High-dose oral vitamin D supplementation and mortality in people aged 65 84 years: the VIDAL cluster feasibility RCT of open versus double-blind individual randomisation

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 08/116/48. The contractual start date was in November 2011. The draft report began editorial review in December 2017 and was accepted for publication in October 2018. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

Irwin Nazareth was on the Health Technology Assessment (HTA) Commissioning Board from 2012 to July 2017. For the duration of the Vitamin D and Longevity (VIDAL) trial, Irwin Nazareth’s PRIMENT Clinical Trials Unit was funded by the National Institute for Health Research (NIHR). He was a member of the HTA Disease Prevention Panel, a member of the HTA Commissioning Sub-board (Expression of Interest) and a member of the HTA Primary Care Themed Call. Benoit Aigret reported that Queen Mary University of London received a grant from the London School of Hygiene & Tropical Medicine to develop the VIDAL online application during the conduct of the study.

Copyright statement

© Queen’s Printer and Controller of HMSO 2020. This work was produced by Rake et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

2020 Queen’s Printer and Controller of HMSO

Why then does the UK need a ‘vitamin D megatrial’ of its own?

We propose two reasons:

1. The UK represents a setting with a high prevalence of vitamin D insufficiency, where supplementation could have maximal impact. Median serum 25(OH)D concentrations among older adults in the UK (37–49 nmol/l, depending on season) are significantly lower than in the countries where large trials are currently being conducted (New Zealand, 66 nmol/l;36 Australia, 69 nmol/l;37 Canada, 70 nmol/l;38 USA, 57 nmol/l39). The efficacy of vitamin D supplementation is likely to depend on the prevalence of inadequate vitamin D status at baseline, so the results of the intervention studies conducted in these settings are likely to underestimate any effects that would be seen in older adults living in the UK and are less likely to achieve statistical significance. International differences in baseline vitamin D status may be partly attributable to the fact that many of the countries listed in Table 1 are situated at lower latitudes than the UK and, therefore, their populations have greater exposure to sunshine of sufficient intensity to stimulate cutaneous vitamin D synthesis; moreover, many of these countries routinely fortify foods with vitamin D (e.g. milk in the USA, Canada and Finland is routinely vitamin D-fortified).

2. Conduct of a further large trial of daily vitamin D supplementation in the UK will add substantially to meta-analysis of these four megatrials to detect and estimate a modest but clinically significant effect of vitamin D on all-cause mortality among participants with low serum 25(OH)D, among whom any effect is likely to be concentrated. Apart from FIND (Finnish Vitamin D Trial)40 in Finland, which stopped recruitment at 2500 participants (target 18,000 participants) with only 830 participants allocated to 3200 IU daily (see Table 1 footnote), the proposed trial would provide the only evidence on the effects of a daily dose of the order of 4000 IU. Three trials (VIDA,41,42 TIPS-3 and D-Health) are testing monthly dosing, which may be less effective. The VITamin D and OmegA-3 TriaL (VITAL), which tested 2000 IU daily, included only 3318 participants aged ≥ 75 years,43 one-third of the number proposed in the VIDAL main trial.

The National Institute for Health Research (NIHR) Health Technology Assessment (HTA) programme did not support our application in 2008 to conduct a large trial of vitamin D supplementation in older adults in the UK, but invited an application for this feasibility study to establish the procedures required to conduct the main trial. Funding was awarded in 2011 and recruitment began in 2013. An application for funding for the main trial was made again in January 2015 when recruitment was complete, but by that time the trials listed in Table 1 were under way and the NIHR HTA programme decided that funding for a large British trial would not be reconsidered until the results of those trials become available. The results of the VITAL trial, which is the only large trial of daily high-dose vitamin D, are inconclusive for all-cause mortality [hazard ratio 2 to 5 years after entry 0.96, 95% confidence interval (CI) 0.84 to 1.11], so further evidence is now needed. 43

Trial design

The Vitamin D and Longevity (VIDAL) feasibility trial was a four-arm multicentre RCT of 2 years’ duration of subjects aged 65–84 years. Twenty general practitioner (GP) practices in England were cluster randomised in matched pairs to either double-blind or open-label study design. The GP practices were assigned to pairs matched approximately on size, whether urban or rural, ethnic mix and ward multiple deprivation index based on practice postcode. The practices in each pair were then randomly assigned to double-blind or open-label individual randomisation. In double-blind practices participants were individually randomised to blind vitamin D (BD) or blind placebo control (BC). In open-label practices individual randomisation of participants was to open-label vitamin D (OD) or untreated open control (OC).

Ethics and regulatory approval and research governance

Ethics approval for the study was given by the London–Chelsea National Research Ethics Service in February 2012 (reference number 11/LO/1989). Clinical trial authorisation for the study was given by the Medicines and Healthcare products Regulatory Agency (MHRA) in March 2012 (reference number 17072/0006/001-0001). Appropriate site-specific assessments were obtained from the primary care trusts to confer the required management permissions for the 20 participating GP practices. The trial was registered with the International Standard Randomised Controlled Trial Register under the reference number ISRCTN 46328341 and also with the European Union Drug Regulating Authorities Clinical Trials (EudraCT) database under the reference number 2011-003699-34.

Eligibility

Members of the general population were recruited from 20 GP practices across England.

Recruitment procedure

Each of the 20 GP practices generated a list of registered patients aged 65–84 years. After excluding ineligible individuals, including those judged by the GP to lack the mental capacity to give informed consent or to be unsuitable for other reasons, study invitations were sent by post to potential participants.

Interested respondents were telephoned by a member of the research team at the GP practice to confirm eligibility and arrange a baseline assessment appointment at the GP practice, at which time informed consent was obtained from those agreeing to participate in the trial. Trial participants then provided information on current medications and conditions, diet (including dietary supplements), skin type and sun exposure, and QoL (see Appendix 1). Virtually all information was entered directly into case report forms (CRFs) accessed via the online clinical data management system [the VIDAL app (online application); see Appendix 2], with paper copies of all CRFs available as a back-up option. Systolic and diastolic blood pressure (BP), height, weight and waist circumference were recorded, and 12.5 ml of blood [9 ml ethylene diamine tetra-acetic acid (EDTA) vacutainer for measurement of 25(OH)D and 3.5 ml serum separator tube (SST) vacutainer for calcium assay] were obtained. The 3.5-ml aliquot was sent for corrected calcium assay at the practice’s local laboratory to verify eligibility. The 9-ml aliquot was sent to the Clinical Trials Service Unit in Oxford for separation of buffy coat and storage in liquid nitrogen. Circulating 25(OH)D was assayed on all stored samples (baseline and 2-year follow-up) at the end of the study on the Cobas 6000 immunoassay (Roche Molecular Diagnostics, Pleasanton, CA, USA).

Informed consent

Informed written consent was obtained during the baseline assessment from eligible participants after an explanation of the aims, methods, anticipated benefits and potential hazards of the study. The original signed and dated consent forms were held at each GP practice, with copies sent to the participant and the Trial Coordination Centre. Patients ineligible for inclusion, based on the corrected blood calcium result, were informed of this by their GP or GP nurse, who also discussed whether or not any treatment was indicated.

Randomisation, concealment and blinding

Follow-up

To obtain information on their vitamin D consumption for an interim report, 121 participants who were randomised to the no treatment arm in open practices (OC arm) before May 2014 were contacted by post, e-mail or telephone in December 2014. There was no other contact after randomisation with the OC participants until they were invited to attend the 2-year final visit. All other study participants (i.e. those allocated to active treatment or placebo) were contacted at least once per month post randomisation by their preferred medium, as described in the following paragraphs. Letters were sent by CFH Docmail Ltd (Radstock, UK) on behalf of the participating GP practices, with responses sent directly back to the study participant’s GP practice.

Interim Data Monitoring Committee reports

A Data Monitoring Committee (DMC) was convened. Interim safety analyses were conducted twice during the feasibility study. The independent statistician conducted an analysis to compare the incidence of SAEs between intervention and control arms for review by the DMC. Had there been a significant difference (p < 0.05) in the incidence of fatal or life-threatening adverse events, the DMC would have been informed and would have discussed whether or not the sponsor and ethics committee should be consulted regarding stopping the trial, but this did not arise.

Trial outcomes

Sample size

The aim of the feasibility study was to recruit 1600 participants aged 65–84 years through 20 GP practices [400 on OD vs. 400 on OC; 400 in the blind vitamin D (BD) arm vs. 400 in the blind placebo control (BC) arm]. The target was to randomise an average of 80 participants aged 65–84 years per GP practice with at least 9% response (number randomised/number invited). If recruitment in some GP practices fell below this target, recruitment in other practices would be continued after 80 participants had been randomised to achieve the overall target of 1600 participants.

Statistical methods

All analyses were performed on an intention-to-treat basis using Stata^® version 15 (StataCorp LP, College Station, TX, USA). Wilcoxon’s signed-rank test was used for comparisons of blind versus open practices within matched pairs.

Recruitment

The CONSORT (Consolidated Standards of Reporting Trials) flow diagram summarising the number of individuals participating at each stage of the trial is shown in Figure 2. In addition to data collected from the participants at study visits, data regarding prescriptions, infections and GP visits were downloaded from each GP database. GP data were obtained for 1554 participants (96.2%) but this varied by GP practice. Some practices did not provide data from patients who had moved GP or who had died, as these data were no longer available on their computer systems. Figure 3 shows cumulative recruitment. The pilot practice began recruiting in April 2013. Main recruitment began in October 2013 and ended in January 2015 with 1615 participants randomised (the target was 1600 participants). The recruitment period for individual practices ranged from 4 to 12 months.

FIGURE 2.

The VIDAL trial CONSORT flow diagram.

FIGURE 3.

Cumulative VIDAL recruitment by month and year.

Table 2 shows the numbers of registered patients in each practice invited, replying, attending the baseline visit and randomised. GP practice codes indicate matched pair (0–9) and whether or not randomisation was open-label or blind (O or B).

The overall recruitment rate (number randomised/number invited) was 14.2%. The rate was higher in open (15.0%, range 8.8–22.4%) than in blind practices (13.4%, range 8.8–26.4%), but this did not approach statistical significance because of the wide variation between practices (Wilcoxon signed-rank test; p = 0.7). Table 3 shows that the recruitment rate was lower (p = 0.002) in participants aged 80–84 years (11.5%) than in participants aged < 80 years (14.6%), and was lower in women than in men (p = 0.002).

Table 4 shows the 1615 randomised participants by age and sex. There were 857 (53.1%) men and 758 (46.9%) women, with similar age distributions. The majority (70.2%) were aged 65–74 years and only 9.7% were 80–84 years. Almost all were white (Table 5: 1600/1615). Table 6 shows numbers by age and sex of those who chose to receive and return quarterly follow-ups by e-mail. The proportion choosing e-mail was higher among men and declined with age, from 77.4% in those aged 65–69 years to 36.5% in those aged 80–84 years. The proportion choosing e-mail for monthly reminders to take their study medication also declined with age, from 70.5% (440/624) of those aged 65–69 years to 33.3% (52/156) of those aged 80–84 years, most of whom (79.5%) requested a monthly telephone call. The proportion requesting text message reminders declined from 40.1% (250/624) in those aged 65–69 years to 9.0% (14/156) in those aged 80–84 years (more than one medium could be chosen).

Compliance

Tables 7 and 8 and Figure 4 show compliance among the 1194 participants randomised to receive study medication (the blind treatment arms or the OD arm). Excluding participants who died during the trial period, 89.9% (1059/1178) were still taking medication at the end of the study, 91.2% in the OD arm and 89.2% in the blind treatment arms. All 24 doses of study medication were taken by 80.2% (625/779) of participants in the blind treatment arms and 83.2% (332/399) of those in the OD arm. Eleven participants in the blind treatment arms (1.5%) and four (1.0%) in the OD arm did not take any study medication from the outset. Compliance was higher among men than among women and among participants aged < 75 years: all 24 doses were taken by 86.6% of men and 79.0% of women aged < 75 years, and by 83.8% of men and 69.9% of women aged ≥ 75 years.

FIGURE 4.

Compliance of participants taking study medication in each quarter (percentage of those randomised taking all three monthly doses vs. at least one dose) by randomisation method (blind or open-label).

Occasional doses were missed for various reasons and so a more useful measure of compliance is the date when a participant last took their study medication. Within 6 months of entry, 16 (4.0%) participants allocated to OD and 34 (4.4%) on blind treatment (BC or BD) had stopped taking study medication. The subsequent rate of decline (Figure 5), which was slightly lower, is likely to provide a better estimate of the proportion who would continue treatment in a longer trial. Among those compliant at 6 months, 6.2% (70/1128) stopped taking medication over the remaining 18 months of the trial [5.0% (19/383) on OD and 6.8% (51/745) on blind treatment; p = 0.21], which is an overall annual attrition of 4.1% per year. If this rate of attrition continued, 81.1% of all randomised participants would still be taking allocated medication at the end of a 5-year trial.

FIGURE 5.

Proportion of participants who were compliant at 6 months still taking study medication over the remainder of the trial. Within 6 months of entry 4.0% of participants allocated to OD and 4.4% on blind treatment had stopped taking study medication.

Contamination

Table 11 shows that almost one-quarter of participants (23.2%) reported taking self-administered daily supplements containing vitamin D at baseline, but the majority of these contained ≤ 200 IU (5 µg). GP notes abstracted at the end of the study revealed that an additional 17 (1.1%) participants were being prescribed medication containing vitamin D at baseline, including six at > 400 IU per day. These six, together with the three individuals who were taking > 400 IU (10 µg) per day (see Table 11 footnote a), were missed by the practice nurse when eligibility was checked at baseline. At follow-up, the proportion reporting self-administered supplements containing vitamin D (16.5% overall) had fallen in the OD arm and in both blind treatment arms but remained unchanged in the OC arm. The proportion of individuals taking high-dose vitamin D increased in the OC arm and the blind treatment arm, but only four individuals reported taking > 1000 IU (25 µg) per day. At 2 years, 3.6% of participants with available data were receiving > 400 IU per day of additional vitamin D (1.1% self-administered, 2.2% prescribed). Contamination data were collected by telephone from 65 participants who did not attend the 2-year visit but remained unknown for 91 participants, including 20 who died during the trial and a further three who died before the telephone follow-up call.

Composite compliance

‘Composite compliance’, as defined in Chapter 2, Methods, was calculated for all participants who did not die within 2 years of randomisation. All between-practice differences were tested using Wilcoxon‘s signed-rank test with 10 matched pairs of practices. Among the control participants, a significantly higher proportion were compliant in blind practices than in open practices (89.7% BC vs. 83.0% OC; p = 0.01). This difference was a result of the lower attendance among untreated OCs at the follow-up visit (94.1% BC vs. 87.8% OC; p = 0.01), as the proportions exceeding the permitted total dose of vitamin D supplements (300,000 IU over 2 years) were similar. There was an opposite but non-significant difference for the composite compliance variable among those allocated to vitamin D, with higher compliance among open practices (91.0% OD vs. 86.9% BD; p = 0.07) owing to slightly higher attendance at the follow-up visit (93.2% OD vs. 91.0% BD) and a slightly higher proportion taking at least 19 doses (92.7% OD vs. 89.2% BD). These opposite effects led to similar overall composite compliance of 88.3% (688/779) in the blind practices and 86.9% (709/816) in the open practices (p = 0.43).

Safety

At least one SAE was reported during the 2-year trial period for 11.6% of participants allocated to study medication (Table 12), the majority of which resulted in hospitalisation. None of the 184 reported SAEs was judged to be associated with the study medication. There were no significant differences between the SAEs reported between the vitamin D arms and the blind placebo arm. Only seven SAEs were not a cancer diagnosis, did not necessitate hospitalisation and did not result in death. The SAEs were not reported during the trial by patients in the untreated OC arm of the study, as they did not return quarterly follow-up forms. Rates of emergency hospitalisation from HES data44 are slightly higher than the self-reported episodes for the patients allocated to study medication (11.6% vs. 10.4%), and much higher for OCs (12.4% vs. 1.7%), as expected. There were no significant differences in the hospitalisation rate between the open and blind arms of the trial (11.8% in both arms spent at least one night in hospital). Very few participants were hospitalised for > 30 days in total during the 2 years following randomisation [8/813 (1.0%) in the control arms and 7/802 (0.9%) in the vitamin D arms]. Table 12 shows that the notification source for 116 (91.3%) of the 127 patients with SAEs other than death was self-report on quarterly follow-up forms (n = 101) or retrospective report at the 2-year follow-up (n = 15). GPs reported only an additional 11 non-fatal SAEs (8.7%), and none in OCs. ARs were rare, with a non-significantly higher rate on OD (2.9%) than on BD (1.1%) (Wilcoxon signed-rank test; p = 0.13). Neither SAEs nor ARs were significantly related to treatment in blind practices.

Quality of life

The self-assessed QoL score at baseline decreased with age, as expected (see lifestyle questionnaire in Appendix 1). At baseline, 23.9% of individuals reported problems with mobility, 3.1% reported problems with self-care, 11.1% reported problems performing usual activities, 41.4% reported some pain and 12.5% reported some anxiety or depression. Overall, 49% of individuals reported no problems with mobility, self-care or performing usual activities and did not report pain, discomfort, anxiety or depression (Table 13). The lowest QoL scores (reflecting some trouble with four or five of the indicators or extreme trouble with two or more indicators) were reported by 4% of the population overall, but this varied by GP practice (0–17%).

Serum 25(OH)D concentrations

Blood samples valid for 25(OH)D analysis were collected for 1608 participants at baseline and 1448 participants at the year 2 visit. (Four baseline samples were not received at the laboratory and three were insufficient.) Overall, 82.6% of participants had baseline 25(OH)D below the 75 nmol/l threshold. Baseline levels by demographic and lifestyle factors reported at baseline are shown in Table 14. Levels decreased with increasing age [p(trend) = 0.01] and were lower in women (p < 0.0001). Average levels were highest in summer and autumn, lower in winter and even lower in spring (p < 0.0001) and increased with skin darkness from the fairest skin to participants who reported that they rarely burn and always tan (olive skin) (p < 0.0001). The small number of participants who described their skin colour as brown or black had lower levels than those with olive skin. Average levels increased with frequency of eating oily fish, although the mean level of 66 nmol/l in participants who reported eating oily fish more than four times a week was still below the adequacy threshold of 75 nmol/l. Levels were lower in participants living at higher latitude (p = 0.03 adjusted for IDAOPI deprivation score). The three right-hand columns in Table 14 show adjusted means and significance levels from a multiple regression including all variables in the table. The trend of reduced 25(OH)D with increasing age was virtually eliminated after adjustment for all variables [p(trend) 0.01 unadjusted, 0.6 adjusted]. The magnitude and significance of trends for other variables were slightly weakened or unaffected by adjustment.

Treatment effects

Figure 6 and Table 15 show blood 25(OH)D levels at baseline and at 2 years in control participants and in those randomised to receive vitamin D. The mean blood 25(OH)D levels at follow-up were 109.6 nmol/l in those allocated to vitamin D and 51.8 nmol/l in control participants. The proportion ≥ 75 nmol/l was 16.4% at baseline and 88.0% at 2 years in the vitamin D treatment arms but remained unchanged in the control arms (18.3% at baseline, 17.9% at 2 years). The proportion ≥ 75 nmol/l at 2 years was 51.7% (15/29) in non-compliant participants allocated to vitamin D. Among those allocated to vitamin D, mean blood levels decreased with time since last dose from 117.6 nmol/l in those tested within a month of last dose (220 participants) to 108.4 nmol/l in those tested 1–3 months after last dose (463 participants) and 80.0 nmol/l in those tested ≥ 4 months after last dose (38 participants).

FIGURE 6.

Baseline blood 25(OH)D levels by treatment arm at (a) baseline and (b) 2 years. The vertical line shows the suboptimal 75 nmol/l threshold.

Figure 7 shows that vitamin D levels increased for almost all (99.3%) of those randomised to receive vitamin D. The estimated mean increase was 57.9 nmol/l (95% CI 56.3 to 59.4 nmol/l) in the vitamin D arms compared with 0.2 nmol/l (95% CI –1.4 to 1.7 nmol/l) in the control arms. Figure 8 shows the change in each arm and Figure 9 shows the smaller change in non-compliant (OD and OB) participants who took vitamin D for < 2 years. A larger proportion of participants in the control arms had their year 2 visit in winter/spring and baseline visit in summer/autumn than those in the vitamin D arms (8.7% vs. 6.7%, respectively) (Table 16), so the change in season was adjusted for yielding an estimated mean increase of 57.8 nmol/l (95% CI 56.1 to 59.5 nmol/l) in the vitamin D arms compared with 0.3 nmol/l (95% CI –1.4 to 2.0 nmol/l) in the control arms. In the control arms, the effect of season change was fairly small with an estimated increase of 5.7 nmol/l (95% CI 3.0 to 8.3 nmol/l) for those whose baseline blood was taken in winter/spring and follow-up was taken in summer/autumn. Conversely, the estimated decrease in vitamin D levels in control participants whose blood samples were taken in summer/autumn at baseline and winter/spring at follow-up was 8.8 nmol/l (95% CI 4.9 to 12.6 nmol/l).

FIGURE 7.

Baseline and follow-up blood 25(OH)D levels by allocated treatment arm in 1444 participants who provided two blood samples. Participants stopping their vitamin D medication early are shown as solid blue points.

FIGURE 8.

Change from baseline to 2 years in blood 25(OH)D levels by treatment arm in 1444 participants who provided two blood samples. No change is denoted by the vertical line at zero.

FIGURE 9.

Change in blood 25(OH)D levels in those allocated to take vitamin D who were compliant and those who stopped taking their medication (non-compliant, n = 41).

Table 17 shows significantly lower blood 25(OH)D levels at baseline in all participants and at follow-up among control participants who did not take supplements containing vitamin D (p < 0.0001). The mean blood 25(OH)D (nmol/l) was 47.0 (95% CI 45.5 to 48.5) among those not taking supplements compared with 65.5 (95% CI 62.8 to 68.1) among those taking supplements at baseline, and similar results were seen at follow-up among the control participants. However, this higher level was still below the optimal threshold of 75 nmol/l. Additional supplementation did not change the blood 25(OH)D levels among those randomised to take vitamin D.

Tables 18 and 19 show the number of infections during the 2-year trial period as reported from GP notes for all patients, comparing control participants (untreated or placebo) with vitamin D participants (open-label or blind). The overall number of infections during the 2-year trial period did not differ between control and vitamin D arms (see Table 19): 28.0% of those in the control arms had at least one infection compared with 26.8% in the vitamin D arms. A slightly higher proportion of control participants with low blood 25(OH)D had at least one infection during the trial period compared with those allocated vitamin D (31.6% vs. 25.9% in the < 25 nmol/l group and 29.1% vs. 28.4% in the 25–49 nmol/l group).

There were no significant differences between the treatment arms in changes from baseline to follow-up for systolic BP and BMI (Table 20). A small increase in diastolic BP and a small decrease in health score were seen among those allocated to take vitamin D (p = 0.03 and p = 0.04, respectively). No significant differences were seen between the treatment arms regarding changes in the five indices of QoL (Table 21). On average, participants lost 0.47 cm in height and 0.55 kg in weight over the 2 years of the trial.

Table 22 shows the number of GP appointments recorded in the GP notes during the year before randomisation and during the 2-year trial period. There was a mean of 5.7 appointments in the year preceding randomisation with no significant differences between the treatment arms (i.e. 5.5 appointments in the OC arm, 5.4 in the OD arm and 5.9 in both blind treatment arms). The mean number of appointments per year increased to 6.3 during the trial, with no significant differences either between year 1 and year 2 or between treatment arms.

Table 23 shows the mortality and cancer incidence for the 1615 participants in the trial. All were flagged with NHS Digital, providing complete follow-up data. The overall death rate was 4.2% in the vitamin D treatment arms and 2.8% in the untreated control arms (p = 0.12), with a cancer incidence rate over 2 years of 2.6% in those allocated to vitamin D and 3.1% among those allocated to the control arms (p = 0.6). Cause of death is shown in the lower part of the table. This is not yet available for the most recent deaths (5/57).

Preliminary procedures

Study approvals occupied almost 3 years (July 2011–May 2014) before recruitment could begin at all GP practices. Research Ethics Committee and MHRA approvals were straightforward, and the major delay was in obtaining approvals for NHS research and development and service support costs in each Clinical Research Network. In view of this experience, a formal review of the costs and benefits of expediting these processes for non-commercial population-based trials of this sort might be evaluated in consultation with the relevant agencies. Protracted correspondence and discussions with NHS Digital also delayed linkage to HES and other databases to obtain follow-up data for death, cancer registration and hospital admissions. Our application to NHS Digital underwent 13 revisions for these linkages. A major focus was the wording of our informed consent, which had been reviewed and approved both by our patient advisory group and by senior staff at the Health and Social Care Information Centre before the trial began. Requested amendments included the wording of the London School of Hygiene & Tropical Medicine’s Data Protection Act registration, and the VIDAL trial website’s description of fair processing, linkage processes, data flow, research outputs, the target audience and benefits to health and social care.

Recruitment

Recruitment in the pilot practice (practice 0B) began in April 2013 and was completed in September 2013. Procedures modified over this period included improvements to the web-based clinical data management system (the VIDAL app, see Appendix 2). The recruitment period in the other 19 practices ranged from 4 to 12 months, and the last participant was randomised in January 2015. Participation was better than expected, with one in seven (14.2% compared with the protocol target of 9%) of all invited eligible patients aged 65–84 years being randomised. Despite the matching of practice pairs on region and social deprivation, recruitment varied widely within matched pairs (see Table 2), and the slightly higher overall recruitment in open practices (15.0% of invited patients randomised) than in blind practices (13.4% randomised) did not approach statistical significance.

Compliance

The proportion of participants allocated to a treatment arm who took all three doses of study medication was 88.7% in the last 3 months of the trial (see Table 7). The proportion was slightly, but not significantly, higher for OD participants than for blind participants throughout (see Figure 4), and this non-significant trend was also seen in the total number of doses taken (see Table 8) and in the proportion of participants compliant at 6 months who were still taking study medication over the remainder of the trial (see Figure 5).

Feasibility of the main trial

The recruitment target in the VIDAL main trial is 20,000 participants aged 65–84 years with equal numbers in each 5-year age group (i.e. 200 practices recruiting an average of 100 participants with 25 in each age group). The recruitment rate was 11.5% at age 80–84 years and higher at < 80 years (see Table 3), so this uniform age distribution could be achieved if the average number of registered patients aged 80–84 years in participating practices was 220 and all were invited. The average number of registered patients aged 80–84 years per practice in England is 190, so this should be easily achievable by targeting larger practices.

The proportions of participants choosing e-mail for quarterly follow-up at ages 65–69, 70–74, 75–79 and 80–84 years were 77.4%, 67.6%, 55.7% and 36.5%, respectively (see Table 6). The corresponding participation rates (see Table 3) imply that in a trial restricted to participants willing to be contacted by e-mail, the participation rates at ages 65–69, 70–74, 75–79 and 80–84 years would be 10.5%, 11.0%, 7.9% and 4.2%, respectively. To recruit an average of 25 patients in each age group, the following numbers of patients would need to be invited on average per practice: 240 patients aged 65–69 years; 230 aged 70–74 years; 320 aged 75–79 years; and 600 aged 80–84 years. The average numbers of registered patients per practice in England in these age groups are 380, 300, 250 and 190, respectively. The main trial could thus be restricted to participants willing to receive and reply to follow-ups by e-mail for those aged < 75 years, or < 80 years if larger practices were targeted, but postal contact would be required for many participants aged 80–84 years to recruit equal numbers in each age range.

A limitation of this study was the virtual absence of recruits of Asian, African or Caribbean origin (see Table 5). Our protocol for the main trial specifies that practices in areas with large numbers of people from these ethnic groups should be invited to participate, but we have no information on the participation rate that could be achieved in such practices.

A further limitation is the monthly dosing regimen, which we would not now recommend. The suggestion that daily dosing may be clinically more effective than our monthly regimen was published during the trial. 47,48 A feasibility study of daily vitamin D₃ supplementation among men and women aged > 65 years at a GP practice in Oxfordshire (the ‘BEST-D’ study)49,50 showed that allocation to 4000 IU of daily vitamin D increased 25(OH)D on average by 80 nmol/l and 88% of participants achieved plasma levels of 25(OH)D > 90 nmol/l. Compliance levels were comparable to those in the current feasibility study, and we would therefore recommend that daily regimen in the main trial.

Our recruitment, compliance and contamination results are based on patients from the 20 participating practices and are inevitably limited to the 2-year treatment duration of the trial. We have no direct evidence of differences between these selected practices and the 200 practices that would be required for the main trial, or on changes that would occur over the 5-year treatment period of the main trial.

Contamination in untreated open control participants

Table 11 shows that only 20 (5.0%) of the 400 untreated control participants who were interviewed at 2 years (366 who attended the 2-year visit and 34 non-attenders who were telephoned) reported taking > 400 IU of vitamin D per day (11 prescribed and nine self-administered), and only one was taking > 1000 IU per day. This very low level of self-reported contamination is confirmed by their blood 25(OH)D levels, which were similar to their baseline results (see Table 15 and Figure 6). Significant contamination was thus negligible over the 2 years since randomisation, during which time they were not contacted. We have no reason to think that contamination was much more common among the 17 (4.1% of 417 2-year survivors allocated to OC) who could not be contacted 2 years after randomisation.

Open versus placebo-controlled trial designs

Recruitment

The overall participation rate (the proportion of invited patients who were randomised) was 14.2% overall (the protocol target was 9%) and 11.5% at age 80–84 years, with no evidence of a difference between open and blind practices. This confirms the feasibility of the main trial as planned.

Compliance and contamination

Treatment compliance was high among participants openly allocated to vitamin D and among those on blind treatment, and contamination was quantitatively negligible in all arms including the open untreated control arm. A trial of 5 years of treatment would thus be equally powerful whether open or blind, so the choice of design depends on other considerations.

For the main end points (overall mortality, cancer diagnosis and reasons for hospital admission), the advantages of open-label randomisation with follow-up by linkage to national registers are:

• A potentially substantial increase in power if those still on medication at 5 years were offered a continued supply of vitamin D, maintaining a large difference in blood 25(OH)D between treated and control participants beyond the initial 5 years of the trial. This might not be feasible in a double-blind trial when treatment is unblinded after 5 years, and in any case results beyond 5 years would not be placebo controlled.

• A lower cost, as control participants would not be recontacted after randomisation and would not be given any medication.

The main advantages of placebo control are:

• Reliable evidence on side effects and any potentially subjective outcomes.

• Lifestyle changes that are affected by treatment allocation might in principle influence health, so any effect observed in an open trial could be a biased estimate of the pharmacological effect of vitamin D.

Research recommendations

1. Having established the feasibility of a large UK vitamin D trial, our main recommendation is that it should begin as soon as possible, irrespective of the pending results of ongoing trials, which are likely to be inconclusive for the reasons shown in Table 1 [bolus dosing or daily regimens ≤ 2000 IU, and higher population 25(OH)D levels than in the UK]. A substantial body of evidence from different populations will be required to decide whether or not, at what dose and for whom, this mass medication should be recommended, and results from a UK trial would be the most relevant. Increasing publicity about the potentially large benefits may make it increasingly difficult to recruit for such a trial. If the evidence from ongoing trials is encouraging but not strong enough to justify recommending mass medication, or to determine both the optimal dose and the blood 25(OH)D level below which it is worthwhile, these issues may never be satisfactorily resolved.

2. The decision on whether randomisation should be open-label, blind or a mixture of the two in a large vitamin D trial with mortality as the primary outcome will depend strongly on a priori assumptions. The NIHR HTA might consider its view on this separately from assessing specific applications to conduct such a trial.

3. The NIHR HTA in consultation with the relevant agencies might review opportunities for reducing delays in Clinical Research Network funding approvals for multicentre population-based prophylactic trials, for simplifying trial regulations for non-prescription treatments, such as vitamin D, for which extensive evidence on safety is already available, and for arranging linkage to HES and other NHS Digital databases for all non-commercial medical research.

Contributions of authors

Christine Rake (Trial Manager) managed all aspects of the conduct of the study and contributed to the design of the study and writing the report.

Clare Gilham (Medical Statistician) performed the majority of the statistical analyses and contributed to the interpretation of data and writing the report.

Laurette Bukasa (Trial Administrator) contributed to the conduct of the study, trial management, data collection and data analysis.

Richard Ostler (Senior Programmer) designed the dedicated online trial management system and contributed to trial oversight via the Trial Management Group.

Michelle Newton (Trial Administrator) contributed to the conduct of the study and data collection.

James Peto Wild (Trial Administrator) contributed to the conduct of the study, data collection and database implementation.

Benoit Aigret (Barts CTU Head) contributed to protocol development and trial oversight via the Trial Management Group and as Operation Lead of Barts CTU.

Michael Hill (Oxford Clinical Trials Service Unit Laboratory Scientific Director) provided technical support in planning and throughout the trial on sample receipt, storage and analysis.

Oliver Gillie (Medical Journalist) was our lay member and contributed to the design of the study and trial oversight via the Trial Management Group.

Irwin Nazareth (Professor of Primary Care and Population Sciences) was delegate clinical chief investigator, hosted the pilot GP practice and assisted with GP practice selection and trial oversight via the Trial Management Group.

Peter Sasieni (Professor of Biostatistics and Cancer Epidemiology) contributed to protocol development and trial oversight via the Trial Management Group and as Director of Barts CTU, which developed the online trial management system.

Adrian Martineau (Clinical Professor of Respiratory Infection and Immunity) was the clinical chief investigator and contributed to the design of the study, trial oversight and writing the report.

Julian Peto (Professor of Epidemiology) was the chief investigator with overall responsibility for the design and supervision of the study, the statistical analysis and writing the report.

Data-sharing statement

All data requests should be submitted to the corresponding author for consideration. Access to available anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health and Social Care.