Thirty subjects, 15 per group, were enrolled. Baseline characteristics of the treatment and placebo groups were similar (Table 1). Three patients in the placebo group died due to CF-related causes before completing the 12-week visit, and one patient in the placebo group was unable to be studied at the 1-week visit. There were no significant differences between the groups in microbiology at the time of hospitalization or the types of antibiotic therapy during hospitalization. The length of IV therapy was similar between the groups. In the vitamin D group, the median length of IV therapy was 5 d and in the placebo group, 6 d (p = 0.2).

The mean serum 25(OH)D concentrations at baseline did not differ between the groups (p = 0.71). At baseline, 60% of subjects in the vitamin D group and 47% in the placebo group were vitamin D sufficient (serum 25(OH)D concentration ≥ 30 ng/mL). At 7 d, all subjects in the vitamin D group were sufficient; there was no change in the proportion of vitamin D sufficiency in the placebo group. The mean serum 25(OH)D concentrations for both groups at all three time points, with p-values for time point-specific mean comparisons, are shown in Table 2. At 1 week, mean serum 25(OH)D increased in the vitamin D group by 27.5 ± 3.8 ng/mL (p < 0.001) and in the placebo group it decreased by 0.2 ± 1.4 ng/mL (p = 0.64). Correspondingly, the week 1 difference in mean concentrations between groups was highly significant (p < 0.001). Figure 1 provides a visual representation of these relationships over the three time points. Adjusting for baseline age and FEV₁ (forced expired volume in 1 second), did not change this relationship (Table 2). The maximum 25(OH)D concentration in an individual subject at 1-week was 83 ng/mL and 47 ng/mL in the vitamin D and placebo groups, respectively.

Subjects exhibited good tolerance of the 250,000 IU dose of cholecalciferol; there were no reported symptoms of vitamin D toxicity as assessed by patient questionnaire at any study visit. There were also no clinical signs of hypercalcemia. As seen in Table 2, there were no significant changes in mean serum calcium or PTH concentrations in either group and no significant differences in these means across groups at any time point.

Subjects were followed up to 12 mo post-randomization or until death for clinical outcomes. Subjects in the vitamin D group exhibited an increase in the number of hospital-free and IV antibiotic therapy-free days as reported in Table 3. Prior to admission, the groups had an equal number of hospital-free days (Table 1). The total number of hospital-free days in 6 mo post-intervention in the vitamin D group was 168.6 ± 3.8 compared with 133.3 ± 14.8 in the placebo group (p = 0.036), a difference of 26.4%. There was also a positive trend in the number of IV antibiotic therapy-free days over 6 mo post-intervention in the vitamin D vs. the placebo group (153.6 ± 7.4 and 121.1 ± 15.4, respectively; p = 0.08).

There was also a trend toward improved lung function in subjects receiving vitamin D compared with placebo assessed by the proportion of subjects who returned to ≥ 95% of baseline lung function. At admission, 18 of the 30 subjects had a ≥ 10% decrease in FEV₁ from baseline. Of the subjects who had a decrease in FEV₁ at admission, 9 out of 10 (90%) subjects receiving vitamin D returned to ≥ 95% FEV₁, compared with 4 out of 8 (50%) subjects in the placebo group (p = 0.12).

Subjects who received vitamin D had a lower, unadjusted one-year mortality rate compared with subjects who received placebo, with 5 deaths in the placebo vs. 1 death in the treatment group at 12 mo (Table 3). Baseline FEV₁, BMI, age, and CF-related diabetes status were assessed as potential confounders, but none were individually significant. Survival analysis via the log-rank test suggested a statistically significantly higher risk for death in the placebo group over the 12 mo follow-up period (p = 0.029; Fig. 2). When adjusted for age and baseline FEV₁, this p-value increased to 0.09. There was no difference between the vitamin D and placebo groups in changes in BMI.

This study was approved by the Emory University Institutional Review Board (project approval number: IRB00007572) before subject recruitment began in August 2008; recruitment was completed in May 2009. The study was registered with clinicaltrials.gov (NCT00788138). Subject recruitment and allocation are summarized in Figure 3, the CONSORT flow diagram.^34,35 All subjects gave written informed consent before the initiation of study protocols.

All adult patients with CF followed at the Emory University CF Center and admitted to Emory University Hospital for treatment of a pulmonary exacerbation were eligible for this study. Pulmonary exacerbations were diagnosed by a pulmonologist (VS or AS) and were characterized by a constellation of symptoms and signs including increased cough, increased sputum production, increased crackles on auscultation, weight loss, and/or a decrease in lung function. Eligibility criteria included: age ≥ 18 y, serum 25(OH)D concentrations 5–75 ng/mL in the past year and current supplemental vitamin D intake < 2,000 IU/day. Subjects were excluded if they had a history of disorders that affect vitamin D, calcium or phosphorous metabolism; a history of organ transplant; currently pregnant or planning to become pregnant.

Within 48 h of admission, subjects were screened for inclusion and exclusion criteria. Subjects were randomized to receive either a single oral dose of 250,000 IU cholecalciferol or an identically matched placebo according to a computer-generated, balanced randomization scheme in blocks of six. All vitamin D capsules were administered by the Emory Investigational Drug Service at Children’s Healthcare of Atlanta. Study personnel, caregivers, subjects and those assessing outcomes were blinded to the allocation of the study drug.

The cholecalciferol dose consisted of five capsules of 50,000 IU of cholecalciferol (total dose 250,000 IU, in a powder vehicle composed of lactose, microcrystalline cellulose and magnesium stearate contained in a capsule of hard gelatin from Tishcon, Inc.) or a matched, identical placebo (five matched capsules containing only the vehicle used in the cholecalciferol capsule). Vitamin D content of the capsules was certified by an independent laboratory (Analytical Research Laboratories) at the end of the study period to contain 96.7% of the expected amount of cholecalciferol. The study drug was ingested without regard to the timing of nutrient or pancreatic enzyme supplementation.

Blood was collected immediately prior to the administration of the study capsules, at 1 week and at 12 weeks post-intervention. Subjects were also followed at 6 mo and 12 mo through their scheduled quarterly CF clinic visits. The primary outcomes included: the change in serum 25(OH)D, calcium, and PTH concentrations. Secondary endpoints included lung function measured by percent of predicted FEV₁; hospitalizations, IV antibiotic therapy and mortality. These outcomes were verified by a review of medical records and local data submitted to the national CF patient registry and accessed via Port CF. Symptoms of vitamin D toxicity (nausea, appetite, thirst, frequent urination, constipation, abdominal pain, muscle weakness, muscle and joint pain, confusion, lethargy and fatigue) were assessed by patient questionnaire at all study visits.

Hospital-free days were calculated as: number of days survival – days hospitalized = hospital-free days. Home IV antibiotic therapy-free days were also evaluated. IV antibiotic therapy-free days were calculated as: number of days survival – days treated with IV antibiotic therapy in the outpatient setting = IV antibiotic therapy-free days. Baseline FEV₁ was determined as the subject’s highest value since their most recent pulmonary exacerbation recorded in the CF clinic records or in the previous 6 mo. FEV₁ recovery to baseline was determined only in those subjects who had a > 10% decrease in FEV₁ from baseline to admission. Recovery FEV₁ was defined as the subject’s highest FEV₁ in the 3 mo following randomization.

Serum 25(OH)D and PTH were analyzed by ELISA (IDS, Ltd. and Immunotopics International, L.L.C., respectively). To ensure accuracy of the serum 25(OH)D measurements, our laboratory participates in the vitamin D external quality assessment scheme (DEQAS, site 606) and the NIST/NIH Vitamin D Metabolites Quality Assurance Program (VitDQAP). Serum calcium, albumin and creatinine were analyzed by standard methods in the Emory University Hospital clinical laboratories. Antibiotic therapy during hospitalization was determined from hospital patient records. Pathogens present at admission were determined by sputum analysis. Vitamin D insufficiency was defined as serum 25(OH)D < 30 ng/mL in accordance with the CF Foundation guidelines.³⁶

All statistical analyses were performed using SAS 9.2 (SAS Institute, Cary, NC). Previous analysis demonstrated that the mean 25(OH)D status of patients from the Emory CF center was 22.6 (± 10) ng/ml.⁴ A 10 ng/ml increase in 25(OH)D would produce mean 25(OH)D > 30 ng/ml; therefore, the enrollment of 30 subjects provided > 90% power to detect a 10 ng/mL difference in serum 25(OH)D between the two groups at a significance of level of α = 0.05. Fisher's exact chi-square tests and Wilcoxon-Mann-Whitney's tests were used to compare frequencies for categorical variables and means of continuous variables at baseline between the treatment groups. Variable assessments at each time point confirmed approximate normality of serum 25(OH)D and calcium concentrations; PTH values were log transformed due to their right-skewed distributions before inclusion in analysis. Paired t-tests were applied to assess within-group changes in mean serum concentrations from baseline.

Mixed effects linear regression models with a random intercept to account for within-subject correlations were used to evaluate the difference in mean serum concentrations of the vitamin D vs. the placebo group at each time point, based on repeated measurements of serum 25(OH)D, calcium, and PTH. Specifically, the following model was fit separately via the SAS MIXED procedure for four outcome variables.

where i indexes subject, j indexes time point, GROUP, WEEK1 and WEEK12 are binary indicator variables, b_0i is a random intercept and ε_ij is a random within-subject error term. Contrasts were constructed and tested in order to compare mean levels between groups at each time point (e.g., the null hypothesis for the week 1 comparison is equivalent to the condition β₁ + β₄ = 0). These tests were assessed for qualitative agreement with simple two-sample t tests at each time point. The same time point-specific contrasts were then assessed after refitting the mixed models while adjusting for potentially influential baseline patient characteristics. The following baseline variables were evaluated as potential confounders: age, race, BMI, FEV₁ and CF-related diabetes status.

The log-rank test was used to compare overall mortality across the two groups over the 12 mo of follow-up. Cox proportional hazards regression was applied univariately to assess the importance of baseline FEV₁, BMI, age and CF-related diabetes status. A multivariable Cox model was then fit to assess the treatment effect after controlling for other variables deemed clinically or univariately statistically significant.