Establishment and Validation of a Machine Learning Model Predicting Post-Radical Prostatectomy Gleason grading group upgrading Author’s information

doi:10.21203/rs.3.rs-4959347/v1

Download PDF

Research Article

Establishment and Validation of a Machine Learning Model Predicting Post-Radical Prostatectomy Gleason grading group upgrading Author’s information

https://doi.org/10.21203/rs.3.rs-4959347/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

Background

Based on the 2014 International Society of Urological Pathology (ISUP) grading system, the study assesses the disparities in gleason grading group between preoperative needle biopsy pathology and post-radical prostatectomy (post-RP) specimens for prostate cancer (PCa). It investigates the risk factors for post-RP gleason grading group upgrading (GGU) and develops and validates a machine learning (ML) model for predicting post-RP GGU in PCa patients.

Methods

A retrospective analysis is conducted on demographic and clinicopathological variables of PCa patients from the Surveillance, Epidemiology, and End Results (SEER) database from 2010 to 2018. Five different ML algorithms, including logistic regression (LR), gradient boosting machine (GBM), neural network (NNET), random forest (RF), and XGBoost (XGB), are utilized. The patients with localized PCa who underwent radical prostatectomy (RP) at Zhongshan People's Hospital from January 2018 to December 2023 were selected as the external validation group. Model performance is evaluated using receiver operating characteristic (ROC) area under the curve (AUC), calibration curve, decision curve analysis (DCA), sensitivity (recall), and specificity. A web-based predictor is developed based on the best-performing model.

Results

This study included a total of 65,574 PCa patients from the SEER database and 98 patients from the external validation group. Among them, there were 11,931 in the training group, 5,112 in the internal validation group, and 24 in the external validation group who experienced post-RP GGU. Risk factors such as patient age, race, preoperative prostate-specific antigen (PSA) level, needle biopsy ISUP grading group, total number of biopsy cores, number of positive cores, and percentage of positive cores were significantly associated with GGU (P < 0.05). Five ML algorithms demonstrated relatively stable consistency, with their AUC values exceeding 0.7. A web-based predictor was developed using the XGB model, which showed the best predictive performance.

Conclusion

The study introduced a ML model and an online predictor designed to assess the risk of post-RP GGU in PCa patients, aiding physicians in customizing clinical decisions and treatment strategies.

machine learning

post-radical prostatectomy gleason grading group upgrading

risk prediction

prostate

Prostate cancer (PCa) is the second most common cancer globally and ranks as the fifth leading cause of cancer-related deaths in men as of 2022, with an estimated 1.5 million new cases and 397,000 deaths worldwide. Variations in PCa diagnostic practices among countries are key factors contributing to the significant differences in PCa incidence worldwide (1). Currently, the diagnostic process for PCa heavily relies on the histological examination of biopsy tissues. The PCa grading and staging system derived from biopsies serves as a primary reference for assessing patient conditions and formulating treatment plans. Ideally, there should be complete consistency between the pathological grading from preoperative biopsies and the final pathological grading reported post-radical prostatectomy (post-RP). However, due to the multifocality and high intralesional heterogeneity of PCa, combined with potential grading errors by pathologists and sampling inaccuracies, biopsy pathology often underestimates the disease compared to post-RP pathology (2, 3). Studies based on the PCa grading and staging system proposed by the International Society of Urological Pathology (ISUP) in 2014 have shown a high rate of inconsistency between post-RP and preoperative biopsy pathological grading, with Gleason grading group upgrading (GGU) being the most common occurrence(4–7). Additionally, GGU has been significantly associated with post-RP biochemical recurrence and a higher likelihood of surgical margin positivity in radical specimens (8–11). Underestimating PCa biopsy pathological grading may lead to an overly conservative selection of initial treatment regimens for prostate cancer, while overestimating patient prognosis, posing significant challenges to clinical treatment.

With advancements in statistical methods and artificial intelligence, predictive models are increasingly utilized in medical diagnostics. Predicting pathological upgrades based on preoperative and biopsy-known variables could be of crucial significance for surgeons when formulating surgical plans and determining postoperative treatment strategies for patients. Scientists have been striving to explore various methods to assess the risk of post-RP GGU more accurately. However, owing to the complexity of medical data and the intricate interplay among various factors, there are significant differences in the calculation methods of the models. Machine learning (ML) is a promising cross-disciplinary method capable of accurately predicting outcomes from multiple unrelated datasets that would otherwise be discrete and challenging to correlate (12). With the rapid development of evidence-based medicine, large and complex medical datasets require more advanced techniques for interpretation, and ML has become a promising choice for the diagnosis and prognostic prediction of many diseases. Previous approaches often relied on the construction of predictive nomograms or analyses based on multiple unfavorable factors (13–15). However, nomogram models are limited to independent predictive variables and risk overfitting when numerous predictors are incorporated, potentially leading to model failure. Consequently, there has been a shift towards more robust predictive classifiers that use ML methods. By processing and analyzing large, heterogeneous, and complex clinical data, practical clinical problems can be addressed, thereby making diagnostic and treatment processes more efficient (16–18). Relevant research has also emerged in the field of urological treatments, for instance, ML algorithms that utilize pathological data have been pivotal in developing and validating automated pathological models for diagnosing and predicting outcomes in patients with bladder cancer or clear-cell renal cell carcinoma (19, 20). Similarly, studies on PCa have employed ML models to predict bone metastasis in patients with PCa or lymph node metastasis in intermediate-to-high-risk PCa patients, aiming for precise diagnosis and treatment guidance (21, 22). Presently, investigations into the risk factors of post-RP GGU are constrained by small sample sizes, single-center methodologies, and predictive models lacking external data validation. In this study, we aimed to identify the risk factors of post-RP GGU, and develop and validate a new ML model to predict post-RP GGU in PCa patients from the Surveillance, Epidemiology, and End Results (SEER) database as training and internal validation groups and PCa patients from Zhongshan People's Hospital as external validation groups using five ML methods.

Data Collection

Data from patients diagnosed with PCa from 2010 to 2018 were retrieved using SEER*Stat software (version 8.4.0.1). The International Classification of Diseases for Oncology, 3rd edition, code 8140/3, was used for data filtering. The inclusion criteria were primary PCa, age ≥ 18 years, and complete clinical information. The exclusion criteria were preoperative neoadjuvant systemic therapy or other local treatments and incomplete clinical information. The SEER database is publicly accessible and anonymizes all patient information, hence exempt from ethical review. The external validation data from this retrospective clinical study waived the requirement for informed consent from the participants, with all procedures conducted in accordance with the Helsinki Declaration.

Inclusion of Analytical Variables

Eight variables potentially affecting GGU in patients with PCa were included: age, race, marital status, preoperative prostate-specific antigen (PSA) level, needle biopsy ISUP grading group, total number of biopsy cores, number of positive cores, and percentage of positive cores.

Statistical Methods

Data were processed using R Studio (version 4.2.3). Normally distributed continuous data were expressed as mean ± standard deviation (x ± s), while non-normally distributed continuous data were presented as median (25th percentile, 75th percentile). Count data were presented as frequency and percentage (%). Intergroup comparisons were conducted using the chi-square test or Fisher's exact test. To determine the variables included in the ML model, a univariate logistic regression analysis of GGU risk factors was first performed, followed by a multivariate logistic regression analysis to identify independent predictive factors associated with GGU, considering P-values < 0.05 as significant.

Machine Learning Classifiers and Model Building and Evaluation

Five ML and decision tree models were constructed based on selected independent risk factors: logistic regression (LR), gradient boosting machine (GBM), neural network (NNET), random forest (RF), and XGBoost (XGB). Initially, LR, one of the most fundamental classifiers, was used to predict the likelihood of a patient achieving a specific outcome based on relevant information or clinical features(23). Next, the GBM method was applied. This ensemble technique constructs multiple decision trees sequentially, each aiming to correct the errors of its predecessors, thus effectively managing complex data and nonlinear relationships (24). NNET was also employed, characterized by its structure of multiple layers of neurons that perform nonlinear transformations, making it particularly adept at capturing intricate patterns and processing large datasets (25). Furthermore, the RF algorithm was implemented to build numerous decision trees using bootstrap samples of the training data and random subsets of features for each split. It combines the outputs of these trees through majority voting to improve prediction accuracy and prevent overfitting (26). Finally, XGB was used, a highly efficient and scalable tree-boosting system that builds robust models even with relatively small sample sizes by creating new models to predict the residuals of previous models and aggregating these for the final prediction (27). The SEER dataset was divided into a 7:3 ratio for the training and internal validation groups, with external validation serving as a separate group. Each model underwent 10-fold cross-validation during training to ensure stability, and the best hyperparameters were selected using a random search method. Receiver operating characteristic (ROC) curves were constructed for both groups. Differences in predictive performance among the ML models were assessed using the area under the ROC curve (AUC), calibration curve, decision curve analysis (DCA), sensitivity, and specificity, with the best-performing model selected as the final model after a comprehensive comparison.

Web-based Predictor

Finally, the accuracy and generalizability of the best predictive model were validated using an independent external validation group. This step involved evaluating the model's performance on a separate dataset not used in the training and internal validation groups. The model's ability to accurately predict outcomes and maintain performance across different datasets and conditions was assessed, ensuring that it was not overfitted to the training data and could generalize well to new, unseen data. To enhance the usability and presentation of the data, a web-based predictor was developed using the Shiny package in R. Shiny is a powerful tool that allows the creation of interactive web applications directly from R software.

Baseline Characteristics and Postoperative Pathological Upgrading

This study included a total of 65,574 PCa patients from the SEER database. They were randomly divided into training (45,903 cases) and internal validation (19,671 cases) groups at a ratio of 7:3. Additionally, 130 patients with PCa who underwent RP at Zhongshan People's Hospital between 2018 and 2023 were included. After excluding 32 patients who did not meet the inclusion criteria, 98 patients were included in the external validation study. The process is illustrated in Fig. 1. In Table 1, the patient characteristics for the training, internal validation, and external validation groups were presented. Among these, 11,931 patients (25.9%) in the training group, 5,112 (25.9%) in the internal validation group, and 24 (24.4%) in the external validation group experienced post-RP GGU.

Table 1

Clinical characteristics of patients.
Characteristics	Training group(n = 45,903)	Internal validation group (n = 19,671)	External validation group (n = 98)
Age (%)
<60	2,610 (5.7)	1,129 (5.7)	7 (7.1)
60 ~ 69	2,973 (6.5)	1,316 (6.7)	36 (36.7)
>=70	40,320 (87.8)	17,226 (87.6)	55 (56.1)
Race (%)
White	35,983 (78.4)	15,462 (78.6)	-
Black	7,075 (15.4)	2,999 (15.2)	-
Asian or Pacific	2,669 (5.8)	1,136 (5.8)	98 (100.0)
American or Alaska	176 (0.4)	74 (0.4)	-
Marital (%)
Married	36,483 (79.5)	15,730 (80.0)	96 (98.0)
Single	4,967 (10.8)	2,085 (10.6)	-
Divorced or Separated	3,462 (7.5)	1,454 (7.4)	2 (2.0)
Other	991 (2.2)	402 (2.0)	-
Gleason Patterns Clinical (%)
Group1	13,245 (28.9)	5,679 (28.9)	10 (10.2)
Group2	16,208 (35.3)	6,902 (35.1)	18 (18.4)
Group3	7,827 (17.1)	3,389 (17.2)	20 (20.4)
Group4	5,268 (11.5)	2,280 (11.6)	30 (30.6)
Group5	3,355 (7.3)	1,421 (7.2)	20 (20.4)
Gleason Patterns Pathological (%)
Group1	8,194 (17.9)	3,503 (17.8)	5 (5.1)
Group2	21,745 (47.4)	9,359 (47.6)	29 (29.6)
Group3	9,489 (20.7)	4,029 (20.5)	24 (24.5)
Group4	2,535 (5.5)	1,139 (5.8)	15 (15.3)
Group5	3,940 (8.6)	1,641 (8.3)	25 (25.5)
PSA (%)
<4	5,106 (11.1)	2,189 (11.1)	5 (5.1)
4 ~ 10	32,416 (70.6)	13,892 (70.6)	31 (31.6)
11 ~ 19	5,476 (11.9)	2,318 (11.8)	23 (23.5)
>=20	2,905 (6.3)	1,272 (6.5)	39 (39.8)
Positive (%)
<5	23,485 (51.2)	10,103 (51.4)	46 (46.9)
5 ~ 9	17,404 (37.9)	7,415 (37.7)	32 (32.7)
>=10	5,014 (10.9)	2,153 (10.9)	20 (20.4)
Examined = >=10 (%)	41,310 (90.0)	17,701 (90.0)	98 (100.0)
Percent = >=33% (%)	27,789 (60.5)	11,819 (60.1)	52 (53.1)

Univariable and Multivariable Logistic Regression Analysis

Subsequently, univariate logistic regression analysis revealed significant associations (P < 0.05) between GGU and variables such as age, race, marital status, preoperative PSA level, needle biopsy ISUP grading group, total number of biopsy cores, number of positive cores, and percentage of positive cores among PCa patients. Multivariate logistic regression analysis was conducted on variables with P-values < 0.05. The results indicated that, except for marital status, all other variables were independent prognostic risk factors for GGU in patients with PCa (Table 2). Additionally, increasing age and preoperative PSA levels were associated with an increased likelihood of pathological upgrading, especially when age was ≥ 70 years and preoperative PSA was ≥ 20 ng/mL (OR 1.85; P < 0.001 and OR 2.70; P < 0.001/ng/mL).

Table 2

Univariate Analysis and Multivariate Logistic Regression Analysis of Variables
Characteristics	Univariate		Multivariate
Characteristics	OR (95%CI)	P-Value	OR (95%CI)	P-Value
Age
< 60	Reference		Reference
60 ~ 69	1.20(1.06–1.35)	0.004	1.41(1.24–1.61)	< 0.001
>=70	1.12(1.02–1.23)	0.016	1.85(1.67–2.05)	< 0.001
Race
White	Reference		Reference
Black	0.93(0.88–0.99)	0.015	0.88(0.83–0.94)	< 0.001
Asian or Pacific	1.06(0.96–1.16)	0.237	1.13(1.02–1.25)	0.017
American or Alaska	1.05(0.74–1.50)	0.774	0.88(0.59–1.29)	0.499
Marital
Married	Reference		Reference
Single	0.97(0.91–1.04)	0.397	0.98(0.91–1.05)	0.523
Divorced Separated	0.91(0.84–0.99)	0.031	0.93(0.85–1.01)	0.097
Other	0.91(0.79–1.06)	0.229	0.98(0.83–1.15)	0.805
Gleason Patterns Clinical
group1	Reference		Reference
group2	0.24(0.22–0.25)	< 0.001	0.20(0.18–0.21)	< 0.001
group3	0.14(0.13–0.16)	< 0.001	0.11(0.10–0.12)	< 0.001
group4	0.21(0.20–0.23)	< 0.001	0.15(0.14–0.16)	< 0.001
PSA
< 4	Reference		Reference
4 ~ 10	1.08(1.00-1.15)	0.037	1.30(1.20–1.40)	< 0.001
11 ~ 19	1.33(1.22–1.46)	< 0.001	2.09(1.89–2.30)	< 0.001
>=20	1.50(1.34–1.67)	< 0.001	2.70(2.40–3.04)	< 0.001
Positive
< 5	Reference		Reference
5 ~ 9	0.85(0.81–0.89)	< 0.001	1.06(0.99–1.14)	0.113
>=10	0.93(0.86-1.00)	0.042	1.25(1.13–1.38)	< 0.001
Examined
< 10	Reference		Reference
>=10	0.83(0.78–0.89)	< 0.001	0.90(0.83–0.98)	0.013
Percent
< 33%	Reference		Reference
>=33%	0.92(0.88–0.96)	< 0.001	1.25(1.16–1.34)	< 0.001

Comparison and Validation of Model Performance

Five supervised ML algorithms were employed to predict post-RP pathological GGU status, establishing binary models where labels were set as upgraded or otherwise. Surprisingly, all our ML models demonstrated relatively stable consistency, including LR, GBM, NNET, RF, and XGB, with their AUC values exceeding 0.7, namely 0.722, 0.726, 0.727, 0.703, and 0.728, respectively, as depicted in Fig. 2A. Despite minor differences in performance, the XGB model exhibited superior and stable results across both the training and internal validation groups. In the training group, the sensitivity, specificity, and AUC were 0.614, 0.760, and 0.728, respectively. Similarly, in the internal validation group, the XGB model maintained consistent performance with an AUC of 0.741, a sensitivity of 0.622, and a specificity of 0.767 (Table 3).

Table 3

Predictive performance of each model.
	Training group			Internal validation group
	Sensitivity	Specificity	AUC	Sensitivity	Specificity	AUC
LR	0.594	0.774	0.722	0.603	0.781	0.736
GBM	0.627	0.746	0.726	0.635	0.752	0.739
NNET	0.617	0.755	0.727	0.626	0.763	0.741
RF	0.573	0.790	0.703	0.577	0.791	0.712
XGB	0.614	0.760	0.728	0.622	0.767	0.741

Calibration curves and DCA were constructed for the five models used in our study (Figs. 2B, C, E, F). In our ML models, except for the RF model which showed a more pronounced deviation from the 45-degree line, the calibration curves fit well. In our DCA, the y-axis represents the net benefit, used to assess whether any specific clinical decision was more beneficial or harmful. Each point on the x-axis represents the threshold probability for distinguishing between patients with and without GGU. The results indicated that all models achieved a net clinical benefit compared to the all-or-none strategy, with the XGB model seemingly having the highest net benefit across the entire range of threshold probabilities, especially when the risk threshold was below 80%. Furthermore, we compared the accuracy of the ML models with that of the nomograms. The ROC curves of the training and internal validation groups showed that the accuracy of the XGB model was higher than that of the nomograms (training group AUC: 0.728 vs. 0.722; internal validation group AUC: 0.741 vs. 0.736), as shown in Fig. 3. Subsequently, we conducted an external validation of the XGB model using an external validation group and constructed the ROC, DCA, and calibration curves. DCA indicated that the XGB model still had a higher net benefit at risk thresholds of 70–90% (Fig. 4).

Development of a Web-Based Predictor for GGU in Prostate Cancer Patients

Finally, based on the XGB model, we developed a web-based predictor of GGU in PCa patients, providing a practical tool for clinicians to assess the risk of post-RP GGU. This predictor enhances clinical decision-making by allowing clinical physicians to quickly and accurately estimate a patient's likelihood of GGU, facilitating personalized treatment planning and potentially reducing the need for unnecessary interventions. By integrating easily accessible clinical data, this web-based predictor can be seamlessly incorporated into routine clinical practice, making it a valuable resource for improving patient outcomes and optimizing resource allocation in PCa management (Fig. 5) (https://alao-riskmodel.shinyapps.io/workrun7/).

In recent years, significant efforts have been made to refine biopsy techniques to improve detection rates and achieve greater consistency with post-RP pathology results. For example, Jiang X et al.(28) introduced regional saturation biopsy, enhancing the detection of clinically significant PCa and accurately identifying index lesions. Additionally, research by Qu L et al.(29) found that transperineal biopsy offers more accurate gleason grading than transrectal biopsy in patients without prior satisfactory magnetic resonance imaging (MRI). Furthermore, image fusion techniques have further increased detection rates, with targeted biopsy combined with ultrasound and multiparametric MRI reducing post-RP GGU compared to systematic biopsy. (30–32). Despite these advances, post-RP GGU remains a common issue, influencing decisions between active surveillance (AS) and aggressive treatment, highlighting the need for early screening models for high-risk PCa patients.

This study utilized the extensive SEER database and ML algorithms to establish a predictive model for identifying PCa patients at risk of post-RP GGU before making treatment decisions. ML offers several advantages, including the ability to handle large and complex datasets, uncover hidden patterns and relationships within the data, and continuously improve predictive accuracy through learning and optimization. These capabilities enable more personalized and precise risk assessments, leading to better-informed clinical decisions. Multifactorial LR analysis identified age, race, preoperative PSA level, biopsy ISUP grade, total number of biopsy cores, number of positive biopsy cores, and the percentage of positive biopsy cores as independent risk factors for post-RP GGU. Results indicated that PCa patients aged ≥ 70 years, with preoperative PSA levels ≥ 20 ng/mL, and over 10 positive biopsy cores were at higher risk of post-RP GGU. Additionally, consistent with previous findings (33, 34), the most common GGU occurred in ISUP GG1 and GG2, with the majority upgraded to GG2, followed by GG3. Notably, these cohorts are often considered for AS, suggesting a risk of missing curative opportunities if adverse pathology is discovered after biopsy. Approximately 30% of patients experience tumor progression during AS, and studies have shown that a significant portion of initially low-risk patients are upgraded to RP, indicating that AS may not suit those at risk of upgrading (35). Thus, identifying the risk factors associated with GGU is crucial to avoid undertreatment.

In this study, five supervised ML algorithms were employed to predict the GGU from biopsy to RP. XGBoost exhibited the best discriminative ability among all cohorts. Comparing our findings with those of other studies investigating post-RP upgrading and downgrading, many of which utilized small historical cohorts, our ML-based models surpassed those of Athanazio et al. (36), based on a cohort of 2529 PCa patients undergoing RP treatment, where their study yielded an AUC value of 0.699. Additionally, compared to Zhuang et al. (37), based on a cohort of 515 individuals who established an ML model predicting pathological consistency using combined systematic and MRI-targeted prostate biopsy, whose best-performing model was XGBoost, yielding an AUC value of 0.71, our models, except for RF, were markedly superior.

The strengths of this study lie in its large cohort and the utilization of nationally registered data. Additionally, multiple models were positively compared to the nomogram model, with XGB demonstrating superior discriminative ability in both the training and internal validation groups. The ML results are more reliable, enhancing the robustness of the outcomes, and the web-based predictor makes clinical application more convenient. However, this study had several limitations that warrant further investigation. Firstly, the overall population of our study was drawn from the SEER database and external validation only used a small sample size from a single center. Secondly, the SEER database provides only partial variables for patients with PCa and lacks crucial data such as prostate volume, PSA density, PI-RADS scores, post-RP surgical margin status, and testosterone levels, thereby restricting the inclusion of various potential risk factors impacting the GGU of patients with PCa. Integrating additional factors could notably augment model performance. Thirdly, the intrinsic opacity of ML algorithms may impede model interpretability. Despite these limitations, we assert the substantive robustness of our primary study findings.

In conclusion, we developed a predictive model for appraising post-RP GGU risk in PCa patients, leveraging the XGB algorithm, and developing a web-based predictor. Individuals at high risk for post-RP GGU are recommended for further detailed screening based on web predictor factors. This may aid clinicians in tailoring personalized treatments for patients with PCa.

active surveillance

AUC

area under curve

DCA

decision curve analysis

GBM

gradient boosting machine

GGU

gleason grade group upgrading

ISUP

International Society of Urological Pathology

logistic regression

machine learning

MRI

magnetic resonance imaging

NNET

neural network

post-RP

post-radical prostatectomy

PCa

prostate cancer

PSA

prostate-specific antigen

random forest

ROC

receiver operating characteristic

radical prostatectomy

SSER

Surveillance, Epidemiology, and End Results

XGB

XGBoost

Competing interests

All authors declare that there are no financial, personal, or other conflicts of interest in this study. The design, implementation, data analysis, and interpretation of the results in this research are entirely based on scientific principles, ensuring the fairness and objectivity of the research process and conclusions.

Ethics approval and consent to participate

The SEER database is an open and publicly accessible resource, thus not requiring approval or informed consent from an agency review committee. This study, which is a retrospective analysis of single-center data, does not involve patient-specific information and therefore does not require ethics committee approval.

Funding

This work was supported by Guangdong Basic and Applied Basic Research Foundation (2022A1515220032), Guangdong Medical Science and Technology Research Foundation(B2023195), Science and Technology Project of Zhongshan City (2020B1073), Zhongshan city people's hospital Major Project of Scientific Research Foundation (BG20228249) and Zhongshan City People's Hospital Outstanding Youth Project (SG2023106).

Author Contribution

JW and JT conceived and designed the study. Administrative support was provided by HH, YZ, RY, and YL. Data collection and assembly were carried out by JW, JT, and ZC. Data analysis and interpretation were performed by YZ, XL, QL, JW, and JT. All authors contributed to the article and approved the submitted version.

Acknowledgements

The authors would like to thank the medical team at Zhongshan People's Hospital for their support and assistance during this study.

Data Availability

The study's data can be accessed from the repository and downloaded using the following link (https://seer.cancer.gov). Additional information can be provided upon reasonable request.

ME JF, Siegel RL, Isabelle Soerjomataram M, Ahmedin Jemal D. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. 2024.
Epstein JI, Feng Z, Trock BJ, Pierorazio PM. Upgrading and downgrading of prostate cancer from biopsy to radical prostatectomy: incidence and predictive factors using the modified Gleason grading system and factoring in tertiary grades. Eur Urol. 2012;61(5):1019–24.
Brassetti A, Lombardo R, Emiliozzi P, Cardi A, Antonio I, Aldo S, et al. Prostate-specific antigen density is a good predictor of upstaging and upgrading, according to the new grading system: the keys we are seeking may be already in our pocket. Urology. 2018;111:129–35.
Çamur E, Coşkun A, Kavukoğlu Ӧ, Can U, Kara Ö, Çamur AD et al. Prostate volume effect on Gleason score upgrading in active surveillance appropriate patients. Archivio Italiano di Urol e Andrologia. 2019;91(2).
Wang G, Wang X, Du H, Wang Y, Sun L, Zhang M, et al. Prediction model of gleason score upgrading after radical prostatectomy based on a bayesian network. BMC Urol. 2023;23(1):159.
Kim H, Kim JK, Hong SK, Jeong CW, Ku JH, Kwak C. Role of multiparametric magnetic resonance imaging to predict postoperative Gleason score upgrading in prostate cancer with Gleason score 3 + 4. World J Urol. 2021;39:1825–30.
Li X, Qu G, Xu N, Xue X, Wei Y, Zheng Q, et al. Risk factors of ISUP Modified Gleason score upgrading after radical prostatectomy. Zhonghua nan ke xue = National. J Androl. 2016;22(5):415–9.
Beauval J-B, Ploussard G, Soulié M, Pfister C, Van Agt S, Vincendeau S, et al. Pathologic findings in radical prostatectomy specimens from patients eligible for active surveillance with highly selective criteria: a multicenter study. Urology. 2012;80(3):656–60.
Cohen MS, Hanley RS, Kurteva T, Ruthazer R, Silverman ML, Sorcini A, et al. Comparing the Gleason prostate biopsy and Gleason prostatectomy grading system: the Lahey Clinic Medical Center experience and an international meta-analysis. Eur Urol. 2008;54(2):371–81.
Imnadze M, Sjoberg DD, Vickers AJ. Adverse pathologic features at radical prostatectomy: effect of preoperative risk on oncologic outcomes. Eur Urol. 2016;69(1):143–8.
Bakavičius A, Drevinskaitė M, Daniūnaitė K, Barisienė M, Jarmalaitė S, Jankevičius F. The impact of prostate cancer upgrading and upstaging on biochemical recurrence and cancer-specific survival. Medicina. 2020;56(2):61.
May M. Eight ways machine learning is assisting medicine. Nat Med. 2021;27(1):2–3.
Moussa AS, Kattan MW, Berglund R, Yu C, Fareed K, Jones JS. A nomogram for predicting upgrading in patients with low-and intermediate‐grade prostate cancer in the era of extended prostate sampling. BJU Int. 2010;105(3):352–8.
Truong M, Slezak JA, Lin CP, Iremashvili V, Sado M, Razmaria AA, et al. Development and multi-institutional validation of an upgrading risk tool for Gleason 6 prostate cancer. Cancer. 2013;119(22):3992–4002.
Lee AJ, Wnorowski A, Ye N, Xu L, Naslund M, Wood BJ, et al. Validation of an MRI-based prostate cancer prebiopsy Gleason score predictive nomogram. Curr Urol. 2022;16(1):38–43.
Kather JN, Pearson AT, Halama N, Jäger D, Krause J, Loosen SH, et al. Deep learning can predict microsatellite instability directly from histology in gastrointestinal cancer. Nat Med. 2019;25(7):1054–6.
Diller G-P, Kempny A, Babu-Narayan SV, Henrichs M, Brida M, Uebing A, et al. Machine learning algorithms estimating prognosis and guiding therapy in adult congenital heart disease: data from a single tertiary centre including 10 019 patients. Eur Heart J. 2019;40(13):1069–77.
Zuo Y, Liu Q, Li N, Li P, Zhang J, Song S. Optimal 18F-FDG PET/CT radiomics model development for predicting EGFR mutation status and prognosis in lung adenocarcinoma: a multicentric study. Front Oncol. 2023;13:1173355.
Chen S, Jiang L, Zheng X, Shao J, Wang T, Zhang E, et al. Clinical use of machine learning-based pathomics signature for diagnosis and survival prediction of bladder cancer. Cancer Sci. 2021;112(7):2905–14.
Chen S, Jiang L, Gao F, Zhang E, Wang T, Zhang N, et al. Machine learning-based pathomics signature could act as a novel prognostic marker for patients with clear cell renal cell carcinoma. Br J Cancer. 2022;126(5):771–7.
Liu W-C, Li M-X, Qian W-X, Luo Z-W, Liao W-J, Liu Z-L et al. Application of machine learning techniques to predict bone metastasis in patients with prostate cancer. Cancer Manage Res. 2021:8723–36.
Wang X, Zhang X, Li H, Zhang M, Liu Y, Li X. Application of machine learning algorithm in prediction of lymph node metastasis in patients with intermediate and high-risk prostate cancer. J Cancer Res Clin Oncol. 2023;149(11):8759–68.
Tolles J, Meurer WJ. Logistic regression: relating patient characteristics to outcomes. JAMA. 2016;316(5):533–4.
Natekin A, Knoll A. Gradient boosting machines, a tutorial. Front Neurorobotics. 2013;7:21.
Rosa JP, Guerra DJ, Horta NC, Martins RM, Lourenço NC, Rosa JP et al. Overview of artificial neural networks. Using Artif neural networks analog Integr circuit Des Autom. 2020:21–44.
Liaw A, Wiener M. Classification and regression by randomForest. R news. 2002;2(3):18–22.
Chen T, Guestrin C, editors. Xgboost: A scalable tree boosting system. Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining; 2016.
Jiang X, Chen M, Tian J, Li X, Liu R, Wang Y, et al. Comparison of Regional Saturation Biopsy, Targeted Biopsy, and Systematic Biopsy in Patients with Prostate-specific Antigen Levels of 4–20 ng/ml: A Prospective, Single-center, Randomized Controlled Trial. European Urology Oncology; 2023.
Qu LG, Al-Shawi M, Howard T, Papa N, Poyet C, Kelly B, et al. Gleason grade accuracy of transperineal and transrectal prostate biopsies in MRI-naïve patients. Int Urol Nephrol. 2021;53:2445–52.
Ahdoot M, Wilbur AR, Reese SE, Lebastchi AH, Mehralivand S, Gomella PT, et al. MRI-targeted, systematic, and combined biopsy for prostate cancer diagnosis. N Engl J Med. 2020;382(10):917–28.
Nassiri N, Beeder L, Nazemi A, Asanad K, Um J, Gill I, et al. Step-by-Step: Fusion-guided prostate biopsy in the diagnosis and surveillance of prostate cancer. Int Brazilian J Urology: official J Brazilian Soc Urol. 2019;45(6):1277.
Andras I, Cata ED, Serban A, Kadula P, Telecan T, Buzoianu M, et al. Combined systematic and MRI-US fusion prostate biopsy has the highest grading accuracy when compared to final pathology. Medicina. 2021;57(6):519.
Demirtaş A, Sönmez G, Tombul ŞT, Demirtaş T, Akgün H. Comparison of the upgrading rates of International Society of Urological Pathology grades and tumor laterality in patients undergoing standard 12-core prostate biopsy versus fusion prostate biopsy for prostate cancer. Urol Int. 2019;103(3):256–61.
Li X, Wang Z-X, Zhu Y-P, Wang J, Yin Y-S, Zeng X-Y. Clinicopathological factors associated with pathological upgrading from biopsy to prostatectomy in patients with ISUP grade group ≤ 2 prostate cancer. Asian J Androl. 2022;24(5):487–93.
Kulkarni GS, Lockwood G, Evans A, Toi A, Trachtenberg J, Jewett MA, et al. Clinical predictors of Gleason score upgrading: implications for patients considering watchful waiting, active surveillance, or brachytherapy. Cancer. 2007;109(12):2432–8.
Athanazio D, Gotto G, Shea-Budgell M, Yilmaz A, Trpkov K. Global Gleason grade groups in prostate cancer: concordance of biopsy and radical prostatectomy grades and predictors of upgrade and downgrade. Histopathology. 2017;70(7):1098–106.
Zhuang J, Kan Y, Wang Y, Marquis A, Qiu X, Oderda M, et al. Machine learning-based prediction of pathological upgrade from combined transperineal systematic and MRI-targeted prostate biopsy to final pathology: a multicenter retrospective study. Front Oncol. 2022;12:785684.

No competing interests reported.

Download PDF

Editorial decision: Revision requested
27 Aug, 2024
Editor assigned by journal
22 Aug, 2024
Submission checks completed at journal
22 Aug, 2024
First submitted to journal
22 Aug, 2024

You are reading this latest preprint version

Establishment and Validation of a Machine Learning Model Predicting Post-Radical Prostatectomy Gleason grading group upgrading Author’s information

Status:

Version 1

Abstract

Background

Methods

Results

Conclusion

Figures

Introduction

Methods

Results

Discussion

Abbreviations

Declarations

Competing interests

Ethics approval and consent to participate

Funding

Author Contribution

Acknowledgements

Data Availability

References

Additional Declarations

Status:

Version 1