|Home | About | Journals | Submit | Contact Us | Français|
Centers for Medicare and Medicaid Services (CMS) considered whether to reimburse stool DNA testing for colorectal cancer screening among Medicare enrollees.
To evaluate the conditions under which stool DNA testing could be cost-effective compared with the colorectal cancer screening tests currently reimbursed by CMS.
Comparative microsimulation modeling study using two independently-developed models.
Derived from literature.
65-year-old (Medicare eligible) individuals; 50-year old individuals as sensitivity analysis.
Stool DNA test every 3 or 5 years in comparison to currently-recommended colorectal cancer screening strategies.
Life expectancy, lifetime costs, incremental cost-effectiveness ratios, threshold costs.
Assuming a cost of $350 per test, strategies of stool DNA testing every 3 or 5 years yielded fewer life-years and higher costs than the currently recommended colorectal cancer screening strategies.
Screening with the stool DNA test would be cost-effective at per-test cost of $40 to $60 for 3-yearly stool DNA testing, depending on the simulation model used. There were no levels of sensitivity and specificity for which stool DNA testing would be cost-effective at its current cost of $350 per test. Stool DNA testing at 3-yearly intervals would be cost-effective at a cost of $350 per test if the relative adherence with stool DNA testing were at least 50% better than with other screening tests.
None of the above mentioned results changed significantly when considering a 50-year old cohort.
We did not model other pathways than the traditional adenoma-carcinoma sequence.
Only if a significant reduction can be made to the test cost or if its availability would entice a large fraction of otherwise unscreened persons to be screened will stool DNA testing be a cost-effective alternative for colorectal cancer screening.
Agency for Healthcare Research and Quality
Recently, new studies have been published on test performance characteristics of stool DNA testing for colorectal cancer screening.(1–2) Although the recent results by Ahlquist et al were not very promising with 20% and 46% sensitivity for screen-relevant neoplasms for two different stool DNA tests,(1) some believe that reasons for optimism remain.(3) Itzkowitz et al. validated a new marker for stool DNA testing and found a sensitivity of 83% for cancer (2) and further technological advances are expected for stool DNA testing. However, stool DNA testing should only be offered if it is both effective and cost-effective compared to the currently-recommended screening tests.
In August 2007, the Centers for Medicare and Medicaid Services (CMS) opened a National Coverage Determination (NCD) to determine whether a stool DNA test, PreGen-Plus™, should be covered as an option for colorectal cancer screening among average-risk Medicare enrollees on a national basis (http://www.cms.gov/mcd/viewdecisionmemo.asp?id=212). PreGen-Plus™ consists of a panel of 23 molecular markers associated with colorectal cancer. The test analyzes the DNA for 21 specific point alterations in the APC, K-ras and p53 genes, a marker for microsatellite instability known as BAT-26, and a novel marker known as DNA Integrity Assay (DIA®), all of which have been associated with the presence of cancer. In response to this NCD, 2 colorectal cancer modeling groups from the National Cancer Institute’s (NCI) Cancer Intervention and Surveillance Modeling Network (CISNET) were asked to perform a cost-effectiveness analysis of screening with this stool DNA test among the average-risk Medicare population. The objective was to identify the reimbursement rate at which this stool DNA test could be cost-effective compared with the colorectal cancer screening tests currently reimbursed by CMS.
We evaluated the cost-effectiveness of stool DNA testing using two existing independently-developed microsimulation models of the CISNET consortium: MISCAN (Microsimulation Screening Analysis, from Erasmus University Medical Center and Memorial Sloan-Kettering Cancer Center), and SimCRC (the Simulation Model of Colorectal Cancer, from the University of Minnesota and Massachusetts General Hospital).(4–7)
Appendix 1 describes the MISCAN and SimCRC models and standardized profiles of each model’s structure, underlying assumptions, and calibration methods are available at http://cisnet.cancer.gov/profiles/. Briefly, each model simulates the life histories of a large population of individuals from birth to death and has a natural history component that tracks the progression of underlying colorectal disease in the absence of screening. As each simulated individual ages, there is a chance that one or more adenomas may develop depending on age, sex, race and individual risk. Adenomas can progress from small (≤5 mm) to medium (6–9 mm) to large (≥10 mm) size, and some may eventually become malignant. A preclinical cancer (i.e., not detected) has a chance of progressing through stages I to IV and may be detected by symptoms at any stage. With screening, adenomas and preclinical cancers may be detected depending on the sensitivity of the test for that lesion and, for endoscopic tests, whether the lesion is within reach of the endoscope.
The natural history model outcomes were calibrated to pre-screening data from autopsy studies (8–18) and clinical incidence data from the Surveillance, Epidemiology, and End-Results (SEER) Program before the introduction of screening (1975–1979).(19) The models use all-cause mortality estimates from the US life tables and colorectal cancer survival data from SEER (1996–1999). Both models have been validated against the long-term reductions in incidence and mortality of colorectal cancer with annual FOBT reported in the Minnesota Colon Cancer Control Study (20–22) and show good concordance with the trial results. The outcomes predicted by the natural history models for individuals at age 65 are compared in Appendix 2.
The National Coverage Determination requested that we evaluate the health effects and costs associated with stool DNA testing every 5 years. In the light of the similarities between FOBT and stool DNA screening we also evaluated a shorter interval of 3 years. We compared the health effects and costs of both strategies with those from the screening strategies currently covered by Medicare (23) and included in most colorectal cancer screening guidelines:(24–27) annual FOBT; flexible sigmoidoscopy every 5 years; flexible sigmoidoscopy every 5 years in conjunction with annual and 3-yearly FOBT; and colonoscopy every 10 years. We considered 3 FOBTs (Hemoccult II, Hemoccult SENSA, and immunochemical FOBT) and 2 strategies for sigmoidoscopy (with and without biopsy). In the strategy of sigmoidoscopy with biopsy, all detected polyps are biopsied and only persons with an adenomatous polyp are referred for a follow-up colonoscopy. With sigmoidoscopy without biopsy, all patients with detected polyps are directly referred for colonoscopy. In our primary or “base-case” analysis, we assumed all individuals begin colorectal cancer screening at age 65 and stop at age 80.
We assumed that an individual with a positive FOBT, sigmoidoscopy, or stool DNA test would be referred for follow-up colonoscopy and, if negative, would undergo subsequent screening with colonoscopy every 10 years. Individuals with adenomas that were detected and removed by colonoscopy (screening or diagnostic) were assumed to undergo colonoscopy surveillance per guidelines (i.e., every 3 years among individuals with an adenoma 10 mm or larger or with 3 or more adenomas of any size detected at the last colonoscopy, and every 5 years otherwise).(28) We assumed that surveillance continued until the diagnosis of colorectal cancer or death. For the base-case analysis, we assumed individuals were 100% adherent with the screening test of interest and with the recommended follow-up and surveillance; alternative adherence assumptions were explored in a sensitivity analysis.
Test characteristics were based on literature review (Table 1). For the stool DNA test, we used the sensitivity for cancer and specificity based on PreGen-Plus (version 1.1).(32) We used studies of the older version of the stool DNA test (version 1.0) to estimate the sensitivity of the test for detecting adenomas (29–31) as the estimates were not available for version 1.1. We further assumed that adenomas smaller than 5 mm were not detectable by the stool DNA test, but could be detected as a false-positive result based on the lack of specificity of the test. Patients undergoing colonoscopy and sigmoidoscopy were assumed to be at risk of serious complications (Table 2). There are no complications from FOBT or stool DNA tests.
The base-case cost-effectiveness analysis was conducted from the CMS perspective. Screening test costs were based on Medicare payments in 2007 (Table 1).(46) For the stool DNA test, we used a private insurer reimbursement of $350 as a base case.(33) The costs of complications were based on the relevant diagnosis-related group codes.(46) (Table 2) Net costs of colorectal cancer-related care were obtained from an analysis of 1998–2003 SEER-Medicare linked data (47) (personal communication, Robin Yabroff, Ph.D. and Martin Brown, Ph.D) and were updated to 2007 dollars using the overall Consumer Price Index. Costs from that period do not reflect the use of the expensive monoclonal antibodies bevacizumab and cetuximab. The costs vary by stage at diagnosis and phase of care (Table 2).
We used the simulation models to calculate the lifetime costs and life expectancy for a previously unscreened cohort of 65-year-old Medicare beneficiaries under 15 competing strategies, including no screening. We conducted an incremental cost-effectiveness analysis from the perspective of CMS and discounted future costs and life years 3% annually to account for time preferences for present over future outcomes.(48) Strategies that were more costly and less effective than another strategy were ruled out by strong dominance. Strategies that were more costly and less effective than a combination of other strategies were ruled out by weak dominance. The relative performance of the remaining strategies was measured using the incremental cost-effectiveness ratio (ICER), defined as the additional cost of a specific strategy, divided by its additional clinical benefit, compared with the next least expensive strategy. All non-dominated strategies represent the set of potentially cost-effective options (depending upon the willingness to pay for a life-year gained) and lie on the efficient frontier.
If the stool DNA strategies were found to be dominated by the currently-reimbursed screening options, for each DNA tool strategy, we calculated the maximum cost per stool DNA test (i.e., the threshold cost) for that strategy to lie on the efficient frontier (i.e., be cost-effective). Second, because the stool DNA test is still evolving,(3) we identified the threshold stool DNA costs for scenarios in which the diagnostic performance of the stool DNA test was improved. The base-case estimates of the sensitivities for small, medium, and large adenomas and for cancer, as well as the estimate for specificity were increased by 10%, 25%, 50%, 75% and 100% of the difference between the base-case values and perfect sensitivity and specificity. Finally, since some have suggested that the availability of a stool DNA test as an option for colorectal cancer screening might entice a previously unscreened individual to undergo screening, we also identified threshold stool DNA costs for scenarios in which we allowed the adherence of stool DNA strategies to be greater than that of all other screening strategies. For this analysis we assumed an overall adherence rate of 57% for each test (i.e., the percent of Medicare-eligible individuals who were adherent with colorectal cancer screening recommendations in the 2005 National Health Interview Survey)(49) with this 57% of the population completely adherent to screening and the remainder completely non-adherent. Modeling adherence in this fashion allowed us to evaluate the impact of enhancing screening participation among a previously unscreened segment of the population. We varied the adherence for a stool DNA strategy from 57% to 100%. Subsequently, threshold costs for stool DNA were calculated comparing overall costs and life-years saved using stool DNA at these higher adherence rates to competing strategies at an adherence rate of 57%.
To evaluate how the estimated stool DNA threshold costs were influenced by our assumptions, we also identified threshold costs in sensitivity analyses with alternative assumptions. First, we explored the effect of a screening interval of 1 year for stool DNA testing. Second, we considered alternative stool DNA test versions. We evaluated version 1.0 (the only stool DNA test that has been evaluated in a general population setting) (1, 29) and a newer version of the stool DNA test (version 2.0)(39) (see Table 1 for test characteristics).
Next, we repeated our base-case threshold analysis for a cohort of 50-year-olds with screening beginning at age 50 as recommended in colorectal cancer screening guidelines (24–27). Finally, we conducted an analysis from a modified societal perspective by including direct costs borne by beneficiaries as well as patient time costs. We did not incorporate productivity costs. Cost inputs for the modified societal perspective are shown in Tables 1 and and22.
This research was supported by the Agency for Healthcare Research and Quality (HHSP233200700123P, HHSP233200700196P, HHSP233200700350P) and the NCI (U01-CA-088204, U01-CA-097426, and U01-CA-115953). The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, and approval of the manuscript.
In the absence of screening, the two models project that 57 out of every 1,000 65-year-old individuals will be diagnosed with colorectal cancer in their lifetimes (Table 3). With screening, many of these colorectal cancer cases can be prevented. Assuming 100% adherence, the reduction in the lifetime risk of colorectal cancer ranged from 32–40% with annual Hemoccult II screening to 53–72% with 10-yearly colonoscopy screening (the reported ranges reflect differences in projections by model). CRC risk reduction with stool DNA testing was similar to that of Hemoccult II and varied from 30–49% depending on the simulation model and the interval used.
Figure 1 shows a plot of total life-years gained (compared with no screening) and total lifetime direct medical costs from the Medicare perspective for each of the 18 screening strategies, and the efficient frontier: the economically rational subset of choices. Five-yearly stool DNA testing was the least effective of all evaluated screening strategies, while the 3-yearly strategy was only slightly more effective than annual Hemoccult II. Assuming reimbursement at $350 per test, both stool DNA testing strategies were the most expensive screening strategies, lying to the far right of the cost-efficient frontier. Although dominated by the currently-recommended screening options, the costs per life-year gained compared with no screening were less than $15,000 for both screening intervals and models (Appendix 3).
Threshold analyses indicated that to be cost-effective, the stool DNA test would need to cost between $34 and $51 when performed every 5 years, or between $40 and $60 when performed every 3 years, depending on the simulation model used. Analysis of 3-yearly and 5-yearly stool DNA testing with the SimCRC model identified no sensitivity and specificity estimates for which the threshold value of the cost of the stool DNA test could be greater than its base-case value of $350 and still lie on the efficient frontier (Figure 2). With the MISCAN model, the cost of the stool DNA test may rise to $364 if the test is perfect with respect to sensitivity and specificity and if offered every 5 years.
Analyses with the MISCAN model showed that adherence with stool DNA testing has to be almost 50% better than with other tests for 3-yearly stool DNA testing to be on the frontier at the base-case cost of $350 (Figure 3). With the SimCRC model the relative adherence with stool DNA testing had to be between 50% and 75% better than adherence with other modalities for 3-yearly stool DNA testing to be on the frontier at the base-case cost of $350.
Offering DNA testing annually did not change the threshold costs ($40–$48). There was no threshold cost at which the stool DNA testing with version 1.0 could be a cost-effective alternative for colorectal cancer screening. For version 2.0 threshold costs were $2–$31, which are lower than the base-case threshold costs. Only from a modified societal perspective did the threshold costs of the stool DNA test (excluding co-payments and patient time costs) increase somewhat compared with the base-case estimate: $88–$134 for the 5-year interval and $73–$116 for the 3-year interval. The higher frequency of FOBT scenarios results in considerably higher additional time costs than with stool DNA screening, allowing for higher per-test costs for the stool DNA test.
None of the above mentioned results changed substantially when considering a 50-year old cohort instead of the 65-year old Medicare eligible cohort. The threshold costs would have to fall to $27–$52 for stool DNA testing to be on the efficient frontier.
Our analysis showed that 3- and 5-yearly stool DNA testing were both more costly and less effective than annual screening with a sensitive FOBT. Screening with a stool DNA test would be an efficient strategy at a per-test cost of $34–$60, depending on the screening interval (3 or 5 years) and model used. Only if the relative adherence with stool DNA testing were 50% better than other screening tests, would the test be cost-effective at current test costs. These results hold both for the Medicare-eligible population, as well as the general screening population. The fact that two independently-developed models come to similar conclusions with respect to cost-effectiveness and threshold costs of stool DNA screening shows the robustness of the results for model uncertainties, particularly pertaining to the natural history of colorectal disease.
We had anticipated that stool DNA testing would be dominated by screening with a sensitive FOBT given that the stool DNA test has similar sensitivity and specificity as Hemoccult SENSA with a cost that is almost 80 times greater. Consequently the aim of our analysis was to explore the conditions under which stool DNA testing could compete with the existing screening tests. We have only explored the potential of stool DNA testing for colorectal cancer screening. The test might also have prognostic or even treatment implications, for example through risk stratification based on genetic markers or by guiding the use of genomic-targeted therapies. However, in this case the test becomes a diagnostic tool that can be administered at the time of diagnosis of colorectal cancer rather than a pre-symptomatic screening test. The evaluation of the cost-effectiveness of a stool DNA test as a diagnostic test is beyond the scope of our analysis.
Because stool DNA testing is still evolving,(3) we evaluated threshold costs for improved test characteristics. Such improvements can be expected, given technological advances in the form of more sensitive polymerase chain reaction strategies.(3) However, the threshold costs for the latest published version of the stool DNA test (version 2.0),(39) are lower than the threshold costs for the base-case stool DNA test (version 1.1) due to the many unnecessary colonoscopies brought about by the considerably lower specificity. Even under the extreme assumption of perfect sensitivity and specificity, the threshold cost for the stool DNA test remained below $350 at intervals of 3 or 5 years.
Substantially higher adherence with stool DNA testing would make stool DNA screening cost-effective at $350. However, while it has been shown that the stool DNA test is acceptable to patients who have already agreed to participate in a screening program,(51–52) there is no evidence that screening adherence with stool DNA testing would be substantially better than with other tests. Adherence needs to be 50% better than even FOBT and since both types of tests are non-invasive, this is unlikely. In the absence of this adherence benefit, stool DNA remains dominated.
Stool DNA testing is currently included in the American Cancer Society (ACS) guidelines for colorectal cancer screening (24). Twelve US states and the District of Columbia have legislative mandates requiring that certain insurers should offer all screening options of the current ACS guidelines. As a consequence, coverage of stool DNA screening is now mandated in these states.(53) For every 65-year-old person switching from Hemoccult SENSA or colonoscopy screening to stool DNA screening, colorectal cancer screening costs would increase on average by $750–$1250 (results not shown), while the life-years saved would on average decrease. In the US, there currently are 2.6 million 65-year olds,(54) so on a national level stool DNA screening could potentially lead to an unnecessary expenditure of $3 billion per year. These figures would be even higher, when considering the complete target population for screening from age 50.
Our findings are comparable to two published cost-effectiveness analysis of stool DNA screening.(55–56) Like ours, both analyses concluded that the stool DNA test was dominated by currently recommended colorectal cancer screening tests. Wu et al. found threshold costs that were slightly higher ($57–$70) than the threshold costs in this analysis. However, those threshold costs were based on a willingness to pay of $13,000 per life-year gained compared to no screening. Song et al. found threshold costs of $195 when comparing 2-yearly stool DNA testing to 10-yearly colonoscopy and assuming considerably higher colonoscopy costs. A similar comparison in the MISCAN and SimCRC models yielded threshold costs of $205–$213.
There are several limitations of the models. First, the models simulate the progression from adenoma to colorectal cancer by increasing the size of the adenomas over time. Because adenoma size and the presence of villous components or high-grade dysplasia are highly correlated,(57) size indirectly represents histology and grade. However, neither model separately simulates the step from adenoma with low-grade dysplasia to an adenoma with high-grade dysplasia. If the advantage of a stool DNA test is detection of a smaller adenoma at the stage of high-grade dysplasia, we may underestimate its effectiveness. Second, we assumed that all colorectal cancers arise through the traditional adenoma-carcinoma sequence, with a linear sequence of mutations in the APC, KRAS and TP53 genes. Recent data indicate the probable existence of at least one alternative pathway to colorectal cancer through a mutation of the BRAF gene.(58) Existence of different pathways will probably not influence the performance of FOBT because bleeding of a lesion is unlikely to be related to the pathway. It may influence sensitivity of endoscopy, as lesions from this pathway are more likely to be proximal and sessile or flat and therefore more difficult to find. However for the stool DNA test, the lesion in question may have acquired a gene mutation not assessed by the test. In this case, a person with a false-negative result on such a test will have a higher than average probability of having a negative test with subsequent screens. Consequently, we may have overestimated the benefit and the threshold cost of this test.
In conclusion, our analysis shows that future developments of the stool DNA test should not only focus on improving test characteristics but also on reducing test cost. The first stool DNA assay that reached the market was expensive ($795); a more recent stool assay for vimentin methylation alone was introduced this year at a cost of $220.(3) These numbers offer hope that further technological refinements will permit significant cost reductions. We are currently lacking good information on the performance characteristics of these new tests. When the performance levels of newer versions of the stool DNA test become available, the results of our sensitivity analysis can be used to determine their cost-effectiveness. If the cost of the test is higher than the threshold costs associated with the level of performance of the new test, it will not be cost-effective. Our analysis shows, that improving tests characteristics alone is insufficient to make stool DNA testing cost-effective. Without further cost reductions, stool DNA screening will not be a cost-effective alternative for average-risk colorectal cancer screening in the Medicare population or the general screening population.
Models are available to approved individuals with written agreement.
This research was supported by the Agency for Healthcare Research and Quality (AHRQ) (HHSP233200700123P, HHSP233200700196P, HHSP233200700350P) and the National Cancer Institute (U01-CA-088204, U01-CA-097426, and U01-CA-115953). The findings and conclusions in this article are those of the authors who are responsible for its contents; the findings and conclusions do not necessarily represent the views of AHRQ. Therefore, no statement in this article should be construed as an official position of AHRQ or of the U.S. Department of Health and Human Services.
We acknowledge Martin Brown, PhD and Robin Yabroff, PhD of the National Cancer Institute (NCI) for their assistance with obtaining cancer treatment costs using SEER-Medicare data; Joan Warren, PhD and Carrie Klabunde, PhD of NCI for sharing their preliminary analysis of SEER-Medicare data on colonoscopy-related complications; John Allen, MD of Minnesota Gastroenterology and Joel Brill, MD of Predictive Health for their assistance in deriving coding for screening and complications; William Larson, Marjorie Baldo, and Marilu Hu of the Centers for Medicare and Medicaid Services (CMS) for providing CMS cost data; Chuck Shih of the Agency of Healthcare Research and Quality for interpreting the CMS cost data; William Lawrence, MD and Kim Wittenberg, MA of AHRQ for contextual and administrative assistance, respectively; and Eric (Rocky) Feuer, PhD of the NCI for continued support of the work and infrastructure of the CISNET consortium.
|Adenoma prevalence, age 65: 39.8%||Adenoma prevalence, age 65: 37.1%|
|Number of adenomas per 1000 by site and size, age 65||Number of adenomas per 1000 by site and size, age 65|
|Proximal colon||121.2||69.9||61.8||Proximal colon||171.8||185.8||23.9|
|Distal colon||134.4||77.4||68.4||Distal colon||123.9||18.3||41.4|
|Distribution of adenomas by site and size, age 65 (%)||Distribution of adenomas by site and size, age 65 (%)|
|Proximal colon||15||9||8||31||Proximal colon||28||31||4||63|
|Distal colon||17||10||8||35||Distal colon||20||3||7||30|
|CRC incidence among cancer-free 65-year-old population, %||CRC incidence among cancer-free 65-year-old population, %|
|Stage 1||Stage 2||Stage 3||Stage 4||Total||Stage 1||Stage 2||Stage 3||Stage 4||Total|
|Strategy||Discounted costs ($)||Discounted LYG||CER ($)||ICER ($)||Discounted costs ($)||Discounted LYG||CER ($)||ICER ($)|
|HS (3y) + SIGB||2,798,156||83.1||1,006||d||2,106,646||77.9||CS||d|
|HS (3y) + SIG||2,857,191||83.7||1,704||d||2,143,225||77.0||CS||d|
|IFOBT (3y) + SIGB||2,932,676||82.9||2,631||d||2,148,703||77.5||CS||d|
|IFOBT (3y) + SIG||2,912,349||83.6||2,367||d||2,187,271||76.6||CS||d|
|HII + SIGB||2,793,800||84.1||941||20,800||2,127,300||79.0||CS||d|
|HII + SIG||2,840,500||84.6||1,488||d||2,113,600||80.2||CS||8,600|
|HS + SIGB||2,863,800||87.1||1,714||23,900||2,187,800||84.7||CS||d|
|HS + SIG||2,909,400||87.1||2,237||d||2,187,000||85.2||CS||14,600|
|IFOBT + SIGB||3,025,600||87.1||3,569||d||2,282,400||84.6||CS||d|
|IFOBT + SIG||2,992,800||87.2||3,190||924,800||2,283,000||85.1||CS||d|
|Stool DNA (3y)*||3,673,500||68.0||14,105||d||3,081,300||64.2||12,233||d|
|Stool DNA (5y)*||3,383,000||58.8||11,375||d||2,814,300||51.4||10,089||d|
LYG = life-years gained vs. no screening; ICER = incremental cost-effectiveness ratio; CER = cost-effectiveness ratio compared with no screening; HII = annual Hemoccult II, HS = annual Hemoccult SENSA; IFOBT = annual immunochemical fecal occult blood test; SIG = 5-yearly sigmoidoscopy without biopsy; SIGB = 5-yearly sigmoidoscopy with biopsy; HS (3y) = 3-yearly Hemoccult SENSA; iFOBT (3y) = 3-yearly immunochemical fecal occult blood test; COL = 10-yearly colonoscopy; d = dominated; --- indicates default strategy (i.e., the least costly and least effective non-dominated strategy); NA = not applicable
None of the investigators has any affiliations or financial involvements related to the material presented in this manuscript.