Computed tomographic colonography (CTC), popularly referred

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2007;5: The Management of Small Polyps Found by Virtual Colonoscopy: Results of a Decision Analysis CHIN HUR,*,, DANIEL C. CHUNG,*, ROBERT E. SCHOEN, and G. SCOTT GAZELLE,,,# *Gastrointestinal Unit, Institute for Technology Assessment, Department of Radiology, Massachusetts General Hospital, Boston, Massachusetts; Harvard Medical School, Harvard School of Public Health, # Department of Health Policy and Management, Boston, Massachusetts; and University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania Background & Aims: There is a firm consensus that larger (>10 mm) colonic polyps should be removed; however, the importance of removing smaller polyps (<10 mm) is more controversial. If computed tomographic colonography (CTC) is used for colorectal cancer screening, the majority of polypoid lesions identified will be less than 10 mm in size. Decision-analytic techniques were used to compare the outcomes of 2 management strategies for smaller (6 9 mm) polyps discovered by CTC. Methods: Hypothetic average-risk patients who had undergone a CTC examination and found to have a small (6 9 mm) polyp were simulated to either: (1) undergo immediate colonoscopy for polypectomy (COLO), or (2) wait 3 years for a repeat CTC examination (WAIT). A Markov model was constructed to analyze outcomes including the number of deaths and cancers after a 3-year follow-up period or time horizon. Values for the model parameters were derived from the published literature and from Surveillance Epidemiology and End Results data, and an extensive sensitivity analysis was performed. Results: The COLO strategy resulted in 14 total deaths per 100,000 patients compared with 79 total deaths in the WAIT strategy, for a difference of 65 deaths. The COLO strategy resulted in 39 cancers per 100,000 patients vs 773 in the WAIT strategy, for a difference of 734 cancers. Sensitivity analysis found that model findings were robust and only sensitive at extreme parameter values. Conclusions: Managing smaller polyps detected on a screening CTC with another CTC examination 3 years later likely will result in more deaths and cancers than immediate colonoscopy and polypectomy. Computed tomographic colonography (CTC), popularly referred to as virtual colonoscopy, is a technique that uses CT data to generate 2- and 3-dimensional images of the colon and rectum. 1 The potential use of CTC for primary colorectal cancer (CRC) screening is currently a topic of great debate because recent studies have yielded widely varied results. 2 5 CTC is not yet reimbursed by Medicare and most other third-party payers for CRC screening, but if reimbursement policies change, a substantial number of patients likely will undergo screening CTC examinations. 6,7 There is a firm consensus that larger ( 10 mm) colonic polyps need to be removed because of the considerable risk of an indwelling cancer or of progression to cancer. Although smaller polyps ( 10 mm) frequently are removed during colonoscopy, the benefit and necessity of resection of these smaller polyps is more controversial. 8 If CTC is used for CRC screening, the majority of polypoid lesions found by these examinations will be less than 10 mm in size. 2 Recently, a published consensus proposal for CTC reporting suggested that it would be reasonable to manage a smaller (6 9 mm) polypoid lesion found on screening CTC examination with another CTC examination 3 years later instead of an immediate colonoscopy for polypectomy, which is recommended for larger polyps that are 10 mm or more in size or if 3 or more smaller polyps are present. 9 Many patients and physicians may not be comfortable with the risks associated with not removing a polyp, which theoretically could harbor malignancy, high-grade dysplasia, or could progress to malignancy within the 3 years before the follow-up CTC examination. On the other hand, the risks of colonoscopy and polypectomy complications could outweigh the benefit of polypectomy and cancer prevention in patients with smaller polyps. The aim of this study was to compare 2 strategies, immediate colonoscopy for polypectomy (COLO strategy) vs repeating another CTC in 3 years (WAIT strategy) for patients with a 6- to 9-mm polypoid lesion found on an initial screening CTC examination. Methods A Markov decision-analytic model was constructed to compare the outcomes of the COLO strategy vs the WAIT strategy. A Markov model is a mathematic model that can be used to predict outcomes for patients as they progress through disease states and undergo diagnostic tests and/or therapeutic interventions. It is particularly useful in modeling clinical situations in which there is a repeated risk over time. 10,11 The primary outcome of the analysis was the total number of deaths. Simulated patients could die from either cancer or colonoscopy complications, however, age-specific all-cause mortality was not included in the model because the rates would have been equivalent in the 2 strategies and simplicity was a goal in our modeling approach. Secondary end points included the total number of cancers (both preclinical and those diagnosed because of signs and symptoms) and their stage distribution. The time horizon or follow-up period for the model was Abbreviations used in this paper: CA, cancer; COLO, undergo immediate colonoscopy for polypectomy; CRC, colorectal cancer; CTC, computed tomographic colonography; SEER, Surveillance Epidemiology and End Results; WAIT, wait 3 years for a repeat computed tomographic colonography examination by the AGA Institute /07/$32.00 doi: /j.cgh

2 238 HUR ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 2 Figure 1. Markov model used for the analysis. The start of the model shows a hypothetic 50-year-old patient who is found to have a 6- to 9-mm polypoid lesion on a screening CTC. Two strategies, immediate colonoscopy for polypectomy (COLO), and repeat the CTC examination in 3 years (WAIT), are modeled. A triangle at the end of a course signifies that the patient will be in this state at the end of the 3-year analysis. A circle with the letter M inside signifies that this is a Markov chain and that the patient will continue to another portion of the model. As stated in the text box, a patient with cancer can die from the cancer at any time in the 3-year analysis based on the annual stage-specific mortality rates stage (see Table 1). 3 years, which was chosen because this is the recommended time interval before follow-up CTC. 9 Model Structure Figure 1 shows a simplified schematic of the model. In the COLO strategy, the hypothetic cohort of 50-year-old patients (age when CRC screening is recommended to commence for average-risk individuals) 12,13 found to have a positive CTC finding are simulated to undergo immediate colonoscopy and possible polypectomy. The simulated patients were at risk for colonoscopy complications and associated death. The analysis also incorporated the possibility of incorrect positive findings (ie, there was no polyp in reality) from the screening CTC and the inability of the endoscopist to locate and remove a polyp seen on CTC. If the polyp was found and removed during the colonoscopy, the simulated patient had no chance of developing cancer in the model. Because patients who had a successful polypectomy could not develop cancer, all the cancers in the COLO strategy were from polyps that could not be found. In the WAIT strategy, the patients with a correct positive CTC finding were modeled to have either a hyperplastic or adenomatous polyp. Although characterization of polyp histology is not possible by a CTC examination, these internal classifications were necessary to make projections of clinical outcomes. Adenomatous polyps were stratified further into low risk, high risk, and cancer within adenoma. Over the time horizon of the simulation (3 y), low-risk adenomas could be stable and remain in the low-risk adenoma state, or, alternatively, progress into

3 February 2007 RISKS AND BENEFITS OF SMALL POLYP REMOVAL 239 Table 1. Parameter Estimates (Model Inputs) Parameter Base Case Estimate, % (range, %) References CTC estimates Incorrect positive rate 31.9 (17 60) See Methods section CTC sensitivity 70 (55 84) Mulhall et al 14 CTC specificity 93 (91 95) Mulhall et al 14 Polyp prevalence (6 9 mm) 17.6 (10 22) Cotton et al, 3 Pickhardt et al, 2 Rockey et al, 5 Yee et al, 41 Pineau et al, 42 Johnson et al 43 Colonoscopy estimates Complication rate 0.1 ( ) Imperiale et al, 44 Viiala et al, 45 Anderson et al 46 Death rate 0.01 (0 0.02) Imperiale et al, 44 Viiala et al, 45 Anderson et al 46 Unable to find 6- to 9-mm polyp 5 (0 13) Pickhardt et al, 17 Rex et al, 15 Hixson et al 16,47 seen on CTC Initial polyp characteristics Hyperplastic polyps 39.3 ( ) Pickhardt et al, 2 DiSario et al, 27 Vatn and Stalsberg, 26 Eide and Stalsberg 28 Adenomas 60.7 ( ) Pickhardt et al, 2 DiSario et al, 27 Vatn and Stalsberg, 26 Eide and Stalsberg 28 Low risk 94.1 See Methods section High risk 5.0 ( ) Tsuda et al, 48 O Brien et al, 49 Butterly et al, 32 Shinya and Wolff, 50 Gschwantler et al 51 Cancer in adenoma 0.9 ( ) van Dam et al 4 Annual transition rates Natural history Low-risk to high-risk adenoma 1.5 See Methods section High-risk adenoma to localized cancer 5 See Methods section Localized to regional CA 45 See Methods section Regional to distant CA 50 See Methods section CRC symptom detection rates Localized CA 22 See Methods section Regional CA 60 See Methods section Distant CA 90 See Methods section CRC mortality rates Localized CA 2 SEER 31 Regional CA 7.5 SEER 31 Distant CA 37 SEER 31 high-risk adenomas. Similarly, a high-risk adenoma could be stable or progress into localized cancer. Localized cancer could progress further into regional cancer and then ultimately to distant cancer (metastatic). This particular cancer staging classification was used because the model was calibrated to the National Cancer Institute s Surveillance Epidemiology and End Results (SEER) data (see Model Inputs section for details). If a cancer was detected because of signs or symptoms during the follow-up period, it was assumed that the patient would receive some form of treatment and that the patient s cancer mortality risk, which was based on the cancer stage at presentation, would remain stable (see Table 1). A commercially available software package (TreeAge Pro 2005 Suite; TreeAge Software, Williamstown, MA) was used to create and run simulations on the model. Model Inputs Model parameters were based on estimates from the published literature. Base-case values are summarized in Table 1 with references. Polyp characteristics including the percentages that were hyperplastic, low-risk adenomas, high-risk adenomas, and malignant were estimated from the literature. All polyps that were not hyperplastic were considered adenomatous. The starting prevalence of low-risk adenomas (94.1%) was derived by subtracting the percentage of adenomas that contained cancer (0.9%) and those that were high-risk adenomas (5%) from 100%. A relatively low estimate for the percentage of adenomas that were high risk was used in the base-case analysis because many of the studies that reported this percentage used enriched cohorts or patients that were not average risk for CRC. The correct positive rate (or positive predictive value) of a CTC examination depends on the sensitivity and specificity of the CTC study and the polyp prevalence in the population studied. The incorrect positive rate of a CTC examination would therefore be the additive inverse, or 1 (correct positive rate). Mulhall et al 14 recently published a comprehensive meta-analysis on CTC performance characteristics that found that for 6- to 9-mm polyps, CTC had a sensitivity of 70% and a specificity of 93%. The 6- to 9-mm polyp prevalence was estimated to be 17.6%, with a range of 10% 22% in the publications cited in Table 1. This screening population polyp prevalence was used only to calculate the incorrect positive rate of a screening CTC examination and independent of polyp characteristics used in other portions of the model. By using these 3 parameters, the CTC incorrect positive rate was 31.9%. Colonoscopy complications and death rates were estimated from the published literature. To estimate the percentage of

4 240 HUR ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 2 polyps visualized on CTC examination that would not be found during follow-up endoscopy, we started with estimates for the percentage of polyps that are missed in screening colonoscopies. Although test characteristics from tandem colonoscopy studies often are used, these studies do not account for the fact that polyps that were missed by the first endoscopist probably were more likely to have been missed by the second endoscopist, raising concerns about the gold standard used in these studies. 15,16 The study performed by Pickhardt et al 17 used a combination of a colonoscopy and the CTC in a segmental unblinding technique to create a better gold standard on which to derive polyp miss rates during colonoscopy. By using this methodology, they estimated the colonoscopy miss rate to be 10% for 6- to 9-mm polyps, however, they did not report the inability to locate any of the polyps missed by colonoscopy once the CTC examination was unblinded, for a presumed 0% rate. In our simulation, in the COLO strategy, the endoscopist would know that a 6- to 9-mm polypoid lesion was noted on CTC and also have the reported location. However, the endoscopist would not know if the finding was a correct positive or an incorrect positive (eg, retained stool or a colonic fold that was interpreted to be a polyp), which would not be uncommon (31.9%, Table 1). We therefore estimated that the polyp miss rate would be greater than 0% but less than 10%, and used the average of these 2 extremes, or 5%, as our base-case estimate. Furthermore, a sensitivity analysis was performed on this parameter to determine how changes in this estimate affected model predictions. Adenoma and Cancer Natural History Transition Probabilities To analyze the outcomes of the WAIT strategy, we needed to model the biological behavior of the 6- to 9-mm polypoid lesion over the follow-up period of 3 years (Figure 1). Our model assumed that patients with incorrect positives or those with hyperplastic polyps were not at risk for developing cancer over the following 3 years, however, adenomas could progress. This clinical scenario is somewhat unique and consequently it is difficult to estimate adenoma and CRC transition probabilities specific to this situation based on published literature. We therefore constructed a separate natural history model based on general US public data and estimated adenoma and CRC transition probabilities (see Methods section) and input these estimates into our final model (see Discussion section regarding possible limitations). In this general natural history model, the starting point was a cohort of 50-year-old patients with an adenomatous polyp prevalence of 20%. By using SEER data and the NCI DevCan software 18 we estimated that a 50-year-old patient had a 5.9% (men, 6.0%; women, 5.8%) chance of developing CRC over the remainder of their lives. Cancer stage distribution was estimated using the earliest SEER data available that included information regarding stage. 19 Earlier SEER data were used to capture cancer stage distribution that was less affected by CRC screening, which has become more widespread in recent decades. We used estimates from previously analyzed and published CRC disease models that specifically provided transition probabilities for adenomas and cancers as starting points for these initial estimates and then varied them to yield model predictions that provided a good visual fit to the SEER data (cancer incidence and stage distribution). SEER data were used as the primary calibration targets for this natural history model because SEER is the most comprehensive source of cancer incidence and survival data in the United States. For cancer, SEER estimated a 5.9% incidence over the remaining lifetime of risk, whereas our natural history model predicted a 5.85% incidence. SEER data estimated that cancer stages at diagnosis were as follows: 39% localized cancer (CA); 39% regional CA; and 20% distant CA. In comparison, our natural history model projections were: 41% localized CA; 33% regional CA; and 25% distant CA. After model calibration, to further verify that the transition probabilities were realistic, specific end points, or predictions (other than the calibration end points of cancer incidence and cancer stage distribution), from our natural history model were analyzed and compared with the published literature for approximate equivalency. Specifically for our model, we examined adenomatous polyp prevalence at various ages and the cancer death rate (the percentage of patients who developed CRC who would die from it). Model outcomes for adenomatous polyp prevalence by age were as follows: 28% at age 60, 35% at age 70, and 41% at age 80. These values are consistent with published autopsy series data that report prevalence estimates ranging from 14% to 50% Although higher prevalences were reported in men compared with women in many of these studies, we chose not to create 2 separate models (see Discussion section for reasoning). The percentage of patients who developed CRC and died from it was projected to be 44%, in line with the SEER data estimate of 43%. 19 Finally, annual stage-specific CRC mortality rates were derived using the most recent SEER mortality data. 31 The most recent data were used to reflect the recent advances in CRC treatment. Analyses Performed First, a base-case analysis using the best estimates for all model parameters was performed. Sensitivity analyses then were performed on key model parameters, including polyp characteristics (percentage hyperplastic, high-risk adenomas, cancers), adenoma and cancer natural history (transition rates and cancer mortality), as well as CTC and colonoscopy characteristics Table 2. Results of Base-Case Analysis WAIT COLO Difference (%) Total deaths (0.07) Cancer deaths 79 4 a 75 a (0.08) Localized CA (0.02) Regional CA (0.03) Distant CA (0.03) COLO complications deaths (0.01) Total cancers a (0.73) Localized CA (0.40) detected b (0.18) Regional CA (0.26) detected (0.17) Distant CA (0.08) detected (0.05) NOTE. Numbers out of a starting pool of 100,000 patients. a Total numbers do not match sums by cancer stage because of rounding. b The number of patients whose cancers were detected because of symptomatic presentation.

5 February 2007 RISKS AND BENEFITS OF SMALL POLYP REMOVAL 241 (CTC incorrect positive rate, colonoscopy complication rates, and inability to find polyp during colonoscopy). For the adenoma and cancer transition rate sensitivity analysis, we varied all 4 transition probabilities (low-risk to high-risk adenoma, high-risk to localized CA, localized CA to regional CA, and regional CA to distant CA) simultaneously using a multiplicative factor that we termed the transition rates factor. We used a similar CA mortality rates factor for the sensitivity analysis of all 3 stage-dependent cancer mortality rates as well. If the model results were sensitive to a particular parameter (ie, the conclusions could be changed by altering the parameter estimate or model input), the threshold value at which the model s predictions changed was determined. Results Base-Case Analysis The results of the base-case analysis are presented in Table 2. Because the hypothetic cohort was an average-risk population and the duration of follow-up evaluation was limited (only 3 y), the percentage of patients who died or developed cancer in the simulations was relatively small. Consequently, much of the results are presented as numbers out of a hypothetic cohort of 100,000 patients for each strategy. The WAIT strategy of repeating a CTC examination in 3 years time resulted in 79 total deaths, all from cancer, whereas the COLO strategy of immediate colonoscopy resulted in 14 deaths, 4 from cancer and 10 as a result of colonoscopy complications. The net difference between the 2 management strategies at the end of 3 years was 65 deaths (of 100,000 patients), or 0.07%. The WAIT strategy resulted in 773 total cancers with the following stage distributions at the end of 3 years: 418 localized, 271 regional, and 86 distant. Of the 773 cancers, 372 (48%) were present initially in the adenomatous polyp, whereas the remaining 401 (52%) progressed to cancer over 3 years. In comparison, the COLO strategy resulted in 39 cancers with the following stage distributions at the end of 3 years: 19 localized, 15 regional, and 5 distant. This total does not include the cancers that were present within the adenomatous polyps that were found and removed during the colonoscopy (353 cancers that were considered cured). Of the 39 cancers, 19 were present initially in the adenomatous polyp, whereas the remaining 20 progressed to cancer from the adenoma over 3 years (these adenomatous polyps were not removed during the colonoscopy because the endoscopist was unable to find the polyp to perform a polypectomy). The resulting net difference in the total number of cancers was 734 more cancers in the WAIT strategy, or 0.73%. Sensitivity Analyses The results of the sensitivity analyses are presented in Table 3. The model was not sensitive to the estimates of the initial percentage of adenomas that were high risk, the initial percentage of adenomas that contained cancer, or the adenoma and cancer transition rates (ie, regardless of the estimate used for these parameters, the model continued to predict that the WAIT strategy would result in more deaths). The model was sensitive to the percentage of polyps that were hyperplastic, but only at values greater than 91.5%, which is well outside the published range. If the cancer mortality rates were decreased almost 8-fold to less than 13% of the base-case values, the COLO strategy resulted in more deaths. Other parameter estimates that the model was sensitive to at extreme values included CTC incorrect positive rates greater than 91%, the inability to find the polyp seen on CTC during colonoscopy at rates greater than 82%, and colonoscopy complication and death rates greater than 7.5-fold higher than the base-case estimate (0.75%). These analyses suggest that the model findings were robust because they were either not sensitive to parameter estimates at any value or only sensitive at extreme values that were well outside the range of the published literature and what seems clinically realistic, further strengthening our confidence in the model results. Discussion Our Markov model decision analysis compared the results of 2 management strategies for small (6 9 mm) polyps found on screening CTC examinations. The model showed that the WAIT strategy of repeating a CTC in 3 years would result in 65 more deaths (0.07%) and 734 (0.73%) more cancers in a cohort of 100,000 patients when compared with the COLO strategy of immediate colonoscopy for polypectomy. Furthermore, our model analysis was limited to 3 years, the interval before a repeat CTC examination in the WAIT strategy. With longer follow-up periods, the substantial difference in cancer prevalence between the 2 hypothetic cohorts could amplify the difference in death rates. An American Gastroenterological Association task force created to assess CTC and the issues that would define its impact reported the results of their findings and underscored the importance of polyp size in the future of CTC implementation. 4 A recently published study by Butterly et al 32 quantified the prevalence of advanced histology in small polyps with the aim of the investigation fueled, at least in part, by the advent of CTC screening. Model-based analyses have examined the cost effectiveness of CTC for screening 21,33,34 and the impact on the number of colonoscopies, 7 our study specifically addresses the management of small polyps found on a screening CTC examination. Although the best efforts were made to make our model as clinically realistic as possible, as with all disease models, ours is a simplification of reality. Many aspects of our model were intentionally kept simple so that our analysis would be readily comprehendible and would maintain a high level of transparency for anyone trying to understand or reproduce our analyses. For example, many reports suggest that polyp prevalence, and to a lesser extent cancer prevalence, is higher in men than women. 35,36 Instead of creating 2 models, male and female, we opted to use average values for both sexes in a single model. Our model also did not incorporate cost and other aspects found in cost-effectiveness analyses in which the primary end point is the incremental cost-effectiveness ratio. 37 The primary end points of our focused analysis, with a relatively short time horizon or follow-up period of only 3 years, were total number of deaths (either from cancer or colonoscopy complications) and cancers (including stage distribution). The aim of our study was to answer a focused clinical question, how to manage a small polyp found on CTC, thereby providing data for patients and care providers. A future model, constructed and analyzed from a societal perspective, 37 may be warranted. To model the specific clinical situation of interest in our study, we used transition probabilities for both adenomas and cancers that were derived from a separate natural history model

6 242 HUR ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 2 Table 3. Results of Sensitivity Analyses WAIT COLO Difference a Parameter/variable Deaths CAs Deaths CAs Deaths (%) CAs (%) Base case (0.07) 734 (0.73) Initial polyp characteristics % hyperplastic polyps (base 39.3%) 10% (0.01) 1089 (1.09) 20% (0.09) 968 (0.97) 60% (0.04) 484 (0.48) 91.5% (death threshold) b (0.10) 100% (0.01) 0 % adenomas high risk (base 5%) 0% (0.05) 460 (0.46) 1% (0.05) 515 (0.52) 10% (0.08) 1009 (1.01) 20% (0.11) 1559 (1.56) % adenoma CA (base 0.9%) 0% (0.01) 382 (0.38) 0.5% (0.04) 581 (0.58) 2% (0.14) 1165 (1.17) 5% (0.33) 2340 (2.34) Polyp/CA natural history Transition rates factor c (base 100%) 1% (0.01) 356 (0.36) 10% (0.01) 384 (0.38) 50% (0.03) 523 (0.52) 150% (0.12) 987 (0.99) 200% (0.20) 1279 (1.28) CA mortality rates factor c (base 100%) 5% (0.01) 13% (death threshold) % (0.03) 150% (0.10) 200% (0.13) CTC and colonoscopy characteristics CTC incorrect positive rate (base 31.9%) 5% (0.10) 1025 (1.03) 20% (0.08) 863 (0.86) 40% (0.06) 647 (0.65) 91% (death threshold) (0.10) 100% (0.01) 0 Colonoscopy complication rate (base 0.1%) 0.01% (0.07) 0.2% (0.06) 0.5% (0.03) 0.75% % (0.03) Unable to find polyp (base 5%) 1% (0.07) 765 (0.77) 10% (0.06) 696 (0.70) 20% (0.05) 619 (0.62) 50% (0.03) 387 (0.39) 82% (death threshold) (0.14) NOTE. All numbers are out of a starting pool of 100,000 patients. a Negative value signifies that the colonoscopy strategy resulted in a greater number. b Parameter value that results in the same number of deaths using both strategies. c Base-case rates used were varied by a multiplicative percentage factor. that was constructed and calibrated using data from the general population. The underlying assumption was that adenomas and cancers in this specific scenario would behave in a biologically similar manner to those found in the general population. So, for example, because all the polyps in our model started out between 6 and 9 mm in size, the underlying assumption was that cancers in smaller polyps behave similarly to those in larger polyps. We are not aware of any data to refute this assumption, however, there are little data regarding the biology of cancers found in smaller polyps.

7 February 2007 RISKS AND BENEFITS OF SMALL POLYP REMOVAL 243 Some effects related to CTC such as the risks of radiation exposure 38 and the effect of extracolonic findings 39,40 were not included in the analysis because these factors have not been proven conclusively to have a clinical impact. Our model did not incorporate de novo colorectal cancers that do not follow the polyp-carcinoma model or the effects of flat adenomas. There are little published data on which to model the natural history of these disease processes. Perhaps even more importantly, if we assume that colonoscopy has relatively little impact on these processes, then these adenomas and cancers would be irrelevant to our results because our primary end points are a relative comparison between the 2 strategies. A comprehensive model that tracks the natural history of polyps from the first dysplastic cell throughout an individual s life would be beneficial in analyzing the effectiveness of management strategies for diminutive ( 5 mm) polyps as well as numerous other CRC screening and surveillance issues. The difficulty in creating such a natural history is the dearth of clinical data on which to calibrate and construct a polyp model. More clinical trials and studies in this area would provide much needed data. Our model suggests that the WAIT strategy will result in more deaths and cancers than the COLO strategy. Although these findings are the results of a disease model and not of a randomized controlled trial, particularly in the absence of contradictory data, these results should be considered by both patients and physicians as they make decisions regarding how to manage a smaller polyp found on a screening CTC examination. Our analysis did not assess the cost effectiveness of the management strategies, and the cost of achieving the improvements in the immediate colonoscopy strategy may be prohibitive from a policy perspective. In conclusion, managing smaller polyps detected on a screening CTC with another CTC examination 3 years later likely will result in more deaths and cancers than immediate colonoscopy and polypectomy. This is a result of the presence of cancer within 6- to 9-mm adenomas at the time of the initial screen, and the presence of advanced pathology that will develop into invasive cancer during the 3-year observation period. References 1. Fenlon HM, Nunes DP, Schroy PC 3rd, et al. A comparison of virtual and conventional colonoscopy for the detection of colorectal polyps. N Engl J Med 1999;341: Pickhardt PJ, Choi JR, Hwang I, et al. Computed tomographic virtual colonoscopy to screen for colorectal neoplasia in asymptomatic adults. N Engl J Med 2003;349: Cotton PB, Durkalski VL, Pineau BC, et al. Computed tomographic colonography (virtual colonoscopy): a multicenter comparison with standard colonoscopy for detection of colorectal neoplasia. JAMA 2004;291: van Dam J, Cotton P, Johnson CD, et al. AGA future trends report: CT colonography. Gastroenterology 2004;127: Rockey DC, Paulson E, Niedzwiecki D, et al. Analysis of air contrast barium enema, computed tomographic colonography, and colonoscopy: prospective comparison. Lancet 2005;365: Angtuaco TL, Banaad-Omiotek GD, Howden CW. Differing attitudes toward virtual and conventional colonoscopy for colorectal cancer screening: surveys among primary care physicians and potential patients. Am J Gastroenterol 2001;96: Hur C, Gazelle GS, Zalis ME, et al. An analysis of the potential impact of computed tomographic colonography (virtual colonoscopy) on colonoscopy demand. Gastroenterology 2004;127: Schoenfeld P. Small and diminutive polyps: implications for colorectal cancer screening with computed tomography colonography. Clin Gastroenterol Hepatol 2006;4: Zalis ME, Barish MA, Choi JR, et al. CT colonography reporting and data system: a consensus proposal. Radiology 2005;236: Beck JR, Pauker SG. The Markov process in medical prognosis. Med Decis Making 1983;3: Sonnenberg FA, Beck JR. Markov models in medical decision making: a practical guide. Med Decis Making 1993;13: Winawer S, Fletcher R, Rex D, et al. Colorectal cancer screening and surveillance: clinical guidelines and rationale update based on new evidence. Gastroenterology 2003;124: Walsh JM, Terdiman JP. Colorectal cancer screening: scientific review. JAMA 2003;289: Mulhall BP, Veerappan GR, Jackson JL. Meta-analysis: computed tomographic colonography. Ann Intern Med 2005;142: Rex DK, Cutler CS, Lemmel GT, et al. Colonoscopic miss rates of adenomas determined by back-to-back colonoscopies. Gastroenterology 1997;112: Hixson LJ, Fennerty MB, Sampliner RE, et al. Prospective blinded trial of the colonoscopic miss-rate of large colorectal polyps. Gastrointest Endosc 1991;37: Pickhardt PJ, Nugent PA, Mysliwiec PA, et al. Location of adenomas missed by optical colonoscopy. Ann Intern Med 2004;141: DevCan: Probability of Developing or Dying of Cancer Software, Version 6.0. Bethesda, MD: Statistical Research and Applications Branch, National Cancer Institute, Ries LAG, Kosary CL, Hankey BF, et al. SEER cancer statistics review, Bethesda, MD: National Cancer Institute. Available: Frazier AL, Colditz GA, Fuchs CS, et al. Cost-effectiveness of screening for colorectal cancer in the general population. JAMA 2000;284: Ladabaum U, Song K. Projected national impact of colorectal cancer screening on clinical and economic outcomes and health services demand. Gastroenterology 2005;129: Chen CD, Yen MF, Wang WM, et al. A case-cohort study for the disease natural history of adenoma-carcinoma and de novo carcinoma and surveillance of colon and rectum after polypectomy: implication for efficacy of colonoscopy. Br J Cancer 2003;88: Blatt LJ. Polyps of the colon and rectum: incidence and distribution. Dis Colon Rectum 1961;4: 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: Johnson DA, Gurney MS, Volpe RJ, et al. A prospective study of the prevalence of colonic neoplasms in asymptomatic patients with an age-related risk. Am J Gastroenterol 1990;85: Vatn MH, Stalsberg H. The prevalence of polyps of the large intestine in Oslo: an autopsy study. Cancer 1982;49: DiSario JA, Foutch PG, Mai HD, et al. Prevalence and malignant potential of colorectal polyps in asymptomatic, average-risk men. Am J Gastroenterol 1991;86: Eide TJ, Stalsberg H. Polyps of the large intestine in Northern Norway. Cancer 1978;42: Williams AR, Balasooriya BA, Day DW. Polyps and cancer of the large bowel: a necropsy study in Liverpool. Gut 1982;23: Rickert RR, Auerbach O, Garfinkel L, et al. Adenomatous lesions of the large bowel: an autopsy survey. Cancer 1979;43: Ries LAG, Eisner MP, Kosary CL, et al. SEER cancer statistics

8 244 HUR ET AL CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 5, No. 2 review, Bethesda, MD: National Cancer Institute. Available: Butterly LF, Chase MP, Pohl H, et al. Prevalence of clinically important histology in small adenomas. Clin Gastroenterol Hepatol 2006;4: Sonnenberg A, Delco F, Bauerfeind P. Is virtual colonoscopy a cost-effective option to screen for colorectal cancer? Am J Gastroenterol 1999;94: Ladabaum U, Song K, Fendrick AM. Colorectal neoplasia screening with virtual colonoscopy: when, at what cost, and with what national impact? Clin Gastroenterol Hepatol 2004;2: Imperiale TF, Wagner DR, Lin CY, et al. Risk of advanced proximal neoplasms in asymptomatic adults according to the distal colorectal findings. N Engl J Med 2000;343: Schoenfeld P, Cash B, Flood A, et al. Colonoscopic screening of average-risk women for colorectal neoplasia. N Engl J Med 2005; 352: Weinstein MC, Siegel JE, Gold MR, et al. Recommendations of the panel on cost-effectiveness in health and medicine. JAMA 1996;276: Brenner DJ, Georgsson MA. Mass screening with CT colonography: should the radiation exposure be of concern? Gastroenterology 2005;129: Gluecker TM, Johnson CD, Wilson LA, et al. Extracolonic findings at CT colonography: evaluation of prevalence and cost in a screening population. Gastroenterology 2003;124: Chin M, Mendelson R, Edwards J, et al. Computed tomographic colonography: prevalence, nature, and clinical significance of extracolonic findings in a community screening program. Am J Gastroenterol 2005;100: Yee J, Akerkar GA, Hung RK, et al. Colorectal neoplasia: performance characteristics of CT colonography for detection in 300 patients. Radiology 2001;219: Pineau BC, Paskett ED, Chen GJ, et al. Virtual colonoscopy using oral contrast compared with colonoscopy for the detection of patients with colorectal polyps. Gastroenterology 2003;125: Johnson CD, Harmsen WS, Wilson LA, et al. Prospective blinded evaluation of computed tomographic colonography for screen detection of colorectal polyps. Gastroenterology 2003;125: Imperiale TF, Ransohoff DF, Itzkowitz SH, et al. Fecal DNA versus fecal occult blood for colorectal-cancer screening in an averagerisk population. N Engl J Med 2004;351: Viiala CH, Zimmerman M, Cullen DJ, et al. Complication rates of colonoscopy in an Australian teaching hospital environment. Intern Med J 2003;33: Anderson ML, Pasha TM, Leighton JA. Endoscopic perforation of the colon: lessons from a 10-year study. Am J Gastroenterol 2000;95: Hixson LJ, Fennerty MB, Sampliner RE, et al. Prospective study of the frequency and size distribution of polyps missed by colonoscopy. J Natl Cancer Inst 1990;82: Tsuda S, Veress B, Toth E, et al. Flat and depressed colorectal tumours in a southern Swedish population: a prospective chromoendoscopic and histopathological study. Gut 2002;51: O Brien MJ, Winawer SJ, Zauber AG, et al. The National Polyp Study. Patient and polyp characteristics associated with highgrade dysplasia in colorectal adenomas. Gastroenterology 1990; 98: Shinya H, Wolff WI. Morphology, anatomic distribution and cancer potential of colonic polyps. Ann Surg 1979;190: Gschwantler M, Kriwanek S, Langner E, et al. High-grade dysplasia and invasive carcinoma in colorectal adenomas: a multivariate analysis of the impact of adenoma and patient characteristics. Eur J Gastroenterol Hepatol 2002;14: Address requests for reprints to: Chin Hur, MD, MPH, 101 Merrimac Street, 10th Floor, Boston, Massachusetts chur@ mgh-ita.org Supported in part by grants from the American Gastroenterological Association s Research Scholar Award (C.H.) and National Cancer Institute grant K07CA (C.H.). The authors wish to thank Dr Amy Knudsen for her thoughtful suggestions, review, and comments on this study.