Multiparametric magnetic resonance imaging can exclude prostate cancer progression in patients on active surveillance: a retrospective cohort study

In this retrospective single-center cohort study, 147 consecutive patients on AS with initially diagnosed PCa at the University of Duesseldorf or at an outside institution received mp-MRI at 3 Tesla between October 2011 and September 2017 at our hospital. Of these patients, 55 were finally included who received a follow-up mp-MRI after a minimum interval of 12 months (median 19 months; IQR 13–33 months) with subsequent targeted MRI/US fusion-guided follow-up biopsy (FUS-GB) and concurrent systematic TRUS-GB after a median interval of 41 days (IQR 32–67 days) (Fig. 1). Inclusion criteria were histologically verified PCa with a GS of 3 + 3 = 6 or 3 + 4 = 7a, i.e., low- or intermediate-risk by D’Amico histological criteria [23], initial mp-MRI, and follow-up mp-MRI (≥ 12 months) with subsequent targeted MR-guided plus systematic 12-core TRUS-guided biopsy. The follow-up mp-MRI and subsequent follow-up biopsies were set to occur 12 to 16 months after the initial, diagnostic biopsy according to in-house standard of AS monitoring. Confirmatory biopsies were regularly scheduled for AS patients independent of the mp-MRI results but imaging is used to detect and locate index lesions of potentially higher GS tumors. Information and PCa localization from the histopathological reports of the initial biopsies were used to correlate index lesions in the initial mp-MRI. If the mp-MRI did not show a PCa lesion, a representative lesion in the region of the pathologically positive results was determined. Exclusion criteria were insufficient MR imaging (n = 1), no follow-up mp-MRI (n = 43), or no follow-up biopsy (n = 35). Of all patients, 13 received definitive curative treatment after GS upgrade in the interim before follow-up mp-MRI and were also excluded. Earlier biopsies compared with standard AS regime were triggered by clinical factors (increase of PSA/PSAD and/or aggravating symptoms). Retrospective assessment of visible radiologic progression was performed, comparing the initial and follow-up mp-MRI. Finally, results of the image analyses were compared with the follow-up biopsies to evaluate the ability of mp-MRI to predict tumor progression. The study was approved by the local ethics committee with a waiver of written informed consent.

Primary endpoint of the study was the negative predictive value (NPV) of the follow-up mp-MRI to exclude tumor progression. Secondary endpoints were positive predictive value (PPV) (1), sensitivity (2), specificity (3), PCa upgrades including GS distribution (4), and benefit of TRUS-GB in addition to FUS-GB (5).

All mp-MRI scans were conducted on 3T MRI scanners (Magnetom TIM Trio, Prisma or Skyra; Siemens Healthcare GmbH) using either 18-channel phased-array surface coil combined with 32-channel spine coil or a 60-channel phased-array surface coil (anterior and posterior part integrate 30 elements each). MR imaging parameters were chosen according to international recommendations [24] and contained T2-weighted sequences in 3 planes (T2WI; turbo spin echo, TSE; axial: voxel size 0.5 × 0.5 × 3.0 mm; FOV 130 mm), diffusion-weighted imaging (DWI; EPI and RESOLVE; voxel size 0.9–1.4 × 0.9–1.4 × 3.0 mm; b values 0, 500, 1000 s/mm² plus ≥ 1400 s/mm²), and dynamic contrast-enhanced imaging (DCE; T1 vibe; voxel size 0.8–1.5 × 0.8–1.5 × 3.0 mm, scan time 3 min, temporal resolution < 8 sec). Apparent diffusion coefficient (ADC) parameter maps were calculated by the scanner using the standard monoexponential model (including b0).

Patients received transrectal targeted follow-up FUS-GB and subsequent systematic 12-core TRUS-GB on a MRI/US fusion-guided biopsy system with elastic registration (Urostation, Koelis or UroNAV, Philips Healthcare) using an 18G fully automatic biopsy gun (Bard Medical). All biopsies were done by three experienced urologists (C.A., A.H., and D.M.) with 9, 8, and 5 years’ experience, respectively. Systematic TRUS-GB were conducted using a standardized biopsy plan which included lateral and midlobar cores at the base, middle, and apex of each prostate lobe. Gleason evaluation was performed by experienced uropathologists according to the recommendations of the International Society of Urological Pathology (ISUP) [25]. Histopathological cancer progression was defined as any increase in GS in any core in either TRUS-GB or FUS-GB.

Image interpretation was done by two radiologists with 5 and 10 years’ experience according to PI-RADS v2.1 in consensus. Prostate volume was measured by software volumetric (DynaCAD, Philips Healthcare) and PSA density (PSAD) was calculated by dividing PSA blood levels by prostate volume.

The analysis of serial mp-MRIs to assess radiologic tumor progression was performed and reported according to the Prostate Cancer Radiological Estimation of Change in Sequential Evaluation (PRECISE) recommendations by the European School of Oncology using a 5-point Likert scale as measure of likelihood of tumor progression on AS [21]. Imaging signs of progression were thus defined as a significant increase in size of an index lesion, measured in 3 planes in T2WI or DWI and/or increase in conspicuity of features suspicious for PCa according to PI-RADS v2.1 [26] in any sequence and/or newly detectable, suspicious lesions. Scores of 4 and 5 were eventually defined as a positive mp-MRI for tumor progression. Scores of 1–3 were classified as stable imaging appearance including if the follow-up MRI continuously did not show a visible PCa lesion. Additionally, the results (radiological progression vs no progression) of the retrospective image analysis using the PRECISE criteria were compared with the results of the original mp-MRI reports in which similar decision criteria had been used. For the retrospective image analysis, the readers were blinded to the biopsy results and to the results of the original mp-MRI reports. Lesion volume was measured in axial and sagittal T2-weighted images (height × width × depths).

Statistical analyses were performed using IBM SPSS® Statistics (version 21, IBM Deutschland GmbH). Data are expressed as mean ± SD and median + IQR. Patient demographic data were reported using descriptive statistics. Performance of the follow-up mp-MRI was assessed by determining PPV, NPV, sensitivity, and specificity relating to Gleason progression compared with subsequent biopsy. Exact binomial confidence limits were used for sensitivity, specificity, PPV, and NPV. Chi-square test was used to test for differences in proportions. Nonparametric data were tested with Mann-Whitney U test. Agreement between GS at TRUS-GB vs FUS-GB was evaluated with McNemar’s test of symmetry. Statistical significance was defined as p value < 0.05.

Of the 55 enrolled patients, 42 had histologically proven PCa with a GS of 3 + 3 = 6 and 13 had PCa with a GS of 3 + 4 = 7. In 32 of all men, the previous, initial, diagnostic biopsy was a combination of targeted MR-guided biopsy plus systematic TRUS-GB, and in the remaining 23 men, the previous, diagnostic biopsy was a systematic TRUS-GB only. The intervals between initial and repeat mp-MRI did not differ significantly for the respective subgroups (p = 0.67 and p = 0.71, respectively). The clinical and demographic characteristics at the time of initial mp-MRI and follow-up mp-MRI are summarized in Table 1.

Of 55 patients, 29 (53%) had histological tumor progression with a GS upgrade after confirmatory biopsy. Detailed Gleason distribution is shown in Table 2. Of 13 patients with initial GS of 3 + 4 = 7a, 5 (38%) had a GS upgrade. Of 42 patients with initial GS of 3 + 3 = 6, 24 (57%) had a GS upgrade. Of 23 patients with previous systematic TRUS-GB only, 14 (61%) had a GS upgrade compared with 15 (47%) patients with a GS upgrade in a group of 32 who had previous FUS-GB plus TRUS-GB. Differences in proportions of GS upgrades between the subgroups of initial GS 6 vs initial GS 7a and subgroups of previous TRUS-GB vs previous FUS-GB plus TRUS-GB, respectively, were not statistically significant (p = 0.24 and p = 0.31, respectively). The subgroup of patients with initial GS of 6 had more often received previous TRUS-GB only (45%) compared with the subgroup of patients with initial GS of 7a (31%); however, the difference was not statistically significant (p = 0.36). Figures 2 and 3 show examples of cases with stable histopathology and with histological progression, respectively. Lesion volumes of patients with GS upgrade were not significantly higher than lesion volumes of patients with stable GS (Mean ± SD (cm³) 0.76 ± 0.64 vs 0.72 ± 0.62, respectively; p = 0.40). The increase of the PSAD between the initial and the follow-up mp-MRI of patients with GS upgrade was not significantly higher than the PSAD increase of patients with stable GS (0.03 ng/ml/ml vs 0.03 ng/ml/ml; p = 0.80).

Overall, 44 men (80%) demonstrated a progression on mp-MRI, of whom 29 had a GS upgrade in the following biopsy. The remaining 11 men (20%) had no signs of progression on mp-MRI and none of them had histological tumor progression. The overall sensitivity and specificity of mp-MRI for histological progression were 100% (CI 0.88–1) and 42% (CI 0.26–0.61), respectively. The NPV and PPV were 100% (CI 0.74–1) and 66% (CI 0.51–0.78), respectively. Overall accuracy was 73% (CI 0.59–0.84). Table 3 illustrates the performance of mp-MRI in predicting GS progression in all patients and in subgroups depending on initial GS and prior biopsy method. Comparison of the results (progression vs no progression) of the follow-up mp-MRI and the original mp-MRI reports was in accordance in all cases.

FUS-GB detected PCa in 46 of 55 (84%) individuals and lead to a GS upgrade in 28 cases (51%) (Table 4). TRUS-GB alone detected significantly fewer PCa with 36 cases (65%; p = 0.007) and lead to significantly fewer GS upgrades with 12 cases (22%; p < 0.001). TRUS-GB detected one GS 4 + 5 = 9 PCa that was not detected in FUS-GB. The combination of both techniques detected 47 PCa (85%) and let to a GS upgrade in 29 cases (53%). The differences in PCa detection and GS upgrade between FUS-GB and the combined approach was not statistically significant (Table 5).