Statistics from Altmetric.com
What is the key question?
Do low vitamin D levels associate with lung function decline and future risk of chronic obstructive pulmonary disease (COPD).
What is the bottom line?
Low plasma levels of 25-hydroxyvitamin D were associated with faster lung function decline and lower lung function. Furthermore, low plasma levels of 25-hydroxyvitamin D were also associated with higher risk of COPD in two separate large general population studies.
Why read on?
This is the first study to show that low plasma levels of 25-hydroxyvitamin D are associated with faster lung function decline and future risk of COPD in the general population.
The most important pathogenic mechanisms involved in development of chronic obstructive pulmonary disease (COPD) are protease-antiprotease imbalance, inflammation, lung remodelling and oxidative stress.1–8 Interestingly, lower vitamin D levels have been related to regulation of each of these processes, that is, higher expression of proteases, modulation of inflammation, modulation of extracellular matrix turnover, and increased oxidative stress as reviewed elsewhere.9–13 Thus, one can speculate that lower vitamin D levels may be linked to increased risk of developing COPD.
In some clinical studies, lower levels of vitamin D, measured as plasma 25-hydroxyvitamin D (25(OH)D), have been associated with lower lung function and faster lung function decline, but the results have been conflicting.14–17
We tested the hypothesis that lower plasma 25(OH)D levels associate with lower lung function, faster lung function decline and increased risk of COPD. For this purpose, we studied 10 116 white individuals from the Copenhagen City Heart Study followed for up to 20 years, and 8391 white individuals from the Copenhagen General Population Study.
The Copenhagen City Heart Study is a prospective cohort study of the Danish general population initiated in 1976–1978 with follow-up examinations in 1981–1983, 1991–1994 and 2001–2003.4 ,18 Individuals 20–100 years of age were drawn randomly from the national Danish Central Person Register to reflect the Danish general population and invited to participate. The present study included 10 116 participants from the 1981–1983 examination (70% participation rate), who had plasma samples available for 25(OH)D measurement and had valid lung function measurements.
The Copenhagen General Population Study is a general population study initiated in 2003 with ongoing enrolment (participation rate 45%), recruited as the Copenhagen City Heart Study.18 The present study included 8391 participants from July 2004 to May 2005, who had plasma samples available for 25(OH)D measurement and had valid lung function measurements.
The Copenhagen City Heart Study and the Copenhagen General Population Study were approved by Danish ethical committees and Herlev Hospital. All participants gave written informed consent.
Measurements of 25-hydroxyvitamin D
Copenhagen City Heart Study plasma samples were collected in 1981–1983 and Copenhagen General Population Study plasma samples were collected in 2004–2005, and were stored at −20°C and −80°C, respectively, until 2009–2010 when 25(OH)D was measured using the DiaSorin Liaison 25(OH)D TOTAL assay.19 Assay precision was tested daily, while assay accuracy was tested monthly using an external quality control programme. The interassay coefficient of variance was 10% for low level controls (∼40 nmol/L) and 8% for high level controls (∼135 nmol/L).
In the Copenhagen City Heart Study, forced expiratory volume in 1 s (FEV1) and forced vital capacity (FVC) were determined using an electronic spirometer (Monaghan N 403; Monaghan, Littleton, Colorado, USA) at the baseline examination in 1981–1983 and with a dry wedge spirometer (Vitalograph; Maids Moreton, Buckinghamshire, UK) at the 1991–1994 and 2001–2003 examinations. In the Copenhagen General Population Study FEV1 and FVC were determined using the dry wedge spirometer.
Instruments were calibrated daily against a 1 L syringe. Each spirometry was performed in triplicate, and results were only accepted if variation between the two best-performing of these was less than 5%. Algorithms for calculation of FEV1 % predicted and FVC % predicted were made using multiple regression with age and height as covariates on a subsample of never-smokers for men and women separately.
Information on covariates was obtained from self-reported questionnaires reviewed together with an examiner at the day of attendance. Daily tobacco consumption (grams per day) was calculated for current smokers. A substantial proportion of smokers used other types of tobacco than cigarettes. Previous analyses of lung function decline showed that the approach of converting cigarettes, cigars, cheroots and pipe into gram of tobacco was useful as all types of smoking were associated with an increased decline of FEV1 and FVC.20 Cumulative tobacco consumption was calculated for former and current smokers in pack years; a pack year was defined as 20 g of tobacco per day for a year. Participants were also asked about duration and intensity of daily physical activities, and about their level of income. Body mass index (BMI) was calculated as measured weight (kilograms) divided by measured height (meters) squared.
In addition to analysis of a possible association of FEV1 and FVC with plasma 25(OH)D, we also tested the association of plasma 25(OH)D levels with spirometric diagnosis of COPD defined in two ways: (1) by the GOLD criteria demanding the presence of FEV1/FVC<0.7 and (2) using the lower limit of normal (LLN) definition for FEV1/FVC calculated as the difference between the predicted value and 1.645 times the SE of the estimate separately for men and women. The associations were evaluated cross-sectionally between spirometrically derived endpoints in both populations. In the Copenhagen City Heart Study, we had up to three sets of spirometry measurements on each individual during a follow-up of up to 20 years. Hence, we also tested the association of baseline plasma 25(OH)D levels with lung function decline and prospective development of spirometrically defined COPD in this population. Participants with spirometrically defined COPD at baseline were excluded from the prospective analyses.
The levels of 25(OH)D was expected to vary according to time of year due to the high latitude geographical position of Denmark. Therefore, we used seasonally adjusted 25(OH)D percentiles in our analyses. We obtained season based percentiles for each population separately using locally weighted regression.21
The association of lung function with seasonally adjusted plasma 25(OH)D levels was evaluated using deciles, whereas to maximise statistical power, quintiles were used in the analyses concerning decline in lung function and risk of COPD. The cross-sectional association of FEV1 % predicted, FVC % predicted and FEV1/FVC with seasonally adjusted plasma 25(OH)D deciles was carried out using Cuzick's non-parametric trend test. We used quantile regression to test for interactions.
Triple measures of lung function were not available for all participants in the Copenhagen City Heart Study (26% had three measurements, 32% had two measurements, and 41% had one measurement), so lung function was analysed in a repeated measures linear mixed model. The number of examinations at the Copenhagen City Heart Study specified the repeated measurements of lung function; an unstructured covariance type for residuals was used as it places no restriction on structure. Identity of each participant was used as a random effect. The interaction term between plasma 25(OH)D percentiles and age was used to test the association of baseline plasma 25(OH)D percentile with prospective change in lung function. Sensitivity analyses using height adjusted FEV1 and FVC were carried out in similar models.22 Model estimates were adjusted for age, sex, height, cumulative tobacco consumption and use of different spirometers between examinations.
Cross-sectional analyses concerning risk of COPD with decreasing levels of seasonally adjusted plasma 25(OH)D were carried out using logistic regression. Cox proportional hazards regression was used for prospective analyses with age as time scale with delayed entry (left truncation). Thus, age differences were automatically adjusted for and referred to in text, tables and figures as age adjusted. Both models included the following covariates that were forced into the models: gender, smoking status (never or ever smoker), cumulative tobacco consumption (pack years), BMI, level of daily physical activity and level of income. Furthermore, age was also included as a variable in logistic regression analyses.
Relative risk estimates and CIs were corrected for regression dilution bias,23 using plasma 25(OH)D from 400 individuals without COPD, cancer, cardiovascular disease and other chronic diseases participating in the 1981–1983, 1991–1994 and 2001–2003 examinations of the Copenhagen City Heart Study; the regression dilution ratio was 0.45.24
The data was 99.8% complete in the Copenhagen City Heart Study and 99.3% complete in the Copenhagen General Population Study in relation to the covariates (see online supplementary table S1); the missing data were imputed using multivariable chained imputation, where age and gender were independent variables and variables with missing observations were dependent variables in the model. We analysed the data with the statistical package Stata V.12.1 (College Station, Texas, USA).
Table 1 summarises baseline characteristics by plasma 25(OH)D levels in both cohorts. Lower levels of 25(OH)D were associated with male gender, current smoking, higher current tobacco consumption, higher cumulative tobacco consumption, higher BMI, lower physical activity and lower income. There was seasonal variation across the months of the year with high levels in August–September and low levels in January–February (see online supplementary table S2). The median 25(OH)D concentration was 41 nmol/L (IQR: 26–58 nmol/L) in the Copenhagen City Heart Study and 44 nmol/L (30–60 nmol/L) in the Copenhagen General Population Study.
Lung function: cross-sectional analyses
In both cohorts, FVC % predicted was 7% lower and FEV1 % predicted was 7–10% lower for lowest versus highest decile of 25(OH)D (ptrend≤1×10−28) (figure 1). However, the association of plasma 25(OH)D with FEV1/FVC was inconsistent. Similar results were seen when stratifying the populations into never smokers, smokers who have smoked up to 20 pack years and smokers having smoked more than 20 pack years (see online supplementary figure S1). Tests for interaction showed a significant interaction of plasma 25(OH)D with cumulative tobacco consumption on FEV1 % predicted (Copenhagen City Heart Study: p=0.001 and Copenhagen General Population Study: p=0.009), whereas the tests regarding FVC % predicted and FEV1/FVC were non-significant. The results suggest that the association of 25(OH)D with FEV1 % predicted is stronger with increasing levels of cumulative tobacco consumption.
Separate analyses for the subgroup of participants with spirometrically defined COPD at the baseline examination showed that FVC % predicted and FEV1 % predicted were both 12% lower for lowest versus highest quintile of 25(OH)D in the pooled cohort (see online supplementary figure S2). Lower plasma 25(OH)D also weakly associated with lower FEV1/FVC in participants with spirometrically defined COPD with FEV1/FVC being 2% lower for lowest versus highest quintile of 25(OH)D. Furthermore, tests for interaction showed a significant interaction of plasma 25(OH)D with spirometrically defined COPD on FEV1 % predicted (p=2×10−6), FVC % predicted (p=1×10−7) and FEV1/FVC (p=0.03). The results suggest that the association is stronger in the subpopulation with COPD compared with the remaining population.
Lung function: prospective analysis
Prospective analysis of change in lung function in the Copenhagen City Heart Study over time showed a significant interaction of age with quintiles of plasma 25(OH)D levels on course of FEV1 % predicted (p=1×10−7) and FVC % predicted (p=1×10−7) (figure 2). Participants in the lowest quintile of seasonally adjusted plasma 25(OH)D had a faster decline in lung function compared with participants in the highest quintile. Based on longitudinal models, average FEV1 % predicted for participants of age 60 years and adjusted for cumulative tobacco consumption was 87% (95% CI 86% to 88%) in the lowest quintile compared with 94% (93% to 95%) in the highest quintile. The corresponding values for FVC % predicted were 92% (91% to 93%) and 98% (97% to 98%). The decline in FEV1/FVC was similar for all five groups. The analyses were adjusted for cumulative tobacco consumption updated at the follow-up examinations.
Sensitivity analyses using height adjusted FEV1 and FVC as the dependent variables showed similar results with a greater decline in lung function in the lowest quintile of seasonally adjusted plasma 25(OH)D compared with the highest quintile (see online supplementary figure S3). Based on the longitudinal models, for male participants of age 60 years, height 175 cm, and adjusted for cumulative tobacco consumption average FEV1 was 2.79 L (95% CI 2.76 to 2.83) in the lowest quintile compared with 3.00 L (2.97 to 3.03) in the highest quintile. The corresponding values for FVC were 3.88 L (3.85 to 3.92) and 3.65 L (3.62 to 3.69).
Using data only from participants with more than one lung function measurement showed a similar trend, that is, higher decline in lung function within the lowest quintile of plasma 25(OH)D, when looking at all participants and at continuous smokers (table 2, see online supplementary figure S4). However, the results were not significant for never smokers or participants who quit smoking during follow-up. Analyses stratified according to spirometrically defined COPD at baseline showed that patterns of lung function decline were similar in both groups: in the lowest quintile of vitamin D, the decline of FEV1 and FVC was faster than among individuals with higher vitamin levels (see online supplementary figure S5). However in the COPD group, only the association between decline in FEV1 % predicted and vitamin D reached statistical significance.
Risk of spirometrically defined COPD
In the cross-sectional analysis, multivariable adjusted ORs for COPD defined by spirometry using GOLD criteria or LLN criteria increased with decreasing levels of 25(OH)D (figure 3). The ORs for lowest versus highest quintile were 2.30 (95% CI 1.55 to 3.41) and 3.06 (1.97 to 4.76) in the Copenhagen City Heart Study for GOLD and LLN criteria, respectively. The corresponding ORs were 1.82 (1.13 to 2.92) and 2.23 (1.35 to 3.69) in the Copenhagen General Population Study.
In the Copenhagen City Heart Study many participants had follow-up spirometry carried out at 10-year (N=5819) and 20-year (N=2631) intervals, which were used for prospective development of COPD as a function of baseline 25(OH)D levels (figure 4). Longitudinal analyses restricted to these participants and excluding participants with spirometrically defined COPD at baseline comprised 5341 and 5527 individuals with at least two measurements for analyses when using GOLD criteria and LLN criteria, respectively. Multivariable adjusted HRs for COPD for lowest versus highest quintile were 1.58 (1.05 to 2.40) and 2.00 (1.19 to 3.36) for GOLD and LLN criteria, respectively.
In this study based on two independent samples of the general population, we present two novel findings showing that lower levels of plasma 25(OH)D were associated with a higher decline in lung function and with a higher risk of spirometrically defined COPD in prospective analyses. Furthermore, lower levels of plasma 25(OH)D were associated with lower FEV1 and FVC and a higher risk of spirometrically defined COPD in cross-sectional analyses.
The latter finding is in accordance with the third National Health and Nutrition Examination Survey.14 One study has shown a similar association only among smokers and another study has shown no association at all.15 ,17 Compared with previous studies our study provides a larger total sample size and independent validation in a second data set that differs with regard to population structure and time of examination, thus extending the external validity of the results. Additionally, we have shown that the association is present among smokers and never smokers alike, although it is strongest among smokers regarding FEV1 % predicted. Furthermore, the association seems to be strongest among the subgroup with spirometrically defined COPD, typically those that smoke the most.
The association of lower plasma 25(OH)D with faster decline of FEV1 % predicted and FVC % predicted in the general population is a novel finding. Previously, the Normative Aging Study found a similar association in a smaller population consisting of men only16; however, the association was restricted to smokers only. Another study examining a cohort of smokers with COPD did not show any association between plasma 25(OH)D and rate of decline in lung function;15 that study was probably underpowered to detect an association given the relatively modest association of plasma 25(OH)D with lung function. In contrast to the latter study, we found that lower plasma 25(OH)D was associated with faster decline in FEV1 % predicted in those with spirometrically defined COPD at baseline. In addition, studies of lung function decline in established COPD may show considerable variation depending on survival bias, regression to the mean and confounding by stage of COPD.25
The prospective association of lower plasma 25(OH)D with increased risk of COPD is another novel finding. This study and previous cross-sectional analyses have shown that COPD patients have lower levels of plasma 25(OH)D.26 ,27 However, cross-sectional associations cannot show whether lower levels of plasma 25(OH)D precedes COPD, or vice versa. Our prospective analyses uniquely indicate that lower plasma 25(OH)D is associated with higher risk of future development of spirometrically defined COPD.
A somewhat puzzling finding was that lower 25(OH)D was associated with increased risk of spirometrically defined COPD, but not with decline in FEV1/FVC. However, our interaction analyses suggest that the association is stronger with higher cumulative tobacco consumption and among those with spirometrically defined COPD, that is, low levels of vitamin D may play a stronger role among individuals who smoke more and/or have a lower lung function. Nevertheless, these observations do not fully explain our findings and additional research is required to answer this question.
A potential limitation of our study is that our cohort consists of only white individuals of Danish descent living in Denmark (55°–58° latitude North), with less sun exposure than closer to the equator; consequently, our findings would be most applicable to individuals with a similar skin colour and a similar level of sun exposure. The delay in measurement from 1981 to 1983 or 2004–2005 to 2009–2010 could raise concern of potential decay of plasma 25(OH)D, but this seems unlikely to have distorted our analyses for several reasons: we noticed the expected seasonal variation of 25(OH)D concentrations in both studies, median levels of plasma 25(OH)D across plasma samples from three different examinations on the same healthy participants from the Copenhagen City Heart Study with storage times of 10 years, 20 years and 30 years were similar,24 ,28 previous studies have shown high stability during storage,29 the median concentration observed in our studies were similar to those in comparable populations,30 and a low sample quality for the 25(OH)D measurement would tend to weaken rather than inflate an association. A limitation is that we did not have information regarding intake of vitamin D supplement use at baseline or during follow-up. Furthermore, we only used one measurements of 25(OH)D, which could fluctuate with time; however, we and others have previously shown that 25(OH)D levels remain relatively stable with follow-up of up to 20 years.24 ,31
Our study has several strengths: independent replication of the results in a second large general population study; up to three spirometry measurements spanning 20 years with no losses to follow-up; information on other major risk factors associated with COPD; and the highest statistical power to date to examine the associations of plasma 25(OH)D levels with lung function variables and risk of COPD. Also, in Denmark UV-B radiation from the sun is only adequate for endogenous vitamin D production in the skin during the summer months and food has never been fortified with vitamin D. Thus, this cohort from the Danish general population allows determination of the natural history of the association of vitamin D levels with lung function and risk of COPD.
In conclusion, we observed an association of lower plasma 25(OH)D levels with lower lung function, a faster decline in lung function and a higher risk of COPD.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.