Pulmonary restriction predicts long-term pulmonary impairment in people with HIV and tuberculosis

Background While tuberculosis is considered a risk factor for chronic obstructive pulmonary disease, a restrictive pattern of pulmonary impairment may actually be more common among tuberculosis survivors. We aimed to determine the nature of pulmonary impairment before and after treatment among people with HIV and tuberculosis and identify risk factors for long-term impairment. Methods In this prospective cohort study conducted in South Africa, we enrolled adults newly diagnosed with HIV and tuberculosis who were initiating antiretroviral therapy and tuberculosis treatment. We measured lung function and symptoms at baseline, 6, and 12 months. We compared participants with and without pulmonary impairment and constructed logistic regression models to identify characteristics associated with pulmonary impairment. Results Among 134 participants with a median CD4 count of 110 cells/μl, 112 (83%) completed baseline spirometry at which time 32 (29%) had restriction, 13 (12%) had obstruction, and 9 (7%) had a mixed pattern. Lung function was dynamic over time and 30 (33%) participants had impaired lung function at 12 months. Baseline restriction was associated with greater symptoms and with long-term pulmonary impairment (adjusted odds ratio 5.44, 95% confidence interval 1.16–25.45), while baseline obstruction was not (adjusted odds ratio 1.95, 95% confidence interval 0.28–13.78). Conclusions In this cohort of people with HIV and tuberculosis, restriction was the most common, symptomatic, and persistent pattern of pulmonary impairment. These data can help to raise awareness among clinicians about the heterogeneity of post-tuberculosis pulmonary impairment, and highlight the need for further research into mediators of lung injury in this vulnerable population.


Background
Among the approximately 8.7 million people who survive tuberculosis disease each year, it is estimated that 30-50%, representing 2.6-4.4 million individuals, will develop chronic lung disease [1][2][3][4][5][6]. Reduced lung function has been associated with increased all-cause mortality and lost productivity and so the potential impact for millions of tuberculosis survivors is profound [7][8][9]. In fact, the majority of tuberculosis-related disability is attributable to chronic pulmonary impairment after cure and not acute disability during active disease [1]. While there is growing awareness of the long-term burden of pulmonary impairment after tuberculosis, there are limited data from prospective cohorts to help us understand the nature of these respiratory sequelae.
Post-tuberculosis lung impairment has traditionally been viewed as an obstructive phenomenon, similar to chronic obstructive pulmonary disease (COPD), where patients have difficulty exhaling as a result of inflammation and damage to the large airways. However, recent data suggest that pulmonary restriction may be as common, if not more so, among tuberculosis survivors [10][11][12]. Pulmonary restriction, which is characterized by the inability to inhale to full lung capacity, typically arises from scarring and fibrosis of the lung parenchyma. Treatment for pulmonary obstruction focuses on relieving airflow limitation with bronchodilators and inhaled corticosteroids. In contrast, there are limited treatments available for fibrotic pulmonary diseases, none of which have been studied in patients with tuberculosis [13].
People with HIV are a uniquely vulnerable population in their potential for lung injury. Not only are people with HIV at increased risk of developing tuberculosis, they also have an increased risk for non-infectious pulmonary diseases such as COPD and pulmonary fibrosis [14][15][16]. Furthermore, while people with HIV and pulmonary tuberculosis often have less lung involvement at diagnosis, immune restoration on antiretroviral therapy (ART) may lead to incident pulmonary inflammation and declines in lung function [17,18]. In a limited number of studies that have focused on lung impairment after tuberculosis in people with HIV, it is unclear whether HIV coinfection alters the risk of incident pulmonary impairment [2,12,19,20].
Given the disproportionate burden of tuberculosis in people with HIV, there is an urgent need to better understand the epidemiology, patterns, and drivers of lung damage in people with HIV and tuberculosis. We sought to determine the prevalence of obstructive versus restrictive lung impairment at baseline and after tuberculosis treatment in a cohort of individuals with HIV and tuberculosis coinfection in order to identify characteristics associated with persistently impaired lung function. A better understanding of pulmonary impairment among tuberculosis survivors will lay the groundwork for future trials to improve lung function after tuberculosis.

Study design and participants
The Lung Function after TB-IRIS (LIFT-IRIS) Study was a prospective cohort study conducted in Gauteng, South Africa that evaluated pulmonary function and symptoms in HIV-infected adults diagnosed and treated for pulmonary tuberculosis [18]. Results are reported here in accordance with STROBE guidelines. We enrolled adults age 18 years or older who were newly diagnosed with HIV and ART-naïve with a CD4 count ≤ 500 cells/ μL. Pulmonary tuberculosis was diagnosed based upon a positive Xpert MTB/RIF ® test. Exclusion criteria included rifampin-resistance, tuberculous meningitis, current immunosuppressive therapy, current incarceration, and pregnancy. All participants were treated with standard first-line tuberculosis treatment and ART (i.e., efavirenz, tenofovir, emtricitabine) as per local guidelines. Participants were clinically screened for additional opportunistic infections at all study visits.

Study procedures
Study visits occurred prior to ART initiation (baseline) and at 6 and 12 months. At baseline, participants were asked about smoking and cigarette pack-years (i.e., cigarettes smoked/day multiplied by years smoked) were calculated. CD4 cell count and HIV viral load were measured at baseline and 1 month; hemoglobin A1c (HbA1c) was measured at baseline. At each visit, pulmonary symptoms were assessed using the COPD Assessment Test (https ://www.cates tonli ne.org) [21,22], a 6 min walk test was conducted [23], and spirometry was performed using the EasyOne Pro ® Spirometer (nDD Medical Technologies, Andover, MA, USA). See supplemental methods for further details of the spirometry testing.
In accordance with the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines, we classified participants as having airflow obstruction on the basis of a forced expiratory volume in 1 s (FEV 1 )/forced vital capacity (FVC) ratio less than 70% [22]. Restriction was defined as an FVC less than 80% of predicted, and participants were classified as having a mixed pattern of both obstruction and restriction if the FEV 1 /FVC ratio and the FVC were below 70% and 80% predicted, respectively. 18 F-fluorodeoxyglucose (FDG)-positron emission tomography/computed tomography (PET/CT) scans were done at baseline on the first 50 participants who agreed to participate in a radiology sub-study. Data regarding the PET findings are reported elsewhere [18]. For this analysis, the CT images from these scans were reviewed for evidence of lung damage, including bronchiectasis, bronchiolitis, and cavitation, using a score for mycobacterial infections developed by Song et al. with possible CT scores ranging from zero (i.e., no abnormal findings) to a maximum of 44. See supplemental methods and Additional file 1: Table S2 for further details of the PET/CT scans and their scoring [24].

Statistical analysis
We determined the relationship between clinical covariates and pulmonary impairment using bivariate analyses and logistic regression. Categorical and continuous variables were compared using a chi-square and Kruskal-Wallis test, respectively, and Pearson coefficients were calculated for correlations between continuous variables. Cohen's kappa coefficient was used to determine agreement between spirometry diagnoses at different time points. Logistic regression models were created to estimate (1) the association between participant characteristics and impaired lung function at baseline and (2) the association between participant characteristics at baseline (including baseline lung function) and impaired lung function at 12 months. All variables included in the models were baseline variables, with the exception of the lung function data and data on the CD4 cell count and viral load (which were characterized both at baseline and as the change from baseline to 4 weeks). Variables were included in the multivariate models on the basis of purposeful selection of clinical covariates, bivariate association, and directed acyclic graphs [25]. Analyses were conducted in SAS version 9.4 (SAS Institute, Cary, NC) and R version 3.6.0.

Ethics
All participants provided written informed consent. The Institutional Review Boards of the University of Pennsylvania, the University of Witwatersrand, and the Health Research Ethics Counsel in South Africa approved this study.

Lung function is dynamic over the course of tuberculosis treatment
There were 93 participants who completed tuberculosis treatment and underwent spirometry testing at six months, 85 (91%) of whom were able to complete satisfactory testing. At that time, there were 51 (60%) participants with normal spirometry, 14 (16%) with obstruction, 17 (20%) with restriction, and three (4%) with a mixed pattern. Six months later, 12 months after diagnosis, spirometry was performed for 95 participants, 92 (97%) of whom completed satisfactory testing; 62 (67%) had normal pulmonary function, 10 (11%) had obstruction, 14 (15%) had restriction, and 6 (7%) had a mixed pattern. Among the 10 participants with an obstructive pattern, the median FEV 1 % predicted was 76.5 (IQR 64-85) and only 3 (30%) participants had an FEV 1 less than 70% of predicted. Among 81 participants with spirometry at both baseline and 12 months, we examined the changes in spirometric patterns over time (Fig. 2). We found that baseline spirometric diagnosis had poor agreement with 12-month diagnosis (κ = 0.26, p value < 0.001) and 6-month diagnosis had only moderate agreement with 12-month diagnosis (κ = 0.50, p value < 0.001). For 44 participants with a normal baseline pattern, 35 (80%) remained normal at 12 months. Among 10 participants with obstruction at baseline, 7 (70%) had normalized at 12 months, while among 22 participants with restriction at baseline, only 12 (55%) had normalized at 12 months.
Over the course of the study, six participants met criteria for TB-IRIS [27]; one participant had worsening dyspnea in the setting of a new pleural effusion and five participants had worsening lymphadenopathy in either the cervical or abdominal regions. There was no apparent association between lung function and the development of clinical IRIS in this limited number of participants. No other opportunistic infections were identified using routine clinical screening.

Table 3 Radiologic abnormalities on CT scan according to the pattern of baseline pulmonary function
CT computed tomography, CT score range = 0-44 All categories of impaired lung function (i.e., obstruction, restriction, and mixed) were compared to normal and an asterisk (*) indicates a p value < 0.05 as compared to normal. All values are represented as median and interquartile range unless otherwise indicated Cavities Pleural effusion 0 (0-0) 0 (0-0) 0 (0-1)* 0 (0-1)

Pulmonary symptoms are uncommon after tuberculosis treatment
Only 12 (13%) participants reported symptoms at 12 months, including seven (58%) participants with normal spirometry at that time, three (25%) with restriction, and two (17%) with a mixed pattern. Although participants with restriction reported more symptoms than those with obstruction or normal spirometry, this difference was not statistically significant (p value 0.10; Additional file 1: Table S4). There was also no significant difference in 6 min walk test distance for those with and without normal pulmonary function at 12 months.

Pulmonary restriction at baseline is associated with long-term pulmonary impairment
In bivariate logistic regression analyses, having either a restrictive or mixed pattern of pulmonary impairment at baseline was associated with having abnormal pulmonary function at 12 months (OR for baseline restriction 3.24, 95% 1.06-9.87; OR for baseline mixed 5.83, 95% CI 0.84-40.32) ( Table 5). After adjustment for age, gender, change in CD4 count from baseline to week 4, and time from TB treatment to ART initiation, baseline restriction and mixed pattern remained significantly associated with impaired pulmonary function at 12 months (aOR for baseline restriction 5.44, 95% CI 1. 16-25.45; aOR for baseline mixed 9.00, 95% CI 1.05-76.87).
In a second set of models that assessed the odds of having either restriction or obstruction at 12 months, impaired pulmonary function at baseline was strongly associated with restriction, but not obstruction at 12 months (OR for obstruction 1.11, 95% CI 0.28-4.40; OR for restriction 16.67, 95% CI 1.99-139.64) (Additional file 1: Table S6). Multivariate models for modeling restriction or obstruction at 12 months were unable to converge on a maximum likelihood estimate. Baseline symptoms were not associated with impaired pulmonary function at 12 months in any of the models. remained normal at 12 months, while 6 (14%) developed incident obstruction, 1 (2%) developed restriction, and 2 (5%) developed a mixed pattern. For 10 participants with obstruction at baseline, 7 (70%) had normalized at 12 months, while 3 (30%) had persistent obstruction. For 22 participants with restriction at baseline, 12 (55%) had normalized at 12 months, while 8 (36%) had persistent restriction, 1 (5%) had developed obstruction in addition to restriction (i.e., a mixed pattern), and 1 (5%) had obstruction alone. For 5 participants who had a mixed pattern both at baseline, 2 (40%) had normalized at 12 months, 2 (40%) had persistent restriction, and 1 (20%) had a persistent mixed pattern

Discussion
In this study we found that half of individuals with tuberculosis and HIV coinfection had impaired pulmonary function at the time of tuberculosis diagnosis and onethird had impaired pulmonary function 12 months later, 6 months after finishing their tuberculosis treatment. At every time point, a restrictive pattern was more common than obstruction. Impaired pulmonary function at baseline, specifically in a restrictive pattern, was significantly associated with greater pulmonary symptoms. Furthermore, pulmonary restriction at baseline was associated with persistently impaired pulmonary function at 12 months, when lung impairment may be permanent.
Our finding that restriction was the most commonly identified spirometric abnormality is similar to what has been reported from a cohort in India, where 52% of individuals had restriction and 24% had obstruction [10]. In contrast to India where obstruction was more symptomatic than restriction, we found that individuals with restriction were more symptomatic than those with obstruction alone. However, HIV prevalence in the cohort from India was only 4% and so it is possible that differences in the HIV coinfection rates may explain some of the differences in our findings. Nevertheless, greater symptoms among the patients with restriction may relate to the distribution and amount of lung involvement at presentation, as those with restriction also had greater radiologic involvement than those with normal spirometry or obstruction. While bronchiectasis was associated with both obstruction and restriction, a number of additional radiologic abnormalities, including bronchiolitis, consolidation, and mosaic perfusion, were significantly more common among those with restriction. In the context of tuberculosis, these findings suggest that airway disease is common, but only a subset of individuals with sufficient parenchymal involvement may be at risk for developing a restrictive ventilatory defect.
While restriction has been reported among tuberculosis survivors without HIV, people with HIV may have a greater propensity to develop parenchymal involvement and, by extension, restrictive lung disease. Higher rates of emphysema and pulmonary fibrosis in people with HIV have been ascribed to the deleterious effects of virally mediated chronic inflammation and oxidative stress [28,29]. Hence, it is possible that those same perturbations increase the risk of post-tuberculosis pulmonary fibrosis. People with HIV also have a relative CD8 cell alveolitis, which may further skew the pulmonary immune response towards fibrosis [30][31][32]. Impaired lung function at baseline, specifically a restrictive or mixed pattern, was a strong predictor for impaired lung function at 12 months. However, 20% of those with normal spirometry at baseline went on to develop incident pulmonary impairment at 12 months. These data suggest that there may be two phases of lung damage in patients with HIV and tuberculosis-a first phase prior to diagnosis and a second phase after treatment initiation, that may represent a pulmonary-specific manifestation of immune reconstitution inflammatory syndrome (IRIS) [17]. Interventions to prevent lung injury may differ according to these phases. Lung damage that occurs prior to diagnosis will best be prevented by continuing to emphasize early diagnosis and active case finding, whereas lung damage that occurs after treatment initiation may be amenable to intervention with antiinflammatory agents or other host-directed therapies. In light of the predominance of restrictive lung disease in this cohort, anti-fibrotic agents, which are increasingly utilized for interstitial lung diseases, represent a promising avenue for preventing pulmonary impairment after tuberculosis [33,34].
Given that respiratory symptoms are often one of the only indications that someone may have decreased lung function, we were surprised to find that few participants reported symptoms at 12 months, irrespective of whether they had impaired pulmonary function. It is possible that these participants may have minimized their pulmonary symptoms, particularly after having been treated for tuberculosis and initiating ART. Another possibility is that these individuals, who were generally quite young, may be able to compensate for their impaired lung function but will become more symptomatic as they age. Yet another consideration is that the COPD Assessment Test is not a sensitive tool for identifying pulmonary symptoms related to tuberculosis. Future studies could assess other measures of exercise capacity, such as the incremental shuttle walk test, among tuberculosis survivors [35].
There are several limitations to this study. Firstly, spirometry is not accurate for diagnosing restrictive lung physiology, with a positive predictive value of 58-64% when compared to the standard of plethysmography [36,37]. However, equipment for plethysmography is relatively expensive and thus is difficult to obtain in resource-limited settings. Lung volume measurements with nitrogen washout or single-breath helium dilution present alternatives that are more cost-effective but may underestimate lung volumes and still require additional equipment [38]. Spirometry is also effort dependent, and so improvements in testing over the course of treatment for tuberculosis and HIV could reflect improvements in overall health, and not simply improvements in lung function. A second limitation is the use of low-dose, thick-section CT scans to assess for radiologic involvement. While these images do not provide the same resolution of pulmonary anatomy as high-resolution diagnostic scans, they do offer substantially greater detail than chest radiographs. Third, we cannot be certain that the observed pulmonary dysfunction was solely due to tuberculosis, as this is an immunocompromised cohort at risk for concurrent infections and because pre-tuberculosis spirometry data were not available. Finally, while we were able to follow participants for up to 6 months after completion of tuberculosis treatment, given the flux in spirometric diagnoses between 6 and 12 months, there could be ongoing pulmonary remodeling even after 12 months.

Conclusions
In light of growing global calls to better understand pulmonary impairment after tuberculosis, our findings from this prospective study make an important contribution to the literature [39]. We found that pulmonary impairment is common in people with HIV coinfection who are successfully treated for pulmonary tuberculosis, even up to 6 months after treatment completion. Further, restriction is not only the most common pattern on spirometry, but is also the pattern most commonly associated with respiratory symptoms and most predictive of long-term pulmonary impairment. While the participants enrolled in this study had advanced HIV disease with a median CD4 count of 110 cells/μL, they also represent a young cohort with minimal smoking history who would otherwise be expected to have little underlying pulmonary disease. Our finding that one-third of these individuals have impaired lung function 6 months after successful completion of tuberculosis treatment is highly concerning, as these impairments will likely impact their quality of life for decades to come. There is an urgent need to better understand pulmonary impairment after tuberculosis-from risk factors and predictors of lung injury to biological pathways that may be amenable to pharmacologic intervention-to lessen the burden of chronic lung impairment in this vulnerable population.
Additional file 1. Additional file contains supplementary methods regarding the spirometry testing and additional Tables as referenced in the manuscript.