You are here

  • Home
  • Archive
  • Volume 77, Issue 11
  • Previous tuberculosis disease as a risk factor for chronic obstructive pulmonary disease: a cross-sectional analysis of multicountry, population-based studies

Article Text

Article menu
Chronic obstructive pulmonary disease
Original research
Previous tuberculosis disease as a risk factor for chronic obstructive pulmonary disease: a cross-sectional analysis of multicountry, population-based studies
  1. Katarina Kamenar1,2,
  2. Shakir Hossen1,2,
  3. Akshay N Gupte2,3,
  4. Trishul Siddharthan1,2,
  5. Suzanne Pollard1,2,
  6. Muhammad Chowdhury4,
  7. Adolfo L Rubinstein5,
  8. Vilma E Irazola5,
  9. Laura Gutierrez5,
  10. J Jaime Miranda6,7,
  11. Antonio Bernabe-Ortiz6,
  12. Dewan Alam8,
  13. Bruce Kirenga9,
  14. Rupert C Jones10,
  15. Frederik van Gemert11,
  16. Robert A Wise1,
  17. William Checkley1,2
  1. 1 Division of Pulmonary and Critical Care, School of Medicine, Johns Hopkins University, Baltimore, Maryland, USA
  2. 2 Center for Global Non-Communicable Disease Research and Training, Johns Hopkins University, Baltimore, Maryland, USA
  3. 3 Division of Infectious Diseases, School of Medicine, Johns Hopkins University, Baltimore, Maryland, USA
  4. 4 Initiative for Noncommunicable Diseases, Health Systems and Population Studies Division, icddr,b, Dhaka, Bangladesh
  5. 5 Institute for Clinical Effectiveness and Health Policy, Buenos Aires, Argentina
  6. 6 CRONICAS Centre of Excellence in Chronic Diseases, Universidad Peruana Cayetano Heredia, Lima, Peru
  7. 7 Departamento de Medicina, Facultad de Medicina, Universidad Peruana Cayetano Heredia, Lima, Peru
  8. 8 School of Kinesiology and Health Science, Faculty of Health, York University, Toronto, Ontario, Canada
  9. 9 Makerere Lung Institute, Makerere University, Kampala, Uganda
  10. 10 Research and Knowledge Exchange, Plymouth Marjon University, Plymouth, UK
  11. 11 University Medical Centre Groningen, University of Groningen, Groningen, The Netherlands
  1. Correspondence to Dr William Checkley, Division of Pulmonary and Critical Care, Johns Hopkins University, Baltimore, MD 21287, USA; wcheckl1{at}jhmi.edu

Abstract

Background Risk factors for COPD in high-income settings are well understood; however, less attention has been paid to contributors of COPD in low-income and middle-income countries (LMICs) such as pulmonary tuberculosis. We sought to study the association between previous tuberculosis disease and COPD by using pooled population-based cross-sectional data in 13 geographically diverse, low-resource settings.

Methods We pooled six cohorts in 13 different LMIC settings, 6 countries and 3 continents to study the relationship between self-reported previous tuberculosis disease and lung function outcomes including COPD (defined as a postbronchodilator forced expiratory volume in one second (FEV1)/forced vital capacity (FVC) below the lower limit of normal). Multivariable regressions with random effects were used to examine the association between previous tuberculosis disease and lung function outcomes.

Results We analysed data for 12 396 participants (median age 54.0 years, 51.5% male); 332 (2.7%) of the participants had previous tuberculosis disease. Overall prevalence of COPD was 8.8% (range 1.7%–15.5% across sites). COPD was four times more common among those with previous tuberculosis disease (25.7% vs 8.3% without previous tuberculosis disease, p<0.001). The adjusted odds of having COPD was 3.78 times higher (95% CI 2.87 to 4.98) for participants with previous tuberculosis disease than those without a history of tuberculosis disease. The attributable fraction of COPD due to previous tuberculosis disease in the study sample was 6.9% (95% CI 4.8% to 9.6%). Participants with previous tuberculosis disease also had lower prebronchodilator Z-scores for FEV1 (−0.70, 95% CI −0.84 to −0.55), FVC (−0.44, 95% CI −0.59 to −0.29) and the FEV1:FVC ratio (−0.63, 95% CI −0.76 to −0.51) when compared with those without previous tuberculosis disease.

Conclusions Previous tuberculosis disease is a significant and under-recognised risk factor for COPD and poor lung function in LMICs. Better tuberculosis control will also likely reduce the global burden of COPD.

  • tuberculosis
  • COPD epidemiology

Data availability statement

Data are available in a public, open access repository. Data are available through BIOLINCC or upon reasonable request.

https://doi.org/10.1136/thoraxjnl-2020-216500

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Key messages

    What is the key question?

    • World Health Organization (WHO) guidelines for Chronic Obstructive Pulmonary Disease (COPD) currently do not include previous tuberculosis disease as a risk factor for COPD. We sought to determine the association between previous tuberculosis disease and lung function outcomes including COPD.

    What is the bottom line?

    • Previous tuberculosis disease was found to be an important risk factor for having COPD and worse lung function. Among persons with COPD, those with previous tuberculosis disease were more likely to experience more severe disease and at an earlier age when compared with those without previous tuberculosis disease.

    Why read on?

    • Our study provides additional evidence that previous tuberculosis disease is an important and under-recognised risk factor for COPD and calls for the inclusion of previous tuberculosis disease as a risk factor in the WHO guidelines for COPD.

    Introduction

    Chronic Obstructive Pulmonary Disease (COPD) is a progressive lung disease characterised by poorly reversible airflow obstruction and is typically accompanied by shortness of breath and worsening cough.1 COPD is currently responsible for approximately 3.2 million deaths a year, making it the third leading cause of death globally.2 With COPD on trend to continue growing in prevalence over the coming decades,2 identifying and addressing its potential risk factors is a public health priority.3 Risk factors for COPD have been well studied in high-income settings; however, less attention has been paid to risk factors that contribute to COPD in low-income and middle-income countries (LMICs).

    Risk factors for COPD in LMICs not only include cigarette smoking but also biomass fuel smoke exposure, poorly controlled asthma, occupational exposures and respiratory infections during early childhood.1–5 Pulmonary tuberculosis, while better controlled in high-income settings, remains a prevalent disease in many LMICs.6 7 Despite higher access to screening and treatment over the last few decades, there were 5.9 million incident cases of active pulmonary tuberculosis in 2019.7 Tuberculosis remains a leading infectious cause of death globally and was responsible for 1.2 million deaths in 2019.7 Yet it is also important to note that survivors of the disease, which vastly outnumber those who die, are likely to suffer from long-term respiratory sequelae. Indeed, a growing body of research indicates that a pulmonary tuberculosis may lead to persistent lung function impairment that manifests as chronic airway obstruction, even in patients who were successfully treated.8–12

    Previous studies have found a positive association between previous tuberculosis disease and worse lung function.8 9 13–16 However, fewer studies have evaluated the relationship between previous tuberculosis disease and lung function outcomes using large-scale, population-based data from exclusively LMICs. We sought to describe the relationship between self-reported previous tuberculosis disease and COPD in LMICs by drawing from a large and geographically diverse study population. The data pooled for these analyses represent a diverse range of rates of urbanisation, educational attainment, smoking prevalence and socioeconomic status in six countries across three different continents.

    Methods

    Study data

    We used pooled data from multiple, contemporary, population-based studies conducted in six countries in Africa, Asia and South America.17–21 The sites in Argentina, Chile and Uruguay were a part of the Pulmonary Risk in South America (PRISA) study.19 The sites in Peru were a part of the CRONICAS cohort study.18 The data from the sites in Bangladesh was collected as part of a longitudinal study.17 The PRISA, CRONICAS and Bangladesh studies underwent variable harmonisation as part of the activities of the Global Alliance for Chronic Diseases. The data from the sites in Uganda were compiled from the Lung Function in Nakaseke and Kampala (LiNK)21 and the Free Respiratory Evaluation and Smoke-exposure reduction by primary Health cAre Integrated gRoups (FRESH AIR) studies.20 Some of these studies are longitudinal cohorts. We only used the baseline (first round of measurements) data for each participant if longitudinal. There are no repeated measurements by individual in this analysis.

    Study design

    We included participants between the ages of 35 and 95 years. We matched the lower age limit among cohorts and thus excluded those under age 35 years. The upper age limit was restricted at 95 years to match reference equations from available normative data.22 A site-stratified simple random sampling strategy outlined by the WHO Expanded Program on Immunization was used for the LiNK study.23 A multilevel sampling approach was used by the FRESH AIR Project.20 24 Random sampling of available census data at each site was used in the Bangladesh study.17 The PRISA study used systematic stratified sampling of multitier clusters by socioeconomic status (SES), age and sex.19 The CRONICAS cohort study used age, sex and site-stratified random sampling of available census data.18 All studies included in the analysis obtained informed consent.17–21 A summary of characteristics of the studies pooled for this analysis in presented in table 1.

    View this table:
    Table 1

    Comparison of the features of the studies pooled for this analysis.

    Spirometry

    All studies used joint American Thoracic Society and European Respiratory Society (ATS/ERS) guidelines for the conduct and interpretation of spirometry,25 and 85%–97% of tests met acceptable criteria. EasyOne spirometers (NDD, Zurich, Switzerland) were used in two studies.17 19 Easy-On PC spirometers (NDD) were used in two studies, and Pneumotrac (Vitalograph, Maids Moreton, UK) spirometers were used in one study.18 20 21 Prebronchodilator readings were ascertained for participants in all studies. Prebronchodilator and postbronchodilator spirometry results were collected from all participants in PRISA and CRONICAS. In the LiNK (forced expiratory volume in one second (FEV1):forced vital capacity (FVC) ratio below fifth percentile of the National Health and Nutrition Examination Survey (NHANES) III African–American population), Bangladesh (FEV1:FVC≤0.7) and FRESH AIR (FEV1:FVC≤0.7) studies, postbronchodilator spirometry was only conducted in participants whose prebronchodilator spirometry results showed evidence of obstruction.

    Definitions

    We defined COPD as having a postbronchodilator FEV1:FVC ratio Z-score 1.64 SDs below the median of the age-adjusted, sex-adjusted and height-adjusted Global Lung Function Initiative 2012 mixed ethnic reference population.22 Individuals were required to have postbronchodilator values to be evaluated for COPD. Severity of COPD was categorised into low, mild, moderate and severe/very severe according to the 2011 Global Initiative for Chronic Obstructive Lung Disease (GOLD) criteria.1 Across all study sites, a history of previous tuberculosis disease was self-reported on a questionnaire by the study participant. Body mass index (BMI) was defined as weight/height2 (kg/m2). We examined two different BMI categories of interest: underweight (BMI ≤19 kg/m2) and obese (BMI ≥30 kg/m2), with BMI of 20–29 kg/m2 as the reference category. We defined daily smokers as current smokers who smoke one or more cigarettes per day. Pack-years of smoking was defined as the number of packs smoked per day multiplied by the number of years of smoking. Biomass fuel smoke exposure was defined as currently using biomass as the primary fuel source for cooking.26 27 Symptomatic COPD was defined as the presence of wheeze, cough or phlegm currently or in the previous 12 months.27

    Biostatistical methods

    The primary aim of this analysis was to evaluate the association between previous tuberculosis disease and lung function outcomes, including COPD. The analysis of lung function was limited to prebronchodilator spirometry. Age-adjusted prevalence rates for previous tuberculosis disease and COPD were calculated using the WHO 2000–2025 Standard Population.28

    We used multivariable logistic regression with random effects to model the odds of having COPD as a function of previous tuberculosis disease adjusted for continuous age in years, sex, daily smoking, biomass fuel use, level of education, and continuous BMI in kilogram per square meter.29 All of the potentially confounding factors were selected a priori. We included a random intercept by study setting. We used adjusted OR estimates to calculate the attributable fraction of COPD due to previous tuberculosis disease in the study sample using Levin’s formula, that is, Embedded Image , where p is the prevalence of previous tuberculosis disease and OR is the adjusted OR between previous tuberculosis disease and COPD.30 We looked at effect modification by sex, smoking status, exposure to biomass fuel smoke, age (≥55 and <55 years) and education. We first evaluated for effect modification through stratified analyses. Additionally, to evaluate whether an interaction was present, we built models with interaction terms and considered an interaction to be present if the p value was less than 0.05. We dichotomised age as ≥55 and <55 years based on the mean age of the study sample. We used multivariable logistic regression models adjusted for a priori potential confounders to estimate site-specific ORs (see online supplemental file for site-specific ORs and 95% CIs).

    Supplemental material

    We used multivariable ordinal logistic regressions with random effects to evaluate the association between previous tuberculosis disease and COPD severity (none, mild, moderate or severe/very severe COPD) or symptomatic COPD (none, asymptomatic COPD and symptomatic COPD) and adjusted for the confounders mentioned previously. We included a random intercept by study setting. We used multivariable ordinal logistic regression models adjusted for a priori potential confounders to estimate site-specific proportional ORs (see online supplemental file for site-specific proportional ORs and 95% CIs).

    We used linear models with random effects to evaluate the association between previous tuberculosis disease and prebronchodilator FEV1, FVC and FEV1:FVC ratio Z-scores.31 We included a random intercept by study setting. The models were adjusted for age, sex, daily smoking, biomass fuel use, level of education and BMI. We used multivariable linear regression models adjusted for a priori potential confounders to estimate site-specific estimates (see online supplemental file for site-specific mean differences in Z-scores and 95% CIs).

    We conducted our analyses with complete data only. A total of 443 participants (3.4% of the data) had missing values. Analyses were performed in Stata V.15.1 and R (www.r-project.org).

    Role of the funding source

    The funders had no role in the data collection, analysis or interpretation, in the writing of the report, or the decision to submit the paper for publication. The corresponding author had full access to the data in the study and had final responsibility for the decision to submit for publication.

    Results

    Participant characteristics

    Of the 13 023 total participants from across the thirteen study sites in Africa, Asia and South America, 12 396 met the inclusion criteria and 332 (2.7%) had previous tuberculosis disease. We summarised sociodemographic characteristics by study setting in table 2. The median age of participants was 54.0 years (IQR 46.5–63.0 years). The study sample was 51.5% male, and 38.3% had completed secondary education or higher. The average pack-years of smoking was 16.2 pack-years with a range of 1.3–35.2 pack-years across settings. The age-adjusted prevalence of COPD ranged widely across settings with the median prevalence at 59.9 cases per 1000 people (IQR 44.7–83.1 cases per 1000 people). The age-adjusted prevalence of previous tuberculosis disease ranged from 5.9 to 74.0 cases per 1000 people with a median prevalence of 14.5 cases per 1000 people (figure 1). We compared sociodemographic differences and lung function Z-scores by COPD status in table 3. As expected, participants with COPD were older, more likely to have previous tuberculosis disease, we more likely to be male, had a lower BMI, a higher average of pack-years smoked, were morel likely to live in a household that used biomass fuels, were less likely to have completed secondary education and had lower lung function values than participants without COPD.

    Figure 1
    Figure 1

    Age-adjusted prevalence of previous tuberculosis disease and COPD by study site. (A) Plot of the age-adjusted prevalence of previous tuberculosis disease and COPD by site. We adjusted for age using the WHO 2000–2025 Standard Population. Sites were arranged by prevalence of COPD from lowest (top) to highest (bottom). (B) Plot of the prevalence of previous tuberculosis disease (x-axis) by prevalence of COPD (y-axis) where each point represents one site. A regression line signifies the positive association between higher prevalence of previous tuberculosis disease and a higher prevalence of COPD (r=0.1, 95% CI 0.08 to 0.11; p<0.001).

    View this table:
    Table 2

    Sociodemographic characteristics stratified by site

    View this table:
    Table 3

    Previous tuberculosis disease, sociodemographic and exposure variables, and prebronchodilator lung function values by COPD status

    Previous tuberculosis disease and COPD

    Overall prevalence of COPD was 8.8%, ranging from 1.7% to 15.4% across sites. Participants with previous tuberculosis disease had a higher prevalence of COPD than those without previous tuberculosis disease (25.9% vs 8.3%, χ2 p<0.001). As expected, the prevalence of COPD in the study sample increased with older age. Moreover, the prevalence of COPD was higher among participants with previous tuberculosis disease when compared with those without previous tuberculosis disease at any age category (figure 2A). The unadjusted odds of having COPD were 4.30 times (95% CI 3.30 to 5.60, Wald test p<0.001) greater among participants with previous tuberculosis disease than in participants without previous tuberculosis disease. In adjusted analyses, the overall OR of having COPD between participants with and without previous tuberculosis disease remained high (OR=3.78, 95% CI 2.87 to 4.98), and the point estimates of the adjusted ORs by site were consistently above 1 (figure 3). We estimated that 6.9% (95% CI 4.8% to 9.6%) of COPD in our study sample was attributable to previous tuberculosis disease, and the attributable fraction for the study sample ranged from 0.63% to 26.1% across settings.

    Figure 2
    Figure 2

    Lung function outcomes stratified by age octiles and previous tuberculosis disease status. (A) Plot of the point prevalence of COPD by age octile and previous tuberculosis disease status. Previous tuberculosis disease status was represented by a filled circle for those with the condition and by an unfilled circle for those without the condition. We also estimated 95% confidence intervals for the point prevalence at each octile. Lines are fitted to describe the cross-sectional relationship between previous tuberculosis disease status and COPD by age. The continuous lines summarise trends for participants with previous tuberculosis disease and broken lines summarise trends for those without the exposure. (B–D) Plot of the mean prebronchodilator FEV1, FVC and FEV1:FVC ratio Z-scores, respectively, by age octile and previous tuberculosis disease status. We fitted lines to describe the cross-sectional relationship between previous tuberculosis disease status and Z-score by age. FEV1, forced expiratory volume in one second; FVC, forced vital capacity.

    Figure 3
    Figure 3

    Interaction effects between previous tuberculosis disease and sex, smoking behaviour, age, education and biomass fuel use on COPD, and the association between previous tuberculosis disease and ordinal outcomes of disease severity and symptoms. The associations between previous tuberculosis disease and COPD were produced by multivariable logistic regression with random effects. (A,B) Plot of the unadjusted and adjusted estimates, respectively, with corresponding 95% CIs using data from all sites, respectively. (C) Plot of site-specific estimates using triangles to designate each site. The overall model evaluated the relationship between previous tuberculosis disease and COPD while adjusting for sex, daily smoking behaviour, age, level of educational attainment and body mass index. We explored potential interaction effects between previous tuberculosis disease and sex, smoking status (daily vs not daily), age (≥55 and <55 years), education (secondary education or higher vs primary education or lower) and use of biomass or clean fuels on COPD outcomes. The last two rows plot the ordinal association between previous tuberculosis disease and COPD severity (none, mild, moderate or severe/very severe COPD) or symptomatic COPD (none, asymptomatic COPD and symptomatic COPD) adjusted for the aforementioned confounders. Site-specific ORs and corresponding 95% CIs are found in the online supplemental file.

    In the stratified analysis, the adjusted odds of having COPD between those with and without previous tuberculosis disease was similar by sex (Wald test for interaction, p=0.55), age category (Wald test for interaction, p=0.43), education category (Wald test for interaction, p=0.90) and type of cooking fuel (Wald test for interaction, p=0.62). There was effect modification by smoking status; participants who smoked cigarettes daily were less likely to have COPD in the setting of previous tuberculosis disease than those who did not smoke (Wald test for interaction, p=0.02). Participants with previous tuberculosis disease also had greater adjusted odds of having more severe airflow obstruction or having symptoms when compared with those without previous tuberculosis disease (figure 3).

    For the 1091 participants with COPD, the average prebronchodilator FEV1 (-1.97 vs −0.25, t-test p<0.001), FVC (-0.49 vs −0.05, t-test p<0.001) and FEV1:FVC ratio Z-scores (-2.82 vs −0.42 t test p<0.001) were lower for those with previous tuberculosis disease compared with those without previous tuberculosis disease, respectively. Additionally, participants with COPD and previous tuberculosis disease were on average 2.7 years younger than their counterparts without previous tuberculosis disease (56.0 vs 58.7 years, respectively; t-test p=0.01). Participants with COPD and previous tuberculosis disease, when compared with those without previous tuberculosis disease, were more likely to be male (72.1% vs 59.4%, respectively; t-test p=0.007), were less likely to be daily smokers (15.1% vs 25.6%, respectively; t-test p=0.006) and were more likely to live in urban settings (69.9% vs 58.8%, respectively; t-test p=0.019).

    Previous tuberculosis disease and lung function

    We plotted mean Z-scores for prebronchodilator FEV1, FVC and FEV1:FVC ratio and stratified by both age octiles and previous tuberculosis disease status in figure 2. Participants with previous tuberculosis disease had lower lung function than those without previous tuberculosis disease (figure 2B–D). Indeed, the difference in prebronchodilator FEV1, FVC and FEV1:FVC ratio Z-scores were consistently lower at any age category for participants with previous tuberculosis disease when compared with those without previous tuberculosis disease, and without an apparent interaction between previous tuberculosis disease and age. In the unadjusted analysis, FEV1, FVC and FEV1:FVC ratio Z-scores were 0.77 (p<0.001), 0.47 (p<0.001) and 0.70 (p<0.001) lower in participants with previous tuberculosis disease than in those without the exposure. Adjusted analyses resulted in similar findings for FEV1 (figure 4), FVC (figure 5) and FEV1:FVC ratio (figure 6), and the point estimates of the adjusted Z-scores by site were consistently below 0 for all lung function parameters (with the exception of one site for FEV1:FVC ratio). Our analysis did not reveal any effect modification between previous tuberculosis disease and sex, smoking behaviour, fuel type, age category and educational category at the 0.05 level.

    Figure 4
    Figure 4

    Interaction effects between previous tuberculosis disease and sex, smoking behaviour, age, education and biomass fuel use on prebronchodilator FEV1 Z-score. The associations between previous tuberculosis disease and FEV1 Z-score were produced by multivariable linear models with random effects. (A,B) Plot of the unadjusted and adjusted estimates, respectively, and corresponding 95% CIs using data from all sites, respectively. (C) Plot of site-specific estimates using triangles to designate each site. The overall model evaluated the relationship between previous tuberculosis disease and FEV1 Z-score while adjusting for sex, daily smoking behaviour, age, level of educational attainment and body mass index. We explored potential interaction effects between previous tuberculosis disease and sex, smoking status (daily vs not), age (≥55 and <55 years), education (secondary education or higher vs primary education or lower) and use of biomass or clean fuels on COPD outcomes. Site-specific mean differences in Z-scores and corresponding 95% CIs are found in the online supplemental file. FEV1, forced expiratory volume in one second.

    Figure 5
    Figure 5

    Interaction effects between previous tuberculosis disease and sex, smoking behaviour, age, education and biomass fuel use on prebronchodilator FVC Z-score. The associations between previous tuberculosis disease and FVC Z-score were produced by multivariable linear models using generalised estimating equations. (A,B) Plot of the unadjusted and adjusted estimates and corresponding 95% CIs using data from all sites, respectively. (C) Plot of site-specific estimates using triangles to designate each site. The overall model evaluated the relationship between previous tuberculosis disease and FVC Z-score while adjusting for sex, daily smoking behaviour, age, level of educational attainment and body mass index. We explored potential interaction effects between previous tuberculosis disease and sex, smoking status (daily vs not), age (≥55 and<55 years), education (secondary education or higher vs primary education or lower) and use of biomass or clean fuels on COPD outcomes. Site-specific mean differences in Z-scores and corresponding 95% CI are found in the online supplemental file. FVC, forced vital capacity.

    Figure 6
    Figure 6

    Interaction effects between previous tuberculosis disease and sex, smoking behaviour, age, education, biomass fuel use and BMI on prebronchodilator FEV1:FVC ratio Z-score. The associations between previous tuberculosis disease and FEV1:FVC ratio Z-score were produced by multivariable linear models using generalised estimating equations. (A,B) Plot of the unadjusted and adjusted estimates and corresponding 95% CIs using data from all sites, respectively. (B) Plot of site-specific estimates using triangles to designate each site. The overall model evaluated the relationship between previous tuberculosis disease and FEV1:FVC ratio Z-score while adjusting for sex, daily smoking behaviour, age, level of educational attainment and BMI. We explored potential interaction effects between previous tuberculosis disease and sex, smoking status (daily vs not), age (≥55 and <55 years), education (secondary education or higher vs primary education or lower), and use of biomass or clean fuels on COPD outcomes. Site-specific mean differences in Z-scores and corresponding 95% CIs are found in the online supplemental file. BMI, body mass index; FEV1, forced expiratory volume in one second; FVC, forced vital capacity.

    Discussion

    Our analysis used contemporary data from 12 396 participants from 13 resource-poor settings in six countries in Africa, Asia and South America to describe the relationship between previous tuberculosis disease and COPD. The study sample comprised sites with varying degrees of urbanisation, smoking prevalence, biomass fuel use and educational attainment and found that previous tuberculosis disease is a significant and under-recognised risk factor for COPD. We observed that participants with previous tuberculosis disease had a higher prevalence of COPD and worse lung function than those without previous tuberculosis disease. Specifically, 6.9% of the COPD in the study sample was attributable to previous tuberculosis disease. This suggests that improved tuberculosis control may also help to reduce the burden of COPD in LMICs.

    The relationship between previous tuberculosis disease and chronic airway obstruction has been well studied; however, estimates of the proportions of individuals with treated tuberculosis who are left with lasting lung function impairment are highly variable across studies. A systematic review of 14 studies conducted by Ravimohan et al found that lung function impairment was common in patients with a history of tuberculosis, with a prevalence of airflow obstruction ranging from 18.4% to 86.8% across studies.9 The results of this review and a study conducted by Gupte et al suggest that duration of disease and timely access to tuberculosis treatment may have important effects on the severity of the lung impairment.9 13 Previous research also indicates that while lung function improved for patients after treatment for their tuberculosis, 18.6% of patients were found to have residual restriction and 16.3% had airflow obstruction.15 In our analysis, participants with previous tuberculosis disease had lower lung function even after adjusting for age, sex and important environmental exposures such as biomass fuel smoke exposure and smoking status. These results are congruent with the findings from earlier studies.9 Together, they provide robustness to the evidence that previous tuberculosis disease is a casual factor of COPD.

    The association we found between previous tuberculosis disease and COPD is consistent with those found in other studies. A systematic review of the association between pulmonary tuberculosis and chronic airway obstruction conducted by Allwood et al found a positive association in 17 out of 19 studies, with ORs ranging from 1.37 to 2.94.8 This previous systematic review included population-based studies such as the PLATINO study in South America, the PREPOCOL study in Colombia and the BOLD study conducted in 19 settings across the world.16 Specifically, the BOLD study included 14 050 participants mainly from high-income countries and found that a history of tuberculosis was associated with 2.51 higher odds (95% CI 1.83 to 3.42) of having airflow obstruction.16 A second systematic review by Byrne et al also found similar results, with a pooled OR of 3.05 (95% CI 2.41 to 3.85) to describe the association between previous tuberculosis disease and COPD.11 We believe this study adds value to the existing body of literature because it was conducted entirely in LMIC settings. Since treatment accessibility and outcome severity for tuberculosis are systematically worse in resource-poor communities and competing causes of COPD may differ with higher-resource settings, it is worthwhile to examine the association between previous tuberculosis disease and COPD exclusively in LMICs.7

    Participants with previous tuberculosis disease also had lower prebronchodilator FEV1, FVC and FEV1:FVC ratio Z-scores than those without previous tuberculosis disease. This finding has some additional potential implications. First, the pathobiology of tuberculosis may involve more endobronchial spread leading to airway fibrosis and narrowing than previously considered.9 Second, previous tuberculosis disease was one of the most important risk factors for impaired lung function. Furthermore, we found that among participants with COPD, prebronchodilator FEV1:FVC ratio, FVC and FEV1 Z-scores were all lower in participants with previous tuberculosis disease when compared with those without previous tuberculosis disease. The lower FVC Z-scores among patients with previous tuberculosis disease are consistent with the findings by Allwood et al and indicate restrictive spirometry.15 Moreover, we also found that the severity of COPD according to the GOLD criteria was worse among those with previous tuberculosis disease. These findings agree with the results of a small-scale conducted by Yakar et al which found that patients with COPD with previous tuberculosis disease performed worse on lung function tests.14 This suggests that COPD related to previous tuberculosis disease may result in more severe disease than COPD from other causes. Furthermore, we found that previous tuberculosis disease-related COPD also affected younger age groups when compared with COPD from other causes and may therefore lead to more disability-adjusted life years lost. These results agree with the findings of Yakar et al, which suggested that among patients with COPD, those that had a history of tuberculosis were more likely be diagnosed and die from COPD at an earlier age than those with no reported history of tuberculosis disease.14 Third, we found that among participants with COPD, those with previous tuberculosis disease were more likely to be male, non-daily smokers and live in urban environments than those without previous tuberculosis disease.

    In this analysis, we expected to see a positive interaction between daily smoking behaviour and previous tuberculosis disease on the odds of having COPD. However, our findings were the opposite of what we expected. A potential explanation for these results is residual confounding due to unmeasured factors such as socioeconomic status, despite controlling for education. In low-resource environments, daily smoking status is commonly associated with higher socioeconomic status.32 Additionally, only 42 of the 12 396 participants reported both exposures of previous tuberculosis disease and daily smoking behaviour. Since the sample size is small, further research is needed before concluding that smoking daily reduces the odds of having COPD among individuals with previous tuberculosis disease. Finally, pack-years smoking was low across cohorts, so smoking exposure in this study sample may not have been sufficient to show an impact on relationship between previous tuberculosis disease and COPD.

    We estimated that 6.9% of COPD in the study sample was attributable to previous tuberculosis disease. In comparison, we found in other analyses that 13.5% of COPD in this study sample was attributable to household air pollution and 12.4% attributable to daily cigarette smoking.26 Furthermore, we found a notable range in the attributable fraction across settings with 0.63%–26.1% of COPD attributable to previous tuberculosis disease in our samples. This observed heterogeneity of the attributable fractions likely reflect the variations in the competing causes of COPD risk factors such as biomass fuel exposure and smoking status and the heterogeneity in previous tuberculosis diseases across sites (0.6%–6.9%). It is important to acknowledge, however, that the attributable fraction calculated here is based on our analytical study sample and thus may not be representative of the populations under study. Overall, what is notable about these results is the low prevalence of previous tuberculosis disease: only 2.7% of the study sample reported previous tuberculosis disease compared with 38% of the sample reporting household air pollution exposure and 12% reporting daily smoking behaviour.26 32 This suggests that even though previous tuberculosis disease is a less prevalent risk factor, it is still a considerable risk factor for COPD. Additionally, we found that the COPD prevalence among the sites (range 1.7%–15.4%) was not explained by any one risk factor. COPD in LMIC settings is not due to a singular risk factor such as tobacco smoking, and there are many important pathways to the development of COPD in these settings.33

    Our analysis has some strengths. We used a large and diverse study sample which allowed for us to adjust for other known risk factors of COPD. Since we pooled data from six different LMICs, the results of this analysis are generalisable to a variety of settings in the global South. This study was not a systematic review or meta-analysis, however. Instead, it is a pooled analysis of data from contemporary studies collected by a group of investigators working together under a US National Institutes of Health and Global Alliance for Chronic Diseases Networks. Second, we used the lower limit of normal to diagnose COPD instead of a fixed cut-off. This means we had a lower likelihood of overdiagnosing COPD in our study sample, especially among older participants.22

    There are some potential shortcomings, however. Our conclusions about the relationships between previous tuberculosis disease and COPD are based on cross-sectional data. This means that we cannot establish a causal effect or a temporal relationship between previous tuberculosis disease and COPD. It remains unclear if pulmonary tuberculosis disease leads to inflammation in the lung tissue which makes individuals more prone to COPD or if chronic obstruction of the airway increases a person’s susceptibility to tuberculosis. Additional cohort studies with repeated assessments of lung function and tracking of pulmonary tuberculosis are needed to establish temporality.34 35 In particular, nested case–cohort studies where lung function data are available before the onset of tuberculosis disease among cases and in a subcohort could be helpful in establishing a temporal relationship. Future studies should examine the relationship between treatment outcomes of tuberculosis, including treatment for multidrug resistance tuberculosis and long-term lung function, such as the likelihood of developing COPD. Additionally, while the studies pooled for this analysis collected analogous data from participants, they did not use identical methods to collect data including different protocols for undertaking postbronchodilator spirometry. Furthermore, quality control methods were not outlined for all studies pooled in this analysis, and thus we were unable to control for the potential biases created by the different methodologies used at study sites. However, given that all pooled studies showed a similar direction in findings, we believe the pooled results are valid. Another potential limitation of this analysis is that we needed to rely on self-reporting of history of tuberculosis at any point in time. Thus, we were not able to account for duration of disease, its severity or reactivation, the type and duration of treatment courses, and the lag time between the tuberculosis episode and development of COPD. There was also potential underestimation in the prevalence of previous tuberculosis disease in the study sample due to under-self-reporting because of stigmatisation of the disease and or a lack of diagnosis in participants. Despite that, we believe that the recall bias for self-reporting is limited in this case because tuberculosis is a significant and serious disease that is not easily forgotten by patients. We were also unable to account for HIV status as a potential confounder in this analysis since HIV status was not available for all pooled studies. The potential effect modification of HIV on the relationship between pulmonary tuberculosis and COPD is worth exploring in future studies.6 7 Additionally, while we adjusted for the level of educational attainment, we were unable to adequately control for socioeconomic status in this analysis. Thus, we were unable to account for this residual confounding and investigate the potential associations between poverty-driven exposures, tuberculosis and COPD. Future longitudinal studies should better control for socioeconomic risk factors and explore the potential effect modification that poverty-driven exposures such as childhood respiratory infections and nutritional deficiencies have on the relationship between pulmonary tuberculosis and COPD. Another limitation is that while the overall study sample was large, with over 12 000 participants, the number of tuberculosis cases was relatively low, with only 332 cases. This means that the comparisons among participants with COPD were based on a small sample size and thus are less reliable than other estimates. Lastly, this analysis did not account for the occupational exposures which may result in residual confounding on the association of interest.

    The WHO currently recognises tobacco use, air pollution and occupation exposures as major risk factors for COPD. Missing from the list is a history of pulmonary tuberculosis, a major public health issue that remains devastating to populations in LMICs. The results of this analysis show that tuberculosis is likely an important and under-recognised risk factor for COPD in these settings. Our findings suggest that health systems may need to plan for a continuum of care for their patients with pulmonary tuberculosis such as improved access to early detection and treatment of COPD. Furthermore, the results of this analysis provide an additional reason for the prioritisation of tuberculosis control and treatment on a global scale to reduce the potential burden of COPD on LMICs in the coming decades.

    Data availability statement

    Data are available in a public, open access repository. Data are available through BIOLINCC or upon reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    The cohorts received ethics approval from internal review boards in the USA (Johns Hopkins School of Medicine, Johns Hopkins Bloomberg School of Public Health, Tulane University); Peru (Universidad Peruana Cayetano Heredia, A.B. PRISMA); Argentina (Comité de Ética de Protocolos de Investigación del Hospital Italiano de Buenos Aires); Chile (Comité de Ética del Servicio de Salud Araucanía Sur, Universidad de la Frontera); Uruguay (Comité de Ética para Proyectos de Investigación de la Universidad de la República); Uganda (Johns Hopkins University-Makerere University, Makerere University School of Medicine Research and Ethics Committee, and Uganda National Council for Science and Technology); and, Bangladesh (Review Committee and the Ethical Review Committee, icddr,b).

    Acknowledgments

    We gratefully acknowledge the study participants and the commitment of funders who supported CRONICAS, PRISA, ACCESS and icddr,b.

    References

    1. ↵
      2. Vestbo J ,
      3. Hurd SS ,
      4. Agustí AG , et al
      . Global strategy for the diagnosis, management, and prevention of chronic obstructive pulmonary disease: gold executive summary. Am J Respir Crit Care Med 2013;187:347–65.doi:10.1164/rccm.201204-0596PP pmid:http://www.ncbi.nlm.nih.gov/pubmed/22878278
      OpenUrlCrossRefPubMedWeb of Science
    2. ↵
      1. GBD Chronic Respiratory Disease Collaborators
      . Prevalence and attributable health burden of chronic respiratory diseases, 1990-2017: a systematic analysis for the global burden of disease study 2017. Lancet Respir Med 2020;8:585–96.doi:10.1016/S2213-2600(20)30105-3 pmid:http://www.ncbi.nlm.nih.gov/pubmed/32526187
      OpenUrlPubMed
    3. ↵
      2. Quaderi SA ,
      3. Hurst JR
      . The unmet global burden of COPD. Glob Health Epidemiol Genom 2018;3:e4.doi:10.1017/gheg.2018.1 pmid:http://www.ncbi.nlm.nih.gov/pubmed/29868229
      OpenUrlPubMed
    4. ↵
      2. Bui DS ,
      3. Lodge CJ ,
      4. Burgess JA , et al
      . Childhood predictors of lung function trajectories and future COPD risk: a prospective cohort study from the first to the sixth decade of life. Lancet Respir Med 2018;6:535–44.doi:10.1016/S2213-2600(18)30100-0 pmid:http://www.ncbi.nlm.nih.gov/pubmed/29628376
      OpenUrlPubMed
    5. ↵
      2. Tai A ,
      3. Tran H ,
      4. Roberts M , et al
      . The association between childhood asthma and adult chronic obstructive pulmonary disease. Thorax 2014;69:805–10.doi:10.1136/thoraxjnl-2013-204815 pmid:http://www.ncbi.nlm.nih.gov/pubmed/24646659
      OpenUrlAbstract/FREE Full Text
    6. ↵
      2. Glaziou P ,
      3. Sismanidis C ,
      4. Floyd K , et al
      . Global epidemiology of tuberculosis. Cold Spring Harb Perspect Med 2014;5:a017798.doi:10.1101/cshperspect.a017798 pmid:http://www.ncbi.nlm.nih.gov/pubmed/25359550
      OpenUrlPubMed
    7. ↵
      1. World Health Organization
      . Global tuberculosis report 2020; 2020.
    8. ↵
      2. Allwood BW ,
      3. Myer L ,
      4. Bateman ED
      . A systematic review of the association between pulmonary tuberculosis and the development of chronic airflow obstruction in adults. Respiration 2013;86:76–85.doi:10.1159/000350917 pmid:http://www.ncbi.nlm.nih.gov/pubmed/23652030
      OpenUrlCrossRefPubMedWeb of Science
    9. ↵
      2. Ravimohan S ,
      3. Kornfeld H ,
      4. Weissman D , et al
      . Tuberculosis and lung damage: from epidemiology to pathophysiology. Eur Respir Rev 2018;27:170077.doi:10.1183/16000617.0077-2017 pmid:http://www.ncbi.nlm.nih.gov/pubmed/29491034
      OpenUrlAbstract/FREE Full Text
    10. ↵
      2. O'Toole RF ,
      3. Shukla SD ,
      4. Walters EH
      . Tb meets COPD: an emerging global co-morbidity in human lung disease. Tuberculosis 2015;95:659–63.doi:10.1016/j.tube.2015.08.005 pmid:http://www.ncbi.nlm.nih.gov/pubmed/26386744
      OpenUrlPubMed
    11. ↵
      2. Byrne AL ,
      3. Marais BJ ,
      4. Mitnick CD , et al
      . Tuberculosis and chronic respiratory disease: a systematic review. Int J Infect Dis 2015;32:138–46.doi:10.1016/j.ijid.2014.12.016 pmid:http://www.ncbi.nlm.nih.gov/pubmed/25809770
      OpenUrlCrossRefPubMed
    12. ↵
      2. Dhar R ,
      3. Singh S ,
      4. Talwar D , et al
      . Bronchiectasis in India: results from the European multicentre bronchiectasis audit and research collaboration (EMBARC) and respiratory research network of India registry. Lancet Glob Health 2019;7:e1269–79.doi:10.1016/S2214-109X(19)30327-4 pmid:http://www.ncbi.nlm.nih.gov/pubmed/31402007
      OpenUrlPubMed
    13. ↵
      2. Gupte AN ,
      3. Paradkar M ,
      4. Selvaraju S , et al
      . Assessment of lung function in successfully treated tuberculosis reveals high burden of ventilatory defects and COPD. PLoS One 2019;14:e0217289.doi:10.1371/journal.pone.0217289 pmid:http://www.ncbi.nlm.nih.gov/pubmed/31120971
      OpenUrlPubMed
    14. ↵
      2. Yakar HI ,
      3. Gunen H ,
      4. Pehlivan E , et al
      . The role of tuberculosis in COPD. Int J Chron Obstruct Pulmon Dis 2017;12:323–9.doi:10.2147/COPD.S116086 pmid:http://www.ncbi.nlm.nih.gov/pubmed/28176901
      OpenUrlPubMed
    15. ↵
      2. Allwood BW ,
      3. Maasdorp E ,
      4. Kim GJ , et al
      . Transition from restrictive to obstructive lung function impairment during treatment and follow-up of active tuberculosis. Int J Chron Obstruct Pulmon Dis 2020;15:1039–47.doi:10.2147/COPD.S219731 pmid:http://www.ncbi.nlm.nih.gov/pubmed/32494129
      OpenUrlPubMed
    16. ↵
      2. Amaral AFS ,
      3. Coton S ,
      4. Kato B , et al
      . Tuberculosis associates with both airflow obstruction and low lung function: BOLD results. Eur Respir J 2015;46:1104–12.doi:10.1183/13993003.02325-2014 pmid:http://www.ncbi.nlm.nih.gov/pubmed/26113680
      OpenUrlAbstract/FREE Full Text
    17. ↵
      2. Alam DS ,
      3. Chowdhury MA ,
      4. Siddiquee AT , et al
      . Prevalence and determinants of chronic obstructive pulmonary disease (COPD) in Bangladesh. COPD 2015;12:658–67.doi:10.3109/15412555.2015.1041101 pmid:http://www.ncbi.nlm.nih.gov/pubmed/26263031
      OpenUrlPubMed
    18. ↵
      2. Jaganath D ,
      3. Miranda JJ ,
      4. Gilman RH , et al
      . Prevalence of chronic obstructive pulmonary disease and variation in risk factors across four geographically diverse resource-limited settings in Peru. Respir Res 2015;16:40.doi:10.1186/s12931-015-0198-2 pmid:http://www.ncbi.nlm.nih.gov/pubmed/25889777
      OpenUrlPubMed
    19. ↵
      2. Rubinstein AL ,
      3. Irazola VE ,
      4. Bazzano LA , et al
      . Detection and follow-up of chronic obstructive pulmonary disease (COPD) and risk factors in the southern cone of Latin America: the pulmonary risk in South America (PRISA) study. BMC Pulm Med 2011;11:34.doi:10.1186/1471-2466-11-34 pmid:http://www.ncbi.nlm.nih.gov/pubmed/21627849
      OpenUrlPubMed
    20. ↵
      2. van Gemert F ,
      3. Kirenga B ,
      4. Chavannes N , et al
      . Prevalence of chronic obstructive pulmonary disease and associated risk factors in Uganda (fresh air Uganda): a prospective cross-sectional observational study. Lancet Glob Health 2015;3:e44–51.doi:10.1016/S2214-109X(14)70337-7 pmid:http://www.ncbi.nlm.nih.gov/pubmed/25539969
      OpenUrlPubMed
    21. ↵
      2. Siddharthan T ,
      3. Grigsby M ,
      4. Morgan B , et al
      . Prevalence of chronic respiratory disease in urban and rural Uganda. Bull World Health Organ 2019;97:318–27.doi:10.2471/BLT.18.216523 pmid:http://www.ncbi.nlm.nih.gov/pubmed/31551628
      OpenUrlPubMed
    22. ↵
      2. Quanjer PH ,
      3. Stanojevic S ,
      4. Cole TJ , et al
      . Multi-Ethnic reference values for spirometry for the 3-95-yr age range: the global lung function 2012 equations. Eur Respir J 2012;40:1324–43.doi:10.1183/09031936.00080312 pmid:http://www.ncbi.nlm.nih.gov/pubmed/22743675
      OpenUrlAbstract/FREE Full Text
    23. ↵
      2. Bostoen K ,
      3. Chalabi Z
      . Optimization of household survey sampling without sample frames. Int J Epidemiol 2006;35:751–5.doi:10.1093/ije/dyl019 pmid:http://www.ncbi.nlm.nih.gov/pubmed/16481364
      OpenUrlCrossRefPubMedWeb of Science
    24. ↵
      2. Chao L-W ,
      3. Szrek H ,
      4. Peltzer K , et al
      . A comparison of EPI sampling, probability sampling, and compact segment sampling methods for micro and small enterprises. J Dev Econ 2012;98:94–107.doi:10.1016/j.jdeveco.2011.08.007 pmid:http://www.ncbi.nlm.nih.gov/pubmed/22582004
      OpenUrlCrossRefPubMed
    25. ↵
      2. Miller MR ,
      3. Hankinson J ,
      4. Brusasco V , et al
      . Standardisation of spirometry. Eur Respir J 2005;26:319–38.doi:10.1183/09031936.05.00034805 pmid:http://www.ncbi.nlm.nih.gov/pubmed/16055882
      OpenUrlAbstract/FREE Full Text
    26. ↵
      2. Siddharthan T ,
      3. Grigsby MR ,
      4. Goodman D , et al
      . Association between household air pollution exposure and chronic obstructive pulmonary disease outcomes in 13 low- and middle-income country settings. Am J Respir Crit Care Med 2018;197:611–20.doi:10.1164/rccm.201709-1861OC pmid:http://www.ncbi.nlm.nih.gov/pubmed/29323928
      OpenUrlPubMed
    27. ↵
      2. Grigsby MR ,
      3. Siddharthan T ,
      4. Pollard SL , et al
      . Low body mass index is associated with higher odds of COPD and lower lung function in low- and middle-income countries. COPD 2019;16:58–65.doi:10.1080/15412555.2019.1589443 pmid:http://www.ncbi.nlm.nih.gov/pubmed/31032662
      OpenUrlPubMed
    28. ↵
      2. Ahmad OB ,
      3. Boschi-Pinto C ,
      4. Lopez AD
      . Age standardization of rates: a new who standard. GPE Discussion Paper Series;31.
    29. ↵
      2. Breslow NE ,
      3. Clayton DG
      . Approximate inference in generalized linear mixed models. J Amer Stat Assoc 1993;88:9–25.
      OpenUrl
    30. ↵
      2. Levin ML
      . The occurrence of lung cancer in man. Acta Unio Int Contra Cancrum 1953;9:531.pmid:http://www.ncbi.nlm.nih.gov/pubmed/13124110
      OpenUrlPubMed
    31. ↵
      2. Pinheiro JC ,
      3. Bates DM
      . Mixed-Effects models in S and S-PLUS. New York: Springer, 2000.
    32. ↵
      2. Grigsby M ,
      3. Siddharthan T ,
      4. Chowdhury MA , et al
      . Socioeconomic status and COPD among low- and middle-income countries. Int J Chron Obstruct Pulmon Dis 2016;11:2497–507.doi:10.2147/COPD.S111145 pmid:http://www.ncbi.nlm.nih.gov/pubmed/27785006
      OpenUrlPubMed
    33. ↵
      2. Checkley W ,
      3. Pollard SL ,
      4. Siddharthan T , et al
      . Managing threats to respiratory health in urban slums. Lancet Respir Med 2016;4:852–4.doi:10.1016/S2213-2600(16)30245-4 pmid:http://www.ncbi.nlm.nih.gov/pubmed/27760724
      OpenUrlPubMed
    34. ↵
      2. Hnizdo E ,
      3. Singh T ,
      4. Churchyard G
      . Chronic pulmonary function impairment caused by initial and recurrent pulmonary tuberculosis following treatment. Thorax 2000;55:32–8.doi:10.1136/thorax.55.1.32 pmid:http://www.ncbi.nlm.nih.gov/pubmed/10607799
      OpenUrlAbstract/FREE Full Text
    35. ↵
      2. Gupte AN ,
      3. Wong ML ,
      4. Msandiwa R , et al
      . Factors associated with pulmonary impairment in HIV-infected South African adults. PLoS One 2017;12:e0184530.doi:10.1371/journal.pone.0184530 pmid:http://www.ncbi.nlm.nih.gov/pubmed/28902919
      OpenUrlPubMed

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Contributors KK, SH and WC conceived the study idea and wrote the first draft of the manuscript. SH and WC performed data analysis and interpretation and designed the figures. ANG, TS, SP, MC, ALR, VEI, LG, JJM, AB-O, DA, BK, RCJ, FvG and RAW contributed to the study design, data collection and manuscript writing. All coauthors were involved in manuscript development, performed a full review of the article, approved the final version of the manuscript and were responsible for all content. WC accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.

    • Funding This study was sponsored and funded by the National Heart, Lung, and Blood Institute, a division of the National Institute of Health in the United States, under contracts HHSN268200900029C, HHSN26820900032C and HHSN268200900033C.

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

    Linked Articles