Patient Preferences for Outcomes Following DCIS Management Strategies: A Discrete Choice Experiment

Abstract

Purpose: Ductal carcinoma in situ (DCIS), a nonobligate precursor of breast cancer, is often aggressively managed with multimodal therapy. However, there is limited research on patients' preferences for trade-offs among treatment-related outcomes such as breast appearance, side effects, and future cancer risk. We sought to investigate whether women consider treatment features aside from cancer risk when making treatment choices for ductal carcinoma in situ and if so, to what degree other features influence these decisions.

Methods: A discrete choice experiment was administered to participants in a comprehensive cancer screening mammography clinic. The experimental design was used to generate constructed health profiles resulting from different management strategies. Health profiles were defined by breast appearance, severity of infection within the first year, chronic pain, hot flashes, and risk of developing or dying from breast cancer within 10 years.

Results: One hundred ninety-four women without a personal history of breast cancer completed the choice task. Across 10 choice questions, 29% always selected the health profile with a lower risk of invasive breast cancer (ie, dominated on cancer risk), regardless of the effects of treatment. For nonrisk dominators, breast cancer risk remained the most important factor but was closely followed by chronic pain (24% [95% CI, 20 to 28]) and infection (22% [95% CI, 18 to 25]). Depending on treatment outcomes, the tolerable increase in breast cancer risk was as high as 3.4%.

Conclusion: Most women were willing to make some trade-offs between invasive cancer risk and treatment-related outcomes. Our findings highlight the importance of shared decision-making weighing risks and benefits between patient and provider management of low-risk disease.

FIG 1.

Preference weights for feature levels. Model estimates are rescaled from 0 to 10, where 0 indicates the least preferred level and 10 indicates the most preferred level of the most important attribute. All the other attribute levels are rescaled accordingly. Rescaled values are referred to as preference weights. Vertical bars represent 95% CIs for the preference weights.

FIG 2.

Importance weights for treatment features. The difference between the least preferred level and the most preferred level within a feature is referred to as a feature's overall importance. The importance weight of each attribute is conditional on the levels employed in the study design. Overall importance weights are calculated for each feature and then weighted proportionally as a fraction of 100, with all importance weights for all features summing to 100. Vertical bars represent 95% CIs for the importance weights.

FIG 3.

Maximum acceptable breast cancer risk and duration of pain for less invasive surgical treatment outcomes. Estimates represent the implicit maximum risk of invasive breast cancer or maximum duration of chronic pain that a woman would accept to avoid more extensive surgery versus less extensive (or no surgery), all else equal.