| Japanese Journal of Clinical Oncology | Pages |
Adjuvant Therapy with Oral Fluoropyrimidines as Main Chemotherapeutic Agents After Curative Resection for Colorectal Cancer: Individual Patient Data Meta-analysis of Randomized Trials
Introduction
Materials And Methods
Search for Radomized Trials
Screening of Individual Information
Statistical Methods
Results
Analysis of Disease-free Survival
Analysis of Survival
Prognostic Factors in the Control Data and Their Distribution Between Groups
Adjusted Analysis
Discussion
Acknowledgment
References
Adjuvant Therapy with Oral Fluoropyrimidines as Main Chemotherapeutic Agents After Curative Resection for Colorectal Cancer: Individual Patient Data Meta-analysis of Randomized Trials
Background: Oral 5-fluorouracil and its prodrugs (tegafur, carmofur) is now being studied for adjuvant chemotherapy of curatively resected colorectal cancers.To evaluate the effect of these oral fluoropyrimidines (o-FPs), an individual patient data (IPD) meta-analysis of randomized clinical trials was performed in Japan as an inter-trialist group study.
Methods: Data from the three clinical trials in which postoperative adjuvant therapy with o-FPs was compared with surgery alone in patients with colorectal cancer were sought. IPD from a total of 4960 patients with follow-up periods of at least 5 years were analyzed.
Results: The results of the meta-analysis on an `intention to treat' basis demonstrated a significant benefit of o-FPs in terms of the disease-free survival (DFS) of the total patients [risk ratio (RR) 0.830, 95% confidence interval (CI) 0.742-0.929, P = 0.001]. o-FPs were also demonstrated to be effective for survival in rectal cancer (RR 0.857, 95% CI 0.734-0.999, P = 0.049) and in Dukes'C colorectal cancer (RR 0.828, 95% CI 0.711-0.965, P = 0.016).
Conclusion: The results suggest the advantage of long term o-FPs, possibly with the injection of mitomycin C, for prognosis for curatively resected colorectal cancer patients.
INTRODUCTION
Cancer of the colon and rectum is the second most common form of cancer in civilized countries. In Japan alone, nearly 56 000 new cases are diagnosed and 36 000 deaths are caused by this disease every year (1,2). Surgical treatment is the primary management of colorectal tumors, with about three quarters of the patients being operable at the time of diagnosis (3). Even when resection is considered curative, the overall 5-year survival is only 50% (4). This has prompted the search for adjuvant treatment in the form of pre- or postoperative adjuvant therapy for colorectal cancers. Unfortunately, most adjuvant treatments for colorectal cancer are not very effective. The improvement in 5-year survival that they produce is typically <10%.Individual clinical trials have an average size of about 400 patients, which is adequate to study a major treatment effect, but probably insufficient to detect small, but medically worthwhile, treatment benefits (5,6). Indeed, assuming that median survival after curative resection of colorectal cancer is 5 years, a 1-year improvement would have an 80% chance of being detected by observing deaths, a level (about 1000; at 5%, two-tailed level) far larger than in any of the published trials (7). Thus, while individual trials are definitely appropriate to assess reliably the toxicities and side-effects of therapy and arguably appropriate to detect large therapeutic benefits, only a combination of all available trials provides a reasonable chance of detecting the small benefits that can be expected from adjuvant therapy in terms of patient survival.
Meta-analysis of adjuvant therapy for colorectal cancer was first performed by Buyse et al. (8). They did not find any statistically significant effect with either radiotherapy or chemotherapy. Recently, another attempt to evaluate the treatment of colorectal cancer was started. Investigators in Oxford, following their prominent success with meta-analysis in early breast cancer (9,10), organized a `Colorectal Cancer Collaborative Group' and has since been trying to evaluate the adjuvant treatments of colorectal cancer in the reports of 154 randomized trials involving 52 000 patients from those studies, started before 1985. In their meta-analysis, studies from Japan, for which detailed individual results are available, will also be included.
Although most of the colorectal cancer patients who underwent curative resection in Western countries were treated by radiotherapy or parenteral chemotherapy, most of the Japanese gastrointestinal cancer patients have been treated by the prolonged administration oral 5-fluorouracil (5-FU) or its derivatives as the main chemotherapeutic agents.Taking the uniqueness of Japanese regimens into account, we believed it worthwhile to evaluate the effect of such oral 5-FU and related agents against colorectal cancer, separately from the worldwide collaborative study.This paper, therefore, describes just the meta-analysis of adjuvant chemotherapies in Japan, utilizing oral 5-FU and related agents against curatively resected colorectal cancers. (Fig. 1).
Figure 1. Overall DFS curves
MATERIALS AND METHODS
Search for Randomized Trials
A search was performed of published and unpublished randomized clinical trials for colorectal cancer mainly utilizing oral 5-FU or oral fluoropyrimidine derivatives started before January 1988 with their patient accrual completed before December 1988. Screening of the formal and informal information for possible randomized trials yielded seven trials in Japan. As far as our investigation is concerned, there were no randomized trials using oral fluoropyrimidines either in the USA or in Europe before 1989. The seven retrieved trials were then reviewed independently by two of the investigators (J.S., C.H) using a standardized summary form. The two forms were compared and, in case of discrepancy, the report was studied again to achieve consensus. Four criteria were applied for checking the trials: (i) randomized trial, (ii) adjuvant chemotherapy by oral fluoropyrimidines with curative intent, (iii) at least a 5-year follow-up and (iv) control arm that consists of surgery alone cases.
Only three trials were considered to match to our pre-specified criteria for the meta-analysis of oral fluoropyrimidine treatment and were included for further analysis (11-13). The four remaining trials had no control arm with `surgery alone' cases and were not considered to be suitable for this meta-analysis.
Screening of Individual Information
Trials were to be included only if they began randomizing before January 1988, but in such trials all patients were to be included, irrespective of whether randomized before or after that date, with follow-up to the date they were last known to be alive. The analysis involved all the deaths or recurrences that have been reported, without any cut-off to exclude recent months for which information is incomplete.
Information for each individual patient was provided by the principal investigator of each trial, including an indentifier and the age, site of tumor, macroscopic and microscopic depth of invasion, nodal status and histological type of the tumor. In all three trials, site of tumor, macroscopic depth of invasion and macroscopic nodal status were considered as the stratification factors for randomization. Information was also sought on the dates of randomization, of first recurrence and of death. Causes of death were sought only if no recurrence was recorded. Data were checked extensively at the Department of Pharmacoepidemiology, Tokyo University, and any implausibilities were queried persistently. Outcome was always to be analyzed according to an assigned treatment, i.e. `intention to treat' basis (14).
Statistical Methods
The following analysis was performed using two end-points (survival and recurrence) for overall results, for colon and rectum and for Dukes' A, Dukes' B and Dukes' C cancer. Analysis of the data was performed by Statistical Analysis System (SAS) procedures, PHREG and LIFETEST (15).
Prognostic FactorsBased on the data for the control group, the importance of several prognostic factors (gender, age, depth of invasion, nodal status and stage of the disease) for survival and recurrence was evaluated, assuming Cox's proportional hazard model. Since all cases enrolled in the trial were randomized based only on the macroscopic findings of the tumor extent, histopathological depth of invasion and lymph node metastasis, which were not used as stratification factors before randomization, were considered to be the most important factors for the accurate comparison of the treated and control groups.
Test for HomogeneityHomogeneity of the prognostic factors between treatment group and control group was investigated through the three studies applying an ANOVA test for the age and a chi-squared test for other factors.
Non-adjusted AnalysisIndividual and common relative risk ratios through the three studies and their 95% confidence limits were provided by a Peto-like method (Cox's proportional hazard model, including only treatment effect stratified by the study) followed by the test for homogeneity of the hazard ratios.
Test for Interaction Between Treatment and Study or Location of TumorA Wald test for interaction was used to examine whether the effect of treatment varied significantly between different groups of patients. The treatment effect of each group was estimated using Cox's regression including only the treatment effect, then the difference among these estimates was evaluated using a Wald type test.
Adjusted AnalysisThe analysis adjusted for the histopathological background of the tumor extent was applied. Here hij(t) was set to the hazard function of the study i (i = 1, 2, 3) and individual j. We utilized the following two models:
Common effect model:
hij(t) = hi(t)·exp([beta]·Dij + [gamma] Zij)
Study-specific model:
hij(t) = hi(t)·exp([beta]i·Dij + [gamma] Zij)
where Dij is a dummy variable with the value 1 if individual ij belongs in the treatment group or 0 if not, [beta] or [beta]i indicates a common effect through the three studies or individual dependent study effect, respectively, Zij corresponds to the vector of prognostic factor and [gamma] is a vector, which shows the covariate effect. Hazard ratios were estimated by exp([beta] hat) and its 95% confidence limits were also calculated. The homogeneity of the effect (the hypothesis [beta] 1 = [beta] 2 = [beta] 3) was tested statistically by Wald test. All statistical tests are were two-tailed and P values were provided.
RESULTS
In the three clinical trials considered in the present meta-analysis, 4960 patients were randomized. The mean and median follow- up was almost 5 years. All patients received resection for colorectal cancer with curative intent between 1984 and 1988. Neither prior chemotherapy nor prior radiotherapy was given before the operation to any of the patients in these trials. 5-FU was used in the SGCCC trial, tegafur in the SGACCS trial and carmofur in the TSGHCFU trial (Table 1). All these fluoropyrimidine-related agents were administered daily per os from 2 weeks to 1 month after the operation. Daily doses of these agents were considered equivalent in terms of effect and side-effect, according to the results of preclinical and clinical Phase I and Phase II studies. Follow-up data were collected on an `intention to treat' basis and were available from 98% of the patients in the SGCCC trial, 96% in the SGACCS trial and 99% in the TSGHCFU trial, with a total of 4787 cases. Although the accumulation of the patient data was passable, considerable imbalance of the number of patients between the treatment and control group was noticed in the SGACCS trial. One of the reasons for this imbalance could be due to the characteristics of Japanese clinical trials in which only a small number of patients can be accrued from each institution. Especially because the envelope method and block randomization were used in the SGACCS and SGCCC trials, there might be a chance that such a considerable imbalance could happen. To decrease the systematic biases in the SGACCS trial, examination of the original randomization record from each institution was performed. This investigation demonstrated that 13 institutions made obvious violations in the random allocation process. For example, one of these institutions allocated six patients to the treatment group, but no patient to the control group. Exclusion of the patient data obtained from these institutions with inadequate randomization was performed. We excluded 137 patients from the treatment group and 48 patients from the control group in the final analysis, resulting in a decrease in eligible cases for meta-analysis in the SGACCS study (Table 2).
Analysis of Disease-free Survival
Overall DFS curves for the three trials together are shown in Fig.1. Objective recurrences were observed in 572 of 1900 (30.1%) patients allocated to the surgery alone control group and 688 of 2702 (25.5%) patients allocated to the adjuvant chemotherapy group.
Table 1. Three randomized clinical trials comparing surgery alone to chemotherapy using oral fluoropyrimidines in patients with curatively resected colorectal cancer - details of the study designs
| Trial | SGCCC (13) | SGACCS:#2 (11) | TSGHCFU:#3 (12) |
| Oral fluoropyrimidines | 5-fluorouracil | tegafur | carmofur |
| Additional chemotherapy | Mitomycin C | Mitomycin C | none |
| Randomization | envelope method | envelope method | telephoning |
| Patient accrual | 23 months | 24 months | 36 months |
| (dates) | 1984-85 | 1984-85 | 1985-88 |
| Follow-up period | 5 years | 5 years | 5 years |
| Eligibility criteria | |||
| Operation | Curative Resection Colorectal Cancer |
Curative Resection Colorectal Cancer |
Curative Resection Colorectal Cancer |
| Stage of cancer | >SS/A1 or Lymph node meta (+) | Unspecified | Depth of invasion from PM to Si/Ai |
| Patient's age | under 75 years | Unspecified | under 75 years |
| Previous CMX/RX | ineligible | ineligible | ineligible |
| Double or multiple cancers | ineligible | ineligible | ineligible |
| Others | no severe complications | no severe complications | no severe complications |
Table 2.
| Trial | Arm | Treatment | No. patients | % | ||
| colon | rectum | total | eligible | |||
| SGCCC | 1. | MMC 12 mg/m2 (i.p.) at operation + MMC 6 mg/m2 (i.v.) 2 weeks, 1, 2, 4, 6 months after operation, AND 5FU 200 mg/body (p.o.) daily from 2 weeks to 6 months after operation | 312 | 342 | 654 | 97 |
| 2. | MMC 6 mg/m2 (i.v.) 2 weeks, 1, 2, 4, 6 months after operation, AND 5FU 200 mg/body (p.o.) daily from 2 weeks to 6 months after operation | 346 | 320 | 666 | 98 | |
| 3. | Surgery alone | 307 | 328 | 635 | 98 | |
| SGACCS:#2 | 1. | MMC 12 mg/m2 (i.v.) 1 p.o.d., 6 mg/m (i.v.) 2, 4, 6, 8, 10, 12 months after operation AND Tegafur 300mg/body (p.o.) daily from 2 weeks to 12 months after operation | 650 | 645 | 1295 | 87a |
| 2. | Surgery alone | 584 | 598 | 1182 | 91 | |
| TSGHCFU:#3 | 1. | Carmofur 3 mg/kg (p.o.) daily from 1 month to 12 months after operation | 50 | 37 | 87 | 99 |
| 2. | Surgery alone | 43 | 40 | 83 | 99 | |
| Total | 2292 | 2310 | 4602 | 93 | ||
Table 3.
| Risk ratio | Confidence interval | P-value | Test for heterogeneity | |
| Mortality | ||||
| Colon cancer | 1.048 | (0.873, 1.259) | 0.612 | 0.803 |
| Rectal cancer | 0.857 | (0.734, 0.999) | 0.049 | 0.096 |
| Total | 0.932 | (0.829, 1.044) | 0.242 | 0.159 |
| Recurrence | ||||
| Colon cancer | 0.936 | (0.787, 1.113) | 0.453 | 0.041 |
| Rectal cancer | 0.767 | (0.656, 0.882) | 0.0003 | 0.139 |
| Total | 0.830 | (0.742, 0.929) | 0.0012 | 0.017 |
Table 4.
| Risk ratio | Confidence interval | P-value | Test for heterogeneity | |
| Mortality | ||||
| Dukes' A | 0.917 | (0.596, 1.412) | 0.695 | 0.527 |
| Dukes' B | 0.899 | (0.732, 1.105) | 0.313 | 0.481 |
| Dukes' C | 0.828 | (0.711, 0.965) | 0.015 | 0.175 |
| Total | 0.932 | (0.829, 1.044) | 0.242 | 0.159 |
| Recurrence | ||||
| Dukes' A | 0.671 | (0.417, 1.080) | 0.100 | 0.663 |
| Dukes' B | 0.741 | (0.609, 0.903) | 0.0029 | 0.268 |
| Dukes' C | 0.761 | (0.659, 0.878) | 0.0002 | 0.113 |
| Total | 0.830 | (0.742, 0.929) | 0.0012 | 0.017 |
DFS RRs are shown for individual trials depending on the site of cancer and overall (Fig. 2). The RRs for individuals trials in colon and rectal cancer ranged from 0.29 to 1.02 and were <1 except in colon cancer patients of the SGACCS trial whose RR was 1.02. The overall DFS RR was 0.83 (95% CI 0.74-0.93). The overall test for treatment effect was significant (P = 0.0012). The effect was distinct in rectal cancer but was not significant in colon cancer (Table 3). Analysis stratified by the Dukes' stage showed both significant results in Dukes' B(P = 0.0029) and in Dukes'C (P = 0.0002) tumors (Fig. 3, Table 4). Significant differences in DFS were observed in rectal cancers both in Dukes'B (P = 0.0001) and Dukes'C (P = 0.0003), whereas no significant difference was detected in colon cancers either in Dukes'B (P = 0.296) or in Dukes'C (P = 0.0863). A test for interaction computed to test the hypothesis that the effect of treatment was the same, except for random fluctuations, between rectal (RR 0.767) and colon (RR 0.936) cancer (Table 3). For DFS this test demonstrated a borderline difference (df = 1; P = 0.075).
Figure 2. DFS RRs of colon cancer and rectal cancer patients in individual trials and overall. Test for treatment effect, P = 0.0012. Figure 3. DFS curves in Dukes' C colorectal cancer patients. Overall survival curves for all the trials together are depicted in Fig. 4. There was no significant survival advantage of oral fluoropyrimidines over surgery alone (P = 0.24). Deaths were recorded in 500 of 1900 (26.3%) patients allocated to the surgery alone control group and 664 of 2702 (24.2%) patients allocated to the adjuvant chemotherapy group. Figure 4. Overall survival curves. Survival RRs are shown for individual trials depending on the site of cancer and overall (Fig. 5). The RRs for individuals trials in colon and rectal cancer ranged from 0.72 to 1.36 and only the SGCCC trial in rectal cancer showed a statistically significant benefit of chemotherapy (RR 0.72; 95% CI 0.57-0.90). Although the overall survival RR of 0.93 (95% CI 0.83-1.05) showed no significant treatment effect, a borderline significant benefit was detected in rectal cancers (Table 3). Figure 5. Survival RRs of colon cancer and rectal cancer patients in individual trials and overall. Test for treatment effect, P = 0.024. Analysis by the Dukes' stage showed significant effect of oral fluoropyrimidines for survival in Dukes'C (P = 0.015) but not in Dukes'B (P = 0.313) (Fig. 6, Table 4). The significant difference in survival between chemotherapy group and control group was remarkable in the Dukes'C rectal cancers (P = 0.0124), whereas the difference in the Dukes'B rectal cancer was not significant (P = 0.1088). No significant difference was detected either in Dukes'B (P = 0.721) or in Dukes'C (P = 0.417) colon cancers. Figure 6. Survival curves in Dukes'C colorectal cancer patients. Test for interaction again demonstrated a tendency that the effect of chemotherapy with oral fluoropyrimidines is more pronounced in patients with rectal cancer (RR 0.857) than in patients with colon cancer (RR 1.048) (P = 0.0974). The analysis of the prognostic value of gender, age (<60 or >60 years of age) and histological types of tumor using control data demonstrated none of these factors to be significant predictors either for DFS or for survival. As expected, histological depth of invasion and nodal status were determined to be important prognostic factors. The relative risk ratio of each category to that of basic depth of invasion (m, sm, pm) and to basic nodal status [n0] demonstrated that the relative risk gains considerably as histological depth of invasion or histologically reconfirmed nodal status proceed. The chi-squared test was applied to evaluate the difference in the distribution of histological depth of invasion and nodal status between the control and the treatment groups. The chi-squared test statistics of the nodal status (df = 4) were 18.546 (P = 0.001) for colon cancer and 16.188 (P = 0.003) for rectal cancer. The statistics of the depth of invasion (df = 5) were 10.729 (P = 0.057) for colon cancer and 22.647 (P = 0.001) for rectal cancer (Table 5). This small imbalance of histopathological tumor extent could be related to the time of randomization, which was performed just after the operation in those trials. Since histopathological data with regard to the extent of the disease were not available at the time of randomization, it is not surprising that such an imbalance of these important prognostic factors exists between the two arms. Evaluation of the influence of pathological depth of invasion and nodal status between the treatment and control groups was performed. The degree of depth of invasion was grouped into three categories [(1) m, sm, pm; (2) ss/al; and (3) s/a2, si/ai] and the degree of lymph node metastasis was also grouped into three categories [(1) n0; (2) n1; and (3) n2, n3, n4]. These two factors plus the site of cancer were evaluated as explanatory variables for the Cox's proportional hazards model in the satellite symposium of the 29th Meeting of the Japan Society for Cancer Therapy and the details of the methodology and results will be published elsewhere (Hamada et al., unpublished reports). Adjusted risk ratios were calculated under the proportional hazard model using eight dummy variables which correspond to the histological depth of invasion (three categories) × nodal status (three categories) cross classification (Table 6). The overall adjusted DFS RR was 0.76 (95% CI 0.68-0.85) and the overall test for treatment effect was highly significant (P < 0.0001).The overall adjusted survival RR was 0.87 (95% CI 0.77-0.98) and the overall test for treatment effect also demonstrated a significant benefit of adjuvant chemotherapy with oral fluoropyrimidines for colorectal cancer patients (P = 0.017) in this adjusted analysis. Table 5. Although a considerable number of 5-FU derivatives have been developed and tested, their clinical development was almost discontinued in Western countries during the 1980s, probably because of their relatively high gastrointestinal toxicity and unclear additional benefit compared with the original 5-FU, which was reported in early phase II studies. Another probable reason for the lack of popularity of these fluoropyrimidine derivatives among Western oncologists could be because they are mostly administered orally on an outpatient basis with a total daily dosage limited to half or even one third of the dose administered through the venous route. In Japan, however, these agents have been widely used for long-term adjuvant chemotherapy against gastrointestinal cancers, sometimes combined with a few injections of low-dose mitomycin C. Although two clinical trials of immunochemotherapy demonstrated a significant effect against minimal residual disease in colorectal cancers (14,16) no adjuvant chemotherapy trial proved its effect and such uncertainty precluded Western oncologists from using heavy chemotherapy in curatively resected colorectal cancers. One of the advantages of oral fluoropyrimidines is that oncologists can prescribe these agents on an outpatient basis for longer periods without too much fear of severe side-effects. Therefore, although those oral 5-FU derivatives were not reported to have discernible benefit in Western high-dose phase II studies, Japanese oncologists continued to use them, with low-dose, long-term administration to outpatients. Reappraisal of these oral fluoropyrimidines have been started again in Western countries since the early 1990s, in the hope that they might be able to give similar or even better results compared with intravenous adminisration of the original 5-FU in an adjuvant therapy to curatively resected gastrointestinal cancer patients (17). The aim of this meta-analysis was to examine randomized clinical trials which had evaluated the effect of those oral fluoropyrimidines against curatively resected colorectal cancer used in adjuvant chemotherapy. Our investigation revealed that only Japanese trials which had started before the end of 1980s had mature 5-year survival and DFS data in terms of the evaluation of oral fluoropyrimidines. We also restricted our meta-analysis to properly conducted randomized trials with surgery alone as control, in order to keep a common basis for comparison of all therapies. Since, in most of the clinical trials at that time, randomization was performed by the envelope method according to the intraoperative macroscopic findings, certain imbalances in number between control and treatment arms were noticed. After checking all the individual patient data and randomization records, we have reason to believe that the imbalances were not caused by intentional exclusions or selection of the patients performed retrospectively. Table 6. By pooling the data on almost 5000 patients from three trials in colorectal cancer, we have shown a significant overall benefit of oral fluoropyrimidines in terms of DFS [attributable risk reduction (ARR) 17%]. Adjustment of the histopathological background by the Cox regression model confirmed our results by showing a more significant effect of these agents (ARR 23.8%).With regard to survival, the overall risk reduction was 6.8% and the advantage was not significant. However, a borderline significant benefit was demonstrated in rectal cancer (ARR 14.3%) and also a significant difference was observed in Dukes' C cancers (ARR 23.9%). The significant effect in rectal cancer could be an issue that might be worth looking at. Interestingly, a previous study of a colorectal cancer meta-analysis by Buyse et al. (8) reported similar findings.They found that the effect of chemotherapy was much more pronounced in patients with rectal cancer (ARR 38%, P = 0.02) than in patients with colon cancer (ARR -8%) and a dependence between treatment effect and tumor site was confirmed by a significant test for interaction (P = 0.02). The result of our meta-analysis is in line with that report, where the effect of oral fluoropyrimidines was obvious in patients with rectal cancer (ARR 14.3%, P = 0.49) but not in patients with colon cancer (ARR -4.8%, not significant). The interaction test for the dependence between treatment effect and tumor site also tended to lend support to the hypothesis (P = 0.097). It is also of interest to consider the effect of the chemotherapy according to the tumor stage. In our meta-analysis, a significant survival advantage was observed in Dukes' C patients (ARR 17.2%, P = 0.0156), but the difference was not as obvious in Dukes' B patients (ARR 8.3%, P = 0.313). It has been claimed by several investigators that the effect of adjuvant chemotherapy is greater in stages of advanced regional spread than in early stages (18-20) and our meta-analysis demonstrated similar findings in the case of chemotherapy with oral fluoropyrimidines. It is possible that the enrollment of patients with less advanced disease in the three trials (Dukes' A:B:C = 1:2:2) might have precluded them from showing a discernible advantage in terms of overall survival. Although extreme caution should be exercised when interpreting results from subsets of the trials, this subset analysis appeared necessary in view of the important indications and target of our adjuvant chemotherapy. Although our results were not obtained in double-blind placebo-controlled trials, the significant effect of oral fluoropyrimidines for survival in rectal cancer and in Dukes' C patients and the significant overall benefit in terms of DFS could be a strong suggestion of the effectiveness of low-dose, long-term administration schedules in an adjuvant setting for curatively resected colorectal cancers. Adjustment of the histopathological background by the Cox regression model confirmed our results by showing a more significant benefit of these agents. Since most of the cancer cells of minimal residual disease are postulated to be in dormant (GO) phase after the radical surgery, strong but intermittent parenteral administration might be less effective, compared with the low-dose, long-term administration of such chemotherapeutic agents. Two important aspects should not be ignored with regard to the results of our meta-analysis. First, since there had been no nationwide registry system for the clinical trials in Japan until April 1997, the possibility still remains that some unreported small trials with insignificant results might have been overlooked and not included in our meta-analysis. Second, although there was no signifcant heterogeneity with regard to the patients' characteristics between the trials, some heterogeneities exist in terms of treatment regimens and schedules and also in terms of trial results. One of the points of interest is the effect of MMC combined with oral fluoropyrimides. In two of the three selected eligible clinical trials (SGCCC and SGACCS), the treatment regimen contained injections of MMC in the treatment arm. Although heterogeneity exist in terms of survival RRs between these two large trials (SGCCC 0.806, SGACCS 1.016), it is possible that the combination of oral fluoropyrimidies and MMC is effective for curatively resected colorectal cancers. An additional sensitivity test demonstrating that the result of this meta-analysis is essentially consistent even after exclusion of the TSGHCFU trial (Hamada et al., unpublished data) supports this assumption. Even though the value of truly effective therapies is ultimately measured in terms of overall survival, the results of our meta-analysis show promise for the use of adjuvant chemotherapy with oral 5-FU and oral fluoropyrimidine derivatives against curatively resected colorectal cancers. Furthermore, considering the fact that not all the patients complied with the treatment assigned, this meta-analysis probably tended to underestimate, or at least did not overestimate, the size of the treatment benefits in patients who did comply. If the oral administration of fluoropyrimidine derivatives is effective (at least effective as an intravenous administration of 5-FU as performed in Western countries), these agents might be very attractive in terms of quality of life of the patient and also in terms of cost-effectiveness of the cancer chemotherapy. In this regard, we look forward to the results of the clinical studies started in the USA and in Europe to confirm the effect of one of these oral fluoropyrimidines in colorectal carcinoma (21,22). This work was partly supported by the Japanese Society of Strategies for Cancer Research and Therapy
Analysis of Survival
Prognostic Factors in the Control Data and Their Distribution Between Groups
Adjusted Analysis
(1) Depth of invasion
Colon cancer
Rectal cancer
Treatment
Control
Treatment
Control
Mucosa (m)
17
14
15
28
Submucosa (sm)
20
28
57
49
Muscularis propria (pm)
124
99
295
243
Subserosa (ss, a1)
626
429
452
306
Serosa (s, a2)
475
290
447
316
Adjacent organ invasion (si, ai)
96
74
78
24
Total
1358
934
1344
966
Chi-square test
10.729
22.647
P-value
0.057
0.001
(2) Lymph node metastasis
Colon cancer
Rectal cancer
Treatment
Control
Treatment
Control
n0
767
599
753
616
n1
358
219
363
201
n2
172
92
187
122
n3
51
23
33
24
n4
10
1
8
3
Total
1358
934
1344
966
Chi-square test
18.546
16.188
P-value
0.001
0.003
DISCUSSION
Risk ratio
Confidence interval
P-value
Test for heterogeneity
Mortality
Colon cancer
0.956
(0.800, 1.143)
0.623
0.740
Rectal cancer
0.812
(0.697, 0.946)
0.007
0.125
Total
0.869
(0.774, 0.976)
0.017
0.241
Recurrence
Colon cancer
0.854
(0.720, 1.012)
0.069
0.241
Rectal cancer
0.702
(0.607, 0.813)
0.0001
0.157
Total
0.762
(0.682, 0.851)
0.0001
0.014
Acknowledgment
References
This article has been cited by other articles:
This page is run by Oxford University Press, Great Clarendon Street, Oxford OX2 6DP, as part of the OUP Journals
Comments and feedback: www-admin{at}oup.co.uk
Last modification: 4 Mar 1999
Copyright© 1999 Foundation for Promotion of Cancer Research, 1999.
CiteULike 
Connotea 
Del.icio.us What's this?
N. Tomita, M. Fukunaga, S. Okamura, H. Narahara, N. Uedo, R. Ishihara, H. Ishida, H. Furukawa, M. Gotoh, and H. Takiuchi
Phase I/II Study of CPT-11 plus UFT in Patients with Advanced/Recurrent Colorectal Cancer: Osaka Gastrointestinal Cancer Chemotherapy Study Group (OGSG): Protocol 0102
Jpn. J. Clin. Oncol.,
July 1, 2007;
37(7):
521 - 527.
[Abstract]
[Full Text]
[PDF]
F. Vigna-Taglianti, P. Vineis, A. Liberati, and F. Faggiano
Quality of systematic reviews used in guidelines for oncology practice
Ann. Onc.,
April 1, 2006;
17(4):
691 - 701.
[Abstract]
[Full Text]
[PDF]
J. Sakamoto, C. Hamada, M. Rahman, S. Kodaira, K. Ito, H. Nakazato, Y. Ohashi, and M. Yasutomi
An Individual Patient Data Meta-analysis of Adjuvant Therapy with Carmofur in Patients with Curatively Resected Colon Cancer
Jpn. J. Clin. Oncol.,
September 1, 2005;
35(9):
536 - 544.
[Abstract]
[Full Text]
[PDF]
A. Figueredo, M. L. Charette, J. Maroun, M. C. Brouwers, and L. Zuraw
Adjuvant Therapy for Stage II Colon Cancer: A Systematic Review From the Cancer Care Ontario Program in Evidence-Based Care's Gastrointestinal Cancer Disease Site Group
J. Clin. Oncol.,
August 15, 2004;
22(16):
3395 - 3407.
[Abstract]
[Full Text]
[PDF]
Efficacy of Oral Adjuvant Therapy After Resection of Colorectal Cancer: 5-Year Results From Three Randomized Trials
J. Clin. Oncol.,
February 1, 2004;
22(3):
484 - 492.
[Abstract]
[Full Text]
[PDF]
T. Tsuji, S. Hidaka, T. Sawai, T. Nakagoe, H. Yano, M. Haseba, H. Komatsu, H. Shindou, H. Fukuoka, M. Yoshinaga, et al.
Polymorphism in the Thymidylate Synthase Promoter Enhancer Region Is Not an Efficacious Marker for Tumor Sensitivity to 5-Fluorouracil-based Oral Adjuvant Chemotherapy in Colorectal Cancer
Clin. Cancer Res.,
September 1, 2003;
9(10):
3700 - 3704.
[Abstract]
[Full Text]
[PDF]
This Article 
Abstract
Alert me when this article is cited
Alert me if a correction is posted
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Add to My Personal Archive
Download to citation manager
Search for citing articles in:
ISI Web of Science (40)
Request Permissions
Google Scholar
Articles by Sakamoto, J
Articles by Ohashi, Y
Search for Related Content
PubMed
PubMed Citation
Articles by Sakamoto, J
Articles by Ohashi, Y
Social Bookmarking
What's this?