Cost-effectiveness and budget impact analyses of a colorectal cancer screening programme in a high adenoma prevalence scenario using MISCAN-Colon microsimulation model

Background The Basque Colorectal Cancer Screening Programme began in 2009 and the implementation has been complete since 2013. Faecal immunological testing was used for screening in individuals between 50 and 69 years old. Colorectal Cancer in Basque country is characterized by unusual epidemiological features given that Colorectal Cancer incidence is similar to other European countries while adenoma prevalence is higher. The object of our study was to economically evaluate the programme via cost-effectiveness and budget impact analyses with microsimulation models. Methods We applied the Microsimulation Screening Analysis (MISCAN)-Colon model to predict trends in Colorectal Cancer incidence and mortality and to quantify the short- and long-term effects and costs of the Basque Colorectal Cancer Screening Programme. The model was calibrated to the Basque demographics in 2008 and age-specific Colorectal Cancer incidence data in the Basque Cancer Registry from 2005 to 2008 before the screening begun. The model was also calibrated to the high adenoma prevalence observed for the Basque population in a previously published study. The multi-cohort approach used in the model included all the cohorts in the programme during 30 years of implementation, with lifetime follow-up. Unit costs were obtained from the Basque Health Service and both cost-effectiveness analysis and budget impact analysis were carried out. Results The goodness-of-fit of the model adaptation to observed programme data was evidence of validation. In the cost-effectiveness analysis, the savings from treatment were larger than the added costs due to screening. Thus, the Basque programme was dominant compared to no screening, as life expectancy increased by 29.3 days per person. The savings in the budget analysis appeared 10 years after the complete implementation of the programme. The average annual budget was €73.4 million from year 2023 onwards. Conclusions This economic evaluation showed a screening intervention with a major health gain that also produced net savings when a long follow-up was used to capture the late economic benefit. The number of colonoscopies required was high but remain within the capacity of the Basque Health Service. So far in Europe, no other population Colorectal Cancer screening programme has been evaluated by budget impact analysis. Electronic supplementary material The online version of this article (10.1186/s12885-018-4362-1) contains supplementary material, which is available to authorized users.

As each simulated person ages, one or more adenomas may develop (Model Appendix Figure 1). These adenomas can be either progressive or non-progressive. Both progressive and non-progressive adenomas can grow in size from small (≤5mm), to medium (6-9mm), to large (≥10mm); however, only progressive adenomas can develop into preclinical cancer. A preclinical cancer may progress through stages I to IV; however, during each stage CRC may be diagnosed because of symptoms. The stage specific survival after the clinical diagnosis of colorectal cancer before age 75 was taken from the Comprehensive Cancer Centre South from 1989-2003 [50]. The survival for individuals aged 75 years and older was adjusted to fit the observed age-increasing mortality/incidence ratio. For individuals with synchronous CRCs at time of diagnosis, the survival of the most advanced cancer is used. The date of death for individuals with CRC is set to the earliest simulated death (either due to CRC or due to another cause (see: 'The demography module')).

Transition Probabilities and Durations in States
An individual's risk of developing adenomas depends on the individual's age and a personal risk index. As a result of the latter most individuals develop no adenomas, whilst some develop many. We assumed that the distribution of adenomas over the colon and rectum equals the distribution of cancers as observed in Basque cancer registry during the period 2005-2008 before the programme started. The age-specific onset of adenomas and the dispersion of the personal risk index were calibrated to data on the prevalence and multiplicity distribution of adenomas as observed in autopsy studies [31-40] and adjusted to the adenoma prevalence observed in the COLONPREV study (Model Appendix Figure 2A and 2B)[20]. The age-specific probability of adenoma-progressivity and the age-and localization-specific transition probabilities between preclinical cancer stages and between preclinical and clinical cancer stages were simultaneously calibrated to Basque cancer registry data on the age-, stage-, and localization-specific incidence of CRC as observed during the period 2005-2008 before the introduction of screening (Model Appendix Figure 3A and 3B).
The average durations between the preclinical cancer stages were calibrated to the rates of screen-detected and interval cancers observed in randomized controlled trials evaluating screening using guaiac fecal occult blood tests [51][52][53]. This exercise has been described extensively in a publication by Lansdorp-Vogelaar and colleagues [23]. The average duration from the emergence of an adenoma (state 2) until progression into preclinical cancer (state 7) (i.e. the adenoma dwell-time) was calibrated to the rates of interval cancers (including surveillance detected cancers) observed in a randomized controlled trial evaluating once-only sigmoidoscopy screening (Model Appendix Figure 4) [54]. We assumed an equal overall dwell-time for adenomas developing into CRC from a medium size (30% of all CRCs) and from a large size (70% of all CRCs). All durations in the adenoma and preclinical cancer phase were drawn from exponential distributions. Durations within the adenoma phase and within the preclinical cancer phase were assumed to be perfectly correlated (i.e. if a small adenoma grows into a medium-sized adenoma rapidly, it will also grow into a large adenoma or develop into CRC rapidly); however, durations in the adenoma phase were assumed to be uncorrelated with durations in the preclinical cancer phase (i.e. a rapidly growing adenoma does not necessarily develop into a rapidly progressing cancer). The proportion of medium sized, non-progressive adenomas growing large and the average duration in the medium size, non-progressive adenoma state (state 5) were calibrated to size-specific adenoma detection rates observed in a Dutch randomized controlled trial on colonoscopy screening [55].

The Screening Module
Screening will alter some of the simulated life histories: Some cancers will be prevented by the detection and removal of adenomas; other cancers will be detected in an earlier stage with a more favorable survival. As the stage-specific survival of screen-detected CRC as observed in randomized controlled trials on guaiac fecal occult blood testing was substantially more favorable than that of clinically detected CRC, even after correcting for lead-time bias [23], we assigned those screen-detected cancers that would have been clinically detected in the same stage the survival corresponding to a one stage less progressive cancer. Hence, a cancer screen-detected in stage II, that would also have been clinically diagnosed in stage II, is assigned the survival of a clinically diagnosed stage I cancer. The only exceptions were screen-detected stage IV cancers. These cancers were always assigned the survival of a clinically diagnosed stage IV cancer.

Integrating Modules
The demography module generates a date of birth and a date of non-CRC death for each individual simulated, creating a life-history without adenomas or CRC. In the first example shown in Model Appendix Figure 5A, the natural history module generates an adenoma. This adenoma progresses into preclinical cancer, which is diagnosed because of symptoms in stage II and results in CRC death before non-CRC death would have occurred. In the screening module a screening examination is simulated, indicated by the blue arrow. During this examination the adenoma is detected, and as a result both CRC and CRC death are prevented. Hence, in this case screening prolongs life by the amount indicated by the green arrow. In the second example (Model Appendix Figure 5B) the patient also develops an adenoma, and although this adenoma does progress into preclinical cancer, this patient would never have been diagnosed with CRC in a scenario without screening. However, during the screening examination simulated in the screening module, again indicated by the blue arrow, CRC is screen-detected in stage I. Hence, in this patient screening results in over-diagnosis of CRC: It detects a cancer that would never have been diagnosed in a scenario without screening. Hence, screening does not prolong life, but it does result in additional LYs with CRC care (over-treatment) as indicated by the red arrow.

The Screening Parameters
We assumed a cecal intubation rate of 95% [56][57][58]. The sensitivity of colonoscopy for each lesion within realized reach was based on back-to-back colonoscopy studies: 75% in adenomas less than or equal to 5 mm, 85% in adenomas 6-9 mm, and 95% in adenomas greater than or equal to 10 mm and cancers [58].
After a positive test, all lesions are removed within a short time. The percentage of the population without adenomas or cancer but with hyperplastic polyps, lipomas, or other lesions that lead to polypectomy and pathology after colonoscopy has been estimated from Kaiser data: [60] 10%. This percentage was assumed to be independent of the screening round.
The stage-specific survival of patients with screen-detected cancer was based on a previous analysis calibrating on three large randomized FOBT-trials [23], and was more favourable than the survival after diagnosis in the same stage without screen-detecting. Removal of an adenoma always prevents development of any subsequent cancer that may have arisen from this adenoma. Risks of complications reported in organized screening programmes [61][62][63] are lower than those reported for general practice colonoscopies [64,65]. The major complications of colonoscopy are perforations (which can occur with or without polypectomy), serosal burns, bleeds requiring transfusion and bleeds not requiring transfusion [61][62][63][64][65]. We estimated a rate of death of 1 per 30,000 for colonoscopies with a polypectomy [66,67].