Study design and subjects
A case-control study was conducted in central Japan in 2006. Three hundred COPD patients referred by respiratory physicians were recruited from the outpatient departments of six hospitals in Aichi, Gifu and Kyoto. According to the protocol of the Global Initiative for Chronic Obstructive Lung Disease (GOLD) [20], diagnosis of COPD was confirmed by spirometry with FEV1/FVC < 0.7, where FEV1 = forced expiratory volume in one second and FVC = forced vital capacity. Predicted FEV1 was calculated using the Japanese Respiratory Society's Guidelines [21].
Inclusion criteria for cases were: (i) age between 50 and 75 years; (ii) had COPD as the primary functionally limiting illness which was diagnosed within the past four years. Subjects were excluded if they had a recent stroke, dementia or other health conditions that prohibited them from being interviewed. Twenty-two eligible patients were subsequently excluded due to missing or incomplete demographic and lifestyle details. The remaining 278 patients (244 men and 34 women) were available for analysis. No statistically significant differences were found between the included and excluded prevalent cases in terms of mean age, BMI and gender distribution (p > 0.05). Permission to recruit patients and access to medical records were granted by the participating hospitals in Japan.
During the same period, 400 community-dwelling adults were recruited from the same catchment areas as the prevalent cases. These controls were approached and interviewed at shopping malls, community centres or when they attended health checks at hospitals. They were selected to be frequency matched to the prevalent cases by age (± 5 years). The same exclusion criteria as cases were applied, resulting in 340 eligible controls (272 men and 68 women). All participants underwent spirometric measurements of respiratory function to avoid misclassification of case-control status. Approval of the study protocol was obtained from the Human Research Ethics Committee of Curtin University (approval number HR 90/2005) and the six hospitals in Japan.
Interview and questionnaire
A structured questionnaire was administered face-to-face by the first author to collect information from each participant. Demographic and lifestyle characteristics solicited included age, gender, weight (kg), height (m), education level (high school or below; college or university), cigarette smoking (never smoker; ex-smoker; current smoker) and alcohol drinking status (non-drinker; drinker). For the prevalent cases, each interview was conducted in the presence of their next-of-kin to minimize recall error, and appointment was made via their respiratory physician. The purpose of the study was explained to each participant before obtaining their formal written consent. Confidentiality of the information provided, and the right to withdraw without prejudice, were ensured and maintained throughout the study. All interviews, averaging 30 minutes in duration, took place in the hospital outpatient departments for COPD patients and their place of recruitment for controls.
Information on habitual food consumption was obtained using a 138-item interviewer-administered food frequency questionnaire taken from the Japan Public Health Center-based prospective study on cancer and cardiovascular disease [22]. Its validity and reproducibility had been established for the Japanese population [23]. The reference recall period for dietary variables was set at 5 years before interview. Therefore, the recall period was shorter among cases (before their diagnoses of COPD) and different from case to case depending on their date of diagnosis. Soy food consumption encompassed tofu (boiled or cold, in miso soup, freeze-dried, deep-fried), natto, bean sprouts, and soy milk which are soybean products. The frequency of intake of tofu, natto and bean sprouts was classified by nine categories: 'almost never', 'once to three times per month', 'once to twice per week', 'three to four times per week', 'five to six times per week', 'once per day', 'twice to three times per day', 'four to six times per day', and '7 or more times per day'. Those participants who answered "monthly" to "daily" were asked their usual amount consumed per meal in terms of standard portion size (tofu, 75 g; tofu in soup, 25 g; dried tofu, 60 g; deep-fried tofu, 2 g; natto, 75 g; bean sprouts, 25 g). The frequency of soy milk drinking was categorized into 'almost never', '1 to 2 times per week', '3 to 4 times per week', '5 to 6 times per week', 'one cup per day', '2 to 3 cups per day', '7 to 9 cups per day', and '10 or more cups per day.'
A further question on 'life-long physical activity involvement' was appended to the questionnaire, defined as "doing active sports or vigorous exercise long enough to get sweaty, at least twice a week", over the entire life course [24]. Response options were dichotomous: 'has never been involved to not any more involved in such activity' and 'has always been involved in such activity'.
Two screening instruments, Medical Research Council's "dyspnoea" scale [25] and the Australian Lung Foundation's "Feeling Short of Breath" scale [26], were used to assess respiratory symptoms of each individual. The latter scale consisted of five simple questions: (i) Do you cough several times most days? (ii) Do you bring up phlegm or mucous most days? (iii) Do you get out of breath more easily than others your age? (iv) Are you over 40 years old? (v) Are you smoker or ex-smoker?
Statistical analysis
Descriptive statistics were first applied to summarise participant characteristics and lung function measures. The daily intake of soy products (g) was derived from the frequency and quantity recorded, accounting for the edible portion of each food. After comparing the dietary consumption pattern between case and control groups, multivariate (unconditional) logistic regression analyses were performed to assess the effects of soy foods on the COPD risk and the prevalence of respiratory symptoms. Soy consumption variables were further categorised based on the corresponding empirical distribution of controls, with the lowest intake or none being the reference category. Besides soy foods consumption, independent variables included in the regression models were age, gender, body mass index (BMI = weight/height2) of five years ago, education level, life-long physical activity involvement, smoking status, smoking pack-years, alcohol drinking status and daily dietary intake of fruits, vegetables, fish, red meat and chicken. These variables were considered potential confounders from the literature. Adequacy of each fitted model was assessed by the Hosmer-Lemeshow statistic as well as area under the ROC curve. All statistical analyses were undertaken using the SPSS for Windows package version 13.
