Method: Women who were at high risk for breast cancer (5 year risk ≥1.66% using Gail Model) received an Internet-administered decision aid describing the use of tamoxifen to prevent a first case of breast cancer. A fractional factorial design was used to vary 5 design factors: 1) use of risk pictographs, 2) presence of contextual information (risk estimates of other diseases), 3) presentation of risks as either additional (i.e., marginal) or total risks, 4) order of presenting risks versus benefits, and 5) risk statistic denominator (100 vs. 1000). After reading through the decision aid, participants answered questions which measured their 1) knowledge about tamoxifen, 2) breast cancer risk perceptions, and 3) intent to act (e.g., search for more information, talk to doctor).

Results: 623 women read through the decision aid and completed the survey. Age range was 40-74, with mean age of 59. Gail score ranged from 1.67 to 17.3, with a mean of 2.5. Receiving risk/benefit information in pictographs was related to women having 1) increased perceived risk of breast cancer (both compared to the average woman and in terms of perceived likelihood of being diagnosed with breast cancer in the next 5 years, p's<.05), 2) decreased likelihood of saying they would take tamoxifen in the next year (p<.02), and 3) increased beliefs regarding how common the side effects of tamoxifen are (p=.01). Other factors had related effects: Both presenting the risks of tamoxifen after its benefits and describing risks using total risk terminology increased women's worries about the side effects of tamoxifen and led women to believe the risks were more common (p's<.02). Receiving contextual information that enumerated comparable risks of other serious diseases (e.g., colon cancer) reduced women's interest in seeking more information about tamoxifen (p = .01), as well as decreased order effects for a number of variables.

Conclusions: Knowledge, risk perceptions, and intention to act can all be significantly affected by the manner in which the risks and benefits of treatments are presented. When designing decision aids, it is critical that developers consider the impact of design decisions on patients' decision making.