Br. J. Surg. 1992, Vol. 79, December, 1372-1 375
J. J. De Cosse, S. BUlow*, K. Neale?, H. JarvinenS, T. Almg, R . Hultcrantzg, F. Moesgaard", C. Costello and the Leeds Castle Polyposis Group N e w York Hospital-Cornell Medical Center, N e w York, USA, *Hvidovre Hospital, Copenhagen, Denmark, t S t Mark's Hospital, London, UK, $Helsinki University Central Hospital, Finland and SKarolinska Hospital, Stockholm, Sweden Correspondence to: Professor J. J. De Cosse, Department of Surgery, New York
Hospital-Cornell Medical Center, 525 East 68th Street, New York, New York 10021, USA
Rectal cancer risk in patients treated for familial adenomatous polyposis Total colectomy with ileorectal anastomosis ( I R A ) in familial adenomatous polyposis ( F A P ) leaves patients at risk f o r rectal cancer. T o assess this risk, the rectal cancer incidence in 297patients with FAP undergoing I R A since 19.51 was determined in the population-based registers of Denmark, Finland and Sweden. A t the same time, detailed data on SO patients with FAP and invasive rectal cancer were obtained from I I international polyposis registries. The cumulative incidence of rectal cancer was 13.1 per cent at 2.5 years. The 5-year survival rate of patients with FAP developing rectal cancer was 71 per cent. Combining both studies, the risk of dying f r o m rectal cancer after I R A was 2.0 per cent at 1.5 years of follow-up. These results justijy I R A as primary treatment f o r most patients; restorative proctocolectomy is preferred f o r some subgroups. The high all-cause mortality rate observed in this relatively young population necessitates lifelong surveillance of patients with FAP.
Two treatment options are currently available for patients newly diagnosed with familial adenomatous polyposis ( F A P ) without cancer of the middle or distal rectum: total colectomy with ileorectal anastomosis (IRA) or restorative total proctocolectomy with ileoanal reservoir (RPC). Preservation of the rectum probably affords a superior quality of life but leaves the patient at risk of rectal cancer and its consequences. The extent of these risks is controversial. At the Mayo Clinic', 46 of 143 patients with F A P (32.2 per cent) developed cancer in the retained rectum during a median follow-up of 17.5 years. In a population-based study in Denmark', the cumulative risk of rectal cancer at 5 and 10 years after IRA was 3.5 and 13.3 per cent respectively. In 174 patients who had undergone IRA at St Mark's Hospital, London3, the cumulative risk of rectal cancer rose to 13 per cent at 25 years. At the Cleveland Clinic4, ten of 133 patients (7.5 per cent) at risk developed rectal cancer during a median follow-up of 5.5 years, ranging up to 37 years. The mortality risk of rectal cancer after IRA has been difficult to examine systematically. In a group of seven patients with rectal cancer, five survived long term5. At St Mark's Hospital, 12 patients had rectal cancer, resulting in four deaths6. Mortality risks from causes other than colorectal cancer have been p ~ b l i s h e d ~ * ' , ~ . In 1991 the Leeds Castle Polyposis Group initiated a two-part study of long-term treatment outcomes in patients with FAP. In one arm of the study, population-based registries in Denmark, Finland and Sweden assessed 297 patients who had undergone IRA over a 40-year period for long-term risk of rectal cancer. In the other arm, data on clinical findings and long-term results of invasive rectal cancer in 50 patients after IRA were collected from 11 polyposis registries in seven countries.
analysed for each patient; privacy was protected. Survival functions and risk of rectal cancer were determined by B M D P Statistical Software (BMDP, Los Angeles, California, USA) according to actuarial methods'. In the registry study, 11 member registries in the Leeds Castle Polyposis Group submitted data on all patients with FAP who had undergone IRA and were diagnosed before mid-1986 with invasive rectal cancer in the retained rectum. The registries applied uniform standards for basic clinical determinants, tumour staging and follow-up. Only patients for whom a minimum of 5 years offollow-up was available were included. Patients with in situ rectal cancer were excluded; privacy was protected. Data were entered into an Epi Info database (USD, Stone Mountain, Georgia, USA)". Analyses were performed on continuous data by Student's t test and on categorical data by the x2 test or, where appropriate, Fisher's exact test. Survival data were analysed with the Lifetest procedure (SAS Institute, Cary, North Carolina, USA)''. Five patients with rectal cancer and 135 who had undergone IRA were common to the Scandinavian and registry components.
Results
The Scandinavian study participants were 297 patients who had undergone total colectomy and IRA for FAP and were followed for at least 1 year before 1991. The patients were identified through the national registers of Denmark (88 patients), Finland (90) and Sweden (119). These patients were unselected over a 40-year span from the entire population of these nations; accession is believed to be virtually complete. Basic demographic and clinical data were collected and
Scandinavian study The median age of the 297 patients at IRA was 27 (range 10-77) years. Thirty-four patients were 350 years of age at the time of IRA. There were 145 male and 152 female patients. Of the 297 patients, 102 were probands with a median age of 36 (range 12-77) years and 195 were call-up patients with a median age of 24 (range 11-63) years. Thirty-five of the 297 patients ( 1 1.8 per cent) had colonic cancer, three (1.0 per cent) rectal cancer found within 1 year of IRA and regarded as a prevalent cancer, and 13 (4.4per cent) subsequently developed rectal cancer. Fourteen patients were lost to follow-up. For the 259 patients without colonic cancer, the median age ?It IRA was 25 (range 11-77) years and survival rates were 97.1 per cent at 5 years, 92.5 per cent at 10 years, 88.8 per cent at 15 years and 88.8 per cent at 20 years, by which time 42 patients remained at risk. For the 35 patients with colonic cancer, the median age at IRA was 44 (range 23-68) years and survival rates were 72.9 per cent at 5 years, 63.8 per cent at 10 years and 55.8 per cent at 15 years, seven patients remaining at risk by this time. During follow-up after IRA, the remaining rectal segment was removed for non-malignant causes (multiple large polyps,
1372
0007 1323/92/121372-04
Patients and methods
'(
1992 Butterworth Heinemann Ltd
Rectal cancer in familial adenomatous polyposis: J. J. De Cosse et al.
Four patients with rectal cancer were treated within 1 year of IRA and were regarded as having a prevalent cancer.
0
5
10
15
20
25
Time after I R A (years1
Figure 1 Cumulative probability of rectal cancer after ileorectal anastomosis ( I R A ) over 25 years of follow-up among patients in the Scandinavian study. Bars are 95 per cent conJdence intervals
severe dysplasia) in 34 patients. The cumulative probability of excision of the remaining rectum for causes other than carcinoma was 4.4 per cent at 5 years, 9.4 per cent at 10 years, 14.4 per cent at 15 years, 17.4 per cent at 20 years and 30.9 per cent at 25 years. The cumulative risk of rectal cancer (Figure 1) after IRA in the 294 patients without rectal cancer was 3.1 per cent at 5 years, 4.5 per cent at 10 years, 5.7 per cent at 15 years, 9.4 per cent at 20 years and 13.1 per cent at 25 years; 21 patients remained at risk. Of the 13 patients who developed rectal cancer, four died from metastatic disease. Sex and the presence of colonic cancer at the time of IRA did not affect the risk of subsequently developing rectal .carcinoma. Inclusion in risk calculation of the three patients diagnosed with rectal cancer within 1 year of IRA did not change these estimates substantially: the cumulative risk was 10.5 per cent at 20 years and 14.2 per cent at 25 years. The effect of age on risk for rectal cancer was examined. The cumulative probability of developing rectal cancer by age 50 years was 5.2 per cent (95 per cent confidence interval 1.2-9.2 per cent). By 60 years this figure was 14.1 per cent (95 per cent confidence interval 4.1-24.1 per cent). Registry study From the 11 polyposis registries, 50 patients with invasive rectal cancer in the retained rectum were identified, constituting 3.7 per cent of 1363 registered patients who were at risk after IRA performed before mid-1986. The median age of these 50 study patients at IRA was 33 (range 10-60) years. There were 27 male and 23 female patients. Eleven of the 50 had an invasive colonic cancer at the time of IRA. The median interval from IRA to the detection of invasive cancer was 11 (mean 13.4, range 0.1-36) years. The median interval was 13 years for women and 8 years for men ( P = 0.29).
Br. J. Surg., Vol. 79, No. 12, December 1992
Determinants at detection of rectal cancer. The median age of the 50 patients at the detection of rectal cancer was 46 (mean 46, range 18-67) years. Lower gastrointestinal symptoms were present in 19 patients, absent in 19 and not known in 12. The interval from the most recent endoscopy to cancer detection was < 6 months in 24 patients, 7-12 months in eight, 13-24 months in seven, > 2 years in four and unknown in seven. In 37 cases, the previous endoscopy had been performed by a member of the participating polyposis centre or by another physician regarded as reliable. Subsequent Dukes’ classification and survival were not affected by the presence of lower gastrointestinal symptoms and had little relationship to the interval from previous endoscopy. All four patients whose cancer was found > 2 years after their last endoscopy had a Dukes’ C lesion. The 50 patients were segregated into quartiles between 1952 and mid-1986 by the date of detection of rectal cancer and the quartiles compared for distribution of Dukes’ classification of the rectal tumour (unknown in three). More Dukes’ A and fewer Dukes’ C cancers were found in the most recent quartile (Figure 2 ) ( P = 0.02, Fisher’s exact test). The distance from the anal verge to the IRA was known in 40 patients; the median was 15.0cm and the mean(s.d.) 14.3(4.3) cm. In 38 patients the distance from the anal verge to the distal edge of the rectal cancer was known. The median was 8.5 cm and the mean(s.d.) 7.5(3-6) cm. The maximum diameter of the rectal cancer was known in 40 patients; the median was 3-9 cm and the mean(s.d.) 3.3( 1.7) cm. The distances from the central point of the rectal tumour to the anal verge and IRA were determined in the 33 patients for whom all measurements were known (Figure 3). Of the 33 tumours, 18 occurred in the most cephalad third of the remaining rectum (1’= 2.21, 1 d.f., P = 014). These 33 patients included ten of 11 who had colonic cancer at the time of IRA. Of the ten, eight had a rectal tumour in the upper third of the rectal segment ( P = 0.08, Fisher’s exact test). No tumours were reported in the most distal 4 cm of the rectum. Of ten patients with a tumour < 2 cm in diameter, only one died from rectal cancer. For larger tumours no correlation of size with survival was evident.
12
10
VI
Y
8
c
.-w
m
n c
r
n
6
0
0
z
4
2
0 1952-1976
1977-1979
1980-1983
1984-1986
Year
Figure 2 Dukes’ classification of invasive rectal cancer by year of diagnosis among patients in ihe registry study. 0 ,Dukes’ A ; 0 ,Dukes’ B and ‘D’(disseminated); R, Dukes’ C. Patients diagnosed in the latest guartile were more likely than those in earlier quartiles to have Dukes’ A rectal cancers ( P = 0.02, Fisher’s exact test)
1373
Rectal cancer in familial adenomatous polyposis: J . J . De Cosse et al
VI
c)
C
._ +. m
a L
0
z
I
Discussion
n 1t-l
2
2
0
4
6
12
10
8
14
16
20
18
D i s t a n c e o f t u m o u r f r o m anal v e r g e (cm)
Figure 3 Distance from the anal verge to midpoint of the rectal tumour in patients in the registry study. 0, Tumours within the cephulad third of retained rectum
Table 1 Dukes’ classification and mortality from rectal cancer by sex ~~
~
F
M ~
Dukes’ classification
No
A
8 5 11
B C
Disseminated Unknown
Total
-~
~~
~~
Deaths
3 0
0 2 I 3 0
21
12
No
Deaths
7
2 I
6
0
0
0
I
3
1
23
4
Treatment. Of the 50 patients, 42 underwent abdominoperineal resection and six low anterior resection; one of the latter was treated palliatively. Two patients with disseminated cancer did not undergo surgery and two died after operation. Results of treatment. Fifteen patients had a Dukes’ A cancer, 12 Dukes’ B and 17 Dukes’ C. Three had disseminated cancer, and classification was unknown in three (Table 1 ). The overall all-cause 5-year survival rate was 66 per cent and the 10-year rate 36 per cent. The 5-year rectal cancer survival rate was 68 per cent, 56 per cent in men and 83 per cent in women ( P = 0.04, log rank test) (Table 1 ). Of the I1 men with a Dukes’ C lesion, eight had full-thickness penetration of the primary tumour (Astler-Coller class C,). Of the six women with regional lymph node involvement, three had full-thickness penetration of the primary tumour. Excluding the three patients treated palliatively and the two who died after operation, the remaining 45 patients had an actual 5-year survival rate of 71 per cent. The 5-year survival rates were 77 per cent for the 15 patients with Dukes’ A cancer, 75 per cent for the 12 with Dukes’ B lesion and 62 per cent for the 17 with Dukes’ C cancer. Of the 17 patients who died during the first 5-year interval, 14 died from rectal cancer and three after operation. The actual 10-year survival rate in the 29 eligible patients was 55 per cent. During this second 5-year interval, six patients died: two from rectal cancer shortly after the 5-year mark and four from other causes. The 10-year rectal cancer survival rate for patients with Dukes’ C lesions was 50 per cent. Of the 13 patients surviving to 10 years, five have since died from conditions not related to colorectal cancer.
1374
This study has determined a cumulative risk for rectal cancer of 13.1 per cent (95 per cent confidence interval 3.3-22.9 per cent) at 25 years in the population-based Scandinavian data. This result is similar to the 13 per cent risk estimate at 25 years in a hospital-based registry3. If these estimates are combined with mortality rates from the registry study, the risk of dying from rectal cancer, given prior IRA with absence of colonic cancer, was 1.0 per cent (3.1 x 0.32) at 10 years and 2.0 per cent (4.5 x 0.45) at 15 years. This is a more favourable result than that of Bussey et aL3. These estimates do not consider the potential benefit of better surveillance and of nutritional modification and chemopreventionL2-’‘. The present observations suggest that retention of the rectum may have less life-threatening consequences than generally believed, and support a general policy of IRA in primary operative management of patients with FAP”. RPC should be the primary procedure for those with middle or upper rectal tumours or with large or confluent adenomas of the rectum, and for those who will not subscribe to regular follow-up. The Scandinavian study may provide a rationale for recommending RPC as the primary procedure in older patients. The benefit of this procedure must be weighed against an increased risk of morbidity at the primary operation and further deterioration during subsequent ageing. Of the 34 Scandinavian patients who underwent IRA at 2 5 0 years of age, only two subsequently developed rectal cancer. The course of F A P in older newly diagnosed patients may be relatively indolent, making a more extensive operation and attendant morbidity difficult to justify. -In the 50 registry patients, the distribution of rectal cancer by Dukes’ classification approximated that observed in patients with sporadic colorectal cancer. Survival by Dukes’ class, however, seemed superior, especially in women. This may be due to greater health awareness, especially among women, leading to more expeditious diagnosis and treatment and earlier substages of rectal cancer. The registry group included more men than women, and the former had shorter time intervals from IRA to rectal cancer, a greater likelihood of being diagnosed Dukes’ C and Astler-Coller C, (these differences were not significant) and a higher mortality rate. Evidence for the benefit of surveillance was mixed. Except for delay in follow-up endoscopy > 2 years, the frequency of endoscopy made no difference in Dukes’ classification at diagnosis or in survival, a finding previously notedI6. Patients diagnosed since 1978 were more likely to have a Dukes’ A tumour. This difference, if real, may be the product of improved health awareness in patients and alertness in physicians. These observations suggest that follow-up endoscopy should be performed at least every 6 months. As in sporadic colorectal cancerL7(except for rectal lesions < 2 c m in diameter) tumour size was not correlated with survival. After discounting the epithelium between the anal verge and the dentate line, the most caudad few centimetres of rectal mucosa appeared spared from rectal cancer. This may be due to the presence of transitional, not columnar, epithelium in the most distal 1-2 cm of rectal mucosaL8.This finding may support mucosal preservation during RPC. Because rectal cancer appeared to occur more frequently in the upper third of the retained rectum, especially in those with colonic cancer at the time of IRA, a low IRA 10-12 cm from the anal verge is recommendedL9. This study does not address the full range of causes of mortality in this comparatively young patient population. Perioperative risks should have decreased over the 40-year span of this study but other genetically based risks have become more p r ~ m i n e n t ~ .In~ .one ~ . registry of 174 patients with IRA, 35 died during follow-up, only seven of them from colorectal cancer3. Another study of 36 patients who died after IRA reported only three deaths from rectal cancer’. In both studies
Br. J. Surg., Vol. 79, No. 12, December 1992
Rectal cancer in familial adenomatous polyposis: J. J. De Cosse et al. the most frequent causes of death were desmoid disease a n d periampullary cancer; these occurred in only 4 a n d 1 per cent respectively of patients of the Danish register*'. During the 353 patient-years of follow-up represented by the registry study, 16 patients died from rectal cancer, 14 within the first 5 years. Twelve patients died from other causes: perioperative deaths (three), upper gastrointestinal cancer (four), suicide (two), desmoid disease (one), thermal t r au m a ( o n e) an d myocardial infarction (one). Nine of these deaths occurred after the 5-year mark. These findings highlight the need for lifelong comprehensive monitoring of these patients. The suicides, noted by others3,*, justify counselling for patients whose lives remain threatened by continuing risk of inherited disease.
5.
6.
7. 8. 9. 10.
11.
Acknowledgements The authors dedicate this paper to H. J. R. Bussey (1907-1991). The participants in the Leeds Castle Polyposis Group Study are: F. A. Macrae and C. Watts, Victoria, Australia; T. Berk and Z. Cohen, Toronto, Canada; S. Biilow and F. Moesgaard, Copenhagen, Denmark; H. Jarvinen, Helsinki, Finland; K. F. Neale and J. P. S. Thomson, London, UK; H. Itoh, Kitakyushu, Japan; S. Baba, Shizuoka, Japan; T. Muto and T. Iwama, Tokyo, Japan; T. Alm and R. Hultcrantz, Stockholm, Sweden; D. G . Jagelman and E. McGannon, Cleveland and Fort Lauderdale, USA; and J. J. De Cosse and C. Costello, New York, USA.
12. 13. 14. 15.
16.
References Bess MA, Adson MA, Elveback LR, Moertel CG. Rectal cancer following colectomy for polyposis. Arch Surg 1980; 115: 460-7. Bulow S. The risk of developing rectal cancer after colectomy and ileorectal anastomosis in Danish patients with polyposis coli. Dis Colon Rectum 1984; 27: 726-9. Bussey HJ, Eyers AA, Ritchie SM, Thomson JP. The rectum in adenomatous polyposis: the St Mark's policy. Br J Surg 1985; 12: S29-31. Sarre RG, Jagelman DG, Beck GJ et al. Colectomy with ileorectal anastomosis for familial adenomatous polyposis: the risk of rectal cancer. Surgery 1987; 101: 20-6.
Br. J. Surg., Vol. 79, No. 12, December 1992
17.
18. 19. 20.
Sener SF, Miller HH, De Cosse JJ. The spectrum of polyposis. Surg Gynecol Obstet 1984; 159: 525-32. Dozois RR (Moderator). Surgical aspects of familial adenomatous polyposis. Int J Colorectal Dis 1988; 3 : 1-16. Jagelman DG, De Cosse JJ, Bussey HJ, the Leeds Castle Polyposis Group. Upper gastrointestinal cancer in familial adenomatous polyposis. Lancet 1988; i: 1149-51. Arvanitis ML, Jagelman DG, Fazio VW, Lavery IC, McGannon E. Mortality in patients with familial adenomatous polyposis. Dis Colon Rectum 1990; 33: 639-42. Cutler SJ, Ederer F. Maximum utilization of the life table method in analyzing survival. J Chron Dis 1958; 8 : 699-712. Dean AD, Dean JA, Button AH, Dicker RC. Epi Info, Version 5: A Word Processing, Database, and Statistics Program for Epidemiology on Microcomputers. Stone Mountain, Georgia: USD, 1990. SAS Institute. SAS Technical Report P-179, Additional SASISTATProcedures,Release 6.03. Cary, North Carolina: SAS Institute, 1988. Bussey HJR, De Cosse JJ, Deschner EE et al. A randomized trial of ascorbic acid in polyposis coli. Cancer 1982; 50: 1434-9. De Cosse JJ, Miller HH, Lesser ML. Effect of wheat fiber on rectal polyps in patients with familial adenomatous polyposis. J Natl Cancer Inst 1989; 81 : 1290-1. LaBayle D, Fischer D, Vielh P e f a / . Sulindac causes regression of rectal polyps in familial adenomatous polyposis. Gastroenterology 1991; 101: 635-9. Madden MV, Neale KF, Nicholls RJ et al. Comparison of morbidity and function after colectomy with ileorectal anastomosis or restorative proctocolectomy for familial adenomatous polyposis. Br J Surg 1991; 18: 789-92. Moertel CG, Hill JR, Adson MA. Management of multiple polyposis of the large bowel. Cancer 1971; 28: 160-4. Wolmark N, Cruz I, Redmond CK, Fisher B, Fisher ER, contributing NSABP investigators. Tumor size and regional lymph node metastasis in colorectal cancer: a preliminary analysis from the NSABP clinical trials. Cancer 1983; 51: 1315-22. Fenger C. The anal transitional zone: location and extent. APMIS 1979; 81: 319-86. Jarvinen HJ. Time and type of prophylactic surgery for familial adenomatosis coli. Ann Surg 1985; 202: 93-7. Biilow S . Familial adenomatous polyposis. Ann Med 1989; 21: 299-301.
Paper accepted 28 June 1992
1375
J. J. De Cosse, S. BUlow*, K. Neale?, H. JarvinenS, T. Almg, R . Hultcrantzg, F. Moesgaard", C. Costello and the Leeds Castle Polyposis Group N e w York Hospital-Cornell Medical Center, N e w York, USA, *Hvidovre Hospital, Copenhagen, Denmark, t S t Mark's Hospital, London, UK, $Helsinki University Central Hospital, Finland and SKarolinska Hospital, Stockholm, Sweden Correspondence to: Professor J. J. De Cosse, Department of Surgery, New York
Hospital-Cornell Medical Center, 525 East 68th Street, New York, New York 10021, USA
Rectal cancer risk in patients treated for familial adenomatous polyposis Total colectomy with ileorectal anastomosis ( I R A ) in familial adenomatous polyposis ( F A P ) leaves patients at risk f o r rectal cancer. T o assess this risk, the rectal cancer incidence in 297patients with FAP undergoing I R A since 19.51 was determined in the population-based registers of Denmark, Finland and Sweden. A t the same time, detailed data on SO patients with FAP and invasive rectal cancer were obtained from I I international polyposis registries. The cumulative incidence of rectal cancer was 13.1 per cent at 2.5 years. The 5-year survival rate of patients with FAP developing rectal cancer was 71 per cent. Combining both studies, the risk of dying f r o m rectal cancer after I R A was 2.0 per cent at 1.5 years of follow-up. These results justijy I R A as primary treatment f o r most patients; restorative proctocolectomy is preferred f o r some subgroups. The high all-cause mortality rate observed in this relatively young population necessitates lifelong surveillance of patients with FAP.
Two treatment options are currently available for patients newly diagnosed with familial adenomatous polyposis ( F A P ) without cancer of the middle or distal rectum: total colectomy with ileorectal anastomosis (IRA) or restorative total proctocolectomy with ileoanal reservoir (RPC). Preservation of the rectum probably affords a superior quality of life but leaves the patient at risk of rectal cancer and its consequences. The extent of these risks is controversial. At the Mayo Clinic', 46 of 143 patients with F A P (32.2 per cent) developed cancer in the retained rectum during a median follow-up of 17.5 years. In a population-based study in Denmark', the cumulative risk of rectal cancer at 5 and 10 years after IRA was 3.5 and 13.3 per cent respectively. In 174 patients who had undergone IRA at St Mark's Hospital, London3, the cumulative risk of rectal cancer rose to 13 per cent at 25 years. At the Cleveland Clinic4, ten of 133 patients (7.5 per cent) at risk developed rectal cancer during a median follow-up of 5.5 years, ranging up to 37 years. The mortality risk of rectal cancer after IRA has been difficult to examine systematically. In a group of seven patients with rectal cancer, five survived long term5. At St Mark's Hospital, 12 patients had rectal cancer, resulting in four deaths6. Mortality risks from causes other than colorectal cancer have been p ~ b l i s h e d ~ * ' , ~ . In 1991 the Leeds Castle Polyposis Group initiated a two-part study of long-term treatment outcomes in patients with FAP. In one arm of the study, population-based registries in Denmark, Finland and Sweden assessed 297 patients who had undergone IRA over a 40-year period for long-term risk of rectal cancer. In the other arm, data on clinical findings and long-term results of invasive rectal cancer in 50 patients after IRA were collected from 11 polyposis registries in seven countries.
analysed for each patient; privacy was protected. Survival functions and risk of rectal cancer were determined by B M D P Statistical Software (BMDP, Los Angeles, California, USA) according to actuarial methods'. In the registry study, 11 member registries in the Leeds Castle Polyposis Group submitted data on all patients with FAP who had undergone IRA and were diagnosed before mid-1986 with invasive rectal cancer in the retained rectum. The registries applied uniform standards for basic clinical determinants, tumour staging and follow-up. Only patients for whom a minimum of 5 years offollow-up was available were included. Patients with in situ rectal cancer were excluded; privacy was protected. Data were entered into an Epi Info database (USD, Stone Mountain, Georgia, USA)". Analyses were performed on continuous data by Student's t test and on categorical data by the x2 test or, where appropriate, Fisher's exact test. Survival data were analysed with the Lifetest procedure (SAS Institute, Cary, North Carolina, USA)''. Five patients with rectal cancer and 135 who had undergone IRA were common to the Scandinavian and registry components.
Results
The Scandinavian study participants were 297 patients who had undergone total colectomy and IRA for FAP and were followed for at least 1 year before 1991. The patients were identified through the national registers of Denmark (88 patients), Finland (90) and Sweden (119). These patients were unselected over a 40-year span from the entire population of these nations; accession is believed to be virtually complete. Basic demographic and clinical data were collected and
Scandinavian study The median age of the 297 patients at IRA was 27 (range 10-77) years. Thirty-four patients were 350 years of age at the time of IRA. There were 145 male and 152 female patients. Of the 297 patients, 102 were probands with a median age of 36 (range 12-77) years and 195 were call-up patients with a median age of 24 (range 11-63) years. Thirty-five of the 297 patients ( 1 1.8 per cent) had colonic cancer, three (1.0 per cent) rectal cancer found within 1 year of IRA and regarded as a prevalent cancer, and 13 (4.4per cent) subsequently developed rectal cancer. Fourteen patients were lost to follow-up. For the 259 patients without colonic cancer, the median age ?It IRA was 25 (range 11-77) years and survival rates were 97.1 per cent at 5 years, 92.5 per cent at 10 years, 88.8 per cent at 15 years and 88.8 per cent at 20 years, by which time 42 patients remained at risk. For the 35 patients with colonic cancer, the median age at IRA was 44 (range 23-68) years and survival rates were 72.9 per cent at 5 years, 63.8 per cent at 10 years and 55.8 per cent at 15 years, seven patients remaining at risk by this time. During follow-up after IRA, the remaining rectal segment was removed for non-malignant causes (multiple large polyps,
1372
0007 1323/92/121372-04
Patients and methods
'(
1992 Butterworth Heinemann Ltd
Rectal cancer in familial adenomatous polyposis: J. J. De Cosse et al.
Four patients with rectal cancer were treated within 1 year of IRA and were regarded as having a prevalent cancer.
0
5
10
15
20
25
Time after I R A (years1
Figure 1 Cumulative probability of rectal cancer after ileorectal anastomosis ( I R A ) over 25 years of follow-up among patients in the Scandinavian study. Bars are 95 per cent conJdence intervals
severe dysplasia) in 34 patients. The cumulative probability of excision of the remaining rectum for causes other than carcinoma was 4.4 per cent at 5 years, 9.4 per cent at 10 years, 14.4 per cent at 15 years, 17.4 per cent at 20 years and 30.9 per cent at 25 years. The cumulative risk of rectal cancer (Figure 1) after IRA in the 294 patients without rectal cancer was 3.1 per cent at 5 years, 4.5 per cent at 10 years, 5.7 per cent at 15 years, 9.4 per cent at 20 years and 13.1 per cent at 25 years; 21 patients remained at risk. Of the 13 patients who developed rectal cancer, four died from metastatic disease. Sex and the presence of colonic cancer at the time of IRA did not affect the risk of subsequently developing rectal .carcinoma. Inclusion in risk calculation of the three patients diagnosed with rectal cancer within 1 year of IRA did not change these estimates substantially: the cumulative risk was 10.5 per cent at 20 years and 14.2 per cent at 25 years. The effect of age on risk for rectal cancer was examined. The cumulative probability of developing rectal cancer by age 50 years was 5.2 per cent (95 per cent confidence interval 1.2-9.2 per cent). By 60 years this figure was 14.1 per cent (95 per cent confidence interval 4.1-24.1 per cent). Registry study From the 11 polyposis registries, 50 patients with invasive rectal cancer in the retained rectum were identified, constituting 3.7 per cent of 1363 registered patients who were at risk after IRA performed before mid-1986. The median age of these 50 study patients at IRA was 33 (range 10-60) years. There were 27 male and 23 female patients. Eleven of the 50 had an invasive colonic cancer at the time of IRA. The median interval from IRA to the detection of invasive cancer was 11 (mean 13.4, range 0.1-36) years. The median interval was 13 years for women and 8 years for men ( P = 0.29).
Br. J. Surg., Vol. 79, No. 12, December 1992
Determinants at detection of rectal cancer. The median age of the 50 patients at the detection of rectal cancer was 46 (mean 46, range 18-67) years. Lower gastrointestinal symptoms were present in 19 patients, absent in 19 and not known in 12. The interval from the most recent endoscopy to cancer detection was < 6 months in 24 patients, 7-12 months in eight, 13-24 months in seven, > 2 years in four and unknown in seven. In 37 cases, the previous endoscopy had been performed by a member of the participating polyposis centre or by another physician regarded as reliable. Subsequent Dukes’ classification and survival were not affected by the presence of lower gastrointestinal symptoms and had little relationship to the interval from previous endoscopy. All four patients whose cancer was found > 2 years after their last endoscopy had a Dukes’ C lesion. The 50 patients were segregated into quartiles between 1952 and mid-1986 by the date of detection of rectal cancer and the quartiles compared for distribution of Dukes’ classification of the rectal tumour (unknown in three). More Dukes’ A and fewer Dukes’ C cancers were found in the most recent quartile (Figure 2 ) ( P = 0.02, Fisher’s exact test). The distance from the anal verge to the IRA was known in 40 patients; the median was 15.0cm and the mean(s.d.) 14.3(4.3) cm. In 38 patients the distance from the anal verge to the distal edge of the rectal cancer was known. The median was 8.5 cm and the mean(s.d.) 7.5(3-6) cm. The maximum diameter of the rectal cancer was known in 40 patients; the median was 3-9 cm and the mean(s.d.) 3.3( 1.7) cm. The distances from the central point of the rectal tumour to the anal verge and IRA were determined in the 33 patients for whom all measurements were known (Figure 3). Of the 33 tumours, 18 occurred in the most cephalad third of the remaining rectum (1’= 2.21, 1 d.f., P = 014). These 33 patients included ten of 11 who had colonic cancer at the time of IRA. Of the ten, eight had a rectal tumour in the upper third of the rectal segment ( P = 0.08, Fisher’s exact test). No tumours were reported in the most distal 4 cm of the rectum. Of ten patients with a tumour < 2 cm in diameter, only one died from rectal cancer. For larger tumours no correlation of size with survival was evident.
12
10
VI
Y
8
c
.-w
m
n c
r
n
6
0
0
z
4
2
0 1952-1976
1977-1979
1980-1983
1984-1986
Year
Figure 2 Dukes’ classification of invasive rectal cancer by year of diagnosis among patients in ihe registry study. 0 ,Dukes’ A ; 0 ,Dukes’ B and ‘D’(disseminated); R, Dukes’ C. Patients diagnosed in the latest guartile were more likely than those in earlier quartiles to have Dukes’ A rectal cancers ( P = 0.02, Fisher’s exact test)
1373
Rectal cancer in familial adenomatous polyposis: J . J . De Cosse et al
VI
c)
C
._ +. m
a L
0
z
I
Discussion
n 1t-l
2
2
0
4
6
12
10
8
14
16
20
18
D i s t a n c e o f t u m o u r f r o m anal v e r g e (cm)
Figure 3 Distance from the anal verge to midpoint of the rectal tumour in patients in the registry study. 0, Tumours within the cephulad third of retained rectum
Table 1 Dukes’ classification and mortality from rectal cancer by sex ~~
~
F
M ~
Dukes’ classification
No
A
8 5 11
B C
Disseminated Unknown
Total
-~
~~
~~
Deaths
3 0
0 2 I 3 0
21
12
No
Deaths
7
2 I
6
0
0
0
I
3
1
23
4
Treatment. Of the 50 patients, 42 underwent abdominoperineal resection and six low anterior resection; one of the latter was treated palliatively. Two patients with disseminated cancer did not undergo surgery and two died after operation. Results of treatment. Fifteen patients had a Dukes’ A cancer, 12 Dukes’ B and 17 Dukes’ C. Three had disseminated cancer, and classification was unknown in three (Table 1 ). The overall all-cause 5-year survival rate was 66 per cent and the 10-year rate 36 per cent. The 5-year rectal cancer survival rate was 68 per cent, 56 per cent in men and 83 per cent in women ( P = 0.04, log rank test) (Table 1 ). Of the I1 men with a Dukes’ C lesion, eight had full-thickness penetration of the primary tumour (Astler-Coller class C,). Of the six women with regional lymph node involvement, three had full-thickness penetration of the primary tumour. Excluding the three patients treated palliatively and the two who died after operation, the remaining 45 patients had an actual 5-year survival rate of 71 per cent. The 5-year survival rates were 77 per cent for the 15 patients with Dukes’ A cancer, 75 per cent for the 12 with Dukes’ B lesion and 62 per cent for the 17 with Dukes’ C cancer. Of the 17 patients who died during the first 5-year interval, 14 died from rectal cancer and three after operation. The actual 10-year survival rate in the 29 eligible patients was 55 per cent. During this second 5-year interval, six patients died: two from rectal cancer shortly after the 5-year mark and four from other causes. The 10-year rectal cancer survival rate for patients with Dukes’ C lesions was 50 per cent. Of the 13 patients surviving to 10 years, five have since died from conditions not related to colorectal cancer.
1374
This study has determined a cumulative risk for rectal cancer of 13.1 per cent (95 per cent confidence interval 3.3-22.9 per cent) at 25 years in the population-based Scandinavian data. This result is similar to the 13 per cent risk estimate at 25 years in a hospital-based registry3. If these estimates are combined with mortality rates from the registry study, the risk of dying from rectal cancer, given prior IRA with absence of colonic cancer, was 1.0 per cent (3.1 x 0.32) at 10 years and 2.0 per cent (4.5 x 0.45) at 15 years. This is a more favourable result than that of Bussey et aL3. These estimates do not consider the potential benefit of better surveillance and of nutritional modification and chemopreventionL2-’‘. The present observations suggest that retention of the rectum may have less life-threatening consequences than generally believed, and support a general policy of IRA in primary operative management of patients with FAP”. RPC should be the primary procedure for those with middle or upper rectal tumours or with large or confluent adenomas of the rectum, and for those who will not subscribe to regular follow-up. The Scandinavian study may provide a rationale for recommending RPC as the primary procedure in older patients. The benefit of this procedure must be weighed against an increased risk of morbidity at the primary operation and further deterioration during subsequent ageing. Of the 34 Scandinavian patients who underwent IRA at 2 5 0 years of age, only two subsequently developed rectal cancer. The course of F A P in older newly diagnosed patients may be relatively indolent, making a more extensive operation and attendant morbidity difficult to justify. -In the 50 registry patients, the distribution of rectal cancer by Dukes’ classification approximated that observed in patients with sporadic colorectal cancer. Survival by Dukes’ class, however, seemed superior, especially in women. This may be due to greater health awareness, especially among women, leading to more expeditious diagnosis and treatment and earlier substages of rectal cancer. The registry group included more men than women, and the former had shorter time intervals from IRA to rectal cancer, a greater likelihood of being diagnosed Dukes’ C and Astler-Coller C, (these differences were not significant) and a higher mortality rate. Evidence for the benefit of surveillance was mixed. Except for delay in follow-up endoscopy > 2 years, the frequency of endoscopy made no difference in Dukes’ classification at diagnosis or in survival, a finding previously notedI6. Patients diagnosed since 1978 were more likely to have a Dukes’ A tumour. This difference, if real, may be the product of improved health awareness in patients and alertness in physicians. These observations suggest that follow-up endoscopy should be performed at least every 6 months. As in sporadic colorectal cancerL7(except for rectal lesions < 2 c m in diameter) tumour size was not correlated with survival. After discounting the epithelium between the anal verge and the dentate line, the most caudad few centimetres of rectal mucosa appeared spared from rectal cancer. This may be due to the presence of transitional, not columnar, epithelium in the most distal 1-2 cm of rectal mucosaL8.This finding may support mucosal preservation during RPC. Because rectal cancer appeared to occur more frequently in the upper third of the retained rectum, especially in those with colonic cancer at the time of IRA, a low IRA 10-12 cm from the anal verge is recommendedL9. This study does not address the full range of causes of mortality in this comparatively young patient population. Perioperative risks should have decreased over the 40-year span of this study but other genetically based risks have become more p r ~ m i n e n t ~ .In~ .one ~ . registry of 174 patients with IRA, 35 died during follow-up, only seven of them from colorectal cancer3. Another study of 36 patients who died after IRA reported only three deaths from rectal cancer’. In both studies
Br. J. Surg., Vol. 79, No. 12, December 1992
Rectal cancer in familial adenomatous polyposis: J. J. De Cosse et al. the most frequent causes of death were desmoid disease a n d periampullary cancer; these occurred in only 4 a n d 1 per cent respectively of patients of the Danish register*'. During the 353 patient-years of follow-up represented by the registry study, 16 patients died from rectal cancer, 14 within the first 5 years. Twelve patients died from other causes: perioperative deaths (three), upper gastrointestinal cancer (four), suicide (two), desmoid disease (one), thermal t r au m a ( o n e) an d myocardial infarction (one). Nine of these deaths occurred after the 5-year mark. These findings highlight the need for lifelong comprehensive monitoring of these patients. The suicides, noted by others3,*, justify counselling for patients whose lives remain threatened by continuing risk of inherited disease.
5.
6.
7. 8. 9. 10.
11.
Acknowledgements The authors dedicate this paper to H. J. R. Bussey (1907-1991). The participants in the Leeds Castle Polyposis Group Study are: F. A. Macrae and C. Watts, Victoria, Australia; T. Berk and Z. Cohen, Toronto, Canada; S. Biilow and F. Moesgaard, Copenhagen, Denmark; H. Jarvinen, Helsinki, Finland; K. F. Neale and J. P. S. Thomson, London, UK; H. Itoh, Kitakyushu, Japan; S. Baba, Shizuoka, Japan; T. Muto and T. Iwama, Tokyo, Japan; T. Alm and R. Hultcrantz, Stockholm, Sweden; D. G . Jagelman and E. McGannon, Cleveland and Fort Lauderdale, USA; and J. J. De Cosse and C. Costello, New York, USA.
12. 13. 14. 15.
16.
References Bess MA, Adson MA, Elveback LR, Moertel CG. Rectal cancer following colectomy for polyposis. Arch Surg 1980; 115: 460-7. Bulow S. The risk of developing rectal cancer after colectomy and ileorectal anastomosis in Danish patients with polyposis coli. Dis Colon Rectum 1984; 27: 726-9. Bussey HJ, Eyers AA, Ritchie SM, Thomson JP. The rectum in adenomatous polyposis: the St Mark's policy. Br J Surg 1985; 12: S29-31. Sarre RG, Jagelman DG, Beck GJ et al. Colectomy with ileorectal anastomosis for familial adenomatous polyposis: the risk of rectal cancer. Surgery 1987; 101: 20-6.
Br. J. Surg., Vol. 79, No. 12, December 1992
17.
18. 19. 20.
Sener SF, Miller HH, De Cosse JJ. The spectrum of polyposis. Surg Gynecol Obstet 1984; 159: 525-32. Dozois RR (Moderator). Surgical aspects of familial adenomatous polyposis. Int J Colorectal Dis 1988; 3 : 1-16. Jagelman DG, De Cosse JJ, Bussey HJ, the Leeds Castle Polyposis Group. Upper gastrointestinal cancer in familial adenomatous polyposis. Lancet 1988; i: 1149-51. Arvanitis ML, Jagelman DG, Fazio VW, Lavery IC, McGannon E. Mortality in patients with familial adenomatous polyposis. Dis Colon Rectum 1990; 33: 639-42. Cutler SJ, Ederer F. Maximum utilization of the life table method in analyzing survival. J Chron Dis 1958; 8 : 699-712. Dean AD, Dean JA, Button AH, Dicker RC. Epi Info, Version 5: A Word Processing, Database, and Statistics Program for Epidemiology on Microcomputers. Stone Mountain, Georgia: USD, 1990. SAS Institute. SAS Technical Report P-179, Additional SASISTATProcedures,Release 6.03. Cary, North Carolina: SAS Institute, 1988. Bussey HJR, De Cosse JJ, Deschner EE et al. A randomized trial of ascorbic acid in polyposis coli. Cancer 1982; 50: 1434-9. De Cosse JJ, Miller HH, Lesser ML. Effect of wheat fiber on rectal polyps in patients with familial adenomatous polyposis. J Natl Cancer Inst 1989; 81 : 1290-1. LaBayle D, Fischer D, Vielh P e f a / . Sulindac causes regression of rectal polyps in familial adenomatous polyposis. Gastroenterology 1991; 101: 635-9. Madden MV, Neale KF, Nicholls RJ et al. Comparison of morbidity and function after colectomy with ileorectal anastomosis or restorative proctocolectomy for familial adenomatous polyposis. Br J Surg 1991; 18: 789-92. Moertel CG, Hill JR, Adson MA. Management of multiple polyposis of the large bowel. Cancer 1971; 28: 160-4. Wolmark N, Cruz I, Redmond CK, Fisher B, Fisher ER, contributing NSABP investigators. Tumor size and regional lymph node metastasis in colorectal cancer: a preliminary analysis from the NSABP clinical trials. Cancer 1983; 51: 1315-22. Fenger C. The anal transitional zone: location and extent. APMIS 1979; 81: 319-86. Jarvinen HJ. Time and type of prophylactic surgery for familial adenomatosis coli. Ann Surg 1985; 202: 93-7. Biilow S . Familial adenomatous polyposis. Ann Med 1989; 21: 299-301.
Paper accepted 28 June 1992
1375
