Researchers evaluate the role of magnetic resonance imaging in the clinical staging of prostate cancer in the definitive and salvage settings, and to identify changes in treatment recommendations.
About the lead author:
Jonathan D. Tward, MD, PhD
Huntsman Cancer Institute at the University of Utah
Salt Lake City, UT
Kevin Stephans, MD
Director of Genitourinary
Why is this article contemporary?
Clinical staging of prostate cancer by digital rectal examination and/or computed tomography scan findings is limited by the incomplete ability to accurately assess disease distribution, volume, and presence of adverse staging findings such as extra-capsular extension (ECE) or seminal vesicle invasion (SVI), which are far more often detected on final surgical pathology. Prostate magnetic resonance imaging (MRI) offers a significant advance in the availability of potential valuable clinical data prior to making a treatment decision.
The challenges in incorporating MRI findings into treatment decision making algorithms include: one, sometimes disparate reports of the sensitivity and specificity of MRI findings for predicting the true extent of disease as reported by surgical pathology; two, understanding the subtleties of MRI versus pathological findings as they relate to prognosis (should MRI find gross but not microscopic ECE or SVI, perhaps MRI findings could represent the more severe cases of these pathological predictors); three, defining the role of MRI in the pre-treatment decision making model; and four, proving that gains in decision making provided by MRI do in fact improve outcomes and are cost-effective.
Answers to each of these questions are needed prior to the routine inclusion of MRI in to the evaluation and decision making algorithm for all patients with prostate cancer. Surgical pathology investigations continue to better define the answers to the first two challenges. Here, the authors present one of the first efforts to address the far more complicated third and fourth question, and begin an investigation that hopefully will be continued and clarified over the next decade.
Background and Purpose
To evaluate the role of magnetic resonance imaging (MRI) in the clinical staging of prostate cancer in the definitive and salvage settings, and to identify changes in treatment recommendations.
Materials and Methods
Between November 2008 and November 2011, 114 patients referred for radiotherapy (RT) consultation underwent a prostate-protocol MRI of the prostate and pelvis. Charts were retrospectively reviewed for demographic and clinical information.
Eighty-six patients were evaluated for definitive treatment, 26 for salvage treatment, and 2 for RT planning in the adjuvant setting. Of the definitive subjects, MRI was performed after RT consultation in 68 patients and before consultation in 18 patients. For patients who underwent an MRI after consultation, MRI led to a change in treatment recommendation for 6 of 68 (9%) patients. Androgen deprivation therapy was added or extended in 3, prophylactic nodal RT was added to 3, and definitive nodal RT was added to 1. Active surveillance was no longer recommended in two persons, and brachytherapy alone was no longer an option in two persons. Of the 86 definitive treatment patients, nodal metastases were identified in 6 (7%). For the 26 salvage patients, imaging did not lead to any changes in planned treatment. MRI evaluation can influence clinical decision making for both physician and patient, but how those decisions change outcome is still unresolved.
In most cases when treatment recommendations were modified based on MRI findings, therapy was escalated to correspond with clinical upstaging. It remains to be seen if this translates into improved disease control, and it is possible that the “Will Rogers phenomenon” (when moving an element from one set to another raises the average values of both) will result from incorporating the high-resolution MRI information into clinical decision making.
Historically, magnetic resonance imaging (MRI) has not been routinely used for the diagnosis and staging of prostate cancer (PC). Accurate staging is important because there is wide variability in recommended treatment options and expected outcomes for patients with low-risk, intermediate-risk, and high-risk PC.1 Although useful for risk stratification, prostate-specific antigen (PSA) measurements can be unreliable correlates of tumor stage and disease burden.2 In addition, while digital rectal examination (DRE) can detect high-risk features such as extracapsular extension (ECE), DRE has been shown to have a low sensitivity.3,4 MRI has several possible roles in the management of PC, and its utility is being investigated in numerous single-institution studies. MRI findings have been correlated to pathologic features after radical prostatectomy.5-7 MRI is being evaluated as a tool for selecting patients who are appropriate active surveillance (AS) candidates, 8 despite some evidence that it may not be useful at predicting adverse pathologic features,9 or for predicting risk of recurrence based on high-risk features.10-14 A better understanding of tumor extent and location within the gland may lead to improvements in therapeutic decision making and in surgical and radiotherapy (RT) treatment planning. Use of prostate MRI could also lead to delivery of more aggressive therapies because of the detection of adverse features that would otherwise remain occult. We reviewed the use of prostate MRI to characterize its effect on the clinical stage and subsequent treatment recommendations for patients referred for RT consultation.After Institutional Review Board approval was obtained, all patients who were evaluated in the radiation oncology department and underwent a prostate protocol MRI between November 2008 and November 2011 were identified. In total, 114 patients met these criteria. Charts were reviewed for age at diagnosis, PSA level, date of biopsy, Gleason score, and pathologic or DRE-based T-stage. The clinical American Joint Committee on Cancer (AJCC 7th edition) Tumour, Node, Metastases (TNM) stage and the National Comprehensive Cancer Network (NCCN) risk groups were documented. Prostate protocol MRI reports were reviewed, and an MRI stage was assigned by the authors based on the extent of disease identified on imaging. Treatment recommendations, per dictated reports, were given at the time of initial RT consultation and again after MRI information was obtained. In situations where the MRI was obtained before the RT consultation, only one set of recommendations was made.MR imaging studies were performed on a whole body 3 Tesla Siemens Magnetom TrioTim or Verio (Siemens Medical Solutions, Erlangen, Germany) with an 8-channel pelvic phased-array coil and spine coil. Routine techniques and sequences were used, including small field of view (FOV) T2-weighted turbo spin-echo sequences in the axial, sagittal, and coronal directions, covering the prostate and seminal vesicles (SVs). Scanning parameters were TR range/TE range, 3000—3300/101-121; slice thickness 3-4 mm; an echo train length of 17-19 and 1-2 averages.
Each plane took 5 to 6 minutes to acquire with a FOV of 160 x 160 mm. A 256 x 256 pixel matrix was used. An axial T1 small FOV series was performed to detect postbiopsy hemorrhage. Whole-pelvis imaging T1 and T2 with fat saturation was performed to look for lymph node (LN) and osseous involvement. Axial whole-pelvis diffusion-weighted imaging was performed with b values of 0 and 1000. All studies were evaluated by one of five abdominal fellowship- trained radiologists with two to 15 years of experience. Imaging was evaluated as part of usual clinical care on a PACS image viewer (iSite, Philips Healthcare, Andover, MA) (Figure 1), with full access to any previous imaging and medical records, including PSA level and biopsy results. No additional review was performed for this study.
Clinical interpretation was focused on providing information for local staging, including an assessment of the prostate capsule, SVs, and pelvic nodes. Criteria used to classify local disease included a focal T2 hypointense mass in the peripheral zone without corresponding T1 hyperintense signal, to suggest postbiopsy hemorrhage. ECE included evaluation on the T2-weighted images for direct extension into the periprostatic fat or an interruption of the T2 dark band that surrounds the prostate, or broad contact of the low-signal intensity focus of disease with the capsule (≥10 mm).Eighty-six patients were evaluated for definitive treatment, 26 for salvage treatment, and 2 patients were evaluated for treatment planning in the adjuvant setting. Patient and tumor characteristics are described in Table 1. For the 86 patients with newly diagnosed PC, AJCC TNM stages before and after MRI acquisition are displayed in Table 2. The most common clinical stage was T1cN0M0 (50%). MRI revealed that 17 (19.8%) had findings concerning for ECE, and 13 (15.1%) had SV invasion (SVI). Of the patients with ECE diagnosed on MRI, the clinical stage was formerly T1c in 7 cases, T2a in 3, T2b in 4, T2c in 1, and T3a in 2. Of the patients with SVI on MRI, the clinical stage was formerly T1c in 4 cases, T2a in 1, T2b in 1, T2c in 3, T3a in 2, and T3b in 2. In 29 cases (34%), MRI findings agreed with the clinical examination findings. In 54 (63%) patients, MRI findings suggested more advanced disease than DRE, and in 3 (3%) cases the MRI findings were less concerning than DRE. Of the 54 who were clinically upstaged, 33 (61%) had clinical T1c disease prior to MRI. Of these, 14 had imaging findings reported as involving both lobes of the prostate gland, 7 had findings concerning for ECE, and 4 had SVI on MRI. Of the 3 who were downstaged, in two cases of clinical T2a disease, the tumor could not be detected on MRI, and in one case of clinical T3a disease, ECE could not be identified on MRI. MRI reported newly found nodal metastases in 6 (7%) patients, and pelvic LNs were “concerning” in an additional two. Distant metastases (DM) were not identified in any patient. MRI was performed after RT consultation in 68 patients and before consultation in 18 patients.
For patients who underwent an MRI after consultation, MRI led to a change in treatment recommendation for 6 of 68 (8.8%) (Table 3). Two patients were initially considered for brachytherapy monotherapy but after MRI they underwent surgery or combined modality treatment with external beam radiation therapy (EBRT) and androgen deprivation therapy (ADT). Two patients were considered for AS, but after MRI underwent definitive treatment. In four patients, EBRT to the prostate and SVs was recommended at the time of diagnosis, but after MRI, the field was extended to include the pelvic LNs. In three cases, ADT was not considered at diagnosis but was recommended after consideration of MRI findings. For the 26 patients evaluated in the salvage setting, the mean time from definitive surgical treatment to MRI was 4.4 years (range 0.5 to 21 years). Retained SVs were identified in seven (26.9%) patients. For the 28 patients evaluated in the salvage and adjuvant settings, no nodal or DM were identified, and imaging did not lead to changes in planned treatment.MRI use for prostate cancer staging, decision making, and treatment planning is supported by most payers and treatment guidelines,1 and has both proponents and opponents. The mantra of “Don’t order a test unless you will action the result” is a refrain most medical students learn early in their training. This study was undertaken specifically to evaluate how the diagnostic MRI altered clinical decision making, despite the absence of level 1 evidence that proves or disproves its merit in changing oncologic outcomes. There are no universally adopted guidelines to inform the clinician on how MRI results should affect patient care. At our institution, in the definitive treatment setting it is our bias that MRI findings of SVI, grossly evident extracapsular extension, lymphadenopathy, or very-high-volume disease should be considered a very-high-risk feature. In these cases, it is our bias that patients should have pelvic lymph nodal radiation, dose-escalation with brachytherapy if possible, as well as ADT as part of their definitive treatment. With the exception of lymphadenopathy, CT scans perform poorly at discriminating these particular features. Conversely, in low-risk and favorable intermediate-risk patients, we consider very-lowvolume or no evidence of disease on MRI to be a favorable prognostic feature that may support the use of monotherapies or AS when taken in context with other clinical parameters. Reports of “capsular abutment concerning for extracapsular extension” are generally considered nonactionable.
Irrespective of how these findings affect the choice of therapy delivered, these authors feel that properly performed MRI sequences can greatly assist in the contour definition of the prostate and surrounding anatomy for both radiation oncologist and surgeon to assist in high-quality treatment planning. In this series of patients, most were found to have more advanced disease on MRI than what DRE implied. Bilateral gland involvement, ECE, and SVI were identified on MRI more often than DRE detected rates of T2c, T3a, and T3b disease. This is consistent with published reports, such as that by Fuchsjager et al,12 who evaluated 224 patients who underwent an endorectal MRI before EBRT for PC. On MRI, bilateral gland involvement, ECE, and SVI were identified in 24.6%, 40.2%, and 7.1%, respectively, compared to 0%, 7.6%, and 5.4% with DRE-based T2c, T3a, and T3b disease, respectively. Numerous studies have demonstrated a correlation between these MRI findings and biochemical failure after RP or EBRT.10,12,13 A study of 977 men demonstrated that PSA failure was predicted by preoperative PSA, MRI T-stage, percentage of positive prostate biopsies, Gleason score, and 1992 AJCC clinical stage in men undergoing RP.10 Riaz et al13 reported on 279 men with intermediate-high risk PC who underwent an endorectal MRI prior to combination EBRT and brachytherapy. ECE on MRI and Gleason score were the only independent predictors of recurrence. Fuchsjager et al found that pretreatment PSA and ECE on MRI were significant predictors of PSA relapse.12 ECE and SVI on MRI predict for poor outcome, and it is therefore appropriate that these features should be considered when treatment recommendations are made.
Use of pelvic MRI for identifying LN involvement is more controversial than the role of MRI in diagnosing locally advanced disease. A meta-analysis of series evaluating the diagnostic accuracy of MRI in diagnosing LN metastases revealed that the sensitivity and specificity of MRI were 0.39 and 0.83, respectively, compared to 0.42 and 0.82, respectively, for CT imaging.15 MRI is limited in detecting LN metastases because the imaging criteria in routine practice depend on a size threshold of ≥10 mm in short axis for elongated nodes or ≥8 mm in short axis for rounded nodes suggesting malignancy. Nodes with malignant deposits often measure less than these thresholds, accounting for the low sensitivity of conventional MRI in malignant LN detection.16
While many studies have analyzed the prognostic significance of the high-risk features detected on MRI, it is more challenging to understand how frequently and significantly this information changes treatment recommendations. In our series of 86 patients with newly diagnosed PC, MRI findings had an impact on treatment decision making in six (8.8%) definitively treated patients. Usually, this resulted in an escalation of therapy to address a higher clinical T-stage than what was previously thought. Whether or not this results in a change in outcome remains to be seen. At best, the practice of changing the treatment recommendation based on the imaging findings may lead to more cures; at worst, these men may suffer additional treatment-related morbidity and cost with no therapeutic gain. Improvements in diagnostic imaging techniques will lead to improved resolution of tumor burden on MRI than can be appreciated on most physical examinations. Hence, it is possible that the “Will Roger’s phenomenon,” which describes the apparent improvement in average outcomes in two cohorts when moving elements of one into another, will be seen by incorporating the high-resolution information that MRI affords. That is, if patients previously thought to be low-risk were now reclassified into the highrisk category with DRE occult disease, both the low- and high-risk groups may demonstrate improved outcome over historical controls. A similar phenomenon was noted when Gleason pattern interpretation shifted over time.17 Therefore, the outcomes of patients diagnosed and treated using MRI information cannot be readily compared to historic controls. One barrier to incorporating MRI findings with clinical stage is the current challenge in diagnosing ECE on MRI. A change in stage from T1/T2 to T3 disease can lead to an increase in RT field size (prostate/SVs vs pelvic field), a change in duration of ADT (none vs months to years), and omission of alternative treatment options such as brachytherapy monotherapy or AS. The detection and interpretation of ECE can be challenging. The diagnostic criteria for ECE include concrete findings such as periprosatic fat infiltration with or without direct extension into the neurovascular bundle, and softer features, including areas of capsular bulging, obliteration of the rectoprostatic angle, and capsular contact of tumor extending ≥10 mm.5,17 A recent study of 151 patients in Korea,18 which compared local staging and detection of ECE at 3.0 T with and without endorectal coils, found a sensitivity of 33.3% versus 31.3%, specificity of 96.6% versus 97.5%, and accuracy of 63.% versus 61.4%, respectively. In no case was the difference statistically significant. The high specificity argues that the prostate MRI protocol is very accurate for detecting disease when it is present, which may account for the change in management decisions that is attributed to the MRI. Recent studies have demonstrated that sensitivity of MRI in local staging improves in poorly differentiated PC.19 This again suggests that the MRI is useful in detecting patients who are at risk of clinical understaging, without increasing the likelihood of overtreatment.
Contrary to patients with newly diagnosed PC, in our limited series of patients evaluated for adjuvant or salvage RT, MRI did not add clinical information that altered our treatment approach. There were no cases of LN involvement or DM, no marginal local recurrences, and the instances of retained SVs seen would have likely been detected on planning CT simulation. In a recent publication by Liauw, endorectal MRI identified a radiographic local recurrence in 21 of 88 (24%) men with clinically undetectable, post-RP biochemical recurrence with a mean pretreatment PSA of 0.30 ng/mL.20 Retained SVs were identified in 14% and LN involvement in 5%. Other institutions have demonstrated that MRI allows for detection of locally recurrent disease in most cases. In a series by Sella et al,5,21 soft tissue local recurrences as identified on MRI were considered confirmed if a positive biopsy was obtained and if the PSA decreased after RT or if the mass progressed on follow-up imaging. However, in the Sella et al series, the indication for MR imaging was not given, and it is possible that imaging was obtained to confirm the presence of a mass or nodule identified on physical examination. In addition, the authors did not state if a retained SV in the setting of an elevated PSA was considered a local recurrence. The discrepancy between these series and ours could lie in the study methodology. In the work published by Sella et al and Liauw et al, MRI images were retrospectively reviewed for study purposes. This has the potential to bias the reading radiologists and lead to a greater documentation of locally recurrent disease. In our study, MRI results were obtained retrospectively from the initial radiologist’s diagnostic report, and therefore the local recurrence detection rate more accurately represents average rates in practice. Limitations of this study include its retrospective nature. Identification of high-risk features on MRI could lead to delivery of more aggressive treatment, yet this may not have been explicitly stated in the dictated notes. In this case, we may be underestimating the effect of MRI on treatment decision making. Change in treatment recommendations based on MRI findings is a subjective measure; however, it is important to recognize the possible impact of MRI on cancer staging and consider how this may affect treatment outcomes. Use of MRI to detect PC in the definitive and salvage settings is significantly different, and each of these topics warrants review. Additionally, the clinical routine prostate MRI protocol at our institution is performed at 3.0 T without an endorectal coil. The literature has been inconclusive as to whether 1.5 T imaging with an endorectal coil is clearly superior to 3.0 T without an endorectal coil when a multiparametric approach is used.5,22This study demonstrates how advanced imaging in PC detection can influence treatment planning and delivery. In most cases, when treatment recommendations were modified based on MRI findings, therapy was escalated to correspond with clinical upstaging. It remains to be seen if this translates into improved disease control. It is possible that the “Will Rogers phenomenon” will result from incorporating the high-resolution MRI information into clinical decision making.
University of Utah, Department of Radiation Oncology, (JT, BT, KK, DS) University of Utah, Department of Radiology (MH) University of Utah, Department of Surgery, Division of Urology (WL) Address correspondence to: Jonathan Tward, University of Utah, Department of Radiation Oncology, 1950 Circle of Hope, Room 1570, Salt Lake City, UT 84112. Tel: 801-581-2396 Fax: 801-585-2666 Email: firstname.lastname@example.org
Address correspondence to: Jonathan Tward, University of Utah, Department of Radiation Oncology, 1950 Circle of Hope, Room 1570, Salt Lake City, UT 84112. Tel: 801-581-2396 Fax: 801-585-2666 Email: email@example.com
Conflicts of Interest: Jonathan D. Tward consulted for Myriad Genetics, provided expert testimony for Garlington, Lohn & Robinson, and received an HCI Directors translational research initiative grant. Dennis C. Shrieve provided expert testimony for a law firm and receives book royalties from Lippincott. Marta E. Heilbrun has a grant from RSNA/ NIBIB and receives book royalties from Amirsys publishing. For the remaining authors, no conflicts of interest were declared.