Clinical utility of MRI in the decision-making process before radical prostatectomy: Systematic review and meta-analysis

Context Magnetic resonance imaging (MRI) is currently the most accurate imaging modality to assess local prostate cancer stage. Despite a growing body of evidence, incorporation of MRI images into decision-making process concerning surgical template of radical prostatectomy, is complex and still poorly understood. Objective We sought to determine the value of MRI in preoperative planning before radical prostatectomy. Materials and methods Systematic search through electronic PubMed, EMBASE, and Cochrane databases from 2000 up to April 2018 was performed. Only studies that used preoperative MRI in decision-making process regarding extension of resection in patients with prostate cancer, in whom radical prostatectomy was an initial form of treatment were included into analysis. Their quality was scored by Risk Of Bias In Non-Randomized Studies of Interventions system. Meta-analysis was performed to calculate the weighted summary proportion under the fixed or random effects model as appropriate and pooled effects were depicted on forest plots. Results The results showed that the preoperative MRI led to the modification of initial surgical template in one third of cases (35%). This occurred increasingly with the rising prostate cancer-risk category: 28%, 33%, 52% in low-, intermediate- and high-risk group, respectively. Modification of neurovascular bundle-sparing surgery based on MRI appeared to have no impact on the positive surgical margin rate. The decision based on MRI was correct on average in 77% of cases and differed across prostate cancer-risk categories: 63%, 75% and 91% in low-, intermediate- and high-risk group, accordingly. Conclusions In summary, MRI has a considerable impact on the decision-making process regarding the extent of resection during radical prostatectomy. Adaptation of MRI images by operating surgeons has at worst no significant impact on surgical margin status, however its ability to decrease the positive surgical margin rates remains unconfirmed.


Objective
We sought to determine the value of MRI in preoperative planning before radical prostatectomy.

Materials and methods
Systematic search through electronic PubMed, EMBASE, and Cochrane databases from 2000 up to April 2018 was performed. Only studies that used preoperative MRI in decisionmaking process regarding extension of resection in patients with prostate cancer, in whom radical prostatectomy was an initial form of treatment were included into analysis. Their quality was scored by Risk Of Bias In Non-Randomized Studies of Interventions system. Meta-analysis was performed to calculate the weighted summary proportion under the fixed or random effects model as appropriate and pooled effects were depicted on forest plots.

Results
The results showed that the preoperative MRI led to the modification of initial surgical template in one third of cases (35%). This occurred increasingly with the rising prostate cancerrisk category: 28%, 33%, 52% in low-, intermediate-and high-risk group, respectively. Modification of neurovascular bundle-sparing surgery based on MRI appeared to have no impact on the positive surgical margin rate. The decision based on MRI was correct on average in 77% of cases and differed across prostate cancer-risk categories: 63%, 75% and 91% in low-, intermediate-and high-risk group, accordingly. PLOS  Introduction to the Preferred Reported Items for Systematic Reviews and Meta-analysis guidelines (PRISMA, http://www.prisma-statement.org).

Data search
A systematic literature search was performed in following databases: MEDLINE via PubMed, Embase via Ovid and the Cochrane Database of Systematic Reviews. The search was restricted to publications in English, dating from January 2000 to April 2018. Over the past two decades there has been a huge improvement in prostate MRI as a result of a combination of high magnetic field strength and multiparametric imaging technique as well as advancements in the standardized interpretation of images. For these reasons we restricted our search form 2000 onwards [7]. Following search terms and their abbreviations were used in all databases: ("prostatic neoplasms" OR "prostate cancer" OR "prostate" OR "prostatic") AND ("magnetic resonance imaging" OR "multiparametric magnetic resonance imaging") AND ("prostatectomy" OR "radical prostatectomy" OR "laparoscopic radical prostatectomy" OR "endoscopic radical prostatectomy" OR "open radical prostatectomy" OR "robot-assissted laparoscopic prostatectomy"). The Mendeley Desktop version 1.17.9 (2008-2016 Mendeley Ltd.) citation manager was used to store records and remove duplicates. Reference lists of included papers and latest review articles were also hand-searched. To provide a completeness of this review, additional papers covering the subject and published during the current year were also searched. The primary research question for this systematic review is: how does the preoperative MRI influence a decision-making process before RP? Eligible studies were those with study cohort of men diagnosed with PCa confirmed on biopsy, evaluated preoperatively with MRI in whom RP was an initial form of treatment. Pathological examination of prostatectomy specimen with respect to PCa stage and postoperative margin status was the reference standard. Special attention was devoted to publications where the decision-making process regarding extension of resection based on MRI was explicitly described.
Three investigators (M.K. W.M. and W.M.) independently screened and assessed the eligibility of the articles based on their title and abstract. Any disagreements were resolved by unanimous decision and accepted by the senior author (J.D.). Subsequently full text articles were reviewed and included in the systematic review or excluded with certain reasons. The data extraction form was designed according to PRISMA (Fig 1). Due to heterogenous study designs and variety of assessed outcomes, extracted data appeared to be suitable for meta-analysis of 9 studies (Tables 1 and 2) [8,10,11,12,13,14,15,16,17]. The 115 studies not fulfilling the inclusion criteria were excluded for following reasons: not reporting the outcomes of interest; being focused only on PCa detection or staging; including experimental MRI protocol; using radiotherapy; using frozen-section as accessory tool influencing final decision; being duplicate reports from the same cohort; comprising cohort of less than 50 men. A total of 1552 men were included with a cohort size ranging from 75 to 353. R program (version 3.2.3, the R foundation for Statistical Computing, http://www.r-project. org) with meta and metafor packages was used to perform statistical analysis. Heterogeneity of the studies was assessed using I 2 statistics and in case of significant heterogeneity (I 2 > 50%), random effects model was favored over fixed effect model. Weighted summary proportions were calculated pooled effects were depicted on forest plots.

Risk of bias assessment
Nine studies were included in the final analysis. Methodologic quality was assessed using "Risk Of Bias In Non-randomized Studies of Interventions" (ROBINS-I) scoring system, which is a new tool for evaluating risk of bias [18]. ROBINS-I views each study as an attempt to simulate an ideal randomized trial, that is expected to answer a particular clinical problem. Seven domains are investigated for potential risk of introducing bias, that are judged with use of signaling questions.
Overall, the risk of bias was moderate in most papers, which is understandable as most studies were non-randomized and had the retrospective design, and as such are subject to confounding and a range of other biases (Fig 2). At the pre-intervention stage, bias due to confounding variables was mainly low, except of three cases, of which in one study serious bias was found because surgeons were not aware of MRI results. Moderate selection bias found in most of the included studies reflects the lack of randomization and control groups. Moreover, highly selective inclusion criteria were spotted in two papers where patients with evidence of EPE or very-high risk PCa were excluded from analysis. Only one study was found to have a serious risk of bias at intervention, because it investigated a presumptive influence of MRI, if it had been implemented. For the same reason this paper was assessed to be seriously biased due to non-adherence to intended intervention. Bias due to missing information about postoperative status of patients was not clearly stated in five studies. Outcomes of interest (NVB-sparing, the PSM rate, the appropriateness of decision) were reported incompletely in four papers, of which one had several of these shortcomings. Biased selection of reported results concerned three papers, in which important outcomes were reported incompletely, only for the part of cohort, which precluded the estimates from being used in our meta-analysis. To date, only one randomized clinical trial concerning the subject of this review was published, however it was assessed with the same tool to maintain coherence. The quality assessment was performed by two independent reviewers.

Evidence synthesis
Main characteristics of studies included into systematic review and meta-analysis are presented in Tables 1 and 2. Outcomes of interest extracted from eligible studies are listed in Table 3.

Impact of MRI on NVB-sparing surgery
Based on preoperative MRI the decision regarding NVB-sparing was changed in 35% (95%CI: 29-41%) of men (n = 525) in the summary of proportions in ten studies included into analysis (Fig 3). In most studies the decision tended to be modified in either of two ways: more aggressive resection (n = 331) or more preserving NVB-sparing (n = 194). In pooled analysis more aggressive resection was chosen in 21% (95%CI: 16-27%) of men (Fig 4), whereas more preserving NVB-sparing was preferred in 16% (95%CI: 13-20%) of patients ( Fig 5). Moreover, in two studies MRI altered the surgical plan exclusively in the direction of more radical excision [8,13].

Differential influence of MRI imaging across PCa risk groups
The rates of patients in whom the surgical plan regarding NVB was modified after MRI varied significantly among PCa-risk groups. Only 4 studies provided sufficient data with this respect.

PSM rate and decision-making process
Most studies adopted positive surgical margins (PSM) as an adequate oncologic end point indicating the appropriateness of decision based on MRI in individual patients. In total PSM   occurred in 11% (95%CI: 8-16%) of men (n = 161; Fig 9). There was no significant difference between groups irrespective of whether the decision was modified (PSM = 14%, 95%CI: 6-29%) based on MRI or remained unchanged (PSM = 15%, 95%CI: 9-23%). Of note, in only 3 studies authors provided results allowing to draw this conclusion [8,12,17].

Appropriateness of change in surgical plan adaptation
In the worst case scenario MRI may misguide a surgeon to unnecessarily remove NVB when dealing with truly organ confined lesion, or conversely, unclear imaging may prompt NVBsparing in case of truly advanced tumor and lead to PSM. This appropriateness of surgical plan adaptation based on MRI was assessed in 4 studies [10,13,14,16], of which two included subgroup analysis [10,14]. When MRI added no additional information and a surgical template remained unchanged, the decision was correct in 93% (95%CI: 84-97%) of cases (Fig 10). If the surgical template set before MRI was eventually modified after imaging, the appropriateness was 77% (95%CI: 72-81%) (Fig 11). The appropriateness varied among PCa risk groups and amounted to 63% (95%CI: 46-77%) in low-, 75% (95%CI: 60-86%) in inter-, and 91% (95%CI: 80-96%) in high-risk PCa group.

Discussion
Our systematic review provides updated summary of studies focused on the influence of MRI on the decision-making process prior to RP. The results indicate that the preoperative MRI leads to the modification of initial surgical template in one third of cases. The phenomenon occurs increasingly with the rising PCa risk category: from 28% in low-risk group, through 33% in intermediate-risk group, to 52% in high-risk PCa group. Modification of NVB-sparing surgery based on MRI appeared to have no impact on the PSM rate. The decision based on MRI is appropriate on average in 77% of cases and differs across various PCa risk categories.

MRI modifies surgical template with respect to NVBs
At present, MRI is considered the most accurate imaging modality in detecting extraprostatic extension (EPE) and provides opportunity for a surgeon to adapt the template for prostate tumor resection [19]. Our meta-analysis has shown preoperative MRI to prompt a change in the decision regarding NVB-sparing in one third (35%) of men underscoring a major role of imaging in decision-making process prior to RP. Intuitively, from the oncological perspective the tendency towards more aggressive resection would be more common than the opposite approach of more preserving surgery. Indeed, across number of studies, the former attitude was observed in 21% of patients, whereas the latter one in 16% of cases. However the rates of adjustments fluctuated across the analysed papers. It was dependent on number of factors including various criteria for nerve-sparing surgery, different MRI specifications (field strengths, MRI sequences, optional endorectal coil, time of MRI) and surgical approaches (open, laparoscopic, robotic) as well (Tables 1 and 2). Not surprisingly, if preoperative MRI indicates the presence of EPE, one would remove the adjacent bundle regardless of other clinical variables. Lee et al have shown that MRI suggestive of EPE was significantly associated with the excision of NVB [20], and those with concern for extracapsular extension (ECE) on MRI had lower rates of NVB-sparing at least on affected side [20]. At the same time, sensitivities and specificities of MRI in the detection of EPE range from 41% to 92% and from 65% to 100%, respectively ( Table 2). These diversities may reflect various accommodation of MRI images in different studies. In the trial by Durskin et al, NVB-sparing rates between MRI and non-MRI group were similar despite suggestion of non-focal EPE in pelvic MRI in selected cases [21]. Therefore, despite fairly high diagnostic accuracy these observations indicate, that MRI before RP is being adopted with great caution. Of note, conservative approach may harm PCa patients being unnecessarily disqualified from the NVBsparing surgery. Retrospective analysis revealed, that one in four patients (26%) who had their bundles resected because of high-risk disease, could probably have one spared due to MRI that suggested organ-confined disease [15].
Discrepancies in the detection of EPE may also be caused by the lack of standardization in MRI reading. In only three studies included in our systematic review the EPE probability scale was used [12,14,15]. To date, several attempts were made to facilitate standardized PCa staging, with PI-RADS and Likert scale as the most commonly used systems. Despite promising preliminary results [22], the systematic assessment of EPE by points did not gain broader acceptance and was later abandoned in the reissued PI-RADS version 2 recommendations, in which the likelihood of EPE was assessed in a descriptive manner giving a binary result [23]. Even though a structured MRI interpretation is done, results may vary considerably between readers [24,25] and heavily depend on readers experience [26]. These observations highlight the importance of standardization when defining explicit EPE criteria and optimal preoperative planning [27]. Our meta-analysis implies, that incorporation of MRI images into surgical templates is appropriate in 77% of cases. In other words, in around a quarter of patients the decision based on MRI was wrong and resulted in PSM or unnecessary bundle removal. The higher appropriateness of the surgical template, when it remained unchanged (93%), may be explained by that, these were the cases, where MRI findings did not modify initial surgical decision, because it was straightforward due to other clinical variables. Imaging played virtually only a confirmatory role in these cases. For example, it could be the case of low risk PCa without any suspicious lesion visible in MRI, where both NVBs might be preserved. On the contrary, bilateral broad resection would be favored in locally advanced high risk PCa irrespective of the imaging. On the other hand, in more difficult cases, where MRI modified the initial surgical approach, the decision was still made with fairly high accuracy (77%). Unfortunately, the majority of analyzed papers do not provide specific percentages of cases, when MRI led to incorrect decision. However, in the only two studies addressing this issue, MRI-based NVB-sparing procedure was appropriate in most cases (91% and 92%), whereas more aggressive surgery due to MRI turned out to be correct in two-thirds of patients (63% and 60%) [14,16]. These observations indicate misguided tendency towards underestimating the MRI result, when it is negative for EPE, which leads to fewer NVB-preservations.
Lastly, the role of MRI as a prognostic marker has been acknowledged. It has been shown that radiologic findings may serve as a surrogate of adverse pathology after RP or even predict biochemical recurrence and therefore guide towards the optimal choice of cancer treatment [28]. MRI, by visualizing intraprostatic lesions, identify men, who are likely to have clinically significant disease and eventually benefit from radical management [29]. At the same time we have learned from the number of studies including recently published ProtecT trial that PCa specific mortality is very low in low risk PCa irrespective of the mode of therapy. However, among those who prefer active monitoring greater risk of disease progression and metastases has to be taken under consideration [2]. Therefore, men with low risk PCa otherwise eligible for active surveillance may benefit from imaging, which in case of a suspicious result would prompt targeted biopsy and radical treatment. Indeed, abnormal MRI is associated with the greater risk of PCa progression in active surveillance cohorts [30]. The probability of undergrading and understaging is significantly greater in those in whom MRI reveals abnormal lesions when compared to those in whom imaging is normal [29]. In the studies included into our research mean preoperative PSA and Gleason score in biopsy were low, which might suggests that studied cohorts largely embraced men with low-risk disease, yet the rates of locally advanced PCa was substantial (range: 9-43%). Therefore, our results confirm the observation that preoperative assessment of clinical stage without imaging is prone to underestimation [31].

MRI has no negative impact on the rate of PSM
Incorporation of MRI images into prostatectomy surgical template followed by more restrictive surgery may raise concerns with respect to the status of surgical margins. Ideally adjustment of surgical confines would lead to decrease in the likelihood of PSM with simultaneous increase in NVB preservation. Our meta-analysis has shown that in those who had preoperative MRI rate of PSMs is no different irrespectively of the direction of surgical technique adaptation (Fig 9). In matched control study, despite the difference in crude numbers, PSM rates in pelvic MRI and non-MRI groups were similar (13.7% vs 19.3%) [21]. This relation was also observed at subgroup analysis in patients with T1c PCa (11.0% vs 18.1%) and more advanced disease as well (21.4% vs 25.0%) [21]. In other retrospective study MRI before surgery was not associated with improvement in the rate of PSM [32]. In the only one randomized trial devoted to the role of preoperative MRI, despite advanced imaging overall risk of PSM was not improved (23% in non-MRI vs 19% in MRI group). Subgroup analysis revealed however possible benefit of imaging in patients with T1c cancers only, in whom the relative reduction of PSM by 41% was observed [8]. At the same time, Brown et al. have shown clear trend towards greater risk of PSM in MRI organ confined disease that was found to be locally advanced after prostatectomy. The rate of PSM in this group mounted to 54% [33]. Furthermore, in the lowest risk group category (T1c, GS 6) the corresponding rate reached 80% if understaging based on MRI was encountered [33]. From the other site, multivariable analysis demonstrated that ECE suspected on MRI had neither protective effect nor increased risk for PSM [34]. It seems that erroneous MRI revealing organ confined disease poses a risk of PSM, therefore the accuracy of imaging in local staging should not be overestimated. However, if ECE is suspected, the bundles are to be removed anyway. To conclude, implementation of images as addition to other clinical data into preoperative planning needs to be viewed with caution and the risk of underestimation of real PCa stage with subsequent risk of PSM even in the realm of so advanced technology has to be taken under consideration. Of note, the location of PSMs usually corresponded to the site of the index lesion visualized by MRI, but in one-third of cases PSM was observed in the remote region of the specimen [21]. This finding indicates that, beyond reader's experience, surgeons' expertise is also of utmost importance to obtain satisfactory oncologic outcome. The existence of learning curve has been proven for different types of radical prostatectomy and it is reasonable to assume that the minimal number of cases is needed to become proficient. A large multicenter study indicated that after reaching the caseload over 200-250 LRPs the reduction of the PSM rate achieves a plateau [35]. Considering NVB-sparing surgery, the rate of PSM in operated organ-confined tumors is proposed as a credible way to assess outcomes [36]. The precise number of prostatectomies performed by each surgeon were provided in three studies included in our review [8,15,16], precluding meaningful conclusion regarding this issue.

PCa risk groups and preoperative MRI
Current guidelines do not recommend mpMRI in low-risk patients for local staging, because of insufficient sensitivity in focal EPE detection, unless it is used as a decision tool to select patients for nerve-sparing procedure [1]. The recommendation underscores the need for selecting the most beneficial group of patients, in whom the MRI would influence the decision making-process with respect to the extent of surgery and most desirably reduce the PSM rate. Our meta-analysis indicates, that indeed the relevance of MRI increases with the PCa risk category, yet it is substantial even in the low-risk group.
Some early studies suggested MRI had no incremental value over standard staging approach [10], and low-risk patients would not benefit from preoperative imaging [37]. However, when combined with low prevalence of EPE in the low-risk subgroup, negative predictive value of MRI in staging is high and facilitate right selection of candidates for nerve sparing surgery [9]. It may be explained by low sensitivity of MRI in the detection of EPE [14,17,,37]. EPE is rather unusual phenomenon in low-risk disease, and if it is encountered, it has focal, microscopic character being beyond the scope of MRI resolution [14]. In RCT closely investigating the influence of MRI on the PSM rate, the modification of initial surgical plan was expressed the most in cT3 disease (83%), however it was the cT1 PCa where more radical excision was chosen in 51% of cases and it was the only subgroup, among which MRI reduced significantly the PSM rate [8]. When referring to the current standards in choosing cases for NVB-sparing, the mpMRI is able to modify the extent of surgery in almost every second man (45%) or even more in the low-risk subgroup (63%) [17]. These numbers imply that MRI, in spite of its low sensitivity in detecting EPE in the low-risk PCa group, provides reassurance, that the decision regarding NVBs was correct [14]. However, the level of its appropriateness in this group is at least moderate (63%). It probably results from overstaging and subsequent unnecessary NVBresection. One may only speculate and weigh the burden of microscopic, focal PSM in MRI understaged PCa against superfluous bundle resection in overstaged disease.

Limitations
Our study has some limitations. First, most studies included in meta-analysis had retrospective design and as such were subjected to selection bias and were prone to data loss. This is reflected by the moderate risk of bias of the majority of included papers and the heterogeneity of the results in some outcomes of interest. Nevertheless, we believe it resembles current practice of MRI use in decision-making process, which is not yet standardized. Second, only two studies included a control group, which makes impossible to credibly answer, whether MRI brings a benefit in terms of lowering the PSM rate. This issue needs further well-designed studies to be resolved. Third, the number of included studies is limited and some of the subanalyses are based on the low number of patients, therefore the results should be treated with caution. Fourth, studies varied considerably in terms of inclusion criteria for NVB-sparing surgery and interpretation of MRI images. Standardized MRI reading, which becomes the standard of practice was lacking in the majority of studies. As a result, these findings may not be straightforwardly reproduced in other centers.

Conclusions
In summary, our meta-analysis showed that MRI exerts significant influence on preoperative planning of the extent of resection during RP. This effect may be spotted among different PCa-risk groups. Modification of the NVB-preservation based on the MRI result seems not to influence the PSM rate. Initial findings regarding influence of imaging on oncological outcomes need to be assessed in further studies.
Supporting information S1