Extracapsular extension of transitional zone prostate cancer miss-detected by multiparametric magnetic resonance imaging

To demonstrate the importance of extracapsular extension (ECE) of transitional zone (TZ) prostate cancer (PCa), examine the causes of its missed detection by Mp-MRI, and develop a new predictive model by integrating multi-level clinical variables. This retrospective study included 304 patients who underwent laparoscopic radical prostatectomy after 12 + X needle transperineal transrectal ultrasound (TRUS)-MRI-guided targeted prostate biopsy from 2018 to 2021 in our center was performed. In this study, the incidence rates of ECE were similar in patients with MRI lesions in the peripheral zone (PZ) and TZ (P = 0.66). However, the missed detection rate was higher in patients with TZ lesions than in those with PZ lesions (P < 0.05). These missed detections result in a higher positive surgical margin rate (P < 0.05). In patients with TZ lesions, detected MP-MRI ECE may have grey areas: the longest diameters of the MRI lesions were 16.5–23.5 mm; MRI lesion volumes were 0.63–2.51 ml; MRI lesion volume ratios were 2.75–8.86%; PSA were 13.85–23.05 ng/ml. LASSO regression was used to construct a clinical prediction model for predicting the risk of ECE in TZ lesions from the perspective of MRI and clinical features, including four variables: the longest diameter of MRI lesions, TZ pseudocapsule invasion, ISUP grading of biopsy pathology, and number of positive biopsy needles. Patients with MRI lesions in the TZ have the same incidence of ECE as those with lesions in the PZ, but a higher missed detection rate.


Introduction
Extracapsular extension (ECE) is important in the clinical staging of prostate cancer. Accurate clinical staging before treatment significantly impacts treatment choice and prognosis in prostate cancer. Assessing ECE before treatment is the key to selecting the best treatment strategy (Gandaglia et al. 2017;Patel et al. 2018).In the current clinical practice, multiparametric magnetic resonance imaging (Mp-MRI) plays a key role in ECE diagnosis (Mason et al. 2019). It is widely acknowledged that the latest prostate imaging report and data system (PI-RADS) guidelines standardized the diagnostic process of Mp-MRI and improved the specificity for ECE, however, the sensitivity was still limited (Turkbey et al. 2019). For example in a meta-analysis of 75 studies involving 9796 patients, the overall sensitivity and specificity for ECE detection were 0.57 (95% CI 0.49-0.65) and 0.91 (95% CI 0.88-0.93), respectively (Rooij et al. 2015). This problem has attracted the attention of many researchers. However, the additional benefits of some attempts such as functional contrast-enhanced imaging, 3D imaging, and PSMA PET remain uncertain (Caglic et al. 2019(Caglic et al. , 2022Muehlematter et al. 2019).
Previously, TZ PCa was often overlooked because of the previously thought low incidence and low malignancy (Ali et al. 2022). But in clinical practice, we found that a large number of patients with TZ PCa pathology suggest the occurrence of ECE but MP-MRI did not find ECE. And, there are few studies on the causes of the low sensitivity of MP-MRI to ECE. Therefore, it is important to demonstrate the importance of ECE of TZ PCa, clarify the reasons for its low ECE sensitivity of Mp-MRI, and develop an ECE prediction tool that fulfills the test requirements for clinical practice, which was the aim of this study.

Study patients
A retrospective study of 217 patients administered laparoscopic radical prostatectomy after 12 + X needle transperineal transrectal ultrasound (TRUS)-MRI-guided targeted prostate biopsy from 2018 to 2020 in our center was performed. The patients were screened based on the following exclusion criteria: (1) PI-RADS < 4 (MRI lesions less likely to be true tumors; in the lesion level analysis, the positive rate of PI-RADS 1 was 2%, versus 4% for PI-RADS 2 and 20% for PI-RADS 3 (Oerther et al. 2021); PI-RADS score was a comprehensive evaluation of T2WI, DWI, and dynamic contrast-enhanced sequences); (2) diffuse MRI lesions or unclear MRI findings. (3) tPSA > 100 ng/ml (inaccurate measurements); (4) neoadjuvant therapy (changes in clinical features). Totally 87 patients who met the same eligibility criteria from January to September 2021 were selected as an external validation cohort.

Mp-MRI and clinical features
All patients were imaged on a 3.0T MRI scanner (MAG-NETOM Skyra; Siemens Healthineers, Erlangen, Germany) with a standard spine array coil and an 18-channel body array coil. The required sequence was obtained as described previously (He et al. 2021). Transperineal 12-needle system puncture and X-needle targeted puncture was performed with the transrectal ultrasound and MRI image fusion software (Huang et al. 2022). All MRI parameters were measured according to PI-RADS Guideline V2.1 (Turkbey et al. 2019) and all volumes were calculated from (maximum AP diameter) × (maximum transverse diameter) × (maximum cranio-caudal diameter) × 0.52. Pathological criteria according to Genitourinary Society Pathology and International Society of Urological Pathology Prostate Cancer Guidelines (Epstein and Kryvenko 2021). All cases were analyzed by two experienced radiologists or pathologist, and their findings were compared. Other experts were involved in case of disagreement.

Statistical analyses
Continuous variables were assessed by the t test. The Chisquare test was used for categorical variables. LASSO regression and cross-validation were used to screen variables. A nomogram model was constructed based on the screened variables and validated. GraphPad Prism (8.0.2) and R (4.1.2) were used for data analysis. P < 0.05 indicated statistical significance.

Descriptive analysis of patients
A total of 304 patients were enrolled. Among them, 217 patients treated from January 2018 to December 2020 were used as the study cohort, and 87 treated from January 2021 to September 2021 were included as an independent validation cohort. The clinical features are shown in Table 1. The ECE detection, sensitivity, and specificity of Mp-MRI in the study cohort are shown in Table 2. TZ patients have a similar incidence of ECE as PZ cases but are more likely to be missed There was no significant difference in the incidence rates of ECE between TZ and PZ patients (PZ VS TZ VS PZ + TZ = 37.19% VS 40.96% VS 38.46%, P = 0.60, P > 0.99, P > 0.99). The missed detection rate was higher in TZ patients compared with PZ cases (PZ VS TZ VS PZ + TZ: 22.22% VS 64.71% VS 40.00%, P < 0.05, P = 0.58 and P = 0.35, respectively). The missed detection rate was not significantly different in the PZ + TZ group compared with PZ and TZ patients, which may be due to the small sample size. So, subsequent analysis did not include the PZ + TZ group. The incidence of PSM of PCa patients with ECE was higher than that of cases without ECE (study cohort, 53.57% (n = 84) VS 12.78% (n = 133); TZ group, 73.53% (n = 34) VS 14.29% (n = 39); PZ group, 37.78% (n = 35) VS 10.53% (n = 76); P < 0.05). Among pECE+ patients, the mECE+ and mECE-subgroups had similar incidence of PSM (study cohort, 52.00% (n = 50) VS 55.88% (n = 34); TZ group, 68.18% (n = 12) VS 83.33% (n = 22); PZ group, 42.86% (n = 35) VS 20.00% (n = 10); P > 0.05). (Fig. 1) Analysis of other various factors revealed the missed detection of ECE in the study cohort can be found in Supplement 1.

Mp-MRI may have a gray area in the detection of ECE in TZ patients
A comparison between mECE+ and mECE− cases in pECE patients of the TZ group showed that the ECE missed detection rate in the TZ group was correlated with smaller LDL, lesion volume, lesion volume ratio, tPSA, PZ volume and PSAD (P < 0.05), but not with NPN, prostate volume, TZ volume, TZ volume ratio, PZ volume ratio, age, ISUP (biopsy) grade, tPSA and mapping of TZ lesions on MRI (P > 0.05) (Fig. 2, Supplement 2).
By comparing factors related to ECE absence and ECE occurrence, it was found that LDL, lesion volume, lesion   Because there were multiple correlations among variables, LASSO regression analysis was performed, with cross-validation to fit the model. With a Lambda of 0.060, LASSO regression and cross validation yielded the optimal solution. Significant coefficients were obtained for LDL (0.161), PCI (1.758), biopsy ISUP grade (0.012), and NPN (0.199). In the study cohort, AUC was 0.938 (P < 0.05) for a sensitivity of 0.864 and a specificity of 0.905, versus 0.913 (P < 0.05), 0.857 and 0.926, respectively, in the validation cohort (Fig. 4).
For the convenience of clinical practice, the variables selected by LASSO regression were used to build a nomogram model. In the study cohort, the AUC, sensitivity and specificity were 0.939 (P < 0.05), 0.864 and 0.905, respectively; in the validation cohort, these values were 0.960 (P < 0.05), 1 and 0.852, respectively (Fig. 4).

Discussion
Assessing ECE in patients before treatment is of great significance for the selection of treatment methods, the quality of life during and after treatment, and postoperative biochemical recurrence. Although Mp-MRI is a powerful tool, its detection rate for ECE is limited (sensitivity of only 0.57) (Rooij et al. 2015). In this study, we found that the missed detection rate of ECE in TZ cases was much higher than that of the PZ group which may be an important reason for the low detection rate of ECE by MP-MRI. Through exploring the causes of missed ECE in TZ patients, we found that Mp-MRI missed more ECE in certain TZ patients, that is, Mp-MRI may have a gray area in the detection of ECE in TZ patients. To address this issue, we developed a clinical risk prediction model for TZ patients, incorporating multi-level clinical variables.
The overall positive rate of pathological ECE in this cohort was 39%, which was consistent with 37-54% reported (Diamand et al. 2021;Gandaglia et al. 2020). The PSMR of the study cohort was 29%, which was almost the same as the reported value of 28% (Rud et al. 2015). The sensitivity and specificity of ECE detection by MP-MRI in this study cohort were 0.59 and 0.83, respectively. These are almost similar to those reported (Rooij et al. 2015), combined with the impact of screening criteria.
In this study, there was no significant difference in ECE incidence between the TZ and PZ groups, which contradicted the previous stereotype. We believe that the low incidence and detection rates of TZ PCa may hamper attention to ECE in TZ patients, although previous findings suggested that some TZ tumors may have a significant risk of ECE, PSM, and biochemical failure (Shannon et al. 2003). In a study by Costa et al., tumor occurrence in TZ or PZ was not related to ECE occurrence, which indicated that TZ and   (Costa et al. 2018). We speculated that ECE in TZ tumors may be related to PCI, because PCI may reflect tumor invasion. In a previous study, it was found that a large TZ tumor would invade the PZ and the anterior fibrous matrix region, which promoted ECE occurrence (McNeal and Haillot 2001). In many studies, TZ and PZ tumors have the same biochemical recurrence rates, and ECE was closely related to biochemical recurrence, which may also suggest that they have the same ECE incidence rates (Chun et al. 2007). The missed detection rate was higher in the TZ group than in the PZ group, which may be related to the difficulty in distinguishing the capsule boundary around the TZ and unobvious ECE on Mp-MRI images. In addition, the neglect of ECE in TZ tumors by imaging professionals may be an important reason. We also found that pECE patients had the same PSMR regardless of whether mECE was detected by MRI or not. In addition, the PSMR of TZ patients with pECE was higher than that of PZ patients with pECE. We believe that this is due to the missed detection of mECE in TZ patients and the inattention of surgeons to TZ tumors.
By comparing the factors related to ECE absence and ECE occurrence, LDL, lesion volume, lesion volume ratio, and tPSA were found to be common. In ROC curve analysis, four cut-off values considered to indicate the boundary value of the possible grey area were identified. The principle of determining the cut-off value is that the upper boundary ensures sufficient specificity while maintaining high sensitivity, and patients with enhanced malignancy but mECEare included as far as possible in the gray area to reduce the possibility of positive margins after surgery. The principle of determining the lower boundary is to ensure sufficient sensitivity to maintain high specificity with as few as possible indicators of low malignancy. PSA ≥ 20 ng/ml was included in the grey area as a commonly used PCa hazard stratification standard, suggesting that the grey area may affect hazard stratification.
To solve the problem that ECE of TZ tumors was largely missed, we constructed a clinical risk prediction model based on LDL, NPN, ISUP level of biopsy, and PCI to predict the risk of ECE in TZ patients with mECE-, where, LDL reflects the size of the tumor on MR images, PCI represents the invasive ability of the tumor on MR scans, NPN means the size of the tumor in the pathological results of biopsy, and ISUP grade indicates the heterogeneity and disorder of the tissue in biopsy pathology.
Preoperative risk stratification is based on multivariate models including clinical variables such as PSA, clinical stage, NPN, and biopsy Gleason score (Tosoian et al. 2017; Memorial Sloan Kettering cancer center. Prostate cancer nomograms pre-radical prostatectom 2022). There were two overlaps between the clinical features of the model and commonly used clinical features, namely NPN and biopsy ISUP grade. ISUP grade reflects tissue heterogeneity in pathological sections, which often reflects the malignant degree of tumors. However, the possible difference between the ISUP level of biopsy pathology and that of postoperative pathology may explain the non-significant correlation between them and the final postoperative ECE findings (Dinh et al. 2015). This defect may depend on the further improvement of the biopsy technology. NPN is the number of positive needles for systematic biopsy, which can also predict the adverse pathological results of PCa (Gandaglia et al. 2019). NPN reflects the tumor size and also is associated with tumor multifocality. Studies have found that up to 80% of patients administered surgery show tumors outside the final targeted lesion (Johnson et al. 2019). Our model did not include PSA levels as a standard for risk stratification, which may be because PSA is significantly correlated with several other variables. In addition, TZ tumors may have higher PSA levels than PZ tumors (Lee et al. 2015). That is to say, the model incorporating PSA levels in the above study is an overall model for PCa, however, for ECE of TZ tumors, PSA levels may not be that significantly affected.
Accumulating studies found that combining Mp-MRI results and clinical features is superior to models based only on clinical parameters in predicting adverse pathological and oncology outcomes (Rayn et al. 2018). In this study, two MP-MRI features, NPN and PCI, were incorporated into our model after the LASSO regression screening of variables. LDL represents the size of a tumor at an MRI angle. Many studies have shown that LDL has a significant correlation with ECE, which is an important feature in predicting tumors before surgery and a useful and repeatable predictor of ECE (Christophe et al. 2020). In addition to LDL, we first incorporated the variable PCI into the prediction model. PCI is an MRI imaging feature described in the PI-RADS guidelines but has not been paid attention to. This may be because TZ tumors and pseudocapsules are often ignored. We believe that PCI may reflect the invasive ability of TZ tumors on Mp-MRI scans. Overall, this is a simple and easy-to-operate model including both imaging and clinical features.
The limitations of this study still deserve attention. First, this was a single-center retrospective study. Although there were some follow-up analyses, large multicenter trials involving different clinical settings are still required to confirm our findings (Tosco et al. 2018). In order to ensure the correspondence between MRI lesions and real lesions, we discarded some patients with PI-RADS 3, which may allow our study to exclude some Pca patients. Secondly, due to the possible learning curve of MP-MRI reading and fusion biopsy, the current results may not apply to all cases (Kasabwala et al. 2019). Whether TZ or PZ is the main origin of a single lesion that penetrates the pseudocapsule and occupies both TZ and PZ remains controversial, but this does not affect the use of the model and the fact that TZ tumors cause ECE. Thirdly, as a high-volume referral medical center, the overall composition of hospitalized patients may be related to the PSA screening of affiliated hospitals, which may lead to changes in the overall composition ratio.

Conclusion
We found that a portion of TZ tumors should also be paid attention to. Most TZ tumors with ECE were missed in clinical practice. Undetected ECE in this portion of TZ tumors because of extracapsular invasion also leads to a high surgical margin positivity rate, thereby affecting patient prognosis. The reason for further exploration is that the detection efficiency of Mp-MRI for ECE caused by TZ tumors may be insufficient, with overt gray areas. Therefore, we preliminarily constructed a simple nomogram to predict the risk of ECE in TZ tumors. Clinicians should be aware that TZ tumors reaching certain conditions may also lead to ECE, especially when they penetrate the pseudocapsule and invade the PZ.
Author contributions XC and WL contributed equally to this work. Conceptualization, XW, YH, JH, and XC; methodology, XW, XC, and WL; data curation, XC, WL, JY, CH, and CZ; formal analysis: XC, WL, JY, YC, and YL; validation, XC, WL, YC, YL, CH and CZ; writing-original draft preparation, all authors; writing-review and editing, all authors; supervision, XW, YH, and JH. All authors have read and agreed to the published version of the manuscript. Data availability The data are available from the corresponding author on reasonable request.

Conflict of interest
The authors report no conflicts of interest in this work.
Ethical approval This study was approved by the Institutional Review Board of First Affiliated Hospital of Soochow University (approval number: NO. 2022 (239)).