Patterns of Pelvic Recurrence Following Definitive Resections of Rectal Cancer

STEPHEN J. PILIPSHEN, MD,' MARTIN HEILWEIL, PHD,f STUART H. Q. QUAN, MD,' STEPHEN S. STERNBERG, MD,+ AND WARREN E. ENKER, MD'

Patterns of local and distant recurrences following resections for rectal cancer provide clinical perspective for multidisciplinary prevention and follow-up programs. From 1968 to 1976 at Memorial Hospital, 41 2 patients with potentially curable rectal cancer were treated by anterior (AR) or abdominoperineal (APR) resections. First sites of recurrences were categorized as pelvis, liver, distant viscera, and intraabdominal/ retroperitoneal sites. Pelvic recurrences were further evaluated according to the location of the tumor, type of resection, and stage of disease. Among the 412 cases, 182 (44.2%)patients developed recurrence, of which 105 (57.6%)were pelvic. Pelvic recurrence was the predominating site either alone (55 of 103) or with concomitant extra-pelvic sites (50/79). In instances of single-site first recurrence, pelvic failure was recognized earliest at 19.1 months, which was significantly earlier than singledistant visceral sites at 34.9 months. Pelvic recurrence was selectively related to various categories of the Dukes and modified Dukes staging systems. Dukes stage significantly predicted pelvic recurrence rates for Dukes A versus B. Astler-Coller stages of B2 and Cl were associated with significantly lower rates of pelvic recurrence (29.7%and 22.196, respectively) than C2 cancers. The incidence of pelvic recurrence was significantly increased for low and mid rectal cancers as compared with cancers at or above 12 cm. The type of resection utilized (APR versus AR) was associated with no difference in the rate of pelvic recurrence, except for the few patients in whom AR was performed for low rectal Dukes C cancers. Patients with pelvic recurrence had an ultimate disease-free survival of only 3.8% as compared with patients with no pelvic recurrence of whom 77% remained alive without disease or went on to die of other causes. The timing and predominance of pelvic failure in rectal cancer with its own treatment-related morbidity and overall dismal survival outcome justifies organized multidisciplinary efforts to prevent such failure and prospective trials of comprehensive follow-up programs to evaluate improved cure rates or palliation. Cancer 53:1354-1362, 1984.

of rectal cancer continues to represent one of the most disproportionate sources of clinical morbidity despite all previous efforts at adequate resection or of multidisciplinary treatment. Documenting the patterns of local and distant recurrence may offer an increased clinical awareness regarding subsequent diagnostic and curative efforts, and may also effect the proper planning of combined surgical and mulELVIC RECURRENCE

tidisciplinary treatment programs. To this end, we report the patterns of local pelvic and distant recurrences in patients undergoing definitive resections of rectal cancer on the Rectum and Colon Surgery Service at Memorial Hospital. Emphasis is placed on the timing and patterns of pelvic recurrence, because it is this particular group in whom the ultimate benefits from integrated adjuvant therapy are most needed. Methods

From the Rectum and Colon Surgery Service, Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York. * Department of Surgery, Memorial Sloan-Kettering Cancer Center. t Bio-Statistician, Data Analysis Unit of Clinical Information Center, Memorial Sloan-Kettering Cancer Center. $ Department of Pathology, Memorial Sloan-Kettering Cancer Center. Address for reprints: Warren E. Enker, MD, Department of Surgery, Memorial Sloan-Kettering Cancer Center. 1275 York Avenue, New York, NY 10021. The authors thank Ms. Lyn Wood for her technical assistance with the manuscript. Accepted for publication November 15, 1983.

From 1968 to 1976 623 cases of primary rectal cancer were treated at the Memorial Sloan-Kettering Cancer Center. Cases presenting during that period with recurrent disease from prior treatment were excluded from this analysis as were tumors coded as anal (112 cases) and rectosigmoid (5I 9 cases) carcinomas. Data accrual utilized a comprehensive protocol with computerized data entry, retrieval, and analysis. Data included epidemiologic, demographic, preoperative, operative, pathologic, and fol-

1354

No. 6

PELVIC RECURRENCE PATTERNS AFTER RECTAL CANCER RESECTIONS*

Pi/ipshen et a/.

1355

low-up information as obtained from hospital and attendings records. Survival statistics were completed through the BDMPPIL program using the product limit method,* with procedures for group comparisons generated by Breslow3 and Mantel.4 Chi-square analyses were used for establishing statistical significance when comparing recurrence patterns. In selective cases where pathology reports were ambiguous, the original pathologic materials were reviewed with an attending pathologist (S.S.S.) at Memorial Hospital for current postsurgical staging. Patterns of recurrence reported in this study represent the first sites of documented recurrent disease after definitive resection. Criteria and classifications for each site of recurrent disease were established in advance of the study. For patients followed at Memorial Hospital, histologic diagnosis of recurrent disease was frequent. Otherwise, recurrence was documented from first clinical or radiologic signs of disease with an unrelenting course leading to tumor progression and/or to death. For patients undergoing follow-up by physicians remote from the Institution, recurrence was established through a status questionnaire. Where present, laparotomy or postmortem data were utilized to confirm clinical findings especially in the case of pelvic disease. Follow-up information from all sources varied with site of recurrence. Follow-up status was obtained in 382 patients regarding pelvic disease, in 355 patients regarding liver metastases, in 338 patients regarding abdominal recurrence, and in 377 cases regarding other viscera. The overall rate of documentation was 88%. Criteria for establishing recurrent pelvic disease included: histologic confirmation; palpable disease or disease evident on radiographic studies with subsequent clinical progression; and supportive biochemical data, i.e., rising level of carcinoembryonic antigen (CEA). Forty-nine percent of patients with pelvic recurrence had histologic verification of the diagnosis. No patient was accepted as having pelvic recurrence by virtue of interval pain, alone. Pelvic recurrences were categorized according to the following patterns:
1. Anastomotic: a suture-line recurrence with histologic verification, and no clinically apparent contiguous extramural disease. 2. Perianastomotic: limited extramural recurrence at the approximate level of the anastomosis without pelvic fixation, i.e., potentially resectable. 3. Pelvic disease without sacral, sidewall, or anterior fixation by nonoperative criteria. 4. Pelvic disease with sacral or sidewall and anterior fixation, precluding resection.

5. Pelvic disease (with or without fixation) presenting through the anastomosis.

Where pelvic recurrence was reported by an outside physician, pelvic sidewall involvement, the most common pattern, was assumed. Metastases to other viscera, i.e., lung, bone, brain, etc., were documented, but were not categorized. Generalized abdominal recurrence was classified as early peritoneal seeding, widespread carcinomatosis, and/or retroperitoneal disease. Patterns of pelvic recurrence were analyzed by the location of the primary tumor, to the nature of the original curative resection, to the stage of disease, to the utilization of preoperative pelvic irradiation, as well as to age, sex, and other clinical features in our patients. The location of the primary tumor heavily affected the selection of the rectal operation. The majority (89%) of low-rectal cancers (from 0-5 cm from the anal verge) were treated by abdominoperineal resection (APR). Similarly, the majority of mid-rectal cancers (6-1 1 cm), i.e., 86770, were treated by low-anterior resection (LAR). Patients with cancers above 12 cm were included for analysis in order to evaluate and to establish the clear difference between low-to-mid rectal and upper-rectal lesions vis-a-vis pelvic recurrence patterns. Postsurgical staging was performed in accordance with three of the established staging methods: the 1932 Dukes classification, the Astler-Coller modified Dukes classification, and the Gastrointestinal Tumor Study Group (GITSG) modification. In the latter staging system, Stages A, B I , and B2 are the same as Astler-Coller. Stage C I denotes one to four positive lymph nodes and C2, five or more positive nodes (Table l).6-8
Results

From 1968 to 1976, 412 patients underwent curative resections, i.e.,LAR or APR for adenocarcinomas of the rectum Dukes stages A, B, or C. The mean age was 62 years; 59% of the patients were men, and 95% of the patients were white. Thirty-four percent of the patients ( 1 4 1 cases) had lowrectal cancers, while 52.6% of the patients (217 cases) had mid-rectal lesions. The remaining 13.4% were at or above 12 cm. One hundred twenty-five of the patients (89%) with low-rectal cancers underwent an APR while 187 patients (86%) with mid-rectal cancers underwent LAR. Of the remaining cases, 53 patients (98%) had an anterior resection. Stage distribution is summarized in Table 1. Overall, 27.4% of the patients had preoperative irradiation (Table 2). In the majority of cases, this amounted to 2,000 to 3,000 rad (56%) with 80% of the patients receiving over 2,000 rad. The overall 5-year survival for Dukes stages A, B, and

1356
TABLE I.

CANCER March 15 1984

Dukes Staging Systems and Staging Distribution System Distribution Survival
3. Five-Year Survival and Incidence TABLE of Pelvic Recurrence by Stage

VOl. 53

Recurrence*
I3.5%? ( I8/ 133) 29.7%$ (32/1 I I ) 39.9%$ (55/138)

Dukes (1932) A Carcinoma limited to the wall of the rectum B Spread of carcinoma into the extrarectal tissues but no nodal involvement C Lymph node involvement is present Astler-Coller (1954) A Mucosal lesions BI Lesion penetrates the muscularis mucosa B2 Lesion through the muscularis propria but no nodal involvement C1 Lesion that has not penetrated through the muscularis propria, but nodes are positive C2 Lesion has penetrated the muscularis propria and lymph nodes are positive Gastrointestinal Tumor Study Group ( I 976) ],Same as B2 Astler-Coller

137 (33%) 120 (29%)

155 (38%)

5 (1%)
132 (32%) 120 (29%) 53 (13%) 102 (25%)

Dukes A B C Astler-Coller A BI 82 CI c2 Gastrointestinal Tumor Study Group** CI c2

= <0.01 $ Ver.su.s $ P = 0. I2 11 V m u s # P = 0.003 ll Versus # P = 0.002 * Vcrsus ** P = 0.25

84% 64%) 39%

100%

0%
14.1% 29.7%ll 22.4907 49.4%#

83% 64% 60% 29%

( 0/ 5 ) (18/128)

(32/11 I )
( I I /49)

(44/89)

43% 27%

36.8%* (39/106) 48.3%** ( I 5/3 I )

* Percent reflects patients with known follow-up regarding the pelvis.

t Versus $ P

CI One to four nodes involved by tumor C2 Five or more nodes involved by tumor

I20 (29%) 32 (8%) Unknown 3*

By virtue of lymph node count.

C was 62%. Survival according to the various staging systems is listed in Table 3. The 5-year survival figures by Dukes stage were 84% for Stage A, 64% for Stage B, and 39% for Stage C. Statistically significant differences in 5-year survival were appreciated by the further segregation of Stage C patients into the C1 or C2 categories of both systems of Dukes stage modification. From among the 412 cases undergoing definitive resection, 182 patients developed recurrent disease (44.2%). The first instances of documented recurrence were single sites in 103 cases (25%) and multiple sites in 79 cases (19%).Overall, there were 289 sites of disease documented in the 182 cases. Of the 289 sites reported, the pelvis was the most common location (105 cases, 36.3%;P < 0.01) followed by liver (69 cases, 23.9%),other viscera (82 cases, 28.3%), and abdominal or retroperitoneal disease (33 cases, 1 1.4%). In the 382 patients with follow-up information regarding pelvic disease, reported in 27.5% of all patients
TABLE 2. Utilization of External Beam Radiation Therapy Preoperative irradiation Oueration APR LAR Total Postoperative irradiation No.
4 13
%

No. of patients
156 256 412

No.
60 53
1 I3

38.4 20.7 27

2.6 4.7 4. I

17

APR: abdominoperineal resection; LAR: low anterior resection.

(105/382), while liver (l9.4%, 69/355), other viscera (2 1.7%, 82/377) and abdominal recurrences were proportionately lower (9.8%,33/338). Of the 103 instances of single-site recurrence, the pelvis predominated (55/103). The pelvis was not only the single most common site of recurrent disease, but the clinical manifestation of pelvic recurrence occurred at a much earlier time than did other sites of disease. On average pelvic recurrence was evident at 19.1 months following resection whereas liver disease was evident at 24.7 months and metastases to other viscera (34.9 months) occurred significantly later (P< 0.01) (Table 4). Dukes staging was predictive for statistically significant differences in pelvic recurrence rates (Table 3), with progressively higher rates according to the stage of the primary cancer. Statistically significant differences in pelvic recurrence rates were evident between Dukes stages A and B (but not between Stages B and C) and between Stages B2, C I , and C2 of the Astler-Coller modified Dukes classification. Using the GITSG modified Dukes staging system, statistically significant differences in survival were evident between Stages B2, C 1 , and C2, but differences in the rates of pelvic recurrence did not achieve significant levels between Stages C1 and C2. The incidence of pelvic recurrence was related to the original site of the primary rectal cancer (Table 5). There was no difference in the overall incidence of pelvic recurrence between low (0.5 cm) or mid (6-1 1 cm) rectal cancers. Tumors originating at or above 12 cm manifested a significantly lower rate of pelvic recurrence, much more comparable to that seen with abdominal as opposed to pelvic primaries (P< 0.002).

No. 6

PELVIC RECURRENCE PATTERNS AFTER RECTAL CANCER RESECTIONS

TABLE 4. Single Sites of Recurrence in 103 Cases
Site Pelvis Liver Distant Visceral lntraabdominal
No.

Pilipshen d a/.

1357

TABLE 5. Incidence of Recurrence by the Location of the Original Rectal Cancer

Time from resection to recurrence 19.1 mo* 24.7 mo 34.9 mo* Low Mid High 0-18 cm

Pelvis* [39/127] 30.7%? [61/203] 30.l%# [5/52] 9.6966

Liver* 19.9% 19.2% 24%

Visceral* 27.6% 17.4% 23%

lntraabdominal 14.3% 8.3% 4.3% 9.990 (33/338)

55 21 23 4

*P

< 0.01.

27.5% 19.4% 21.7% (105/382) (69/355) (82/377)

The patterns of pelvic recurrence were analyzed for their potential resection either for cure or for palliation. There was no difference in the rate of pelvic recurrence according to the original operation, i.e., APR versus LAR. The pelvic recurrences that followed an APR were characterized by sacral, sidewall or combined patterns of involvement in 90% of cases, i.e.,patterns which were considered unresectable and incurable (Table 6). Of the pelvic recurrences that followed LAR, 67.6% were in proximity to the anastomosis and had various degrees of perianastomotic (4.5%),or pelvic involvement (48.5%). Potentially true anastomotic recurrences were seen in 9 of the 66 cases of LAR complicated by pelvic recurrence ( 14.6%),or 9/240 cases of LAR overall, i.e., 3.8% of cases. Seven of the nine cases were Dukes Stage C. Despite the early and apparently limited nature of the recurrence, only one patient ultimately survived after radiation and chemotherapy (Table 7). The effect of radiation therapy was evaluated. Numerically, preoperative radiation therapy predominated during this time period. There was no specified protocol influencing its utilization, and it was used selectively on more clinically advanced cancers. In this study, no attempt is made to critically comment on it. Other than total treatment dose, no other critical radiation factors were analyzed. Patients receiving preoperative radiotherapy greater than 1000 rad were compared with those patients with either no therapy or doses less than 2000 rad. For cancers located within 0 to 1 1 cm from the anal verge, no significant differences in either subsequent pathologic stage distribution or evidence of pelvic recurrence were noted (Table 8). Pelvic recurrence bore significant implications regarding survival. At the close of this study, 89% of patients with pelvic disease were dead of cancer, 7% were alive with known disease, and only 3.8% were clinically free of disease, despite consideration of re-resection wherever possible. This was in contrast to patients without pelvic recurrence where 22% of patients were dead of disease, 22%were dead of other causes, and 56%were alive without evidence of recurrence. The location of the primary rectal cancer, the type of resection, and Dukes stage were analyzed in relation to pelvic recurrence. In patients with mid-rectal cancers there

* All sites reported. i,e.. single or multiple. t and # versus 9 P < 0.002.

were no significant differences in pelvic recurrence rates whether they were treated by LAR or by APR (Table 9). Low-rectal cancers treated by LAR, although very few in number, had the highest incidence of pelvic recurrence (43%, Table lo), but this did not differ significantly from recurrence rates in patients treated by APR. Where sphincter preservation was attempted in Dukes C low-rectal cancers, recurrence rates were unacceptably high (83%, P = 0.02, Table 10). Nevertheless, AR was attempted in only 16 patients with low-rectal cancers, 13 of whom were women. Despite this apparent difference of pelvic recurrence between low-rectal Dukes C patients treated by APR versus AR, there was no difference in disease-free survival (APR, 26%, AR, 17%, Table 1 I ) at 5 years. The APR patients did achieve better local control, with most deaths being due to extrapelvic recurrence. The liver was recorded as the first site of manifested recurrence in 19.4% of cases available for review. Sixtynine instances of liver metastases were recorded. Only nine of these cases (13%) were considered unilobar at first presentation, and seven of these (10%)were considered solitary or unifocal disease, i.e., potentially curable by resection (Table 12). Other viscera accounted for 2 1.7% of the first recurrence sites reported. Of the 82 reported instances of metastases, 38 were pulmonary (46%), I2 were bone (l4%),and 6 were to brain (7%)(Table 12). A small proportion of patients presented with intraabTABLE 6. Pelvic Patterns of Recurrence
LAR* Anastomotic Perianastomotic Pelvic with fixation Pelvic without fixation Pelvic with anastomotic (with or without fixation) 14.6%(9) 4.5%(3) 29% (19) 4.5%(3) 48.5%(32) APRt

89.7%(35) 10.3%(4) -

One hundred five of 382 patients (27.5%) with follow-up information reported pelvic recurrence. * 66/240 (27.5%). t 39/142 (27.4%). LAR: low anterior resection; A P R abdominopenneal resection.

1358

CANCER March 15 1984

TABLE 7. Anastornotic Suture-Line Recurrences in Sphincter-Preserving Operations
Location Dukes stage*

Vol. 53

No.
Low rectum Mid rectum 212 cm Total
(0-5 cm) (6-11 cm) (0- I8 cm)
0/14 9/175 0/5 I 9/240

(%)

No.
A

("/.)

(5.1%)

B C

2/86 0/68
7/86

(2.3%)
(8.190)

(3.8%)

* Disease-free (5-year) survival

1/9 patients ( I 1%)

dominal disease, the majority of which were either unresectable carcinomatosis or retroperitoneal recurrence (Table 12). In patients in whom the first sites of recurrence were multiple, pelvic recurrence was a significant component of the pattern in 63.3% of cases. Discussion The criteria for reporting recurrent colorectal cancers and pelvic disease, in particular, will vary, compounding the problems of comparing reported incidences. This confusion further reduces the value of comparisons between various surgical procedures, between various institutions, and after various attempts at multidisciplinary therapy unless criteria for declaring recurrence are a prospective part of established protocols or programs. Diagnostic procedures, methods of documentation, acknowledgement or confirmation of diagnosis, presence or absence of histology, the use of interval pain, anatomic definitions of the rectum, first site of recurrence, or cumulative recurrence data, etc., will all affect an analysis of incidence rates, timing, and patterns reported. Similarly, as at our own Institution, some patients cared for

at a tertiary center will obtain follow-up care closer to home, necessitating the acceptance of reports from a wider pool of physicians. Despite these problems, it is important to review the patterns of treatment failure resulting after rectal cancer management. These data may serve as the foundation for efforts at improving the results of multidisciplinary care. They may heighten the alertness of physicians who may recognize recurrent disease when palliation can be most easily attempted.' Our data support the fact that after resection for rectal cancer the predominant site of recurrent disease, whether alone or in combination with other sites, is the pelvis. The pelvis is also the earliest site of recurrent disease. The implications of pelvic recurrence are ominous with 96% of the patients either dead of disease or dying of disease despite our utmost attention to the possibilities of salvage procedures. These data are consistent with those reported by Rao and coworkers." In their series, 40% of the patients developing local recurrences had no distant metastases. Welch and Donaldson, reporting on an autopsy series of 145 patients demonstrated that of the 56 patients with rectal cancer and rectosigmoid cancers, 25% of the cases had local spread while 25% spread to distant sites. The remaining 50% demonstrated combined regional and distant disease. Cass in reviewing recurrences in 15 I patients noted contiguous or regional recurrence in 42 patients (27.8%)." Gunderson and Sosin noted local or regional-nodal failure alone in 50% of patients, and as some component of the overall pattern of recurrence in 92%of cases undergoing second-look laparotomy." In our experience, recurrence is related to the site of disease (rectum versus rectosigmoid), to Dukes stage, and in some cases to modifications of the Dukes staging system which segregates early and late Stage C disease (Astler-

''

TABLE 8. Pelvic Recurrence and Preoperative Radiation Therapy for Rectal Cancers (0 to 1 I cm)

No Preoperative RT or RT less than lo00 rad

No. Dukes A Dukes B Dukes C Total Pelvic recurrence No.

Preoperative RT (greater than 1000 rad) Pelvic recurrence

7

Preoperative RT but Rad unspecified

No.

Pelvic recurrence

89 (3890)
59 (25%)

10
(11%)

24 (29%) 28 (34%) 3 1 (37%) 83

( loo%)

(29%)

0 2 0 2

20 (34%) 39 (44%) 69 (29%)

1 0 (36%)
1 4 (45%) 31 (37%)

5
2
II

88 (37%) 236
( 100%)

R T radiation therapy.

No. 6

PELVIC RECURRENCE PATTERNS AFTER RECTALCANCER RESECTIONS

Pilipshen et al.

1359

Coller modification). While such figures would tend to support the liberal need for adjuvant therapy, we are unable to comment from these data regarding the value of extended operations, radiation therapy, chemotherapy, or combinations thereof on the reduction of pelvic recurrence rates. In addition to Dukes stage, a marked contrast in the incidence of pelvic recurrence was dictated by the height of the primary tumor from the anal verge. While the overall incidence of pelvic recurrence was 30% in tumors located between 0 to I 1 cm from the anal verge, lesions at or above 12 cm had a markedly lower rate of local recurrence, i.e., 9.6%. Wilson and Beahrs and Dionne14 have previously reported similar contrasts between rectal cancers with abdominal colonic carcinomas based on defining the site of the primary lesion. These data would seem to confirm that definition of the rectum at 0 to 12 cm, a definition which has been commonly adopted by some of the multidisciplinary cooperative groups.8 Within the range of 0 to 12 cm there was no overall difference in the incidence of pelvic recurrence (Table 5). Our data did not demonstrate any difference in pelvic recurrence rates for the mid-rectum (6 to 1 I cm) whether treated by APR or by L A R . These results are consistent with previous reports by Deddish and Stearns from this Institution. l 5 Nevertheless, attempts at anterior resection for lowrectal cancers (0 to 5 cm) were associated with a higher than average incidence of pelvic recurrence (43%). In the very small number of cases where Dukes C low-rectal cancers were treated by sphincter-preserving operations, an unacceptably high rate of local recurrence occurred, i.e., 83% as opposed to 35% where APR was properly chosen (Table 9). The patterns of pelvic recurrence were different after APR or L A R . In the case of pelvic recurrence following APR, only 10% initially presented without clinical evidence of pelvic side wall or combined side wall with anterior-posterior involvement precluding resection. Following anterior resection, however, approximately 3.8% of the patients presented with anastomotic (suture line) disease of an initially salvageable nature.16 Despite aggressive treatment, only one such patient survived 5 years (Table 7). These data suggest a worse prognosis than the selective experience of Vassilopoulos and coworkers in which a 49% 5-year survival was seen in patients referred for reoperation for anastomotic recurrence whose initial resections were performed outside of their specialty institutions. The poor survival witnessed in our own patients undoubtedly reflects the natural history of Dukes C rectal cancer (seven of nine initial anastomotic recurrences) while some of Vassilopoulos patients had resec-

TABLE 9. Pelvic Recurrence in Cancers of the Mid Rectum (6 to I 1 cm) Dukes stage A B C
0veraI1

APR*
0%(O/I I) 10%(2/1 I ) 61% (4/6)

LARt 15% ( l0/65) 36% (I 7/47) 44% (28/63) 3 I .4%

Overall 5-year survival 84% 56% 45% 65%

2 I .4%

t One hundred seventy-five patients. APR: abdominoperineal resection; LAR: low anterior resection.
tions for recurrences which surrounded intraabdominal anastomoses. In general, however, our predictive factors for pelvic recurrence, i.e., location below 12 cm, invasion of perirectal fat and involvement of lymph nodes are in agreement with other studies (Table 13). The overall pattern of recurrence following resection for rectal cancer differed appreciably from data reported on patterns of recurrence after resection of intraabdominal colonic carcinoma. Data reported by the GITSG for abdominal colonic carcinomas (above 12 cm) demonstrate that the majority of recurrences are distant, with the liver being the most common site of recurrence in 45% of cases. As we have indicated previously, from the data at hand, we cannot report on the value of adjuvant therapy. Nevertheless, the extreme morbidity of pelvic recurrence suggests that major efforts to eliminate pelvic recurrence are justified. These might include extended pelvic lymphadenectomy at the time of primary resection for rectal cancer, or adjuvant radiation therapy with or without radiation-enhancing drugs. Neither chemotherapy nor immunotherapy has been reported to create any recognizable impact on this problem. The roles of pelvic lymphadenectomy and/or radiotherapy are discussed elsewhere. Suffice to say that both subjects constitute important bases of interest for various clinicians who are
TABLE 10. Pelvic Recurrence in Cancers of the Low Rectum (0 to 5 cm) Dukes stage
A B C

* Twenty-eight patients.

APR* 14% (7/36) 32% (10/31) 35% (16146) 29%

LARt
0%(0/5) 33% ( l / 3 ) 83% (5/6)

Overall 5-year survival 82% 61% 32% 5590

Overall

4 3%

t Fourteen patients.

One hundred thirteen patients.

APR: abdominoperineal resection; LAR: low anterior resection.

1360

CANCER March I5 1984

TABLE I I. The Outcome of Pelvic Recurrences in Dukes C Low-Rectal Cancers (0 to 5 cm) Recurrence rate No. APR 16/46.
9 0

VOl. 53

Five-year survival status

No.
9 0

Condition Alive, no cancer Dead of cancer Dead other causes Alive, no cancer Alive with cancer Dead of cancer

Sites of first failure 16 pelvic (55%) 13 other sites no pelvic

35%

12/46? 29/46 5/46 1/6t 116 416

26% 63% I I%
17%

LAR

5/6*

8390

I790 66%

5 pelvic ( 100%)

= 0.02.
=

APR: abdominoperineal resection; LAR: low anterior resection.

tP

not significant.

interested in improving the end results of cancer treatment. These data on the incidence, the timing, and the patterns of recurrence following resection for rectal cancer raise serious questions about the role of intensive followup after adequate resection. The stated goals of followup include the detection of recurrent tumor at the time when cure may still be possible, the detection of metachronous disease at its earliest stages, the detection and the elimination of premalignant precursors, i.e., adeno-

TABLE 12. Extra-Pelvic Recurrences No. Live?? Unilobar, single Unilobar, multiple Multilobar Multimets, unspecified Total Visceral*.$ Lung Bone Brain Miscellaneous, other single distant sites Multiple visceral Total Intraabdominal** Peritoneal seeding Carcinomatosis Retroperitoneal Retroperitoneal with seeding carcinomatosis Percent

13% 32 69 87%
100%

38 12 6 14 12 82

46% 15% 1% 17% 15%

100%

2 16 14
1

6% 49% 42% 3%
100%

Total

33

* Does not preclude involvement of other sites or pelvic occurring concomitantly. t Sixty-nine of 355 patients (19.4%). 4 Eighty-two of 377 patients (21.7%). 8 Thirty-three of 338 patients (9.9%).

mas, and the development of cohorts such as groups that might benefit from adjuvant therapy, detection of associated cancers at other sites, genetic syndrome evaluation, etc.20The economics of follow-up depend on the orientation of the follow-up schema and on the frequency with which certain tests are employed. Colonic mucosal precursors of malignant disease and colonic metachronous disease can be detected for the most part by the prospective use of fecal occult blood testing, rigid or flexible sigmoidoscopy, and colonoscopy or air-contrast barium enema at appropriately timed intervals. The long interval between the onset of primary and metachronous disease and the relativelyslow growth ofadenomas would indicate that expense in this area could be minimized without seriously compromising the detection of new cancers.2'*22 Less can be said, however, in support of the frequent testing of patients for the discovery of recurrent disease at a point of salvageable intervention. Our data would indicate that despite aggressive attention to the management of recurrent disease at a tertiary specialty center, 96% of patients with pelvic recurrence were dead or dying of disease at the close of this study, and fewer than 10% of the patients with metastatic disease to other viscera would be considered resectable for cure by extraordinary efforts.While earlier detection of disease using serial CEA assays may focus our attention on more patients who are considered salvageable,the natural history of rectal cancer would suggest ~ t h e r w i s e . ~Furthermore, ~.*~ reports continue to appear in the literature suggesting that the reappearance of symptomatic recurrent disease either coincides with, appears at intervals between, or even after, the onset of CEA elevation, making comprehensive follow-up more useful in the detection of recurrent disease than CEA elevation. Beart and coworkers have reported no salvage from recurrent disease based on comprehensive follow-up including CEA at 12-to-15-week intervals.25 While Martin and coworkers would advocate more frequent CEA determination and would advocate the con-

No. 6

PELVIC RECURRENCE PATTERNS AFTER RECTALCANCER RESECTIONS

TABLE 13. Pelvic Recurrence in Other Surgical Series
Reference No. of patients
204
151

Pilipshen et al.

1361

Percent of Dukes C
27% 30% 43%

Type of operation APR 46% LAR 54% APR

Location of tumor Rectum, rectosigmoid, and sigmoid Rectum, rectosigmoid Rectum

Pelvic recurrence
21% 27.8% Low: 32% Mid: 9% Upper: 3.6% Overall: 20.4% APR: 46% LAR: 55% 0-5 cm: 56% 6-10 cm: 55% > I 0 cm: 57% Overall alone: 50% With other sites: 92% APR: A (6%). B (25%),C (38%) LAR: A (2%).B (23%),C (33%)

Rao er al. (1981)' Cass el a/. (1976)" Moosa er al. (1975)''

I52

Steams ( 1974)16 Gunderson and Sosin ( 1974)12

86* 75t

APR LAR APR LAR

75 I1 67 8

6-11 cm (mid rectum) Rectum

Slanetz et a/. ( 1971)''

514

42%

APR 211 LAR 247

8-18 cm

* Selected cases with follow-up. t Selected second-look laparotomies.

tinued use of CEA for the detection of recurrent disease,26 our data would suggest little long-term likelihood of cure even by aggressive surgical intervention. It is unlikely that a timing interval of follow-up alone can explain these differences. These data raise serious doubts as to the value of intensive follow-up programs which are predominantly oriented towards the detection of recurrent disease with the aim of aggressive surgical intervention. An alternative remains the application of extraordinary efforts to those patients who do present with evidence of resectable metastatic disease while evaluating the expense of follow-up programs deliberately designed to detect such cases as a separate issue. A further alternative remains the segregation of follow-up effortsinto long-range and short-range goals with more intensive application of long-range programs for metachronous disease or for other vulnerable sites of new cancers in those patients who are either young or who have early stages of disease.
Conclusions

APR: abdominoperineal resection: LAR: low anterior resection.

definition of the rectum for multidisciplinary studies, i.e.. in adjuvant therapy. Pelvic recurrence rates are extraordinarily high when sphincter-preservingoperations are attempted in low-rectal cancer especially in Dukes stage C disease. Such efforts are not appropriate in view of the extraordinary pelvic recurrence rates.' The overall survival following pelvic recurrence is dismal and the clinical course of patients with pelvic recurrence is characterized by extreme morbidity. In view of these findings, organized efforts to eliminate pelvic recurrence are justified by these data. In view of the poor survival of patients with pelvic recurrence and distant metastatic disease following adequate resection, the efforts and the expense of comprehensive follow-up or of serial CEA evaluations deserve prospective evaluation in terms of patient welfare and expense.
REFERENCES
I. Dixon WJ,ed. BMDP Statistical Software. Berkeley: UCLA Press, 1981. 2. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958; 53:457-48 I . 3. Breslow N. Generalized Kruskal-Wallis test for comparing k samples subject to unequal patterns of censorship. Biomefrika 1970 30:89-99. 4. Mantel N. Evaluation of survival data and two new rank-order statistics arising in its consideration. Cancer Chemother Rep 50: 163170. 5. Wilson SM, Beahrs OH. The curative treatment of carcinoma of the sigmoid, rectosigmoid, and rectum. Ann Surg 1976; 183:556-565. 6. Dukes CE. The classification of cancer of the rectum. J Parhol Bucteriol 1932: 35:323.

The pelvis either alone or in combination with other sites is the earliest and most predominant site of first recurrence after adequate resection of low-rectal or midrectal cancer (0-1 1 cm). The incidence of pelvic recurrence directly relates to Dukes stage and in the case of Astler-Coller staging, to the further stratification of B2, C1, and C2. A statistically significant difference exists in the incidence of pelvic recurrence for lesions situated between 0 to 11 cm from the anal verge or those at or above 12 cm. These data support the use of 0 to 12 cm as a working

I362

CANCER March 15 1984

VOl. 53

7. Astler VB, Coller FA. The prognostic significanceof direct extension of carcinoma of the colon and rectum. Ann Surg 1954; 139:846-852. 8. Holyoke E, Lokich J. Wright H. Adjuvant therapy of adenocarcinoma of the colon following clinically curative resection. Proceedings of the Gastrointestinal Tumor Study Group, Division of Cancer Treatment, National Institutes of Health, National Cancer Institute, Bcthesda, Maryland, July, 1975. 9. Polk HC, Spratt JS. Recurrent colorectal carcinoma: Detection, treatment, and other considerations. Surgery 197 I ; 69:9-23. 10. Rao AR, Kagan AR, Chan PM, Gilbert HA, Nussbaum H, Hintz BL. Patterns of recurrence following curative resection alone for adenocarcinoma of the rectum and sigmoid colon. Cuncer 1981; 48: 14921495. 1 1. Welch JP, Donaldson GA. The clinical correlation of an autopsy study of recurrent colorectal cancer. Ann Surg 1979; 189:496-502. 12. Cass AW, Million RR, PfalT WW. Patterns of recurrence following surgery alone for adenocarcinoma of the colon and rectum. Cuncer 1976; 3712861-2865. 13. Gunderson LL, Sosin H. Areas of failure found at reoperation (second or symptomatic look) following Curative Surgery for adenocarcinoma of the rectum. Cuncer 1974; 34: 1278- 1292. 14. Dionne L. The pattern ofblood-borne metastasis from carcinoma of the rectum. Cuncer 1965; 18:775-780. 15. Deddish MR, Stearns MW. Anterior resection for carcinoma of the rectum and rectosigmoid area. Ann Surg 1961; 154:961-966. 16. Pihl E, Hughes ESR, McDermott FT,Price AB. Recurrence of carcinoma of the colon and rectum at the anastomotic suture line. Surg Gynecol Obstet I98 1 ; 153:495-496. 17. Vassilopoulos PP, Yoon JM, Ledesma EJ, Mittelman A. Treatment of recurrence of adenocarcinoma of the colon and rectum at the anastomotic site. Surg Gynecol Ohstet 198 I ; I52:777-780. 18. Mittelman AM, Knowlton AH, Weiland L. Phase Ill study of

adjuvant therapy of surgically operable rectal carcinoma, Stages 8 2 and C. Proceedings Gastrointestinal Tumor Study Group, National Cancer Institute, 1975. 19. Enker WE, Shank B, Kemeny N, Rotstein L. Defining the needs for adjuvant therapy of rectal and colonic cancer. Surg Clin North Am I98 I ; 6 I : 1295-1 3 10. 20. Enker WE, Kramer RG. The following of patients after definitive resections for large bowel cancer. World J Surg 1982; 6578-584. 2 I . Winawer SJ, Miller DG.Screening approaches based on risk in colorectal cancer prevention, epidemiology, and screening. In: Winawer SJ, Schottenfeld D, Sherlock P, eds. Progress in Cancer Research and Therapy. New York: Raven Press, 1980; 159-165. 22. Gilbertsen BA, McHugh R, Schuman L, WilliamsSE. The earlier detection of colorectal cancers: A preliminary report of the results of the occult blood study. Cuncer I 9 8 0 45:2899-290 I . 23. Stearns MW Jr. The choice among anterior resection, the pullthrough and abdominoperineal resection of the rectum. Cuncer 1974; 34:969-97 I . 24. Moossa AR, Ree PC, Marks JE, Levin B, Platz CE, Skinner DB. Factors influencing local recurrence after abdominoperineal resection for cancer of the rectum and rectosigmoid. Br J Surg 1975; 62:727730. 25. Beart RW Jr, Metzger PP, OConnell MJ, Schutt AJ. Postoperative screening of patients with carcinoma of the colon. Dis Colon Rectum 1981; 24585. 26. Martin EW Jr, Cooperman M, Carey LC, Minton JD. Sixty second-look procedures indicated primarily by rise in serial carcinoembryonic antigen. J Surg Res 1980 28:389. 27. Slanetz CA, Herter FP, Grinnell RS. Anterior Resection versus abdominoperineal resection for cancer of the rectum and rectosigmoid. A m J S u r g 1972; 123:110-117.