The predictive value of multiparametric magnetic resonance imaging in enhancing prostate-specific antigen assessment for prostate cancer: a cross-sectional study

Introduction

Prostate cancer (PCa) poses a serious health concern for men globally, being the second most prevalent cancer and the fifth leading cause of cancer-related fatalities. According to the Global Cancer Observatory (GLOBOCAN) 2020 data, over 1.4 million new cases of PCa were diagnosed worldwide, and over 375,000 deaths were attributed to the disease, highlighting the urgent need for improved diagnostic strategies and risk stratification tools. The diagnosis of PCa typically involves a combination of methods, including prostate biopsy, prostate-specific antigen (PSA) testing, digital rectal examinations (DRE), magnetic resonance imaging (MRI), and routine health screenings. However, traditional diagnostic pathways often lead to overdiagnosis and overtreatment, particularly in men with indolent tumors, while potentially missing clinically significant cancers due to the limitations of systematic biopsies and non-specific biomarkers (1). For low-risk cases, active surveillance (AS) has become the preferred approach. Traditional views suggest that a biopsy should be performed as soon as PCa is suspected. However, with advancements in risk stratification, MRI and functional imaging technologies, and the emergence of biomarkers, the identification and management of PCa have become increasingly precise (2).

Given that PCa often exhibits an indolent course, active surveillance—defined as continuous monitoring of disease progression with the goal of treatment—has become the preferred method for men with less aggressive forms of PCa. This paradigm shift reflects a growing emphasis on risk-adapted management, wherein imaging and molecular tools are increasingly used to differentiate clinically significant cancers that require intervention from indolent tumors that may be safely monitored (3).

Multiparametric magnetic resonance imaging (mp-MRI) incorporates T2-weighted sequences along with various functional imaging techniques, including diffusion-weighted imaging (DWI), dynamic contrast-enhanced imaging (DCEI), and magnetic resonance spectroscopy (MRS) (3). This advanced imaging method has been proven to be an effective approach for diagnosing, localizing, and assessing the risk of PCa by combining T2-weighted images with functional sequences like DWI and DCEI. mp-MRI also shows potential in guiding prostate biopsies, assessing cancer aggressiveness, and detecting local recurrence after treatment (4). The active surveillance period based on mp-MRI is a promising follow-up approach that helps differentiate between indolent and aggressive tumors, addressing the limitations of traditional detection methods like PSA testing and random biopsies, thereby reducing the number of repeat biopsies during active surveillance (5,6). Additionally, the standardized Prostate Imaging Reporting and Data System (PI-RADS) scoring system enhances the clinical value of mp-MRI in PCa. Parameters such as prostate volume, maximum lesion diameter, maximum lesion volume, and total lesion volume assessed via mp-MRI are closely correlated with biopsy results (7,8).

This study aims to construct and validate a predictive model for PCa based on mp-MRI findings in patients clinically suspected of having PCa who underwent standard transrectal ultrasound (TRUS)-guided biopsies. Although mp-MRI and the PI-RADS scoring system have significantly improved the detection and risk stratification of PCa, current tools still face limitations in diagnostic performance and generalizability. By integrating imaging features with routinely available clinical parameters such as PSA and prostate volume, this study seeks to develop an individualized, evidence-based model to enhance pre-biopsy risk assessment and reduce unnecessary invasive procedures, thereby improving clinical decision-making. We present this article in accordance with the TRIPOD reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-aw-851/rc).

Methods

Study design

This study is a cross-sectional analysis involving male patients who provided written informed consent and underwent mp-MRI before a standard TRUS-guided biopsy. The results were graded using the PI-RADS criteria and compared to biopsy outcomes. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was reviewed and approved by the Institutional Review Board of the Ningxiang People’s Hospital (No. 202504). Written informed consent was obtained from all participants or their legal representatives prior to enrollment. A priori sample size calculation was not performed; instead, the study size was determined by the number of consecutive eligible patients who underwent mp-MRI and TRUS-guided biopsy at our institution during the study period.

Study subjects

Inclusion criteria: (I) Individuals with aberrant DRE results and increased PSA values (>4 ng/mL) who are clinically suspected of having PCa; (II) undergoing mp-MRI; (III) PSA levels ≤20 ng/mL and DRE results not suggesting extraprostatic disease. Exclusion criteria included prior prostate biopsies, contraindications to MRI, and inability to obtain pathological tissue. Patients without mp-MRI or without histopathological confirmation of prostate status were excluded from the analysis.During the study period, mp-MRI was not universally performed for all patients with suspected PCa due to institutional resource limitations and variability in referral patterns. At our tertiary care center, mp-MRI was typically reserved for patients with abnormal DRE findings or elevated PSA levels above 4 ng/mL, in accordance with institutional protocols and European Association of Urology (EAU) guidelines. However, the decision to proceed with mp-MRI also depended on urologist discretion and patient factors, such as age, comorbidities, or preference. Additionally, some patients were directly referred for biopsy without MRI due to scheduling constraints or limited availability of imaging slots. These operational and clinical factors contributed to the relatively small number of patients included in this study over the 5-year period. This reflects the real-world practice patterns of our institution and may explain the observed patient flow and selection. In the final analytic dataset, there were no missing values for PSA or mp-MRI-derived variables, so a complete-case analysis was undertaken and no data imputation was performed.

Baseline variables recorded before biopsy included age, serum total PSA level, prostate volume measured on mp-MRI, PI-RADS category, maximum lesion diameter, total lesion volume and lesion density. In patients with multiple lesions classified as PI-RADS ≥3, the maximum lesion diameter was defined as the largest single diameter among all suspicious lesions; the total lesion volume was calculated as the sum of the volumes of all suspicious lesions; and lesion density was derived by dividing the total lesion volume by the prostate volume. For other analytical variables, such as PI-RADS score, prostate volume, and lesion count, patient-level values were used consistently: PI-RADS classification reflected the highest score among all lesions, prostate volume was recorded as a single total measurement, and lesion count represented the total number of PI-RADS ≥3 lesions. These aggregated patient-level values were used in all subsequent statistical analyses to ensure consistency and to represent the overall tumor burden. These variables, together with the presence or absence of a positive mp-MRI (PI-RADS ≥3), were considered as candidate predictors for the multivariable prediction model.

mp-MRI and PI-RADS grading

Multiparametric MRI was performed using a 1.5-T or 3.0-T scanner equipped with a pelvic phased-array coil, with fasting for 4 hours prior to the examination and microenema administered 2 hours before imaging. mp-MRI was conducted in accordance with PI-RADS v2 guidelines, incorporating T2-weighted imaging (T2WI), DWI, and dynamic contrast-enhanced (DCE). For DWI, a 4-minute scan was performed to obtain b-values of 0 (b0), 100 (b1), 1,000 (b2), and 1,500 (b3) s/mm², using three orthogonal encoding directions and a six-fold averaging along each. To reduce distortion, two-fold parallel acceleration was used. B0 was not included in the apparent diffusion coefficient (ADC) mapping. DCEI required acquisition of repeated gradient echo (GE) images before and after gadolinium-based contrast injection. DCE-MRI typically employed T1-weighted GE sequences to ensure cranio-caudal coverage matched that of T2-weighted and DWI images. Using a power injector, a dose of 0.1 mmol/kg of gadolinium contrast agent (Clariscan, 0.5 mmol/mL, GE Healthcare, Chicago, IL, USA) was given at a rate of 3 mL/s. DCE pictures with a temporal resolution of 10 seconds were acquired during a period of 2.5 minutes.

Three radiologists, each with over 5 years of experience in prostate MRI, independently interpreted the results while remaining blinded to PSA levels, medical history, and other imaging or biopsy information. To address inter-reader variability, we assessed agreement using Cohen’s kappa statistics across radiologists for PI-RADS scoring, which demonstrated substantial agreement (κ=0.72), thereby supporting the reliability of mp-MRI interpretations in this study. Based on findings from T2WI, DWI, and DCE-MRI, on a 5-point rating system, the likelihood of PCa was ranked as follows: 1 for very low probability, 2 for low probability, 3 for intermediate or ambiguous probability, 4 for moderate probability, and 5 for very high probability. For patients with multiple lesions, the maximum lesion diameter was defined as the largest single diameter among all suspicious lesions measured manually on axial T2WI images. Prostate volume was calculated using the ellipsoid formula: width × height × length × π/6, based on axial, sagittal, and coronal T2WI images. For cases with multiple lesions, the volume of each lesion was calculated by contouring visible regions on DWI and T2WI images and summing across slices; the sum of all lesion volumes was used for analysis. mp-MRI metrics included prostate volume, number of lesions, maximum lesion diameter, maximum lesion volume, total lesion volume, and PI-RADS scores. Lesion density was calculated as the total lesion volume divided by the prostate volume, regardless of the number of lesions. The highest Gleason score among all positive biopsy cores was used for classification when multiple lesions were present.

Standard TRUS-guided biopsy

Biopsies were performed by experienced operators using standard transrectal techniques (TRUS). Biopsy samples were taken from 10 to 12 cores located in the peripheral zones of the prostate, specifically from the base, mid-gland, and apex regions. Histopathological assessment of biopsy cores was conducted by pathologists with over 10 years of experience in prostate pathology. Gleason scoring was employed to assess malignancy, with a pathological result classified as malignant if the highest Gleason score was ≥7.

The primary outcome for the prediction model was the presence of clinically significant PCa, defined as a highest Gleason score ≥7 in any biopsy core obtained at the standard TRUS-guided biopsy performed after mp-MRI. The biopsy and histopathological assessment therefore provided a single baseline outcome measurement for each participant. Pathologists assessing the biopsy cores were not provided with mp-MRI reports or serum PSA levels when reading the specimens. Although transperineal biopsy is increasingly adopted in contemporary urologic practice due to its reduced infection risk and improved anterior zone sampling, this study exclusively utilized TRUS-guided biopsy to ensure methodological uniformity and facilitate direct comparison with existing literature. Moreover, TRUS remains the predominant technique at our institution during the study period, supporting the practical relevance of the findings within our clinical context.

Statistical analysis

For continuous variables, analyses included means (standard deviations); for categorical variables, they included frequencies and percentages. Analysis of variance (ANOVA) or Welch’s t-test were used to compare continuous variables between groups. If anticipated frequencies were less than 5, Fisher’s exact test was applied for categorical data comparisons; if not, the chi-square test was applied. To evaluate the diagnostic value of various metrics for PCa, receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC) was computed. This allowed for the determination of ideal cut-off values as well as sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) at these cut-off points. After evaluating the relationships between each metric and the result variable using univariate logistic regression analysis, multivariate logistic regression analysis was used to find independent factors that were significantly correlated with the outcome while controlling for potential confounders. Based on minimizing the Akaike information criterion (AIC) value, the optimal set of variables was chosen for the multivariate analytic model using a backward stepwise regression technique. R software (version 4.2.2) was used for statistical analysis and graphical representations.

Model development and performance assessment

We developed a multivariable logistic regression model to predict clinically significant PCa (Gleason score ≥7). All candidate predictors listed above were first evaluated in univariable logistic regression models. To reduce multicollinearity, we examined correlations between variables and excluded PSA density and lesion density because they are derived from PSA and lesion volume, respectively.

Next, we applied least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation to select the most informative predictors. Variables with non-zero coefficients in the optimal LASSO model were entered into a multivariable logistic regression model, yielding the final prediction model. Continuous predictors were retained on their original scales, whereas mp-MRI findings were analysed both as PI-RADS category and as a binary variable (positive vs. negative mp-MRI, defined as PI-RADS ≥3).

Model discrimination was quantified using the AUC with 95% confidence intervals (CIs). For clinical interpretation, we also calculated sensitivity, specificity, PPV and NPV at the optimal cut-off value determined by the Youden index. Apparent model performance was evaluated in the same dataset that was used for model development; no additional internal validation procedures (such as bootstrapping or use of a separate validation sample) and no model updating or recalibration were performed.

Results

From February 2018 to August 2023, we enrolled a total of 258 patients who were clinically suspected of having PCa. After excluding 61 patients who did not undergo mp-MRI and 105 who lacked pathological examinations, we ultimately analyzed a cohort of 192 subjects. Among these patients, 95 (49.5%) had a pathological Gleason score of ≥7, while 108 (56.3%) had a PI-RADS grade of ≥3. Baseline demographic and clinical characteristics of the 192 included patients are summarised in Table 1. The cohort had a mean age of 70 years and a median PSA level of 5.98 ng/mL. There were no missing data for the main predictors or the biopsy outcome in the analytic dataset.

Overall, 192 of the 258 eligible men (74.4%) were included in the final analysis. Among them, 95 patients (49.5%) had clinically significant PCa (Gleason score ≥7) and 97 (50.5%) had a Gleason score <7, while 108 patients (56.3%) had a PI-RADS grade ≥3.

All analyses were performed in this single development dataset; no separate validation cohort was available, and therefore no comparison between development and validation samples could be undertaken.

We subsequently grouped the subjects based on Gleason score and PI-RADS grade. Analysis revealed that patients with a Gleason score <7 exhibited significantly lower levels of PSA (mean difference: −4.2 ng/mL, 95% CI: −5.1 to −3.3), PSA density (mean difference: −0.08, 95% CI: −0.10 to −0.06), maximum lesion diameter (mean difference: −6.7 mm, 95% CI: −8.2 to −5.2), total lesion volume (mean difference: −1.9 cm³, 95% CI: −2.4 to −1.4), and lesion density (mean difference: −0.12, 95% CI: −0.15 to −0.09) compared to those with a Gleason score ≥7.

Similarly, when grouped by PI-RADS grade, patients with a PI-RADS grade ≤2 had significantly lower PSA (mean difference: −3.7 ng/mL, 95% CI: −4.5 to −2.9) and PSA density (mean difference: −0.07, 95% CI: −0.09 to −0.05) than those with a PI-RADS grade ≥3. A comparative overview of clinical data based on Gleason score and PI-RADS classification is provided in Table 2.

All ROC and regression analyses were based on the 192 patients with complete data, among whom 95 (49.5%) had clinically significant PCa (Gleason score ≥7).

To further evaluate the diagnostic value of mp-MRI for PCa, we used a Gleason score of ≥7 as the gold standard. We calculated the sensitivity, specificity, and AUC for PSA, PSA density, maximum lesion diameter, sum of lesion volumes, lesion density, and PI-RADS grade. Among the various diagnostic metrics, PI-RADS grade demonstrated the highest diagnostic value (AUC =0.777, 95% CI: 0.725–0.829), followed by PSA (AUC =0.745, 95% CI: 0.678–0.813). The diagnostic metrics for each indicator are detailed in Table 3, and the corresponding ROC curves are illustrated in Figure 1.

Figure 1 ROC curves of different indicators. AUC, area under the curve; CI, confidence interval; mp-MRI, multiparametric magnetic resonance imaging; PSA, prostate-specific antigen; ROC, receiver operating characteristic.

The combined model including PSA and mp-MRI showed good discrimination with an AUC of 0.854 (95% CI: 0.806–0.902), outperforming mp-MRI alone (AUC 0.777, 95% CI: 0.725–0.829) and PSA alone (AUC 0.745, 95% CI: 0.678–0.813). Sensitivity, specificity, PPV and NPV at the optimal cut-off are reported in Table 3.

Next, we examined the factors influencing PCa based on Gleason score, including age, PSA, prostate volume, maximum lesion diameter, total lesion volume, and mp-MRI findings. To avoid multicollinearity, we excluded PSA density and lesion density from the analysis, as they are secondary calculated metrics. Univariate analysis indicated that elevated PSA levels, larger maximum lesion diameters, and positive mp-MRI results were associated with a higher risk of PCa. Multivariate analysis further confirmed that high PSA levels and positive mp-MRI findings were significant risk factors for PCa. On univariable analysis, higher PSA level, higher PSA density, larger lesion volume, higher lesion density and higher PI-RADS grade were each significantly associated with an increased risk of clinically significant PCa (Table 4).

The regression coefficients and odds ratios for the final multivariable model are presented in Table 4. The predicted probability p of clinically significant PCa for an individual patient can be calculated using the logistic function:

p=1/{1+exp[−(β0+β1×PSA+β2×positivemp−MRI)]}[1]

where β₀, β₁ and β₂ are the regression coefficients shown in Table 4.

In clinical practice, a patient’s serum PSA value and mp-MRI result (coded as positive for PI-RADS ≥3 and negative for PI-RADS ≤2) are entered into this equation to obtain an individual predicted probability of clinically significant PCa. For descriptive purposes we report model performance at the optimal probability cut-off derived from the Youden index, but other cut-offs can be chosen according to the desired balance between sensitivity and specificity.

Given the discrepancies observed between univariate and multivariate logistic regression results, we conducted multicollinearity diagnostics and excluded PSA density and lesion density due to their derived nature and high correlation with PSA and lesion volume, respectively. To further optimize predictor selection, we applied LASSO regression with 10-fold cross-validation (Figure 2). This approach retained two key variables—PSA and mp-MRI—as optimal predictors. The final logistic regression model combining these two factors achieved a high diagnostic performance with an AUC of 0.854 (95% CI: 0.806–0.902), as illustrated in Figure 3.

Figure 2 LASSO path for feature selection. LASSO, least absolute shrinkage and selection operator; mp-MRI, multiparametric magnetic resonance imaging; PSA, prostate-specific antigen.

Figure 3 ROC curves of PSA combined with mp-MRI. AUC, area under the curve; CI, confidence interval; mp-MRI, multiparametric magnetic resonance imaging; PSA, prostate-specific antigen; ROC, receiver operating characteristic.

To explore the relative contribution of each variable, we constructed a random forest classification model. Feature importance ranking based on Mean Decrease Gini revealed that mp-MRI was the most influential predictor (Mean Decrease Gini =42), followed by PSA level (Mean Decrease Gini =35), total lesion volume (Mean Decrease Gini =18), and lesion density (Mean Decrease Gini =16). These results underscore the synergistic value of combining imaging and biochemical markers in the diagnosis of clinically significant PCa (Gleason score ≥7).

No model updating procedures, such as recalibration or addition or removal of predictors, were performed after the initial model had been derived.

Axial multiparametric MRI images of the prostate in a patient with biopsy-proven clinically significant PCa (Figure 4A-4F). These images exemplify the typical mp-MRI appearance of lesions used for risk assessment in our study. The lesion appears as an ill-defined low-signal-intensity area on T2-weighted images and shows marked signal abnormality on fat-suppressed and diffusion-weighted sequences, consistent with restricted diffusion (PI-RADS ≥3). Notably, the lesion demonstrates imaging characteristics suggestive of extracapsular extension (ECE), such as capsular bulging and loss of capsule integrity. Although ECE was not specifically quantified in our analysis, its presence in this representative case highlights the imaging spectrum of high-risk disease within the study cohort.

Figure 4 Representative multiparametric MRI findings in biopsy-proven prostate cancer. Axial multiparametric MRI images of the prostate in a patient with clinically significant, biopsy-proven prostate cancer. (A-F) A focal lesion in the prostate gland on different sequences and at different magnifications. On T2-weighted images, the lesion appears as an ill-defined low-signal-intensity area within the prostate, while on fat-suppressed and diffusion-weighted images it shows conspicuous signal abnormality consistent with restricted diffusion and high suspicion for malignancy (PI-RADS ≥3). Notably, the lesion demonstrates imaging characteristics suggestive of ECE, including capsular bulging and disruption of capsule integrity. These features are often associated with higher-grade disease. Although ECE was not specifically quantified in the present analysis, its presence in representative images reflects the advanced spectrum of disease observed in the study cohort. ECE, extracapsular extension; MRI, magnetic resonance imaging; PI-RADS, Prostate Imaging Reporting and Data System.

Discussion

In this study, we conducted mp-MRI examinations prior to TRUS-guided biopsies in 192 patients suspected of having PCa. Despite the widespread use of mp-MRI and existing stratification systems, there remains variability in diagnostic accuracy and clinical decision-making. Thus, despite the availability of existing mp-MRI-based risk stratification tools, there is a justified need to develop an additional predictive model that integrates imaging and biochemical markers to enhance diagnostic accuracy, address variability in current tools, and improve individualized risk assessment across diverse patient populations. A pathological Gleason score of ≥7 was present in 95 patients (32.5%) and a PI-RADS grade of ≥3 was present in 108 patients (37.0%). A higher risk of PCa was linked to both positive mp-MRI results and high PSA levels, according to multivariate logistic regression analysis. mp-MRI had an AUC of 0.777 (95% CI: 0.725–0.829) with a sensitivity and specificity of 73.7% and 81.7%, respectively, for PCa diagnosis. The sensitivity was 80.0%, the specificity was 74.6%, and the AUC was 0.745 (95% CI: 0.678–0.813) for PSA (cutoff =6.87). When combined, the AUC for diagnosing PCa using positive mp-MRI and PSA was 0.854 (95% CI: 0.806–0.902).

In our study, the sensitivity of PSA (cutoff =6.87) for diagnosing PCa was 80.0%, specificity was 74.6%, and AUC was 0.745 (95% CI: 0.678–0.813), indicating that its diagnostic value is higher than that of PSA density. PSA remains the most clinically valuable and widely accepted tumor marker for detecting PCa (9,10). Previous research has shown that total PSA with a cutoff value of ≤3.14 exhibits 93% sensitivity and 82% specificity, aligning with our findings (11). Lotfi et al. suggested that PSA density is more effective than PSA alone, particularly for patients with PSA levels between 4 and 10 ng/mL, recommending a PSA density cutoff of 0.1 to better detect clinically significant cancer (12). However, in our study, the sensitivity, specificity, and AUC for PSA density were slightly lower than those for PSA.

Within these constraints, our findings indicate that combining serum PSA with mp-MRI information provides substantially better discrimination for clinically significant PCa than either test alone. The strong association between positive mp-MRI findings and high-grade disease observed in our cohort is consistent with previous reports and supports the concept of imaging-based risk stratification in men with elevated PSA. In addition, the model showed a high negative predictive value, suggesting that patients classified as low risk by the model are unlikely to harbour clinically significant cancer and might safely avoid immediate biopsy.

mp-MRI is a crucial method for detecting PCa, and multiple guidelines recommend it as a first-line examination for suspected cases suitable for radical treatment. PCa risk stratification is possible using the PI-RADS score system, which ranks T2WI, DWI, and DCE imaging results on a scale of 1 to 5, serving as a standardized reporting method for mp-MRI. Studies have indicated that higher PI-RADS scores correlate with higher cancer detection rates (13,14). A meta-analysis reported that mp-MRI is a sensitive tool for diagnosing PCa, with an overall AUC of 0.87 (95% CI: 0.84–0.90), although the fact that the included studies were very heterogeneous (15). Incorporating MRI into diagnostic models significantly enhances the accuracy of PCa detection, with a weighted AUC of 0.88 (95% CI: 0.86–0.90), consistent with our findings of 0.854 (95% CI: 0.806–0.902) (16). The EAU guidelines suggest that mp-MRI can reduce the need for targeted repeat biopsies, as mp-MRI-guided biopsies, either via TRUS or MRI, are particularly effective at identifying cancers located in the anterior region of the prostate, which traditional TRUS-guided biopsies may miss (17).

Notably, it has been demonstrated that TRUS biopsies underestimate clinical risk even at the time of diagnosis by underestimating Gleason scores and the local area of cancer (18,19). For men at clinical risk of PCa who have not undergone prior biopsies, utilizing MRI and MRI-targeted biopsies for risk assessment is superior to standard TRUS-guided biopsies (20). A large-scale study involving 1,032 suspected PCa patients reported that, in addition to standard biopsies, targeted fusion prostate biopsies were performed on lesions with a PI-RADS score of 3 or higher. The standard prostate biopsy missed 12 out of 272 clinically significant PCa (4.5%), while targeted fusion biopsies missed 44 cases (16.2%) (21).

This study has several limitations that should be acknowledged. First, it was conducted in a single tertiary centre and included a relatively small sample of 192 men with 95 cases of clinically significant PCa, which may limit the statistical power and the precision of the estimated effects. Second, only patients who underwent both mp-MRI and standard TRUS-guided biopsy were enrolled, which may introduce selection bias and may not fully represent men managed in primary care or those who are not referred for imaging. Third, all predictors were measured at a single time point before biopsy, and temporal changes in PSA levels or lesion characteristics were not considered. Fourth, the prediction model was evaluated only in the development dataset without additional internal or external validation, so the apparent performance is likely to be somewhat optimistic. Finally, several potentially relevant clinical variables, such as detailed DRE findings, family history and novel serum or urinary biomarkers, were not incorporated into the model and might further improve prediction in future work.

Because our model was assessed only in the original development cohort, its transportability to other settings remains uncertain. External validation in independent, preferably multicentre cohorts is therefore required before the model can be recommended for widespread clinical use.

From a clinical perspective, the proposed prediction model could serve as a practical decision-support tool to guide which patients with suspected PCa should proceed to biopsy. By quantifying individual risk based on routinely available variables, it has the potential to reduce unnecessary biopsies and procedure-related complications while still identifying patients with clinically significant disease. Before routine implementation, however, the model should undergo robust external validation and formal evaluation of its clinical utility, for example, using decision-curve analysis or prospective impact studies. Additionally, although our model demonstrated high diagnostic performance, its accuracy and predictive performance may have been influenced by the use of TRUS-guided biopsy as the reference standard. TRUS biopsy is known to underestimate clinical risk by undergrading Gleason scores and missing anterior or small-volume lesions. This limitation may have led to misclassification of some clinically significant cancers as insignificant, potentially affecting model calibration and reducing its generalizability to settings using MRI-targeted or transperineal biopsy approaches. Future studies should assess model performance using more accurate biopsy techniques, such as MRI-targeted or transperineal biopsies, to validate its robustness across different diagnostic standards.

This is especially important given the current lack of universally accepted models that can effectively integrate mp-MRI findings with standard clinical indicators in diverse patient populations. Future research should also explore integration of this model into user-friendly risk calculators or electronic health record systems and examine whether incorporating additional clinical or biomarker information can further enhance its performance.

Conclusions

For men at clinical risk of PCa who have not undergone prior biopsies, pre-biopsy risk assessment using mp-MRI is superior to standard TRUS-guided biopsy.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-aw-851/rc

Data Sharing Statement: Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-aw-851/dss

Peer Review File: Available at https://tau.amegroups.com/article/view/10.21037/tau-2025-aw-851/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tau.amegroups.com/article/view/10.21037/tau-2025-aw-851/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study protocol was reviewed and approved by the Institutional Review Board of the Ningxiang People’s Hospital (No. 202504). Written informed consent was obtained from all participants or their legal representatives prior to enrollment.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Kimura S, Kimura T. Diagnosis and Treatment of Prostate Adenocarcinoma. Cancers (Basel) 2021;13:3660. [Crossref] [PubMed]
2. de Vos II, Luiting HB, Roobol MJ. Active Surveillance for Prostate Cancer: Past, Current, and Future Trends. J Pers Med 2023;13:629. [Crossref] [PubMed]
3. Litwin MS, Tan HJ. The Diagnosis and Treatment of Prostate Cancer: A Review. JAMA 2017;317:2532-42. [Crossref] [PubMed]
4. Yoo S, Kim JK, Jeong IG. Multiparametric magnetic resonance imaging for prostate cancer: A review and update for urologists. Korean J Urol 2015;56:487-97. [Crossref] [PubMed]
5. Durmus T, Baur A, Hamm B. Multiparametric magnetic resonance imaging in the detection of prostate cancer. Aktuelle Urol 2014;45:119-26. [Crossref] [PubMed]
6. Kirlik G, Gullapalli R, D'Souza W, et al. A Supervised Learning Tool for Prostate Cancer Foci Detection and Aggressiveness Identification using Multiparametric magnetic resonance imaging/magnetic resonance spectroscopy imaging. Cancer Inform 2018;17:1176935118786260. [Crossref] [PubMed]
7. Thompson JE, Moses D, Shnier R, et al. Multiparametric magnetic resonance imaging guided diagnostic biopsy detects significant prostate cancer and could reduce unnecessary biopsies and over detection: a prospective study. J Urol 2014;192:67-74. [Crossref] [PubMed]
8. Ueno Y, Tamada T, Bist V, et al. Multiparametric magnetic resonance imaging: Current role in prostate cancer management. Int J Urol 2016;23:550-7. [Crossref] [PubMed]
9. Zou BZ, Wen H, Luo HJ, et al. Value of serum free prostate-specific antigen density in the diagnosis of prostate cancer. Ir J Med Sci 2023;192:2681-7. [Crossref] [PubMed]
10. Yazdani M, Saberi N, Baradaran A, et al. Diagnostic value of total serum/free prostate specific antigen and prostate cancer antigen-3 levels in prostate cancer. Am J Clin Exp Urol 2023;11:414-9.
11. Ochiai A, Okihara K, Kamoi K, et al. Clinical utility of the prostate cancer gene 3 (PCA3) urine assay in Japanese men undergoing prostate biopsy. BJU Int 2013;111:928-33. [Crossref] [PubMed]
12. Lotfi M, Assadsangabi R, Shirazi MAM, et al. Diagnostic value of prostate specific antigen and its density in Iranian men with prostate cancer. Iranian Red Crescent Medical Journal 2009;11:170-5.
13. Hofbauer SL, Maxeiner A, Kittner B, et al. Validation of Prostate Imaging Reporting and Data System Version 2 for the Detection of Prostate Cancer. J Urol 2018;200:767-73. [Crossref] [PubMed]
14. Park KJ, Choi SH, Lee JS, et al. Risk Stratification of Prostate Cancer According to PI-RADS® Version 2 Categories: Meta-Analysis for Prospective Studies. J Urol 2020;204:1141-9. [Crossref] [PubMed]
15. Zhen L, Liu X, Yegang C, et al. Accuracy of multiparametric magnetic resonance imaging for diagnosing prostate Cancer: a systematic review and meta-analysis. BMC Cancer 2019;19:1244. [Crossref] [PubMed]
16. Saatchi M, Khatami F, Mashhadi R, et al. Diagnostic Accuracy of Predictive Models in Prostate Cancer: A Systematic Review and Meta-Analysis. Prostate Cancer 2022;2022:1742789. [Crossref] [PubMed]
17. Cornford P, van den Bergh RCN, Briers E, et al. EAU-EANM-ESTRO-ESUR-ISUP-SIOG Guidelines on Prostate Cancer-2024 Update. Part I: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol 2024;86:148-63. [Crossref] [PubMed]
18. Ahmed HU, El-Shater Bosaily A, Brown LC, et al. Diagnostic accuracy of multi-parametric MRI and TRUS biopsy in prostate cancer (PROMIS): a paired validating confirmatory study. Lancet 2017;389:815-22. [Crossref] [PubMed]
19. Jin BS, Kang SH, Kim DY, et al. Pathological upgrading in prostate cancer patients eligible for active surveillance: Does prostate-specific antigen density matter? Korean J Urol 2015;56:624-9. [Crossref] [PubMed]
20. Kasivisvanathan V, Rannikko AS, Borghi M, et al. MRI-Targeted or Standard Biopsy for Prostate-Cancer Diagnosis. N Engl J Med 2018;378:1767-77. [Crossref] [PubMed]
21. Pepe P, Garufi A, Priolo GD, et al. Is it Time to Perform Only Magnetic Resonance Imaging Targeted Cores? Our Experience with 1,032 Men Who Underwent Prostate Biopsy. J Urol 2018;200:774-8. [Crossref] [PubMed]