In our model, annual mammography plus MRI, compared to annual mammography alone, has an ICER of $50,900 per QALY. This ICER was estimated using local cost and treatment data, with input from clinicians and decision-makers on the project’s advisory panel, in an effort to most accurately depict the context of breast cancer screening and treatment for BRCA1/2 mutation carriers in British Columbia. These results suggest that the cost-effectiveness of the MRI screening program for BRCA1/2 mutation carriers is finely balanced, with sensitivity to input parameters and statistical uncertainty. The BCCA does not use a cost-effectiveness threshold, but the ICER falls within the generally accepted range for funded programs.
The mammography plus MRI strategy of the model differs from the mammography alone strategy in four key ways: the cost of MRI screening, increased screening sensitivity, a more favourable stage distribution among MRI-detected cancers, and more false positive screens due to decreased screening specificity. The cost-effectiveness of MRI screening is highly dependent on the cost of an MRI scan, as indicated in one-way sensitivity analysis. In situations where MRI scans are costlier than at the BCCA, MRI screening for breast cancer may not be a cost-effective option. However, these results also suggest that improvements in technical efficiency leading to reductions in the per-scan cost of MRI may reduce the cost-effectiveness of MRI screening to more acceptable levels.
The time horizon of this model, as with any model of preventive or screening techniques, also has an impact on the findings. The costs of MRI screening accrue from the beginning of model, while the benefits arising from MRI screening, such as lower treatment costs for cancers detected at an earlier stage, appear much later in the model, particularly as the cohort ages and cancer incidence rises. Consequently, the model is very sensitive to discounting assumptions for cost and QALYs.
The ICER calculated in this study is higher than previously published cost-effectiveness estimates from the UK, but lower than those from the US [20–22, 24]. In the UK study by Norman et al., women were screened for only 10 years, beginning at age 30 or 40 years, giving ICERS of approximately CAD$17,600 and $30,600 per QALY . By contrast, our model includes MRI screening from age 25 to 65 years. In the US studies, the cost of MRI was much higher than in this study, around USD$1000 for a bilateral screen, which is a potential reason why the reported ICERs are also higher. Moore et al. found in their sensitivity analysis that reducing the cost of MRI to below USD$315 resulted in an ICER of under USD$50,000/QALY, down from the base-case ICER of nearly USD$180,000/QALY, which is more consistent with the findings of this study . Both Moore’s model and this study highlight the fact that MRI screening for breast cancer may be cost effective, when the cost of MRI scans is low.
A limitation of this model is that it represents an idealized screening program, with all women entering at age 25 and participating until age 65 or until they develop cancer. The MRI screening program operated by the BCCA has a dynamic population. Women join the program at various ages when they are deemed to be eligible, and leave after undergoing prophylactic surgery, after developing cancer, or for other reasons. The timing of screening also varies: women who must travel to Vancouver for screening often have both MRI and mammography done concurrently, and the interval between MRI screens may exceed 12 months. Consequently, the cost-effectiveness of the screening program, if it were to be measured using real-world, comparative effectiveness program data, may be different. Although our goal was to use as much local data as possible, the challenge of acquiring comparative effectiveness data to inform the model was a further limitation of this study. We had insufficient sample size and follow-up to fully evaluate the effectiveness of the BCCA’s MRI screening program. We instead relied on the literature for screening effectiveness data. A further limitation of the model is that we were unable to include the risk of overdiagnosis from additional screening with MRI. Estimates of overdiagnosis attributable to mammography screening vary widely, from under 10% to as high as 50% [43–46]; however, overdiagnosis from MRI screening has not been assessed, nor has the rate of overdiagnosis in the BRCA1/2 population.
The model that we constructed to assess the cost-effectiveness of MRI screening lays the foundation to potentially address other questions related to breast cancer screening. For example, as more data become available the model could be adapted to find the optimal start time and duration of MRI screening from a cost-effectiveness perspective, or to investigate the relationship between lifetime breast cancer risk and cost-effectiveness of MRI screening, exploring the feasibility of expanding MRI screening to other high-risk groups.