Several authors have investigated molecular non-invasive screening tests for early detection of CRC. DNA, RNA and other molecules derived by tumour in stool, as well as their concentrations in blood have been studied extensively. Colorectal process of carcinogenesis is characterized by genetic and epigenetic alteration transforming normal cells into cancer cells. Most studies concerning molecular markers in stool have focused on the detection of tumour DNA. These investigations have concentrated on the detection of mutated KRAS, TP53, APC and markers for microsatellite instability (MSI)[7-9]. A faecal DNA test targeted at molecular biomarkers has been commercially available for twelve years, with reported sensitivity for cancer ranging from 25% up to 92% for the latest tests based on BEAMing technology, and 94%-98% specificity[10-12]. Apart from genetic alterations, the DNA promoter hypermethylation silencing the tumour suppressor genes has been widely investigated. Epigenetic changes, depending on the markers or their combinations evaluated, have been detected in CRC patients with 70%-96% sensitivity and 72%-96% specificity[9,13,14]. Many combinations of genetic and epigenetic markers have been studied, but until now, the results have not endorsed their use in clinical practice. Using blood instead of stool as a screening material could offer some obvious advantages. Several studies have evaluated potential plasma DNA genetic and epigenetic biomarkers in CRC detection. The overall sensitivity ranges from 30% to 87%, with specificity of up to 96%. The use of RNA biomarkers in stool has not been investigated as extensively as was the case for DNA biomarkers, mainly because stool environment is responsible for mRNA degradation, although improving laboratory retrieval methods seems to solve this problem. Koga et al[15] analysed mRNA expression of MMP7, PTGS2, TP53 and MYBL2 in colonocytes isolated from stool by quantitative real-time RT-PCR, to find out that these markers can identify CRC patients with 58% sensitivity and 88% specificity. Sensitivity was found to depend on tumour size and tumour location, but not cancer stage[15]. Most recently, the so called transcriptomic studies have investigated the expression of miRNAs - short, non-coding 18-22 nucleotide RNA molecules in stools of CRC patients. The most extensively studied miR21, miR106a, miR135, miR17-92 were found to be overexpressed in CRC patients compared with healthy individuals[16,17]. As was the case with RNA markers in stool, many studies have been evaluating mRNA of different tumour genes in whole blood, plasma or circulating tumour cells to identify new CRC screening markers. Most of them investigated mRNA molecules of CK19, CK20, or Carcinoembryonic antigen (CEA). The overall sensitivity of these markers was up to 72%, specifically when combinations of these markers were used[18,19]. The specificity was very high with healthy control samples or much lower when compared to other cancer or inflammatory bowel diseases samples[20]. Recent studies have indicated that circulating miRNAs may be involved in the process of oncogenesis. The use of miRNA as a biomarker is now being evaluated. A large number of miRNA molecules have been assessed, with a focus on miR145, miR143, miR135, miR17-92. More specifically, Huang et al[21] has found that plasma miR29a and miR92a demonstrated a significant diagnostic value for advanced neoplasia with 83% and 84% sensitivity and specificity, respectively, in discriminating CRC patients. These studies need to be validated in randomised trials to define their value in CRC screening.

Protein markers for screening and early detection of CRC can be divided into tumour TAAs, antibodies against TAAs, and other CRC-relevant proteins. CEA was discovered almost 50 years ago, in 1965, and it still remains the only tumour marker of recognised efficacy in monitoring CRC patients’ therapy[22]. CEA was first considered specific for CRC, but elevated CEA levels were later detected in other neoplasms too, e.g. gastric and pancreatic cancers, and in inflammatory conditions. Elevated CEA concentrations are only rarely identified in CRC stage I. Moreover, CEA does not differentiate benign versus malignant polyps. According to The European Group on Tumor Markers, European Society of Medical Oncology and American Society of Clinical Oncology guidelines[2,23,24], CEA is not recommended for use in screening tests. Recently, some studies have investigated the advantages of mRNA molecules encoding CEA for the detection of CRC, but the results were not superior to those of CEA[19].

In some studies, high CEA concentrations in patients with CRC stage II and III were found to be potentially indicative of more aggressive types of cancer[25,26]. Earlier, the Colorectal Working Group of American Joint Committee on Cancer proposed to include CEA baseline concentration to the traditional TNM classification as the so-called C-stage. C-stage was proposed to be divided into Cx, C0 (CEA < 5 ng/mL) and C1 (CEA > 5 ng/mL) substages[27]. The meaning of CEA as an independent prognostic factor was also confirmed in a recent retrospective analysis of 17910 patients with CRC, with a mean 27-mo follow-up, with longer survival periods for patients with IIA C0 and IIIA C0 vs I C1, IIIA C0 vs IIA C1, and IIIB C0 vs IIB-C C1, respectively[28]. No study, however, has shown that CEA concentration level can be used to select those patients with stage II CRC who would benefit from adjuvant chemotherapy. From a prognostic point of view, it appears reasonable to determine CEA levels before surgery in patients with disseminated CRC. The roles of CEA in determining life expectancy was confirmed in several studies on patients with liver metastases[29,30]. Recent study proved that combined use of CEA and serum amyloid A (SAA) is able to identify patients with favourable and poor prognosis. In addition to tumour baseline parameters, routine analysis of CEA together with SAA provides improved prognosis value on cancer specific survival and disease-free survival in resected rectal cancers[31]. CEA half-life is known to last approximately 7 d. After R0 resection surgery, CEA levels should return to normal within 4 to 6 wk. Sustained elevated CEA levels can be indicative of infiltration or metastases. Slow increase in CEA concentrations after surgery is a typical sign of local recurrence, whereas dynamically increasing levels can be symptomatic of metastases, most probably in the liver[24,32]. Testing CEA levels is considered most cost-effective in detecting post-surgery recurrences[24]. Please note that CEA levels tested every 3 mo for the first 3 years and thereafter every 6 mo for subsequent 2-3 years is a golden follow-up standard after CRC therapy recommended by a number of scientific associations[2,23,33]. It appears particularly important in asymptomatic patients, in whom chemotherapy can be used, with a much longer life expectancy as compared to treatment administered after the onset of symptoms of recurrence. CEA is a marker of choice in monitoring disseminated disease during systemic therapy. Constant increase in CEA levels is typically associated with a progression of the disease, even though radiological tests may prove otherwise[23,24]. However, chemotherapy can also result in temporary increase in CEA concentration, which must be also taken into account. Therefore, it is not recommended to test CEA levels within 2 wk of chemotherapy, whereas in patients on oxaliplatin, tests can be carried out after 4 to 6 wk.

Cancer antigen 19-9 (CA 19-9) is a glycoprotein whose relevance in CRC diagnosis still remains an issue. The majority of researchers arrived at the conclusion that CA 19-9 sensitivity is much inferior to that of CEA, and that elevated CA 19-9 levels is a poor prognostic factor[2,23,34-36]. Other carbohydrate antigens: CA 195, CA 50 have been also investigated, but with comparatively disappointing results. CA 72-4 is a biomarker with poor sensitivity ranging from 9% to 31% and better specificity ranging from 89% to 95% in patients screened for CRC. The diagnostic information in recurrent CRC provided by CA 72-4 has borderline significance, by far worse than CEA. All authors conclude that CA 72-4 sensitivity is rather low and specificity incomplete in screening and following up in patients with CRC. On the other hand, an algorithm based on combination of CEA, CA 19-9, CA 72-4, CA 242, CYFRA21-1 improves the diagnostic accuracy compared with these biomarkers alone[34-39]. Among other protein markers examined for screening purposes, two have been extensively investigated: the tumour specific M2 isoform of pyruvate kinase (M2-PK) in stool and tissue inhibitor of matrix metalloproteinase 1 (TIMP1). M2-PK measured in stool showed relatively high sensitivity for CRC up to 91%, and much lower for adenomas[40,41]. Plasma level of TIMP1 is reported to be elevated in CRC patients and prospective studies have been carried to assess its utility as biomarker. The results of the study included more than 4500 patients screened by endoscopy for CRC demonstrated that TIMP1 is not significantly superior to CEA marker in cancer screening and is not suitable for the detection of premalignant lesions[42]. Tissue polypeptide-specific antigen (TPS) and tissue polypeptide antigen (TPA) which detects the fragments of cytokeratines 8, 18 and 19 due to lack of sensitivity and specificity can not to be recommended in CRC screening. The majority of investigators have found that increased levels of TPA and TPS are observed in metastatic stage of CRC. A further studies has suggested that combination of TPA and CEA rises the sensitivity of these biomarkers in identifying the patients with CRC recurrence[34,37,43,44]. Other biomarkers, such as: thymidine phosphorylase (TP), DNA ploidy were determined to be insignificant in detecting, staging and following-up of patients with CRC[23].

MSI denotes changes in coding and non-coding sequences of microsatellite chromosomes, i.e. repeated DNA sequences. These sequences are particularly exposed to errors in the mutation repair system that consist in the loss or multiplication of nucleotide sequence repetitions, which results in shortening or extension of microsatellite regions in neoplastic cells. Mutations arising out of these processes are eliminated by mismatch repair genes (MMR) such as MSH2, MSH6, PMS2 and MLH1, which makes some researchers believe that MSI can be caused by mutations in these genes[45]. Microsatellite instability can be classified into microsatellite instability-high (MSI-H), and microsatellite instability-low (MSI-L), depending on the percentage of loci that correlate to MSI characteristics. Tumour cells that lack MSI features are designated as MSS.

In retrospective studies and meta-analyses in patients with CRC stage II and III, MSI-H was shown to be a predictive factor that improved overall survival (OS), irrespective of the progression (stage) of cancer. A lower incidence of lymph node metastases and distant metastases as compared to MSI-L or MSS cancer cells was also observed[46-49]. MSI status is currently recommended in the WHO classification of mucinous-type CRC - MSI-H indicates good prognosis, MSI-L or MSS - poor outcome. However, MSI should be considered more of a prognostic rather than predictive factor. This conclusion is based on equivocal results of studies evaluating the efficacy of 5-FU-based chemotherapy in groups of patients with MSI-H and MSI-L or MSS. Ribic et al[48] examined tumour specimens collected from 570 patients with CRC stage II and III and correlated the test results with chemotherapy outcomes in these patients to reveal a tendency for shorter overall survival in patients with MSI-H on adjuvant therapy. Significant improvement was observed in patients with MSS tumours. A recent pooled analysis of randomized clinical studies revealed significant decrease in the overall five-year survival rate for patients with CRC stage II and MSI-H on 5-FU-based chemotherapy. 5-FU-based chemotherapy was found to improve therapeutic outcomes only in patients with CRC stage III and MSI-L or MSS[50]. Some studies indicated potentially negative effects of 5-FU-based chemotherapy in patients with MSI-H. A longer survival rate as compared to patients on 5-FU-based adjuvant chemotherapy was observed in a reference group of patients undergoing surgical treatment. Resistance of MSI-H tumours to 5-FU was also confirmed in in vitro studies[51]. A completely different conclusion can be drawn from earlier studies of Elsaleh et al[52,53], which confirmed the efficacy of 5-FU in patients with CRC stage III and MSI-H. Recent study also proved that prognostic value of MMR mutation was similar in the presence or absence of fluorouracil and folinic acid chemotherapy[54]. Beragnolli et al[55] revealed that a higher rate of overall 5-year progression-free survival was observed in patients with CRC stage III and MSI-H on 5-FU and irinotecan vs 5-FU-based chemotherapy. To recap, the results of MSI studies and clinical experience in patients with CRC stage II indicate that the degree of microsatellite instability may be of significance as a prognostic factor. Also, adjuvant 5-FU-based chemotherapy was proved to provide no benefits (or potentially cause adverse reactions) in patients with MSI-H. Further research is needed to investigate whether the MSI status can predict benefit (in high-risk patients) from irinotecan-based treatment or oxaliplatinum-based therapy.

A number of studies were dedicated to another prognostic factor in patients with CRC stage II and III - chromosome 18q loss of heterozygosity in the coding place of, inter alia, SMAD 4 proteins specific to CRC. In these studies, the overall 5-year survival was poorer for patients with CRC stage III and 18qLOH as compared to non-18qLOH patients[56]. A meta-analysis of data from 27 studies and 2189 patients by Popat et al[57] confirmed that poorer survival was correlated with 18q chromosome deletion. Two years later, the same research team questioned these findings after re-examining the same data[58]. Likewise, no correlation was identified between the presence of 18qLOH and 5-year survival in patients with non-MSI-H phenotype[59]. The role of 18qLOH in predicting response to standard chemotherapy has not been yet fully confirmed. Watanabe et al[60] demonstrated better response to 5-FU-based chemotherapy in patients with CRC stage III and MSS and with the absence of 18q chromosome deletion vs. patients in whom 18q chromosome deletion was present. The recently published results of the same research team can be a proof that in patients with CRC stage II and III and MSS-H (> 33%), the level of LOH of four chromosomes, including 18, is correlated with significantly poorer survival rate as compared to patients with MSS and LOH-L or MSI-H phenotype[61].

Based on the available data, 18q chromosome deletion cannot be the sole basis for any therapeutic decisions, however, it is being more closely examined under ECOG 5202 study, featuring molecular markers identified so far in selecting the most proper adjuvant post-surgery treatment, by prospectively analysing the role of MSI and 18qLOH in prognosis and therapeutic decisions in patients with CRC stage II. Patients with good prognosis (with MSI-H and w/o 18qLOH) were followed-up, and patients with poor prognosis (with MSI-L or MSS and 18qLOH) were randomized to one of two groups on chemotherapy (FOLFOX alone or FOLFOX and bevacizumab). The results of E5202 are expected in the next few years. No conclusion can be drawn from this study about the possible inefficacy of chemotherapy in patients with MSI-H, however, the study will include a multifactor analysis of biomarkers that can assist in taking therapeutic decisions in other groups of patients[62].

Mutation in the tumour suppressor gene p53 (chromosome region 17p13) occur in 50%-70% of all CRC and is associated with worse outcomes, including disease free survival and overall survival[63]. Results obtained from a study that included more than 3500 CRC patients confirm the prognostic value of p53 mutation, which seems to be determined by the primary tumour site. Patients with p53 mutation and tumour of proximal colon had better OS when treated with adjuvant chemotherapy compared to those treated by surgery alone[64].

A number of currently tested markers have been discovered in the course of studies on epidermal growth factor receptor (EGFR) signalling pathways. KRAS gene mutation on short arm of chromosome 12 at codon 12 (80% of patients) or, to a lesser extent, codon 13 is believed to be of use as a biomarker in patients on cetuximab or panitumumab[65]. These mutations are one of the most common in proliferative diseases (37% and 13%, respectively), and their significance in CRC carcinogenesis has been examined in much detail[66]. As these mutations are present in EGFR signalling pathway, they can be a predictive factor for therapy with anti-EGFR antibodies. In studies performed so far, KRAS mutation was found to be correlated with non-responsiveness to cetuximab and panitumumab[67,68]. CRYSTAL and OPUS data indicate that the effectiveness of FOLFOX or FOLFIRI alone is no inferior to that of cetuximab in patients with KRAS in combination with chemotherapy according to FOLFIRI and FOLFOX regimen, respectively. However, in non-KRAS patients, cetuximab improves the therapeutic outcome[69,70]. The same conclusions can be drawn from the results of other large clinical studies: COIN, NORDIC VII or PRIME[71-73]. Yet, the effects of KRAS mutation at codon 12 or 13 on tumour biology were found to differ. In two studies, the survival rate was higher in patients with an uncommon G13D mutation at codon 13 on cetuximab vs patients with other mutations, and similar to patients with no KRAS mutations identified[65,74]. It is presently believed that anti-EGFR antigens should not be used in patients with tumours indicative of G12V mutation of KRAS at codon 12. For bevacizumab, KRAS mutation was found to be of no use as a predictive factor[75].

The same applies to BRAF mutations found in 8%-13% of patients with CRC, which are mutually exclusive with KRAS mutations. The most frequently observed BRAF mutation is V600E mutation. BRAF mutations make the tumour to a large extent resistant to anti-EGFR monoclonal antibodies, and significantly worsen prognosis, especially in patients with MSI-L and MSS[66,70,76-78]. Based on the available data, National Comprehensive Cancer Network (NCCN) suggests considering BRAF mutation testing when KRAS is mutation negative. Interestingly, good prognosis was reported even in those MSI-H CRC patients who had coincident BRAF mutations[78]. In one of studies, the OS period was shown to be slightly longer in patients on cetuximab even if the BRAF mutation was present[69]. Very limited response to vemurafenib, recently approved for metastatic melanoma patients harboring BRAF (V660E) mutation, was demonstrated in CRC patients. Researchers reported that by adding cetuximab strongly synergistic reaction with BRAF inhibitors was observed[79]. NRAS is another member of RAS proto-oncogenes which was found to be rarely mutated, while BRAF is mutually exclusive with KRAS mutations. Since NRAS mutation can predict resistance to EGFR therapy, NCCN suggests considering NRAS mutation testing when KRAS is mutation negative. To date, NRAS mutation does not appear to be associated with the prognosis[80].

Phosphatidylinositide-3-kinases (PI3K) are kinases that promote cellular proliferation. Mutations in PIK3CA gene encoding p110α catalytic subunit of PI3K have been identified in different human solid tumours, including CRC. PIK3CA gene is mutated in 10%-20% of CRC tumours. PIK3CA gene encodes the kinase that regulates, alongside with KRAS, downstream signalling pathways of EGFR. Moreover, PI3K-initiated signalling is inhibited by phosphatase and tensin homologue deleted on chromosome 10 (PTEN). Recent studies have revealed an increase in colon cancer-specific mortality in patients with PIK3CA-mutated tumours, as compared with patients with PIK3CA wild-type tumours[81,82]. However only the coexistence of PIK3CA exon 9 and 20 mutations but not PIK3CA mutation in either exon 9 or 20 alone has been reported to be associated with the worse prognosis[82]. Among patients with KRAS wild-type tumours, the presence of PIK3CA mutation correlated with a significant increase in CRC specific mortality. In contrast, PIK3CA mutation did not significantly affect mortality among patients with KRAS-mutated tumours. Thus, the effect of PIK3CA mutation may be potentially limited to patients with KRAS wild-type tumours[81]. Following the fact that only patients with KRAS-wild type CRC may respond to anti-EGFR antibodies, several studies have investigated the role of PIK3CA mutations on CRC cells response to cetuximab or panitumumab. The data collected so far indicate that CRC with PIK3CA mutations are significantly resistant to anti-EGFR antibodies. When only KRAS wild-type tumours are analyzed, the correlation is even stronger[83-85]. Changes in PIK3 signalling and loss of PTEN expression have been generally linked with the lack of response to EGFR-targeted therapy[86,87]. Recent studies have found that inhibition of cyclooxygenase-2 by regular use of aspirin after CRC diagnosis was associated with longer cancer specific survival time among patients with mutated as opposed to wild-type PIK3CA. The authors conclude that PIK3CA mutations may serve as a predictive biomarker for adjuvant aspirin therapy[88]. Further studies involving KRAS mutated CRC patients are necessary to establish the role of aspirin in PI3K pathway.

Irinotecan is a chemotherapeutic agent that inhibits topoisomerase I, thereby inhibiting replication and stimulating cell apoptosis. Advanced neutropenia and intensive diarrhoea caused by damaged intestinal epithelium are the most common adverse effects of irinotecan, which significantly limit its use. UGT1A1 gene polymorphism is a very useful biomarker of the potential toxicity of irinotecan. It appears that the use of genetic tests is reasonable before treatment initiation with irinotecan to avoid severe adverse effects - mainly neutropenia in women. Genotyping for UGT1A1 can be carried out to select a group of sensitive patients with UGT1A1*28 allele, of whom lower initial doses would be recommended. Hopefully, it will also allow to administer a higher accumulated dose of the drug, divided into smaller portions, to limit its toxicity[89,90]. However, according to a recent meta-analysis, genotyping for UGT1A1 has no predictive value in terms of responsiveness to various doses of irinotecan among patients with CRC[91]. On the other is recommended by ESMO for patients with several toxicity reaction in whom irinotecan in high doses should be used[33]. Furthermore, homozygosity for the UGT1A1*28 has been linked with improved efficacy of FOLFIRI[92].

Since the vascular endothelial growth factor (VEGF) - targeted therapy has been integrated into CRC treatment protocols, some anti-angiogenic drugs have been introduced (bevacizumab, regorafenib, aflibercept). However, a patient selection strategy to identify those patients who benefit most from this therapy has yet to be developed. To date, a predictive biomarker for bevacizumab - the most commonly administered anti-angiogenic drug in CRC therapy - has not yet been identified. Several studies on the identification of predictive biomarkers of bevacizumab have been performed. Jürgensmeier et al[93] evaluated retrospectively, using samples from randomised trial HORIZON III, the prognostic/predictive value of VEGF and soluble VEGF receptor-2. High baseline values of VEGF were associated with worse progression free survival (PFS) and overall survival. These data have revealed that baseline VEGF levels were not predictive of PFS or OS outcome in bevacizumab-treated patients[93]. Other studies have demonstrated that plasma VEGF-A may serve as a prognostic marker, but is unable to predict response to VEGF-targeted therapy in advanced CRC[94,95]. At the same time, KRAS mutation was found to be of no use as a predictive factor for bevacizumab[75].

Ezrin protein, a part of ezrin/radixin/moesin family may play an important role in tumour invasion process. Recent studies has found that overexpression of ezrin protein correlates with CRC aggressiveness, its metastatic potential and worse prognosis. High ezrin expression was also identified as marker of early local recurrence of rectal cancer[96,97]. Although further investigation is needed, ezrin may represent a relevant biomarker and target for personalized anti-metastatic therapies.