Population-based observational studies have found that low serum 25-hyroxyvitamin D (25(OH)D) concentration is associated with poor lung function [1
]. However, the observational design prevents one from knowing whether these relationships are causal, or whether they could be reversed by increasing 25(OH)D. To investigate the causality and reversibility of these associations, randomized controlled trials (RCTs) of vitamin D supplementation are required.
A limited number of RCTs have investigated the effect of vitamin D supplementation on lung function in adults [3
]. However, vitamin D efficacy remains unclear, due to the conflicting findings of these trials: some reported beneficial changes [7
], others found no effects [3
], and another reported mixed results [9
]. Most of these studies had relatively small sample sizes (n
≤ 130) [3
] and short follow-up periods (<1 year), which limited their ability to assess long-term efficacy [3
]. Nearly all of these studies were restricted to patients with respiratory conditions such as asthma [6
] or chronic obstructive pulmonary disease (COPD) [4
]. However, vitamin D trials should include other groups of people, too, in order to study the role of vitamin D supplementation on lung health in general [13
]. Trials should also investigate vitamin D-deficient people and smokers, as vitamin D supplementation could potentially be more effective in these people. This is because non-linear relationships between 25(OH)D and health outcomes suggest that adverse effects associated with low vitamin D status are greatest in vitamin D-deficient people [14
], while observational studies suggest that the relationship between 25(OH)D and lung function could be stronger in smokers than in non-smokers [1
Given the above knowledge gaps, we used an RCT design to investigate the effect of long-term (≥1 year on average), high-dose vitamin D supplementation on lung function in a population-based sample of >400 adults. We performed pre-specified subgroup analyses among participants who had vitamin D deficiency or asthma/COPD, or were smokers.
The study flowchart is shown in Figure 1
. From the 5110 participants randomized in the main ViDA study, 517 (10%) were randomly selected and invited to partake in the current sub-study. Of these, 74 declined and 1 withdrew consent (data analysis prohibited), and were thus not included in subsequent analyses. Out of the remaining 442, a complete set of both baseline and 1-year follow-up measurements was available in 366 participants (83%). The 76 participants with missing data comprised 49 who did not attend the follow-up interview (could not attend, uncontactable or moved overseas) and 27 who had unobtainable spirometry data (could not obtain a reading). Per intention-to-treat, all 442 people were included in the total-sample analysis. The proportion of the total sample with missing follow-up data did not differ across the two treatment groups (p
= 0.96, χ2
test). Further, baseline lung function of those with missing follow-up data did not differ from those without missing follow-up data (p
-values ranging from 0.47 to 0.57; analysis of variance.)
Of the total sample, 226 received vitamin D and 216 received the placebo. Table 1
shows the baseline characteristics of these participants by treatment group. The follow-up period (randomization—follow-up) averaged 1.1 years (mean and median) and ranged from 0.9 to 1.5 years. The mean age was 65 years (range: 50–84 years), 58% were male and just over three-quarters were of European/Other ethnicity (with 96% having European ancestry). Fourteen percent had asthma (77% well-controlled—ATS score ≥ 20) and 17% had COPD (mostly mild or moderate—GOLD stages 1 and 2). Nearly 30% had a deseasonalised 25(OH)D of <50 nmol/L (vitamin D deficiency). Almost one-half (49%) had smoked (82% of whom were ex-smokers). Supplementary analyses (not tabulated) showed that ex-smokers had quit a median of 30 years earlier (interquartile range: 13 to 40 years), and the median number of cigarettes smoked per day among current smokers was approximately just under 10 (54% smoked ≤10 per day). Further supplementary analyses showed that the proportion of participants with spirometry grades A or B were similar in the vitamin D (30%) and placebo (28%) groups.
The deseasonalised 25(OH)D concentrations at baseline and follow-up visits by treatment group are illustrated in Figure 2
. In the total sample, the change (95% confidence interval) from baseline in the vitamin D group compared to placebo at 6 and 12 months follow-up, respectively, was +51 (45, 58) and +57 (51, 64) nmol/L (p
< 0.001). No cases of hypercalcemia were detected.
FEV1 at baseline and follow-up by intervention group is shown in Table 2
. All effects were in the positive direction for the vitamin D supplemented group compared to placebo. The effects were non-significant in the total (n
= 442), vitamin D-deficient (n
= 130) and asthma/COPD (n
= 113) samples. This was true, too, among participants with both asthma/COPD and vitamin D-deficiency, despite the effect being large (109 mL) and borderline significant (p
= 0.08). Conversely, among all ever-smokers, FEV1 significantly increased in the vitamin D group with respect to placebo (p
= 0.03), with a mean (95% confidence interval) change of 57 (4, 109) mL. This effect more than doubled when restricted to ever-smokers who also had vitamin D deficiency (β = 122 mL, p
= 0.04) or asthma/COPD (β = 160 mL, p
= 0.004). Similar patterns were observed when we modelled FEV1 as a z
-score, with net (placebo-controlled) vitamin D effects of 0.13 (0.01, 0.24; p
= 0.03) and 0.35 (0.11, 0.59; p
= 0.005) z
-scores among all ever-smokers and ever-smokers with asthma/COPD, respectively (Table S1
When these analyses were repeated with FVC as the response variable (Table 3
), all effects were smaller (compared to FEV1 effects). The effect of vitamin D compared to placebo was minimal and non-significant (p
> 0.05) in the total sample. Across subgroup samples, the effect was larger (in nearly all cases) and consistently in the positive direction, although still not statistically significant. Similar patterns were observed with FVC as a z
-score (Table S2
When these analyses were repeated with FEV1/FVC as the dependent variable, all effects (vitamin D compared to placebo) were in the positive direction (Table 4
). The effect among all ever-smokers was 1.1% (p
= 0.05) and almost tripled when confined to ever-smokers who also had asthma/COPD (β = 3.0%, p
= 0.01). Similarly, all FEV1/FVC z
-score effects were in the positive direction, with the largest being among ever-smokers with asthma/COPD (β = 0.37, p
= 0.01; Table S3
Further analysis showed that vitamin D (with respect to placebo) significantly improved FEV1 more in ever-smokers than in never-smokers (p = 0.02 for three-way interaction between smoking, treatment group, and time). We confirmed this interaction when we restricted this analysis to people with vitamin D deficiency (p = 0.048) and asthma/COPD (p = 0.0005).
Correlations between changes in observed 25(OH)D concentration and changes in lung function measures are shown in supplementary Table S4
. All FEV1 and FEV1/FVC correlations were in the positive direction. Among ever-smokers and their subgroups, 25(OH)D change was positively correlated (r
= 0.17 to 0.34) with change in FEV1, which mirrors the FEV1 increases in these samples shown in Table 2
This randomized, double-blinded, placebo-controlled trial showed that monthly, high-dose vitamin D supplementation for just over 1 year did not affect lung function in the total sample, nor in subgroups defined by either vitamin D deficiency (<50 nmol/L) or having asthma/COPD. However, vitamin D supplementation did result in larger, statistically significant increases in FEV1 and FEV1 z-score among ever-smokers, especially those with vitamin D deficiency (FEV1 only) or asthma/COPD.
To our knowledge, this is the first study to show that vitamin D supplementation (compared to placebo) increases FEV1 and FEV1 z
-score in ever-smokers. The restriction of this effect to ever-smokers only is consistent with observational research, which has shown stronger 25(OH)D-lung function associations among smokers [1
] and a stronger smoking-FEV1 relationship in vitamin D-deficient people [2
]. Taken together, these findings suggest that vitamin D supplementation may mitigate smoking-associated lung function damage, although smoking avoidance and cessation remain paramount for preserving lung health.
Smoking decreases the production of 1,25-dihydroxyvitamin D in lung epithelial cells [29
] and may affect expression levels of the vitamin D receptor [30
]. Smoking-related lung destruction is partly mediated through inflammation, oxidative stress, and increased proteases [31
], and these pathophysiological changes may persist even after smoking cessation [33
]. However, vitamin D could mitigate these processes [34
]. Further, there is increased activity of these processes in asthma [37
] and COPD [31
]. Collectively, these observations could explain our finding that vitamin D effects were limited to ever-smokers and were the largest in ever-smokers with asthma/COPD.
The effects of vitamin D among all asthma/COPD participants were non-significant (although in the positive direction), which concurs with some prior RCTs of patients with asthma [9
] or COPD [4
]. We build on these past trials both by showing that the effects on asthma/COPD participants were stronger in ever-smokers (Table 2
and Table 3
) and because only one of these studies used the same dosing regimen we administered (monthly ≥100,000 IU dosing for ≥1 year) [5
]. That study differed from ours in that it comprised largely men (80%) with both mostly severe or very severe COPD and a history of recent exacerbations [5
]; in contrast, our COPD cases were primarily mild or moderate, and were combined with predominantly well-controlled asthma cases.
The intervention effect for FEV1 as a percentage of the average lung function parameter value in the vitamin D group (both in Table 2
) was modest (3%) among all ever-smokers (57 mL as a percentage of 2241 mL), larger (5%) among vitamin D-deficient ever-smokers (122 mL as a percentage of 2348 mL), and sizeable (10%) among asthma/COPD ever-smokers (160 mL as a percentage of 1538 mL). A change in FEV1 of at least 100 mL is considered to be clinically relevant [38
], suggesting that the net vitamin D effects on FEV1 among vitamin D-deficient ever-smokers (122 mL) and asthma/COPD ever-smokers (160 mL) in our study (Table 2
) represent, by definition, clinically meaningful improvements. Our FEV1 z
-score results, which account for spirometric influences of demographics and height using a different statistical approach, provide further support of a benefit, with net vitamin D effects of 0.13 among all ever-smokers and 0.35 among ever-smokers with asthma/COPD (Table S1
). Because there is a paucity of information on the size of the association between FEV1 z
-scores and health outcomes [40
], more such research is required to quantify the clinical impact of these z
-score results. As for FEV1/FVC, the effects on this parameter were meaningful: given that FEV1/FVC declines by ~0.2% per year [26
], the net vitamin D effects (increases) in FEV1/FVC of 1.1% among all ever-smokers and 3.0% among ever-smokers with asthma/COPD (Table 4
) would correspond to changes that typically occur over 5.5 years and 15 years, respectively. Finally, as vitamin D effects on FEV1 were markedly more positive when analyses in asthma/COPD or vitamin D-deficient participants were restricted to ever-smokers than when they were not, this suggests that future RCTs of asthma, COPD or vitamin D-deficient people should carry out subgroup analyses among smokers to capture a potential difference in treatment effects.
Our study sample was population-based, as the vast majority of New Zealand residents (94%) are registered with family practices [42
]. This augments the external validity of our findings. Regarding limitations, the missingness of the intention-to-treat sample (Figure 1
) renders our study findings prone to selection bias. However, as mentioned, this missingness did not differ across the treatment groups, and did not predict baseline lung function. The data analyst was not blinded to the treatment group, although we did include pre-specified analyses in our statistical analysis plan (mentioned above). The equations used to calculate our z
-score results [26
] may have limited applicability to our Maori and Pacific participants, as these ethnic groups were not included in the data that these z
-scores are based on. A longer follow-up period may have allowed us to better evaluate the long-term efficacy of the intervention. Although our total sample size was large, relative to previous RCTs of vitamin D and lung function [3
], our statistical power was limited (particularly for the subgroup analyses), which may explain why at least some treatment effects were not statistically significant. Finally, the multiplicity of statistical tests we performed raises the possibility that at least some of our significant findings may have been due to chance. However, we observed positive dose-response relationships between change in 25(OH)D and change in lung function (Table S4
), which supports a true effect (biological gradient). Further support includes the fact that, as reported, the FEV1 and FEV1 z
-score treatment effects were consistent with observational research, biologically plausible, and were all unidirectional (across samples; Table 2
). Also, if study conclusions are based on the primary outcome (FEV1) results only, far fewer comparisons are involved.
Our analyses were based on subsamples of an RCT (Figure 1
). Although a limitation, we do not expect there to be marked, systematic differences in baseline participant characteristics, for the following reasons: Firstly, the selection of our total analysis sample from the main ViDA study was random. Second, as everybody was randomized in the same way, the selection of subgroups from the total sample should not differ across treatment groups. Third, since the analyses controlled for age, sex, ethnicity and height, effects of any imbalances in these demographic variables would have been minimized. Fourth, stratifying the study randomization by our subgroup variables (vitamin D deficiency, asthma/COPD and ever-smokers) could have reduced any baseline imbalances within these subgroups [43
]. Although we did not do this, this effect would have been partially captured, as we stratified randomization by age and ethnicity, which are associated with these subgroup variables [44
]. Further, some have proposed that stratification of randomization is not required for pre-specified subgroup analyses (such as ours) [46
In summary, monthly high-dose vitamin D supplementation over an average of 1.1 years, which increased serum 25(OH)D concentration by >50 nmol/L with respect to placebo, did not improve lung function in the overall study population. In subgroup analyses, we found that vitamin D supplementation improved lung function (FEV1 and FEV z-score) in ever-smokers, particularly those with vitamin D deficiency (FEV1 only) or asthma/COPD. We encourage similar RCTs in smokers to assess the efficacy of different dosing regimens (e.g., daily or weekly supplementation). Additional RCTs are needed to investigate whether the observed beneficial effects translate into improvements in lung function-related health, such as improved asthma/COPD control.