1

Table S1: Vitamin D supplement use stratified by selected factors related to bone health.

Table S1a. Vitamin D supplement use by menopausal status

Vitamin D Supplement Use (IU/day) / Menopausal Status* / Total
Pre/Peri-menopausal / Post-menopausal
0 / 10 (23.3%) / 16 (12.6%) / 26 (15.3%)
>0-1000 / 11 (25.6%) / 42 (33.1%) / 53 (31.2%)
>1000 / 22 (51.2%) / 69 (54.3%) / 91 (53.5%)
Total / 43 (100%) / 127 (100%) / 170 (100%)

*unknown for 3 participants

Table S1b. Vitamin D supplement use by AI use

Vitamin D Supplement Use (IU/day) / Aromatase Inhibitor Use* / Total
No / Yes
0 / 16 (13.6%) / 9 (17%) / 25 (14.6%)
>0-1000 / 39 (33.1%) / 14 (26.4%) / 53 (31%)
>1000 / 63 (53.4%) / 30 (56.6%) / 93 (54.4%)
Total / 118 (100%) / 53 (100%) / 171 (100%)

*unknown for 2 participants

Table S1c. Vitamin D supplement use by bisphosphonate use

Vitamin D Supplement Use (IU/day) / Bisphosphonate Use / Total
No / Yes
0 / 26 (16.6%) / 1 (6.2%) / 27 (15.6%)
>0-1000 / 49 (31.2%) / 4 (25%) / 53 (30.6%)
>1000 / 82 (52.2%) / 11 (68.8%) / 93 (53.8%)
Total / 157 (100%) / 16 (100%) / 173 (100%)

Table S2: Sample size calculation examples showing numbers needed to screen.

HR / # to enroll (M) / # to screen (N) using pre-specified criteria / # to screen (N) using revised criteria
0.9 / 21864 / 172158 / 110985
0.85 / 9423 / 74197 / 47832
0.8 / 5132 / 40410 / 26051
0.75 / 3175 / 25000 / 16117
0.7 / 2122 / 16709 / 10772

Calculations to determine sample size (M) and number of patients screened (N) to identify suitable candidates were performed for a hypothetical RCT of vitamin D supplementation using several different effect sizes (HR). The pre-specified (i.e. serum 25-OH Vitamin D <72 nmol/L, current supplementation <1000IU/day and willing to participate; 12.7% of women), and revised (i.e serum 25-OH Vitamin D <72 nmol/L, current supplementation <4000IU/day and willing to participate; 19.7% of women) eligibility criteria were considered. A study with three years of accrual and a further three years of follow-up was modeled with 1:1 randomization to vitamin D vs. placebo. A 5-year recurrence rate in the control arm of 15% was assumed. To have 80% power to detect a difference between the two arms using a log-rank test with two-sided type I error of 0.05, M=5132 patients would be required to detect a hazard ratio of 0.8. To obtain 5132 eligible patients, N= M*100/12.7 = 40410 patients would need to be screened (pre-specified criteria). Sample calculations for other scenarios are included in Table S1. These calculations do not take into consideration other eligibility factors, which would further increase the number of screened participants required.