Interval Cancers in Screening with Fecal Occult Blood Test for Colorectal Cancer B . MBLLER JENSEN, 0. KRONBORG & C. FENGER Dept. of Surgical Gastroenterology and Dept. of Pathology, Odense University Hospital, Odense, Denmark
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
Jensen BM, Kronborg 0, Fenger C. Interval cancers in screening with fecal occult blood test for colorectal cancer. Scand J Gastroenterol 1992;27:77%782. Interval cancers, which are cancers diagnosed in spite of one or more negative screening tests, were studied in a randomized population with Hemoccult-I1 for colorectal cancer in 61,938 persons between 45 and 74 years old. Three biannual screenings were performed from 1985 to 1991, and 52% of all the cancers detected after doing at least one Hemoccult-I1 test were interval cancers (81 persons). These were more advanced than cancers diagnosed after a positive Hemoccult-I1 test, of larger size, less frequently of Dukes stage A , more often invading neighboring organs, and less often resectable for cure. They were located in the rectum more often than cancers diagnosed by screening and cancers in controls. Otherwise, interval cancers did not differ from cancers in controls or cancers in non-responders, and all characteristics suggested that no delay in diagnosis resulted from one or more negative HemoccultI1 tests, compared with controls. However, even if screening with Hemoccult-I1 demonstrates a reduction in mortality from colorectal cancer, the present high number of interval cancers makes it necessary to look for other methods of screening populations for colorectal cancer.
Key words: Colonoscopy; colorectal cancer; fecal occult blood test; interval cancers; population screening Ole Kronborg, M . D.,Dept. of Surgical Gastroenterology, Odense Universiiy Hospital, DK-5000 Odense C, Denmark
Negative screening tests may result in false security in several persons having a diagnosis of colorectal cancer before further screening (interval cancers), and it has been suggested that patients with interval cancer may have a mortality from cancer which may be even higher than that in persons subjected to no screening (1). The present study describes the interval cancer occurring in a randomized population study with biannual Hemoccult-I1 testing. Possible ways of reducing the number of interval cancers are discussed. METHODS AND PATIENTS
In 1985, 30,970 persons were allocated at random to screening with Hemoccult-I1 (H-11) every 2nd year and 30,968 to a control group. The samples were generated from two EDP files, The Central Person Register (CPR) and the Patient File of the County of Funen, which has 140,000 inhabitants between 45 and 75 years old. In the first screening 20,965 invited persons completed the test (2). At the second screening 18.779 of those initially screened participated (3). In the third screening 17,284 persons performed the H-I1 test (4). The H-TI test used has been described in detail earlier (5). Dietary restrictions were prescribed, but no rehydration of the test was used. Persons with a positive H-I1 were invited for full colonoscopy. An incomplete examination was followed by a double-contrast barium enema (DCBE) (2). The cancers were described in accordance with previously
published characteristics (2). Differences between proportions were evaluated with the chi-square test. RESULTS Eighty-one interval cancers were diagnosed from August 1985 to August 1991 in 35 men and 46 women. Thirty-four were diagnosed after the first screening, 26 after the second screening, and 21 after the third screening. The numbers of cancers detected by the three screenings were 37, 13, and 24, the detection rates of cancer per 1000 persons screened being 1.8, 0.7, and 1.4, respectively. The distributions of cancers within the 2-year intervals were rather similar after 1 and 2 screenings (16 cancers within the 1st year and 18 within the 2nd year after the first negative H-I1 test, the corresponding figures being 11 and 15 after two negative tests). The third interval of 2 years will not end until August 1992. Rectal cancers were more frequent among interval cases than among persons with screen-detected cancer ( p < 0.05) (Table I). Sigmoid cancer was less frequent among interval cases than among persons with a positive H-I1 test ( p < 0.001). The distribution of interval cancers also differed from that in controls, more rectal cancers ( p < 0.05) and fewer sigmoid cancers ( p < 0.05) being diagnosed in the former. The distribution in non-responders did not differ significantly from that in controls (Table I). The distribution of cancers on the basis of Dukes stage
780
B. M . Jensen et al.
Table I. Anatomic distribution of colorectal cancer
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
Rectum Sigmoid colon Descending colon Transverse colon Ascending colon Cecum Appendix More than one location Total
Interval cancers, no. of patients (%)
Screendetected cancers, no. of patients (96)
Cancers in nonresponders, no. of patients (%)
Cancers in controls, no. of patients (%)
43 (53) 10 (12) 3 9 5 6 2 3 81
25 (34) 27 (36) 3 8 4 4 0 3 74
51 (46) 26 (23) 1 11 8 10
98 (40) 60 (24) 13 28 22 22 1 2 246
I 3 111
Table 11. Dukes stage of colorectal cancer
Dukes A Dukes B Dukes C Distant spread Not classified Total
Interval cancers, no. of patients (%)
Screendetected cancers , no. of patients 1%)
Cancers in nonresponders, no. of patients (%)
Cancers in controls, no. of patients 1%)
13 (16) 21 (26) 23 (28) 16 (20) 8 (10) 81
37 (50) 21 (28) 10 (14) 4 (5) 2 (3) 74
12 (11) 35 (32) 17 (15) 44 (40) 3 (3) 111
22 (9) 83 (34) 71 (20) 56 (23) 14 (5) 246
was similar in interval cases and controls, but screen-detected cancers were more often Dukes stage A ( p < 0.00001). Nonresponders had distant spread more frequently than interval cases ( p < 0.01) (Table 11). Interval cancers tended to be larger than screen-detected cancers (Table 111), and cancers 1 2 c m in diameter were more seldom than among those detected by screening ( p = 0.01). Large cancers 2 5 cm in diameter were significantly more frequent among controls than among interval cases ( p = 0.05). However, the size could not be measured in 39 patients who had no removal of the cancer. No difference in tumor size was found between interval cancers and cancers in non-responders. Ninety-six per cent of the cancers were histologically classified. Among these, 97% were adenocarcinomas, 2% carcinoids, and 1% squamous carcinomas. Poorly differentiated cancers were not significantly more frequent among interval cases than among the three other groups, but the missing classification in 13 interval cancers compared with 1 screen-detected cancer makes it possible that there is a tendency towards more poorly differentiated tumors among interval cases (Table IV). Invasion of neighboring organs did not occur in screendetected cancers and was not more frequent among interval cases (5 of 81) than among controls (20 of 246) and nonresponders (7 of 111). Curative resections were performed in 62% of the interval
-
cases, a frequency similar to that among controls (60%) and among non-responders (50%) but lower than that among screen-detected cancers (93%) ( p < 0.00001). So far, the number of deaths from colorectal cancer have been 34 of 81 interval cases, 6 of 74 screen-detected cases, 116 of 246 controls, and 44 of 111non-responders, suggesting that the long-term prognosis for interval cases is similar to that of controls and non-responders but not as good as that for screen-detected cases. DISCUSSION The proportion of interval cancers amounted to 52% of all cancers detected in persons having at least one H-I1 test. This high percentage could easily compromise the benefit from detecting more favorable cancers in persons with positive tests. The high proportion of rectal cancers among interval cases may be explained by insufficiently hemolyzed blood during the short passage through the rectum. There was a slight tendency towards more right-sided cancers among interval cases; this has also been found in the English population trial (6) and has been explained by a total breakdown of hematin during the long passage of blood. The distribution of interval cancers on the basis of Dukes stage and tumor size was not worse than that among controls,
Interval Cancers in Screening
781
Table 111. Size of cancers
patients (%)
cancers, no. of patients (%)
Cancers in nonresponders, no. of patients (%)
patients (%)
9 (11) 11 14 40 (49) 7 81
21 (28) 12 14 26 (35) 1 74
8 (7) 10 13 65 (59) 15 111
12 (5) 19 30 153 (62) 32 246
Screendetected cancers,
Cancers
Cancers
Screen-
Largest diameter, cm 52
3 4
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
25 Not measured Total
Interval cancers, no. of
detected
Cancers in
controls, no. of
Table IV. Degree of differentiation in cancers Interval
Well diff. Moderately diff. Poorly diff. No classification Total
in non-
in
responders,
controls,
cancers, no. of patients (%)
no. of
no. of
no. of
patients (%)
patients (%)
patients (%)
7 (9) 38 (47) 23 (28) 13 81
12 (16) 45 (60) 16 (21) 1 74
8 (7) 60 (54) 20 (18) 23 111
12 (5) 133 (54) 65 (26) 36 246
and similar numbers could have curative resections (Tables
I1 and HI), suggesting that the former will have no delay in diagnosis because of the negative H-I1 test. The present death rates seem to confirm this, but it must be admitted that patients with interval cancers have been given a false security. The distribution of interval cancers also tended to be more favorable than among non-responders. The number of interval cancers was highest in the first interval, but the number within the third interval has not reached its final value, since only 1year has passed since the third negative H-I1 test in half the persons being screened. However, the present figures do not suggest that the number of interval cases will decrease with time. The distribution of interval cancers in the periods between screenings suggests that less than half could have been detected by screening if the intervals between screenings had been reduced from 2 to 1 year. However, acceptability might also have been reduced by such a policy. Poorly differentiated cancers are generally believed to be fast-growing tumors and therefore difficult to detect with any screening method, unless the intervals are very short. The present figures suggest that a poor differentiation may be commoner in interval cancers than in screen-detected cancers and that the proportion may be as high as 2844%, depending on the proportion present among the unclassified tumors (Table IV). The present number of interval cancers, which is even higher than in a similar English study (6), may be reduced by rehydrating the H-I1 test and thereby increasing the
sensitivity (7). However, this cannot be achieved without a concomitant decrease in specificity, making it necessary to subject a larger number of persons to unnecessary further examinations (8). Increasing the number of days with unhydrated H-I1 testing from 3 to 6 may also reduce the number of interval cancers (9), but the acceptability may decrease. The specificity for cancer detection may possibly be increased by use of immunochemical tests for occult blood, but these are expensive and not office procedures. Some of them need to be further evaluated in major controlled studies (10, ll), but it is uncertain whether they will result in any major increase in the cancer detection rate. A high sensitivity could probably be obtained by rehydrating the H-I1 test, and further testing of H-11-positive persons could be done with an immunochemical test, thereby avoiding the large fall in specificity and the large number of persons needing colonoscopy. The H-I1 test could also be combined with flexible sigmoidoscopy, being performed at intervals of several years. An initial 60-cm sigmoidoscopy reaching above the sigmoid colon would reduce the number of interval cancers considerably, since most of them are located in the rectum and sigmoid colon. However, flexible sigmoidoscopy is expensive, and only scant experience with this instrument has been gained within preventive medicine in many countries. A large controlled trial will be necessary to evaluate the possible effect on mortality from colorectal cancer, by adding sigmoidoscopy to fecal occult blood testing with H-11. Even if the ongoing randomized population trials with H-I1 (8) result in a reduction of mortality from colorectal cancer
782
B. M . Jensen et al,
by 10-20% and prevent morbidity from the disease in a substantial number of persons, we cannot be satisfied with such an achievement when better screening strategies may be available.
ACKNOWLEDGEMENTS
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
Support was given by The Danish Cancer Society, The County of Funen, Sygekassernes Helsefond, The Astrid Thaysen Foundation, the Danish Medical Research Council, and others. REFERENCES
1. Anderson I, Aspegren K, Janzon L, et al. Mammographic screening and mortality from breast cancer: the Malmo mammographic screening trial. Br Med J 1988;297:943-8. 2. Kronborg 0, Fenger C, S0ndergaard 0, Pedersen KM, Olsen J . Initial mass screening for colorectal cancer with fecal occult blood test. Scand J Gastroenterol 1987;22:677-86. 3. Kronborg 0, Fenger C, Olsen J, Bech K, Smdergaard 0. Repeated screening for colorectal cancer with fecal occult blood Received 27 January 1992 Accepted 20 March 1992
test. A prospective randomized study at Funen, Denmark. Scand J Gastroenterol 1989;24:59!9-606. 4. Kronborg 0, Fenger C, Worm J, et al. Causes of death during the first 5 years of a randomized trial of mass screening for colorectal cancer with fecal occult blood test. Scand J Gastroenterol 1991 ;27:47-52. 5. Klaaborg K, Madsen MS, Stindergaard 0, Kronborg 0. Participation in mass screening for colorectal cancer with fecal occult blood test. Scand J Gastroenterol 1986;21:1180-4. 6. Hardcastle JD, Chamberlain J, Sheffield J, et al. Randomised controlled trial of faecal occult blood screening for colorectal cancer. Results for first 107349 subjects. Lancet 1989;1:1160-4. 7. Kewenter J, Bjork S, Haglind E, Smith L, Svanvik J, Ahren C . Screening and rescreening for colorectal cancer. A controlled trial of fecal occult blood testing in 27,700 subjects. Cancer 1988;62:645-51. 8. Kronborg 0. Population screening for colorectal cancer, the goals and means. Ann Med 1991;23:373-9. 9. Thomas WM, Pye G, Hardcastle JD, Mangham CM. Faecal occult blood screening for colorectal neoplasia: a randomized trial of three days or six days of test. Br J Surg 1990;77:277-9. 10. Hakkinen I, Paasivuo R, Partanen P. Screening of colorectal tumours using an improved faecal occult blood test. Quantitative aspects. Gut 1988;29:1194-7. 11. St. John DJB. Faecal occult blood tests-a critical review. In: Hardcastle JD, editor. Screening for colorectal cancer. Englewood (NJ): Normed Verlag, 1990:54-68.
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
Jensen BM, Kronborg 0, Fenger C. Interval cancers in screening with fecal occult blood test for colorectal cancer. Scand J Gastroenterol 1992;27:77%782. Interval cancers, which are cancers diagnosed in spite of one or more negative screening tests, were studied in a randomized population with Hemoccult-I1 for colorectal cancer in 61,938 persons between 45 and 74 years old. Three biannual screenings were performed from 1985 to 1991, and 52% of all the cancers detected after doing at least one Hemoccult-I1 test were interval cancers (81 persons). These were more advanced than cancers diagnosed after a positive Hemoccult-I1 test, of larger size, less frequently of Dukes stage A , more often invading neighboring organs, and less often resectable for cure. They were located in the rectum more often than cancers diagnosed by screening and cancers in controls. Otherwise, interval cancers did not differ from cancers in controls or cancers in non-responders, and all characteristics suggested that no delay in diagnosis resulted from one or more negative HemoccultI1 tests, compared with controls. However, even if screening with Hemoccult-I1 demonstrates a reduction in mortality from colorectal cancer, the present high number of interval cancers makes it necessary to look for other methods of screening populations for colorectal cancer.
Key words: Colonoscopy; colorectal cancer; fecal occult blood test; interval cancers; population screening Ole Kronborg, M . D.,Dept. of Surgical Gastroenterology, Odense Universiiy Hospital, DK-5000 Odense C, Denmark
Negative screening tests may result in false security in several persons having a diagnosis of colorectal cancer before further screening (interval cancers), and it has been suggested that patients with interval cancer may have a mortality from cancer which may be even higher than that in persons subjected to no screening (1). The present study describes the interval cancer occurring in a randomized population study with biannual Hemoccult-I1 testing. Possible ways of reducing the number of interval cancers are discussed. METHODS AND PATIENTS
In 1985, 30,970 persons were allocated at random to screening with Hemoccult-I1 (H-11) every 2nd year and 30,968 to a control group. The samples were generated from two EDP files, The Central Person Register (CPR) and the Patient File of the County of Funen, which has 140,000 inhabitants between 45 and 75 years old. In the first screening 20,965 invited persons completed the test (2). At the second screening 18.779 of those initially screened participated (3). In the third screening 17,284 persons performed the H-I1 test (4). The H-TI test used has been described in detail earlier (5). Dietary restrictions were prescribed, but no rehydration of the test was used. Persons with a positive H-I1 were invited for full colonoscopy. An incomplete examination was followed by a double-contrast barium enema (DCBE) (2). The cancers were described in accordance with previously
published characteristics (2). Differences between proportions were evaluated with the chi-square test. RESULTS Eighty-one interval cancers were diagnosed from August 1985 to August 1991 in 35 men and 46 women. Thirty-four were diagnosed after the first screening, 26 after the second screening, and 21 after the third screening. The numbers of cancers detected by the three screenings were 37, 13, and 24, the detection rates of cancer per 1000 persons screened being 1.8, 0.7, and 1.4, respectively. The distributions of cancers within the 2-year intervals were rather similar after 1 and 2 screenings (16 cancers within the 1st year and 18 within the 2nd year after the first negative H-I1 test, the corresponding figures being 11 and 15 after two negative tests). The third interval of 2 years will not end until August 1992. Rectal cancers were more frequent among interval cases than among persons with screen-detected cancer ( p < 0.05) (Table I). Sigmoid cancer was less frequent among interval cases than among persons with a positive H-I1 test ( p < 0.001). The distribution of interval cancers also differed from that in controls, more rectal cancers ( p < 0.05) and fewer sigmoid cancers ( p < 0.05) being diagnosed in the former. The distribution in non-responders did not differ significantly from that in controls (Table I). The distribution of cancers on the basis of Dukes stage
780
B. M . Jensen et al.
Table I. Anatomic distribution of colorectal cancer
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
Rectum Sigmoid colon Descending colon Transverse colon Ascending colon Cecum Appendix More than one location Total
Interval cancers, no. of patients (%)
Screendetected cancers, no. of patients (96)
Cancers in nonresponders, no. of patients (%)
Cancers in controls, no. of patients (%)
43 (53) 10 (12) 3 9 5 6 2 3 81
25 (34) 27 (36) 3 8 4 4 0 3 74
51 (46) 26 (23) 1 11 8 10
98 (40) 60 (24) 13 28 22 22 1 2 246
I 3 111
Table 11. Dukes stage of colorectal cancer
Dukes A Dukes B Dukes C Distant spread Not classified Total
Interval cancers, no. of patients (%)
Screendetected cancers , no. of patients 1%)
Cancers in nonresponders, no. of patients (%)
Cancers in controls, no. of patients 1%)
13 (16) 21 (26) 23 (28) 16 (20) 8 (10) 81
37 (50) 21 (28) 10 (14) 4 (5) 2 (3) 74
12 (11) 35 (32) 17 (15) 44 (40) 3 (3) 111
22 (9) 83 (34) 71 (20) 56 (23) 14 (5) 246
was similar in interval cases and controls, but screen-detected cancers were more often Dukes stage A ( p < 0.00001). Nonresponders had distant spread more frequently than interval cases ( p < 0.01) (Table 11). Interval cancers tended to be larger than screen-detected cancers (Table 111), and cancers 1 2 c m in diameter were more seldom than among those detected by screening ( p = 0.01). Large cancers 2 5 cm in diameter were significantly more frequent among controls than among interval cases ( p = 0.05). However, the size could not be measured in 39 patients who had no removal of the cancer. No difference in tumor size was found between interval cancers and cancers in non-responders. Ninety-six per cent of the cancers were histologically classified. Among these, 97% were adenocarcinomas, 2% carcinoids, and 1% squamous carcinomas. Poorly differentiated cancers were not significantly more frequent among interval cases than among the three other groups, but the missing classification in 13 interval cancers compared with 1 screen-detected cancer makes it possible that there is a tendency towards more poorly differentiated tumors among interval cases (Table IV). Invasion of neighboring organs did not occur in screendetected cancers and was not more frequent among interval cases (5 of 81) than among controls (20 of 246) and nonresponders (7 of 111). Curative resections were performed in 62% of the interval
-
cases, a frequency similar to that among controls (60%) and among non-responders (50%) but lower than that among screen-detected cancers (93%) ( p < 0.00001). So far, the number of deaths from colorectal cancer have been 34 of 81 interval cases, 6 of 74 screen-detected cases, 116 of 246 controls, and 44 of 111non-responders, suggesting that the long-term prognosis for interval cases is similar to that of controls and non-responders but not as good as that for screen-detected cases. DISCUSSION The proportion of interval cancers amounted to 52% of all cancers detected in persons having at least one H-I1 test. This high percentage could easily compromise the benefit from detecting more favorable cancers in persons with positive tests. The high proportion of rectal cancers among interval cases may be explained by insufficiently hemolyzed blood during the short passage through the rectum. There was a slight tendency towards more right-sided cancers among interval cases; this has also been found in the English population trial (6) and has been explained by a total breakdown of hematin during the long passage of blood. The distribution of interval cancers on the basis of Dukes stage and tumor size was not worse than that among controls,
Interval Cancers in Screening
781
Table 111. Size of cancers
patients (%)
cancers, no. of patients (%)
Cancers in nonresponders, no. of patients (%)
patients (%)
9 (11) 11 14 40 (49) 7 81
21 (28) 12 14 26 (35) 1 74
8 (7) 10 13 65 (59) 15 111
12 (5) 19 30 153 (62) 32 246
Screendetected cancers,
Cancers
Cancers
Screen-
Largest diameter, cm 52
3 4
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
25 Not measured Total
Interval cancers, no. of
detected
Cancers in
controls, no. of
Table IV. Degree of differentiation in cancers Interval
Well diff. Moderately diff. Poorly diff. No classification Total
in non-
in
responders,
controls,
cancers, no. of patients (%)
no. of
no. of
no. of
patients (%)
patients (%)
patients (%)
7 (9) 38 (47) 23 (28) 13 81
12 (16) 45 (60) 16 (21) 1 74
8 (7) 60 (54) 20 (18) 23 111
12 (5) 133 (54) 65 (26) 36 246
and similar numbers could have curative resections (Tables
I1 and HI), suggesting that the former will have no delay in diagnosis because of the negative H-I1 test. The present death rates seem to confirm this, but it must be admitted that patients with interval cancers have been given a false security. The distribution of interval cancers also tended to be more favorable than among non-responders. The number of interval cancers was highest in the first interval, but the number within the third interval has not reached its final value, since only 1year has passed since the third negative H-I1 test in half the persons being screened. However, the present figures do not suggest that the number of interval cases will decrease with time. The distribution of interval cancers in the periods between screenings suggests that less than half could have been detected by screening if the intervals between screenings had been reduced from 2 to 1 year. However, acceptability might also have been reduced by such a policy. Poorly differentiated cancers are generally believed to be fast-growing tumors and therefore difficult to detect with any screening method, unless the intervals are very short. The present figures suggest that a poor differentiation may be commoner in interval cancers than in screen-detected cancers and that the proportion may be as high as 2844%, depending on the proportion present among the unclassified tumors (Table IV). The present number of interval cancers, which is even higher than in a similar English study (6), may be reduced by rehydrating the H-I1 test and thereby increasing the
sensitivity (7). However, this cannot be achieved without a concomitant decrease in specificity, making it necessary to subject a larger number of persons to unnecessary further examinations (8). Increasing the number of days with unhydrated H-I1 testing from 3 to 6 may also reduce the number of interval cancers (9), but the acceptability may decrease. The specificity for cancer detection may possibly be increased by use of immunochemical tests for occult blood, but these are expensive and not office procedures. Some of them need to be further evaluated in major controlled studies (10, ll), but it is uncertain whether they will result in any major increase in the cancer detection rate. A high sensitivity could probably be obtained by rehydrating the H-I1 test, and further testing of H-11-positive persons could be done with an immunochemical test, thereby avoiding the large fall in specificity and the large number of persons needing colonoscopy. The H-I1 test could also be combined with flexible sigmoidoscopy, being performed at intervals of several years. An initial 60-cm sigmoidoscopy reaching above the sigmoid colon would reduce the number of interval cancers considerably, since most of them are located in the rectum and sigmoid colon. However, flexible sigmoidoscopy is expensive, and only scant experience with this instrument has been gained within preventive medicine in many countries. A large controlled trial will be necessary to evaluate the possible effect on mortality from colorectal cancer, by adding sigmoidoscopy to fecal occult blood testing with H-11. Even if the ongoing randomized population trials with H-I1 (8) result in a reduction of mortality from colorectal cancer
782
B. M . Jensen et al,
by 10-20% and prevent morbidity from the disease in a substantial number of persons, we cannot be satisfied with such an achievement when better screening strategies may be available.
ACKNOWLEDGEMENTS
Scand J Gastroenterol Downloaded from informahealthcare.com by McMaster University on 10/31/14 For personal use only.
Support was given by The Danish Cancer Society, The County of Funen, Sygekassernes Helsefond, The Astrid Thaysen Foundation, the Danish Medical Research Council, and others. REFERENCES
1. Anderson I, Aspegren K, Janzon L, et al. Mammographic screening and mortality from breast cancer: the Malmo mammographic screening trial. Br Med J 1988;297:943-8. 2. Kronborg 0, Fenger C, S0ndergaard 0, Pedersen KM, Olsen J . Initial mass screening for colorectal cancer with fecal occult blood test. Scand J Gastroenterol 1987;22:677-86. 3. Kronborg 0, Fenger C, Olsen J, Bech K, Smdergaard 0. Repeated screening for colorectal cancer with fecal occult blood Received 27 January 1992 Accepted 20 March 1992
test. A prospective randomized study at Funen, Denmark. Scand J Gastroenterol 1989;24:59!9-606. 4. Kronborg 0, Fenger C, Worm J, et al. Causes of death during the first 5 years of a randomized trial of mass screening for colorectal cancer with fecal occult blood test. Scand J Gastroenterol 1991 ;27:47-52. 5. Klaaborg K, Madsen MS, Stindergaard 0, Kronborg 0. Participation in mass screening for colorectal cancer with fecal occult blood test. Scand J Gastroenterol 1986;21:1180-4. 6. Hardcastle JD, Chamberlain J, Sheffield J, et al. Randomised controlled trial of faecal occult blood screening for colorectal cancer. Results for first 107349 subjects. Lancet 1989;1:1160-4. 7. Kewenter J, Bjork S, Haglind E, Smith L, Svanvik J, Ahren C . Screening and rescreening for colorectal cancer. A controlled trial of fecal occult blood testing in 27,700 subjects. Cancer 1988;62:645-51. 8. Kronborg 0. Population screening for colorectal cancer, the goals and means. Ann Med 1991;23:373-9. 9. Thomas WM, Pye G, Hardcastle JD, Mangham CM. Faecal occult blood screening for colorectal neoplasia: a randomized trial of three days or six days of test. Br J Surg 1990;77:277-9. 10. Hakkinen I, Paasivuo R, Partanen P. Screening of colorectal tumours using an improved faecal occult blood test. Quantitative aspects. Gut 1988;29:1194-7. 11. St. John DJB. Faecal occult blood tests-a critical review. In: Hardcastle JD, editor. Screening for colorectal cancer. Englewood (NJ): Normed Verlag, 1990:54-68.
