The association between diet and chronic obstructive pulmonary disease in subjects selected from general practice

Sample size and power calculations

For 80% power and 95% confidence using Nquery (Statisical Solutions Ltd, Cork, Ireland), assuming one third of controls had a high intake of antioxidant rich food that this is protective with an odds ratio (OR) of 0.4, 115 cases and 115 controls were required in an unmatched design. This figure was doubled to allow for 50% nonresponse. This figure was then inflated by a further 50% to allow for low uptake up of spirometry testing.

Participants

Figure 1⇓ explains the recruitment process. Subjects with and without a general practioner's (GP) diagnosis of COPD, aged >45 yrs and with >10 pack-yrs of smoking were randomly selected from eight practice lists. Spirometry was used to confirm status as either a case or control using the British Thoracic Society Guidelines 2. Cases were defined as subjects with an FEV₁ ≤80% predicted and an FEV₁/forced vital capacity (FVC) ≤70% pred. Controls were those with an FEV₁ >80% pred and an FEV₁/FVC ratio >70% pred. Exclusion criteria included all subjects with a history of ischaemic heart disease, diabetes or gastrointestinal disease, where subjects may have been advised to modify their diets. Permission was obtained from the local research ethics committees and subjects gave their informed consent.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.—

Flow diagram of the recruitment process.

Instruments

After exclusions, the remaining subjects were mailed the HEA3 dietary food frequency questionnaire, validated for use in general practice 18. This questionnaire asks the frequency (day, week or month) and portion size (small, medium, large) of intake of foods belonging to the six main food groups: fruit and vegetables (including juice), meats and proteins, dairy produce, starchy foods, snack items and fats. Patients were asked to assess their food intake over the proceeding year. The questionnaire, self-administered at home, has proven to be an accurate measure of dietary intake when compared to a 7‐day weighed record and does not overestimate intake of food consumption. It is, however, limited to food groups as opposed to including details on individual food items, such as apples, oranges, etc. Therefore, subjects were also asked to complete the EPIC dietary questionnaire 19, which provides more detailed information regarding specific foods but can over estimate consumption. Results presented are from the HEA3 questionnaire unless otherwise stated. These questionnaires allowed ranking of intake of the major food groups and of foods of primary interest, i.e. those containing antioxidants.

Subjects were also sent the American Thoracic Society (ATS) 1978 respiratory questionnaire validated for self-completion at home 20. This questionnaire measures variables associated with smoking history, symptomatic prevalence and sociodemographical variables, such as employment history, education, and residency. Two reminders were sent to nonresponders.

Spirometry

M. Dorward performed all spirometric tests to assess FEV₁ and FVC, using a hand held Vitalograph 2120 spirometer (Vitalograph, Maids Moreton, UK). All subjects were asked to refrain from taking any inhaled or oral respiratory medication for 4–12 h prior to the test. Each subject was asked to perform three forced expiratory manoeuvres and the maximum value of FEV₁ and FVC recorded was used in these analyses. Subjects had their standing height measured using a portable stadiometer and were weighed on portable calibrated scales. Patients whose FEV₁/FVC were <70% and FEV₁ was <80% and who had symptoms associated with COPD according to the ATS 1978 questionnaire 20, such as cough and phlegm, but who were not originally recorded as having COPD by their GP, were recategorised as COPD patients for the purpose of analysis. It was thought that these patients were likely to have COPD but had failed to attend their GP for diagnosis. Thirty-four such incident cases were found.

Statistical analysis

Food variables were divided into equal thirds and then multiple stepwise logistic regression analysis for an unmatched case-controlled study was performed comparing each category with the reference category, i.e. the lowest third of intake. The model was adjusted for body mass index (BMI); some cases of COPD lose body weight due to lower calorie intake, and thus controlling for BMI controlled for the effect of COPD on total calorie intake. Known risk factors for COPD, such as sex, age and social class, were also tested in the model and retained if significance was at 5%.

Selection and response bias

The authors have no reason to believe that important bias due to selection and response occurred. Information collected on nonresponders indicated that although responders were more likely to be male and slightly older, there was no evidence of a differentiated response bias between those with and without COPD. The slightly larger number of male cases doubtless reflects the current male dominance of the disease, due to higher smoking rates among males in the past, although in this analysis, sex did not predict COPD status. Previous literature does not support sex and dietary interactions in association with COPD.

Recall bias

Recall bias in the completion of dietary questionnaires is a recognised problem when collecting nutritional data 22. Social desirability may have played a part in the responses of subjects and those foods considered healthy, e.g. fruit, may have been over-reported. However, HEA3 has previously been shown to accurately assess dietary intake of such foods against a 7‐day weighed record 18. Despite any possible over-reporting, there is no reason why cases should report diet differently to controls and subjects who had a diagnosis likely to be associated with dietary change, such as ischaemic heart disease or diabetes, were excluded. If the effect observed was due to differentiated social desirability between cases and controls, then a consistent effect would be seen for foods considered less healthy, e.g. dairy products, fats and snacks, which was not the case. Similarly, a differential affect on appetite or overall calorie intake would be expected to affect all food groups. Under-reporting of all foods is particularly common in the overweight 22, and weight reduction due to reduced calorie intake is a well-recognised feature of some cases, leading to reduced BMI. However, controlling for BMI did not significantly alter the estimates, i.e. the adjusted and unadjusted estimates were very similar. Finally, the estimates from incident cases for the association between fruit and vegetables and COPD were very similar to the prevalent cases, i.e. not explained by changes in diet subsequent to diagnosis. Therefore, recall bias and changes in diet due to having the disease are unlikely explanations of the results.

Sample size and study design

The sample size was powered to detect an OR of 0.4 but a lesser protective effect from other dietary components in a large population cannot be ruled out.

Due to the cross-sectional design of this study, exposure to diet has been established retrospectively. Prospective cohorts would be required to confirm these findings.

Risk factors for chronic obstructive pulmonary disease

This study shows that a dietary intake of vegetables (≥93 g·day⁻¹) and fruit (≥121 g·day⁻¹) is inversely associated with COPD. To the authors' knowledge, no other study has attempted to measure the differences in diet between subjects with a defined smoking pack history who have developed COPD and subjects with the same exposure to cigarette smoke who have not developed the disease. Miedema et al. 9 found fruit and vegetables to be protective against incidence of lung obstruction (defined as episodes of symptoms, such as regular coughing and phlegm for 3 months or a diagnosis by a clinical specialist) in a population of both smokers and nonsmokers, although spirometry was not used to ensure correct diagnosis. In the Seven Countries Study, Tabak et al. 5 found fruit and vegetables to be positively related to pulmonary function as well as protecting against mortality from COPD in a random population 7. However, the instruments used for dietary data collection differed between the participating countries, making accurate estimation of intake difficult. A number of observations made by other groups assessing lung function in relation to fruit intake, all indicate that high fruit intake leads to an increase in FEV₁ 4, 6, 9, 10. These results show that consuming apples (≥3·week⁻¹) was inversely associated with COPD in univariate analysis (OR 0.47, 95% CI 0.24–0.95), although not in multivariate analysis, which corroborates with the recent findings of Butland et al. 4, which indicated that those who ate ≥5 apples·week⁻¹ had higher FEV₁. Miedema et al. 9 postulated that it could be the antioxidant effect of the flavonoid quercetin, found in high concentrations in apples, onions and red wine, that protects against lung tissue injury. However, red wine and onions did not show any association with COPD in this study. No other individual fruits showed an independent association with COPD, including citrus fruits, which have high levels of vitamin C, previously suggested as a protective antioxidant in association with lung function 23–25. The findings in this study suggest that it is more likely to be the combined effect of various fruits and vegetables that protect against lung obstruction.

One previous study found that vitamin A from milk was protective against COPD, but this was not seen in this study 15. Shahar et al. 11 found a protective effect for fish in relation to lung obstruction but this was not confirmed in this study. No other food groups were associated with COPD, so it would appear that fruit and vegetable consumption is the main dietary factor.

Nondietary risk factors for chronic obstructive pulmonary disease

Age is shown to be a significant risk factor for COPD and is known to be one of the principle risk factors for developing the disease 2. Subjects >70 yrs were 13-times more likely to develop COPD in this population of smokers. Pack-yrs of smoking did not remain a significant risk factor when allowing for age and BMI in this model, as both cases and controls were selected based on their cigarette-smoking history and age also contributes considerably to pack years. BMI was significantly associated with COPD. This may be due, in part, to the reduction in calorie intake for some patients with the disease. However, the inverse association between fruit and vegetables and COPD cannot be explained by this since adjusting for BMI did not alter the estimates. Reduced total calorie intake and malnutrition is common in subjects with severe COPD 2 and the lack of significance of the underweight category when compared to the reference group can be explained by the small number of subjects.

Despite antismoking campaigns, many subjects continue to smoke and as they age, greatly increase their risk of developing COPD. If, as it appears here, fruit and vegetables can provide a protective mechanism against the oxidants contained within cigarette smoke, a clear health promotion message would be to encourage all subjects, but particularly smokers, to increase their level of fruit and vegetable intake to at least ≥1 portions·day⁻¹. However, this hypothesis ultimately needs to be tested in a large-scale intervention study.

Conclusion

This study indicates that low vegetable intake is likely to be an important risk factor for chronic obstructive pulmonary disease in subjects with >10 pack-yrs of smoking. Higher fruit consumption is also inversely associated with chronic obstructive pulmonary disease. These findings require confirmation in prospective cohorts but suggest that health promotion regarding dietary intake, particularly of fruit and vegetables, should be aimed at those most at risk of developing chronic obstructive pulmonary disease, namely smokers and exsmokers.