A total of 332 women who had been diagnosed with either osteoporosis or who were postmenopausal at risk of fracture were recruited from eleven global sites in countries, such as the USA (n = 64), Canada (n = 61), Mexico (n = 36), Japan (n = 35), Argentina (n = 30), Belgium (n = 30), the UK (n = 29), Spain (n = 28), and the Netherlands (n = 19). Participants in the USA were recruited from two different sites, one in California (n = 35) and one in Washington (n = 29), and those recruited in Canada were interviewed by sites in Ontario (n = 31) and Quebec (n = 30). Sites recruited participants from a number of sources, including the local community, the site investigator’s clinical rheumatology or endocrinology practice, or an osteoporosis patient network, such as the Canadian Osteoporosis Patient Network (COPN) or the Royal Osteoporosis Society, UK (ROS). Given that this study was observational and did not pose a specific hypothesis, a sample size calculation was not applicable. Our study suggested a pragmatic target sample of 30 participants per site.

Participants at risk of fracture were defined as women with elevated FRAX score, elevated bone mineral density, or for whom pharmacologic and/or non-pharmacologic osteoporosis management was being considered by their site clinician. In order to ensure the diversity of participant perspectives, it was recommended that sites utilize a convenience sample of (1) at least ten participants who had sustained a fracture since the age of 40, (2) at least ten participants who had completed at least some education at the university level, (3) at least ten participants who had decided to take a prescription osteoporosis medication, and (4) at least ten participants who did not meet each of these criteria (Appendix Table 2).

At each site, a study coordinator guided participants through a structured survey either online via Zoom video conference or in-person, depending on participant preference and local COVID-19 guidelines. In both settings, Microsoft PowerPoint was used to display questions and all visuals to each participant. An extensive scoping review was conducted by Beaudart et al. to inform the design of this exercise, including the use of icon arrays as a visual aid to enhance patient understanding of fracture probability [6]. Icon arrays have previously been found to yield a similar understanding of risk presentation among patients with various levels of numeric literacy and reduce the impact of denominator neglect when considering the relative benefit of treatment in a population [7, 8].

Although the survey was originally developed in English, site investigators were able to request translation of participant-facing materials into locally spoken languages by certified professionals.

At the beginning of each survey, participants took a short, five-question assessment—extracted from an experimentally designed Graph Literacy Scale [7]—with which investigators gained a rough understanding of their ability to interpret graphs or visual aids. Two of these questions were Likert-type questions that were used to gauge the participant’s self-perception of numeric literacy and preference for numeric versus verbal presentation of information [9]. However, the remaining three questions were an objective assessment of numeric literacy and were used to stratify participants into high- or low-numeric literacy groups [7] (Appendix Fig. 3).

Within the survey, participants were presented with a maximum of eight hypothetical scenarios and asked whether they would be willing to accept a prescription for osteoporosis medication with minimal, transient side effects. To promote simplicity and participant understanding, this referred to any class of antiresorptive therapy. Each hypothetical scenario utilized two icon arrays to depict various fracture risk levels; the first icon array demonstrated the participant’s hypothetical risk of MOF over 10 years or FRAX probability, and the second showed how much of that risk could be mitigated upon taking a medication (Fig. 1). An optimistic 40% reduction in fracture risk was assumed with medication use given recent literature on antiresorptive osteoporosis therapies [10].

The first of eight hypothetical scenarios, in which participants were asked whether they would willing to accept a prescription of osteoporosis medication with minor, transient side effects given that their risk of MOF was 5% over a 10-year period and that medication would reduce this to a 3% fracture risk over 10 years

The eight hypothetical risk scenarios presented 10-year FRAX probabilities ranging from 5 to 40%, increasing in increments of 5%. The lowest fracture probability at which the participant agreed to accept a prescription osteoporosis medication was determined to be their FRDP. If they did not agree to medication by the final scenario (40% risk of MOF), then their FRDP was considered to be > 40%.

Institutional Review Board (IRB) approval was obtained on a site-by-site basis given local or university-wide requirements. Advarra, a central IRB, provided approval for all sites in the USA and Canada.

The primary outcome measured medication acceptance by the percentage of participants in each country who reported that they would hypothetically accept a prescription osteoporosis medication given a fracture probability less than or equal to their national FRAX-based treatment threshold. This threshold varied between countries as well as within countries in those that use an age-dependent or hybrid threshold. This outcome was measured specifically in countries where FRAX intervention thresholds are widely used in a clinical setting. The Netherlands does not and thus was excluded from this analysis [2]. Secondary outcomes detailed the impact of numeric literacy level, age, and fracture history on participant FRDP.

Site coordinators uploaded source documents and recorded all data in an electronic data capture system called “Online Clinical Trials” (http://www.essaionline.com), which allowed the lead study team to ensure sites’ adherence to study protocol and accuracy of data entry.

In order to validate the statistical significance of the difference in median FRDP among those with variable levels of numeric literacy and those with or without fracture history, Mood’s analyses were conducted. The Pearson correlation coefficient was calculated to understand the significance of the relationship between age and FRDP. Chi-squared tests were used to analyze preference for risk timeframe given fracture history and age. p-values were compared to an alpha level of 0.05 to determine statistical significance, and all data was analyzed with use of the statistical software, QI Macros.