All investigators participated in study design and reviewed the final article.
The role of preoperative endorectal magnetic resonance imaging in the decision regarding whether to preserve or resect neurovascular bundles during radical retropubic prostatectomy
Article first published online: 11 MAY 2004
Copyright © 2004 American Cancer Society
Volume 100, Issue 12, pages 2655–2663, 15 June 2004
How to Cite
Hricak, H., Wang, L., Wei, D. C., Coakley, F. V., Akin, O., Reuter, V. E., Gonen, M., Kattan, M. W., Onyebuchi, C. N. and Scardino, P. T. (2004), The role of preoperative endorectal magnetic resonance imaging in the decision regarding whether to preserve or resect neurovascular bundles during radical retropubic prostatectomy. Cancer, 100: 2655–2663. doi: 10.1002/cncr.20319
- Issue published online: 2 JUN 2004
- Article first published online: 11 MAY 2004
- Manuscript Accepted: 31 MAR 2004
- Manuscript Revised: 26 MAR 2004
- Manuscript Received: 18 FEB 2004
- U.S. National Institutes of Health. Grant Number: R01 CA76423
- magnetic resonance imaging (MRI);
- endorectal magnetic resonance imaging (eMRI);
- prostate neoplasm;
- erectile dysfunction;
Because the recovery of erectile function and the avoidance of positive surgical margins are important but competing outcomes, the decision to preserve or resect a neurovascular bundle (NVB) during radical prostatectomy (RP) should be based on the most accurate information concerning the location and extent of the tumor. In the current study, the authors determined the incremental value of endorectal magnetic resonance imaging (eMRI) in making this decision.
eMRI was performed in 135 patients preoperatively. For each NVB, tumor extension to the NVB and the need for NVB resection was judged by a surgeon on a scale from 1 (definite preservation) to 5 (definite resection) before and after reviewing eMRI with a radiologist. Histopathologic findings were used as the standard of reference. The value of eMRI was assessed using binormal receiver operating characteristic (ROC) analysis adjusted for multiple observations per patient, and a mixed effects ordinal regression model was used for risk stratification.
Histopathologic examination determined that NVB resection was warranted in 44 of 270 NVBs (16%) because of posterolateral extracapsular extension (n = 29), positive surgical margins (n = 7), or both (n = 8). The areas under the ROC curves (AUC) were 0.741 for pre-MRI and 0.832 for post-MRI surgical planning (P < 0.01). MRI findings suggested altering the surgical plan in 39% of NVBs (106 of 270 NVBs). When the surgeon judged that the NVB resection was definitely not necessary (165 NVBs), MRI confirmed that decision in 138 NVBs (84%); the concordant decision was correct in 96% of the cases (133 of 138 NVBs). In 36 high-risk patients (≥ 75% probability of extracapsular extension), MRI findings changed the surgical plan for 28 NVBs (78%); the change was found to be appropriate in 26 cases (93%).
MRI was found to significantly improve the surgeon's decision to preserve or resect the NVB during radical prostatectomy. Cancer 2004. © 2004 American Cancer Society.
Radical retropubic prostatectomy (RP) provides excellent long-term disease control for patients with clinically localized prostate carcinoma.1 Erectile dysfunction is one of the possible complications after RP. Recovery of erectile function after RP is quantitatively related to the preservation of the neurovascular bundles (NVBs).2 However, prostate carcinoma most commonly arises in the peripheral zone of the prostate, often posteriorly, just beneath the capsule. Among patients with extracapsular extension (ECE), the tumor is most often present posterolaterally in the region of the NVBs.3–5 Therefore, if ECE is present in the area of the NVB, preservation of an NVB may compromise disease control and result in a positive surgical margin.4, 5 A positive surgical margin reportedly increases the probability of disease recurrence.1, 5 Consequently, surgeons often resect widely in areas of suspected ECE to avoid a positive surgical margin.4, 6, 7 In spite of all precautions, positive surgical margins are still present in approximately 23–31% of RP specimens.4, 7
The presence and location of ECE is difficult to predict from clinical stage, serum prostate-specific antigen (PSA) level, and the biopsy Gleason grade, even when quantitative assessment of the extent of tumor in systematic biopsies is considered.8 Previous studies of endorectal magnetic resonance imaging (eMRI) compared the MRI interpretations of independent readers with pathologic findings, including seminal vesicle invasion (SVI) and the presence of ECE overall and in the area of the NVB.9–14 To our knowledge, none of these studies has prospectively examined the impact of preoperative MRI on the surgical management decision to preserve or resect an NVB during RP. We undertook the current prospective study to determine the incremental value of preoperative eMRI on the decision to preserve or resect NVBs during RP, using step-section histopathologic findings as the standard of reference.
MATERIALS AND METHODS
This prospective, single-institution, cross-sectional study was conducted between March 2001 and February 2003. A total of 144 consecutive patients with biopsy-proven prostate carcinoma who were scheduled to undergo RP based on clinical assessment of the stage of the tumor were referred for eMRI by one of two attending urologists (P.T.S. and J.E.). All patients were scheduled to undergo RP before referral for MRI. Scheduling for MRI depended on the MRI schedule as well as the availability of the surgeon and radiologist because a joint prospective reading was required prior to surgery. Patients who had previously undergone hormonal therapy, radiation therapy, or chemotherapy were excluded from the final analysis. Patients were recruited as part of an ongoing National Institutes of Health-funded study of MRI for prostate carcinoma. Our institutional review board approved the study and informed consent was obtained from the patients. All patients underwent standard preoperative evaluation for prostate carcinoma, including history, physical examination, digital rectal examination (DRE), measurement of the serum PSA level, and sextant biopsy. Clinical stage was assigned using the 1992 International Union Against Cancer (UICC) TNM staging system.15
Of the 144 patients referred for eMRI, 135 fulfilled the study inclusion criteria. Nine patients were excluded from the final analysis because of presurgical hormonal treatment (six patients), radiation therapy (one patient), or combined chemotherapy and radiation therapy (two patients). Table 1 lists the clinical findings of the 135 patients included in the data analysis. The median age of the patients was 58 years (range, 40–73 years). The median PSA level was 5.78 ng/mL (mean, 6.17 ng/mL; range, 1.06–76.84 ng/mL). Sextant biopsy of the prostate was performed at the study institution (n = 28 patients) or at an outside institution (n = 107 patients). A single pathologist at the study institution center reviewed all biopsy specimens. The median interval from biopsy to MRI was 8 weeks (range, 1.4–84 weeks). The median interval between MRI and radical prostatectomy was 6.3 weeks (range, 0.1–29 weeks).
|Age (yrs)||PSA (ng/mL)||Percent probability organ-confined tumor based on the staging nomogram of Partin staging nomogram16, 17|
|< 4||4–10||> 10||≤ 25||> 25 & < 75||≥ 75|
|No. of patients (n = 135)||23||95||17||18||93||24|
|Median||58||2.99||5.78 (mean, 6.17)||12.71||22||63||77|
|B. Distribution of Clinical Stage (2002 TNM Classification) and Gleason Grade (Primary plus Secondary) on Each Side of the Prostate|
|Clinical stage (by side)||Gleason grade (by side)|
|T0||T1c||T2a||T2b||T2c||T3a||0||3 + 3||3 + 4||3 + 5||4 + 3||4 + 4||4 + 5|
|No. of patients||18||190||30||15||14||3||81||129||31||1||10||13||5|
MRI was performed on a 1.5-Tesla whole body MR scanner (Signa®; GE Medical Systems, Milwaukee, WI). Patients were examined in the supine position, using the body coil for excitation and a pelvic-phased array coil (GE Medical Systems) in combination with a commercially available balloon-covered expandable endorectal coil (Medrad Inc., Pittsburgh, PA) for signal reception. T1-weighted, axial spin-echo images (Repetition Time/Echo Time [TR/TE] = 700/8 milliseconds [msec], slice thickness of 5 mm, interslice gap of 1 mm, field of view [FOV] of 24 cm, matrix of 256 × 192, frequency direction transverse, 1 excitation) were obtained from the aortic bifurcation to the symphysis pubis, and T2-weighted axial and coronal fast spin-echo images (TR/effective/TE = 5000/96 msec, echo train length of 16, slice thickness of 3 mm, interslice gap of 0 mm, FOV of 14 cm, matrix of 256 × 192, frequency direction anteroposterior, 3 excitations) were obtained through the prostate and seminal vesicles. All MR images of each patient in the current study were reviewed by one of two independent readers (H.H. and F.V.C.).
MRI interpretation included analysis of tumor location and evaluation of tumor ECE in the region of the NVB (posterolateral). In addition, the image was evaluated for the presence of any ECE, SVI, or lymph node metastasis. Tumor was identified on T2-weighted imaging when there was a nodular, mass-like area of low signal intensity. The findings used to diagnose ECE on eMRI have been described previously and include: 1) an irregular bulge of the gland margin, 2) a contour deformity with a stepoff or angulated margin, 3) a breech of the capsule with direct tumor extension, 4) the obliteration of the rectoprostatic angle, and 5) asymmetry and/or tumor envelopment of the NVBs.10–14 Both readers used the same MR criteria. The MRI likelihood for ECE was scored on a scale of 1 to 5 as follows: 1: no ECE; 2: probably no ECE; 3: possible ECE; 4: probable ECE; and 5: definite ECE. Only a single reading was performed for each patient. The choice of the MRI reader was determined by the availability of the radiologist at the time of interpretation. Although the readers differed with regard to years of experience, both had read > 500 eMRIs of the prostate.
Determination of Pre-MRI and Post-MRI Surgical Plans
A data sheet was prepared for each patient by one of two independent coauthors (L.W. and D.C.W.) not involved in the image interpretation or clinical decision making. The data presented were DRE findings, serum PSA level, and sextant biopsy results (including Gleason primary and secondary grade, tumor location, percentage of positive biopsy cores, and the percentage of tumor-involved core tissue). Published Partin tables (2001 version)16 were used to estimate (based on the preoperative stage, Gleason grade, and PSA level) the probability that the tumor was histopathologically confined to the prostate gland (“organ-confined”).16, 17 Using all available clinical information, the attending urologist formulated a surgical plan for each NVB on a scale from 1 (definite preservation) to 5 (definite resection), designated the “pre-MRI” plan. The attending urologist then reviewed the MR images in consultation with an attending radiologist and formulated the “post-MRI” surgical plan on the same 5-point scale. Therefore, the post-MRI plan was based on a combination of the clinical and MRI findings. Intraoperative management was guided by the post-MRI plan, occasionally modified by intraoperative findings.
RP and Histologic Evaluation
RP was performed as previously described,18 with nerve-sparing (n = 223 patients), partial nerve resection (n = 7 patients), or complete nerve resection (n = 40 patients). Once the prostate was removed, the specimen was coated with India ink (the right and left sides were inked in 2 colors) and fixed in 10% buffered formaldehyde.19 The apical prostate was truncated and axial step-sections were obtained at 3–4-mm intervals in a plane perpendicular to the long axis of the prostate. The presence of tumor cells beyond the capsular margin defined ECE.
Standard of Reference
The standard of reference for the incremental value of MRI was based on histopathologic findings from the surgical specimen. Regardless of the actual surgery performed, the “optimal” (appropriate) surgical plan was determined based on histopathologic findings of ECE or positive surgical margins in the region of the NVB, posterolateral to the prostate. For each NVB, preservation was considered appropriate if the tumor did not extend outside the capsule in the posterolateral region of the prostate and the adjacent surgical margin was negative. NVB resection was considered appropriate if there was ECE or if there was a positive surgical margin in the posterolateral region of the prostate.
A total of 270 NVBs were evaluated in 135 patients. The effect of eMRI on NVB management was examined for the frequency and the appropriateness of change between the pre-MRI and post-MRI surgical plans. The surgical plan was recorded as more conservative or more aggressive according to whether it included NVB preservation or NVB resection, respectively. To compare the accuracy of the surgeon's pre-MRI and post-MRI judgments, we used a receiver operating characteristic (ROC) curve approach. Specifically, we estimated the parameters of the ROC curves using a latent variable binormal model. The model included a random effect to account for the two observations per patient, right and left NVBs, both pre-MRI and post-MRI. Model estimates were obtained using maximum likelihood and the accuracy of pre-MRI and post-MRI were compared using a likelihood ratio test.20
To investigate whether MRI provided a benefit for both high-risk and low-risk patients, the Partin nomogram score was used as a continuous variable and the nomogram-MRI interaction was used as covariate in the ordinal regression model. To better visualize the results of the risk-group analysis we created low-risk and high-risk groups. If a patient's chance of having organ-confined disease was > 75% as predicted by the published tables,16 he was classified as being at low risk; if the patient had a < 75% chance of having organ-confined disease, he was classified as high risk. To assess the management change, we furthermore dichotomized the pre-MRI and post-MRI ratings as negative (1, 2, or 3) or positive (4 or 5).
Because Partin staging nomograms are not universally used, we performed an additional analysis using a simplified, clinically defined, low-risk group (PSA level < 10 ng/ML, Gleason grade of 6, and clinical T1c or T2a disease). Furthermore, because the percentage of positive biopsy cores is often used as an indicator for NVB resection, correlation of the percentage of positive biopsy cores with the incremental value of MRI was performed as well.
Based on the histopathologic findings, NVB resection was considered the appropriate surgery for 44 of 270 NVB sides (16%) because of posterolateral ECE (n = 29), positive surgical margins (n = 7), or both (n = 8). Overall, 32 patients had ECE; in 7 of these patients (22%) the ECE was located only anteriorly or anterolaterally and away from the NVB and therefore did not warrant NVB resection.
The post-MRI surgical plan, including the decision to preserve or resect the NVB, agreed with the pre-MRI clinical surgical plan for 164 of 270 NVBs (61%). The surgeon judged that NVB resection was definitely not necessary in 165 NVBs and the eMRI confirmed that decision in 138 NVBs (84%). Of the 138 NVB sites in which the clinical decision concurred with the post-MRI decision not to resect an NVB, the decisions were correct for 133 of the sites (96%). MRI findings suggested a change in the surgical plan for 106 of 270 NVBs (39%). Thirty-nine NVBs had a more conservative surgical plan post-MRI; in 35 cases (90%), this change was appropriate. Overall, 67 NVBs had a more aggressive surgical plan post-MRI; in 45 cases (67%), this change was appropriate. Table 2 compares the pre-MRI and post-MRI surgical plans and shows the influence of MRI findings and its appropriateness. Two representative cases are illustrated in Figures 1 and 2. Table 3 lists the incremental effect of MRI stratified by the likelihood of extraprostatic disease from published staging tables.16, 17 For patients with a high probability of extraprostatic disease (probability of organ-confined disease of ≤ 25%), MRI changed the NVB management in 28 of 36 sites (78%); in 26 of these 28 sites (93%), the change was appropriate. In one patient, MRI suggested unilateral ECE with the involvement of NVB, prompting the more aggressive post-MRI surgical plan. At the time of surgery, no ECE was found in this patient, although there was extensive neural invasion within the prostate and a moderate-volume tumor (Fig. 3).
|Pre-MRI plan||Post-MRI plana|
|No change||More conservative||More aggressive|
|NVB||Score||No. of “NVB”||No. of NVB||Appropriate||No. of NVB||Appropriate||No. of NVB||Appropriate|
|270||164||157 (96%)||39||35 (90%)||67||45 (67%)|
|Percent probability of organ-confined tumor||No change post-MRI plan||More conservative post-MRI plan||More aggressive post-MRI plan||Appropriate post-MRI altered surgical plans|
|Probability||No. of NVB||No. of NVB||Appropriate||No. of NVB||Appropriate||No. of NVB||Appropriate||No. of NVB (%)|
|> 75%||47||38||37||0||0||9||6||6/9 (67%)|
|< 25%||36||8||8||12||10||16||16||26/28 (93%)|
ROC curves for pre-MRI and post-MRI surgical plans are plotted in Figure 4. The area under the curve (AUC) was 0.741 for pre-MRI and 0.832 for post-MRI. The difference between the two curves was found to be statistically significant (P < 0.01), suggesting that MRI substantially improves the accuracy of the surgeon's judgment with regard to the appropriateness of preserving or resecting NVBs.
In the ordinal regression model, MRI retained its significance (P < 0.01) and the nomogram also was found to be significant (P < 0.01). The nomogram-MRI interaction had a negative sign, suggesting that the benefit of MRI dissipates as the probability of having organ-confined disease increases, implying less benefit for low-risk patients, although this finding did not reach statistical significance (P = 0.08).
The analysis using a clinically defined low-risk group (PSA level < 10 ng/mL, a Gleason grade of 6, and clinical stage T1c or T2a disease) confirmed the Partin staging nomogram analysis.16, 17 and showed that, for low-risk patients, the incremental value of MRI in the decision to change the surgical approach is minimal (Table 4). Similarly, when the data were dichotomized for the percentage of positive biopsy cores at 25% (observed median), MRI was found to have a greater incremental value for the high-risk group of patients (Table 5).
|Risk category||Pre-MRI AUC||Post-MRI AUC|
|Category||Pre-MRI AUC||Post-MRI AUC|
Analysis of the management change (no resection [score of 1–3] to resection [score of 4–5]) in the post-MRI surgical plan is presented in Table 6. The management change was initiated in 32 NVBs, and was found to be appropriate in 29. Although the overall impact of MRI was found to be significant, the benefit was greatest in high-risk patients. In 24 high-risk patients, the pre-MRI and post-MRI plans disagreed; in 22 of these patients, the post-MRI classification was correct. These results indicate that adding MRI was useful for patients who had a low probability of organ-confined disease.
|Pre-MRI plan||Post-MRI plan||No. of NVBs||Appropriate|
|MRI affected surgical management||1, 2, 3||4, 5||15||14 (93%)|
|3, 4, 5||1, 2||17||15 (88%)|
|MRI change in score did not affect management||1||2||18|
The role of eMRI in the staging of and surgical planning for prostate carcinoma continues to provoke controversy.18, 21 A large multiinstitutional trial published in 1990 suggested that the accuracy of MRI for detecting ECE or SVI was low.22 However, in the last decade MRI has improved considerably with technologic refinements and increased reader experience.9 Among other technologic improvements, and pertinent to the evaluation of the prostate gland, was the introduction of eMRI. eMRI technology provides a more detailed anatomic image of the prostate compared with body coil MRI technology.10, 23 The MRI data obtained in the late 1980s used MRI technology that is now obsolete; therefore, the data are not representative of current state-of-the-art imaging capabilities. Furthermore, the 1990 multicenter study as well as most of the subsequent MRI studies focused on imaging pathologic correlation and MRI technology assessment.9–11, 21 We are unaware of any studies investigating the effect of preoperative eMRI on surgical management. However, the prognostic ability of preoperative MRI has been reported recently.21, 23–27 In a retrospective study of 1025 patients undergoing RP, T-stage classification by MRI was found to be an additional independent predictor of postoperative PSA failure (P < 0.001) for patients at intermediate and high risk.24 Although the study had a different endpoint from the current report, it is interesting to note that both studies suggest that MRI is more helpful in patients with unfavorable prognostic features.
The results of the current study suggest two important management benefits of eMRI prior to RP: improved surgical planning in high-risk patients and the provision of appropriate reassurance for preserving NVB in other patients. In the current study, we used the 2001 version of the Partin staging tables16, 17 for risk stratification because it is a validated predictive instrument that is widely used for patient counseling.28, 29 Patients with a probability of organ-confined tumor of ≤ 25% were considered to be high risk in the current study. MRI findings changed the surgical plan in 78% of these high-risk patients; in 93% of these cases, the changes were appropriate. Despite the high clinical probability of extraprostatic disease in these patients, MRI correctly favored NVB preservation in 10 of 12 NVBs (83%). In low-risk and intermediate-risk groups, MRI helped to provide appropriate reassurance in those cases in which NVB preservation was favored on the pre-MRI surgical plan. Overall, the strength of MRI for low-risk and intermediate-risk patients lies in a high negative predictive value (i.e., demonstration of the absence of tumor in the region of interest). In the high-risk group, MRI has incremental value in the clinical assessment additive to the Partin staging nomogram16, 17 a simplified clinical risk stratification (using PSA level, Gleason grade, and clinical disease stage), or the percentage of positive biopsy cores.
The current study has several limitations. First, the study design and results are based on personal consultation between experienced urologists and radiologists, all of whom were aware of the clinical findings, such as the location and percentage of positive biopsy cores and the percentage of tumor-involved core tissue. This multidisciplinary approach should optimize the interpretation of MRI findings and allow an optimal assessment of the incremental effect of MRI; however, it might not be representative of widespread daily clinical practice. Second, reflective of current surgical practice, a high percentage of patients had favorable pathologic results. With such a small prior probability, and therefore a small number of patients who required NVB resection, statistical validation of incremental change was difficult to ascertain and will require further analysis.
eMRI prior to RP was found to improve the validity of the surgical decision to preserve or resect the NVB. Thus, preoperative eMRI can provide additional, critical information that allows treatment to be adjusted to an optimal, patient-specific therapy.
The authors wish to thank Dr. James Eastham for patient referral.
- 9Critical analysis of the ability of the endorectal coil magnetic resonance imaging scan to predict pathologic stage, margin status, and postoperative prostate-specific antigen failure in patients with clinically organ-confined prostate cancer. J Clin Oncol. 1996; 14: 1770–1777., , , et al.
- 18Radical prostatectomy for clinical stage T1 and T2 prostate cancer. In: VogelzangNJ, ScardinoPT, ShipleyWU, CoffeyDS, editors. Comprehensive textbook of genitourinary oncology. 2nd edition. Philadelphia: Lippincott Williams and Wilkins, 2000: 722., :
- 20Generalized, linear, and mixed models. New York: John Wiley & Sons, 2001., .