When to stop screening for colorectal cancer

Abstract

Lin OS, Kozarek RA, Schembre DB, Ayub K, Gluck M, Drennan F, Soon MS, Rabeneck L. (Gastroenterology Section, Virginia Mason Medical Center, Seattle, Washington). Screening colonoscopy in very elderly patients: prevalence of neoplasia and estimated impact on life expectancy. JAMA 2006;295:2357–2365.Colorectal cancer (CRC) represents the second leading cause of cancer death in the United States. Although multiple strategies are recommended to decrease mortality, guidelines currently do not entertain a maximum age beyond which screening is not endorsed (CA Cancer J Clin 1997;47:154–160, Gastroenterology 2003;124:544–560). The incidence of CRC increases with advancing age; however, the competing risk of mortality from other causes results in a progressively smaller proportion of deaths that may be attributed to CRC. It stands to reason that there exists a threshold age beyond which the risk and resources required to continue screening exceed the benefit. To address this issue, the investigators compared the mean extension in life expectancy of screening colonoscopy in very elderly versus younger persons.The study methods combine prospective data regarding the prevalence of neoplasia in a large and diverse screening population with mathematical modeling techniques to estimate the gain in life expectancy achieved through screening colonoscopy. Consecutive asymptomatic patients who received screening colonoscopy at a single center were enrolled in the study. Exclusion criteria included the presence of gastrointestinal symptoms (bleeding, diarrhea, change in bowel habits, abdominal pain), previous history of CRC or adenomas, inflammatory bowel disease, or familial forms of CRC (familial adenomatous polyposis or hereditary nonpolyposis cancer syndromes), although a family history of sporadic CRC was allowed. The prevalence of neoplasia, advanced neoplasia (adenoma with >25% villous features or ≥1 cm in diameter), adenomas with high-grade dysplasia, and adenocarcinoma were tabulated among this cohort. These data were used as inputs in a mathematical model to estimate the extension in life expectancy gained through the performance of screening colonoscopy. The model chosen by the investigators was the Declining Exponential Approximation of Life Expectancy (DEALE). This technique was developed as a simple yet elegant method to separate the impact of a single disease process, such as CRC, from the competing risk of mortality from other processes (Am J Med 1982;73:883–888, Am J Med 1982;73:889–897). For this study, the incidence and mortality from CRC were extrapolated from the prevalence of neoplasia detected in the population sample, and the DEALE was used to estimate the extension in life expectancy with screening colonoscopy. For the primary outcome of the analysis, the investigators examined various age groups and estimated differences in life expectancy extension between very elderly (age ≥80) and younger patients (ages 50–54 years).The results revealed that the prevalence of neoplasia increased with age; compared with an adenoma prevalence of 13.9% in patients 50–54 years old, 26.7% of patients 75–79 years old and 28.2% of those ≥80 years harbored adenomas. Advanced neoplasia (including high-grade dysplasia and cancer) were found in 3.2%, 4.7%, and 14% of the cohorts aged 50–54, 75–79, and ≥80 years of age, respectively. Based on these findings, mathematical modeling predicted that the mean extension in life expectancy was greater for those 50–54 years old (0.85 years) than for older age groups (0.17 years and 0.13 years for those 75–79 years old and those ≥80 years old, respectively). The model was sensitive to assumptions regarding the lag time between polyp discovery and development of CRC (in the absence of polypectomy); longer lag times decreased the extension in life expectancy with colonoscopy among all ages, but this was more evident among the older age groups. The investigators concluded that although the incidence of CRC increases with age, so too does the risk of mortality from other diseases. As a result, patients ≥80 years of age experience only 15% of the expected gain in life expectancy achieved by screening those aged 50–54 years.CommentPrevious studies have demonstrated differences in benefit from CRC screening between younger and older patients (Comput Biomed Res 1999;32:13–33, Gastrointest Endosc 2000;51:517–523, Gastroenterology 2005;129:1163–1170, J Natl Cancer Inst 2000;92:557–563). A common theme to these studies is the finding that the risk of death from competing disease tends to offset the benefit achieved from CRC screening as the population ages. This is true of any disease process for which screening is conducted to decrease mortality; older persons benefit less from reduction of any 1 disease because of the increasing risk of death from other diseases. This study is innovative in its explicit focus on comparing the benefit of screening among “very elderly” patients (≥80 years of age) to younger age groups (50–54 years old). Thus, this paper has the ability to address a fundamental decision that must be made for every patient: when to stop tests that are designed to reduce mortality from CRC.A brief comment should be made regarding the differences between incidence and prevalence, and between mortality and life expectancy. Prevalence is the proportion of the population with a particular condition at a fixed moment in time, and incidence is the proportion of the population who develop a condition between multiple time points. Mortality is the rate of death, which may be an overall rate encompassing all forms of death or specific to the proportion of persons dying of a particular disease over a period of time. Life expectancy is the inverse of mortality (see DEALE, below) that incorporates the risk of mortality from multiple diseases including those associated with “old age” to describe the amount of time remaining until death of an individual or cohort. An intervention may decrease mortality from a specific disease and yet have little or no impact on life expectancy of a population if that disease is very rare, if excess mortality from other disease is caused by the intervention, or if the population has a high risk of death from competing disease. This concept explains in part the discrepancy between increased incidence of CRC yet decreased extension in life expectancy with screening colonoscopy in older age groups.The methods used in this study are worthy of comment. The DEALE assumes that the survival of a population declines in an exponential fashion as a function of time. The equation is quite simple:S(t)=S0×e−[μ(age)+μ(CRC)]×t where S(t) is the surviving population at time t, S0 is the initial population and the additional variables represent the mortality rates owing to CRC [μ(CRC)] and age-related non-CRC factors [μ(age)]. The DEALE requires only these few inputs to describe the survival of a cohort and separate the effect of mortality from a single (or several) disease states from overall mortality. The data required can be obtained from the vital statistics recorded by multiple countries, including the United States, and the only assumptions required are those built into the DEALE itself. The life expectancy of an individual patient can be estimated by the inverse of the total mortality:Life⁢expectancy=1/μ(total)=1/[μ(age)+μ(CRC)].The DEALE technique has been compared with other, more complex mathematical derivations of life expectancy developed by Gompertz or Makeham (Am J Med 1982;73:883–888, Philos Trans R Soc [Lond] 1825;115:513–585). The maximal error in life expectancy estimation using the DEALE is limited to 12%, and the ability to discriminate changes in life expectancy owing to single diseases has been shown to be robust as long as the overall mortality from this disease for the population being studied is <10%. The DEALE as a mathematical model has been largely usurped by more complex statistical methods such as Markov models and linear programming that use the capabilities of modern computers (Med Decis Making 1983;3:419–458). The disadvantage of using more complex modeling techniques is that the clinical inputs required to populate the models are often erroneous or poorly described; thus, multiple assumptions regarding the values of these inputs have to be made, potentially rendering the results invalid. Having used both techniques, this reviewer believes that the qualitative comparisons presented by the investigators in this study are adequately and transparently provided using the DEALE.In their conclusions, the authors fall short of advocating cessation of CRC screening for patients beyond a cutoff age of 75 or 80 years. This may be because there was a persistent benefit derived from CRC screening even among very elderly patients, albeit less than among younger patients. CRC screening will fail to provide benefit when the risk of mortality from the screening test is greater than the risk of mortality from CRC. Up to that point, however, the decision to continue screening should be based on patient preference and societal resource allocation that requires a cost-effectiveness analysis, which was not conducted in this study. Multiple economic studies of CRC screening have been conducted, all of which affirm the benefit and cost effectiveness of screening colonoscopy in average risk people to decrease mortality from CRC (Am J Med 2001;111:593–601, Ann Intern Med 2000;133:573–584, JAMA 2000;284:1954–1961). However, little consensus has been derived concerning a specific age beyond which screening should be curtailed. Certainly we have all seen 50-year-old patients who would be unlikely to benefit from screening based on the risk of death from other disease; conversely, it is reasonable to continue screening in an 80-year-old patient whose genealogy expresses longevity or whose individual life expectancy exceeds 10 years. From the perspective of resource utilization and health policy, it would be reasonable to advocate continued screening in elderly patients as long as the cost effectiveness does not exceed the cost effectiveness of interventions society commonly recommends for other diseases. For example, the incremental cost effectiveness for cervical cancer screening ($250,000 per life-year saved) surveillance among patients with Barrett’s esophagus without dysplasia compared with Barrett’s with dysplasia ($350,000 per quality-adjusted life year saved) or automobile airbags ($1,000,000 per life-year saved) may exceed the resource expenditures required to continue CRC screening for appropriate patients ≥80 years of age. As long as those practices are condoned and paid for by society, it is reasonable to advocate continued CRC screening even for very elderly patients who are reasonable candidates for colonoscopy. If one is to “stop the madness” then one should look elsewhere. Lin OS, Kozarek RA, Schembre DB, Ayub K, Gluck M, Drennan F, Soon MS, Rabeneck L. (Gastroenterology Section, Virginia Mason Medical Center, Seattle, Washington). Screening colonoscopy in very elderly patients: prevalence of neoplasia and estimated impact on life expectancy. JAMA 2006;295:2357–2365. Colorectal cancer (CRC) represents the second leading cause of cancer death in the United States. Although multiple strategies are recommended to decrease mortality, guidelines currently do not entertain a maximum age beyond which screening is not endorsed (CA Cancer J Clin 1997;47:154–160, Gastroenterology 2003;124:544–560). The incidence of CRC increases with advancing age; however, the competing risk of mortality from other causes results in a progressively smaller proportion of deaths that may be attributed to CRC. It stands to reason that there exists a threshold age beyond which the risk and resources required to continue screening exceed the benefit. To address this issue, the investigators compared the mean extension in life expectancy of screening colonoscopy in very elderly versus younger persons. The study methods combine prospective data regarding the prevalence of neoplasia in a large and diverse screening population with mathematical modeling techniques to estimate the gain in life expectancy achieved through screening colonoscopy. Consecutive asymptomatic patients who received screening colonoscopy at a single center were enrolled in the study. Exclusion criteria included the presence of gastrointestinal symptoms (bleeding, diarrhea, change in bowel habits, abdominal pain), previous history of CRC or adenomas, inflammatory bowel disease, or familial forms of CRC (familial adenomatous polyposis or hereditary nonpolyposis cancer syndromes), although a family history of sporadic CRC was allowed. The prevalence of neoplasia, advanced neoplasia (adenoma with >25% villous features or ≥1 cm in diameter), adenomas with high-grade dysplasia, and adenocarcinoma were tabulated among this cohort. These data were used as inputs in a mathematical model to estimate the extension in life expectancy gained through the performance of screening colonoscopy. The model chosen by the investigators was the Declining Exponential Approximation of Life Expectancy (DEALE). This technique was developed as a simple yet elegant method to separate the impact of a single disease process, such as CRC, from the competing risk of mortality from other processes (Am J Med 1982;73:883–888, Am J Med 1982;73:889–897). For this study, the incidence and mortality from CRC were extrapolated from the prevalence of neoplasia detected in the population sample, and the DEALE was used to estimate the extension in life expectancy with screening colonoscopy. For the primary outcome of the analysis, the investigators examined various age groups and estimated differences in life expectancy extension between very elderly (age ≥80) and younger patients (ages 50–54 years). The results revealed that the prevalence of neoplasia increased with age; compared with an adenoma prevalence of 13.9% in patients 50–54 years old, 26.7% of patients 75–79 years old and 28.2% of those ≥80 years harbored adenomas. Advanced neoplasia (including high-grade dysplasia and cancer) were found in 3.2%, 4.7%, and 14% of the cohorts aged 50–54, 75–79, and ≥80 years of age, respectively. Based on these findings, mathematical modeling predicted that the mean extension in life expectancy was greater for those 50–54 years old (0.85 years) than for older age groups (0.17 years and 0.13 years for those 75–79 years old and those ≥80 years old, respectively). The model was sensitive to assumptions regarding the lag time between polyp discovery and development of CRC (in the absence of polypectomy); longer lag times decreased the extension in life expectancy with colonoscopy among all ages, but this was more evident among the older age groups. The investigators concluded that although the incidence of CRC increases with age, so too does the risk of mortality from other diseases. As a result, patients ≥80 years of age experience only 15% of the expected gain in life expectancy achieved by screening those aged 50–54 years. CommentPrevious studies have demonstrated differences in benefit from CRC screening between younger and older patients (Comput Biomed Res 1999;32:13–33, Gastrointest Endosc 2000;51:517–523, Gastroenterology 2005;129:1163–1170, J Natl Cancer Inst 2000;92:557–563). A common theme to these studies is the finding that the risk of death from competing disease tends to offset the benefit achieved from CRC screening as the population ages. This is true of any disease process for which screening is conducted to decrease mortality; older persons benefit less from reduction of any 1 disease because of the increasing risk of death from other diseases. This study is innovative in its explicit focus on comparing the benefit of screening among “very elderly” patients (≥80 years of age) to younger age groups (50–54 years old). Thus, this paper has the ability to address a fundamental decision that must be made for every patient: when to stop tests that are designed to reduce mortality from CRC.A brief comment should be made regarding the differences between incidence and prevalence, and between mortality and life expectancy. Prevalence is the proportion of the population with a particular condition at a fixed moment in time, and incidence is the proportion of the population who develop a condition between multiple time points. Mortality is the rate of death, which may be an overall rate encompassing all forms of death or specific to the proportion of persons dying of a particular disease over a period of time. Life expectancy is the inverse of mortality (see DEALE, below) that incorporates the risk of mortality from multiple diseases including those associated with “old age” to describe the amount of time remaining until death of an individual or cohort. An intervention may decrease mortality from a specific disease and yet have little or no impact on life expectancy of a population if that disease is very rare, if excess mortality from other disease is caused by the intervention, or if the population has a high risk of death from competing disease. This concept explains in part the discrepancy between increased incidence of CRC yet decreased extension in life expectancy with screening colonoscopy in older age groups.The methods used in this study are worthy of comment. The DEALE assumes that the survival of a population declines in an exponential fashion as a function of time. The equation is quite simple:S(t)=S0×e−[μ(age)+μ(CRC)]×t where S(t) is the surviving population at time t, S0 is the initial population and the additional variables represent the mortality rates owing to CRC [μ(CRC)] and age-related non-CRC factors [μ(age)]. The DEALE requires only these few inputs to describe the survival of a cohort and separate the effect of mortality from a single (or several) disease states from overall mortality. The data required can be obtained from the vital statistics recorded by multiple countries, including the United States, and the only assumptions required are those built into the DEALE itself. The life expectancy of an individual patient can be estimated by the inverse of the total mortality:Life⁢expectancy=1/μ(total)=1/[μ(age)+μ(CRC)].The DEALE technique has been compared with other, more complex mathematical derivations of life expectancy developed by Gompertz or Makeham (Am J Med 1982;73:883–888, Philos Trans R Soc [Lond] 1825;115:513–585). The maximal error in life expectancy estimation using the DEALE is limited to 12%, and the ability to discriminate changes in life expectancy owing to single diseases has been shown to be robust as long as the overall mortality from this disease for the population being studied is <10%. The DEALE as a mathematical model has been largely usurped by more complex statistical methods such as Markov models and linear programming that use the capabilities of modern computers (Med Decis Making 1983;3:419–458). The disadvantage of using more complex modeling techniques is that the clinical inputs required to populate the models are often erroneous or poorly described; thus, multiple assumptions regarding the values of these inputs have to be made, potentially rendering the results invalid. Having used both techniques, this reviewer believes that the qualitative comparisons presented by the investigators in this study are adequately and transparently provided using the DEALE.In their conclusions, the authors fall short of advocating cessation of CRC screening for patients beyond a cutoff age of 75 or 80 years. This may be because there was a persistent benefit derived from CRC screening even among very elderly patients, albeit less than among younger patients. CRC screening will fail to provide benefit when the risk of mortality from the screening test is greater than the risk of mortality from CRC. Up to that point, however, the decision to continue screening should be based on patient preference and societal resource allocation that requires a cost-effectiveness analysis, which was not conducted in this study. Multiple economic studies of CRC screening have been conducted, all of which affirm the benefit and cost effectiveness of screening colonoscopy in average risk people to decrease mortality from CRC (Am J Med 2001;111:593–601, Ann Intern Med 2000;133:573–584, JAMA 2000;284:1954–1961). However, little consensus has been derived concerning a specific age beyond which screening should be curtailed. Certainly we have all seen 50-year-old patients who would be unlikely to benefit from screening based on the risk of death from other disease; conversely, it is reasonable to continue screening in an 80-year-old patient whose genealogy expresses longevity or whose individual life expectancy exceeds 10 years. From the perspective of resource utilization and health policy, it would be reasonable to advocate continued screening in elderly patients as long as the cost effectiveness does not exceed the cost effectiveness of interventions society commonly recommends for other diseases. For example, the incremental cost effectiveness for cervical cancer screening ($250,000 per life-year saved) surveillance among patients with Barrett’s esophagus without dysplasia compared with Barrett’s with dysplasia ($350,000 per quality-adjusted life year saved) or automobile airbags ($1,000,000 per life-year saved) may exceed the resource expenditures required to continue CRC screening for appropriate patients ≥80 years of age. As long as those practices are condoned and paid for by society, it is reasonable to advocate continued CRC screening even for very elderly patients who are reasonable candidates for colonoscopy. If one is to “stop the madness” then one should look elsewhere. Previous studies have demonstrated differences in benefit from CRC screening between younger and older patients (Comput Biomed Res 1999;32:13–33, Gastrointest Endosc 2000;51:517–523, Gastroenterology 2005;129:1163–1170, J Natl Cancer Inst 2000;92:557–563). A common theme to these studies is the finding that the risk of death from competing disease tends to offset the benefit achieved from CRC screening as the population ages. This is true of any disease process for which screening is conducted to decrease mortality; older persons benefit less from reduction of any 1 disease because of the increasing risk of death from other diseases. This study is innovative in its explicit focus on comparing the benefit of screening among “very elderly” patients (≥80 years of age) to younger age groups (50–54 years old). Thus, this paper has the ability to address a fundamental decision that must be made for every patient: when to stop tests that are designed to reduce mortality from CRC. A brief comment should be made regarding the differences between incidence and prevalence, and between mortality and life expectancy. Prevalence is the proportion of the population with a particular condition at a fixed moment in time, and incidence is the proportion of the population who develop a condition between multiple time points. Mortality is the rate of death, which may be an overall rate encompassing all forms of death or specific to the proportion of persons dying of a particular disease over a period of time. Life expectancy is the inverse of mortality (see DEALE, below) that incorporates the risk of mortality from multiple diseases including those associated with “old age” to describe the amount of time remaining until death of an individual or cohort. An intervention may decrease mortality from a specific disease and yet have little or no impact on life expectancy of a population if that disease is very rare, if excess mortality from other disease is caused by the intervention, or if the population has a high risk of death from competing disease. This concept explains in part the discrepancy between increased incidence of CRC yet decreased extension in life expectancy with screening colonoscopy in older age groups. The methods used in this study are worthy of comment. The DEALE assumes that the survival of a population declines in an exponential fashion as a function of time. The equation is quite simple:S(t)=S0×e−[μ(age)+μ(CRC)]×t where S(t) is the surviving population at time t, S0 is the initial population and the additional variables represent the mortality rates owing to CRC [μ(CRC)] and age-related non-CRC factors [μ(age)]. The DEALE requires only these few inputs to describe the survival of a cohort and separate the effect of mortality from a single (or several) disease states from overall mortality. The data required can be obtained from the vital statistics recorded by multiple countries, including the United States, and the only assumptions required are those built into the DEALE itself. The life expectancy of an individual patient can be estimated by the inverse of the total mortality:Life⁢expectancy=1/μ(total)=1/[μ(age)+μ(CRC)]. The DEALE technique has been compared with other, more complex mathematical derivations of life expectancy developed by Gompertz or Makeham (Am J Med 1982;73:883–888, Philos Trans R Soc [Lond] 1825;115:513–585). The maximal error in life expectancy estimation using the DEALE is limited to 12%, and the ability to discriminate changes in life expectancy owing to single diseases has been shown to be robust as long as the overall mortality from this disease for the population being studied is <10%. The DEALE as a mathematical model has been largely usurped by more complex statistical methods such as Markov models and linear programming that use the capabilities of modern computers (Med Decis Making 1983;3:419–458). The disadvantage of using more complex modeling techniques is that the clinical inputs required to populate the models are often erroneous or poorly described; thus, multiple assumptions regarding the values of these inputs have to be made, potentially rendering the results invalid. Having used both techniques, this reviewer believes that the qualitative comparisons presented by the investigators in this study are adequately and transparently provided using the DEALE. In their conclusions, the authors fall short of advocating cessation of CRC screening for patients beyond a cutoff age of 75 or 80 years. This may be because there was a persistent benefit derived from CRC screening even among very elderly patients, albeit less than among younger patients. CRC screening will fail to provide benefit when the risk of mortality from the screening test is greater than the risk of mortality from CRC. Up to that point, however, the decision to continue screening should be based on patient preference and societal resource allocation that requires a cost-effectiveness analysis, which was not conducted in this study. Multiple economic studies of CRC screening have been conducted, all of which affirm the benefit and cost effectiveness of screening colonoscopy in average risk people to decrease mortality from CRC (Am J Med 2001;111:593–601, Ann Intern Med 2000;133:573–584, JAMA 2000;284:1954–1961). However, little consensus has been derived concerning a specific age beyond which screening should be curtailed. Certainly we have all seen 50-year-old patients who would be unlikely to benefit from screening based on the risk of death from other disease; conversely, it is reasonable to continue screening in an 80-year-old patient whose genealogy expresses longevity or whose individual life expectancy exceeds 10 years. From the perspective of resource utilization and health policy, it would be reasonable to advocate continued screening in elderly patients as long as the cost effectiveness does not exceed the cost effectiveness of interventions society commonly recommends for other diseases. For example, the incremental cost effectiveness for cervical cancer screening ($250,000 per life-year saved) surveillance among patients with Barrett’s esophagus without dysplasia compared with Barrett’s with dysplasia ($350,000 per quality-adjusted life year saved) or automobile airbags ($1,000,000 per life-year saved) may exceed the resource expenditures required to continue CRC screening for appropriate patients ≥80 years of age. As long as those practices are condoned and paid for by society, it is reasonable to advocate continued CRC screening even for very elderly patients who are reasonable candidates for colonoscopy. If one is to “stop the madness” then one should look elsewhere.