The Effects of Four Doses of Vitamin D Supplements on Falls in Older Adults : A Response-Adaptive, Randomized Clinical Trial

Abstract

Background: Vitamin D supplementation may prevent falls in older persons, but evidence is inconsistent, possibly because of dosage differences.

Objective: To compare the effects of 4 doses of vitamin D₃ supplements on falls.

Design: 2-stage Bayesian, response-adaptive, randomized trial. (ClinicalTrials.gov: NCT02166333).

Setting: 2 community-based research units.

Participants: 688 participants, aged 70 years and older, with elevated fall risk and a serum 25-hydroxyvitamin D [25-(OH)D] level of 25 to 72.5 nmol/L.

Intervention: 200 (control), 1000, 2000, or 4000 IU of vitamin D₃ per day. During the dose-finding stage, participants were randomly assigned to 1 of the 4 vitamin D₃ doses, and the best noncontrol dose for preventing falls was determined. After dose finding, participants previously assigned to receive noncontrol doses received the best dose, and new enrollees were randomly assigned to receive 200 IU/d or the best dose.

Measurements: Time to first fall or death over 2 years (primary outcome).

Results: During the dose-finding stage, the primary outcome rates were higher for the 2000- and 4000-IU/d doses than for the 1000-IU/d dose, which was selected as the best dose (posterior probability of being best, 0.90). In the confirmatory stage, event rates were not significantly different between participants with experience receiving the best dose (events and observation time limited to the period they were receiving 1000 IU/d; n = 308) and those randomly assigned to receive 200 IU/d (n = 339) (hazard ratio [HR], 0.94 [95% CI, 0.76 to 1.15]; P = 0.54). Analysis of falls with adverse outcomes suggested greater risk in the experience-with-best-dose group versus the 200-IU/d group (serious fall: HR, 1.87 [CI, 1.03 to 3.41]; fall with hospitalization: HR, 2.48 [CI, 1.13 to 5.46]).

Limitations: The control group received 200 IU of vitamin D₃ per day, not a placebo. Dose finding ended before the prespecified thresholds for dose suspension and dose selection were reached.

Conclusion: In older persons with elevated fall risk and low serum 25-(OH)D levels, vitamin D₃ supplementation at doses of 1000 IU/d or higher did not prevent falls compared with 200 IU/d. Several analyses raised safety concerns about vitamin D₃ doses of 1000 IU/d or higher.

Primary funding source: National Institute on Aging.

Figure 1.

Overview of Dose-finding and Confirmatory STURDY Trial (Panel A) and Eligibility, Randomization, and Follow-up of Participants (Panel B). During dose-finding (October 2015-March 2018), eligible and consenting participants were randomized to supplemental vitamin D dose (200IU/day, 0.5 probability throughout; 1000, 2000, or 4000IU/day, varying probability per the posterior probability of the dose being best). After dose-finding ended, participants who had been randomized to 2000 or 4000IU/day were switched to 1000IU/day at the next study pill resupply time (resupply every 3 months) and new enrollees were randomized to 200 or 1000IU/day with equal probability. Abbreviation: IU/d = international units per day.

Figure 1.

Overview of Dose-finding and Confirmatory STURDY Trial (Panel A) and Eligibility, Randomization, and Follow-up of Participants (Panel B). During dose-finding (October 2015-March 2018), eligible and consenting participants were randomized to supplemental vitamin D dose (200IU/day, 0.5 probability throughout; 1000, 2000, or 4000IU/day, varying probability per the posterior probability of the dose being best). After dose-finding ended, participants who had been randomized to 2000 or 4000IU/day were switched to 1000IU/day at the next study pill resupply time (resupply every 3 months) and new enrollees were randomized to 200 or 1000IU/day with equal probability. Abbreviation: IU/d = international units per day.

Figure 2.

Kaplan-Meier Analyses of the Composite Primary Outcome of First Fall or Death During Dose-finding (Panel A) and Over all Follow-up (Panel B) in the STURDY Trial. Panel A displays a time-to-event analysis of the primary outcome during dose-finding, showing the data on which 1000IU/day was declared the best of the non-control doses for preventing falls; the median follow-up in the combined non-control groups was 4.5 months. Panel B displays the time-to-event primary efficacy analysis. 1000* denotes the combined 1000, 2000, and 4000IU/day groups with their observation time limited to time on best dose; the 1000* group excludes 41 participants randomized to 2000 or 4000IU/day who were never issued a bottle of 1000IU/day due to dying, completing the trial, dropping out of the trial, or stopping study pills without ever being switched to best dose. For those randomized to 2000 or 4000IU/day, at-risk time is measured and events are counted from the date of their switch to 1000IU/day. Median time to primary outcome or end of follow-up, whichever occurred first, among all 647 participants in the experience on best dose analysis was 6.5 months. 150 participants (median follow-up, 10.2 months) in the Experience on Best Dose group and 125 participants (median follow-up, 20.3 months) in the 200IU/day group who did not fall or die during follow-up were censored at their date of last contact. The hazard ratios and 95% confidence limits were derived from Cox regression models with dose group as the single model variable. In each panel, the number of participants remaining at risk of the event is shown below each time point. The events and event rates shown between pairs of time points are the number of events occurring in that time interval and the event rates over that interval. Abbreviations: CI = confidence interval; PY = person-years.

Figure 2.

Kaplan-Meier Analyses of the Composite Primary Outcome of First Fall or Death During Dose-finding (Panel A) and Over all Follow-up (Panel B) in the STURDY Trial. Panel A displays a time-to-event analysis of the primary outcome during dose-finding, showing the data on which 1000IU/day was declared the best of the non-control doses for preventing falls; the median follow-up in the combined non-control groups was 4.5 months. Panel B displays the time-to-event primary efficacy analysis. 1000* denotes the combined 1000, 2000, and 4000IU/day groups with their observation time limited to time on best dose; the 1000* group excludes 41 participants randomized to 2000 or 4000IU/day who were never issued a bottle of 1000IU/day due to dying, completing the trial, dropping out of the trial, or stopping study pills without ever being switched to best dose. For those randomized to 2000 or 4000IU/day, at-risk time is measured and events are counted from the date of their switch to 1000IU/day. Median time to primary outcome or end of follow-up, whichever occurred first, among all 647 participants in the experience on best dose analysis was 6.5 months. 150 participants (median follow-up, 10.2 months) in the Experience on Best Dose group and 125 participants (median follow-up, 20.3 months) in the 200IU/day group who did not fall or die during follow-up were censored at their date of last contact. The hazard ratios and 95% confidence limits were derived from Cox regression models with dose group as the single model variable. In each panel, the number of participants remaining at risk of the event is shown below each time point. The events and event rates shown between pairs of time points are the number of events occurring in that time interval and the event rates over that interval. Abbreviations: CI = confidence interval; PY = person-years.