Available raw data from the five randomized trials were reviewed by the primary author (L.L.G.) and Mayo Clinic Cancer Center statisticians (D.J.S., C.A.) to determine what data items were consistently available. Surgery and pathology parameters included degree of rectal wall invasion, invasion of or adherance to adjacent organs or structures, and number of involved nodes; number of total nodes was available except in NCCTG 79-47-51. Chemotherapy data included agent, dates, and objective status. Radiation therapy data included site, dose, number of fractions, and elapsed days. Follow-up information included relapse status (date, site) and survival status (alive, dead, date). Five-year follow-up information was available in 94% of surviving patients and 8-year follow-up information was available in 62%. All patients were censored at 8 years post–random assignment.

Patient demographics included age, sex, and performance status. Patient age ranged from 20 to 99 years, with a median of 61 years. There were 2,360 male patients and 1,416 female patients. Performance status (PS) was PS 0 in 2,616 patients; PS 1 in 1,049 patients; PS 2 in 50 patients; PS 3 in 1 patient; unknown in 5 patients; and missing in 70 patients.

The data from all arms of the five studies were pooled for the primary analyses. Kaplan-Meier curves were used to estimate the distribution of OS and DFS.²⁵ Time to local and distant relapse was estimated using the method of cumulative incidence.²⁶ Log-rank tests were used to compare the survival distributions between patient subgroups.²⁷ All P values reported are two-sided, with P < .05 used to denote statistical significance. For analyses adjusting for treatment effect, Cox proportional hazards models were used.²⁸ The six treatment groups that were created for the adjustment analyses are listed in Table 1.

Analyses of treatment effects within patient risk groups are presented as descriptive only. This is due to the large time span these trials included (1979-1992) and to possible changes in patient populations due to differences in patient selection and treatment. Formal testing indicated between-trial heterogeneity for the same treatment regimen in different studies.

Before publication of the prior adjuvant rectal cancer pooled analysis,¹⁷ there was only minimal data in the colorectal cancer literature examining the impact of NT stage (T substage within N-stage) on either survival or disease relapse. In view of the large patient numbers available in the current pooled analysis, results were analyzed in relation to both N and NT stage (Table 3 and Fig 2).

Survival and relapse rates by TN stage of disease are presented in Table 4. Patients in the current analysis are separated into three risk groups (intermediate-risk, moderately high-risk, high-risk) based on data from the prior¹⁷ and current pooled analysis, and OS, DFS, and relapse rates are from the current analysis. For patients with intermediate-risk lesions (T1-2/N1, T3/N0), 5-year OS was 79% and 75%, respectively; 5-year DFS was 73% and 65%, respectively; 5-year local relapse (LR) rates were 7% and 9%, respectively; and 5-year distant metastasis (DM) rates were 15% and 20%, respectively. Patients with moderately high-risk lesions (T1-2/N2, T3/N1, T4/N0) had a 5-year OS range of 60% to 67%, a DFS of 48% to 58%, an LR range of 8% to 13%, and a DM range of 28% to 37%. Those with high-risk lesions (T3/N2, T4/N1-2) had a 5-year OS range of 37% to 44%, a DFS of 30% to 36%, an LR range of 14% to 23%, and a DM range of 39% to 53%.

Five-year OS by treatment method is shown in Figure 3 for each of the six treatment methods evaluated. Surgery alone and adjuvant radiation alone had inferior survival when compared with the other four treatment methods (P < .001). These differences are also demonstrated in Tables 5, 6, 7, 8, 9, 10, 11, and 12, for most TN stages of disease.

For each of the six treatment methods previously described, survival and relapse rates at 5-years were analyzed in descriptive fashion as a function of both TN and NT stage. Tables 5 through 10 show 5-year OS by TN (y-axis) and NT stage (x-axis) for each treatment method. For many of the treatment methods, there is a steady decrease in OS, with a progressive increase in both TN and NT stage. Patients were separated into three risk groups (intermediate, moderately high, high-risk) for analysis of survival (Table 11) and relapse rates (Table 12) by TN stage and treatment method.

Patterns of relapse after surgery alone for primary rectal cancer have been previously defined in clinical, reoperative and autopsy series.^2,5–8 The rate of local (pelvic) relapse after “curative resection” is related both to the degree of tumor extension beyond the rectal wall and to nodal spread. For patients with a single high-risk factor of either direct tumor extension beyond the wall, nodes negative (T3/N0), or positive nodes but primary tumor confined to the wall (T1-2/N1-2), local relapse rates published in older surgical series vary from 20% to 40%.^2,5–8 For patients with both positive nodes and extension beyond the wall (T3-4/N1-2), the risk of pelvic relapse was nearly additive (40% to 65% in clinical series and 70% in a reoperative series).^2,5–8 When lesions extend beyond the muscularis propria, the amount of uninvolved perirectal tissue on microscopic review (circumferential or radial margin) may be equally or more important than the degree of extrarectal extension in determining the risk of local relapse.²⁹ If lesions with surgical adherence to unresected structures have the sites of adherence marked by the surgeon, the pathologist can then determine whether the margins of resection are either narrow or microscopically involved.

The rate of systemic metastases seems to be significantly higher for patients with both high-risk pathologic factors (extension beyond rectal wall and positive nodes; T3-4/N1-2), as opposed to those with only a single risk factor (T3-4/N0, T1-2/N1). In published data from adjuvant rectal cancer patients irradiated at either Massachusetts General Hospital (MGH)¹³ or the Mayo Clinic,¹⁶ the incidence of subsequent systemic relapse was approximately 20% with T3-4/N0 and T1-2/N+ lesions versus 40% to 60% with T3-4/N+ lesions.

In the current pooled analysis, a progressive increase in T stage (T1-2 v T3 v T4) negatively impacted both survival and disease relapse. Within each T stage, substaging based on the amount of nodal involvement (N stages of N0, N1, N2) resulted in a decrease in both OS and DFS and an increase in both local and distant relapse rates. The differences in local relapse by TN stage did not reach statistical significance, except for the T3/N0-2 group of patients, at 9% v 12% v 14% (P < .002).

When results in the current pooled analysis were evaluated by N stage and NT stage (substaging within N stage based on the amount of primary disease extension or T stage), there was, again, variable survival and disease relapse by both N stage and by NT stage. Single-institution analyses from the Mayo Clinic¹⁶ and MGH¹³ had previously suggested that N1-2/T1-2 patients treated with postoperative irradiation alone or combined with chemotherapy had outcomes similar to N0/T3 and N0/T4 patients; however, the patient numbers in each stage subset were small. In the current analysis, 581 patients had N1-2/T1-2 lesions, 1,060 patients had N0/T3 lesions, and 111 had N0/T4 lesions. The 79% 5-year OS rates observed for the N1/T1-2 patients (n = 355) in the current analysis was similar to that for N0/T3 patients (75% 5-year OS). It was also similar to results found in a “patterns of care” analysis.³⁰ The 67% 5-year survival seen with N2/T1-2 patients (n = 226) in the current analysis was higher than expected confirming findings in the earlier rectal pooled analysis.¹⁷ For the N2/T1-2 patients, 5-year OS was similar to that for patients with N0/T4 or N1/T3 lesions at 67% (N2/T1-2) v 65% (N0/T4) and 60% (N1/T3). Results by N2 stage were rarely available before the earlier pooled analysis,¹⁷ and results by T sub-stage within the N2 group (ie, T1-2/N2, T3/N2, T4/N2) were nonexistent.

Survival and disease relapse after surgery alone^1,2,4–12 or combined with adjuvant treatment^{1,5,13–22,30–49} for rectal cancer patients are a function of both degree of bowel wall penetration of the primary lesion and nodal status, as suggested for nearly 50 years.^50,51 Nodal involvement alone does not determine survival and relapse rates.

In the current pooled analysis of survival and relapse rates by TN stage (Table 4), patients were separated into four risk groups (low-, intermediate-, moderately high-, and high-risk). The percentages listed for low-risk patients were based on single-institution^{2,5–16,31–33} or “patterns of care” analyses,³⁰ since the low-risk patients (T1-2N0) were not eligible for the trials included in this analysis because of low relapse risks and excellent survival (approximate 90% 5-year OS). The 5-year OS, DFS, and relapse rates for the intermediate-, moderately high-, and high-risk patients are data from the current pooled analysis. For patients with intermediate-risk lesions (T1-2/N1, T3/N0), 5-year OS was 79% and 75%; those with moderately high-risk lesions (T1-2/N2, T3/N1, T4/N0) had a 5-year OS range of 60% to 67%, and patients with high-risk lesions (T3/N2, T4/N1-2) had 5-year OS of 37% to 44%. The 5-year DFS by risk group had similar trends: intermediate-risk, 73% and 65%; moderately high-risk, 48% to 58%; high-risk, 30% to 36%.

The advantage of TN substaging over MAC staging in predicting both survival and disease relapse rates for node-positive patients is seen for MAC C1 and C2 stages (T1-3/N+) in the pooled analysis data (Table 4) but not with MAC C3 (T4/N+). For patients with T1-2 and T3 lesions, those with N1 disease (1 to 3 nodes involved) have statistically improved survival (Table 2) and lower rates of disease relapse (Table 2) when compared with their N2 counterparts (4 or more nodes involved). For patients with T4/N+ lesions, the number of involved nodes (N1 versus N2) did not affect either survival or the rate of local relapse, but did have an impact on the rate of distant metastasis (Table 2).

Data from the current and prior rectal cancer pooled analysis¹⁷ strongly supports sub-staging, of tumor-node-metastasis system stages II and III, as accomplished in the sixth edition (2002) of tumor-node-metastasis system staging.³ As presented in Table 4, the continued use of Dukes' stages A, B, C or TNM stage I, II, III without the benefit of sub-staging could be considered insufficient, since patient prognosis is not accurately reflected. Within Dukes' C or TNM stage III, three separate prognostic subgroups of lesions exist (intermediate-risk, T1-2/N1; moderately high-risk, T1-2/N2 and T3/N1; high-risk, T3/N2 and T4/N1-2). To combine or merge all of these patients into Dukes' C or TNM stage III does not provide full information for patients, regarding prognosis relative to proposed treatment.

Patients with T1-2/N1-2 disease should receive a more favorable discussion regarding prognosis. Accurate data from large numbers of patients are now available to discuss the prognosis of T1-2/N1 and T1-2/N2 patients relative to other substages of disease.

For patients with N2 disease, information from the current and prior pooled rectal cancer analyses shows that N2 disease does not by itself confer poor prognosis. Sub-staging by T stage influenced survival in the current analysis (5-year OS: N2/T1-2, 67%; N2/T3, 44%; and N2/T4, 37%; P < .001; 5-year DFS: 58% v 36% and 30%, P < .001). Therefore, placement of all N2 patients within AJCC IIIC substage does not reflect the markedly different prognosis of N2 patients found in the current and prior pooled rectal cancer analyses.¹⁷

In the prior rectal cancer pooled analysis¹⁷ no attempt was made to evaluate treatment outcome relative to treatment method whereas for the present analysis, descriptive data is presented regarding the impact of treatment method on survival and relapse by TN stage. The decision to present the data as descriptive rather than comparative is based on several factors. The studies spanned a 13-year accrual period during which surgical technique, treatment method, and sequencing evolved. No single treatment arm was used in more than two trials, thus limiting the ability to compare the same regimen in different trials. When this was possible, significant between-trial heterogeneity was observed (ie, the results of external beam radiation therapy plus FU–based arms in NCCTG 86-47-51, INT 0114, and NSABP R02 differed significantly, as did the results of semustine-based chemotherapy in NSABP R01 and R02).

At the present time, most clinicians concur that patients with low-risk lesions (T1-2/N0) are best managed with surgical resection alone, except for those with select lesions that may be amenable to less extensive treatment. Such patients are usually not referred for adjuvant treatment unless surgical resection margins are compromised.

More controversy exists for patients with intermediate-risk lesions (T1-2/N1, T3/N0), who have been candidates for all of the trials included in the current pooled analysis and were included in the mandate for combined chemoradiotherapy in the 1990 National Institutes of Health consensus statement.¹ It has been suggested for several decades that properly selected patients with T1-2/N1-2 or T3/N0 lesions may have relatively low risks of local and distant relapse with surgery alone.^52,53 This may include those with upper rectal T1-2/N1 lesions who have only a few positive nodes adjacent to the primary lesion or those with negative nodes and minimal extension beyond the muscularis propria (T3/N0) who have adequate radial or circumferential resection margins.^54,55 One of the challenges is adequate pathological evaluation of specimens to provide the necessary information concerning patient prognosis. The other challenge is to have enough data to determine what constitutes an adequate radial or circumferential resection margin. Collection of radial or circumferential margin data was encouraged in NCCTG 86-47-51 and INT 0114, but the percentage of patients with such information was so low that it could not be evaluated.^19,20–22

In the current pooled analysis, patients with T1-2/N1 and T3/N0 lesions had a 5-year OS of 85% and 84% when treated with surgery followed by chemotherapy (S plus CT) as a single adjuvant (Tables 7 to 11). These results were similar to those achieved with trimodality combinations of surgery and chemo-radiation which resulted in a 5-year OS range of 78% to 83% for T1-2/N1 lesions and 74% to 80% for T3/N0 lesions (Table 11). An evaluation of 5-year OS data suggests that use of trimodality postoperative chemoradiation for all patients with T1-2/N1 and T3/N0 lesions may be excessive treatment. However, since this is a descriptive analysis of results across different trials and eras, definitive conclusions cannot be made regarding the necessity of adjuvant radiation. When 5-year DFS and relapse rates were evaluated (Tables 11 and 12), S plus CT had good results for patients with T1-2/N1 lesions with DFS of 78%, local relapse (LR) of 5%, and distant relapse of 16% (trimodality treatment: DFS range, 75% to 78%; LR range, 5% to 7%). For patients with T3/N0 lesions, however, 5-year DFS fell to 69% with S+CT and LR increased to 11% (distant relapse similar at 18%). For trimodality treatment, the 5-year DFS range with T3/N0 lesions was 63% to 75%, and LR ranged from 5% to 10%. The DFS and relapse data suggests that further improvements in outcome should be feasible, even for patients with “intermediate risk” rectal cancers. Such improvements in outcome might be achieved with the use of preoperative chemo-radiation in properly selected patients (T3 or N+ on the basis of endoscopic ultrasound), by the addition of new agents to current “standard regimens” or with the evaluation of targeted therapy based on molecular markers.

For patients with moderately high-risk lesions (T1-2/N2, T3/N1, T4/N0), the current analysis of results by treatment method showed definite room for improvement in outcome. Five-year OS ranged from 43% to 70% with S+CT and 44% to 80% with S plus RT plus CT. It was uncommon to have 5-year DFS ≥ 60%, LR was 9% to 18% even with trimodality treatment arms, and distant relapse ranged from 20% to 60% with CT arms.

In the high-risk patients (T3/N2, T4/N1-2 lesions), 5-year OS and DFS were less than 50% for most combinations of TN stage and treatment method (5-year OS range of 25% to 45% with S plus CT and 29% to 57% with S plus RT plus CT). Local relapse ranged from 0% to 33% with trimodality treatment, and distant relapse ranged from 22% to 78% even in CT arms, demonstrating much room for improvement in both therapy and outcomes.

Different treatment strategies may be indicated for intermediate-risk (T3/N0, T1-2/N1) compared with moderately high or high-risk patients in view of variable survival and rates of relapse. For future trial design, it would be appropriate to perform separate studies for the various risk subsets of patients. Alternatively, stratification by risk factor could allow a planned statistical analysis of treatment effect within each prognostic group.

For carefully selected patients with intermediate risk lesions, it may be appropriate to evaluate surgery alone or plus postoperative chemotherapy if careful pathology evaluation suggests a low risk of LR after a confirmed total mesorectal excision by an experienced surgeon.^53,56–58 However, initiation of new trials that evaluate radiation as a component of treatment for select intermediate-risk patients should be done with caution, using stringent surgical and pathological guidelines,^54–57 since preoperative radiation reduced LR rates when combined with total mesorectal excision (8% v 2%, P < .001) in a Dutch phase III trial.⁵⁶

Patients with T1-2/N1 and T3/N0 lesion are still appropriate candidates for adjuvant chemoradiotherapy. Trials evaluating adjuvant preoperative chemo-radiation would be reasonable for patients with distal lesions, in whom down-staging would improve the chance for rectal preservation and reanastamosis.^59–65 Postoperative chemo-radiation trials should include patients without indications for preoperative treatment who have an inadequate evaluation of nodes (< 10 to 15 nodes examined by pathology)²¹ and those at higher risk for local relapse based on amount of primary tumor extension beyond the muscularis propria (≥ 2 mm) or a narrow radial margin (≤ 4 mm).⁵³

For patients with moderately high- or high-risk lesions, the high rates of both local and distant relapse support continuing use of both concurrent chemo-radiation and adjuvant chemotherapy or other systemic therapy. Further testing of the sequencing of chemo-radiation and surgery^66,67 will probably not be indicated, since the completed German preoperative versus postoperative phase III trial resulted in an improvement in both sphincter preservation and local control for patients randomized to receive the preoperative chemoradiotherapy.^68,69 Evaluation of preoperative chemoradiotherapy combined with more aggressive chemotherapy (ie, irinotecan or oxaliplatin) or targeted therapy (vascular endothelial growth factor or epidermal growth factor receptor inhibitors)^70–74 is indicated in these patients in view of the modest efficacy of postoperative adjuvant therapy.