Colorectal cancer screening guidelines clearly recommend. Comparing Risks and Benefits of Colorectal Cancer Screening in Elderly Patients

GASTROENTEROLOGY 2005;129:1163 1170 Comparing Risks and Benefits of Colorectal Cancer Screening in Elderly Patients CYNTHIA W. KO* and AMNON SONNENBERG *Department of Medicine, University of Washington, Seattle, Washington; and Portland VA Medical Center and Department of Medicine, Oregon Health and Science University, Portland, Oregon See editorial on page 1342. Background & Aims: In patients with limited life expectancy, the risks of colorectal cancer screening may outweigh the benefits. The aim of this study was to quantify risks and benefits of different screening strategies in elderly patients with varying life expectancies. Methods: We examined risks and benefits of screening in patients aged 70 94 years with differing health status using 3 strategies: annual fecal occult blood tests, flexible sigmoidoscopy every 5 years, or colonoscopy every 10 years. We compared the number needed to screen to prevent one cancer-related death and the number needed to encounter one screening-related complication for different strategies. Results: The potential benefit from screening varied widely with age, life expectancy, and screening modality. One cancer-related death would be prevented by screening 42 healthy men aged 70 74 years with colonoscopy, 178 healthy women aged 70 74 years with fecal occult blood tests, 431 women aged 75 79 years in poor health with colonoscopy, or 945 men aged 80 84 years in average health with fecal occult blood tests. Colonoscopy screening had the greatest benefit but the highest risk of complications. The potential for screening-related complications was greater than estimated benefit in some population subgroups aged 70 years and older. At all ages and life expectancies, the potential reduction in mortality from screening outweighed the risk of colonoscopy-related death. Conclusions: The potential benefits and risks of screening vary in elderly patients of different life expectancies. For any individual patient, the potential for harm from screening must be weighed against the likelihood of benefit, especially with shorter life expectancy. Colorectal cancer screening guidelines clearly recommend screening in average-risk patients beginning at age 50 years, based on evidence from randomized controlled trials of fecal occult blood testing and observational studies of lower endoscopy. 1 8 Patients in the published randomized controlled trials of fecal occult blood testing have been younger than 75 years, making it difficult to generalize the effectiveness of screening to older patients. 3 5 Observational studies of lower endoscopy have not explicitly examined the effectiveness of screening in different age groups, again making it difficult to apply results to older patients. 2,6,7 Such lack of evidence has led to uncertainty about the necessity for cancer screening in the elderly population, and current recommendations do not provide clear guidance about an appropriate upper limit of age for colorectal cancer screening. 9 12 The potential benefits of colorectal cancer screening may decrease with age and increasing comorbidity, as competing risks for mortality become more prominent. In addition, the potential for harm related to screening, such as colonoscopyrelated complications, may increase in older patients and those with significant comorbidity. 13 Therefore, in some patient subgroups, the potential harm related to screening may outweigh any potential benefit in reducing cancer-specific mortality. Few studies have quantitatively addressed the benefits and harms of colorectal cancer screening, especially in older individuals or those with significant comorbidity, 14 and prior cost-effectiveness models of colorectal cancer screening have generally focused on younger individuals. 15 20 However, cost-effectiveness of screening is maintained, even in older patients. 18 Cost-effectiveness models are useful in formulating health policy but often do not provide information useful in decision making for individual patients. The aim of this study was to examine explicitly the probability of colorectal cancer screening related benefits and harms in older patients with varying health Abbreviations used in this paper: NNC, number needed to encounter one screening-related complication; NNS, number needed to screen to prevent one cancer-related death; SEER, Surveillance, Epidemiology, and End Results. 2005 by the American Gastroenterological Association 0016-5085/05/$30.00 doi:10.1053/j.gastro.2005.07.027

1164 KO AND SONNENBERG GASTROENTEROLOGY Vol. 129, No. 4 status to provide important information to aid screening decisions for individual patients. Materials and Methods The framework for our study was similar to that of Walter and Covinsky. 14 We developed a model to estimate the risk of dying of colorectal cancer, the potential reduction in mortality with colorectal cancer screening, the number of colonoscopies needed, and the potential for colonoscopy-related complications in elderly patients at average risk for colorectal cancer. We quantified these potential benefits in patients aged 70 94 years, with varying life expectancies based on underlying health status. We examined 3 different screening strategies: annual fecal occult blood tests, flexible sigmoidoscopy every 5 years, and colonoscopy every 10 years. General Assumptions We assumed that fecal occult blood tests, flexible sigmoidoscopy, and colonoscopy reduce colorectal cancer mortality by 18%, 40%, and 75%, respectively. 2 6 Because randomized controlled trials of fecal occult blood tests did not show a significant reduction in colorectal cancer mortality until 5 years after screening began, 3 5 we also assumed that cancer-specific mortality would not decrease significantly for at least 5 years of screening. Because of this, we assumed that screening would be unlikely to benefit patients whose estimated life expectancy was 5 years. We assumed that screening-related harms would occur solely because of colonoscopyrelated complications, because the risk of complications after screening flexible sigmoidoscopy is low. 21 The risk of a serious colonoscopy-related complication, including postpolypectomy bleeding or perforation, was estimated to be 0.3%, and the risk of death was estimated to be 0.01%. 13,22 25 We did not vary these estimates based on age or comorbidity. We estimated the rate of a positive fecal occult blood test result at 3%. The prevalence of an adenomatous polyp in these age groups was estimated at 30%. 26 28 We assumed that all patients with a positive fecal occult blood test result would subsequently undergo colonoscopy. We also assumed that all patients with an adenomatous polyp would undergo surveillance colonoscopy every 5 years and not undergo future screening with fecal occult blood tests or flexible sigmoidoscopy. Last, our model was based on full compliance with all recommended screening tests, because this would maximize the potential benefits of screening. Because full compliance with any recommended strategy would also increase the total number of colonoscopies required, this assumption would also lead to the largest estimates of potential screening-related harm. Estimates of Life Expectancy and Cancer-Specific Mortality We obtained age- and sex-specific estimates of median life expectancy and the distribution of life expectancy from US life tables. 29 Patients with poorer underlying health status would be expected to have lower-than-average life expectancy and were estimated to have life expectancy in the lowest quartile for their age- and sex-specific group. Patients with better underlying health status would be expected to have higher-than-average life expectancy and were estimated to have life expectancy in the highest quartile for their age- and sex-specific group. We estimated age- and sex-specific colorectal cancer related mortality rates using data from the Surveillance, Epidemiology, and End Results (SEER) program and the method of Walter and Covinsky. 14,30 Because colorectal cancer screening rates in older patients are fairly low, we assumed that the mortality rates in the SEER program approximate cancer mortality rates in the absence of prior screening. Estimates of Screening Benefit and Harm We assumed a 5-year time lag between initiation of screening and reduction in mortality. Thus, we assumed that individuals with an estimated life expectancy 5 years would not undergo screening. To estimate the cancer-specific reduction in mortality from screening, we applied the screening strategy specific reduction in mortality to the estimated colorectal cancer mortality beginning 5 years after initiation of screening. We calculated the estimated cancer-specific mortality for the first 5 years using age- and sex-specific rates from SEER. To estimate future cancer-specific mortality, we again applied cancer-specific mortality rates to the remainder of the patient s life span, reduced by the assumed mortality benefit for the various screening strategies. The sum of these 2 mortality estimates equaled the estimated colorectal cancer specific mortality rates for various age-, sex-, and health status specific groups. The absolute risk reduction for cancer-specific mortality was calculated by subtracting the risk of colorectal cancer death with screening from the risk of colorectal cancer death without screening. The number needed to screen to prevent one cancer-related death (NNS) was calculated as 1/absolute risk reduction. We also estimated the number of colonoscopies needed for the different screening strategies in the different age-, sex-, and health status specific groups and based on the previously described assumptions about diagnostic and surveillance colonoscopies. We multiplied the number of colonoscopies in each subgroup by the estimated complication rate to determine the absolute risk of colonoscopy complications. The number needed to encounter one screening-related complication (NNC) was calculated as 1/absolute risk of colonoscopy complications. Sensitivity Analysis Because data are lacking regarding the effectiveness of screening and the risk of colonoscopy complications specifically in elderly patients, we performed a sensitivity analysis to examine the effects of this uncertainty. In these analyses, we varied the reduction in mortality for fecal occult blood test screening from 15% to 21%, for flexible sigmoidoscopy from 30% to 50%, and for colonoscopy from 60% to 90%. The risk of colonoscopy complications was varied from 0.1% to 0.5%.

October 2005 COLORECTAL CANCER SCREENING IN THE ELDERLY 1165 This, for example, would encompass the risk of a variety of complications, including the risk of postpolypectomy bleeding and perforation. Results The estimated life expectancy and risk of colorectal cancer death for men and women aged 70 years and older are shown in Tables 1 and 2. Similar estimates for men and women aged 50 54 years are included for comparison. Although colorectal cancer incidence rates increase with age, the cumulative lifetime risk of colorectal cancer death actually decreases in men and women with advancing age and increasing comorbidity. For example, the risk of colorectal cancer death is 1.8% in men aged 80 84 years with average health status, compared with 2.3% for similar men aged 50 54 years. In addition, the cumulative lifetime risk of colorectal cancer death is generally higher in men than in women, even after accounting for differences in life expectancy. We assumed that colorectal cancer screening is unlikely to benefit individuals whose life expectancy is 5 years. Based on this assumption and published estimates of life expectancy, the following groups would not benefit from colorectal cancer screening: men aged 75 84 years with poor health, men aged 85 years and older with average or poor health, women aged 80 89 years with poor health, and women aged 90 years and older with average or poor health (Tables 1 and 2). The NNS varied with the screening regimen chosen, age, and risk of colorectal cancer death (Tables 1 and 2). In the NNS analysis, small numbers are generally preferred to large numbers, indicating a more effective screening regimen. In men aged 70 74 years in good health, the NNS varied from 42 with colonoscopy screening to 177 with fecal occult blood test screening. In similar women, the NNS varied from 43 with colonoscopy screening to 178 with fecal occult blood test screening. Even within a given sex and age group, the NNS varied widely depending on underlying health status and the screening modality chosen. For example, in women aged 70 74 years, the NNS varied from 43 (colonoscopy screening for women in good health) to 1046 (fecal occult blood test screening for women in poor health). Because the cumulative lifetime risk of colorectal cancer death decreased with age, the NNS increased with age in both men and women. For men aged 90 years and older, screening would likely be beneficial only for individuals in good health. Even in these individuals, the NNS is large, ranging from 482 (colonoscopy screening) to 2008 (fecal occult blood test screening). Similar results were found in women aged 90 years and older. Given the increased effectiveness of colonoscopy screen- Table 1. Expected Colorectal Cancer Mortality and NNS in Men Age group (y) Life expectancy (y) a cancer death (%) Risk of colorectal Fecal occult blood test NNS b Flexible sigmoidoscopy Colonoscopy 50 54 36 4.1 138 (116 166) 62 (50 83) 33 (28 42) 28.5 2.3 255 (219 306) 115 (92 153) 61 (51 77) 19.6 1.0 629 (539 754) 283 (226 377) 151 (126 189) 70 74 18 3.8 177 (152 212) 80 (75 106) 42 (35 53) 12.4 2.1 380 (326 456) 171 (137 228) 91 (76 114) 6.7 0.9 1877 (1609 2252) 845 (676 1126) 450 (375 563) 75 79 14.2 3.6 207 (177 248) 93 (75 124) 50 (41 62) 9.3 1.9 525 (450 630) 236 (189 315) 126 (105 157) 4.9 0.9 80 84 10.8 3.2 277 (237 332) 125 (100 166) 66 (55 83) 6.7 1.8 945 (810 1134) 425 (340 567) 227 (189 283) 3.3 0.8 85 89 7.9 2.7 554 (475 665) 249 (199 332) 133 (111 166) 4.7 1.6 2.2 0.8 90 or older 5.8 2.0 2008 (1721 2409) 903 (723 1205) 482 (402 602) 3.2 1.1 1.5 0.5 a Life expectancy was estimated from US life tables 29 and represent the 75th, 50th, and 25th percentiles for given age- and sex-specific groups. Patients with good health would have greater anticipated life expectancy (75th percentile), and patients with poor health would have lower anticipated life expectancy (25th percentile). b The numbers in parentheses represent sensitivity analyses for the estimated range of efficacy for fecal occult blood testing (15% 21%), flexible sigmoidoscopy (30% 50%), and colonoscopy (60% 90%), respectively. The higher NNS is estimated from the lower estimate of efficacy, and the lower NNS is estimated from the higher estimate of efficacy.

1166 KO AND SONNENBERG GASTROENTEROLOGY Vol. 129, No. 4 Table 2. Expected Colorectal Cancer Mortality and NNS in Women Age group (y) Life expectancy (y) a cancer death (%) Risk of colorectal Fecal occult blood test NNS b Flexible sigmoidoscopy Colonoscopy 50 54 40 3.8 147 (126 176) 66 (53 88) 35 (29 44) 33 2.2 263 (226 316) 119 (95 158) 63 (53 79) 24.5 1.0 577 (495 693) 260 (208 346) 139 (115 173) 70 74 21.3 3.5 178 (153 214) 80 (64 107) 43 (36 53) 15.7 2.0 340 (291 408) 153 (122 204) 82 (68 102) 9.5 0.9 1046 (897 1255) 471 (377 628) 251 (209 314) 75 79 17 3.3 204 (175 245) 92 (74 123) 49 (41 61) 11.9 2.0 408 (349 489) 183 (147 245) 98 (82 122) 6.8 0.9 1797 (1540 2156) 808 (647 1078) 431 (359 539) 80 84 13 3.0 262 (224 314) 118 (94 157) 63 (52 78) 8.6 1.8 581 (498 698) 262 (209 349) 140 (116 174) 4.6 0.8 85 89 9.6 2.5 455 (390 456) 205 (164 273) 109 (91 137) 5.9 1.6 2326 (1994 2791) 1047 (837 1396) 558 (465 698) 2.9 0.8 90 or older 6.8 1.8 1163 (997 1396) 523 (419 698) 279 (23 349) 3.9 1.0 1.5 0.4 a Life expectancy was estimated from US life tables 29 and represent the 75th, 50th, and 25th percentiles for given age- and sex-specific groups. Patients with good health would have greater anticipated life expectancy (75th percentile), and patients with poor health have lower anticipated life expectancy (25th percentile). b The numbers in parentheses represent sensitivity analyses for the estimated range of efficacy for fecal occult blood testing (15% 21%), flexible sigmoidoscopy (30% 50%), and colonoscopy (60% 90%), respectively. The higher NNS is estimated from the lower estimate of efficacy, and the lower NNS is estimated from the higher estimate of efficacy. ing compared with other modalities, the NNS with colonoscopy was consistently lower than with either flexible sigmoidoscopy or fecal occult blood tests. The results of our sensitivity analysis for the effectiveness of screening are also presented in Tables 1 and 2. In these analyses, a lower NNS would represent the higher estimated screening efficacy. The results of these sensitivity analyses generally parallel the results with our base-case assumptions and show wide variations in the NNS depending on the chosen estimate of screening effectiveness. We next estimated the required number of colonoscopies during a patient s lifetime for different screening regimens (Table 3). The required number of colonoscopies was generally highest with colonoscopy-based screening, intermediate with flexible sigmoidoscopy based screening, and lowest with fecal occult blood test based screening. The expected number of colonoscopies decreased with increasing age and comorbidity. For instance, with fecal occult blood test screening, women in good health aged 70 74 years would be expected to undergo an average of 0.8 colonoscopies during their lifetime, while women in poor health aged 70 74 years would undergo an average of 0.4 colonoscopies during their lifetime. Women in good health aged 85 89 years would undergo an average of 0.4 colonoscopies in their lifetime with fecal occult blood test screening. Last, we estimated the risk of screening-based complications for different combinations of sex, age, life expectancy, and screening regimen (Tables 4 and 5). In the NNC analysis, large numbers are preferred to small numbers, indicating a potentially less harmful screening regimen. For these estimates, we assumed that all harms from screening would come from colonoscopy-related complications, because the risk of complications with screening flexible sigmoidoscopy is extremely small and there is minimal risk associated with fecal occult blood testing itself. Because the required number of colonoscopies is lowest with fecal occult blood test based screening, the risk of screening-related complications is lowest and the NNC highest. The risk of screeningrelated complications is highest and the NNC lowest with colonoscopy-based regimens. Because the required number of colonoscopies decreases with age and comorbidity, the likelihood of screening-related complications decreases and the NNC value increases. The results of our sensitivity analyses for risk of complications are also presented in Tables 4 and 5. In these analyses, a lower NNC represents the higher estimate of complication risk. These analyses parallel our results with our basecase assumption and show that the risk of screeningrelated complications can be substantial in some population subgroups.

October 2005 COLORECTAL CANCER SCREENING IN THE ELDERLY 1167 Table 3. Estimated Number of Colonoscopies Needed Over Remaining Life Span With Different Screening Strategies Men Women Colonoscopies Colonoscopies Age group (y) Life expectancy (y) FOBT FS Colon Life expectancy (y) FOBT FS Colon 50 54 36 0.9 2.1 4.9 40 1.0 2.4 5.2 28.5 0.9 1.5 3.6 33 0.9 1.8 3.9 19.6 0.7 0.9 2.3 24.5 0.8 1.2 2.6 70 74 18 0.7 0.9 2.3 21.3 0.8 1.2 2.6 12.4 0.5 0.6 1.3 15.7 0.6 0.9 2.3 6.7 0.2 0.3 1.0 9.5 0.4 0.3 1.0 75 79 14.2 0.6 0.6 1.3 17 0.7 0.9 2.3 9.3 0.4 0.3 1.0 11.9 0.5 0.6 1.3 4.9 6.8 0.2 0.3 1.0 80 84 10.8 0.4 0.6 1.3 13 0.6 0.6 1.3 6.7 0.2 0.3 1.0 8.6 0.3 0.3 1.0 3.3 4.6 85 89 7.9 0.2 0.3 1.0 9.6 0.4 0.3 1.0 4.7 5.9 0.1 0.3 1.0 2.2 2.9 90 94 5.8 0.1 0.3 1.0 6.8 0.2 0.3 1.0 3.2 3.9 1.5 1.5 FOBT, fecal occult blood testing annually; FS, flexible sigmoidoscopy every 5 years; colon, colonoscopy every 10 years. Comparing the likelihood of benefit (NNS) with the likelihood of a screening-related complication (NNC) for different screening regimens will allow some estimate of the absolute benefits and risks of screening. For most age, sex, and life expectancy groups, the potential for avoiding colorectal cancer death outweighed the risk of screening-related complications. However, the potential for a screening-related complication was higher than the screening benefit for certain subgroups (shown in bold in Tables 4 and 5). For example, 1046 women aged 70 74 years in poor health would have to be screened with fecal occult blood testing to avoid one colorectal cancer death, while 1 in 909 would encounter screening-related harm. Similarly, 450 men aged 70 74 years in poor health Table 4. NNC in Men Age group (y) Life expectancy (y) FOBT FS Colon 50 54 36 352 (211 1057) 159 (95 476) 68 (41 204) 28.5 375 (225 1126) 222 (133 667) 93 (87 435) 19.6 447 (268 1340) 370 (222 1111) 145 (87 435) 70 74 18 462 (277 1385) 370 (222 1111) 145 (87 435) 12.4 643 (386 1928) 556 (333 1667) 256 (154 769) 6.7 1996 (1197 5987) 1111 (667 3333) 333 (200 1000) 75 79 14.2 556 (334 1669) 556 (333 1667) 256 (154 769) 9.3 909 (545 2727) 1111 (667 3333) 333 (200 1000) 4.9 80 84 10.8 790 (474 2369) 556 (333 1667) 256 (154 769) 6.7 1996 (1197 5987) 1111 (667 3333) 333 (200 1000) 3.3 85 89 7.9 1390 (834 4171) 1111 (667 3333) 333 (200 1000) 4.7 2.2 90 94 5.8 3817 (2290 1145) 1111 (667 3333) 333 (200 1000) 3.2 1.5 NOTE. Numbers in parentheses represent results for sensitivity analysis of complication rates. The lower estimates of NNC represent higher estimated complication rates, while the higher estimates of NNC represent lower estimated complication rates. Estimates that exceed the number of cases needed to avoid one cancer-related death are shown in bold. FOBT, fecal occult blood testing annually; FS, flexible sigmoidoscopy every 5 years; colon, colonoscopy every 10 years. NNC

1168 KO AND SONNENBERG GASTROENTEROLOGY Vol. 129, No. 4 Table 5. NNC in Women Age group (y) Life expectancy (y) FOBT FS Colon 50 54 40 346 (209 1039) 139 (83 417) 64 (28 192) 33 359 (215 1076) 185 (111 556) 85 (51 256) 24.5 397 (238 1192) 278 (167 833) 128 (77 385) 70 74 21.3 423 (254 1268) 278 (167 833) 128 (77 385) 15.7 526 (315 1577) 370 (222 1111) 145 (87 435) 9.5 909 (545 2727) 1111 (667 3333) 333 (200 1000) 75 79 17 480 (288 1439) 370 (222 1111) 145 (87 435) 11.9 705 (423 2116) 556 (333 1667) 256 (154 769) 6.8 1996 (1197 5987) 1111 (667 3333) 333 (200 1000) 80 84 13 595 (357 1784) 556 (333 1667) 256 (154 769) 8.6 1089 (653 3266) 1111 (667 3333) 333 (200 1000) 4.6 85 89 9.6 909 (545 2727) 1111 (667 3333) 333 (200 1000) 5.9 3817 (2290 11451) 1111 (667 3333) 333 (200 1000) 2.9 90 94 6.8 1996 (1197 5987) 1111 (667 3333) 333 (200 1000) 3.9 1.5 NOTE. Numbers in parentheses represent results for sensitivity analysis of complication rates. The lower estimates of NNC represent higher estimated complication rates, while the higher estimates of NNC represent lower estimated complication rates. Estimates that exceed the number of cases needed to prevent one cancer-related death are shown in bold. FOBT, fecal occult blood testing annually; FS, flexible sigmoidoscopy every 5 years; Colon, colonoscopy every 10 years. NNC would have to be screened with colonoscopy to avoid one colorectal cancer death, while 1 in 333 would encounter screening-related harm with this screening strategy. For flexible sigmoidoscopy based screening, there was not an age-, sex-, and comorbidity-related subgroup in which the risk of screening-related complications outweighed the potential mortality benefit. For colonoscopy-based screening, the potential for screening-related complications outweighed the potential benefit in the following groups: women aged 75 79 years in poor health, women aged 85 89 years in average health, men aged 70 74 years in poor health, and men aged 90 94 years in good health. At all ages and life expectancies, the expected reduction in mortality from screening was 0.01%, the estimated risk of screening-related death. Discussion Screening for colorectal cancer has become an accepted medical practice, and current guidelines recommend that average-risk individuals begin screening at age 50 years. However, considerable uncertainty exists for both patients and providers in determining whether to offer screening to older individuals whose life expectancy is relatively limited. In this study, we evaluated the potential benefits and risks of colorectal cancer screening in elderly individuals. Our results show that there is considerable variation in the estimated absolute benefit from colorectal cancer screening, depending on age, life expectancy, and screening strategy chosen. The potential for reduction in cancer-related mortality generally decreased with increasing age and comorbidity. We also considered the potential harms from screening, which are primarily due to colonoscopy complications. The risk of colonoscopy complications appears to increase with patient age and comorbidity, 13 situations in which the potential benefits of screening also decrease. Therefore, we found that the potential for an adverse event related to screening may outweigh the potential reduction in colorectal cancer mortality in certain individuals with advanced age or significant comorbidity. This direct, quantitative comparison of the relative risks and benefits of screening may be helpful in discussing screening options with such individuals. Prior cost-effectiveness studies of colorectal cancer screening have not explicitly quantified the expected benefits and risks in elderly individuals, and most models have focused on younger individuals. Screening may have the greatest medical benefit, as measured by life years saved, in younger individuals (aged 60 65 years), whereas the greatest economic benefit is seen in older individuals (aged 70 years and older). 18,31 However, costeffectiveness is often not a prominent factor in patients decision making about medical procedures. In contrast, patients knowledge, values, and preferences about disease risk and medical care are often extremely important. For example, Salkeld et al found that a majority of study

October 2005 COLORECTAL CANCER SCREENING IN THE ELDERLY 1169 participants interviewed would trade off potential harms and benefits from colorectal cancer screening in decision making about screening. 32 Our results can provide quantitative estimates to be integrated with patients values and preferences to help in decisions about screening. The tables provided in the present report could be used to estimate the cost-effectiveness of screening among different age and sex groups. For instance, to save one life among men aged 70 74 years through colonoscopy, 42 subjects need to be screened by 2.3 colonoscopies. Under the assumption that death would occur, on average, during the middle of the remaining 18-year life expectancy, a simple calculation yields 10.7 colonoscopies per life year saved ([42 2.3]/9). Similarly, it can be calculated that 8 colonoscopies ([80 0.9]/9) would be needed per life year saved in the flexible sigmoidoscopy screening group in addition to the original 26.7 screening flexible sigmoidoscopies (8/30%). Cost-effectiveness analyses and incremental cost-effectiveness ratios are well suited for the comparison of different screening techniques and the assessment of their impact on public health policy. However, the results of such analyses, expressed in dollar amounts per life year saved, are less helpful in deciding against or in favor of screening in the individual patient. It has been shown that screening remains cost-effective with increasing age, even in the oldest age groups. 18 In such cost-effectiveness analysis, the impact of reduced life expectancy in the elderly is compensated by the higher diagnostic yield of screening endoscopy. Calculations of cost-effectiveness alone, therefore, do not provide an unambiguous cutoff point of when to stop screening. The data shown in Tables 1 and 2 are more congruent with the thought process involved in decision making pertaining to the individual subject. Our method of calculating the NNS or the NNC is relatively simple and transparent and can be readily reproduced or expanded as screening modalities evolve. However, our results provide a relatively crude measure of the gains from colorectal cancer screening, focusing exclusively on prevention of cancer-related death. We did not quantify the number of life years saved or prevention of cancer occurrence with screening, and we did not examine costs, quality of life, or cost-effectiveness. This approach may be partially justified by our focus on cancer screening in older individuals whose life expectancy is already relatively limited. This approach may be less applicable to younger individuals, where the number of life years saved and quality-of-life considerations may require greater consideration. We assumed a constant risk of colonoscopy complications that did not vary with age or comorbidity because no such detailed data are currently available from the literature. If the risk of a colonoscopy complication increases with age or comorbidity, the balance of harms and benefits may lean more heavily toward harm for more frail individuals. We assumed that screening would provide equal reduction in colorectal cancer mortality regardless of age. However, randomized trials of fecal occult blood testing included mostly individuals aged 75 years or younger, and we do not know the absolute benefit of screening in individuals older than 75 years. In general, screening tends to identify less aggressive cancers, which may not become clinically significant over the life span of elderly patients. We also assumed that a reduction in cancer-related mortality would not be seen for at least 5 years for any screening modality. This assumption is supported by data from the randomized trials for fecal occult blood testing. This may not be true for flexible sigmoidoscopy or colonoscopy, and the validity of this assumption will need to be reassessed after current randomized trials of sigmoidoscopy screening are completed. 33,34 We also assumed 100% compliance with screening tests and any recommended follow-up diagnostic and surveillance tests. This approach would maximize the reduction in cancer-related mortality but also maximize the potential for screening-related complications as the number of colonoscopies increases. In conclusion, the potential to reduce colorectal cancer related mortality with screening varies widely with patient age, comorbidity, and screening modality chosen. The potential for screening-related harm, primarily from colonoscopy complications, is small but not insignificant and in some cases may be greater than the estimated mortality benefit. This is especially true in elderly individuals with limited life expectancy and with more invasive screening options. These results may provide additional quantitative information to be integrated with patient preferences and provider judgment, which may be useful in decision making about colorectal cancer screening in individual elderly patients. References 1. Selby JV, Friedman GD, Collen MF. Sigmoidoscopy and mortality from colorectal cancer: the Kaiser Permanente Multiphasic Evaluation Study. J Clin Epidemiol 1988;41:427 434. 2. Selby JV, Friedman GD, Quesenberry CP, Weiss NS. A casecontrol study of screening sigmoidoscopy and mortality from colorectal cancer. N Engl J Med 1992;326:653 657. 3. Mandel JS, Bond JH, Church TR, Snover DC, Bradley GM, Schuman LM, Ederer F. Reducing mortality from colorectal cancer by screening for fecal occult blood. Minnesota Colon Cancer Control Study. N Engl J Med 1993;328:1365 1371. 4. Hardcastle JD, Chamberlain JO, Robinson MHE, Moss SM, Amar SS, Balfour TW, James PD, Mangham CM. Randomised controlled trial of faecal-occult-blood screening for colorectal cancer. Lancet 1996;348:1472 1477.

1170 KO AND SONNENBERG GASTROENTEROLOGY Vol. 129, No. 4 5. Kronborg O, Fenger C, Olsen J, Jorgensen OD, Sondergaard O. Randomised study of screening for colorectal cancer with faecaloccult-blood test. Lancet 1996;348:1467 1471. 6. Muller AD, Sonnenberg A. Protection by endoscopy against death from colorectal cancer: a case-control study among veterans. Arch Intern Med 1995;155:1741 1748. 7. Newcomb PA, Norfleet RG, Storer BE, Surawicz T, Marcus PM. Screening sigmoidoscopy and colorectal cancer mortality. J Natl Cancer Inst 1992;84:1572 1575. 8. Winawer SJ, Flehinger BJ, Schottenfeld D, Miller DG. Screening for colorectal cancer with fecal occult blood testing and sigmoidoscopy. J Natl Cancer Inst 1993;85:1311 1318. 9. Screening for colorectal cancer: recommendation and rationale. Ann Intern Med 2002;137:129 131. 10. Rex DK, Johnson DA, Lieberman DA, Burt RW, Sonnenberg A. Colorectal cancer prevention 2000: screening recommendations of the American College of Gastroenterology. American College of Gastroenterology. Am J Gastroenterol 2000;95:868 877. 11. Winawer S, Fletcher R, Rex D, Bond J, Burt R, Ferrucci J, Ganiats T, Levin T, Woolf S, Johnson D, Kirk L, Litin S, Simmang C. Colorectal cancer screening and surveillance: clinical guidelines and rationale update based on new evidence. Gastroenterology 2003;124:544 560. 12. Smith RA, Cokkinides V, Eyre HJ. American Cancer Society guidelines for the early detection of cancer, 2003. CA Cancer J Clin 2003;53:27 43. 13. Gatto NM, Frucht H, Sundararajan V, Jacobson JS, Grann VR, Neugut AI. Risk of perforation after colonoscopy and sigmoidoscopy: a population-based study. J Natl Cancer Inst 2003;95: 230 236. 14. Walter LC, Covinsky KE. Cancer screening in elderly patients: a framework for individualized decision making. JAMA 2001;285: 2750 2756. 15. Ness RM, Holmes AM, Klein R, Dittus R. Cost-utility of one-time colonoscopic screening for colorectal cancer at various ages. Am J Gastroenterol 2000;95:1800 1811. 16. Sonnenberg A, Delco F, Inadomi JM. Cost-effectiveness of colonoscopy in screening for colorectal cancer. Ann Intern Med 2000;133:573 584. 17. Vijan S, Hwang EW, Hofer TP, Hayward RA. Which colon cancer screening test? A comparison of costs, effectiveness, and compliance. Am J Med 2001;111:593 601. 18. Sonnenberg A, Delco F. Cost-effectiveness of a single colonoscopy in screening for colorectal cancer. Arch Intern Med 2002; 162:163 168. 19. Frazier AL, Colditz GA, Fuchs CS, Kuntz KM. Cost-effectiveness of screening for colorectal cancer in the general population. JAMA 2000;284:1954 1961. 20. Pignone M, Saha S, Hoerger T, Mandelblatt J. Cost-effectiveness analyses of colorectal cancer screening: a systematic review for the U.S. Preventive Services Task Force. Ann Intern Med 2002; 137:96 104. 21. Levin TR, Conell C, Shapiro JA, Chazan SG, Nadel MR, Selby JV. Complications of screening flexible sigmoidoscopy. Gastroenterology 2002;123:1786 1792. 22. Viiala CH, Zimmerman M, Cullen DJ, Hoffman NE. Complication rates of colonoscopy in an Australian teaching hospital environment. Intern Med J 2003;33:355 359. 23. Korman LY, Overholt BF, Box T, Winker CK. Perforation during colonoscopy in endoscopic ambulatory surgical centers. Gastrointest Endosc 2003;58:554 557. 24. Sieg A, Hachmoeller-Eisenbach U, Eisenbach T. Prospective evaluation of complications in outpatient GI endoscopy: a survey among German gastroenterologists. Gastrointest Endosc 2001; 53:620 627. 25. Levin TR, Zhao W, Conell C, Seeff LC, Manninen DL, Leighton PC, Shapiro JA, Schulman J. Complications of colonoscopy in a community practice setting (abstr). Gastroenterology 2004;126 (Suppl):A25. 26. Lieberman DA, Weiss DG, Bond JH, Ahnen DJ, Garewal H, Chejfec G. Use of colonoscopy to screen asymptomatic adults for colorectal cancer. Veterans Affairs Cooperative Study Group 380. N Engl J Med 2000;343:162 168. 27. Rickert RR, Auerbach O, Garfinkel L, Hammond EC, Frasca JM. Adenomatous lesions of the large bowel: an autopsy survey. Cancer 1979;43:1847 1857. 28. Arminski TC, McLean DW. Incidence and distribution of adenomatous polyps of the colon and rectum based on 1,000 autopsy examinations. Dis Colon Rectum 1964;19:249 261. 29. National Center for Health Statistics. Life tables of the United States, 1997. Volume 2005. Hyattsville, MD: 1997. 30. Surveillance, Epidemiology, and End Results (SEER) Program, 1973 1996. In: Ries LAG, Eisner MP, Kosary CL, Hankey BF, Miller BA, Clegg L, Manotto A, Feuger EJ, Edwards BK, eds. SEER Cancer Statistics Review, 1975-2001. Bethesda, MD: National Cancer Institute, 2004. 31. Delco F, Sonnenberg A. At what age should a one-time only colonoscopy for screening of colorectal cancer be performed? Eur J Gastroenterol Hepatol 1999;11:1319 1320. 32. Salkeld G, Solomon M, Short L, Ryan M, Ward JE. Evidence-based consumer choice: a case study in colorectal cancer screening. Aust N Z J Public Health 2003;27:449 455. 33. Atkin WS, Hart A, Edwards R, McIntyre P, Aubrey R, Wardle J, Sutton S, Cuzick J, Northover JM. Uptake, yield of neoplasia, and adverse effects of flexible sigmoidoscopy screening. Gut 1998; 42:560 565. 34. Segnan N, Senore C, Andreoni B, Aste H, Bonelli L, Crosta C, Ferraris R, Gasperoni S, Penna A, Risio M, Rossini FP, Sciallero S, Zappa M, Atkin WS. Baseline findings of the Italian multicenter randomized controlled trial of once-only sigmoidoscopy SCORE. J Natl Cancer Inst 2002;94:1763 1772. Received March 29, 2005. Accepted June 30, 2005. Address requests for reprints to: Cynthia W. Ko, MD, MS, Division of Gastroenterology, Box 356424, University of Washington, Seattle, Washington 98195. e-mail: cwko@u.washington.edu; fax: (206) 685-8684.