Prostate health index enhances prostate cancer management in Chinese populations: evidence from the nation’s large cohort and transnational multicohort harmonization

Abstract

Background

The escalating global burden of prostate cancer (PCa) demands accurate diagnostics to optimize management. Conventional prostate-specific antigen (PSA) testing exhibits low specificity and overdiagnosis, compromising its reliability. While the Prostate Health Index (PHI) may enhance precision, large-scale Asian clinical evidence and systematic evaluations of PHI’s standalone/combined diagnostic efficacy remain limited, hindering its broader clinical adoption.

Methods

A retrospective analysis was conducted on hospitalized patients initially diagnosed with PCa or BPH at Wuhan Tongji Hospital over the past three years. Receiver operating characteristic (ROC) curves assessed PHI’s diagnostic performance across clinical scenarios. Binary diagnostic tests simulating real-world clinical settings were performed to assess the practical utility of PSA, PHI, and PI-RADS scores. Furthermore, a multiparameter pre-biopsy diagnostic model integrating PHI with complementary methods was developed. Finally, a systematic review and meta-analysis of PHI-related studies were conducted to comprehensively assess its diagnostic accuracy and robustness.

Results

The study included 2,091 patients, forming the largest Asian cohort for PHI-based diagnosis. ROC analysis revealed that PHI significantly outperformed PSA in diagnosing both PCa and clinically significant PCa (csPCa) (P < 0.001), with enhanced diagnostic superiority in the PSA gray zone (4–10 ng/mL). At a cutoff value of 30, PHI achieved 90% sensitivity for csPCa while maintaining superior performance to PSA in binary diagnostic testing. A risk prediction model integrating PHI with PI-RADS scores was constructed through regression analysis, effectively reducing unnecessary biopsy referrals. Meta-analysis further confirmed the robust diagnostic performance of PHI and its combined use with PI-RADS scores, with better performance observed in Asian populations compared to Western countries.

Conclusion

This large-scale cohort study and updated meta-analysis validate PHI’s clinical utility for PCa diagnosis in Asian populations. The integration of PHI with PI-RADS scores demonstrates significant clinical benefit, positioning PHI as a precise, reliable tool to advance PCa care in Asia.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-15577-3.

Background

The escalating incidence of prostate cancer (PCa) has emerged as a critical global health concern for men, particularly in Asia where epidemiological trends demonstrate accelerated progression [1–3]. Traditional screening protocols and biopsy indications predominantly rely on digital rectal examination (DRE) and prostate-specific antigen (PSA) testing [4, 5]. While operationally convenient, these methods are diagnostically suboptimal. DRE is highly subjective, has poor patient acceptance, and often misses early-stage lesions. Meanwhile, PSA’s limited specificity frequently leads to overdiagnosis and overtreatment [6, 7]. Consequently, developing more reliable diagnostic or risk stratification strategies is imperative.

The traditional PSA test measures total PSA (tPSA), whose serum concentration correlates positively with malignancy risk. Refinements of this approach include free PSA (fPSA), unbound to plasma proteins and inversely associated with prostate cancer [8], and [-2]proPSA (p2PSA), a proteolytically stable precursor isoform with enhanced prostate specificity [9–11]. The Prostate Health Index (PHI), a multivariable algorithm integrating tPSA, fPSA, and p2PSA, demonstrates superior diagnostic accuracy to tPSA alone [12, 13]. Approved for men aged ≥ 50 years with serum PSA 4.0–10.0 ng/mL and negative DRE findings, PHI aims to reduce unnecessary prostate biopsies [14, 15].

Numerous studies have investigated PHI’s diagnostic performance, either independently or in combination with other strategies [16–19]. However, evidence supporting PHI’s clinical utility in Asian populations remains scarce, particularly lacking large-scale cohort data, which hampers its regional adoption [20, 21]. Furthermore, comprehensive syntheses of existing evidence and systematic evaluations of PHI’s combinatorial diagnostic potential are notably absent.

Here, we present the latest findings from a large single-center Chinese cohort study evaluating PHI’s diagnostic value and its synergistic potential with PI-RADS scores. Additionally, we conduct an updated meta-analysis to consolidate existing evidence and assess PHI’s clinical applicability across diverse scenarios.

Methods

Study cohort

This study included patients with suspected PCa or benign prostatic hyperplasia (BPH) who were initially evaluated at Tongji Hospital in Wuhan between January 2022 and December 2024 (Fig. 1). Upon admission, the patients first underwent PHI testing (including tPSA and fPSA%), followed by needle biopsy the next day. Indications for biopsy included: suspicious lesions identified on MRI or PSMA PET/CT, tPSA > 10 ng/mL, or tPSA > 4 ng/mL accompanied by fPSA% < 0.16 and/or PSAD > 0.15 and/or prostate-specific antigen velocity (PSAV) > 0.75 µg/(L·year) and/or long-term typical clinical symptoms such as dysuria, or other abnormal prostate-related test results. The biopsy procedure consisted of transperineal ultrasound-guided systematic 12-core sampling combined with MRI‑fusion targeted biopsy. Patients lacking definitive pathological diagnoses, concurrent infections, or other malignancies were excluded. Demographic and clinical data, including age, tPSA, fPSA percentage (fPSA%), p2PSA, PHI, Prostate Imaging-Reporting and Data System (PI-RADS) scores, prostate volume, and Gleason scores, were collected. Based on Gleason scores, patients with PCa were stratified into two groups: (1) PCa (Gleason score ≥ 6, with BPH as controls) and (2) csPCa (Gleason score > 6, with non-csPCa as controls).

Fig. 1.

Flowchart of the study cohort construction. PHI: prostate health index; PSA: prostate-specific antigen; MRI: magnetic resonance imaging; PI-RADS: Prostate Imaging Reporting and Data System

Meta analysis

The systematic review and meta-analysis was registered with PROSPERO on April 5, 2025 (https://www.crd.york.ac.uk/PROSPERO/view/CRD420251024921.). The work process was carried out and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline (Supplementary methods) [22].

Statistical analysis

Baseline characteristics were compared using the Mann-Whitney U test (continuous variables) and Fisher’s exact test (categorical variables). Prostate volume was calculated as 0.52 × (anteroposterior diameter × transverse diameter × longitudinal diameter). PSA density (PSAD) was calculated as tPSA/ prostate volume. ROC curve analysis was performed in SPSS 26 to determine the maximum Youden index and the corresponding cut-off values. Univariate and multivariate regression analyses were conducted using SPSS 26. For the multivariate regression, variables with a p-value < 0.05 were considered statistically significant and were included as the final covariates. AUC comparisons utilized DeLong’s test. Regression modeling and decision curve analysis (DCA) were conducted using R software (version 4.4.3; packages: stats for regression, rmda for DCA).

Results

A total of 2,091 patients were included in this study, among whom 724 (35% of the cohort) were pathologically diagnosed with PCa via biopsy. Of these, 607 cases (29% of the cohort) exhibited a pathological score ≥ 3 + 4, defined as csPCa. Additionally, 667 patients had PSA levels within the “gray zone” (4–10 ng/mL) (Table 1). Overall, compared to non-cancer patients, patients with PCa demonstrated significantly elevated levels of tPSA, p2PSA, and PHI (33.36 in BPH vs. 135.62 in PCa), alongside significantly reduced fPSA% (Table S1). A consistent trend was observed in the csPCa subgroup, where the mean PHI was 34.44 in non-csPCa group versus 152.71 in csPCa group (Table S2). Among the cohort, 726 patients underwent magnetic resonance imaging (MRI). PCa group exhibited higher PSAD and PI-RADS scores compared to BPH group, with similar findings in the csPCa subgroup (Tables S3–4).

Table 1.

Patients’ characteristics

	Total group	Gray-zone subgroup	MRI subgroup
Patients (n)	2091	667	726
PCa, n (%)	724 (35)	186 (28)	263 (36)
csPCa, n (%)	607 (29)	132 (20)	228 (31)
Median age, yr (IQR)	67.93 (62.10–73.30)	67.37 (61.62–72.42)	68.39 (62.48–73.74)
Median tPSA, ng/ml (IQR)	14.91 (3.70–15.70)	6.76 (5.29–8.08)	19.07 (6.14–20.42)
Median fPSA, % (IQR)	19.02 (10.94–24.40)	18.81 (12.57–23.80)	16.93 (9.77–21.86)
Median p2PSA (IQR)	47.04 (8.36–31.18)	18.51 (10.05–22.27)	56.30 (11.48–37.78)
Median PHI score (IQR)	68.77 (22.60–64.20)	40.42 (24.60–48.00)	75.84 (25.68–76.22)
Median PSAD (IQR)			0.42 (0.11–0.44)
PI-RADS
1, n (%)			8 (1)
2, n (%)			244 (34)
3, n (%)			212 (29)
4, n (%)			98 (14)
5, n (%)			164 (23)

Clinically, the diagnostic utility of conventional tPSA is limited in the gray zone (4–10 ng/mL). In our cohort, 27.90% of gray-zone patients were diagnosed with PCa, and 19.8% with csPCa (Table S5). Table 2 and S6 summarize PCa detection rates across PHI and PI-RADS categories. Notably, PHI > 55 or PI-RADS ≥ 4 was associated with a sharply increased probability of PCa and csPCa.

Table 2.

PCa and CsPCa with different PHI cut-off values

	Cutoff				Total	p
	25	25–35	35–55	>55
PCa	66/646	49/365	121/444	488/636	724/2091	<0.01
	10.20%	13.40%	27.30%	76.70%	34.60%
csPCa	41/646	30/365	81/444	455/636	607/2091	<0.01
	6.30%	8.20%	18.20%	71.50%	29.00%

The area under the curve (AUC) was used to evaluate the diagnostic performance of PSA-derived markers for PCa and csPCa. Compared to tPSA (AUC = 0.75), fPSA% exhibited inferior diagnostic accuracy for PCa (AUC = 0.73), while p2PSA (AUC = 0.78) and PHI (AUC = 0.85) demonstrated superior performance (Fig. 2a). Statistical analyses confirmed that PHI significantly outperformed other PSA-related markers in discriminating PCa from non-cancer cases (P < 0.001) (Fig. 2b). At an optimal cutoff of 45.55 (maximizing Youden’s index), PHI achieved a sensitivity of 76.66% and specificity of 83.10% (Figure S1a). For csPCa detection, PHI also exhibited the highest diagnostic efficacy (AUC = 0.88), surpassing tPSA (AUC = 0.78), fPSA% (AUC = 0.74), and p2PSA (AUC = 0.81) (Fig. 2c and d). At a cutoff of 47.15, PHI yielded a sensitivity of 0.81 and specificity of 0.83 for csPCa diagnosis (Figure S1b). These results indicate that PHI holds greater potential than PSA for precise identification of PCa and csPCa.

Fig. 2.

(a-b) ROC curves (a) and corresponding AUC values (b) for age, tPSA, fPSA%, p2PSA, and PHI in predicting PCa; (c-d) ROC curves (c) and corresponding AUC values (d) for age, tPSA, fPSA%, p2PSA, and PHI in predicting csPCa. PSA: prostate-specific antigen; p2PSA: [-2]proPSA; PHI: prostate health index; PI-RADS: Prostate Imaging Reporting and Data System. 95% CI: 95% confidence intervals. ****, p < 0.0001

Further analyses explored PHI’s diagnostic value in different clinical scenarios. In the gray-zone subgroup (Group 2), tPSA showed poor predictive utility for PCa and csPCa (AUC < 0.6) (Fig. 3a and b). PHI retained robust diagnostic performance, with AUCs of 0.78 for PCa and csPCa. In the MRI subgroup (Group 3), PI-RADS and PSAD were analyzed. PI-RADS demonstrated the second-highest diagnostic accuracy after PHI (AUCs: 0.86 for PCa, 0.88 for csPCa), closely followed by PSAD. (Fig. 3c and d).

Fig. 3.

(a-b) ROC curves for age, tPSA, fPSA%, p2PSA, and PHI in predicting PCa (a) and csPCa (b) within the PSA gray zone (4–10 ng/mL); (c-d) ROC curves for age, tPSA, fPSA%, p2PSA, PHI, PI-RADS score, and PSAD in predicting PCa (c) and csPCa (d). The AUCs and their corresponding 95% confidence intervals are presented in parentheses. PSA: prostate-specific antigen; p2PSA: [-2]proPSA; PHI: prostate health index; PI-RADS: Prostate Imaging Reporting and Data System

Given clinical reliance on absolute thresholds, binary diagnostic tests were conducted using predefined cutoffs: PSA > 10 ng/mL and PI-RADS ≥ 3 as positive indicators. Table 3 details PHI’s sensitivity and specificity for PCa/csPCa prediction at varying thresholds in Group 1. Consistent with prior Asian population studies, a PHI cutoff of 30 achieved ~ 90% sensitivity for PCa/csPCa detection. Thus, PHI cutoff was set at 30. Comparative analysis in Group 1 revealed that PHI outperformed PSA in sensitivity, negative predictive value, and Youden’s index but had lower specificity and accuracy (Fig. 4a-c). In Group 3, PI-RADS surpassed both PSA and PHI, exhibiting higher Youden’s index, sensitivity, and accuracy (Fig. 4d-f).

Table 3.

Sensitivity and specificity of PCa and CsPCa prediction at different PHI cutoffs

Group		Cutoff
		25	30	35	40	45	50	55
PCa	Se	91%	88%	84%	81%	77%	71%	67%
	Sp	42%	56%	66%	75%	82%	87%	89%
csPCa	Se	93%	90%	88%	86%	82%	78%	75%
	Sp	41%	54%	63%	73%	80%	85%	88%

Fig. 4.

(a-b) Sensitivity, specificity, positive likelihood ratio (LR+), negative likelihood ratio (LR−), positive predictive value (PPV), and negative predictive value (NPV) for PSA > 10 ng/mL and PHI > 30 in predicting PCa (a) and csPCa (b). c Youden’s index and accuracy of PSA and PHI in predicting PCa and csPCa. d-e Sensitivity, specificity, LR+, LR−, PPV, and NPV for PSA > 10 ng/mL, PHI > 30, and PI-RADS ≥ 3 in predicting PCa (d) and csPCa (e). f Youden’s index and accuracy of PSA, PHI, and PI-RADS in predicting PCa and csPCa. PSA: prostate-specific antigen; p2PSA: [-2]proPSA; PHI: prostate health index; PI-RADS: Prostate Imaging Reporting and Data System

Multivariate regression identified PHI and PI-RADS as the strongest independent predictors of PCa/csPCa (P < 0.001) (Figure S2). A predictive model integrating PHI and PI-RADS was developed using Group 3 data (263 PCa and 463 BPH cases), split into training (n = 582, 211 PCa) and validation (n = 144, 52 PCa) cohorts (Fig. 5a). The calibration curve indicated good model calibration (Fig. 5b), and the combined model achieved AUC > 0.90 in both cohorts (Figure S3). Similarly, a csPCa prediction model integrating PHI with PI-RADS scores was developed using an analogous methodology, demonstrating comparable diagnostic accuracy (Fig. 5c-d and S4). DCA demonstrated superior net clinical benefit for the PHI-PI-RADS model compared to tPSA, PHI, or PI-RADS alone, in guiding biopsy decisions (Fig. 5e and f). These findings highlight the potential of PHI combined with PI-RADS as a precise diagnostic tool for PCa/csPCa, with high clinical utility for biopsy guidance. Furthermore, the diagnostic performance of PHI was investigated within different PI-RADS score intervals (Figure S5 and Table S7). ROC analysis revealed that in the low PI-RADS score interval (< 3), PHI achieved AUCs of 0.70 for PCa and 0.76 for csPCa, with a corresponding cut-off value of 59.40. The diagnostic performance of PHI improved with increasing PI-RADS scores. In the PI-RADS 3 subgroup, using a cut-off value of 56.60, PHI reached AUCs of 0.81 for PCa and 0.82 for csPCa. These results further demonstrate the complementary diagnostic value of PHI to PI-RADS scores.

Fig. 5.

a-b Nomogram (a) and calibration curve (b) of the PCa predictive model combining the PHI and PI-RADS scores; (c-d) Nomogram (c) and calibration curve (d) of the csPCa predictive model combining the PHI and PI-RADS scores; (e-f) DCA comparing clinical benefit of tPSA, PHI, PI-RADS, and the combined PHI + PI-RADS model as biopsy indications for PCa (e) and csPCa (f). PSA: prostate-specific antigen; PHI: prostate health index; PIRADS: Prostate Imaging Reporting and Data System; DCA: Decision curve analysis

In order to further investigate the clinical value of PHI, we conducted a systematic review and meta-analysis of existing relevant clinical studies (Figure S6). A total of 102 articles were identified for meta-analysis (Table S8). Data from the Tongji Prostate Cancer Early Screening Cohort were integrated, resulting in a pooled analysis of 85 cohorts (n = 29,193) for PCa outcomes and 66 cohorts (n = 24,331) for csPCa outcomes. Funnel plot asymmetry analysis suggested publication bias within the csPCa group (p = 0.04) (Figure S7), and the bivariate boxplot analysis revealed a significant correlation between sensitivity and specificity in both PCa and csPCa group, suggesting the presence of a threshold effect (Figure S8). To address threshold effects and inter-study heterogeneity, the hierarchical summary receiver operating characteristic (HSROC) model was selected to more accurately evaluate the overall performance of the diagnostic test.

For PCa, the pooled sensitivity and specificity were 0.77 (0.73–0.80) and 0.63 (0.58–0.67), respectively, as shown in the forest plot (Figure S9). The HSROC model showed a pooled sensitivity and specificity of 0.77 (0.73–0.80) and 0.63 (0.58–0.67), respectively (Figure S10a). In the PSA level ≤ 10 ng/ml group, the pooled sensitivity and specificity were 0.76 (0.71–0.81) and 0.63 (0.56–0.68) (Figure S10b). For csPCa, as shown in the forest plot, the pooled sensitivity and specificity were 0.82 (0.79–0.85) and 0.62 (0.57–0.66), respectively (Figure S11). The HSROC model showed a pooled sensitivity and specificity of 0.82 (0.79–0.85) and 0.62 (0.57–0.66), respectively (Fig. 6a). In the PSA level ≤ 10 ng/ml group, the pooled sensitivity and specificity were 0.81 (0.75–0.86) and 0.59 (0.51–0.66) (Fig. 6b). To mitigate potential bias from the large sample size of this cohort, a supplementary meta‑analysis restricted to external cohorts was performed. Consistency with the overall meta‑analysis was observed (Figure S12‑14), suggesting the reliability of our cohort findings and providing further robust data supporting the practical value of PHI.

Fig. 6.

HSROC model-based assessment of PHI’s diagnostic performance for csPCa (a) and gray-zone csPCa (b), alongside PHI-MRI combined model performance for PCa (c) and csPCa (d). Se: Sensitivity; Sp: Specificity

Subgroup analyses by continent showed consistent trends for both PCa and csPCa. Diagnostic efficacy varied geographically, with the highest performance observed in Asian populations (Youden index: 0.42 for PCa and 0.48 for csPCa), followed by European (0.38 for PCa and 0.41 for csPCa) and North American populations (0.30 for PCa and 0.33 for csPCa) (Figure S15 and S16).

Furthermore, seven cohorts reported the diagnostic value of PHI with MRI for PCa and eight cohorts for csPCa. The pooled sensitivity and specificity were 0.82 (0.77–0.86) and 0.82 (0.76–0.87) for PCa, 0.86 (0.82–0.90) and 0.75 (0.60–0.86) for csPCa (Fig. 6c and d). This combination demonstrated exceptional diagnostic performance, which can significantly reducing unnecessary biopsies.

Discussion

PSA testing plays the most important role in the early screening of prostate cancer [23]. However, in clinical practice, there are still many prostate cancer patients with PSA levels less than 10 ng/ml [24, 25]. PHI was introduced to reduce overdiagnosis in patients within the diagnostic gray zone [26]. Currently, PHI lacks an established consensus on its diagnostic threshold, and previous studies have often been limited by small sample sizes. Therefore, investigating the diagnostic performance of PHI in a large-scale cohort holds substantial clinical significance.

This study represents the largest single-center experimental evaluation of diagnostic modalities for prostate cancer involving 2,091 patients. In the Tongji Prostate Cancer Early Screening Cohort, 724 patients were confirmed to have prostate cancer through biopsy and 607 were classified as csPCa. PHI demonstrated superior diagnostic accuracy compared to PSA, with an AUC of 0.85 versus 0.75 for PCa and 0.88 versus 0.78 for csPCa. In the gray zone, PHI maintained robust diagnostic accuracy, demonstrating an AUC of 0.78, outperforming conventional PSA testing (AUC = 0.54). The result from this cohort supported the evidence that PHI was a reliable biomarker for patients to reduce unnecessary biopsies [26, 27]. Nevertheless, it should be noted that the relatively high proportion of patients with PSA > 10 ng/ml in this cohort may call for due consideration when interpreting the value of PHI across a wide PSA spectrum.

PHI currently lacks standardized diagnostic thresholds due to its heterogeneous performance across diverse populations [18, 28]. This cohort study provides evidence-based thresholds for Chinese populations: 45.55 (Se: 0.77, Sp: 0.83) for initial PCa detection and 47.15 (Se: 0.81, Sp: 0.83) for csPCa. Notably, lower thresholds reached the best Youden index in the PSA diagnostic gray zone (4–10 ng/mL), with 39.95 (Se: 0.73, Sp: 0.73) for PCa and 40.10 (Se: 0.77, Sp: 0.69) for csPCa.

MRI is recommended for prostate cancer screening due to its diagnostic accuracy [29, 30]. However, it shares a key limitation with traditional PSA testing: in patients with PIRADS ≤ 3 lesions, biopsy decisions rely on subjective clinical judgment and patient preference, leading to unnecessary procedures [31]. This study demonstrates that combining PHI with MRI significantly improves diagnostic performance to reduce unnecessary biopsies. In the validation group (n = 144), the sensitivity and specificity of this multi-model method for predicting csPCa were 0.84 and 0.86. 92 biopsies could be avoided while only 7 (15.56%) patients with csPCa were missed.

In the meta-analysis, the diagnostic efficacy of PHI was synthesized for PCa and csPCa, which remained high in the gray zone. We found a difference across each continent that PHI performs the best in the Asia population, then in the European population, the worst in North America. The public awareness of prostate cancer and popularization of PSA testing make more Americans detect prostate cancer at an early stage, which makes it harder to diagnose [32]. The combination of PHI and MRI presented a highly sensitive and specific effect in diagnosing PCa and csPCa [20]. Thus, the multi-model method showed a promising future in clinical practice.

As the threshold effects exists, the sensitivity is negatively correlated with the specificity. The cutoff setting may directly influence on the diagnostic effect of PCa and csPCa. Our study provided a threshold that was suitable for patient populations in Central China. Furthermore, we proposed a multi-model diagnostic strategy for PCa and csPCa. Given that both PHI and MRI are recognized as valuable tools within the PSA gray zone, this study specifically demonstrated that their combination yields a superior and more interpretable diagnostic performance.

Notably, biopsy-free radical prostatectomy is being explored clinically to eliminate diagnostic biopsies for high-risk patients with significant serological and imaging abnormalities, thereby reducing treatment burden and complications [33, 34]. Current biopsy-free strategies mainly rely on expert interpretation of PSA, MRI and PET-CT findings, which is high cost and may miss early-stage or clinically subtle cases [35, 36]. Our findings demonstrate that the combined PHI-PI-RADS model substantially improves diagnostic reliability for csPCa at a risk threshold exceeding 4%. This enhancement in pre-biopsy risk stratification provides a stronger quantitative foundation to inform clinical decision-making within evolving diagnostic pathways.

Furthermore, while existing PHI-MRI diagnostic models are feasible and reliable, they require integrating parameters from serological (e.g., PHI) and imaging (e.g., PI-RADS) sources, complicating data harmonization and introducing potential errors. Recent efforts aim to identify blood-based biomarkers for streamlined, precise PCa detection [37–39]. On this basis, future integration of PHI with novel liquid biopsy markers may enable a unified diagnostic platform, enhancing clinical practicality and diagnostic accuracy.

Several limitations of this study should be acknowledged. First, the retrospective design and restriction to biopsy-proven cases may introduce selection bias and limit generalizability to a broader screening population, potentially affecting the estimated diagnostic performance of PHI and PI-RADS. Nevertheless, within our single-center pathway, the use of standardized biopsy indications, a dedicated expert team, and a large sample size may enhance the internal consistency of the results. Second, DRE was excluded from the multivariable model given its subjective and operator-dependent assessment, which contrasts with our focus on quantitative biomarkers. Nonetheless, its exclusion may be a comparative limitation. Also, the PI-RADS assessments, though conducted by experienced radiologists, are subject to inter-reader variability. Finally, although we present one of the largest PHI cohorts from China, prospective, multi-center studies with standardized protocols are warranted to further elevate the level of evidence, and conclusions from the meta-analysis should be interpreted with caution due to potential publication bias.

Conclusion

In summary, the study provides the largest Chinese cohort-based evaluation of the PHI to date, confirming its robust diagnostic performance across diverse PCa diagnostic scenarios, particularly within the PSA gray zone. Meanwhile, the predictive model integrating PHI with MRI demonstrates clear clinical benefit in guiding biopsy decision-making. Furthermore, the updated meta-analysis, the largest to date on PHI, further validates its diagnostic accuracy. This facilitates the standardization of PHI thresholds for Chinese and broader Asian populations, further validating its clinical utility in PCa diagnosis across these demographics.

Abbreviations

AUC

Area under the curve

BPH

Benign prostatic hyperplasia

csPCa

Clinically significant PCa

DCA

Decision curve analysis

DRE

Digital rectal examination

fPSA

Free PSA

fPSA%

FPSA percentage

HSROC

Hierarchical summary receiver operating characteristic

MRI

Magnetic resonance imaging

p2PSA

[-2]proPSA

PCa

Prostate cancer

PHI

Prostate Health Index

PI-RADS

Prostate Imaging-Reporting and Data System

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-analyses

PSA

Prostate-specific antigen

PSAD

PSA density

ROC

Receiver operating characteristic

Se

Sensitivity

Sp

Specificity

tPSA

Total PSA

Authors’ contributions

Conceptualization: X.Z., Q.X., S.W.; Formal analysis: X.Z., Y.X., F.Y.; Investigation: Y.Y., X.G.; Project administration: Q.X., S.W.; Writing – original draft: X.Z., Y.X., F.Y.; Writing – review & editing: Q.X., S.W.

Funding

This research is supported by grants from Hubei Urology Minimally Invasive Therapy Clinical Medical Center (SCZ2023017).

Data availability

All data generated or analysed during this study are included in this published article and its supplementary information files.

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the Declaration of Helsinki. Informed consent and approval were obtained from all the patients and the Ethics Committee of Tongji Hospital (TJ-IRB202407023).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.