|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Originally published as JCO Early Release 10.1200/JCO.2004.09.059 on April 5 2004 © 2004 American Society of Clinical Oncology. Pooled Analysis of Fluorouracil-Based Adjuvant Therapy for Stage II and III Colon Cancer: Who Benefits and by How Much?From the Mayo Clinic and Mayo Foundation, Rochester, MN; British Columbia Cancer Agency, Vancouver, British Columbia; National Cancer Institute of Canada–Clinical Trials Group, Queens University, Kingston, Ontario, Canada; University of Pennsylvania Cancer Center, Philadelphia, PA; Southwest Oncology Group Statistical Center, Seattle, WA; University of Siena, Siena; Ospedali Riuniti, Bergamo, Italy; University of the Mediterranean, Marseilles and Fédération Francophone de Cancérologie Digestive, Dijon, France; University of North Carolina at Chapel Hill, Chapel Hill, NC. Address reprint requests to Charles L. Loprinzi, MD, Medical Oncology, Mayo Clinic, 200 First St Southwest, Rochester, MN 55905; e-mail: cloprinzi{at}mayo.edu
PURPOSE: Although it is well-established that fluorouracil- (FU-) based adjuvant therapy improves survival for patients with resected high-risk colon cancer, the magnitude of adjuvant therapy benefit across specific subgroups and for individual patients has been uncertain. PATIENTS AND METHODS: Using a pooled data set of 3,302 patients with stage II and III colon cancer from seven randomized trials comparing FU + leucovorin or FU + levamisole to surgery alone, we performed an analysis based on a Cox proportional hazards regression model. Treatment, age, sex, tumor location, T stage, nodal status, and grade were tested for both prognostic and predictive significance. Model derived estimates of 5-year disease-free survival and overall survival (OS) for surgery alone and surgery plus FU-based therapy were calculated for a range of patient subsets. RESULTS: Nodal status, T stage, and grade were the only prognostic factors independently significant for both disease-free survival and OS. Age was significant only for OS. In a multivariate analysis, adjuvant therapy showed a beneficial treatment effect across all subsets. Treatment benefits were consistent across sex, location, age, T-stage, and grade. A significant stage by treatment interaction was present, with treatment benefiting stage III patients to a greater degree than stage II patients. CONCLUSION: Patients with high-risk resected colon cancer obtain benefit from FU-based therapy across subsets of age, sex, location, T stage, nodal status, and grade. Model estimates of survival stratified by T stage, nodal status, grade, and age are available at http://www.mayoclinic.com/calcs. This information may improve patients' and physicians' understanding of the potential benefits of adjuvant therapy.
In 2003, an estimated 106,000 Americans were diagnosed with a primary colon cancer.1,2 Approximately 40% had lymph node involvement and 15% to 20% had node-negative, T3 or T4 disease.3 Thus, more than 60,000 patients were candidates for adjuvant chemotherapy following surgery. Six months of fluorouracil (FU) and leucovorin following resection of stage III colon cancer has become standard adjuvant therapy, credited with an estimated one-third reduction in the risk of colon cancer recurrence.4 Despite several large randomized trials conducted over the past two decades, controversies remain. For stage II colon cancer, the role of chemotherapy is still debated. It is unclear whether the magnitude of adjuvant therapy benefit is consistent across different patient subsets. It has been suggested that the benefit obtained by FU-based chemotherapy may be preferentially attributed to subsets of female patients and patients with right-sided colon tumors.5 Unfortunately, data from individual clinical trials are not typically precise enough to evaluate treatment within small patient subsets. Ideally, decisions regarding adjuvant chemotherapy should be based on accurate estimates of baseline prognosis and the magnitude of incremental benefit that can be achieved, using data from prior studies. Additionally, the risks of therapy and alternative options should be enumerated.6 Newly diagnosed cancer patients consistently report that they desire information regarding their likelihood of cure,7,8 and prefer both qualitative and quantitative information when making management decisions.9-11 At the same time, variations and uncertainty exist when estimating cancer prognosis,12 particularly when translating the results of clinical trials for adjuvant therapy to predictions of benefit for individual patients.13,14 Estimates of prognosis and benefit are often not provided to patients, although this information might significantly affect a patient's decision to pursue treatment.15 In the absence of such information, patients may have misperceptions about the risks of their disease16 and an incomplete understanding of the rationale for adjuvant treatment.15 We therefore performed a pooled analysis to improve our understanding of the magnitude of adjuvant therapy benefit observed for specific subgroups of patients. Recognizing that adjuvant systemic therapy tools for resected breast cancer and melanoma (http://www.mayoclinic.com/calcs) have been well accepted,13,17-19 we provided individualized estimates of baseline prognosis and adjuvant therapy benefit for patients with resected colon cancer based on the results of this pooled analysis.
A pooled analysis of adjuvant chemotherapy trials for patients with resected colon cancer was performed to quantify FU-based chemotherapy benefit for predefined clinical and pathologic subgroups. This data set was previously described in a 2001 analysis of the effects of chemotherapy in the elderly.20 Individual patient data were pooled for analysis from seven selected randomized trials (Table 1). Patients with multiple primary tumors or rectal tumors were excluded. All trials included random assignment to either FU-based chemotherapy or no therapy following surgical resection of stage II or III colon cancer.
The primary end points were disease-free survival (DFS), defined as time from randomization to first of either confirmed recurrence or death, and overall survival (OS), defined as the time from randomization to death. Data were analyzed for up to 8 years from the date of randomization. Patient age, sex, tumor location (left v right, defined respectively as distal or proximal to the splenic flexure), histologic grade, depth of tumor invasion (T stage), and nodal status were recorded. To account for inconsistencies in grade classification systems across different studies, tumors described as well or moderately differentiated, or grade 1 or 2 tumors, were categorized as low-grade, while poorly differentiated, anaplastic, and grade 3 or 4 tumors were categorized as high-grade. Ordinal categories were defined for T stage (T1/T2, T3, and T4) and for the number of positive lymph nodes as per the nodal groupings utilized by the original studies (none, 1-4,  5). Data regarding the absolute number of positive lymph nodes or total nodes examined were not consistently available. DFS and OS estimates for the raw data were determined according to the Kaplan-Meier method.26 A log-rank test stratified for the patient's original protocol comparing patients randomly assigned to FU-based adjuvant treatment to surgery alone was performed within each variable. The primary analysis for efficacy was an intent-to-treat analysis based on a Cox proportional hazards regression model,27 stratified by study, and including terms for treatment, age, sex, location, T stage, grade, and nodal status. Hazard ratios (HRs) and 95% CIs for recurrence and death among treated patients compared with controls were computed for univariate and multivariate models. The appropriateness of the proportional hazards assumption was verified using graphical methods and testing as per Grambsch and Therneau.28 Likely as a result of the large sample size, several covariates tested significant for nonproportional hazards. Visual examination of the estimated time-varying coefficient28 revealed these to be minor variations over time and were mainly as a result of variations in years 5 to 8 of follow-up. Therefore, we proceeded to use the Cox model. A nested likelihood ratio test for a treatment effect interaction was performed to evaluate the predictive significance of each factor considered. Power to detect the various interactions depends on the prevalence of each factor, and the number of factor levels. As a general guide, the sample size of 3,302 patients provided 80% power to detect an interaction reflected by a HR of 1.4 for a two-level factor with a prevalence split of 75% versus 25%. Factors with higher prevalence would have greater power for an interaction to be detected, conversely for factors with a lower prevalence. All statistical tests were two-sided, with P values of less than .05 used to denote statistical significance. Complete data was available on age, sex, nodal status, and location on all cases. As a small proportion of patients were missing T stage (8%) and grade (2%) data, multiple imputation was used to allow inclusion of these cases.29 There was no reason to believe that missing grade or T stage values were dependent on the unobserved grade and stage themselves, however logistic regression suggested that the missingness of these two variables did depend on nodal status. Therefore, a missing-at-random mechanism for the missing data was assumed and implemented based on multiple imputation using IVEware (University of Michigan Survey Research Center, Ann Arbor, MI).30,31 Five imputation data sets were created and five sets of analyses were combined per Rubin's formula.29 The Cox proportional hazards model was used to estimate treatment effects within specific patient subsets. This prediction model was derived using the full data set (N = 3,302). The heuristic shrinkage estimator was used to test for model overfitting,32 with values close to 1 indicating no overfitting.32 The internal validation of the final model was assessed using bootstrap resampling.33 The c-statistic was used to measure discriminating ability with the bootstrap used to correct for overfit. Calibration of the prognostic model was evaluated by comparing the predicted baseline prognosis and adjuvant therapy benefit to the observed Kaplan-Meier survival probabilities for patients sorted by grade, nodal status, T stage, and treatment among groups exceeding 30 patients. Three factors exhibited significant between-trial heterogeneity as prognostic factors of DFS: grade (P = .005), age (P = .0006), and tumor location (P = .008). For tumor grade, the heterogeneity was quantitative, that is, all effects were in the same direction but of varying magnitude. Thus, a single term for grade was applied in the multivariate models. For location and age, however, the heterogeneity were qualitative—location was significantly prognostic for DFS in a single study and had little effect in the six remaining studies. A likewise finding was observed for age. Location and age were thus not included in the final predictive model for DFS as a result of both the observed heterogeneity and the lack of a univariate prognostic effect for either factor.
Patient Characteristics Of the 3,302 patients included from the seven trials (Table 1), 51% (1,681 patients) were assigned to FU-based adjuvant therapy and 49% (1,621 patients) to no therapy. As presented in Table 2, 1,440 patients (44%) were node-negative while 1,399 (42%) had 1 to 4 positive lymph nodes, and 463 (14%) had five or more positive nodes. The distribution of patient characteristics was similar for the treatment and control groups.
Factors Affecting Baseline Prognosis In univariate analyses of DFS, nodal status was the most significant prognostic factor, with a relative risk of 2.11 for patients with 1 to 4 positive nodes and 4.23 for patients with five or more nodes, compared to those patients with no positive nodes.(Table 3) Deeply penetrating tumors and high grade also correlated with inferior prognosis. Tumor location, sex, and age were not independently prognostic for DFS. Similar conclusions were observed for OS, with the exception of a poorer survival for patients age 60 years or older relative to those less than 60 years (HR 1.20; 95% CI, 1.03 to 1.40).
Efficacy of Adjuvant Therapy According to Prognostic Subsets Patients randomly assigned to treatment achieved a 5-year DFS of 67% as compared to 55% (corresponding to a 30% proportional reduction in risk of recurrence [HR, 0.70; range, 0.63 to 0.78]). OS improved from 64% to 71% at 5 years with adjuvant chemotherapy (a 26% proportional reduction in risk of death [HR 0.74; range, 0.66 to 0.83]). Using the index of 60-day mortality as a measure of treatment-related mortality, there were two deaths in 1,621 untreated patients compared with 12 deaths in 1,681 treated patients (0.1% v 0.7%).
By univariate analysis, a statistically significant benefit was seen with FU-based adjuvant therapy for DFS and OS across all subsets of age, sex, and primary location (Table 4). The relationship between T stage, treatment, and survival was consistent, with treatment providing a significant benefit in T3 and T4 patients and a similar trend toward improvement in T1/T2 patients, although statistical significance was not reached for this smaller subset. A significant improvement in DFS and OS was observed among patients with low-grade tumors. In the smaller subset of patients with high-grade tumors, chemotherapy did not improve survival in univariate analysis. However, a test for interaction between treatment and grade was not significant (P = .16), nor was interaction testing between treatment and any of the following: age, sex, location, or T stage (P > .10 for each factor). With respect to treatment effect within subgroups of nodal status, a statistically significant treatment benefit was observed in each nodal subgroup for DFS. For OS, a treatment benefit was observed among patients with node-positive disease, with a smaller improvement (P = .1127) seen in node-negative disease (Table 4). Although benefit exists across all nodal subgroups, a significant interaction was observed between treatment and nodal status (P = .01), indicating that the magnitude of benefit varied depending on nodal status. For DFS, the HRs comparing treatment to control by nodal subgroup were as follows: 0 nodes, HR = 0.831; 1 to 4 nodes, HR = 0.605,
Building on these univariate results, we constructed multivariate models to estimate the individual baseline prognosis for patients with resected stage II and III colon cancer. The pooled analysis identified three factors that independently predict baseline prognosis: nodal status (0 v 1-4 v 5), depth of invasion (T1/T2 v T3 v T4) and grade (low-grade v high-grade). For DFS, these three factors were used as the basis for the prognostic model as the other factors considered (age, sex, location) had HRs close to 1 and univariate P values for DFS > 0.20. Age was included for the end point of OS. Calculation of treatment benefits for individual T stage, nodal status, and grade strata was performed by adjusting the baseline survival probability, defined as estimated survival with surgery alone, by the HR for adjuvant therapy. The HR reduction with treatment is assumed to remain constant over time. Given the previously described differential treatment benefit by nodal status, the final model includes a term to account for the interaction between treatment effect and nodal status. We subsequently tested a model that also accounted for the suggestive interaction between grade and treatment, as was illustrated in Table 4. This did not yield improved performance characteristics and, since grade was not uniformly defined across studies, an interaction term for treatment and grade was not included in the final model. Table 5 presents computed estimates for 5-year DFS among patients treated with surgery alone compared with patients treated with surgery plus adjuvant FU-based chemotherapy. These are accompanied by 95% CIs to reflect the precision of the point estimates. Age-specific estimates for OS were computed similarly (Table 6).
Using the end point of 5-year DFS, the model was evaluated for discriminative ability, calibration, and overfitting. The c-statistic with the full data set was 0.660 (range, 0 to 1.0) reflecting a moderate measure of discrimination, and remained stable after bootstrap correction (0.655). The calibration of the model was assessed by comparing the predicted 5-year DFS rates for surgery alone and for surgery plus adjuvant therapy with the corresponding observed or actual Kaplan-Meier estimates probabilities (provided in Table 7). A calibration plot of actual versus predicted probabilities is presented in Figure 1 and illustrates acceptable calibration. Finally, the shrinkage estimate of the final model is very close to 1 (0.9797), thus indicating no significant overfitting.
Development of a Web-Based Adjuvant Therapy Tool for Colon Cancer To consolidate this information as an easily accessible tool, an Internet program was developed which would allow physicians to input a patient's age, T stage, lymph node status, and tumor grade. The output of the tool presents tailored estimates of 5-year DFS and OS probabilities with surgery alone and with surgery plus adjuvant FU-based chemotherapy. This format assists physicians in providing prognostic and predictive information for an individual patient so that the physician, patient, and family can obtain the most applicable estimates regarding the risks and benefits of adjuvant therapy.
A pooled analysis was performed based on data from seven clinical trials that included patients randomly assigned to FU-based adjuvant chemotherapy or surgery alone, with the goal of estimating treatment effects in subsets of patients for whom individual study results had remained controversial. The prognostic and predictive model derived from this analysis was developed to provide clinically useful data regarding the utility of adjuvant therapy for stage II and III colon cancer in an accessible format. The meaning of a proportional reduction of risk, a commonly utilized statistical approach, is not readily grasped by most patients and many physicians,13 and the availability of 5-year survival statistics may provide a more instructive metric for understanding the potential benefits of adjuvant systemic therapy. Our analysis confirms that significant prognostic factors include nodal status, histologic grade, and depth of tumor invasion into the bowel wall.34,35 Baseline prognosis is not influenced by patient sex or tumor location. While age does have an impact on OS, elderly patients with colon cancer do not have inferior DFS when compared to younger patients.20,36 Due to data availability, there are potential prognostic factors not considered in this analysis. Previous studies have indicated that obstruction of the colon by tumor, the total number of lymph nodes examined, preoperative carcinoembryonic antigen levels, and DNA ploidy and proliferative index may have prognostic value for patients with resected colon cancer.37-41 While these studies adjusted for stage, the histologic grade and the degree of lymph node involvement were not uniformly evaluated and, hence, the additive contribution of these factors to our model is unknown. Although some oncologists may consider such factors when evaluating patients, it is unclear whether their inclusion would meaningfully alter baseline prognoses beyond our current estimates. Similarly, molecular markers including microsatellite instability, allelic loss of chromosome 18q, and tumor expression of thymidylate synthase were not included; their clinical prognostic and predictive utility remains undefined and their use is presently limited to research applications.42-44 The current analysis also evaluated the question of a differential treatment efficacy by location and sex, as had been previously suggested. Elsaleh et al5 retrospectively reviewed 656 non–randomly assigned patients with Dukes' C colon and rectal cancers and reported a more favorable survival benefit from adjuvant chemotherapy among patients with right-sided tumors and among women. As depicted in Figure 2 and Table 4, our analysis fails to support any evidence of a differential treatment benefit by location or sex. The observed differences in the Elsaleh analysis5 were ascribed to a correlation of this phenotype with a higher frequency of microsatellite instability (MSI), leading the authors to conclude that adjuvant FU-based therapy may predominantly benefit patients with tumors exhibiting a high level of MSI (MSI-H). This conclusion was recently contradicted by a molecular analysis of 570 patients enrolled in FU-based adjuvant therapy trials for stage II and III colon cancer.45 While patients with MSI-H tumors exhibited overall improved survival outcomes, those treated with chemotherapy did not show benefit. These data are pending confirmation from future prospective studies. Tumors with MSI are associated with distinct features including proximal location, an earlier stage distribution, and poor differentiation.44,46,47 It is of interest that our univariate analysis suggested a lack of treatment benefit for poorly differentiated tumors (Table 4). It could be hypothesized that poor differentiation may serve as a surrogate marker for MSI, and a reported greater proportion of MSI-H tumors in high-grade disease44,48 may contribute to the attenuated treatment benefit suggested in our series. However, without supporting prospective data, this hypothesis is uncorroborated and in our multivariate analysis, a statistically significant interaction between treatment and grade was not observed. Hence a differential treatment benefit by grade of disease was not applied in our model.
The model-derived estimates pertaining to node-negative disease deserves special comment. In the International Multicenter Pooled Analysis of Colon Cancer Trials (IMPACT) B2 meta-analysis of stage II subgroups from five adjuvant trials, nonsignificant trends for improvements in DFS (73% v 76%) and OS (80% v 82%) were reported.49 On the other hand, a separate pooled-analysis of four consecutive National Surgical Adjuvant Breast and Bowel Project adjuvant trials demonstrated consistent treatment benefits in both stage II and stage III patients, but interpretation was limited because of the heterogeneity of the treatment and control arms in the included trials.50 Our current data set includes 1,440 patients with node-negative disease, utilizing patients pooled from the five IMPACT B2 trials plus stage II patients from two additional trials.21,23 In our univariate analysis, improvements with adjuvant chemotherapy did reach statistical significance for 5-year DFS (72% v 76%; P = .0490) but did not for OS (80% v 81%; P = .1127). It has been estimated that a prospective adjuvant trial with a no-treatment control arm designed to detect a significant survival benefit among stage II colon cancer patients with a baseline prognosis of 80% would likely require a sample size of at least 5,000 patients.51 Furthermore, 5-year survival may be an insufficient measure of long-term outcomes for patients with node-negative disease. Recognizing these limitations, how can existing data be reconciled to provide today's patients with the best available information? While treatment benefit trends are maintained across all nodal subgroups, a differential magnitude of benefit was observed in our analysis with proportional reductions in risk of recurrence and death of 17% and 15%, respectively, for node-negative disease as compared with 40% and 35%, respectively, for node-positive disease. Thus, nodal subgroup-specific estimates are applied in our model to derive predicted survival probabilities. A biologic explanation for a differential effectiveness of adjuvant FU-based therapy between node-negative and node-positive colon cancers has not been elucidated. It may again be postulated that this effect is associated with MSI, as a greater proportion of node-negative colon cancers are observed with the MSI-H phenotype.44 In clinical practice, adjuvant chemotherapy is considered for node-negative patients with high-risk features, including obstructing, adherent, or perforated T4 tumors. The decision to treat patients with T3N0 colon cancers remains a highly debated issue in the realm of adjuvant management. Treatment is being administered to a significant proportion of patients with low-risk stage II colon cancer, despite uncertainty regarding the significance or magnitude of survival benefit, as documented in a recent US Surveillance, Epidemiology, and End Results Medicare population-based study.52 The decision to offer adjuvant therapy for stage II disease needs to be individualized to the circumstances of each specific patient, and should be balanced against the possible risks of treatment-related toxicity. The stratified estimates of baseline prognosis and potential treatment benefit provided from this analysis may assist oncologists in presenting better prognostic information to patients and facilitate more informed decision-making regarding the potential utility of adjuvant therapy in stage II disease. The use of an accessible tool for presentation of estimated risks is not without pitfalls. Historically, only a select proportion of cancer patients participate in clinical trials, and thus the group of patients on whom this analysis was based may not be representative of all patients seen by physicians, even after accounting for similar stage of disease. One could also question whether it is more appropriate to apply observed data from individual prognostic subsets (Table 7) versus estimates derived from statistical modeling. While it may seem ideal to use actual data, this would require several thousands of patients to accurately predict individual outcomes. Such a database does not exist. The proposed model is derived from pooled data of prospective randomized trials, collectively representing the largest data set available, to provide patients with as reliable an estimate as is currently feasible. The model is reasonably accurate and valid in its predictions of 5-year DFS and OS and can be modified in the future as new prognostic information and new adjuvant therapies become available. For example, if a combination regimen containing oxaliplatin or irinotecan is proven to confer a further proportional reduction of risk of recurrence and death, then this proportional benefit can be applied to the existing model. This tool is intended for use as an aid for adjuvant systemic therapy decisions and, as with all prognostic tools, is to be interpreted within the context of individual patient preferences and a careful discussion between the oncologist, patient, and their family. The results of the model estimates from this study are summarized and available for use by physicians at http://www.mayoclinic.com/calcs.
The authors indicated no potential conflicts of interest.
Supported by grant CA-25224. Presented in part at the 39th Annual Meeting of the American Society of Clinical Oncology, Chicago, IL, May 31-June 3, 2003. Authors' disclosures of potential conflicts of interest are found at the end of this article.
1. SEER Cancer Statistics Review 1973-1999. National Cancer Institute: Bethesda, MD. http://seer.cancer.gov/csr/1973_1999
2. Jemal A, Murray T, Samuels A, et al: Cancer statistics, 2003. CA Cancer J Clin 53:5-26, 2003 3. Eisenberg B, Decosse JJ, Harford E, et al: Carcinoma of the colon and rectum: The natural history reviewed in 1704 patients. Cancer 49:1131-1134, 1982[CrossRef][Medline] 4. Macdonald JS: Adjuvant therapy for colon cancer. CA Cancer J Clin 49:202-219, 1999[Abstract] 5. Elsaleh H, Joseph D, Grieu F, et al: Association of tumour site and sex with survival benefit from adjuvant chemotherapy in colorectal cancer. Lancet 355:1745-1750, 2000[CrossRef][Medline]
6. Leighl N, Gattellari M, Butow P, et al: Discussing adjuvant cancer therapy. J Clin Oncol 19:1768-1778, 2001 7. Degner LF, Kristjanson LJ, Bowman D, et al: Information needs and decisional preferences in women with breast cancer. JAMA 277:1485-1492, 1997[Abstract] 8. Davison BJ, Degner LF, Morgan TR: Information and decision-making preferences of men with prostate cancer. Oncol Nurs Forum 22:1401-1408, 1995[Medline] 9. Cassileth BR, Zupkis RV, Suttan-Smith K, et al: Information and participation preferences among cancer patients. Ann Intern Med 92:832-836, 1980 10. Ravdin PM, Siminoff IA, Harvey JA: Survey of breast cancer patients concerning their knowledge and expectations of adjuvant therapy. J Clin Oncol 16:515-521, 1998[Abstract] 11. Sutherland HJ, Llewellyn-Thomas HA, Lockwood GA, et al: Cancer patients: Their desire for information and participation in treatment decisions. J R Soc Med 82:260-263, 1989[Abstract] 12. Loprinzi CL, Ravdin PM, de Laurentiis M, et al: Do American oncologists know how to use prognostic variables for patients with newly diagnosed primary breast cancer? J Clin Oncol 12:1422-1426, 1994[Abstract]
13. Loprinzi CL, Thome SD: Understanding the utility of adjuvant systemic therapy for primary breast cancer. J Clin Oncol 19:972-979, 2001
14. Rajagopal S, Goodman PJ, Tannock IF: Adjuvant chemotherapy for breast cancer: Discordance between physicians' perception of benefit and the results of clinical trials. J Clin Oncol 12:1296-1304, 1994 15. Siminoff LA, Fetting JH, Abeloff MD: Doctor-patient communication about breast cancer adjuvant therapy. J Clin Oncol 7:1192-1200, 1989[Abstract] 16. Mackillop WJ, Stewart WE, Ginsburg AD, et al: Cancer patients' perceptions of their disease and its treatment. Br J Cancer 58:355-358, 1988[Medline] 17. Thome SD, Loprinzi CL, Heldebrant MP: Determination of potential adjuvant systemic therapy benefits for patients with resected cutaneous melanomas. Mayo Clin Proc 77:913-917, 2002[Medline] 18. Landro L: Going Online to Make Life-and-Death Decisions. The Wall Street Journal, 2002, p. D1
19. Sonpavde G: Communicating the Value of Adjuvant Chemotherapy. J Clin Oncol 21:948-949, 2003
20. Sargent DJ, Goldberg RM, Jacobson SD, et al: A pooled analysis of adjuvant chemotherapy for resected colon cancer in elderly patients. N Engl J Med 345:1091-1097, 2001 21. Laurie JA, Moertel CG, Fleming TR, et al: Surgical adjuvant therapy of large-bowel carcinoma: An evaluation of levamisole and the combination of levamisole and fluorouracil. The North Central Cancer Treatment Group and the Mayo Clinic. J Clin Oncol 7:1447-1456, 1989[Abstract]
22. O'Connell MJ, Laurie JA, Kahn M, et al: Prospectively randomized trial of postoperative adjuvant chemotherapy in patients with high-risk colon cancer. J Clin Oncol 16:295-300, 1998 23. Moertel CG, Fleming TR, Macdonald JS, et al: Levamisole and fluorouracil for adjuvant therapy of resected colon carcinoma. N Engl J Med 322:352-358, 1990[Abstract] 24. Francini G, Petrioli R, Lorenzini L, et al: Folinic acid and 5-fluorouracil as adjuvant chemotherapy in colon cancer. Gastroenterology 106:899-906, 1994[Medline] 25. Efficacy of adjuvant fluorouracil and folinic acid in colon cancer. International Multicentre Pooled Analysis of Colon Cancer Trials (IMPACT) investigators. Lancet 345:939-944, 1995[CrossRef][Medline] 26. Kaplan E, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958[CrossRef] 27. Cox D: Regression models and life-tables. J R Stat Soc B 34:187-202, 1972
28. Grambsch PM, Therneau TM: Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 81:515-526, 1994 29. Little RJ, Rubin D: Statistical analysis with missing data. John Wiley and Sons, Inc, New York, NY, 2002 30. Raghunathan TE, Solenberger PW, Van Hoewyk J: IVEware: Imputation and variance estimation software. Institute for Social Research, University of Michigan, Ann Arbor, MI, 1998 31. Raghunathan TE, et al: A Multivariate Technique for Multiplying Imputing Missing Values Using a Sequence of Regression Models. Surv Methodol 27:85-95, 2001 32. Van Houwelingen JC, Le Cessie S: Predictive value of statistical models. Stat Med 9:1303-1325, 1990[Medline] 33. Harrell FE, Lee KL, Mark DB: Tutorial in biostatistics multivariable prognostic models: Issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med 15:361-387, 1996[CrossRef][Medline] 34. Chapuis PH, Dent OF, Fisher R, et al: A multivariate analysis of clinical and pathological variables in prognosis after resection of large bowel cancer. Br J Surg 72:698-702, 1985[Medline] 35. Cohen AM, Tremiterra S, Candela F, et al: Prognosis of node-positive colon cancer. Cancer 67:1859-1861, 1991[CrossRef][Medline]
36. Iwashyna TJ, Lamont EB: Effectiveness of adjuvant fluorouracil in clinical practice: A population-based cohort study of elderly patients with stage III colon cancer. J Clin Oncol 20:3992-3998, 2002 37. Wolmark N, Fisher B, Wieand HS, et al: The prognostic significance of preoperative carcinoembryonic antigen levels in colorectal cancer. Results from NSABP (National Surgical Adjuvant Breast and Bowel Project) clinical trials. Ann Surg 199:375-382, 1984[Medline] 38. Wolmark N, Wieand HS, Rockette HE, et al: The prognostic significance of tumor location and bowel obstruction in Dukes B and C colorectal cancer. Findings from the NSABP clinical trials. Ann Surg 198:743-752, 1983[Medline] 39. Steinberg SM, Barkin JS, Kaplan RS, et al: Prognostic indicators of colon tumors. The Gastrointestinal Tumor Study Group experience. Cancer 57:1866-1870, 1986[CrossRef][Medline] 40. Witzig TE, Loprinzi CL, Gancheroff NJ, et al: DNA ploidy and cell kinetic measurements as predictors of recurrence and survival in stages B2 and C colorectal adenocarcinoma. Cancer 68:879-888, 1991[CrossRef][Medline]
41. Le Voyer TE, Sigurdson ER, Hanlon AL, et al: Colon cancer survival is associated with increasing number of lymph nodes removed—A secondary analysis of intergroup trial INT-0089. J Clin Oncol 21:2912-2919, 2003
42. Allegra CJ, Paik S, Colangelo LH, et al: Prognostic value of thymidylate synthase, Ki-67, and p53 in patients with Dukes' B and C colon cancer: A National Cancer Institute-National Surgical Adjuvant Breast and Bowel Project collaborative study. J Clin Oncol 21:241-250, 2003
43. Watanabe T, Wu TT, Catalano PJ, et al: Molecular predictors of survival after adjuvant chemotherapy for colon cancer. N Engl J Med 344:1196-1206, 2001
44. Gryfe R, Kim H, Hsieh ET, et al: Tumor microsatellite instability and clinical outcome in young patients with colorectal cancer. N Engl J Med 342:69-77, 2000
45. Ribic CM, Sargent DJ, Moore MJ, et al: Tumor Microsatellite-Instability Status as a Predictor of Benefit from Fluorouracil-Based Adjuvant Chemotherapy for Colon Cancer. N Engl J Med 349:247-257, 2003 46. Kim H, Jen J, Vogelstein B, et al: Clinical and pathological characteristics of sporadic colorectal carcinomas with DNA replication errors in microsatellite sequences. Am J Pathol 145:148-156, 1994[Abstract]
47. Boland CR, Thibodeau SN, Hamilton SR, et al: A National Cancer Institute workshop on microsatellite instability for cancer detection and familial predisposition: Development of international criteria for the determination of microsatellite instability in colorectal cancer. Cancer Res 58:5248-5257, 1998
48. Halling KC, French AJ, McDonnell SK, et al: Microsatellite instability and 8p allelic imbalance in stage B2 and C colorectal cancers. J Natl Cancer Inst 91:1295-1303, 1999
49. Efficacy of adjuvant fluorouracil and folinic acid in B2 colon cancer. International Multicentre Pooled Analysis of B2 Colon Cancer Trials (IMPACT B2) Investigators. J Clin Oncol 17:1356-1363, 1999
50. Mamounas E, Wieand S, Wolmark H, et al: Comparative efficacy of adjuvant chemotherapy in patients with Dukes' B versus Dukes' C colon cancer: Results from four National Surgical Adjuvant Breast and Bowel Project adjuvant studies (C-01, C-02, C-03, and C-04). J Clin Oncol 17:1349-1355, 1999 51. Buyse M, Piedbois P: Should Dukes' B patients receive adjuvant therapy? A statistical perspective. Semin Oncol 28:20-24, 2001 (1 suppl 1)
52. Schrag D, Rifas-Shiman S, Saltz L, et al: Adjuvant chemotherapy use for Medicare beneficiaries with stage II colon cancer. J Clin Oncol 20:3999-4005, 2002 Submitted September 12, 2003; accepted January 6, 2004. Related Correspondence
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
INTRODUCTION
