INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Cross-sectional associations have been demonstrated between fresh fruit consumption and lung function in national surveys of British children (1) and adults (2), and between antioxidant vitamin intake (3, 4) or nutrient status (5) and ventilatory function in adults (3, 4) and children (5). It is not clear whether these associations reflect current dietary habits or nutrition during some critical period in the past, for instance in childhood.

In this report we extend the work of Strachan and colleagues (2) who published cross-sectional data from a national survey of British adults. Using data from the 7-yr follow up examination of these individuals we assess the longitudinal effects of fresh fruit consumption on lung function within a healthy subset of an adult population. More specifically we investigate the roles played by both average level and change in the level of fresh fruit consumed in a weekly diet in determining change in forced expiratory volume in one second (FEV₁) over this same period.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The Health and Lifestyle Survey

The Health and Lifestyle Survey (HALS1) consisted of a random sample of 9,003 adults resident in England, Scotland, or Wales, who were interviewed and measured initially during 1984 and 1985. A follow-up study (HALS2) 7 yr later managed to trace 5,352 (59.4%) of the original sample. Further details of the sample selection and data collection for both studies have been published (6, 7).

The structure and method used in both surveys were similar. First an interview that collected a wide range of information including socioeconomic status, self-reported health, dietary habits, and smoking history was carried out in the respondent's home. Individuals were assigned a "household socioeconomic group" based on the 1980 Registrar General's classification (8). This interview was followed by a home visit by a nurse who carried out an array of physiological measurements, including height and respiratory function.

Fresh Fruit Consumption Score

At both HALS1 and 2 subjects were asked "How often do you eat these foods?" for over 30 different items, including fresh fruit in summer and winter, salads or raw vegetables in summer and winter, and pure fruit juice. For each food item they were asked to rate their consumption as: more than once a day, once a day, most days, once or twice a week, less than once a week, or never. These responses were assigned a score from 5 to 0 accordingly, following the methods of Cook and coworkers (1).

An overall fresh fruit consumption score was calculated by taking the mean of the winter and summer consumption levels. This was then divided into five categories for the purposes of presentation (0-0.5, 1-1.5, 2-3, 3.5-4, and 4.5-5), with the two extreme groups representing "never" and "more than once a day" on an annual basis. Change in fresh fruit consumption was calculated by the absolute difference in scores.

Spirometry

Standing height was measured with a portable stadiometer. FEV₁ was measured with an electric turbine spirometer (Micro Medical Instruments, Rochester, UK). Calibration of each instrument was carried out at the start of each wave of data collection and rechecked at the end by placing it in series with a Vitalograph spirometer. Measurements were discarded in the few cases in which significant calibration drift had occurred.

Following suitable instructions and a practice attempt, each subject performed three forced expiratory maneuvers. The maximal value of FEV₁ attained is used in the analysis here. Subjects with an acute respiratory infection had their measurements discarded, as did any whose tests were deemed unsatisfactory by the nurse (6).

As previously noted (2) there was probably a systematic underestimation of forced vital capacity (FVC) within the Heath and Lifestyle Survey, so we do not present any longitudinal analyses of FVC here.

Exclusions

Analyses in this report were restricted to Caucasians who were between 18 and 73 yr old at the first survey (HALS1) with satisfactory spirometric measurements at both examinations (n = 2,171). Data were too sparse among subjects age 74 or older at entry. Further, all subjects with a self-reported history of respiratory disease (asthma, bronchitis, or other chest problems, past or present) at either study were also excluded to reduce the variance of FEV₁ and to eliminate possible bias arising from changes in fresh fruit consumption as a response to respiratory symptoms or illness (n = 1,406). Finally, subjects were excluded if their change in FEV₁ (FEV) lay outside the middle 99% of the distribution (n = 22).

Statistical Analysis

For the purposes of analyses, subjects were grouped into five categories based on their smoking history: lifetime never smokers, ex-smokers (ex-smokers at HALS1 and not smoked since), current smokers (at both HALS1 and HALS2), quitters (smokers at HALS1 but ex-smoker at HALS2), and other smokers (any other permutation).

We used a two-step approach to our regression analyses. First we adjusted FEV₁ and FEV measurements (both in liters) for the effects of age (in years) and height (in centimeters). Cross-sectionally we used a model that included terms for age, age squared, height, and an age-height interaction. This model was chosen through an ad hoc process that involved looking at R-squares and compatibility with its implied longitudinal form. This was done for each sex separately at both sets of examinations using only lifetime never-smoking subjects with available data (246 males and 497 females). The resulting cross-sectional equations were as follows:

Males (1984):

Females (1984):

Males (1991):

Females (1991):

Longitudinally we used a model with annual change in FEV₁ (FEV/ AGE) as the dependent variable as some subjects had 6 or 8 yr between studies. The model contained terms for mean age (AGEM) and height at first examination, which is compatible with the form of the cross-sectional model with height assumed constant (see Appendix ). The resulting equations were as follows:

Males:

Females:

The regression coefficients from the above equations were used to calculate a set of predicted values and residuals (observed minus predicted). The mean residuals for FEV₁ for the 2,171 subjects were 77 ml in 1984 and 82 ml in 1991 (nonzero and negative as the model has been extended to include smokers). In the longitudinal model, for the purposes of presentation, we multiplied the residuals by 7 to represent the predicted loss between examinations (FEV, mean = 6 ml). The residuals were used as the outcome variable in our analyses.

To investigate the effect of change in fresh fruit consumption, we regressed the FEV residuals on change in fruit score using PROC GLM in SAS (SAS Institute, Cary, NC). The following were included as potential confounding variables: social class (six household socioeconomic groups plus an "other" category), region of residence (nine English regions, Scotland, and Wales), mean pack-years (estimated as the average current consumption times years of smoking at each examination), change in pack-years (increase from HALS1 to HALS2, set to zero if negative), and average level of fresh fruit consumption. Change in fresh fruit consumption score was treated as a five factor categorical variable in the regressions, except when testing for the significance of linear trends.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Fresh Fruit Consumption

Table 1 shows changes in fresh fruit consumption between examinations. Almost half (44%) showed no change in their fresh fruit scores between studies, while slightly more showed an overall increase in consumption (29%) than a decrease (27%).

Table 2 shows the relationship between change in fruit consumption and other factors. The small increase in consumption represented a significant overall shift (p = 0.044), with the largest increases seen in women, the middle age groups, social class III (skilled workers), in the Midlands and Wales and in ex-smokers at HALS2, although change in consumption was not significantly related to any of these factors.

Cross-sectional Effects of Fresh Fruit Consumption

Table 3 summarizes the cross-sectional regressions that were carried out at each examination on the 2,171 subjects with longitudinal data. The effect of fresh fruit on FEV₁ is consistent at both time points (p = 0.002 and p < 0.001 for HALS1 and HALS2, respectively), with the group consuming no fresh fruit having the lowest lung functions (some 105 ml and 188 ml less than the "daily" group at HALS1 and HALS2, respectively).

The effect is strongest in smokers at both examinations (though not statistically different from never smokers, p = 0.09 and 0.08 at HALS1 and HALS2), with the group consuming no fresh fruit showing deficits of 181 ml and 161 ml at HALS1 and HALS2, respectively, in comparison with the "daily" consumers. Although in the subset of lifelong nonsmokers with longitudinal data the effects of fruit on lung function cross-sectionally were nonsignificant (p = 0.723), when all the cross-sectional data available at HALS1 on never smokers (n = 1,590) are used, a borderline significant effect is found (p = 0.053, not shown).

Changes in FEV₁ versus Average and Change in the Levels of Fresh Fruit Consumption

In order to investigate whether these cross-sectional associations represented short-term or long-term effects, we analyzed the relationship between average fresh fruit consumption and change in FEV₁ (Table 4), and change in fresh fruit intake and change in FEV₁ (Table 5).

Average level of fresh fruit consumption was not a significant predictor of change in FEV₁ (p = 0.446, Table 4). By contrast, significant trends were seen in relation to change in fruit consumption (Table 5), with the effect most striking among those subjects who decreased their consumption by two or more categories. This trend is seen overall in all healthy subjects (p = 0.002), though of greater magnitude, albeit not significantly so, among females (p < 0.001) and current smokers (p = 0.013). For instance, males and females who decreased their fruit consumption greatly had losses of FEV₁ some 89 ml/yr and 133 ml/yr, respectively, greater than those whose fruit consumption did not change. For never smokers this difference was 102 ml while for current smokers it was approximately 167 ml.

Table 6 shows FEV in relation to fresh fruit at HALS1 and HALS2 after adjusting for all other potential confounding variables. Among the majority of subjects (on the diagonal) whose fruit intake did not change, FEV₁ was similar at HALS1 and HALS2 and FEV varied little with level of fruit consumption. Groups whose fruit consumption declined experienced a loss of FEV₁, whereas those whose intake increased generally showed an increase in FEV₁.

Other Dietary Items

The analysis was extended to include changes in consumption of other dietary items included on the food frequency questionnaires (not shown), such as salads, fruit juice, and green vegetables (with correlations r = 0.17, r = 0.11, and r = 0.09 with change in fresh fruit intake, respectively). Changes in intake of green vegetables showed a similar, but weaker trend with change in FEV₁ (p = 0.06), which was lessened by including change in fresh fruit in the model (p = 0.10), while salads and fruit juice failed to show any (p = 0.45 and p = 0.82, respectively). These items had all shown much stronger cross-sectional trends at HALS1 (p < 0.001, p = 0.01, and p = 0.07 for fruit juice, salads, and green vegetables, respectively) than at HALS2 (p = 0.13, p = 0.02, and p = 0.67 similarly).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This is the first study to consider the longitudinal effects of fresh fruit consumption on lung function. These results from a national survey of British adults indicate that changes in the level of fresh fruit consumption, as estimated from a food frequency questionnaire, are associated with change in FEV₁ over a 7-yr period, while average level of intake is not related to rate of loss of ventilatory function. This suggests that the effects of fresh fruit intake on level of FEV₁ are reversible, and not progressive.

Bias in the data collection process is unlikely to explain these results. At both time points spirometric measurements and dietary interviews were carried out at separate home visits by different observers who were generally unaware of this hypothesis. Changes in fresh fruit intake in response to poor health were ruled out by restriction to healthy subjects.

The magnitudes of the effects found were modest but potentially of epidemiological importance. Subjects who showed the greatest decrease in their fresh fruit consumption over 7 yr lost FEV₁ more rapidly than would be expected if their intake had remained the same (given their age, height, and smoking profile). We estimate this additional loss to be 167 ml (95% confidence interval, 0 to 334 ml) for smokers and 101 ml (95% confidence interval, 16 to 219 ml) for never smokers over this period. It is unlikely that both the crudity of the food frequency questionnaire and score will have contributed to any true effects being underestimated. Recent studies showing cross-sectional relationships of spirometric indices with fresh fruit (1, 2) and also vitamin C intake (3) have been based on the hypothesis that increasing the strength of antioxidant defenses in the lung may protect against oxidative tissue damage and development of airflow obstruction. It has been suggested that current increasing trends in the prevalence of asthma may be partially due to declining amounts of fresh fruit (9) and vitamin C (10) in the diet. These cross-sectional studies provide no insight into the timing or reversibility of the protective effect.

The protective effect of fresh fruit may be due to: (1) a fixed effect derived during growth and carried through adult years; (2) a progressive effect, such that low fruit intake is related to increased lung function decline; (3) a fixed reversible effect that can be modified by changing intake. Our longitudinal findings show that changing intake is related to changes in FEV₁, providing direct support for the third suggestion of a fixed reversible effect. We also demonstrated the lack of any effect of average intake on decline, which argues against the progressive effect. Since the magnitude of the longitudinal effects in our analyses are of sufficient size to explain the consistent cross-sectional effects seen in both adults and children (arguing against the first suggestion of a fixed effect derived during growth and carried through adult years), we believe that the last suggestion, a fixed reversible effect that can be modifying by changing intake is the likeliest explanation of the nature of the cross-sectional relationship.

Both the cross-sectional and longitudinal effects of fresh fruit were greater in smokers, with the longitudinal effect showing a trend with pack-years (not shown). Smokers are known to have lower circulating vitamin C concentrations than nonsmokers (11), and this difference may not simply be due to differences in vitamin C intake (12) as cigarette smoke is known to increase the total body catabolism of vitamin C (13). There may also be some confounding within smokers, as changes in fresh fruit consumption were inversely related to changes in smoking pack-years. There may be a selection effect too, as a previous analysis using all the cross-sectional data at HALS1 found greater effects in nonsmokers (2).

The main sources of vitamin C in the UK diet are thought to be (in order) vegetables (in particular potatoes), beverages (such as fruit juice), and fresh fruit (14, 15). Cross-sectional trends with FEV₁ in line with fresh fruit but of smaller magnitude were present for fruit juice and green vegetables, more so at HALS1. In our analysis looking at change, green vegetables showed a similar but weaker trend than fresh fruit; however, there were no relationships found for changes in intake of potatoes, salads, root vegetables, or fruit juice. It may be that reported intake of fruit and vegetables is more sensitive to individual differences than these other measures (16), especially when changes in their intake over time are considered.

Other nutrients may be involved in explaining this effect besides vitamin C because fruit intake is correlated with vitamin A, vitamin E, carotene, and magnesium (1, 4, 17). A recent study of subjects 70 yr and older found a relationship between FEV₁ and vitamin E (but not vitamin C); however, fruit intake was not correlated with lung function (18).

In conclusion, we have shown that changes in the level of fresh fruit consumption over 7 yr are positively associated with changes in FEV₁, irrespective of average level of intake, and this effect on lung function is of sufficient magnitude to explain cross-sectional effects seen in both children and adults.