Chronic obstructive pulmonary disease (COPD) is a leading cause of morbidity and mortality worldwide [1], with cigarette smoking being established as the principle risk factor [2–4]. However, while 95% of COPD patients are, or have been, cigarette smokers, only 20% of smokers develop COPD [5]. Therefore, other factors such as dietary and environmental exposures may protect against, or contribute to, disease development. Because of the high burden and societal cost of COPD, studies on potential new methods of prevention through an appropriate diet are important. An extensive literature review suggested that fruits, vegetables, fish, and meat may be associated with the risk of COPD and respiratory symptoms [6].

High fruit intake is inversely related to the COPD risk [7–9] and respiratory symptoms [10–12]. A cohort study of 49,140 Singaporean men observed that non-citrus fruits (apples, pears and grapes) were independently associated with reduced cough with phlegm [12]. Similarly, a cross-sectional study among 13,651 adults aged 20–59 years in The Netherlands reported that solid fruits (apples and pears) were inversely associated with the prevalence of COPD symptoms such as chronic cough and breathlessness [10]. The same authors also noted the favourable effect of fruits (> 180 g/day) in lowering the prevalence of these respiratory symptoms [11]. Increased vegetable consumption is likely to reduce the risk of COPD [7, 8, 13]. However, two large studies conducted in The Netherlands and Singapore observed no relationship between vegetable intake and the prevalence of respiratory symptoms [11, 12].

One cross-sectional study from Norway involving 4,300 subjects found fish intake beneficial against developing a night cough but not other symptoms [14]. On the other hand, epidemiological evidence from several prospective studies indicated that red meat or cured meat consumption had detrimental effects on the risk of COPD [15–18]. In particular, preserved red meat was positively associated with respiratory symptoms [19] and could lead to a decline in lung function [17].

Little is known in relation to soy foods, apart from one cohort study in Singapore which suggested that soy foods may reduce the development of chronic productive cough [12], and conclusions cannot be drawn because of the limited evidence [6].

Soybean products, a primary source of isoflavones, are consumed worldwide. In Japan, a wide variety of soy foods are available including tofu (soybean curd), natto (fermented soybeans), miso (fermented soybean paste) soup, bean sprouts and soy milk. The aim of this study was to investigate the relationship between soy consumption and the risk of COPD and respiratory symptoms.

The overall prevalence of respiratory symptoms was: cough (n = 105, 17%); sputum (n = 193, 31%); breathlessness (n = 217, 35%), with significantly higher prevalence in the COPD group (p < 0.001), especially for breathlessness (66% in cases versus 12% in controls). The binary breathlessness outcome was validated against the Medical Research Council's dyspnoea score. The observed spearman rank correlation of 0.7 confirmed a good agreement between the two scales. Distribution of the dyspnoea scale was: none, 46.9%; Grade 1, 23%; Grade 2, 18.5%; Grade 3, 7.8%; Grade 4, 3.1%; Grade 5, 0.7%.

The relationship with lung function was next examined. [see Figure S1, Additional file 1] presents the scatter plot of FEV₁ % predicted against total soy intake. All observed lung function measures (FVC, FEV₁, FEV₁/FVC and FEV₁ % predicted) were positively correlated with total soy consumption. Linear regression results in [Table S1, Additional file 1] further confirmed the significant positive association between lung function and soy intake even after adjusting for the effects of confounding factors.

In this case-control study, the effects of soy bean products on lung function, COPD risk and the prevalence of respiratory symptoms were documented. All participants were carefully screened for respiratory symptoms and spirometric measurements of lung function were taken to ensure correct classification. Another strength of the study was the inclusion of mainly patients (80%) with COPD diagnosed within the past two years and all 278 patients were interviewed within four years of confirmed COPD diagnosis, which provided better accuracy in capturing their dietary exposure information. Moreover, the total intake of soy foods by our community-dwelling controls (mean 59.98 g/day, SD 50.23) was comparable with the average consumption level of 60.4 g/day by the general Japanese population in 2006 [27].

Soy consumption was found to be positively correlated with lung function and inversely associated with the risk of COPD. The epidemiological evidence also indicated an inverse association between total soy intake and breathlessness. Our new findings were consistent with a previous cohort study in Singapore that reported soy foods may reduce the development of chronic respiratory symptoms, especially productive cough [12]. It has been suggested that flavonoids from soy foods act as an anti-inflammatory agent in the lung, and can protect against tobacco carcinogens for smokers [12]. However, further research is needed to understand the underlying biological mechanism.

Several limitations should be considered in conjunction with the findings. Habitual dietary assessment was based on self-report using a validated and reliable questionnaire. Nevertheless, responses from the older adults inevitably incurred some recall error due to possible memory and cognitive loss. Therefore, face-to-face interviews were conducted in the presence of patients' next-of-kin to increase the response rate and to improve the accuracy of their answers. In addition, all interviews were conducted by the same investigator (first author) to eliminate inter-interviewer bias. Although the control subjects were recruited during the same period and from the same domain as the prevalent cases and should be representative of the elderly population from which the cases arose, the existence of selection bias could not be ruled out because of their voluntary participation in the study. All participants were blinded to the study hypothesis while the protective effects of soy foods were not yet established for COPD. Reverse causation should not pose a serious threat to the study validity as demonstrated by our sensitivity analysis of the subgroup of patients with COPD diagnosed within one year. Nevertheless, there was still the possibility of residual confounding even though established confounding factors were controlled for in the multivariate analyses.

Another limitation was the exclusion of miso soup and soy sauce towards the calculation of total soy intake. The concentration of miso soup varies substantially according to personal taste, whereas soy sauce is mainly added during cooking and thus difficult to quantify the exact amount consumed. Data were also lacking on less common soy foods such as yuba (soy milk skin) and okara (bean curd byproduct) from the food frequency questionnaire. Finally, evidence concerning the beneficial effects of soy products may be regarded as preliminary, in view of the relatively small number of female participants recruited into the study, even though the gender ratio was typical of the elderly COPD population in Japan.

Smoking is acknowledged as the major cause of COPD, but the present case-control study has suggested an inverse association between soy products and COPD risk for Japanese adults. More research and/or replications are required to ascertain whether the observed findings can be generalized to other populations, before incorporating these foods into dietary guidelines so as to encourage consumption. Besides experimental studies, long-term prospective cohort studies collecting detailed dietary exposure information are recommended to provide epidemiological evidence on both morbidity and mortality due to COPD.