The study design and pre-specified data analysis plan for the BEST-D trial have been previously reported . In summary, BEST-D is a double-blind, randomized placebo-controlled, parallel group trial comparing the effects of daily supplementation with either 4000 IU (100 μg) or 2000 IU (50 μg) cholecalciferol (vitamin D3) or placebo for 12 months on biochemical markers of vitamin D status, cardiovascular risk factors and clinical tests of physical function . About 300 men and women aged 65 years or older were recruited from a single general practice in Banbury, Oxfordshire, UK. Individuals who were ambulatory, living in the community and not currently taking more than 400 IU (10 μg) vitamin D3 daily were eligible to participate. All participants provided written informed consent, and the study was approved by the National Research Ethics Service Committee (Oxford B), UK. All trial investigators, trial staff and participants were kept blinded to the treatment allocation until the trial had been completed, the data analysis plan had been finalized and the database had been locked for analysis.
Randomization and interventions
A research nurse visited participants at home and recorded a history of vascular disease or presence of vascular risk factors, falls and fractures and daily dietary intake of calcium; assessed symptoms of muscle and joint pain (using a ten-point visual scale with 10 being the most severe); and measured height, weight, blood pressure, arterial stiffness and handgrip strength (average of 3 measures on each hand) with a Jamar dynamometer . Blood pressure and arterial stiffness were also measured at each visit after 10-min rest in the seated position . First, a finger probe (PulseTrace PCA 2) was placed on the right forefinger to record the digital volume pulse using photoplethysmography over 30–60 s. This was followed by recording the mean of two blood pressure and brachial artery arterial stiffness measurements (aortic pulse wave velocity [PWV] and aortic augmentation index) made over 2 min using a TensioClinicTM® Arteriograph . Blood was collected into vacutainers containing either lithium heparin or ethylenediaminetetraacetic acid (EDTA) for analyses. Randomization to study treatment (vitamin D3 4000 or 2000 IU or placebo daily) was by telephone to the coordinating centre where a computer-based minimization algorithm was used to ensure matching of the allocated groups by age, body mass index [BMI], smoking history, ethnicity and history of fracture. Vitamin D3 and matching placebo soft gel capsules were provided by Tishcon Corporation (Westbury, NY, USA).
All participants were visited again in their homes by the study nurse at 6 and 12 months after randomization. In addition, one third of the participants were randomly selected to have a blood sample collected at 1 month. At the 6- and 12-month visits, information was recorded on compliance, serious adverse events, non-serious adverse events leading to discontinuation of study treatment, symptoms of muscle and joint pain, measures of physical function (handgrip strength) [34, 35] and mood (using the geriatric depression four-point scale) [36, 37], and a blood sample was collected. The participants were asked to attend a single assessment centre at their local general practice after their 12-month assessment to have clinical tests of physical function (chair rises, tests of balance and a 3-m walk) [34, 35] and for bone density at the heel and wrist bones measured using an OsteoSys EXA-300 scanner (OsteoSys, Seoul, Korea).
Plasma levels of 25(OH)D were selected as the best measure of vitamin D status rather than 1,25-dihydroxy-vitamin D because of its long half-life and direct relationship to intake and synthesis. Plasma 25(OH)D levels and plasma PTH were measured using an Access 2 immunoassay analyzer (Beckman Coulter Ltd., High Wycombe, England). The laboratory participated in the international DEQAS scheme for 25(OH)D and had a mean (SD) bias of −11.8% (7.5) from the target value over the period of the study. The performance target for the scheme was ±25% of the target value, which is assigned by the NIST reference measurement procedure. Plasma levels of albumin, calcium, phosphate and alkaline phosphatase were measured using a UniCel DxC 800 Synchron clinical system (Beckman Coulter Ltd., High Wycombe, England), and plasma levels of high-sensitivity C-reactive protein were measured using a BN ProSpec system (Siemens, Frimley, England); all assays used the suppliers’ reagents and calibrators. Further details of assays and performance characteristics are recorded in the Online Resource.
All efficacy and safety assessments were conducted according to the intention-to-treat principle and followed the pre-specified data analyses set out in the trial Statistical Analysis Plan . The co-primary outcomes were mean plasma 25(OH)D levels and the percentage of participants with plasma 25(OH)D levels >90 nmol/L at 12 months. The primary assessment of these outcomes was to compare participants allocated 4000 IU with those allocated 2000 IU daily (with secondary assessments including comparisons of each dose with placebo). Tertiary assessments of the co-primary outcomes involved comparison of the two active doses of vitamin D in the following subgroups: sex, age (<70, ≥70 years), body mass index (BMI <25, ≥25 to <30, ≥30 kg/m2), plasma 25(OH)D level at baseline (≤50, >50 nmol/L), dietary calcium intake (≤700, >700 mg/day), estimated glomerular filtration rate (eGFR ≤75, >75 ml/min/1.73 m2) and history of cardiovascular disease (CVD). A range of secondary and tertiary outcomes were also pre-specified, including mean 25(OH)D and percentage of participants with 25(OH)D >90 nmol/L at 1 and 6 months; percentage of participants with PTH in the reference interval at 1, 6 and 12 months; percentage of participants with calcium above the reference interval at 1, 6 and 12 months; mean values of all other biochemical tests and measures of vascular function at 6 and 12 months; and physical function at 12 months. Comparisons of mean follow-up values between treatment arms involved analysis of covariance (ANCOVA) adjusted, where possible, for the baseline value (with multiple imputation used to impute the few missing data). ANCOVA provides a more powerful test of the null hypothesis than either a comparison of mean follow-up values in isolation or a comparison of mean changes from baseline . Comparisons of dichotomous outcomes were done using standard methods for 2 × 2 contingency tables. All p values were two sided and considered statistically significant, without allowance for multiple testing, if they were <0.05 . Analyses were conducted using SAS version 9.3 and R version 2.11.1. The study was designed to have >90% power (at 2p=0.01) to detect a true difference in mean 25(OH)D between the two active doses at 12 months of 11 nmol/L (assuming an SD of 20 nmol/L). In the two active dose arms, it also had >80% power at 2p=0.05 to detect an increase in the proportion achieving a 12-month concentration >90 nmol/L from 60% to 80%. All of the analyses were conducted independently of the sources of support.
Figure 1 shows the outcome of the 1122 people who were assessed for eligibility for invitation to participate in the trial. Between 24 September 2012 and 14 March 2013, 305 participants were randomly allocated to take either 4000 IU D3 (n = 102), 2000 IU D3 (n = 102) or placebo (n = 101) daily.
The mean age was 72 years, 51% were men, and 7% were current smokers (Table 1). About 12% reported taking low-dose vitamin D3 (400 IU daily or less), and 3% reported using calcium supplements. While the self-reported physical activity ratings were high, one third of participants reported muscle aches and pains and two thirds reported joint aches and pains.
Among those allocated to 4000 IU, 2000 IU or placebo, 93%, 93% and 87%, respectively, reported taking their capsules on all or most days at 6 months, while 90, 92 and 85% reported doing so at the 12-month visit. Overall, only 5%, 4% and 6% of those allocated to 4000 IU, 2000 IU or placebo were unable to attend their scheduled final visits (Fig. 1).
Effects on plasma 25(OH)D levels
Mean (SD) plasma 25(OH)D levels were about 50 (18) nmol/L at baseline and increased to 137 (39), 102 (25) and 53 (16) nmol/L respectively, after 12 months of treatment among those allocated 4000 IU, 2000 IU or placebo, respectively (Fig. 2, Table 2), and 88%, 70% and 1%, respectively, achieved a 25(OH)D level >90 nmol/L at 12 months. Figure 2 shows the mean plasma levels of 25(OH)D when measured on three to four visits over 12 months. In the 100 participants evaluated at 1 month, mean plasma 25(OH)D levels were already significantly elevated (Fig. 2). By 6 months, 86% and 64% of participants allocated to 4000 and 2000 IU vitamin D, respectively, had plasma 25(OH)D >90 nmol/L (compared with 2% of placebo-allocated participants). Between 6 and 12 months, the 25(OH)D mean levels continued to increase by 11 nmol/L and 5 nmol/L, respectively.
Effects in pre-specified subgroups
The effects of 4000 IU daily versus 2000 IU daily of vitamin D on plasma levels of 25(OH)D after 12 months were similar in all pre-specified subgroups, except when grouped by BMI, where the further increase from the higher dose on plasma 25(OH)D levels was smaller among those with higher baseline BMI (Fig. 3: p for trend <0.0001). The effects of 4000 versus 2000 IU dose of vitamin D on plasma levels of 25(OH)D were attenuated by one third in those who were overweight and by two thirds in those who were obese, compared with those with normal BMI. In a post hoc analysis of the effects of 2000 IU vitamin D versus placebo, however, the achieved difference in plasma 25(OH)D levels was similar in every subgroup, including by baseline BMI (Online Resource Fig. 1). The achieved differences in plasma levels of 25(OH)D at 6 and 12 months between those allocated 4000 versus 2000 IU daily and versus placebo were broadly similar when subdivided by the quartiles of plasma 25(OH)D levels at baseline (Online Resource Table 1).
Effects on plasma PTH, calcium levels and alkaline phosphatase
Plasma levels of PTH at baseline were balanced by treatment allocation but were outside the laboratory normal range (1.1–6.8 pmol/L) in 22 participants. During follow-up, the proportion of participants with plasma PTH levels within the normal range was similar between those allocated 4000 IU daily, 2000 IU daily or placebo (respectively, 89%, 95% and 88% at 6 months and 91%, 95% and 87% at 12 months). Compared with placebo, mean PTH decreased significantly (p < 0.0001) with both doses of vitamin D (Fig. 2, Table 2), with lower mean plasma PTH levels at 6 and 12 months among participants allocated 4000 IU compared with participants allocated 2000 IU daily (p = 0.01 and p = 0.03, respectively). The reduction from baseline in plasma PTH levels at 12 months was 22% in the 4000 IU group and 14% in the 2000 IU group. Mean plasma levels of albumin-corrected calcium were also significantly increased by both vitamin D3 doses (Fig. 2, Table 2), but the absolute increases were small and not clinically apparent. The differences in 6- and 12-month plasma levels of albumin-corrected calcium between participants allocated 4000 IU daily and those allocated 2000 IU daily were also small (Table 2) and did not vary by baseline plasma levels of albumin-corrected calcium (Online Resource Table 2). Likewise, vitamin D3 had no significant effect on mean plasma levels of alkaline phosphatase at 12 months (Table 2). In post hoc analyses, there was no evidence that baseline plasma levels of plasma 25(OH)D modified the effect of allocation to vitamin D on 12-month alkaline phosphatase levels (p for interaction 0.89).
Effects on cardiovascular risk factors and clinical tests of physical function
Table 3 shows the effects of vitamin D3 (both doses combined) on cardiovascular risk factors and on the physical function measurements at 12 months; Online Resource Table 3 gives the estimates at 6 months. Allocation to either dose of vitamin D3 had no significant effect on the physical measurements or measures of arterial stiffness at 12 months or on 12-month levels of total or LDL cholesterol, triglycerides, apolipoprotein B, NT-proBNP, phosphate, creatinine or urinary albumin/creatinine ratio (Table 3). Mean plasma levels of HDL cholesterol, apolipoprotein A1, C-reactive protein and albumin at 12 months were all slightly, but significantly, lower among participants allocated vitamin D3 (without allowing for multiple testing).
There were no significant differences in the reported clinical or adverse events with any vitamin D3 intake versus placebo on clinical events at 12 months or on self-reported fractures, falls, muscle pain severity, joint pain severity and physical activity ratings compared with those allocated to placebo (Table 3). Geriatric depression scores were also similar in both groups, as were the results of measures of physical function including handgrip strength, chair rises, balance and a 3-m walk. Moreover, the results of these clinical events and physical measures were similar at 6 months (Online Resource Table 3) and when the three treatment arms were considered separately (Online Resource Table 4). Bone density T-scores at the heel and wrist after 12 months were also broadly similar between the two groups (the p value of 0.03 for a lower heel T-score associated with allocation to vitamin D3 would not be statistically significant if allowance was made for multiple testing and levels at baseline were not measured to allow for any baseline imbalance to be assessed).
Safety and tolerability
At randomization, albumin-corrected calcium was mildly elevated (>2.55 mmol/L) in eight participants (six who were subsequently allocated vitamin D3 and two who were subsequently allocated placebo). By 12 months, there were no new cases with elevated plasma levels of albumin-corrected calcium. Among the eight participants who had marginally elevated plasma levels of albumin-corrected calcium at baseline, six (five allocated vitamin D3 and one allocated placebo) still had elevated levels at 12 months, but none was considered clinically significant. Mean plasma levels of creatinine and phosphate in each of the treatment groups at 6 and 12 months were similar and similar to baseline values (data not shown).
At least one serious adverse event (SAE) was reported by 29 participants allocated 4000 IU, 30 allocated 2000 IU of vitamin D and 25 allocated placebo (Online Resource Table 5), respectively, and included three deaths (all deaths were among those allocated to placebo). None of the SAEs was considered treatment-related events. Study treatment tolerability was good and was discontinued before the scheduled end of the study by 17 participants (5 allocated 4000 IU, 5 allocated 2000 IU and 7 allocated placebo). A SAE was attributed as the reason for discontinuation in 3 participants (one in each group), and a non-SAE was the reason for discontinuation in 4 participants (1 allocated 4000 IU and 3 allocated placebo).