Current Colorectal Cancer Reports

, Volume 7, Issue 3, pp 227–240

Biological Markers in Patients with Early-Stage Colon Cancer: Consensus and Controversies

Authors

    • Department of Medical OncologyRoyal Melbourne Hospital
    • Ludwig Colorectal Cancer InitiativeLudwig Institute for Cancer Research Melbourne - Parkville Branch
    • Division of Cancer MedicinePeter MacCallum Cancer Centre, St. Andrews Place
    • Department of MedicineUniversity of Melbourne
Article

DOI: 10.1007/s11888-011-0102-5

Cite this article as:
Field, K.M. & Zalcberg, J.R. Curr Colorectal Cancer Rep (2011) 7: 227. doi:10.1007/s11888-011-0102-5
  • 53 Views

Abstract

Colorectal cancer is one of the most common malignancies in the Westernized world. Particularly in the setting of metastatic disease, advances in the use of biomarkers have resulted in “personalized therapy” being an aspect of routine care. However, their role is less clear-cut in the context of adjuvant treatment for early-stage colon cancer. We still largely rely on conventional TNM staging to determine prognosis and treatment options in early-stage disease. Ongoing research is elucidating alternative measures, in the form of prognostic and predictive biomarkers, which may better assist in understanding the wide spectrum of outcomes in early-stage disease; as well as the potential to select and tailor appropriate therapeutic options for an individual rather than simply treating a disease and a “cancer stage”. This review outlines key discoveries and controversies in the application of biomarkers to tailor treatment in early-stage colon cancer.

Keywords

Colorectal cancerBiomarkersK-rasB-rafPrognosticPredictiveMicrosatellite instabilityMismatch repairColon cancerAdjuvantChemotherapyChromosomal instabilityCIMP

Introduction

The 21st century is now a decade old, and with it we have seen striking developments in the understanding of tumor biology. Targeted therapies in cancer medicine have become routine for some malignancies as our understanding of molecular pathways and their significance in oncogenesis and proliferation has continued to grow. The “holy grail” for oncology practice is an individualized approach to cancer management, recognizing that there may be many different subtypes of a particular malignancy. This is critically important to understand, given that defined prognostic and predictive factors can not only help to estimate a patient’s outcome and survival, but also select the most appropriate therapy for that person. Ideally, treatment should be selected that is most likely to benefit an individual, while avoiding unnecessary harm in another who is unlikely to derive benefit from that same therapy.

Despite an exponential learning curve in the understanding of the importance of biomarkers in the last decade, there is perhaps more controversy than consensus in their application for early-stage colon cancer, a situation in which biomarkers as prognostic and predictive factors could be the most useful. In this context, reliably identifying those patients most likely to derive benefit from adjuvant treatment (and selecting the most appropriate therapy from what is now a range of options) while sparing those who are unlikely to require adjuvant therapy from unnecessary toxicity is synonymous with optimal cancer management. However, despite significant progress in the understanding and application of biomarkers to routine clinical practice in metastatic colorectal cancer and in other malignancies, their use and significance in the adjuvant disease setting is clouded by uncertainty and there remain many unanswered questions. To date, the American Joint Committee on Cancer (AJCC) TNM (tumour-node-metastasis) staging system which clearly separates stage II and III colon cancer remains the main determinant of whether a patient is offered adjuvant therapy, and there is still considerable dissent in the literature regarding the degree of benefit that patients with stage II disease obtain from adjuvant chemotherapy. This is a situation where ongoing translational research to better understand and develop reliable biomarkers is crucial.

Stage II and III Colon Cancer: An Overlapping Spectrum of Outcomes

The conventional stratification of patients with early-stage colon cancer is the AJCC/TNM staging system. This is the current “gold standard” by which risk of recurrence or future development of metastatic disease is determined in the clinic, and the principal determinant of whether or not adjuvant chemotherapy is offered. Together with an assessment of the patient’s clinical status, age, and comorbidities, and pathologic features such as lymph node yield, differentiation grade, lymphovascular and perineural invasion, this allows for a decision regarding the role of adjuvant chemotherapy to be made.

In general, patients with stage III (node-positive) disease have a higher risk of recurrence than many patients with stage II (node-negative) disease. Nodal status is an independent prognostic factor for both disease-free survival and overall survival [1]. However, some stage II cancers present with “high-risk” features which render the risk of recurrence higher. A patient with a T4N0 tumor, for example, has a worse disease-free and overall survival than a patient with a T3N1 tumor [1], despite being classified as stage II. The current 7th edition of the AJCC staging system [2], available since late 2009, recognizes this in strategic changes that have been made in colorectal cancer staging compared with the earlier 6th edition, and reflects these emerging new data in considering current survival expectations. Substaging of stage II and III disease has been delineated, as well as subdividing T4 disease and further subcategorizing nodal disease [3].

Even considering stage-specific treatment options, the dilemma when trying to understand the benefits, or otherwise, of adjuvant chemotherapy for stage II disease lies in the conflicting literature surrounding this topic. The benefit from chemotherapy in stage II disease is generally quoted to be an approximately 3%–5% improvement in overall survival; however, the actual benefit differs considerably between trials and analyses. A recently published pooled analysis from the National Surgical Adjuvant Breast and Bowel Project (NSABP) for stage II and III colon cancer (C-01 to C-05), comparing outcomes from those who received post-operative 5FU/LV to those receiving surgery for stage II disease, concluded that those with stage II disease had significantly improved OS with adjuvant treatment (HR 0.58, 95% CI 0.48–0.71, P < 0.0001) [4]. This striking improvement is in contrast with other trials which report a considerably smaller benefit for adjuvant chemotherapy in this setting. The pooling of five trials meant that patients were not all randomized to chemotherapy versus surgery alone and this diversity in the control versus treatment arms is in contrast with randomized phase 3 adjuvant studies such as QUASAR, PETACC-3, and individual NSABP trials, where more reliable comparison between treatment arms can be made (Table 1). Additionally, differences between trials in terms of the degree of benefit of chemotherapy in stage II disease may arise from the differential inclusion of patients with high-risk features (T4 disease, perforated tumor, lymphovascular or perineural invasion).
Table 1

Benefit of chemotherapy in stage II colon cancer: key trials and reviews

Author, year

Study

Description

Findings for stage II

Statistically significant positive studies (stage II analyses only)a

 Figueredo et al. [68•], 2008

Cochrane review

Review of randomized trials or meta-analysis including stage II patients: adjuvant therapy vs surgery alone

Chemo vs surgery alone:

OS: Pooled relative risk ratio: 0.95 (0.88–1.05)

33 trials and 17 meta-analyses included

DFS: Pooled relative risk ratio: 0.83 (0.75–0.92)

 Wilkinson et el. [4•], 2010

Pooled analysis of NSABP trials

Analysis across 5 trials (NSABP C-01-C05)

Chemo vs surgery alone:

DFS: HR 0.68 (0.57–0.81), P < 0.0001

Comparison of surgery alone vs surgery + 5-FU/LV

OS: HR 0.58 (0.48–0.71), P < 0.0001

Stage II and III

Recurrence-free interval:

Stage II n = 1255

HR = 0.70 (0.54–0.91), P < 0.0068

 Gill et al. [1], 2004

Pooled analysis

Pooled data from 7 randomized trials of 5-FU–based therapy vs surgery alone

5-year DFS (chemo vs surgery alone):

Stage II and III

76% vs 72% (P = 0.0490)

Stage II n = 1440

5-year OS:

81% vs 80% (P = 0.1127)

 Gray et al. [69], 2007

QUASAR

Randomized phase 3

Relative risk of death (chemo vs surgery alone):

5-FU/LV vs surgery alone

0.82 (0.70–0.95), P = 0.008

n = 3239, all stage II

Relative risk of recurrence:

(2291 colon, 948 rectal)

0.78 (0.67–0.91), P = 0.0001

Absolute improvement in survival = 3.6%

Nonstatistically significant studies (stage II analyses only)a

 Sargent et al. [78•], 2009

ACCENT dataset

Analysis of dataset from 18 phase 3 adjuvant trials

8-y OS (chemo vs surgery):

Multiple different treatments vs surgery alone

72.2% vs 66.8%, P = 0.076

Stage II and III

Stage II n = 6896

 IMPACT investigators [70], 1999

IMPACT B2

Analysis from 5 separate trials

5-y event-free survival (chemo vs surgery):

1016 patients with stage B2 colon cancer

76% vs 73%

5-FU/LV vs surgery alone

HR 0.83 (90% CI 0.68–1.01), P = 0.061

5-year OS:

82% vs 80%

HR 0.81 (90% CI 0.64–1.01), P = 0.057

 Kuebler et al. [54], 2007

NSABP C-07

Randomized phase 3

4-y DFS (5-FU/LV vs FLOX):

Bolus 5-FU/LV vs FLOX

81% vs 84.2% (3.2% absolute increase)

Stage II and III

 Van Cutsem et al. [10], 2009

PETACC-3

Randomized phase 3

DFS (LV5FU2 vs FOLFIRI)

LV5FU2 vs FOLFIRI

3-y 82.5% vs 84.6%

Stage II and III

5-y 76.9% vs 80.9%

(stage II = 934)

HR 0.81 (0.61–1.08), P = 0.158

OS:

3-y 93.5% vs 95.1%

5-y 88.8% vs 90.0%, P = 0.344

 Allegra et al. [71], 2010

NSABP C-08

Randomized phase 3

3-y DFS:

mFOLFOX6 ± bevacizumab

87.4% vs 84.7%

Stage II and III

HR 0.82 (0.54–12.5), P = 0.35

(stage II = 666)

 Figueredo et al. [72], 2004

Systematic review (Cancer Care Ontario)

37 trials and 11 meta-analyses

Mortality risk ratio:

Adjuvant therapy vs observation

HR 0.87 (0.75–1.01), P = 0.07

aIncludes subgroup analyses from trials that have involved stage II and III patients (may not have been powered for stage II analyses alone)

Available level 1 evidence strongly supports the administration of adjuvant chemotherapy to improve disease-free and overall survival in stage III colon cancer. However, for both stage II and III disease, regardless of the ongoing uncertainty over the level of benefit for stage II disease as a whole, there will be some patients who receive chemotherapy who did not need it—that is, their tumor was cured by surgery alone; and some who receive chemotherapy but still develop recurrence or metastatic disease. In between this is the relatively smaller group for whom chemotherapy actually contributes to the patient’s cure or long-term disease-free survival. To be able to delineate these three groups and to better estimate in which of these groups an individual patient best fits, clearly requires more than the current TNM staging system which fails to reflect the innate heterogeneity of these diseases.

The ultimate aim of “personalized medicine” is to tailor treatment to an individual based not just on age or pathologic grouping, but by ever more selective molecular characterization of their tumor type and likelihood of response to a particular treatment. We need to become more sophisticated in our ability to assess a patient’s individual risk of developing disease recurrence, or of benefitting from an available treatment. The most likely way to achieve this is to describe early-stage colon cancer by its molecular phenotype rather than its disease stage.

What is a Biomarker?

A biomarker is a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to therapeutic intervention [5]. Hence, the use of a “biomarker” in colorectal cancer (CRC) can be applied broadly, ranging from molecular and genetic tests (such as K-ras), to routine blood tests (such as CEA or serum albumin), and clinical status (such as node positivity) (Fig. 1). Biomarkers can be either prognostic or predictive. A prognostic factor can be used to estimate a patient’s prognosis by determining the likely behavior of the disease, irrespective of treatment given, and reflects the natural history of the cancer. A predictive factor can estimate the likelihood of response to a certain treatment or therapy.
https://static-content.springer.com/image/art%3A10.1007%2Fs11888-011-0102-5/MediaObjects/11888_2011_102_Fig1_HTML.gif
Fig. 1

Examples of prognostic and predictive biomarkers of interest in early-stage colon cancer

Despite increasing interest in the applicability of biomarkers in understanding disease behavior and guiding clinical practice, only a small number are routinely used in guiding practice. This is at least partly for two reasons: firstly, validating and standardizing the testing procedure and determining the reliability of the biomarker requires significant time, effort, and large sample sizes; and secondly, the test must be cost-effective to have applicability in routine practice.

Molecular Biomarkers—The Way Forward?

There are many candidate molecular prognostic and predictive biomarkers that are of relevance in CRC (Table 2). A number of the candidate markers arise from the concept of the adenoma–carcinoma sequence, in which sequential genetic events are thought to eventually lead to the development of malignant disease. Although complex and inter-related, a simplified categorization of the adenoma–carcinoma sequence is as follows. There are two predominant pathways leading to sporadic colorectal carcinogenesis: the microsatellite instability (MSI) pathway (in around 15%) and the chromosomal instability (CIN) pathway (in up to 85%). Additionally, the CpG island methylator phenotype (CIMP), an epigenetic phenomenon, is thought to be another subset or pathway of CRC carcinogenesis, closely related to MSI. There are associated genetic mutations in oncogenes and tumor suppressor genes that may occur together with these pathway abnormalities. Tumors which arise from these pathways may have differing prognosis and/or sensitivities to particular anti-cancer agents, based on the genetic imbalances that govern their biology. There may be overlap between these pathways, with varying impact on their use as prognostic (or predictive) biomarkers. As we develop further understanding into the intricacies of these molecular subtypes, considering one biomarker in isolation will be considerably less useful than appreciating the “bigger” picture—which will allow the complexity and heterogeneity of any single tumor to be adequately considered.
Table 2

Prognostic and predictive biomarkers in early-stage colon cancer

Biomarker

Discussion

References

Prognostic biomarkers

 MSI

Most advanced prognostic biomarker in adjuvant disease setting

 

Most studies indicate MSI-H is associated with good prognosis

[11••, 12••, 13]

Some studies do not associate MSI-H with good prognosis

[1417]

 CIN—eg, 18q LOH

Associated with poor prognosis in two meta-analyses

[21, 22•], [23]

Not all studies associate 18q LOH with poor prognosis

[2426, 73]

 K-ras

Adjuvant disease: more recent studies have found no prognostic value of k-ras mutation

[11••, 32••]

Some studies have found K-ras mutation is negative prognostic factor

[12••, 34, 35]

 B-raf

Strong prognostic factor for inferior overall survival in adjuvant disease setting (as well as metastatic disease)

[11••, 30, 37, 38•]

Not prognostic for recurrence risk in QUASAR patient cohort

[12••]

 TP53

Conflicting evidence about prognostic effect

[36, 74, 75]

 Thymidylate synthase

Conflicting evidence about prognostic effect

[58, 7577]

 CIMP phenotype

Conflicting evidence about prognostic effect

[29, 30]

Predictive biomarkers

 MSI

Most studies indicate MSI-H is predictive for lack of benefit from fluoropyrimidine chemotherapy

[42, 43••]

Some studies do not associate MSI-H with lack of benefit from fluoropyrimidines

[12••, 14, 45]

MSI-H tumors may be more sensitive to irinotecan (not seen in all studies)

[49•, 50]

Studies with oxaliplatin: small numbers of MSI-H patients studied, conflicting evidence, no real conclusions can be drawn

[5153]

 K-ras

Mutation associated with lack of benefit for anti-EGFR monoclonal antibodies in metastatic disease

[31•]

 P53

Conflicting evidence regarding whether p53 is predictive for inferior survival in 5-FU–treated patients

[74, 79, 80]

 Thymidylate synthase

High expression associated with improved survival with 5-FU–based therapy in some studies, but not others

[76, 81, 82]

Prognostic Biomarkers: A Better Way to Stratify Risk?

Microsatellite Instability/Mismatch Repair Pathway

A microsatellite is a repeated sequence of DNA, which can become abnormally long or short in the setting of MSI. MSI occurs when mismatch repair genes are either mutated (for example in Lynch syndrome) or epigenetically silenced, causing an inability to repair single nucleotide DNA mismatches. Epigenetic inactivation by promotor CpG island hypermethylation of mismatch repair genes (largely MLH1) is seen in up to 15%–20% of sporadic colorectal cancer [6]. The frequency of MSI depends on tumour stage, with approximately 22% of stage II colon cancers being MSI-H, 12% in stage III, and only 2% in metastatic disease [7]. There is a relationship between MSI-H status and proximal location, poorly differentiated tumors, and lymphocytic infiltration [8, 9]. As such, MSI-H tumors may be regarded as a distinct subset of CRC—both phenotypically and also from their clinical behavior.

MSI has been shown to be a significant and well-validated prognostic biomarker in early-stage colon cancer in multiple studies, with MSI-H tumors having a better prognosis than those without mismatch repair dysfunction. Recent large translational research sub-studies from prospective clinical trials to demonstrate this include the PETACC-3 and QUASAR studies. PETACC-3, a large adjuvant (stage II and III) colon cancer trial comparing infusional 5-FU/LV alone or with irinotecan [10], incorporated a prospective translational research component such that over 1500 tissue blocks were available for analysis [11••]. Multivariable survival analysis showed that patients with MSI-H tumors had a significantly better prognosis than those with MSI-L/S tumors. The hazard ratio for OS (MSI-H vs MSI-L/S) for stage II and III cancers was 0.39 (95% CI 0.24–0.65, P = 0.00023). This was even more striking for stage II cancers (HR 0.14, 95% CI 0.03–0.64, P = 0.011). In PETACC-3 there was no untreated control arm and all patients had been treated with chemotherapy, so no comparison was possible for patients receiving chemotherapy versus surgery alone. The QUASAR study, on the other hand, compared adjuvant chemotherapy versus observation in stage II CRC, and with over 1900 patients with tissue available for analysis similarly found that recurrence rate for MSI-H cancers was half that for MSI L/S cancers (risk ratio 0.53, 95% CI 0.40–0.70, P < 0.001) [12••]. An earlier meta-analysis including 32 studies and totaling over 7600 patients (including stage II disease) estimated the HR for OS associated with MSI was 0.65 (95% CI 0.59–0.71), indicating an overall prognostic advantage in patients with MSI-H tumors [13].

However, not all studies have concluded that patients with MSI-H tumors have a better prognosis. An analysis of 548 archival tissues from patients with stage II and III colon cancer participating in four NSABP trials found no difference in OS between MSI-H and MSI-L/S patients, irrespective of treatment arm [14]. The advantage of the NSABP analysis is that a comparison between treated and untreated patients was possible; however, this retrospective analysis is complicated by the fact that not all four trials compared chemotherapy with surgery alone; and less than a third of patients had tissue available for analysis. The authors acknowledged that retrospective studies have inherent biases, but argued that their rate of detection of MSI-H tumors (18.1%) was similar to other studies, and characteristics of those patients with representative blocks were similar to those who were not included. An Italian study of 893 consecutive patients with all stages of CRC found that although cancer-specific survival was higher in MSI-H cancer, this was not independent of tumor stage in their multivariate analysis [15]. An analysis from Belgium of 241 early-stage colon tumors similarly found no survival benefit for MSI-high tumors [16]. The retrospective nature of the study and the relatively small sample size may partly explain this; however, a larger, prospectively designed study of 416 patients with stage I–IV CRC found no difference between MSI-H and MSI/L/S groups in terms of disease-free and overall survival [17].

Overall, most recent prospectively designed studies indicate that MSI status is a valid prognostic biomarker in early-stage colon cancer, albeit with some dissent in the literature. The mechanism by which MSI may confer an improved survival outcome is not well understood, but theories include its ability to induce lymphocytic infiltration thereby enhancing the host immune response to the tumor [18].

Chromosomal Instability Pathway

The CIN pathway is the mechanism thought to be responsible in the etiology of the majority of CRC. The loss or gain of portions of chromosomes, resulting in karyotypic variability between cells, results in imbalances in chromosome number (aneuploidy) and loss of heterozygosity (LOH)—the loss of function of one allele of a gene in which the other was already inactivated. There are multiple mechanisms that may lead to chromosomal instability, including defects in chromosome segregation, telomere regulation, and DNA damage response [19].

LOH is regarded as the “hallmark” of CIN-positive tumors, although the mechanism by which this occurs is not well understood. The most common example is 18q LOH (also termed 18q allelic imbalance [AI/CIN]), seen in up to 70% of colorectal tumors [20]. Genes found in this region include SMAD2 and SMAD4 (both mediators of the TGF-β pathway), and “deleted in colorectal carcinoma” (DCC). Other genes which may be affected by LOH include APC and TP53 (both tumor suppressor genes which may become inactivated by this method).

18q LOH has been associated with poor prognosis in some studies, but not others. Two meta-analyses support the association between CIN and worse prognosis. A meta-analysis of 27 studies (both adjuvant and metastatic CRC), published in 2005, showed inferior OS in patients with 18q LOH including the adjuvant disease setting (for stage II and III: HR 1.69, 95% CI 1.13–2.54) [21]. A subsequent meta-analysis of studies stratifying for survival by CIN status (63 studies and over 10,000 patients in total) also concluded that CIN was associated with a worse prognosis (for stage II and III: HR 1.45, 95% CI 1.27–1.65, P < 0.001) [22•]. More recently, Bertagnolli et al. [23] analyzed the role of 18qLOH as a prognostic marker in patients with stage II colon cancer treated in the CALGB9581 trial. This study randomized 1738 patients to postoperative antibody therapy (MoAb 17-1A, no longer in use for CRC) or observation alone (no patients received chemotherapy). In the 156 patients who had 18qLOH results available, a significantly decreased DFS and OS were observed in patients with 18qLOH (5-year OS 0.85 vs 0.98, HR 0.25, 95% CI 0.07–0.83, P = 0.01).

Despite these data indicating that 18q LOH is an adverse prognostic marker, there is dissent in the literature. Popat et al. [24] in 2007 analyzed 255 tumors from patients participating in a molecular substudy of an adjuvant colon cancer clinical trial, finding no difference in 5-year survival based on 18q LOH, even after correcting for covariates. Although not yet published, the results from the PETACC-3 translational substudy described earlier suggest that in this patient cohort, 18q LOH was not associated with inferior overall survival (HR 0.84, 95% CI 0.60–1.24, P = 0.43) [25] in stage II and III colon cancer combined. Interestingly though, the group had previously reported that the prognostic value of 18qLOH for relapse-free survival differed between stage II (significant) and stage III (nonsignificant) cancers—indicating a potential stage-specific prognostic value. However, in a multivariate analysis, 18q LOH did not remain significant [26].

CpG Island Methylator Phenotype Pathway

CIMP is characterized by gene silencing due to hypermethylation of CpG islands. Normally methylation does not occur within the CpG islands in the promoter regions of genes; aberrant methylation in these regions results in epigenetic silencing. This is the mechanism by which sporadic MSI occurs (in particular MLH1); other genes together with MLH1 appear susceptible to aberrant methylation and CIMP is defined as methylation of three of more specific loci [27]. Not all CIMP-high tumors are MSI-high, however. CIMP is seen in approximately 15% of CRC [28]. There is conflicting evidence about the prognostic effect of CIMP status. Presence or absence of other markers has an effect; for example, in tumors without MSI, increasing CIMP (from negative to high) conferred a worse prognosis in one population-based series of 582 cases [29]. This is in contrast with Ogino et al. [30], who reported in a cohort study of 649 patients with stage I–IV colon cancer that CIMP-high was an independent predictor for low cancer-specific mortality (independent of MSI status).

Oncogenes and Tumor Suppressor Genes as Biomarkers

There are numerous oncogenes and tumor suppressor genes which, when activated or deactivated respectively, can contribute to carcinogenesis. The activation or deactivation may be caused by the pathway mechanisms described above; for example, loss of heterozygosity or silencing of mismatch repair genes may “switch on” an oncogene such as k-ras or b-raf, or “switch off” a tumor suppressor gene such as TP53 or SMAD4. Some of these genes are also proving to have a role as prognostic and/or predictive biomarkers, in both the adjuvant and metastatic disease settings for CRC. Two of the most extensively studied oncogenes—k-ras and b-raf—are discussed below.

K-ras is a GTPase protein and one of the members of the Ras subfamily of proto-oncogenes involved in cellular signal transduction from the epidermal growth factor receptor (EGFR). Mutations in k-ras are common, in around 40% of colorectal cancers, and result in a constitutively activated signaling pathway, which ultimately leads to gene transcription and cell proliferation. The constitutively activated pathway thereby confers resistance to anti-EGFR therapies [31•]. In the metastatic disease setting it is used as a predictive biomarker in this context. The frequency of mutations is similar in stage II versus III disease [32••].

There is uncertainty as to whether K-ras mutations are prognostic biomarkers in both the metastatic and adjuvant disease settings. In patients with metastatic CRC treated in the CO.17 randomized trial of cetuximab versus supportive care, k-ras mutations were not found to be of prognostic significance in the supportive care arm, a cohort where the effect of a biomarker can be analyzed independently of any potential treatment effect [31•]. In the PETACC-3 translational substudy, K-ras mutations did not have a major prognostic value for recurrence-free or overall survival [11••]. This finding is supported by translational research arising from the NCI CALGB 89803 study in stage III colon cancer comparing 5-FU/LV with or without irinotecan [33], where 508 patients had tissue available for K-ras testing [32••]. There was no difference in DFS or OS between patients with wild-type and mutant tumors. This contrasts with the QUASAR trial’s translational research findings, where risk of recurrence was significantly higher for K-ras mutant compared with K-ras wild-type tumors (risk ratio 1.40, 95% CI 1.12–1.74, P = 0.002) [12••]. Earlier findings from the RASCAL studies [34, 35] similarly showed an increased risk of recurrence and death associated with K-ras mutations, as well as a number of other retrospective studies which seemed to indicate the same [36]. Although K-ras mutational status is not currently routinely used in early-stage colon cancer as either a prognostic or predictive biomarker, ongoing translational research is crucial due to these conflicting results from recent studies.

B-raf is a proto-oncogene downstream from k-ras in this signaling pathway. Mutations in the B-raf gene, most commonly the V600E mutation, are mutually exclusive from k-ras mutations. They are less common than k-ras mutations, occurring in around 8% to 10% of CRCs, and their frequency is similar in stage II versus III disease [11••]. In both the metastatic and adjuvant disease settings, studies indicate that the B-raf mutation is a valid adverse prognostic biomarker. The PETACC-3 translational study found that BRAF mutations were prognostic for inferior overall survival, in particular in MSI-L/S patients [11••]. This has been seen in other retrospective studies [32••, 37, 38•]. Interestingly, the QUASAR translational study did not find an association between B-raf mutations and risk of recurrence; but this may be at least partly because in this cohort 57% of B-raf mutant tumors were MSI-high [12••].

For both these mutations and others, it seems likely that looking at the single mutations alone without the context of other information, such as MSI status, is less useful. For example, the B-raf mutation occurs in around 10% of colorectal cancers overall, but in up to 50% of older female patients with right-sided tumors that are MSI-H [39•]. In addition, examining the association between particular biomarkers can facilitate a greater understanding of disease biology. In the PETACC-3 translational study, k-ras mutations were more common in MSI L/S tumors and b-raf mutations more common in MSI-H tumors, with some differing effects on survival depending on these subgroups, supporting the concept that these are in fact different forms of colon cancer. While it may seem counterintuitive that sporadic MSI-H tumors are associated with a better prognosis, given that B-raf mutations (an adverse prognostic biomarker) are seen more commonly in these tumors, could MSI-H b-raf mutant and MSI-H b-raf wild-type tumors be two distinct molecular types of colon cancer? These interactions and associations are obviously complex and our understanding of them may be marred by the effects of subgroup analysis and relatively small patient numbers when attempting to drill down to these subgroups. However, such analyses have facilitated our understanding that there appears to be distinct biological and molecular subtypes of colon cancer, which have different behaviors and may respond differently to particular anti-cancer agents. This is the cornerstone of tailoring treatment to an individual.

Predictive Molecular Biomarkers in Early-Stage Disease: Toward Better Tailoring of Treatment

MSI

In recent years there has been intense interest regarding the role of MSI status as a predictive biomarker for the use of 5-FU (and other agents) in the adjuvant disease setting. As early as the early 1990s, in vitro data have suggested reduced sensitivity to fluoropyrimidine and other drugs in MMR-deficient tumors, based on their genetic destabilization due to methylation [40, 41]. The hypothesis is that the 15% to 20% of sporadic CRCs that are presumed due to the MSI pathway are distinct from the remaining 85% that are caused by the CIN pathway—not just in terms of pathologic features or prognosis, but also with respect to their resistance to chemotherapy. As chemotherapy, for the most part, affects DNA replication, it makes sense that alterations in DNA in distinct tumor subtypes may confer differential sensitivity to anti-cancer agents.

MSI and 5-FU

Multiple studies have indicated that patients with MSI-H tumors do not derive benefit from adjuvant 5-FU. Ribic et al. [42] examined 570 specimens from patients previously enrolled in five separate randomized trials for stage II or III colon cancer, including groups receiving no treatment, allowing a comparison of survival between patients receiving 5-FU or not. Although MSI-H was a significant prognostic factor for improved survival overall, the group found that patients with MSI-H tumors did not derive any additional benefit from adjuvant chemotherapy, whereas those with MSI-L/S tumors did [42]. In fact, for patients with MSI-H tumors, treatment was associated with possibly a worse outcome for both stage II and stage III cancers. A subsequent international collaboration using data from 457 patients participating in randomized trials of 5-FU–based adjuvant therapy versus no therapy was published in 2010 [43••]. The authors reported similar findings in that patients with MSI-H tumors receiving chemotherapy had no improvement in disease-free survival compared with those receiving surgery alone. Pooling this analysis with Ribic’s earlier cohort, for patients with MSI-H stage II disease, 5-FU–based chemotherapy was associated with inferior overall survival (HR 2.95, 95% CI 1.02–8.54, P = 0.04). The authors did acknowledge that patients included in this analysis were entered into trials as far back as 20–30 years previously; and that only the role of 5-FU was examined in an era where oxaliplatin-based regimens are now regarded as standard of care for stage III colon cancer. Nevertheless, when looking at stage II disease, the results are somewhat compelling.

At this stage, with what we know, should we be using MSI status to determine whether patients should, or should not, receive adjuvant 5-FU–based chemotherapy? Although MSI status appears to be both a prognostic and also a predictive biomarker for chemotherapy effect and benefit in more recent studies, there is obvious controversy and conflict regarding this. Nevertheless, MSI status is beginning to factor into management decisions for stage II disease in some centers. The National Comprehensive Cancer Network (NCCN) clinical practice guidelines now suggest that stage II MSI-H patients “may have a good prognosis and do not benefit from 5-FU adjuvant therapy” [44]. However, there is dissent in the literature and, as always, debate regarding how much can be taken from retrospective cohort studies. In metastatic disease, in a study of 244 patients non-randomly allocated to treatment, Liang et al. [45] found that MSI-H tumors responded significantly better to 5-FU chemotherapy compared with MSI-L/S tumors. The NSABP collaborative study, comparing patients treated with 5-FU or observation, found no interaction between MSI status and treatment for RFS or OS [14]. Analysis of tissue from patients participating in the QUASAR trial, comparing post-surgery 5-FU/FA versus observation alone in stage II disease, indicated that MMR status did not predict for benefit from chemotherapy; nor did k-ras or b-raf mutations [12••]. These differences are perhaps related to the retrospective nature of some studies (with the inherent biases that follow); differing techniques for measuring MSI that different studies have used; and the heterogeneity of MSI tumors (CIMP or CIN backgrounds behaving differently, depending on the pathway).

We may learn more from the Eastern Cooperative Oncology Group (ECOG) 5202 trial, which closed to accrual in February 2011 with over 3000 patients enrolled. Patients with resected stage II colon cancer were assigned to treatment based on biomarker information [46]. Patients classified as “high risk” (MSI-L/MSS tumors and chromosome 18q LOH) received adjuvant chemotherapy (FOLFOX ± bevacizumab), whereas “low-risk” patients (MSI-H and no 18q LOH) received no adjuvant therapy. This trial was one of the first in CRC to stratify patients on the basis of biomarker status and was a large step forward in trial design, although we will not learn whether chemotherapy affects outcomes within each risk category because there were no “control” arms for each risk group.

MSI and Irinotecan

Irinotecan, to date, does not have a role in early-stage colon cancer, although its use in metastatic disease is well-established. Both the PETACC-3 and CALGB 89803 studies failed to demonstrate a survival benefit for adding irinotecan to a 5-FU regimen [10, 33]. Both trials included biomarker substudies. This is a significant strength of the trial designs, in that we may ultimately come to understand the reasons for the finding of a lack of benefit for irinotecan for early-stage disease, in contrast with its role in stage IV CRC.

More than a decade ago, preclinical data from xenograft models had suggested that MSI-H tumors may be more sensitive to irinotecan [47]. While the PETACC-3 translational substudy did not find an effect for the addition of irinotecan depending on MSI status [48], in the CALGB 89803 study, patients with MSI-H tumors treated with irinotecan had an improved 5-year DFS compared with MSI-L/S (hazard ratios 0.76 vs 0.59, P = 0.03) [49•]. This difference was not seen in patients treated with 5-FU/LV alone. In metastatic disease, a small retrospective study (n = 72) found that MSI-H was a predictive factor for response to irinotecan [50]. Unfortunately, definitive conclusions about the effect of MSI on irinotecan responsiveness are limited by relatively small sample sizes.

MSI and Oxaliplatin

In vitro, MMR deficient and proficient cells are equally sensitive to oxaliplatin. Small studies are emerging regarding the role of MSI status on response to oxaliplatin. A study of 135 stage II–IV patients with CRC, all treated with adjuvant FOLFOX, found no difference in outcomes based on MMR status, suggesting that oxaliplatin may overcome the negative impact of 5-FU on outcome [51]. This study did not include a “control” group without oxaliplatin, however. The same research group, in a larger cohort of 171 patients with metastatic disease treated with either CAPOX or FOLFOX, similarly found no difference in PFS or OS depending on MMR status [52]. Only 6% (n = 10) of the cohort were MSI-H, and again, all patients received oxaliplatin, so this makes conclusions somewhat limited. Zaanan et al. [53], in a retrospective study of 233 patients with stage III colon cancer treated with either 5-FU/LV or FOLFOX (not randomized), found on univariate analysis that MSI-H tumors (total n = 32) had significantly improved DFS with oxaliplatin compared with 5-FU/LV alone (HR 0.17, 95% CI 0.4–0.68, P = 0.03). The impressive HR must be tempered by the fact this was not a prospective study, and patients were not randomized to treatment. The median age of MSI-H patients was considerably lower in patients receiving FOLFOX compared with 5-FU/LV. Given that oxaliplatin-based chemotherapy is currently the standard of care in the treatment of stage III colon cancer based on results from the NSABP C-07 and MOSAIC studies [54, 55], MMR status alone is currently unlikely to influence a decision whether or not to give oxaliplatin-based treatment to these patients.

CIN and CIMP

Could CIN be used as a predictive as well as a prognostic biomarker? The ECOG 5202 study will not give us this information, as all patients with 18q LOH will receive chemotherapy. Although the influence of CIN on response to 5-FU, irinotecan, and oxaliplatin has not been well studied, CIN is thought to be associated with taxane resistance, as the mechanism of action of taxanes requires an intact spindle, and CIN inherently can interfere with their function. This is postulated to be the reason why taxanes are not active in CRC [56].

As a predictive biomarker, one recently published study has examined a population-based cohort of 302 patients with CRC and found that those with CIMP + CRC did not benefit from 5-FU–based adjuvant therapy, whereas 5-FU–based therapy significantly increased DFS in those with CIMP-ve tumors [57]. Further validation studies are needed to further explore these associations.

Pharmacogenomic Predictive Biomarkers

An individual’s genetic makeup can influence response to chemotherapy and other drugs by polymorphisms in critical enzymes involved in the drug’s metabolism. As such, genetic variation between individuals could be applied as a predictive biomarker to determine the likelihood of response to, or toxicity from, particular anti-cancer agents. The ability to tailor therapy based on pharmacogenomics to maximize efficacy while reducing toxicity is a prime example of personalized therapy; however, due to multiple factors including cost and in some circumstances a low prevalence of the polymorphism in question, this approach is not currently used routinely to guide treatment. Examples include thymidylate synthetase expression, a target for fluoropyrimidine therapy, which in itself may be a prognostic biomarker [58], and dihydropyrimidine dehydrogenase (DPD) deficiency, which predicts for severe 5-FU toxicity [59].

Much of the focus in the adjuvant disease setting is related to the role of fluoropyrimidines. However, there is emerging interest in potential predictive biomarkers for the role of oxaliplatin and irinotecan, which are generally combined with fluoropyrimidines. Polymorphisms in ERCC1 (a DNA excision repair protein) have been reported to be associated with oxaliplatin resistance in some studies, ranging from cell line studies [60] to cohort studies [61]. A recently published meta-analysis (incorporating 1787 patients in total) of a number of gastric and colorectal cancer studies found that particular ERCC1 and ERCC2 polymorphisms were significant predictive biomarkers for poor outcomes (response rate, PFS, and OS) with oxaliplatin-based therapies [62]. Largely these studies involved patients with metastatic disease, but the emerging role of ERCC1 polymorphisms may also be of value in the adjuvant disease setting in the future. For irinotecan, UGT1A1 polymorphisms (the commonest of which manifests as Gilbert’s syndrome) may affect irinotecan metabolism as the drug is inactivated by glucuronidation catalyzed by UGT1A1 [63] and hence predict for a different therapeutic ratio. However, testing for these polymorphisms is not currently in routine use even in metastatic disease, where irinotecan is largely used.

Tumor Gene Expression Profiling

As technology has advanced over recent years, it is now becoming possible to analyze sequences of genes from an individual’s tumor that may assist in predicting the chance of recurrence. This has become most advanced in the setting of early-stage breast cancer, where tests such as Oncotype DX and MammaPrint are used to assist in predicting recurrence risk and assessing the potential benefit of adjuvant chemotherapy. One advantage of this technique is that it combines the evaluation of multiple genetic markers rather than a test for a single molecular marker.

Numerous studies have correlated gene expression profiles in colon cancer with recurrence risk. The Oncotype Dx colon cancer assay has been developed based on four large and independent studies where seven recurrence-risk genes were selected from a panel of 761 candidate genes [64••]. The assay was validated using tissue and clinical details from patients with stage II colon cancer from the QUASAR trial as a prospectively designed validation study [65]. Similarly to MammaPrint, a ColoPrint gene assay has also been recently developed using a set of 18 genes from 188 patients. This has been validated in an independent set of over 200 samples from patients with stage I–III CRC. The signature appears robust in a multivariate analysis as an independent prognostic factor for recurrence [66••]. A separate study, recently presented, independently validated the use of ColoPrint in a sample of 233 patients with stage II and III disease, finding that the test was able to predict the development of distant metastases independently from clinical parameters such as T4 stage, nodal yield, and differentiation grade [67].

It will be interesting to monitor and note the application of this relatively new technology into routine care in the coming years, and its effect on cancer-specific outcomes in the setting of early-stage CRC. Notably, although both Oncotype Dx and ColoPrint appear to be useful prognostic tools, there remain limitations as to how to incorporate them into decision-making as to date they do not assist in predicting benefit from adjuvant therapy. During the development of the Oncotype Dx panel, six genes associated with a benefit from 5-FU were also selected [64••] but later removed as during the validation testing, the predictive treatment score was not found to be valid [65]. Given that gene expression profiles are in relatively initial stages of development for early-stage colon cancer, their ability to be used as predictive markers may be further clarified in coming years.

Conclusions

While our appreciation of the significance of prognostic and predictive biomarkers continues to grow, there remain gaps in the knowledge of how to apply them most appropriately in routine clinical practice. There are no molecular biomarkers in stage II and III colon cancer that are routinely used in guiding treatment, although this is starting to change with the assessment of MSI in stage II disease. We have been hampered along the way by small sample sizes, dissent regarding the most appropriate laboratory techniques for particular tests, lack of control samples for some hypotheses, and the necessary uncertainty that must come with the often complex decision about whether potentially curative treatment should be offered to a patient. Ongoing translational research will continue to enrich this area of medicine and our treatment of early-stage colon cancer, as we move toward our ultimate goal of “personalized medicine” for all patients.

Disclosure

No potential conflicts of interest relevant to this article were reported.

Copyright information

© Springer Science+Business Media, LLC 2011