Abstract
The incidence and prevalence of nontuberculous mycobacterial pulmonary disease (NTM-PD) have been increasing worldwide. The risk of NTM-PD may be higher in undernourished populations. In this study, we tried to elucidate the impact of body mass index (BMI) and its change on development of NTM-PD.
We performed a retrospective cohort study including South Koreans aged >40 years who underwent biennial National Health Insurance System (NHIS) health check-ups in both 2005 and 2009 or 2006 and 2010. We monitored eligible individuals from the study initiation date (NHIS health check-up date in 2009 or 2010) until the diagnosis of NTM-PD or until December 31, 2017. Enroled individuals were classified based on BMI at initiation date. By calculating hazard ratios, we compared NTM-PD incidence per 100 000 person-years by BMI group and by BMI change.
A total of 5 670 229 individuals were included in the final analysis. Compared with the BMI <18.5 kg·m−2 group, the incidence of NTM-PD gradually decreased with increased BMI (adjusted hazard ratio 0.38, 95% confidence interval (CI) 0.35–0.42 for BMI 18.5–22.9; 0.17, 95% CI 0.15–0.19 for BMI 23–24.9; 0.1, 95% CI 0.09–0.11 for BMI 25–29.9; and 0.1, 95% CI 0.07–0.13 for BMI ≥30). A BMI decrease of ≥1 kg·m−2 over 4 years increased the incidence of NTM-PD (adjusted hazard ratio 1.08, 95% CI 1.01–1.16) whereas a BMI increase of ≥1 kg·m−2 decreased the incidence of NTM-PD (adjusted hazard ratio 0.77, 95% CI 0.71–0.83).
In conclusion, BMI was inversely related to development of NTM-PD and weight loss increased the risk of NTM-PD.
Abstract
Higher body mass index was protective against development of nontuberculous mycobacterial pulmonary disease (NTM-PD) among Korean people. Weight loss increased the risk of NTM-PD. https://bit.ly/2PDawxn
Introduction
Nontuberculous mycobacteria (NTM) are ubiquitous microorganisms that have an environmental origin and can affect various human organs. Typically, NTM infection causes progressive inflammatory damage to the lung, resulting in nontuberculous mycobacterial pulmonary disease (NTM-PD). The incidence and prevalence of NTM-PD are increasing worldwide [1]. The reasons for this increase in prevalence are unclear, but may include increased recognition and testing, increased exposure to NTM from home water heaters or shower aerosols, a decrease in population immunity to mycobacteria owing to decreasing tuberculosis (TB) infection, or an increased number of people with risk factors for NTM-PD [2, 3]. Risk factors proposed for the development of NTM-PD encompass malignant diseases of the thorax [4], chronic lung disease [5], thoracic skeletal abnormalities [6], rheumatic diseases [7] and medications including immunosuppressants [8] and inhaled corticosteroids (ICS) [5, 9].
Several nutritional indicators, such as subcutaneous fat [10] and triceps skinfold thickness [11], have also been reported to be associated with NTM-PD. Patients with NTM-PD tend to have lower body weight or lower body mass index (BMI) [11, 12]. In addition, poor nutritional status is correlated with the progression of NTM-PD [13] and higher all-cause mortality among patients with Mycobacterium avium complex pulmonary disease [14].
Although lower BMI is common among patients with NTM-PD and is associated with worse prognosis, it is unclear whether it is a risk factor for the development of the disease. The aim of this study was to elucidate, using a nationwide database, the association between BMI and the development of NTM-PD in the South Korean population.
Materials and methods
Data source
We used the database of the National Health Insurance Service (NHIS), which includes the entire population of South Korea (approximately 49.8 million people) as well as registered foreign-nationality residents. The NHIS is a compulsory insurance system, with exceptions for individuals who qualify for the National Medical Aid programme or foreign military personnel [15]. The NHIS recommends that insurance subscribers undergo a standardised health check-up biennially. The annual number of subscribers who participate in this examination is more than 13 million [16].
The NHIS database contains information on demographics and all medical services rendered (with diagnostic codes as per the International Statistical Classification of Diseases and Related Health Problems, 10th edition (ICD-10)), as well as all prescription medications dispensed. The database also contains information of subscribers’ residence and income. Residence is classified according to residence in one of six metropolitan cities, Seoul, or Sejong City versus other areas. A low income level is defined as the lowest quintile of the annual income level. The NHIS database has been used in previous studies [17, 18].
NHIS health check-up programme
The standardised biennial health check-up programme provided by the NHIS includes the following components: self-reporting questionnaires, physical measurements, physical examinations, chest radiographs and blood tests at officially designated centres. Standardised self-reporting questionnaires are used to collect data at the time of examination for the following variables: alcohol consumption (non-drinker, mild drinker with a mean consumption of <30 g·day−1 or heavy drinker with a mean consumption of ≥30 g·day−1), smoking status (never smoker, ex-smoker or current-smoker) by pack-year, comorbidities and physical activity. Individuals who met one of the following criteria were defined as physically active: ≥3 days of vigorous activity for ≥20 min·day−1 or ≥5 days of moderate intensity activity or walking for ≥30 min·day−1. During the physical examination, each individual's height and weight were measured and blood specimens were collected after an overnight fast of at least 8 h. Chest radiographs were interpreted by radiologists and the results were classified according to four categories: normal, inactive TB, active TB and other abnormalities.
Study design
We conducted a retrospective cohort study based on the NHIS database. Individuals aged >40 years who underwent a biennial NHIS health check-up in both 2005 and 2009 (“Cohort 2009”) or in 2006 and 2010 (“Cohort 2010”) were included in the analysis. Individuals given an ICD-10 code for NTM-PD between January 01, 2002 and the initiation date were excluded from the cohort. Enroled individuals were classified based on BMI (in kg·m−2) at the initiation date as follows: BMI <18.5 (underweight), ≥18.5 BMI <23 (normal weight), ≥23 BMI <25 (overweight), ≥25 BMI <30 (obese I) and BMI ≥30 (obese II–III). These five categories are based on the recommendations of the World Health Organization (WHO) for Asia-Pacific populations [19]. Eligible individuals were monitored from the study initiation date until diagnosis of NTM-PD or until December 31, 2017.
This study was approved by the official review committee of the NHIS (study number: NHIS-2018-1-460) and the institutional review board of Seoul National University Hospital (IRB number: E-1808-112-967).
Definitions
Development of NTM-PD
Within the cohort, we identified individuals with NTM-PD who were: 1) newly registered with ICD-10 code A31.0 after the initiation date; and 2) had at least two ambulatory visits or hospital admissions with an A31.0 diagnosis code during 1 year.
Change in body mass index
Change in BMI was defined as the difference in BMI over a 4-year period. We calculated the difference between BMI measured in 2005 and 2009 for Cohort 2009 and between that measured in 2006 and 2010 for Cohort 2010. Categorisation of the 4-year change in BMI was defined as stable (BMI change <1 kg·m−2), increased (≥1 kg·m−2) or decreased (≥1 kg·m−2).
Comorbidities
Comorbidities were defined based on ICD-10 codes registered within 1 year prior to the initiation date. Details are provided in supplementary table S1. Exceptionally, diabetes mellitus was defined using the following criteria: 1) at least one claim for prescription of an anti-diabetic agent under ICD-10 code E11-14 within 1 year prior to the study initiation date; or 2) a fasting glucose level of ≥7 mmol·L−1 (126 mg·dL−1).
Statistical analysis
Baseline characteristics of the study population, including demographics and comorbidities, are described using frequency (n (%)), median (interquartile range (IQR)) and mean±sd.
We calculated the incidence rate of NTM-PD per 100 000 person-years and compared amongst BMI groups. Candidates for confounding variables were selected based on risk factors for the development of NTM-PD reported in previous studies [8, 9], as well as factors that could affect the diagnosis of NTM-PD. They included age, sex, residence, income, diabetes, respiratory diseases, solid cancers, transplantation history, haematologic malignancy, HIV infection, gastroesophageal reflux disease (GERD), number of outpatient clinic visits or hospital admissions and chest radiograph findings. Cox proportional hazards models were constructed to obtain adjusted hazard ratios with 95% confidence intervals (CIs) after adjusting for variables with p<0.05 in univariable analyses, as well as biologically plausible risk factors. The analysis included the unadjusted model (Model 1) and adjusted for: 1) age and sex (Model 2); 2) all variables included in Model 2 plus residence, income, diabetes, respiratory disease, solid cancer, haematologic malignancy, transplantation, AIDS, GERD, smoking status and number of hospital visits (Model 3); and 3) all variables included in Model 3 plus four categories of chest radiograph findings during health screening (Model 4).
To elucidate the impact of changes in BMI on the incidence of NTM-PD, we calculated the incidence rate of NTM-PD per 100 000 person-years in the groups with stable, decreased or increased BMI, respectively. Cox proportional hazards models were constructed to obtain adjusted hazard ratios with 95% CIs in a stepwise pattern, as described above.
To exclude the possibility that development of NTM-PD caused a decrease in BMI, we performed a sensitivity analysis ignoring a diagnosis of NTM-PD for 1 year after the initiation date. For Cohort 2009, the incidence rates and hazard ratios were calculated based on a diagnosis of NTM-PD after 2010 while, for Cohort 2010, they were calculated based on a diagnosis of NTM-PD after 2011.
Statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA) and R version 3.2.3 (The R Foundation for Statistical Computing, Vienna, Austria; www.rproject.org). A two-sided p-value of less than 0.05 was considered to indicate statistical significance.
Results
We identified 4 101 614 individuals who had undergone a biennial health check-up in both 2005 and 2009 (Cohort 2009) and 5 450 550 individuals who received a health check-up in both 2006 and 2010 (Cohort 2010). After excluding individuals aged <40 years, those with an ICD-10 code for NTM-PD (A31.0) between January 01, 2002 and the initiation date, those with incomplete data and those duplicated in both cohorts, 5 670 229 individuals were included in the final analysis (figure 1).
Flow diagram of the study population. ICD-10: International Statistical Classification of Diseases and Related Health Problems, 10th edition; NTM-PD: nontuberculous mycobacterial pulmonary disease; NHIS: National Health Insurance Service; BMI: body mass index.
Demographic and clinical characteristics of individuals
The median age of the 5 670 229 individuals included in the analysis was 54 years (IQR 47–63 years) and 53.6% were male. Of the total, 19.3% were current smokers and 17.8% were ex-smokers, 667 664 (11.8%) had diabetes, 953 733 (16.8%) had respiratory comorbidities, 166 249 (2.94%) had connective tissue diseases and 118 535 (2.1%) had solid organ or haematologic malignancies. Chest radiographs were normal in 87.7% of individuals and active pulmonary TB was suspected in 0.25% of cases (table 1).
Demographic and clinical characteristics of individuals according to body mass index (BMI)
Based on BMI at the study initiation date, 121 481 individuals (2.1%) were classified as underweight (BMI <18.5), 2 035 239 (35.9%) were normal weight (≥18.5 BMI <23), 1 571 445 (27.7%) were overweight (≥23 BMI <25), 1 777 062 (31.3%) were class I obese (≥25 BMI <30) and 165 002 (2.9%) were class II–III obese (BMI ≥30). Median age was higher in the underweight group (56 years) than in the other groups. The distribution of BMI categories by age is provided in supplementary table S2. There were more male individuals in the class I obese group (56.2%) and the class II–III obese groups (58.5%). The prevalence of diabetes increased steadily with increased BMI. Chronic obstructive pulmonary disease (COPD), bronchiectasis and TB sequelae were most common in the underweight group. Inactive TB, active TB and other abnormalities on chest radiograph were also most common in the underweight group. In addition, there were more current smokers in the underweight group. The number of hospital visits increased with increased BMI (table 1).
Incidence of NTM-PD according to body mass index categories
The median duration of observation among individuals was 7.9 years (95% CI 7.4–8.4 years). During observation, NTM-PD was diagnosed in 5029 of 5 670 229 individuals (0.09%). The incidence rate of NTM-PD was highest amongst the underweight group (71.5 per 100 000 person-years) and decreased with increased BMI as follows: normal weight group (18.0 per 100 000 person-years), overweight group (7.4 per 100 000 person-years), class I obese group (4.2 per 100 000 person-years), class II–III obese group (4.0 per 100 000 person-years) (p<0.0001). The same trends were observed in both males and females (table 2).
Incidence of nontuberculous mycobacterial pulmonary disease (NTM-PD) according to body mass index (BMI)
The inverse relationship between incidence of NTM-PD and BMI was maintained after adjusting for potential confounding factors. In Model 4, when compared with the underweight group, adjusted hazard ratios for developing NTM-PD were as follows: normal weight group (0.38, 95% CI 0.35–0.42), overweight group (0.17, 95% CI 0.15–0.19), class I obese group (0.10, 95% CI 0.09–0.11), class II–III obese groups (0.10, 95% CI 0.07–0.13) (table 2). This inverse association was maintained when treating BMI as a continuous variable (supplementary table S3).
Development of NTM-PD according to changes in body mass index
The incidence of NTM-PD differed according to changes in BMI. The incidence rate of NTM-PD was 11 per 100 000 person-years in the group with stable BMI, 14 per 100 000 person-years in the group with decreased BMI and nine per 100 000 person-years in the group with increased BMI. In Model 4, compared with the stable BMI group, the adjusted hazard ratios for NTM-PD development were as follows: decreased BMI group (1.08, 95% CI 1.01–1.16), increased BMI group (0.77, 95% CI 0.71–0.83) (table 3).
Incidence of nontuberculous mycobacterial pulmonary disease (NTM-PD) with reference to a body mass index (BMI) change of ≥1 kg·m−2
Sensitivity analysis
The results of sensitivity analyses ignoring development of NTM-PD within 1 year following the study initiation date showed similar results. The higher incidence of NTM-PD in the decreased BMI group and lower incidence of NTM-PD in the increased BMI group were confirmed, although differences in some comparisons were absent after adjustment for potential confounding variables (table 4).
Sensitivity analysis for nontuberculous mycobacterial pulmonary disease (NTM-PD) incidence with reference to a body mass index (BMI) change of ≥1 kg·m−2, while ignoring a diagnosis of NTM-PD for 1 year after the initiation date
Discussion
In this study, based on the nationwide health insurance database of South Korea, we showed that higher BMI was protective against development of NTM-PD. In addition, the 4-year change in BMI was inversely correlated with the development of NTM-PD. These results suggest that lower BMI per se, as well as weight loss, could be risk factors for NTM-PD development.
The protective effect of higher BMI (or increased risk of lower BMI) in other infectious diseases, including TB, is well known. Epidemiologic studies have consistently reported the association between higher BMI and lower incidence of TB. One study of an older adult population in Hong Kong reported lower active TB incidence amongst obese and overweight individuals [20], while an analysis based on the US National Health and Nutrition Examination Survey found a similar result [21]. A meta-analysis of six population-based studies has confirmed the inverse log-linear relationship between BMI and TB incidence within a BMI range of 18.5–30 kg·m−2 [22]. This protective effect of obesity against TB is known as the “obesity paradox” [23] and has also been observed in other conditions, such as sepsis [24], pneumonia [25] and Chagas disease. At the same time, higher risk of community-acquired pneumonia [26], upper respiratory infection [27] and fungal infection [27] have been reported in underweight populations.
Although patients with NTM-PD tend to have lower BMI and this has been reported to be associated with worse prognosis of NTM-PD [13, 14], the question of whether lower BMI is a risk factor or results from the development of NTM-PD has not yet been answered. The results of our study supported the former hypothesis. In our analysis, higher BMI was associated with decreased development of NTM-PD in a dose-dependent manner and weight loss over a period of 4 years was associated with increased incidence of NTM-PD. The results of a sensitivity analysis showing the same trends after ignoring the development of NTM-PD within 1 year following the initiation date strengthen this observation.
Lower BMI as a risk factor for NTM-PD can be understood based on the interleukin-12 (IL-12)/interferon-γ (IFNγ) pathway, which is pivotal in host immunity against NTM infection [28]. Impaired production of IFNγ, IL-12 and tumour necrosis factor-α (TNF-α) has been reported amongst patients with NTM-PD [11]. Meanwhile, lower BMI is reportedly associated with decreased leptin levels and increased adiponectin levels [11]. Decreased leptin secretion owing to malnutrition is known to cause a blunted T-helper 1 cell (Th1) response, making the host vulnerable to infection [29]. Furthermore, elevation of adiponectin decreases Toll-like receptor (TLR) mediated IFNγ secretion in natural killer cells [30]. Given these observations, adipokine dysregulation in malnourished individuals could be a biological mechanism that contributes to increased risk of NTM-PD [31].
To correctly interpret our results, the strengths and limitations of this study must be recognised. The main strength of the study is the inclusion of a large nationwide study population with standardised health check-up data. This advantage provides good statistical power and minimises the possibility of selection bias. However, the fact that the diagnosis of NTM-PD was made based on ICD-10 codes and was not based on clinical and bacteriological diagnostic criteria is a limitation. To overcome this issue, we tried to minimise the possibility of over-diagnosis by including patients who had at least two hospital visits with a diagnosis of NTM-PD (A31.0).
In conclusion, BMI was inversely related to the development of NTM-PD and weight loss increased the risk of NTM-PD. Clinicians should understand the importance of the nutritional status of individuals with predisposing factors for NTM-PD.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-00454-2020.Supplement
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-00454-2020.Shareable
Footnotes
This article has supplementary material available from erj.ersjournals.com
Conflict of interest: J.H. Song has nothing to disclose.
Conflict of interest: B.S. Kim has nothing to disclose.
Conflict of interest: N. Kwak has nothing to disclose.
Conflict of interest: K-D. Han has nothing to disclose.
Conflict of interest: J-J. Yim has nothing to disclose.
- Received February 28, 2020.
- Accepted August 1, 2020.
- ©ERS 2021. For reproduction rights and permissions contact permissions{at}ersnet.org