Development and internal validation of a noninvasive predictive model based on iron and fat measurements for pathologic risk stratification in prostate cancer

Introduction

Prostate cancer (PCa) is a major threat to men’s health, and its epidemiological traits and clinical management strategies have attracted much attention. The latest statistical data from the World Health Organization show that PCa is the second most common cancer and has become the sixth leading cause of cancer-related death in men worldwide (1).

The Gleason score (GS) has been an important cornerstone for histological grading of PCa since its establishment in 1966 (2). However, the traditional GS system has gradually shown limitations in clinical application due to its inability to adequately capture the heterogeneity within prostate tumors. To address this challenge, in 2014, the International Society of Urological Pathology (ISUP) introduced a landmark grading system that categorized GS combinations into grade groups (GGs) ranging from 1 to 5, providing a clearer guide to the degree of differentiation of tumor tissue and patient prognosis (3). Currently, PCa is classified clinically into clinically significant (csPCa, ISUP GG ≥2) and non-clinically significant (ncsPCa, ISUP GG =1) based on pathologic grade. However, the latest cohort study revealed that patients with GS =3+4 (ISUP GG =2) have significantly better survival rates and therapeutic effect than those with GS =4+3 (ISUP GG =3) (4,5). Based on this evidence, we propose an innovative risk stratification model: defining ISUP GG ≤2 [GS ≤7 (3+4)] as the pathological low-risk (PL) group and ISUP GG ≥3 [GS ≥7 (4+3)] as the pathological high-risk (PH) group (3). The multiparametric magnetic resonance imaging (mpMRI)-based PI‑RADS v2.1 assessment is widely used for PCa risk stratification. However, its diagnostic performance varies across studies. A systematic review by Annamalai et al. concluded that although PIRADS v2.1 marginally improved interobserver agreement compared to version 2.0, reader experience continues to be the important determinant of diagnostic accuracy, and the clinical impact of these improvements may be limited (6). Moreover, a head‑to‑head meta‑analysis directly comparing PI‑RADS v2.1 with version 2.0 found that the newer version exhibited lower specificity (62% vs. 66%, P=0.02) (7). These specificity issues are particularly pronounced in the intermediate‑risk population, where the distinction between favorable and unfavorable disease has direct therapeutic implications. Focusing on the transition zone, Agrotis et al. demonstrated that upgrading a PI‑RADS 3 lesion to category 4 detects approximately 37% of clinically significant PCas; however, the falsepositive rate remains as high as 54%, suggesting that more than half of such upgraded lesions are benign or clinically insignificant (8). Therefore, developing a non-invasive method to predict the pathological risk level of PCa is crucial for developing individualized treatment strategies. In recent years, obesity has attracted widespread attention as an independent risk factor for the progression of PCa (9,10). Excessive fat accumulation can interfere with normal insulin, leptin, and free insulin-like growth factor-I (IGF-1) levels. In addition, it can induce chronic inflammation in periprostatic adipose tissue (PPAT), thereby promoting the massive release of inflammatory mediators. These actors combine to establish a microenvironment conducive to the proliferation of neoplasm cells (10-12). Mechanistically, it has been found that tributyltin-induced dysregulation of PPAT leads to enhanced secretion of leptin and C-C motif chemokine ligand 7 (CCL7) from adipocytes, which in turn increases PCa cell proliferation, migration, and survival (13). Specifically, CCL7 promotes PCa cell migration via the CC motif chemokine receptor 3 (CCR3), a process that is enhanced under obese conditions (9,13). Moreover, Shen et al. further demonstrated that tumors induce “browning” of PPAT, characterized by the conversion of white adipose tissue to brown adipocytes (BAT) expressing UCP1, and that BAT-related adipokines promote epithelial-mesenchymal transition (EMT) and invasiveness in human PCa cells (14). Magnetic resonance imaging (MRI)-proton density fat fraction (PDFF) precisely quantifies fat fraction (FF) by considering factors that affect MRI signal intensity (15). This quantitative value indicates the amount of fat within a tissue. The PDFF has been successfully applied to organs including the liver (15), pancreas (16), bone marrow (17), and muscle (18).

Iron is an essential nutrient for tumor proliferation and mitochondrial metabolism. As a key cofactor for multiple metabolic enzymes, iron supports oxidative phosphorylation by participating in the mitochondrial electron transport chain, and serves as a core component of ribonucleotide reductase involved in DNA synthesis, thereby driving tumor cell proliferation (19,20). Consequently, cancer cells require substantially more iron than normal cells and actively increase iron bioavailability through various mechanisms to fuel tumor growth (21). In PCa, this iron dependency is driven by specific molecular dysregulations. PCa cells can significantly upregulate hepcidin expression, and the abnormal secretion of this key regulatory factor in iron metabolism leads to intracellular iron homeostasis imbalance by binding to the cell membrane ferroportin (22). As a key nutrient for tumor development, iron promotes tumor cell proliferation by activating enzymes that regulate androgen receptor transcriptional activity (22). The iron stored in the tissue has a paramagnetic effect, which can significantly change the uniformity of the local magnetic field, thereby affecting the MRI signal. Based on this principle, the R2* map technology can non-invasively quantify the iron content of tissues and has been previously applied to the evaluation of iron deposition in the liver (23), heart (24), and prostate (25). The iron content in the tissue is indirectly reflected by measuring the change in T2* value [R2* (1/s) =1,000/T2* (ms)], providing an important non-invasive means for the diagnosis and monitoring of related diseases (23-25).

Previous studies usually rely on body mass index (BMI) to assess body fat content (26). Although some studies have indicated that PPAT has a strong correlation with PCa progression (27-29), no study has yet examined the pathological risk grading of PCa lesions by directly measuring both the iron and fat content within those lesions. The primary purpose of this study was to noninvasively estimate the iron and fat content in PCa lesions using mpMRI to evaluate the practical value of integrating these two estimations in grading the pathological risk of PCa. Through this innovative approach, we hope to provide a more accurate foundation for the early pathology risk classification and personalized PCa treatment strategies. We present this article in accordance with the TRIPOD reporting checklist (available at https://tau.amegroups.com/article/view/10.21037/tau-2026-0256/rc).

Methods

Participants

This study prospectively enrolled 247 patients with suspected PCa who underwent MRI and radical prostatectomy (RP) surgery in The First Affiliated Hospital of Soochow University between December 2019 and March 2022 (Figure 1). Clinical, pathological, and imaging data were collected, including age, BMI, preoperative prostate-specific antigen (PSA) level, blood lipid level, periprostatic fat thickness (PPFT), prostate volume (PV), and subcutaneous fat thickness (SFT), along with the average apparent diffusion coefficient (ADC), FF and T2* values, Prostate Imaging Reporting and Data System (PI-RADS) scores, and postoperative GS and ISUP GG of lesions. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The First Affiliated Hospital of Soochow University (project number 2024-418). Written informed consent was obtained from all patients undergoing the examination.

Figure 1 Flow diagram of patient selection. BPH, benign prostatic hyperplasia; mpMRI, multiparametric magnetic resonance imaging; MRI, magnetic resonance imaging; PCa, prostate cancer; PH, pathological high-risk; PL, pathological low-risk.

The inclusion criteria were as follows: (I) PCa confirmed by surgery (laparoscopic or transurethral RP) pathology; (II) comprehensive clinical, pathological, and imaging data; (III) no prostate biopsy or treatment before the MRI examination; and (IV) 3 weeks between the MRI examination and RP surgery. The exclusion criteria were as follows: (I) patients with severe invasion of PPAT that could not be measured; (II) MRI images with artifacts affecting image analysis; (III) patients with other diseases affecting iron metabolism (e.g., hepatitis) in addition to PCa; and (IV) patients with urinary catheter placement. Finally, 109 patients with PCa were enrolled, comprising 42 individuals in the PL group and 67 individuals in the PH group.

MRI parameters

Patients were scanned using a 3.0 T MR system (MAGNETOM Skyra, Siemens Healthcare, Erlangen, Germany) with 18- and 32-channel phased array coils. During the scan, the patient lay supine, entered the scanner head first, and the scanning center position was set 2 cm above the pubic symphysis. The 2D non-fat-suppressed turbo spin-echo T2-weighted imaging (T2WI) in axial, coronal, and sagittal planes was performed. The parameters for the axial T2WI were as follows: repetition time/echo time (TR/TE) =6,980/140 ms, field of view (FOV) =200 mm × 200 mm, matrix =384×384, resolution =0.5 mm × 0.5 mm × 3 mm, slice thickness =3 mm, slices =23, slice gap =0 mm, and acquisition time =3 min 22 s.

The axial diffusion-weighted imaging (DWI) scanning parameters were as follows: TR/TE =5,640/96 ms, FOV =220 mm × 220 mm, matrix =130×130, resolution =1.7 mm × 1.7 mm × 3 mm, slice thickness =3 mm, slices =25, slice gap =0 mm, gradient direction =1, and b-values =50, 1,000, 1,500 mm²/s. The acquisition time was 4 min 12 s. ADC maps were generated using DWI images with two b values (0 and 1,000 mm²/s).

The axial gradient echo Q-Dixon scanning parameters were as follows: TR/TE =9/1.12, 2.46, 3.69, 4.92, 6.15, and 7.38 ms; FOV =360 mm × 280 mm; matrix =160×98; slices =25; slice thickness =3 mm; slice gap =0 mm; flip angle =15°; resolution =1.1 mm × 1.1 mm × 3 mm; and acquisition time =16 s.

Axial Q-DIXON and DWI scans were aligned to maximize agreement with T2WI.

Image assessment

Two radiologists (Z.Y. and L.G., with 5 and 10 years of prostate MRI diagnosis experience, respectively) independently completed the imaging analysis in the Radiology Department using the Siemens MR post-processing workstation (Syngo.via). When there was a disagreement in PI-RADS scores or region of interest (ROI) delineation between the two radiologists, the determination was referred to a senior radiologist (L.Y., with 20 years of experience in prostate imaging). The Q-DIXON sequence was fused and aligned to obtain six sets of images, namely in-phase, out-of-phase, water, fat, FF, and T2* maps (30-32).

All suspicious lesions were defined by three radiologists using the PI-RADS (version 2.1) (33). The boundaries of the maximum cross-sectional area of each suspicious lesion were manually outlined as the ROI (excluding necrotic, hemorrhagic, and cystic areas). Lesions within the peripheral zone were primarily delineated using ADC maps, with supplementary information from T2WI and DWI sequences. Conversely, T2WI, augmented by DWI and ADC maps, served as the primary delineating tool for lesions in the transitional zone. Each radiologist repeated the measurements three times and calculated the average as the final result, deriving the mean values of FF, T2*, ADC, and T2 signal intensities at the largest cross-sectional area of the lesion (Figure 2).

Figure 2 Measurement diagrams of fat fraction, T2*, and apparent diffusion coefficient values of prostate cancer lesions with different pathological groups. (A-E) Measurement diagrams of PCa lesions in the high-risk group: 60-year-old patient (preoperative PSA =31.7 ng/mL) with a suspicious lesion in the left transitional zone, PI-RADS score =4, postoperative GS =4+4. (A) A uniform region of low signal intensity is observed on T2-weighted images, with a long diameter of 1.2 cm. (B) A high signal on DWI. (C) A low signal on ADC (ADC value =0.64×10⁻³ mm²/s). (D) Schematic of the prostate fat fraction measurements (FF value =3.4%). (E) Schematic of the prostate T2* measurements (T2* value =35.46 ms). (F-J) Measurement diagrams of PCa lesions in the low-risk group: 72-year-old patient (preoperative PSA =23.2 ng/mL) with a suspicious lesion in the left transitional zone, PI-RADS score =5, postoperative GS =3+4. (F) A uniform region of low signal intensity is observed on T2-weighted images, with a long diameter of 1.7 cm. (G) A high signal on DWI. (H) A low signal on ADC (ADC value =0.865×10⁻³ mm²/s). (I) Schematic of the prostate fat fraction measurements (FF value =1.67%). (J) Schematic of the prostate T2* measurements (T2* value =50.14 ms). ADC, apparent diffusion coefficient; DWI, diffusion-weighted imaging; FF, fat fraction; GS, Gleason score; PCa, prostate cancer; PI-RADS, Prostate Imaging Reporting and Data System; PSA, prostate-specific antigen.

After RP surgery, the surgical specimens were sectioned at 5-mm intervals. The pathologist marked six anatomical locations on the pathological sections, namely the base, apex, peripheral zone, central gland, and urethra. These locations correspond to the positions on the MRI images to precisely localize the lesions. After pathological analysis and GS determination of all lesions, we retrospectively selected the imaging measurements and PI-RADS scores corresponding to the lesions with the highest GS for final analysis. If multiple lesions shared the highest GS, the data of the largest lesion were used for the final evaluation.

To calculate the PV, the product of the largest diameters measured in the axial, coronal, and sagittal planes was multiplied by a scaling factor of 0.52 (34). The measurements of SFT (the shortest distance from the pubic symphysis to the skin) and PPFT (the shortest vertical distance from the pubic symphysis to the prostate on the mid-sagittal image) were approximately based on cross-sectional and sagittal T2-weighted images (29) (Figure 3).

Figure 3 Measurement diagram of SFT and PPFT. (A) Subcutaneous fat thickness: the shortest distance from the pubic symphysis to the skin in axial images (red straight line). (B) Periprostatic fat thickness: the shortest vertical distance from the pubic symphysis to the prostate on the median sagittal plane (red straight line). PPFT, periprostatic fat thickness; SFT, subcutaneous fat thickness.

Collection of pathological data

After prostatectomy, the specimen was cut into 5-mm-thick slices. A pathologist with 11 years of experience marked six anatomical points on the prostate pathological image corresponding to the MRI images, namely the base, apex, peripheral zone, central gland, apex, and urethra. These markings were used to align the image slices with the pathology slices, allowing for the confirmation and sampling of the lesion’s location as described by MRI (35). Two uropathologists (with 11 and 13 years of experience) reviewed the tissue sections to determine the location and boundaries of the lesions and reported the GS and ISUP GG for each lesion according to ISUP guidelines (36).

Statistical analysis

Statistical analyses were performed using SPSS version 25.0 (IBM Corporation, Armonk, NY, USA) and R software (version 4.3.2). The chi-square test was used for counting data, whereas the independent t-test or Mann-Whitney U test was used for continuous data. Spearman correlation analysis was used to evaluate the association between ISUP GG and an array of clinical and imaging variables. Predictive risk factors were selected by screening variables with nonzero coefficients through Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis, thus avoiding multicollinearity among covariates (37). All P values were two-sided, with P values <0.05 considered to indicate statistical significance. Ten-fold cross-validation was implemented to verify the model efficacy. Briefly, the dataset was randomly divided into 10 mutually exclusive subsets, with each partition serving once as the test dataset and nine times as the training dataset, guaranteeing the stability and reliability of the two models across varying datasets. After identifying the optimal LASSO regression model and optimizing the regularization parameter λ based on the Akaike information criterion, we constructed the final improved fusion (IF) model and established a risk factor-based nomogram to predict the probability of high-risk PCa. Internal validation was performed using a bootstrap resampling validation method with 1,000 iterations.

Based on the IF model, a conventional fusion (CF) model, which only retained clinical and traditional imaging variables, was constructed. The goodness-of-fit of the two models was evaluated using the Hosmer-Lemeshow goodness-of-fit test (38). We compared two predictive models using receiver operating characteristic (ROC) curve analysis. The area under the curve (AUC) values, sensitivity and specificity values of the two models were calculated and compared using the DeLong test. The calibration curve was used to evaluate the alignment between the model prediction results and the actual results, and decision curve analysis (DCA) was used to evaluate the clinical applicability of the prediction model.

Results

Clinical data and tumor imaging characteristics of the two groups

A total of 109 patients were included in this study (average age: 71.08±5.99 years), comprising 42 patients in the PL group (ISUP GG ≤2; three patients with GG =1 and 39 patients with GG =2) and 67 patients in the PH group (ISUP GG ≥3; 29 patients with GG =3, 17 patients with GG =4, and 21 patients with GG =5). The PH group had higher preoperative PSA, PPFT, and FF values and PI-RADS scores than the PL group, along with lower T2*, ADC values, and PV. However, no statistically significant differences were observed in BMI (P=0.07), age (P=0.11), SFT (P=0.94), low-density lipoprotein (LDL) (P=0.83), high-density lipoprotein (HDL) (P=0.054), triglycerides (TG) (P=0.63), or total cholesterol (TC) (P=0.39) (Table 1).

Correlation between variables and grade of ISUP

The ISUP grade was significantly correlated with the lesion’s T2* value (r=−0.702), FF value (r=0.670), ADC value (r=−0.410), PSA level (r=0.477), PI-RADS score (r=0.419), PPFT (r=0.286), and age (r=0.192) (all P values <0.05). No significant correlation was found between BMI (r=−0.136, P=0.16), PV (r=−0.181, P=0.06), SFT (r=−0.286, P=0.78), LDL (r=−0.025, P=0.79), HDL (r=0.182, P=0.06), TG (r=0.046, P=0.64), and TC (r=0.036, P=0.71) with ISUP GG.

Construction of the prediction model

A vertical line was drawn at the values chosen through 10-fold cross-validation, allowing us to select the optimal lambda value (37). We selected six traits with nonzero coefficients, namely the PV, PSA, and PI-RADS score and the average T2*, FF, and ADC values of the lesions, to construct the IF model. The EPV was 11.2 (67/6). The CF model was constructed with the clinical and traditional imaging variables only, excluding the FF and T2 values of the lesions (Figure 4; Table 2).

Figure 4 Construction of the IF model. (A) Selection of the optimal lambda value for the LASSO model used 10-fold cross-validation based on minimizing a specific criterion. (B) Creation of coefficient profiles for the logarithmic transformation of the lambda series. The plot includes a vertical line representing the lambda values selected using 10-fold cross-validation that produced six candidates with nonzero coefficients at lambda =0.05. IF, improved fusion; LASSO, Least Absolute Shrinkage and Selection Operator.

Comparison between the IF and CF models

The Hosmer-Lemeshow goodness-of-fit test showed that both models fit the data well (P>0.05). The AUC value of the IF model was significantly higher than that of the CF model (DeLong test: P=0.002), as well as having a higher cutoff value and sensitivity (cutoff value =0.722, sensitivity =0.952, specificity =0.761) (Table 3; Figure 5A).

Figure 5 Performance comparison of the two models. (A) The ROC curve of IF and CF models. (B) The DCA curve of IF and CF models. AUC, area under the curve; CF, conventional fusion; CI, confidence interval; DCA, decision curve analysis; IF, improved fusion; ROC, receiver operating characteristic.

Figure 5B shows the net gain of the two models in distinguishing pathological risk stratification of prostate lesions. The net gain of the IF model was higher than that of the CF model over a wide range of threshold probabilities, suggesting that the IF model has a higher clinical gain in pathological risk stratification of PCa lesions. A nomogram derived from the IF model was created to distinguish between the low- and high-risk groups of PCa (Figure 6A). Furthermore, internal validation was performed using a bootstrap resampling validation method with 1,000 iterations. The resulting calibration curve indicated good agreement between the IF model and the actual probabilities (Figure 6B), with a calibration slope of 0.984 and intercept of 0.01 (Table 4).

Figure 6 IF model visualization nomogram and its calibration curve. (A) The nomogram plot of the IF model. (B) The calibration curve of the IF model. ADC, apparent diffusion coefficient; FF, fat fraction; IF, improved fusion; PI-RADS, Prostate Imaging Reporting and Data System; PSA, prostate-specific antigen; PV, prostate volume.

Discussion

Several studies have previously investigated the utility of the Q-Dixon sequence in PCa, including the biochemical recurrence of PCa (35) and the assessment of differential iron metabolism between patients with benign prostatic hyperplasia and PCa (25). Li et al. (39) measured the FF of PCa lesions, with the results indicating a notable correlation with PCa aggressiveness and pathological risk. However, the heterogeneity of the tumor microenvironment is driven by multiple biophysical processes. Abnormal iron metabolism associated with rapid tumor growth and hypoxia may also play a significant role in tumor aggressiveness. This study overcomes the limitations of analyzing a single fat parameter by combining FF and T2* values as quantitative indicators to explore their combined diagnostic value in PCa risk stratification, which may improve the diagnostic accuracy. The IF model, which incorporates lesion FF and T2* values, along with conventional clinical and imaging parameters, showed encouraging predictive capabilities in this exploratory study. Moreover, the FF and T2* values of lesions were identified as independent risk factors for high-risk PCa in our cohort. DCA further suggested that the IF model may provide net clinical benefit, indicating its potential for predicting high-risk PCa.

Previous studies have mostly focused on identifying csPCa to minimize unnecessary biopsies in patients with low pathological-grade PCa. However, studies have shown that patients with ISUP GG =3 have significantly lower overall survival rates than those with ISUP GG =2 (4). Another study (5) showed that GS =4+3 (ISUP GG =3) was associated with a three-fold higher risk of PCa‑specific mortality compared with GS =3+4 (ISUP GG =2). Based on these findings, patients with ISUP GG ≤2 were categorized into the low-risk group, while those with ISUP GG ≥3 were classified into the high-risk group.

mpMRI has become an indispensable tool in the staging and management of PCa, especially in active surveillance (40,41). PDFF sequences can reliably quantify the fat content level in non-alcoholic fatty liver disease (42) and can effectively distinguish benign from malignant bone lesions (42,43). In this study, we found that the FF value of lesions and PPFT in the PL group exhibited lower magnitudes than those in the PH group, with both parameters demonstrating a statistically significant association with the ISUP GG of the lesions (r=0.670, P=0.005; r=0.286, P=0.02). The result may be due to the increase in inflammatory mediators in the periprostatic tissues caused by PPAT, alongside elevated insulin, IGF-1, and estrogen levels due to excess fat mass, contributing to a microenvironment that promotes PCa cell growth while reducing the apoptosis rate (11,12). PCa also differs in the mode of energy supply, preferring to use fatty acids for energy, which enhances lipid absorption and accumulation within cells, leading to invasive progression of the cancer cells (44-46). Meanwhile, PCa cells can regulate the secretion of tumor driver factors, such as IL-6, IL-8, IL-10, IFN, TNF, and VEGF, from the surrounding adipose tissue, which further promote tumor progression (47). Taken together, these findings highlight why the FF value is an independent risk factor for high-risk PCa. The findings of this study are consistent with those of several previous studies. For example, Randall EC (48) and Andersen MK (49) analyzed lipids in prostatic tissues by mass spectrometry and found increased lipid intensity in both the prostate tumor region and its adjacent normal tissues, and the magnitude of the change correlated well with the GS. Moreover, Xiong et al. (50) measured the adipose characteristics of 901 men before undergoing biopsy, including the PPFT and SFT, and found that increased PPFT was significantly associated with a positive biopsy result. Another study (51) also demonstrated that combining PPFT with PI-RADS v2 scores significantly enhanced the prediction of PCa presence compared with using the PI-RADS v2 alone. In contrast, no statistically significant differences in BMI, SFT, or blood lipid content were observed among the groups, nor were there any notable correlations with ISUP grade. This may be because these factors are primarily associated with subcutaneous peripheral fat, which may not adequately reflect the impact of visceral fat, whereas visceral fat, known for its higher metabolic activity, plays a more crucial role in the progression of PCa (29,52).

R2* mapping is a validated MRI technique based on the Q-Dixon sequence, which can noninvasively quantify the iron content in tissues by detecting the paramagnetic effect of stored iron (31,53,54). In this study, the T2* value in the PH group was significantly lower than that in the PL group and was negatively correlated with the ISUP grade. A previous study (25) found that both serum hepcidin and ferritin levels, along with the ISUP grade of prostate lesions, were negatively correlated with T2* values. Furthermore, we reviewed relevant literature and discovered that PCa cells reduce the ferritin content on the cell membrane by promoting hepcidin expression, thereby leading to an increase in intracellular iron content (15). Therefore, the T2* value may reflect the local iron metabolism of the prostate gland. This study further highlighted the vital role of iron in PCa progression and confirmed that the increase in intracellular iron content not only enhances invasive ability but also facilitates the proliferation of tumor cells (15), providing strong support for the malignant biological behavior of PCa.

The study revealed a significant difference in the ADC values of the lesions between the PL and the PH group, and the ADC value was negatively correlated with the ISUP grade. This phenomenon can be reasonably explained by the histopathological traits of the neoplasm: as the pathological grade increases (ISUP GG ≥3), the tumor cells show a more dense proliferation pattern, and the glandular structure gradually becomes disordered or even completely loses the normal glandular architecture. This pathological alteration leads to a significant narrowing of the extracellular space and impaired cell membrane integrity, thus limiting the Brownian motion of water molecules in the lesion area through a dual mechanism of increased cell density and extracellular matrix viscoelasticity (55). Spadarotto’s study (56) verified the findings of our study in a larger multicenter study with a larger sample size (including 244 patients with pathologically confirmed PCa). The study revealed that not only was the absolute value of ADC negatively correlated with the ISUP GG classification, but the ADC/PSAD ratio adjusted by PSAD also showed a significant negative association. These consistent findings suggest that the ADC value, as a functional imaging parameter, can reflect the microstructural traits of the neoplasm tissue, provide an important basis for establishing a radiomics model of PCa based on mpMRI, and provide a dynamic monitoring index for predicting the aggressiveness of the neoplasm and disease progression.

Our results demonstrated a significant difference in PV between the PL and PH groups. Similarly, a previous study highlighted PV as a key prognostic variable in PCa, associating a smaller PV with poorer pathologic features and higher rates of biochemical recurrence, confirming that a reduction in PV is strongly associated with more aggressive tumor biological behavior (57). From the perspective of pathophysiological mechanisms, androgens regulate the growth and differentiation of prostate tissues, and their abnormal levels may affect the malignant phenotype of tumors through a dual mechanism: first, a smaller prostate volume may reflect the loss of balance of the androgen metabolic microenvironment, and such a defect in hormonal regulation may contribute to the evolution of tumor cells toward de-differentiation; and second, tumors surviving in a relatively androgen-deficient microenvironment tend to have stronger proliferative activity and anti-apoptotic ability and, thus, exhibit higher invasive potentials (58,59). Although this finding was consistent with some previous studies, further evaluation of its clinical utility as a risk predictor is required. However, no significant correlation was observed between PV and ISUP grade in this study, which may be because we did not directly measure the lesion volume. The relationship between tumor load and invasiveness has been assessed by directly measuring tumor volume through three-dimensional pathologic reconstruction or imaging histology methods, or by calculating the tumor volume/PV ratio (tumor/prostate ratio). These parameters have been found to have a strong correlation with tumor pathologic grading and envelope invasion (60,61). This approach will also be the focus of our next study.

Despite its strengths, there are also some limitations in this study. First, the quantitative analysis method based on manually drawing the ROI on a single imaging slice may lead to parameter measurement bias due to the inhomogeneity of iron deposition and fat content inside the lesion. In the future, a multi-plane fusion ROI drawing method needs to be developed to improve the global characterization ability of measurement parameters. Additionally, our study was conducted at a single center using a single scanner and field strength from a specific vendor, which may limit the generalizability of the findings. Finally, the most significant limitation of this study is the lack of independent external validation. Our sample size was relatively small (n=109). The EPV of the IF model was 11.2, which met the minimum recommended threshold of 10 but remained modest. Internal validation using 10‑fold cross‑validation quantifies model optimism but does not prove generalizability to other clinical settings, patient populations, or MRI protocols. Therefore, multicenter prospective studies with external validation cohorts are necessary before our predictive model can be considered for clinical implementation.

Conclusions

In this study, we used the Q-Dixon sequence to noninvasively evaluate the iron and fat contents in PCa lesions. For the first time, these variables were combined with traditional clinical and imaging parameters to establish a prediction model for the pathological risk stratification of PCa, providing a new method for guiding the selection of personalized treatment plans for patients.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tau.amegroups.com/article/view/10.21037/tau-2026-0256/rc

Data Sharing Statement: Available at https://tau.amegroups.com/article/view/10.21037/tau-2026-0256/dss

Peer Review File: Available at https://tau.amegroups.com/article/view/10.21037/tau-2026-0256/prf

Funding: This work was mainly supported by the Key Program of Jiangsu Commission of Health (No. K2023027), the Medicine Plus X Project from Suzhou Medical School of Soochow University (No. ML12203423), a grant from the Infectious and Inflammatory Radiology Committee of Jiangsu Research Hospital Association (No. GY202301), the Jiangsu Province Capability Improvement Project through Science, Technology and Education (Jiangsu Provincial Medical Key Discipline Cultivation Unit, No. JSDW202242), and Suzhou Key Laboratory of Medical Imaging (No. SZS2024032).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tau.amegroups.com/article/view/10.21037/tau-2026-0256/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. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Medical Ethics Committee of The First Affiliated Hospital of Soochow University (project number 2024-418). Written informed consent was obtained from all patients undergoing the examination.

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. Schafer EJ, Laversanne M, Sung H, et al. Recent Patterns and Trends in Global Prostate Cancer Incidence and Mortality: An Update. Eur Urol 2025;87:302-13. [Crossref] [PubMed]
2. 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.
3. Mottet N, Bellmunt J, Bolla M, et al. EAU-ESTRO-SIOG Guidelines on Prostate Cancer. Part 1: Screening, Diagnosis, and Local Treatment with Curative Intent. Eur Urol 2017;71:618-29.
4. Lau WK, Blute ML, Bostwick DG, et al. Prognostic factors for survival of patients with pathological Gleason score 7 prostate cancer: differences in outcome between primary Gleason grades 3 and 4. J Urol 2001;166:1692-7.
5. Stark JR, Perner S, Stampfer MJ, et al. Gleason score and lethal prostate cancer: does 3 + 4 = 4 + 3? J Clin Oncol 2009;27:3459-64. [Crossref] [PubMed]
6. Annamalai A, Fustok JN, Beltran-Perez J, et al. Interobserver Agreement and Accuracy in Interpreting mpMRI of the Prostate: a Systematic Review. Curr Urol Rep 2022;23:1-10. [Crossref] [PubMed]
7. Lee CH, Vellayappan B, Tan CH. Comparison of diagnostic performance and inter-reader agreement between PI-RADS v2.1 and PI-RADS v2: systematic review and meta-analysis. Br J Radiol 2022;95:20210509. [Crossref] [PubMed]
8. Agrotis G, Pooch EP, Marsitopoulos K, et al. Detection rates for prostate cancer using PI-RADS 2.1 upgrading rules in transition zone lesions align with risk assessment categories: a systematic review and meta-analysis. Eur Radiol 2025;35:6454-65. [Crossref] [PubMed]
9. Laurent V, Guérard A, Mazerolles C, et al. Periprostatic adipocytes act as a driving force for prostate cancer progression in obesity. Nat Commun 2016;7:10230. [Crossref] [PubMed]
10. Wilson RL, Taaffe DR, Newton RU, et al. Obesity and prostate cancer: A narrative review. Crit Rev Oncol Hematol 2022;169:103543. [Crossref] [PubMed]
11. Di Sebastiano KM, Pinthus JH, Duivenvoorden WC, et al. Elevated C-Peptides, Abdominal Obesity, and Abnormal Adipokine Profile are Associated With Higher Gleason Scores in Prostate Cancer. Prostate 2017;77:211-21. [Crossref] [PubMed]
12. Hoda MR, Popken G. Mitogenic and anti-apoptotic actions of adipocyte-derived hormone leptin in prostate cancer cells. BJU Int 2008;102:383-8. [Crossref] [PubMed]
13. Feijó M, Casanova CR, Fonseca LRS, et al. Tributyltin-dysregulated periprostatic adipose tissue enhances prostate cell survival and migration, implicating the C-C motif chemokine receptor 3. J Environ Sci (China) 2026;165:225-38. [Crossref] [PubMed]
14. Shen CY, Pan JK, Lin WD, et al. Integrating Magnetic Resonance Chemical Shift Imaging for Localized Prostate Cancer Risk Stratification on the Basis of the Impact of Periprostatic Brown Adipocytes Within Tumor Microenvironment. Ann Surg Oncol 2025;32:6962-73. [Crossref] [PubMed]
15. Tang A, Desai A, Hamilton G, et al. Accuracy of MR imaging-estimated proton density fat fraction for classification of dichotomized histologic steatosis grades in nonalcoholic fatty liver disease. Radiology 2015;274:416-25. [Crossref] [PubMed]
16. Schawkat K, Eshmuminov D, Lenggenhager D, et al. Preoperative Evaluation of Pancreatic Fibrosis and Lipomatosis: Correlation of Magnetic Resonance Findings With Histology Using Magnetization Transfer Imaging and Multigradient Echo Magnetic Resonance Imaging. Invest Radiol 2018;53:720-7. [Crossref] [PubMed]
17. Martel D, Leporq B, Saxena A, et al. 3T chemical shift-encoded MRI: Detection of altered proximal femur marrow adipose tissue composition in glucocorticoid users and validation with magnetic resonance spectroscopy. J Magn Reson Imaging 2019;50:490-6. [Crossref] [PubMed]
18. Schlaeger S, Inhuber S, Rohrmeier A, et al. Association of paraspinal muscle water-fat MRI-based measurements with isometric strength measurements. Eur Radiol 2019;29:599-608. [Crossref] [PubMed]
19. Torti SV, Manz DH, Paul BT, et al. Iron and Cancer. Annu Rev Nutr 2018;38:97-125. [Crossref] [PubMed]
20. Hsu MY, Mina E, Roetto A, et al. Iron: An Essential Element of Cancer Metabolism. Cells 2020;9:2591. [Crossref] [PubMed]
21. Bu X, Wang L. Iron metabolism and the tumor microenvironment: A new perspective on cancer intervention and therapy Int J Mol Med 2025;55:39. (Review). [Crossref] [PubMed]
22. Tesfay L, Clausen KA, Kim JW, et al. Hepcidin regulation in prostate and its disruption in prostate cancer. Cancer Res 2015;75:2254-63. [Crossref] [PubMed]
23. Hankins JS, McCarville MB, Loeffler RB, et al. R2* magnetic resonance imaging of the liver in patients with iron overload. Blood 2009;113:4853-5. [Crossref] [PubMed]
24. Brendel JM, Kratzenstein A, Berger J, et al. T2* map at cardiac MRI reveals incidental hepatic and cardiac iron overload. Diagn Interv Imaging 2023;104:552-9. [Crossref] [PubMed]
25. Tian Z, Li YG, Li GZ, et al. A correlative study of iron metabolism based on q-Dixon MRI in benign prostatic hyperplasia and prostate cancer. Asian J Androl 2022;24:671-4. [Crossref] [PubMed]
26. Discacciati A, Orsini N, Wolk A. Body mass index and incidence of localized and advanced prostate cancer--a dose-response meta-analysis of prospective studies. Ann Oncol 2012;23:1665-71. [Crossref] [PubMed]
27. Altuna-Coy A, Ruiz-Plazas X, Sánchez-Martin S, et al. The lipidomic profile of the tumoral periprostatic adipose tissue reveals alterations in tumor cell's metabolic crosstalk. BMC Med 2022;20:255. [Crossref] [PubMed]
28. Cao H, Wang Y, Zhang D, et al. Periprostatic Adipose Tissue: A New Perspective for Diagnosing and Treating Prostate Cancer. J Cancer 2024;15:204-17. [Crossref] [PubMed]
29. Woo S, Cho JY, Kim SY, et al. Periprostatic fat thickness on MRI: correlation with Gleason score in prostate cancer. AJR Am J Roentgenol 2015;204:W43. [Crossref] [PubMed]
30. Schaapman JJ, Tushuizen ME, Coenraad MJ, et al. Multiparametric MRI in Patients With Nonalcoholic Fatty Liver Disease. J Magn Reson Imaging 2021;53:1623-31. [Crossref] [PubMed]
31. Sheth S, Allen CJ, Farrell DE, et al. Measurement of the liver iron concentration in transfusional iron overload by MRI R2* and by high-transition-temperature superconducting magnetic susceptometry. Clin Imaging 2019;55:65-70. [Crossref] [PubMed]
32. von Falkenhausen M, Meyer C, Lutterbey G, et al. Intra-individual comparison of image contrast in SPIO-enhanced liver MRI at 1.5T and 3.0T. Eur Radiol 2007;17:1256-61. [Crossref] [PubMed]
33. Turkbey B, Rosenkrantz AB, Haider MA, et al. Prostate Imaging Reporting and Data System Version 2.1: 2019 Update of Prostate Imaging Reporting and Data System Version 2. Eur Urol 2019;76:340-51. [Crossref] [PubMed]
34. Buie J, De Riese W, Sharma P. Smaller prostate volume is associated with adverse pathological features and biochemical recurrence after radical prostatectomy. Urol Sci 2019;30:255-61.
35. Zhao Y, Li G, Tian Z, et al. Quantitative Estimation of Iron and Fat Content in Prostate Cancer by Multiparametric MRI and Its Application in Optimizing D'Amico Score. J Magn Reson Imaging 2025;61:2223-33. [Crossref] [PubMed]
36. van Leenders GJLH, van der Kwast TH, Grignon DJ, et al. The 2019 International Society of Urological Pathology (ISUP) Consensus Conference on Grading of Prostatic Carcinoma. Am J Surg Pathol 2020;44:e87-99. [Crossref] [PubMed]
37. Sauerbrei W, Royston P, Binder H. Selection of important variables and determination of functional form for continuous predictors in multivariable model building. Stat Med 2007;26:5512-28. [Crossref] [PubMed]
38. Wang L, Mazaheri Y, Zhang J, et al. Assessment of biologic aggressiveness of prostate cancer: correlation of MR signal intensity with Gleason grade after radical prostatectomy. Radiology 2008;246:168-76. [Crossref] [PubMed]
39. Li G, Ding H, Tian Z, et al. Application of proton density fat fraction imaging in risk stratification of prostate cancer. Transl Androl Urol 2024;13:1878-90. [Crossref] [PubMed]
40. Huch Böni RA, Boner JA, Debatin JF, et al. Optimization of prostate carcinoma staging: comparison of imaging and clinical methods. Clin Radiol 1995;50:593-600. [Crossref] [PubMed]
41. Klotz L. Active surveillance for prostate cancer: for whom? J Clin Oncol 2005;23:8165-9. [Crossref] [PubMed]
42. Kim BK, Bernstein N, Huang DQ, et al. Clinical and histologic factors associated with discordance between steatosis grade derived from histology vs. MRI-PDFF in NAFLD. Aliment Pharmacol Ther 2023;58:229-37.
43. Schmeel FC, Luetkens JA, Wagenhäuser PJ, et al. Proton density fat fraction (PDFF) MRI for differentiation of benign and malignant vertebral lesions. Eur Radiol 2018;28:2397-405. [Crossref] [PubMed]
44. Fontaine A, Bellanger D, Guibon R, et al. Lipophagy and prostate cancer: association with disease aggressiveness and proximity to periprostatic adipose tissue. J Pathol 2021;255:166-76. [Crossref] [PubMed]
45. Sena LA, Denmeade SR. Fatty Acid Synthesis in Prostate Cancer: Vulnerability or Epiphenomenon? Cancer Res 2021;81:4385-93. [Crossref] [PubMed]
46. Yue S, Li J, Lee SY, et al. Cholesteryl ester accumulation induced by PTEN loss and PI3K/AKT activation underlies human prostate cancer aggressiveness. Cell Metab 2014;19:393-406. [Crossref] [PubMed]
47. Islami F, Goding Sauer A, Gapstur SM, et al. Proportion of Cancer Cases Attributable to Excess Body Weight by US State, 2011-2015. JAMA Oncol 2019;5:384-92. [Crossref] [PubMed]
48. Randall EC, Zadra G, Chetta P, et al. Molecular Characterization of Prostate Cancer with Associated Gleason Score Using Mass Spectrometry Imaging. Mol Cancer Res 2019;17:1155-65. [Crossref] [PubMed]
49. Andersen MK, Høiem TS, Claes BSR, et al. Spatial differentiation of metabolism in prostate cancer tissue by MALDI-TOF MSI. Cancer Metab 2021;9:9. [Crossref] [PubMed]
50. Xiong T, Cao F, Zhu G, et al. MRI-measured adipose features as predictive factors for detection of prostate cancer in males undergoing systematic prostate biopsy: a retrospective study based on a Chinese population. Adipocyte 2022;11:653-64. [Crossref] [PubMed]
51. Cao Y, Cao M, Chen Y, et al. The combination of prostate imaging reporting and data system version 2 (PI-RADS v2) and periprostatic fat thickness on multi-parametric MRI to predict the presence of prostate cancer. Oncotarget 2017;8:44040-9. [Crossref] [PubMed]
52. Vieira DFM, Torres CVS, Secaf AF, et al. Comparison of Morphological and Functional MRI Assessments of Periprostatic Fat for Predicting Prostate Cancer Aggressiveness. Int Braz J Urol 2025;51:e20240318. [Crossref] [PubMed]
53. Tang H, Jensen JH, Sammet CL, et al. MR characterization of hepatic storage iron in transfusional iron overload. J Magn Reson Imaging 2014;39:307-16. [Crossref] [PubMed]
54. Verlhac S, Morel M, Bernaudin F, et al. Liver iron overload assessment by MRI R2* relaxometry in highly transfused pediatric patients: an agreement and reproducibility study. Diagn Interv Imaging 2015;96:259-64. [Crossref] [PubMed]
55. Yan X, Ma K, Zhu L, et al. The value of apparent diffusion coefficient values in predicting Gleason grading of low to intermediate-risk prostate cancer. Insights Imaging 2024;15:137. [Crossref] [PubMed]
56. Spadarotto N, Sauck A, Hainc N, et al. Quantitative Evaluation of Apparent Diffusion Coefficient Values, ISUP Grades and Prostate-Specific Antigen Density Values of Potentially Malignant PI-RADS Lesions. Cancers (Basel) 2023;15:5183. [Crossref] [PubMed]
57. Kim DHS, Sonni I, Grogan T, et al. Quantitative 3-T Multiparametric MRI Parameters as Predictors of Aggressive Prostate Cancer. Radiol Imaging Cancer 2025;7:e240011. [Crossref] [PubMed]
58. Freedland SJ, Isaacs WB, Platz EA, et al. Prostate size and risk of high-grade, advanced prostate cancer and biochemical progression after radical prostatectomy: a search database study. J Clin Oncol 2005;23:7546-54. [Crossref] [PubMed]
59. Hong SK, Poon BY, Sjoberg DD, et al. Prostate size and adverse pathologic features in men undergoing radical prostatectomy. Urology 2014;84:153-7. [Crossref] [PubMed]
60. Avolio PP, Hassan T, Addar A, et al. The Role of Tumor Volume Ratio in Predicting Clinically Significant Prostate Cancer on Transperineal Biopsy. J Endourol 2025;39:S47-53. [Crossref] [PubMed]
61. Martorana E, Pirola GM, Scialpi M, et al. Lesion volume predicts prostate cancer risk and aggressiveness: validation of its value alone and matched with prostate imaging reporting and data system score. BJU Int 2017;120:92-103. [Crossref] [PubMed]