MOLECULAR MARKERS FOR METASTATIC COLORECTAL CANCER

Prognostic Markers

Patient-related Prognostic Markers. Patient-related markers correlating with disease prognosis have been extensively investigated, are well documented, and are readily incorporated into the stratification process of modern clinical trials.

Performance status (PS) has consistently emerged as a prognostic marker, from early trials with 5-FU alone to trials incorporating biological targeted agents.[12–18] Most recent clinical trials do not incorporate patients with ECOG PS ≥2, making a true comparison difficult. A recently published pooled analysis by Sargent et al.[19] incorporated 6286 mCRC patients from nine controlled trials, of whom 509 had a PS = 2. PS was found to be a prognostic marker for OS, progression free survival (PFS) and response rate (RR), despite a similar relative benefit from treatment. Higher rates of toxicity in patients with a PS = 2 vs PS = 0/1 were also demonstrated.

The importance of age as a prognostic marker has been controversial. Elderly patients are often an underrepresented group with little incorporation in clinical trials and 'ageism' may be a common practice, depicted by less frequent use of combined treatment and less invasive interventions in the elderly.[20]

More than 70% of deaths from CRC in the Western world occur in patients ≥65 years old. The fact that the outcome in this group of patients is often worse may be attributed to more advanced stage at diagnosis and often, less aggressive treatment and decreased use of second- and third-line therapies.[20–21] Elderly patients fit enough to be eligible for clinical trials appear to derive a similar benefit from the treatment. In general, fit elderly patients tolerate the chemotherapy quite well, but some studies reported that elderly patients may have more specific toxicity from cytotoxics: e.g. increased neutropenia in patients ≥75 years when treated with irinotecan[22] and an increase in bevacizumab-related arterial thrombosis in patients ≥70 year.[23] Co-morbidity and socio-economic status further impact treatment availability and prognosis. Several additional patient-related factors that have been shown to produce improved prognosis include female sex[12, 24] and limited weight loss at diagnosis.

Tumor-related Prognostic Markers. Tumor-related biomarkers and factors remain the most important prognostic variables. The stage of disease currently described by TNM, as published by the American Joint Committee on Cancer (AJCC) and American Cancer Society (ACS)[25] is one of the strongest prognostic factors.[24]

A colonic primary tumor is a positive prognostic marker compared to a rectal tumor.[13–14] A mucinous histology is usually accepted as being related to a worse outcome. Metachronous metastases have a better outcome compared to synchronous metastases. Metastatic spread confined to the liver shows improved OS rates as opposed to spread to multiple organ sites.[12] In contrast, patients with liver metastases show a shorter survival compared to lung metastases.[14, 17] Peritoneal metastases are related to a worse outcome. The higher the number of metastatic sites, the worse the prognosis is.[14] In patients with mCRC, a prior adjuvant treatment is considered as a negative prognostic factor, especially when treated with oxaliplatin-based adjuvant chemotherapy.[12, 24] The number of previous lines of chemotherapy is also accepted as a negative prognostic factor.[16]

Center-related Factors. The center in which treatment is directed influences the prognosis, with experience and volume[26–27] of patients playing a role, as well as possible deviation from standard clinical practice[28] for a variety of reasons. The treatment center is therefore a common stratification factor in multi-centric and especially in international trials.

Biochemical Prognostic Markers. Biochemical markers should optimally be combined together and not be seen as single factors. Many of these factors have a prognostic as well as a predictive significance. Pretreatment blood count values with low hemoglobin levels[13, 15] and high white blood cell (WBC) counts[13, 17] depict poor prognosis. Increased alkaline phosphatase baseline levels are considered one of the strongest poor prognostic factors for survival in mCRC.[12–14, 18] Lactate dehydrogenase (LDH) participates in anaerobic glycolytic metabolism in tumor cells and is postulated as a biomarker for high angiogenic tumors.[29] High base line LDH measurements predict poor prognosis.[12, 17] High levels of serum bilirubin and low levels of serum albumin also represent poor prognosis. Preoperative CEA has been advocated as prognostic for disease free survival as well as persistent disease after surgery.[30] Baseline levels of CEA prior to treatment in the metastatic setting also seem to have some prognostic relevance.[12, 17] A multivariate analysis,[14] on a total of 3825 patients from 19 prospective trials treated with 5-FU-based therapy for mCRC distinguished three risk groups based on the main prognostic markers; a low risk group described as ECOG 0/1 and only one tumor site; an intermediate risk group defined as patients with ECOG 0/1 and more than one tumor site and alkaline phosphatase <300 U/l or patients with ECOG >1, WBC count <10 × 109/l and only one tumor site; and a high risk group of patients with ECOG 0/1 and more than one tumor site and alkaline phosphatase of ≥300 U/l or patients with ECOG >1 and more than one tumor site or WBC count >10 × 109/l. The median survivals were 15 months, 10.7 months and 6.1 months, respectively. A recommendation to stratify future trials based on these prognostic factors is sometimes made.

Molecular Prognostic Markers. Real evidence as to the validity of the many proposed prognostic markers published in the literature is scarce. Two molecular markers seem to have a consistent prognostic value.

Microsatellite Instability: The development of CRC through microsatellite instability (MSI) is attributed to the hereditary HNPCC syndrome patients and to about 15% of sporadic CRC patients.[31] MSI-H patients have a phenotype characterized by right-sided location and a relatively early stage at diagnosis. In a study [32] in 607 patients at ages ≤50 diagnosed with CRC, 17% were found to have high MSI that was associated with a significant survival advantage regardless of standard prognostic factors, including stage. An Italian study in 1263 CRC patients detected 20% of the tumors as MSI-H showing a favorable prognosis.[33] A subset analysis according to stage, though, showed this advantage was not maintained in stage IV patients. A meta-analysis by Popat et al.[34] spanning across 32 studies with 7642 cases of whom 1277 (16.7%) were MSI-H reported the hazard ratio estimate for overall survival associated with MSI as 0.65 (95% CI, 0.59 to 0.7) maintained through stages II–IV.

BRAF Mutations: The V600E mutation in BRAF has been implicated in several trials as a poor prognostic marker in CRC. Ogino et al.[35] analyzed 649 colon cancers of all stages. Analysis adjusted for other prognostic markers revealed that BRAF mutation was associated with a significantly high cancer-specific mortality (multivariate HR 1.97; 95% CI 1.13 to 3.42). Roth et al.[36] reported also that BRAF is a negative prognostic factor in stage II/III colon in the framework of the PETACC-3 study and Van Cutsem et al.[37] reported the negative prognostic significance of BRAF mutations in metastatic CRC treated in the Crystal trial.[37] The relationship between BRAF mutation and MSI was assessed by Somawitz et al[38] in 911 CRC patients. The V600E mutation was found in 9.5% of the overall population, in 5% (40/803) of microsatellite-stable (MSS) cancers and in 52% (43/83) of MSI tumors. The BRAF V600E mutation was associated with a significantly worse survival in stages II to IV MSS colon cancers. No effect on prognosis in MSI tumors was detected.

Several conflicting data have been published on the prognostic significance of KRAS mutations. Two recent large studies, however, showed no prognostic role of KRAS mutations in stage III and II/III colon cancer.[36, 39]
Predictive Markers

Most if not all proclaimed predictive markers have not been validated in prospectively designed trials and must therefore be used with caution. Some of these markers are, however, sometimes used in every day clinical practice to guide the clinicians.

A distinction must also be made between markers that can be assessed prior to the start of treatment as opposed to those revealed only once treatment has started. The first might help determine if treatment should be started while the other might help determine if treatment should be stopped.

Patient-related Predictive Markers for Outcome. Markers predicting outcome have been investigated for specific treatment regimens and as such can only be considered when these protocols are implemented.

General Clinical and Biochemical Factors: Patient and tumor variables are strong prognostic factors and as such often predict efficacy of the chemotherapeutic treatment mCRC. Several trials using regimens based on 5FU, irinotecan and/or oxaliplatin reported that baseline hemoglobin and WBC levels, LDH levels, time since diagnosis, low number of organs involved, performance status, and age are predictive of response and survival.[40–45] The data in these studies usually do not clearly distinguish between the prognostic and predictive value of these markers.

Extensive and symptomatic peritoneal carcinomatosis is often related to a lower chance of response to an antitumoral therapy, on top of its prognostic significance. However, peritoneal carcinomatosis is often difficult to evaluate properly.

The ability of carcinoembryonic antigen (CEA) kinetics to predict response to chemotherapy had been debated. Clearly absolute correlation does not occur as patients presenting with decreased markers may still show progression on imaging. Strimpakos et al. [46] retrospectively examined 670 patients receiving first-line chemotherapy and having baseline CEA measurements plus at least two more measurements during treatment. CEA levels were described as flare in 78 (12%) patients (flare defined as a ≥15% rise from baseline with a minimum 4 μg/l rise followed by subsequent ≥15% decrease from baseline). Those were compared to patients with increasing CEA and were found to have a significantly better ORR (73% versus 11%; respectively, P < 0.001), similar significant differences were seen with PFS and OS. Other series support this observation.[47] According to a few small trials, a CEA response after treatment, at least in patients who had pretreatment elevated levels, correlated with RR.[48–50] The use of CEA is, however, not recommended to routinely replace imaging in assessing response.

Skin Rash: One of the more consistent clinical parameters is the association found between the development of skin toxicity, mainly rash, and response on EGFR-targeting agents: the more severe the rash (usually acneiform) is, the better the RR, PFS and/or survival is for patients treated with EGFR inhibitors.

This correlation has been shown in early phase II trials, as well as in phase III trials.[51, 52] It is reported for cetuximab and panitumumab in colorectal cancer as well as in other disease sites and for the small molecule tyrosine kinase inhibitors: e.g. erlotonib in pancreatic cancer.[53] Starting from this information, the Everest study has explored whether dose increase of cetuximab in patients with no rash or a slight rash can increase the activity of cetuximab.[54] Although the RR was slightly higher in patients receiving a dose increase of cetuximab, with the aim to induce a more pronounced rash and a higher activity, this strategy cannot be recommended outside of clinical trials before further validation studies have been reported.

Hypertension: Preliminary data of a small phase II study in 39 patients treated with bevacacizumab/irinotecan/5-FU has suggested the correlation of activity and early hypertension.[55] This correlation has also been suggested in pancreatic cancer patients treated with axitinib, a multitarget tyrosine kinase inhibitor.[56] The data on the correlation of hypertension and benefit from an angiogenesis inhibitor are at this moment weak without clear therapeutic implications in clinical practice.[57]

Clinical and Biochemical Predictive Markers for Toxicity. Renal Function and Capecitabine: Pharmacokinetic studies of capecitabine showed that moderate impairment of creatinine clearance causes an increase in the AUC of the key metabolite 5′-DFUR (the immediate precursor to 5-FU) and that this correlates with more adverse events.[58] A combined safety analysis on two phase III trials[59] with capecitabine revealed an increase in grade 3–4 toxicity in patients with moderate renal impairment at baseline (defined as creatinine clearance 30–50 ml/min calculated by the Cockcroft and Gault formula[60]). More dose reductions were needed during the treatment and a 75% starting dose was therefore recommended in patients with a moderately impaired renal function without compromising activity. An increase in toxicity was also observed in patients with impaired renal function treated with 5-FU/LV.

Bilirubin Levels and UGT Polymorphism and Irinotecan: Irinotecan's active metabolite SN-38 is inactivated through glucuronidation by uridine diphosphate glucuronosyltransferases (UGTs) mainly in the liver and is excreted through the bile ducts. Bilirubin being a substrate of UGT1A1 is a possible indirect marker of its activity. An analysis of four phase II trials with irinotecan revealed that increased bilirubin levels correlate with a high risk for grade 3–4 neutropenia.[45] A meta-analysis demonstrated that UGT1A1*28 genotype is moderately predictive of severe irinotecan induced hematologic toxicity at moderate doses of irinotecan and strongly predictive at high doses, but at low doses these patients have a comparable incidence of toxicity to other patients [61–62]. In 2005 the FDA added a warning to the irinotecan packaging label indicating that patients with the UGT1A1*28 genotype were at increased risk for neutropenia and that a reduced initial irinotecan dosage should be considered. However, the precise implications of UGT1A1 genotype on the safety, efficacy, and development of individualized patient dosing of irinotecan are not yet clear. Significant unexplained variability exists within the UGT1A1 subgroups, and other characteristics of the patients have been understudied. For example, Miya et al.[63] identified many additional clinical characteristics that affected SN38 distribution and levels such as sex, BSA, and age. Thus possible confounders to the UGT1A1*28 allele-irinotecan adverse event relationship include: other enzymes and transporters that play a role in irinotecan disposition as well as other genetic variation in the UGT1A1 gene; age, gender, and ethnicity; comorbidity and performance status; pretreatment serum bilirubin concentrations; use of other drugs that share the irinotecan metabolizing pathway. Important questions still remain before implementing UGT1A1 testing in clinical practice such as how accurately does genotyping of UGT1A1 predict total irinotecan metabolic capacity.

Cardiovascular Disease as Risk for Toxicity on Bevacizumab: Patients treated with bevacizumab are more likely to develop an arterial thrombosis.[23, 64] Patients with a history of arterial thrombotic disorders such as angina pectoris, myocardial infarction, or stroke and elderly patients above 65 years are especially at risk. In these patients the potential benefit of an antiangiogenic treatment should always be put into balance with the potential adverse events.[65]
Molecular Predictive Markers for Efficacy

Predictive Markers for Anti-EGFR Antibodies. KRAS mutations (most frequently analyzed those in codon 12 and 13) are strongly associated with resistance to the anti-EGFR antibodies cetuximab and panitumumab. This has been consistently shown in many small retrospective data sets, retrospective analyses of larger phase III studies, and also in a few prospective studies.[52, 66–71] Approximately 40% of patients with CRC harbor a KRAS mutation and will derive no benefit from cetuximab or panitumumab. Of the remaining patients with a KRAS wild-type tumor, approximately 15–20%, will have an objective response when treated with an antibody alone and approximately 35–40% when treated with cetuximab plus irinotecan in irinotecan refractory CRC, leading to the statement that KRAS mutations predict resistance to cetuximab and panitumumab as single agent and in combination with chemotherapy in pretreated as well as in early lines of treatment in combination with a cytotoxic partner. It has even been reported that a KRAS mutation is associated with a deleterious effect when patients are treated with an oxaliplatin based backbone in combination with an anti-EGFR antibody. Several small studies have reported that there is a more than 95% concordance between the presence of a KRAS mutation in the primary tumor and in the metastases.[72] This consistent association has led the major regulatory bodies to limit the use of cetuximab and panitumumab to KRAS wild-type tumors with mandatory evaluation of mutations prior to the administration.

Emerging Potential Predictive Markers for Anti-EGFRs. BRAF and KRAS mutations are considered mutually exclusive and apparently mutated at a similar phase of the tumorigenesis.[73] The BRAF mutation (mainly V600E) is detected in 6–10% of patients with mCRC.[37, 75] Several studies have reported consistently that the presence of the BRAF mutation V600E correlates with resistance to cetuximab and panitumumab in chemorefractory mCRC.[74–77] In the updated analysis of the Crystal trial[37] 9% of the K-ras wild-type patients had a BRAF mutation (6% of all patients with CRC). This analysis showed that the presence of BRAF mutations had a strong negative prognostic significance. Whether BRAF mutations have also a predictive significance for cetuximab in combination with a standard cytotoxic combination in the first line treatment of mCRC cannot be shown, but cannot be excluded either in view of the important prognostic significance of BRAF mutations, the potential interaction with chemotherapy, and the relatively low frequency of BRAF mutations. Similar findings were reported in the Opus study.[78] Similar to KRAS, a retrospective analysis showed a high concordance between the presence of B-raf mutations in the primary tumor and in metastases.[72, 79]

Amphiregulin and epiregulin are ligands of the EGFR. Preliminary evidence from small series showed correlation between ligand expression levels and efficacy under anti-EGFR treatment. Recently larger retrospective studies have confirmed the potential predictive role of these ligands in patients with a KRAS wild-type tumor. High epiregulin gene expression was found to be associated with improved median PFS (5.4 versus 1.9 months; P < 0.0001, respectively), and median OS (9.8 versus 5.1 months; P < 0.001, respectively) in KRAS wild-type patients treated with cetuximab in chemorefractory CRC in the NCIC CTG CO.17 phase III trial.[80] A correlation between epiregulin and amphiregulin expression was also reported in 200 KRAS wild-type chemorefractory CRC by Jacobs et al..[81] Further validation of the ligand data is however needed in prospective trials, as well as in earlier lines of treatment in combination with chemotherapy. Moreover the linear correlation between the level of amphi- and epiregulin may render clinical decisions more difficult.

Additional effectors participating in the EGFR downstream signal transduction are PI3K, PTEN and NRAS. It is biologically reasonable to assume that further mutations might preclude the efficacy of anti-EGFR antibodies. Preliminary evidence to this effect is emerging.[76, 82–84] According to some recent analysis NRAS mutations may also play an important role in determining resistance to anti-EGFR antibodies. However, the data on PI3K mutations are less clear. According to a European Consortium analysis PI3K mutations are not clearly associated with resistance to cetuximab.[77] The actual data on the role of PTEN immunohistochemistry as a predictive marker are not so convincing.[85] PTEN was only positive in the metastatic site making it a difficult marker to use in the clinic.

Other Potential Markers for Chemotherapy Efficacy and Toxicity. ERCC1 for Oxaliplatin Efficacy: The mechanism of action of platinum agents is via creating cross-links in the DNA strands. The nucleotide excision repair mechanism removes these cross-links and results in resistance to the platinums. One of the key proteins in this process is the ERCC1. Several retrospective analyses in patients with mCRC treated with oxaliplatin based regimens, showed an association between ERCC1-118 T/T polymorphism and treatment efficacy.[86–88] A similar trend was found with ERCC1 expression assessed by IHC[89] or mRNA.[90]

Potential Markers for 5FU Efficacy and Toxicity: Thymidylate synthase (TS) is the key enzyme targeted by 5-FU. Numerous studies were published discussing the role of TS levels and/or the role of TS polymorphisms, either by IHC or genetic analysis in predicting efficacy. Most of the accumulated evidence was performed on small population sizes and evidence is conflicting.[91–93] This precludes any official recommendation.

Dihydropyrimidine dehydrogenase (DPD) is the rate-limiting enzyme in the catabolism of 5-FU. A correlation between the DPD expression levels and outcome was reported in several studies [94], but the data are not strong and consistent enough to determine DPD routinely in clinical practice. More important are the data of several retrospective series that demonstrated the importance of DPD activity level in developing toxicity from 5-FU. Low activity levels of DPD appear to cause more rapid development of severe toxicity and even higher mortality rates than normal DPD levels. The most frequent mutation leading to decreased activity is IVS14+1G>A.[95–97] The current recommendation is to look for DPD deficiency in patients with severe fluoropyrimidine related toxicity in order to avoid further exposure to 5-FU that may lead to major toxicity.

Expert Panel Recommendations for Molecular Predictive Markers Testing. KRAS Mutation Status – for Anti-EGFR Treatment: Analysis for KRAS mutation should be routinely done before chemotherapy for metastatic CRC is given. Analysis can be done on paraffin-embedded tumor block of the primary or metastatic site.

BRAF Mutation Status – for Anti-EGFR Treatment: BRAF mutations have a strong prognostic significance. In chemorefractory mCRC, the evidence is accumulating for a predictive role of BRAF mutations as a marker for resistance and testing can be considered before an anti-EGFR antibody is considered. However, in earlier lines of treatment routine testing of BRAF mutations cannot currently be recommended.

MSI: In the adjuvant setting, an emerging role of MSI in stage II colon cancer has been shown. In young patients and in families with clinical features of Lynch syndrome, diagnosed with mCRC, MSI testing is part of the genetic evaluation in the exploration of genetic predisposition for Lynch syndrome and is mandatory for genetic counseling. There is however, not sufficient evidence to guide treatment choice in patients with mCRC based on MSI or MSS status.
Treatment Strategy for Metastatic Colorectal Cancer

Patients suspected to have metastatic spread should be thoroughly examined with available imaging techniques. An assessment of the tumor burden as well as the relevant prognostic and predictive markers should be done in order to clearly define the treatment goal and strategy.

The optimal treatment strategy of patients with mCRC should be discussed in a multidisciplinary expert team, at least for those patients with metastases limited to the liver and/or the lungs. The experience of the team has a clear impact on the outcome of the patient.

The majority of patients with metastatic disease have lesions that initially are not suitable for resection. It is important to select patients in whom the metastases are suitable for resection and those with initially unresectable disease in whom the metastases can become suitable for resection after a major response has been achieved with combination chemotherapy. The aim of the treatment in the last group of patients may therefore be to reverse initially unresectable metastatic CRC to resectable CRC, i.e. conversion therapy. Some patients may initially have a 'technically' resectable disease, but it is biologically not sound to do a resection. A response to therapy in these patients may tell that it later is biologically sound to do even extensive surgery. The determination of the goal of therapy (palliative versus potentially curative) will therefore determine the selected treatment option

Patients Needing or Desiring an Aggressive Approach. The primary decision directing strategic planning is the goal of the treatment. An aggressive approach with combination chemotherapy is indicated when the patient can withstand such treatment and falls under one of the following groups, since a maximal tumor shrinkage or maximal control of tumor progression may be waranted:

* patients with potentially resectable metastases;
* patients with clearly symptomatic disease in whom tumor regression is needed;
* patients with the risk of rapid deterioration due to an aggressive tumor biology and/or extensive disease.

In these patients a cytotoxic doublet in combination with a biological is proposed. As options for cytotoxic combination FOLFOX, FOLFIRI and capecitabine/oxaliplatin are usually proposed. The combination of three cytotoxics is usually not proposed as backbone, but remains an alternative for the experts when a high shrinkage of the tumor is needed and there is contraindication to biologic targeted therapies. In this situation the determination of KRAS mutation status is important. In patients with a KRAS mutant tumor the obvious choice is a bevacizumab-containing regimen. In patients with a KRAS wild-type tumor bevacizumab and cetuximab in combination with a cytotoxic doublet are valid options. The data for panitumumab is less mature, still not published, but the preliminary data show efficacy results in the same direction of cetuximab and this may become an option in patients with KRAS wild-type tumors. There are no data available of randomized studies comparing the activity of a doublet of cytotoxics plus bevacizumab with a doublet plus cetuximab/panitumumab in the first line treatment. Toxicity considerations related to the targeted agents are important, as well as considerations that the evidence for the activity of anti-EGFR antibodies – even administered as single agents – is stronger in later lines of treatment than for bevacizumab. It may be very relevant for a patient in good condition to use the available agents in the continuum of care of his disease in order to maximize the chances for a prolongation of survival.

The combination of a doublet of cytotoxics plus cetuximab has led to higher resection rates (although still low in absolute numbers) in patients with KRAS wild-type, liver-limited unresectable mCRC. The combination of FOLFOX/cetuximab and FOLFIRI/cetuximab has led to similar response rates and resection rates in KRAS wild-type tumors. The combination of a fluoropyrimidine/oxaliplatin/bevacizumab has led to a non-significant trend in an increased resection rate compared to the cytotoxic backbone alone, although no increase in response rate was shown.

Patients not Needing or Desiring an Aggressive Approach. In patients with clearly unresectable disease that is unlikely to become resectable, without relevant tumor-related symptoms, without an immediate risk of rapid progression and with important comorbidity, a sequential therapy starting with a fluoropyrimidine monotherapy in combination with bevacizumab (if no cardiovascular contraindications are present) remains a valid option. It is, however, not easy to select this group of patients, as there is no data regarding biomarkers for predicting the optimal strategy. It is estimated that up to 15%–20% of patients with mCRC belong to this group. If this less aggressive approach is considered as the suitable option for this population of patients, a close follow-up with frequent evaluations should be mandated, as if the selected treatment is not active the patients may render to a situation of rapid deterioration due to tumor burden and no further treatment options may be then considered. This effect has been suggested as one of the potential explanations why some sequential treatment strategies starting with a non-doublet fluoropyrimidine may have a worse outcome in OS as less patients have been treated at the end with all the available drugs in the continuum of care concept.

Later Lines of Therapy. In patients for whom first-line treatment has failed, the cytotoxic backbone should be changed. In the second-line treatment the systematic use of biologic targeted agents is less established. If bevacizumab was not used in the first line, it can be recommended to add it at this stage, as there is one study showing OS advantage. The data with the anti-EGFR antibodies suggest a similar relative benefit in the second- and third-line treatment. In the third line the combination of cetuximab plus irinotecan is clearly more active than cetuximab single agent and is therefore the preferred option in fit patients who can tolerate it. The combination of panitumumab/FOLFIRI has also been shown to be more active than FOLFIRI alone in second line.

Additional Issues. How Long to Treat?: It is recommended to continue first line combination with biologic targeted therapies until disease has progressed, treatment is intolerable by the patient, or metastases have become resectable. In the event that one of the cytotoxic partners (oxaliplatin, irinotecan) is stopped for safety reasons, it is usually recommended to continue with the biological in combination with a fluoropyrimidine until disease progression or non-acceptable toxicity appears.

Biological Agents Beyond Progression: To date, there is no clear evidence as to the advantage of continuing a biological agent after disease progression while exposed to this agent.

The Role of Rash: Although the severity of rash correlates with the efficacy of anti-EGFR antibodies, this has no immediate implications in clinical practice. A lack of rash is not currently recommended as reason for discontinuing anti-EGFR therapy.

Choice of First-line Chemotherapy after Adjuvant: If an oxaliplatin-based combination was administered in the adjuvant setting, the choice of an irinotecan-based combination is recommended in first-line treatment. This recommendation is stronger if the recurrence of disease occurred <12 months after stopping the adjuvant treatment.

Treatment Refractory Disease: In patients with chemorefractory disease and a good performance status, adequate organ function and a clear wish for further treatment, combination of a fluoropyrimidine with mitomycin can be considered, although the level of evidence for activity is low as well as the level of agreement between experts. The experts prefer in this situation to propose to patients a well-designed clinical trial.
Treatment of Resectable Metastatic Colorectal Patients

Resectable Liver or Lung Metastases. The perioperative or postoperative administration of a cytotoxic doublet is the recommended approach in patients with resectable metastases. There is no evidence yet that adding a biological to a cytotoxic doublet improves the outcome in resectable metastases compared to a cytotoxic doublet alone in combination with resection of the metastases.