Cost-effectiveness of family history-based colorectal cancer screening in Australia

With 14.234 diagnoses and over 4047 deaths reported in 2007, colorectal cancer (CRC) is the second most common cancer and second most common cause of cancer-related mortality in Australia [1]. The direct treatment cost has recently been estimated to be around AU$1.2 billion for the year 2011, which corresponds to a four-fold increase compared the cost reported in 2001 [2, 3].

Approximately 10-15% of the population have a family history of CRC, which increases the disease personal risk by two-four fold (excess familial risk) depending on the number of relatives affected, the degree of relationship of the affected relatives and the age of diagnosis [4]. Even after excluding CRCs due to known genetic disorders such as Lynch syndrome and familial adenomatous polyposis (FAP), the cause of about 25-50% of this excess familial risk is unknown [4, 5] but could be due to specific genetic variants or shared environmental risk factors that are yet to be characterised [6].

The focus of this study is people at increased risk of CRC because of a strong family history of CRC. Here, we define strong family history as people with one first-degree relative diagnosed before the age of 55 years, or with two first-degree relatives or one first- and one second-degree relative on the same side of the family diagnosed at any age.

On average, these people have an estimated risk of developing CRC between three and six times higher compared to the average risk population [7, 8].

A targeted screening strategy addressed to this segment of the population is a potentially effective approach to reduce the burden of CRC.

Whilst three randomised controlled trials have demonstrated that screening with faecal occults blood test (FOBT)—followed with a diagnostic colonoscopy in case of positive test—is effective in reducing CRC incidence and mortality in the average risk population, no direct evidence exists to support the effectiveness of screening in people in this increased risk category. Based on expert opinion, current guidelines generally advise that persons with a strong family history of CRC should initiate screening at an earlier age or undergo more intensive screening compared to the average risk population [9‐11]. For example, current guidelines in the United States recommend colonoscopy screening every 10 years starting at age 40 or ten years younger than the age of the first diagnosis of CRC in the family [11, 12] Similarly, the Australian National Health and Medical Research Council (NHMRC) guidelines designate persons in this risk category as being at “moderately increased-risk” of CRC and recommend colonoscopic screening every five years from the earlier of age 50 years or ten years younger than the age of the first diagnosis of CRC in the family [10].

Similar to the clinical effectiveness, the cost-effectiveness of CRC screening in the average-risk population, based on various strategies and using different screening modalities, has been established by several studies [13‐18]. Pignone and colleagues identified six Australian studies published between 1996 and 2010. All of these studies concluded that annual and biennial CRC screening by guaiac-based or immunochemical faecal occult blood test (iFOBT) is cost-effective, with cost per life-year gained under $55,000 per year in 2010 Australian dollars [19]. Conversely, very few economic evaluation studies have focused on the economic aspects of screening in people at increased risk of CRC due to a strong family history, despite the fact that up to a quarter of all CRC cases are diagnosed in this segment of the population. We identified only two recent studies, conducted in Spain and the United States, assessing the economic implication of implementing a family history-based CRC screening programme [20, 21]. Both studies found colonoscopic screening every five years starting at age 40 years to be a cost-effective strategy for people with a family history of CRC.

There is a strong rationale for designing risk-based CRC screening policies (i.e. intensiveness of screening based on risk) as this would, theoretically at least, permit a more efficient allocation of limited health care resources [22]. However, such policies, based on a personalised approach of CRC screening, would also have an important economic impact on the health care system as a whole (e.g. familial risk assessment programmes, higher number of screening colonoscopies and more specialists able to perform them) and overall public health implications which need to be investigated.

The main objective of this analysis was to provide an economic evaluation for people at increased risk of CRC, based on Australian data assessing the cost-effectiveness of biennial iFOBT, five-yearly colonoscopic screening starting at age 50 years, and ten-yearly colonoscopic screening starting at age 50 years to the current program of the National Bowel Cancer Screening programme of Australia (NBCSP) which currently consists of a one-off iFOBT screening offered at age 50, 55, 60 and 65 years (9) irrespective of family history.

We developed a Markov microsimulation model to simulate the natural history of CRC and to evaluate costs and outcomes of screening. Figure 1 summarises the model structure and assumptions. Four CRC screening strategies were superimposed on this model to simulate the evolution of a hypothetical population of 10,000 individuals at “moderately increased risk” of CRC due to a strong family history-as defined by the current NHMRC guidelines [10].

Based on current understanding of the variability in the natural history of the disease, we created nine mutually exclusive, possible health states (from normal to death due to CRC or other causes) and specified transition probabilities for movements between these health states. An individual can remain in the same health state or move to another health state according to the predetermined transition probabilities. Costs and health outcomes (including life expectancy) were assigned to each state and transition. The software TreeAge Pro was used to implement the model [23].

Table 3 presents the results of the base-case model. Under the Usual Care screening scenario (one-off iFOBT screening offered at age 50, 55, 60 and 65), 6,491 out of a cohort of 10,000 “moderately increased-risk” 50 year-old people would develop CRC over 40 years (to age 90) with an average lifetime cost of AU$3,441 per person. Compared to the NBCSP, the iFOBT₂, COLO₁₀ and COLO₅ screening strategies reduced the number of CRC cases by 27%, 35% and 60% and CRC mortality by 15%, 26% and 46% respectively. All three approaches resulted in more LYG and were more costly than Usual Care. Five-yearly colonoscopy screening starting at age 50 was the most costly option with a lifetime cost estimated to be AU$8,734 per person but also provided the highest LYG of all the screening strategies.

The incremental cost-effectiveness ratios of the different screening strategies compared to Usual Care are shown in Table 4. These ratios were AU$8,306 and AU$12,405 per LYG for COLO₁₀ and COLO₅ respectively. iFOBT₂ was dominated by both screening approaches in the model (Figure 2). The COLO₁₀ strategy had the lowest ICER compared to usual care. However, with a superior clinical effect (longer life expectancy and 60% reduction of incident CRC cases over the 40-year period) and the lowest number of colonoscopy procedures needed to save one life, COLO₅ screening appeared as the most cost-effective strategy of the models compared.

Sensitivity analyses showed that above 20% participation rate the level of colonoscopy screening uptake was not a very influential parameter on the ICER (Table 5). For example, in a direct comparison with the NBCSP scenario, with a 20% screening participation rate, the ICERs for the COLO₅ and COLO₁₀ strategies were $9,108 and $7,835 per LYG respectively. The cost of both strategies were positively correlated with the level of screening uptake but increased by only 13% and 24% respectively when screening participation was set to 90%.

Effects of varying the level of CRC risk (i.e. 3 and 6 fold risk increase compared to 4 fold increase used in the base-case model) associated with a strong family history on the different outcomes of the model are presented in Additional file 1: Table S1-S4.

Our aim was to provide an insight into the cost-effectiveness of CRC screening for people at increased risk of the disease due to having a strong family history. We compared three different screening alternatives to the current NBCSP using Australian data on CRC incidence, adenoma distribution and cancer stage for invasive carcinoma. We chose the current NBCSP screening strategy as the base case scenario in our model instead of a no screening scenario as we consider this approach to be more realistic and more relevant to the Australian context where a population-based screening programme has been in place since 2006 and will be expended to cover all individuals aged 50 to 75 years by 2017 [30].

All three strategies had an ICER under $50,000 per LYG, which is regarded as the upper limit of acceptable cost-effectiveness in the Australian health system [24, 31]. An incremental cost of AU$12,405 per LYG, meant colonoscopy screening every five-year appears to be the most cost effective strategy of the three tested. In comparison, the Australian study conducted by Tran and colleagues (from which we obtained the CRC survival rates and treatment costs for our modelling) found the NBCSP, as implemented in 2008, to be cost-effective for the general population at AU$38,216 per LYG [3].

Ladabaum et al. conducted a cost-effectiveness analysis including only direct medical costs and using data from a screening pilot programme implemented in Aragón (Spain) targeting first-degree relatives of patients with CRC. The authors analysed the cost-effectiveness of CRC screening starting at age 40 years comparing colonoscopy every ten years and colonoscopy every five years, to no screening. Both screening strategies dominated the “no screening” option. Similar to our findings, five-yearly colonoscopic screening was found to be the most cost-effective strategy [20].

In a study conducted in the United States, Ramsey and colleagues used a validated microsimulation model [32] adopting a societal perspective (i.e. analysis including direct and indirect costs of screening) to evaluate the clinical and economic implications of implementing a CRC screening programme based on family history in the United States. Similar to our approach, the authors used a very specific definition of family history, based on current CRC screening guidelines. In the model, persons with a positive family history (i.e. having one FDR diagnosed with CRC before age 60 or two FDRs diagnosed at any age) had three alternative colonoscopy-based screening scenarios, which varied by frequencies (five or ten years) and age at which the screening began (age 40 to age 50). Results from each one of these screening strategies were then compared to a “usual care” scenario where colonoscopic screening was offered every ten years to the entire population (average- and increased-risk persons) from age 50. The analysis found all screening strategies to have acceptable ICERs (including the usual care option when compared to no screening) with five-yearly colonoscopy screening from age 40 years being the most cost-effective option (US$18,678 per LYG). The study concluded that a combination of a more aggressive CRC screening strategy targeting people with a family history, with a programme addressed to the general population may be a valuable approach to prevent CRC in the population [21].

In this study we attempted to conduct a similar analysis in the Australian context where an iFOBT-based screening programme has been in place since 2006. To our knowledge, this is the first analysis attempting to measure the economic aspect of different CRC screening strategies (including NHMRC recommendations) that target persons at “moderately increased risk” of CRC in the Australian context. It should be noted that this category does not include all individuals with a family history of CRC. Our model is particularly relevant to a segment of the population at a substantially increased risk of CRC—namely, individual whom CRC family history characteristics place them in the risk category 2 of the current Australian guidelines [10]. Persons in this category are more vulnerable to CRC due to their familial risk but also due to the fact that the current NBCSP strategy, by definition, does not address their specific needs in terms of screening modality and intensity. We believe that there is an urgent need to adjust the existing screening programme in order to address the specific screening requirements for this population. Our estimates of the number of colonoscopies needed for each screening scenario represents a proxy for resource utilization as well as adverse events from screening. Currently, over 500,000 colonoscopies are performed each year in Australia [33], a substantial proportion of these procedures occur outside of the national screening programme [19] and are undertaken by individuals at average risk of CRC for whom colonoscopy is not recommended as a screening procedure [34]. In this context, adjusting the current NBCSP to appropriately address the screening needs of people above average risk of CRC is likely to mainly consist of a more efficient use of already existing resources.

The modelling presented here has several limitations. For example the model assumed that family history of CRC was known for all participants and did not include administrative costs associated with the implementation of a family history assessment.

While the model accounted for the higher incidence of polyps observed in people with family history of CRC as a cause of a higher CRC incidence [35], it did not include information on polyp sojourn time (i.e. preclinical phase), which determines the rate of cancerous transformation. Polyp behaviour is those with family history is still not well characterised, particularly the malignancy transformation rate.

Our analysis was conducted from a third-party payer perspective and was limited to direct costs only. Indirect costs such as production loss due to CRC treatment need be taken into account for a more exhaustive economic assessment of a family history-based CRC screening programme. We did not take into account the costs involved with potential complications from screening tests, and the combination of screening and administrative costs. Also, our costs parameters rely heavily on estimates reported in Bishop et al. review, which was published in 2008. It is possible that some of those estimates might be out-of-date given the constant progress and changes in CRC treatment protocols.

In conclusion, our results are consistent with findings from previous research and present informative preliminary estimates in the perspective of familial risk-based CRC screening strategies within the context of a national screening programme addressed to the average risk population. Given limited colonoscopy capacity and budget, there is an increasing number of investigators calling for a better inclusion of personal disease risk in the design of CRC screening policies. The rational of this approach being that more screening resources should be allocated to those who would benefit most from more screening while at the same time reducing the intensity of screening among those who have less to gain. Several risk prediction tools have been developed [36, 37] and future advances in genetic testing for CRC variants will facilitate this process, to allow tailored screening strategies. Our findings suggest that health benefits can be generated by implementing more intensive screening to those at increased risk at a relatively modest incremental cost.