopsy sessions avoided at the cost of missing 6% csPCa. Lowering the threshold to ≥0.12 ng/mL
2 would result in 26% of biopsy sessions avoided, missing no csPCa [
28]. Therefore, PSA-D may represent a good index to decide which PI-RADS 3 men should undergo a biopsy [
29]. The risk stratification of PI-RADS 3 cases could further be improved by a model-based approach in which PSA-D, largest index lesion diameter, and age are combined in a multivariable prediction model that predicts the risk of csPCa of a PI-RADS 3 man, as shown by our findings. To the best of our knowledge, our study is one of the first studies, next to the work of Di Trapani et al., to show the high added value of such a model-based approach in safely avoiding TBx sessions and low-risk PCa diagnoses in specifically PI-RADS 3 cases [
29].
Next to the most often used definitions for csPCa, we studied the prevalence of PI-RADS 3 lesions related to the presence of CR and IDC in TBx specimens. CR and IDC are prognostic drivers in cancer-specific survival, even more than other Gleason 4 patterns [
30,
31]. Although ISUP grade ≥2 PCa was our primary outcome measure for csPCa, the incorporation of this secondary growth pattern information into the risk stratification could further improve the selection of men who will benefit from treatment, especially because almost half of the detected ISUP grade ≥2 PCa in our cohort was ISUP grade 2 without CR and/or IDC PCa. Therefore, we may argue that the threshold for csPCa should be ISUP grade ≥2 with CR and/or IDC PCa to save even more TBx sessions in men with PI-RADS 3 index lesions and thereby avoid the (over)detection of ISUP grade 2 PCa, which potentially could never harm a patient if left undetected [
32].
The strength of our study is the inclusion of data from two centers, resulting in one of the largest series of men with an overall PI-RADS score 3 undergoing first-time TBx. This makes our study results more generally representative by giving more an overall view of the real-world setting of PI-RADS 3 lesions in daily practice, compared to reporting single center results. It must, however, be noted that every institution should know their own test performance statistics when making clinical decisions based on prostate MRI findings, because of existing differences in radiology, fusion biopsy and pathology learning curves per institution [
33]. Furthermore, our analysis of different csPCa definitions including the presence of secondary growth patterns is of high added value for further clinical decision-making.
Some limitations of our study should be highlighted. First, our study has a retrospective design and could thereby introduce a selection bias. However, our study represents a cohort of consecutive men. Second, men included were treated over a long time frame in which changes in the PI-RADS classification also occurred. However, the newer PI-RADS versions may not necessarily be better regarding diagnostic accuracy than the original PI-RADS version [
34,
35,
36]. Third, we did not include SBx or prostatectomy outcomes as a reference standard in our analyses. Some literature suggests performing a combined biopsy strategy in PI-RADS 3 cases [
7]. However, since our primary objective was to establish directions for the lesion characteristic-specific management of PI-RADS 3 lesions in csPCa diagnosis, SBx outcomes would not have been of added value in answering our research questions. Furthermore, our results are similar to studies investigating PSA-D as a stratification tool but reporting on TBx and SBx outcomes [
17]. This suggests that the TBx-only strategy could be similar in csPCa detection to the combined biopsy strategy in PI-RADS 3 cases. Nevertheless, when considering a biopsy in PI-RADS 3 men, we advise that adding SBx to TBx should be discussed at an individual level taking into account the benefits and harms. Fourth, although we have found potential predictors of csPCa in men with a PI-RADS 3 index lesion, the constructed prediction model is not (yet) usable in clinical practice for TBx decision management. To construct a more robust prediction model for TBx decision in PI-RADS 3 cases, more data are necessary and an external validation of the model is advised before its application in clinical practice. Lastly, the lesion measurements, although measured according to the PI-RADS recommendations, were not standardized. We acknowledge that standardized MRI lesion measurement should be the gold standard [
37,
38]. However, as long as this is not implemented in routine clinical practice, lesion measurement according to the PI-RADS guidelines is the daily workflow in most hospitals.