Value of resource modelling
The resource modelling described in this paper could be viewed as a particular type of budget impact analysis, with costs expressed in terms of resources and harms rather than monetary units. As in budget impact analyses , we focussed on the short-term health service impact of adopting technologies and generated results specific to the actual target population. Like budget impact analyses, the types of analyses shown here are intended to address the practical needs of both decision-makers and health service planners. They are also intended to complement more conventional cost-effectiveness analyses. In fact, these analyses did inform decision making in Ireland: a population-based colorectal cancer screening programme in 2013 commenced (see below) , and the estimates were used in planning the implementation of the programme. Our findings suggested that the annual endoscopy requirements associated with screening the 55–74 age-group by biennial FIT would be difficult, if not impossible, for the health services to meet in the short-term. However, the magnitude of the health gains that could be achieved within five years of screening implementation meant that FIT-based screening remained the preferred option of policy-makers, despite the lower resource requirements for gFOBT and FSIG. Therefore, the policy decision was to introduce biennial FIT screening; initially the programme will invite individuals aged 60–69 to be screened with the intention of eventually rolling-out screening to the full 55–74 age-range.
Delivering sufficient capacity
The few previous analyses of the health service impact of colorectal cancer screening [12, 52, 53] all noted that the ability of the health services to deliver sufficient endoscopy services was a crucial determinant of programme feasibility. In England, for example, although screening based on once-only FSIG was the most cost-effective option , the associated requirements for endoscopy facilities and staff were considered impossible to meet in the short-term: instead, a programme based on gFOBT was implemented.
In this study, we found that a programme based on FIT the option considered most cost-effective  – would require substantially more resources for colonoscopy, CT colonography, and other diagnostic procedures, than one based on gFOBT (with FIT as the second-line test in those who test weakly positive) or FSIG. FIT-based screening will generate demand for 11,000-15,000 additional colonoscopies annually in Ireland. In Ireland, in 2008, 44,000 publically-funded colonoscopies were performed, 2,000 of these under the National Treatment Purchase Fund. This fund provides for procedures to be undertaken in private facilities in the event of long-waiting times in the public system. In April 2009 more than 2,300 individuals had been waiting for a colonoscopy for more than three months . These figures suggest there is little spare capacity in the service, at least under current working patterns. They also justify the assumption, inherent in our analysis, that meeting the demand for “screening-generated” colonoscopies would require additional capacity over and above existing services.
If not prepared for, increases in colonoscopy demand can increase waiting times . Studies of screening programmes for other cancers have found that the period between having a positive screening test and waiting for diagnostic follow-up can be a time of considerable anxiety for the individuals concerned (see, for example, ). Although this issue does not appear to have been investigated to any great extent for colorectal screening, it is entirely plausible that many people will be distressed or anxious in the time between receiving a positive colorectal screening test result and attending for colonoscopy. Since anxiety and distress can be considered a cost of screening, serious consideration should be given to ensuring that as many people as possible undergo diagnosis and treatment (if required) as quickly as possible. Thus, while recent studies continue to build the clinical case for screening with FIT in preference to other tests [56–59], the implication of our results is that any FIT-based programme – in Ireland or elsewhere – will have to plan carefully how to deliver sufficient services for initial diagnosis and ongoing surveillance. The types of analyses illustrated here, and elsewhere , could usefully inform that planning process.
Strategies that might be used to help address endoscopy capacity and other resource challenges include restricting screening to a narrower age-range, and/or gradually phasing in screening across the full age-range. Although restricting FIT-based screening to those aged 55–64 would reduce resource requirements by around 40%, our cost-effectiveness analyses concluded that, when fully implemented this strategy would be slightly less desirable than screening the 55–74 age group . This suggests that gradual roll-out across the full age-range would be preferable, and indeed this is what has been implemented in Ireland, albeit in a different age-group initially to that considered here . However, it is worth remembering that the short-term health gains achieved with such phased-in screening will be lower than if the full age-group was screened from the outset.
Safety is a key consideration in the implementation of any screening programme . The chances of bowel perforation and death are, in fact, the major reasons why some – including the NCSS in Ireland – consider colonoscopy unsuitable as a primary screening tool . Colonoscopy is, however, the main option for diagnostic follow-up of those with positive screening tests. Our analysis suggest that the much higher numbers of colonoscopies generated by FIT-based screening than the other options will result in a higher number of adverse events. A higher adverse event rate was also evident in our cost-effectiveness analysis . However, the advantage of the results presented here is that they express the actual numbers of events that might be expected to occur in the real-world eligible population: these values are likely to be more informative for the purposes of assessing and managing risk.
A screening test must be acceptable to the population  and one of the main measures of acceptability is screening uptake. As might be expected, the resource requirements and health gains attainable with screening were highly dependent on estimates of uptake. For all screening tests, if actual uptake was higher than our base-case estimates more cancers would be averted; however, resource requirements would be higher and there would be more colonoscopy-related adverse events. On the other hand, if uptake was lower than the base-case value, resource requirements would be lower and there would be fewer adverse events, but there would also be lower health gains.
Since it is impossible to be entirely certain what uptake might be attained for any screening test in any population, base-case values were chosen to be conservative. For the faecal tests, the base-case estimate of 53% was based on the second round of the UK pilot programmes [34, 35]. Although this is lower than considered desirable by the NCSS in Ireland (60%) , it is higher than in some other European programmes and pilot programmes [58, 63–65]. In Ireland, 51% of almost 10,000 Dublin residents invited to complete a FIT did so , suggesting the base-case estimate may be reasonably realistic. For FSIG, there was no local data to inform what level of uptake might be realistic. Although participation rates of 55% and higher have been achieved in some settings [40, 67], generally uptake has been lower [58, 63]. We decided to derive our base-case value from the large UK Flexible Sigmoidoscopy Trial  to provide some consistency with the source of the estimates of uptake for the faecal tests.
Since we conducted our analysis evidence has accrued that uptake may be higher with FIT than gFOBT , although this has not been a universal finding . If FIT is more acceptable than gFOBT, our estimates of the resource requirements for FIT relative to gFOBT may be conservative. However, it is worth noting that the difference in numbers of screening-related harms may also be under-estimated.
Although parameter values were based on review of the best available evidence, as with all modelling studies, some uncertainty remains. Even for gFOBT, which has been shown to reduce mortality in several trials , performance characteristics remain uncertain. This is particularly true for newer versions of the test . For FIT, the available tests have heterogeneous performance characteristics  and for FSIG, information on sensitivity and specificity is limited [18, 20, 33]. For all tests, assessment of true sensitivity and specificity is further complicated by the fact that the reference (gold) standard, colonoscopy, misses lesions . We addressed this uncertainty through sensitivity analyses, and found that varying specificity, in particular, had a notable impact on requirements for diagnostic investigations. Hence, if the true test specificity differs from the values used here, the resource requirements and health outcomes might differ from those estimated.
We assumed, for gFOBT and FIT, that sensitivity and specificity remained constant across all screening rounds. Recent evidence suggests that the ability of gFOBT-based programmes to identify cancers might decline with repeated screenings . The same may be true for FIT. If so, this would suggest that we may have over-estimated the health gains associated with screening programmes based on FIT and gFOBT, although the extent of this over-estimation is difficult to quantify from currently available evidence.
We evaluated a strategy that combined gFOBT with reflex FIT, instead of the more conventional approach of reflex gFOBT. This was because second-line FIT has been shown to limit the number of referrals to colonoscopy [72, 73]. Finding effective strategies to “adjust” screening when colonoscopy capacity is limited is very topical . Although gFOBT with reflex FIT is currently not widely used, our results indicate that it could be an attractive option for delivering population screening in settings where capacity for diagnostic investigation is expected to be limited.
We did not estimate resources required for adjuvant chemotherapy in screen-detected colon cancers, or for follow-up after resection (which can involve annual colonoscopy or CT colonography for five years) since responsibility for providing these would most likely fall outwith the screening programme/NCSS. Facilities would, however, need to be provided elsewhere in the health services, so this distinction is somewhat arbitrary.