Risk of reclassification and novel tools in active surveillance

Abstract

Objective

Low-risk prostate cancer usually has an indolent course, and in this scenario, active surveillance (AS) is currently the preferred management, reducing overtreatment without compromising oncologic outcomes in well-selected patients. Two pillars of this strategy are patient selection and adequate monitoring strategy to detect early progression, leading to active treatment in the window of curability. The aim of this review is to assess the novel available tools and their role in AS.

Methods

We conducted a comprehensive review of the current literature addressing the risk of reclassification and studies evaluating traditional and emerging diagnostic and prognostic tools, including imaging modalities, biomarkers, and genomic classifiers, which could decrease reclassification rates and minimize the burden of serial exams during follow-up and, especially, the frequency of biopsies, which are still necessary to evaluate progression.

Results

The most significant follow-up cohorts demonstrate long-term treatment rates ranging from 30% to 45%, and 36% of low-risk prostate cancer patients have an upgrade in radical prostatectomy specimens. Despite the emergence of new diagnostic and prognostic tools, few of them have been validated or included in the assessment of eligible patients for AS. In the current guidelines, a combination of cT stage, Gleason score, prostate-specific antigen value, prostate-specific antigen density, and number of positive cores in biopsy is used to select patients for AS. However, various risk factors have been associated with the risk of reclassification, which reveals the need to incorporate better tools that may contribute to better risk stratification of patients eligible for AS.

Conclusion

Although AS is feasible and safe, the risk of progression is of great concern for patients and physicians, and monitoring is also an important part of the therapeutic strategy. Novel risk stratification tools are promising but need further validation to improve results and decrease the burden and anxiety that monitoring can bring to patients.

1. Introduction

Prostate cancer (PCa) is the most common solid tumor in men, accounting for more than 1 400 000 cases in 2020 worldwide [1]. The adoption of prostate-specific antigen (PSA) in the 1990s enabled early detection of PCa and increased diagnosis of low-risk disease [2]. It allowed a reduction in cancer-specific mortality [3]; however, as previous studies have demonstrated the long natural history of PCa and the relatively indolent course of low-risk disease [4,5], it was associated with overtreatment. Different tools were integrated into PCa screening, aiming to reduce the diagnosis of non-clinically significant tumors and, consequently, the morbidity associated with the recommended treatments. Still, despite the risk-stratification strategies, an essential proportion of all diagnoses is of indolent disease.

In the low-risk scenario, active surveillance (AS) emerged as an option for these patients with the primary goal of maintaining quality of life by deferring curative treatment of patients with non-clinically significant PCa according to guideline criteria, reducing overtreatment without compromising oncologic outcomes [6]. Two pillars of this strategy are patient selection to characterize the disease better and decrease the risk of reclassification and an adequate monitoring strategy to detect early progression, leading to active treatment in the window of curability.

Different risk stratification and monitoring strategies are recommended in the guidelines, which include PSA testing, digital rectal exams, multiparametric MRI, genomic classifiers, and biopsies. However, since the implementation of AS as the preferred treatment for selected patients, only a few novel tools have been validated or included in the assessment of eligible patients, which could decrease reclassification rates and minimize the burden of serial exams during follow-up and, especially, the frequency of biopsies, which are still paramount to evaluate progression.

In this review, we aim to examine predictive factors of reclassification, the newly available tools that might be integrated into the assessment and monitoring process, and current guideline recommendations on AS.

2. Risk of reclassification

The rationale for AS is to avoid overtreatment of low-risk PCa and monitor these patients for early detection of disease progression so they can be treated in the window of curability. The key to AS' safety and success is the ability to properly select patients and tailor the follow-up and intensity of monitoring. The largest follow-up cohorts demonstrate long-term treatment rates ranging from 30% to 45% [[7], [8], [9], [10]], and some studies show that patient characteristics, PSA, biopsy, and genomic factors are associated with outcomes in AS.

Various institutions and research groups have developed specific AS criteria, each incorporating factors such as Gleason score, PSA levels, PSA density, clinical staging, and the extent of tumor involvement in biopsy cores. Table 1 presents a comparative overview of the most widely recognized AS criteria, including the Epstein criteria [4], Prostate Cancer Research International Active Surveillance (PRIAS) [11], Memorial Sloan Kettering Cancer Center [12], University of California San Francisco [13], University of Miami [14], University of Toronto [7], Yonsei [15], and Johns Hopkins [9] protocols. Each criterion aims to balance the accuracy of risk stratification with a pragmatic approach to minimizing unnecessary treatment. The GAP3 study, a large-scale international registry encompassing 26 999 patients across 25 AS cohorts, reported a median follow-up period of 4.95 years. The data indicated that 55% of men remained free of treatment at the 10-year mark. Cancer-specific survival was highly favorable at 99.1% over 10 years, demonstrating the safety of AS as a management strategy. Overall survival for the same period was 78%, underscoring the long-term viability of AS in managing low--risk and favorable intermediate-risk PCa, reducing overtreatment while maintaining excellent oncological outcomes [16].

Table 1.

Contemporary AS criteria.

Protocol	Gleasonscore	PSA,ng/mL	PSAdensity, ng/mL/cm3	Clinicalstaging	Maximumpositivecore, n	Maximum tumorinvolvementper core, %
Epstein criteria [4]	≤6	≤10	<0.15	T1c–T2a	≤2	<50
PRIAS [11]	≤6	≤10	<0.2	T1c–T2	≤2	NS
MSKCC [12]	≤6	≤10	<0.15	T1c–T2a	≤3	NS
UCSF [13]	≤6	≤10	<0.15	T1c–T2	≤33%of biopsycores	≤50
University of Miami [14]	≤6	≤10	<0.15	T1c–T2a	≤2	<50
University of Torontoa [7]	≤6	≤10	NS	T1c–T2	≤3	≤20
Yonsei [15]	≤6	≤10	NS	T1c–T2	≤1	≤50
Johns Hopkins [9]	≤6	≤10	≤0.15	T1c	≤2	<50

AS, active surveillance; PSA, prostate-specific antigen; PRIAS, Prostate Cancer Research International Active Surveillance; MSKCC, Memorial Sloan Kettering Cancer Center; UCSF, University of California San Francisco; NS, not specified.

^aAS was also offered to patients with favorable intermediate-risk disease (PSA 10 ng/mL to 20 ng/mL and/or Gleason score 3+4) with significant comorbidities and a life expectancy of less than 10 years.

A meta-analysis evaluated the diagnostic accuracy of seven contemporary AS criteria for identifying clinically insignificant PCa and the Epstein criteria emerged as the most effective in predicting insignificant PCa, with the pooled diagnostic accuracy of insignificant PCa of 0.44 (95% CI 0.31–0.58). The study highlights the need for a globally consistent protocol in AS to optimize patient selection, suggesting confirmatory biopsies and potential use of MRI to reduce misclassification rates. Although new biomarkers, such as prostate cancer antigen 3 (PCA3), may offer additional predictive value, the analysis underscores the Epstein criteria as the optimal AS criteria currently, pending further validation and improvements in AS diagnostic standards worldwide [17].

Studies suggested that PSA density, free PSA percentage, and extent of cancer on biopsy cores are valuable predictors for progression on AS. Notably, African American men were shown to have higher progression risks, indicating a need for tailored criteria for this group. PSA kinetics, although widely studied, provided limited short-term predictive value [18]. Also, high PSA density, PSA kinetics, and high genomic score were associated with AS failure [19], and a recent study has indicated that small prostate volume and low testosterone would also be predictors for reclassification [20].

Tohi et al. [21] analyzed the outcomes of radical prostatectomy (RP) in 136 Japanese patients who were part of the PRIAS-JAPAN AS cohort, focusing on differences between those who underwent surgery according to protocol-defined criteria (on-criteria) and those who did not (off-criteria). Of these patients, 25.7% were treated off-criteria, often influenced by elevated PSA levels rather than biopsy-confirmed progression. The analysis showed that pathological findings, such as rates of unfavorable pathology, did not significantly differ between the on-criteria and off-criteria groups. Additionally, 3-year PSA recurrence-free survival rates were comparable (96.7% for on-criteria and 95.5% for off-criteria), highlighting that elevated PSA alone could lead to overtreatment, as it was not consistently linked to worse pathology outcomes. The findings emphasize careful clinical decision-making to avoid unnecessary surgery, promoting balanced AS protocols that minimize overtreatment while preserving curative opportunities [21].

The American Urological Association guidelines address the concern of patients at higher risk of progression during AS and highlight that active treatment may be offered for selected patients, such as those with a significant family history, germline mutations, high-volume disease, and high PSA density [22]. Carter et al. [23] explored the association between germline mutations in DNA repair genes (BRCA1/2 and ATM) and grade reclassification among men undergoing AS for PCa. The analysis included 1211 men from Johns Hopkins' and North Shore cohorts. Results indicated that carriers of BRCA2 mutations had a significantly higher risk of upgrading, with a five-fold increase in risk for reclassification from Grade Group 1 to Grade Group 3 or higher, suggesting that men with these mutations may have a more aggressive cancer phenotype, informing the decision-making process between AS and curative treatment [23].

Moreover, PCa incidence and mortality vary worldwide due to biological, racial, and environmental variations that can influence PCa behavior. Kim et al. [24] evaluated the ability of contemporary AS protocols in selecting eligible Korean men and found that, among 1006 patients, the rates of Gleason score upgrading to ≥7 ranged from 41.6% to 50.6% across the six AS criteria. Extracapsular extension rates varied between 4.1% and 8.5%, and seminal vesicle invasion ranged from 0.5% to 1.6%. The overall misclassification rates ranged from 44.5% to 54.8%, indicating that the current AS criteria may have limitations when applied to Korean men [24]. In line with these findings, Yamada et al. [25] validated the AS criteria in Asian men and found that predictive rates were lower in Asian men compared to men in the United States, ranging from 27% to 51%. In this study, prostate volume was a significant predictive factor and including it in the criteria improved prediction rates up to 66.7% [25]. AS protocols should incorporate ethnic-specific clinical characteristics to enhance predictive accuracy, highlighting the importance of developing tailored AS protocols for more precise patient selection across diverse demographic groups.

Finally, AS criteria of major cohorts were established before the recommendation for the use of MRI to guide targeted biopsies. Bloom et al. [26] compared systematic biopsy alone with combined biopsy, which included systematic and targeted biopsy, prior to RP. It revealed lower rates of adverse pathology in the MRI and combined biopsy group compared to systematic biopsy alone [26]. Furthermore, Baboudjian et al. [27] found no statistical association between tumor burden at biopsy and adverse pathology in intermediate-risk patients with International Society of Urological Pathology (ISUP) Grade 2, while tumor size ≥15 mm on multiparametric MRI was significantly associated with adverse outcomes and biochemical recurrence.

3. Novel tools for patient selection and monitoring

Despite the variation in the current guidelines, a combination of cT stage, Gleason score, PSA value, PSA density, and the number of positive cores in biopsy is used to select patients for AS, as shown in Table 2. Although they are recognized as prognostic features, studies have demonstrated that 36% of low-risk PCa patients have an upgrade in RP specimens [28], which reveals the need for better tools for patient selection. Imaging, serum and urinary biomarkers, and digital histopathology are being studied in this setting and may contribute to better risk stratification of patients eligible for AS.

Table 2.

Guidelines' recommendations on AS.

Guideline	Inclusion criteria	Confirmatory testing	FU	MRI	PSMA PET/CT	Genomic classifier
NCCN [29]	-Life expectancy ≥10 yr -Very low-risk PCaa,low-risk PCab, or favorable intermediate-risk PCac	-Confirmatory prostate biopsy for all Pts within 1–2 yr -Within 6–12 m if no pre-biopsy MRI, MRI with PSAd, and/or molecular tumor analysis	• Monitoring should be performed no more frequently than - PSA every 6 m - DRE every 12 m - MRI every 12 m - prostate biopsyevery 12 m	-Prior biopsy for all Pts -For AS monitoringevery 12 m	NR for men eligible to AS	• May be used for Pts eligible for AS but with higher risk for progression in - low-risk PCa if high PSAd or high-volume disease - intermediate-risk PCa •Should not be repeated during AS monitoring
AUA [22]	-Low-risk PCa or favorable intermediate-risk PCa	-If no pre-biopsy MRI, performing MRI, and proceeding to target biopsy (if PI-RADS 4 or 5) or repeat biopsy within 12 m (if PI-RADS 1, 2 or 3)	-Serial PSA every 6 m -DRE every 1–2 yr -Periodic prostate bio-psy	-For all Pts selected for AS -During monitoring if PSA increases or abnormal DRE -Should not replace surveillance biopsy	NR for men eligible to AS	•Should not be routinely used • May be used to improve risk stratification if - high-volume GS 6 - favorable intermediate-risk PCa •Do not perform serial genomic testing
EAU [72]	-ISUP Grade 1, cT1c–T2a, PSA <10 ng/mL, and PSAd <0.15 ng/mL/cm3	-Confirmatory biopsy if no pre-biopsy MRI within 6–12 m	-PSA every 6 m -DRE every 12 m -Repeated biopsy every 2–3 yr	-Perform if PSA is rising (PSA DT <3 yr) -Serial MRI can improve adverse pathology during FU	NR for men eligible to AS	•Should not be used in standard clinical practice

AS, active surveillance; NCCN, National Comprehensive Cancer Network; AUA, American Urological Association; EAU, European Association of Urology; yr, years; PCa, prostate cancer; GG, Grade Group; PSA, prostate-specific antigen; PSAd, PSA density; m, months; PI-RADS, Prostate Imaging-Reporting and Data System; DRE, digital rectal exam; DT, doubling time; FU, follow-up; GS, Gleason score; Pts, patients; NR, not recommended; ISUP, International Society of Urological Pathology; PSMA, prostate-specific membrane antigen; PET, positron emission tomography.

^acT1c, PSA <10 ng/mL, GG1, <3 biopsy cores, <50% core involvement, and PSAd <0.15 ng/mL/cm³.

^bcT1c–T2a, PSA <10 ng/mL, GG1, <3 biopsy cores, and low PSAd.

^cLow percentage of pattern 4, low tumor volume, low PSAd, and/or low genomic risk.

3.1. Prostate MRI

Although not recommended for PCa screening, MRI is recommended before prostate biopsy since it increases clinically significant PCa detection [29]. The Prostate Imaging-Reporting and Data System (PI-RADS) score is correlated with high-grade lesions, and it is able to predict unfavorable pathology [30], including worse oncologic outcomes after treatment in intermediate-risk patients with PI-RADS 5 [31,32]. Drost et al. [33] demonstrated in a meta-analysis that the MRI pathway, when MRI is used to help decide whether to do a biopsy, avoided 66% of Grade Group 1 PCa and detected 12% more ISUP 2 or higher PCa.

Klotz et al. [34] demonstrated that among patients undergoing AS, pre-biopsy MRI is associated with fewer AS failures and lower progression rates during monitoring. For patients eligible for AS whose biopsy was performed without MRI, the addition of MRI-targeted confirmatory biopsy is recommended immediately if MRI demonstrates suspicious findings [22] or in 6–12 months if it demonstrates PI-RADS 1, 2, or 3 [22,35] since studies found ISUP ≥2 upgrading rates from 10% to 48% [36,37] and reclassification in 59% of patients considered for AS [38].

However, the role of MRI for AS monitoring remains controversial. It provides relevant information, but studies have shown insufficient accuracy in replacing periodic biopsies since follow-up biopsies triggered just by MRI progression would miss 12% of histologic progression [39,40]. In 2017, Moore et al. [41] published the Prostate Cancer Radiological Estimation of Change in Sequential Evaluation recommendations standardizing MRI reports and identifying patients who progress during AS, and a recent upgrade has been published [42]; however, some limitations have to be addressed with regard to variability in measuring lesions, lack of standardization of scan timing and growth thresholds, and interobserver variability [43].

In recent years, there has been a growing interest in radiomics. This quantitative assessment is an image analysis method that extracts many features missed by the human eye from scans through algorithms. Radiomics has been studied in all PCa scenarios, mainly in tumor detection and characterization, with some studies in AS patients demonstrating superior accuracy in identifying clinically significant PCa compared to the PI-RADS score [44,45]. Sushentsev et al. [46] compared the Prostate Cancer Radiological Estimation of Change in Sequential Evaluation score with MRI-derived delta radiomics, computed as the arithmetic difference between the final and baseline features, demonstrating comparable performance for histopathologic progression, with the advantage of being a more standardized method. In 2023, Sushentsev et al. [47] developed a predictive model combining PSA density with time series radiomics, which is based on longitudinal changes in radiomics features extracted from all MRI scans during follow-up. This study's time-series radiomics outperformed delta radiomics and showed comparable performance to expert radiologists for predicting PCa progression on AS [47]. Despite being a quantitative and interpretable solution, as mentioned by the authors, radiomics in this setting still requires validation for its use in clinical practice.

3.2. Prostate-specific membrane antigen (PSMA) positron emission tomography (PET)/CT

PSMA PET/CT has a well-defined role in biochemical recurrence and the evaluation of high-risk patients, leading to a change in management [48]. Among patients eligible for AS, the favorable intermediate-risk patients considered for this modality are a very heterogeneous group, and novel risk stratification methods could improve patient selection.

The PRIMARY study, which tested the association of prostate MRI with PSMA PET/CT for the non-invasive diagnosis of PCa, demonstrated an association between the maximum standardized uptake value and ISUP grade. The probability of clinically significant PCa increased exponentially with increasing maximum standardized uptake value [49]. In a post hoc PRIMARY study analysis, Emmett et al. [50] proposed the PRIMARY score, which considers the pattern and intensity of intra-prostatic PSMA. They developed a 5-point scale analogous to PI-RADS to report the primary tumor's features. In the analysis, the score demonstrated the potential for high accuracy in diagnosing clinically significant PCa, with the PRIMARY score of 5 having a positive predictive value of 100% in this scenario [50]. Akcay et al. [51] also demonstrated a significant correlation between the PRIMARY score and ISUP grade, while no substantial correlation was found between the PI-RADS and the ISUP grade. However, it is essential to note that the PRIMARY score requires validation before its implementation in clinical practice.

Heetman et al. [52] evaluated the addition of PSMA PET/CT for patient selection for AS and found upgrading of 9% of the patients when PSMA PET/CT was used for targeted biopsy. Pepe et al. [53], who only enrolled low-risk patients, did not show improvement in clinically significant PCa detection with PSMA PET/CT; however, with a better false positive rate compared to MRI (20% vs. 43.3%) and a better negative predictive value (85.7% vs. 57.1%), PSMA would have avoided 80% of the scheduled biopsies. There is no evidence of the use of PSMA PET/CT for patients on AS, and trials are currently in progress [54,55].

3.3. Blood biomarkers

Blood and urinary biomarkers are primarily used in the PCa screening scenario in a pre-diagnosis stage, but due to their characteristics, they also show promise in AS patients. Prostate Health Index (PHI), which is a blood test that combines free PSA, total PSA, and isoform [−2]proPSA, was retrospectively evaluated in 167 patients on an AS protocol and correlated with biopsy reclassification, with [−2]proPSA being the strongest predictor of change in Gleason score throughout follow-up [56]. Moreover, Hirama et al. [57] demonstrated the ability of PHI to discriminate significant cancer and demonstrated that it may be a predictor of reclassification during AS, and other studies showed the superiority of PHI over PCA3 in clinically significant PCa in men on AS [58,59]. A prospective cohort study conducted at 19 Japanese institutes evaluated the predictive power of [−2]proPSA-related parameters for reclassification at the first-year protocol biopsy in patients on AS. Multivariate analyses identified %p2PSA and PHI at inclusion and before the 1-year biopsy as significant predictors. Receiver operating characteristic analysis established optimal cutoff points, with PHI demonstrating the highest clinical utility according to decision curve analysis and the most effective marker for guiding clinical decisions in AS [60].

The 4Kscore®, a blood test that uses the dosage of four kallikreins with clinical features to improve risk stratification in PCa screening, was shown to predict high-grade cancer in the first biopsy after diagnosis but not in subsequent biopsies in AS patients [61]. Furthermore, Borque-Fernando et al. [62] and Bryant et al. [63] suggest that 4Kscore® can play a role in deciding whether to perform a confirmatory biopsy in AS since it is a tool to identify clinically significant PCa [64].

3.4. Urinary biomarkers

PCA3 and TMPRSS2:ERG are the most studied urinary biomarkers in the patient selection setting for AS. PCA3 is an mRNA overexpressed in PCa. While Tosoian et al. [65] associated higher PCA3 with grade reclassification, other studies demonstrated a modest association with reclassification in the first biopsy but not in subsequent biopsies [66,67].

TMPRSS2:ERG gene fusion is the most frequent genomic alteration in PCa [68], and it has been correlated with disease volume [69], aggressive cancer [69], and disease progression [70] in patients on AS. However, Newcomb et al. [66], evaluating biomarkers to predict reclassification, did not find an association with TMPRSS2:ERG levels.

In this scenario, despite being recommended in specific situations during prostate screening as they are predictors of clinically significant PCa, they still have no defined role in AS. Therefore, there is no recommendation in this setting.

3.5. Tissue biomarkers

High PSMA expression on biopsies was demonstrated to be an independent prognostic marker at the time of initial diagnosis [71]. Ahmadi et al. [72] examined PSMA expression with immunohistochemistry in baseline biopsies. They found that overexpression of PSMA in malignant glands was an independent predictor of early PCa progression in patients managed on AS [72].

Tissue biomarkers based on gene or protein expression can also contribute to risk stratification for men eligible for AS who would benefit from refining the risk assessment. Commercially available tissue biomarkers (Oncotype Dx®, Prolaris®, Decipher®, and ProMark®) were validated in retrospective studies and demonstrated prognostic accuracy to better identify men who would benefit most from the AS strategy [29,73].

Oncotype Dx® is a 17-gene RNA expression assay with a Genomic Prostate Score (GPS) ranging from 0 to 100. Different weights are given to each gene, and higher scores indicate the risk of more aggressive disease [74]. It has been demonstrated that Oncotype Dx® can predict adverse surgical pathology and recurrence after RP [75,76]. In men on AS, a higher GPS predicted biopsy upgrading [77,78] and adverse pathology [75]. However, a prospective cohort of men eligible for AS was evaluated by Lin et al. [79], who retrospectively performed GPS testing in biopsy tissue. There was no significant association of GPS with biopsy upgrading during AS monitoring or adverse pathology at RP after AS [79]. Moreover, GPS was not correlated with disease volume among men with National Comprehensive Cancer Network (NCCN) very low- and low-risk PCa classification [80].

The Prolaris® test calculates a score by measuring the mRNA expression of 31 cell cycle progression genes and 15 housekeeping genes. It was developed to predict biochemical recurrence and death from PCa [81] and was validated as a risk-stratification tool [82,83]. Lin et al. [84] evaluated the role of Prolaris® in patient selection for AS. They showed an improvement in risk stratification, suggesting it could also be used in this setting [84].

Decipher® is a 22-gene RNA biomarker assay that generates a score from 0 to 1. It was initially developed in an RP cohort to predict metastasis [85] and was validated as a superior prognostic factor compared to clinical and pathologic features [86]. Later, Kim et al. [28] assessed Decipher® in biopsies of patients who underwent RP but who would be eligible for AS. Decipher® was evaluated as a risk stratification tool for the absence of adverse pathology; it was found to be an independent predictor of adverse pathology, adding value to the Cancer of the Prostate Risk Assessment score, which was not a significant predictor alone. Press et al. [87] also associated the Decipher® score with short-term biopsy upgrades during AS.

ProMark® is a platform that analyzes eight protein-based biomarkers. It was developed to be used in prostate biopsy specimens and has the advantage of detecting high-grade molecular features in small tissue samples [88]. The score ranges from 0 to 100, and it was validated as an additional tool to improve the D'Amico or NCCN risk stratification, helping in the therapeutic decision [89]. No further studies evaluated the prognostic value of ProMark® in the selection and monitoring of patients on AS.

However, as the oncological impact of the tests has yet to be demonstrated in prospective studies, they should only be used in the group of patients for whom they can provide additional information, such as those with low-risk or favorable intermediate-risk PCa who opt for AS. Moreover, the use of these molecular classifiers is still limited by their high cost, the need for adequate integrity of biopsy specimens, and the fact that, despite being based on tumor tissue, they may not adequately represent the heterogeneity of the tumor.

3.6. Artificial intelligence (AI)

AI has been applied in many areas of medicine, from diagnosis to assistance in cancer therapy [90]. In localized PCa, Esteva et al. [91] created a multimodal AI, a deep learning and self-supervised learning system that uses clinical and digital histopathology imaging data. The model validated prognostic biomarkers from the data of five prospective randomized phase 3 studies. The NCCN risk stratification was used as a baseline comparator. The multimodal AI performed better in all results, confir-ming that it is a prognostic factor capable of predicting long-term outcomes of biochemical recurrence, distant metastasis, and cancer-specific mortality in PCa [91].

NCCN guidelines describe AI digital pathology as a tool that may be used only if the results impact the therapeutic decision. In AS, it would not be recommended for very low-risk patients. Still, it could have a role in low-risk patients with criteria associated with reclassification, such as high PSA density and high tumor volume, as well as for favorable intermediate-risk patients [29].

AI, compared to urinary, blood, and tissue biomarkers, does not require collecting or processing biological samples. Furthermore, with the increasing adoption of digital histopathology, multimodal AI has the potential to be used on a global scale, but its availability is still limited.

4. Future directions and conclusions

AS is currently the preferred management option for an important portion of patients diagnosed with localized PCa. Although AS has been shown to be feasible and safe, the risk of progression is of great concern for patients and physicians, and monitoring is also an important part of the therapeutic strategy.

The well-known clinical and pathologic risk factors, such as PSA, digital rectal exam, and Grade Group, used to stratify patients need to be improved or combined with novel stratifications and monitoring tools for better results in AS. The evolving role of imaging, with better acquisition methods, more tailored classification systems, quantitative imaging analysis, and inclusion of new imaging methods, has shown promising results but still needs validation in this setting. The molecular analysis of tumors, the use of AI, and new biomarkers may increase the precision of risk stratification and decrease the burden and anxiety that monitoring can bring to the patients.

Author contributions

Study concept and design: Bárbara Vieira Lima Aguiar Melão, Karim A. Touijer.

Data acquisition: Bárbara Vieira Lima Aguiar Melão.

Data analysis: Bárbara Vieira Lima Aguiar Melão, Lucas Mendes Nogueira.

Drafting of manuscript: Bárbara Vieira Lima Aguiar Melão, Lucas Mendes Nogueira.

Critical revision of the manuscript: Bárbara Vieira Lima Aguiar Melão, Lucas Mendes Nogueira, Karim A. Touijer.

Conflicts of interest

The authors declare no conflict of interest.