This review confirms the findings of the earlier USPSTF review (4
)—that colorectal cancer screening is cost-effective compared with no screening, irrespective of the screening modality used. Moreover, it shows that there is a tendency toward more favorable costs per LYG with colorectal cancer screening in more recent years. However, as in the previous review, no single strategy is consistently found to be the most effective or to be preferred for a given willingness to pay per LYG.
Web Table 2
indicates several reasons for the disparate findings between studies. Despite recommendations from the Panel on Cost-Effectiveness in Health and Medicine (56
), this table shows that included studies still differ widely with respect to perspective, population, time horizon, and discount rate. These differences potentially explain the differences in model results. However, even 2 models evaluating the same strategy of annual FOBT, both from a third-party-payer perspective, looking at a cohort aged 50 years followed for a lifetime, discounting costs and results at 3%, arrived at very different results (37
A closer look at these studies (Web Table 2
Continued) reveals that the costs for FOBT are 2–3 times higher in the first study than in the second, while FOBT sensitivity is 60% and 40%, respectively. These differences alone may explain a large portion of the observed differences, and we have not yet considered differences in modeling of the natural history of disease. Comparing the natural history assumptions between the models proves difficult. The Sonnenberg et al. study reports an adenoma incidence of 1% per year but does not report on progression parameters from adenoma to cancer or on progression from early to late cancer. The models differ with respect to how the effectiveness of screening is modeled: in the Sonnenberg et al. model, benefit of screening is an explicit model parameter of 18% for early detection and 75% for adenoma removal; in the Khandker et al. model, the screening benefit is the result of the improved stage distribution with screening and prevention of colorectal cancer with removal of adenomas. Thus, even when considering just 2 models, it proves impossible to determine exactly which model parameters are responsible for the disparity in model outcomes. For more models, it would become only more difficult.
The model differences in adenoma dwell time and unit test costs are most striking. The former difference seems to occur from disparities in modeling philosophy between studies, where short adenoma dwell times are based on assumption or expert opinion and long adenoma dwell times were calibrated to observed adenoma prevalence and polyp growth rates. Although the calibrated dwell times have a more empirical basis, these estimates often do not allow for heterogeneity in adenoma dwell times, implying a very long adenoma duration for all adenomas. The fact that there was only very little attenuation of the protective effect of sigmoidoscopy screening even 12 years after having the test (57
) indicates that an adenoma dwell time of 12 years or less is unlikely and that adenoma dwell times are indeed long for most adenomas.
In an attempt to reconcile differences between models and determine what causes model differences, the Institute of Medicine (58
) and the National Cancer Institute's Cancer Intervention and Surveillance Modeling Network (59
) have offered platforms for modelers to cooperate with each other and standardize selective model inputs such as costs and test characteristics of their models. These exercises showed that with standardization of these model inputs, the model outcomes become more similar. Although the models continue to differ on the absolute costs and benefits of colorectal screening, the relative costs and effects of one screening strategy compared with another become similar (58
). Although no more than 2 of 5 models could agree on the preferred colorectal cancer screening strategy at a willingness-to-pay threshold of $50,000 with original model assumptions in the Institute of Medicine workshop, all 5 models recommended the same screening strategy with standardized model assumptions (58
). However, these standardized assumptions were not evidence based, so no conclusions concerning the optimal colorectal cancer screening strategy can be reached based on these results.
A joint analysis of CTC screening by the 3 colorectal cancer models in the Cancer Intervention and Surveillance Modeling Network group found disparate estimates for costs per LYGs of all screening strategies. However, all 3 models were consistent in their estimated costs per LYG of CTC screening relative to the other tests (22
). In addition, the models reached similar conclusions concerning what level of test costs would make CTC screening cost-effective.
The fact that different models reach similar conclusions when model inputs are standardized indicates uncertainty regarding what the values of these inputs should be. The differences between models (or model inputs, such as costs) may very well reflect the fact that they were developed for different settings, for example, different countries or payer settings. If cost-effectiveness studies publish sufficient model outcomes, local decision makers should be able to judge at least the face value of the models and translate the outcomes to their own specific setting in which they operate.
An example of such an outcomes table has recently been suggested (61
). Cost-effectiveness of a colorectal cancer screening test is only one of the factors influencing the decision to implement a colorectal cancer screening program. An outcomes table would provide decision makers with additional factors such as the number of colonoscopies required and the number of harms, such as the number of false-positive test results and overdiagnosed cases (i.e., detection of cases that would not have been detected without screening). Furthermore, with sufficient details, policy makers could apply local costs to the model outputs on tests and colorectal cancer diagnoses and estimate cost-effectiveness of colorectal cancer screening in their particular setting. However, since no single colorectal cancer screening strategy emerges as the most cost-effective, it is likely that the cost-effectiveness of all established colorectal cancer screening strategies is comparable and that other factors such as patient acceptability, screening compliance, capital versus operational costing, and colonoscopy and human resources required will determine the final verdict with respect to which colorectal cancer screening program to implement.
Differences in local preferences for colorectal cancer screening are already reflected in the different colorectal cancer screening policies in place throughout the world. In Australia and Japan, colorectal cancer screening programs are based on FIT testing. The European Union recommends only guaiac FOBT screening and none of the other tests because the effectiveness of these tests was not yet established by randomized controlled trials at the time of recommendation. Despite this recommendation, the colorectal cancer screening strategies currently being implemented in Europe differ widely between countries. Of 17 countries with colorectal cancer screening, 10 have adopted only FOBT, 6 use both FOBT and endoscopy, and only 1 uses colonoscopy (9
). In the United States, all established screening strategies, and in some guidelines even the newly developed screening strategies of stool DNA and CTC, are recommended for the general population (5
). The FOBT and endoscopic tests are currently reimbursed by Medicare and most other health care insurers (62
), so US individuals can choose their preferred screening strategy.
Availability of (high-quality) resources for colonoscopy and population preferences are expected to be the most important determinants of the final decision on which colorectal cancer screening program to implement. Even without a (colonoscopy) screening program, several countries already have a waiting list for colonoscopy, with waiting times of more than 6 months reported (63
). Implementing a guaiac FOBT screening pilot has been shown to increase colonoscopy activity by 21%–31% (66
). Screening programs with FIT or colonoscopy will only further increase the demand for colonoscopy. However, even when sufficient colonoscopy capacity is available, some countries may still opt out of implementing an invasive colonoscopy screening program, despite colonoscopy being the most accurate test. Some countries might prefer to offer a 2-step approach, with a less accurate, but also less invasive strategy—for example, FOBT—to stratify the population before offering invasive colonoscopies to only those at higher risk (i.e., with a positive result on the noninvasive test). Population preferences are important to consider when offering screening because, in the end, any screening test can be effective and thus cost-effective only if the population adheres to it.
The modeling of population adherence is an important limitation of all of the included cost-effectiveness analyses. Thirty of 32 models assumed the same adherence for all screening tests. The 2 cost-effectiveness analyses that looked at differential adherence among screening tests both assumed a higher adherence for FOBT compared with endoscopy (40
). This assumption is supported by recent results from a randomized controlled trial in the Netherlands that showed a 61.5% initial uptake with FIT, a 49.5% uptake with guaiac FOBT, and only a 32.4% uptake with sigmoidoscopy (67
). However, these results are for initial adherence only. FOBT is generally repeated every other year, whereas the recommended interval for endoscopy screening is 5–10 years. To be effective, high adherence with repeat FOBT is necessary. Experience with mammography screening has shown that, with frequent testing, adherence may decline over time (68
). It is possible that a similar pattern may occur for FOBT.
Even in a country such as the United States, where individuals can generally choose the colorectal cancer screening test they prefer, adherence to colorectal cancer screening is far from perfect, with an estimated 50% of the population having had an FOBT within the past year and/or endoscopy within the past 10 years (69
). This lack of adherence is one of the reasons that new tests such as FIT, stool DNA, CTC, and capsule endoscopy are being developed.
The included studies clearly show that, from a cost-effectiveness point of view, only FIT may currently be ready for widespread implementation in the general population. There are many caveats regarding this conclusion. First, there is no such thing as a FIT test. Instead, several FITs are currently available, each with its own performance characteristics and costs (70
). Cost-effectiveness analyses generally evaluate just one of the available FITs or a hypothetical FIT, with test characteristics that average those available in the literature. In this review, we simply paired the studies evaluating FIT together without considering which FIT was evaluated. Similar problems occur with stool DNA testing, with several versions of the test being available and new ones being developed (71
CTC performance characteristics will highly depend on the scan used as well as the expertise of the radiologist. However, even when considering the same machine and radiologist, several other aspects drastically influence the cost-effectiveness of CTC, such as the criterion for referral for a follow-up colonoscopy, the screening interval, and the cost setting (e.g., public vs. private insurer). Again, it was beyond the scope of this review to extensively examine the details concerning these settings between studies, but these factors will influence the study results considerably. Reassuringly, irrespective of the referral threshold or screening interval considered, CTC was shown not to be cost-effective compared with colonoscopy in the majority of studies.
Despite not being cost-effective compared with the established tests, a situation exists where implementation of the newly developed colorectal cancer screening tests can still be considered. If the tests would entice a previously unscreened segment of the population to adhere to screening, the no-screening strategy would be the relevant comparator for these people. Since the costs per LYG of these new tests are favorable compared with no screening, the new tests could, in that case, be recommended. Although there is some evidence that patients prefer CTC or stool DNA testing over the established screening strategies, no evidence currently shows that any of the newly developed tests increase colorectal cancer screening uptake among subjects unwilling to perform any of the established tests.
In conclusion, this review shows that colorectal cancer screening is cost-effective compared with no screening, but no screening method can be identified as the most effective or is the preferred strategy for a given willingness to pay per LYG. This finding indicates that the cost-effectiveness of the established colorectal cancer screening tests is likely to be comparable and that factors other than cost-effectiveness, such as population preferences and colonoscopy resources, might be more important in the decision about which colorectal cancer screening program to introduce. The newly developed screening tests of stool DNA testing, CTC, and capsule endoscopy are not yet cost-effective compared with the established screening options.