2.1 Study population

This study was approved by the Ethics Committee of Taizhou Hospital of Zhejiang Province, affiliated with Wenzhou Medical University (approval number: K20221103). The requirement for informed consent was waived due to the retrospective nature of the analysis. Clinical and examination data were retrospectively reviewed for 606 patients who underwent transrectal ultrasound (TRUS)-guided transperineal prostate biopsy following mpMRI between March 2020 and January 2022. Among them, patients with pre-biopsy PSA levels between 4 and 10 ng/mL were selected, and the age range was limited to 56–80 years. This age criterion was established based on previous research suggesting that PSAMR is more stable within this group [8]. Patients were excluded if they met any of the following criteria: (a) history of prostate biopsy or PCa treatment; (b) biopsy performed more than two months after mpMRI; (c) blurred mpMRI images or non-compliance with PI-RADS version 2 (v2) technical requirements; or (d) incomplete clinical data. Based on these criteria, 208 patients were included for analysis, comprising 62 patients with biopsy-confirmed PCa and 146 patients without PCa. The patient selection process is summarized in Fig. 1.

Fig. 1.Flowchart of the study selection process. TRUS: transrectal ultrasound; PSA: prostate-specific antigen; PI-RADS v2: Prostate Imaging Reporting and Data System version 2; mpMRI: multiparametric magnetic resonance imaging.

2.2 Calculation of related parameters

Height and weight were routinely recorded before performing biopsy. PV was calculated using mpMRI images. Axial T2-weighted imaging (T2WI) was used to measure the maximum transverse and anteroposterior diameters of the prostate, while sagittal T2WI was used to determine the maximum longitudinal diameter. PV (mL) was then calculated using the following formula: PV (mL) = maximum transverse diameter (cm) × maximum anteroposterior diameter (cm) × maximum longitudinal diameter (cm) × 0.52. Body surface area (BSA) was calculated based on the Stevenson formula: BSA (m²) = 0.0061 × height (cm) + 0.0128 × weight (kg) − 0.1529. Plasma volume was calculated using the following formula: plasma volume (L) = BSA (m²) × 1.670. PSAMR was calculated according to the following formula: PSAMR (μg/mL) = PSA (ng/mL) × plasma volume (L)/PV (mL) [8]. Finally, PSAD (ng/mL) was calculated as follows: PSA (ng/mL)/PV (mL).

2.3 MRI acquisition and image analysis

MRI examinations were performed using a 3.0-T system equipped with an 8-channel body phased-array coil (Discovery MR750 3.0T, GE Healthcare, Florence, SC, USA). The scanning protocol followed PI-RADS v2 guidelines. Sequences included axial, coronal and sagittal T2WI, axial diffusion-weighted imaging (DWI), and dynamic contrast-enhanced imaging. The scanning area covered the pelvis, prostate, and seminal vesicles.

The lesions were scored in accordance with PI-RADS v2 criteria by two radiologists who were blinded to the clinical and biopsy results. If discrepancies in the scoring were observed, the radiologists reached a consensus through discussion to establish the final PI-RADS score.

2.4 Prostate biopsy and pathology

All patients underwent TRUS-guided transperineal 12-core systematic biopsy. In addition, 2–3 targeted cores were obtained from suspicious lesions identified on MRI. These procedures were performed by a urologist with more than three years of experience in prostate biopsy. Pathological diagnosis was conducted by a senior pathologist with expertise in prostate pathology. Tissue specimens were evaluated using routine histopathology and immunohistochemistry. The pathological results were classified according to the 2014 International Society of Urological Pathology (ISUP) grading system [17]. Specifically, Gleason score ≤6 was defined as ISUP grade 1, Gleason 3 + 4 as grade 2, Gleason 4 + 3 as grade 3, Gleason 4 + 4, 3 + 5 or 5 + 3 as grade 4, and Gleason 9–10 as grade 5.

2.5 Statistical analysis

Statistical analyses were conducted using SPSS version 25.0 (IBM Corp., Armonk, NY, USA) and MedCalc version 20.022 (MedCalc Software Ltd, Ostend, Belgium). Continuous variables were expressed as mean ± standard deviation if normally distributed or as median (interquartile range) if not. Categorical variables were reported as counts (percentages). Normality was assessed using the Kolmogorov-Smirnov test. Comparisons of normally distributed continuous variables between PCa and non-PCa groups were performed using the t-test, while the Mann-Whitney U test was used for non-normally distributed data. The Chi-square test was used to compare categorical variables. To ensure normality, %fPSA, PSAD and PSAMR were logarithmically transformed before analysis. Univariable and multivariable logistic regression analyses were conducted to evaluate the diagnostic value of these variables in predicting PCa. Model performance was assessed using receiver operating characteristic (ROC) curve analysis and expressed as the area under the curve (AUC). Comparisons between AUCs were performed using the Z-test. The optimal cut-off points for prediction models were determined using the maximum Youden index to calculate sensitivity and specificity. Furthermore, Spearman’s correlation analysis was used to assess the relationship between ISUP grade and the AUC values of each model. A two-tailed p-value < 0.05 was considered statistically significant.

4.1 Clinical implications

Since its introduction in 1979, PSA has been the most widely used tumor marker for PCa, with extensive applications in diagnosis, treatment planning and prognosis assessment. However, the limited specificity of PSA as a screening tool can lead to overdiagnosis and overtreatment [18]. This limitation is particularly evident in men with mildly elevated PSA levels, for whom there is a pressing need for non-invasive diagnostic strategies with improved predictive accuracy. To address this, several PSA-derived parameters, such as %fPSA and PSAD, have been adopted in clinical settings to enhance diagnostic precision. While these parameters have modestly improved the sensitivity and specificity of PSA, the PSAMR has emerged as a more promising alternative, as it accounts for the influence of confounding variables that affect PSA levels [19].

Our findings demonstrate that PSAMR offers significantly better predictive value for PCa among individuals with PSA levels between 4 and 10 ng/mL compared with PSA alone, corroborating the results reported by Hong et al. [20]. Building on these findings, we further compared the diagnostic performance of PSAMR to that of other PSA-related parameters. The AUC for PSAMR was 0.738, which was significantly higher than that of %fPSA, though not statistically different from that of PSAD. Notably, PSAMR achieved a specificity of 70.5% at the optimal diagnostic threshold, which exceeded the corresponding values for both PSAD and %fPSA. This improved specificity is clinically meaningful, as it may contribute to reducing unnecessary prostate biopsies.

Although PSAMR adjusts for both PV and plasma volume, its diagnostic capacity remains limited when used alone. To further enhance predictive performance, we evaluated the combined use of PSAMR and the PI-RADS v2 score. Our results confirmed a high diagnostic value of the PI-RADS v2 score in patients within the PSA gray area, with an AUC of 0.824 and high specificity. These findings are consistent with those of a meta-analysis by Li Zhang et al. [21], which reported a sensitivity and specificity of 0.85 and 0.71, respectively, for PI-RADS v2 in detecting PCa. Similar results were reported by Woo et al. [22]. In a cohort of 235 patients with PSA levels in the gray area, the PI-RADS v2 score (threshold ≥4) yielded an AUC of 0.853, with sensitivity, specificity, positive predictive value, and negative predictive value of 60.0%, 89.7%, 50.0% and 92.9%, respectively [23]. Although the AUC in our study was slightly lower, this may be attributed to differences in mpMRI technical parameters and the level of radiologist experience.

We also analyzed the relationship between prediction models and clinical outcomes using the ISUP grading system. Both PSAMR and PI-RADS v2 scores were significantly associated with ISUP grades, and their combination demonstrated an even stronger correlation with PCa severity. These findings suggest that integrating PSAMR with PI-RADS v2 could enhance risk stratification, support prognosis estimation, and inform individualized treatment decisions. Although emerging markers such as PCa antigen 3 (PCA3) [24] and the Prostate Health Index [25] have shown potential for improving the detection of clinically significant PCa in patients with mildly elevated PSA levels, the high cost of developing and validating these novel biomarkers currently limits their widespread clinical application. Therefore, PSA and its derived parameters, including PSAMR, remain essential and practical tools for PCa screening.

4.2 Clinical practice

In this study, the combination of PSAMR and PI-RADS v2 score demonstrated significantly higher diagnostic performance than either variable alone. This combined model significantly improved the detection rate of PCa among patients with PSA levels in the 4–10 ng/mL range. By applying this predictive model, clinicians can estimate the likelihood of PCa before proceeding with prostate biopsy, thereby facilitating more informed decision-making. Furthermore, Model A, which incorporated both PSAMR and PI-RADS v2 score, achieved superior sensitivity and specificity compared with models that included only one of the parameters. Given that PSAMR is derived from PSA, its predictive accuracy was also compared with that of PSAD to assess its added value as a diagnostic tool.

4.3 Limitations

Several limitations of this study should be acknowledged. First, the study was conducted at a single center using a retrospective design, which may have introduced selection bias. Second, no universally accepted diagnostic threshold for PSAMR currently exists, and assessing its diagnostic performance based solely on sensitivity and specificity at a single ROC curve cut-off may be insufficient. Further prospective studies are needed to validate PSAMR and determine an optimal threshold. Third, our study focused only on subjects in the gray zone of PSA and did not consider that prostate cancer can also develop in subjects with PSA values less than 4 ng/mL. In our study, there were only four patients with PSA values less than 4 ng/mL. It is not only very difficult to further explore the predictive rate of prostate cancer in the normal population using PSAMR, but also to indicate whether PSAMR could be extended to all subjects as a complement of screening. In the future, we can establish a database, including research data on outpatients with PSA values less than 4 ng/mL. Fourth, due to the limited sample size, only a restricted number of variables could be included in the multivariable prediction models. Finally, there is currently no database capturing patients who underwent PSA testing but did not proceed to biopsy. Consequently, it is not possible to determine PSAMR values for this population or to track the proportion of these individuals subsequently diagnosed with PCa during follow-up. Future studies should aim to explore whether PSAMR can be used to identify patients who may safely avoid biopsy or to guide the extension of follow-up intervals.