*Review* **Prostate Cancer Liquid Biopsy Biomarkers' Clinical Utility in Diagnosis and Prognosis**

**Milena Matuszczak <sup>1</sup> , Jack A. Schalken <sup>2</sup> and Maciej Salagierski 1,\***


**Simple Summary:** In prostate cancer, overdiagnosis and overtreatment is a common problem for clinicians. Accurate diagnosis and prognosis are essential to avoid unnecessary biopsy and to increases the effectiveness of treatment. A new, easy-to-use and non-invasive test based on liquid biopsy biomarkers such as Progensa PCA3, MyProstateScore, ExoDx, SelectMDx, PHI, 4K, Stockholm3 and ConfirmMDx have been developed to improve diagnosis, prognosis and to help guide the decision-making process. This article provides an overview of the above-mentioned commercial tests. The performance and financial aspects of the tests have been compared using available studies. Then the application of biomarker tests as an adjunct to multiparametric MRI in the diagnosis, prognosis and monitoring of prostate cancer has been discussed.

**Abstract:** Prostate cancer (PCa) is the most common cancer in men worldwide. The current gold standard for diagnosing PCa relies on a transrectal ultrasound-guided systematic core needle biopsy indicated after detection changes in a digital rectal examination (DRE) and elevated prostate-specific antigen (PSA) level in the blood serum. PSA is a marker produced by prostate cells, not just cancer cells. Therefore, an elevated PSA level may be associated with other symptoms such as benign prostatic hyperplasia or inflammation of the prostate gland. Due to this marker's low specificity, a common problem is overdiagnosis, which leads to unnecessary biopsies and overtreatment. This is associated with various treatment complications (such as bleeding or infection) and generates unnecessary costs. Therefore, there is no doubt that the improvement of the current procedure by applying effective, sensitive and specific markers is an urgent need. Several non-invasive, cost-effective, high-accuracy liquid biopsy diagnostic biomarkers such as Progensa PCA3, MyProstateScore ExoDx, SelectMDx, PHI, 4K, Stockholm3 and ConfirmMDx have been developed in recent years. This article compares current knowledge about them and their potential application in clinical practice.

**Keywords:** cancer biomarkers; prostate cancer; liquid biopsy; prognosis; diagnosis; early detection

#### **1. Introduction: Prostate Cancer Diagnosis**

Prostate cancer (PCa) is the most common cancer in men and the second most common cause of mortality in this population in the United States, with 191,930 new cases and 33,330 deaths in 2020 [1]. Globally, there are approximately 1,276,106 new cases and 358,989 deaths each year [2]. The lifetime risk of being diagnosed with prostate cancer is estimated to be 1 in 9 men, while the risk of death is, fortunately, not as high at around 2% [1].

There is an emerging role for liquid biopsy in PCa, which has excellent potential in preoperative medicine. It is a minimally invasive procedure, analysing even small numbers of targets, which allows its usefulness in screening, diagnosis, prognosis, follow-up and therapeutic management [3]. This review compares the diagnostic and prognostic utility of prostate cancer tests. Good clinical outcomes can be achieved by accurate diagnosis

**Citation:** Matuszczak, M.; Schalken, J.A.; Salagierski, M. Prostate Cancer Liquid Biopsy Biomarkers' Clinical Utility in Diagnosis and Prognosis. *Cancers* **2021**, *13*, 3373. https:// doi.org/10.3390/cancers13133373

Academic Editors: Fabrizio Bianchi and Craig N. Craig Robson

Received: 1 April 2021 Accepted: 29 June 2021 Published: 5 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

followed by acute treatment or active surveillance in patients with disease located within the gland. There is an unmet clinical need for non-invasive, easily performed diagnostic tests to assess whether a prostate biopsy is indicated. The EAU 2020 guidelines [4] recommend mpMRI before the first biopsy in men with a clinical suspicion of prostate cancer (PCa). Indeed, when mpMRI shows lesions suspicious for PCa (i.e., PI-RADS ≥ 3), targeted biopsy (TBx) and systemic biopsy (SBx) are recommended in patients who have not had a previous biopsy. It therefore represents an important diagnostic tool, and its combination with biomarkers further improves the accuracy of the initial diagnosis of PCa.

The traditional diagnosis of PCa (Figure 1) is based on the assessment of serum prostate-specific antigen (PSA) levels, digital rectal examination (DRE), followed by biopsy under the guidance of transrectal ultrasonography (TRUS). In screening programmes, high PSA levels, despite a normal DRE, lead to the diagnosis of PCa in more than 60% of asymptomatic patients. Serum PSA levels are commonly used for detection, risk stratification and monitoring of PCa [5]; unfortunately, it results in a high number of unnecessary biopsies and detection of asymptomatic cancers with low clinical risk [6]. The reason may be that PSA has a low positive predictive value (~30%) and poor specificity, being organ rather than cancer-specific. This highlights the need to develop more precise methods to identify clinically relevant PCa, such as liquid biopsy-derived biomarkers.

**Figure 1.** Suggested workflow for utilisation of prostate cancer biomarkers.

As prostate cancer is a heterogeneous disease, urologists, after identifying the presence of disease during the baseline assessment, focus primarily on assessing the risk group. Risk groups have been classified since 2014 using a classification system with five distinct Grade Groups based on modified Gleason score groups. Group 1 = Gleason score ≤ 6, Group 2 = Gleason score 3 + 4 = 7, Group 3 = Gleason score 4 + 3 = 7, Group 4 = Gleason score 4 + 4 = 8, Group 5 = Gleason score 9 and 10.

Currently, the gold-standard test to confirm all of the above clinical situations is the histopathological result of a prostate biopsy.

Unfortunately, this invasive procedure is painful, expensive and may pose a risk of complications (e.g., infection or sepsis). Furthermore, the procedure is prone to significant sampling error. It is therefore important to avoid unnecessary biopsies [7,8].

Liquid biopsy biomarkers are proving to be a promising new diagnostic and prognostic approach to help optimise the pre-biopsy decision and stratify whether the patient requires treatment or can be monitored under active surveillance.

#### **2. Material and Methods**

A literature review was performed by searching MEDLINE/PubMed, Google Scholar and and CrossRef electronic databases to identify articles published from January 2000 to October 2020 whose methods included commercially available prognostic and diagnostic prostate cancer liquid biopsy biomarkers or contain information about the characteristics of a relevant biomarker. The search terms included ConfirmMDx, ExoDx, MiPS, PCA3, PHI, SelectMDx, Stockholm3, 4Kscore and and prostate cancer liquid biopsy using search terms database = specific—medical subject headings terms in various combinations appropriate to the research objective. Articles on biomarkers not available in clinical practice or studies based on less than 40 patients were excluded.

#### **3. Urine Biomarkers**

Urine is obtained non-invasively and contains fluid excreted from the prostate gland, which may contain products from prostate cancer cells [9]. For many urinary biomarkers, performing a DRE is crucial as it increases the excretion of fluid from the prostate. To date, four tests are available with proven clinical utility.

#### *3.1. Prostate Cancer Antigen 3 (PCA3)*

The Progensa PCA3 test (Hologic Inc., Marlborough, MA, USA) is a test that measures PSA messenger RNA (mRNA) and PCA3 mRNA detectable in the first catch urine sample after DRE.

Prostate cancer antigen 3 (PCA3, previously called "DD3") is a long, non-coding RNA (lncRNA) that is overexpressed in 95% of prostate cancers [9]. The test is based on the fact that 60–100 times more PCA3 gene mRNA is detected in prostate cancer cells compared to non-cancerous prostate tissue.

The PCA3 score is calculated using the Progensa PCA3 method. The test result represents the PCA3/mRNA PSA ×1000 ratio [10].

It is the first urine biomarker test to be approved in 2006 by the European Union, Canada [11] and in 2012 by the FDA. The FDA recommends its use in men ≥ 50 years old to support repeat biopsy decision-making in whom one or more previous prostate biopsies have been negative and for whom repeat biopsy is recommended based on current standards of care [12]. However, some clinical studies [13,14] report the benefits of using the test as early as the first biopsy.

Although the FDA recommends a cut-off PCA3 score = 25, many studies [12,13] suggest a cut-off score of 35 as a more optimal cut-off point. Establishing a cut-off point appears to be of vital importance.

A study [12] evaluated different PCA3 score cut-off points: 10 and 35. For these values, the sensitivity was 87% and 58%, respectively, and the specificity was 28% and 72%, respectively. The results showed that a PCA3 score cut-off of 35 could provide an optimal

balance between sensitivity (58%) and specificity (72%) for the diagnosis of PCa and was superior to PSA (Table 1).

Although the study [12] demonstrated high sensitivity and specificity, the ability to improve prostate cancer detection was not shown. For this reason, Wei et al. conducted a prospective validation trial on 859 men [14] to assess whether the PCA3 score could improve the PPV for initial biopsy and NPV for repeat biopsy. The results were PPV = 80% for detecting any PCa at initial biopsy and NPV = 88% at repeat biopsy. This showed that at initial biopsy, a PCA3 score > 60 increases the likelihood of detecting PCa, and at repeat biopsy, a PCA3 score < 20 indicates a low risk of detecting PCa at biopsy [14].

A systematic review and meta-analysis of studies (with a threshold of 35) [13] yielded the following overall values: AUC = 0.734, sensitivity 69% and specificity 65%. These results support the greater clinical utility of cut-off point = 35 than 25 (FDA approved).

Determining the best cut-off value is controversial, especially for primary biopsy—the available studies are very heterogeneous. Several have highlighted that PCA3 does not perform well at a single threshold, showing a high NPV below the low cut-off and a high PPV above the high cut-off, with a grey zone in between—reflecting prostate cancer specificity [14].

Roobol et al., in a publication [15], highlight men with a PCA3 score ≥ 100 and no PCa in a biopsy. This study combines data from the initial and re-biopsies that provided a PPV of 52.2% in men with PCA3 ≥ 100, resulting in almost 50% unexplained high results. To date, there is no explanation why PCA3 scores can be excessively high despite the absence of biopsy-detectable PCa.

Publications [16–18] do not show a relationship between PCA3 value and prostate cancer aggressiveness (Gleason score). A high PCA3 level, due to its low specificity, does not help assess prognostic parameters and is therefore of low utility in clinical practice, as it does not provide an answer to how to proceed with the patient. For this reason, to detect patients who require rapid and radical treatment, it is reasonable to use newer, more sensitive and specific diagnostic tools, e.g., SelectMDx (MDxHealth, Inc., Irvine, CA, USA).

Numerous studies [14,19] have shown that the diagnostic value of the test increases when adding other predictors (i.e., age, PSA value, DRE result or prostate volume). Therefore, the producer recommends its use in combination with standard diagnostic parameters [20].

To determine the clinical utility of the PCA3 test in African Americans, Feibus et al. conducted a study [21] (Table 2) on a racially diverse group of men, where 60% of the participants were African American. They demonstrated that the PCA3 test in African Americans also improves the ability to predict the presence of any prostate cancer and high malignancy.

Ochiai et al. [22] (Table 2) examined the diagnostic utility of PCA 3 in Japanese men undergoing prostate biopsy. They achieved a similar diagnostic value to that obtained in men in Europe and the USA. The PCA3 score for men with prostate cancer was significantly higher than for men with negative biopsy results. Furthermore, they showed that also in Japanese men, PCA3 was significantly better than PSA in predicting PCa.

The reported clinical utility of the study mentioned above on the Japanese population and the desire to verify the promising reports of Shen et al. [23] (Table 2) on a small group of Chinese men (prostate cancer patient group (*n* = 35), BPH patient group (*n* = 64)), inspired other researchers to study the Chinese population. Wang et al. conducted a study on a cohort of 500 Chinese men [24] (Table 2). This study showed a moderate improvement in diagnostic accuracy using PCA3 during the initial prostate biopsy. In patients qualified for initial biopsy (PSA ≥ 4 and/or suspicious DRE), the Progensa test was not used, but the RC-PCR-based PCA3 test was used. The values obtained were sufficient to distinguish positive from negative prostate biopsy results but were not correlated with PCa aggressiveness.

In a study [25] (Table 2) involving Latino Americans, results were comparable to those obtained for other populations, indicating its potential use in Latino Americans with persistently elevated PSA and previous negative biopsies.



PCa—prostate cancer, PCA3—PCa antigen 3, PPV—positive predictive value, PSA—prostate-specific antigen, RASSF1—ras association domain family member 1, Ref—references, Se—sensitivity, Sp—specificity.


PCA3 shows more significant diagnostic and prognostic potential when combined with other biomarkers, such as TMPRSS2 fusion: ERG [36] hK2, PSA [37] and PSAD [38]. Currently, some researchers are making efforts to develop more precise detection methods for PCA3 [39,40].

#### *3.2. Mi-Prostate Score (MiPS)*

MyProstateScore (MPS, LynxDx, Ann Arbor, MI, USA (previously known as MiPS— Michigan Prostate Score)) is an algorithm that measures mRNA, PCA3 and TMPRSS-ERG (abnormal fusion of TMPRSS2 and ERG (T2-ERG)) expression in urine from the first collection after DRE and serum PSA.

More than 50% of patients with PCa have an ERG gene fusion with TMPRSS2 [41]. The presence of this translocation has been shown to be associated with poor patient prognosis—an increased risk of recurrence and mortality from PCa.

The test is indicated for men with suspicious PSA levels who are being considered for initial or repeat biopsy. The test result validates the need for biopsy and predicts the risk of high-grade prostate cancer (GS > 7) in a diagnostic needle biopsy [42–44]. Values range from 0 to 100, and the higher the score, the greater the risk of aggressive cancer.

In 2013, Salami et al. [44] showed that the MPS test was significantly more accurate than any single variable (TMPRSS2-ERG AUC = 0.77 compared with 0.65 for PCA3 and 0.72 for serum PSA alone), AUC was 0.88, with specificity and sensitivity of 90% and 80%, respectively (Table 3).

A pivotal study published by Tomlins et al. [43] in 2016 indicated the high diagnostic value of MPS, AUC = 0.751 for detecting PCa on biopsy and AUC = 0.772 for detecting clinically significant PCa (defined as Gleason ≥ 7), which was significantly better than for PSA alone (AUC = 0.651) (Table 3).

In a prospective study [45] involving 1077 men, MPS was shown to increase the detection of aggressive prostate cancers compared with PSA alone. When the cut-off point was set at 95% sensitivity, the specificity of detecting HG PCa increased from 18% (PSA alone) to 39%. The authors further demonstrated that if biopsies were performed in patients with positive urine PCA3 (score > 20) or T2-ERG (score > 8) or with serum PSA > 10 ng/mL, 42% of unnecessary biopsies could be avoided.

In a study including a validation cohort of 1525 men, the MPS test was confirmed to improve the detection of csPCa. The authors also intended to set a threshold to exclude GG cancer ≥ 2. An MPS threshold of ≤10 was recommended. At this value, sensitivity (96%) and NPV (97%) were obtained, avoiding 32% of unnecessary biopsies while missing 3.7% of GG cancer cases ≥ 2 (Table 3) [46].

#### *3.3. ExoDx Prostate ® (IntelliScore) (EPI)*

The ExoDx Prostate (IntelliScore) (EPI, Exosome Diagnostics, Waltham, MA, USA) assesses the exosomal RNA expression of three genes (ERG, PCA3 and SPDEF) involved in the initiation and progression of PCa. ExoDx prostate is a test performed from a urine sample that does not require prior DRE testing. Exosomal RNA is derived from exosomes, which are small membrane vesicles secreted by several cell types, including immune and cancer cells [47]. The high potential of exosomes as biomarkers is due to their structure—a lipid bilayer protects the contents from degradation by proteases.

The test scores range from 1 to 100 and a cut-off point of 15.6 indicates men at increased risk of HG PCa (≥GG2) at subsequent biopsy, making the test helpful in validating the need for biopsy in men at risk. The test is recommended for men aged ≥50 years who are in the PSA "grey zone" (2–10 ng/mL) to distinguish a benign (HG1; when the test value < 15.6) from high-grade PCa (HG2 ≥ 15.6) [48].


Sp. = 34% [55]

**Table 3.** Predictive capacity of prostate cancer prebiopsy biomarkers.




164



**Table 3.** *Cont.*

Decision (LCD) L37733, which covers the ExoDx Prostate test before an initial prostate biopsy. 2—SOC: prostate-specific antigen level, age, race, family history) 3—Predictors in S3m include prostate examination (DRE and prostate volume), clinical variables (first-degree family history of PCa, and a previous biopsy, age) \*—Similar to PCA3, the utility for prognosis remains controversial. \*\*—AUC for ERSPC risk calculator plus PCA3 plus TMPRSS2-ERG \*\*\*—The initial S3M version had also included intact PSA, but due to interference between kallikreins in the immunosorbent assay with the allergen chip, it was removed from the S3M. Recently, a novel biomarker, HOXB13 SNP, a rare germline mutation of the HOXB13 gene with a high impact on prostate cancer risk, was included. Abbreviations: AUC—area under the curve, csPCa—clinically significant prostate cancer, DLX1—distal-less homeobox 1, DRE—digital rectal exam, ERG—estrogen-regulated gene, fPSA—free PSA, GG—grade group, GS—Gleason score, HG—high grade, hK2—human kallikrein-related peptidase 2, HOXC6- homeobox C6, iPSA—intact PSA, KLK3—kallikrein-related peptidase 3, LG—low grade, *n*- number of patients participating in study, mRNA—messenger RNA, NPV—negative predictive value, PCa—prostate cancer, PHI—Prostate Health Index, PPV—positive predictive value, PSA—prostate-specific antigen, PSAd—PSA density, p2PSA—(−2) proPSA, Ref—references, SNPs—single-nucleotide polymorphisms, Se—sensitivity, Sp—specificity, STHLM3—Stockholm-3, SPDEF-SAM pointed domain-containing ETS transcription factor, T2-ERG—transmembrane protease serine 2-ERG, tPSA: total PSA, 4K—four-kallikrein panel.

A validation study [48] conducted in 2016 on a (training) cohort of 255 patients initially and a separate validation cohort of 519 patients tested the ability of the ExoDx test in combination with SOC (PSA level, age, race and family history) to identify PCa GS (Gleason score) ≥7 in men aged ≥50 awaiting their first biopsy (PSA 2–20 ng/mL and/or suspicious DRE). With a cut-off value of 15.6, ExoDx alone demonstrated a sensitivity of =91.89%, NPV = 91.3% and AUC of 0.71 for distinguishing GS ≥ 7from GS 6 and benign PCa. When the test was combined with SOC (AUC = 0.73), the ExoDx test outperformed SOC alone (AUC = 0.63) and the PCPTRC risk calculator (AUC = 0.62) in differentiating PCa GS ≥ 7od GS 6 and benign PCa.

In a study published 2 years later [61], McKiernan et al. evaluate the clinical utility of ExoDx in comparison with standard clinical parameters for distinguishing grade (GG) ≥ 2 PCa from GG1 PCa and benign disease (in men eligible for the first biopsy) and conducted a prospective study of 503 patients aged ≥50 years with PSA "grey zone" (2–10 ng/mL). The results obtained were similar to previous studies. The combined model of ExoDx and SOC achieved the highest value (AUC = 0.71), and ExoDx alone (AUC = 0.70) was better at predicting GG2 PCa at initial biopsy than SOC (AUC = 0.62). A value of 15.6 was confirmed as the recommended cut-off point to distinguish patients at high risk of GG2 PCa at their initial biopsy. At a cut-off point of 15.6, a high negative predictive value of NPV = 89% was achieved (Table 3), preventing 26% of unnecessary biopsies and 20% of all biopsies (with only 7% of ≥GS7 PCa missed).

These two prospective studies validating over 1000 patients [48,61] showed that ExoDx (AUC 0.71 and AUC 0.70, respectively) was better at predicting clinically significant PCa at first biopsy than existing risk calculators and PCPT-RC (AUC 0.63), ERSPC-RC (AUC 0.58) and PSA alone (AUC 0.58). Both studies show that this test is useful for risk stratification of ≥GG2 due to GG1 cancer and benign disease and improves identification of patients at higher risk of advanced prostate cancer and helps avoid unnecessary biopsies.

Other work [48,61] confirmed the utility of ExoDx for primary biopsy but lacked confirmation for use on repeat biopsy. McKiernan et al. conducted a study [62] in 229 patients qualified for repeat biopsy; an AUC of 0.66 and an NPV of 92% (irrespective of other clinical features) were achieved at a previously validated cut-off point of 15.6, which would avoid 26% of unnecessary biopsies, omitting only 2.1% of patients with HG PCa (Table 3). Furthermore, in this study, AUC curves and net health benefits analyses showed better performance of ExoDx than the ERSPC and PSA risk calculator in predicting HG-PCa in men with a prior negative prostate biopsy. A total of 71.6% of patients were Caucasian, 14.4% African American and the study was completed on the most ethnically diverse group. The vast majority of publications are from the USA, and no studies have been completed on Asian or African populations or, more widely, on African Americans. It is not known what the cut-off values should be and what the diagnostic and prognostic accuracy is for a multiethnic population.

The clinical utility of ExoDx Prostate was recently evaluated in 1094 patients scheduled for their first biopsy (with PSA 2–10 ng/mL). This first study [63] of PCa biomarkers with a blinded control arm showed that ExoDx helped avoid unnecessary biopsies when the test was negative and increased the detection of HG PCa by 30% compared with a control arm without ExoDx (SOC alone). Compared to SOC, the test missed 49% fewer HG PCa. The study showed that ExoDx improved patient stratification and influenced the decisions made by (68%) urologists about biopsy (with rising PSA being the main reason for not following ExoDx results).

#### *3.4. SelectMDx*

SelectMDx (MDxHealth, Inc., Irvine, CA, USA) is a urine-based test after DRE that measures three biomarkers: DLX1 (progression gene), HOXC6 (cell proliferation gene), KLK3 (reference gene) and clinical risk factors (age, DRE, PSA and prostate volume, which can be calculated from the TRUS measurements substituted into the formula: height × width × length × 0.523). HOXC6 and DLX1 mRNA levels are assessed to estimate the risk of PCa on biopsy and the presence of high-risk cancer.

Men with elevated PSA levels in the "grey zone" (4–10 ng/mL) and/or an abnormal DRE result are subjects for whom an initial biopsy is considered. The result determines whether the patient is at high or low risk of PCa. It supports clinical decision-making and stratifies patients into those who may benefit from biopsy and early cancer detection and others for whom it is better to avoid this invasive procedure and continue with routine screening or active surveillance.

In a study [64], 386 men with an elevated PSA (≥3 ng/mL), abnormal DRE or family history of PCa, awaiting initial or repeat biopsy were studied. The predictive model (which included DRE as an additional risk factor) achieved an AUC = 0.86 in predicting high-grade cancer (after biopsy). Moreover, it was shown that with a cut-off point of −2.8, a 98% NPV with a sensitivity of 96%, the risk of GS ≥ 7 PCa was very low. For GS = 7 PCa, a 53% reduction in unnecessary biopsies was achieved while missing only 2% of cases with csPCa.

A study [65] was performed on a multinational (Netherlands, France, Germany) group of 715 patients with PSA < 10 ng/mL, before the initial prostate biopsy. SelectMDx achieved very high predictive values (AUC = 0.82 with 89% sensitivity, 53% specificity and NPV = 95% (Table 3), outperforming the PCPTRC 2.0. risk calculator (AUC = 0.70). This supports the use of the SelectMDx (MDxHealth, Inc., Irvine, CA, USA) test for the detection of HG PCa prior to the initial prostate biopsy.

To evaluate the clinical utility of SelectMDx, 418 patients who had an initial biopsy were studied. A total of 165 of them were positive. The number of biopsies performed within 3 months of the test was reviewed. For patients with a positive result, 71 patients (43%) were biopsied—27 of these patients were identified as having cancer, including 10 with a grade > 2. During this time, 9 patients with negative SelectMDx test results (3.6%) were biopsied—4 were identified as having cancer—all with a grade ≤ 2. SelectMDx has been shown to have a significant impact on decisions about the frequency and timing of biopsies. When the test was positive, the time period was shorter (median: 2 months) and the number of biopsies was five times higher than when SelectMDx was negative (median: 5 months). The test assisted urologists in their decision-making and is, therefore, a useful tool in daily urological practice [71].

A typical dilemma for the urologist deciding on a repeat prostate biopsy was presented in a case report of two men [72]. Both patients had already had their first negative biopsy, with normal DRE results, serum PSA levels of 3–10 ng/mL, no family history of PCa (and a negative ERSPC RC4 risk score). In these considerations, the European Association of Urology (EAU) recommendations [4] suggest the inclusion of mpMRI, RC and/or liquid biopsy tests. The mpMRI is the most accurate tool for localisation of PCa, but this imaging modality performed in the second patient did not show the presence of a tumour. The SelectMDx test showed the presence of PCa and therefore played a key role in individualising the need for repeat biopsy. In the mentioned report, NPV = 98%, and the risk score correlates with the mpMRI results, but it describes only two cases; therefore it suggests and indicates the need for further studies in risk stratification for repeat biopsy using the SelectMDx test.

It is unknown what the cut-off values should be and what the diagnostic and prognostic accuracy is for a multi-ethnic population. There is a lack of studies on Asian or African American populations.

#### **4. Serum Biomarkers**

Serum biomarkers, determined from blood samples, are produced by healthy and abnormal cells. PSA is undoubtedly the most widely studied cancer biomarker, but its clinical utility due to low specificity and specificity raises the need to find a test with better diagnostic values. Three tests that may have a positive impact on clinical practice are described below.

#### *4.1. Prostate Health Index (PHI)*

The Prostate Health Index (PHI; Beckman Coulter Inc., Brea, CA, USA), determined from a serum test, includes total PSA (tPSA), free, non-protein-bound PSA (fPSA) and (–2)proPSA (the fPSA isoform resulting from incomplete processing of the PSA precursor).

Determination of PHI values is indicated in men with PSA levels in the "grey zone" (4–10 ng /mL) and an unsuspected digital rectal examination (DRE) result [53], at age ≥50 years. The PHI score is calculated from the formula: ((–2)proPSA/fPSA) × √ PSA.

The above formula indicates that men with lower fPSA, higher total PSA and (– 2)proPSA are at an increased risk of development of clinically significant PCa [73].

This was confirmed by a study [74] in which the authors demonstrated that a low fPSA with a high total PSA indicates a risk of more aggressive PCa.

High PHI values indicate an increased likelihood of detecting prostate cancer, so when a biopsy (initial or repeat) is recommended, consideration should be given to using this less invasive method.

The PHI score has a greater diagnostic value than considering each of the indices (tPSA, fPSA) separately [49,51], improves the detection of PCa [75], improves clinical decision-making and predicts PCa aggressiveness [49,76,77]. Although the PHI score mainly provides information on overall PCa risk, studies [49,53] show an association between PHI value and prediction of PCa GS ≥ 7. A study [49] reported AUC = 0.72 to distinguish PCa GS ≥ 4 + 3 from GS ≤ 3 + 4 or no PCa.

Teams of researchers Lepor et al. [78] and Loeb et al. [53] showed that PHI is more specific in detecting csPCa than tPSA and/or fPSA. Furthermore, they concluded that this test might be useful in active surveillance and prediction of adverse outcomes after prostatectomy. Guazzoni et al. [52] showed that this was due to (–2)proPSA, as at GS ≥7, both PHI and (–2)proPSA were significantly elevated.

De la Calle et al., based on a multicentre study [50], showed that PHI is a predictor of PCa GS ≥ 7 (AUC = 0.78–0.82). When the PHI cut-off value of 24 is taken, 36% of unnecessary biopsies are avoided, while only 2.5% of high-grade cancers are missed. With a PHI cut-off point of 25, 40% of biopsies would be missed, and detection of lower grade PCa cases (GS = 6) would be reduced by 25%. However, this is associated with an underdetection of approximately 5% of clinically significant cancer cases.

A study [79] also found a significant effect of PHI on biopsy decisions. The study included 506 men diagnosed using PHI score and 683 without PHI determination, who were the control group. In both groups, men had PSA in the range of 4–10 ng/mL and unsuspecting DRE results. PHI score influenced medical management in 73% of patients; when the score was low, biopsy was postponed, and when it was high or moderate (PHI ≥ 36), biopsy was performed. Men who had a PHI test had fewer biopsies than the control group: 36.4% vs. 60.3%, respectively.

In response to these publications, Ehdaie and Carlsson [80] expressed concern about excluding men from a biopsy on the basis of PHI values and the risk of overlooking aggressive cancer, pointing out that the rate of an omitted PCa was 30%.

The authors of the paper [79], in response [81] to [80], maintain that the biopsy was safely postponed. They cite NCCN and AUA recommendations that men without biopsy who are in the diagnostic grey area will be monitored more closely or with additional methods. In a second response [82], it was shown that due to the small number of highgrade cancers, the study would not allow drawing firm conclusions.

In a study [33] (Table 2) involving a European (*n* = 503) and an Asian (*n* = 1652) population, the use of PHI established the recommended cut-off points for the above ethnic groups. More biopsies were avoided in the Asian group (56% vs. 40%). This study also identified the need to establish differential cut-off points for different ethnic groups. The authors of the publication recommended cut-off points for csPCa: PHI > 40 for European men and PHI > 30 for men of Asian origin. This result is not surprising, as Asians have a four times lower risk of prostate cancer than Europeans.

To assess the ability of PHI to detect Gleason grade 2–5 (GGG) PCa in Black men, 158 patients with elevated PSA levels and 135 controls were recruited [32] (Table 2). With PSA ≥ 4.0 and PHI ≥ 35.0, 33.0% of unnecessary biopsies were avoided, but 17.3% of GGG 2–5 PCa were missed. With PSA ≥ 4.0 and PHI ≥ 28.0, 17.9% were avoided, and the sensitivity of detecting GGG 2–5 PCa was 90.4%. These results indicate that PHI ≥ 28.0 can be safely used to avoid unnecessary biopsies in Black men, although it is associated with a risk of missed detection of GGG 2–5 PCa.

Currently, some researchers are considering the use of the PHI density index (PHID calculated as PHI divided by prostate volume) in diagnosis to identify csPCa. Tosoian et al. [83] showed that the prevalence of csPCa is associated with higher PHID and has a higher diagnostic value compared to PHI (AUC= 0.84 vs. 0.76). Their study indicates that PHID can prevent 38% of unnecessary biopsies while failing to detect only 2% of csPCa.

Another article [84] examined the diagnostic efficacy of PHI and PHID in terms of avoiding unnecessary biopsies. The results indicate that PHI (AUC = 0.722) and PHID (AUC = 0.739) have a higher diagnostic value than PSA, f-PSA% and PSAD (AUC = 0.595, 0.612 and 0.698, respectively). The combined sensitivity of PHI and PHID was 98%, avoiding 20% of biopsies in the non-diagnosis of only one patient with csPCa. Therefore, the use of the PHID density index may be a promising tool in the evaluation of PCa.

#### *4.2. 4Kscore*

The 4Kscore (Opko Health Inc., Miami, FL, USA) is a test developed to identify HG-PCa in patients with a suspicious DRE or elevated serum PSA. The test measures the levels of four kallikreins (4K): total PSA (tPSA), free PSA (fPSA), intact PSA (iPSA) and serum levels of human kallikrein 2 (hK2). It then compares the values obtained in the algorithm with the patient's age, DRE and results of previous prostate biopsies. Based on this information, the algorithm generates a percentage probability score to predict HG PCa even years in advance. This assessment allows further management to be determined depending on the outcome of the test and a decision to perform an initial or subsequent biopsy. This test is recommended primarily for men with a genetic family history. However, it can be performed by any man over 35 years of age who wants to assess his personalised risk of disease in the future.

The 4K test, although not designed to assess the predicted course of already diagnosed prostate cancer, has also been used in patients with csPCa to identify candidates for more intensive therapy. It has also been used to improve treatment selection and thus increase the chance of cure in patients suspected of having an underestimated malignancy. The 4Kscore provides an estimate of a patient's risk of developing distant metastases within 10 years.

Parekh and colleagues [56] on a validation cohort of 1012 indicated that the 4Kscore was better at predicting clinically significant PCa than the Prostate Cancer Prevention Trial Risk Calculator 2.0 (PCPT RC) (AUC 0.82 vs. 0.74) (Table 3). This study also indicates that, depending on the cut-off point, 30–58% of biopsies were reducible, while missing only 1.3–4.7% of HG PCa. A threshold of 1–7.5% is considered low risk, allowing safe delay of biopsy and continued follow-up with PSA. A cut-off of 9% reduces the number of biopsies to 43%, with 2.4% of csPCa cases missed [56,58]. At a cut-off of 15%, this test avoids prostate biopsies performed for indolent cancer by up to 58% and misses 4.7% [56]. A cut-off score of ≥20% indicates the need for biopsy due to the high risk of csPCa.

A comprehensive systematic review [57] including 12 studies (11,134 patients) showed, almost identically to the above study, an AUC = 0.81 for the 4Kscore in detecting csPCa (Table 3).

In a study [85], 43,692 asymptomatic men (unscreened, PCa-free, with low PSA values) were followed for 20 years, and the 4Kscore was evaluated for early detection of malignant prostate cancer. This work aimed to estimate the risk of prostate cancer metastasis or death by analysing the 4Kscore and PSA. It turned out that already at the time of blood collection, the 4Kscore indicated in whom an aggressive form of prostate cancer would appear. The

4Kscore significantly improved the detection of HG PCa in men with moderately elevated PSA. The authors concluded that men with an elevated PSA but a low 4Kscore could be safely observed by performing blood marker tests instead of direct biopsy. They indicated that men with a low 4Kscore have a very low long-term risk of death from prostate cancer or metastasis.

In a study [58] involving 611 patients, the 4Kscore test was ordered to assess the risk of aggressive prostate cancer in men with abnormal PSA and/or DRE results. Patients were divided into three risk groups: low, medium and high. The test results influenced biopsy decisions in 88.7% of men, where the biopsy avoidance rates were: 94.0%, 52.9% and 19.0% for the low, intermediate and high-risk groups, respectively. The risk category assessed by the 4Kscore was closely related to biopsy outcome, confirming the usefulness of the test in clinical practice.

A case–control study [35] (Table 2) evaluated the 4Kscore in 1667 prostate cancer cases and 691 control men with PSA ≥2 ng/mL. The men were from a variety of ethnic groups, including African American, Hispanic, Japanese, Native Hawaiian and Caucasian men. Results showed that across all ethnic groups, the 4Kscore was better at detecting both general and aggressive prostate cancer than tPSA or tPSA + fPSA. Therefore, the 4Kscore has broad clinical applicability and can be used for prostate cancer screening in a multiethnic population.

A study [56] conducted in the USA, evaluating the efficacy of the 4Kscore, examined 1012 men scheduled for prostate biopsy. The diagnostic performance in detecting HG-PCa was evaluated, showing an AUC of 0.82. African American (AA) patients comprised only 8.1% of the study group, which meant that the results were not representative of AA. For this reason, a validation study [34] (Table 2) was conducted on a population with a higher proportion of AA patients. The study included 366 men, 205 of whom were African American. The results of the study showed no significant difference in predicting tumour aggressiveness in this population, showing AUC = 0.81; therefore, the 4Kscore can be used to make biopsy decisions in both African Americans and non-African Americans.

A study [60] aimed at reducing unnecessary biopsies and overdiagnosis of benign PCa used the 4Kscore and the RPCRP risk calculator to predict csPCa at biopsy. A study of 2873 men showed that RPCRP and 4Kscore had very similar performance (AUC = 0.868 vs. AUC = 0.876), and their combination gave even better results (AUC = 0.888). This indicates that adding further predictors is a compromise between clinical utility, cost and patient burden.

#### *4.3. Stockholm3 Model*

Stockholm 3 Model (S3M) combines serum biomarkers (total PSA, free PSA, free/total PSA ratio, hK2, MIC1 and MSMB with genetic markers (254 single nucleotide polymorphisms [SNPs] and an unclassified variable for SNP HOXB13). The test also takes into account clinical data (age, previous prostate biopsy—family history, use of 5-alpha reductase inhibitors) and prostate examination (DRE, prostate volume). The S3M available in Sweden, Norway, Denmark and Finland is in clinical use for predicting the risk of aggressive prostate cancer and assessing the need for biopsy. The S3M research team at Karolinska Institute is currently working with two major laboratories in Europe, as well as laboratories in the US and Canada, to introduce the test in additional countries around the world. Additional validation studies have been conducted in Germany, the Netherlands and the UK. Studies on non-Caucasian populations (e.g., Hispanics, African Americans, Asians) are also planned. If the S3M is negative, the man has a low or normal risk of prostate cancer and is recommended to be followed up in 6 years. If the test is positive, it is recommended that the man is referred to a urologist. The urologist measures the volume of the prostate gland and carries out a DRE. If the prostate volume and/or the DRE test is abnormal, a biopsy is recommended. Otherwise, a Stockholm3 test in 2 years is recommended.

A study involving 59,159 men [67] compared S3M with PSA ≥ 3 ng/mL, a screening test for prostate cancer. The study was designed so that both tests detected the same number of Gleason score (GS) ≥ 7 tumours, and the tests were graded on the number of biopsies needed to achieve this. The results showed that S3M, in detecting tumours with a Gleason score of at least 7, has significantly higher specificity, sensitivity and AUC (0.74 versus 0.56 for PSA) for csPCa. Patients with a S3M score ≥ 11% were recommended to be referred to a urologist for further diagnosis. With a retained GS sensitivity ≥ 7, S3M avoided 32% of prostate biopsies (Table 3). In benign tumours, the level of biopsies avoided was 44%. In addition, the authors indicated that S3M could detect aggressive cancer even in men with PSA levels of 1.5–3 ng/mL, and the number of tumours with a Gleason score ≤ 6 was reduced by 17%, reducing overdiagnosis.

A study [68] described how, after fitting S3M to more data, the updated S3M slightly improved the AUC in predicting prostate cancer GS ≥ 7 compared with previously published results [67] (0.75 vs. 0.74). Each additional predictor (including DRE, previous biopsies and prostate volume) increased the AUC by up to one unit. The combination of predictors helps to increase the accuracy of diagnosis while reducing the number of unnecessary biopsies.

Studies [67–70,86] prove that S3M reduces overdiagnosis and the number of prostate biopsies while maintaining sensitivity for clinically significant prostate cancer.

In a short report [86], the authors evaluated how the S3M threshold affects the number of cancers detected and the number of biopsies performed. They collected data from a validation cohort of 47,688 men (with PSA ≥ 1 ng/mL) and then calculated the percentage of biopsies avoided and the percentage of cancer detections for different cut-off points of the S3M test. They noted that as the cut-off point increased, the number of cancers detected and biopsies performed decreased. They considered it reasonable to use S3M test values between 7% and 14% for the cut-off point for biopsy decisions, where cut-off values below 10% would increase sensitivity for Gleason score tumours ≥ 7 compared with PSA ≥ 3 ng/mL. They noted that the threshold could be selected to fit different health systems and even individual men.

Long-term follow-up of the replacement of PSA (as part of the standard prostate cancer diagnostic procedure) with Stockholm3 in prostate cancer detection in primary care in the Stavanger region of Norway showed that the implementation was beneficial. Compared with PSA, S3M reduced the proportion of clinically insignificant PCa (from 58% to 35%) and the number of biopsies performed (from 29.0% to 20.8%). In addition, it increased the proportion of biopsies positive for csPCa from 42% to 65%. This management may also lead to a reduction in healthcare costs. It has been estimated that direct healthcare costs decreased by 23–28% per male studied [87].

S3M is not suitable for men who have previously been diagnosed with or treated for prostate cancer or who are under follow-up after prostate cancer. It has no proven value for men diagnosed with prostate cancer or who have undergone a biopsy or other examination by a urologist within the last 6 months. It does not replace biopsy in men under active monitoring. This test was not evaluated on men younger than 50 years or older than 70 years and was restricted to an ethnically homogeneous population (Stockholm County, Sweden). The S3M was shown to be superior to prostate-specific antigen (PSA) as a screening tool for prostate cancer in all men aged 50–70 years. Furthermore, the S3M test can be performed in cases where the PSA value is > 1.5. The S3M has been shown to be superior in detecting, now often overlooked, aggressive cancer in men with PSA levels of 1.5–3 ng/mL. The S3M may reduce unnecessary biopsies without compromising the ability to diagnose prostate cancer with a Gleason score of at least 7.

#### **5. Tissue Biomarkers**

Tissue biomarkers analyse changes in nucleic acid expression and composition of tissue collected during needle core biopsy of the prostate. The main concept is to detect

changes in the histologically normal field neighbouring prostate cancer. This helps to verify whether the patient requires an additional biopsy.

#### *ConfirmMDx (MDxHealth)*

ConfirmMDx (MDxHealth, Inc., Irvine, CA, USA) is a tissue-based epigenetic assay that uses methylation-specific PCR (MSP) to analyse three prostate-cancer-related changes in DNA methylation patterns of suppressor genes (GSTP1, APC and RASSF1) in biopsy tissue (formalin-fixed and paraffin-embedded). All these biomarkers were isolated from biopsy-positive tissue.

ConfirmMDx is a molecular test clinically validated for predicting prostate cancer risk in men who have had a traditional thick-needle prostate biopsy that did not reveal the presence of cancer cells in collected histopathological material. In many cases, prostate biopsy results are falsely negative. The biopsy specimen may not be cancerous, and the histopathological result will not reveal the presence of cancer. However, due to the "halo effect", tissue with a normal morphological appearance will show epigenetic changes, indicating the presence of cancer. Using the test, histopathological material that has already been taken—during a prostate biopsy—can be re-examined in a detailed epigenetic analysis quantifying the level of methylation of promoter regions of three genes in benign prostate tissue, assessing with high accuracy the presence of cancer cells in neighbouring areas.

ConfirmMDx offers the opportunity to avoid unnecessary repeat biopsies. It allows a decision to be made on whether to include (rule-in) or exclude (rule-out) therapy. High-risk men with a previously negative biopsy may have undetected cancer after the test. Such patients with a previous "false negative" biopsy result should be included for repeat biopsy and appropriate treatment.

It also allows low-risk men to be excluded from repeat biopsies, which protects the patient from unnecessary stress and possible complications and reduces healthcare costs. This test increases the negative predictive value.

In the MATLOC study [28] involving 498 men with histopathologically negative prostate biopsies who had repeat biopsies within 30 months, positive results (cases) and negative results (controls) were reported. The clinical impact of a panel of epigenetic markers was assessed, showing for all cancers: NPV, sensitivity and specificity of 90%, 68% and 64%, respectively (Table 1). The results showed that in a multivariate model including patient age, PSA, DRE and histopathological features of the first biopsy, the epigenetic test was a significant independent predictor. At the same time, it was shown that the addition of this test could improve the diagnostic process for prostate cancer and reduce the number of unnecessary biopsies.

This was confirmed in the multicentre DOCUMENT study [29], which validated the clinical ability of ConfirmMDx to predict negative histopathological results in repeat prostate biopsies. For this purpose, archived core tissue samples from prostate biopsies with negative prostate cancer from 350 patients were evaluated. All patients had repeat biopsies after 24 months with negative (control) or positive (cases) histopathological results. The epigenetic test was shown to be a significant independent predictor of PCa detection after repeat biopsy and showed an NPV of 88%, with a sensitivity = 62% and specificity= 64% (Table 1).

Van Neste et al. [30] conducted a study on a cohort of 803 men, stratified according to their general methylation status (positive or negative) as defined in MATLOC [28] and DOCUMENT [29]. This study demonstrated an NPV of 96% for csPCa, and that methylation intensity was strongly correlated with the cancer stage. In assessing the prediction of GS ≥ 7 PCa after repeat biopsy, ConfirmMDx reached an AUC = 0.76 (Table 1). The decision curve analysis indicated the high clinical utility of the risk score as a decision tool in repeat biopsy. This indicates that ConfirmMDx is a much better predictor compared to currently used indicators such as PSA and risk calculator (PCPT-RC).

A population of 211 African American men undergoing repeat biopsy was studied to compare the accuracy of predicting repeat biopsy outcomes with previous studies

conducted in predominantly Caucasian populations [31] (Table 2). The specificity of this epigenetic assay was 60.0% and the sensitivity was 74.1% for detecting PCa at repeat biopsy. For detection of all PCa and GS ≥ 7 PCa, the NPV was 78.8% and 94.2%, respectively (Table 1). This study showed no significant differences in sensitivity and specificity between this test and previously described validation studies involving predominantly Caucasian populations and indicates usefulness for African Americans in risk stratification after an initially negative biopsy.

Wojno et al. [88] in 2014 already noted a reduced number of biopsies in clinical practice in centres using ConfirmMDx. They studied 138 patients with a median PSA level of 4.7 ng/mL and previous negative biopsies. They indicated a 4.4% repeat biopsy rate in ConfirmMDx-negative men, compared with a 43% prior repeat biopsy rate, indicating a potential 10-fold reduction.

A later study [89] in 2019 confirmed the impact of ConfirmMDx on biopsy decisionmaking. A total of 605 men with a median PSA level of 6.8 ng/mL and previous negative biopsies were studied. There was a six times higher repeat biopsy rate in ConfirmMDx positive men than in men with a negative test result.

ConfirmMDx enables a higher degree of accuracy (previously unattainable by prostate biopsy procedures alone) and has clinical, financial [90] and health benefits by reducing the number of medically unnecessary and expensive repeat biopsies that are part of the current standard of care.

#### **6. The Financial Aspect**

A prostate MRI costs between 500 and 2500 USD in the United States, depending on whether the patient is insured. Approximately 1 million American men are currently referred for prostate biopsy each year. If all of these men underwent an MRI instead, costs could reach 3 billion USD per year.

In a paper [91] addressing the costs associated with prostate biopsy and its potential complications, the authors analysed charges for the procedure and related claims for all Medicare Fee-for-Service patients over a 2-year period (January 2014–December 2015). The study included 234,819 prostate biopsy cases and associated costs.

Uncomplicated biopsies cost about 1750 USD, those with one complication were already more expensive at 4060 USD, and for patients requiring hospital admission, the cost was as high as 13,840 USD (average cost was 2020 USD). The most common complication of biopsy is bleeding and infection, which can be prevented using biomarker tests from urine or blood. The cost of tests based on these is higher than the commonly used PSA but lower than biopsy, which makes it a cost-effective option.

In a paper [92], Santhianathen et al. conducted a cost-effectiveness analysis of biomarkers for 2018. Costs were obtained directly from pharmaceutical companies (these were as reported by Prostate Cancer Markers): PHI 499 USD, 4Kscore 1185 USD, SelectMDx 500 USD and ExoDx 760 USD the cost of ExoDx was estimated using data from the CMS (Centers for Medicare and Medicaid Services) Clinical Laboratory Fee Schedule). Discounted QALYs and costs were estimated; for example, a 50-year-old male with an elevated PSA level (3 ng/mL or greater). The cost of the current SOC strategy of ultrasound-guided transrectal biopsy was 3863 USD and the discounted QALY (an indicator of an individual's or group's health status expressing quality-adjusted life expectancy) was 18.0853. Each of the biomarkers tested improved the QALY compared with SOC. The ExoDx index provided the highest QALY with an incremental cost-effectiveness ratio of 58,404 USD per QALY. The study showed that before biopsy in men with elevated PSA levels, the use of SelectMDx (MDxHealth, Inc., Irvine, CA, USA) or EPI (Exosome Diagnostics, Waltham, MA, USA) assesses were cost-effective, PHI (Beckman Coulter Inc., Brea, CA, USA), was found to be more expensive and less efficient.

In an economic evaluation, Nicholson et al. [93], comparing diagnostic value for money, found that the PHI test and PCA3 were no more cost-effective than clinical evaluation, which also generates more QALYs.

Reports [94] on SelectMDx support data from four European countries [95], which showed that SelectMDx in the initial diagnosis of prostate cancer saves healthcare costs and increases QALYs compared with the current standard of care based on prostate biopsy for elevated prostate-specific antigen [95].

This was confirmed in a study by Govers et al. [96] on a population of US men with elevated PSA. Results were related to QALYs and cost of care from a payer (Medicare) perspective. Routine use of SelectMDx to guide biopsy decisions was shown to be beneficial—gaining an average of 0.045 QALYs and saving 1694 USD per patient.

Based on studies [92,95,96], it can be concluded that a SelectMDx-based strategy improves health outcomes and reduces costs.

The publication [90] focused on determining the impact of ConfirmMDx on the healthcare budget. It examined whether costs are recovered by avoiding unnecessary biopsies.

The implementation of ConfirmMDx created a hypothetical commercial health plan in which direct costs were calculated over a 1-year horizon using 2013 Medicare fee-forservice rates. The study concluded that the net cost of a commercial health plan with 1 million members would be reduced by approximately 500,000 USD if patients with histopathologically negative biopsies were screened using an epigenetic test to distinguish between patients who should undergo repeat biopsies and those who should not. The use of this genetic test may reduce healthcare costs and improve clinical management.

STHLM3 is not a commercially available test, for which reason its price is unknown. However, it is expected to be similar to other biomarkers currently available (>224 USD). These tests are more expensive than the common PSA, but are more reliable and can be performed less frequently due to their better diagnostic value. It also avoids biopsies, reduces overdiagnosis and allows a treatment plan to be customised to the patient and thus also reduces costs.

#### **7. Guidelines**

The National Comprehensive Cancer Network (NCCN) guidelines—version 2.2020 recommend considering the use of biomarkers for the early detection of prostate cancer, indicating that the specificity of screening can be improved in assessing the indication for biopsy (Grade C recommendation). They indicate the possibility of using the Prostate Health Index (PHI), SelectMDx, 4Kscore and ExoDx to assess the likelihood of high-grade cancer (Gleason score ≥ 3 + 4, GG ≥ 2).

The NCCN guidelines also address post-biopsy management. They indicate the possibility of using tests to improve specificity in high-risk patients despite a negative prostate biopsy result: 4Kscore, PHI, percentage free PSA, PCA3 and ConfirmMDx (included/added from 2020). The recommendations for the management of benign biopsy results themselves have changed "PSA and DRE 6–24 months apart and consideration of per cent free PSA, 4Kscore, PHI, PCA3 or ConfirmMDx and/or mpMRI and/or improved prostate biopsy techniques. Repeat prostate biopsy, depending on risk". However, the guidelines note that the extent to which tests are validated in different populations varies and that it is unclear what the optimal combination of tests with MRI would be. In the current NCCN guidelines, MPS is listed as a biomarker requiring additional testing.

The EAU gives a strong recommendation for the use of risk-calculators and imaging in asymptomatic men with PSA levels of 2–10 ng/mL, while giving a weak recommendation (strength rating—weak) for urine and blood biomarkers to avoid biopsy [4].

The FDA has approved PCA3 and PHI. ExoDx received FDA Breakthrough Design recognition in June 2019. SelectMDx has not been reviewed by the FDA due to the agency determining that such approval is not necessary but includes CAP (College of American Pathologists) and CLIA (Clinical Laboratory Improvement Amendments) accreditations.

ConfirmMDx and 4Kscore do not have FDA recommendations but are accredited by CAP and CLIA.

#### **8. Biomarkers and mpMRI**

Multiparametric magnetic resonance imaging (mpMRI) is a promising new tool for the diagnosis, prognosis and monitoring of PCa.

The European Society of Urology guidelines on PCa recommends the use of mpMRI before prostate biopsy in previously untreated patients with suspected PCa [4].

In addition to the high sensitivity of mpMRI in detecting hg-PCa, mpMRI also has disadvantages, i.e., low specificity, high cost, the need for expensive, specialised equipment, low sensitivity in predicting the presence of extra-urethral expansion and the requirement for an expert review. Current research focuses on comparing biomarker tests with mpMRI and also on the extent to which they can complement each other.

A study [66] on 172 men showed promising correlation results between SelectMDx and mpMRI. There was a statistically significant difference in the SelectMDx score between PI-RADS 3 and 4 (*p* < 0.01) and between PI-RADS 4 and 5. The SelectMDx score was better than the PCA3 score in predicting outcome for suspected PCa on mpMRI (AUC = 0.83 for SelectMDx versus 0.65 for PCA3), suggesting the possibility of using the SelectMDx test to stratify patients for mpMRI.

The combination of 4KScore (AUC = 0.70) with mpMRI (AUC = 0.74) resulted in a prognostic improvement (AUC = 0.82 for 4KScore and mpMRI combined) in detecting aggressive PCa [97,98].

In a recent article [94], the authors demonstrated that the 4KScore, used in addition to mpMRI, can reduce unnecessary SBx (without worsening the diagnosis of csPCa) and identify patients who would benefit from undergoing TBx alone. An evaluation of 408 men showed a reduction (39.5%) in unnecessary biopsies and a reduction in detection (33.9%) of GG1 disease, with 5.2% (diagnosed with SBx) and 1.1% (diagnosed with SBx combined with TBx) missing.

In another study [99], 266 men who were not biopsied underwent three strategies using 4Kscore, mpMRI and combination PSA density (PSAD) to determine the safest method to skip biopsy. The first strategy starts by assessing the 4Kscore value. If it was >7.5, indicating an intermediate or high risk of csPCa, mpMRI was performed. If it was negative and the 4Kscore value was above 7.5 but below 18 (intermediate risk), the patient remained under clinical observation, but in case of a positive mpMRI result, a biopsy was performed. The second strategy started with mpMRI and was similar thereafter. In the third strategy, PSAD was calculated in case of a positive mpMRI result. The results confirmed that 4Kscore combined with mpMRI gave a better AUC = 0.82 than each method alone: 4Kscore (AUC = 0.70), mpMRI (AUC = 0.74). The best strategy seems to be an initial biopsy if the 4Kscore was >7.5%, followed by mpMRI and another biopsy for those with positive mpMRI (PIRADS ≥ 3) or 4Kscore >18%. This would avoid 34.2% of prostate biopsies while missing 2.7% of clinically significant PCa. However, this model is more expensive and requires external validation in a multicentre study, but it gives us an idea of how we can improve the selection of men for biopsy using biomarkers and mpMRI.

PHI, total PSA, PSAD and the ability of mpMRI to identify csPCa were compared in a group of 395 men [100]. In detecting csPCa for PSA, PSAD, Pi-RADS and PHI, the AUCs were as follows, respectively: 59.5, 64.9, 62.5 and 68.9 in patients undergoing biopsy, and for patients with a previous negative biopsy: 55.4, 69.3, 64.4 and 71.2. This indicates that PHI had comparable results to mpMRI and outperformed other indices.

Adding PHI to mpMRI leads to increased predictive accuracy of csPCa and a reduction of up to 50% in unnecessary biopsies (for men with PI-RADS 3–5 and PHI ≥ 30). Moreover, combination AUC outperforms PHI and mpMRI alone (AUC were 0.87, 0.73 and 0.83, respectively [101].

A study [102] performed prostate cancer diagnosis using a combination of Stockholm3 and mpMRI. Targeted biopsies or mpMRI were performed only in men at higher risk as assessed by S3M. When maintaining the number of detected FG cancers ≥2, there was a 42% saving of biopsies and a 46% reduction in FG1 detection. Using a combination of S3M and MRI TBx, the detection of GG 1 tumours and the number of biopsies needed were almost halved, with no reduction in sensitivity in detecting GG 2 cancers, compared with using SBx.

#### **9. Discussion**

PSA is a highly sensitive screening test. However, it lacks specificity, resulting in a high rate of unnecessary prostate biopsies. The liquid biopsy tests are more expensive than the commonly used PSA, but because of their better diagnostic value, they can be performed less frequently and avoid other more costly procedures such as biopsy or mpMRI.

It seems important to differentiate the tests in terms of their advantages and disadvantages and to demonstrate which biomarker may be most useful in a given clinical situation.

PHI, 4Kscore, SelectMDx and ExoDx offer better specificity than PSA and can help identify men with GS ≥ 7 PCa. MPS also outperforms PSA and each of its components in HG PCa detection, and its performance in men with suspicious PSA levels helps to validate the need for initial or repeat biopsy.

STHLM3 is also significantly superior to PSA and can detect HG PCa even in men with PSA levels ≥ 1.5 ng/mL. However, this test has not yet been validated on multiethnic groups, nor have tests comparing it with other liquid biopsy tests been developed.

In men at increased risk of PCa with a previous negative biopsy, additional information can be obtained with the Progensa-PCA3 urine test, MPS, ExoDx and the 4Kscore, PHI and STHLM3 serum tests or the tissue-based epigenetic test (ConfirmMDx).

PCA3 reduces prostate biopsy rates in men undergoing repeat biopsy, but there is still no consensus on the cut-off value.

As PCA3 increases with cancer aggressiveness, tests based on it—Progensa PCA3, MiPS and ExoDx—show the ability to distinguish between cancers with high and low Gleason scores, indicating high utility in therapeutic decision-making.

As ExoDx uses an algorithm independent of PSA and its derivatives, clinical factors (features) and standard of care (SOC), it is feasible (in the US) to perform at home. The patient takes a sample, hands it over to a courier and then discusses the result with the doctor via telehealth. This novelty (ExoDx Prostate At-Home Collection) seems particularly useful in times of coronavirus pandemics and for people living far from medical care.

PHI is significantly better than SelectMDx in diagnosing any PCa, while SelectMDx is significantly better than PHI in diagnosing csPCa.

The 4Kscore assesses the risk of detecting HG PCa if a biopsy is performed. It has been shown to have a better detection rate for HG PCa than the modified PCPTRC and SOC. In addition, 4Kscore can predict HG PCa even years in advance and assess the risk of distant metastasis, e.g., in genetically burdened men. It thus helps to non-invasively avoid prostate biopsy for men in whom it is not necessary and identifies men at higher risk for whom an early intervention is beneficial.

Hendriks [103] and colleagues undertook a comparison of the diagnostic values of two FDA-approved tests, PHI and PCA3, for primary and repeated biopsy. Unfortunately, after compiling all studies published before 2017, they were unable to draw clear conclusions due to the conflicting results of the articles analysed. Study [104] notes that although in a double-blind study of PCA3 vs. PHI, PCA3 is superior to PHI in cancer prediction accuracy, when considering only significant PCa, PHI remains the most accurate predictor. For this reason, the authors recommend using PHI instead of PCA3 in population-based screening.

In a study [54] on 531 men (PSA 3–15 ng/mL) who underwent an initial biopsy, 4Kscore and PHI had similar AUCs in predicting PCa (AUC = 0.69 and 0.74, respectively) and csPCa (0.72 vs. 0.71, respectively).

Russo et al. indicated in their systematic review [55] the high diagnostic accuracy of PHI and 4Kscore. Both tests were tested on multiethnic groups and showed high diagnostic value in them. Although both biomarkers provide similar diagnostic accuracy in the detection of general and high-grade PCa and reduce the number of unnecessary biopsies, it should be borne in mind that there are disturbing reports on PHI [80–82].

Furthermore, PHI should not be interpreted as absolute proof of the presence or absence of prostate cancer. Elevated PSA and PHI can be observed not only in patients with prostate cancer but also with benign diseases. PHI results should be interpreted taking into account clinical factors or family history, and individual clinical decisions should be made based on them.

Vedder et al. [105] added PCA3 and 4Kscore to the ERSCPC risk calculator and compared performance. They showed that 4Kscore was better than PCA3 in predicting PCA in men (with PSA ≥ 3.0 ng/mL) (AUC 0.78 and 0.62, respectively). However, when no PSA limit was set, PCA3 performed better than 4Kscore (AUC 0.63 vs. 0.56). When added to ERSPC, both biomarkers slightly improved the prediction of PCa, with no significant differences (in performance) between them.

Additionally, the previously mentioned study [60] confirmed that adding ERSPC to the 4Kscore improves diagnostic value. However, it is worth recalling that the 4Kscore is the most expensive of the tests compiled in our review.

In addition, it is important to remember that drugs such as 5-alpha-reductase inhibitors: finasteride, dutasteride and anti-androgen therapy can affect the levels of PSA and other biomarkers. Such medications should be discontinued for at least 6 months prior to the study. Samples for the test should be taken when the clinician is satisfied that the prostate tissue has recovered, normally no less than 6 months after the date of the last biopsy or any other prostate procedure. The impact of these procedures on the performance of the test has not yet been assessed.

#### **10. Conclusions**

Recently, molecular characterisation of PCa has become increasingly important, and a wide range of biomarker-based liquid biopsy tests are commercially available to assist urologists in clinical decision-making. The prostate cancer liquid biopsy biomarkers listed above have a high NPV and therefore help prevent unnecessary biopsies. As mentioned earlier, numerous publications [16–18] have not shown a correlation between PCA3 values and prostate cancer aggressiveness (Gleason score). Given this fact and reports of unexplained PCA3 well above the cut-off [15] without cancer on biopsy, it is reasonable to use newer, more sensitive and specific diagnostic tools to detect patients requiring prompt and radical treatment. For example, PCA3 in combination with other biomarkers such as TM-PRSS2: ERG fusion [36] in Mi-Prostate Score [41–46] or ERG and SPDEF in ExoDx Prostate IntelliScore [47,48,61–63], where it shows better diagnostic and prognostic potential.

From a clinical point of view, it is critical to identify assays for the early detection of aggressive PCa subtype when it can still be treated effectively. Recent years have led to the development of totally non-invasive tests i.e., (ExoDx Prostate At-Home Collection) where first catch, nondigital rectal examination urine specimens appeared helpful in identifying aggressive (Gleason score 7–10) PCa in a racially diverse patient cohort. Similarly, the four-kallikrein panel showed effectiveness in identifying aggressive PCa in a multiethnic population.

It seems that in the near future, molecular biomarkers, clinical and histopathological features and diagnostic imaging will have to be used in a complementary rather than a competitive manner to ensure the best possible selection of patients for mpMRI and eventual biopsy.

**Author Contributions:** All authors made substantial contributions to this work; writing—original draft preparation: M.M.; supervision and review of the paper: M.S. and J.A.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Collegium Medicum University of Zielona Góra.

**Conflicts of Interest:** J.A.S. is a consultant for MDxHealth. M.M. and M.S declare no conflict of interest.

#### **Abbreviations**


#### **References**


Antigen for Prostate Cancer Detection in the 2.0 to 10.0 ng/ml Prostate Specific Antigen Range. *J. Urol.* **2011**, *185*, 1650–1655. [CrossRef]


## *Review* **The Roadmap of Colorectal Cancer Screening**

**Enea Ferlizza 1,\* , Rossella Solmi <sup>1</sup> , Michela Sgarzi <sup>1</sup> , Luigi Ricciardiello <sup>2</sup> and Mattia Lauriola <sup>1</sup>**


**Simple Summary:** Colorectal cancer (CRC) is the third most common form of cancer in terms of incidence and the second in terms of mortality worldwide. CRC develops over several years, thus highlighting the importance of early diagnosis. Fecal occult blood test screening reduces incidence and mortality. However, the participation rate remains low and the tests present a high number of false positive results. This review provides an overview of CRC screening globally and the most recent approaches aimed at improving accuracy and participation in CRC screening, while also considering the need for gender and age differentiation. New fecal tests and markers such as DNA methylation, mutation or integrity, proteins and microRNAs are explored, including recent investigations into fecal microbiota. Liquid biopsy approaches, involving novel markers, such as circulating mRNA, micro-RNA, DNA, proteins and extracellular vesicles are discussed. The approaches reported are based on quantitative PCR methods or arrays and sequencing assays that identify candidate biomarkers in blood samples.

**Abstract:** Colorectal cancer (CRC) is the third most common form of cancer in terms of incidence and the second in terms of mortality worldwide. CRC develops over several years, thus highlighting the importance of early diagnosis. National screening programs based on fecal occult blood tests and subsequent colonoscopy have reduced the incidence and mortality, however improvements are needed since the participation rate remains low and the tests present a high number of false positive results. This review provides an overview of the CRC screening globally and the state of the art in approaches aimed at improving accuracy and participation in CRC screening, also considering the need for gender and age differentiation. New fecal tests and biomarkers such as DNA methylation, mutation or integrity, proteins and microRNAs are explored, including recent investigations into fecal microbiota. Liquid biopsy approaches, involving novel biomarkers and panels, such as circulating mRNA, micro- and long-non-coding RNA, DNA, proteins and extracellular vesicles are discussed. The approaches reported are based on quantitative PCR methods that could be easily applied to routine screening, or arrays and sequencing assays that should be better exploited to describe and identify candidate biomarkers in blood samples.

**Keywords:** fecal immunochemical test (FIT); colonoscopy; flexible sigmoidoscopy; liquid biopsy; mRNA; microRNA; ctDNA; proteins; extracellular vesicles

#### **1. Introduction**

Colorectal cancer (CRC) develops over time from modifications of the normal intestinal mucosa to benign precancerous adenomas, carcinoma, and eventually aggressive metastatic cancer [1]. This transition is a complex, multifactorial process that has been characterized over the years. Adenomatous polyposis coli (APC) gene mutations or deletions leading to chromosomal instability represents one of the pathways that drives the development of CRC [2–4]. Activating mutations of the KRAS oncogene and inactivating mutations of the TP53 tumor suppressor gene further promote adenoma–carcinoma

**Citation:** Ferlizza, E.; Solmi, R.; Sgarzi, M.; Ricciardiello, L.; Lauriola, M. The Roadmap of Colorectal Cancer Screening. *Cancers* **2021**, *13*, 1101. https://doi.org/10.3390/ cancers13051101

Academic Editor: Fabrizio Bianchi

Received: 9 February 2021 Accepted: 27 February 2021 Published: 4 March 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

progression. Microsatellite instability (MSI), aberrant CpG island methylation phenotype (CIMP), chromosomal instability (CIN) and BRAF mutations are also associated with the transition and development of CRC. Colorectal cancer is also linked to different risk factors such as older age, male sex, adverse lifestyle habits (smoking, increased consumption of red meat and alcohol), chronic intestinal diseases, clinical history of polyps, genes, and heredity [5].

The slow growth of this cancer makes the identification of precancerous lesions and early detection of cancer fundamental for defeating the disease. Screening is, thus, essential to reduce the incidence and mortality of CRC. In fact, CRC mortality is gradually decreasing in industrialized countries due to the widespread adoption of screening programs [5]. Today, the implementation of screening opportunities is crucial, and the research in this field is prolific globally.

This review is divided in two parts. The first focuses on the CRC status, screening methodologies, national screening programs and their application worldwide. The second part firstly examines the main drawbacks of the fecal occult blood test (FOBT), which is the golden standard screening test worldwide, and then focuses on the strategies aimed at improving CRC screening using liquid biopsy approaches and suitable candidate biomarkers (mRNA, miRNA, ctDNA, microvesicles).

#### **2. Colorectal Cancer Status in Europe and in the World**

CRC is the third most common form of cancer in terms of incidence and the second in terms of mortality worldwide, with 1.9 million new cases and 930,000 deaths reported in 2020 [6]. There are important geographical discrepancies regarding the incidence and mortality of CRC (Figure 1, Tables S1 and S2). Australia and New Zealand show the highest incidence, followed by Europe and North America [7–10]. The highest reported mortality rates are in central Eastern Europe. The lowest CRC incidence is registered in South Asia and in Africa, where also the lowest mortality rates are recorded, although in these areas the highest mortality to incidence ratio is recorded. *Cancers* **2021**, *13*, x 3 of 20

**Figure 1.** World and colorectal cancer in 2020. (**a**) Estimated age standardized incidence rate (100,000) for world countries; (**b**) Estimated age standardized mortality rate (100,000) for the world countries. Modified from Global Cancer Observatory (GBO) 2020, International Agency for Research on Cancer, World Health Organization [6]. In Europe, CRC is the second most common oncological disease in terms of incidence and mortality, with 519,820 new cases and 244,824 deaths registered in 2020 [6]. The highest incidence was observed in central Eastern Europe and there are substantial differences **Figure 1.** World and colorectal cancer in 2020. (**a**) Estimated age standardized incidence rate (100,000) for world countries; (**b**) Estimated age standardized mortality rate (100,000) for the world countries. Modified from Global Cancer Observatory (GBO) 2020, International Agency for Research on Cancer, World Health Organization [6].

between European countries (Table S3; Figure 2), with respect to both risk factors, linked

In Italy, CRC is the third most common oncological disease in terms of incidence and the second in terms of mortality, with 49,327 new cases and 19,258 deaths reported in 2018

to different lifestyles, and screening policies [5,9,11].

In Europe, CRC is the second most common oncological disease in terms of incidence and mortality, with 519,820 new cases and 244,824 deaths registered in 2020 [6]. The highest incidence was observed in central Eastern Europe and there are substantial differences between European countries (Table S3; Figure 2), with respect to both risk factors, linked to different lifestyles, and screening policies [5,9,11]. 4 of 20

**Figure 2.** Colorectal cancer in Europe in 2020. (**a**) Estimated age standardized incidence rate (100,000) for European countries; (**b**) Estimated age standardized mortality rate (100,000) for the European countries. Modified from Global Cancer Observatory 2020, International Agency for Research on Cancer, World Health Organization [6]. **Figure 2.** Colorectal cancer in Europe in 2020. (**a**) Estimated age standardized incidence rate (100,000) for European countries; (**b**) Estimated age standardized mortality rate (100,000) for the European countries. Modified from Global Cancer Observatory 2020, International Agency for Research on Cancer, World Health Organization [6].

**3. Colorectal Cancer Screening**  *3.1. Advantages of Screening in Terms of Incidence and Mortality*  CRC screening programs have been shown to reduce incidence and mortality [7,13– In Italy, CRC is the third most common oncological disease in terms of incidence and the second in terms of mortality, with 49,327 new cases and 19,258 deaths reported in 2018 (Table S4). The 5-year survival is 66%, with no differences between men and women [12].

#### 16]. Soon after the activation of screening programs, the incidence showed an increase in **3. Colorectal Cancer Screening**

#### the short-term, which tended to decrease over the subsequent years [13,17] and the cancers that were detected were more often diagnosed at earlier stages [16]. Notably, popu-*3.1. Advantages of Screening in Terms of Incidence and Mortality*

lations with active screening programs have shown an impressive reduction in mortality from 22 to 68% [16–21]. *3.2. Screening Tests*  3.2.1. Stool-Based Tests CRC screening programs have been shown to reduce incidence and mortality [7,13–16]. Soon after the activation of screening programs, the incidence showed an increase in the short-term, which tended to decrease over the subsequent years [13,17] and the cancers that were detected were more often diagnosed at earlier stages [16]. Notably, populations with active screening programs have shown an impressive reduction in mortality from 22 to 68% [16–21].

There are currently three types of screening for detecting CRC: stool-based, imaging,

#### and endoscopic tests [5]. Stool-based tests (fecal occult based test, FOBT) shows the pres-*3.2. Screening Tests*

3.2.2. Imaging Tests

#### ence of hem (gFOBT) or human globin (FIT) of hemoglobin in stool samples. The gFOBT 3.2.1. Stool-Based Tests

has a long history and consists of a colorimetric assay which uses the guaiac reaction [22]. FIT is an immunochemical test, which exploits a specific antibody. It has replaced gFOBT because it is more sensitive and accurate at detecting CRC (sensitivity 69–95% vs. 25–38) and does not require dietary restrictions [23–25]. Patients with positive FIT tests are referred for a colonoscopy for further investigations. In order to respond to the best cost–benefit strategy, numerous studies have tried to fix the optimal cut-off of FIT. The most commonly used value currently appears to be There are currently three types of screening for detecting CRC: stool-based, imaging, and endoscopic tests [5]. Stool-based tests (fecal occult based test, FOBT) shows the presence of hem (gFOBT) or human globin (FIT) of hemoglobin in stool samples. The gFOBT has a long history and consists of a colorimetric assay which uses the guaiac reaction [22]. FIT is an immunochemical test, which exploits a specific antibody. It has replaced gFOBT because it is more sensitive and accurate at detecting CRC (sensitivity 69–95% vs. 25–38%) and does not require dietary restrictions [23–25].

100 ng/mL, corresponding to 20 μg of Hb per g of stool [24]. A high variability has been recorded in FIT screening between different centers and kits, with the analytical performance depending on antibody characteristics (mono or polyclonal), buffer volume or composition of collection vials [7,26]. Other drawbacks of FIT are related to the less than optimal level of enrollment in screening programs, the high number of false positive results (15–30%), the poor ability to detect serrated polyps, and the low sensitivity for adenomas [5,7,27]. Patients with positive FIT tests are referred for a colonoscopy for further investigations. In order to respond to the best cost–benefit strategy, numerous studies have tried to fix the optimal cut-off of FIT. The most commonly used value currently appears to be 100 ng/mL, corresponding to 20 µg of Hb per g of stool [24]. A high variability has been recorded in FIT screening between different centers and kits, with the analytical performance depending on antibody characteristics (mono or polyclonal), buffer volume or composition of collection vials [7,26]. Other drawbacks of FIT are related to the less than optimal level of enrollment

Imaging tests include the double-contrast barium enema (DCBE), computed tomo-

methods are often used rather than DCBE [5]. CTC was introduced about 20 years ago and provides endoluminal images of the air-distended colon, reconstructed by computed tomography or magnetic resonance [5]. CCE is recognized by the European Society of Gastrointestinal Endoscopy as an acceptable screening method for CRC (with a sensitivity in screening programs, the high number of false positive results (15–30%), the poor ability to detect serrated polyps, and the low sensitivity for adenomas [5,7,27].

#### 3.2.2. Imaging Tests

Imaging tests include the double-contrast barium enema (DCBE), computed tomographic colonography (CTC), and colon capsule endoscopy (CCE). Today, novel imaging methods are often used rather than DCBE [5]. CTC was introduced about 20 years ago and provides endoluminal images of the air-distended colon, reconstructed by computed tomography or magnetic resonance [5]. CCE is recognized by the European Society of Gastrointestinal Endoscopy as an acceptable screening method for CRC (with a sensitivity of 84% and specificity of 93%) [28]. However, these methods require intensive bowel preparation and are more expensive than colonoscopy and biopsies cannot be performed [5,28].

#### 3.2.3. Endoscopic Tests

Endoscopic tests consist of flexible sigmoidoscopy (FS) and colonoscopy (CS). FS visualizes only the distal gastrointestinal tract, but does not detect lesions in the proximal colon. The advantages of FS include the fact that no dietary restrictions are required and it involves minimal bowel preparation [5,19,20]. Colonoscopy represents the gold standard for diagnosis, with a high sensitivity and specificity for detecting cancerous and precancerous lesions (97–98%) in the entire large bowel and the distal part of the small bowel [18]. During the procedure, it is also possible to perform biopsies for histological evaluation. However, colonoscopy is an expensive and risky method, since complications such as bleeding or bowel perforation occur in approximately 0.1–0.2% of patients [5,7,19].

#### *3.3. Screening Status in Europe and the World*

In 2012, the European Union drew up guidelines on CRC screening and diagnosis, recommending the use of national screening programs based on FIT, FS or CS [29].

In Italy, the national screening program is FIT-based, which is recommended every two years and carried out for the population deemed at risk (50–69 years) [7,12,30–32]. According to the most recent data, screening coverage was between 90 and 96%, depending on the geographic area; however, the overall participation was still low, ranging from 60% in the north to 23% in the south (Tables S5 and S6).

Several European countries, including the extra-EU, have developed an ongoing or planned national or regional screening program, with an invitation system for the population considered at risk (Table 1). The majority use the fecal occult blood test (gFOBT or FIT), with wide differences in participation rates.


**Table 1.** Colorectal cancer (CRC) screening in Europe [15,33–39].


**Table 1.** *Cont.*

<sup>1</sup> Guaiac fecal occult blood test (gFOBT), fecal immunochemical test (FIT), colonoscopy (CS), flexible sigmoidoscopy (FS), not available (NA). <sup>2</sup> Cut-off for FIT in µg Hb/g feces <sup>3</sup> . Target age and interval screening according to the national programs. <sup>4</sup> Percentage of people of the target age invited to participate in the screening. <sup>5</sup> Percentage of invited people that participated in the screening. <sup>6</sup> Regional or national differences.

> National and regional organized screening programs using the fecal test (gFOBT or FIT) have also been reported for Canada, Brazil, Argentina, Chile and Uruguay, which obtained high participation rates in the pilot studies (90.1–79.7%) [10,36].

> On the other hand, in the USA, the U.S. Preventive Service Task Force recommends that asymptomatic adults aged from 50–75 have a screening test on a voluntary basis, and a national screening program is still not available. The choice for the average risk population in the USA is between stool-based tests (gFOBT, FIT, FIT-DNA), direct visualization tests (FS, CS, CTC) or the serological DNA test (SEPT9). As of 2018, 68.8% of people aged 50–75 with health insurance are reported to be up-to-date with colorectal cancer screening [40].

> In 2015, recommendations for CRC screening in the Asia Pacific region were updated [41] and the Asia Cohort Consortium focused also on health outcomes in Asian populations [42]. Few countries (Australia, China, Japan, New Zealand, South Korea, Thailand and Taiwan) have national or regional screening programs and these are mainly based on the fecal test (FIT or gFOBT). Additionally, in these regions the national participation rates were rather low (13–41.3%) [10,36].

> Among the countries in the eastern Mediterranean, only Israel reported an organized screening program based on FIT designed for people aged 50–74 years, while in other countries, only opportunistic screening has been adopted (Jordan, Qatar and the United Arabic Emirates). Finally, regarding Africa, the adoption of organized screening may have a limited impact due to the relatively low incidence of CRC and the limited economic resources [10,36].

#### **4. Disadvantages of Fecal Tests (FIT/gFOBT), Room for Improvement**

In order to fully benefit from screening programs, the participation rate should be higher than 80%, thus, low take-up is one of the main drawbacks of all the screening programs, with large differences among and within countries [10,16,17,19,31,36,43,44]. Numerous studies have addressed the pitfalls of FIT, including the high rate of false positives and negatives. A false positive FIT can create unnecessary psychological distress and superfluous requests for colonoscopies, with associated healthcare costs. Between 8 and 32% of FIT positive participants do not have significant lesions [19,45,46]. On the other hand, false-negative FIT results can delay CRC diagnosis and dissuade participants from subsequent evaluations [47–49].

The FIT sensitivity for CRC ranges from 91 to 71%, according to the Hb cut-off, with specificity ranging from 90 to 95%. For advanced adenoma (AA), the sensitivity falls from 40 to 25%, with specificity ranging from 90 to 95% [50,51]. There are several risk factors for CRC: gender, age, obesity, alcohol consumption, current or former smoking and the use of drugs, such as non-steroidal anti-inflammatory drugs (NSAIDs) or anticoagulants [24,27,47]. Differences in FIT performance by sex and age have been described. The pooled sensitivity of FIT for advanced neoplasia (AN) was higher in women than in men, with pooled specificities of 92 and 94%, respectively. Accordingly, De Klerk et al., found that the highest risk of a false positive was found for females and the use of NSAIDs [47]. Interestingly, low-dose aspirin was associated with a higher risk of false positives (FPs), suggesting a possible effect on bleeding of early lesions. Several factors appear to be associated with an increased risk of false-positive FIT: male sex, older age (>65), obesity, and current smoking [27,52].

The importance of improving FIT screening has led to various strategies aimed at finding a balance between resources, participation rates and large populations. Some countries have decided to increase the FIT positivity threshold, with the hope of reducing the false positive results and optimizing colonoscopy performance. However, a high threshold leads to a decreased sensitivity and an increased specificity only for advanced neoplasia [53]. This solution reduces the number of false positives, but also increases the false negatives. The decision to increase the FIT cut-off value "simply" to reduce the number of colonoscopies seems more "cost-effective" than "patient-effective", and the adjustment of the FIT cut-off value cannot be the only viable solution.

A novel approach to FIT screening was developed by Senore et al., who evaluated the sum of quantitative FIT results during consecutive negative screening rounds [54]. Subjects with a cumulative fecal Hb level ≥20 µg/g showed an 18-fold increase in their cumulative AN (CRC and AA) risk over the subsequent two rounds. This is an interesting approach; however, the number of false positive FITs would still be very high. Another option is to select gender-specific or age-specific cut-offs in FIT screening [53,55,56]. In a stratification model, patients could be assigned to different risk levels of finding AN, by combining different risk factors (such as sex, age or Hb value), or considering FIT separately from the prediction model. The risk-stratification based on prediction models might be better at predicting neoplastic outcomes, including all FIT results, and might enlarge the eligible population including younger subjects (<50 years) and/or people with a history of familiarity for CRC.

#### **5. New Tests**

#### *5.1. Fecal Tests*

In recent years, new tests have been developed to optimize CRC screening and diagnosis. Since the first studies on *RAS* oncogene mutations [57], DNA alterations in stools have been investigated, as well as proteins and microRNAs.

A recent systematic review evaluated the performance of FIT combined with other stool markers, including DNA methylation, mutation or integrity markers (PHACTR3, APC, p53, KRAS and BRAF), proteins (transferrin, calprotectin and calgranulin) and microRNA (miR-106a) [58]. Notably, the largest increase in sensitivity for CRC was found

with long DNA as a measure of DNA integrity in the APC gene and p53, with a specificity of 98%. Among the protein markers, combining transferrin or calgranulin C tests to FIT yielded a slight increase in sensitivity [58,59]. However, calprotectin led to a significant increase in sensitivity for all adenomas, from 53 to 86%, but the specificity decreased from 68 to 26% [58,60].

Following the first studies on DNA and the feasibility of the prototype of a multitarget panel assay in stools [61], Imperiale et al. further developed and tested the multitarget stool DNA ColoGuard (MT-sDNA; Exact Science) in a large screening setting. This test is currently an alternative stool test, which was approved by the Food and Drug Administration (FDA) and is currently employed in the USA [62]. The MT-sDNA test, ColoGuard, combining stool DNA markers (methylated BMP3 and NDRG4 promoter regions, mutant KRAS) with the results of FIT, undoubtedly represented a milestone of stool-based molecular testing in CRC screening. It was tested in nearly 10,000 people, showing a significantly better sensitivity than FIT in predicting any stage of CRC (92.3 vs. 73.8%, respectively) and AA (42.2 vs. 23.8%), and retaining a high specificity (89.9 vs. 96.4%). Similar results were obtained by Bosch et al., in a cohort of 1014 people [63]. By comparing the single-application performance of the MT-sDNA test with FIT, MT-sDNA showed a greater sensitivity for AA than FIT at the lowest cut-off tested (10 µg Hb/g of feces), with a slight decrease in specificity (94 vs. 98%). No significant difference was highlighted to distinguish between proximal and distal AA.

A similar FIT-DNA test kit, ColoClear, manufactured by New Horizon Health Technology Corporation Limited in Hangzhou, China, calculates a risk prediction by combining the FIT test with the detection of the KRAS gene mutation, NDRG4 and BMP3 methylation. ColoClear was tested in 839 subjects, obtaining a sensitivity for CRC and AA of 97.5 and 53.1%, respectively, with a combined sensitivity for predicting AN (CRC and AA) of 88.9% and a specificity of 89.1% [64]. Moreover, no significant difference was highlighted for proximal or distal colon CRC, while sensitivity for distal AA was higher than for proximal AA (61 vs. 30%).

Another stool test, developed in Italy, with the collaboration of Diatech Pharmacogenetics, is based on the evaluation of stool DNA integrity [65,66]. The authors carried out a quantitative evaluation based upon fluorescence amplification of different genomic DNA targets called fluorescence long DNA (FL-DNA). FL-DNA showed a 70% sensitivity and 87% specificity in detecting CRC in stool samples from subjects recruited by a regional screening program based on FIT positivity.

Microbiome-based tests could represent a new frontier in the CRC detection. Grobbee et al. measured fecal microbiota in FIT positive subjects. An overall increase in total bacterial content (16S) was associated with patients affected by high grade dysplasia and CRC [67]. Similarly, a new non-invasive CRC screening test based on microbiome data was employed to reduce the false positive rate of FIT [68]. The authors targeted specific genomic DNA bacterial sequences: Eubacteria (EUB) as the total bacterial load, *Faecalibacterium prausnitzii* (B10), *Subdoligranulum variabile* (B46), *Ruminococcus*, *Roseburia* and *Coprococcus* (B48), *Roseburia intestinalis* (RSBI), *Gemella morbillorum* (GMLL), *Peptostreptococcus stomatis* (PTST), *Bacteroides fragilis* (BCTF), *Collinsella intestinalis* (CINT), and *Bacteroides thetaiotaomicron* (BCTT). GMLL, PTST and BCTF correlated significantly with AN. Although the sensitivity values for bacterial markers alone were much lower than FIT performance, a final algorithm consisting of the combination of FIT with three ratios between bacterial markers (PTST/EUB, BCTF/EUB, BCTT/ EUB) decreased the number of false positive results by 50%, obtaining a sensitivity of 80% and a specificity of 90%. Finally, panels of proteins were tested in stool samples to identify AA or AN subjects, however, the levels of sensitivity and specificity were quite low (54 vs. 13%) (Hp, LRG1, RBP4, and FN1; 62 vs. 40%) [69].

Nevertheless, despite these efforts to improve fecal-based screening, the main drawbacks remain the low participation rate and the costs [70,71].

#### *5.2. A New Alternative: Liquid Biopsy*

The analysis of tumor-derived biomarkers in biological fluids has the potential to increase the participation rate [72]. Peripheral blood is one of the most studied biological fluids, and an accurate blood test could be an attractive alternative for asymptomatic, average-risk individuals who are reluctant to undergo screening by a stool test or endoscopy. Two independent surveys [73,74] showed that a blood sample would be preferred to a stool sample in a screening setting. In a clinical trial, 12% of people who refused to enroll in a stool-based screening, agreed to perform the blood-based test [71]. Thus, blood CRC biomarkers remain very attractive and are under investigation, including several molecules from nucleic acids such as DNA and various types of RNA (messenger, mRNA; micro, miRNA; long non-coding, lncRNA) to proteins, from circulating tumor cells to microvesicles. There is also growing interest in biomarker combination, which could obtain a higher sensitivity than single biomarker-based tests [72,75,76]. *Cancers* **2021**, *13*, x 9 of 20

> Figure 3 summarizes the blood-based tests discussed in this review. The next part of this review focuses, above all, on mRNA, including microRNAs and lncRNA. DNA, proteins and microvesicles are also briefly discussed. Figure 3 summarizes the blood-based tests discussed in this review. The next part of this review focuses, above all, on mRNA, including microRNAs and lncRNA. DNA, proteins and microvesicles are also briefly discussed.

**Figure 3.** Routine and new tests used or proposed for colorectal cancer screening. Guaiac fecal occult blood test (gFOBT), fecal immunochemical test (FIT), double-contrast barium enema (DCBE), computed tomographic colonography (CTC), colon capsule endoscopy (CCE), colonoscopy (CS), flexible sigmoidoscopy (FS), fluorescent long DNA (FL-DNA), not available (NA), false positive (FP). 1 The reported percentages of sensitivity and specificity refer to colorectal cancer. 2 Refers only to distal colorectal cancer. 3 Refers to advanced neoplasia. 4 Refers to the average value among 8 cancer types. **Figure 3.** Routine and new tests used or proposed for colorectal cancer screening. Guaiac fecal occult blood test (gFOBT), fecal immunochemical test (FIT), double-contrast barium enema (DCBE), computed tomographic colonography (CTC), colon capsule endoscopy (CCE), colonoscopy (CS), flexible sigmoidoscopy (FS), fluorescent long DNA (FL-DNA), not available (NA), false positive (FP). <sup>1</sup> The reported percentages of sensitivity and specificity refer to colorectal cancer. <sup>2</sup> Refers only to distal colorectal cancer. <sup>3</sup> Refers to advanced neoplasia. <sup>4</sup> Refers to the average value among 8 cancer types.

> next generation sequencing (NGS) is promising for the diagnosis, prognosis and therapy 192

The emergence of RNA sequencing (RNA-seq) technologies with the evolution of

In 2016, Rodia et al., [78] used a novel bioinformatic approach to search for specific RNAs with high differential gene expression between CRC and normal blood. The genes showing the highest significant difference were analyzed by qRT-PCR in blood samples of healthy and CRC patients. The authors reported that *CEACAM6*, *LGALS4*, *TSPAN8* and

of cancers including CRC. However, this method is very expensive and, paradoxically, provides too much information that is not yet fully exploitable for the purpose of early

5.2.1. mRNA

diagnosis [77].

#### 5.2.1. mRNA

The emergence of RNA sequencing (RNA-seq) technologies with the evolution of next generation sequencing (NGS) is promising for the diagnosis, prognosis and therapy of cancers including CRC. However, this method is very expensive and, paradoxically, provides too much information that is not yet fully exploitable for the purpose of early diagnosis [77].

In 2016, Rodia et al. [78] used a novel bioinformatic approach to search for specific RNAs with high differential gene expression between CRC and normal blood. The genes showing the highest significant difference were analyzed by qRT-PCR in blood samples of healthy and CRC patients. The authors reported that *CEACAM6*, *LGALS4*, *TSPAN8* and *COL1A2* (known as CELTiC) discriminated between the two groups with a sensitivity and specificity of 92 and 67%, respectively. The CELTiC panel was subsequently analyzed in a population of FIT positive subjects, confirming its ability to identify patients with high-risk lesions (CRC and AA), and appeared able to discriminate false positive FIT and low risk patients (non-advanced adenoma and polyps) [79]. In 2020, the CELTiC panel was measured in blood samples from healthy FIT negative subjects, reporting significant gender differences for *CEACAM6* and *COL1A2*, thus highlighting the importance of gender as a potential factor in the comparison between healthy and FIT false positive subjects. The CELTiC panel obtained high AUCs when comparing healthy to AN, low risk, FIT false positive subjects or a combination of these groups, with good sensitivities and specificities ranging from 83 to 90% and from 76 to 81%, respectively. These results confirmed the need for additional studies to better define gender- and age-specific reference intervals for the early diagnosis of CRC [80].

A similar approach was applied using ColonSentry, a panel of seven mRNAs [81,82]. Six out of seven genes (*ANXA3*, *CLEC4D*, *LMNB1*, *PRRG4*, *TNFAIP6* and *VNN1*) were overexpressed in the blood of CRC patients, and one (*IL2RB*) was under expressed with a blinded validation test set resulting in 72% sensitivity and 70% specificity, with similar predictive values for left- and right-sided CRC.

A similar test is COLOX [83,84], a panel of 29 mRNAs (BCL3, IL1B, PTGS2, MAP2K3, PTGES, PPARG, MMP11, CCR1, EGR1, CACNB4, CES1, IL8, S100A8, CXCL11, ITGA2, NME1, JUN, TNFSF13B, CXCR3, MAPK6, CD63, ITGB5, GATA2, LTF, MMP9, CXCL10, MSL1, RHOC, FXYD5) measured in the peripheral blood mononuclear cells. Few individual genes showed significant differences among age classes, but the whole panel was not affected by the age of the patient. Twelve mRNAs (BCL3, IL1B, PTGS2, PTGES, PPARG, MMP11, CCR1, EGR1, CACNB4, CES1, IL8, S100A8) were able to differentiate between the control group and CRC, and five mRNAs (CES1, CXCL11, IL1B, ITGA2, NME1) identified large adenomas. The authors also found seven markers specifically able to differentiate between large adenomas and CRC (BCL3, PTGES, PPARG, MMP11, IL8, TNFSF13B, CXCR3).

Other mRNAs have also been investigated as blood markers of CRC [85–87]. In 2018, Alamro et al. reported significantly higher mRNA expression of inflammatory genes (COX-2, TNF-α, NF-κB, IL-6) in blood samples of 20 CRC compared to 15 healthy controls without significant association with gender, age or tumor localization [85]. A case–control study performed on 83 CRC patients and 11 healthy donors resulted in significantly higher levels of circulating HMGA2 mRNA in CRC patients with an AUC of 0.932 and a sensitivity of 86.8%. The authors highlighted also a significant association with tumor localization, reporting a greater expression in patients with colon cancer and right-sided CRC, but not with age or gender of the patients [86]. Hamm et al. performed a genome-wide expression analysis on RNA obtained from peripheral blood monocytes collected from 329 subjects (128 healthy, 160 CRC, 41 other gastric diseases) divided in different cohorts [87]. Twentythree genes showed differential expression between healthy and CRC. By testing different statistical models, the authors reported sensitivity values from 80 to 100%, specificity from 92.3 to 93.3%, and AUCs from 0.86 to 0.99. However, the panel was not tested for the evaluation of preneoplastic lesions (i.e., polyps) or tumor localization [87].

#### 5.2.2. miRNA

MicroRNAs (miRNAs) are small non-coding RNAs (~20–22 nucleotides) that regulate gene expression through repression or degradation of mRNAs. miRNAs seem to be promising plasma biomarkers associated with the onset of CRC, and several studies have searched for specific panels of miRNA capable of increasing both the sensitivity and specificity of screening [4,7,88–91]. miR-7, miR-17-3p, miR-18, miR-21, miR-29a, miR-31, miR-92a, miR-93, miR-155, miR-181b, miR-200c, miR-221, miR-409-3p, let-7g are some of the miRNAs tested, either individually or as a panel in plasma or serum of patients affected by CRC. However, only a few have been confirmed as diagnostic CRC biomarkers by more than one study [89–93]. Among the most studied, mir-21 and mir-29 family (mir-29a, mir-29b and mir-29c) are overexpressed in CRC and associated with CRC progression and metastasis [7,91]. miR-21 is one of the most investigated diagnostic markers of CRC, identified in several biological fluids (plasma, serum, whole blood) [72,94].

Clusters of miRNA have been taken into consideration and the mir-17–92 cluster, also called oncomiR-1, is one of the most studied clusters in relation to CRC. This cluster contains different members such as miR-17, miR-18a, miR19a, miR-19b, miR-20a and miR-92a, and evidence suggests that miR-29a and miR-92a may have a good sensitivity (69 to 89%) and specificity (70 to 89.1%) in CRC detection. In addition, the combination of miR-92a and miR-29a appears to increase the performance of single miRNAs for detecting AN [75,91,95]. Other members, such as miR-19a and miR-19b, were upregulated in plasma from CRC patients compared to healthy individuals, and their combination obtained an AUC of 0.82. The combination with another four plasma miRNAs (miR-18a, miR-29a, miR-15b and miR-335) showed promising results in differentiating between controls and CRC (AUC 0.95) or AA (AUC 0.91) and with similar performances for proximal (AUC 0.97) and distal (AUC 0.95) CRC [75,96]. These results highlight the importance of combinatorial approaches involving specific panels of miRNAs. Other panels of miRNA have recently been reported [97–99]. A panel of seven miRNAs (miR-103a-3p, miR-127-3p, miR-151a-5p, miR-17-5p, miR-181a-5p, miR-18a-5p and miR-18b-5p) was identified and evaluated in a four-stage experiment (screening, training, testing and external validation) involving a total of 139 CRC patients and 132 controls. The performances of the panel obtained an AUC of 0.895 with a sensitivity and specificity of 76.9 and 86.7%, respectively, without significant associations between serum levels of the analyzed miRNAs and age, gender or location [99].

However, the use of mRNA and miRNA is still limited due to the lack of extensive clinical validations.

#### 5.2.3. DNA

In addition to RNA, DNA has also been widely studied in liquid biopsies searching for CRC biomarkers. Cell-free DNA (cfDNA) and the tumor-derived fraction termed as circulating tumor DNA (ctDNA) are of great interest. cfDNA mutations in genes frequently associated with tumorigenesis have been assessed for the early detection of the most common tumor types, including CRC [72,100,101]. KRAS mutations were detected in plasma from CRC patients, however it was also reported that 0.45–20% of healthy individuals may carry genomic alterations in cfDNA, with particular regard to TP53 and KRAS variants [72]. The low abundance of tumor-derived DNA is one of the main challenges for early detection, as well as cancer-associated mutations accumulated with age. Aberrant DNA methylation is a feature of most solid cancers and is a promising biomarker for early diagnosis [75,100,102]. Septin 9 (SEPT9), a GTP-binding protein belonging to the Septin family, is one of the most widely studied DNA markers in blood in relation to CRC. In fact, CRCs show an atypical methylation status of SEPT9 gene. The EpiProcolon assay, which detects circulating methylated SEPT9 (mSEPT9), was recently approved by the FDA [102,103]. The values of sensitivity obtained from independent studies for mSEPT9 ranged from 48.2 to 95.6%, with specificity ranging from 79.1 to 99.1%. In the most recent studies, the sensitivity of the EpiproColon test 2.0 ranged from 61.2 to 82.2% and

the specificity from 83.6 to 95.1%, showing a better performance than carcinoembryonic antigen (CEA) and/or FIT tests in the screening of asymptomatic populations [75,102–104]. However, mSEPT9 is not able to distinguish between CRC and polyps or adenomas and seems not affected by tumor localization, but may be affected by age or sex, suggesting that age- and sex-specific cut-offs are required to better optimize the screening and diagnostic procedures [75,102,104]. The combination of mSEPT9 with the FIT test seems to improve the sensitivity for CRC and AA detection obtaining 94 and 43%, respectively, but at the cost of losing the specificity [76].

Other ctDNA markers, such as BCAT1 and IKZF1, have been studied. BCAT1 and IKZF1 methylation obtained a sensitivity of 66% and a specificity of 94% for CRC detection in a prospective study analyzing more than 2000 individuals, including 129 people with CRC [75,105]. A different approach was recently applied by Cohen et al. [106] who described a multi-analyte test (CancerSEEK) to identify eight common cancers, including CRC, by determining the levels of circulating proteins and mutations in ctDNA. The median sensitivity of CancerSEEK was 73 and 78% for stage II and III cancers, respectively, and 43% for stage I cancers. In particular, 14 out of the 16 genes tested (AKT1, APC, BRAF, CDKN2A, CTNNB1, FBXW7, FGFR2, GNAS, KRAS, NRAS, PIK3CA, PPP2R1A, PTEN, TP53) were detected in plasma samples of CRC patients. CRC was also the type of cancer detected with the highest prediction accuracy.

Another approach [107] analyzed cfDNA and ctDNA by applying targeted error correction sequencing (TEC-Seq) for the sensitive and specific detection of low-abundance sequence alterations using NGS in commonly altered cancer genes. In plasma samples from 44 healthy individuals and 194 patients affected by CRC (n = 42), lung (n = 65), ovarian (n = 42) or breast (n = 45) cancers, the authors analyzed a panel of 55 cancer driver genes. cfDNA was significantly higher in cancer patients than in healthy individuals and, within CRC patients, stage IV showed significantly higher cfDNA than stages I to III. In addition, 83% of CRC patients had detectable alterations in driver genes (ctDNA). These detection rates were higher in patients with stages II, III and IV, (89, 90 and 94%, respectively) and were also detected in half of the patients with stage I cancer, suggesting that larger panels of ctDNA may improve the ability to detect small tumors and pre-neoplastic lesions.

cfDNA in the blood samples of CRC patients has also been studied using NGS, machine-learning approaches, genome sequencing and digital sequencing technologies [108–111]. The various models applied by Wan et al. [108] obtained a variable sensitivity (71–85%) with 85% specificity, showing promising preliminary results.

#### 5.2.4. Proteins

Carcinoembryonic antigen (CEA) and carbohydrate antigen (CA19-9) are two of the most studied gastrointestinal tumor-associated proteins in blood (or plasma/ serum) [90,112–114]. Serum CEA and/or CA19-9 levels are significantly higher in CRC patients compared to healthy subjects and are well-known cancer markers. However, CEA and CA19-9 concentrations may also be high in other conditions or tumors and their usefulness as CRC screening biomarkers is still an open issue. However, today CEA and CA19-9 are used and approved in clinical practice to detect metastatic disease, recurrence, or to monitor response to treatments [114–120].

Proteomic approaches have recently been applied to blood samples (or plasma/serum) of CRC patients to search for new biomarkers in screening or diagnosis [121–124]. Chen and colleagues performed protein profiling quantifying tumor-associated protein biomarkers in CRC and healthy control plasma samples. Seventeen proteins showed significantly different concentrations between CRC and controls, nine were overexpressed (CEA, GDF-15, AREG, IL-6, CXCL10, CXCL9, PSA, TNFα, cathepsin-D), and eight were downregulated (HGF receptor, CXCL5, ERBB4, FLT3L, CD69, EMMPRIN, VEGFR-2, Caspase-3). Carcinoembryonic antigen (CEA), growth differentiation factor 15 (GDF-15), and amphiregulin (AREG) were the most significant. In addition, applying a logistic regression model, the authors constructed a multi-marker prediction algorithm including eight markers (IFNg, EMMPRIN, ERBB4, PSA, CD69, AREG, HGF receptor and CEA) reporting moderate sensitivities (44–65%) at high specificities (80–90%) [117].

Finally, an additional panel of eight plasma proteins including AFP, CA19-9, CEA, hs-CRP, CyFra21-1, Ferritin, Galectin-3 and TIMP-1 was tested in 4698 subjects including CRC, AA, non-advanced adenomas and extracolonic cancers [115,125]. All the individual biomarkers significantly identified AN (CRC + AA) and the multivariable model including all the biomarkers and age and gender obtained an AUC of 0.76, with 80% sensitivity and 50% specificity. However, the various models tested, including a combination of the eight proteins, showed moderate performances (90% specificity, 19% sensitivity) in discriminating AA from other conditions (non-advanced adenomas, non-colonic tumors, healthy).

#### 5.2.5. Extracellular Vesicles

Extracellular vesicles (EVs), such as exosomes (EXOs), microvesicles (MVs) and large oncosomes, may contain promising biomarkers. Three main categories divide EVs on the basis of biogenesis and approximate size: EXOs (~40–100 nm) derive from multivesicular bodies within the cells; MVs (~100 nm–1 µm) are formed from the outward budding of the plasma membrane; apoptotic bodies (APs) (~1–5 µm) arise from dying cells undergoing apoptosis [72]. In addition to these classes, some cancer-specific subtypes of EVs have been identified: oncosomes (~100–400 nm) produced by non-transformed cells, whose contents can determine oncogenic effects, and large oncosomes (~1–10 µm) derived from malignant cells [126]. EVs contain proteins, RNA, DNA and lipids, which reflect in part the composition of the cell of origin. By protecting nucleic acids from degradation, EVs could also be considered a better source for tumor molecular profiling compared with cell-free nucleic acids [126,127]. EVs and EXOs are also secreted by cancer cells and in a greater amount than normal cells, therefore, increasing the transfer of RNAs, growth factors and chemokines participating in cancer progression [128–130]. Examples of molecules identified in EVs at increased levels include surface proteins detected by flow cytometry, such as the epithelial cell adhesion molecule (EpCAM), CD9, CD81, CD63 and CD147 in the bloodstream of CRC patients [126,131–133].

One of the drawbacks of studying EVs and exosomes is the lack of a standardized protocol to isolate them from blood and to extract their content or surface material [133–137]. Despite the differences in EV isolation and although most of the studies are case–controls, miRNA is one of the classes most studied as a biomarker in EVs.

Ogata-Kawata et al. [138] evaluated 88 CRC patients and 11 controls to assess the ability of serum EV-miRNAs. In particular, miR-21, miR-23a, and miR-1246 differentiated CRC patients (all stages) from controls [138]. By comparing serum EV-miRNA from CRC patients to healthy controls, Yan et al., found that miR-486 was significantly upregulated, while miR-548c was significantly downregulated [139]. Liu et al. also reported an increase in miR-486 levels in the serum EVs of CRC patients compared to healthy subjects [140]. Peng et al. further confirmed the downregulation of serum EV miR-548c in CRC patients, also finding an association with shorter survival and liver metastases [141].

Other authors have evaluated panels of EV-miRNA. Min et al. [142] analyzed EVmiRNA from blood samples of early-stage CRC patients and non-cancerous controls. The authors found 38 miRNAs upregulated and 57 downregulated in CRC patients compared to healthy controls, some of which, such as Let-7b-3p, miR-150-3p, miR-145-3p, miR-139-3p, had already been reported in the plasma of CRC patients. ROC curve analysis of the single miRNAs reached AUCs of 0.792, 0.686, 0.692, and 0.679, respectively. On the other hand, a logistic model including let-7b-3p, miR-139-3p, and miR-145-3p, confirmed the increased potential of panels of EV-miRNAs compared to individuals ones with an AUC of 0.927 [142].

Cha et al. [143] evaluated eight mRNA markers (MYC, VEGF, CDX2, CD133, CEA, CK19, EpCAM, and CD24) extracted from plasma EVs. Of the eight mRNAs, the combination of VEGF and CD133 showed statistically significant differences between healthy

and CRC, and obtained an AUC of 0.96 with 100% sensitivity and 80% specificity in discriminating between the two groups.

EVs also contain other types of RNA, such as long non-coding RNA (lncRNA) or mRNA which are dysregulated in CRC. The presence and performance of lnc-RNA and mRNA has been assessed in APs, MVs and EXOs [144]. In a first screening, in sera of CRC patients and healthy subjects, 21 lncRNAs and 16 mRNAs showed significant differences between EXOs of healthy and CRC samples. In the subsequent validation phase, tested in 30 CRC, 20 adenoma and 30 healthy subjects, the combination of lncRNA breast cancer antiestrogen resistance 4 (BCAR4) with two mRNAs (keratin-associated protein 5–4, KRTAP5-4, and melanoma antigen family A3, MAGEA3) provided the greatest predictive ability, with an AUC of 0.877.

#### **6. Conclusions**

There is a long history of CRC screening tests and several studies have attempted to discover cancer biomarkers in stool or blood samples. However, most of the identified biomarkers, (mRNAs, miRNAs, ctDNA, EVs) have only been evaluated in preliminary case–control studies.

In order to improve the screening and the diagnosis of CRC, large-scale randomized studies are needed to confirm the clinical benefits and the usefulness of these tests. In particular, RNA-seq and NGS, could be used to describe and characterize the evolution and development of CRC, in order to discover new and earlier biomarkers, thus improving outcomes. On the other hand, qRT-PCR may be simpler and cheaper when applied to panels of biomarkers aimed at higher levels of performance in terms of sensitivity, specificity, accuracy and speed of execution.

In addition to the differences in sensitivity and specificity between tests, and sometimes the lack of extensive investigating trials, the main drawback remains the low participation rate. However, the use of blood samples may change this trend. Liquid biopsy could be also used to assess the prognosis, response to therapies and during follow-up.

Finally, the development of algorithms, including those derived with artificial intelligence, which associate outcome-influencing parameters such as gender and age with candidate markers, will be a further tool to improve the current efficacy of CRC screening.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2072-669 4/13/5/1101/s1, Table S1: World and colorectal cancer in 2020. New cases, incidence and mortality rates estimation in continents. Obtained from Global Cancer Observatory (GBO) 2020, International Agency for Research on Cancer (http://gco.iarc.fr/ (accessed on 21 January 2021) [6]), World Health Organization, Table S2: World and colorectal cancer in 2020. New cases, incidence and mortality rates estimation in world areas. Obtained from Global Cancer Observatory (GBO) 2020, International Agency for Research on Cancer (http://gco.iarc.fr/ (accessed on 21 January 2021) [6]), World Health Organization, Table S3: Europe and colorectal cancer in 2020. New cases estimation, incidence and mortality rates in European countries for males and females. Obtained from Global Cancer Observatory (GBO) 2020, International Agency for Research on Cancer (http://gco.iarc.fr/ (accessed on 21 January 2021) [6]), World Health Organization, Table S4: Italy and colorectal cancer. New cases estimation, incidence and mortality for regions of Italy for males and females, Table S5: Colorectal cancer screening in Italy. Invitation rates of the target population and participation rates of the invited people in northern central and southern of Italy, Table S6: Colorectal cancer screening in Italy. Population, percentage of people older than 65 years and participation in FIT screening of the invited target population for each region of Italy.

**Author Contributions:** Writing—original draft preparation, E.F. and R.S.; figures and editing, M.S.; writing—review, L.R. and M.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by Spin-Off Projects (J34I19003100002 Finanziamento regione Emilia Romagna per Assegni Call for Business Plan 2019) and POC (Proof of Concept) grants from the University of Bologna and the Emilia-Romagna regional administration (co-funded with EU resources FSE 2014/2020).

**Data Availability Statement:** Data are contained within the article or supplementary material. The data presented in this study are available in Supplementary Tables S1–S6.

**Conflicts of Interest:** R.S. and M.L. have intellectual property rights on an international patent pending (WO2016/185451: method and kit for diagnosis of colorectal cancer; USA patent number 10900085; European patent office number EP3298165). R.S., L.R. and M.L. have intellectual property rights on an international patent pending (WO/2019/138303: new prognostic method). The remaining authors (E.F. and M.S.) declared no conflict of interest. The funders had no role in the design of the study; in the collection, analysis, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**

