Validation of a new classification system for curatively resected colorectal adenocarcinoma

782

Validation of a New Classification System for Curatively Resected Colorectal Adenocarcinoma Ahud Sternberg, M.D.1 Ohn Sibirsky, M.D.1 David Cohen, M.D.2 Leslie E. Blumenson, Ph.D.3 Nicholas J. Petrelli, M.D.3

BACKGROUND. The parameters within which colorectal adenocarcinoma is currently staged are often insufficient for decisions regarding therapy after potentially curative surgery. Consequently, oncologists make frequent use of additional prognostic indicators when assessing individual prognosis and selecting patients for adjuvant systemic treatment. Follow-up programs are generally uniform for all patients, regardless of disease stage and prognosis. As a result, patients with a

1

Department of Surgery, Sackler School of Medicine, Tel Aviv University, Tel Aviv, Israel.

favorable prognosis are needlessly subjected to stressful, costly follow-up too early

2

Department of Pathology, Sackler School of Medicine, Tel Aviv University, Tel Aviv, Israel.

system that is a superior predictor of individual prognosis following curative

3

management and follow-up.

Division of Surgical Oncology, Roswell Park Cancer Institute, Buffalo, New York.

and too frequently. This study was conducted to validate a new classification surgery and may serve as a guide for personalized, cost-effective postoperative

METHODS. A total of 231 American colorectal carcinoma patients who underwent curative resection were retrospectively staged according to a new classification (containing 4 stage-groups) for curatively resected colorectal adenocarcinoma. This classification is based on statistical analysis of the impact on prognosis of numerous characteristics of 363 consecutive Israeli colorectal carcinoma patients who underwent curative resection. All the patients in both cohorts had had surgery at least 5 years previously. The new classification is based on three histologic variables (venous invasion, depth of primary tumor penetration, and regional lymph node status) and a scoring system that correlates higher numeric score with worse prognosis. In both cohorts, the new classification was compared with the Dukes, Astler–Coller, and TNM staging systems for patient distribution and survival (both disease free and cancer-related survival).

RESULTS. In both cohorts, the 4 stage-groups of the new classification differed significantly in both the rate of and the time to first recurrence and cancer-related death, with progression from Group 4 to Group 1. Groups of high risk lymph node negative patients were defined, and lymph node positive patients were subdivided according to prognosis. It is suggested that, by using this new classification as a guide, selection for adjuvant systemic treatment may be refined, and postoperative Presented at the First World Congress of Surgical Oncology and the 51st Annual Cancer Symposium of the Society of Surgical Oncology, San Diego, California, March 1998; and the First Anglo-Israeli Oncology Conference, Eilat, Israel, December 1998. The authors thank Ms. A. Drezna for her statistical analysis of the Israeli cohort.

follow-up may be personalized and therefore more cost-effective.

CONCLUSIONS. The new classification for curatively resected colorectal adenocarcinoma, based on an analysis of the Israeli cohort and validated in the American cohort, is superior to the Dukes, Astler–Coller, and TNM staging systems as a predictor of individual prognosis, most probably because it incorporates the microscopic forerunner of distant, hematogenous spread (i.e., venous invasion) with the locoregional parameters of extent of disease (i.e., T and N values). It is suggested that the new classification may serve as a guide for more refined

Address for reprints: Ahud Sternberg, M.D., Department of Surgery, Hillel Jaffe Medical Center, Hadera 38100, Israel.

selection of patients for adjuvant systemic treatment and for individualized and

Received December 7, 1998; revision received March 17, 1999; accepted March 17, 1999.

any health care system worldwide. Cancer 1999;86:782–92.

© 1999 American Cancer Society

more cost-effective postoperative follow-up. The new classification is simple and easy to use, requires no sophisticated equipment or tests, and can be applied in © 1999 American Cancer Society.

A New Staging System for Colorectal Carcinoma/Sternberg et al.

783

KEYWORDS: adjuvant systemic treatment, classification, colorectal adenocarcinoma, curative resection, personalized postoperative follow-up, prognosis, staging, venous invasion.

T

he principal value of a staging system for malignancy is its ability to predict prognosis accurately and to guide appropriate clinical decisions regarding postoperative management and follow-up. For over 60 years, the name of C. E. Dukes and his classification for rectal carcinoma have been intimately associated with the staging of colorectal carcinoma. The Dukes staging system is simple and easy to use, but its three stages group together heterogeneous subsets of patients, therefore decreasing its power as an indicator of prognosis for the individual patient. Over the years, numerous modifications have been proposed in an attempt to improve the prognostic predictive ability of the original Dukes classification.1–9 This multitude of Dukes modifications led to considerable confusion and difficulty in interpreting and comparing data from different sources. The need for a universal terminology for solid tumors led to the TNM staging system, which is in fact yet another modification of the Dukes classification (because TNM Stages I, II, and III correlate with Dukes stages A, B, and C) and therefore not superior to it as a predictor of prognosis for the individual patient.10 Many investigators have studied the impact of numerous clinical, biologic, and histologic variables on the prognosis associated with colorectal carcinoma.11– 41 Based on their findings, several of these authors have proposed new and supposedly improved staging systems for colorectal carcinoma, but these have failed to achieve widespread acceptance, mainly because some are complicated, others require sophisticated equipment and additional work, and still others are based on criteria that are largely subjective. The prognosis, the various treatment options, and the prospects that the chosen modalities may improve the prognosis of patients with metastatic colorectal carcinoma are sufficiently well defined, according to the sites and extent of the metastatic disease. Although the decision to treat or not is pursuant to medical contraindications and the personal desire of the patient, the need for treatment is absolute, and further refinement of Stage D or IV is not required. This is not the case for patients who have undergone potentially curative resection of colorectal carcinoma, for they are a heterogeneous population with regard to their personal prognosis; some are cured, whereas others will have a recurrence. These latter patients differ significantly with regard to both personal risk and timing of tumor recurrence. It is for these patients that refinement of staging and prediction of personal

prognosis is needed, to improve their selection for adjuvant treatment. In addition, protocols for postoperative follow-up after curative resection of colorectal carcinoma are usually uniform for all patients, regardless of personal stage. As a result, patients with a favorable prognosis are needlessly subjected to stressful, costly follow-up too early and too frequently in their postoperative course. The study described herein was undertaken in an attempt to create, validate, and generalize a new classification system that would be a superior predictor of prognosis for the individual following potentially curative resection of colorectal adenocarcinoma, and could serve as a guide for personalized, cost-effective postoperative management and follow-up.

PATIENTS AND METHODS Israeli Cohort, Scoring System, and the New Classification Personal, familial, clinical, and histologic variables for 363 consecutive patients, who had undergone potentially curative resection of Dukes A, B, or C colorectal adenocarcinoma at least 5 years prior to the time of data analysis, were retrospectively analyzed with regard to their impact on cancer recurrence and cancerrelated death. In this analysis, the variables for the 105 patients (28.9%) whose tumors recurred were compared with the variables for the 258 patients (71.1%) who remained disease free throughout the study period. All pathology slides were reviewed, and revised when appropriate, by one of the authors (D.C.). Microscopic diagnosis of venous invasion on hematoxylin and eosin (H&E)–stained tissue was based on the identification of tumor cells within an endotheliumlined space, surrounded by a rim of smooth muscle, or containing red blood cells. When present, an adjacent artery of similar size verified that the involved vessel was indeed a vein. When these criteria were not met, the involved vessel was regarded as a lymphatic channel. Elastic tissue stains were not used for identification of blood vessels. Patients with inflammatory bowel disease, intestinal polyposis syndromes, perforated tumors, or metachronous colorectal carcinomas of a higher stage than the reference tumor were excluded from the study. Local excision and polypectomy were considered curative for tumors confined to the mucosa/submucosa only; treatment by fulguration only was excluded. Tumor locations were the rectum in 119 cases (32.7%) and the sigmoid colon in

784

CANCER September 1, 1999 / Volume 86 / Number 5

TABLE 1 Variables That Had a Statistically Significant Effect on Tumor Recurrence and/or Cancer-Related Death in Univariate Analysis (These Variables Were Entered into the Multivariate Analysis) Variable Tumor configuration (polypoid vs. infiltrating) Adenocarcinoma vs. other histology Dukes stage Astler–Coller stage Depth of tumor invasion No. of involved lymph nodes Ratio: involved to examined lymph nodes Venous invasion Lymphatic invasion Perineural invasion Tumor grade

Effect on tumor recurrence

Effect on cancerrelated death

TABLE 2 Scoring System for the New Classification Variable

Score

a. Venous invasion

Yes—1 No—0 Subserosa/serosa/pericolic fat—1

b. Penetration of tumor in bowel wall P 5 0.0029 P 5 0.018 P , 0.0001 P , 0.0001 P 5 0.0006 P , 0.0001 P , 0.05 P , 0.0001 P , 0.0001 P , 0.0001 N.S.

P 5 0.0002 N.S. P , 0.0001 P , 0.0001 P , 0.0001 P , 0.0001 P , 0.05 P , 0.0001 P , 0.0001 P , 0.0001 P , 0.05

N.S.: not significant.

112 (30.9%), and the remaining 132 (36.4%) were all other colonic sites. Patients were followed every 3 months for the first 3 years and every 6 months thereafter. The mean follow-up was 12.6 years. Four patients (1.1%) were lost to follow-up during the first 5 years (3 patients during the fourth and 1 patient during the fifth year). At 5 years, overall survival and survival adjusted for cancer deaths only were 61.7% and 76.3%, respectively. Venous invasion increased the incidence of recurrence from 18% to 50% and was associated with decreased survival (P , 0.0001). Increasing depth of penetration of the primary tumor in the bowel wall, higher Dukes and Astler–Coller stage (P , 0.001), and higher tumor grade (P 5 0.013) were all associated with increased incidence of venous invasion.

Statistical methods Cancer recurrence and cancer-related death were considered end points. Patients whose tumors recurred and patients who remained disease free were initially compared by univariate analysis, using the following tests as appropriate: the Student t test, the Mann–Whitney U test, the Wilcoxon rank sum test, and the chi-square test. Variables that showed a statistically significant effect on cancer recurrence or on cancer-related death at the P , 0.05 level (Table 1) were then used in a multivariate stepwise regression analysis (Cox model),42 to identify the variables with an independent effect on prognosis at the P , 0.01 level, arranged according to the magnitude of their prognostic impact. Survival curves were calculated using adjusted actuarial life tables and compared by the

c. Regional lymph node status

Mucosa/muscularis propria—0 Involved—1 Not involved—0

Total score

Prognostic group

Prognosis

0 1 2 3

1 2 3 4

Excellent Intermediate-good Intermediate-poor Poor

log rank test. BMDP, SAS, and SPSS statistical software was used (BMDP Statistical Software, Inc., Los Angeles, CA; SAS Institute Inc., Cary, NC; SPSS Inc., Chicago, IL). The three pathologic variables identified in multivariate analysis as bearing the strongest independent effect on both disease free and cancer-related survival were (in order of decreasing prognostic impact) venous invasion, depth of tumor penetration in the bowel wall, and regional lymph node status. These three variables were selected as the basis for the new classification. A scoring system was designed to correlate higher score with worse prognosis by giving the unfavorable feature of each of the 3 variables a score of 1 point, whereas the favorable features scored nil. Patients were thus classified to Group 1, 2, 3, or 4, respective to their total numeric score (0, 1, 2, or 3 points) (Table 2). These 4 stage-groups expressed the severity of disease and differed significantly with regard to rate of recurrence, time to recurrence, cancer-related death, time to cancer-related death, and 5-year survival. The mean times to first recurrence were as follows: Group 4, 1.3 years; Group 3, 1.8 years; Group 2, 2.0 years; Group 1, 3.1 years. Groups 1 and 2 and Groups 3 and 4 differed significantly with regard to the time to cancer-related death (3.6 years and 2.4 years, respectively). Based on these data, a separate schema for postoperative management and follow-up was proposed for each of the 4 stage-groups. It is noteworthy that the site of the primary tumor in the colon or rectum did not emerge as a statistically significant independent indicator of prognosis. Consequently, the resulting new classification may be applied to all curatively resected colonic and rectal adenocarcinomas.

A New Staging System for Colorectal Carcinoma/Sternberg et al.

The new classification system was compared with the Dukes, Astler–Coller, and TNM systems.

American Cohort: Validation and Generalization of the New Classification The medical records of 231 consecutive patients who had undergone potentially curative resection of Dukes A, B, or C colorectal adenocarcinoma at Roswell Park Cancer Institute in Buffalo, New York, at least 5 years prior to the time of data analysis were retrospectively reviewed, and the patients were then staged by the new classification as well as by the Dukes, Astler– Coller, and TNM systems. The 4 systems were compared for distribution of patients and for survival (disease free and adjusted for cancer-related deaths only). The times by which 5% and 10% of recurrences had occurred were noted for each stage-group of the 4 systems. Possible association between preoperative carcinoembryonic antigen (CEA) levels and stagegroup of the new classification system was examined. Venous invasion was diagnosed on H&E–stained tissue according to the same criteria that were used for the Israeli cohort, and, as in the Israeli series, elastic tissue stains were not used for identification of blood vessels.

Statistical methods All survival curves (disease free, overall, and adjusted for cancer-related deaths only) were calculated using the Kaplan–Meier method,43 and the statistical significance of differences among survival curves was analyzed using the log rank test.44 Analyses of the crosstabulations were performed using the chi-square test. The logarithmic transformation was used to equalize variances of CEA values within each stage of a staging system. Comparisons of the logarithms of CEA values among different stages were made using both one-way analysis of variance and the Kruskal–Wallis test.45 All analyses were performed using the appropriate (1990) programs from BMDP.

RESULTS Overall survival and survival adjusted for cancer-related deaths only for the American patients were 68.0% and 80.2% at 5 years and 51.0% and 73.7% at 10 years, respectively; 5- and 10-year disease free survival were 74.0% and 72.3%, respectively. Staging the American patients by the new classification yielded distributions and survival curves (disease free and adjusted for cancer-related deaths only) that were very similar to those of the Israeli patients. As in their Israeli counterparts, the times to first recurrence and cancer-related death clearly differed among the 4 stage-groups, progressing from Group 4 to Group 1 (and

785

TABLE 3 American Patients: Correlation between the Dukes/TNM and the New Classifications (A: Distribution of Dukes/TNM Stages among the 4 Stage-Groups of the New Classification; B: Distribution of Dukes/TNM Stages within Each of the 4 Stage-Groups) (All Numbers Are Percentages) Groups Dukes/TNM

1

2

3

4

Total

A/I B/II C/III

93.1 0.0 0.0

6.9 92.3 15.8

0.0 7.7 56.8

0.0 0.0 27.4

100.0 100.0 100.0

Total

23.4

39.4

26.0

11.3

100.0

A/I B/II C/III

100.0 0.0 0.0

4.4 79.1 16.5

0.0 10.0 90.0

0.0 0.0 100.0

25.1 33.8 41.1

Total

100.0

100.0

100.0

100.0

100.0

A

B

could therefore serve as the basis for a graduated, group specific follow-up plan). Distribution of the patients between the 4 groups was as follows: Group 1, 54 (23.4%); Group 2, 91 (39.4%); Group 3, 60 (26.0%); Group 4, 26 (11.3%). Tables 3 and 4 present the distribution of the Dukes/TNM and the Astler–Coller stages, respectively, of the patients, as they are categorized according to the 4 stage-groups of the new classification. Figures 1 a, b, and c show the disease free survival curves for the new classification, Dukes/TNM, and Astler–Coller staging systems, respectively. Figures 2 a, b, and c present the survival curves (adjusted for cancer-related deaths only) for these staging systems. Table 5 gives the 5- and 10year survival rates (cancer-related deaths only) for the 4 stage-groups of the new classification. Table 6 presents the 5 year disease free survival rates, as well as the times in months by which 5% and 10% of recurrences had occurred, for each of the 4 stage-groups of the new classification. Preoperative CEA levels showed correlation with the stage-groups of the new classification (increasing levels with higher group), but with wide variance and considerable overlapping among the 4 groups, especially in the low levels of CEA. By using the logarithms of the CEA values, we found a statistically significant correlation with the 4 groups (which was expected, because higher CEA levels correlate with worse prognosis). Nevertheless, due to the variance and overlapping, we do not recommend that preoperative CEA levels be used as indicators for tentative staging of an individual patient to a certain stage-group.

786

CANCER September 1, 1999 / Volume 86 / Number 5

TABLE 4 American Patients: Correlation between the Astler–Coller and the New Classifications (A: Distribution of Astler–Coller Stages among the 4 Stage-Groups of the New Classification; B: Distribution of Astler– Coller Stages within Each of the 4 Stage-Groups) (All Numbers Are Percentages) Groups Astler– Coller

1

2

3

4

Total

A B1 B2 B3 C1 C2 C3

100.0 90.5 0.0 0.0 0.0 0.0 0.0

0.0 9.5 91.9 100.0 88.2 0.0 0.0

0.0 0.0 8.1 0.0 11.8 67.1 60.0

0.0 0.0 0.0 0.0 0.0 32.9 40.0

100.0 100.0 100.0 100.0 100.0 100.0 100.0

Total

23.4

39.4

26.0

11.3

100.0

A B1 B2 B3 C1 C2 C3

29.6 70.4 0.0 0.0 0.0 0.0 0.0

0.0 4.4 74.7 4.4 16.5 0.0 0.0

0.0 0.0 10.0 0.0 3.3 81.7 5.0

0.0 0.0 0.0 0.0 0.0 92.3 7.7

6.9 18.2 32.0 1.7 7.4 31.6 2.2

Total

100.0

100.0

100.0

100.0

100.0

A

B

DISCUSSION Numerous investigators have reported on the adverse effect of venous invasion on the prognosis of patients with colorectal carcinoma (i.e., increased incidence of recurrence, especially visceral metastases, and decreased survival).4,31,46 – 63 Several of these authors have recommended considering venous invasion as an indicator for the administration of systemic adjuvant treatment.58,59,62,63 Our findings correlated well with these findings and led us to similar conclusions. The reported incidence of diagnosis of venous invasion varies considerably in different studies. Obviously, this variance may reflect differences in stage at diagnosis, and it may also indicate currently unidentified inherent differences between patient populations; however, it is probably due mainly to interobserver differences and variance in specimen processing. Although some investigators advocate the use of elastic tissue stains to aid in the diagnosis of venous invasion,56,57,60 – 62 most authors have not found these stains to be helpful, and they identify venous invasion on H&E–stained paraffin sections as tumor cells within an endothelium-lined space, surrounded by a rim of smooth muscle, or containing red blood cells.53,55,58,59 We used these criteria to evaluate both the Israeli and the American cohorts. Talbot et

al.53 and Dirschmid et al.62 emphasized the importance of cutting multiple (up to 5) tangential blocks from the periphery of the tumor to facilitate the diagnosis of extramural venous invasion. The number of blocks is directly related to the size of the tumor; therefore, minimum standards for declaring a specimen negative for venous invasion have not been clearly defined in the literature. In both cohorts, differentiation between intramural and extramural venous invasion was not routinely made. Nevertheless, the statistically significant results of our study indicate that differentiating between tumors with and without venous invasion, without differentiating according to the site of the involved veins, suffices as an indicator of prognosis. Moreover, in a recent study of venous invasion in 81 Stage IV colorectal carcinoma specimens, sites of involved veins were as follows: only intramural in 31.6%, only extramural in 45.6%, and both intramural and extramural in 22.8%. These findings indicate that, although extramural veins are involved more frequently, involvement of intramural veins is also an indicator of poor prognosis that may lead to the establishment of clinically significant distant metastases (Sternberg A, Amar M, Alfici R, Groisman G, unpublished data). Until the day when we are able to prevent colorectal carcinoma, our main goal will be to cure as many afflicted patients as possible. To do so in a cost-effective manner today, a classification system for patients following curative resection is needed— one that will predict an individual patient’s prognosis more accurately than existing systems and thus serve as a guide for a personalized plan for postoperative management and follow-up. The new system, based on prognostic indicators identified in Israeli patients and validated in American patients treated at a major comprehensive cancer center, appears to fulfill this need. This classification incorporates the microscopic forerunner of distant, hematogenous metastasis (namely, venous invasion) with the parameters of locoregional extent of disease (i.e., T and N values), thereby giving expression to all three modes of tumor spread (local extension, lymphatic spread, and hematogenous dissemination). Conversely, according to the Dukes, Astler–Coller and TNM staging systems, curatively resected patients are staged by T and N values alone, for these classifications are based on macroscopic evidence of distant metastases only, which by definition is M 5 0 for all curatively resected patients. This difference is most probably the main contributor to the superiority of the new classification over existing staging systems as a predictor of prognosis for the individual patient. Adopting this new classification would definitely improve staging; understaging can-

A New Staging System for Colorectal Carcinoma/Sternberg et al.

787

FIGURE 1. Disease free survival (determined by the Kaplan–Meier method) is shown for the American patients, staged by (a) the new classification, (b) Dukes/TNM, and (c) Astler–Coller. Compared with the Dukes/TNM and Astler– Coller staging systems, the new classification defines a group of curatively resected colorectal carcinoma patients whose prognosis is extremely unfavorable (stage-group 4). This is an illustration of the new classification’s superiority in predicting an individual’s prognosis. The disorderly nature and crossover of survival curves of the stage-groups of the Astler–Coller classification are contrasted with those of the new classification. As a result, the new classification may serve as a guide to postoperative management and follow-up (see text).

not occur, for even in the event of a false-negative diagnosis of venous invasion, the T and N values would still stage the patient appropriately according to the principles of existing staging systems. Positive diagnosis of venous invasion would upstage the patient, and for some patients it would indicate the need for systemic adjuvant treatment, which would not have been indicated by existing staging systems and currently accepted criteria for the administration of adjuvant treatment. Tables 3 and 4 demonstrate how the new classification takes patients of the same Dukes, Astler–Coller, or TNM stage and emphasizes the severity of disease by upstaging some patients, whereas others are downstaged, indicating that their disease is of a less aggressive nature. The new classification defines groups of high risk lymph node negative colorectal carcinoma patients and subdivides lymph node positive patients according to their relative prognosis.

Using the New Classification as a Guide for Systemic Adjuvant Treatment The National Institutes of Health Consensus Development Statement regarding adjuvant therapy for colo-

rectal carcinoma patients recommends continuation of the search for prognostically relevant risk factors, to define better the patient population that is most likely to benefit from additional therapy.64 Regional lymph node status is no doubt a strong indicator of prognosis, but do lymph node positive colorectal carcinoma patients constitute a prognostically homogeneous group, and do they all require the same adjuvant treatment? Based on clinical experience and the molecular biology of colorectal carcinoma (e.g., DCC gene data), our answer to these questions must be negative; we know that some curatively resected, lymph node positive colorectal carcinoma patients do extremely well even with no adjuvant treatment, whereas others die of their disease despite having received aggressive locoregional and systemic treatment. In other words, by adhering to currently accepted criteria for the administration of adjuvant systemic treatment, we may be overtreating some lymph node positive colorectal carcinoma patients and undertreating other lymph node positive as well as some lymph node negative patients. Tables 3 and 4 show how the new staging method divides lymph node positive patients between 3 stage-

788

CANCER September 1, 1999 / Volume 86 / Number 5 TABLE 5 American Patients Staged by the New Classification: 5- and 10-Year Survival Rates (Adjusted for Cancer-Related Deaths Only) Groups

n

5-yr rate (%)

SE (%)

10-yr rate (%)

SE (%)

1 2 3 4

54 91 60 26

97.6 86.6 68.6 48.4

2.4 3.8 6.6 10.1

94.0 80.0 60.1 43.6

4.2 4.7 7.4 10.2

Total: 231 SE: standard error.

FIGURE 2. Survival (determined by the Kaplan–Meier method, adjusted for cancer deaths only) is shown for the American patients staged by (a) the new classification, (b) Dukes/TNM, and (c) Astler–Coller. groups according to the severity of their disease. This prognostic diversity among lymph node positive colorectal carcinoma patients may be comprehended by studying the disease free and cancer-related survival

curves of the lymph node positive American patients, stratified for venous invasion, which show wide and significant separation (Figs. 3a and b). Venous invasion also explains the surprisingly favorable survival curves of Astler–Coller C1 patients (Figs. 1c, 2c): 92.9% of the lymph node positive patients who also had venous invasion were in Astler–Coller Stages C2 and C3, and only 7.1% were in Astler–Coller Stage C1. Conversely, of all Astler–Coller C1 patients in this cohort, only 11.8% had venous invasion, hence their favorable prognosis. By using the new classification as a guide for the administration or withholding of systemic adjuvant treatment, we may definitely improve the selection of patients for such therapy. Considering the 5- and 10-year survival rates (adjusted for cancer-related deaths only) and the 5-year disease free survival rates for each of the 4 stagegroups of the new classification (Tables 5 and 6), we believe that adjuvant systemic treatment is categorically indicated for all patients in Groups 3 and 4 (comprised of tumors with high T value and venous invasion, lymph node positive tumors with venous invasion, or lymph node positive tumors with high T value), and definitely not indicated for all patients in Group 1 (comprised of lymph node negative tumors with low T value and no venous invasion). As for Group 2, with a 5-year disease free survival rate of 81.3% and 5- and 10-year cancer-related survival rates of 86.6% and 80.2%, respectively, we feel that there is no clear-cut and categoric indication for systemic adjuvant treatment. The lymph node positive cases would most probably be selected for adjuvant systemic therapy, according to currently accepted criteria. Lymph node negative tumors with low T value but positive venous invasion should probably all be treated as well. As for the remaining patients in Group 2 (high T value with neither venous invasion nor lymph node involvement), pros and cons should be considered for each case, the option should be pre-

A New Staging System for Colorectal Carcinoma/Sternberg et al.

789

TABLE 6 American Patients Staged by the New Classification: 5-Year Disease Free Survival, and Times by Which Approximately 5% and 10% of Recurrences Had Occurred Time when approximately 5% recurred

Time when approximately 10% recurred

5-yr DFS Groups

n

Rate (%)

SE (%)

Mos

Actual rate (%)

Mos

Actual rate (%)

1 2 3 4

54 91 60 26

95.7 81.3 56.2 35.1

3.0 4.0 7.0 10.0

19.0 9.7 5.4 1.6

4.3 4.7 5.3 4.0

NA 19.1 12.2 4.9

NA 10.0 10.7 11.5

DFS: disease free survival; SE: standard error; NA: not applicable.

FIGURE 3. Lymph node positive American patients, stratified for venous invasion (BVI) (positive vs. negative): (a) disease free survival (P 5 0.016), (b) survival (adjusted for cancer deaths only) (P 5 0.007). (Kaplan–Meier method.)

sented to the patient and discussed, and the final decision should be reached on an individual basis. It is noteworthy that patients with rectal tumors that qualify for adjuvant radiation treatment are treated according to currently accepted criteria. These patients would obviously be staged to Groups 2, 3, or 4 of the new classification.

Using the New Classification as a Guide for Individualized, Cost-Effective Postoperative Follow-Up Although data from randomized trials comparing minimalistic and intensive postoperative follow-up of colorectal carcinoma patients have not yet been published, it is generally accepted that postoperative follow-up after curative resection of colorectal carcinoma is worthwhile, for it has been shown that timely, aggressive treatment of locoregional and/or distant

metastases may achieve a significant percentage of cures, prolonged survival, or at least significant palliation and improved quality of life.65–72 To be cost-effective, follow-up should be carried out only during periods when the chances of positive diagnosis justify it. Ideally, follow-up should be personalized according to the individual patient’s prognosis, so that patients who are at a low risk of recurrence are not subjected to unnecessary follow-up procedures. We believe that the survival statistics and survival curves of the 4 stage-groups of the new classification can guide the decision regarding if and when patients in each stage-group should be followed for the identification of tumor recurrence. When should postoperative follow-up commence? We believe that it is not unethical to have the few patients with early-recurring tumors refer them-

790

CANCER September 1, 1999 / Volume 86 / Number 5

selves for symptoms, rather than start follow-up unnecessarily early for the vast majority of patients who are in the same stage-group. It is appropriate to reiterate here that data proving that early detection of metastases clearly improves survival have yet to be published. Accordingly, we suggest that active follow-up for each stage-group be started only after the first 5% of recurrences have occurred. Table 6 shows the times by which approximately 5% and 10% of recurrences were diagnosed for each of the 4 stagegroups of the new classification. Immediately after surgery, all patients should receive explanations regarding the follow-up plan and the signs and symptoms of recurrence. It is obvious that for patients in Group 4 (with 4.0% of recurrences already occurring by 1.6 months) and Group 3 (with 5.3% of recurrences occurring by 5.4 months), follow-up should start immediately following surgery. This is also in accordance with our recommendation for mandatory adjuvant systemic treatment for all patients in these two groups. Considering the rates of recurrence and cancer-related death, which are apparent from the survival curves (Figs. 1a, 2a), intensive follow-up seems appropriate for the first 2 years. From the third year and thereafter, the rates of recurrence and cancer-related death decrease, and the frequency of follow-up procedures may be lowered accordingly. Follow-up for patients in Group 2 who receive no adjuvant systemic treatment (comprising 31.2% of the entire series) may be deferred until 10 –12 months postoperatively, which is the approximate time by which 5% of the recurrences had occurred (Table 6). Follow-up should be intensive during the first year of follow-up, considering the steep slope of the disease free survival curve for this period; from the second year of follow-up (the third postoperative year), the intensity of follow-up may be lowered. Group 2 patients who receive adjuvant treatment are obviously followed closely from the immediate postoperative period. The first recurrence in Group 1 (comprising 23.4% of the patients) was diagnosed at 13 months and the last at 19 months, bringing the total recurrence rate to 4.3%. With a 5-year disease free survival rate of 95.7% and no indication for adjuvant therapies, this stagegroup is the ideal example of a subset of patients for whom (from the standpoint of cost-effectiveness) a very solid case can be presented against any follow-up at all. At most, minimal follow-up (e.g., starting at 18 months and consisting of semiannual monitoring of CEA level and annual follow-up visits) may be performed for patients who insist on active follow-up. It is noteworthy that earlier detection (for exam-

ple, as a result of effective screening, public education, etc.) should increase the percentage of patients for whom active follow-up is not indicated (Group 1) or delayed (Group 2), thereby increasing the cost-effectiveness of this follow-up policy. According to the survival curves (Figs. 1a and 2a), all recurrences and cancer-related deaths occurred within 8 years of potentially curative surgery. Accordingly, active follow-up for the detection of metastasis need not be continued beyond 8 years for any patient. Notwithstanding this personalized follow-up program for the detection of metastases, colonic screening for metachronous colorectal tumors should be carried out for all patients, as is customary. In conclusion, we recommend the new classification for improved staging of patients after potentially curative resection of colorectal adenocarcinoma and as a guide to personalized, cost-effective postoperative follow-up and adjuvant systemic treatment, as outlined in this article. This new classification is simple and easy to use, and its application requires no sophisticated equipment or tests. It may be applied in any health care system worldwide, provided that all surgical specimens are evaluated by competent pathologists who are aware of the crucial importance of diagnosing and reporting venous invasion.

REFERENCES 1.

Gordon-Watson C, Dukes CE. The treatment of carcinoma of the rectum with radium. Br J Surg 1929 –30;17:365–9. 2. Dukes CE. The classification of cancer of the rectum. J Pathol Bacteriol 1932;35:323–32. 3. Gabriel WB, Dukes CE, Bussey HJR. Lymphatic spread of cancer in the rectum. Br J Surg 1935;23:395– 413. 4. Dukes CE, Bussey HJR. The spread of rectal cancer and its effect on prognosis. Br J Cancer 1958;12:309 –20. 5. Kirklin JW, Dockerty MB, Waugh JM. The role of the peritoneal reflection in the prognosis of carcinoma of the rectum and sigmoid colon. Surg Gynecol Obstet 1949;88:326 –31. 6. Astler VB, Coller FA. The prognostic significance of direct extension of carcinoma of the colon and rectum. Ann Surg 1954;139:846 –51. 7. Gunderson LL, Sosin H. Areas of failure found at reoperation (second or symptomatic look) following “curative surgery” for adenocarcinoma of the rectum: clinicopathologic correlation and implications for adjuvant therapy. Cancer 1974; 34:1278 –92. 8. Turnbull RB, Kyle K, Watson FR, Spratt J. Cancer of the colon: the influence of the no-touch isolation technic on survival rates. Ann Surg 1967;166:420 –7. 9. Gastrointestinal Tumor Study Group. Prolongation of the disease-free interval in surgically treated rectal carcinoma. N Engl J Med 1985;312:1465–72. 10. American Joint Committee on Cancer. Cancer staging manual. 5th edition. Philadelphia: Lippincott-Raven, 1998. 11. Ponz de Leon M, Sant M, Micheli A, Sacchetti C, DiGregorio C, Fante R, et al. Clinical and pathologic prognostic indicators in colorectal cancer: a population-based study. Cancer 1992;69:626 –35.

A New Staging System for Colorectal Carcinoma/Sternberg et al. 12. Michelassi F, Ayala JJ, Balestracci T, Goldberg R, Chappell R, Block GE. Verification of a new clinico-pathologic staging system for colorectal adenocarcinoma. Ann Surg 1991;214: 11– 8. 13. Crucitti F, Sofo L, Doglietto GB, Bellantone R, Ratto C, Bossola M, et al. Prognostic factors in colorectal cancer: current status and new trends. J Surg Oncol 1991;12(Suppl): 76 – 82. 14. Jass JR, Mukawa K, Goh HS, Love SB, Capellaro D. Clinical importance of DNA content in rectal cancer measured by flow cytometry. J Clin Pathol 1989;42:254 –9. 15. Kokal WA, Gardine RL, Sheibani K, Morris PL, Prager E, Zak IW, et al. Tumor DNA content in resectable, primary colorectal carcinoma. Ann Surg 1989;209:188 –93. 16. Shepherd NA, Saraga EP, Love SB, Jass JR. Prognostic factors in colonic cancer. Histopathology 1989;14:613–20. 17. Hermanek P, Guggenmoos-Holzmann I, Gall FP. Prognostic factors in rectal carcinoma: a contribution to the further development of tumor classification. Dis Colon Rectum 1989;32:593–9. 18. Halvorsen TB, Seim E. Degree of differentiation in colorectal adenocarcinomas: a multivariate analysis of the influence on survival. J Clin Pathol 1988;41:532–7. 19. Halvorsen TB, Seim E. Influence of mucinous components on survival in colorectal adenocarcinomas: a multivariate analysis. J Clin Pathol 1988;41:1068 –72. 20. Wiggers T, Arends JW, Volovics A. Regression analysis of prognostic factors in colorectal cancer after curative resections. Dis Colon Rectum 1988;31:33– 41. 21. Wiggers T, Arends JW, Schutte B, Volovics A, Bosman FT. A multivariate analysis of pathologic prognostic indicators in large bowel cancer. Cancer 1988;61:386 –95. 22. Hannisdal E, Thorsen G. Regression analyses of prognostic factors in colorectal cancer. J Surg Oncol 1988;37:109 –12. 23. Michelassi F, Block GE, Vannucci L, Montag A, Chappell R. A 5- to 21-year follow-up and analysis of 250 patients with rectal adenocarcinoma. Ann Surg 1988;208:379 – 89. 24. Jass JR, Love SB, Northover JMA. A new prognostic classification of rectal cancer. Lancet 1987;1:1303– 6. 25. Jass JR, Morson BC. Reporting colorectal cancer. J Clin Pathol 1987;40:1016 –23. 26. Griffin MR, Bergstrahl EJ, Coffey RJ, Beart RW Jr., Melton LJ. Predictors of survival after curative resection of carcinoma of the colon and rectum. Cancer 1987;60:2318 –24. 27. Schmitz-Moormann P, Himmelmann GW, Baum U, Nilles M. Morphological predictors of survival in colorectal carcinoma: univariate and multivariate analysis. J Cancer Res Clin Oncol 1987;113:586 –92. 28. Umpleby HC, Bristol JB, Rainey JB, Williamson RC. Survival of 727 patients with single carcinomas of the large bowel. Dis Colon Rectum 1987;27:803–10. 29. Jass JR, Atkin WS, Cuzick J, Bussey HJR, Morson BC, Northover JM, et al. The grading of rectal cancer: historical perspectives and a multivariate analysis of 447 cases. Histopathology 1986;10:437–59. 30. Fielding LP, Phillips RKS, Fry JS, Hittinger R. Prediction of outcome after curative resection for large bowel cancer. Lancet 1986;2:904 – 6. 31. Steinberg SM, Barkin JS, Kaplan RS, Stablein DM. Prognostic indicators of colon tumors: the gastrointestinal tumor study group experience. Cancer 1986;57:1866 –70. 32. Khubchandani M. Relationship of symptom duration and survival in patients with carcinoma of the colon and rectum. Dis Colon Rectum 1985;28:585–7.

791

33. Chapuis PH, Dent OF, Fisher R, Newland RC, Pheils MT, Smyth E, et al. A multivariate analysis of clinical and pathological variables in prognosis after resection of large bowel cancer. Br J Surg 1985;72:698 –702. 34. McDermott FT, Hughes ES, Pihl E, Johnson WR, Price AB. Local recurrence after potentially curative resection for rectal cancer in a series of 1008 patients. Br J Surg 1985;72:34 –7. 35. Wied U, Nilsson T, Knudsen JB, Sprechler M, Johansen A. Postoperative survival of patients with potentially curable cancer of the colon. Dis Colon Rectum 1985;28:333–5. 36. Freedman LS, Macaskill P, Smith AN. Multivariate analysis of prognostic factors for operable rectal cancer. Lancet 1984; 2:733– 6. 37. deMello J, Struthers L, Turner R, Cooper EH, Giles GR. Multivariate analyses as aids to diagnosis and assessment of prognosis in gastrointestinal cancer. Br J Surg 1983;48:341– 8. 38. Knudsen JB, Nilsson T, Sprechler M, Johansen A, Christensen N. Venous and nerve invasion as prognostic factors in postoperative survival of patients with resectable cancer of the rectum. Dis Colon Rectum 1983;26:613–7. 39. Kemeny N, Braun DW Jr. Prognostic factors in advanced colorectal carcinoma: importance of lactic dehydrogenase level, performance status, and white blood cell count. Am J Med 1983;74:786 –96. 40. Falterman KW, Hill CB, Markey JC, Fox JW, Cohn I Jr. Cancer of the colon, rectum, and anus: a review of 2313 cases. Cancer 1974;34:951–9. 41. Maetani S, Onodera H, Nishikawa T, Tobe T. Systematic computer-aided search of optimal staging system for colorectal cancer. J Clin Epidemiol 1991;44:285–91. 42. Cox DR. Regression models and life tables. J R Stat Soc B 1972;34:187–220. 43. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457– 81. 44. Mantel N. Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 1966;50:163–70. 45. Kruskal WH, Wallis WA. Use of ranks in one-criterion variance analysis. J Am Stat Assoc 1952;47:583– 621. 46. Dukes CE, Bussey HJR. Venous spread in rectal cancer. Proc R Soc Med 1941;34:571– 81. 47. Brown CE, Warren S. Visceral metastasis from rectal carcinoma. Surg Gynecol Obstet 1938;66:611–21. 48. Sunderland DA. The significance of vein invasion by cancer of the rectum and sigmoid: a microscopic study of 210 cases. Cancer 1949;2:429 –37. 49. Grinnell RS. Lymphatic metastases of carcinoma of the colon and rectum. Ann Surg 1950;131:494 –506. 50. Swinton NW. Cancer of the colon and rectum: a statistical study of 608 patients. Surg Clin N Am 1959;39:745–53. 51. Spratt JS Jr., Spjut HJ. Prevalence and prognosis of individual, clinic and pathologic variables associated with colorectal carcinoma. Cancer 1967;20:1976 – 85. 52. Copeland EM, Miller LD, Jones RS. Prognostic factors in carcinoma of the colon and rectum. Am J Surg 1968;116: 875– 81. 53. Talbot IC, Ritchie S, Leighton MH, Hughes AO, Bussey HJR, Morson BC. The clinical significance of invasion of veins by rectal cancer. Br J Surg 1980;67:439 – 42. 54. Talbot IC, Ritchie S, Leighton MH, Hughes AO, Bussey HJR, Morson BC. Spread of rectal cancer within veins: histologic features and clinical significance. Am J Surg 1981;141:15–7.

792

CANCER September 1, 1999 / Volume 86 / Number 5

55. Talbot IC, Ritchie S, Leighton MH, Hughes AO, Bussey HJR, Morson BC. Invasion of veins by carcinoma of rectum: method of detection, histological features and significance. Histopathology 1981;5:141– 63. 56. Minsky BD, Mies C, Recht A, Rich TA, Chaffey JT. Resectable adenocarcinoma of the rectosigmoid and rectum: the influence of blood vessel invasion. Cancer 1988;61:1417–24. 57. Minsky BD, Mies C, Rich TA, Recht A, Chaffey JT. Potentially curative surgery of colon cancer: 2. The influence of blood vessel invasion. J Clin Oncol 1988;6:119 –27. 58. Horn A, Dahl O, Morild I. The role of venous and neural invasion on survival in rectal adenocarcinoma. Dis Colon Rectum 1990;33:598 – 601. 59. Horn A, Dahl O, Morild I. Venous and neural invasion as predictors of recurrence in rectal adenocarcinoma. Dis Colon Rectum 1991;34:798 – 804. 60. Inoue T, Mori M, Shimono R, Kuwano H, Sugimachi K. Vascular invasion of colorectal carcinoma readily visible with certain stains. Dis Colon Rectum 1992;35:34 –9. 61. Ouchi K, Sugawara T, Ono H, Fujiya T, Kamiyama Y, Kakugawa Y, et al. Histologic features and clinical significance of venous invasion in colorectal carcinoma with hepatic metastases. Cancer 1996;78:2313–7. 62. Dirschmid K, Lang A, Mathis G, Haid A, Hansen M. Incidence of extramural venous invasion in colorectal carcinoma: findings with a new technique. Hum Pathol 1996;27: 1227–30. 63. Blumberg D, Paty PB, Picon AI, Guillem JG, Klimstra DS,

64.

65. 66.

67.

68.

69. 70.

71.

72.

Minsky BD, et al. Stage I rectal cancer: identification of high-risk patients. J Am Coll Surg 1998;186:574 – 80. National Institutes of Health Consensus Development Statement. Adjuvant therapy for patients with colon and rectal cancer. JAMA 1990;264:1444 –50. Sardi A. Multiple operations for recurrent colorectal cancer. Semin Surg Oncol 1991;7:146 –56. Ovaska J, Jarvinen H, Kujari H, Perttila I, Mecklin JP. Follow-up of patients operated on for colorectal carcinoma. Am J Surg 1990;159:593– 6. Makela J, Huikipuro K, Laitenen S, Kairaluoma MI. Surgical treatment of recurrent colorectal cancer: five-year followup. Arch Surg 1989;124:1029 –32. Ovaska JT, Jarvinen HJ, Mecklin JP. The value of a follow-up programme after radical surgery for colorectal carcinoma. Scand J Gastroenterol 1989;24:416 –22. Wanebo HJ, Stevens W. Surgical treatment of locally recurrent colorectal cancer. Probl Gen Surg 1987;4:115–29. Schiessel R, Wunderlich M, Herbst F. Local recurrence of colorectal cancer: effect of early detection and aggressive surgery. Br J Surg 1986;73:342– 4. Minton JP, Hoen JL, Gerber DM, Horsely JS, Connolly DP, Salwan F, et al. Results of a 400-patient carcinoembryonic antigen second-look colorectal cancer study. Cancer 1985; 55:1284 –90. Martin EW, Minton JP, Carey LC. CEA-directed second look surgery in the asymptomatic patient after primary resection of colorectal carcinoma. Ann Surg 1985;202:310 –7.