- Email: [email protected]
Colorectal cancer screening: Clinical guidelines and rationale
October 1997
CORRESPONDENCE 1423
deaths. If the smoking cessation recommendation was repeated in connection with biannual screenings, the smoking quit rate might increase proportionally with the number of screenings.6 These extrapolations suggest that the incorporation of an antismoking activity in the screening program may avoid more deaths than the screening itself, e.g., 3–10 times the number of avoided colorectal carcinoma deaths resulting from the screening. From the point of public health, it is questionable that the recent version of clinical guidelines for colorectal cancer screening endorsed by the American Cancer Society did not mention the smoking issue.1 Neither was it mentioned in a 17-point list of topics for additional research.1 The substantial cost involved with wide-scale implementation of colorectal cancer screening should be seen in relation to other options for improved public health. For instance, a smoking cessation intervention program is likely to reduce the burden of avoidable deaths more than colorectal carcinoma screening.6 Accordingly, combined prevention and screening strategies regarding colorectal carcinoma seem preferable to a screening-only intervention for a wider effort to reduce the number of avoidable deaths in Western societies.5,6 FINN EDLER VON EYBEN
Department of Internal Medicine Herning Central Hospital CARSTEN THORDAL
Holstebro Central Hospital Ringkøbing County Denmark 1. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening. Clinical guidelines and rationale. Gastroenterology 1997;112: 594–642. 2. Marshall JR, Fay D, Lance P. Potential costs of flexible sigmoidoscopy–based colorectal cancer screening. Gastroenterology 1996; 111:1411–1417. 3. Kronborg O, Fenger C, Olsen J, et al. Randomised study of screening for colorectal cancer with faecal occult–blood test. Lancet 1996;348:1467–1471. 4. Mandel JS, Bond JH, Church TR, et al., for the Minnesota Colon Cancer Control Study. Reducing mortality from colorectal cancer by screening for fecal occult blood. NEJM 1993;328:1365–1371. 5. Ford BJ. Smokescreen. A guide to the personal risks and global effects of the cigarette habit. North Perth, Australia: Halcyon 1994. 6. Smoking Cessation Guideline Panel. Clinical Practice Guideline Number 18. Smoking cessation. Washington. U.S. Department of Health and Human Services. Public Health Service. Agency for Health Care Policy and Research 1996.
Dear Sir: In the background paper on colorectal cancer screening,1 a model is used to project the cumulative rate of detection of neoplasms by FOBT. The results of this model are driven by four features: the stage of a lesion, the ‘‘dwell time’’ of a lesion in each stage, the sensitivity of FOBT for each stage, and an assumption that the sensitivity of the FOBT for cancer is ‘‘evenly distributed’’ over all the lesions. Inherent in this latter assumption is that lesions not detected at one application of the test may, at a subsequent testing, be detected. For example, if at any one application of FOBT, the rate of detection (sensitivity) were, e.g., even as low as 30%, at subsequent testing, remaining new lesions would still be detected at a rate of 30% without difference between the two categories. Such an assumption that the sensitivity rate is spread over all neoplasms (i.e., that, biologically,
/ 5e21$$0039
09-08-97 11:50:54
gasas
all lesions bleed at one time or another but only 30% at any one time) has the effect in a model such that repeated testing causes the majority of neoplasms to be detected over time despite the low sensitivity. For example, a test with a sensitivity of 30% would, with repeated testing during a 4-year period, detect 76% (Å1 0 [0.7 1 0.7 1 0.7 1 0.7]) of all the lesions. Although this assumption of ‘‘equal chance to bleed’’ may be correct, we are concerned that it is incorrect and that some neoplasms do not bleed enough over their lifetime to ever be detected, at least in an early stage, by FOBT. It is clear that some neoplasms do bleed and do bleed enough to be detected by FOBT, but it is not at all clear how many do and how many may not.2,3 In any case, this assumption has an important effect on results, and it should be subjected to sensitivity analysis in the process of deriving recommendations and should be subject to validation in future studies. The committee concluded from a literature review that ‘‘[a]bout two thirds of cancers bleed in the course of a week.’’1 This conclusion is apparently derived from reference.4 However, what that article shows is average blood loss in cancer patients during a 8 – 10-day period; it does not give day-by-day blood loss. Therefore, the validity of the estimate of two-thirds has little objective support in that article, or in other literature of which we are aware. Additionally, other data suggest that at least some advanced cancers may bleed infrequently even when assessed repeatedly over time.2 It is clear that periodic FOBT screening can reduce colorectal cancer mortality. However, the reported reduction in mortality (on an intention-to-treat basis) is relatively small. The maximum mortality reduction of FOBT screening has yet to be determined and, in the published trials, is compromised by less than complete compliance with screening. An analysis of persons compliant with screening might provide important data about the ‘‘upper bound’’ of mortality reduction.5 The results of modeling, as a means to project mortality reduction, are critically dependent on the assumptions used, and those assumptions may be correct or incorrect. It is clear that how many cancers bleed and how often they bleed at detectable levels are critical features affecting the success of these and newer tests. DAVID F. RANSOHOFF, M.D.
Department of Medicine Chapel Hill, North Carolina CHRISTOPHER A. LANG, M.D.
St. Joseph Hospital Denver, Colorado GRAEME P. YOUNG, M.D.
Department of Medicine The Queen Elizabeth Hospital University of Adelaide Adelaide, Australia 1. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997;112: 594–642. 2. Ahlquist DA, McGill DB, Fleming JL, et al. Patterns of occult bleeding in asymptomatic colorectal cancers. Cancer 1989;63:1826– 1830. 3. Ahlquist DA, Wieand HA, Moertel CG, et al. Accuracy of fecal occult blood screening for colorectal neoplasia: a prospective study using Hemoccult and HemoQuant tests. JAMA 1993; 269: 1262 – 1267.
WBS-Gastro
1424 CORRESPONDENCE
GASTROENTEROLOGY Vol. 113, No. 4
4. Young GP, St John DJB. Selecting an occult blood test for use as a screening tool for large bowel cancer. In: Rozen P, Reich CB, Winawer SJ, eds. Large bowel cancer: policy, prevention, research and treatment. Front Gastrointest Res 1991; 18:135 – 156. 5. Ransohoff DF, Lang CA. Screening for colorectal cancer with the fecal occult blood test: a background paper. Ann Intern Med 1997; 126:811–822.
Reply. Drs. Ransohoff, Lang, and Young are essentially correct in their analysis of the model used. However, several factors affect the model with respect to FOBT screening. First, the test is assumed to be insensitive to lesions that have not become cancerous. Only sensitivity to cancerous lesions was modeled. Therefore, the first two features, the stage of a lesion and its dwell time are important only insofar as they cause lesions to be cancerous or not. The third feature, sensitivity of FOBT, was varied between 40% and 80%, resulting in substantial changes in specific clinical outcomes such as number of tests, follow-up test complications, cancer cases, and deaths. However, no change was found in the overall result—that screening using FOBT, given appropriate follow-up, has the potential to reduce cancer incidence and deaths. The sensitivity analysis performed showed that cancer incidence was reduced by nearly 40% in the worst case and by more than 60% in the best case. In addition, cancer deaths were reduced by 46% in the worst case and 62% in the best case. This may have been due to
the fourth feature cited—that sensitivity of FOBT for cancer is evenly distributed over all the lesions. However, it is more likely that the performance of FOBT as a screening methodology was substantially more affected by the total screening program, in which all positive FOBTs were followed up by a full colonoscopy. Because FOBT is somewhat notorious for its rate of false positives, this resulted in an average of approximately 2.8 colonoscopies per person in the original simulation cohort over a 35-year period. In terms of adjusting the model, one feasible alternative to the assumption that FOBT sensitivity for cancer is evenly distributed over all lesions is to assume that a fixed percentage of lesions will never bleed. While I suspect that this change will affect performance only slightly, I am in the process of revising the model to account for the possibility. In the meantime, absent clinical evidence, I suggest that the minimum value obtained from the simulation, slightly more than 46%, would be a good upper bound for the potential reduction in colorectal cancer mortality for a fully compliant population using annual FOBT with complete colonoscopy as a follow-up to any positive FOBT. As always, this estimate must be tempered with the knowledge that patients may or may not comply with a screening regimen and that actual mortality reduction will likely be significantly lower.
LAURA M. MILLER, Ph.D.
Image of the Month Answer Answer to Image of the Month Question (page 1062): Appendiceal and sigmoid diverticulosis shown in the Figure was encountered during single-contrast barium enema performed for evaluation of abdominal pain. Diverticula appear as barium-filled globular protrusions from the external surface of the appendix and the sigmoid colon. Barium enema studies are not indicated in the context of acute diverticulitis because of the risk of exacerbation or barium peritonitis, and it is preferable to allow the acute attack to subside. Right lower quadrant graded, compressed ultrasound and/or computerized axial tomography scanning are considered preferable imaging modalities. The patient’s right lower abdominal pain was attributed to appendiceal diverticulitis. He has since remained asymptomatic while consuming a high-fiber diet and antispasmodics. Diverticula of the appendix are rare, have a similar pathogenesis to diverticula elsewhere in the alimentary tract, and are usually found incidentally. Appendiceal diverticulitis may resemble acute appendicitis or right-sided (cecal or ascending colon) diverticulitis. In contrast to acute appendicitis, it affects older people more commonly, and its clinical presentation is not severe. However, the diagnosis is frequently made at urgent laparotomy. Rapid resolution with medical therapy may obviate the need for surgery. For submission instructions, please fax the Editorial Office at (301) 951-0757, or write to: Editorial Office, 7910 Woodmont Avenue, 7th Floor, Bethesda, Maryland 02814-3015.
/ 5e21$$0039
09-08-97 11:50:54
gasas
WBS-Gastro
CORRESPONDENCE 1423
deaths. If the smoking cessation recommendation was repeated in connection with biannual screenings, the smoking quit rate might increase proportionally with the number of screenings.6 These extrapolations suggest that the incorporation of an antismoking activity in the screening program may avoid more deaths than the screening itself, e.g., 3–10 times the number of avoided colorectal carcinoma deaths resulting from the screening. From the point of public health, it is questionable that the recent version of clinical guidelines for colorectal cancer screening endorsed by the American Cancer Society did not mention the smoking issue.1 Neither was it mentioned in a 17-point list of topics for additional research.1 The substantial cost involved with wide-scale implementation of colorectal cancer screening should be seen in relation to other options for improved public health. For instance, a smoking cessation intervention program is likely to reduce the burden of avoidable deaths more than colorectal carcinoma screening.6 Accordingly, combined prevention and screening strategies regarding colorectal carcinoma seem preferable to a screening-only intervention for a wider effort to reduce the number of avoidable deaths in Western societies.5,6 FINN EDLER VON EYBEN
Department of Internal Medicine Herning Central Hospital CARSTEN THORDAL
Holstebro Central Hospital Ringkøbing County Denmark 1. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening. Clinical guidelines and rationale. Gastroenterology 1997;112: 594–642. 2. Marshall JR, Fay D, Lance P. Potential costs of flexible sigmoidoscopy–based colorectal cancer screening. Gastroenterology 1996; 111:1411–1417. 3. Kronborg O, Fenger C, Olsen J, et al. Randomised study of screening for colorectal cancer with faecal occult–blood test. Lancet 1996;348:1467–1471. 4. Mandel JS, Bond JH, Church TR, et al., for the Minnesota Colon Cancer Control Study. Reducing mortality from colorectal cancer by screening for fecal occult blood. NEJM 1993;328:1365–1371. 5. Ford BJ. Smokescreen. A guide to the personal risks and global effects of the cigarette habit. North Perth, Australia: Halcyon 1994. 6. Smoking Cessation Guideline Panel. Clinical Practice Guideline Number 18. Smoking cessation. Washington. U.S. Department of Health and Human Services. Public Health Service. Agency for Health Care Policy and Research 1996.
Dear Sir: In the background paper on colorectal cancer screening,1 a model is used to project the cumulative rate of detection of neoplasms by FOBT. The results of this model are driven by four features: the stage of a lesion, the ‘‘dwell time’’ of a lesion in each stage, the sensitivity of FOBT for each stage, and an assumption that the sensitivity of the FOBT for cancer is ‘‘evenly distributed’’ over all the lesions. Inherent in this latter assumption is that lesions not detected at one application of the test may, at a subsequent testing, be detected. For example, if at any one application of FOBT, the rate of detection (sensitivity) were, e.g., even as low as 30%, at subsequent testing, remaining new lesions would still be detected at a rate of 30% without difference between the two categories. Such an assumption that the sensitivity rate is spread over all neoplasms (i.e., that, biologically,
/ 5e21$$0039
09-08-97 11:50:54
gasas
all lesions bleed at one time or another but only 30% at any one time) has the effect in a model such that repeated testing causes the majority of neoplasms to be detected over time despite the low sensitivity. For example, a test with a sensitivity of 30% would, with repeated testing during a 4-year period, detect 76% (Å1 0 [0.7 1 0.7 1 0.7 1 0.7]) of all the lesions. Although this assumption of ‘‘equal chance to bleed’’ may be correct, we are concerned that it is incorrect and that some neoplasms do not bleed enough over their lifetime to ever be detected, at least in an early stage, by FOBT. It is clear that some neoplasms do bleed and do bleed enough to be detected by FOBT, but it is not at all clear how many do and how many may not.2,3 In any case, this assumption has an important effect on results, and it should be subjected to sensitivity analysis in the process of deriving recommendations and should be subject to validation in future studies. The committee concluded from a literature review that ‘‘[a]bout two thirds of cancers bleed in the course of a week.’’1 This conclusion is apparently derived from reference.4 However, what that article shows is average blood loss in cancer patients during a 8 – 10-day period; it does not give day-by-day blood loss. Therefore, the validity of the estimate of two-thirds has little objective support in that article, or in other literature of which we are aware. Additionally, other data suggest that at least some advanced cancers may bleed infrequently even when assessed repeatedly over time.2 It is clear that periodic FOBT screening can reduce colorectal cancer mortality. However, the reported reduction in mortality (on an intention-to-treat basis) is relatively small. The maximum mortality reduction of FOBT screening has yet to be determined and, in the published trials, is compromised by less than complete compliance with screening. An analysis of persons compliant with screening might provide important data about the ‘‘upper bound’’ of mortality reduction.5 The results of modeling, as a means to project mortality reduction, are critically dependent on the assumptions used, and those assumptions may be correct or incorrect. It is clear that how many cancers bleed and how often they bleed at detectable levels are critical features affecting the success of these and newer tests. DAVID F. RANSOHOFF, M.D.
Department of Medicine Chapel Hill, North Carolina CHRISTOPHER A. LANG, M.D.
St. Joseph Hospital Denver, Colorado GRAEME P. YOUNG, M.D.
Department of Medicine The Queen Elizabeth Hospital University of Adelaide Adelaide, Australia 1. Winawer SJ, Fletcher RH, Miller L, et al. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology 1997;112: 594–642. 2. Ahlquist DA, McGill DB, Fleming JL, et al. Patterns of occult bleeding in asymptomatic colorectal cancers. Cancer 1989;63:1826– 1830. 3. Ahlquist DA, Wieand HA, Moertel CG, et al. Accuracy of fecal occult blood screening for colorectal neoplasia: a prospective study using Hemoccult and HemoQuant tests. JAMA 1993; 269: 1262 – 1267.
WBS-Gastro
1424 CORRESPONDENCE
GASTROENTEROLOGY Vol. 113, No. 4
4. Young GP, St John DJB. Selecting an occult blood test for use as a screening tool for large bowel cancer. In: Rozen P, Reich CB, Winawer SJ, eds. Large bowel cancer: policy, prevention, research and treatment. Front Gastrointest Res 1991; 18:135 – 156. 5. Ransohoff DF, Lang CA. Screening for colorectal cancer with the fecal occult blood test: a background paper. Ann Intern Med 1997; 126:811–822.
Reply. Drs. Ransohoff, Lang, and Young are essentially correct in their analysis of the model used. However, several factors affect the model with respect to FOBT screening. First, the test is assumed to be insensitive to lesions that have not become cancerous. Only sensitivity to cancerous lesions was modeled. Therefore, the first two features, the stage of a lesion and its dwell time are important only insofar as they cause lesions to be cancerous or not. The third feature, sensitivity of FOBT, was varied between 40% and 80%, resulting in substantial changes in specific clinical outcomes such as number of tests, follow-up test complications, cancer cases, and deaths. However, no change was found in the overall result—that screening using FOBT, given appropriate follow-up, has the potential to reduce cancer incidence and deaths. The sensitivity analysis performed showed that cancer incidence was reduced by nearly 40% in the worst case and by more than 60% in the best case. In addition, cancer deaths were reduced by 46% in the worst case and 62% in the best case. This may have been due to
the fourth feature cited—that sensitivity of FOBT for cancer is evenly distributed over all the lesions. However, it is more likely that the performance of FOBT as a screening methodology was substantially more affected by the total screening program, in which all positive FOBTs were followed up by a full colonoscopy. Because FOBT is somewhat notorious for its rate of false positives, this resulted in an average of approximately 2.8 colonoscopies per person in the original simulation cohort over a 35-year period. In terms of adjusting the model, one feasible alternative to the assumption that FOBT sensitivity for cancer is evenly distributed over all lesions is to assume that a fixed percentage of lesions will never bleed. While I suspect that this change will affect performance only slightly, I am in the process of revising the model to account for the possibility. In the meantime, absent clinical evidence, I suggest that the minimum value obtained from the simulation, slightly more than 46%, would be a good upper bound for the potential reduction in colorectal cancer mortality for a fully compliant population using annual FOBT with complete colonoscopy as a follow-up to any positive FOBT. As always, this estimate must be tempered with the knowledge that patients may or may not comply with a screening regimen and that actual mortality reduction will likely be significantly lower.
LAURA M. MILLER, Ph.D.
Image of the Month Answer Answer to Image of the Month Question (page 1062): Appendiceal and sigmoid diverticulosis shown in the Figure was encountered during single-contrast barium enema performed for evaluation of abdominal pain. Diverticula appear as barium-filled globular protrusions from the external surface of the appendix and the sigmoid colon. Barium enema studies are not indicated in the context of acute diverticulitis because of the risk of exacerbation or barium peritonitis, and it is preferable to allow the acute attack to subside. Right lower quadrant graded, compressed ultrasound and/or computerized axial tomography scanning are considered preferable imaging modalities. The patient’s right lower abdominal pain was attributed to appendiceal diverticulitis. He has since remained asymptomatic while consuming a high-fiber diet and antispasmodics. Diverticula of the appendix are rare, have a similar pathogenesis to diverticula elsewhere in the alimentary tract, and are usually found incidentally. Appendiceal diverticulitis may resemble acute appendicitis or right-sided (cecal or ascending colon) diverticulitis. In contrast to acute appendicitis, it affects older people more commonly, and its clinical presentation is not severe. However, the diagnosis is frequently made at urgent laparotomy. Rapid resolution with medical therapy may obviate the need for surgery. For submission instructions, please fax the Editorial Office at (301) 951-0757, or write to: Editorial Office, 7910 Woodmont Avenue, 7th Floor, Bethesda, Maryland 02814-3015.
/ 5e21$$0039
09-08-97 11:50:54
gasas
WBS-Gastro