Multivariate Analysis of Recurrence After Anterior Resection for Colorectal Carcinoma Stephan Rinnert-Gongora MD, Paul Ian Tartter MD, FACS, NOW York, NOW York

We performed a multivariate analysis of survival data from 278 patients who underwent potentially curative anterior resection with hand-sewn anasto- mosis for nonobstructing colorectal carcinoma to evaluate the interaction of the resection margin with distance from the anal verge and their contri- butions to local and distant recurrence. Cumulative 5-year disease-free survival was 66 percent for the 258 patients with complete follow-up. Forty-nine patients ( 19 percent) had local recurrence and 42 ( 16 percent) developed initial distant metastases. Local recurrence rates increased with increasing age and with more advanced Dukes’ stage. It developed in twice as many patients with colostomies as witb- out colostomies. Distant metastases developed sig- nificantly more often in patients with nodal involve- ment and in patients with resection margins exceeding 3.5 cm. Forty-four percent of patients with lesions within 14 cm of the anal verge resected with margins of at least 3.5 cm developed distant recurrence. This study suggests that aggressive pel- vic dissection to achieve resection margins greater than 3.5 cm may contribute to tumor dissemination and subsequent distant metastases.

R esection margin and distance from the anal verge are important variables affecting local recurrence after

anterior resection for colorectal carcinoma. The minimal adequate margin of resection continues to be a controver- sial issue. Numerous investigators believe that more is better, 5 cm being the absolute minimum acceptable re- sected length of tumor-free rectum [l-4]. However, sev- eral studies have not found high local recurrence rates with short margins [5-71; indeed, one large study found increasing risk of local recurrence rates with longer mar- gins of resection [a].

Controversy also surrounds the relationship between local recurrence and distance of the lesion from the anal verge. Several investigators have observed higher local recurrence rates with lesions close to the anal verge,

From the Department of Surgery, The Mount Sinai Medical Center, New York, New York.

Requests for reprints should be addressed to Paul Ian Tartter, MD, Department of Surgery, Annenberg 25-60, Mount Sinai Medical Cen- ter, 1 Gustave L. Levy Place, New York, New York 10029.

whereas others have denied the prognostic significance of proximity to the anus [2,5-7,9-111. Since the length of the distal resection margin is limited by the proximity of the lesion to the anal verge, the prognostic importance of each of these variables must be evaluated in the context of the other.

Published studies examining the prognostic signifi- cance of distance and margin rarely consider both, and we are not aware of any study that has analyzed local recurrence or survival in relation to both resection margin and distance from the anal verge simultaneously. The interaction of resection margin with distance from the anal verge may explain the divergent findings of these studies with respect to their prognostic importance.

MATERIAL AND METHODS The study group consisted of 278 patients who under-

went anterior resection for Dukes’ stages A, B,, B2, Cl, and Cz adenocarcinoma of the rectum at the Mount Sinai Medical Center between 1973 and 1982. Tumors were staged by a modification of the original Dukes’ staging system: stage A indicated disease limited to mucosa and submucosa; stage Bt, invasion of muscularis with unin- volved nodes; stage Bz, serosa involved with uninvolved nodes; stage Cl, less than four involved nodes; and stage Cz, four or more involved nodes. The original charts of these patients were obtained from the hospital medical records department in September 1986. Information was recorded on age, sex, distance of the lesion from the anal verge, stage, differentiation, size ([length/21 X [width/ 21 X ?r), margin of resection, use of proximal diverting colostomy, and postoperative therapy. Follow-up infor- mation was gathered by letters and telephone contact with patients, family members, private physicians, and review of tumor registry records. Complete follow-up in- formation was available for 257 of the patients (93 per- cent), and these patients became the subjects of this study. Since the majority of patients were followed by private physicians, a consistent protocol for detecting re- currences was not followed.

The goal of the analysis was to quantify the relation- ship between disease-free survival and potential prognos- tic factors. This was accomplished using the maximum partial likelihood ratio method based on the Cox propor- tional hazards regression model [ 121, available through the City University of New York Computer Center [ 131. This program investigates the relation between survival as a dependent variable and a set of independent variables which may be categorical (such as diverting colostomy, sex, and postoperative therapy) or interval-scaled vari- ables (such as age, margin of resection, and tumor size). The p values for Table I are based on chi-squares calcu- lated from the relationship of each independent variable

THE AMERICAN JOURNAL OF SURGERY VOLUME 157 JUNE 1989 573

TABLE I Multlvarlate Analysis of Recurrence (p values)*

Variable Local Distant

Age 0.0269 0.7734 Sex 0.2007 0.9561 Stage 0.0002 0.0119 Margin 0.3122 0.0040 Distance 0.9677 0.0350 Differentiation 0.5974 0.5723 Sk8 0.3513 0.1161 colostomy 0.0060 0.9602

* p values from chi-squares calculated from the relation of each Independent variable to time to disease recurrence as described in

“Material and Methods.”

to disease recurrence [13]. The life table method of Cut- ler and Ederer was utilized to calculate the survival rates portrayed in Figure 1. Significance of the observed differ- ences was evaluated using the chi-square analysis for categorical variables and the Student’s t test for continu- ous variables.

RESULTS The overall 5-year disease-free survival was 66 per-

cent. Forty-nine patients (19 percent) subsequently pre- sented with local recurrences and 42 (16 percent) pre- sented with distant metastases with 19 hepatic, 12 pulmonary, 7 simultaneous hepatic and other distant site, and 4 simultaneous hepatic and local. The latter were counted as distant recurrences for statistical purposes because the outcome of analysis was no different when these patients were counted as local recurrences. The mean disease-free survival was 65 months, compared with 26 months in the local recurrence group and 24 months in the distant metastasis group (Figure 1).

The most significant prognostic factor was Dukes’ stage (p <O.OOOl, Table I). More advanced disease was

associated with both higher local and distant recurrence rates (Table II). However, distant recurrence rates were low (11 percent or less) when nodes were not involved (stages A, Bt, and Bz), whereas local recurrence rates were low only for tumors limited to the mucosa (stage A). Local and distant recurrence rates were also related to the administration of postoperative adjuvant therapy (p = 0.005, Table II). The highest recurrence rates were seen in the 33 patients (13 percent) who received adjuvant therapy because patients with more advanced stage were significantly more likely to receive such therapy (chi- square = 27.39, p = 0.007). Since the association of recurrence with adjuvant therapy was present in a small number of patients and due to the association of therapy with disease stage, we will not consider it further as a prognostic factor.

The other significant variable independently related to distant recurrence was the margin of resection: 26 percent of the 78 patients with margins exceeding 3.5 cm developed distant recurrences (20 patients) compared with 12 percent of the 178 patients with shorter margins (21 patients (p) = 0.008). Distance from the anal verge was significantly related to distant recurrence when the margin of resection was taken into consideration.

Multivariate analysis identified a subset of patients at high risk for distant recurrence: 44 percent of patients with lesions within 14 cm of the anal verge resected with margins of at least 3.5 cm developed distant recurrences compared with 14 percent of all other patients combined (chi-square = 10.55, p = 0.001; Table III). The risk of local recurrence decreased from 21 percent for margins less than 5 mm to 15 percent for margins exceeding 5 cm, although this relationship was not statistically significant.

Local recurrence rates increased progressively with age, from 6 percent for patients less than 50 years of age to 30 percent for patients over 80, and distant recurrence rates decreased with age as local recurrence rates in- creased, but these relationships were not statistically sig-

I I I I I I

y 100 cl E 3 90 x

IO 20 30 40 50 60 MONTHS

Figure 1. Local, distant, and overall disease-free survival after anterior resectlcn In all 278 patlents.

574 THE AMERICAN JOURNAL OF SURGERY VOLUME 157 JUNE 1989

nificant. Distance from the anal verge was not signifi- cantly related to local recurrence because lesions close to the anus were, in fact, less likely to recur locally than lesions further away. This may have been due to the conservative policy followed by this department during the study period: patients with lesions of borderline suit- ability for anterior resection received abdominoperineal resections.

Proximal diverting colostomies were significantly re- lated to local recurrence because patients with colosto- mies were twice as likely to develop local recurrence as patients without colostomies (p = 0.006; Table II). Colos- tomies were used significantly less frequently in very young and very old patients: 63 percent of the 211 pa- tients between 50 and 80 years of age (134 patients) had proximal diverting colostomies compared with 32 percent of the 37 patients less than 50 or over 80 (12 patients) (p = 0.0015). Patients with lesions close to the anal verge (distance less than 10 cm) were significantly more likely to receive proximal diverting colostomies than patients with higher lesions (chi-square = 24.18, p = 0.0001). Multivariate analysis revealed that age, disease stage, and use of proximal diverting colostomy were significant- ly and independently related to local recurrence.

In summary, advanced disease stage and resection margins exceeding 3.5 cm were significantly and inde- pendently related to distant recurrence whereas advanced disease stage, advanced patient age, and colostomy were significantly and independently associated with local re- currence. Low lesions resected with margins exceeding 3.5 cm had an unusually high distant recurrence rate.

COMMENTS The association of Dukes’ stage with both local and

distant recurrences after anterior resection for colorectal cancer observed in this study has also been noted in nu- merous previous studies. Recent Scandinavian studies, where advanced stage was not associated with increased local recurrence rates, attributed their observations to less extensive hiitopathologic examination of the resected specimen and to the use of the EEA stapler [7,14]. The latter study had small numbers of patients and unusually high local recurrence rates for Dukes’ stage A lesions. We found that local recurrence rates increased significantly with invasion of the muscularis (stage Bt, or more ad- vanced) whereas distant recurrence increased significant- ly with nodal involvement (stage Ci or more advanced).

We observed progressively increasing local recurrence rates with increasing patient age. A similar observation was made by Stower and Hardcastle [15], who observed that crude survival of colorectal cancer patients declines with age, and McDermott et al [I6], who noted signifi- cantly prolonged cancer-free survival in patients 30 years of age or less. However, Phillips et al [17] found the opposite: local recurrence declined with patient age in their study of 370 anterior resections. We did not find a relationship between patient age and overall disease-free survival, only between age and local recurrence. This may have reflected the surgeons’ reluctance to perform ab- dominoperineal resection for low rectal lesions in high-

TABLE II Relation of Potential Prognosttc Varlabtea lo

Locailm of Recurrw*

RECURRENCEAFTRRANTRRlORRRSECTION

P Variable Disease-Free LoCal Distant Value

Patients 165 (64) 49 (19) 42 (16)

Age (mearh yr) 67 70 66 0.6564 Sex

Male 66 (64) 26 (21) 21 (15) 0.756 Female 60 (65) 20 (16) 22 (19)

Distance (cm) 14 14 13 0.2522 Margin (cm) 3.06 2.89 4.09 0.2267 colostomy

Yes 96 (62) 37 (24) 21 (14) 0.0357 No 71 (69) 12 (12) 20 (19)

Dukes’ stage A 20 (90) 0 (0) 2 (10) Sl 33 (70) 9 (19) 5 (11) S2 70 (71) 19 (19) 10 (IO) 0.0017

Cl 34 (53) 12 (19) 16 (28)

c2 IO (40) 9 (36) 8 (24) Tumor dlfferentlatlon

Well 56 16 18 Moderate 102 27 22 0.885 Poor 6 2 2

Tumor size (cm? 14 13 12 Postoperative therapy

Radiotherapy 2 (29) 5 (71) 0 (0) MmotheWJY 12 (55) 4 (18) 6 (27) 0.005 Seth 1 (25) 2 (50) t (25) None 151 (67) 36 (17) 35 (16)

l Values in Parentheses are percentages.

TABLE III Interactlon of Resection Margin and Distance From the

Anal Verge In Relation to Dlstant Reoumnce*

I Distant P

n Recurrence Value

close to anus (< 14 cm)

Long resection margin 27 12 (44) (> 3.5 cm) 0.000

Short resection margin 103 10 (10) (< 3.5 cm)

Far from anus (> 14 cm)

Long resection margin 33 5 (15) (> 3.5 cm) 0.542

Short resection margin 54 11 (25) (< 3.5 cm)

l Values in Parentheses are Percentages.

risk elderly patients. These elderly patients may present with early local recurrences, but few live long enough to develop late distant recurrences. The decrease in distant recurrences observed with advancing age supports this hypothesis. These data indicate that advanced age should not be a deterrent to performing adequate resection for rectal carcinoma.

To our knowledge, the value of proximal transverse colostomy in protecting difficult anastomoses has been limited to analysis of postoperative morbidity and mortal-

THE AMERICAN JOURNAL OF SURGERY VOLUME 157 JUNE 1989 575

ity due to anastomotic leaks. In the present study, techni- cal difficulties with low pelvic dissections and anatomoses that caused surgeons to perform protective colostomies may have contributed to local tumor dissemination at the time of surgery and resulted in high local recurrence rates. Local recurrence after anterior resection may de- crease in incidence if the present practice of avoiding proximal diverting colostomy continues. Several studies have cited the increased morbidity, mortality, and costs of proximal diverting colostomy and the lack of anastomotic protection afforded [ 281.

The interaction of resection margin with distance from the anal verge observed in this study may explain the divergent findings of previous studies with respect to the prognostic importance of these variables. Several studies have cited resection margin and distance from the anal verge as important factors affecting local recur- rence, but few have related either to the development of distant metastases [I-6,9,20]. Although local recurrence rates increased progressively with shorter resection mar- gins in our study, the results were not statistically signifi- cant. On the other hand, distant recurrences were signifi- cantly increased after resection with at least 3.5 cm margins. Patients with resected lesions less than 14 cm in diameter with margins of at least 3.5 cm had very high distant recurrence rates in this study (44 percent), which suggests that aggressive pelvic dissection to achieve mar- gins of resection exceeding 3.5 cm may contribute to tumor dissemination and subsequent distant recurrence. There may be an optimal resection margin below 3.5 cm that is associated with minimal pelvic dissection and low local and distant recurrence rates.

REFERENCES 1. Cole WH. Recurrence in carcinoma of the colon and proximal rectum following resection for carcinoma. Arch Surg 1952; 65: 264-10. 2. Vandertoll DJ, Baehrs OH. Carcinoma of rectum and low sig- moid. Arch Surg 1965; 90: 793-8.

3. Copeland EM, Miller LD, Jones RS. Prognostic factors in carci- noma of the colon and rectum. Am J Surg 1968; 116: 875-81. 4. Enker WE, Laffer UT, Block GE. Enhanced survival of patients with colon and rectal cancer upon wide anastomotic resection. Ann Surg 1979; 190: 350-60. 5. Wilson SM, Beahrs OH. The curative treatment of carcinoma of the sigmoid, rectosigmoid, and rectum. Ann Surg 1975; 183: 556- 65. 6. McDermott FT. Hughes ESR, Pihl E, Johnson WR, Price AB. Local recurrence after potentially curative resection for recta1 can- cer in a series of 1008 patients. Br J Surg 1985; 72: 34-7. 7. Malmberg M, Graffner H, Ling L, Olson S-A. Recurrence and survival after anterior resection of the rectum using the end to end anastomotic stapler. Surg Gynecol Obstet 1986; 163: 231-4. 8. Phillips RKS, Hittinger R, Blesovsky L, Fry JS, Fielding LP. Local recurrence following ‘curative’ surgery for large bowel can- cer: II. The rectum and sigmoid. Br J Surg 1984; 71: 17-20. 9. Morson BC, Vaughan EG, Bussey HJR. Pelvic recurrence after excision of rectum for carcinoma. Br Med J 1963; 13-8. 10. Whittaker M, Goligher JC. The prognosis after surgical treat- ment for carcinoma of the rectum. Br J Surg 1976; 63: 384-8. 11. Pilpshen SJ, Heilweil M, Quan SHQ, et al. Patterns of pelvic recurrence following definitive resection of rectal cancer. Cancer 1984; 53: 1354-62. 12. Cox DR. Regression models and life tables. J Roy Stat See (series B) 1972; 34: 187-220. 13. Hopkins A. Regression with incomplete survival data. In: Dix- on WJ ed. Stepwise logistic regression. BMDP statistical software. Berkeley: University of California Press, 1981: 576-94. 14. Pahlman L, Glimelius B. Local recurrence after surgical treat- ment for rectal carcinoma. Acta Cbir Stand 1984; 150: 331-5. 15. Stower MJ, Hardcastle JD. The result of 1115 patients with colorectal cancer treated over an I-year period in a single hospital. Eur J Surg Oncol 1985; 11: 119-23. 16. McDermott ET, Hughes SR, Pihl EA, et al. Comparative results of surgical management of carcinoma of the colon and rectum. A series of 1939 patients treated by one surgeon. Br J Surg 1981; 68: 850-5. 17. Phillips RKS, Hittinger R, Blesovsky L, Fry JS, Fielding LP. Local recurrence following ‘curative’ surgery for large bowel can- cer: I. The overall picture. Br J Surg 1984; 71: 12-6. 18. Fielding LP, Stewart-Brown S, Hittinger R, Blesovsky L. Cov- ering stoma for elective anterior resection of the rectum: an out- moded operation? Am J Surg 1983; 147: 524-30.

EDITORIAL COMMENT

John S. Sprat& MD, Louisville, Kentucky

The preceding study attempts resolution of the question, does aggressive pelvic dissection to achieve margins of resection greater than 3.5 cm contribute to tumor dissemination and subsequent distant metastases? The authors address this com- plicated question through a retrospective analysis of 278 patients with cancers of assorted stages operated on between 1973 and 1982 by a large number of sur- geons, with follow-up conducted with no consistent protocol. The cases are strati- fied by the Dukes’ system, which excludes a wide array of significant variables. With Tartter as one of the coauthors, I am disappointed that he did not stratify cases according to the receipt of perioperative blood transfusions.

From the Department of Surgery, University of Louisville, Louisville, Kentucky.

576 THE AMERICAN JOURNAL OF SURGERY VOLUME 157 JUNE 1989