Keywords

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Epidemiology, Staging and Biology of CRC

Epidemiology

Colorectal cancer (CRC) is the third commonest cancer in the Western world, with an estimated 142,570 cases diagnosed in the USA in 2010 (SEER database). Worldwide, approximately 1.23 million new cases are diagnosed each year and 608,000 deaths from CRC occurred in 2008 [1]. The lifetime risk for CRC in the Western world is roughly 5–6%. Although familial cases—patients with two or more first- or second-degree relatives with CRC—comprise approximately 20% of disease, high-penetrance Mendelian conditions are responsible for a relatively small proportion (3–4%). Hereditary non-polyposis colorectal cancer (HNPCC), also known as Lynch syndrome, results from germline mutation in mismatch repair proteins [2]. Defective DNA mismatch repair leads to accumulation of mutations with development of CRC typically in the 5th decade of life [2]. Carriers have a risk of 22–69% for CRC development before age 70 [3] and are also at significantly increased risk of tumours of the endometrium, ovary, stomach, uroepithelium, small bowel and bile duct [4]. In familial adenomatous polyposis (FAP), germline mutation of the APC tumour suppressor results in the development of hundreds to thousands of adenomatous polyps within the colorectum, with inevitable malignant change by the 4th to 5th decade of life [2]. Recent large-scale genome-wide association studies have begun to demonstrate the low-penetrance, common genetic variants that underlie many sporadic CRCs [5, 6].

Several dietary factors appear to modify the risk of CRC. An inverse association between folate intake and the incidence of colorectal adenomas and carcinomas has been shown in epidemiological studies [7], but trials of folate supplementation after CRC diagnosis demonstrated no reduction in risk of adenoma or carcinoma recurrence [8, 9]. Evidence also supports a role for calcium supplementation in the reduction of CRC incidence and adenoma recurrence [10–12]. Additional risk factors for CRC include smoking [13, 14] and high BMI [15], while physical exercise [16, 17] and aspirin use [18–20] reduce risk.

Staging

The anatomical and histological progression of CRC is well understood. The classic model is the development of an adenomatous polyp, with malignant conversion leading to local invasion through submucosa, muscularis propria, and, eventually through the outer layers of the colon and into surrounding fat or adjacent structures. Invasion of lymphatic vessels facilitates spread to locally draining mesenteric lymph nodes and access to blood vessels enables spread to distant organs, most commonly the liver. The extent of local invasion, nodal and distant metastases remains the best prognosticators in CRC, as codified in the staging systems used clinically—the AJCC TNM [21] and the Dukes’ systems. These are summarised in Table 9.1 together with the approximate 5-year survival associated with each stage. Further prognostication is provided by additional factors such as bowel obstruction or perforation at presentation, tumour vascular and lymphatic invasion, tumour grade and patient performance status [22]. However, even accounting for these, there remains substantial heterogeneity in outcome within stage groups, indicating additional differences in tumour and patient biology not revealed by these indicators.

Table 9.1 AJCC/Dukes’ staging systems for colorectal cancer and 5-year overall survival
Full size table

Biology

The adenoma–carcinoma sequence of CRC development described above is estimated to take place over 10–20 years. As a result of this stepwise progression, and the technical feasibility of obtaining pathological material from colonic biopsy and tumour resection, the molecular biology of CRC is better understood than that of many cancers. CRC can be divided into two principal pathological categories. The majority (65–70%) of cases are characterised by aneuploidy and chromosomal instability (CIN), while tumours from patients with HNPCC and approximately 15% of sporadic cancers [23, 24] demonstrate an alternative molecular phenotype of microsatellite instability (MSI), and typically retain diploid chromosome complement.

In a seminal paper published over two decades ago, Vogelstein et al. proposed a model of sequential mutations to account for adenoma–carcinoma progression in CIN CRC [25]. In this, inactivating mutations in tumour suppressor genes and activating mutations in oncogenes each confer a proliferative, survival or metastatic advantage to the tumour, enabling progression, invasion and ultimately metastasis. Early lesions were shown to lack APC, a tumour suppressor that encodes a negative regulator of Wnt signalling—a key developmental signalling pathway that promotes cell division [26]. Subsequent events include activating mutations in KRAS, a small GTPase downstream of many receptor tyrosine kinases (RTKs) including EGFR, which leads to constitutive activation of the MAPK pathway, resulting in further mitogenic effects, loss of chromosome 18q-containing SMAD4—a transcriptional regulator with tumour suppressive effects and loss of TP53, which results in resistance to apoptosis. Though this basic model has been validated by subsequent studies [27], it has also been substantially enriched by the demonstration that there are many more mutations and epigenetic changes in CRC than previously realised [28]. A summary of the schematic proposed by Vogelstein et al. and an updated version are shown in Fig. 9.1.

Fig. 9.1
figure 1_9

Original and updated genetic models of adenoma–carcinoma sequence. (a) Model of carcinoma development proposed by Vogelstein and Kinzler. The first event is mutation of the tumour suppressor APC, resulting in constitutive activation of the Wnt pathway and the development of crypt dysplasia and early adenoma. Subsequent activation of the small GTPase KRAS by mutation at codons 12, 13 or 61 leads to activation of the MAPK pathway, while 18q deletion results in loss of SMAD4, a component of the transforming growth factor-β (TGFβ) pathway, leading to adenoma progression. Inactivating mutations in TP53 result in progression to invasive carcinoma. (b) The refined model encompasses subsequent data since the pivotal work of Vogelstein and reflects the demonstration that CRC development can be broadly divided into two subclasses, depending on the underlying pattern of molecular aberrations—chromosomal instability (CIN) (upper figure) and microsatellite instability (MSI) (lower figure). CIN tumours largely develop according to the schematic proposed in (a), though further common mutations have been demonstrated, including inactivating mutations in the ubiquitin ligase FBW7 and activating mutations in PIK3CA—encoding the catalytic subunit of PI3K. MSI tumours result from defective mismatch repair apparatus and are characterised by an alternative spectrum of mutations. Early lesions demonstrate Wnt pathway activation, while mutation in BRAF appears to substitute for KRAS mutation in promotion of adenoma progression. Inactivating mutations in tumour suppressors TGFβR2, IGF2R and PTEN characterise progression to carcinoma. It should be noted that both models are significant oversimplifications—recent work suggests CRC typically contains over ten driver mutations—and that many tumours demonstrate ­features of both

Full size image

MSI results from defective function of the DNA mismatch repair (MMR) proteins MLH1, MSH2, MSH6 and PMS1 MMR leading to slippage and duplication of repetitive microsatellite DNA elements during DNA replication. This results in mutation of genes that contain such repeats, including TGFBR2 [29], IGF2R [30] and BAX [31], in addition to point mutations in other genes such as BRAF [32, 33]. Though this may result from germline mutation (HNPCC, discussed above), the commonest cause in sporadic carcinomas is the downregulation of MLH1 by promoter methylation [34, 35]. Consequently, MSI tumours demonstrate a different mutational spectrum and histological features (proximal to splenic flexure, mucinous, poorly differentiated, lymphocytic invasion) to CIN tumours [36–38]. The potential utility of CIN and MSI status in tumours as prognostic and predictive markers is discussed below.

To summarise, recent data indicate that the simple model of CRC progression associated with predictable genetic changes requires embellishment. Though tumours are characterised by common themes of pathway activation, tumour suppressor loss and acquisition of metastatic ability, there exists a substantial variation between tumours in the actual genes mutated. A summary of the commonly deregulated pathways and mutations is shown in Fig. 9.2. The increased knowledge of the molecular biology of CRC has provided many new therapeutic targets, of which some have been successfully validated, as discussed below.

Fig. 9.2
figure 2_9

Aberrant MAPK and PI3K pathway activation in colorectal cancer results from mutation. The MAPK and PI3K-AKT pathways are essential in cellular homeostasis, and their activation drives colorectal carcinoma (CRC) development via multiple mechanisms. Binding of extracellular ligand to RTKs including the epidermal growth factor receptor (EGFR) typically causes receptor dimerisation and autophosphorylation at tyrosine residues. These serve as docking moieties for adaptor proteins including GRB2 and the regulatory p85 subunit of PI3K (not shown). Though precise mechanisms vary, these typically act to recruit substrates to the plasma membrane where they are activated. In the case of the MAPK pathway, membrane localisation of the small GTPase KRAS by the GRB2-SOS complex causes it to exchange GDP for GTP, converting it from its inactive to active form. Activating KRAS mutations occur in 40% of CRC and abrogates the GTPase activity of KRAS, rendering it constitutively “switched on.” BRAF is a serine–threonine kinase immediately downstream of KRAS and is mutated in 10% of CRC, particularly in MSI tumours. BRAF mutations substitute for KRAS mutation in activating the MAPK pathway, and the two events appear mutually exclusive. Downstream of BRAF, MEK and ERK activation leads to changes in transcriptional regulators such as Fos, Jun and EGR1 with net result of increased cellular proliferation, growth and migration. The PI3K pathway is also activated by RTKs via p85 recruitment, which causes the p110α subunit of PI3K to phosphorylate the membrane lipid phosphatidylinositol-4,5-bisphosphate (PIP2) at the 3′ position to generate phosphatidylinositol-3,4,5-trisphosphate (PIP3). By removal of the 3′ phosphate from PIP3, the tumour suppressor PTEN catalyses the reversal of this reaction. PIP3 recruits the serine–threonine kinase AKT to the plasma membrane, where it activated and proceeds to phosphorylate multiple substrates. Sequential phosphorylation events downstream of AKT include activation of mTOR, a key regulator of cell growth, activation of glycolytic pathways and inactivation of multiple anti-apoptotic proteins. The PI3K pathway is activated in the majority of CRC by multiple mechanisms, including activating mutations in PIK3CA, loss of PTEN and mutation in the ubiquitin ligase FBW7. Approximate mutational frequencies are indicated for pathway intermediates (PTEN is also silenced by promoter methylation). It should be noted that extensive crosstalks and feedbacks exist between the MAPK and PI3K (and additional) pathways, not shown for reasons of space

Full size image

Current Standard of Care and Novel Biotargets

Multimodality Management of Colorectal Cancer

The management of CRC is multimodal and optimum therapy is determined by the tumour stage. Surgery to remove the primary tumour and draining lymph nodes is the only treatment required in stage I and low-risk stage II disease, and resection followed by adjuvant chemotherapy is generally recommended for high-risk stage II and all stage III tumours [39]. Although historically, patients with stage IV disease were managed with palliative chemotherapy alone, an expanding indication for surgery is the removal of hepatic metastases, which permits 5-year survival rates of 20–30% in appropriately staged patients [40, 41]. Radiotherapy, though seldom used in management of colonic tumours, is an essential component of rectal tumour therapy either delivered as a short course prior to surgery or as a longer fractionation schedule combined with chemotherapy [42].

Chemotherapy in CRC

Metastatic CRC

The median survival of stage IV CRC with no treatment is less than 7.5 months [43]. Early clinical trials of chemotherapy in metastatic CRC (mCRC) demonstrated that 5-fluorouracil (FU), combined with leucovorin (LV), which improves fluorouracil (FU) response rate through modulation of thymidylate synthetase activity resulted in the extension of survival by a median of 3.7 months [44]. Irinotecan is an inhibitor of topoisomerase I, an enzyme required for the unwinding of DNA prior to replication or repair, and has activity in FU-refractory CRC. In a pivotal clinical trial for first-line therapy of mCRC, the addition of irinotecan to 5-FU (FOLFIRI) was shown to improve response rate (RR) (39% vs. 21%, P  <  0.001) and extend median overall survival (OS) (14.8 vs. 12.6 months, P  =  0.04) compared to FU alone [45]. Another agent with anti-tumour activity in FU-refractory CRC is oxaliplatin, a third-generation platinum that mediates cytotoxicity though addition of platinum adducts to DNA. Oxaliplatin improved RR (50.7% vs. 22.3%; P  =  0.0001), progression-free survival (PFS) (median, 9.0 vs. 6.2 months; P  =  0.0003) and OS (median, 16.2 vs. 14.7 months; P  =  0.12) when combined with FU (FOLFOX) in first-line therapy for advanced CRC compared to FU alone [46]. The clinical benefit of the addition of oxaliplatin and irinotecan to the therapeutic armamentarium is reflected in the improved overall survival of patients with stage IV CRC treated with all three drugs when compared with patients who were not able to receive such therapy [47]. Capecitabine (Xeloda™) is an oral fluoropyrimidine with at least equivalent efficacy to bolus FU in mCRC as monotherapy [48] or in combination with oxaliplatin (XELOX) [49]. It provides a convenient alternative to intravenous FU and has been widely adopted in clinical practice. Response rates and overall survival have been further improved, albeit modestly by the addition of targeted therapies to the above regimens in the last decade, as described below.

Adjuvant Therapy

The benefit of adjuvant chemotherapy in stage III disease was demonstrated in pivotal clinical trials, which showed that FU treatment reduced the absolute risk of disease relapse by approximately 15% [50]. The activity of irinotecan and oxaliplatin in the metastatic setting led to their evaluation in adjuvant therapy of stage II and III disease. Results were mixed; while the addition of oxaliplatin to FU-based therapy led to significant improvement in disease-free survival (DFS) and OS [51, 52], combination of irinotecan and FU failed to demonstrate superiority over FU alone in several large studies [53, 54]. Though the benefits of adjuvant therapy in stage II disease are smaller, the QUASAR trial demonstrated that FU-based therapy resulted in a 3.6% absolute benefit in survival compared to observation [55]. Subsequent data indicate that the addition of oxaliplatin to FU in stage II disease does not significantly improve this benefit, though the study was not powered for this analysis [51]. As in the metastatic setting, capecitabine can be substituted for FU monotherapy [56] and may be combined with oxaliplatin [57].

Novel Targets in CRC

Vascular Endothelial Growth Factor

Like other solid tumours, CRC is reliant on the development of new vasculature to maintain cellular viability and was therefore a logical target for antiangiogenic therapy. Bevacizumab (Avastin®), is a monoclonal antibody that targets vascular endothelial growth factor (VEGF)—an extracellular ligand that promotes angiogenesis and endothelial cell survival [58]. In a large phase III clinical trial, the addition of bevacizumab to FOLFIRI as first-line therapy of mCRC was associated with substantial improvement in PFS (10.6 vs. 6.2 months, P  <  0.001) and OS (20.3 vs. 15.6 months, P  <  0.001) [59]. Following these results, bevacizumab was widely adopted as part of first-line therapy for mCRC. Subsequent studies combining bevacizumab with FOLFOX or XELOX either as first-line therapy or following FOLFIRI also demonstrated improvements in PFS and OS [60, 61], though the OS improvements of 6–8 weeks were modest compared to the pivotal FOLFIRI trial. Unfortunately, despite efficacy in the metastatic setting, the addition of bevacizumab to FOLFOX for adjuvant therapy in the NSABP C-08 trial did not translate into improved 3-year DFS (77.4% vs.75.5% P  =  NS) [62]. Additionally, the AVANT study, though as yet unpublished, was discontinued early by the data monitoring committee due to lack of efficacy [63].

Epidermal Growth Factor Receptor

The epidermal growth factor receptor (EGFR) is a cell surface receptor tyrosine kinase expressed on epithelial cells, including CRC cells. Activation of EGFR by ligand binding induces a cascade of downstream phosphorylation events through the MAPK and PI3K pathways (Fig. 9.2). The resulting changes in gene transcription, protein localisation and protein activity act in concert to promote cell division, cell growth and cell survival [64]. Cetuximab and panitumumab are monoclonal antibodies that target the EGFR extracellular domain, and both have proven efficacy in advanced CRC. Several RCTs have demonstrated significant benefit from cetuximab or panitumumab treatment either as monotherapy [65, 66] or in combination with cytotoxics [67–69]. Notably, response to therapy in these studies was not correlated with EGFR expression [70]—highlighting the need for alternative biomarkers predictive of response. In light of the encouraging data in advanced disease, anti-EGFR therapy in the adjuvant setting was examined in a recently reported study. Disappointingly, the addition of cetuximab to FOLFOX chemotherapy showed a trend to inferiority to FOLFOX alone [71]. Thus, EGFR targeting cannot be recommended following resection of localised disease. The reasons for these discordant results are presently unclear, though they may reflect disparate biology of micro- and macrometastases.

Other Emerging Targets

PIK3CA, which encodes the catalytic subunit of PI3K, is activated by mutation in 13% of CRC [72]. This results in activation of the PI3K-AKT pathway and increased sensitivity to PI3K inhibition [73]. Several compounds targeting PI3K have been developed and are presently in early-phase clinical trials with promising results [74]. BRAF is a serine–threonine kinase downstream of KRAS and is mutated in approximately 10% of CRC [75, 76]. This mutation—a substitution of glutamic acid for valine at residue 600 (V600E)—results in a constitutively active protein with downstream pathway activation. Recently, PLX4032, an inhibitor of mutated BRAF, has shown remarkable activity in metastatic melanoma [77]. An early trial of PLX4032 in BRAF V600E mutant pre-treated mCRC has demonstrated a more modest benefit (3.7 month PFS), though this remains a target worthy of further validation in the clinic [78]. Insulin-like growth factor receptor-1 (IGF1R) mediates signalling from insulin-like growth factors (IGF1 and IGF2). IGF1R is overexpressed in many solid tumours and is an emerging therapeutic target in breast and ovarian cancers [79]. Unfortunately, a phase II study of an anti-IGF1R monoclonal antibody in cetuximab-/panitumumab-refractory CRC demonstrated no anti-tumour activity [80]. Identification of additional therapeutic targets in CRC is the focus of intensive investigation.

Biomarkers in Clinical Practice

The generally accepted definition of a biomarker is a measurable variable that either varies categorically (present or absent) or continuously (low level to high level). Our definition in this chapter is broad—it may be a single biomarker with prognostic or predictive import or a multivariable prediction model (such as DNA microarray data). Broadly, biomarkers can be divided into three categories:

  1. 1.

    Prognostic biomarkers

    This is a biomarker, the level of which has implications for patient outcome independent of the treatment used.

  1. 2.

    Predictive biomarkers

    These comprise biomarkers’ presence or absence or the relative level of which predicts response to therapy. If therapy is universally used, then a predictive biomarker will also have prognostic import.

  1. 3.

    Biomarkers correlating with treatment response

    This is a measurable substance, the levels of which correlate with response to therapy. The commonest example in CRC is the use of serum carcinoembryonic antigen (CEA) to monitor disease burden in response to palliative chemotherapy.

Though prognostic and predictive biomarkers have been the subject of many published papers, unfortunately, the majority of studies are small retrospective studies with poorly defined protocols for sample collection and analysis. These studies have a high risk of bias and false-positive results. In view of these difficulties, guidelines for assessing the level of evidence (LOE) for biomarkers have been published [81] as summarised below:

Categories that constitute levels of evidence determination for biomarker studies (from Simon et al. [81]):

  1. A

    Prospective, controlled trial designed to address tumour marker. Specimens collected, processed and assayed in real time. Study powered to answer tumour marker question.

  2. B

    Prospective trial not designed to address tumour marker, but incorporating tumour marker utility. Specimens collected, processed and archived prospectively using generic SOPs. Assayed after trial completion. Study powered to answer therapeutic question, but not marker question.

  3. C

    Prospective observational registry, treatment and follow-up not dictated. Specimens collected, processed and archived prospectively using generic SOPs. Not prospectively powered.

  4. D

    No prospective aspect to study. Specimens collected, processed and archived with no SOPs. Not prospectively powered.

Though it is preferable that it is subsequently confirmed, validation of a category A study is not required to obtain Level I evidence, while category B studies require validation to become LOE I and merit incorporation of results into clinical practice. Category B studies without validation and validated category C studies both fall into LOE II, while category C studies without validation merit LOE III. Category C studies are unlikely to change practice in the absence of strong supporting data. Category D studies are very likely to be due to the play of chance, are assigned LOE IV or V, and are hypothesis generating rather than practice changing.

Biomarkers in CRC: An Unmet Need

Despite accurate staging, CRC remains a heterogeneous disease, with 5-year survival varying from 72% to 83% in stage II disease and 44% to 83% in stage III disease [21]. Adjuvant chemotherapy reduces risk of relapse by 3–5% and 12–15% in stage II and III disease, respectively. However, most patients who receive treatment fail to benefit from therapy, while all are exposed to the toxicities and suffer the inconvenience of treatment. Stage II disease, in particular, provides an illustrative example of the need for improved predictors of outcome. Typically, patients with high-risk features (T4 tumour, vascular and lymphatic invasion, high tumour grade) are offered chemotherapy. Even with an optimistic estimation of therapeutic efficacy, 19 of every 20 such patients gain no benefit, as their disease was low risk and cured by surgery alone, or inherently chemotherapy resistant. The availability of accurate prognostic biomarkers would spare the former group chemotherapy, while predictive biomarkers would identify the latter group, enabling either omission of therapy or the use of an alternative regimen. Improved prognostic and predictive markers would also assist in individualising therapy for patients with stage III and IV disease.

A particularly important area is the identification of biomarkers that predict response to targeted therapies. These may be as simple as the presence of the target on or within the cell (e.g. overexpression of the HER2 receptor and benefit from trastuzumab in breast cancer) or the presence or absence of mutations—for instance, detectable aberrations downstream of the agent that interfere with therapeutic effect. The pressing need for these is underlined by the relatively modest clinical benefits associated with targeted therapies and their high cost compared to conventional cytotoxics.

Current and Emerging Biomarkers in CRC

Some biomarkers, such as the CEA tumour marker are long established in clinical practice, while others, such as KRAS mutation testing for prediction of response to anti-EGFR therapy, have entered use more recently. Yet other biomarkers remain unvalidated or demonstrate mixed results in studies. Here, we summarise the current status of established and emerging biomarkers, together with the LOE underlying their use. As previously discussed, distinction must be made between prognostic markers—which predict relapse or progression independent of future treatment effects and predictive markers—which predict response or resistance to a particular therapy [82, 83]. As several prognostic biomarkers also have predictive significance, the two are combined in the following section.

Tumour-Associated Proteins

CEA

CEA is a mucoprotein that is secreted by CRC cells and detectable within the serum. Although the specificity of CEA for CRC is high, its sensitivity is low, and therefore, CEA is not recommended by the American Society of Clinical Oncology (ASCO) or the European Group on Tumour Markers (EGTM) for screening of CRC [84]. However, measurement of CEA prior to resection of primary disease is recommended by ASCO and EGTM as it provides prognostic information independent of other variables [85, 86] and acts as baseline for subsequent assay [84, 87]. CEA levels prior to resection of liver metastases were also shown to be prognostic for risk of relapse in two large case series [88, 89], while a further study demonstrated that CEA of <30 ng/mL before resection of liver metastases was associated with median survival of 34.8 months, while patients with CEA  >  30 had median survival of 22 months [90].

Regular determination of CEA is also recommended in follow-up of patients following resection of CRC [84, 87] as a sensitive and cost-effective indicator of recurrence [91–94]. CEA provides no indication of the likelihood of tumour response to therapy and, therefore, has no value as a predictive marker.

Tumour DNA Repair and Chromosomal Instability

Mismatch Repair/Microsatellite Instability

MSI is defined as instability in at least two of five microsatellite markers within the tumour [95]. An association between MSI and favourable prognosis has been detected in several randomised clinical therapeutic trials (RCTs) [96–98] and confirmed in a meta-analysis comprising 7,642 patients, 1,277 of whom had MSI tumours. This demonstrated conclusively that patients with MSI tumours have better overall survival than those with microsatellite stable (MSS) tumours, with hazard ratio for death of 0.65 (95%; CI 0.59–0.71) [99]. Thus, defective MMR is a confirmed prognostic marker in CRC.

Incorporation of FU metabolites into DNA resulting in FU/G mispairs is one of the mechanisms by which the drug exerts its effects. These can be recognised and repaired by proficient MMR machinery [100, 101]. Of four prospective RCTs comparing adjuvant FU with no chemotherapy, two found that chemotherapy benefit was limited to patients with MSI-L or MSS tumours [102, 103], while two found no difference in outcome by tumour MSI status [96, 104]. Provocatively, both the meta-analysis referred to above and more recent meta-analyses have found no evidence of benefit from adjuvant FU chemotherapy in patients with MSI tumours, though small numbers and methodological heterogeneity limit firm conclusions [99, 105, 106]. However, patients with microsatellite unstable advanced CRC respond to FU chemotherapy [107, 108], and currently, the ability of microsatellite instability to predict FU efficacy is unclear. Further complication is created by the incorporation of oxaliplatin and irinotecan into treatment regimens, as preclinical and clinical data indicate that sensitivity to these agents is unaffected by perturbations in MMR apparatus [109–111]. Interestingly, a recent subgroup analysis of the CALGB 89803 adjuvant trial demonstrated improved outcome in patients with MSI-positive resected stage III CRC treated with FOLFIRI compared to patients with MSS tumours [112]. This difference was not seen in patients who received FU. These interesting results merit prospective investigation.

The Eastern Cooperative Oncology Group (ECOG) E5202 trial (ClinicalTrials.gov Identifier: NCT00217737) selects patients with resected stage II CRC according to MSI status and 18q loss of heterozygosity (LOH) (see below). Patients with MSI tumours without LOH are managed with observation alone. This trial has completed accrual and will provide valuable information of the safety of omission of chemotherapy in this patient group.

The 2006 ASCO panel on colorectal tumour markers concluded that although evidence suggested a favourable prognosis in MSI CRC compared to MSS, current data were insufficient to recommend the use of MSI status as a prognostic or predictive marker [84]. However, based upon the available current evidence, MSI is clearly a prognostic biomarker, though its potential role as a predictive biomarker requires clarification.

Chromosomal Instability

Chromosomal instability (CIN) is usually defined as loss and gain of chromosome complement, or structural changes in chromosomes, and is typically measured by flow cytometry. The 2006 ASCO tumour marker guidelines recommended that tumour aneuploidy could not be recommended for prognostication of CRC [84]. Subsequently, a meta-analysis of 10,126 patients in 63 studies has demonstrated unequivocally that CIN is a poor prognostic factor in CRC [113]. CIN was detected in 60% of CRC and was associated with HR for death of 1.45 (95%; CI of 1.35–1.55, P  <  0.001). The effect was present in both stage II and III disease and independent of adjuvant FU therapy [113].

CIN results in tumour heterogeneity and rapid mutation which could predict for early selection of resistant clones in response to anti-neoplastic therapy [114]. CIN is associated with multidrug resistance in colorectal cell lines, and analysis of informative studies of adjuvant FU is consistent with reduced benefit from therapy in CIN-positive tumours [115]. However, prospective data linking CIN with decreased therapeutic efficacy are currently lacking.

Chromosome 18q Deletion

Deletions of the long arm of chromosome 18 are common in CRC [25] and have been associated with poor prognosis [116]. 18q loss can be detected either by loss of heterozygosity for polymorphic markers [97] or loss of DCC protein by IHC [117]. In addition to DCC, a netrin receptor involved in apoptosis, other candidate tumour suppressors on 18q include the transcription factor SMAD4 and SMAD22. However, the above data are confounded by the association of 18q deletion as a marker of CIN, rather than an independent prognostic variable. 18q loss was used in ECOG E5202 to stratify patients into the poor-risk group who received adjuvant therapy—the results from this study should indicate whether prognosis in patients with stage II disease and retained 18q is sufficiently good to be spared chemotherapy.

Data on 18q loss as predictor of adjuvant chemotherapy benefit are conflicting; studies have suggested that compared with patients with 18q-positive tumours, outcomes are improved [118], no difference [96] or worse [98]. Thus, conclusions regarding the prognostic and predictive role of 18q loss are limited by confounding from CIN and the variable methodology used between studies. 18q loss is not currently recommended for either by the ASCO guidelines [84].

Tumour Apoptosis

P53

TP53 is the most commonly mutated tumour suppressor in human cancer (http://www-p53.iarc.fr/). The incidence of TP53 mutations in CRC is approximately 40–50% [119], and it is thought to be a late event in adenoma to carcinoma transition [120]. Unfortunately, despite over 100 research papers involving over 18,000 patients purporting to determine the prognostic effect of TP53 mutation, no firm conclusions can be drawn from the systematic reviews of the literature [121, 122]. One systematic review showed that patients with abnormal TP53 measured either by immunohistochemistry or mutation were at increased risk of death with relative risk of 1.32 (95%; CI 1.23–1.42) and 1.31 (95%; CI 1.19–1.45), respectively [122]. This study also suggested that the risk associated with abnormal TP53 was greater in patients at lower risk of relapse. However, studies were heterogeneous, and TP53 status is not presently recommended in prognostication by most authorities [84]. The largest meta-analysis showed no effect of TP53 status on benefit from chemotherapy [122].

Tumour Proliferation

Ki67 Proliferative Index

A number of studies have assayed the role of CRC proliferative index in prognostication. Results are mixed; of ten studies using flow cytometry identified by the 2006 ASCO guidelines, five found that the percentage of cells in S phase was an independent prognostic factor, and five did not [84]. Thus, proliferative index in primary CRC cannot be recommended as prognostic at present.

Tumour Signalling Transduction

Like many other malignancies, activation of the MAPK and PI3K-AKT pathways due to mutation of multiple intermediates is near universal in CRC. Consequently, the prognostic and predictive import of these mutations has been the subject of intensive study.

EGFR

EGFR is the cell surface receptor for extracellular ligands from the neuregulin family [64] and is the target of the monoclonal antibodies cetuximab and panitumumab. EGFR is overexpressed in 40–80% of CRC, and overexpression generally appears to correlate with poor prognosis, though as studies are small and retrospective, the potential for publication bias is high [123–128]. Data linking an R497K polymorphism in EGFR—which results in decreased ligand binding and downstream pathway activation—to improved outcome require validation [129].

Interestingly, despite predictions, cetuximab efficacy does not correlate with levels of EGFR expression [66, 67, 130]. While increased EGFR copy number does correlate with response to EGFR targeting [131–134], its predictive ability is not sufficient to recommend its use as a biomarker at present [135]. Additionally, in contradistinction to non-small cell lung cancer, EGFR mutations do not predict anti-EGFR efficacy in CRC [136]. Recent retrospective data suggest that an EGFR A61G polymorphism may correlate with cetuximab efficacy in KRAS wild-type tumours, with increased response rate and overall survival in A/G heterozygote patients compared to homozygotes [137], though these data require confirmation.

EGFR overexpression cannot currently be recommended in prognostication and has no role in predicting cetuximab efficacy. The prognostic and predictive value of EGFR polymorphisms requires prospective validation.

KRAS

Point mutations in KRAS at codons 12, 13 and 61 occur in 30–40% of CRCs and are an early event in the adenoma–carcinoma sequence [75, 138]. These result in abrogation of KRAS GTPase activity with constitutive activation of the protein and downstream MAPK and PI3K-AKT pathways [139].

The role of KRAS mutations in CRC prognosis has been extensively investigated with mixed results. Though a large early study suggested that KRAS mutation is associated with poorer outcome, this finding was restricted to the G12V substitution and stage III disease only [138]. Although a subsequent study has also linked KRAS mutation with poor prognosis [140], several large RCTs have found no correlation with survival [33, 121, 141, 142]. Thus, presently, there is no clear evidence that KRAS mutation is an independent prognostic factor in CRC, though emerging data suggest that it may predict for lung metastases at relapse [143].

KRAS mutations are predicted to cause constitutive activation of downstream pathways irrespective of RTK activation. Consequently, therapeutic approaches targeting RTKs may be futile in the presence of KRAS mutation. Conclusive demonstration that this is the case for the anti-EGFR monoclonal antibodies cetuximab and panitumumab in CRC has been provided by several RCTs [68, 69, 144–147]. In a recent systematic review, anti-EGFR therapy in patients with KRAS mutant tumours was shown to confer no significant benefit in progression-free or overall survival, in contrast to patients with KRAS wild-type tumours, in whom anti-EGFR treatment resulted in significant improvement in both parameters [148]. KRAS mutation is a highly specific negative biomarker of response to anti-EGFR therapy (specificity 93%), though its sensitivity is limited (47%), indicating the existence of additional resistance mechanisms [149]. Consequently, KRAS mutation testing is indicated for all patients being considered for anti-EGFR therapy [150]. Interestingly, a small number of patients with KRAS-mutated tumours respond to anti-EGFR therapies, possibly due to the particular codon or residues substituted [145, 151]. Though this does not affect current guidelines, further research is warranted. KRAS mutation does not predict response to cytotoxic therapy in the absence of anti-EGFR agents [152].

BRAF

BRAF is a serine–threonine kinase directly downstream of KRAS in the MAPK pathway. Mutations in BRAF, almost universally a valine to glutamic acid substitution at residue 600 (V600E), activate the MAPK pathway and are mutually exclusive with KRAS mutation [32, 76]. BRAF mutations are found in 5–10% of CRC overall, but at much higher frequency (40–60%) in MSI tumours [75, 76, 140, 153, 154]. BRAF mutation appears to be associated with shorter survival in the context of advanced disease [76, 140, 154–157], though data on MSI tumours are conflicting [156]. Similar to KRAS, mutant BRAF is associated with lack of response to anti-EGFR therapy, indicating that it may be MAPK pathway activation that confers resistance [75, 158]. BRAF mutation (V600E) testing is likely to become commonplace in clinical practice over the coming years [135]. As discussed previously, although the remarkable results of targeting V600E mutant BRAF in melanoma have not been reproduced in early-phase studies of CRC, further research is ongoing.

NRAS

NRAS is a further member of the RAS family of oncogenes that also functions as a signal transducer downstream of EGFR. NRAS mutations have been detected in 2–5% of CRC [159, 160]. Emerging data indicate that NRAS mutation confers resistance to anti-EGFR therapy, though patient numbers are small [75].

PIK3CA

PIK3CA encodes the p100α catalytic subunit of PI3K [161]. Mutations typically occur at three hotspots (two in exon 9, one in exon 20) and render the protein constitutively active, resulting in activation of the PI3K-AKT pathway. PIK3CA mutations are found in 13–15% of CRC [72, 75, 162]. Data on the prognostic significance of PIK3CA mutation are limited, though a recent prospective study showed an association of mutation with poor prognosis in patients who had undergone curative resection. Interestingly, this effect was stronger in patients with KRAS wild-type tumours [162]. These interesting data require further analysis.

An emerging body of literature indicates that PIK3CA mutation in the context of wild-type KRAS and BRAF is predictive of lack of response to anti-EGFR therapy. However, data are not uniform and the relationship is not as strong as for KRAS or BRAF mutation [75, 163, 164]. Interestingly, in one study, only mutations in exon 20 predicted resistance to anti-EGFR therapy [75].

PTEN

PTEN is a lipid phosphatase and key negative regulator of the PI3K pathway. PTEN is mutated or epigenetically silenced in approximately 20% of CRC [155]. Data regarding the significance of PTEN as prognostic marker are conflicting [144, 155]. Though results are not universal, loss of PTEN by IHC appears to predict lack of benefit from anti-EGFR therapy [144, 164, 165]. Unfortunately, PTEN IHC is challenging, and standardisation of methodology is likely to be required for prospective studies in order to draw firm conclusions.

IGF2

IGF2 is an imprinted, paternally expressed growth factor which signals via IGF1R to activate the PI3K-AKT and MAPK pathways [166]. Regulation of IGF2 expression is complex and involves an imprinting control region (ICR) and additional differentially methylated regions (DMRs) within the gene locus [167]. Loss of imprinting (LOI) of IGF2 with biallelic expression is common in CRC and is detectable in the normal adjacent mucosa [168, 169]. A recent study has shown that IGF2 DMR0 hypomethylation correlates with LOI and also demonstrated that DMR0 hypomethylation was associated with higher mortality in 1,033 patients in two prospective cohort studies [170]. These interesting results require validation in a prospective RCT.

Pharmacogenetics and Therapeutic Efficacy

The three principal cytotoxic drugs used in the management of CRC are all metabolised with differing efficiencies depending on germline polymorphisms present in the population. As these result in varying exposure of patients to active drug, they may lead to differences in therapeutic efficacy and toxicity. As it has been in clinical practice for the longest duration, most of the available data pertain to FU, the metabolism of which is complex and summarised in Fig. 9.3. Most information on targeted therapies has focused on tumour resistance mechanisms, though emerging studies suggest that germline variants may also contribute to response.

Fig. 9.3
figure 3_9

5-Fluorouracil metabolism and variants. 5-Fluorouracil (FU) is converted into three principal active metabolites: fluorodeoxyuridine monophosphate (FdUMP), fluorodeoxyuridine triphosphate (FdUTP) and fluorouridine triphosphate (FUTP). Inhibition of thymidylate synthetase (TS) by FdUMP is an important mechanism of FU action and is enhanced by the presence of the reduced folate 5,10-methylenetetrahydrofolate.High intratumoral concentrations of TS are associated with poor prognosis and decreased benefit from FU. Incorporation of FdUTP into DNA also contributes to cytotoxicity, though may require an intact mismatch repair system to detect the modified base. Incorporation of FUTP into RNA disrupts normal processing and function and leads to toxicity through several mechanisms. The main mechanism of FU catabolism is by conversion to dihydrofluorouracil (DHFU) via the action of dihydropyrimidine dehydrogenase (DPD). Germline variants in the DPD activity are associated with FU toxicity. dTMP deoxythymidine monophosphate, dTTP deoxythymidine triphosphate, dUMP deoxyuridine monophosphate, FdUDP fluorodeoxyuridine diphosphate, FUDP fluorouridine diphosphate, FUdR fluorodeoxyuridine, FUMP fluorouridine monophosphate, FUR 5-fluorouridine, OPRT uridine monophosphate synthetase, RR ribonucleotide reductase, TK thymidine kinase, TP thymidine phosphorylase, UK uridine kinase, UP uridine phosphorylase

Full size image

Thymidylate Synthase

Thymidylate synthetase (TS) catalyses the formation of thymidylate, required for DNA replication and repair, and is thought to be the principal target for the main active metabolite of FU, fluorodeoxyuridine monophosphate (5-fdUMP). A systematic review and meta-analysis showed poorer overall survival patients with high tumour TS expression in both the adjuvant and metastatic settings. Interestingly, the prognostic effect of high TS was greatest in patients who did not receive chemotherapy [171]. TS expression is determined by germline polymorphisms—a variable tandem repeat in the promoter and a 6 base pair (6 bp) insertion and deletion polymorphism in the 3′ UTR. Approximately one quarter of the Caucasian population is homozygous for a double repeat (2R/2R) in the promoter, one quarter homozygous for a triple repeat (3R/3R) and half are heterozygous [172]. Although the 3R/3R repeat does not affect the level of TS mRNA, it is associated with significantly higher levels of TS protein [173]. The 6-bp deletion polymorphism in the 3′ UTR decreases mRNA stability and results in lower intratumoral TS expression [174]. Both preclinical and clinical evidence indicate that high TS correlates with decreased benefit from FU [171, 172, 175, 176].

Methyltetrahydrofolate Reductase

Methyltetrahydrofolate reductase (MTHFR) catalyses the irreversible conversion of 5-10-methylenetetrahydrofolate (5,10-mTHF) to 5-methyltetrahydrofolate (5-mTHF), thus reducing levels of the former, an essential cofactor in the conversion of deoxyuridine monophosphate to deoxythymidine monophosphate by TS. Two polymorphisms within MTHFR have been demonstrated to have functional significance: a C677T polymorphism associated with reduced enzymatic activity [177] and increased sensitivity to FU [178] and to a lesser degree an A1298C polymorphism. Though initial data suggested that MTHFR 677T polymorphism was associated increased sensitivity to FU [179, 180], other studies have detected no difference in outcome [181, 182]. Interestingly, a recent study suggests that favourable MTHFR polymorphisms predict response to FOLFOX in advanced CRC [183]. Both polymorphisms appear to predict capecitabine and FOLFOX toxicity [184, 185].

Thymidine Phosphorylase

Thymidine phosphorylase (TP) catalyses the conversion of FU to the active metabolite FUdR. Data on its role as a biomarker are conflicting, however, as studies have suggested that high TP expression may either decrease [186] or not affect FU activation [187]. Further complication is afforded by the additional function of TP in promotion of angiogenesis—reflected in its alternative name, platelet-derived endothelial cell growth factor [188–191]. Though higher levels of TP in tumours have been associated with more extensive angiogenesis [192] and poorer outcome [191, 193, 194], firm conclusions cannot be drawn at present.

Dihydropyrimidine Dehydrogenase

Dihydropyrimidine dehydrogenase (DPD) is the principal enzyme in FU catabolism [195]. Levels of DPD in the population vary—3–5% of people are partially and 0.2% are completely DPD deficient. DPD deficiency results in the accumulation of active drug and is associated with FU toxicity [196]. However, the association of over 30 polymorphisms with DPD deficiency precludes screening for this prior to FU therapy [197]. Several studies have shown an inverse correlation between tumour DPD expression and survival following adjuvant FU treatment for CRC [198–200], while another study has shown that high FU clearance predicts poorer outcome in this context [201]. However, this requires confirmation.

Oxaliplatin Sensitivity

Reduced sensitivity to oxaliplatin has been linked to decreased tumour penetration, increased detoxification and increased removal of platinum DNA adducts by proficient repair mechanisms. Glutathione S-transferases are a family of enzymes that target drugs for excretion by conjugation with glutathione. GSTP1 targets platinum derivatives, including oxaliplatin for this process. Polymorphisms in GSTP1—I105V and A114V—result in decreased activity and have been associated with improved outcome and increased neuropathy after oxaliplatin treatment [202–205], though other studies have shown no correlation with either toxicity or outcome [206, 207]. Enhanced removal of platinum DNA adducts by nucleotide excision repair (NER) machinery would be predicted to result in decreased therapeutic efficacy. A recent systematic review of 17 published studies comprising 1,787 patients showed that polymorphisms in two NER genes, ERCC1 C11615T and ERCC2 T13181G, were predictive of substantial reduction in oxaliplatin effect (HR for survival 2.03 and 1.42, respectively) [208]. These interesting results require prospective validation.

Irinotecan Sensitivity

The active metabolite of irinotecan, SN38 is conjugated and detoxified by UDP-glucuronosyl­transferase (UGT1A1). The UGT1A1 promoter is polymorphic, with variation in the number of repeats of a TATA element, with increasing repeat number associated with decrease in enzyme activity—homozygosity for the 7-repeat allele, referred to as UGT1A1*28, is associated with significantly increased risk of toxicity from irinotecan, particularly at higher doses [205, 209, 210]. Though testing for the UGT1A1*28 polymorphism was approved by the FDA in patients prior to irinotecan therapy, use has been patchy, possibly as a result of the decreased toxicity associated with lower irinotecan doses in combination regimens.

Determinants of Response/Toxicity to Targeted Therapies

Activation of antibody-dependent cell-mediated cytotoxicity (ADCC) contributes substantially to the therapeutic effect of trastuzumab and rituximab [211]. FcγR polymorphisms modify the killing function of effector immune cells and are associated with tumour response in patients treated with these agents. Although it appears clear that the predominant mechanisms of resistance to anti-EGFR monoclonal antibodies are tumour intrinsic, small studies have shown that FcγR polymorphisms (FCGR2A-H131R and FCGR3A-V158F) are associated with cetuximab response [212, 213]. Although these interesting results suggest that ADCC contributes to cetuximab efficacy, they require prospective confirmation.

Emerging Platforms in Biomarker Analysis

Though yet to fully translate into advances in the clinic, high-throughput arrays and other emerging technologies provide hugely powerful platforms for analysis of tumours and facilitate a shift from hypothesis-driven research to unbiased interrogation of the whole genome, transcriptome and proteome (Fig. 9.4) [214–217]. With the rapidly decreasing cost and increasing capacity of next generation sequencing, it is likely in the not too distant future both mutational and expression analysis of tumours will be feasible on an individual patient basis [218]. The enormous amounts of data generated by these technologies pose significant logistical and statistical challenges to analysis and require careful experimental design for accurate biomarker identification and validation.

Fig. 9.4
figure 4_9

High-throughput methodologies in colorectal cancer. Emerging platforms in the diagnosis and investigation of CRC include assays of change in genetic sequence, gene expression and protein expression. Most advanced in clinical development are gene-expression signatures (Oncotype DX®, ColoPrint®), which use either RT-qPCR or microarray technology to assay the relative expression of hundreds to thousands of genes within tumour samples. By generation of prognostic and predictive signatures, these seek to enable individualised therapy by stratification of patients according to risk of relapse and likelihood of benefit from therapy. The rapid progress in sequencing technology is likely to permit the mutational analysis of tumours in the next few years, while RNAseq combines expression analysis with mutational detection of the transcriptome. Genome-wide association studies (GWAS) have been used to identify single nucleotide polymorphisms (SNPs) that confer increased risk of developing CRC and, more recently, to predict outcome. Proteomics provides several powerful platforms for tumour analysis, including high-throughput immunohistochemistry (IHC) via tissue microarrays (TMA) and mass-spectrometric approaches

Full size image

Gene-Expression Signatures

The first published gene signature in CRC was published in 2004 [219]. Based on analysis of 31 relapses in 71 patients, Wang et al. proposed a 23-gene prognostic signature as identifying patients likely to develop recurrent disease. However, validation of this set showed that this performed little better than chance (67% positive predictive value), and the same group demonstrated in a larger cohort that a reduced signature of seven genes performed better in an enlarged patient cohort [220]. Further small studies have also generated prognostic signatures in stage II and III CRC [221, 222]. Escherisch et al. analysed adenomas and stage II–IV CRC by microarray. They generated a 43-gene signature that was superior to TNM staging in prediction 36-month overall survival and validated this in an ­independent cohort of 95 patients [223]. However, again, this awaits large-scale validation.

Recently, O’Connell et al. have published a recurrence score generated from 1,851 formalin-fixed, paraffin-embedded tumour samples from patients with stage II/III CRC enrolled in NSABP adjuvant trials C-01/C-02/C-04/C-06 and a cohort of untreated patients from the Cleveland Clinic. They performed reverse transcriptase quantitative PCR (RT-qPCR) for 761 candidate genes and found 48 genes significantly associated with recurrence and 66 predictive of FU benefit [224]. From these, they selected seven recurrence genes, six FU-benefit genes and five internal reference genes, and validated them in 1,436 patients with stage II CRC from the QUASAR trial. This 12-gene recurrence signature (commercially available as Oncotype DX®) predicted recurrence risk (P  =  0.004) and retained significance (P  =  0.008) in multivariate analysis independent of MSI status, T-stage, tumour grade and lymphovascular invasion [225]. This recurrence score has also been shown to be prognostic in stage III disease [226]. Another group analysed fresh frozen tissue from 188 patients with stage I–IV CRC by microarray and generated an 18-gene signature (ColoPrint®). This was then validated in an independent dataset or 206 tumours. Classification of patients into low and high risk by the signature correlated with 5-year relapse-free survival of 87.6 and 67.2%, with hazard ratio of 2.5 (95% CI, 1.33–4.73; P  =  0.005), and the signature retained significance in multivariate analysis (HR  =  3.34; P  =  0.017) [227]. The PARSC trial seeks to compare the ColoPrint® signature with current prognostic factors in prediction of relapse in patients with resected stage II CRC and is currently recruiting [228]. However, practically speaking, the collection and storage of good quality fresh frozen tissue that is viable for this screening assessment is a huge challenge. A 33-gene signature has been generated from the NSABP C-07 trial using Illumina arrays and division of the study population into equally sized training and validation subsets. Classification of patients into low- and high-risk groups by the index predicated recurrence at 5 years (82.6% disease free vs. 64.3% disease free, P  <  0.001). Notably, the authors demonstrated that although the relative benefit from the addition of oxaliplatin to FU was similar in both risk categories, the absolute risk reduction from combination therapy was small enough in the low-risk group to be of questionable benefit [229]. These interesting results await further validation.

High-throughput platforms have substantial promise in the identification of patients likely to benefit from individual therapies, though this remains challenging due to the large sample sizes required in studies. In addition to the 6-gene FU benefit signature predictor referred to above, small studies have attempted to generate signatures predictive of therapeutic response. A small study generated a 14-gene signature predictive of FOLFIRI response in mCRC by microarray profiling of snap-frozen tissue [230]. Though this identified all responding patients, the study lacked an independent validation set, and, therefore, requires confirmation. Microarray analysis has also been used to predict response to cetuximab [231], and interestingly, the response signature overlaps significantly with the KRAS mutation signature [232].

Genome-Wide Association Studies

SNP array-based genome-wide association studies (GWAS) have proven a powerful platform for the discovery of CRC susceptibility loci [5, 6]. Functional SNPs in genes encoding kinases and cell cycle-associated proteins may be predicted to alter tumour phenotype and thus outcome. Though large numbers are required to gain sufficient power to detect modifiers, one study in CRC has been reported in abstract form. In this, samples from 947 patients with stage II/III CRC were typed at 309,200 SNPs and analysed by Cox regression. Thirty-three SNPs were further analysed in three additional patient cohorts comprising 2,213 patients. In the final meta-analysis, one SNP met the significance level. This polymorphism, located near a gene that regulates cell motility and invasion conferred an HR for relapse of 1.46 (95%; CI 1.10–1.94) [233]. Based upon this analysis, it is unlikely that SNPs conferring HR  >  2.0 exist, though further studies may define lower risk loci.

Proteomics

Proteomics provides a powerful platform for the simultaneous analysis of multiple tumour-associated proteins. Mass spectrometry may be performed on serum or primary tumour tissue, and antibody-based approaches include IHC and reverse-phase protein array (RPPA). With the exception of IHC, methodologies are in a relatively early stage of development. Small studies using mass spectrometry have demonstrated characteristic tumour-associated and serum signatures [234, 235] that have promise in early CRC diagnosis. However, these require validation in large prospective studies. The advent of tissue microarrays (TMA) and automated image analysis enables high-throughput analysis of tumour protein expression [236]. These technologies have been utilised in CRC to demonstrate that intratumoral T-cell infiltration is associated with favourable prognosis, with superior prognostic value than TNM staging [237, 238]. The same group have recently published an additional retrospective study confirming these results [239], and this finding merits prospective confirmation. RPPA enables evaluation of pathway activation by quantification of phosphoproteins within tumours and has been utilised in CRC in small studies [240, 241].

While incorporation of proteomic analysis of tumours into routine clinical practice is unlikely to occur within the next few years, the technology has substantial promise to contribute to CRC detection and management in the future.

Conclusions

The duration for which validated biomarkers have been clinically utilised in CRC varies from greater than three decades in the case of CEA to less than 5 years for KRAS mutation testing. During this period, myriad other biomarkers have been postulated in small, retrospective studies, without confirmation in larger retrospective or prospective cohorts. This emphasises the requirement for evaluation of biomarkers in well-designed prospective clinical trials, where the potential for bias is minimised. Several approaches have been postulated to incorporate biomarker studies into clinical trial design in order to expedite biomarker validation and adoption into clinical practice [242]. A summary of the LOE supporting the use of selected established and emerging biomarkers is presented in Table 9.2. Currently, the evidence supports the use of only three of the many biomarkers discussed above in routine CRC clinical practice. A further two categories of biomarker are promising, but require additional validation before adoption into patient care (Table 9.3, Fig. 9.5).

Table 9.2 Levels of evidence supporting biomarker studies
Full size table
Table 9.3 Biomarkers supported by current clinical evidence and likely to be validated in the near future
Full size table
Fig. 9.5
figure 5_9

Recommendations for biomarker-informed management of CRC. Evidence-based algorithm incorporating biomarkers for management of CRC. Patients with stage I disease have less than 10% chance of relapse and require no further therapy. Patients with stage II disease have variable prognosis; T4 stage of the primary tumour is the most significant pathological factor, and patients with T4 disease should be offered adjuvant chemotherapy if appropriate. Patients with MSI-high T3 tumours without nodal involvement have good prognosis, appear less likely to benefit from adjuvant chemotherapy and should be observed only (level I evidence). MSS T3 tumours should be stratified according to further pathological criteria—grade, vascular invasion—and are good candidates for gene-expression analysis of tumour to determine prognosis and inform therapy (level II evidence). All patients at significant risk of relapse (excluding stage I disease and low-risk stage II) require regular CEA measurement as part of follow-up, with the exception of patients who are not candidates for therapy at relapse (level I evidence). All patients with stage IV disease being considered for anti-EGFR therapy require KRAS mutation testing on tumours, with therapy not indicated if KRAS mutant (level I evidence)

Full size image

Furthermore, the significance of individual CRC biomarkers has been reappraised and refined in light of increased understanding of tumour biology over recent years. For example, 18q deletion appears to be a surrogate for CIN rather than an independent prognostic factor, while the precise prognostic significance of BRAF V600E mutation requires clarification given its association with microsatellite instability. It is plausible that ­additional markers will be discovered to co-segregate when analysed in toto.

Building on the bedrock outlined above, high-throughput technologies promise to enhance biomarker discovery and validation and have the potential to enable truly individualised therapies for patients with CRC. Perhaps the greatest challenge will be the integration of data from multiple analyses and platforms—germline and tumour genomic, proteomic, immunologic—to facilitate this. High-quality collaborations between basic scientists and clinicians and well-designed clinical trials are essential over the coming years if we are to achieve this aim.