- Email: [email protected]
Combined clinical characteristics and multiparametric MRI parameters for prediction of cribriform morphology in intermediate-risk prostate cancer patients
ARTICLE IN PRESS
Urologic Oncology: Seminars and Original Investigations 000 (2019) 1−9
Original Article
Combined clinical characteristics and multiparametric MRI parameters for prediction of cribriform morphology in intermediate-risk prostate cancer patients Jie Gaoa,1, Qing Zhanga,1, Yao Fub,1, Wei Wanga, Chengwei Zhanga, Yanshen Kana, Haifeng Huanga, Danyan Lic, Jiong Shib, Hongqian Guoa,*, Bing Zhangc,* a
Department of Urology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Institute of Urology Nanjing University, Nanjing, Jiangsu, China b Department of Pathology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing, Jiangsu, China c Department of Radiology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing, Jiangsu, China Received 22 June 2019; received in revised form 20 August 2019; accepted 4 September 2019
Abstract Purpose: To develop a risk model with combined clinical characteristics and multiparametric MRI parameters for prediction of cribriform morphology in intermediate-risk prostate cancer (CaP) patients. Methods: We retrospectively included 215 CaP patients received multiparametric MRIexamination, targeted biopsy (TB) combined with systematic biopsy (SB), radical prostatectomy and final Gleason group 2 or 3. Cribriform status was confirmed on both biopsy slices and wholemount sections. Characteristics were stratified by cribriform status. Mann Whitney U test was performed for continuous variables and the x2 test for categorical variables. Univariate and multivariate logistic regression analyses were performed for significant predictors, followed by cribriform-risk nomogram construction. Receiver operating characteristic analysis was used for internal discrimination validation with corresponding area under the curve. Calibration curves were plotted and decision curve analysis was performed for clinical benefit exploration. Results: Cribriform morphology was identified in 51.2% (110/215) patients. Cribriform-positive CaP demonstrated significantly higher prostate-specific antigen level, higher prostate-specific antigen density , larger lesion dimension on MRI, higher Prostate Imaging Reporting and Data System score, larger tumor dimension, higher Gleason score, higher pT stage, higher pN stage and more positive surgical margin (all P < 0.01). Sensitivities of TB, SB, TB + SB for detecting cribriform morphology were 28.2% (31/110), 22.7% (25/110), and 36.4% (40/ 110), respectively. Further, prostate-specific antigen density (P = 0.003), Prostate Imaging Reporting and Data System score (P < 0.001), and maximal biopsy Gleason score (P = 0.004) were independent predictors for positive cribriform morphology. Cribriform-risk nomogram was constructed with the 3 parameters and demonstrated considerable discrimination (area under the curve = 0.887, sensitivity 79.2%, specificity 84.0%) and calibration (mean absolute error 0.021), harboring net benefits with threshold probabilities range from 0 to 0.88. Conclusion: A cribriform-risk nomogram was developed and well predicted aggressive cribriform morphology in intermediate-risk CaP patients. Ó 2019 Elsevier Inc. All rights reserved.
Keywords: Prostate cancer; Cribriform morphology; Nomogram; Multiparametric MRI; Biopsy
1. Introduction Prostate cancer (CaP), the most common cancer type and a major cause of cancer mortality for men in Europe and the 1
These authors contributed equally to this work. *Corresponding authors. Tel.: 862583106666; fax:862568182863. E-mail addresses: [email protected] (H. Guo), [email protected] (B. Zhang). https://doi.org/10.1016/j.urolonc.2019.09.002 1078-1439/Ó 2019 Elsevier Inc. All rights reserved.
United States, is a histologically heterogeneous and frequently multifocal disease [1,2]. The original scoring system described by Gleason in 1966 for grading CaP on the basis of histological morphology was later found to accurately predict cancer-specific death [3]. Cribriform morphology, a subtype of Gleason pattern 4, is recognized as a more aggressive and often more lethal architecture than noncribriform morphologies of Gleason pattern (GP) 4 [4,5]. Over the past few years, many studies have emerged
ARTICLE IN PRESS 2
J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
demonstrating that cribriform-positive CaP is associated with increased risk of lymph node invasion [6], distant metastasis [7], biochemical recurrence [8] and cancer-specific death [9]. Hence, correct identification of the cribriform status within CaP is crucial for clinical decision. Currently, the diagnosis of cribriform morphology is most commonly made after transrectal ultrasonography (TRUS)-guided prostate biopsy. Besides, prostate multiparametric MRI (mpMRI) is gaining popularity for the detection of clinically significant CaP (csCaP), and MRI/ US fusion targeted biopsy (TB) of suspicious mpMRI lesions improve the detection of csCaP by 30% [10]. Despite advances in biopsy techniques, the inherent sampling limitations of needle biopsy are still problematic for accurate grading of CaP. Matthew T et al. revealed that MRI/US fusion TB alone, systematic biopsy (SB) and the combined approach (TB + SB) demonstrated low sensitivity and specificity for cribriform morphology detection using radical prostatectomy (RP) as the reference standard [11]. Thus, additional techniques or models for diagnosing cribriform morphology is necessary prior of treatment. Generally, cribriform-positive CaP harbored relatively larger tumor size compared with cribriform-negative CaP (predominantly poorly formed or fused glands) [12,13], together with larger lesion dimension on mpMRI [13]. With Prostate Imaging Reporting and Data System v2 (PIRADS v2), the most current mpMRI scoring system, PIRADS score 5 was significantly associated with positive cribriform morphology on biopsy pathology [13]. This implies that cribriform morphology is more likely to be present in tumors greater than 1.5 cm or those with an invasive appearance on mpMRI. Given the potential differences between tumors of different cribriform status, we hypothesized the feasibility constructing a diagnostic model with
combined clinical characteristics and mpMRI parameters for prediction of aggressive cribriform morphology in CaP, which has never been studied as of yet. In this study, we aimed to identify significant predictors for cribriform morphology and further construct a risk model for prediction of aggressive cribriform-positive CaP with whole-mount sections of RP. 2. Materials and methods 2.1. Patients Between August 2016 and December 2018, data of patients underwent RP at our single center were retrospectively collected. The inclusion criteria were as follows: (a) preoperative mpMRI within 3 months from surgery; (b) mpMRI/US-guided TB and transperineal SB were performed and patients were pathologically diagnosed as prostate adenocarcinoma; (c) all biopsy slices and whole-mount sections were available; (d) final Gleason score (GS) 3 + 4 and 4 + 3; (e) no neoadjuvant therapy before surgery (such as androgen deprivation therapy and transurethral resection of the prostate. Such patients harbored incomplete or unrecognizable pathology, which might disturb the correlation between clinic-radiological parameters and actual pathology). Initially, 856 patients underwent RP were included, and specific selection flowchart is shown in Fig. 1. Finally, a total of 215 patients were enrolled. The institutional review board approval was obtained, while all patients were informed and signed consent. 2.2. MRI examination and interpretation Prostate mpMRI examinations were routinely performed and reported in consensus by 2 radiologists according to the
Fig. 1. Study flowchart of patient selection. ADT = androgen deprivation therapy; mpMRI = multiparametric magnetic resonance imaging; RP = radical prostatectomy; TURP = transurethral resection of the prostate.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
3
PI-RADS v2 standard [14]. Specific descriptions were available in supplementary files.
21.0 (IBM, USA). All tests were 2-sided, with statistical significance set at P < 0.05.
2.3. Biopsy protocol
3. Results
For lesions suspicious of CaP (PI-RADS score ≥3) on mpMRI, TB was performed as recommended by PI-RADS v2 [14]. All biopsies were conducted in transperineal approach with MRI-US image registration system (Esaote and RVS) that provides real-time fusion of TRUS images and mpMR images as previously described [15]. Briefly, lesions identified on mpMRI were semiautomatically displayed on the real-time TRUS image. The biopsy started with TB to the center of suspicious lesions using free-hand transperineal technique. Afterwards, standard 12-core transperineal SB was carried out.
Characteristics of the total 215 patients are summarized in Table 1. After pathologic slices evaluation, these patients were divided into cribriform-positive CaP (110/215, 51.2%) and cribriform-negative CaP (105/215, 48.8%) according to the cribriform status on whole-mount sections. Further, characteristics stratified by cribriform status were compared (Table 1). Patients with positive cribriform morphology demonstrated significantly higher prostate-specific antigen (PSA) level (P = 0.001), higher prostate-specific antigen density (PSAD) (P < 0.001), larger lesion dimension on MRI (P < 0.001), higher PI-RADS score (P < 0.001), more positive biopsy cores (TB + SB) (P < 0.001), larger tumor dimension on final pathology (P < 0.001), higher GS (P < 0.001), higher pT stage (P = 0.001), higher pN stage (P = 0.019) and more positive surgical margin (P = 0.004). No significant differences were found for age and prostate volume between 2 groups (P = 0.145 and P = 0.093, respectively). Further, using whole-mount pathology as gold standard, we evaluated the sensitivities of different biopsy techniques for detecting any GP 4, noncribriform GP 4 and cribriform GP 4. As shown in Fig. 2A, 167 out of 215 (77.7%) patients of GS > 3 + 3 harbored biopsy GS > 3 + 3, with 18 patients (10.8%) demonstrating biopsy GS > 3 + 3 only in TB cores, 39 patients (23.4%) demonstrating biopsy GS > 3 + 3 only in SB cores, and 110 patients (65.8%) demonstrating biopsy GS > 3 + 3 in both TB cores and SB cores. Thus, sensitivities of TB, SB, TB + SB for detecting GP 4 were 59.5% (128/215), 69.3% (149/215) and 77.7% (167/215), respectively. As for 105 noncribriform CaP patients, 8 patients demonstrating GP 4 only in TB cores, 22 patients demonstrating GP 4 only in SB cores, and 38 patients demonstrating GP 4 in both TB cores and SB cores. Thus, sensitivities of TB, SB, TB + SB for detecting noncribriform GP 4 were 43.8% (46/105), 57.1% (60/105) and 64.8% (68/105), respectively. In terms of 110 cribriform-positive patients, 36.4% (40/110) harbored cribriform morphology on biopsy pathology, while 63.6% (70/110) were missed (Fig. 2B). Representative cases illustrating presence and absence of cribriform morphology on biopsy slices were shown in Figs. 3 and 4, respectively. Among the 40 positive cases, 15 (37.5%) demonstrated cribriform morphology only in TB cores, 9 (22.5%) demonstrated cribriform morphology only in SB cores, and 16 (40.0%) demonstrated cribriform morphology in both TB cores and SB cores. Thus, sensitivities of TB, SB, TB + SB for detecting cribriform morphology were 28.2% (31/110), 22.7% (25/110) and 36.4% (40/110), respectively. Given the relatively low sensitivities of different techniques for detecting cribriform morphology (Fig. 2B) and the differences of clinic-radiological characteristics between
2.4. Biopsy and whole-mount histopathology Biopsy specimens were routinely disposed and reported by 2 urological pathologists. After prostatectomy, intact prostate specimens were sliced from apex to base at 3-mm intervals, following by paraffin embedding and hematoxylin-eosin stain. All whole-mount slices were reviewed in consensus by 2 urological pathologists according to the 2014 International Society of Urological Pathology (ISUP) modified criteria of CaP [4]. Each lesion was outlined and assigned a GS. Notably, cribriform morphology was specifically identified. For patients with positive cribriform morphology on whole-mount sections, corresponding biopsy slices including both TB cores and SB cores were reviewed separately. Any presence of cribriform morphology on biopsy sections was recorded with the specific core number. 2.5. Statistical analysis Mann-Whitney U test was performed for continuous variables and the x2 test for categorical variables. Univariate and multivariate logistic regression analyses were performed for significant parameters to predict the presence of cribriform morphology on final pathology. Significant predictors were used for nomogram construction with R software version 3.5.2 (https://www.r-project.org/) using the “rms” package. Receiver operating characteristic (ROC) analysis for the internal discrimination validation was done by bootstrapping with 1,000 iterations using the “pROC” package. Corresponding area under the curve (AUC) with 95% confidence interval, sensitivity and specificity for differentiation were derived. The extent of over- or underestimation of predicted probabilities relative to observed probabilities of cribriform-positive CaP was explored graphically by calibration plot using the “rms” package, which was internally validated using bootstrapping with 1,000 iterations. Decision curve analysis (DCA) was performed using the “rmda” package as previously described [16]. Statistical analysis was performed with software SPSS
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
4
Table 1 Characteristics of 215 patients and stratified by cribriform status of CaP. Characteristic
Age (y) PSA level (ng/ml) Prostate volume (ml) PSAD (ng/ml/ml) Lesion dimension on MRI (cm) PI-RADS score 3 4 5 Positive biopsy cores (TB + SB) Maximal biopsy GS (TB + SB) 3+3 3+4 4+3 4+4 Tumor dimension on final pathology (cm) Gleason score (RP) 3+4 4+3 pT stage T2 T3a T3b pN stage N0 N1 Surgical margin Positive Negative
n = 215
Cribriform morphology Positive (n = 110)
Negative (n = 105)
69 (64−74) 10.9 (7.3−17.9) 29.9 (22.5−41.7) 0.39 (0.23−0.64) 1.5 (1.1−2.2)
71 (63−75) 13.2 (8.6−21.5) 28.4 (22.5−38.8) 0.48 (0.29−0.77) 2.0 (1.5−2.5)
68 (64−73) 9.3 (6.7−15.5) 33.1 (22.9−43.9) 0.30 (0.20−0.51) 1.2 (1.0−1.5)
23 (10.7) 86 (40.0) 106 (49.3) 3 (5−7)
4 (3.6) 27 (24.6) 79 (71.8) 6 (4−8)
19 (18.1) 59 (56.2) 27 (25.7) 4 (2−6)
44 (20.5) 50 (23.3) 63 (29.3) 58 (26.9) 1.8 (1.4−2.5)
7 (6.4) 22 (20.0) 38 (34.5) 43 (39.1) 2.3 (1.7−2.9)
37 (35.2) 28 (26.7) 25 (23.8) 15 (14.3) 1.5 (1.1−2.1)
127 (59.1) 88 (40.9)
44 (40.0) 66 (60.0)
83 (79.0) 22 (20.9)
99 (46.0) 96 (44.7) 20 (9.3)
39 (35.5) 55 (50.0) 16 (14.5)
60 (57.1) 41 (39.1) 4 (3.8)
205 (95.3) 10 (4.7)
101 (91.8) 9 (8.2)
104 (99.0) 1 (1.0)
58 (27.0) 157 (73.0)
39 (35.5) 71 (64.5)
19 (18.1) 86 (81.9)
P
0.145 0.001 0.093 <0.001 <0.001 <0.001
<0.001 <0.001
<0.001 <0.001
0.001
0.019y
0.004
Continuous variables are presented as median (interquartile range, IQR), while categorical variables are presented as patients (%). Fisher’s exact test was done for P value calculation. GS = Gleason score; MRI = magnetic resonance imaging; PI-RADS = Prostate Imaging Reporting and Data System; PSA = prostate-specific antigen; PSAD = prostate- specific antigen density; RP = radical prostatectomy; SB = systematic biopsy; TB = targeted biopsy.
y
different cribriform statuses (Table 1), we tried to find significant predictors for cribriform morphology. With univariate and multivariate logistic regression model analyses, PSAD (odds ratio 1.97, 95% confidence interval 1.36
−2.85, P = 0.003), PI-RADS score (odds ratio 10.89, 95%CI 4.02−29.51, P < 0.001) and maximal biopsy GS (P = 0.004) and were identified as independent predictors of cribriform morphology within CaP (Table 2).
Fig. 2. Sensitivities of different biopsy techniques for detecting any GP 4, noncribriform GP 4 and cribriform GP 4. Biopsy slices of TB cores and SB cores were separately evaluated to identify any GP 4, noncribriform GP 4 and cribriform GP 4. Further, corresponding sensitivities of different biopsy techniques for detecting any GP 4, noncribriform GP 4 and cribriform GP 4 were derived. GP = Gleason pattern; SB = systematic biopsy; TB = targeted biopsy.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
5
Fig. 3. Representative radio-pathological matching case of cribriform-positive CaP with positive cribriform morphology on biopsy pathology. Patient summary: a 71-year-old man with CaP in right peripheral zone, PSA level 15.38 ng/ml, prostate volume 29.1 ml; PSAD 0.53 ng/ml/ml; lesion dimension on MRI 1.5cm; PI-RADS score 5, positive biopsy cores 4 out of 14, final Gleason score 4 + 3, tumor dimension 1.7 cm. (A) axial T2-weighted MRI; (B) axial diffusion-weighted MRI; (C) axial ADC map; (D) representative biopsy pathology with positive cribriform morphology; (E) whole-mount section of radical prostatectomy specimen; (F) representative cribriform morphology derived from whole-mount section. Lesion was outlined with solid line on corresponding MR images and with dashed line on whole-mount section, respectively. ADC = apparent diffusion coefficient; CaP = prostate cancer; MRI = magnetic resonance imaging; PI-RADS = prostate imaging reporting and data system; PSA = prostate-specific antigen; PSAD = prostate-specific antigen density.
Further, a cribriform-risk nomogram incorporating these 3 predictors was constructed (Fig. 5A). The nomogram was internally validated by bootstrapping. ROC analysis was used for the nomogram discrimination, with an AUC of 0.887 (95% CI 0.8440.930) (Fig. 5B). Corresponding sensitivity and specificity were 79.2% and 84.0%, respectively. Moreover, bootstrapped calibration curve of the nomogram demonstrated that there was no untoward deviation of predicted risk from observed risk of cribriform morphology, with a mean absolute error of 0.021 (Fig. 5C). Further, the bootstrapped DCA was performed for clinical benefit exploration (Fig. 6A). With threshold probabilities within the range from 0 to 0.88, the nomogram adds more net benefits than taking intervention to either all or no patients. DCA was also performed in subset patients for whom GS ≤ 3 + 4 and > 3 + 4 were identified at biopsy (Supplementary Figure 1). 4. Discussion To our knowledge, this is the first study to construct a novel risk model with combined clinical characteristics and mpMRI parameters for prediction of aggressive cribriform morphology in CaP. Regarding the cribriform-risk
nomogram, corresponding sensitivity and specificity for cribriform status were 79.2% and 84.0%, together with considerable calibration and net benefits under specific threshold probabilities. Cribriform morphology, which was described by Gleason as the architecture having “moderately differentiated glands, range from small to large, growing in spaced-out infiltrative patterns” [17], has drawn much attention recently. Dong et al. revealed that the presence of cribriform architecture was an independent predictor for biochemical recurrence as well as metastasis after RP [18]. In 2014, the ISUP conference reached a consensus that cribriform gland should be assigned Gleason pattern 4, regardless of morphology [4]. Since the 2014 ISUP meeting, a rapidly growing body of evidence indicated the aggressive behavior of cribriform morphology as described above [5−9]. In our cohort, cribriform-positive CaP was significantly associated with increased PSA level, PSAD and tumor dimension. Moreover, cribriform morphology was more frequently present in carcinoma with relatively higher GS, T stage and N stage, consistent with a recent study of RP cohort [19]. As we know, mpMRI is currently the best imaging method to noninvasively identify and characterize CaP [14]. With PI-RADS v2, mpMRI demonstrated good
ARTICLE IN PRESS 6
J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
Fig. 4. Representative radio-pathological matching case of cribriform-positive CaP with negative cribriform morphology on biopsy pathology. Patient summary: a 60-year-old man with CaP in left peripheral zone, PSA level 5.84 ng/ml, prostate volume 26.6 ml; PSAD 0.22 ng/ml/ml; lesion dimension on MRI 0.9 cm; PI-RADS score 4, positive biopsy cores 3 out of 14, final Gleason score 4 + 3, tumor dimension 1.3 cm. (A) axial T2-weighted MRI; (B) axial diffusion-weighted MRI; (C) axial ADC map; (D) representative biopsy pathology with negative cribriform morphology; (E) whole-mount section of radical prostatectomy specimen; (F) representative cribriform morphology derived from whole-mount section. Lesion was outlined with solid line on corresponding MR images and with dashed line on whole-mount section, respectively. ADC = apparent diffusion coefficient; CaP = prostate cancer; MRI = magnetic resonance imaging; PI-RADS = Prostate Imaging Reporting and Data System; PSA = prostate-specific antigen; PSAD = prostate-specific antigen density.
performance for detecting CaP, with pooled sensitivity of 89% and specificity of 73%, respectively [20]. MRI/US fusion TB of suspicious mpMRI lesions improved the detection of csCaP by 30% [10]. As for csCaP (GS > 3 + 3), TB combined with SB demonstrated a satisfying sensitivity of 77.7% for detecting GP 4 in our cohort. However, compared to TB, SB showed better sensitivity for detecting
GP 4 in our cohort. The experiences of radiologists at their institution or quality of MR registration might account for the difference. Additionally, TB + SB showed acceptable sensitivity (64.8%) for noncribriform GP 4 detection. However, for cribriform GP 4, several biopsy techniques including TB, SB, TB + SB demonstrated poor performance on cribriform morphology identification, with sensitivities of
Table 2 Univariate and multivariate logistic regression analyses for the prediction of cribriform-positive CaP. Variable and intercept
Age (per 5 yr) PSA level (interquartile OR) Prostate volume (interquartile OR) PSAD (interquartile OR) Lesion dimension on MRI PI-RADS (5 vs. 3/4) Maximal biopsy GS 3 + 4 vs. 3 + 3 4 + 3 vs. 3 + 3 4 + 4 vs. 3 + 3 Positive biopsy cores (interquartile OR)
Univariate logistic regression
Multivariate logistic regression
OR (95% CI)
P
1.15 (0.96−1.39) 1.80 (1.45−2.24) 0.85 (0.69−1.04) 2.23 (1.77−2.81) 4.46 (2.74−7.27) 15.83 (8.04−31.17)
0.133 <0.001 0.112 <0.001 <0.001 <0.001 <0.001
5.79 (1.95−17.2) 8.96 (3.13−25.67) 12.2 (4.17−33.72) 2.18 (1.66−2.86)
<0.001
OR (95% CI)
0.71 (0.46-1.31) 1.97 (1.36−2.85) 1.24 (0.58−2.63) 10.89 (4.02−29.51) 2.96 (0.82−10.70) 5.22 (1.44−18.94) 8.69 (2.38−31.73) 1.34 (0.92−1.97)
P
0.342 0.003 0.578 <0.001 0.004
0.129
Significant P values were presented in bold text. CI = confidence interval; GS = Gleason score; MRI = magnetic resonance imaging; OR = odds ratio; PIRADS = Prostate Imaging Reporting and Data System; PSA = prostate-specific antigen; PSAD = prostate-specific antigen density.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
7
Fig. 5. Cribriform-risk nomogram construction and validation. (A) Risk nomogram including PSAD, PI-RADS score and maximal biopsy GS for predicting cribriform risk of CaP. (B) ROC curve analysis for the performance of the nomogram in cribriform discrimination with AUC, corresponding sensitivity and specificity. (C) Calibration curves depicting the calibration of the nomogram in terms of agreement between the predicted risk of cribriform morphology and observed pathological cribriform status. The 45˚ blue line represents a perfect prediction, the red dashed line represents the predictive performance of the nomogram, together with a bias-corrected black solid line. The closer the dashed line to the ideal line, the better the predictive accuracy of the nomogram is. AUC = area under the curve; CaP = prostate cancer; GS = Gleason score; PI-RADS = Prostate Imaging Reporting and Data System; PSAD = prostate-specific antigen density; ROC = receiver operating characteristic.
20.7%, 28.6%, and 37.1% in 47 RP patients, respectively [11]. Similarly, with 110 out of total 215 patients harboring cribriform morphology in our cohort, corresponding sensitivities for TB, SB, TB + SB were 28.2% (31/110), 22.7% (25/110), and 36.4% (40/110), respectively. Notably, among the 40 positive patients, 15 patients (37.5%) demonstrated cribriform morphology only in TB cores, 9 patients (22.5%) demonstrated cribriform morphology only in SB cores. As for the discrepant detection rates between noncribriform GP 4 and cribriform GP 4, relatively smaller volume of specific cribriform GP 4 fraction might partially attribute to the difference (In cribriform GP 4 cohort, minority patients were pure cribriform GP 4, most pattern 4
were consist of both noncribriform and cribriform GP 4), although tumor volume was larger for cribriform GP 4 patients. Besides, cribriform morphology was less visible on MRI [11,12], which might decrease cribriform GP 4 detection in targeted biopsy cores. Clinically, TB + SB rather than just TB or SB alone should be performed for maximal cribriform morphology detection. There is one thing that need to point out, all patients in our cohort received transperineal biopsy rather than transrectal biopsy. Recently, many studies aimed to compare the efficiency between these 2 techniques. According to Yaxley AJ et al., there was no significant difference in the ability to detect CaP or significant CaP using transperineal or transrectal
ARTICLE IN PRESS 8
J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
Fig. 6. Decision curve analysis for the cribriform-risk nomogram. Red line represents the nomogram, blue line represents the hypothesis that all patients harbor cribriform morphology, black line represents the hypothesis that no patients harbor cribriform morphology. The decision curve shows that if the threshold probability is between 0 and 0.88, then using the cribriform-risk nomogram to predict cribriform status adds more benefit than taking intervention to either all or no patients.
biopsy [21]. However, a recent meta-analysis revealed that the pooled diagnostic sensitivity of transperineal route (86%; 95% CI, 77−96) was better than transrectal route (73%, 62−88%) in diagnosing csCaP [22]. However, most studies were retrospective. Large prospective randomized or head-to-head comparison studies comparing the 2 approaches are warranted. Given the poor performance of different biopsy techniques for cribriform morphology, additional model for predicting cribriform morphology is necessary. Recent studies compared tumor size and PI-RADS score of tumors with different cribriform statuses [12,13]. Moreover, considering the significant differences of specific clinical characteristics and mpMRI parameters between cribriform-positive CaP and cribriform-negative CaP in our cohort, we intended to develop a risk model for prediction of aggressive cribriform morphology, which has never been proposed before. With multivariate logistic regression analysis, parameters including PSAD, PI-RADS score and maximal biopsy GS were identified for nomogram construction. The cribriform-risk nomogram demonstrated excellent discrimination and calibration with internal validation, together with considerable net benefits under specific threshold probabilities. With this nomogram, corresponding cribriform-positive risk could be calculated for patients diagnosed as CaP. Cribriform risk derived from the nomogram helps with the decision-making on ruling out candidates for active surveillance or focal therapy. Additionally, patients suspicious of cribriformpositive CaP should receive more radical treatment (such as enlarged pelvic lymph node dissection and extrafascial resection) than cribriform-negative CaP, because such patients are more frequent harboring positive lymph node and surgical margin.
Of course, our study has some limitations. First, data were retrospectively collected, and we used final RP pathology as the reference standard. Hence, a selection bias might occur. Nonetheless, the final RP specimen, especially whole-mount sections, is the most accurate arbiter to determine presence or absence of cribriform morphology. Second, we only enrolled patients with lesions of PI-RADS score ≥3 on mpMRI, which might restrict the nomogram popularized to general population. However, mpMRI could successfully detect 85−95% of index lesions and csCaP using RP specimens as reference standard [23,24]. In this respect, the majority of invisible lesions were clinical insignificant CaP. Third, due to the study aim (cribriform morphology), we specifically enrolled patients with GS 3 + 4/ 4 + 3 in RP, which might restrict clinical application to the general population. However, GS 6 and GS 8 to10 are 2 extremes of prostate cancer. With the nomogram, we believe that characteristics of such patients correspond to noncribriform CaP and cribriform-positive CaP, respectively. Fourth, we just conducted internal validation of the cribriform-risk nomogram on account of sample size. Further external validation is necessary. 5. Conclusions In summary, considering the poor performance of different biopsy techniques for identifying cribriform morphology, we developed a nomogram with combined clinical characteristics and mpMRI parameters for cribriform morphology prediction. The cribriform-risk nomogram demonstrated excellent discrimination and calibration with internal validation, together with considerable net benefits under specific threshold probabilities. Of course, further external validation is necessary for the nomogram. Conflict of interest No conflict of interest to declare. Funding This work was supported by the National Natural Science Foundation of China (grant nos. 81772710, 81572519, 81802535 and 81602221), the Project of Invigorating Health Care through Science, Technology and Education, Jiangsu Provincial Key Medical Discipline (Laboratory, ZDXKB2016014), the National Natural Science Foundation of Jiangsu Province (BK20160117), China Postdoctoral Fund (223427) and Nanjing Medical Science and technique Development Foundation (YKK 18064). Supplementary materials Supplementary material associated with this article can be found in the online version at https://doi.org/10.1016/j. urolonc.2019.09.002.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
References [1] Siegel RL, Miller KD, Jemal A. Cancer statistics, 2017. CA Cancer J Clin 2017;67:7–30. [2] Le JD, Tan N, Shkolyar E, et al. Multifocality and prostate cancer detection by multiparametric magnetic resonance imaging: correlation with whole-mount histopathology. Eur Urol 2015;67:569–76. [3] Gleason DF, Mellinger GT. Prediction of prognosis for prostatic adenocarcinoma by combined histological grading and clinical staging. J Urol 1974;111:58–64. [4] Epstein JI, Egevad L, Amin MB, Delahunt B, Srigley JR, Humphrey PA. The 2014 International Society of Urological Pathology (ISUP) consensus conference on gleason grading of prostatic carcinoma: definition of grading patterns and proposal for a new grading system. Am J Surg Pathol 2016;40:244–52. [5] Truong M, Frye T, Messing E, Miyamoto H. Historical and contemporary perspectives on cribriform morphology in prostate cancer. Nat Rev Urol 2018;15:475–82. [6] Kryvenko ON, Gupta NS, Virani N, et al. Gleason score 7 adenocarcinoma of the prostate with lymph node metastases: analysis of 184 radical prostatectomy specimens. Arch Pathol Lab Med 2013;137:610–7. [7] Dong F, Yang P, Wang C, et al. Architectural heterogeneity and cribriform pattern predict adverse clinical outcome for Gleason grade 4 prostatic adenocarcinoma. Am J Surg Pathol 2013;37:1855–61. [8] Kir G, Sarbay BC, Gumus E, Topal CS. The association of the cribriform pattern with outcome for prostatic adenocarcinomas. Pathol Res Pract 2014;210:640–4. [9] Harding-Jackson N, Kryvenko ON, Whittington EE, et al. Outcome of Gleason 3 + 5 = 8 prostate cancer diagnosed on needle biopsy: prognostic comparison with Gleason 4 + 4 = 8. J Urol 2016;196: 1076–81. [10] Siddiqui MM, Rais-Bahrami S, Turkbey B, et al. Comparison of MR/ ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA 2015;313:390–7. [11] Truong M, Feng C, Hollenberg G, et al. A Comprehensive analysis of cribriform morphology on magnetic resonance imaging/ultrasound fusion biopsy correlated with radical prostatectomy specimens. J Urol 2018;199:106–13. [12] Truong M, Hollenberg G, Weinberg E, Messing EM, Miyamoto H, Frye TP. Impact of Gleason subtype on prostate cancer detection using multiparametric magnetic resonance imaging: correlation with final histopathology. J Urol 2017;198:316–21.
9
[13] Prendeville S, Gertner M, Maganti M, et al. Role of magnetic resonance imaging targeted biopsy in detection of prostate cancer harboring adverse pathological features of intraductal carcinoma and invasive cribriform carcinoma. J Urol 2018;200:104–13. [14] Weinreb JC, Barentsz JO, Choyke PL, et al. PI-RADS Prostate Imaging - Reporting and Data System: 2015, Version 2. Eur Urol 2016;69:16–40. [15] Zhang Q, Wang W, Yang R, et al. Free-hand transperineal targeted prostate biopsy with real-time fusion imaging of multiparametric magnetic resonance imaging and transrectal ultrasound: single-center experience in China. Int Urol Nephrol 2015;47:727–33. [16] Kerr KF, Brown MD, Zhu K, Janes H. Assessing the clinical impact of risk prediction models with decision curves: guidance for correct interpretation and appropriate use. JClin Oncol 2016;34:2534–40. [17] Gleason DF. Classification of prostatic carcinomas. Cancer Chemother Rep 1966;50:125–8. [18] P Y, C W, S W, et al. Architectural heterogeneity and cribriform pattern predict adverse clinical outcome for Gleason grade 4 prostatic adenocarcinoma.. Am J Surg Pathol 2013;37:1855–61. [19] Masoomian M, Downes MR. Concordance of biopsy and prostatectomy diagnosis of intraductal and cribriform carcinoma in a prospectively collected data set. Histopathology. 2019: 74:474-82 [20] CH S, SY K, JY C, SH K. Diagnostic performance of Prostate Imaging Reporting and Data System version 2 for detection of prostate cancer: a systematic review and diagnostic meta-analysis. Eur Urol 2017;72:177–88. [21] Yaxley AJ, Yaxley JW. Comparison between target magnetic resonance imaging (MRI) in-gantry and cognitively directed transperineal or transrectal-guided prostate biopsies for Prostate Imaging-Reporting and Data System (PI-RADS) 3-5 MRI lesions. 2017: 120Suppl 3:43-50 [22] Tu X, Liu Z, Chang T, et al. Transperineal magnetic resonance imaging-targeted biopsy may perform better than transrectal route in the detection of clinically significant prostate cancer: systematic review and meta-analysis. Clin Genitourin Cancer 2019. https://doi.org/ 10.1016/j.clgc.2019.05.006. [23] Radtke JP, Schwab C, Wolf MB, et al. Multiparametric magnetic resonance imaging (MRI) and MRI-transrectal ultrasound fusion biopsy for index tumor detection: correlation with radical prostatectomy specimen. Eur Urol 2016;70:846–53. [24] Baco E, Ukimura O, Rud E, et al. Magnetic resonance imaging-transectal ultrasound image-fusion biopsies accurately characterize the index tumor: correlation with step-sectioned radical prostatectomy specimens in 135 patients. Eur Urol 2015;67:787–94.
Urologic Oncology: Seminars and Original Investigations 000 (2019) 1−9
Original Article
Combined clinical characteristics and multiparametric MRI parameters for prediction of cribriform morphology in intermediate-risk prostate cancer patients Jie Gaoa,1, Qing Zhanga,1, Yao Fub,1, Wei Wanga, Chengwei Zhanga, Yanshen Kana, Haifeng Huanga, Danyan Lic, Jiong Shib, Hongqian Guoa,*, Bing Zhangc,* a
Department of Urology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Institute of Urology Nanjing University, Nanjing, Jiangsu, China b Department of Pathology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing, Jiangsu, China c Department of Radiology, Nanjing Drum Tower Hospital, The Affiliated Hospital of Nanjing University Medical School, Nanjing, Jiangsu, China Received 22 June 2019; received in revised form 20 August 2019; accepted 4 September 2019
Abstract Purpose: To develop a risk model with combined clinical characteristics and multiparametric MRI parameters for prediction of cribriform morphology in intermediate-risk prostate cancer (CaP) patients. Methods: We retrospectively included 215 CaP patients received multiparametric MRIexamination, targeted biopsy (TB) combined with systematic biopsy (SB), radical prostatectomy and final Gleason group 2 or 3. Cribriform status was confirmed on both biopsy slices and wholemount sections. Characteristics were stratified by cribriform status. Mann Whitney U test was performed for continuous variables and the x2 test for categorical variables. Univariate and multivariate logistic regression analyses were performed for significant predictors, followed by cribriform-risk nomogram construction. Receiver operating characteristic analysis was used for internal discrimination validation with corresponding area under the curve. Calibration curves were plotted and decision curve analysis was performed for clinical benefit exploration. Results: Cribriform morphology was identified in 51.2% (110/215) patients. Cribriform-positive CaP demonstrated significantly higher prostate-specific antigen level, higher prostate-specific antigen density , larger lesion dimension on MRI, higher Prostate Imaging Reporting and Data System score, larger tumor dimension, higher Gleason score, higher pT stage, higher pN stage and more positive surgical margin (all P < 0.01). Sensitivities of TB, SB, TB + SB for detecting cribriform morphology were 28.2% (31/110), 22.7% (25/110), and 36.4% (40/ 110), respectively. Further, prostate-specific antigen density (P = 0.003), Prostate Imaging Reporting and Data System score (P < 0.001), and maximal biopsy Gleason score (P = 0.004) were independent predictors for positive cribriform morphology. Cribriform-risk nomogram was constructed with the 3 parameters and demonstrated considerable discrimination (area under the curve = 0.887, sensitivity 79.2%, specificity 84.0%) and calibration (mean absolute error 0.021), harboring net benefits with threshold probabilities range from 0 to 0.88. Conclusion: A cribriform-risk nomogram was developed and well predicted aggressive cribriform morphology in intermediate-risk CaP patients. Ó 2019 Elsevier Inc. All rights reserved.
Keywords: Prostate cancer; Cribriform morphology; Nomogram; Multiparametric MRI; Biopsy
1. Introduction Prostate cancer (CaP), the most common cancer type and a major cause of cancer mortality for men in Europe and the 1
These authors contributed equally to this work. *Corresponding authors. Tel.: 862583106666; fax:862568182863. E-mail addresses: [email protected] (H. Guo), [email protected] (B. Zhang). https://doi.org/10.1016/j.urolonc.2019.09.002 1078-1439/Ó 2019 Elsevier Inc. All rights reserved.
United States, is a histologically heterogeneous and frequently multifocal disease [1,2]. The original scoring system described by Gleason in 1966 for grading CaP on the basis of histological morphology was later found to accurately predict cancer-specific death [3]. Cribriform morphology, a subtype of Gleason pattern 4, is recognized as a more aggressive and often more lethal architecture than noncribriform morphologies of Gleason pattern (GP) 4 [4,5]. Over the past few years, many studies have emerged
ARTICLE IN PRESS 2
J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
demonstrating that cribriform-positive CaP is associated with increased risk of lymph node invasion [6], distant metastasis [7], biochemical recurrence [8] and cancer-specific death [9]. Hence, correct identification of the cribriform status within CaP is crucial for clinical decision. Currently, the diagnosis of cribriform morphology is most commonly made after transrectal ultrasonography (TRUS)-guided prostate biopsy. Besides, prostate multiparametric MRI (mpMRI) is gaining popularity for the detection of clinically significant CaP (csCaP), and MRI/ US fusion targeted biopsy (TB) of suspicious mpMRI lesions improve the detection of csCaP by 30% [10]. Despite advances in biopsy techniques, the inherent sampling limitations of needle biopsy are still problematic for accurate grading of CaP. Matthew T et al. revealed that MRI/US fusion TB alone, systematic biopsy (SB) and the combined approach (TB + SB) demonstrated low sensitivity and specificity for cribriform morphology detection using radical prostatectomy (RP) as the reference standard [11]. Thus, additional techniques or models for diagnosing cribriform morphology is necessary prior of treatment. Generally, cribriform-positive CaP harbored relatively larger tumor size compared with cribriform-negative CaP (predominantly poorly formed or fused glands) [12,13], together with larger lesion dimension on mpMRI [13]. With Prostate Imaging Reporting and Data System v2 (PIRADS v2), the most current mpMRI scoring system, PIRADS score 5 was significantly associated with positive cribriform morphology on biopsy pathology [13]. This implies that cribriform morphology is more likely to be present in tumors greater than 1.5 cm or those with an invasive appearance on mpMRI. Given the potential differences between tumors of different cribriform status, we hypothesized the feasibility constructing a diagnostic model with
combined clinical characteristics and mpMRI parameters for prediction of aggressive cribriform morphology in CaP, which has never been studied as of yet. In this study, we aimed to identify significant predictors for cribriform morphology and further construct a risk model for prediction of aggressive cribriform-positive CaP with whole-mount sections of RP. 2. Materials and methods 2.1. Patients Between August 2016 and December 2018, data of patients underwent RP at our single center were retrospectively collected. The inclusion criteria were as follows: (a) preoperative mpMRI within 3 months from surgery; (b) mpMRI/US-guided TB and transperineal SB were performed and patients were pathologically diagnosed as prostate adenocarcinoma; (c) all biopsy slices and whole-mount sections were available; (d) final Gleason score (GS) 3 + 4 and 4 + 3; (e) no neoadjuvant therapy before surgery (such as androgen deprivation therapy and transurethral resection of the prostate. Such patients harbored incomplete or unrecognizable pathology, which might disturb the correlation between clinic-radiological parameters and actual pathology). Initially, 856 patients underwent RP were included, and specific selection flowchart is shown in Fig. 1. Finally, a total of 215 patients were enrolled. The institutional review board approval was obtained, while all patients were informed and signed consent. 2.2. MRI examination and interpretation Prostate mpMRI examinations were routinely performed and reported in consensus by 2 radiologists according to the
Fig. 1. Study flowchart of patient selection. ADT = androgen deprivation therapy; mpMRI = multiparametric magnetic resonance imaging; RP = radical prostatectomy; TURP = transurethral resection of the prostate.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
3
PI-RADS v2 standard [14]. Specific descriptions were available in supplementary files.
21.0 (IBM, USA). All tests were 2-sided, with statistical significance set at P < 0.05.
2.3. Biopsy protocol
3. Results
For lesions suspicious of CaP (PI-RADS score ≥3) on mpMRI, TB was performed as recommended by PI-RADS v2 [14]. All biopsies were conducted in transperineal approach with MRI-US image registration system (Esaote and RVS) that provides real-time fusion of TRUS images and mpMR images as previously described [15]. Briefly, lesions identified on mpMRI were semiautomatically displayed on the real-time TRUS image. The biopsy started with TB to the center of suspicious lesions using free-hand transperineal technique. Afterwards, standard 12-core transperineal SB was carried out.
Characteristics of the total 215 patients are summarized in Table 1. After pathologic slices evaluation, these patients were divided into cribriform-positive CaP (110/215, 51.2%) and cribriform-negative CaP (105/215, 48.8%) according to the cribriform status on whole-mount sections. Further, characteristics stratified by cribriform status were compared (Table 1). Patients with positive cribriform morphology demonstrated significantly higher prostate-specific antigen (PSA) level (P = 0.001), higher prostate-specific antigen density (PSAD) (P < 0.001), larger lesion dimension on MRI (P < 0.001), higher PI-RADS score (P < 0.001), more positive biopsy cores (TB + SB) (P < 0.001), larger tumor dimension on final pathology (P < 0.001), higher GS (P < 0.001), higher pT stage (P = 0.001), higher pN stage (P = 0.019) and more positive surgical margin (P = 0.004). No significant differences were found for age and prostate volume between 2 groups (P = 0.145 and P = 0.093, respectively). Further, using whole-mount pathology as gold standard, we evaluated the sensitivities of different biopsy techniques for detecting any GP 4, noncribriform GP 4 and cribriform GP 4. As shown in Fig. 2A, 167 out of 215 (77.7%) patients of GS > 3 + 3 harbored biopsy GS > 3 + 3, with 18 patients (10.8%) demonstrating biopsy GS > 3 + 3 only in TB cores, 39 patients (23.4%) demonstrating biopsy GS > 3 + 3 only in SB cores, and 110 patients (65.8%) demonstrating biopsy GS > 3 + 3 in both TB cores and SB cores. Thus, sensitivities of TB, SB, TB + SB for detecting GP 4 were 59.5% (128/215), 69.3% (149/215) and 77.7% (167/215), respectively. As for 105 noncribriform CaP patients, 8 patients demonstrating GP 4 only in TB cores, 22 patients demonstrating GP 4 only in SB cores, and 38 patients demonstrating GP 4 in both TB cores and SB cores. Thus, sensitivities of TB, SB, TB + SB for detecting noncribriform GP 4 were 43.8% (46/105), 57.1% (60/105) and 64.8% (68/105), respectively. In terms of 110 cribriform-positive patients, 36.4% (40/110) harbored cribriform morphology on biopsy pathology, while 63.6% (70/110) were missed (Fig. 2B). Representative cases illustrating presence and absence of cribriform morphology on biopsy slices were shown in Figs. 3 and 4, respectively. Among the 40 positive cases, 15 (37.5%) demonstrated cribriform morphology only in TB cores, 9 (22.5%) demonstrated cribriform morphology only in SB cores, and 16 (40.0%) demonstrated cribriform morphology in both TB cores and SB cores. Thus, sensitivities of TB, SB, TB + SB for detecting cribriform morphology were 28.2% (31/110), 22.7% (25/110) and 36.4% (40/110), respectively. Given the relatively low sensitivities of different techniques for detecting cribriform morphology (Fig. 2B) and the differences of clinic-radiological characteristics between
2.4. Biopsy and whole-mount histopathology Biopsy specimens were routinely disposed and reported by 2 urological pathologists. After prostatectomy, intact prostate specimens were sliced from apex to base at 3-mm intervals, following by paraffin embedding and hematoxylin-eosin stain. All whole-mount slices were reviewed in consensus by 2 urological pathologists according to the 2014 International Society of Urological Pathology (ISUP) modified criteria of CaP [4]. Each lesion was outlined and assigned a GS. Notably, cribriform morphology was specifically identified. For patients with positive cribriform morphology on whole-mount sections, corresponding biopsy slices including both TB cores and SB cores were reviewed separately. Any presence of cribriform morphology on biopsy sections was recorded with the specific core number. 2.5. Statistical analysis Mann-Whitney U test was performed for continuous variables and the x2 test for categorical variables. Univariate and multivariate logistic regression analyses were performed for significant parameters to predict the presence of cribriform morphology on final pathology. Significant predictors were used for nomogram construction with R software version 3.5.2 (https://www.r-project.org/) using the “rms” package. Receiver operating characteristic (ROC) analysis for the internal discrimination validation was done by bootstrapping with 1,000 iterations using the “pROC” package. Corresponding area under the curve (AUC) with 95% confidence interval, sensitivity and specificity for differentiation were derived. The extent of over- or underestimation of predicted probabilities relative to observed probabilities of cribriform-positive CaP was explored graphically by calibration plot using the “rms” package, which was internally validated using bootstrapping with 1,000 iterations. Decision curve analysis (DCA) was performed using the “rmda” package as previously described [16]. Statistical analysis was performed with software SPSS
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
4
Table 1 Characteristics of 215 patients and stratified by cribriform status of CaP. Characteristic
Age (y) PSA level (ng/ml) Prostate volume (ml) PSAD (ng/ml/ml) Lesion dimension on MRI (cm) PI-RADS score 3 4 5 Positive biopsy cores (TB + SB) Maximal biopsy GS (TB + SB) 3+3 3+4 4+3 4+4 Tumor dimension on final pathology (cm) Gleason score (RP) 3+4 4+3 pT stage T2 T3a T3b pN stage N0 N1 Surgical margin Positive Negative
n = 215
Cribriform morphology Positive (n = 110)
Negative (n = 105)
69 (64−74) 10.9 (7.3−17.9) 29.9 (22.5−41.7) 0.39 (0.23−0.64) 1.5 (1.1−2.2)
71 (63−75) 13.2 (8.6−21.5) 28.4 (22.5−38.8) 0.48 (0.29−0.77) 2.0 (1.5−2.5)
68 (64−73) 9.3 (6.7−15.5) 33.1 (22.9−43.9) 0.30 (0.20−0.51) 1.2 (1.0−1.5)
23 (10.7) 86 (40.0) 106 (49.3) 3 (5−7)
4 (3.6) 27 (24.6) 79 (71.8) 6 (4−8)
19 (18.1) 59 (56.2) 27 (25.7) 4 (2−6)
44 (20.5) 50 (23.3) 63 (29.3) 58 (26.9) 1.8 (1.4−2.5)
7 (6.4) 22 (20.0) 38 (34.5) 43 (39.1) 2.3 (1.7−2.9)
37 (35.2) 28 (26.7) 25 (23.8) 15 (14.3) 1.5 (1.1−2.1)
127 (59.1) 88 (40.9)
44 (40.0) 66 (60.0)
83 (79.0) 22 (20.9)
99 (46.0) 96 (44.7) 20 (9.3)
39 (35.5) 55 (50.0) 16 (14.5)
60 (57.1) 41 (39.1) 4 (3.8)
205 (95.3) 10 (4.7)
101 (91.8) 9 (8.2)
104 (99.0) 1 (1.0)
58 (27.0) 157 (73.0)
39 (35.5) 71 (64.5)
19 (18.1) 86 (81.9)
P
0.145 0.001 0.093 <0.001 <0.001 <0.001
<0.001 <0.001
<0.001 <0.001
0.001
0.019y
0.004
Continuous variables are presented as median (interquartile range, IQR), while categorical variables are presented as patients (%). Fisher’s exact test was done for P value calculation. GS = Gleason score; MRI = magnetic resonance imaging; PI-RADS = Prostate Imaging Reporting and Data System; PSA = prostate-specific antigen; PSAD = prostate- specific antigen density; RP = radical prostatectomy; SB = systematic biopsy; TB = targeted biopsy.
y
different cribriform statuses (Table 1), we tried to find significant predictors for cribriform morphology. With univariate and multivariate logistic regression model analyses, PSAD (odds ratio 1.97, 95% confidence interval 1.36
−2.85, P = 0.003), PI-RADS score (odds ratio 10.89, 95%CI 4.02−29.51, P < 0.001) and maximal biopsy GS (P = 0.004) and were identified as independent predictors of cribriform morphology within CaP (Table 2).
Fig. 2. Sensitivities of different biopsy techniques for detecting any GP 4, noncribriform GP 4 and cribriform GP 4. Biopsy slices of TB cores and SB cores were separately evaluated to identify any GP 4, noncribriform GP 4 and cribriform GP 4. Further, corresponding sensitivities of different biopsy techniques for detecting any GP 4, noncribriform GP 4 and cribriform GP 4 were derived. GP = Gleason pattern; SB = systematic biopsy; TB = targeted biopsy.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
5
Fig. 3. Representative radio-pathological matching case of cribriform-positive CaP with positive cribriform morphology on biopsy pathology. Patient summary: a 71-year-old man with CaP in right peripheral zone, PSA level 15.38 ng/ml, prostate volume 29.1 ml; PSAD 0.53 ng/ml/ml; lesion dimension on MRI 1.5cm; PI-RADS score 5, positive biopsy cores 4 out of 14, final Gleason score 4 + 3, tumor dimension 1.7 cm. (A) axial T2-weighted MRI; (B) axial diffusion-weighted MRI; (C) axial ADC map; (D) representative biopsy pathology with positive cribriform morphology; (E) whole-mount section of radical prostatectomy specimen; (F) representative cribriform morphology derived from whole-mount section. Lesion was outlined with solid line on corresponding MR images and with dashed line on whole-mount section, respectively. ADC = apparent diffusion coefficient; CaP = prostate cancer; MRI = magnetic resonance imaging; PI-RADS = prostate imaging reporting and data system; PSA = prostate-specific antigen; PSAD = prostate-specific antigen density.
Further, a cribriform-risk nomogram incorporating these 3 predictors was constructed (Fig. 5A). The nomogram was internally validated by bootstrapping. ROC analysis was used for the nomogram discrimination, with an AUC of 0.887 (95% CI 0.8440.930) (Fig. 5B). Corresponding sensitivity and specificity were 79.2% and 84.0%, respectively. Moreover, bootstrapped calibration curve of the nomogram demonstrated that there was no untoward deviation of predicted risk from observed risk of cribriform morphology, with a mean absolute error of 0.021 (Fig. 5C). Further, the bootstrapped DCA was performed for clinical benefit exploration (Fig. 6A). With threshold probabilities within the range from 0 to 0.88, the nomogram adds more net benefits than taking intervention to either all or no patients. DCA was also performed in subset patients for whom GS ≤ 3 + 4 and > 3 + 4 were identified at biopsy (Supplementary Figure 1). 4. Discussion To our knowledge, this is the first study to construct a novel risk model with combined clinical characteristics and mpMRI parameters for prediction of aggressive cribriform morphology in CaP. Regarding the cribriform-risk
nomogram, corresponding sensitivity and specificity for cribriform status were 79.2% and 84.0%, together with considerable calibration and net benefits under specific threshold probabilities. Cribriform morphology, which was described by Gleason as the architecture having “moderately differentiated glands, range from small to large, growing in spaced-out infiltrative patterns” [17], has drawn much attention recently. Dong et al. revealed that the presence of cribriform architecture was an independent predictor for biochemical recurrence as well as metastasis after RP [18]. In 2014, the ISUP conference reached a consensus that cribriform gland should be assigned Gleason pattern 4, regardless of morphology [4]. Since the 2014 ISUP meeting, a rapidly growing body of evidence indicated the aggressive behavior of cribriform morphology as described above [5−9]. In our cohort, cribriform-positive CaP was significantly associated with increased PSA level, PSAD and tumor dimension. Moreover, cribriform morphology was more frequently present in carcinoma with relatively higher GS, T stage and N stage, consistent with a recent study of RP cohort [19]. As we know, mpMRI is currently the best imaging method to noninvasively identify and characterize CaP [14]. With PI-RADS v2, mpMRI demonstrated good
ARTICLE IN PRESS 6
J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
Fig. 4. Representative radio-pathological matching case of cribriform-positive CaP with negative cribriform morphology on biopsy pathology. Patient summary: a 60-year-old man with CaP in left peripheral zone, PSA level 5.84 ng/ml, prostate volume 26.6 ml; PSAD 0.22 ng/ml/ml; lesion dimension on MRI 0.9 cm; PI-RADS score 4, positive biopsy cores 3 out of 14, final Gleason score 4 + 3, tumor dimension 1.3 cm. (A) axial T2-weighted MRI; (B) axial diffusion-weighted MRI; (C) axial ADC map; (D) representative biopsy pathology with negative cribriform morphology; (E) whole-mount section of radical prostatectomy specimen; (F) representative cribriform morphology derived from whole-mount section. Lesion was outlined with solid line on corresponding MR images and with dashed line on whole-mount section, respectively. ADC = apparent diffusion coefficient; CaP = prostate cancer; MRI = magnetic resonance imaging; PI-RADS = Prostate Imaging Reporting and Data System; PSA = prostate-specific antigen; PSAD = prostate-specific antigen density.
performance for detecting CaP, with pooled sensitivity of 89% and specificity of 73%, respectively [20]. MRI/US fusion TB of suspicious mpMRI lesions improved the detection of csCaP by 30% [10]. As for csCaP (GS > 3 + 3), TB combined with SB demonstrated a satisfying sensitivity of 77.7% for detecting GP 4 in our cohort. However, compared to TB, SB showed better sensitivity for detecting
GP 4 in our cohort. The experiences of radiologists at their institution or quality of MR registration might account for the difference. Additionally, TB + SB showed acceptable sensitivity (64.8%) for noncribriform GP 4 detection. However, for cribriform GP 4, several biopsy techniques including TB, SB, TB + SB demonstrated poor performance on cribriform morphology identification, with sensitivities of
Table 2 Univariate and multivariate logistic regression analyses for the prediction of cribriform-positive CaP. Variable and intercept
Age (per 5 yr) PSA level (interquartile OR) Prostate volume (interquartile OR) PSAD (interquartile OR) Lesion dimension on MRI PI-RADS (5 vs. 3/4) Maximal biopsy GS 3 + 4 vs. 3 + 3 4 + 3 vs. 3 + 3 4 + 4 vs. 3 + 3 Positive biopsy cores (interquartile OR)
Univariate logistic regression
Multivariate logistic regression
OR (95% CI)
P
1.15 (0.96−1.39) 1.80 (1.45−2.24) 0.85 (0.69−1.04) 2.23 (1.77−2.81) 4.46 (2.74−7.27) 15.83 (8.04−31.17)
0.133 <0.001 0.112 <0.001 <0.001 <0.001 <0.001
5.79 (1.95−17.2) 8.96 (3.13−25.67) 12.2 (4.17−33.72) 2.18 (1.66−2.86)
<0.001
OR (95% CI)
0.71 (0.46-1.31) 1.97 (1.36−2.85) 1.24 (0.58−2.63) 10.89 (4.02−29.51) 2.96 (0.82−10.70) 5.22 (1.44−18.94) 8.69 (2.38−31.73) 1.34 (0.92−1.97)
P
0.342 0.003 0.578 <0.001 0.004
0.129
Significant P values were presented in bold text. CI = confidence interval; GS = Gleason score; MRI = magnetic resonance imaging; OR = odds ratio; PIRADS = Prostate Imaging Reporting and Data System; PSA = prostate-specific antigen; PSAD = prostate-specific antigen density.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
7
Fig. 5. Cribriform-risk nomogram construction and validation. (A) Risk nomogram including PSAD, PI-RADS score and maximal biopsy GS for predicting cribriform risk of CaP. (B) ROC curve analysis for the performance of the nomogram in cribriform discrimination with AUC, corresponding sensitivity and specificity. (C) Calibration curves depicting the calibration of the nomogram in terms of agreement between the predicted risk of cribriform morphology and observed pathological cribriform status. The 45˚ blue line represents a perfect prediction, the red dashed line represents the predictive performance of the nomogram, together with a bias-corrected black solid line. The closer the dashed line to the ideal line, the better the predictive accuracy of the nomogram is. AUC = area under the curve; CaP = prostate cancer; GS = Gleason score; PI-RADS = Prostate Imaging Reporting and Data System; PSAD = prostate-specific antigen density; ROC = receiver operating characteristic.
20.7%, 28.6%, and 37.1% in 47 RP patients, respectively [11]. Similarly, with 110 out of total 215 patients harboring cribriform morphology in our cohort, corresponding sensitivities for TB, SB, TB + SB were 28.2% (31/110), 22.7% (25/110), and 36.4% (40/110), respectively. Notably, among the 40 positive patients, 15 patients (37.5%) demonstrated cribriform morphology only in TB cores, 9 patients (22.5%) demonstrated cribriform morphology only in SB cores. As for the discrepant detection rates between noncribriform GP 4 and cribriform GP 4, relatively smaller volume of specific cribriform GP 4 fraction might partially attribute to the difference (In cribriform GP 4 cohort, minority patients were pure cribriform GP 4, most pattern 4
were consist of both noncribriform and cribriform GP 4), although tumor volume was larger for cribriform GP 4 patients. Besides, cribriform morphology was less visible on MRI [11,12], which might decrease cribriform GP 4 detection in targeted biopsy cores. Clinically, TB + SB rather than just TB or SB alone should be performed for maximal cribriform morphology detection. There is one thing that need to point out, all patients in our cohort received transperineal biopsy rather than transrectal biopsy. Recently, many studies aimed to compare the efficiency between these 2 techniques. According to Yaxley AJ et al., there was no significant difference in the ability to detect CaP or significant CaP using transperineal or transrectal
ARTICLE IN PRESS 8
J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
Fig. 6. Decision curve analysis for the cribriform-risk nomogram. Red line represents the nomogram, blue line represents the hypothesis that all patients harbor cribriform morphology, black line represents the hypothesis that no patients harbor cribriform morphology. The decision curve shows that if the threshold probability is between 0 and 0.88, then using the cribriform-risk nomogram to predict cribriform status adds more benefit than taking intervention to either all or no patients.
biopsy [21]. However, a recent meta-analysis revealed that the pooled diagnostic sensitivity of transperineal route (86%; 95% CI, 77−96) was better than transrectal route (73%, 62−88%) in diagnosing csCaP [22]. However, most studies were retrospective. Large prospective randomized or head-to-head comparison studies comparing the 2 approaches are warranted. Given the poor performance of different biopsy techniques for cribriform morphology, additional model for predicting cribriform morphology is necessary. Recent studies compared tumor size and PI-RADS score of tumors with different cribriform statuses [12,13]. Moreover, considering the significant differences of specific clinical characteristics and mpMRI parameters between cribriform-positive CaP and cribriform-negative CaP in our cohort, we intended to develop a risk model for prediction of aggressive cribriform morphology, which has never been proposed before. With multivariate logistic regression analysis, parameters including PSAD, PI-RADS score and maximal biopsy GS were identified for nomogram construction. The cribriform-risk nomogram demonstrated excellent discrimination and calibration with internal validation, together with considerable net benefits under specific threshold probabilities. With this nomogram, corresponding cribriform-positive risk could be calculated for patients diagnosed as CaP. Cribriform risk derived from the nomogram helps with the decision-making on ruling out candidates for active surveillance or focal therapy. Additionally, patients suspicious of cribriformpositive CaP should receive more radical treatment (such as enlarged pelvic lymph node dissection and extrafascial resection) than cribriform-negative CaP, because such patients are more frequent harboring positive lymph node and surgical margin.
Of course, our study has some limitations. First, data were retrospectively collected, and we used final RP pathology as the reference standard. Hence, a selection bias might occur. Nonetheless, the final RP specimen, especially whole-mount sections, is the most accurate arbiter to determine presence or absence of cribriform morphology. Second, we only enrolled patients with lesions of PI-RADS score ≥3 on mpMRI, which might restrict the nomogram popularized to general population. However, mpMRI could successfully detect 85−95% of index lesions and csCaP using RP specimens as reference standard [23,24]. In this respect, the majority of invisible lesions were clinical insignificant CaP. Third, due to the study aim (cribriform morphology), we specifically enrolled patients with GS 3 + 4/ 4 + 3 in RP, which might restrict clinical application to the general population. However, GS 6 and GS 8 to10 are 2 extremes of prostate cancer. With the nomogram, we believe that characteristics of such patients correspond to noncribriform CaP and cribriform-positive CaP, respectively. Fourth, we just conducted internal validation of the cribriform-risk nomogram on account of sample size. Further external validation is necessary. 5. Conclusions In summary, considering the poor performance of different biopsy techniques for identifying cribriform morphology, we developed a nomogram with combined clinical characteristics and mpMRI parameters for cribriform morphology prediction. The cribriform-risk nomogram demonstrated excellent discrimination and calibration with internal validation, together with considerable net benefits under specific threshold probabilities. Of course, further external validation is necessary for the nomogram. Conflict of interest No conflict of interest to declare. Funding This work was supported by the National Natural Science Foundation of China (grant nos. 81772710, 81572519, 81802535 and 81602221), the Project of Invigorating Health Care through Science, Technology and Education, Jiangsu Provincial Key Medical Discipline (Laboratory, ZDXKB2016014), the National Natural Science Foundation of Jiangsu Province (BK20160117), China Postdoctoral Fund (223427) and Nanjing Medical Science and technique Development Foundation (YKK 18064). Supplementary materials Supplementary material associated with this article can be found in the online version at https://doi.org/10.1016/j. urolonc.2019.09.002.
ARTICLE IN PRESS J. Gao et al. / Urologic Oncology: Seminars and Original Investigations 00 (2019) 1−9
References [1] Siegel RL, Miller KD, Jemal A. Cancer statistics, 2017. CA Cancer J Clin 2017;67:7–30. [2] Le JD, Tan N, Shkolyar E, et al. Multifocality and prostate cancer detection by multiparametric magnetic resonance imaging: correlation with whole-mount histopathology. Eur Urol 2015;67:569–76. [3] Gleason DF, Mellinger GT. Prediction of prognosis for prostatic adenocarcinoma by combined histological grading and clinical staging. J Urol 1974;111:58–64. [4] Epstein JI, Egevad L, Amin MB, Delahunt B, Srigley JR, Humphrey PA. The 2014 International Society of Urological Pathology (ISUP) consensus conference on gleason grading of prostatic carcinoma: definition of grading patterns and proposal for a new grading system. Am J Surg Pathol 2016;40:244–52. [5] Truong M, Frye T, Messing E, Miyamoto H. Historical and contemporary perspectives on cribriform morphology in prostate cancer. Nat Rev Urol 2018;15:475–82. [6] Kryvenko ON, Gupta NS, Virani N, et al. Gleason score 7 adenocarcinoma of the prostate with lymph node metastases: analysis of 184 radical prostatectomy specimens. Arch Pathol Lab Med 2013;137:610–7. [7] Dong F, Yang P, Wang C, et al. Architectural heterogeneity and cribriform pattern predict adverse clinical outcome for Gleason grade 4 prostatic adenocarcinoma. Am J Surg Pathol 2013;37:1855–61. [8] Kir G, Sarbay BC, Gumus E, Topal CS. The association of the cribriform pattern with outcome for prostatic adenocarcinomas. Pathol Res Pract 2014;210:640–4. [9] Harding-Jackson N, Kryvenko ON, Whittington EE, et al. Outcome of Gleason 3 + 5 = 8 prostate cancer diagnosed on needle biopsy: prognostic comparison with Gleason 4 + 4 = 8. J Urol 2016;196: 1076–81. [10] Siddiqui MM, Rais-Bahrami S, Turkbey B, et al. Comparison of MR/ ultrasound fusion-guided biopsy with ultrasound-guided biopsy for the diagnosis of prostate cancer. JAMA 2015;313:390–7. [11] Truong M, Feng C, Hollenberg G, et al. A Comprehensive analysis of cribriform morphology on magnetic resonance imaging/ultrasound fusion biopsy correlated with radical prostatectomy specimens. J Urol 2018;199:106–13. [12] Truong M, Hollenberg G, Weinberg E, Messing EM, Miyamoto H, Frye TP. Impact of Gleason subtype on prostate cancer detection using multiparametric magnetic resonance imaging: correlation with final histopathology. J Urol 2017;198:316–21.
9
[13] Prendeville S, Gertner M, Maganti M, et al. Role of magnetic resonance imaging targeted biopsy in detection of prostate cancer harboring adverse pathological features of intraductal carcinoma and invasive cribriform carcinoma. J Urol 2018;200:104–13. [14] Weinreb JC, Barentsz JO, Choyke PL, et al. PI-RADS Prostate Imaging - Reporting and Data System: 2015, Version 2. Eur Urol 2016;69:16–40. [15] Zhang Q, Wang W, Yang R, et al. Free-hand transperineal targeted prostate biopsy with real-time fusion imaging of multiparametric magnetic resonance imaging and transrectal ultrasound: single-center experience in China. Int Urol Nephrol 2015;47:727–33. [16] Kerr KF, Brown MD, Zhu K, Janes H. Assessing the clinical impact of risk prediction models with decision curves: guidance for correct interpretation and appropriate use. JClin Oncol 2016;34:2534–40. [17] Gleason DF. Classification of prostatic carcinomas. Cancer Chemother Rep 1966;50:125–8. [18] P Y, C W, S W, et al. Architectural heterogeneity and cribriform pattern predict adverse clinical outcome for Gleason grade 4 prostatic adenocarcinoma.. Am J Surg Pathol 2013;37:1855–61. [19] Masoomian M, Downes MR. Concordance of biopsy and prostatectomy diagnosis of intraductal and cribriform carcinoma in a prospectively collected data set. Histopathology. 2019: 74:474-82 [20] CH S, SY K, JY C, SH K. Diagnostic performance of Prostate Imaging Reporting and Data System version 2 for detection of prostate cancer: a systematic review and diagnostic meta-analysis. Eur Urol 2017;72:177–88. [21] Yaxley AJ, Yaxley JW. Comparison between target magnetic resonance imaging (MRI) in-gantry and cognitively directed transperineal or transrectal-guided prostate biopsies for Prostate Imaging-Reporting and Data System (PI-RADS) 3-5 MRI lesions. 2017: 120Suppl 3:43-50 [22] Tu X, Liu Z, Chang T, et al. Transperineal magnetic resonance imaging-targeted biopsy may perform better than transrectal route in the detection of clinically significant prostate cancer: systematic review and meta-analysis. Clin Genitourin Cancer 2019. https://doi.org/ 10.1016/j.clgc.2019.05.006. [23] Radtke JP, Schwab C, Wolf MB, et al. Multiparametric magnetic resonance imaging (MRI) and MRI-transrectal ultrasound fusion biopsy for index tumor detection: correlation with radical prostatectomy specimen. Eur Urol 2016;70:846–53. [24] Baco E, Ukimura O, Rud E, et al. Magnetic resonance imaging-transectal ultrasound image-fusion biopsies accurately characterize the index tumor: correlation with step-sectioned radical prostatectomy specimens in 135 patients. Eur Urol 2015;67:787–94.