Economic potential of abbreviated breast MRI for screening women with dense breast tissue for breast cancer

The superior accuracy of breast MRI in comparison with conventional imaging techniques in screening women at high risk for breast cancer has been suggested by a number of studies; several recent prospective multi-center trials have compared breast MRI to conventional imaging techniques, such as conventional mammography (XM) and ultrasound (US), underlining a higher sensitivity at comparable specificity levels [1‐3]. However, current international guidelines still do not recommend breast MRI as a screening tool outside of high-risk screening collectives [4‐6].

Apart from high-risk collectives, the superior diagnostic performance has also been confirmed in screening women at intermediate risk of breast cancer due to their elevated breast density [7, 8]. Dense breast tissue is associated with an increased risk of breast cancer along with a decreased sensitivity in conventional breast imaging (ACR BI-RADS categories C or D) [9].

In order to cope with the scarcity in resource allocation and to make breast MRI more affordable, abbreviated protocols (AB-MRI) have been introduced offering similar sensitivity in comparison to full protocol breast MRI (FB-MRI), yet at the expense of decreased specificity. Initially, AB-MRI was defined as solely pre- and post-contrast sequences with subtracted and maximum-intensity projection images [10]. Recently, a variety of abbreviated protocols have been evaluated. These included additional T2-weighted images, ultrafast contrast-enhanced sequences, or diffusion-weighted images as a contrast-free alternative — all reporting varying specificities [11‐14].

The prospective trials of Comstock and Bakker et al have reported similar sensitivities of AB-MRI and FB-MRI in women of intermediate risk. Comstock et al described a specificity of AB-MRI of 87%, whereas Bakker et al found a specificity of 92% for the full protocol, in line with prior findings [10].

The lower specificity of AB-MRI is associated with a higher number of false positive findings and increased recall rates. Both techniques have been described to be cost-effective in screening women at intermediate risk with an incremental cost-effectiveness ratio (ICER) below commonly accepted willingness to pay thresholds of $100,000 per quality-adjusted life year (QALY) gained [15‐17].

The lack of a broadly accepted definition for both techniques — abbreviated as well as full-scale protocols — makes the evaluation of AB-MRI in general difficult as every approach leads to a different diagnostic accuracy.

Based on recent data, reductions in specificity that are associated with abbreviating breast MRI protocols were assumed as the key determinant to model the differences between full-scale and abbreviated protocols in the present study.

Consecutively, we aim to investigate the role of AB-MRI in screening women with dense breast tissue in relation to FB-MRI with a focus on the following objectives:

The superior diagnostic performance of breast MRI in screening women with dense breast tissue for breast cancer in comparison to x-ray-based techniques has been investigated and confirmed in recent prospective multi-center studies [7, 8].

A significant reduction in interval cancer rates as an accepted surrogate marker for mortality [38] was observed in the DENSE trial when screening women with elevated breast tissue densities with breast MRI compared to XM [8]. A recent publication demonstrated that additionally to its diagnostic benefit, FB-MRI may also be considered cost-effective in this setting, evaluating long-term costs and outcomes of screening women at intermediate risk of breast cancer due to their elevated breast density [15].

AB-MRI was introduced in 2014 in order to reduce both examination as well as image reading time, i.e., the costs per examination, while maintaining an acceptable diagnostic performance for reasons of cost-effectiveness [10]. The initial concept by Kuhl et al comprised only one pre- and one post-contrast acquisition with computation of a first post-contrast subtracted image and a maximum-intensity projection-image. In the following years, various protocols with and without contrast agent, some only based on diffusion-weighted imaging or ultrafast contrast-enhanced sequences, have been proposed in order to make breast MRI supposedly less expensive and thereby more cost-effective [11‐14]. However, the reduction in examination costs at the expense of reduced specificity must economically be considered a trade-off.

Recently, Comstock et al demonstrated a significantly higher rate of breast cancer detection for AB-MRI compared to DBT [7]. Instead of applying only one method of abbreviation, Comstock et al accepted various abbreviated protocols in their multi-center study with 48 participating imaging centers, with pre- and post-contrast sequences and T2-weighted imaging as long as a total examination time of 10 minutes was not exceeded. A recent publication demonstrated that AB-MRI may be cost-effective compared to DBT in screening women with dense breast tissue [17].

From an economical perspective, however, the degree of abbreviation and the resulting cost-effectiveness of the different forms of abbreviating protocols is of major interest when extending MR-based breast cancer screening to a population level. This led to the main goal of the present study as to determine the optimal balance of price vs. specificity of AB-MRI — assuming a standard Medicare reimbursement for breast MRI of $ 314.

A reduction of imaging specificity resulted in an increase of false positive findings and follow-up scan rates as well as a reduction in quality of life of the affected women. These downsides need to be put in relation to these cost reductions achieved by AB-MRI. However, AB-MRI has only rarely been investigated within clinical trials investigating intermediate-risk collectives. Up to this day, there is no broadly accepted definition as to the effective reduction of costs by applying different sets of abbreviated protocols. We therefore varied the costs per examination of AB-MRI in our study within our sensitivity analyses.

Comparing the specificities for breast MRI in a screening setting for women with dense breast tissue, we were able to determine a cut-off value of approximately 80% of the Medicare reimbursement of a FB-MRI examination ($ 259).

In other words: Assuming a reduced specificity of AB-MRI in an intermediate-risk population, the cost of AB-MRI should not exceed $ 259 in order to be preferred in terms of cost-effectiveness. This analysis may be considered the first step towards an understanding of the economic potential of AB-MRI on a population-based level.

However, some limitations of the model-based economic evaluation need to be considered:

The selected WTP thresholds of $ 50,000–$ 100,000 have been subject to scientific debate, yet have been applied to the U.S healthcare system by the majority of economic evaluations and represent the commonly accepted standard.

A Markov model can only represent an approximation of clinical reality and is characterized by its input parameters. The values of the diagnostic performance are based on two randomized prospective multi-center studies that were conducted in slightly different populations with different baseline risks: the DENSE trial included women with breast density category D, whereas Comstock et al included women with categories C and D. Breast tissue density hereby represents certainly only one risk factor among others that can contribute to an intermediate risk profile.

Importantly, sensitivity and specificity of breast MRI are independent from mammographic breast tissue density [39]. Therefore, the absolute values of diagnostic performance reported by Comstock and Bakker could reasonably be applied to this study population. An average pre-test probability of malignancy as reported by Comstock and Bakker and an average incidence rate were included into the economic model for comparing both screening strategies in order to deal with the heterogeneity of the two underlying studies. Further, the study design was independent from breast density as diagnostic performances of both FB-MRI and AB-MRI are unaffected by breast tissue density; the goal was merely to compare different specificity levels of the underlying methods of examination. Comparing different risk-stratified screening strategies across various screening populations with different grades of breast tissue density was not an aim of this analysis. The comparison of MR-based techniques with XM and DBT was not feasible since performance measures of x-ray-based techniques were acquired from different populations with divergent breast tissue densities.

In the DENSE concept biennial screening with conventional mammography was supplemented with breast MRI. Comstock et al on the contrary applied annual screening. Evidence-based recommendations on the selection of appropriate screening intervals for screening women with dense breasts with MRI are so far unavailable. We therefore selected a biennial screening interval for our economic evaluation, until prospective data comparing annual with biennial in terms of diagnostic performance and effects on mortality are scientifically explored further. Recent literature has suggested the potential economic value of a 4-year screening interval [40]. However, we did not include overly distant screening intervals of 4 years or more into our analysis since empirical outcome data on the interval cancer rates and stages of screening-detected breast cancer are unavailable as of today.

Additionally, both studies only describe the results of only the first screening round. However, recent findings of the second screening round of the DENSE trial indicate a decrease of false positives when breast MRI is routinely applied as a screening tool [41]. Even though the initial definition of protocol abbreviation as described by Kuhl et al suggested a FB-MRI examination after an initially suspicious finding, abbreviated protocols have not been clearly defined or standardized so far. Comstock et al included varying protocols as defined above and did not consider appending a FB-MRI examination in case of unclear findings.

Just as FB-MRI is merely defined by its fixed reimbursement and not ultimately by the composition of its protocol, further research as to the optimal degree of abbreviation and set of sequences is needed.

Further studies are required in order to analyze the role of other sources of varying specificity, such as reader experience.

At the same time, technical advances, e.g., by parallel imaging, allow for shorter examination times without possible detriments to diagnostic performance [42]. Hence, the gap between abbreviated protocols and full protocols may diminish to some extent in the future. Based on these developments, we deem recommendations on the role of AB-MRI in screening women with dense breasts premature as of today.

In conclusion, our preliminary findings offer a first delineation of the economic value of AB-MRI compared to FB-MRI and indicate certain cost thresholds that should not be exceeded in order to maintain a preferable cost-effectiveness of AB-MRI over FB-MRI.