Adjuvant Therapy for completely resected Stage II Colon Cancer

Abstract

Background

: Colon cancer is potentially curable by surgery. Although adjuvant chemotherapy benefits patients with stage III disease, there is uncertainty of such benefit in stage II colon cancer. A systematic review of the literature was performed to better define the potential benefits of adjuvant therapy for patients with stage II colon cancer.

Objectives

: To determine the effects of adjuvant therapy on overall survival and disease‐free survival in patients with stage II colon cancer.

Search methods

: Ovid MEDLINE(R) (1986‐2007), EMBASE (1980‐2007), and EBM Reviews ‐ Cochrane Central Register of Controlled Trials ( to 2007) were searched using the medical headings "colonic neoplasms", "colorectal neoplasms", "adjuvant chemotherapy", "adjuvant radiotherapy" and "immunotherapy", and the text words "colon cancer" and "colonic neoplasms". In addition, proceedings from the annual meetings of the American Society of Clinical Oncology and the European Society of Medical Oncology (1996 to 2004) as well as personal files were searched for additional information.

Selection criteria

: Randomized trials or meta‐analyses containing data on stage II colon cancer patients undergoing adjuvant therapy versus surgery alone.

Data collection and analysis

: Three reviewers summarized the results of selected studies. The main outcomes of interest were overall and disease‐free survival, however, data on toxicity and treatment delivery were also recorded.

Main results

: With regards to the effect of adjuvant therapy on stage II colon cancer, the pooled relative risk ratio for overall survival was 0.96 (95% confidence interval 0.88, 1.05). With regards to disease‐free survival, the pooled relative risk ratio was 0.83 (95% confidence interval 0.75, 0.92).

Authors' conclusions

: Although there was no improvement in overall survival in the pooled analysis, we did find that disease‐free survival in patients with stage II colon cancer was signficantly better with the use of adjuvant therapy. It seems reasonable to discuss the benefits of adjuvant systemic chemotherapy with those stage II patients who have high risk features, including obstruction, perforation, inadequate lymph node sampling or T4 disease. The co‐morbidities and likelihood of tolerating adjuvant systemic chemotherapy should be considered as well. There exists a need to further define which high‐risk features in stage II colon cancer patients should be used to select patients for adjuvant therapy. Also, researchers must continue to search for other therapies which might be more effective, shorter in duration and less toxic than those available today.

Plain language summary

Adjuvant therapy for completely resected stage II colon cancer

Colon cancer is the second most common cause of cancer deaths in the Western world. A large proportion of colon cancer patients can be cured by surgical resection alone. For those patients with lymph node positive (stage III) disease, the recurrence rate can exceed 50% and adjuvant chemotherapy has been shown to significantly reduce the risk of recurrence. In patients without lymph node involvement (stage I and II), the prognosis is quite good with surgery alone, with survival rates of 75% to 95% at 5 years. However, some patients with high risk stage II disease have a relapse rate approaching that of stage III colon cancer patients. Due to the effectiveness of systemic chemotherapy in stage III disease, a similar approach has been considered for patients with stage II disease. We performed a systematic review looking at all randomized clinical trials evaluating stage II colon cancer patients and adjuvant therapy versus surgery alone. Our review found that adjuvant therapy ‐either systemic or regional chemotherapy or immunotherapy‐ can improve the outcomes of stage II patients. In counselling individual patients, the advice given should be conditioned by the patient's age and comorbidities. In addition, the high risk features of the tumour should also be considered when contemplating the benefits of systemic therapy in patients with stage II colon cancer. Further investigation is needed to elucidate which patient and tumour factors can be used to select stage II colon cancer patients for adjuvant therapy. There also exists a need to continue to search for other adjuvant therapies which might be more effective, shorter in duration and less toxic than those available today.

Background

Colon cancer affects 50 to 60 people per 100,000 population and is the second most common cause of cancer death after lung cancer in North America. Approximately half of patients diagnosed with colon cancer ultimately die of the disease. The prognosis of newly diagnosed colon cancer patients is determined by the clinico‐pathologic stage of the disease. In stage II disease there is tumour penetration through the bowel wall beyond the submucosa, but there is no involvement of the regional lymph nodes or other distant sites (Table 1). Overall survival for this group of patients is on average 75% five years after surgery alone. However, this is a heterogeneous group of patients and some have poorer outcomes, similar to that of patients with stage III disease with a 5‐year overall survival of 50% (O'Connell 2004). High risk features include tumours which penetrate the bowel wall, adhere or invade surrounding structures, perforation, obstruction, poorly differentiated tumours, extramural vascular invasion and aneuploidy (O'Connell 2004; Galandiuk 1992). The number of lymph nodes examined can also affect prognosis and fewer than 12 examined lymph nodes may result in a worse outcome (Baxter 2005; Berger 2005). Extensive genomic alterations in colon cancer (Grady 2004) may have prognostic consequences, such as chromosome 18q mutation and microsatellite stability (Halling 1999; Graziano 2003; Popat 2005). Adjuvant therapy has been recommended for stage III colon cancer patients because it decreases relapse and mortality rates by 30% to 40% when compared to observation alone after surgery (NIH 2000). Although there is no biological reason why adjuvant therapy will not have a similar effect on stage II patients, ongoing controversy exists as to whether adjuvant therapy should also be advised for patients with stage II colon cancer. For this reason, numerous clinical trials have been performed in this setting. Results of these trials have been inconclusive to date. The explanation for this in the past has been that insufficient numbers of patients have been enrolled in a single trial to demonstrate any overall survival benefit (Buyse 2001; Gill S, 2004). In view of the low power of the individual studies, a systematic review of the literature was undertaken to collect data on stage II colon cancer patients investigated in randomized clinical trials comparing adjuvant therapy to observation alone after surgery.

1. Staging of Colorectal Cancer.

TMN	Stage	Extension To	Dukes
T1N0M0	I	Mucosa	A
T2N0M0	I	Muscularis	B1
T3N0M0	II	Subserosa/Perciloic Tissue	B2
T4N0M0	II	Visceral Per/Other Organs	B3
T1‐2N1M0	IIIA	N1 1‐3 lymph nodes	C
T3‐4N1M0	IIIB	as above	C
Tany N2M0	IIIC	N2 >3lymph nodes	C
TanyNanyM1	IV	Metastatic Disease	D

Objectives

1.‐ To determine if adjuvant therapy improves the survival of stage II colon cancer patients compared with surgery alone.

2.‐ To determine if the group of patients with poor prognostic factors (high‐risk stage II) benefit more in regards to overall survival than those without poor prognostic features.

3.‐ To determine if improvements in disease‐free survival parallel those of overall survival.

4.‐ To determine compliance with treatment and associated toxicity.

Methods

Criteria for considering studies for this review

Types of studies

Randomized controlled clinical trials (RCTs) and meta‐analyses of RCTs comparing adjuvant therapy to surgery alone.

Types of participants

Patients with completely resected stage II colon carcinoma of any age or sex.

Types of interventions

Adjuvant therapy may include any form of chemotherapy, radiotherapy and/or immunotherapy compared to observation alone immediately before or after surgery.

Inclusion criteria

Articles were selected for inclusion in this systematic review if the following criteria were met: 
 1. Randomized clinical trials or meta‐analyses of RCTs; 
 2. Include patients with completely resected stage II colon cancer. This includes studies that explore adjuvant therapy in patients with other stages of colon cancer. In addition it includes studies in which a subset of patients have rectal cancer. 
 3. Adjuvant therapy is compared with observation alone after surgery; 
 4. Data on patient survival at 5 years is available, with a median follow‐up of at least 48 months.

Exclusion criteria

1. Trials published before 1987. Buyse et al (Buyse 1988) summarized the results of randomized trials of adjuvant therapy for colorectal cancer up to that year. Major methodological concerns about the design of the trials, lack of standardization of surgery and staging and therapy not presently used would make further delving into those data uninformative. 
 2. Trials without a surgery only control arm. 
 3. Trials that do not allow for the separate calculation of survival data for stage II colon cancer patients. 
 4. Studies with less than 48 months of median follow‐up.

Types of outcome measures

1.‐ Number of deaths at 5 years. (Mortality rate at 5 years) 
 2.‐ Number of tumour recurrences at 5 years. (Recurrence rate at 5 years) 
 3.‐ Number of patients experiencing adverse events related to treatment: deaths, hospitalizations, neutropenic sepsis, grade 3 or higher gastrointestinal toxicity and other grade 3 or higher toxicities. 
 4.‐ Number of patients receiving >80% of the prescribed protocol treatment. 
 5.‐ Overall planned dose intensity of the adjuvant treatment. 
 6.‐ Number of patients who received cointervention or control patients who received adjuvant treatment. 
 7.‐ Number of patients with high‐risk factors such as T4 disease, perforation, obstruction, poorly differentiated tumours.

Search methods for identification of studies

Ovid MEDLINE(R) (1950 through March Week 1 2007), EMBASE (1980 through 2007 Week 11) and EBM Reviews ‐ Cochrane Central Register of Controlled Trials ( through 1st Quarter 2007) were searched using the medical headings "colonic neoplasms", "colorectal neoplasms", "adjuvant chemotherapy", "adjuvant radiotherapy" and "immunotherapy", and the text words "colon cancer" and "colonic neoplasms". In addition, proceedings from the annual meetings of the American Society of Clinical Oncology and the European Society of Medical Oncology (1996 to 2004) were searched for reports of recently completed or ongoing trials on stage II colon cancer patients. Personal reprint files, consultation with colleagues and reference lists of relevant studies were also searched.

CCTR, EMBASE, Ovid MEDLINE(R) search strategy

1. Colonic Neoplasms/ [MeSH] 
 2. Colorectal Neoplasms/ [MeSH] 
 3. (colon OR colorectal OR colonic).ti.[title] 
 4. (malignan$ OR Neoplas$ OR cancer OR carcinoma OR adenocarcinoma).ti. 
 5. 3 AND 4 
 6. 1 OR 2 OR 5 
 7. Chemotherapy, Adjuvant/ [MeSH] 
 8. radiotherapy, adjuvant/ [MeSH] 
 9. Immunotherapy/ [MeSH] 
 10. adjuv$.mp. 
 11. 7 OR 8 OR 9 OR 10 
 12. 6 AND 11 
 13. (clinical AND trial).ab.ti. [abstract or title] 
 14. Clinical Trials/ [MeSH] 
 15. clinical trial.pt [publication type] 
 16.random$.mp. 
 17.Random Allocation/ [MeSH] 
 18.therapeutic use$.mp. 
 19. 13 OR 14 OR 15 OR 16 OR 17 OR 18 
 20. 12 and 19

Data collection and analysis

Development of the systematic review

This review is an update of a previous guideline on the use of adjuvant therapy in stage II colon cancer produced by the Gastrointestinal Cancer Disease Site Group of the Cancer Care Ontario Program on Evidence‐based Care (Figueredo 1997) and a more recent overview published as the background for the American Society of Clinical Oncology guideline on the same topic (Figueredo 2004). While the previous reviews followed the methodology of the practice guideline cycle described elsewhere (Browman 1995), the present systematic review follows the Cochrane Collaboration methodology.

Data Collection 
 
 The three reviewers summarized articles of selected trials using a paper data collection form. Although mortality at 5 years was the main outcome of interest, data relating to disease‐free survival, patient toxicities and treatment delivery were also recorded. Significant toxicity was categorized for the present review as that causing death, hospitalization, major infection, major gastrointestinal disorder or other major events. Toxicity criteria were assessed using the World Health Organization toxicity criteria. Treatment delivery was measured by compliance with the treatment and the projected dose intensity of the regimen.

Assessment of Study Quality 
 
 Study quality was assessed according to the four types of biases described by Feinstein (Feinstein 1985). Studies comparing adjuvant therapy to observation do not have blinding for treatment assignment or outcome measurement. Given this significant deficit of adjuvant trials, we concentrated on considering other areas of biases as follows:

1.Selection bias. In considering only randomized controlled trials, a major source of selection bias is eliminated. We also recorded whether or not allocation concealment took place. Finally, we also compared patient characteristics with known or potential prognostic significance (age, sex, disease stage, comorbidity). 
 2.Performance bias: Both contamination (control group patients receiving the intervention) and cointervention (different care for the two randomized groups such as variation in follow‐up) were evaluated for each selected study. 
 3.Attrition bias: This is an important source of bias in adjuvant trials. Although some loss of trial participants is expected, we looked for a clear description of these losses. We considered losses in the range of <10% to be reasonable. 
 4.Detection bias: As outcome assessments in adjuvant trials are not blinded, we have chosen the most objective of the outcomes: whether the patient is alive or not. Other outcomes are more subjective. Disease recurrence diagnosis depends on patient contact with physician and the use of screening tests for recurrences. Although follow‐up schedule and tests are the same for treated and control patients after treatment completion, there is more intensive testing and follow‐up during adjuvant treatment and this may introduce some bias in early recurrence diagnosis.

To classify the studies according to the biases present, the following considerations will be taken into account:

Major biases: 
 ‐ No concealment of randomization 
 ‐ More than 10% of randomized patients lost for analysis 
 ‐ Control group gets treatment in >10% of patients 
 ‐ For determination of disease‐free survival, detection of recurrences during period of adjuvant treatment.

Minor biases: 
 ‐ Doubt in regard to concealment of randomization 
 ‐ Differences in baseline characteristics (p <0.05) 
 ‐ Control group gets treatment in <10% of patients 
 ‐ Cointervention favours one group

Studies will be classified as follows:

Risk of Bias Interpretation Bias Assessment

Low unlikely to alter results No major or one minor bias 
 Moderate may alter results One major or up to 3 minor biases 
 High likely to alter results One or more major or minor biases 
 
 Studies with moderate and high biases will be considered separately in a sensitivity analysis.

Analysis

Results were recorded with regards to an intention to treat analysis.

Pooled analysis was performed using the RevMan version 4.2.7software. Results were expressed as relative risks or risk ratios (RR) with 95% confidence intervals. RR less than 1.0 favours the experimental group, indicating that patients on the experimental group (adjuvant therapy) experienced fewer events than the control group (observation after surgery) (Sinclair 1994). Data were analyzed using the random effects model (DerSimonian 1986).

Meta‐analysis results were investigated for clinical and statistical heterogeneity. In the case of significant statistical heterogeneity, outlier results were looked for and cumulative meta‐analysis was used to detect trials that caused the heterogeneity.

To avoid treatment heterogeneity, the analysis of trials was performed according to the adjuvant treatments given. For example, patients receiving common chemotherapy such as 5‐fluorouracil (5‐FU) plus either levamisole or leucovorin, were considered together. The same analysis was carried out for patients receiving regional portal vein infusion chemotherapy or adjuvant immunotherapy.

Similar analyses were performed for disease recurrence, toxicity and treatment compliance.

The issue of Hazard Ratio (HR) versus Relative Risk (RR): 
 Hazard ratios typically give a better description and measure of the difference in survival between two randomized groups. Hazard ratios also describe the total time experience of the groups rather than at a single time post. However, this single point measurement relates to the total number of patients over the total time of the trial and it can give a good estimate of the difference. Our group selected RR over HR for the purposes of this meta‐analysis given the significant difficulty in obtaining HR for many trials. We thought that by selecting out trials for review according to this criterium, this would result in substantial bias. We did obtain HR's for some of the larger adjuvant trials, including those using 5‐FU plus leucovorin and/or levamisole. We found that the HR's were in fact very similar to RR's.

Results

Description of studies

The original literature search identified 1732 abstracts. An initial screening by title reduced the number of relevant reports to 312. Of these 312 abstracts, 159 abstracts appeared to deal with curatively resected stage II colorectal cancer patients who had been randomized to adjuvant therapy or observation. Upon review of the full publications, 117 articles were considered potentially relevant according to the inclusion criteria. These included 84 publications reporting on clinical trials of which 32 were accessories to the main trials. Furthermore, there were 33 systematic reviews, with most of these including meta‐analyses within them. Of the meta‐analyses, 4 contained data to corroborate or update trials also identified using the search strategy. Figure 1 summarizes the overall results of the search strategy employed. Five additional publications were found in personal files (Gray 1988; Gray 1991; LIMAG 1997;Schlag 1990; Harris 2000). Figure 1 (see Additional Figures) depicts the trials classified according to the adjuvant therapy received as well as the total number of patients and the percentage of patients with stage II disease. Tables 2 to 6 provide more details about the trials reviewed, showing both the wide range of therapies tested and the forthcomings in reporting significant data.

1.

Table 1.

Included studies 
 Twenty‐five publications reporting on 33 trials and 17 meta‐analyses are shown in the Table of Included Studies. We only included randomized phase III clinical trials (RCTs) and meta‐analyses of RCTs with a control arm consisting of surgery, and other arms consisting of adjuvant therapy. The validity of the trial results is measured by the risk of bias. This risk was determined to be low in 26 studies, moderate in 9 and high in none.

The participants in the trials were patients with completely resected colon (n=7731) or rectal (n=7600) cancers. There were 4 trials reported that included stage II colon cancer patients alone (n=3384). Patients were excluded if there was any evidence of previous malignancy, prior treatment with chemotherapy or radiation, abnormal blood counts or liver or renal function or severe co‐morbidities. Most trials were performed in one country (n=16), although some were international (n=16) or data from single institutions (n=3). Patient accrual in these trials occurred between the years 1976 and 2002. The number of participants in each study ranged from 78 to 1738 and had a median age of approximately 62 years and a male/female ratio of 1.11. The median follow‐up of patients was more than 60 months in 21 trials.

The interventions investigated included a variety of treatments: systemic chemotherapy (n=16), regional chemotherapy (n=6) and immunotherapy (n=7). A description of the treatment(s) is given for each trial. Six trials had 2 treatment arms.

All included studies had risk ratios and 95% confidence intervals calculated for all stage II colon cancer patients. 5‐year overall survival and disease free survival were recorded. In view of the variety of treatments tested, a pooling of trial results was deemed to be reasonable only within certain treatment groups. These groups were determined a priori and included 5‐fluorouracil (5‐FU) plus nitrosoureas, mitomycin C and fluoro‐pyrimidines, 5‐FU plus levamisole and/or folinic acid, portal vein infusion (PVI) of 5‐FU alone or with mitomycin C and specific versus non‐specific immunotherapy.

In addition to the randomised clinical trials, 19 meta‐analyses were also included. Two of these include primary results of 10 RCTs (IMPACT B2 1999; Glimelius 2005). Another meta‐analysis provides data on stage II patients not available in the original publication (Mamounas 1999). The other meta‐analyses utilize either published reports or directly obtained individual patient data. These latter trials investigate a variety of factors in regard to the benefits and harms of the adjuvant therapies and ways to predict prognosis.

Excluded Studies

Clinical trials were excluded if there was no observation arm or separate data on DFS or OS for stage II patients or if they were accessories to the main trial publication. For completeness of the review the available data of excluded RCTs with an observation arm are depicted in Tables 1 through 6 together with data from the included studies. Meta‐analyses were excluded if they were similar to previous publications by authors of the same group. All of the above information is listed in the Table of Excluded Studies.

Risk of bias in included studies

Twenty‐five publications reporting on 33 trials and 17 meta‐analyses are shown in the Table of Included Studies. All included studies were either randomized phase III clinical trials (RCTs) or meta‐analyses of RCTs, with a control arm consisting of surgery, and other arms consisting of adjuvant therapy. The validity of the trial results was measured by the risk of bias. This risk was determined to be low in 26 studies, moderate in 9 and high in none.

Effects of interventions

Adjuvant systemic chemotherapy (Figure 2, Figure 3, Figure 4 & Figure 5 in Additional Figures)

2.

Table 2.

Adjuvant Systemic Chemotherapy with Early Regimens

3.

Table 2b.

4.

Table 3.

Adjuvant Systemic Chemotherapy with commonly used regimens

5.

Table 3b.

This form of adjuvant treatment had the largest number of trials. The trials were grouped into categories of early chemotherapy regimens, more contemporary chemotherapy regimens, and those using miscellaneous treatments no longer in use (intraluminal 5‐FU, radiation to the liver and razoxane). 
 
 Early chemotherapy trials tested adjuvant therapy with a fluoro‐pyrimidine plus either a nitrosourea or mitomycin C (Figure 2(see Additional Figures)). Four trials tested a combination of 5‐FU and nitrosoureas, and one of these (Wolmark 1988) was the first to demonstrate significant benefits in DFS and OS for combined stage II and III colon cancer patients. There was no significant interaction between treatment effect and stage. However, two other similar trials did not confirm these findings (Panettiere 1988; Abdi 1989) and 10‐year results of the positive trial found that the apparent benefits disappeared over time (Smith 2004). All of these trials reported separate data for stage II patients. A pooled analysis revealed a RR for death of 0.93; (95% CI 0.71, 1.10). Compliance with treatment (>80% of planned dose) was noted in 41% of patients in the only trial describing it (Wolmark 1988). The same trial also described 3 toxic deaths due to acute leukemia and severe GI toxicity in 15% of patients. Renal failure occurred in one patient each of 2 trials (Abdi 1989; Wolmark 1988).

The other group of 9 trials, using a variety of fluoro‐pyrimidines ‐mostly administered orally and often combined with mitomycin C‐, showed benefits in both DFS and OS when considering all patients, and one trial (Watanabe 2006) demonstrated significant benefit in DFS for stage II patients. The published results did not allow for pooling of results in stage II patients (Figure 2 and Figure 3(see Additional Figures)). However, the use of oral fluoro‐pyrimidines as adjuvant therapy in Japan was the topic of 4 pooled analyses (Sakamoto 1999; Sakamoto 2001; Sakamoto 2004; and Sakamoto 2005). The 2001 report by Sakamoto et al. was excluded since only one of these trials compared adjuvant therapy with observation after surgery (Ito 1996). This trial and two others (Yasutomi 1997; the other not reviewed here), which tested carmofur (HCFU) as adjuvant therapy, either alone or with mitomycin C, were pooled in an individual patient data meta‐analysis (Sakamoto 2005). When all patients were considered, there was a significant benefit for those receiving treatment: DFS of 77% vs. 71% (p=0.003); OS 80% vs. 76% (p=0.043). For patients with stage II colon cancer, the absolute difference favouring treatment for DFS was 6.6% (p=0.032) and for OS was 6.8% (p=0.086). The two other pooled analyses, using similar methodology, investigated the value of adjuvant oral HCFU or 5‐FU for one year on the outcome of patients with both colon and rectal cancer. One of the analyses (Sakamoto 2004) considered trials where the randomization was performed centrally. In patients with stage II disease receiving adjuvant therapy there was a hazard ratio reduction (21%±SE 9%) in both colon and rectal cancer patients. In a previous analysis (Sakamoto 1999), where patient randomization was by the closed envelop method, there was no benefit for stage II patients receiving adjuvant oral fluoro‐pyrimidines. Compliance with planned drug doses was 71% and 76%, and toxicity was reported as mild in one trial (CCCSG Japan 1995). In another trial (Watanabe 2006) compliance with drug doses was 59% for mitomycin C, 93% for 5‐FU and 59% for HCFU.

The contemporary chemotherapy consisted of 5‐FU plus either levamisole and/or folinic acid (Figure 4 and Figure 5(see Additional Figures)).

Six individual trials have tested levamisole alone or combined with 5‐FU, and all except one (Dahl 2003) have reported separate data for stage II patients (Figure 4). None of the individual trials have demonstrated a benefit for adjuvant therapy in stage II patients. Pooled results from 4 trials reporting on stage II patients (Windle 1987; Laurie 1989; Moertel 1995 and Taal 2001) showed a RR for mortality of 0.88; (95% CI 0.71, 1.10). Compliance with 5‐FU + levamisole was 80% at 6 months and 69% at 12 months in one trial (Taal 2001) and in another trial (Moertel 1995) 30% of patients discontinued treatment, mostly due to severe nausea, at a median of 5 months. No deaths were reported in either trial but later reports of a severe leukoencephalophaty (Hook 1992) which affects up to 5% of treated patients (Figueredo 1995), led to the abandonment of levamisole as adjuvant treatment.

Thirteen trials testing the combination of 5‐FU plus folinic acid, with or without levamisole, have been reported either individually or in pooled analyses involving 1016 (IMPACT B2 1999) and 1520 (Glimelius 2005) colon cancer patients (Figure 4 and Figure 5(see Additional Figures)). Separate data for stage II patients was available in two individual trials (Francini 1994; Zaniboni 1997) which are also included in the pooled analysis of 5 trials (IMPACT B2 1999). Neither the individual trials nor the meta‐analyses demonstrated significant benefits for adjuvant therapy. This adjuvant therapy was well tolerated: 85% to 98% of patients received the planned doses of the drugs (IMPACT B2 1999). Toxicity did not include deaths but severe nausea/vomiting (4%), stomatitis (11%) and diarrhea (8%) did occur. Glimelius et al. (Glimelius 2005) also noted that 67% of patients received the planned number of treatment courses of 5‐FU ± levamisole ± folinic acid. In a pooled analysis of 7 trials using 5‐FU + levamisole or 5‐FU + folinic acid (Sargent 2001) it was noted that grade 3 or higher gastrointestinal toxicity was more common with 5‐FU + levamisole than with 5‐FU + folinic acid. With regards to leucopenia, the reverse was true. However, these comparisons were carried out between rather than within trials. Of interest, it should be noted that patients >70 years of age appeared to tolerate systemic adjuvant treatment similar to their younger counterparts.

A later pooled analysis of 3302 patients with stage II and III colon cancer investigated in 7 clinical trials comparing observation to adjuvant therapy with 5‐FU combined with levamisole or folinic acid has been performed using individual patient data (Gill 2004). Applying a Cox proportional hazard regression model, the authors estimated 5‐year DFS and OS according to T stage, nodal status and histological tumour grade. As expected, the larger benefits were observed in patients with node positive disease. In patients without nodal involvement, regardless of T stage and tumour grade, the absolute percent benefit in DFS for treated versus observed patients was 4% to 6%, with wide confidence intervals. The OS benefit for these stage II patients was smaller than the DFS, and the results were consistent across the various age and tumour grade groups.

Regional Chemotherapy (Figure 6 and Figure 7(see Additional Figures))

6.

Table 4.

Adjuvant regional chemotherapy

7.

Table 4b.

Given that most recurrences of colon and rectal cancer occur within the area of the portal vein system, particularly the liver, regional chemotherapy delivers a higher concentration of drug to the portal/hepatic circulation. Several trials evaluated this delivery route. Enthusiasm for this approach followed the significant survival benefits, especially in stage II patients, reported by Taylor et al. (Taylor 1985). This regional chemotherapy treatment was well tolerated and delivered one week after surgery.

At least 13 other trials have attempted to reproduce these early results, but only one has demonstrated a benefit in DFS (Wolmark 1990). All these trials used 5‐FU except one using floxiridine (Schlag 1990). Some of these trials also tested the hypothesis that either heparin (Gray 1987; Fielding 1992; Nitti 1997) or urokinase (Wereldsma 1990), which was used to maintain patent the portal vein access, may have some anti‐tumour effect. However, no benefit was detected in these studies. Four trials also intensified the chemotherapy by adding mitomycin C to 5‐FU, and, again, no benefit was demonstrated. Four published trials (Wolmark 1990; Beart 1990; Fielding 1992 and Laffer 1995) reported separate data for stage II patients.

Considering that the effect of chemotherapy by portal vein infusion is small and that the reported trial samples have a small to moderate number of patients, two meta‐analyses have been published. The initial one (Gray 1991) pooled results of 6 trials of chemotherapy by portal vein infusion and found significant benefits in overall survival, but did not give separate data for stage II patients. The Liver Infusion Meta‐analysis Group (LIMAG 1997) used individual patient data obtained from 10 trials (9 excluding the original Taylor's study) and had a total of 3088 patients, most of whom had colon cancer. The odds of death were reduced for the treated patients by 11%±5% (p=0.04) for the entire sample and by 5%±2% (p=0.01) for Dukes' A and B stages combined.

Three trials investigated peritoneal chemotherapy (Figure 7 ‐ additional figures). One of them only reported preliminary findings (Graf 1994). Another trial combined intraperitoneal chemotherapy combined with intravenous chemotherapy for a prolonged period of time and was therefore included in the 5‐FU plus folinic acid trials (Scheithauer 1995). The only trial that truly tested the intraperitoneal route with mature results was the trial reported by Vaillant et al. (Vaillant 2000). There was a trend towards improved outcomes, including overall survival in stage II patients.

A single study has investigated chemotherapy via hepatic artery infusion (Sadahiro 2004) (Figure 7). The reported results for the entire sample, including colon cancer patients with stages I, II and III, were significant for both DFS and OS; and there was a trend noted for improved overall survival in stage II colon cancer patients.

Immunotherapy (Figure 8 and Figure 9(see Additional Figures))

8.

Table 5.

Adjuvant immunotherapy

9.

Table 5b.

The interest in non‐toxic, immune‐stimulating treatments to facilitate rejection of residual cancer cells after apparent complete surgical excision has been widely investigated in colorectal adjuvant clinical trials. The immune therapy tested was categorized as being either specific, using altered tumour cells or antibodies, or non‐specific, using known immune stimulants such as BCG, interleukin‐2, interferon and others.

Our search identified 10 trials of non‐specific immunotherapy (Figure 8(see Additional Figures)): three trials tested BCG (Prohm 1988; Wolmark 1988; Abdi 1989), 3 tested interleukin‐2 (Nichols 1992; Nichols 1993, Brivio 2006), and one each tested PAPUA ‐polyadenyl‐polyuridylic acic‐ (Lacour 1992), PSK ‐a protein‐bound polysaccharide‐ (Nio 1992) and PBG KP‐45 ‐propionibacterium granulosum ‐ (Isenberg 1994).

Two trials reported conflicting results with the use of BCG. One trial (Wolmark 1988) showed benefit in overall survival (51% vs. 56%; p=0.02) whereas the other (Abdi 1989) reported poorer outcomes for treated patients. Of the 3 trials testing interleukin‐2, only one (Monga 2006) reported overall survival results favouring treatment. This benefit was not observed in patients with stage II disease. The only other trial reporting survival data (Isenberg 1994) tested PAPUA and demonstrated a significant trend favouring treatment and especially in stage II patients.

Among the 7 trials of adjuvant specific immunotherapy (Figure 8 and Figure 9(see Additional Figures)), four have tested the combination of altered tumour cell plus BCG administered by intradermal inoculations over 1 to 18 months (Vermorken 1999; Harris 2000; Hoover 1993; Gray 1989). Only one individual trial (Vermorken 1999) found an improvement for treated patients compared to untreated controls. This benefit was particularly prominent and significant in stage II patients. Pooling the results of 3 of these trials (excluding Hoover 1993) showed a RR for death of 0.91 (95% CI 0.70 ‐ 1.18). A critical review of 3 trials excluding Gray, 1989 (Hanna 2001) performed one meta‐analysis for all randomized patients and another for a selected group of "analyzable patients". The latter group required definite protocol eligibility and vaccines with >70% of viable cells and a dose of 107 tumour cells per immunization. The meta‐analysis of all 723 patients demonstrated a significant improvement in DFS (Odds Ratio 0.75; CI 0.56, 1.00). The pooled analysis of analyzable patients also demonstrated a significant delay in recurrences (p=0.028) which also occurred in stage II patients (80% vs. 63%; p=0.032). Most of the benefit was seen in patients who developed an induration of >5 mm at the site of the inoculum. Based on these findings, the investigators concluded that specific immunotherapy is an important adjuvant modality, particularly valuable for stage II patients. The different results of the 3 trials was attributed to variability in the immunogenicity of the vaccine, status of the host immune system and whether patients developed systemic cell‐mediated immunity. These 3 trials prepared the vaccine by initial enzymatic dissociation of tumour fragments, freezing the cells in liquid nitrogen for 30 days, thawing, washing and then irradiating the cells which were then added to BCG prior to administration. The fourth trial of specific immunotherapy (Gray 1989) prepared the tumour cells similarly, but used Vibrio‐cholera neuraminidase prior to freezing and irradiation. The tumour cells were not mixed with BCG, but instead they were applied to the site where BCG was previously inoculated. This trial did not report benefit for the treatment in spite of 18 months of immunotherapy.

Two trials reported initial results on the use of edrecolomab adjuvant therapy (Hartung 2005; Colacchio 2004). Edrecolomab is an antibody against glycoprotein 17‐1A found on the surface of colorectal tumour cells. It was successfully tested in patients with stage III colorectal cancer (Riethmuller 1994). The 2 trials reported results after median follow‐up times of 34 and 44 months, respectively. No significant difference in survival between treated and observed patients was noted. A meta‐analysis of the results at this early stage showed a RR for death of 0.88; 0.65, 1.20). Longer follow‐up is required before a final judgement can be made about this form of adjuvant therapy.

Another specific antibody 105AD7, mimicking the glycoprotein CD55 on the tumour cell surface, was tested together with BCG administered intradermally or with alum administered intramuscularly (Ullenhag 2006). These vaccinations were continued for 2 years. The available publications reported only immune responses, and did not report survival outcomes.

Other treatments (Figure 10(see Additional Figures))

10.

Table 6.

Miscellaneous therapies

Four placebo‐controlled trials tested histamine‐2 receptor antagonists (H‐2RA) in the adjuvant treatment of resected colorectal cancer (Svendsen 1995; Kelly 1999; Nielsen 2002; Kapoor 2005). Although the exact mechanism of action is unknown, cimetidine demonstrated cytotoxicity to colorectal cancers and stimulation of lymphocyte function in vitro. Other investigators noted improved survival for patients with rectal cancer containing lymphocytic infiltrates. Two of the trials administered cimetidine or ranitidine for 24 to 60 months whereas the other two used a higher dose of cimetidine or famotidine for 5 to 7 days. These trials showed a trend toward better survival for treated patients but none of them reached statistical significance. It was noted that the treatment effect was found only in patients who did not receive blood transfusions in the peri‐operative period (Nielsen 2002).

One report described a proposal for an adjuvant trial involving the COX‐2 inhibitor, Rofecoxib (Pendlebury 2003). It has been suggested that COX‐2 inhibition leads to a reduction in cancer cell invasion and metastasis. This VICTOR trial was stopped when serious cardiovascular side‐effects of the COX‐2 inhibitors were described. A recent report indicates that the incidence of cardiovascular events was significant in patients on rofecoxib compared to those receiving a placebo (estimated relative risk 2.66; p=0.04) (Kerr 2007).

Hazard Ratios versus Relative Risk Ratios 
 
 We calculated hazard ratios for several of the trials included in the meta‐analysis to illustrate that there was very little difference between calculated hazard ratios and relative risk ratios. See Figure 11 and Figure 12 in Additional Figures.

11.

Table 7.

12.

Table 8.

Other published meta‐analyses

Following the ground‐breaking pooled analysis by Buyse and collaborators (Buyse 1988a) the next two meta‐analyses continued to investigate results of colorectal cancer adjuvant therapy for all stages together. In one study (Zalcberg 1996) it was found that the planned total dose of 5‐FU during the first 3 months of treatment predicted for the size of survival benefit. The odd ratios for death were 0.71 for >10 gm, 0.79 for 8 to 10 gm, 0.93 for <8 gm and 1.04 for oral 5‐FU. Further, the value of adding levamisole was questioned because its benefit disappeared when the 5‐FU dose was considered. Dube (Dube 1997) performed a meta‐analysis of 39 clinical trials published between 1959 and 1993 and found the odds ratio for death predicted an absolute 5‐year overall survival benefit of 5% for colon cancer and 9% for rectal cancer.

A guideline developed by the Cancer Care Ontario, Provincial Guideline Initiative, Gastrointestinal Disease Site Group was the first to focus on stage II colon cancer (Figueredo 1997). It pooled results of 11 trials published after 1987 comparing adjuvant treatment versus observation after surgery. It found that neither systemic nor portal vein infusion chemotherapy provided significant survival benefits. Therefore, it recommended no standard adjuvant therapy for stage II colon cancer but discussion of potential benefits in patients at high risk of recurrence and participation in trials with an observation arm.

In 1999, two meta‐analyses investigated the value of adjuvant therapy in stage II colon cancer. The individual patient data meta‐analysis of 5 RCTs using 5‐FU plus folinic acid demonstrated the 5‐year absolute OS benefit of treated patients was 2% (p=0.398) (IMPACT B2 1999). The second study (Mamounas 1999) pooled results of 4 consecutive adjuvant trials in colon cancer by the National Surgical Adjuvant Breast & Bowel Project (NSABP). Two of the trials tested adjuvant therapy versus observation (Wolmark 1988 and Wolmark 1990) but the two other trials compared two adjuvant therapies. The results of the analysis suggested a 30% reduction in mortality for stage II patients. Further, this finding was supported by the lack of interaction between treatment and stage described in each of the individual trials.

The Cancer Care Ontario Guideline for the adjuvant treatment of stage II colon cancer was updated (Figueredo 2004). Eighteen trials and several meta‐analyses were considered and the pooled results for the different forms of adjuvant therapy remained inconclusive. The recommendations given, together with those of the accompanying article (Benson 2004), discussed the uncertainty of the benefits and suggested these be discussed with patients, particularly those at high risk of recurrence. Another meta‐analysis (Wan 2005) included the 4 trials considered by Mamounas, the IMPACT‐B2 data, that of Taal (Taal 2001) and also included results of a non‐randomized trial. The results showed a significant benefit of adjuvant therapy for stage II colon cancer.

In a statistical analysis of the question of adjuvant therapy in stage II colon cancer Buyse (Buyse 2001) noted that meta‐analysis may not be the most efficient tool searching for true benefits of treatment as it considers only a portion of the trial population. Either the lack of interaction between treatment and stage or an individual patient data with subgroup analysis may give the best answer short of a massive clinical trial. A group of investigators (Sargent, 1999) had already begun to collect individual patient information in a large data‐base of contemporary adjuvant systemic chemotherapy trials using 5‐FU plus either levamisole or folinic acid. The data was used then to determine patients who may benefit from the adjuvant treatment (Gill 2004). A Cox proportional hazard regression model was used to predict DFS and OS considering as variables treatment, age, sex, tumour location, T stage, nodal status and histological grade of malignancy. Treatment benefits were observed across age, sex, tumour location, T stage and grade. The benefit in 5‐year DFS for treated patients, compared to observation after surgery, was 4% to 6% regardless of T stage and grade. The 5‐year OS benefit for adjuvant therapy ranged from 2% to 4% regardless of age and grade, with a 1% larger benefit for T4 compared to T3 stage. Neither of the values were individually significant. Moertel 1995 had already performed a multivariate analysis with data from the reported trial as well that from Laurie 1989, where the end‐points were DFS or OS and the variables were age ±61 years, sex, delay of treatment after surgery (±21 days), adhesion or invasion to other organ, obstruction, perforation and location of the primary tumour within the colon. Neither of the investigated variables had an interaction with the 5‐FU plus levamisole adjuvant treatment on disease recurrence or survival.

Another study (Sargent 2005) tested whether DFS at 3 years could be used as a surrogate end‐point for 5‐year OS. The correlation obtained for the two was 0.89. This correlation was explained because 80% of recurrences occur within 3 years and 91% of these recurrences lead to death within 5 years. This correlation was found for both the treatment and control arms of the trials. The author cautioned, however, that marginally significant benefits in DFS may not translate into significant advantages in 5‐year OS.

One of the initial studies of the above group (Sargent 2001) explored whether adjuvant chemotherapy benefits extended to elderly patients. For this pooled analysis 7 clinical trials involving 3351 patients were tested for death and toxicity according to 4 age groups (<50, 51‐50, 61‐70 and >70 years). No significant interaction was detected for overall survival (p=0.61) or for recurrence (p=0.33). Deaths without evidence of recurrence, however, were more common with increasing age (2% in patients <50 years old; 13% for those >70 years old). Toxicity was not related to age except for severe leucopenia. There was no separate data for treatment effect on stage II patients.

Two other publications have examined the outcomes of African‐Americans compared to Caucasians in adjuvant trials in attempting to explain the poorer prognosis for African‐American patients in the population at large. While the first study (Dignam 1999) including patients in 5 trials lead by NSABP found an 8% higher recurrence rate (p=0.27) and a 21% higher risk of death (p=0.004) for African‐Americans, the other study (McCollum 2002) investigating a large single trial of several 5‐FU adjuvant regimens did not find differences in either disease‐free or overall survival. In the second study, however, the toxicity for African‐Americans was significantly lower than for Caucasians, including diarrhea, nausea, vomiting, stomatitis and overall (p<0.005).

Discussion

To better understand the extensive literature on the heterogeneous adjuvant treatments for stage II colon cancer, the discussion will be divided into three sections: systemic chemotherapy, regional chemotherapy and immunotherapy.

Systemic chemotherapy 
 The systemic administration of chemotherapy has been the most intensely investigated and 5‐FU and its analogs the main components of the regimens. The nitrosoureas are no longer used due to toxicities which include acute leukemias (Wolmark 1988). Mitomycin C has remained active in Japanese adjuvant trials, mainly as an induction agent followed by 5‐FU or analogs (HCFU or carmofur, UFT or uracil plus ftorafur). These regimens have demonstrated in meta‐analyses (Sakamoto 2005 and Sakamoto 2004) significant benefits for stage II colon cancer in DFS (6%; p=0.003) and OS (4%; p=0.043). 
 Most trials in the Western world have used 5‐FU by bolus intravenous injection, only recently substituted by continuous intravenous infusion or by the oral analog capecitabine (5‐deoxy‐5‐fluorouridine or 5'DFUR). Initially 5‐FU was combined with levamisole and later with folinic acid. The benefit of levamisole was questioned early in a meta‐analysis (Zalcberg 1996) when the dose of 5‐FU was found the determinant factor in improving survival. Further, its delayed toxicity which included a leukoencephalopathy and the demonstration of the value of folinic acid led to its abandonment. In the reviewed trials only 5‐FU by IV bolus injection, combined with levamisole and/or folinic acid has been used. These regimens of adjuvant treatment are considered contemporaneous because have been standard therapy for stage III colon cancer until recently and have been advocated for stage II high‐risk patients. The definite beneficial advantage of adjuvant therapy with these regimens in stage III have not been conclusive for stage II disease patients. A meta‐analysis of 5‐FU plus folinic acid adjuvant trials (IMPACT B2 1999) revealed minimal increase in OS (absolute risk difference 2%; p=0.08). A more ambitious pooled analysis (Gill 2004) attempted to determine benefits for subgroup of patients with different stage of disease, histological malignancy grade and age. In stage II patients the benefits in DFS and OS were small (4% to 6%) and not statistically significant. A more recent massive adjuvant trial with 5‐FU±folinic acid±levamisole including over 3000 patients (Gray 2004) confirmed this non‐significant benefit. It is probable that pooling the results of this trial with the data‐base used by Gill will provide statistically significant results albeit the benefit will still be in the range of 1% to 6%. This information as well as early results from a trial 5‐FU+ folinic acid alone or combined with oxaliplatin opened again the debate of advising adjuvant therapy for stage II colon cancer (Midgley 2005; Sobrero 2006; Andre 2006; Monga 2006). However, an update of the oxaliplatin trial (De Gramont 2007) showed again significant benefits for stage III patients but not for stage II patients. For this latter group of patients the DFS risk difference was 3.8% (p=0.258) and for OS 0.1% (p=0.996).

Regional chemotherapy 
 The attempt to increase drug concentration in the organs of the portal vein system, the most affected by first site of recurrence, chemotherapy by portal vein, peritoneal and hepatic artery infusion have been tested. 
 Ten published adjuvant trials have administered chemotherapy by portal vein infusion for one week immediately after surgery. Only 4 trials provide separate data for stage II patients. A meta‐analysis of the trials revealed no benefit for this treatment. A previous meta‐analysis (Gray, 1991) pooled results of 6 trials but did not report separate data for stage II patients. A more inclusive study (LIMAG, 1997) used individual patient data and included 4000 patients participating in 10 trials. It demonstrated a small but significant decrease in the risk of death for patients with stages I and II colon cancer combined (5%±2%; p=0.01). 
 Chemotherapy by peritoneal infusion (Vaillant 2000) as well as by hepatic artery infusion (Sadahiro 2004), as single modalities, have also reported to induce a trend toward improved survival in stage II patients. 
 Regional chemotherapy as an adjuvant for colon cancer has been mostly abandoned because the technical difficulties of the procedures and similar results from systemic chemotherapy.

Immunotherapy 
 Attempts to boost the immune response against residual cancer cells after apparent complete surgery have been widely investigated. The non‐specific immune stimulation has utilized a variety of agents. BCG was initially investigated alone (Prohm 1988; Wolmark 1988; Abdi 1989) or combined iwith chemotherapy (Panettiere 1988). The only study demonstrating a persistent benefit in survival, even after 12‐year follow‐up did not show an improvement in DFS (Wolmark 1988; Smith 2004). Other non‐specific immune stimulants (Interferon, lnterleukin‐2, PAPUA, etc.) have been tested but separate data for stage II colon cancer patients are not available except for one study (Isenberg 1994). This small study, using propiobacterium granulosum preparation KP‐ 45 intravenously in the perioperative period observed a significant decrease in mortality and recurrence for treated patients with stage II disease. These results should be considered preliminary and would require confirmation in a larger trial. 
 Specific immunotherapy has been included in 4 trials using altered tumour cells plus BCG and 2 trials using the antibody edrecolomab. Among RCTs testing active specific immunotherapy (Hoover 1993; Gray 1988; Vermorken 1999; Harris 2000) only one has demonstrated a survival benefit for stage II colon cancer (Vermorken 1999). A critical review and pooling the results of three of those trials (Hanna 2001) has demonstrated overall survival benefits, particularly when including only patients who were eligible and received and responded to active vaccination. The recurrence‐free survival in stage II patients favoured treatment (80% vs 63%; p=0.0032). No data is given for overall survival. Although this active specific tumour cell vaccination appears to have a larger benefit than other forms of treatment and requires only a short duration of treatment witn only minor side‐effects, it is a complicated procedure requiring a dedicated laboratory to obtain an active vaccine and a responsive host. An easier form of specific immunotherapy utilizes an antibody to a glycoprotein of the cancer cell surface. Edrecolomab, produced commercially, was found effective in reducing risk in stage III colorectal cancer (Riethmuller 1994) and was then tested in stage II colon cancer (Hartung 2005; Colacchio 2004). Neither of these latter trials have shown improved outcomes for treated patients but follow‐up has been shorter than the usual 5‐year mark. One of the trials has included the measure of various prognostic factors in tumour tissues which may be relevant for future selection of patients and therapeutic approaches.

Authors' conclusions

Implications for practice.

In a statistical perspective about whether patients with stage II colon cancer should receive adjuvant therapy (Buyse 2001) it is stated that to investigate benefits, four statistical approaches can be used short of a massive RCT: 1) consider overall treatment effects regardless of the stage; 2) estimate benefits for stage II patients only in a meta‐analysis; 3) compare the relative benefits among patients with stage II and stage III disease; and 4) perform tests of interaction between treatment and stage. All these tests have been done as well as a massive clinical trial. A benefit has been demonstrated by most procedures but the benefit is small even when statistically significant. The lack of major treatment effect in stage II colon cancer can be attributed to the good survival of patients treated with only surgery (75% to 80% at 5 years) and the competing mortality from causes other than colon cancer (Buyse 2001). This phenomenon had already been reported (Nauta 1989). These investigators compared the survival of patients with stage II colon cancer participating in 2 Gastro Intestinal Tumour Study Group (GITSG) trials to age, sex and race adjusted cohorts of healthy people. Further, they subdivided both cohorts into four age groups (<50, 50‐60, 60‐70 and > 70 years). The survival of patients with stage II colon cancer was similar for all 4 age groups. This has also been observed in the pooled analyses of Sargent and Gill (Sargent 2001; Gill 2004). The survival of the healthy cohorts, however, decreased steadily from the <50 to the >70 year groups. The difference between expected survival for the four age groups was 23% for those <50 years and declined to almost nil for those >70 years. As most patients with stage II colon cancer are over 60 years, it is clear that the potential benefit in survival must be small as has been found in all other trials and pooled analyses.

It is also important that the definition of stage II colon cancer be restricted to those with at least 10 to 13 lymph nodes examined (Baxter 2005; Berger 2005) and to those with T4 disease which have been resected in block and who did not have a perforated tumour. Otherwise, it is possible that lax criteria for stage II disease may include patients at higher risk of undetected lymph node metastases or spillage of cancer cells in the peritoneal cavity at surgery. These and other potential prognostic factors have bee described (Nathanson 1994; Hase 1998).

Patients with resected colon cancer should be staged appropriately, restricting the label stage II for those who had no tumour involvement in at least 10 to 13 regional lymph nodes, had real en block resection for T4 tumours and had no evidence of tumour perforation into the peritoneal cavity. These patients have an excellent prognosis from the cancer and, if over age 60 years, unlikely to benefit from adjuvant therapy. Patients less than 60 years old with co‐morbidities should be assessed in regard to the potential to survive the next 5 and possibly 10 years. If their probability of survival is lesser than a corresponding age cohort, then adjuvant therapy might not be beneficial.

All these factors and uncertainty of the benefit of adjuvant therapy must be clearly explained to patients stressing that to save one life treatment must be given to 17 to 100 patients. Or inversely that 16 to 99 patients will receive treatment for no gain. If systemic chemotherapy is planned it may include 5‐FU + leucovorin or oral capecitabine for 6 months. There is not enough evidence yet to support the use of additional drugs, such as oxaliplatin in patients with stage II colon cancer. Toxicity occurrence must also enter this discussion as potential non‐toxic therapies may become available and this small gain may then be more acceptable.

Implications for research.

Clinical adjuvant trials involving patients with stage II colon cancer must be restrictive in their definition of the stage. Further investigation of healthy cohorts are needed as the ones cited by Nauta are from the 1980's. Other prognostic factors may have to be included in the selection of patients considered at high risk such as genetic changes. Randomization of patients to an observation arm is still ethical if patients are well informed.

What's new

Date	Event	Description
14 May 2008	Amended	Converted to new review format.

History

Protocol first published: Issue 3, 2005
 Review first published: Issue 3, 2008

Date	Event	Description
1 May 2008	New citation required and conclusions have changed	Substantive amendment

Data and analyses

Comparison 1. Deaths.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Deaths for Stage II Colon Cancer ‐ By subcategory	28	18975	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.91, 1.02]
1.1 Adjuvant Systemic Chemotherapy	16	7097	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.87, 1.05]
1.2 Adjuvant Regional Chemotherapy	25	8775	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.89, 1.08]
1.3 Adjuvant Immunotherapy	8	3103	Risk Ratio (M‐H, Random, 95% CI)	0.90 [0.77, 1.06]

1.1. Analysis.

Comparison 1 Deaths, Outcome 1 Deaths for Stage II Colon Cancer ‐ By subcategory.

Comparison 2. Recurrences.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Recurrences for Stage II Colon Cancer ‐ by subcategory	18	11772	Risk Ratio (M‐H, Random, 95% CI)	0.84 [0.77, 0.92]
1.1 Adjuvant Systemic Chemotherapy	18	8642	Risk Ratio (M‐H, Random, 95% CI)	0.83 [0.75, 0.91]
1.2 Adjuvant Regional Chemotherapy	2	456	Risk Ratio (M‐H, Random, 95% CI)	0.77 [0.46, 1.30]
1.3 Adjuvant Immunotherapy	6	2674	Risk Ratio (M‐H, Random, 95% CI)	0.88 [0.70, 1.11]

2.1. Analysis.

Comparison 2 Recurrences, Outcome 1 Recurrences for Stage II Colon Cancer ‐ by subcategory.

Comparison 3. Systemic Chemotherapy Deaths.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Systemic Chemotherapy Deaths	15	4457	Risk Ratio (M‐H, Random, 95% CI)	0.99 [0.89, 1.11]
1.1 5‐FU and Nitrosourea Chemotherapy Regimens	3	470	Risk Ratio (M‐H, Random, 95% CI)	0.97 [0.71, 1.33]
1.2 Miscellaneous Treatments	1	374	Risk Ratio (M‐H, Random, 95% CI)	1.19 [0.93, 1.53]
1.3 Mitomycin C and Mostly Oral Fluoropyrimidines	1	302	Risk Ratio (M‐H, Random, 95% CI)	1.11 [0.59, 2.09]
1.4 5‐FU and Levamisole	5	749	Risk Ratio (M‐H, Random, 95% CI)	0.94 [0.74, 1.19]
1.5 5‐FU and Folinic Acid Regimens	5	2562	Risk Ratio (M‐H, Random, 95% CI)	0.94 [0.80, 1.10]

3.1. Analysis.

Comparison 3 Systemic Chemotherapy Deaths, Outcome 1 Systemic Chemotherapy Deaths.

Comparison 4. Regional Chemotherapy Deaths.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Regional Chemotherapy Deaths	6	1393	Risk Ratio (M‐H, Random, 95% CI)	1.11 [0.74, 1.68]
1.1 Portal Vein Infusion	4	1101	Risk Ratio (M‐H, Random, 95% CI)	1.29 [0.76, 2.18]
1.2 Intraperitoneal Infusion	1	151	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.56, 1.92]
1.3 Hepatic Artery Infusion	1	141	Risk Ratio (M‐H, Random, 95% CI)	0.50 [0.20, 1.23]

4.1. Analysis.

Comparison 4 Regional Chemotherapy Deaths, Outcome 1 Regional Chemotherapy Deaths.

Comparison 5. Immunotherapy Deaths.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Immunotherapy Deaths	8	3091	Risk Ratio (M‐H, Random, 95% CI)	0.95 [0.78, 1.14]
1.1 Non‐specific Immunotherapy	3	439	Risk Ratio (M‐H, Random, 95% CI)	1.00 [0.34, 2.95]
1.2 Specific Immunotherapy	5	2652	Risk Ratio (M‐H, Random, 95% CI)	0.94 [0.78, 1.14]

5.1. Analysis.

Comparison 5 Immunotherapy Deaths, Outcome 1 Immunotherapy Deaths.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Abdi 1989.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. Alberta, Canada 4. Years: 1976‐83 5. N = 253 6. Age: median 57.1 years 7. Sex M/F: 143/ 110 8. Follow‐up: ‐schedule & tests described ‐ duration: mean 83.4 months
Interventions	1. Observation 2a. Semustine 130mg/m2 po day 1 & 5‐FU 325 mg/m2 days 1‐5 and 375 mg/m2 days 36‐40; every 10 weeks; for 20 months. 2b. BCG 120 mg po every 28 days for 5 years. 3. Duration: 20 months 4. Delay after surgery: up to 6 weeks
Outcomes	1. Dose given: N/A 2. Toxicity: ‐ no deaths ‐ grade 3 GI: N/A ‐ grade 3 other: 1 renal failure
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Beart 1990.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. US; NCCTG 4. Years: N/A 5. N = 224 6. Age: 66 median 7. Sex M/F: 103/116 8. Follow‐up: ‐ schedule: yes ‐ tests: yes 9: Duration: 66 months median
Interventions	1. Observation 2. 5‐FU 500 mg/m2/day x 7
Outcomes	1. Dose given: >80% in 100 pts. <50% in 10 pts. 2. Toxicity: none significant 3. % 5‐year OS: ‐all pts: 68 /68 ‐stage II: 82/82 4. % 5‐year DFS: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Benson 2004.

Methods	1. Recommendations based on meta‐analyses
Participants	1. Colon cancer 2. Stage II 3. International 4. Years: 1987‐2004 5. N=N/R 6. Age: N/R 7. Sex: N/R
Interventions	1. Observation 2. Various chemotherapies including 5‐FU
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. % 5‐yr OS: N/R 4. % 5‐yr DFS: N/R
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

CCCSG Japan 1995.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer (here we consider only colon). 2. Stages II & III 3. Japan, CCCSG 4. Years: 1984‐85 5. N = 899 6. Age: mean 59 7. Sex M/F: 458/441 8. Follow‐up: ‐schedule & tests described ‐median duration: >60 months
Interventions	1. Observation 2a. Mitomycin C 12 mg intraportally day 0 and 6 mg days 7 & 14 & at 2, 4 & 6 months; plus 5‐FU 200 mg po daily from day 14. 2b. As above except no intraportal chemotherapy. 3. Duration: 6 months 4. Delay after surgery: 0
Outcomes	1. Dose prescribed: overall 71% to 76% of drug doses given. 2. Toxicity: reported as mild. 3. % 5‐yr OS: ‐ all pts: 80/80/82 ‐ Stg II: 90/84/88 4. % 5‐yr DFS: ‐ all pts: 76/77/81 ‐ Stg II: 88/88/89
Notes	Patients with rectal cancer not considered.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Chlebowski 1988.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. US, WCSG 4. Years 1976‐78 5. N = 78 6. Age: median 57 years 7. Sex M/F: 42/36 8. Follow‐up: ‐ schedule & tests described ‐ duration: median 90 months
Interventions	1. Placebo 2. Levamisole 2.5 mg/K/day on days 1 & of each week. 3. Duration: 18 months 4. Delay after surgery: N/A 3.
Outcomes	1. Dose given: all 52 had >80% of planned dose 2. Toxicity: 1 death due to agranulocytosis. 3. % 5‐year OS: ‐ all colon pts: 67/60 ‐ Stg II colon: 84/79 4. % 5‐year DFS: ‐ all pts: N/A ‐ Stg II: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Colacchio 2004.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon 2. Stage II 3. US, Canada; CALGB 4. Years: 1997‐2002 5. N = 1738 6. Age: N/A 7. Sex M/F: N/A 8. Follow‐up: ‐schedule & tests: N/A ‐ duration: 32 to 34 months, median
Interventions	1. Observation 2. Edrecolomab 500 mg and then 100 mg IV every month, times 4. 3. Duration: 4 months 4. Delay after surgery: N/A
Outcomes	1. Dose given: N/A 2. Toxicity: ‐ 0 deaths ‐ Grade 3‐4 GI toxcity: 13% 3. % 5‐year OS: ‐ all pts: 93/94 ‐ Stg II: 93/94 4. % 5‐year DFS: ‐ all pts: 88/88 ‐ Stg II: 88/88
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Dignam 1999.

Methods	1. Meta‐analysis of 5 RCTs 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages I to III 3. US, NSABP 4. Years: 1977‐1994 5. N=5969 whites & 663 blacks 6. Age: 52.6% > 60 yrs 7. Sex M/F: 3606/3026 8. Follow‐up: N/R
Interventions	1. Observation or systemic chemotherapy 2. Systemic or PVI chemotherapy 3. Duration: variable 4. Delay after surgery: variable
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. % 5‐yr OS: 21% excess risk of death for blacks compared with whites (p=0.004) 4. % 5‐yr DFS: 8% excess risk of recurrence for blacks compared with whites (p=0.27)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Dube 1997.

Methods	1. Meta‐analysis of 29 RCTs 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. International 4. Years: 1950‐90 5. N=N/R 6. Age: N/R 7. Sex M/F: N/R 8. Follow‐up: N/R
Interventions	1. Observation 2. Various chemotherapies including 5‐FU 3. Duration: N/R 4. Delay after surgery: N/R
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. % 5‐yr OS: OR for death for colon cancer 0.81 (95% CI: 0.69, 0.94) or an absolute 5% increase in OS 4. % 5‐yr DFS: N/R
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Fielding 1992.

Methods	1. RCT 2. Risk of bias: moderate (>10% drop‐outs; difference in well differentiated tumours ‐p= 0.04‐)
Participants	1. Colon & rectal cancer 2. Stages I, II, III 3. UK, LBCP 4. Years: 1981‐84 5. N=398 6. Age: N/A 7. Sex M/F: N/A 8. Follow‐up: ‐schedule described, but not the tests. ‐duration: >84 months
Interventions	1. Observation 2a. 5‐FU 1000mg plus heparin 10,000 u by PVI daily. 2b. Heparin 10,000 u by PVI daily. 3. Duration: 7 days 4. Delay after surgery: 0
Outcomes	1. Dose given: treatment was incomplete in 11% on heparin & 18% on 5‐FU/heparin 2. Toxicity: ‐ 1 death due to neutropenic sepsis ‐other: N/A 3. % 5‐year OS: ‐ all pts: 77/73/82 ‐ Stg II: 93/81/90 4. % 5‐year DFS: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Figueredo 1997.

Methods	1. Meta‐analysis of 11 RCTs 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stage II 3. International 4. Years: 1987‐96 5. N=N/R 6. Age: N/R 7. Sex M/F: N/R 8. Follow‐up: N/R
Interventions	1. Observation 2. Various treatments 3. Duration: variable 4. Delay after surgery: N/R
Outcomes	1. Dose given: N/R 2. Toxicity: described 3. % 5‐yr OS: OR for death for systemic chemotherapy 1.02 (95% CI: 0.80, 1.30) and for PVI chemotherapy 0.62 (95% CI: 0.35, 1.11)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Figueredo 2004.

Methods	1. Meta‐analysis of 13 RCTs and 1 M‐A 2. Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. International 4. Years: 1987‐2004 5. N=4187 6. Age: N/R 7. Sex M/F: N/R 8. Follow‐up: N/R
Interventions	1. Observation 2. Various treatments 3. Duration: N/R 4. Delay after surgery: variable
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. % 5‐yr OS: RR for death 0.87 (95% CI: 0.75, 1.01); also reported for groups receiving 5‐FU + levamisole, 5‐FU + folinic acid and PVI chemotherapy
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Francini 1994.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. Siena, Italy 4. Years: 1985‐90 5. N = 398 6. Age: N/A 7. Sex M/F: N/A 8. Follow‐up: ‐schedule & tests described ‐Duration: 65 months
Interventions	1. Observation 2. 5‐FU 400mg/m2 and folinic acid 200mg/m2 IV days 1‐5, every 4 weeks 3. Duration: 1 year 4. Delay after surgery: up to 3 weeks
Outcomes	1. Dose given: N/A 2. Toxicity: ‐ 0 deaths ‐ Grade 3 GI: 21 diarrhea, 20 stomatitis ‐ Grade 3 other: 3 dermatitis, 13 conjunctivitis 3. % 5‐yr OS: ‐ all pts: 65/79 (p=0.0044) ‐ Stg II pts: 86/89 (p=0.005) 4. % 5‐yr DFS: ‐ all pts: 59/74 (p=0.005) ‐ Stg II pts: 77/83
Notes	Included in IMPACT B2, 1999
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Gill 2004.

Methods	1. Meta‐analysis of 7 RCTs 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. International 4. Years: 1989‐98 5. N=3302 6. Age: 56%>60 years 7. Sex M/F: 1753/1649 8. Follow‐up: N/R
Interventions	1. Observation 2. 5‐FU plus either levamisole or folinic acid 3. Duration: 6 to 12 months 4. Delay after surgery: N/R
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. % 5‐yr DFS and OS: calculated according to prognostic subsets
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Glimelius 2005.

Methods	1. Meta‐analysis of 5 RCTs 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. Scandinavia 4. Years: 1991‐97 5. N = 2170 (colon 1520) 6. Age: 62‐66 7. Sex M/F: 1212/999 8. Follow‐up: ‐schedule & tests not described, variable for each institution ‐duration: >60 months
Interventions	1. Observation 2. Either of 5 regimens: ‐ 5‐FU & levamisole as in Moertel, 1995 ‐ 5‐FU & folinic acid as in O'Connell, 1997 ‐ 5‐FU 500mg/m2 plus folinic acid 60mg/m2 IV days 1 & 2, every 2 weeks, 10 times (Nordic regimen) ‐ 5‐FU & folinic acid as in O'Connell, 1997 plus levamisole as in Moertel, 1995 ‐ 5‐FU & folinic acid as in Nordic plus levamisole as in Moertel, 1995 3. Duration: 5‐12 months 4. Delay after surgery: 30‐42 days
Outcomes	1. Dose given: 67% of pts received the planned number of treatments 2. Toxicity: N/A 3. % 5‐year OS: ‐ all colon pts: N/A ‐ Stg II colon: 79/79 4. % 5‐year DFS: ‐ all pts: N/A ‐ Stg II: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Gray 1989.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. Australia & New Zealand 4. Years 1978‐81 5. N = 293 6. Age: mean 63‐64 years 7. Sex M/F: 160/133 8. Follow‐up: ‐schedule & tests described ‐duration: >60 months
Interventions	1. Observation 2. Autologous tumour cells plus BCG by scarification every 2 weeks for 2 months; monthly 10 times; then every 3 monts. ‐Duration: 2 years ‐Delay after surgery: 2 weeks
Outcomes	1. Dose given: N/A 2. Toxicity: 1 death due to renal failure (immune complexes in the mesangium of the kidneys) 3. % 5‐year OS: ‐ all pts: 54/55 ‐ Stg II: 64/63 4. % 5‐year DFS: ‐ all pts: 56/55 ‐ Stg II: 61/65
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Gray 1991.

Methods	1. Meta‐analysis of 6 RCTs of PVI chemotherapy 2. Risk of bias: Moderate
Participants	1. Colon & rectal cancer 2. Stages I to III 3. International 4. Years: pre 1990 5. N = 2201 6. Age: N/A 7. Sex: N/A 8. Follow‐up: ‐ schedule: N/A ‐ tests: N/A ‐ duration: N/A
Interventions	1. Observation 2. 5‐FU +/‐ mitomycin C by PVI 3. Duration: 7 days 4. Delay after surgery: none
Outcomes	1. Dose given: N/A 2. Toxicity: N/A 3. % 5‐yr OS: ‐ all pts: OR for death 31% +/‐ 8% ‐ stg II: N/A 4. % 5‐yr DFS: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Gray 2007.

Methods	1. RCT 2. Risk of bias: low
Participants	1. Colon and rectal cancer 2. Stages II and III
Interventions	1. Observation 2. 5‐FU or levamisole
Outcomes	1. Dose given: N/A 2. Toxicity: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Hanna 2001.

Methods	1. Meta‐analysis of 3 RCTs 2. Risk of bias: moderate
Participants	1. Colon & rectal cancer 2. Stages II & III 3. US & Holland 4. Years: 1980‐99 5. N = 728 6. Age: N/A 7. Sex: N/A 8. Follow‐up: ‐schedule N/A ‐tests: N/A ‐duration: >60 months
Interventions	1. Observation 2. Autologour tumour cells + BCG intradermally 3. Duration: 2 to 6 months. 4. Delay postop.: N/A
Outcomes	1. Dose given: N/A 2. Toxicity: mild, local 3. % 5‐year OS: N/A except for subgroups 4. % 5‐year DFS: ‐all pts: 66 /77 HR=0.75, p=0.05 ‐stage II: 63 /80 ‐stage II:
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Harris 2000.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. US, Canada; ECOG 4. Years: 1986‐93 5. N = 412 6. Age: 66 years, median 7. Sex M/F: 216/194 8. Follow‐up: ‐schedules & tests described ‐ Duration: median 80 months
Interventions	1. Observation 2. Intradermal irradiated tumour cells plus BCG at weeks 0, 1 & 3, the latter without BCG 3. Duration: 1 month 4. Delay after surgery: 4‐5 weeks
Outcomes	1. Dose given: N/A 2. Toxicity: ‐ 0 deaths ‐ ulceration, draining and crusting in 79 patients 3. % 5‐yr OS: ‐ all pts: 64/62 ‐ Stg II pts: 70/69 4. % 5‐yr DFS: ‐ all pts: 59/58 ‐ Stg II pts: 66/64
Notes	see also Hanna review
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Hartung 2005.

Methods	1. RCT 2. Risk of bias: Moderate (17% drop‐out)
Participants	1. Colon cancer 2. Stage II 3. Manheim, Germany 4. Years: 1997‐2000 5. N = 312 6. Age: median 64 years 7. Sex M/F: 118/186 8. Follow‐up: ‐schedule & tests described ‐duration: median 42 months
Interventions	1. Observation 2. Edrecolomab 500mg IV and then 100mg IV every month, 4 times. 3. Duration: 4 months 4. Delay after surgery: up to 6 weeks.
Outcomes	1. Dose given: N/A 2. Toxicity: ‐0 deaths ‐no Grade 3 toxicity 3. % 5‐year OS: ‐ all pts: 87/90 ‐ Stg II: 87/90 4. % 5‐year DFS: ‐ all pts: 76/76 ‐ Stg II: 76/76
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

IMPACT B2 1999.

Methods	1. Meta‐analysis of 5 RCTs 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stage II only 3. International 4. Years: 1982‐91 5. N = 1016 6. Age: 62 years 7. Sex M/F: 559/457 8. Follow‐up: ‐schedule & tests not described ‐ Duration: median 62 to 102 months
Interventions	1. Observation 2. 5‐FU 370‐425 mg/m2 plus folinic acid 20‐200 mg/m2 days 1‐5, every month 3. Duration: 6 months except one 12 months. 4. Delay after surgery: 3‐6 weeks.
Outcomes	1. Dose given: median relative dose‐intensity (given/planned) 85%‐98% 2. Toxicity: ‐ 0 deaths ‐ Grade 3 GI: nausea/vomiting 4%; stomatitis 11% and diarrhea 8%.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Isenberg 1994.

Methods	1. RCT 2. Risk of bias: Moderate ( uncertain allocation concealment)
Participants	1. Colon & rectal cancer 2. Stages I, II & III 3. Koln, Germany 4. Years: 1983‐86 5. N = 97 6. Age: mean 61 & 64 years 7. Sex M/F: 45/56 8. Follow‐up: ‐schedule & tests: standard in the clinic. ‐duration: 96 months
Interventions	1. Observation 2. Propiobacterium granulosum KP‐45 10mg IV on days ‐7 to +3 3. Duration 10 days 4. Pre‐ & post‐surgery
Outcomes	1. Dose given: all >80% 2. Toxicity: fever/chills in 43, headaches in 28 vomiting in 16 3. % 6‐year OS: ‐ all pts: 34/54 ‐ Stg II: 45/90 4. % 5‐year DFS: ‐ all pts: 34/54 ‐ Stg II: 45/90
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Kimura 1995.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. Tottori Univ., Japan 4. Years: 1985‐88 5. N = 110 6. Age: mean 63 years 7. Sex M/F: 52/58 8. Follow‐up: ‐schedule & tests: N/A ‐ duration: >60 months
Interventions	1. Observation 2. 5‐FU 3mg/K/day by PVI for 14 days plus Mitomycin C 10mg by PVI or IV on days 0, 7, 14, plus 5‐FU 150mg/day po for 6 months after infusion 2. Duration: 6 months 3. Delay after surgery: 0
Outcomes	1. Dose given: N/A 2. Toxicity: N/A 3. % 5‐year OS: N/A 4. % 5‐year DFS: ‐all pts: 74/81 ‐ Stg II: 82/92
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Kingston 1993.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. England, UK 4. Years 1981‐83 5. N = 603 6. Age: median 66‐67 years 7. Sex M/F: 341/262 8. Follow‐up: ‐schedule & tests described ‐duration: >60 months
Interventions	1. Observation 2. Razoxane 125mg twice daily, 5 days per week 3. Duration: 2 years 4. Delay after surgery: up to 8 weeks
Outcomes	1. Dose given: >80% in 17%, 50‐80% in 18%
Notes	Trial cut short due to worse outcome for treatment
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Laffer 1995.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages I, II & III 3. Switzerland, SAKK 4. Years: 1981‐87 5. N = 505 6. Age: 62 years 7. Sex M/F: 277/228 8. Follow‐up: ‐schedule & tests described ‐ Duration: 96 months
Interventions	1. Observation 2. 5‐FU 500mg/m2 plus heparin 5000 u, days 1‐7 and mitomycin C 10 mg/m2 on day 1 only, all by PVI. 3. Duration: 7 days 4. Delay after surgery: 0
Outcomes	1. Dose given: 69% of patients received >80% planned doses. 2. Toxicity: N/A 3. %5‐yr OS: ‐ all pts: 55/66 (p=0.026) ‐ all colon: 61/72 ‐ Stg II pts: 69/77 4. % 5‐yr DFS: ‐ all pts: 48/57 (p=0.051) ‐ all colon: 55/66 ‐ Stg II: 63/68
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Laurie 1989.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. US, NCCTG 4. Years 1980s 5. N = 408 6. Age: median 61 years 7. Sex M/F: 204/204 8. Follow‐up: ‐schedule & tests described ‐duration: 93 months
Interventions	1. Observation 2a. Levamisole 150mg days 1, 2 & 3 every 2 weeks. 2b. Levamisole as above plus 5‐FU 450mg/m2 days 1‐5 and beginning day 28, weekly. 3. Duration: 1 year 4. Delay after surgery: up to 5 weeks.
Outcomes	1. Dose given: >80% of planned dose for levamisole in 95% and for 5‐FU+levamisole 83% of patients. 2. Toxicity: 1 case of severe dermatitis due to levamisole. 3. % 5‐year OS: ‐all pts: 57/59/62 ‐ Stg II: 58/76/74 4. % 5‐year DFS: ‐all pts: 45/58/59 ‐Stg II: 58/76/74
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

LIMAG 1997.

Methods	1. Meta‐analysis of 10 RCTs 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages I to III 3. International 4. Years: 1979‐96 5. N = 3499 6. Age: 65 years median 7. Sex M/F: 1696/1390 8. Follow‐up: ‐ schedule: N/A ‐ tests: N/A ‐ duration: > 60 months
Interventions	1. Observation 2. 5‐FU +/‐ mitomycin C by PVI x 7 days 3. Delay after surgery: nil
Outcomes	1. Dose given: N/A 2. Toxicity: N/A 3. % 5‐yr OS: ‐ all pts: 57/61 ‐ stg II: 71/76 (p=0.01) 4. % 5‐yr DFS: N/A ‐ all pts:
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Mamounas 1999.

Methods	1. Meta‐analysis of 4 RCTs 2. Risk of bias: moderate
Participants	1. Colon cancer 2. Stages II & III 3. US 4. Years: 1977‐90 5. N = 4006 6. Age: mid60's 7. Sex M/F: 2081/1739 8. Follow‐up: ‐schedule: described ‐tests: described ‐duration: .60 months
Interventions	1a. Observation vs. MOF or PVI 5‐FU in 2 trials 1b. MOF or FU/Lev vs. FUFA in 2 trials
Outcomes	1. Dose given: N/A 2. Toxicity: N/A 3. % 5‐year OS: ‐all pts: Obs. vs MOF: 60/67 Obs vs. PVI: 67/74 MOF vs. FUFA: 66/76 (p<0.001) FU/Lev vs. FUFA: 70/75 ‐stage II pts: Obs. vs. MOF: 72/75 Obs. vs PVI: 76/88 (p=0.01) MOF vs FUFA: 84/92 (p=0.05) FU/Lev vs. FUFA: 81/85 4. % 5‐year DFS: all pts: N/A ‐stage II: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

McCollum 2002.

Methods	1. Meta‐analysis of arms in an RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. US, Intergroup 0089 4. Years: 198?‐92 5. N=3759 6. Age: median 64 years 7. Sex M/F: 1846/1534 8. Follow‐up: N/R
Interventions	Systemic chemotherapies 1, 2, 3 & 4 3. Duration: 6 to 12 months 4. Delay after surgery: N/R
Outcomes	1. Dose given: N/R 2. Toxicity: ‐Grade 3+ diarrhea: 23% whites, 8% blacks ‐Grade 3+ stomatitis: 12% whites, 9% blacks ‐other Grade 3+ toxicities not different 3. %5‐yr OS: ‐blacks/whites: 65%/66% ‐blacks/whites: 57%/58%
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

McDermott 2003.

Methods	1. RCT 2. Risk of bias: Moderate (rectal cancer pts receive RT; other characteristics not described in abstract)
Participants	1. Colon & rectal cancer 2. Stages II & III 3. Belfast, UK 4. Years: 1994‐97 5. N = 254 6. Age: N/A 7. Sex M/F: N/A 8. Follow‐up: ‐schedule & tests: N/A ‐duration: median 80 months
Interventions	1. Observation 2. 5‐FU 400mg/m2 plus folinic acid 200mg/m2 by IV boluses on day 1 and 5‐FU 400mg/m2 by IV infusion over 22 hours for 2 days; every 2 weeks. 3. Duration: 4 months 4. Delay after surgery: N/A
Outcomes	1. Dose given: N/A 2. Toxicity: ‐1 death ‐Grade 3 diarrhea in 0.8% 3. % 5‐year OS: ‐ all pts: N/A ‐ stage II: 73/76 4. % 5‐year DFS: ‐ all pts: N/A ‐ stage II: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Moertel 1995.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stage II 3. US, Canada, Intergroup 4. Years 1984‐87 5. N = 318 6. Age: 61‐62 median 7. Sex M/F: 183/135 8. Follow‐up: ‐schedule & tests described ‐duration: median 74 months
Interventions	1. Observation 2. Levamisole 150mg days 1, 2 & 3 every 2 weeks and 5‐FU 450mg/m2 days 1‐5 and starting on day 28, weekly. 3. Duration: 1 year 4. Delay after surgery up to 5 weeks
Outcomes	1. Dose given: NR 2. Toxicity: NR 3. % 5‐year OS: ‐Stg II: 61/60 4. % 5‐year DFS: ‐ Stg II: 56/61
Notes	For Dose given & Toxicity see Moertel, 1990
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Panettiere 1988.

Methods	1. RCT 2. Risk of bias: Moderate (drop‐outs 12%)
Participants	1. Colon & rectal cancer 2. Stages II & III 3. USA ‐ SWOG 4. Years 1977‐80 5. 37 months 6. N = 279 7. Age: median 60 years 8. Sex M/F: 156/123 9. Follow‐up: ‐schedule & tests described ‐duration: median 84 months
Interventions	1. Observation 2a. Semustine 175mg/m2 po day 1 & 5‐FU 400mg/m2 days 1, 8 & 15; all every 8 weeks 2b. As above plus BCG 600 M org. orally q 2weeks 3. Duration: 1 year 4. Delay after surgery: 2‐6 weeks
Outcomes	1. Dose given: N/A 2. Toxicity: N/A 3. % 5‐year OS: ‐all pts: 61/62/50 ‐Stg II: 88/66/58 4. % 5‐year DFS: ‐all pts: 47/47/45 ‐ Stg II: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Sadahiro 2004.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages I, II & III 3. Tokai, Japan 4. Years: 1994‐2001 5. N = 305 6. Age: mean 60 years 7. Sex M/F: 185/120 8. Follow‐up: ‐schedule & tests described ‐duration: 59 months
Interventions	1. Observation 2. 5‐FU 250mg/day plus dexamethasone 4mg by hepatic artery infusion. 3. Duration: 3 weeks 4. Delay after surgery: from week before to 2 weeks after surgery
Outcomes	1. Dose given: N/A 2. Toxicity: ‐0 deaths ‐0 significant Grade 3 3. % 5‐year OS: ‐all pts: 82/94 (p=0.0005) ‐Stg II: 84/92 4. % 5‐year DFS: ‐all pts: 72/86 (p=0.0022) ‐Stg II: 79/88
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Sakamoto 2004.

Methods	1. Meta‐analysis of 3 RCTs 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages I to III 3. Japan 4. Years: 1986‐90 5. N = 5233 6. Age: median 59 7. Sex M/F: 2944/2289 8. Follow‐up: ‐schedule: N/A ‐tests: N/A ‐duration: >60 months
Interventions	1. Observation 2. IV mitomycin C or 5‐FU followed by oral 5‐FU, UFT or HCFU 3. Duration: 1 year 4. Delay after surgery N/A
Outcomes	1. Dose given: N/A 2. Toxicity: N/A 3. % 5‐year OS: For all colon cancer, HR 0.86 (0.73, 1.00). For stage II: decrease by 21% SE 9%; absolute OS difference 4.3%. 4. % 5‐year DFS: HR 0.87 (075, 0.99); absolute DFS difference 4.7%.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Sakamoto 2005.

Methods	1. Meta‐analysis of 3 RCTs 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages Dukes B & C 3. Japan 4. Years: 1987‐90 5. N = 2152 6. Age: N/A 7. Sex: N/A 8. Follow‐up: ‐schedule: N/A ‐tests: N/A ‐duration: >60 months
Interventions	1. Observation 2. 5‐FU or mitomycin C IV followed by oral HCFU (Carmofur) 3. Duration: 6 weeks to 12 months 4. Delay after surgery: N/A
Outcomes	1. Dose given: N/A 2. Toxicity: N/A 3. % 5‐year OS: ‐all pts: 76/80 (HR = 0.82 (0.68, 0.99)) ‐stage II: 82/89 4. % 5‐ year DFS: ‐all pts: 71/77 (HR = 0.77 (0.65, 0.91) ‐stage II: 78 / 85
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Sargent 2001.

Methods	1. Meta‐analysis of 7 RCTs 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. International 4. Years: 1978‐2000 5. N=3351 6. Age: 53%>61 yrs 7. Sex M/F: N/R 8. Follow‐up: N/R
Interventions	1. Observation 2. Systemic chemotherapy with 5‐FU + levamisole or folinic acid 3. Duration: 6 to 12 months 4. Delay after surgery: 4 to 7 weeks
Outcomes	1. Dose given: N/R 2. Toxicity: ‐leucopenia and N/V more common with 5‐FU + levamisole ‐stomatitis and diarrhea more common with 5‐FU + leucovorin
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Sargent 2005.

Methods	1. Meta‐analysis of 18 RCTs 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages I, II & III 3. International 4. Years: 1977‐99 5. N=20,898 6. Age: 58%>60 yrs 7. Sex M/F: 55%/45% 8. Follow‐up: N/R
Interventions	1. Observation or chemotherapy 2. Systemic chemotherapy 3. Duration: variable 4. Delay after surgery: variable
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. 3‐yr DFS predicts 5‐yr OS (correlation for hazard ratios 0.92)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	D ‐ Not used

Taal 2001.

Methods	1. RCT 2. Risk of bias: Moderate (RT in rectal & some colon patients)
Participants	1. Colon & rectal cancer 2. Stages II & III 3. Netherlands, NACCP 4. Years: 1990‐96 5. N = 1029 6. Age: 60 7. Sex M/F: 599/439
Interventions	1. Observation 2. Levamisole 150 mg days 1, 2 & 3 every 2 weeks plus 5‐FU 450mg/m2 days1‐5 and starting on day 28, weekly. 3. Duration: 1 year 4. Delay after surgery: up to 8 weeks
Outcomes	1. Dose given: compliance with treatment 80% at 6 months and 69% at 1 year 2. Toxicity: N/A 3. % 5‐yr OS: ‐ all pts: 58/60 ‐ stageII: 73/79 4. % 5‐year DFS: ‐ all pts: 51/56 ‐ stage II: 65/71
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Vaillant 2000.

Methods	1. RCT 2. Risk of bias: Moderate (16% drop‐out, 2 co‐intervention)
Participants	1. Colon cancer 2. Stages II & III 3. France 4. Years: 1986‐91 5. N = 267 6. Age: mean 63 years 7. Sex M/F: 124/143 8. Follow‐up: ‐schedule not described but tests yes ‐duration: 58 months
Interventions	1. Observation 2. 5‐FU 600mg/m2 by intraperitoneal infusion over 3 hours, days 1 to 6. 3. Duration: 6 days 4. Delay after surgery: 0
Outcomes	1. Dose given: >80% for 103; 50‐80% for 8 and <50% for 12 patients 2. Toxicity: ‐0 deaths ‐0 Grade 3 3. % 5‐year OS: ‐all pts: 69/74 ‐Stg II: 79/88 4. % 5‐year DFS: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Vermorken 1999.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. Amsterdam, Netherlands 4. Years: 1987‐96 5. N = 254 6. Age: 65, 66 7. Sex M/F: 136/118 8. Follow‐up: ‐schedule & tests described ‐duration: 64 months
Interventions	1. Observation 2. Autologous tumour cell preparation plus BCG 1 M organisms intradermal on day 0, weeks 1 and 3, and at 6 months 3. Duration: 6 months 4. Delay after surgery: 4‐5 weeks
Outcomes	1. Dose given: all had >80% 2. Toxicity: mild 3. % 5‐year OS: ‐all pts: 71/75 ‐Stg II: 73/83 (p=0.01) 4. % 5‐year DFS: ‐all pts: 61/70 ‐Stg II: 63/77 (p=0.018)
Notes	see also Hanna, 2001 review
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Wan 2005.

Methods	1. Meta‐analysis of 8 studies (11 RCTs and 1 prospective group) 2. Risk of bias: High
Participants	1. Colon & rectal cancer 2. Stage II 3. International 4. Years: 1993‐2002 5. N=6518 6. Age: N/R 7. Sex M/F: N/R 8. Follow‐up: N/R
Interventions	1. Observation or systemic chemotherapy 2. Systemic or PVI chemothearpy with 5‐FU +/‐ levamisole +/‐ folinic acid 3. Duration: N/R 4. Delay after surgery: N/R
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. % 5‐yr OS: Stg II: RR 0.79 (95% CI: 0.7, 0.9)
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Watanabe 2006.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. Japan, JFMTC 4. Years: 1989‐90 5. N = 984 6. Age: median 62 years 7. Sex M/F: 507/494 8. Follow‐up: ‐schedule & tests not described ‐duration: >60 months
Interventions	1. Observation 2. Mitomycin C 6mg/m2 days 0, 7 & 28 ans every month to 6 months plus 5‐FU 250mg IV days 1‐8 plus HCFU (Carmofur) 300 mg daily from day 14 to one year. 3. Duration: 1 year 4. Delay after surgery: 0
Outcomes	1. Dose prescribed: ‐MitoC: 59.6% ‐5‐FU: 93.2% ‐HCFU: 59.5% 2. Toxicity: no deaths; not graded 3. % 5‐year OS: ‐all pts: 78/82 ‐Stg II: N/A 4. % 5‐year DFS: ‐all pts: 72/78 ‐Stg II: 85/93
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Windle 1987.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon & rectal cancer. 2. Stages I, II & III 3. Leicester, UK 4. Years: 1979‐82 5. N = 131 6. Age: median 60 years 7. Sex M/F: 67/64 8. Follow‐up: ‐schedule & tests described ‐Duration: >60 months
Interventions	1. Observation 2a. 5‐FU 1gm IV days 0, 1 & 2, then weekly. 2b. 5‐FU as above plus levamisole 150mg po days 0‐2 3. Duration: 6 months 4. Delay after surgery: 0
Outcomes	1. Dose prescribed: all patients had > 80% 2. Toxicity: N/A 3. % 5‐yr OS: ‐all pts: 56/48/68 ‐Stg II pts: 60/64/64 4. % 5‐yr DFS: ‐all pts: 56/48/68 ‐ Stg II pts: N/A
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Wolmark 1988.

Methods	1. RCT 2. Risk of bias: Moderate (difference in patient characteristics)
Participants	1. Colon cancer 2. Stages II & III 3. US & Canada, NSABP 4. Years 1977‐83 5. N = 1116 6. Age: median 65 years 7. Sex M/F: 692/424 8. Follow‐up: ‐schedule & tests described ‐duration: mean 77.3 months
Interventions	1. Observation 2a. Semustine 130 mg/m2 day 1 plus 5‐FU 325 mg/m2 days 1‐5 and 375mg/m2 days 36‐40; every 10 weeks. 2b. BCG 600 M org. by scarification, weekly 12 times and every other week 33 times. 3. Duration: 20 months 4. Delay after surgery: 3‐6 weeks
Outcomes	1. Dose prescribed: >80% in 159/383 2. Toxicity: ‐3 toxic deaths; acute leukemia ‐ Grade 3 GI: 15% ‐ Grade 3 other: 1 case nephrotoxicity 3. % 5‐year OS: ‐all pts: 59/67/67 ‐Stg II: 72/75/NR 4. % 5‐year DFS: ‐all pts: 51/58/56 ‐Stg II: N/A
Notes	for data on Stg II pts see Mamounas, 1999
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Wolmark 1990.

Methods	1. RCT 2. Risk of bias: Moderate (22% drop‐out due to pre‐randomization)
Participants	1. Colon cancer 2. Stages I, II & III 3. US, Canada; NSABP 4. Years: 1984‐88 5. N = 898 6. Age: 65 7. Sex M/F: 527/374 8. Follow‐up: ‐schedule & tests described ‐duration: >60 months
Interventions	1. Observation 2. 5‐FU 600 mg/m2 plus heparin 5000 u by PVI daily 3. Duration: 7 days 4. Delay after surgery: 0
Outcomes	1. Dose given: N/A 2. Toxicity: ‐ 0 deaths ‐ 0 Grade 3 3. % 5‐year OS: ‐all pts: 73/81 ‐Stg II: 88/76 4. % 5‐year DFS: ‐all pts: 64/72 (p=0.02) ‐Stg II: N/A
Notes	for data on Stg II pts see Mamounas, 1999
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Zalcberg 1996.

Methods	1. Meta‐analysis of 17 RCTs 2. Risk of bias: Low
Participants	1. Colon & rectal cancer 2. Stages II & III 3. International 4. Years: 1976‐90 5. N=3221 6. Age: N/A 7. Sex M/F: N/A 8. Follow‐up: ‐schedule & tests: N/A ‐ duration: N/A
Interventions	1. Observation 2. Various chemotherapy regimens including 5‐FU 3. Duration: variable 4. Delay after surgery: variable
Outcomes	1. Dose given: N/R 2. Toxicity: N/R 3. % 5‐yr OS: OR for death 0.82 (95% CI: 0.74, 0.91) 4. % 5‐yr DFS: N/G 5. The larger the planned dose in the first 3 months, the larger the benefit in survival.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Zaniboni 1997.

Methods	1. RCT 2. Risk of bias: Low
Participants	1. Colon cancer 2. Stages II & III 3. Italy, GIVIO 4. Years: 1989‐91 5. N = 869 6. Age: median 62 years 7. Sex M/F: 451/437 8. Follow‐up: ‐schedule & tests not described ‐duration: median 64 months
Interventions	1. Observation 2. 5‐FU 370mg/m2 plus folinic acid 200 mg/m2 on days 1‐5, every 28 days. 3. Duration: 6 months 4. Delay after surgery: up to 5 weeks.
Outcomes	1. Dose given: 5% patients discontinued because side‐effects.
Notes	Study included in IMPACT‐B2, 1999. QoL measured.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

OS = overall survival 
 DFS = disease‐free survival

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Ahn 2001	not an RCT
Alikhan 2004	comment
Alliot 2004	multiple letters about a trial with no observation arm
Arnaud 1988	Stage III only
Arnaud 1989	Stage III only
Boussen 1999	no observation arm; median follow‐up 25.8 months
Brivio 2006	Trial of Obs. vs. Interleukin‐2 preoperatively. No separate data for stage II patients.
Bukowski 1987	Review
Buyse 1988	Meta‐analysis of trials published before 1987 of observation vs. Radiotherapy +/‐ chemotherapy in colorectal cancer. No separate data for stage II patients.
Buyse 1989	French version of Buyse, 1988
Chlebowski 1994	Accessory to Chlebowski, 1998
Cremona 1997	Accesory to AXIS, 2003
Dahl 2003	Trial of Obs. vs. 5‐FU + levamisole Abstract: no separate data for stage II patients
De Gramont 2003	abstract only
Fata 2002	not an RCT; analysis of registry data
Fietkau 1998	Norwegian commentary about a Norwegian trial in rectal cancer
Focan 2000	no observation arm
Franchi 1991	Trial of Obs. vs. 5‐FU + folinic acid + Mitomycin C Preliminary report on toxicity; no survival data.
Gambill 2001	Accesory to Hartung, 2005
Garcia 1996	Trial of Obs. vs. Mitomycin C + Ftorafur No separate data for Stage II patients
Gennatas 2002	no observation arm
Graf 1994	Trial of 5_FU infusion to peritoneal cavity. Early report on Tolerance and toxicity. No survival data.
Gray 1988	Accesory to Gray, 1989
Gray 2004	QUASAR trial of Obs. vs. 5‐FU + folinic acid +/‐ levamisole. Abstract: no separate data for stage II patients.
Grem 2004	commentary
Grundmann 1988	Acesory to Isenberg, 1994
GTSG 1991	Trial of Obs. vs. radiotherapy to liver plus 5‐FU. No separate data for stage II patients.
Hafstrom 1988	abstract about a review
Hafstrom 1990	Stage III data only
Haller 2001	Review; not an RCT
Harder 1990	Accesory to SAKK, 1995 Data on transfusion effect
Hashimoto 1994	Neocarzinostatin into bowel wall. No RCT
Hess 1997	Trial comparing two treatments; no observation arm.
Hoover 1989	Accesory to Hoover, 1993
Hoover 1993	Specific immunotherapy vs. observation. No separate data for stage II patients.
IMPACT 1995	Pooling of results of 5 RCTs comparing Obs. vs. 5‐FU + folinic acid, and including patients with colon cancer, stages II and III. Later report, IMPACT‐B2, 1999 gives results for stage II only.
Ito 1996	Trail of Obs. vs. carmofur (HCFU) No separate data for stage II patients
Kapoor 2005	Trial of Obs. vs Famotidine. No survival data.
Kato 2002	Trial of Obs. vs. UFT. No separate data for stage II patients
Kelly 1999	Trial of Obs. vs. Cimetidine. No separate data for stage II patients.
Kerr 2000	Review; not an RCT
Knuth 1988	Commentary; not an RCT
Kodaira 1989	Accessory to Kunii, 1995 (CCCSG of Japan)
Kodaira 1997	Review and meta‐analysis
Kunii 1987	Accesory to Matsuda, 1991
Kunii 1995	Trial of observation vs. mitomycin C + UFT. Noseparate data for stage II patients.
Lacour 1992	Trial of Obs. vs. PAPUA (polyadenylic polyuridylic acid). No data for stage II patients.
Laffer 1987	Accesory to SAKK, 1995
Laffer 1992	Accesory to SAKK, 1995
Laffer 1992a	Accesory to SAKK, 1995
Lise 1987	Review and meta‐analysis; no separate Stage II data
Marangolo 1989	Trial of Obs. vs. 5‐FU + lomustine No data for stage II patients
Marsoni 2001	Accessory to IMPACT‐B, 1999
Matsuda 1991	Trial of Obs. vs. Ftorafur vs. ACNU + Ftorafur No separate data for stage II patients
Matsumoto 2002	no observation arm
Metzger 1987	Accesory to SAKK 1995
Metzger 1989	Accesory to SAKK 1995 Data on effect of transfusion
Metzger 1990	Accesory to SAKK 1995
Mitomi 1990	no observation arm
Mitomi 1993	no observation arm
Moertel 1995	Stage III data only
Moertel 1990	Accessory to Moertel CG, 1995
Moertel 1993	Accessory to GTSG, 1991
Mori 2002	not an RCT; review
Mukai 2003	not an RCT
Nichols 1992	Trial of Interleukin‐2 vs. observation Report on immune responses. No survival data.
Nichols 1993	Trial of Interleukin‐2 + Interferon vs observation. Report on immune response. No survival data.
Nielsen 2002	Trial of Obs. vs Ranitidine. No separate data for stage II patients.
Niimoto 1988	Accesory to Matsuda, 1991
Nio 1992	Trial of Obs. vs. PSK orally. No survival data.
Nishida 1993	Accesory to Kunii, 1995
Nitti 1997	Trial of Obs. vs. 5‐FU + heparin by PVI. No searate data for stage II patients.
O DH Jr. 2000	Commentary; not an RCT
O'Connell 1997	Obs. vs 5‐FU + folinic acid No separate data for stage II patients. Data included, however, in the IMPACT‐B2 meta‐analysis
Pater 1994	Background for the IMPACT meta‐analysis
Pendlebury 2003	Protocol for a trial of Obs. vs. Rofecoxib. Trial abandoned after demonstration of increased cardiovascular morbidity with COX‐2 inhibitors (Kerr, 2007).
Petrasch 1997	Commentary; not an RCT
Piedbois 2000	Accessory to Buyse, 1988
Prohm 1988	Trial of Obs. vs. BCG Abstract: no survival data.
Ragnhammar 2001	Review of chemotherapy for colorectal cancer
Reddy 2004	No observation arm; stage III data only; median follow‐up < 48 months
Reddy 2005	summary; not an RCT
Riethmuller 1994	Trial of Edrecolomab vs. observation. Only patients with stage III disease.
Rougier 1993	not an RCT; review and trial proposal
Rougier 1998	Trial of Obs. vs. 5‐FU + heparin by PVI. No separate data for stage II patients.
Sakamoto 1999	Earlier report on adjuvant carmofur ( HCFU) in colon cancer. See Sakamoto, 2005
Sakamoto 2001	Earlier report of adjuvant fluorpyrimidines in colorectal cancer. See Sakamoto, 2004
SAKK 1997	Accessory to SAKK, 1995
Sargent 2004	Abstract of data presented in Sargent, 2005
Scheithauer 1995	Trial of Obs. vs. 5‐FU + folinic acid both intraperitoneal and intravenous. No separate data for stage II patients; it is stated thewas no significant difference.
Seitz 1994	Initial report on the IMPACT meta‐analysis
Smith 2004	Accesory to Wolmark, 1988 10‐year results of that trial.
Suzuka 1999	Preliminary report; Japanese
Svendsen 1995	Trial of Obs. vs. Cimetidine. No separate data for stage II patients
The QCG 2004	Trial of Obs. vs. 5‐FU + folinic acid +/‐ levamisole. Early report; abstract; no separate data for stage II patients.
Thomas 1987	Accesory to GITSG, 1991
Ullenhag 2006	Trial of Obs. vs. antibody 105AD7. Preliminary report on immune responses. No survival data.
Ulsperger 1993	not an RCT
Uyl‐de Groot 2005	Accesory to Vermorken, 1999
Vaillant 1998	Accessory to Vaillant, 2000
Van den Eertwegh 200	Accesory to Vermorken, 1999
Wan 1988	Trial of Obs. vs. 5‐FU intraluminal + IV No survival data
Watanabe 2004	Accesory to Watanabe, 2006
Weber 1993	Accesory to SAKK, 1995
Weber 1995	Accesory to SAKK, 1995
Wein 2004	Stage III data only; median follow‐up 45.3 months
Wereldsma 1990	Trial of Obs. vs. 5‐FU + heparin vs. urokinase by PVI. No report of stage II data.
Wiesenfeld 1995	Trial of Obs. vs. Interferon gamma. No separate data for stage II patients.
Wurschmidt 1995	Stage III data only
Yamamoto 1998	Tiral of Obs. vs. 5‐FU + folinic acid. No separate data for stage II patients.
Yanagi 2003	Stage III data only; abstract
Yasutomi 1997	Trial of Obs. vs. Mitomycin C + Carmofur (HCFU). No separate data for stage II patients.

Contributions of authors

AF: Design of protocol, selection of publications, data collection, and writing of the text and design of tables

SM: selection of publications, data collection, performance of meta‐analysis, embedding tables, editing text & tables

MC: execution of search strategy, selection of publications, data collection, importation and management of references, editing text and tables

Declarations of interest

None

Edited (no change to conclusions)