INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Antioxidant vitamins (provitamins), such as vitamins C, E, and carotenoids may beneficially affect a permanent loss of lung function over time (1) by scavenging endogenous and/ or environmental oxidants.

Several carotenoids such as -carotene, -carotene, lycopene, lutein/zeaxanthin, and -cryptoxanthin show antioxidant capacity in vitro (5, 6) and have been measured in reasonable amounts in lung tissue (7). Individual carotenoids may differ in their role in lung physiology. These levels of carotenoids in lung tissue have correlated more strongly to blood levels of carotenoids than to dietary intake of carotenoids measured by questionnaire (7). One of the first epidemiological studies investigating the dietary intake of total carotene in relation to lung function did not show an association (8). However, other cross-sectional studies found positive associations between lung function, levels of serum -carotene (9), plasma -carotene (10) and dietary intake of -carotene (11). Other carotenoids as found in lung tissue have not been studied in relation to lung function.

Two studies among adults did not show associations between blood levels of -tocopherol (vitamin E) and the development of airway obstruction (12) or lung function (10). In addition, dietary vitamin E was not independently associated with lung function in adults (11, 13). However, in elderly subjects dietary vitamin E was positively associated with lung function (14).

Elderly individuals may be particularly vulnerable for low levels of antioxidant vitamins in relation to lung function as exposure to free radicals is increased and obstructive pulmonary disease is an important cause of disability and death in old age (15). Therefore, we studied the association between lung function and the serum levels of the six major serum carotenoids and -tocopherol among elderly subjects.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The present study used data derived from a survey conducted in 1991-1992 on life-style and health among noninstitutionalized Dutch elderly living in the city of Arnhem, age 65 to 85 yr. The selection of this study population has been described in detail elsewhere (16). Between October 28, 1991 and April 6, 1992 a random sample of elderly, prestratified on gender and age, participated in a health survey, including home interviews (n = 1,012) and a physical examination (n = 685). Written informed consent was obtained from the subjects before the physical examination. The physical examination included measurements of height, weight, and pulmonary function. In addition, nonfasting blood samples (n = 641) were drawn. In 1998, serum levels of carotenoids and -tocopherol could be determined in a total of 638 subjects. Of these 638 subjects, 528 had technically acceptable and reproducible lung function measurements that were used in the data analysis.

Nonfasting serum samples had been stored for a mean time of 6 yr at 80° C before vitamin analyses. After extraction (17), the concentrations of the carotenoids -carotene, -carotene, lycopene, lutein, zeaxanthin, and -cryptoxanthin and of -tocopherol were measured in serum, by reverse-phase high performance liquid chromatography (HPLC) [adapted from Hess and coworkers (18) and Craft and Wise (19)]. Detection after separation was carried out using two ultraviolet (UV) detectors, one for determination of carotenoids (UV 2000) and one (UV 1000) for determination of -tocopherol. The coefficient of variation (CV = SD/mean) of 34 pooled serum samples (from duplicate measurements in every run; n = 17) was 12.3% for -carotene, 10.2% for -carotene, 23.9% for lycopene, 10.7% for -cryptoxanthin, 14.9% for zeaxanthin, 7.2% for lutein, and 5.1% for -tocopherol.

Pulmonary function measurements were performed according to the protocol of the European Community of Coal and Steel (ECCS) guidelines of 1983 (20) by well-trained technicians. A Vicatest 5 dry rolling seal spirometer (Mnhardt, Bunnik, The Netherlands) was coupled to a microcomputer for storage of the data. Calibration of the spirometers was done every day with a 3-L syringe. All pulmonary function maneuvers were adjusted for body temperature and pressure-saturated with water vapor (BTPS). Analyses were based on the maximal value of the reproducible FEV₁ and FVC. Of the 638 subjects with serum levels of antioxidants, 68 subjects could not fulfill at least three technically acceptable maneuvers according to ECCS criteria (20), such as no hesitant start, no cough, and no early termination of the maneuvers. Forty-two subjects could not fulfill the reproducibility criteria of the ECCS of 1983, that is, the highest FVC should not be more than 300 ml different from the second highest FVC.

The interview, which was mainly precoded, included questions on demographic and lifestyle characteristics. No data on dietary intake were collected. Present and past tobacco consumption and duration was assessed. Pack-years were calculated for current and former cigarette smokers as the number of cigarettes smoked times the number of years smoked divided by 20. Zero pack-years was given to never, cigar, and pipe smokers. Alcohol consumption was classified into four categories (0, 0-1, 1-3, > 3 glasses per day). Physical activity was assessed by a validated questionnaire on household activities, sports, and other physically active leisure time activities, developed for free-living elderly. From the responses a total activity score was calculated using an intensity code based on net energetic costs of the specific activities (21). Subjects reported the presence or absence of respiratory disease including asthma, chronic bronchitis, emphysema, or the use of respiratory medication. The use of prescribed medication was asked with a reference period of 3 months before the interview.

All data analyses were performed with the statistical software package SAS (version 6.12). First, the mean FEV₁ and FVC were calculated for the different quintiles of carotenoids and -tocopherol after adjustment for age, height, gender, and pack-years of smoking using analysis of covariance (ANCOVA) (22). Adjustments were based on a basic model using multiple linear regression for FEV₁ and FVC including important predictors of lung function, such as age, height, gender, and different powers of height and age. The choice of the "best" basic model was based on assessment of model simplicity, analysis of residuals, and the percentage of variance explained by the model.

In the multiple regression model, the serum levels of the carotenoids and -tocopherol were classified into quintiles, with the lowest quintile as a reference. Pack-years of smoking was included in the model as a confounder in the relation between antioxidant vitamins and lung function. Evaluated as possible confounders were educational level, smoking status, pack-years of smoking, body mass index (BMI), alcohol consumption, antioxidant supplement use, self-reported lung disease, and serum level of total cholesterol.

Test for trend was performed by fitting the serum antioxidants as continuous variables in the multiple linear regression model. We were not able to perform statistical evaluation of the effect modification by smoking status on the relation between the serum antioxidants and lung function because of the small numbers in each group.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 1 shows the demographic, life-style, and pulmonary function characteristics of the 528 study subjects. The study population had a mean age of 74 yr and 55% were men. Men were more likely to be highly educated, smokers, alcohol consumers, and physically active compared with women. The mean FEV₁ was 2.63 L (10 to 90% range of 1.58 to 3.60 L) for men and 1.83 L (10 to 90% range of 1.16 to 2.53 L) for women. The mean FVC for men was 3.88 L (10 to 90% range of 2.91 to 4.86 L) and for women 2.60 L (10 to 90% range of 1.79 to 3.38 L). The percentage of subjects with airway obstruction (FEV₁/ FVC < 70%) was 47%.

Of the original population of 638 elderly individuals, 110 had no technically acceptable and reproducible lung function measurements and were excluded from the analyses. These excluded subjects were on average older and a higher percentage were women. No differences in antioxidant vitamin serum levels were observed between the two groups.

Except for lycopene, all serum carotenoid concentrations were higher in women than in men, ranging from 12% higher for lutein to 69% higher for -cryptoxanthin. Most of the individual carotenoids were inversely associated with pack-years of smoking and BMI, with Spearman correlations ranging from 0.11 to 0.22 (p < 0.05). Except for lycopene (r = 0.16; p = 0.0002) and -tocopherol (r = 0.12; p = 0.007), none of the other vitamins were significantly associated with age. No associations for the vitamins with physical activity, lung disease, or lung obstruction were observed. Higher levels of serum vitamins were found for higher levels of educational level, but this was not statistically significant.

The individual carotenoids were highly correlated (p = 0.0001), with the sum of these individual carotenoids ranging from Spearman correlation of r = 0.50 for lutein to r = 0.80 for -carotene. Serum concentrations of -tocopherol showed a Spearman correlation of r = 0.30, p = 0.0001 with total carotenoids. Correlating the carotenoids among themselves, the highest correlations for the carotenoids were seen for - and -carotene and for lutein and zeaxanthin, r = 0.78 and r = 0.67, respectively. Lowest correlations were observed for lycopene with the other carotenoids ranging from r = 0.14 with zeaxanthin to r = 0.41 for -carotene.

FEV₁ and FVC were higher among men than women (Table 1). FEV₁ was highly correlated with age (r = 0.32; p = 0.0001), height (r = 0.59; p = 0.0001), physical activity score (r = 0.31; p = 0.0001), serum cholesterol (r = 0.11; p = 0.01) but not with BMI (r = 0.07; p = 0.13). As 58% of women were never smokers, pack-years of smoking was correlated with FEV₁ only in men (r = 0.22, p = 0.0001). For FVC the correlations were somewhat stronger but similar in magnitude. Lower levels of FEV₁ and FVC were observed in current and ex-smokers compared with nonsmokers and in subjects with lung disease. Higher levels of FEV₁ and FVC were observed with higher levels of educational level and higher consumption of alcohol.

The mean FEV₁ and FVC and the upper 95% confidence limit of each quintile of serum antioxidant concentration after adjustment for age, height, gender, and pack-years of smoking are shown in Figure 1. The median levels of the antioxidant vitamins for each quintile are given within each bar. The mean FEV₁ and FVC increased from the first until the last quintile for most of the serum carotenoids although there were some fluctuations between the middle quintiles. The mean FEV₁ and FVC differed significantly between the first and fifth quintile of serum total carotenoids, -carotene, -carotene, and lycopene. The median serum total carotenoid concentration was 1.08 µmol/L; -carotene (30%), lutein (23%), and -cryptoxanthin (21%) were the three major contributors. -Carotene had a small contribution to the total carotenoid concentration with 4%.

View larger version (53K): [in this window] [in a new window]   Figure 1.   Mean FEV1 and FVC adjusted for age, height, gender, and pack-years of smoking and upper 95% confidence interval for quintiles of the serum antioxidant vitamins with the median values of the antioxidant vitamins in the bars (µmol/L).

Table 2 shows the associations between antioxidants in quintiles and FEV₁ adjusted for age, height, gender, and pack-years of smoking. The FEV₁ was mostly higher in the second to fourth quintile of all carotenoids and -tocopherol compared with the first quintile of these antioxidants but this was often not statistically significant. In addition, the FEV₁ was higher in the fifth quintile of serum lutein, zeaxanthin, and -tocopherol compared with the first quintile but this was not statistically significant. The FEV₁ in the fifth quintile of -carotene, -carotene, and lycopene were significantly higher compared with the first quintile of these serum antioxidants. Significant (p < 0.05) positive trends were observed between FEV₁ and serum -carotene, -carotene, and lycopene (Table 2).

For FVC similar results can be observed as for FEV₁ (see Figure 1). After additional adjustment for pack-years of smoking, subjects in the fifth quintile of -carotene had a 290 ml (95% confidence interval [(95% CI]: 131 to 450 ml) higher FVC and subjects in the fifth quintile of -carotene had a 271 ml (95% CI: 113 to 429 ml) higher FVC compared with subjects in the first quintile of these carotenoids. Significant positive trends were observed between FVC and -carotene (p = 0.004) and -carotene (p = 0.001). The fourth and fifth quintile of lycopene were higher (respectively, 104 ml [95% CI: 55 to 263 ml] and 133 ml [95% CI: 26 to 292 ml]) but these quintiles were not significantly different compared with the first quintile. Although the FVC seems to increase with increasing quintiles of lycopene, a trend between FVC and lycopene was not statistically significant (p = 0.11).

The variance for FEV₁ and FVC was explained for 43% and 62%, respectively, by an ANCOVA model including quintiles of total carotenoids, age, gender, and height. Including also pack-years of smoking changed the explained variance from 43% to 46% for FEV₁ and from 62% to 64% for FVC. Additional inclusion of a squared term of pack-years or smoking status did not change the explained variance for FEV₁ or FVC. Similar figures were found for all the individual carotenoids and -tocopherol.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The serum carotenoids, -carotene and -carotene were positively associated with FEV₁ and FVC. Lycopene was associated with FEV₁ but not statistically significantly with FVC. The lung function of the third to fifth quintile of most of the other carotenoids and -tocopherol was higher compared with the first quintile but these associations and the trends over the quintiles were not statistically significant.

Of our elderly (65 to 85 yr) study population, prior to analysis we had to exclude 10% of the subjects as they were not able to perform technically acceptable lung function measurements and 7% of the subjects were not able to produce reproducible lung function measurements. These percentages of subjects were not considerably different compared with the percentages in a Dutch study among younger adults (20 to 59 yr), respectively, 7% and 6% (11).

Because subjects were excluded for not meeting acceptability and reproducibility criteria, selection bias may have affected the associations between lung function and serum carotenoids. This issue of selection bias can be partly disentangled by evaluating the associations when, on one hand, a less strict definition of lung function reproducibility was used, that is, including 42 subjects in the analyses who did not meet reproducibility criteria of ECCS 1983, and, on the other hand, a stricter definition of lung function reproducibility was used in our study, that is, the 1993 European Respiratory Society criteria (23). The associations between lung function and the serum levels of carotenoids were not essentially changed after including subjects not fulfilling 1983 criteria or excluding subjects who did not fulfil these stricter reproducibility criteria (n  = 123, 23% for FEV₁ and n = 105, 20% for FVC measurements) so selection bias probably did not occur in these associations. However, selection bias can not be totally excluded because it remains unclear whether the associations would change when subjects without technically acceptable lung function measurements were included in the analysis.

We took into account all known confounding factors, such as age, gender, height, and pack-years of smoking in the association between serum antioxidants and lung function. Other demographic and life-style factors, such as alcohol consumption, were evaluated as possible confounding factors but did not affect the association between serum carotenoids, -tocopherol, and lung function. However, we can not exclude that due to measurement error in possible confounders some residual bias might have occurred. Because this study was cross-sectional in design, we can not exclude the possibility that reverse causality may have occurred in our associations. However, we have no reason to believe from the current literature that serum antioxidant levels are decreased as a consequence of a decline in lung function.

Our results on plasma carotenoids and lung function were probably not affected by storage time because all serum samples were stored at 80° C for 6 yr, which was observed not to affect serum concentrations of carotenoids and -tocopherol (24, 25).

In a study comparing individual carotenoids among elderly subjects from cities in the United States and 10 different European countries (including one Dutch center), absolute median serum concentrations of -carotene, lutein, and -cryptoxanthin were mostly of similar magnitude as in our study (26). Only serum lycopene concentrations varied considerably among study populations. In our population absolute lycopene concentrations were fourfold less than in the elderly populations from the United States, France, Ireland, and Italy (26). As lycopene is mainly derived from the intake of tomato and tomato products, differences in diet between these countries are likely to explain the observed differences in serum.

We observed positive associations between serum -carotene, -carotene, lycopene, and FEV₁ and between serum -carotene, -carotene, and FVC. The association between serum lycopene and FVC was not statistically significant. A validation study (n = 21) in which most carotenoids were measured in serum and in lung tissue observed that the proportion in the lung tissues of lycopene was slightly lower and -carotene was higher compared with the proportions of these carotenoids in serum levels (7). The other carotenoid levels had approximately similar proportions in serum and lung tissue. Therefore, the results of this small validation study suggest that lycopene and -carotene concentrations in blood may not represent the actual concentrations of that in lung tissue. Thus, the associations between these serum carotenoids and lung function may not be similar when extrapolating this to the lung tissue.

One of the first studies on dietary total carotene in relation to lung function did not observe an association (8). The authors of that study pointed out, however, that different carotenoids might have individual effect on lung physiology. Since then, two other cross-sectional studies were published on dietary -carotene in relation to lung function. One of the two studies (n = 6,555) (11) did show an association between dietary -carotene and lung function whereas the other study did not show such an association (n = 798) (9). Recently, Michaud and coworkers suggested that measurement errors occurred in the old carotenoid database because the new database showed better correlations between carotenoids measured in plasma and by questionnaire in two American cohorts (27). Therefore, blood levels of carotenoids are important to study in relation to lung function. In addition, a validation study observed that blood levels of carotenoids were probably better markers for the concentrations in the lung tissue than carotenoids measured by questionnaire (7). We found positive associations between serum -carotene and lung function. This was consistent with the only two other published studies investigating the associations between blood levels of -carotene and lung function (9, 10). To our knowledge, the present study is the only one reporting other carotenoids than -carotene in relation to lung function.

The magnitude of the association between serum -carotene and lung function in our study can be compared with the two other cross-sectional studies. One study comprised 798 men age 45 to 74 yr and a difference in the concentration of serum -carotene of 0.29 µmol/L was significantly associated with a 90 ml higher FEV₁ and an 82 ml higher FVC (9). In another study among Dutch adults (n = 367) age 20 to 59 yr, a similar difference of 0.29 µmol/L plasma -carotene would be associated with a 46 ml higher FEV₁ and a 93 ml higher FVC. These differences in lung function, in particular for FVC, for a difference of 0.29 µmol/L serum -carotene were in line with our study, with a 84 ml higher FEV₁ and a 114 ml higher FVC.

In the present study, the difference in FEV₁ between the fifth quintile compared with the first quintile of serum total carotenoids was 210 ml. The magnitude of this difference in FEV₁ is equivalent to eight times the annual decline of FEV₁ (25 ml) in nonsmoking Dutch adults (28). Furthermore, the difference was approximately equivalent to the adverse effects of 14 yr of heavy smoking ( 25 cigarettes per day) on FEV₁ decline in Dutch adults (29). The present study did not show a significant association between serum -tocopherol and lung function, although there was a difference in FEV₁ of 94 ml between the fifth and first quintile of -tocopherol. These results are consistent with a small follow-up study in 83 men. In this study, serum levels of -tocopherol were not associated with the development of airway obstruction (FEV₁  75% FVC) during five subsequent years (12). In addition, our results are also consistent with a cross-sectional study among 367 Dutch adults in which no association was observed between plasma -tocopherol and lung function (10). Although dietary vitamin E was independently associated with lung function in 178 elderly subjects, this could not be confirmed in two large studies (n > 2,500) among adults (11, 13). So far, there is no clear evidence that vitamin E (-tocopherol) is related to lung function. We carefully speculate that serum levels of carotenes may better indicate lung tissue levels than serum -topherol as suggested by a small validation study (n = 21) (7). In this study the correlation between serum and lung tissue -carotene (r = 0.72) was higher than for serum and tissue -topherol levels (r = 0.47). Furthermore, it can, of course not be ruled that the observed associations between carotenoids and lung function may be linked to other protective constituents or factors related to fruit and vegetable intake (30).

In conclusion, the results suggest that serum -carotene and additionally -carotene were positively associated with lung function. Serum lycopene was positively associated with FEV₁ but not significantly with FVC. The other serum carotenoids and -tocopherol were not associated with lung function.