Recruitment and follow-up
CIPRIS (Cholecalciferol Intervention to Prevent Respiratory Infections Study) was a double-blind randomised placebo-controlled trial of 20,000 IU/week cholecalciferol supplementation, undertaken in Hobart, Australia (latitude 42.9°S). This dose was chosen with a goal to give the equivalent of roughly 3,000 IU per day to realise a replete 25(OH)D status. The study was registered at http://www.clinicaltrials.gov (NCT01549938) and the Australia-New Zealand Clinical Trials Register (ACTRN12612000054819). The study was approved by the Southern Tasmania Human Research Ethics Committee, conforming to the principles embodied in the Declaration of Helsinki. All participants provided written informed consent. The study protocol was developed and finalised prior to the start of recruitment. No alterations to protocol were made during the study recruitment or follow-up periods.
Study methods, results and interpretations thereof are presented as per the CONSORT guidelines (Additional files 1 and 2).
Participants were recruited from the student and staff and their friends and families of the Menzies Research Institute Tasmania and the University of Tasmania School of Medicine, the Royal Hobart Hospital and the TasTAFE Campbell Street Campus, all essentially healthy controls representative of the general population. All clinics were conducted at the Menzies Research Institute Tasmania, however. Recruitment was in March-June 2012 conducted via flyers, email invitation and word-of-mouth communication. Exclusion criteria were 1) age <18 or >60 years; 2) using immunomodulatory medication; 3) diagnosed with an autoimmune, immune-deficiency or chronic respiratory condition; 4) diagnosis of hypercalcemia or malabsorption syndrome, or any parathyroid, liver or kidney disorder; 5) using vitamin D supplements (>1,000 IU/day) in the preceding 3 months; 6) being pregnant or planning to so within the study period; 7) inability to swallow capsules; or 8) inability to provide informed consent. Interested persons were invited to complete an online assessment questionnaire querying the exclusion criteria, with acceptable participants asked to provide contact information for the study investigators to contact them on. Of the 173 persons who accessed the assessment questionnaire, 52 were ineligible and 89 either did not provide contact information or declined to participate on discussing the study protocol, leaving 32 participants.
Follow-up was during the winter months, from May to October 2012, much of which is when ambient UV is too low to generate vitamin D [7].
At baseline assessment, participants were queried for demographic information and diet and behaviour affecting infection or vitamin D. At weekly updates, participants were queried about changes in health status, medication, supplements, health status, or time outside or physical activity.
Participants were also queried whether they had travel out of the state in the interval since the last weekly update. Excluding these people did not materially affect any analyses (data not shown).
Sample size for this pilot was set at 32 to allow a balance between treatment arms, with the number selected to test methods and principals in anticipation of a larger study. Participants were a convenience sample of healthy adults and did not represent any groups at risk of vitamin D deficiency or infection susceptibility.
Randomisation and treatment
Participants were randomised simply 1:1 to parallel treatment using a computerised randomisation program (http://www.randomization.com), using the first generator. A person outside the study was asked to run the randomisation and to affix the treatment labels with study IDs to the respective bottles of treatment and placebo.
Cholecalciferol and placebo capsules were obtained from Dartnells Pharmacy in Victoria, Australia. Both cholecalciferol and placebo were identical white capsules. Treatment allocations were dispensed at weekly clinics.
Blinding
As a double-blind RCT, all CIPRIS staff (including nurses and database entry personnel), investigators and participants were blinded to treatment allocation until the conclusion of follow-up.
Biological specimens
At each monthly review, a blood sample was taken from participants for measuring 25(OH)D. The first two monthly specimens were also tested for corrected calcium and phosphate. Blood samples were centrifuged and the serum aliquoted and stored at -80°C until analysis.
Biological assays
Serum corrected calcium and phosphate were assessed using standard methods by the Pathology Department of the Royal Hobart Hospital each week.
After the conclusion of the study, serum 25(OH)D was assessed by tandem mass spectroscopy by Canterbury Health Laboratories in New Zealand.
Infection assessment
Participants completed daily online questionnaires querying the occurrence and magnitude (0–5, where 0 is no presence of that symptom and 5 is most severe) of acute infection symptoms, including respiratory, gastrointestinal, urinary tract, eye, ear, skin and cold sore infections, as well as nonspecific symptoms. Online questionnaire responses were monitored each day by the chief investigator. When a participant reported symptoms of magnitude greater than 2/5 (thought to be of sufficient magnitude for a single day’s occurrence to warrant assessment in clinic), or two successive days of symptoms of any magnitude, the participant was invited to come into clinic for objective assessment by our study nurse.
After the conclusion of the study, infection reports from daily online surveys, and from clinic assessments, were reviewed by the chief investigator and the study nurse, classifying infections and infection type. In the event of disagreement between reviewers, infection status was agreed upon by discussion. Respiratory, gastrointestinal, urinary tract, eye, skin and cold sore infections were defined by symptoms reflective of these infection types. Ear infections were defined as earache symptoms in the absence of RTI symptoms. Systemic/nonspecific infections were defined as nonspecific symptoms like fever, malaise, fatigue, headache, and/or arthromyalgia, in the absence of other infection symptom types.
Infection duration was defined from the first day of reported symptoms and concluded at the resolution of symptoms. Where infections were confluent, a judgement was made on the basis of symptoms and symptom severities as to the end of one infection and start of another.
Infection severity for each infection was evaluated in three fashions. One was to take the maximum reported infection symptom severity for that infection. We also summated the infection severity scores for all infection symptoms for each day of the infection and summated this for a total infection severity score. Finally, we averaged the daily total infection symptom severities. For RTIs, only RTI-specific symptoms were used.
While the majority of infections were seen in clinic, a subset (11.8%) was only reported on online daily questionnaire but not seen in clinic. Sub-analyses excluding these are reported.
Adverse events
Serum collected at the first 2 months was evaluated for the levels of corrected calcium and phosphate. These reports were reviewed by an external monitor, who could terminate a participant’s participation if levels exceeded safe levels.
In addition, on daily questionnaire and at weekly and infection clinics, participants were queried about the occurrence and severity (0–5) of symptoms potentially indicative of hypercalcemia or hyperphosphatemia, including excessive thirst, abnormally high urine output, change in urine colour, bone pain, groin-area pain, confusion, irritability or other neurological symptoms. Any sustained occurrence of these could have resulted in that subject’s participation being stopped.
Statistical analyses
Analysis was by intention-to-treat.
Primary outcomes were time to infection. Secondary outcomes were infection severity and duration. Tertiary outcomes were change in serum 25(OH)D and the occurrence of adverse events.
Predictors of time-to-acute infection were assessed by survival analysis, using Cox proportional hazard models for repeated events [23]. All covariates satisfied the proportional hazard assumption.
Significance of differences in study follow-up duration and inter-review interval duration were evaluated by Kruskal-Wallis rank test.
Predictors of having an infection during the study were evaluated by logistic regression.
Predictors of infection duration, infection severity and longitudinally measured 25(OH)D were assessed by multilevel mixed effects linear regression. Predictors of infection count during the study were assessed by linear regression. Predictors of baseline-measured 25(OH)D were assessed by linear regression. As the distribution of all these variables was skewed, transformation was applied as required to satisfy homoscedasticity; however, all coefficients are reported on the scale of the original value.
Where interaction was assessed, a product term containing the primary predictor and the interaction covariate was generated and included in the model.
For all instances where data was missing, analyses were restricted to persons with complete data.
All analyses were performed using STATA/SE for Windows.