Low BMI and weight loss aggravate COPD mortality in men, findings from a large prospective cohort: the JACC study

To clarify how low BMI and weight loss were associated with risk of chronic obstructive pulmonary disease (COPD) mortality, in a large prospective cohort of the general population across Japan, the Japan Collaborative Cohort Study, conducted between 1988 and 2009. A total of 45,837 male residents were observed for a median period of 19.1 years. Self-administered questionnaires, collecting information on BMI, weight loss since the age of 20, lifestyles, history of diseases, as well as records of COPD mortality, were analysed at 2019. During follow-up, 268 participants died from COPD. The multivariate-adjusted hazard ratio (95% confidence interval) of COPD mortality associated with a 1-SD increment of body mass index (BMI) was 0.48 (0.41–0.57), while for weight change from age of 20 (+ 2.0 kg) it was 0.63 (0.59–0.68). These associations were persistently observed after stratifications with smoking status, excluding those having airway symptoms in the baseline survey, and excluding early COPD deaths within 5, 10 and 15 years. Our study suggests that BMI and weight change since the age of 20 could be markers for COPD prognosis, indicated by risk of COPD mortality.


Results
shows participants' age-adjusted baseline characteristics according to BMI. Compared to the control group whose BMI was 20.0 to < 22.0 kg/m 2 , those with BMI < 18.5 kg/m 2 were more likely to be older and current smokers, to walk 1 h or more per week, and to have histories of diabetes. Those with BMI ≥ 22.0 kg/m 2 were likely to be younger and current drinkers, to walk 1 h or more per week, to have higher education, and histories of diabetes and hypertension, and less likely to be current smokers. Over the 741,706 person-years of follow-up (median follow-up period was 19.1 years) for 45,837 male subjects, we found that a total of 268 participants died from COPD. Table 2 represents age-adjusted and multivariate-adjusted HRs (95% CI) of mortality from COPD according to BMI categories. The multivariate-adjusted excess risk of COPD death was 3.2 times higher among subjects in the lowest BMI category (BMI < 18.5 kg/m 2 ), compared to those with a BMI between 20.0 and < 22.0 kg/m 2 . www.nature.com/scientificreports/ There was a 42% reduction in risk for subjects in the highest BMI category (BMI ≥ 22.0 kg/m 2 ). Further, the multivariate HR and 95% CI for COPD mortality associated with a 1-SD decrement of BMI (2.80 kg/m 2 ) was 0.48 (0.41 to 0.57). The inverse associations were not attenuated after adjusting for potential confounding variables. In addition, when COPD deaths within 5, 10, and 15 years of the baseline survey were excluded, the inverse association remained statistically significant (Table 2) and the HR trends were similar (Fig. 1). This inverse association was also detected regardless of expectoration reported in the baseline survey (Supplementary Table S1 online). Of 268 deaths, 6.3% were of never smokers. When COPD deaths among smokers were analysed the inverse association between lower BMI and greater HR remained significant (Supplementary Table S2 online). Table 3 shows participants' demographic characteristics according to changes in weight from age 20. Compared to the control group whose weight change was > − 5.0, < 5.0 kg, those who lost ≥ 10.0 kg were more likely to be older, current smokers, to do exercise more, and to have lower mental stress and histories of diabetes and hypertension, and less likely to have higher education. On the other hand, those who gain ≥ 5.0 kg were more likely to be younger, never smokers, to walk 1 h or more per week, to have high mental stress, higher education, and history of hypertension. Table 4 shows the HR for COPD death according to weight changes from the age of 20. The multivariateadjusted HR in the first weight change category (participants who had lost 10 kg or more) was 3.3 times greater than in the reference category (participants who had lost or gained less than 5 kg). The age-adjusted HR in the fourth category (participants who had gained 5 kg or more) was 47% lower than that of the reference category (Table 4). After adjusting for confounding variables, the multivariate-adjusted HR and 95% CI of COPD mortality associated with a 1-SD decrement of weight change was 0.63 (0.59-0.68).
There were significant inverse trends in COPD mortality associated with weight changes. These inverse associations remained unchanged after the exclusion of participants who died within 5, 10 and 15 years of the baseline survey (  www.nature.com/scientificreports/       Table S6 online). Figure 2 presents the association of COPD mortality-represented by age-adjusted HRs-with both BMI and weight change categories, referenced to those with BMI between 20.0 and < 22.0 kg/m 2 and weight change between > − 5.0 and < 5.0 kg. The greatest risk of COPD mortality was found among participants in the lowest BMI and greatest weight loss groups (BMI < 18.5 kg/m 2 and weight loss ≥ 10 kg). The HR (95% CI) for these participants was 11.2 (6.33-19.7). Participants in the highest BMI and largest weight gain groups (BMI ≥ 22.0 and weight change ≥ 5) had the lowest risk for COPD mortality. The HR (95% C.I.) for this category of participant was 0.28 (0.12-0.63). These results shown in Fig. 2 suggest that weight change from age of 20 and BMI values have an additive and inverse effect on COPD mortality. Of note, even the participants with BMI > 22.0 kg/m 2 and weight loss ≥ 10.0 kg had significantly greater HR compared to the reference.

Discussion
In the present large-scale prospective community-based cohort study, lower BMI values and greater weight loss were independently associated with a greater risk of COPD mortality. The associations were observed even after exclusion of 5-, 10-, and 15-year early mortality incidents and participants with expectoration symptoms. Furthermore, by analysing COPD mortalities together with a non-COPD population (after excluding those with a history of tuberculosis), our study convincingly determined that low BMI and weight loss are associated with the increased risk of COPD mortality.
The previous Chinese study implied that lower BMI was associated with COPD mortality partially due to the comorbid chronic tuberculous infection 11 . However, our present analyses found an inverse association between BMI and COPD mortality regardless of a history of tuberculosis, suggesting that tuberculosis per se may not be responsible for the association. Nonetheless, the association between lower BMI and higher mortality from COPD even after exclusion of first 5, 10 and 15 years in the present study may suggest that being underweight may have a direct effect on the manifestation of COPD mortality. Furthermore, the inverse associations between BMI and COPD mortality did not differ by smoking status in the present study, although smoking may lower body weight 14 and increase risk of COPD 15 . This suggests that smoking is unlikely to be an intermediate factor for the presented association.
There could be several possible explanations for the risk elevation of COPD mortality among underweight men. Firstly, underweight men may be less resistant to infections, such as pneumonia 16 , and pneumonia is known to initiate fatal exacerbations of COPD. In addition, underweight COPD patients are more likely to suffer from other comorbidities, such as cardiovascular diseases, osteoporosis, depression, and lung cancer 15,17,18 , in association with a worse prognosis 19 . Secondly, underweight patients with anorexia nervosa showed emphysematous changes , a typical finding of COPD, in CT images 20 , as well as an old description of the residents with restricted intake of protein and calorie by physicians in the Warsaw ghetto 21 , suggesting that being underweight or having poor nutrition may develop COPD. Finally, the association between low BMI and COPD mortality may be explained by the accelerating effect that chronic cigarette smoke exposure has on the lung ageing process [22][23][24] . Both COPD patients (mostly smokers) and the elderly have been shown to suffer from low body weight due to reduced free fat mass 5,25-28 . A reduction in Sirtuin-1 activity and cellular senescence has also been associated with the development of COPD and cigarette smoke exposure [22][23][24] . Thus, chronic smoke exposure artificially induces ageing processes that result in both low body weight and lung ageing, thus potentially leading to the strong association between having low BMI and higher COPD mortalities, as observed in the present study. www.nature.com/scientificreports/ A recent analysis of the ECLIPSE cohort comprising COPD patients with GOLD stages 2 to 4 showed that COPD mortality was predicted with the presence of cachexia, which is defined as rapid weight loss > 5% or > 2% if BMI < 20 kg/m 2 within 12 months 29 . This study raises two issues that require discussion. Firstly, the association between weight reduction and greater COPD mortality in our study may be accounted for by cachexia. Unfortunately, it was not assessed whether or not the present study included the participants with cachexia, who developed weight loss rapidly within a year 30 , however weight reduction could occur in the early stages of COPD, since the association remained statistically significant, even after excluding COPD death in the initial 5, 10 and 15 years of our study ( Table 4), suggesting that weight reduction is risk for COPD mortality.
Secondly, the definition of cachexia in the ECLIPSE study 29 suggests that a weight reduction > 2% in the patients with lower BMI has the same impact on mortality as that > 5% in those with normal BMI. However, our study showed that HRs with 1-SD decrement of weight change were not so different between the four BMI categories (Supplementary Table S6 online), and therefore the impact of percentage of weight change to BMI on COPD mortality is independent from the four BMI categories. This implies that the weight reduction since the age of 20 in the present study may not due to cachexia.

Political implication.
There are remarkable disparities between Asian nations in the proportion of different types of respiratory diseases and mortality, including COPD, tuberculosis, asthma and pneumonia for total respiratory mortality [11][12][13] . A large cohort study of 220,000 Chinese men aged 40 to 79 year with 15 years followup showed that 5 kg/m 2 decrement of BMI was associated with 31% higher risk of mortality from COPD 11 . This Chinese cohort study had the smaller excess risk of COPD mortality than did our present study, which estimated 60% increase in COPD mortality per 2.8 kg/m 2 decrement of BMI, equivalent to 107% increase per 5.0 kg/m 2 decrement.
A study described the harmful effects of cigarette smoking, and found that it was the largest cause of noncommunicable disease (NCD) mortality-deaths caused by cardiovascular and respiratory diseases, and cancer 31 . Cigarette smoking accounted for 15% of total NCD deaths in Japanese men, followed by high blood pressure 31 . Confirming this, our study found that the majority of COPD mortalities occurred in smokers. Weight reduction and smoking cessation have long been the mainstays in efforts to prevent NCD. The present study did not collect information regarding whether the weight loss resulted from voluntary activity or involuntarily from disease processes. There are several voluntary means of weight reduction, including diet 32 , exercise 33 and behavioural modification 34 , whereas involuntary processes in COPD patients include imbalance in energy, impaired oxygenation and appetite loss-linked reduced dietary intake 35,36 , as well as cachexia 37 and muscle wasting (sarcopenia) 38 . These different causes of weight change in COPD patients can indirectly affect the association between the weight change and mortality from COPD. Nevertheless, the present study may provoke a possible dilemma as voluntary weight loss may reduce mortality due to cardiovascular diseases, but increase COPD mortality (i.e. the obesity paradox) 39 .

Limitations.
There are several limitations in this study; first, weight change was defined as the difference between body weight at the baseline survey and that at the age of 20. Because the latter value depends on participants' memory, the possible bias was present in reported values. Specifically, thin men may have overestimated their past weight, while overweight men may have underestimated it 40 . However, the recalled and measured data were well correlated 40,41 . Moreover, our previous study 16 had confirmed the results of other studies in terms of the association between weight loss and mortality from all causes 12,42,43 . Second, our study only included men, which might restrict the generalizability of the results to women or individuals of other ethnicities. On the other hand, the prevalence of asthma and tuberculosis in Japan is not as high as in other Asian countries. Thus, our study could provide comparative evidence to that carried out in Western countries, where tuberculosis is not a leading cause of death. Third, there is a potential for reverse causation between having low BMI, weight loss, and higher COPD mortality risk. However, the possible reverse causation may be small in the present study, since these inverse associations remained unchanged after the exclusion of mortality events occurring within the first 5, 10 and 15 years. Finally, we did not make any lung function assessment of the participants, so we are not able to discuss the severity of COPD. However, analyses with stratification by smoking exposure (≥ 20 pack year vs < 20) at enrollment showed that the inverse association between BMI and mortality from COPD was observed independently from smoking. This indirectly suggested that lower BMI may predict worse mortality regardless of any reduction in lung function, since COPD patients with more impaired lung fucntion are more likely to belong to the group with a higher smoking index .
Our study presents evidence of an inverse and additive association between both weight change from the age of 20 and BMI, and COPD mortality (Fig. 2). Low BMI and weight loss may be noticeable long before COPD is diagnosed, or COPD-linked symptoms are observed. We may thus be able to advocate for low body weight and weight loss as possible predictive markers of COPD mortality, even if they might not be a pathogenic cause of disease development.
Strengths of the study. A major strength of our study was that the JACC study followed a large prospective well-characterized cohort of the general population for a long period 16,[44][45][46][47] . This enabled us to analyse the data using the combined four BMI, and four weight change categories after adjustment with many variables, as well as to conduct further stratified analyses by smoking status and initial symptoms.

Conclusion
In conclusion, our large-scale prospective community-based epidemiological study showed that lower BMI and weight loss are potential independent disease-specific markers for COPD mortality. The individuals who have lost more than 10 kg since the age of 20, and those with a BMI < 18.5 kg/m 2 had highest risk for mortality from COPD.

Methods
Study population and mortality surveillance. The study population was drawn from participants in the Japan Collaborative Cohort Study for Evaluation of Cancer Risk (JACC) study 44 , which was sponsored by Monbu-kagaku-sho (the Ministry of Education, Science, Sports and Culture, Japan). This study's baseline survey and participant follow-up methodology has been described in detail elsewhere 44 . Briefly, a comparative study was conducted from 1988 to 1990 with 110,585 individuals (46,395 men and 64,190 women) aged 40 to 79 years who were living in 45 communities throughout Japan. Participants completed self-administered questionnaires that asked about their current weight and height, their weight at the age of twenty, their lifestyles, and any medical history of cardiovascular disease (CVD) and cancer 44 and tuberculosis. Informed consent was obtained from participants before they completed the questionnaire or sometimes from community leaders instead of individuals, as this had been a common practice for informed consent at that time in Japan.
Since mortality due to COPD was rather rare among the female participants, we focused on male JACC study participants. We excluded 8094 male participants who had a history of lethal diseases including stroke, myocardial infarction, and cancer, as well as tuberculosis, leaving a total of 45,837 male subjects enrolled in the study. The analyses were conducted at 2019.
Of 45 municipalities, participants were followed until the end of 2009 to determine COPD-caused mortality in 35 areas. Of remaining 10 areas, four areas terminated follow-up in 1999, four in 2003, and two in 2008. Mortality data were centralized at the Ministry of Health and Welfare, and the underlying cause of death was coded according to the International Classification of Diseases (ICD), 9th version, for deaths between 1988 and 1994, and the 10th revision for deaths from 1995 and thereafter.
Registration of death is required by the Family Registration Law of Japan, and thus all deaths in the cohort were ascertained with death certificates, under the permission of the Director General of the Prime Minister's Office (Ministry of Internal Affairs and Communications). The only exceptions were subjects who died after they moved away from their original community. These were treated as censored cases. In this study, COPD mortality is defined as that from emphysema (ICD10 J43) and from other diseases or conditions in association with COPD (ICD10 J44). The relevant outcome was defined as death attributed to COPD during the study period. For each participant, the follow up period (person-year) was the period between completing the baseline survey and death, emigration, or the end of the study, i.e., the end of 1999, 2003, 2008, or 2009.
The study population was drawn from participants in the Japan Collaborative Cohort Study for Evaluation of Cancer Risk (JACC) study 48 and COPD mortality is defined as that from emphysema (ICD10 J43) and from other diseases or conditions in association with COPD (ICD10 J44) in death certificates.
Statistical analysis. Participants were divided into four categories according to their BMI (kg/m 2 ): (1) BMI less than 18.5, (2) BMI from 18.5 to < 20.0, (3) BMI from 20.0 to < 22.0, and (4) BMI of 22.0 or above, based on a previously published study 47 , so that tthese groupings presented a similar number of COPD deaths across BMI categories (Table 1). Category 3 (20.0 to < 22.0) was treated as the reference group. The analyses of weight change from the age of 20 were based on our previous studies 45,49 . Briefly, weight changes were defined as the difference between body weight at the baseline survey and that at the age of 20, and the participants were divided into four groups as in our previous studies 45,49 : (1) participants who had lost 10 kg or more (≤ − 10), (2) those who had lost between 5 to < 10 kg (> − 10, ≤ − 5), (3) those who had either lost or gained less than 5 kg (> − 5, < + 5) (reference), and (4) those who had gained 5 kg or more (≥ + 5 ).
Means with standard deviations (SD) and proportions of selected COPD risk factors were determined, according to categories of BMI, as well as to weight change. To test the difference between these categories, t-test and chi square test were performed. Age-adjusted and multivariate adjusted hazard ratios (HRs) and their 95% confidence intervals (95% CI), as well as 1-standard deviation (SD) increments of BMI or weight change, were determined using Cox proportional hazard models. Then, participants were further divided into 16 groups based on the combination of BMI and weight change categories, and each HR (95% CI) was determined and referenced to the BMI (20.0 to < 22.0) and weight change (> − 5, < + 5) groups. The crude death ratio (/100,000) and age-adjusted HR using the Cox proportional hazard model were also determined.
Potential variables in the multivariate adjusted model included age, BMI, weight change from age of 20, ethanol intake, hours of walking per week, hours of exercise per week, education level, perceived mental stress, history of hypertension, and history of diabetes (all values as recorded in the baseline survey). We conducted the stratified analyses by smoking status and the presence or absence of initial symptoms of expectoration, as well as by BMI categories (< 18.5, ≥ 18.5 to < 20.0, ≥ 20.0 to < 22.0, ≥ 22.0 kg/m 2 ) at base line survey. Early deaths caused by COPD-those that took place within the first 5, 10, and 15 years of enrolment-and participants with self-reported expectoration at the baseline survey were excluded in order to reduce the chances of a reverse causal relationship between BMI or weight reduction, and COPD mortality.
All analyses were performed using SAS version 9.3 (SAS Institute, Cary, North Carolina, USA). Two-tailed probability p-values lower than 0.05 were considered statistically significant.
All the study was carried out in accordance with the ethics guideline set out by the Ministry of Health, Labour and Welfare, Japan, based on the fundamental principle in the Declaration of Helsinki 50 , and all of these processes were reviewed and approved by the Ethics Committees of the Hokkaido University, Osaka University