Multiparametric magnetic resonance imaging underestimates extraprostatic extension in higher risk tumors
Stephen Schmit, BS1, Sai Allu, BA1, Borivoj Golijanin, BS2, Joshua Ray Tanzer, PhD3, Gyan Pareek, MD1, Elias Hyams, MD1.
1The Warren Alpert Medical School of Brown University, Providence, RI, USA, 2The Minimally Invasive Urology Institute at The Miriam Hospital, Providence, RI, USA, 3Rhode Island Hospital Biostatistics Core, Providence, RI, USA.
BACKGROUND: Multiparametric magnetic resonance imaging (mpMRI) is increasingly used for risk stratification and pre-operative staging of prostate cancer. Concern for extraprostatic extension (EPE) is a key radiographic feature that suggests wider dissection may be appropriate. Prior studies have demonstrated less reliable mpMRI findings in higher risk tumors (i.e. by D'Amico risk criteria). It remains unclear how Grade Group (GG) alone and more comprehensive multivariable risk assessments may interact with the ability of mpMRI to determine the presence of EPE on surgical pathology.
METHODS: A retrospective review of a robotic assisted laparoscopic radical prostatectomy (RALP) database from a single academic institution from 2016-2020 was performed. Radiology mpMRI reports were assessed for positive or possible EPE findings and compared with surgical pathology reports. The data were stratified by GG and a multivariable cluster analysis was performed to incorporate the following variables: age at diagnosis, family history of prostate cancer, BMI, prostate volume estimated by mpMRI, PSA, number of positive biopsy cores, Gleason score, and preoperative grade group based on biopsy. Furthermore, risk ratios were calculated to determine how mpMRI findings and radiographic EPE relate to positive surgical margins.
RESULTS: A total of 297 patients underwent at least one mpMRI prior to RALP. Preoperative mpMRI demonstrated a sensitivity of 39.2% and specificity of 89.3% for pathological EPE, and had a negative predictive value (NPV) of 50.2%, and positive predictive value (PPV) of 84.1%. The PPV of GG 5 tumors increased to 96.0% but NPV decreased to 12.0% (Table 1). On multivariable analysis, a cluster of patients with the most positive biopsy cores (8.64; 95% CI: 8.09-9.23), demonstrated significantly lower NPV relative to the other clusters (Table 2). Additionally, positive EPE on preoperative mpMRI was associated with a significantly decreased risk of positive surgical margins (RR: 0.655; 95% CI: 0.557-0.771).
CONCLUSIONS: Because negative mpMRI findings for EPE have low NPV in GG5 tumors and a cluster of patients with high volume disease, urologists should exercise caution in aggressive nerve sparing approaches in these patients. Findings of EPE on mpMRI seem to favorably influence margin status, likely by prompting a wider ipsilateral dissection.
| Biopsy Grade Group | n | EPE Concern on MRI | EPE Prevalence on Pathology | Sensitivity | Specificity | Negative Predictive Value | Positive Predictive Value |
| GG 1-5 | 297/297 (100%) | 82/297 (27.6%) | 176/297 (59.3%) | 39.2% | 89.3% | 50.2% | 84.1% |
| GG 3-5 | 143/297 (48%) | 48/143 (33.6%) | 109/143 (76.2%) | 40.4% | 88.2% | 31.6% | 91.7% |
| GG 4-5 | 69/297 (23%) | 32/68 (46.4%) | 61/69 (88.4%) | 49.2% | 75.0% | 16.2% | 93.8% |
| GG 5 | 50/297 (17%) | 25/50 (50%) | 46/50 (92%) | 52.2% | 75.0% | 12.0% | 96.0% |
Table 1: mpMRI detection of EPE by increasing tumor grade.
-
| Trait | Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 |
| Demographics | ||||
| Age at Diagnosis | 62 [60-63] | 57 [54-60] | 61 [60-62] | 62 [61-63] |
| Family History of Prostate Cancer | 100% [100%-100%] | 0% [0%-0%] | 0% [0%-0%] | 15% [9%-23%] |
| BMI | 28 [27-29] | 35 [29-42] | 28 [27-29] | 28 [27-29] |
| Disease State | ||||
| Preoperative Prostate Volume | 44 [36-55] | 48 [20-231] | 53 [44-65] | 41 [34-51] |
| PSA | 8 [7-10] | 7 [3-35] | 8 [6-9] | 10 [8-13] |
| Number of Positive Cores | 5.04 [4.59-5.54] | 7.00 [4.56-10.74] | 3.43 [3.09-3.81] | 8.64 [8.09-9.23] |
| Gleason Score | 7.74 [7.55-7.93] | 8.00 [6.87-9.13] | 7.28 [7.13-7.43] | 7.45 [7.29-7.6] |
| Grade Group | 2.49 [2.37-2.6] | 2.67 [2.01-3.32] | 2.22 [2.09-2.34] | 2.29 [2.18-2.4] |
| Group Frequencies | 86 | 3 | 102 | 103 |
| Predictive Value of MRI and Pathology Reports | ||||
| MRI % Positive | 23% [14%-33%] | - | 16% [10%-25%] | 37% [27%-47%] |
| Pathologist % positive | 65% [53%-75%] | - | 48% [38%-59%] | 71% [61%-79%] |
| Sensitivity | 38% [26%-51%] | - | 27% [17%-41%] | 49% [39%-60%] |
| Specificity | 91% [80%-96%] | - | 93% [86%-97%] | 79% [65%-88%] |
| Positive predictive value | 87% [73%-94%] | - | 75% [56%-88%] | 86% [76%-92%] |
| Negative predictive value | 48% [36%-59%] | - | 63% [53%-72%] | 37% [26%-49%] |
Table 2: Clustering analysis of preoperative mpMRI accuracy in the detection of EPE.
Back to 2023 Abstracts