Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Diagnostic Accuracy of Predictive Models in Prostate Cancer: A Systematic Review and Meta-Analysis

Diagnostic Accuracy of Predictive Models in Prostate Cancer: A Systematic Review and Meta-Analysis Hindawi Prostate Cancer Volume 2022, Article ID 1742789, 10 pages https://doi.org/10.1155/2022/1742789 Review Article Diagnostic Accuracy of Predictive Models in Prostate Cancer: A Systematic Review and Meta-Analysis 1,2 1 1 1 Mohammad Saatchi , Fatemeh Khatami , Rahil Mashhadi , Akram Mirzaei , 1 1 1 Leila Zareian , Zeinab Ahadi , and Seyed Mohammad Kazem Aghamir Urology Research Center, Tehran University of Medical Sciences, Tehran, Iran Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran Correspondence should be addressed to Seyed Mohammad Kazem Aghamir; mkaghamir@tums.ac.ir Received 10 January 2022; Revised 17 May 2022; Accepted 20 May 2022; Published 8 June 2022 Academic Editor: Cristina Magi-Galluzzi Copyright © 2022 Mohammad Saatchi et al. �is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Aim. Accurate diagnosis of prostate cancer (PCa) has a fundamental role in clinical and patient care. Recent advances in diagnostic testing and marker lead to standardized interpretation and increased prescription by clinicians to improve the detection of clinically signi‡cant PCa and select patients who strictly require targeted biopsies. Methods. In this study, we present a systematic review of the overall diagnostic accuracy of each testing panel regarding the panel details. In this meta-analysis, using a structured search, Web of Science and PubMed databases were searched up to 23 September 2019 with no restrictions and ‡lters. �e study’s outcome was the AUC and 95% con‡dence interval of prediction models. �is index was reported as an overall and based on the WHO region and models with/without MRI. Results. �e thirteen ‡nal articles included 25,691 people. �e overall AUC and 95% CI in thirteen studies were 0.78 and 95% CI: 0.73–0.82. �e weighted average AUC in the countries of the Americas region was 0.73 (95% CI: 0.70–0.75), and in European countries, it was 0.80 (95% CI: 0.72–0.88). In four studies with MRI, the average weighted AUC was 0.88 (95% CI: 0.86–0.90), while in other articles where MRI was not a parameter in the diagnostic model, the mean AUC was 0.73 (95% CI: 0.70–0.76). Conclusions. �e present study’s ‡ndings showed that MRI signi‡cantly improved the detection accuracy of prostate cancer and had the highest discrimination to distinguish candidates for biopsy. So, new PCa biomarkers have been proposed to im- 1. Introduction prove the accuracy of PSA in the management of early PCa Prostate cancer (PCa) is the second most frequent cancer in [5–8]. �e diagnostic panels include PSA isoforms, PSA men worldwide, and its incidence and mortality correlate density and velocity, age-adjusted PSA, free PSA to total with increasing age [1]. �e accurate PCa diagnosis is a PSA ratio (fPSA/tPSA), PSA density (PSAd), PSA doubling problematic issue because it is essential to identify which time (PSADT), Prostate health index (Phi), 4K score (in- PCa are destined to progress and which would bene‡t from clude kallikrein-related peptidase 2/hK2, intact PSA, fPSA, early radical treatment [2]. PCa has traditionally been di- and tPSA), advanced MRI (mpMRI and bpMRI), PCA3 agnosed by digital rectal examination (DRE) and prostate- mRNA, PSA glycoforms, TMPRSS2:ERG fusion gene, speci‡c antigen (PSA) blood test, followed by transrectal microRNAs, circulating tumor cells (CTCs), and androgen ultrasound (TRUS) guided biopsy [3]. �eir limited speci- receptor variants [5, 9–12]. AUC is an e§ective way to ‡city and an elevated rate of overdiagnosis are the main summarize the overall diagnostic accuracy of each testing problems associated with PCa testing. Benign prostatic panel. hyperplasia (BPH) has similar symptoms to PCa, and most No comprehensive study represents the most accurate PCa patients are diagnosed as asymptomatic patients with ones, and the heterogeneity of all clinical trials is too high in normal DRE and elevated PSA [4]. both the panel components and AUC. �is systematic 2 Prostate Cancer review summarizes all PCa diagnostic panels and compares 3. Results their AUC to find the most accurate ones. In this systematic review, 4188 articles were identified, of which 4185 articles were extracted from the search of 2. Methods electronic databases, and three articles were extracted from the search of the list of selected articles and other sources. )is systematic review and meta-analysis were designed After deleting duplicate articles, the title and abstract of 3228 according to the latest version of the PRISMA checklist, and articles were screened, and according to the exclusion cri- it was registered on Prospero with registration number: CRD teria, 3186 articles were removed. Finally, 13 articles were 149417. )e summary major was AUC with a 95% confi- used in the final analysis (Figure 1). dence interval (CI). )e thirteen final articles included 25,691 people. )e characteristics of the studies include the names of the 2.1. Search Strategy. We searched Scopus, Web of Science, authors, the country, the WHO region, sample size, mean and PubMed databases on 23 September 2019. )e search or median age, AUC and 95% confidence interval, model query was as follows: “Prostate Neoplasms” OR “Prostatic parameters, quality assessment score, and model name. Neoplasm” OR “Prostate Cancer” OR “Prostatic Cancer” Based on the findings of our study, the highest AUC was AND “ risk score” OR “prognostic score” OR “prognostic observed in the study of Boesen et al. [14] (0.89 model” OR “prognostic panel” OR “prognostic score (0.87–0.92)) and Dwivedi et al. [15] (0.89 (0.83–0.95)). In model”. Duplicate studies were removed prior to download. both studies, MRI played an important role in increasing After that, we included articles with these inclusion criteria: AUC. In the study of Roobol et al. [16], the lowest AUC articles that provided a model/panel for prostate cancer was observed in the GOTEBORG-R2–6 cohort and PSA prediction. Exclusion criteria included the following: (1) DRE-model, which included only PSA, DRE, and Prior articles that investigate genetic factors, (2) articles that did biopsy (Table 1). not report AUC (with 95% confidence interval) for their In the final analysis, most articles are from European and model, and (3) articles studying the treatment, recurrence, or American regions. As shown in Figure 2, the weighted metastasis of prostate cancer. average AUC in the countries of the American region was 0.73 (95% CI: 0.70–0.75), and in European countries, it was 0.80 (95% CI: 0.72–0.88). A study from Southeast Asia and a 2.2. Data Collection. )ree reviewers, RM, AM, and LZ were study from the Asia-Pacific region were also in the final independently involved in the title and abstract and read and analysis. )e overall AUC and 95% CI in thirteen studies was determined the eligibility of the studies. All three authors, 0.78 (95% CI: 0.73–0.82). Data from the previously pub- RM, AM, and LZ, independently extracted all relevant data, lished meta-analysis indicated that PI-RADS are superior in including the year of publication, first author, country, diagnosing PCa with high sensitivity, specificity, and AUC sample size, mean/median or range of age, AUC (95% CI), than PHI and PCA3 [22]. and model contents. )e disagreement was resolved by Figure 3 shows the AUC of studies based on the presence discussion, and when necessary, two reviewers (SMKA and or absence of MRI in the final model. In four studies with FKh) assisted in adjudicating a final decision. MRI, the average weighted of AUCs was 0.88 (95% CI: 0.86–0.90), while in other articles where MRI was not a parameter in the diagnostic model, the mean AUCs were 2.3. Methodological Quality Assessment. )e New- 0.73 (95% CI: 0.70–0.76). castle–Ottawa Scale (NOS) assessment tool was used to Figure 4 shows the funnel plot to investigate publication evaluate the quality of the articles by three authors [1]. )e bias. )e Begg (P value � 0.428) and Egger (P value � 0.780) scoring was based on the assignment of stars from 1 to 9. tests showed no significant publication bias in our study. According to the NOS score, the selected studies were di- vided into high quality (≥6) and low quality (<6). 4. Discussion 2.4. Statistical Analysis. )e chi-square test at a significant )e present study assessed the predictive models for PCa level of 5% was used for the qualitative assessment of het- detection to find the models that had the highest discrim- erogeneity across studies. Based on the Higgins categori- ination in distinguishing candidates for biopsy. In the zation, an I-square of more than 75% was considered current study, the highest AUCs were observed for two heterogeneity. )e index of interest in this study was AUC models; one of them is based on age, PSA density, DRE, and which was calculated as the proportion using the ROC curve bpMRI (AUC: 0.89, 95% CI: 0.87–0.92) [23], and the second method with 95% confidence intervals. )e weights for the one developed with PSA, MRSI, and DW-MRI (mpMRI) weighted average AUC calculation were calculated in ac- (AUC: 0.89, 95% CI: 0.83–0.95) [14]. )e present study’s cordance with the methods described by Zhou et al. [13]. findings showed that the best predictive models for PC Data analysis was performed using the Stata version 11 detection were based on the combination of clinical pa- (StataCorp, College Station, TX, USA) statistical software. rameters and bpMRI or mpMRI. By adding the MRI to Also, the random effect model at a confidence level of 95% clinical parameters, the predictive accuracy improved sig- was used in the data analysis. nificantly. Also, the AUCs of most models based on only Prostate Cancer 3 Number of Records identified Number of Additional records through database searching identified through other sources (n = 4185) (n = 3) Number of Records after duplicates removed No. of Records excluded (n = 3186)with reasons: (n = 3228) Genetic studies (n=1806); Studies that have examined the treatment, recurrence and No. of Records screened metastasis of the disease (n = 3228) (n=1114); Studies without AUC with 95% confidence interval (n=224); Book or book chapter (n=6); Conference paper (n=12); review article (n=19); No. of full-text articles assessed for eligibility Letter to the editor (n=5) (n = 42) No. of full-textarticles excluded (Inadequate No. of Studies included in information provided) meta-analysis (n=29) (n = 13) Figure 1: Flow of information through di§erent steps of the systematic review and meta-analysis. MRI, along with clinical parameters, can be utilized to clinical variables were lower than the AUCs of models with the incorporation of imaging [16–21, 24, 25]. decrease the number of unnecessary biopsies. Otherwise, MRI can ensure information about cancer location, staging, Previous documents assessing the e¨cacy of prostate cancer detection have highlighted the need to decrease in- and the volume for target biopsies. At present, both the signi‡cant prostate cancer’s overdiagnosis [26–28]. Hence, a American and European associations of urology (AUA and novel diagnostic panel is required to decrease the number of EAU) recommend using mpMRI as a useful diagnostic tool unneeded biopsies and recognition of insigni‡cant prostate before repeat biopsy and for men enrolled in active sur- cancer. So, recently numerous nomograms and predictive veillance [32, 33]. A recent systematic review reported models with various parameters, varying degrees of accu- clinically signi‡cant disease detection rates, the sensitivity, racy, generalizability, and validation were developed to and the negative predictive value (NPV) of mpMRI ranged improve the accuracy of PC diagnosis. Recently, the evidence from 44 to 87%, 58–97%, and 63% to 98%, respectively. In 2022, Futterer et al. had shown that mpMRI could be applied showed that when mpMRI or bpMRI is added to the standard clinical factors, the predictive accuracy enhances to rule out signi‡cant disease because of its extraordinary NPV [34]. �e use of radiomics and kallikreins failed to [14, 23, 29, 30]. A meta-analysis showed that bpMRI o§ers similar test accuracy to mpMRI in identifying prostate outperform PI-RADSv2.1/IMPROD bpMRI Likert, and cancer, but heterogeneity does not allow de‡nitive recom- their combination did not lead to further performance gains. mendations to be made [31]. Boesen et al. [14] showed that �e high expenses of mpMRI are debating using the mpMRI by adding bpMRI to clinical parameters (age, PSAd, ctDRE), to detect prostate cancer. Despite the high cost of mpMRI at the AUC of the model improved signi‡cantly from 0.85 to ‡rst look, it is generally considered a cost-e§ective method in 0.89 for predicting PC and achieved the highest discrimi- PC diagnosis because it reduces unnecessary biopsies costs, nation power. Also, they showed that the AUC of the model prevents unnecessary therapies, and increases the quality of based on the only bpMRI was 0.84 and demonstrated that life in the long term [35, 36]. �e prostate-speci‡c mem- brane antigen (PSMA) PET/CT and mpMRI have compa- the MRI-derived score as a PC detection is the most pow- erful single predictor. In line with this, Dwivedi et al. [23] rable diagnostic accuracy in the discovery and intraprostatic found that the model’s accuracy is higher with mpMRI than localization of prostate cancer foci whereas mpMRI makes without (0.89 vs. 0.66). van Leeuwen et al. showed that the better in the assessment of extracapsular extension (ECE) addition of mpMRI to commonly used clinical elements and seminal vesicle invasion (SVI) [37]. However, the ad- enhanced the predictive accuracy by 9% [29]. As a result, vantage of systematic biopsy (SBx) added to combined MRI/ Included Eligibility Screening Identification 4 Prostate Cancer Table 1: Baseline characteristics for studies included in meta-analysis. Mean age/ Author WHO Sample ID Country range/ AUC Model content Score Model name Name region size median Total PSA 0.69 Family history (0.65–0.74) DRE Free PSA Prostate Cancer Prevention Ankerst et 1 USA Americas 575 63.4 Total PSA 7 Trial Risk Calculator al. [17] Family history (PCPTRC) model 0.64 DRE (0.65–0.74) Free PSA [-2] Pro PSA Age 0.89 PSA density Advanced imaging model (0.87–0.92) cTDRE bpMRI 0.78 PSA Boesen et Baseline model 2 Denmark Europe 876 65 (0.75–0.82) cTDRE 8 al. [14] 0.84 bpMRI Imaging model (0.81–0.86) Age 0.85 PSA density Advanced model (0.83–0.88) cTDRE Age ADC 0.66 (NA) Original PSA Metabolic ratio Age ADC 0.78 (NA) PSA Original Metabolic ratio DW-MRI Dwivedi et Southeast Age 3 India 137 65 9 al. [15] Asia ADC 0.83 (NA) PSA Original Metabolic ratio MRSI Age ADC 0.89 PSA Developed (0.83–0.95) Metabolic ratio model mpMRI (MRSI + DW-MRI) Age at biopsy Abnormality on DRE Family history 0.71 Previous negative Predicting PHI (0.64–0.77) biopsy Total PSA Foley et al. Free PSA 4 Ireland Europe 250 63.7 7 [18] p2PSA Age at biopsy Abnormality on DRE 0.62 Family history Predicting PSA (0.55–0.69) Previous negative biopsy PSA Prostate Cancer 5 Table 1: Continued. Mean age/ Author WHO Sample ID Country range/ AUC Model content Score Model name Name region size median Age Family history Ethnicity Urinary voiding Sunnybrook nomogram- 0.67 Symptom score based prostate cancer risk (0.65–0.69) DRE calculator (SRC) Nam et al. Median age PSA 5 Canada Americas 2130 8 [19] 63 free: total PSA ratio Age Family history Prostate Cancer Prevention 0.61 Ethnicity Trial (PCPT)-based risk (0.59–0.64) DRE calculator (PRC) PSA Age BMI Diabetes status Smoking status Emphysema Multi parameterized Roffman et 0.73 Asthma 6 USA Americas 1672 67 9 artificial neural network al. [20] (0.71–0.75) Race (ANN) Ethnicity Hypertension Heart disease Exercise habits History of stroke Age Low-risk PSA (class via PCa DRE) DRE-model 0.70 Abnormal DRE (0.68–0.72) Roobol et Prostate volume 7 Netherland Europe 3580 68 9 al. [21] PSA Low-risk Age PCa Abnormal DRE TRUS model 0.73 Prostate volume (0.70–0.75) Abnormal TRUS 6 Prostate Cancer Table 1: Continued. Mean age/ Author WHO Sample ID Country range/ AUC Model content Score Model name Name region size median PSA GOTEBORG-R1 Median age 0.77 DRE 740 cohort 61 (0.73–0.81) Prostate volume DRE vol-RC model Prior biopsy PSA GOTEBORG-R1 Median age 0.71 740 DRE cohort 61 (0.67–0.76) Prior biopsy PSA DRE-model PSA GOTEBORG-R2–6 Median age 0.60 DRE 1241 cohort 63 (0.57–0.64) Prostate volume DRE vol-RC model Prior Biopsy PSA Median age 0.56 GOTEBORG-R2–6 cohort 1241 DRE 63 (0.52–0.60) PSA DRE-model Prior biopsy PSA DRE ROTTERDAM-R1 Median age 0.74 2895 Prostate volume cohort 66 (0.72–0.79) Prior biopsy DRE vol-RC model Family history PSA DRE ROTTERDAM-R2-3 Median age 0.65 1494 Prostate volume cohort 67 (0.62–0.69) Prior biopsy DRE vol-RC model Family history Roobol et PSA 8 Netherlands Europe 8 ROTTERDAM-R2-3 al. [16] Median age 0.60 DRE 1494 cohort 67 (0.57–0.63) Prior biopsy PSA DRE-model Family history PSA DRE Prostate volume Median age 0.66 Biopsy Gleason CCF cohort 64 (0.64–0.68) grade DRE vol-RC model Family history African origin Prior biopsy PSA DRE Biopsy Gleason Median age 0.62 CCF cohort 2631 grade 64 (0.60–0.64) PSA DRE-model Family history African origin Prior biopsy PSA Median age 0.72 DRE Tyrol cohort 63 (0.70–0.73) Prostate volume DRE vol-RC model Prior biopsy PSA Median age 0.67 Tyrol cohort 4199 DRE 63 (0.65–0.69) PSA DRE-model Prior biopsy Abbreviations. PSA: prostate-specific antigen, DRE: digital rectal examination, PCPTRC: prostate cancer prevention trial risk calculator, PRC: prostate cancer prevention trial (PCPT)-based risk calculator, ANN: artificial neural network, TRUS: transrectal ultrasound, DW-MRI: diffusion-weighted magnetic resonance imaging, BMI: body mass index, SRC: Sunnybrook nomogram–based prostate cancer risk calculator, MRSI: magnetic resonance spectroscopic imaging, ADC: apparent diffusion coefficients, and PHI: prostate health index. ultrasound fusion targeted biopsy (TBx) is mainly limited to model was developed with age, PSAd, DRE, prostate volume, smaller PI-RADS score 3–4 lesions [38]. and PSA [39]. )e determining PSAd requires an accurate In the current study, the highest AUC of the model assessment of prostate volume, and in this model, prostate developed based on only clinical variables was 0.83, and this volume was estimated using transrectal ultrasound (TRUS) Prostate Cancer 7 Study ID ES (95% CI) %Weight Americas Ankerst 2012 0.69 (0.65, 0.74) 7.57 Mehralivand 2018 0.84 (0.79, 0.89) 7.42 Mam 2011 0.67 (0.65, 0.69) 8.09 Roffman 2018 0.73 (0.71, 0.75) 8.09 Roobol 2017 0.73 (0.70, 0.75) 8.02 Vickers 2011 0.75 (0.72, 0.77) 8.02 Vickers 2013 0.71 (0.68, 0.74) 7.92 Subtotal (I = 89.2%, p = 0.000) 0.73 (0.70, 0.76) 55.13 Europe Boesen 2019 0.89 (0.87, 0.92) 8.02 Foley 2016 0.71 (0.64, 0.77) 6.95 Roobol 2012 0.77 (0.73, 0.81) 7.70 Stojadinovi 2019 0.83 (0.77, 0.88) 7.27 Subtotal (I = 93.0%, p = 0.000) 0.80 (0.72, 0.88) 29.94 South-east asia Dwivedi 2017 0.89 (0.83, 0.95) 7.12 Subtotal (I = .%, p = .) 0.89 (0.83, 0.95) 7.12 Western Pacific van Leeuwen 2017 0.88 (0.84, 0.91) 7.82 Subtotal (I = .%, p = .) 0.88 (0.85, 0.92) 7.82 Overall (I = 96.0%, p = 0.000) 0.78 (0.73, 0.82) 100.00 NOTE: Weights are from random effects analysis -0.95 0 0.95 Figure 2: Forest plot of AUC (95% CI) of predictive models in the di§erent regions. Study ID ES (95% CI) %Weight without MRI Ankerst 2012 0.69 (0.65, 0.74) 7.57 Foley 2016 0.84 (0.64, 0.77) 6.95 Nam 2011 0.67 (0.65, 0.69) 8.09 Roffman 2018 0.73 (0.71, 0.75) 8.09 Roobol 2012 0.77 (0.73, 0.81) 7.70 Roobol 2017 0.75 (0.70, 0.77) 8.02 Stojadinovi 2019 0.83 (0.77, 0.88) 7.27 Vickers 2011 0.75 (0.72, 0.77) 8.02 Vickers 2013 0.71 (0.68, 0.74) 7.92 Subtotal (I = 85.4%, p = 0.000) 0.73 (0.70, 0.76) 69.63 with MRI Boesen 2019 0.89 (0.87, 0.92) 8.02 Dwivedi 2017 0.89 (0.83, 0.95) 7.12 Mehralivand 2018 0.84 (0.79, 0.89) 7.42 van Leeuwen 2017 0.88 (0.84, 0.91) 7.82 Subtotal (I = 5.4%, p = 0.366) 0.88 (0.86, 0.90) 30.37 Overall (I = 96.0%, p = 0.000) 0.78 (0.73, 0.82) 100.00 NOTE: Weights are from random effects analysis -0.95 0 0.95 Figure 3: Forest plot of AUC (95% CI) of predictive models according to with/without MRI. 8 Prostate Cancer Funnel plot with pseudo 95% confidence limits another limitation of the study was the search date. We 0.0 searched databases until September 2019, and the studies published after this date were not included in the review and 0.1 meta-analysis. 0.2 5. Conclusion )e present study confirmed that mpMRI and bpMRI were 0.3 the strong predictive markers to improve the detection accuracy of models and could decrease the rate of unnec- 0.4 essary biopsies and decrease the overdetection of insignif- icant prostate cancer. 0.5 1 1.5 2 2.5 3 Data Availability expAUC )e data will be provided by the corresponding author on Figure 4: Funnel plot to assess the publication bias. request. in the three dimensions. )e accuracy assessment of prostate Ethical Approval volume using DRE has been found to be insufficient [40]. Not applicable. Furthermore, TRUS has been the first choice among imaging modalities for a long time [41]. However, low intraoperator Consent reproducibility and the poor interoperator agreement could affect the accuracy of TRUS [42]. Not applicable. Additionally, prostate volume might be underestimated using RTUS, particularly in patients with prostatic hyper- Conflicts of Interest plasia [43–45]. Recently, MRI has played a critical role in detecting PC and is considered the greatest accurate and All authors claim that there are no potential conflicts of reliable imaging for prostate volume estimation [46, 47]. For interest. example, Boesen et al. achieved the AUC of 0.85 and ex- cellent discrimination by the combination of only three Authors’ Contributions clinical parameters, including age, PSAd, and ctDRE [14], and in this study, the prostate volume was evaluated using SMK.A. was the principal investigator and supervisor of the bpMRI and indicated that MRI has excellent accuracy in project. M.S. was the epidemiologist who runs the statistical estimating prostate volume. analysis of data and provides figures. R.M., L.Z., and A.M. )ere were significant differences in the AUCs of pre- individually screened the data and made the data extraction dictive models between South East Asia and the Western sheet and tables of the article. Z.A. and F.KH wrote the Pacific with the Americas region (0.89 vs. 0.73, 0.88 vs. 0.73, manuscript. Z.A. edited the manuscript as well. respectively). Europe, the Western Pacific, and Southeast Asia regions had similar AUCs. )e highest AUCs were Acknowledgments observed in studies where models were developed based on MRI and clinical markers in all four regions. )e variance of Special thanks are due to the Urology Research Center, AUCs can probably be explained by inherent differences in Tehran University of Medical Sciences, Tehran, Iran. study design, calculation methods of AUCs, various pa- rameters of models, and validation methods of AUCs. References )e overall AUC of models that were based on the combination of MRI and clinical parameters was 0.88, and [1] P. Rawla, “Epidemiology of prostate cancer,” World Journal of the AUC of models that were developed with only clinical Oncology, vol. 10, no. 2, pp. 63–89, 2019. variables was 0.73. )ere was a significant difference between [2] X. Filella, J. Alcover, R. Molina, J. M. Corral, P. Carretero, and A. M. Ballesta, “Measurement of complexed PSA in the the accuracy of models with and without MRI. )e findings differential diagnosis between prostate cancer and benign indicated that the addition of MRI could improve the ac- prostate hyperplasia,” .e Prostate, vol. 42, no. 3, pp. 181–185, curacy of predictive models by 15%. [3] J.-L. Descotes, “Diagnosis of prostate cancer,” Asian Journal of Limitation. )ere was wide heterogeneity across the studies Urology, vol. 6, no. 2, pp. 129–136, 2019. in terms of study design, calculation methods of AUCs, [4] H. A. Guess, “Benign prostatic hyperplasia and prostate various parameters of models, and validation methods of cancer,” Epidemiologic Reviews, vol. 23, no. 1, pp. 152–158, AUCs. )erefore, these factors might assemble highly het- erogeneous studies, and the finding of our study showed the [5] X. Filella, E. Fernandez-Gal ´ an, ´ R. Fernandez ´ Bonifacio, and high heterogeneity and this subject is inevitable. Also, L. Foj, “Emerging biomarkers in the diagnosis of prostate S.E. of expAUC Prostate Cancer 9 cancer,” Pharmacogenomics and Personalized Medicine, “Improving the Rotterdam European randomized study of vol. 11, pp. 83–94, 2018. screening for prostate Cancer risk calculator for initial [6] F. Khatami, S. M. K. Aghamir, and S. M. Tavangar, “Onco- prostate biopsy by incorporating the 2014 International So- metabolites: a new insight for oncology,” Molecular Genetics ciety of Urological Pathology Gleason Grading and Cribri- and Genomic Medicine, vol. 7, no. 9, 2019. form growth,” European Urology, vol. 72, no. 1, pp. 45–51, [7] F. Khatami, S. M. K. Aghamir, A. Salmaninejad, S. Shivarani, and M. H. Khorrami, “Biomarkers for prostate cancer diag- [22] S. M. K. Aghamir, G. Shafiee, M. Ebrahimi et al., “Comparison nosis from genetic perspectives,” Translational Research in on diagnostic accuracy of prostate cancer detection tools: a Urology, vol. 2, no. 2, pp. 51–58, 2020. systematic review and meta-analysis,” Translational Research [8] M. Saadati, S. Tamehri, M. P. Kamali, and D. Taheri, In Urology, vol. 1, no. 1, pp. 31–44, 2019. “Phosphatase and tensin gene associated with features of [23] D. K. Dwivedi, R. Kumar, A. K. Dwivedi et al., “Prebiopsy aggressive prostate cancer,” Translational Research In Urology, multiparametric MRI-based risk score for predicting prostate vol. 3, no. 1, pp. 32–37, 2021. cancer in biopsy-naive men with prostate-specific antigen [9] K. M. Koo, P. N. Mainwaring, S. A. Tomlins, and M. Trau, between 4-10 ng/mL,” Journal of Magnetic Resonance Imag- “Merging new-age biomarkers and nanodiagnostics for pre- ing, vol. 47, no. 5, pp. 1227–1236, 2018. cision prostate cancer management,” Nature Reviews Urology, [24] A. J. Vickers, A. M. Cronin, M. J. Roobol et al., “A four- vol. 16, no. 5, pp. 302–317, 2019. kallikrein panel predicts prostate cancer in men with recent [10] W. A. Cieslikowski, J. Budna-Tukan, M. Swierczewska et al., screening: data from the European randomized study of “Circulating tumor cells as a marker of disseminated disease screening for prostate cancer, rotterdam,” Clinical Cancer in patients with newly diagnosed high-risk prostate cancer,” Research, vol. 16, no. 12, pp. 3232–3239, 2010. Cancers, vol. 12, no. 1, p. 160, 2020. [25] A. J. Vickers, A. Gupta, C. J. Savage et al., “A panel of kal- [11] S. S. Tamehri Zadeh, D. Taheri, S. Shivarani, F. Khatami, and likrein marker predicts prostate cancer in a large, population- R. Kazemi, “Liquid biopsy in prostate cancer diagnosis and based cohort followed for 15 years without screening,” Cancer prognosis: a narrative review,” Translational Research In Epidemiology Biomarkers and Prevention, vol. 20, no. 2, Urology, vol. 2, no. 4, pp. 139–146, 2020. pp. 255–261, 2011. [12] F. Khatami and M. Hasanzad, “Circulating tumor cells as a [26] F. H. Schroder, ¨ J. Hugosson, M. J. Roobol et al., “Screening novel prostate cancer diagnostic tool,” Translational Research and prostate cancer mortality: results of the European in Urology, vol. 2, no. 3, pp. 93–95, 2020. randomised study of screening for prostate cancer (ERSPC) at [13] X.-H. Zhou, D. K. McClish, and N. A. Obuchowski, Statistical 13 years of follow-up,” Lancet (London, England), vol. 384, Methods in Diagnostic Medicine, John Wiley and Sons, no. 9959, pp. 2027–2035, 2014. Hoboken, NJ, USA, 2009. [27] D. Ilic, M. M. Neuberger, M. Djulbegovic, and P. Dahm, [14] L. Boesen, F. B. )omsen, N. Nørgaard et al., “A predictive “Screening for prostate cancer,” Cochrane Database of Sys- model based on biparametric magnetic resonance imaging tematic Reviews, vol. 2013, no. 1, 2013. and clinical parameters for improved risk assessment and [28] E. A. M Heijnsdijk, E. M. Wever, A. Auvinen et al., “Quality- selection of biopsy-na¨ıve men for prostate biopsies,” Prostate of-life effects of prostate-specific antigen screening,” New Cancer and Prostatic Diseases, vol. 22, no. 4, pp. 609–616, England Journal of Medicine, vol. 367, no. 7, pp. 595–605, [15] D. K. Dwivedi, R. Kumar, G. S. Bora et al., “Stratification of the [29] P. J. van Leeuwen, A. Hayen, J. E. )ompson et al., “A aggressiveness of prostate cancer using pre-biopsy multi- multiparametric magnetic resonance imaging-based risk parametric MRI (mpMRI),” NMR in Biomedicine, vol. 29, model to determine the risk of significant prostate cancer no. 3, pp. 232–238, 2016. prior to biopsy,” BJU International, vol. 120, no. 6, [16] M. J. Roobol, F H Schroder, ¨ J. Hugosson et al., “Importance of pp. 774–781, 2017. prostate volume in the European Randomised Study of [30] S. Mehralivand, J. H. Shih, S. Rais-Bahrami et al., “A magnetic Screening for Prostate Cancer (ERSPC) risk calculators: re- resonance imaging-based prediction model for prostate bi- sults from the prostate biopsy collaborative group,” World opsy risk stratification,” JAMA Oncology, vol. 4, no. 5, Journal of Urology, vol. 30, no. 2, pp. 149–155, 2012. pp. 678–685, 2018. [17] D. P. Ankerst, T. Koniarski, Y. Liang et al., “Updating risk [31] E J Bass, A. Pantovic, M. Connor et al., “A systematic review prediction tools: a case study in prostate cancer,” Biometrical and meta-analysis of the diagnostic accuracy of biparametric Journal, vol. 54, no. 1, pp. 127–142, 2012. prostate MRI for prostate cancer in men at risk,” Prostate [18] R. W. Foley, L. Gorman, N. Sharifi et al., “Improving mul- Cancer and Prostatic Diseases, vol. 24, no. 3, pp. 596–611, tivariable prostate cancer risk assessment using the prostate health index,” BJU International, vol. 117, no. 3, pp. 409–417, [32] N. Mottet, J. Bellmunt, M. Bolla et al., “EAU-ESTRO-SIOG guidelines on prostate cancer. Part 1: screening, diagnosis, and [19] R. K. Nam, M. W. Kattan, J. L. Chin et al., “Prospective multi- local treatment with curative intent,” European Urology, institutional study evaluating the performance of prostate vol. 71, no. 4, pp. 618–629, 2017. cancer risk calculators,” Journal of Clinical Oncology, vol. 29, [33] A. B. Rosenkrantz, S. Verma, P. Choyke et al., “Prostate no. 22, pp. 2959–2964, 2011. magnetic resonance imaging and magnetic resonance imaging [20] D. A. Roffman, G. R. Hart, M. S. Leapman et al., “Devel- targeted biopsy in patients with a prior negative biopsy: a opment and validation of a multiparameterized artificial consensus statement by AUA and SAR,” Journal of Urology, neural network for prostate cancer risk prediction and stratification,” JCO Clinical Cancer Informatics, vol. 2, no. 2, vol. 196, no. 6, pp. 1613–1618, 2016. [34] J. J. Futterer, ¨ A. Briganti, P. De Visschere et al., “Can clinically pp. 1–10, 2018. [21] M. J. Roobol, J. F. M. Verbeek, T. van der Kwast, significant prostate cancer be detected with multiparametric I. P. Kummerlin, ¨ C. F. Kweldam, and G. J. L. H. van Leenders, magnetic resonance imaging? A systematic review of the 10 Prostate Cancer literature,” European Urology, vol. 68, no. 6, pp. 1045–1053, [35] M. de Rooij, S. Crienen, J. A. Witjes, J. O. Barentsz, M. M. Rovers, and J. P. C. Grutters, “Cost-effectiveness of magnetic resonance (mr) imaging and MR-guided targeted biopsy versus systematic transrectal ultrasound-guided biopsy in diagnosing prostate cancer: a modelling study from a health care perspective,” European Urology, vol. 66, no. 3, pp. 430–436, 2014. [36] A. Mirzaei, L. Zareian Baghdadabad, M. H. Khorrami, and S. M. K. Aghamir, “Arsenic Trioxide (ATO), a novel thera- peutic agent for prostate and bladder cancers,” Translational Research In Urology, vol. 1, no. 1, pp. 1–6, 2019. [37] I. Sonni, E. Felker, A. T. Lenis et al., “Head-to-head com- parison of 68ga-psma-11 pet/ct and mpmri in the detection, intra-prostatic localization, and local extension of primary prostate cancer: a single-center imaging study with histopa- thology gold-standard,” Journal of Clinical Oncology, vol. 39, no. 6_suppl, p. 193, 2021. [38] Y. Hou, K.-W. Jiang, J. Zhang et al., “A clinical available decision support scheme for optimizing prostate biopsy based on mpMRI,” Prostate Cancer and Prostatic Diseases, vol. 25, pp. 1–8, 2022. [39] M. Stojadinovic, M. Stojadinovic, and D. Pantic, “Decision tree analysis for prostate cancer prediction,” Srpski Arhiv Za Celokupno Lekarstvo, vol. 147, no. 1-2, pp. 52–58, 2019. [40] D. R. Christie, J. Windsor, and C. F. Sharpley, “A systematic review of the accuracy of the digital rectal examination as a method of measuring prostate gland volume,” Journal of Clinical Urology, vol. 12, no. 5, pp. 361–370, 2019. [41] M. R. E. Harris, E. J. Harding, T. S. Bates, and M. J. Speakman, “Gland volume in the assessment of prostatic disease: does size matter?” BJU International, vol. 100, no. 3, pp. 506–508, [42] V. Narayanamurthy, K. Mishra, K. Mishra et al., “Inter-im- aging accuracy of computed tomography, magnetic resonance imaging, and transrectal ultrasound in measuring prostate volume compared to the anatomic prostatic weight,” Tu¨rk Uroloji Dergisi/Turkish Journal of Urology, vol. 46, no. 1, pp. 50–56, 2020. [43] P. J. MacMahon, A.-M. Kennedy, D. T. Murphy, M. Maher, and M. M. McNicholas, “Modified prostate volume algorithm improves transrectal US volume estimation in men presenting for prostate brachytherapy,” Radiology, vol. 250, no. 1, pp. 273–280, 2009. [44] M. K. H. Hong, H. H. I. Yao, K. Rzetelski-West et al., “Prostate weight is the preferred measure of prostate size in radical prostatectomy cohorts,” BJU International, vol. 109, pp. 57– 63, 2012. [45] E. Rodriguez, D. Skarecky, N. Narula, and T. E. Ahlering, “Prostate volume estimation using the ellipsoid formula consistently underestimates actual gland size,” Journal of Urology, vol. 179, no. 2, pp. 501–503, 2008. [46] D. R. H. Christie and C. F. Sharpley, “How accurately can prostate gland imaging measure the prostate gland volume? Results of a systematic review,” Prostate Cancer, vol. 2019, Article ID 6932572, 12 pages, 2019. [47] I. Karademir, D. Shen, Y. Peng et al., “Prostate volumes derived from MRI and volume-adjusted serum prostate- specific antigen: correlation with Gleason score of prostate cancer,” American Journal of Roentgenology, vol. 201, no. 5, pp. 1041–1048, 2013. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Prostate Cancer Hindawi Publishing Corporation

Diagnostic Accuracy of Predictive Models in Prostate Cancer: A Systematic Review and Meta-Analysis

Download PDF
10 pages

Loading next page...
 
/lp/hindawi-publishing-corporation/diagnostic-accuracy-of-predictive-models-in-prostate-cancer-a-wW28jil0g9

References (47)

Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2022 Mohammad Saatchi et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
2090-3111
eISSN
2090-312X
DOI
10.1155/2022/1742789
Publisher site
See Article on Publisher Site

Abstract

Hindawi Prostate Cancer Volume 2022, Article ID 1742789, 10 pages https://doi.org/10.1155/2022/1742789 Review Article Diagnostic Accuracy of Predictive Models in Prostate Cancer: A Systematic Review and Meta-Analysis 1,2 1 1 1 Mohammad Saatchi , Fatemeh Khatami , Rahil Mashhadi , Akram Mirzaei , 1 1 1 Leila Zareian , Zeinab Ahadi , and Seyed Mohammad Kazem Aghamir Urology Research Center, Tehran University of Medical Sciences, Tehran, Iran Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, Iran Correspondence should be addressed to Seyed Mohammad Kazem Aghamir; mkaghamir@tums.ac.ir Received 10 January 2022; Revised 17 May 2022; Accepted 20 May 2022; Published 8 June 2022 Academic Editor: Cristina Magi-Galluzzi Copyright © 2022 Mohammad Saatchi et al. �is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Aim. Accurate diagnosis of prostate cancer (PCa) has a fundamental role in clinical and patient care. Recent advances in diagnostic testing and marker lead to standardized interpretation and increased prescription by clinicians to improve the detection of clinically signi‡cant PCa and select patients who strictly require targeted biopsies. Methods. In this study, we present a systematic review of the overall diagnostic accuracy of each testing panel regarding the panel details. In this meta-analysis, using a structured search, Web of Science and PubMed databases were searched up to 23 September 2019 with no restrictions and ‡lters. �e study’s outcome was the AUC and 95% con‡dence interval of prediction models. �is index was reported as an overall and based on the WHO region and models with/without MRI. Results. �e thirteen ‡nal articles included 25,691 people. �e overall AUC and 95% CI in thirteen studies were 0.78 and 95% CI: 0.73–0.82. �e weighted average AUC in the countries of the Americas region was 0.73 (95% CI: 0.70–0.75), and in European countries, it was 0.80 (95% CI: 0.72–0.88). In four studies with MRI, the average weighted AUC was 0.88 (95% CI: 0.86–0.90), while in other articles where MRI was not a parameter in the diagnostic model, the mean AUC was 0.73 (95% CI: 0.70–0.76). Conclusions. �e present study’s ‡ndings showed that MRI signi‡cantly improved the detection accuracy of prostate cancer and had the highest discrimination to distinguish candidates for biopsy. So, new PCa biomarkers have been proposed to im- 1. Introduction prove the accuracy of PSA in the management of early PCa Prostate cancer (PCa) is the second most frequent cancer in [5–8]. �e diagnostic panels include PSA isoforms, PSA men worldwide, and its incidence and mortality correlate density and velocity, age-adjusted PSA, free PSA to total with increasing age [1]. �e accurate PCa diagnosis is a PSA ratio (fPSA/tPSA), PSA density (PSAd), PSA doubling problematic issue because it is essential to identify which time (PSADT), Prostate health index (Phi), 4K score (in- PCa are destined to progress and which would bene‡t from clude kallikrein-related peptidase 2/hK2, intact PSA, fPSA, early radical treatment [2]. PCa has traditionally been di- and tPSA), advanced MRI (mpMRI and bpMRI), PCA3 agnosed by digital rectal examination (DRE) and prostate- mRNA, PSA glycoforms, TMPRSS2:ERG fusion gene, speci‡c antigen (PSA) blood test, followed by transrectal microRNAs, circulating tumor cells (CTCs), and androgen ultrasound (TRUS) guided biopsy [3]. �eir limited speci- receptor variants [5, 9–12]. AUC is an e§ective way to ‡city and an elevated rate of overdiagnosis are the main summarize the overall diagnostic accuracy of each testing problems associated with PCa testing. Benign prostatic panel. hyperplasia (BPH) has similar symptoms to PCa, and most No comprehensive study represents the most accurate PCa patients are diagnosed as asymptomatic patients with ones, and the heterogeneity of all clinical trials is too high in normal DRE and elevated PSA [4]. both the panel components and AUC. �is systematic 2 Prostate Cancer review summarizes all PCa diagnostic panels and compares 3. Results their AUC to find the most accurate ones. In this systematic review, 4188 articles were identified, of which 4185 articles were extracted from the search of 2. Methods electronic databases, and three articles were extracted from the search of the list of selected articles and other sources. )is systematic review and meta-analysis were designed After deleting duplicate articles, the title and abstract of 3228 according to the latest version of the PRISMA checklist, and articles were screened, and according to the exclusion cri- it was registered on Prospero with registration number: CRD teria, 3186 articles were removed. Finally, 13 articles were 149417. )e summary major was AUC with a 95% confi- used in the final analysis (Figure 1). dence interval (CI). )e thirteen final articles included 25,691 people. )e characteristics of the studies include the names of the 2.1. Search Strategy. We searched Scopus, Web of Science, authors, the country, the WHO region, sample size, mean and PubMed databases on 23 September 2019. )e search or median age, AUC and 95% confidence interval, model query was as follows: “Prostate Neoplasms” OR “Prostatic parameters, quality assessment score, and model name. Neoplasm” OR “Prostate Cancer” OR “Prostatic Cancer” Based on the findings of our study, the highest AUC was AND “ risk score” OR “prognostic score” OR “prognostic observed in the study of Boesen et al. [14] (0.89 model” OR “prognostic panel” OR “prognostic score (0.87–0.92)) and Dwivedi et al. [15] (0.89 (0.83–0.95)). In model”. Duplicate studies were removed prior to download. both studies, MRI played an important role in increasing After that, we included articles with these inclusion criteria: AUC. In the study of Roobol et al. [16], the lowest AUC articles that provided a model/panel for prostate cancer was observed in the GOTEBORG-R2–6 cohort and PSA prediction. Exclusion criteria included the following: (1) DRE-model, which included only PSA, DRE, and Prior articles that investigate genetic factors, (2) articles that did biopsy (Table 1). not report AUC (with 95% confidence interval) for their In the final analysis, most articles are from European and model, and (3) articles studying the treatment, recurrence, or American regions. As shown in Figure 2, the weighted metastasis of prostate cancer. average AUC in the countries of the American region was 0.73 (95% CI: 0.70–0.75), and in European countries, it was 0.80 (95% CI: 0.72–0.88). A study from Southeast Asia and a 2.2. Data Collection. )ree reviewers, RM, AM, and LZ were study from the Asia-Pacific region were also in the final independently involved in the title and abstract and read and analysis. )e overall AUC and 95% CI in thirteen studies was determined the eligibility of the studies. All three authors, 0.78 (95% CI: 0.73–0.82). Data from the previously pub- RM, AM, and LZ, independently extracted all relevant data, lished meta-analysis indicated that PI-RADS are superior in including the year of publication, first author, country, diagnosing PCa with high sensitivity, specificity, and AUC sample size, mean/median or range of age, AUC (95% CI), than PHI and PCA3 [22]. and model contents. )e disagreement was resolved by Figure 3 shows the AUC of studies based on the presence discussion, and when necessary, two reviewers (SMKA and or absence of MRI in the final model. In four studies with FKh) assisted in adjudicating a final decision. MRI, the average weighted of AUCs was 0.88 (95% CI: 0.86–0.90), while in other articles where MRI was not a parameter in the diagnostic model, the mean AUCs were 2.3. Methodological Quality Assessment. )e New- 0.73 (95% CI: 0.70–0.76). castle–Ottawa Scale (NOS) assessment tool was used to Figure 4 shows the funnel plot to investigate publication evaluate the quality of the articles by three authors [1]. )e bias. )e Begg (P value � 0.428) and Egger (P value � 0.780) scoring was based on the assignment of stars from 1 to 9. tests showed no significant publication bias in our study. According to the NOS score, the selected studies were di- vided into high quality (≥6) and low quality (<6). 4. Discussion 2.4. Statistical Analysis. )e chi-square test at a significant )e present study assessed the predictive models for PCa level of 5% was used for the qualitative assessment of het- detection to find the models that had the highest discrim- erogeneity across studies. Based on the Higgins categori- ination in distinguishing candidates for biopsy. In the zation, an I-square of more than 75% was considered current study, the highest AUCs were observed for two heterogeneity. )e index of interest in this study was AUC models; one of them is based on age, PSA density, DRE, and which was calculated as the proportion using the ROC curve bpMRI (AUC: 0.89, 95% CI: 0.87–0.92) [23], and the second method with 95% confidence intervals. )e weights for the one developed with PSA, MRSI, and DW-MRI (mpMRI) weighted average AUC calculation were calculated in ac- (AUC: 0.89, 95% CI: 0.83–0.95) [14]. )e present study’s cordance with the methods described by Zhou et al. [13]. findings showed that the best predictive models for PC Data analysis was performed using the Stata version 11 detection were based on the combination of clinical pa- (StataCorp, College Station, TX, USA) statistical software. rameters and bpMRI or mpMRI. By adding the MRI to Also, the random effect model at a confidence level of 95% clinical parameters, the predictive accuracy improved sig- was used in the data analysis. nificantly. Also, the AUCs of most models based on only Prostate Cancer 3 Number of Records identified Number of Additional records through database searching identified through other sources (n = 4185) (n = 3) Number of Records after duplicates removed No. of Records excluded (n = 3186)with reasons: (n = 3228) Genetic studies (n=1806); Studies that have examined the treatment, recurrence and No. of Records screened metastasis of the disease (n = 3228) (n=1114); Studies without AUC with 95% confidence interval (n=224); Book or book chapter (n=6); Conference paper (n=12); review article (n=19); No. of full-text articles assessed for eligibility Letter to the editor (n=5) (n = 42) No. of full-textarticles excluded (Inadequate No. of Studies included in information provided) meta-analysis (n=29) (n = 13) Figure 1: Flow of information through di§erent steps of the systematic review and meta-analysis. MRI, along with clinical parameters, can be utilized to clinical variables were lower than the AUCs of models with the incorporation of imaging [16–21, 24, 25]. decrease the number of unnecessary biopsies. Otherwise, MRI can ensure information about cancer location, staging, Previous documents assessing the e¨cacy of prostate cancer detection have highlighted the need to decrease in- and the volume for target biopsies. At present, both the signi‡cant prostate cancer’s overdiagnosis [26–28]. Hence, a American and European associations of urology (AUA and novel diagnostic panel is required to decrease the number of EAU) recommend using mpMRI as a useful diagnostic tool unneeded biopsies and recognition of insigni‡cant prostate before repeat biopsy and for men enrolled in active sur- cancer. So, recently numerous nomograms and predictive veillance [32, 33]. A recent systematic review reported models with various parameters, varying degrees of accu- clinically signi‡cant disease detection rates, the sensitivity, racy, generalizability, and validation were developed to and the negative predictive value (NPV) of mpMRI ranged improve the accuracy of PC diagnosis. Recently, the evidence from 44 to 87%, 58–97%, and 63% to 98%, respectively. In 2022, Futterer et al. had shown that mpMRI could be applied showed that when mpMRI or bpMRI is added to the standard clinical factors, the predictive accuracy enhances to rule out signi‡cant disease because of its extraordinary NPV [34]. �e use of radiomics and kallikreins failed to [14, 23, 29, 30]. A meta-analysis showed that bpMRI o§ers similar test accuracy to mpMRI in identifying prostate outperform PI-RADSv2.1/IMPROD bpMRI Likert, and cancer, but heterogeneity does not allow de‡nitive recom- their combination did not lead to further performance gains. mendations to be made [31]. Boesen et al. [14] showed that �e high expenses of mpMRI are debating using the mpMRI by adding bpMRI to clinical parameters (age, PSAd, ctDRE), to detect prostate cancer. Despite the high cost of mpMRI at the AUC of the model improved signi‡cantly from 0.85 to ‡rst look, it is generally considered a cost-e§ective method in 0.89 for predicting PC and achieved the highest discrimi- PC diagnosis because it reduces unnecessary biopsies costs, nation power. Also, they showed that the AUC of the model prevents unnecessary therapies, and increases the quality of based on the only bpMRI was 0.84 and demonstrated that life in the long term [35, 36]. �e prostate-speci‡c mem- brane antigen (PSMA) PET/CT and mpMRI have compa- the MRI-derived score as a PC detection is the most pow- erful single predictor. In line with this, Dwivedi et al. [23] rable diagnostic accuracy in the discovery and intraprostatic found that the model’s accuracy is higher with mpMRI than localization of prostate cancer foci whereas mpMRI makes without (0.89 vs. 0.66). van Leeuwen et al. showed that the better in the assessment of extracapsular extension (ECE) addition of mpMRI to commonly used clinical elements and seminal vesicle invasion (SVI) [37]. However, the ad- enhanced the predictive accuracy by 9% [29]. As a result, vantage of systematic biopsy (SBx) added to combined MRI/ Included Eligibility Screening Identification 4 Prostate Cancer Table 1: Baseline characteristics for studies included in meta-analysis. Mean age/ Author WHO Sample ID Country range/ AUC Model content Score Model name Name region size median Total PSA 0.69 Family history (0.65–0.74) DRE Free PSA Prostate Cancer Prevention Ankerst et 1 USA Americas 575 63.4 Total PSA 7 Trial Risk Calculator al. [17] Family history (PCPTRC) model 0.64 DRE (0.65–0.74) Free PSA [-2] Pro PSA Age 0.89 PSA density Advanced imaging model (0.87–0.92) cTDRE bpMRI 0.78 PSA Boesen et Baseline model 2 Denmark Europe 876 65 (0.75–0.82) cTDRE 8 al. [14] 0.84 bpMRI Imaging model (0.81–0.86) Age 0.85 PSA density Advanced model (0.83–0.88) cTDRE Age ADC 0.66 (NA) Original PSA Metabolic ratio Age ADC 0.78 (NA) PSA Original Metabolic ratio DW-MRI Dwivedi et Southeast Age 3 India 137 65 9 al. [15] Asia ADC 0.83 (NA) PSA Original Metabolic ratio MRSI Age ADC 0.89 PSA Developed (0.83–0.95) Metabolic ratio model mpMRI (MRSI + DW-MRI) Age at biopsy Abnormality on DRE Family history 0.71 Previous negative Predicting PHI (0.64–0.77) biopsy Total PSA Foley et al. Free PSA 4 Ireland Europe 250 63.7 7 [18] p2PSA Age at biopsy Abnormality on DRE 0.62 Family history Predicting PSA (0.55–0.69) Previous negative biopsy PSA Prostate Cancer 5 Table 1: Continued. Mean age/ Author WHO Sample ID Country range/ AUC Model content Score Model name Name region size median Age Family history Ethnicity Urinary voiding Sunnybrook nomogram- 0.67 Symptom score based prostate cancer risk (0.65–0.69) DRE calculator (SRC) Nam et al. Median age PSA 5 Canada Americas 2130 8 [19] 63 free: total PSA ratio Age Family history Prostate Cancer Prevention 0.61 Ethnicity Trial (PCPT)-based risk (0.59–0.64) DRE calculator (PRC) PSA Age BMI Diabetes status Smoking status Emphysema Multi parameterized Roffman et 0.73 Asthma 6 USA Americas 1672 67 9 artificial neural network al. [20] (0.71–0.75) Race (ANN) Ethnicity Hypertension Heart disease Exercise habits History of stroke Age Low-risk PSA (class via PCa DRE) DRE-model 0.70 Abnormal DRE (0.68–0.72) Roobol et Prostate volume 7 Netherland Europe 3580 68 9 al. [21] PSA Low-risk Age PCa Abnormal DRE TRUS model 0.73 Prostate volume (0.70–0.75) Abnormal TRUS 6 Prostate Cancer Table 1: Continued. Mean age/ Author WHO Sample ID Country range/ AUC Model content Score Model name Name region size median PSA GOTEBORG-R1 Median age 0.77 DRE 740 cohort 61 (0.73–0.81) Prostate volume DRE vol-RC model Prior biopsy PSA GOTEBORG-R1 Median age 0.71 740 DRE cohort 61 (0.67–0.76) Prior biopsy PSA DRE-model PSA GOTEBORG-R2–6 Median age 0.60 DRE 1241 cohort 63 (0.57–0.64) Prostate volume DRE vol-RC model Prior Biopsy PSA Median age 0.56 GOTEBORG-R2–6 cohort 1241 DRE 63 (0.52–0.60) PSA DRE-model Prior biopsy PSA DRE ROTTERDAM-R1 Median age 0.74 2895 Prostate volume cohort 66 (0.72–0.79) Prior biopsy DRE vol-RC model Family history PSA DRE ROTTERDAM-R2-3 Median age 0.65 1494 Prostate volume cohort 67 (0.62–0.69) Prior biopsy DRE vol-RC model Family history Roobol et PSA 8 Netherlands Europe 8 ROTTERDAM-R2-3 al. [16] Median age 0.60 DRE 1494 cohort 67 (0.57–0.63) Prior biopsy PSA DRE-model Family history PSA DRE Prostate volume Median age 0.66 Biopsy Gleason CCF cohort 64 (0.64–0.68) grade DRE vol-RC model Family history African origin Prior biopsy PSA DRE Biopsy Gleason Median age 0.62 CCF cohort 2631 grade 64 (0.60–0.64) PSA DRE-model Family history African origin Prior biopsy PSA Median age 0.72 DRE Tyrol cohort 63 (0.70–0.73) Prostate volume DRE vol-RC model Prior biopsy PSA Median age 0.67 Tyrol cohort 4199 DRE 63 (0.65–0.69) PSA DRE-model Prior biopsy Abbreviations. PSA: prostate-specific antigen, DRE: digital rectal examination, PCPTRC: prostate cancer prevention trial risk calculator, PRC: prostate cancer prevention trial (PCPT)-based risk calculator, ANN: artificial neural network, TRUS: transrectal ultrasound, DW-MRI: diffusion-weighted magnetic resonance imaging, BMI: body mass index, SRC: Sunnybrook nomogram–based prostate cancer risk calculator, MRSI: magnetic resonance spectroscopic imaging, ADC: apparent diffusion coefficients, and PHI: prostate health index. ultrasound fusion targeted biopsy (TBx) is mainly limited to model was developed with age, PSAd, DRE, prostate volume, smaller PI-RADS score 3–4 lesions [38]. and PSA [39]. )e determining PSAd requires an accurate In the current study, the highest AUC of the model assessment of prostate volume, and in this model, prostate developed based on only clinical variables was 0.83, and this volume was estimated using transrectal ultrasound (TRUS) Prostate Cancer 7 Study ID ES (95% CI) %Weight Americas Ankerst 2012 0.69 (0.65, 0.74) 7.57 Mehralivand 2018 0.84 (0.79, 0.89) 7.42 Mam 2011 0.67 (0.65, 0.69) 8.09 Roffman 2018 0.73 (0.71, 0.75) 8.09 Roobol 2017 0.73 (0.70, 0.75) 8.02 Vickers 2011 0.75 (0.72, 0.77) 8.02 Vickers 2013 0.71 (0.68, 0.74) 7.92 Subtotal (I = 89.2%, p = 0.000) 0.73 (0.70, 0.76) 55.13 Europe Boesen 2019 0.89 (0.87, 0.92) 8.02 Foley 2016 0.71 (0.64, 0.77) 6.95 Roobol 2012 0.77 (0.73, 0.81) 7.70 Stojadinovi 2019 0.83 (0.77, 0.88) 7.27 Subtotal (I = 93.0%, p = 0.000) 0.80 (0.72, 0.88) 29.94 South-east asia Dwivedi 2017 0.89 (0.83, 0.95) 7.12 Subtotal (I = .%, p = .) 0.89 (0.83, 0.95) 7.12 Western Pacific van Leeuwen 2017 0.88 (0.84, 0.91) 7.82 Subtotal (I = .%, p = .) 0.88 (0.85, 0.92) 7.82 Overall (I = 96.0%, p = 0.000) 0.78 (0.73, 0.82) 100.00 NOTE: Weights are from random effects analysis -0.95 0 0.95 Figure 2: Forest plot of AUC (95% CI) of predictive models in the di§erent regions. Study ID ES (95% CI) %Weight without MRI Ankerst 2012 0.69 (0.65, 0.74) 7.57 Foley 2016 0.84 (0.64, 0.77) 6.95 Nam 2011 0.67 (0.65, 0.69) 8.09 Roffman 2018 0.73 (0.71, 0.75) 8.09 Roobol 2012 0.77 (0.73, 0.81) 7.70 Roobol 2017 0.75 (0.70, 0.77) 8.02 Stojadinovi 2019 0.83 (0.77, 0.88) 7.27 Vickers 2011 0.75 (0.72, 0.77) 8.02 Vickers 2013 0.71 (0.68, 0.74) 7.92 Subtotal (I = 85.4%, p = 0.000) 0.73 (0.70, 0.76) 69.63 with MRI Boesen 2019 0.89 (0.87, 0.92) 8.02 Dwivedi 2017 0.89 (0.83, 0.95) 7.12 Mehralivand 2018 0.84 (0.79, 0.89) 7.42 van Leeuwen 2017 0.88 (0.84, 0.91) 7.82 Subtotal (I = 5.4%, p = 0.366) 0.88 (0.86, 0.90) 30.37 Overall (I = 96.0%, p = 0.000) 0.78 (0.73, 0.82) 100.00 NOTE: Weights are from random effects analysis -0.95 0 0.95 Figure 3: Forest plot of AUC (95% CI) of predictive models according to with/without MRI. 8 Prostate Cancer Funnel plot with pseudo 95% confidence limits another limitation of the study was the search date. We 0.0 searched databases until September 2019, and the studies published after this date were not included in the review and 0.1 meta-analysis. 0.2 5. Conclusion )e present study confirmed that mpMRI and bpMRI were 0.3 the strong predictive markers to improve the detection accuracy of models and could decrease the rate of unnec- 0.4 essary biopsies and decrease the overdetection of insignif- icant prostate cancer. 0.5 1 1.5 2 2.5 3 Data Availability expAUC )e data will be provided by the corresponding author on Figure 4: Funnel plot to assess the publication bias. request. in the three dimensions. )e accuracy assessment of prostate Ethical Approval volume using DRE has been found to be insufficient [40]. Not applicable. Furthermore, TRUS has been the first choice among imaging modalities for a long time [41]. However, low intraoperator Consent reproducibility and the poor interoperator agreement could affect the accuracy of TRUS [42]. Not applicable. Additionally, prostate volume might be underestimated using RTUS, particularly in patients with prostatic hyper- Conflicts of Interest plasia [43–45]. Recently, MRI has played a critical role in detecting PC and is considered the greatest accurate and All authors claim that there are no potential conflicts of reliable imaging for prostate volume estimation [46, 47]. For interest. example, Boesen et al. achieved the AUC of 0.85 and ex- cellent discrimination by the combination of only three Authors’ Contributions clinical parameters, including age, PSAd, and ctDRE [14], and in this study, the prostate volume was evaluated using SMK.A. was the principal investigator and supervisor of the bpMRI and indicated that MRI has excellent accuracy in project. M.S. was the epidemiologist who runs the statistical estimating prostate volume. analysis of data and provides figures. R.M., L.Z., and A.M. )ere were significant differences in the AUCs of pre- individually screened the data and made the data extraction dictive models between South East Asia and the Western sheet and tables of the article. Z.A. and F.KH wrote the Pacific with the Americas region (0.89 vs. 0.73, 0.88 vs. 0.73, manuscript. Z.A. edited the manuscript as well. respectively). Europe, the Western Pacific, and Southeast Asia regions had similar AUCs. )e highest AUCs were Acknowledgments observed in studies where models were developed based on MRI and clinical markers in all four regions. )e variance of Special thanks are due to the Urology Research Center, AUCs can probably be explained by inherent differences in Tehran University of Medical Sciences, Tehran, Iran. study design, calculation methods of AUCs, various pa- rameters of models, and validation methods of AUCs. References )e overall AUC of models that were based on the combination of MRI and clinical parameters was 0.88, and [1] P. Rawla, “Epidemiology of prostate cancer,” World Journal of the AUC of models that were developed with only clinical Oncology, vol. 10, no. 2, pp. 63–89, 2019. variables was 0.73. )ere was a significant difference between [2] X. Filella, J. Alcover, R. Molina, J. M. Corral, P. Carretero, and A. M. Ballesta, “Measurement of complexed PSA in the the accuracy of models with and without MRI. )e findings differential diagnosis between prostate cancer and benign indicated that the addition of MRI could improve the ac- prostate hyperplasia,” .e Prostate, vol. 42, no. 3, pp. 181–185, curacy of predictive models by 15%. [3] J.-L. Descotes, “Diagnosis of prostate cancer,” Asian Journal of Limitation. )ere was wide heterogeneity across the studies Urology, vol. 6, no. 2, pp. 129–136, 2019. in terms of study design, calculation methods of AUCs, [4] H. A. Guess, “Benign prostatic hyperplasia and prostate various parameters of models, and validation methods of cancer,” Epidemiologic Reviews, vol. 23, no. 1, pp. 152–158, AUCs. )erefore, these factors might assemble highly het- erogeneous studies, and the finding of our study showed the [5] X. Filella, E. Fernandez-Gal ´ an, ´ R. Fernandez ´ Bonifacio, and high heterogeneity and this subject is inevitable. Also, L. Foj, “Emerging biomarkers in the diagnosis of prostate S.E. of expAUC Prostate Cancer 9 cancer,” Pharmacogenomics and Personalized Medicine, “Improving the Rotterdam European randomized study of vol. 11, pp. 83–94, 2018. screening for prostate Cancer risk calculator for initial [6] F. Khatami, S. M. K. Aghamir, and S. M. Tavangar, “Onco- prostate biopsy by incorporating the 2014 International So- metabolites: a new insight for oncology,” Molecular Genetics ciety of Urological Pathology Gleason Grading and Cribri- and Genomic Medicine, vol. 7, no. 9, 2019. form growth,” European Urology, vol. 72, no. 1, pp. 45–51, [7] F. Khatami, S. M. K. Aghamir, A. Salmaninejad, S. Shivarani, and M. H. Khorrami, “Biomarkers for prostate cancer diag- [22] S. M. K. Aghamir, G. Shafiee, M. Ebrahimi et al., “Comparison nosis from genetic perspectives,” Translational Research in on diagnostic accuracy of prostate cancer detection tools: a Urology, vol. 2, no. 2, pp. 51–58, 2020. systematic review and meta-analysis,” Translational Research [8] M. Saadati, S. Tamehri, M. P. Kamali, and D. Taheri, In Urology, vol. 1, no. 1, pp. 31–44, 2019. “Phosphatase and tensin gene associated with features of [23] D. K. Dwivedi, R. Kumar, A. K. Dwivedi et al., “Prebiopsy aggressive prostate cancer,” Translational Research In Urology, multiparametric MRI-based risk score for predicting prostate vol. 3, no. 1, pp. 32–37, 2021. cancer in biopsy-naive men with prostate-specific antigen [9] K. M. Koo, P. N. Mainwaring, S. A. Tomlins, and M. Trau, between 4-10 ng/mL,” Journal of Magnetic Resonance Imag- “Merging new-age biomarkers and nanodiagnostics for pre- ing, vol. 47, no. 5, pp. 1227–1236, 2018. cision prostate cancer management,” Nature Reviews Urology, [24] A. J. Vickers, A. M. Cronin, M. J. Roobol et al., “A four- vol. 16, no. 5, pp. 302–317, 2019. kallikrein panel predicts prostate cancer in men with recent [10] W. A. Cieslikowski, J. Budna-Tukan, M. Swierczewska et al., screening: data from the European randomized study of “Circulating tumor cells as a marker of disseminated disease screening for prostate cancer, rotterdam,” Clinical Cancer in patients with newly diagnosed high-risk prostate cancer,” Research, vol. 16, no. 12, pp. 3232–3239, 2010. Cancers, vol. 12, no. 1, p. 160, 2020. [25] A. J. Vickers, A. Gupta, C. J. Savage et al., “A panel of kal- [11] S. S. Tamehri Zadeh, D. Taheri, S. Shivarani, F. Khatami, and likrein marker predicts prostate cancer in a large, population- R. Kazemi, “Liquid biopsy in prostate cancer diagnosis and based cohort followed for 15 years without screening,” Cancer prognosis: a narrative review,” Translational Research In Epidemiology Biomarkers and Prevention, vol. 20, no. 2, Urology, vol. 2, no. 4, pp. 139–146, 2020. pp. 255–261, 2011. [12] F. Khatami and M. Hasanzad, “Circulating tumor cells as a [26] F. H. Schroder, ¨ J. Hugosson, M. J. Roobol et al., “Screening novel prostate cancer diagnostic tool,” Translational Research and prostate cancer mortality: results of the European in Urology, vol. 2, no. 3, pp. 93–95, 2020. randomised study of screening for prostate cancer (ERSPC) at [13] X.-H. Zhou, D. K. McClish, and N. A. Obuchowski, Statistical 13 years of follow-up,” Lancet (London, England), vol. 384, Methods in Diagnostic Medicine, John Wiley and Sons, no. 9959, pp. 2027–2035, 2014. Hoboken, NJ, USA, 2009. [27] D. Ilic, M. M. Neuberger, M. Djulbegovic, and P. Dahm, [14] L. Boesen, F. B. )omsen, N. Nørgaard et al., “A predictive “Screening for prostate cancer,” Cochrane Database of Sys- model based on biparametric magnetic resonance imaging tematic Reviews, vol. 2013, no. 1, 2013. and clinical parameters for improved risk assessment and [28] E. A. M Heijnsdijk, E. M. Wever, A. Auvinen et al., “Quality- selection of biopsy-na¨ıve men for prostate biopsies,” Prostate of-life effects of prostate-specific antigen screening,” New Cancer and Prostatic Diseases, vol. 22, no. 4, pp. 609–616, England Journal of Medicine, vol. 367, no. 7, pp. 595–605, [15] D. K. Dwivedi, R. Kumar, G. S. Bora et al., “Stratification of the [29] P. J. van Leeuwen, A. Hayen, J. E. )ompson et al., “A aggressiveness of prostate cancer using pre-biopsy multi- multiparametric magnetic resonance imaging-based risk parametric MRI (mpMRI),” NMR in Biomedicine, vol. 29, model to determine the risk of significant prostate cancer no. 3, pp. 232–238, 2016. prior to biopsy,” BJU International, vol. 120, no. 6, [16] M. J. Roobol, F H Schroder, ¨ J. Hugosson et al., “Importance of pp. 774–781, 2017. prostate volume in the European Randomised Study of [30] S. Mehralivand, J. H. Shih, S. Rais-Bahrami et al., “A magnetic Screening for Prostate Cancer (ERSPC) risk calculators: re- resonance imaging-based prediction model for prostate bi- sults from the prostate biopsy collaborative group,” World opsy risk stratification,” JAMA Oncology, vol. 4, no. 5, Journal of Urology, vol. 30, no. 2, pp. 149–155, 2012. pp. 678–685, 2018. [17] D. P. Ankerst, T. Koniarski, Y. Liang et al., “Updating risk [31] E J Bass, A. Pantovic, M. Connor et al., “A systematic review prediction tools: a case study in prostate cancer,” Biometrical and meta-analysis of the diagnostic accuracy of biparametric Journal, vol. 54, no. 1, pp. 127–142, 2012. prostate MRI for prostate cancer in men at risk,” Prostate [18] R. W. Foley, L. Gorman, N. Sharifi et al., “Improving mul- Cancer and Prostatic Diseases, vol. 24, no. 3, pp. 596–611, tivariable prostate cancer risk assessment using the prostate health index,” BJU International, vol. 117, no. 3, pp. 409–417, [32] N. Mottet, J. Bellmunt, M. Bolla et al., “EAU-ESTRO-SIOG guidelines on prostate cancer. Part 1: screening, diagnosis, and [19] R. K. Nam, M. W. Kattan, J. L. Chin et al., “Prospective multi- local treatment with curative intent,” European Urology, institutional study evaluating the performance of prostate vol. 71, no. 4, pp. 618–629, 2017. cancer risk calculators,” Journal of Clinical Oncology, vol. 29, [33] A. B. Rosenkrantz, S. Verma, P. Choyke et al., “Prostate no. 22, pp. 2959–2964, 2011. magnetic resonance imaging and magnetic resonance imaging [20] D. A. Roffman, G. R. Hart, M. S. Leapman et al., “Devel- targeted biopsy in patients with a prior negative biopsy: a opment and validation of a multiparameterized artificial consensus statement by AUA and SAR,” Journal of Urology, neural network for prostate cancer risk prediction and stratification,” JCO Clinical Cancer Informatics, vol. 2, no. 2, vol. 196, no. 6, pp. 1613–1618, 2016. [34] J. J. Futterer, ¨ A. Briganti, P. De Visschere et al., “Can clinically pp. 1–10, 2018. [21] M. J. Roobol, J. F. M. Verbeek, T. van der Kwast, significant prostate cancer be detected with multiparametric I. P. Kummerlin, ¨ C. F. Kweldam, and G. J. L. H. van Leenders, magnetic resonance imaging? A systematic review of the 10 Prostate Cancer literature,” European Urology, vol. 68, no. 6, pp. 1045–1053, [35] M. de Rooij, S. Crienen, J. A. Witjes, J. O. Barentsz, M. M. Rovers, and J. P. C. Grutters, “Cost-effectiveness of magnetic resonance (mr) imaging and MR-guided targeted biopsy versus systematic transrectal ultrasound-guided biopsy in diagnosing prostate cancer: a modelling study from a health care perspective,” European Urology, vol. 66, no. 3, pp. 430–436, 2014. [36] A. Mirzaei, L. Zareian Baghdadabad, M. H. Khorrami, and S. M. K. Aghamir, “Arsenic Trioxide (ATO), a novel thera- peutic agent for prostate and bladder cancers,” Translational Research In Urology, vol. 1, no. 1, pp. 1–6, 2019. [37] I. Sonni, E. Felker, A. T. Lenis et al., “Head-to-head com- parison of 68ga-psma-11 pet/ct and mpmri in the detection, intra-prostatic localization, and local extension of primary prostate cancer: a single-center imaging study with histopa- thology gold-standard,” Journal of Clinical Oncology, vol. 39, no. 6_suppl, p. 193, 2021. [38] Y. Hou, K.-W. Jiang, J. Zhang et al., “A clinical available decision support scheme for optimizing prostate biopsy based on mpMRI,” Prostate Cancer and Prostatic Diseases, vol. 25, pp. 1–8, 2022. [39] M. Stojadinovic, M. Stojadinovic, and D. Pantic, “Decision tree analysis for prostate cancer prediction,” Srpski Arhiv Za Celokupno Lekarstvo, vol. 147, no. 1-2, pp. 52–58, 2019. [40] D. R. Christie, J. Windsor, and C. F. Sharpley, “A systematic review of the accuracy of the digital rectal examination as a method of measuring prostate gland volume,” Journal of Clinical Urology, vol. 12, no. 5, pp. 361–370, 2019. [41] M. R. E. Harris, E. J. Harding, T. S. Bates, and M. J. Speakman, “Gland volume in the assessment of prostatic disease: does size matter?” BJU International, vol. 100, no. 3, pp. 506–508, [42] V. Narayanamurthy, K. Mishra, K. Mishra et al., “Inter-im- aging accuracy of computed tomography, magnetic resonance imaging, and transrectal ultrasound in measuring prostate volume compared to the anatomic prostatic weight,” Tu¨rk Uroloji Dergisi/Turkish Journal of Urology, vol. 46, no. 1, pp. 50–56, 2020. [43] P. J. MacMahon, A.-M. Kennedy, D. T. Murphy, M. Maher, and M. M. McNicholas, “Modified prostate volume algorithm improves transrectal US volume estimation in men presenting for prostate brachytherapy,” Radiology, vol. 250, no. 1, pp. 273–280, 2009. [44] M. K. H. Hong, H. H. I. Yao, K. Rzetelski-West et al., “Prostate weight is the preferred measure of prostate size in radical prostatectomy cohorts,” BJU International, vol. 109, pp. 57– 63, 2012. [45] E. Rodriguez, D. Skarecky, N. Narula, and T. E. Ahlering, “Prostate volume estimation using the ellipsoid formula consistently underestimates actual gland size,” Journal of Urology, vol. 179, no. 2, pp. 501–503, 2008. [46] D. R. H. Christie and C. F. Sharpley, “How accurately can prostate gland imaging measure the prostate gland volume? Results of a systematic review,” Prostate Cancer, vol. 2019, Article ID 6932572, 12 pages, 2019. [47] I. Karademir, D. Shen, Y. Peng et al., “Prostate volumes derived from MRI and volume-adjusted serum prostate- specific antigen: correlation with Gleason score of prostate cancer,” American Journal of Roentgenology, vol. 201, no. 5, pp. 1041–1048, 2013.

Journal

Prostate Cancer – Hindawi Publishing Corporation

Published: Jun 8, 2022

There are no references for this article.