DRAFT: Not for Citation
We used three existing microsimulation models validated against the best available data (Loeve 1999, 2000, Frazier 2000, Knudsen 2005) to inform CMS and AHRQ in assessing the effectiveness and cost-effectiveness of CT colonography, in comparison with the currently-recommended CRC screening strategies. Although randomized controlled trials are the preferred method for establishing effectiveness of (screening) interventions, they are expensive and require long follow-up. Accordingly, well-validated microsimulation models may be used to estimate the required resources and expected benefits from different screening policies and inform decision making. The validity of the models is based on clinical incidence data before the introduction of screening (1975-1979 SEER data) and the size distribution of adenomas in colonoscopy and autopsy studies (Clark 1985, Blatt 1961, Arminski 1964, Vatn 1982, Jass 1992, Johannsen 1989, Bombi 1988, Williams 1982, Rickert 1979, Chapman 1963, Rutter 2007). The external validity has further been tested on the results of large (randomized) screening and surveillance studies, such as the Minnesota Colon Cancer Control Study (Mandel 1993), the CoCap sigmoidoscopy study (Doria-Rose 2004), and the National Polyp Study (Loeve 2000). The models also use common all-cause mortality estimates from the US life tables and colorectal cancer survival data from SEER (2004). Finally, the models were able to explain observed incidence and mortality trends in the US when accounting for risk factor trends, screening practice and chemotherapy treatment (Vogelaar 2006, Knudsen 2004, 2005). Using three models (i.e., a comparative modeling approach) adds credibility to the modeling results and serves as a sensitivity analysis on the underlying structural assumptions of the models, particularly pertaining to the natural history of colorectal disease. Through the NCI CISNET consortium, standardized profiles of the each model's structure and underlying assumptions are available at http://cisnet.cancer.gov/profiles/.
We used the MISCAN, SimCRC, and CRC-SPIN simulation models to calculate the lifetime costs (discounted and undiscounted) and life expectancy (discounted and undiscounted) for a cohort of 65-year-old individuals residing in the US (i.e., eligible for Medicare benefits) under 14 strategies plus no screening. The 14 CRC screening strategies vary by screening test or combination of tests and screening interval. We conducted an incremental cost-effectiveness analysis from the perspective of CMS and discounted future costs and life years 3% annually (Gold 1996). Strategies that were more costly and less effective were ruled out by simple dominance. Strategies that were more costly and less effective than a combination of other strategies were ruled out by weak dominance. In this report, dominance refers to either simple or weak dominance. The relative performance of the remaining strategies was measured using the incremental cost-effectiveness ratio, 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 (efficient) strategies define the efficient frontier and may be cost-effective depending on the willingness to pay for a life-year gained.
The MISCAN, SimCRC, and CRC-SPIN models simulate the life histories of a large population of individuals from birth to death. Each model has a natural history component that tracks the progression of underlying disease in the absence of screening. The models share many characteristics; they use similar model inputs and are calibrated to the same data regarding adenoma prevalence, cancer incidence, and stage distribution. These data were collected and processed as part of CISNET and can be considered the best-available data for informing the simulation models. As each simulated individual ages, there is a chance that an adenomatous polyp—a benign precursor lesion that may lead to CRC—develops. One or more adenomas can occur in any individual and each can develop into preclinical CRC (Figure 1). The risk of developing an adenoma depends on age, sex, genetic and other propensity factors. The models track the location in the colon and the size of each adenoma, which influence disease progression and the chance of being found by screening.
Adenomas can grow in size over time. Some adenomas eventually become malignant, transforming to stage I preclinical cancer. A preclinical cancer (i.e., not detected) has a chance of progressing through the stages (from stages I to IV) and may be detected by symptoms at any stage. We assume that adenomas are asymptomatic and can only be detected by a screening test.
To project the effectiveness of a screening strategy, the models incorporate a screening component together with the natural history model. The effectiveness of each screening test is modeled through each test's ability to detect lesions (i.e., adenomas, preclinical cancer). Once screening is introduced, a simulated person who has an underlying adenoma or preclinical cancer has a chance of having it detected during a screening year depending on the sensitivity of the test for that lesion. For screened persons without an underlying lesion we apply the false-positive rate (1—specificity) to determine whether or not that person will undergo an unnecessary follow-up examination. Hyperplastic polyps are not modeled explicitly but are reflected in the specificity of the test. In addition, a percentage of individuals with false-negative test results (i.e., adenoma or preclinical cancer present but not detected) will be referred to colonoscopy because of the detection of a hyperplastic polyp. Flexible sigmoidoscopy can only detect lesions located in the distal colon or rectum, while other tests have the ability to detect lesions in any part of the colorectal tract. Colonoscopy and to a lesser extent, CT colonography, are associated with a small mortality risk due to the risk of perforation during the procedure.
The models include the possibility of multiple adenomas or preclinical cancers. An individual with multiple adenomas, especially multiple adenomas of a larger size, would be more likely on average to be detected by screening than an individual with a single small adenoma. Consequently multiplicity and size of the adenomas, or whether there is a preclinical cancer, are included in estimates of sensitivity and specificity.
Although the models are calibrated to the same data on adenoma prevalence and cancer incidence, the underlying distributions of dwell times (i.e., the total time spent with adenoma and preclinical cancer prior to symptom detection) differ among the three models. A key assumption in the MISCAN model is that there are two types of adenomas: progressive adenomas (adenomas that eventually can become cancer) and non-progressive adenomas (adenomas that cannot become cancer). In the SimCRC and CRC-SPIN models all adenomas have the ability to progress to cancer (although most will not during the lifespan of the individual). An additional difference is that CRC-SPIN models continuous size rather than discrete stages of adenoma size. Although all three models predict similar estimates of adenoma prevalence and CRC incidence, the difference in the adenoma growth assumptions results in different dwell time estimates among the models. In the MISCAN model adenomas and preclinical cancer have been present for 10 years on average before clinical diagnosis, while the estimate is approximately 22 years for SimCRC and 25 years for CRC-SPIN. Little is known about how fast this progression truly occurs. It is estimated that 30% to 50% of the population have one or more adenomas, but it is difficult to measure dwell time in a real population because, by definition, it is the period during which the condition is undiagnosed. As a result of the difference in dwell time, more life-years are gained from screening in the SimCRC and CRC-SPIN models than in the MISCAN model. In the MISCAN model the additional benefit of increasing screening frequency will be greater than that in SimCRC and CRC-SPIN. A summary of each model is in Appendix 1.
Another key difference among the models is the distribution of adenomas in the colorectal tract (Appendix 2). In the MISCAN model, adenomas are assumed to have the same distribution as CRCs, while the SimCRC and CRC-SPIN models are calibrated to the distribution of adenomas from autopsy studies. Approximately 30% of CRCs are located in the rectum, while data from autopsy studies suggest that 8-10% of adenomas are located in the rectum. As a result of this difference, the MISCAN model finds strategies involving sigmoidoscopy to be more effective than the SimCRC and CRC-SPIN models, because a larger proportion of adenomas are within the reach of the sigmoidoscope.
We used the natural history models to estimate the distribution of underlying disease for the 65-year-old US population in 2005 in terms of the presence, location, size, and type (adenoma vs. preclinical cancer) of lesions (Appendix 2 for comparison of natural history models). We conducted an analysis of the effect of different screening strategies among a 65-year-old cohort of individuals who have never been screened as our base case. However this cohort with no prior screening represents a higher-risk group than a cohort of previously-screened 65-year-old individuals. As a comparison, we conduct a sensitivity analysis for a 50-year-old cohort.
In consultation with AHRQ and CMS, we compared CT colonography screening to the basic strategies of screening with FOBT every year, flexible sigmoidoscopy (SIG) every five years, combinations of FOBT and SIG, and colonoscopy every 10 years, which are recommended by the USPSTF (USPSTF 2008); the American Cancer Society (Smith 2006, Levin 2008), and the Multi-Society Task Force (Winawer 1997, 2003, 2006, Levin 2008). No screening was also considered. Although barium enema was included in the older screening recommendations for the USPSTF, it was not included in the newer recommendations and is not considered in this analysis. We evaluated three FOBTs: Hemoccult II (HII), Hemoccult SENSA (HS) and immunochemical FOBT (FIT) and two strategies for SIG (with and without biopsy).
Strategy |
Abbreviation |
Interval, test 1 (y) |
Interval, test 2 (y) |
Biopsy @ SIG? |
---|---|---|---|---|
No screening |
-- |
-- |
-- |
-- |
Hemoccult II |
HII |
1 |
-- |
-- |
Hemoccult SENSA |
HS |
1 |
-- |
-- |
Fecal immunochemical test |
FIT |
1 |
-- |
-- |
Flexible sigmoidoscopy |
SIGB |
5 |
-- |
yes |
Flexible sigmoidoscopy |
SIG |
5 |
-- |
no |
Hemoccult II, SIG |
HII + SIGB |
1 |
5 |
yes |
Hemoccult II, SIG |
HII + SIG |
1 |
5 |
no |
Hemoccult SENSA, SIG |
HS + SIGB |
1 |
5 |
yes |
Hemoccult SENSA, SIG |
HS + SIG |
1 |
5 |
no |
Fecal immunochemical test, SIG |
FIT + SIGB |
1 |
5 |
yes |
Fecal immunochemical test, SIG |
FIT + SIG |
1 |
5 |
no |
Colonoscopy |
COL |
10 |
-- |
-- |
-- indicates not applicable
We compared these screening strategies to CT colonography screening based on the test parameters of the DoD study (Pickhardt 2003) using 3-dimensional imaging as the primary read and the NCTC trial (Johnson 2008) using both 2D and 3D reads. Subjects with lesions 6 mm or larger detected by CT colonography were referred to colonoscopy. Those with no 6 mm or larger polyps detected had a repeat CT colonography in 5 years. The request for the NCD did not specify a repeat screening interval; we used a 5-year to rescreen (Levin 2008). In addition to these two base-case scenarios for CT colonography, we conducted a sensitivity analysis in which we explored CT colonography scenarios using primary 2D reads, referral of individuals with 10 mm or larger lesions for colonoscopy, and a 10-year interval for repeat screening (Table 2). We also considered a hypothetical worst-case scenario for CT colonography.
CT colonography strategy abbreviation |
Study |
Primary read |
Colonoscopy referral threshold (mm) |
Screening interval (y) |
---|---|---|---|---|
Strategies evaluated in the base-case analysis |
||||
CTC DoD 3D 6mm 5y |
DoD |
3D |
6 |
5 |
CTC NCTC 2D/3D 6mm 5y |
NCTC |
2D/3D |
6 |
5 |
Strategies evaluated in sensitivity analyses |
||||
CTC DoD 3D 6mm 10y |
DoD |
3D |
6 |
10 |
CTC DoD 3D 10mm 5y |
DoD |
3D |
10 |
5 |
CTC DoD 3D 10mm 10y |
DoD |
3D |
10 |
10 |
CTC DoD 2D 6mm 5y |
DoD |
2D |
6 |
5 |
CTC DoD 2D 6mm 10y |
DoD |
2D |
6 |
10 |
CTC DoD 2D 10mm 5y |
DoD |
2D |
10 |
5 |
CTC DoD 2D 10mm 10y |
DoD |
2D |
10 |
10 |
CTC NCTC 2D/3D 6mm 10y |
NCTC |
2D/3D |
6 |
10 |
CTC NCTC 2D/3D 10mm 5y |
NCTC |
2D/3D |
10 |
5 |
CTC NCTC 2D/3D 10mm 10y |
NCTC |
2D/3D |
10 |
10 |
CTC J 3D 10mm 5y |
J |
3D |
10 |
5 |
CTC J 3D 10mm 10y |
J |
3D |
10 |
10 |
CTC J 2D 10mm 5y |
J |
2D |
10 |
5 |
CTC J 2D 10mm 10y |
J |
2D |
10 |
10 |
CTC WC 2D/3D 6mm 5y |
WC |
2D/3D |
6 |
5 |
CTC WC 2D/3D 6mm 10y |
WC |
2D/3D |
6 |
10 |
CTC WC 2D/3D 10mm 5y |
WC |
2D/3D |
10 |
5 |
CTC WC 2D/3D 10mm 10y |
WC |
2D/3D |
10 |
10 |
CTC = computed tomography colonography; DoD = Department of Defense study (Pickhardt 2003, 2007a); NCTC = National CT Colonography Trial (Johnson 2008); J = Johnson study (Johnson 2007); WC = hypothetical worst case scenario
For the purposes of this report, we assumed that all individuals begin CRC screening at age 65 (i.e., the age at which Medicare eligibility begins) and end at age 80. Those with adenomas or colorectal cancer detected are assumed to have colonoscopic surveillance according to the Multi-Society guidelines (Winawer 2006, Levin 2008) and continue surveillance with no stopping age. The cohort was followed for their lifetimes to a maximum of age 100. The USPSTF has now recommended a stop age for CRC screening of age 75 (USPSTF 2008; Zauber 2008a). We used the stopping age of 80 in this report to be consistent with the DNA stool report and because we assume that screening doesn't begin until age 65. We would expect similar ranking of strategies for stop age of 75 as well as 80 given comparable adherence.
We assumed that any individual with a positive FOBT or a positive CT colonography (defined as the visualization of a lesion of size ≥6 mm) is referred for a follow-up colonoscopy. We evaluated two scenarios for flexible sigmoidoscopy: (1) all detected polyps are biopsied and any person with an adenomatous polyp is referred for a follow-up colonoscopy, and (2) all persons with detected polyps are directly referred for colonoscopy (i.e., no biopsy is performed). For the year in which both FOBT and flexible sigmoidoscopy are due, the FOBT is performed first and if positive, the subject is referred for colonoscopy. Flexible sigmoidoscopy is done only for those with a negative FOBT. If a follow-up colonoscopy is negative, then the subject is assumed to undergo subsequent screening with colonoscopy with a 10-year interval (as long as the repeat colonoscopy is negative) and does not return to the initial screening schedule, as is the recommendation of the US Multi-Society Task Force (Winawer 2006) and ACS (Levin 2008). In other words, once a person has a colonoscopy, the individual remains on a colonoscopy schedule.
If adenomas are detected on colonoscopy then the individual begins surveillance with colonoscopy per the 2006 guidelines from the joint publication of the US Multi-Society Task Force and the American Cancer Society (Winawer 2006; Rex 2006; Levin 2008). Individuals found with one or two adenomas that are both less than 10 mm in size will undergo colonoscopy surveillance every 5-10 years (5 years was used). Individuals with at least one adenoma greater than or equal to 10 mm in size or with 3 or more adenomas will undergo colonoscopy surveillance every 3 years unless the surveillance colonoscopy is normal or only detects one or two adenomas of size <1.0 cm, then the next surveillance colonoscopy would be at 5 years.
For the base-case analysis we assumed that all individuals are 100% adherent with screening, follow-up, and surveillance procedures. In sensitivity analysis we examined less than optimal adherence to determine if differences in adherence affect our results (see section on sensitivity analyses).
We specified a stop age of 80 for screening but allowed all individuals with an adenoma detected to continue to have surveillance colonoscopies until a diagnosis of CRC or death from other causes. All simulated individuals were followed until death (or age 100). The life-years gained per scenario were derived relative to no screening.
Table 3 contains an overview of test characteristics used in our analyses. For all strategies other than CT colonography, test characteristics were taken from those derived for our previous report on stool DNA screening (Zauber 2007). Test parameters are given by person for the FOBTs and by lesion for CT colonography, colonoscopy, and flexible sigmoidoscopy. We assume that the test performance characteristics for FOBTs and CT colonography are based on assessment of the whole colorectum. For sigmoidoscopy and colonoscopy, the test characteristics apply to the portion of the colorectum reached by the scope. We assumed that 80% of sigmoidoscopy examinations reach the junction of the sigmoid and descending colon and 40% reach the beginning of the splenic flexure. None reach beyond. For colonoscopy, we assumed that an average of 1.05 colonoscopies is performed per subject to obtain a "complete" assessment of the colorectum and that the cecum is reached in 98% of subjects.
The test characteristics for CT colonography (Table 3) are based on the literature review described above. As CT technology has changed rapidly, we used the sensitivity and specificity estimates from the two recent large-scale CT colonography screening trials (Pickhardt 2003, 2007a; Johnson 2008) for our base-case estimates. We did not combine the estimates from these two studies because of significant heterogeneity in the estimates for sensitivity for adenomas size 6-9 mm and for specificity. Other estimates were evaluated in sensitivity analyses (see section on sensitivity analyses below).
Test |
Sensitivitya by adenoma size or CRC (%) |
Specificity (%) | |||
---|---|---|---|---|---|
≤5 mm |
6-9 mm |
≥10 mm |
CRC |
||
Hemoccult II |
2.0 |
5.0 |
12.0 |
40.0 |
98.0 |
Hemoccult SENSA |
7.5 |
12.4 |
23.9 |
70.0 |
92.5 |
Fecal immunochemical test |
5.0 |
10.1 |
22.0 |
70.0 |
95.0 |
Sigmoidoscopyb |
75.0 |
85.0 |
95.0 |
95.0 |
92.0c |
Colonoscopy |
75.0 |
85.0 |
95.0 |
95.0 |
90.0c |
CTC DoD 3D 6mm |
-- |
83.6 |
92.2 |
92.2 |
79.6d |
CTC NCTC 2D/3D 6mm |
-- |
57.0 |
84.0 |
84.0 |
88.0d |
-- indicates sensitivity is not provided because size is smaller than the colonoscopy referral
threshold of 6mm.
a. Sensitivity is provided per individual for
stool-based tests and per lesion for endoscopy and CT tests.
b. Test characteristics for sigmoidoscopy apply only
to lesions in the distal colon and rectum.
c, The lack of specificity with sigmoidoscopy and colonoscopy
reflects the detection of non-adenomatous lesions. With sigmoidoscopy, the
presence of non-adenomatous lesions induces biopsy costs (in the case of
sigmoidoscopy with biopsy) or results in referral for diagnostic colonoscopy
(in the case of sigmoidoscopy without biopsy). With colonoscopy,
non-adenomatous lesions are removed and therefore induce polypectomy and biopsy
costs.
d. The lack of specificity with CT colonography
reflects the detection of non-adenomatous polyps, artifacts, and adenomas
smaller than the colonoscopy referral threshold of 6mm.
We assumed conditional independence for all screening tests. In other words, the sensitivity for detecting an adenoma or cancer depended only on the disease status at the time of the screen and did not depend on the test results from previous screening tests.
The base-case cost-effectiveness analysis was conducted from the payer (CMS) perspective. We also conducted an analysis from a modified societal perspective by including direct costs borne by beneficiaries as well as estimated patient time costs, but excluding costs due to lost productivity caused by early death or disability. Screening costs were based on information provided by CMS on Medicare payments in 2007 for procedures and tests associated with CRC screening and complications of screening. Net costs of CRC-related care were obtained from an analysis of SEER-Medicare linked data.
The screening test costs are provided in Table 4. The costs for FOBT, flexible sigmoidoscopy, colonoscopy, complications of screening, pathology, and of colorectal cancer treatment are those used for the cost-effectiveness analysis of the DNA stool test for CMS (Zauber 2007) https://www.cms.hhs.gov/mcd/viewtechassess.asp?from2=viewtechassess.asp&id=212&. Briefly, screening-related costs were based on the set of current procedural terminology (CPT) codes relevant to CRC screening (refer to Zauber 2007 for CPT codes used) in conjunction with the points of service for the procedures. For procedures with polypectomy or biopsy, we included the associated pathology costs. We assumed that in 5% of exams, a repeat colonoscopy is necessary in order to adequately visualize the colorectum. Instead of modeling incomplete colonoscopies, we increased the costs of a colonoscopy without polypectomy by 5%. For colonoscopy with polypectomy we added the same absolute difference in cost ($25) based on the assumption that polyps were only removed at one of the two colonoscopies. The cost of sedation was included in the cost of colonoscopy, assuming that it is not administered by an anesthesiologist.
Screening test |
CMS cost, $ |
Modified societal cost,b $ |
---|---|---|
Guaiac Hemoccult (II or SENSA) |
4.54 |
21.54 |
Fecal immunochemical test |
22.22 |
39.22 |
Flexible sigmoidoscopy |
160.78 |
270.30 |
Flexible sigmoidoscopy with biopsy |
348.19 |
497.37 |
Colonoscopy without polypectomy |
497.59 |
794.94 |
Colonoscopy with polypectomy or biopsy |
648.52 |
979.28 |
CT colonographya |
488.29 |
643.64 |
a. Based on CMS reimbursement for CT of the abdomen (CPT 74150), CT of the pelvis (CPT 72192), and image
processing on an independent workstation (CPT 76377).
b. Modified societal costs
include beneficiary costs (co-payments) and time costs in addition to the payer
costs.
Given that this report was written in conjunction with the NCD for CT colonography for CRC screening in the Medicare population, there is no national CMS reimbursement rate for a screening CT colonography at this time. Accordingly, we use as a proxy the national average CMS reimbursement (excluding patient co-pays) for an abdominal CT without contrast (CPT code 74150), a pelvic CT without contrast (CPT code 72192) and image processing on an independent workstation (CPT 76377). We obtained estimates of the 2008 rates for these procedures and converted them to 2007 dollars using a decrease of 3.5% in medical care costs to be compatible with the 2007 cost estimates obtained for other screening tests, complications, and colorectal cancer care. This process yielded a base-case cost for CT colonography of $488.29. Note that this is similar to the average reimbursement (excluding beneficiary co-payments) for a diagnostic CT colonography among carriers in the NY area ($486) (personal communication, Bill Larson, Paul Deutch).
There are essentially no complications from the stool-based screening tests (Hemoccult II, SENSA, or FIT) from the tests themselves. However patients undergoing colonoscopy and, to a lesser extent, flexible sigmoidoscopy and CT colonography are at risk of experiencing complications from the procedures. Because individuals with a positive sigmoidoscopy, CT colonography or stool-based tests are referred for a follow-up colonoscopy, the complications and the associated costs are relevant and accounted for in all of the screening strategies. We used the risks and associated costs of complications with sigmoidoscopy and colonoscopy that we derived for the stool DNA report (Table 5) (Zauber 2007). The costs of complications were based on the relevant DRG codes. For CT colonography we assumed a risk of perforation of 4.56 per 100,000 (Pickhardt 2006a). Although perforations from CT colonography may be less severe than those from colonoscopy we conservatively assumed that 5.19% of those who have a perforation die as a result (Gatto 2003), regardless of which test caused the perforation.
Complication |
Rate per 1000 |
CMS cost, $ |
Modified societal cost, $ |
---|---|---|---|
With colonoscopy |
|||
Perforation |
0.7 |
12,446 |
12,712 |
Serosal burn |
0.3 |
5,208 |
5,474 |
Bleed with transfusion |
0.4 |
5,208 |
5,474 |
Bleed without transfusion |
1.1 |
320 |
586 |
With flexible sigmoidoscopy |
|||
Perforation |
0.02 |
12,446 |
12,712 |
With CT colonography |
|||
Perforation |
0.0456 |
12,446 |
12,712 |
The costs of CRC treatment were also the same as those used in the DNA stool test report (Zauber 2007). Briefly, these costs were derived from comparison of costs for CRC cases relative to those of matched controls in the SEER-Medicare files for the years 1998-2003 (personal communication, Robin Yabroff, Ph.D. and Martin Brown, Ph.D; Yabroff 2008) and vary by phase of care (Table 6).
AJCC Stage |
Initial Phase |
Continuing Phase |
Last Year of Life | |
---|---|---|---|---|
Died from CRC |
Died from Other Causes |
|||
Direct medical costs |
||||
I |
25,487 |
2,028 |
45,689 |
11,257 |
II |
35,173 |
1,890 |
45,560 |
9,846 |
III |
42,885 |
2,702 |
48,006 |
13,026 |
IV |
56,000 |
8,375 |
64,428 |
34,975 |
Modified societal costs |
||||
I |
32,720 |
2,719 |
56,640 |
17,408 |
II |
43,752 |
2,561 |
56,417 |
15,740 |
III |
53,003 |
3,573 |
59,481 |
19,413 |
IV |
68,853 |
10,743 |
78,227 |
44,384 |
a. The initial phase of care is the first 12 months following diagnosis, the last-year–of-life phase is the final 12 months of life, and the continuing phase is all the months between the initial and last-year-of-life phases. Cancer-related costs in the continuing phase of care are an annual estimate.
We did not include the additional medical costs nor potential benefits to follow up of extracolonic findings detected by CT colonography. Although the prevalence of extracolonic findings has been reported (Levin 2008) as well as costs (Pickhardt 2008a), the long-term benefit of working up the various extracolonic findings is not well documented. The implicit assumption that we are making by not formally incorporating these costs and benefits is that, conditional on a CT colonography examination being done, cost-effective approaches to follow-up care of extracolonic finding are being adopted.
In a sensitivity analysis we added beneficiary costs (co-payments) and time costs to the payer costs for a modified societal perspective. We label this perspective a "modified societal perspective" because while we include the above costs, we do not incorporate productivity costs.
Beneficiary costs associated with screening tests were based on the CMS co-payment per point of service and type of CPT code. To incorporate patient time costs associated with CRC screening we assumed that the value of patient time was equal to the median US wage rate in 2007 from the Bureau of Labor Statistics, $16.64 per hour. We assumed that endoscopy screening requires preparation and recovery. We assumed that the time associated with a colonoscopy procedure was 8 hours, 4 hours with flexible sigmoidoscopy, and 2 hours with CT colonography. Patient time requirements for stool-based screen tests (e.g., Hemoccult II, Hemoccult SENSA, and FIT) were assumed to be 1 hour. For treatment of complications with colonoscopy, sigmoidoscopy, and CT colonography, we assumed that patient time requirements would be on average 16 hours. Modified societal costs for screening are given in the right-hand side of Table 4.
The beneficiary costs for treatment were also derived based on the copayment and time costs. Estimated patient deductibles and coinsurance expenses were added by adjusting Part A and Part B payments with Medicare reimbursement ratios provided by the CMS Office of the Actuary. Estimates of time costs for cancer care were from a recently published analysis of the SEER-Medicare linked data (Yabroff 2007) and updated to 2007 dollars using the Consumer Price Index. The treatment costs that were used as model inputs for the modified societal perspective are shown in the bottom half of Table 6.
Using the base-case inputs, we used each model to project a number of outcomes for each screening strategy. These outcomes include the number of cancers detected, number of cancer deaths averted, life expectancy (discounted and undiscounted) and the lifetime CMS costs (discounted and undiscounted). Differences in results across models reflect the different underlying natural history models.
For each model, we ranked the 14 screening strategies (no screening, 12 non-CTC screening strategies, 1 candidate CT colonography strategy) by increasing effectiveness (i.e., discounted number of life-years gained compared with no screening). Strategies that were more costly and less effective than another strategy were ruled out by simple dominance. Strategies that were more costly and less effective than a combination of other strategies were ruled out by extended dominance. Remaining strategies were then rank ordered by increasing costs and effectiveness, and incremental cost-effectiveness ratios (ICERs) were calculated by dividing the incremental discounted cost by the incremental discounted life-years gained, relative to the next least expensive option. These strategies represent the set of efficient options. On a plot of costs vs. life-years gained, a line that connects the efficient strategies is called the efficient frontier, and all dominated strategies (simple or extended) lie below this line. If the CT colonography strategy did not lie on the efficient frontier, we then determined the degree to which each of the following parameters would have to change in order for the CT colonography strategy to reach the frontier: unit cost of the CT scan, or relative adherence with CT colonography compared with other screening tests. Because the two base-case CT colonography scenarios do not represent competing options for CT colonography screening but rather two different estimates for test performance, we repeated this process separately for each CT colonography strategy.
For each CT colonography strategy, we calculated the maximum cost of a single CT scan for the strategy to be part of the efficient frontier. There were three possible situations to consider when including a CT colonography strategy as an efficient strategy: (1) the CT colonography strategy was less effective than the least effective strategy on the efficient frontier, (2) the CT colonography strategy was more effective than the most effective strategy on the efficient frontier, and (3) the effectiveness of CT colonography strategy was intermediate to the least effective and most effective strategies on the efficient frontier.
In the first case the threshold cost of a CT scan was calculated such that the total cost for the CT colonography strategy was the same as the next least effective efficient strategy (yielding an ICER of 0 for that non-CTC strategy). In the second case the threshold test cost was calculated such that the ICER for the CT colonography strategy compared with the most effective efficient strategy was equal to $50,000 per life-year gained. In the third case we identified the efficient strategy with lowest life-years gained that would still have more life-years gained than the CT colonography strategy. Subsequently the threshold cost was calculated such that the ICER of the CT colonography strategy was equal to the ICER of that selected strategy.
We also considered three sensitivity analyses for the threshold costs. First, we calculated the cost of a single CT scan that would result in the same discounted lifetime cost as no screening. Second, we determined threshold costs for a CT colonography scan such that the test strategy has the same average cost-effectiveness ratio (ACER) as the non-CT colonography strategy with the highest ACER value. ACERs represent the incremental cost per life-year saved of each strategy relative to no screening. Third, we calculated the per-test cost that would allow a CT colonography strategy to have the same ACER as the colonoscopy ACER.
We first conducted sensitivity analyses where we evaluated alternative scenarios of CT colonography in terms of test performance according to the primary reading approach (2D, 3D, or both 2D and 3D) and the minimum size polyp detected on CT colonography that will trigger a referral for optical colonoscopy. The test parameters for these sensitivity analyses are given in Table 7 and are based on data reported in the DoD, NCTC, and Johnson 2007 studies. We also considered a hypothetical worst-case scenario that had slightly lower test characteristics than all other scenarios evaluated.
CT colonography scenario |
Sensitivity by adenoma size or CRC, % |
Specificitya (%) |
|||
---|---|---|---|---|---|
≤5 mm | 6-9 mm | ≥10 mm |
CRC |
||
CTC DoD 3D 10mm |
-- | -- |
92.2 |
92.2 |
96.0 |
CTC DoD 2D 6mm |
-- | 31.9 |
75.0 |
75.0 |
93.4 |
CTC DoD 2D 10mm |
-- | -- |
75.0 |
75.0 |
98.0 |
CTC NCTC 2D/3D 10mm |
-- | -- |
84.0 |
84.0 |
86.0 |
CTC J 3D 10mm |
-- |
-- |
73.1 |
73.1 |
97.6 |
CTC J 2D 10mm |
-- |
-- |
72.0 |
72.0 |
98.1 |
CTC WC 2D/3D 6mm |
-- |
30.0 |
64.0 |
64.0 |
78.0 |
CTC WC 2D/3D 10mm |
-- | -- |
64.0 |
64.0 |
84.0 |
-- indicates sensitivity is not provided because size is smaller than the colonoscopy referral
threshold of either 6mm or 10mm; DoD = Department of Defense study (Pickhardt 2003, 2007a); NCTC = National CT Colonography Trial (Johnson 2008); J = Johnson
study (Johnson 2007); WC = hypothetical worst-case scenario.
a. The lack of specificity with CT colonography
reflects the detection of non-adenomatous polyps, artifacts, and adenomas smaller
than the colonoscopy referral threshold.
We also conducted sensitivity analyses where we varied relative adherence of CT colonography relative to the other CRC screening strategies. Some have suggested that CT colonography might entice a previously unscreened individual to undergo screening because it is non-invasive (Levin 2008). Our base-case analysis assumes that 100% of participants adhere to recommendations for the screening tests. To test the impact of differential adherence rates on the threshold CT colonography test cost, we conducted a sensitivity analysis on adherence. We first started with a more realistic 50% adherence rate for all tests (Shapiro 2008). We assumed that 50% of the population would be 100% adherent with a screening strategy and the other 50% would be non-adherent. The impact of modeling adherence in this fashion is that it does not alter the ICERs and it allows us to evaluate the impact of enhancing screening with CT colonography in a previously unscreened segment of the population. We then allowed the overall adherence with the CT colonography strategy to increase from 50% to 55% and 62.5% (a 10% and 25% increase respectively), and identified the corresponding CT colonography threshold costs per scan.