Cost-effectiveness of colorectal cancer screening in Ukraine

We modified a model structure of a previously-published Markov cohort model [7] to evaluate the cost-effectiveness of implementing a national screening protocol in Ukraine compared with the current situation of no screening (Fig. 1). We assumed that screening started for 50-year-old members of the population with an average cancer risk and that screening continued to the age of 75 years. We adopted the third party perspective and modeled clinical events and costs over a lifetime time horizon. The cycle length of the Markov model was 1-year. We implemented the model in TreeAge Pro (TreeAge Software Inc, Williamston, MA) and audited calculations via MS Excel. Table 1 lists the model parameters and key assumptions.

Persons entered the model at the age of 50 and could reside in any of the following health states: normal mucosa, low risk polyp (< 1 cm), high-risk polyp (> 1 cm), preclinical CRC (localized, regional, and disseminated). They could transition between states in yearly time intervals, stay in their current state or develop clinical CRC (localized, regional, and disseminated). Persons could either die from cancer or from other causes or be in surveillance after treatment for colorectal cancer.

We used data from the published literature to obtain estimates of the prevalence of the polyps and age-specific transition probabilities [12‐19] (Table 1). We adjusted the estimates of prevalence of the polyps to the incidence based on age of CRC in Ukraine. We used 2013 Ukrainian life tables from the World Health Organization for background mortality. We estimated stage-specific CRC mortality with treatment using US data from the Surveillance, Epidemiology and End Results (SEER) database, given that the data on CRC mortality in Ukraine doesn’t represent mortality with treatment since large proportion of patients to not receive cancer specific treatment [19]. We used the National Cancer Registry of Ukraine database to estimate current Ukrainian CRC incidence, mortality, and treatment adherence [8].

In this analysis we included three different screening strategies that would be reasonable to implement in a LMIC with an established medical infrastructure: (1) yearly FOBT, followed by confirmatory colonoscopy if positive; (2) flexible sigmoidoscopy every 5 years with FOBT yearly, also followed by colonoscopy if positive; (3) colonoscopy every 10 years [20‐30]. Data on detection rate of these three screening scenarios, data on adherence with screening, screening complications rates, and utilities were obtained from screening studies conducted in Western countries and are displayed in Table 1 [17‐22, 31]. We assumed no polyp removal during sigmoidoscopy, but instead that polyps are removed as part of a full colonoscopy procedure.

We calculated costs from the third party payer perspective, in this case, the Ukraine Ministry of Health, and included direct treatment and screening costs. Indirect costs that are incurred by the patients such as transportation and lost productivity were not included. Screening costs were based on the manufacturer’s price for FOBT and on prices charged by private medical institutions in Ukraine (Table 1). We estimated treatment costs based on chemotherapy third party prices that the Ministry of Health of Ukraine pays for medications, as documented on the Ministry’s website [32]. We assumed doses appropriate for a 70 kg male and calculated costs for regional cancer [5-fluorouracil + Leukovorin + Oxaliplatin (FOLFOX for 12 cycles)] and disseminated cancer (FOLFOX + Bevacizumab for 12 cycles) regimens. We obtained surgical and radiation therapy costs from private clinics and modified them to reflect what would be reasonable in the public sector based on expert opinion from personnel at the National Cancer Institute of Ukraine (Personal communication with Dr. Kolesnik and Dr. Lukashenko). All costs were translated to 2012 US dollars.

We compared the four strategies by conducting a cost-effectiveness analysis. Strategies that were more costly and less effective than an alternative option were considered dominated and therefore excluded from the final cost-effectiveness calculations. For the remaining strategies, the incremental cost-effectiveness ratio (ICER) was calculated as the additional costs divided by the additional health benefit of the strategy compared with the next best non-dominated strategy. The most cost-effective strategy was then identified by comparing the ICERs against the willingness-to-pay (WTP) threshold for an additional QALY. We chose a WTP threshold of three times the per-capita gross domestic product (GDP), which is $11,700 US dollars per QALY (Ukraine’s annual per capita GDP is 3900 US dollars) [30]. The most cost-effective option was the strategy with the highest ICER below this WTP threshold. We discounted future costs and effectiveness at an annual rate of 3%, consistent with recommendations in the field of health economics [30].

We conducted sensitivity analyses to determine the impact of different variables on the cost-effectiveness results. We tested the uncertain variables in the model in one-way sensitivity analyses. The variables that we varied were adherence with screening, test characteristics, rate of complications, and costs of screening strategies and treatment. The ranges were chosen based on existing literature for variation on natural history, test characteristics, and adherence parameters and based on expert opinion for cost parameters (Table 1). We conducted two and three-way sensitivity analyses on the assumptions that had the greatest effects on the cost-effectiveness results identified in the one-way sensitivity analyses.

Figure 2 shows the annual incidence of CRC per 100,000 persons by age for the model predictions and data from the National Cancer Registry of Ukraine in 2013–2014. The figure illustrates that the observed incidences were very similar to what the model predicted.

The lifetime costs, effects, and cost-effectiveness of the different scenarios are presented in Table 2. All screening programs were less costly and more effective compared with the no screening program. FOBT yearly, sigmoidoscopy with FOBT every 5 years, and no screening were all more costly and less effective than colonoscopy every 10 years (Fig. 3).

Mortality estimates (Table 2) demonstrate benefits with all three screening scenarios compared to no screening, with colonoscopy every 10 years producing the greatest benefits in terms of decreased mortality. Colonoscopy screening every 10 years decreased colon cancer mortality by 73% compared to no screening. FOBT decreased mortality by 61.6% and flexible sigmoidoscopy with FOBT decreased mortality by 64%.

The results of the one-way sensitivity analyses show that the variable cost of colonoscopy and decreased compliance with colonoscopy have an impact on the most optimal screening scenario. The optimal screening scenario changes from colonoscopy every 10 years to sigmoidoscopy with FOBT when the costs of colonoscopy exceed the threshold value of $236 and the probability of compliance with colonoscopy falls below the threshold value of 0.66. The cost of cancer treatment or surveillance in the range tested did not impact the preferred strategy (Fig. 4).

In the scenario with decreased adherence to treatment, currently present in Ukraine, all the screening strategies were still cost effective, and the colonoscopy was still the most cost effective strategy with an incremental cost effectiveness ratio of $843.