Chronic inflammation underlies many pathological processes including several types of cancer( Reference Elinav, Nowarski and Thaiss 1 ), CVD( Reference Ross 2 ) and autoimmune disease( Reference Munoz, Janko and Schulze 3 ). Animal and in vitro studies have identified a number of mechanisms by which the most biologically active form of vitamin D, 1,25-dihydroxyvitamin D (1,25(OH)2D), may reduce inflammation( Reference van Etten and Mathieu 4 ). For example, treatment with 1,25(OH)2D suppresses the NF-κB pathway( Reference D’Ambrosio, Cippitelli and Cocciolo 5 ), thereby reducing downstream expression of pro-inflammatory cytokines. By acting as a ligand-activated transcription factor, the vitamin D receptor (VDR) can also directly regulate cytokine gene transcription( Reference Alroy, Towers and Freedman 6 ). In addition, the VDR is expressed in adipocytes that are responsible for the production of the adipokines adiponectin and leptin. These have anti- and pro-inflammatory properties, respectively, but investigations of the influence of 1,25(OH)2D on their secretion are scarce( Reference Menendez, Lage and Peino 7 , Reference Sun and Zemel 8 ).
Serum concentrations of the intermediate product 25-hydroxyvitamin D (25(OH)D) are used to measure a person’s vitamin D status. Many observational studies suggest that 25(OH)D concentration is inversely associated with pro-inflammatory markers and positively associated with anti-inflammatory markers( Reference Bellia, Garcovich and D’Adamo 9 – Reference Husemoen, Skaaby and Martinussen 14 ). For example, studies have reported an inverse relationship between serum 25(OH)D levels and circulating concentrations of C-reactive protein (CRP)( Reference Bellia, Garcovich and D’Adamo 9 , Reference van de Luijtgaarden, Voute and Hoeks 10 ), IL-6( Reference Bellia, Garcovich and D’Adamo 9 ) and leptin( Reference Vilarrasa, Vendrell and Maravall 11 ), and a positive association with adiponectin levels( Reference Kim, Na and Sohn 12 – Reference Husemoen, Skaaby and Martinussen 14 ). Other studies suggest a less straightforward relationship. Both pro- and anti-inflammatory cytokines levels were found to be down-regulated in summer in association with increased serum 25(OH)D levels( Reference Khoo, Chai and Koenen 15 ). In contrast, an observational analysis of data from a trial suggested that supplementation with vitamin D increased the levels of both pro- and anti-inflammatory cytokines in people with initially low serum 25(OH)D levels( Reference Barker, Rogers and Levy 16 ).
In some trials, supplementation with vitamin D has been associated with significant reductions in circulating concentrations of CRP( Reference Asemi, Samimi and Tabassi 17 , Reference Witham, Dove and Khan 18 ) and IL-6( Reference Abou-Raya, Abou-Raya and Helmii 19 , Reference Grossmann, Zughaier and Liu 20 ), and significant increases in IL-10( Reference Schleithoff, Zittermann and Tenderich 21 ) and adiponectin( Reference Neyestani, Nikooyeh and Alavi-Majd 22 ). However, most trials with significant results were carried out in patients with a specific disease condition rather than in ‘healthy’ participants. Results from trials that used people without apparent disease were largely null, although these used relatively small sample sizes (n 23–307)( Reference Barnes, Horigan and Cashman 23 – Reference Carrillo, Flynn and Pinkston 29 ). A pooled analysis of data from four trials, none of which was population-based, found no significant change in CRP following high-dose supplementation with vitamin D( Reference Jorde, Strand Hutchinson and Kjaergaard 30 ).
Many people without a specific disease may still have chronic, low-grade inflammation, especially if exposed to toxic chemicals (e.g. tobacco smoke), or if overweight or obese( Reference Medzhitov 31 , Reference Trinchieri 32 ). Given the possible consequences for the prevention of cancer and CVD, it is worth investigating whether vitamin D supplementation can dampen subclinical inflammation in apparently healthy people. We used secondary data from our randomised controlled trial (RCT) of vitamin D supplementation in a population-based sample of older Australians, which demonstrated large effects of supplementation on serum 25(OH)D( Reference Tran, Armstrong and Carlin 33 ), to assess the effect of supplementary vitamin D on circulating inflammatory markers (CRP, leptin, adiponectin, leptin:adiponectin ratio (LAR), IL-6 and IL-10). To the best of our knowledge, this is one of the largest and, hence, most highly powered studies of the possible effect of vitamin D on inflammation, with relevance extending beyond highly specific patient groups.
Study population and design
The methods and protocols of the study have been described( Reference Tran, Armstrong and Carlin 33 ). Briefly, the Pilot D-Health trial 2010/0423 was a population-based, randomised, placebo-controlled, double-blind chemoprevention trial of vitamin D3 in older adults. This study was conducted according to the guidelines laid down in the 1964 Declaration of Helsinki, and all procedures involving human subjects were approved by The Human Research Ethics Committee at the QIMR Berghofer Medical Research Institute. Written informed consent was obtained from all participants. Participants consented to blood samples being stored indefinitely and for them to be used in studies related to vitamin D. The QIMR Berghofer Medical Research Institute Human Research Ethics Committee approved the use of samples for this study.
We recruited 644 people aged 60–84 years who were residents of one of the four eastern states of Australia, using the Australian Electoral Roll as the sampling frame. Participants were sampled in strata of 5-year age bands, sex, state and location of residence (capital city and elsewhere). We excluded anybody who was taking ≥10 μg of vitamin D/d or who had a history of kidney stones, hyperparathyroidism, osteomalacia, osteoporosis or sarcoidosis. Recruitment took place between October 2010 and March 2011, and 10 % of people invited were randomised. Participants were randomly allocated to one of three trial arms: placebo (n 214), 750 ��g (n 215), or 1500 μg (n 215) vitamin D3 (cholecalciferol), each taken once per month for 12 months. Participants, study investigators and staff were unaware of the treatment allocation until analysis of the primary outcome (change in serum 25(OH)D level) was complete. Approximately 95 % of participants completed the study and 96 % reported taking at least ten of their twelve study tablets.
Blood samples and laboratory analyses
Non-fasting blood samples were collected in serum separator tubes at study entry and after the intervention period and transported cold by overnight courier to the Queensland University of Technology (Brisbane, Australia). Participants were requested to provide a post-intervention blood sample within 2 weeks of the last tablet taken. Samples were centrifuged at 4°C for 15 min at 2000 g , and the serum was stored at –80°C.
Pre- and post-intervention serum 25(OH)D concentrations were measured in a single batch using a commercial chemiluminescent immunoassay (LIAISON® 25(OH)D Vitamin D TOTAL Assay; DiaSorin, Inc.). Intra- and inter-assay variabilities were 3–6 and 6–9 %, respectively. The laboratory undertaking the testing was certified by the international Vitamin D External Quality Assessment Scheme.
We measured non-fasting serum levels of five inflammatory markers in post-intervention blood samples: IL-6, IL-10, leptin, adiponectin and CRP. Levels of IL-6, IL-10, leptin and adiponectin were measured using commercial sandwich ELISA kits (Merck Millipore) using a Multiskan Primary EIA V. 2.3 device with Labsystems Genesis V3.03 software, and levels of CRP were measured using the high-sensitivity CardioPhase hsCRP assay on the BNTM II system (Siemens Healthcare Diagnostics). The inter- and intra-assay CV were <10 and <15 % for all biomarkers. The minimum detectable concentrations for IL-6, IL-10, leptin, adiponectin and CRP were 1·6 pg/ml, 2 pg/ml, 0·78 ng/ml, 0·1 µg/ml and 0·175 mg/l, respectively. All assays for inflammatory markers were performed by Sullivan Nicolaides Pathology (Indooroopilly, Australia) according to the manufacturers’ instructions.
Within 2 weeks of having blood drawn at study entry and exit, participants were asked to complete questionnaires in which they reported information about their height, weight, skin phenotype, smoking and alcohol consumption, history of cancer, diabetes or CVD, physical activity, time outdoors, overall health and dietary intake of vitamin D.
We excluded participants for whom no blood sample was available for the analysis of inflammatory markers. Cytokine levels below the minimum detectable concentration were assigned a value equal to the minimum detectable concentration for that biomarker. Adipokine levels outside the appropriate range of the assay were assigned the lower or upper bound as appropriate. The LAR( Reference Satoh, Naruse and Usui 34 ) was calculated by dividing the leptin level (ng/ml) by the adiponectin level (µg/ml).
Comparisons of post-intervention levels of serum 25(OH)D were made between intervention groups using two-sample t tests. As biomarker distributions were right-skewed, we present their descriptive statistics as median (interquartile range (IQR)).
We used quantile regression (QR) to evaluate the association between intervention group and each biomarker level. QR is used to model a specified percentile of a response variable, conditional on one or more predictors. It is therefore similar to linear regression, which models the conditional mean of the response variable. The distribution of some biomarkers remained right-skewed following a log transformation. QR makes no assumptions regarding the distribution of the regression residuals and therefore offered an advantage over linear regression. Further, QR was preferable to Wilcoxon’s rank sum test because it permitted the inclusion of interaction terms.
We compared median levels of CRP, leptin, adiponectin and LAR, and 75th percentile levels of IL-6 and IL-10. These were meaningful comparisons given the shapes of biomarker distributions. In particular, as IL-6 and IL-10 levels were very right-skewed, we were interested in whether vitamin D supplementation might have a dampening effect on more ‘extreme’ levels of these biomarkers.
Intervention group was treated as a categorical predictor in QR models, with the placebo group used as the reference level. For CRP, leptin, adiponectin and LAR, QR coefficients represent the estimated difference in medians between an actively supplemented group (either the 750 or 1500 μg group) and the placebo group. For IL-6 and IL-10, the coefficients represent the estimated difference in 75th percentiles. 95 % confidence intervals for each coefficient are interpreted in the usual way; a CI that does not contain 0 indicates that the percentile differs significantly between the placebo and supplemented groups. Wald’s statistic was used to assess the overall significance of the intervention group. To test for linear trends, we fitted models with intervention group treated as a continuous variable, with the values 0, 1 and 2 corresponding to the placebo, 750 and 1500 μg groups, respectively.
To test whether the effect of supplementation was modified by sex, age (<70, ≥70 years), baseline BMI (<25, ≥25 kg/m2), baseline serum 25(OH)D level (<30, ≥30 nmol/l) or time between last tablet and collection of post-intervention blood sample (≤7, >7 d), we included interaction terms for intervention group by each predictor. The cutoff point of 30 nmol/l was selected based on the Institute of Medicine’s 2011 report( 35 ). Analyses were repeated within strata of sex, age, baseline BMI, baseline serum 25(OH)D level and timing of blood collection.
Analyses were performed in SAS version 9.2 (SAS Institute, Inc.). All P values are two-sided, and we used a statistical significance level of P<0·05.
The trial was designed to have adequate power to assess changes in serum 25(OH)D levels( Reference Tran, Armstrong and Carlin 33 ). The final number enrolled was greater than planned because of an increase in response rate in later waves of recruitment.
Given strong recommendations against their use( Reference Hoenig and Heisey 36 ), post hoc power calculations have not been performed for this secondary analysis of inflammatory markers. This is, however, one of the largest randomised placebo-controlled trials conducted to date that investigates the effect of vitamin D supplementation on inflammatory markers.
Characteristics of participants
We randomised 644 participants, 615 of whom completed the study. Post-intervention blood samples for the analysis of inflammatory markers were available for 613 participants. The median time between the last tablet taken and collection of the post-intervention blood sample was 7 (IQR 6–11) d and, within each treatment group, 93 % of people provided the sample within 2 weeks of taking the final tablet. The average age of participants included in this analysis was 72 (sd 7) years, and 46 % were women. Baseline characteristics are presented in Table 1. Baseline serum 25(OH)D levels were similar for the three groups (Table 1), whereas average post-intervention serum 25(OH)D levels were highest in the 1500 μg group (78 (SD 20) nmol/l), intermediate in the 750 μg group (64 (SD 17) nmol/l) and lowest (unchanged) in the placebo group (42 (SD 15) nmol/l)( Reference Tran, Armstrong and Carlin 33 ); P values for all pairwise comparisons were <0·0001. The percentage of participants with serum 25(OH)D levels >75 nmol/l also significantly differed between groups (1500 μg group, 54 %; 750 μg group, 24 %; placebo group, 1 %).
25(OH)D, 25-hydroxyvitamin D; MET, metabolic equivalent task.
* Missing data for one participant in the placebo group.
† Missing data for one participant in the 1500 μg group.
Analysis of inflammatory markers
Each of CRP, leptin, adiponectin and LAR had one missing observation, and data were missing for IL-6 and IL-10 in four and twelve cases, respectively. The percentages of observations with a biomarker level below the limit of detection were 1 % (CRP), 32 % (IL-6), 58 % (IL-10), 3 % (leptin) and 2 % (adiponectin); the percentages did not differ between treatment groups. The 75th percentile IL-6 level was 2·8 pg/ml higher (95 % CI 0·4, 5·8 pg/ml) in the 1500 μg group than in the placebo group (75th percentiles: 11·0 v. 8·2 pg/ml) (Table 2). There was a somewhat smaller and non-significant difference (in 75th percentile IL-6 levels) between the 750 μg and placebo groups (Table 2). There were no statistically significant differences in median levels of CRP, leptin, adiponectin or LAR, or in 75th percentile levels of IL-10 levels, between the placebo group and either the 750 or the 1500 μg group (Table 2).
IQR, interquartile range; CRP, C-reactive protein; LAR, leptin:adiponectin ratio.
* For CRP, leptin, adiponectin and LAR, quantile regression coefficients (denoted b) represent the estimated difference in medians between the actively supplemented group (750 or 1500 μg) and the placebo group. For IL-6 and IL-10, b represents the estimated difference in 75th percentiles.
† P value from the Wald test for intervention group treated as a categorical predictor (with 2 df), with the placebo group used as the reference level.
‡ P trend from the Wald test for intervention group treated as a continuous variable (1 df) with values 0, 1 and 2 corresponding to the placebo, 750 and 1500 μg groups, respectively.
§ Missing data for one participant in the placebo group.
|| Missing data for three participants in the placebo group and one participant in the 750 μg group.
¶ Missing data for four, three and five participants in the placebo, 750 and 1500 μg groups, respectively.
There was a significant interaction between the intervention group and age with respect to the association between supplementation and median levels of leptin and LAR (P interaction=0·03 and 0·02, respectively) (online Supplementary Table S1). For people aged <70 years, the median leptin and LAR levels were higher in the 750 μg group and lower in the 1500 μg group compared with the placebo group. The opposite pattern was observed for those aged at least 70 years. There were no other significant interactions.
Despite large differences in serum 25(OH)D levels between the three groups after 12 months of vitamin D supplementation, we found little evidence of an effect of supplementation on levels of cytokines and adipokines, with the possible exception of IL-6. The significantly higher 75th percentile IL-6 level in the 1500 μg group compared with the placebo group was in contrast to our expectations.
There have been at least ten RCT examining the effect of vitamin D supplementation on IL-6 or IL-10 concentrations. Of the RCT that included participants either without a specific health condition( Reference Barnes, Horigan and Cashman 23 – Reference Yusupov, Li-Ng and Pollack 26 ) or that only included overweight or obese participants( Reference Jorde, Sneve and Torjesen 27 – Reference Carrillo, Flynn and Pinkston 29 ), the results for IL-6 and IL-10 were all essentially null. In one of these, there was a suggestion of a greater change in median IL-10 levels in the supplemented group (−70 %) than in the placebo group (−50 %), but the difference was not significant after adjustment for multiple testing( Reference Yusupov, Li-Ng and Pollack 26 ). Possible explanations for these null results include that the sample sizes were too small, the duration or dose of supplementation was insufficient to elicit a response or that the people studied had adequate vitamin D status at baseline. However, for almost all of these trials, supplementation significantly increased serum 25(OH)D levels( Reference Pittas, Harris and Stark 24 – Reference Carrillo, Flynn and Pinkston 29 ), and in two trials( Reference Wood, Secombes and Thies 25 , Reference Wamberg, Kampmann and Stodkilde-Jorgensen 28 ) participants had low baseline 25(OH)D levels.
It is not clear why, in the present study, the upper quartile IL-6 level was greater for those receiving the highest dose of vitamin D supplementation than in the placebo group, and it is possible that this is a chance finding. It is also possible that vitamin D supplementation might be detrimental at higher doses. However, in addition to having pro-inflammatory effects, there is also evidence that IL-6 exhibits anti-inflammatory properties( Reference Tilg, Dinarello and Mier 37 , Reference Steensberg, Fischer and Keller 38 ). Further, in some in vitro studies using human cells, treatment with 1,25(OH)2D3 increased IL-6 production by differentiated adipocytes( Reference Sun and Zemel 8 ), and induced a subpopulation of IL-6-producing T cells when administered to T lymphocytes( Reference Willheim, Thien and Schrattbauer 39 ).
Results from trials measuring CRP have been mixed, with no consistent evidence of any clinically relevant effect. In many RCT in which participants did not have a specific disease, CRP levels did not significantly change( Reference Barnes, Horigan and Cashman 23 – Reference Wood, Secombes and Thies 25 , Reference Jorde, Sneve and Torjesen 27 – Reference Carrillo, Flynn and Pinkston 29 , Reference Gepner, Ramamurthy and Krueger 40 ). Vitamin D supplementation was associated with a significant reduction in circulating concentrations of CRP in healthy pregnant women( Reference Asemi, Samimi and Tabassi 17 ) and in patients with a history of myocardial infarction( Reference Witham, Dove and Khan 18 ), but CRP levels increased in another trial among people who were overweight or obese( Reference Beilfuss, Berg and Sneve 41 ).
With respect to the adipokines, some observational studies have reported positive associations between serum 25(OH)D level and adiponectin levels( Reference Kim, Na and Sohn 12 , Reference Vaidya, Williams and Forman 13 ), with one study using a Mendelian randomisation approach to demonstrate that the association was possibly causal( Reference Husemoen, Skaaby and Martinussen 14 ). In contrast, an inverse correlation has been observed between serum 25(OH)D levels and circulating concentrations of leptin( Reference Vilarrasa, Vendrell and Maravall 11 ). Trial data concerning adipokines are scarce and those that have been conducted had small sample sizes( Reference Neyestani, Nikooyeh and Alavi-Majd 22 , Reference Wamberg, Kampmann and Stodkilde-Jorgensen 28 , Reference Breslavsky, Frand and Matas 42 ). To the best of our knowledge, only one trial has reported a significant increase in adiponectin levels following vitamin D supplementation( Reference Neyestani, Nikooyeh and Alavi-Majd 22 ). The study was restricted to participants with type 2 diabetes, and the end-of-study adiponectin levels reported were much lower than in our participants. We found no difference in levels of either adipokine between placebo and active supplementation groups. The significant interaction between the intervention and age with respect to leptin and LAR levels is mostly likely a chance finding, as there was no dose–response relationship in stratified analyses.
Our results were largely null. For a given biomarker, this means that either that the effect due to supplementation was too small to be of biological importance or that the sample size was insufficient to detect a small but potentially important effect. The CI provide a range of plausible values for the actual effect size.
Cytokine regulation may be dependent on the serum 25(OH)D levels achieved by supplementation. The authors of a study that showed that IL-6 production by human monocytes was inhibited by treatment with ≥75 nmol/l (but not 37·5 nmol/l) 25(OH)D3, concluded that maintaining a serum 25(OH)D level ≥75 nmol/l was necessary for optimal anti-inflammatory effects( Reference Zhang, Leung and Richers 43 ). A small observational study of young adults (25–42 years) also noted that levels of selected pro-inflammatory cytokines were higher in those with serum 25(OH)D levels ≤80 nmol/l when compared with people with higher levels( Reference Barker, Martins and Hill 44 ). Approximately one half of our participants in the 1500 μg group achieved a post-intervention serum 25(OH)D level ≥75 nmol/l, compared with just 1 % in the placebo group. Thus, for comparisons between the 1500 μg group and the placebo group, it seems unlikely that our mostly null results are because of inadequate serum 25(OH)D levels following supplementation.
This study has some limitations. Diet influences systemic inflammation, and thus it was not ideal to have used non-fasting blood samples. Although this has the potential to influence our summary statistics, the randomisation eliminates any risk of biased estimates of treatment effect that might have otherwise resulted. Because we did not measure baseline levels of inflammatory markers, we could not assess baseline equivalence and could not analyse changes in levels. The lack of baseline data is extremely unlikely to have led to any bias in these results. The trial groups were extremely well balanced at baseline( Reference Tran, Armstrong and Ebeling 45 ), and it would be extraordinary if the distribution of any biomarkers differed before the intervention. Thus, any differences after the intervention are an indication of change from baseline. We used a monthly bolus dose of vitamin D, which some have argued might confer suboptimal benefits (for some endpoints) when compared with daily dosing( Reference Bergman, Lindh and Bjorkhem-Bergman 46 – Reference Hollis and Wagner 48 ). However, two studies found that daily or monthly dosing did not produce significantly different serum 25(OH)D levels after 2 months( Reference Ish-Shalom, Segal and Salganik 49 ) or 1 year( Reference Binkley, Gemar and Engelke 50 ) of vitamin D3 supplementation. Given the relatively long half-life of vitamin D and 25(OH)D( Reference Vieth 51 ), variability in timing of post-intervention blood samples should also not be responsible for our largely null findings. Although non-differential follow-up could introduce bias, our extremely high completion rate (96 %) means that any bias would be negligible and would not have materially influenced the results. Caution should also be exercised in generalising our results, as our population was older and predominantly (95 %) Caucasian.
Our study also has strengths. It is one of the largest randomised placebo-controlled trials with measurement of inflammatory markers conducted to date, and, importantly, our participants were recruited from the general population. Consequently, our results may be more relevant to the general elderly population than those from studies focusing on people with specific diseases. The duration of our intervention and the doses administered were sufficient to elicit measurable increases in serum 25(OH)D levels, and our study had high rates of retention and adherence (96 % of participants reported taking at least ten of the twelve study tablets).
With the possible exception of IL-6, we found little evidence of an effect of vitamin D supplementation on cytokine or adipokine levels. These results highlight the uncertainty surrounding the effectiveness of vitamin D supplementation in improving inflammatory biomarker profiles and the need to better understand the effects of vitamin D supplementation before recommending that it is routinely used in the general population.
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S0007114515002366
This project was funded by the National Health and Medical Research Council (NHMRC) of Australia (grant no. 613655). P. M. W., R. E. N., A. J. V., R. M. L. and D. C. W. are supported by Fellowships from the National Health and Medical Research Council. The investigational product was supplied free of charge by Sanofi-Aventis Healthcare Pty Ltd trading as Sanofi Consumer Healthcare, Virginia, QLD, Australia.
M. W. performed the statistical analysis. B. T. and M. W. wrote the manuscript. R. E. N. coordinated the study, contributed to writing the manuscript and provided overall supervision. All authors contributed to study design and data interpretation and read and approved the final version of the manuscript.
P. R. E. has received prior funding from Sanofi-Aventis Healthcare. All other authors have no conflicts of interest to disclose.