Prediction of Post-Tuberculosis Lung Function Impairment
Prediction of Post-Tuberculosis Lung Function Impairment
Prediction of Post-Tuberculosis Lung Function Impairment
Email address:
*
Corresponding author
Received: December 9, 2018; Accepted: December 25, 2018; Published: January 18, 2019
Abstract: Objective: The aim of this study was to establish prediction equations for post-tuberculosis residual lung function
in patients successfully treated for pulmonary tuberculosis (PTB). Methods: This study took place at the Yaounde Jamot
Hospital of Yaounde (YJH) and used data from three cross-sectional studies conducted from January to July 2015 (7 months),
December 2015 to May 2016 (6 months) and from January to May 2017 (5 months). Adults successful treated for
bacteriologically proven pulmonary TB were included. Spirometric indices including forced expiratory volume in 1s (FEV1),
forced vital capacity (FVC) and FEV1/FVC ratio were measured using standard methods. Predicted values were estimated
using the reference spirometric equations of the Global Lung Initiative equations (GLI) 2012. General linear models were used
to establish prediction equations of post-tuberculous residual lung function. Internal validation of the derived models used the
bootstrap resampling procedures. A difference was considered significant if p < 5%. Results: In this study, 400 patients (53.5%
men) were included. The median age (25th -75th percentiles) of men was 40 (31-50) years and that of women was 36(27.8-46)
years (p=0.002). Determinants of the post-tuberculosis spirometric indices vary according to each indice and include age,
weight, height, body mass index, smoking, duration of symptoms before TB treatment, persistent of respiratory symptoms after
TB treatment, persistent of cavity lesions and extension of lung sequelae. The prediction equations of the spirometric indices
have been established separately for men and women to account for significant differences in the absolute values of spirometric
parameters in men and women. The prediction equations of residual lung function parameters were in the form: lung function
parameters = Intercept + β1*P1 + β2*P2 +…βn*Pn; βn is regression coefficient for corresponding predictor (Pn), for categorical
variables Pn is 1 if the modality is present and 0 if the modality is absent. For each of the spirometric variable, differences in
performance measures (optimism) were mostly marginal. Conclusion: The equations developed and validated in this study
could help the selection of patients in whom spirometry should be a priority after TB treatment. Like any newly developed
model, results from this study are just preliminary findings. Models will require independent external validation to establish the
performance both in the study setting and in other settings.
171 Eric Walter Pefura-Yone et al.: Prediction of Post-Tuberculosis Lung Function Impairment
Keywords: Tuberculosis, Residual Lung Function, Spirometry, Sequelae, Obstructive Lung Disease, Restrictive Pattern
heels joined together and arms along the body. Body mass references value [20]. Quality control was done by regularly
index in kg/m2 was calculated as the ratio of weight (kg) to supervising investigators who performed spirometry.
height squared (squared meter). Spirometric curves were reviewed weekly by one of the
Standard full-size postero-anterior chest X-ray (CXR) was experienced chest physicians working in YJH and feedback
used to quantify the extent of lung remodeling in patients. was made to investigators.
The nature, distribution, and extent of abnormalities were
noted for initial and final CXR. Type of lesions collected 2.3. Data Analysis
include: cavities, infiltrates, nodules, fibrotic lesions, miliary The data was analyzed using IBM-SPSS software for
lesions, patchy or confluent consolidation and associated Windows version 23 (IBM, Chicago, US). Categorical
pleural effusion. Extension of radiographic lesions was variables were expressed as counts and proportions.
appreciated by using total percentage of lung affected by any Quantitative variables were summarized in terms of mean
pathology on the initial and final chest X-ray (at the end of (standard deviation) when the distribution was considered
treatment). To grade the percentage of affected lung, visual normal; otherwise they were expressed in terms of median
estimation of the extent of opacification, cavitation or other (25th -75th percentiles). The groups comparison used Chi-2 or
pathology as a percentage of visible lungs was made. To Fisher exact test for proportions and Student t-test or non-
evaluate this, each lung field was subdivided into three zones parametric equivalent for continuous variables (Mann-
(upper, middle and lower zone) according to the method Whitney U test or median test). Bivariate Pearson’s or
employed by Ralph et al [18]. The delimitation was done by Spearman’s correlation tests and matrix graph were used to
two horizontal lines: the first is the horizontal line passing investigate the shape and significance of the associations of
through the upper border of the chondro-costal junction of quantitative candidate predictors with each of the spirometric
the anterior arch of the second rib to the lateral thoracic wall; indices (FEV1, FVC, FEV1/FVC ratio, % predicted FEV1
the second is the horizontal line passing through the upper and % predicted FVC). General linear models were used to
border of the chondro-costal junction of the anterior arch of establish prediction equations of post-tuberculous spirometric
the 4th rib to the lateral thoracic wall. Dense opacification of parameters.
an entire zone was graded as 100% of that zone, while patchy
consolidation within a zone attracted scores <100% 2.3.1. Prediction Models Development
depending on the extent of consolidation. An average Significant predictors in bivariate association (p-
percentage of lesions per lung field was then evaluated and value<0.10 for the correlation test and for association
an average of the two lung fields deduced to determine the between categorical variables and spirometric parameters)
total percentage of affected lung (% of pulmonary were entered all together in the same multivariable general
tuberculosis lung sequelae). linear models and backward elimination procedures were
used to retain the most significant predictors in the final
2.2.2. Spirometric Data model (based on a p-value<0.05). General linear models were
Spirometry was performed on eligible participants as per used to relate each of the five targeted spirometric variables
standard [19], using a turbine pneumotachograph (spiro USB, with potential predictors, separately in men and women.
care fusion, Yorba Linda-USA), thereby respecting the Candidate continuous predictors included age, weight,
American Thoracic Society (ATS) 1994 standards. The exam height which are all known to be related with the spirometric
was done by pulmonary function test (PFT) technicians or by indices, duration of symptoms and % of lung sequelae.
well-trained final year medical student. Candidate categorical variables included residency, level of
All measurements were performed after at least 15 min formal education, marital status, smoking, persistent of
rest, with the participant in a seated position, the back respiratory symptoms, persistent of cavity lesions.
straight up, and nose clipped to allow air flow only by mouth. Appropriate procedures were used to check the adequacy of
The American Thoracic Society/European Respiratory the final models. These procedures included residuals
Society (ATS/ERS) acceptability and reproducibility criteria analysis (plots analysis), Cook statistics and multicolinearity
were applied [19]. Instructions were explained before every detection (inflation factor of variance).
maneuver; at least three tests were done per participant with
the maximum being eight tests, while observing a rest period 2.3.2. Validation of Prediction Models
of at least one minute between consecutive tests to establish Internal validation of the derived models used the
the forced vital capacity (FVC) curve. Spirometry variables bootstrap resampling procedures, which was based on 2000
measured included: forced expiratory volume in 1s (FEV1), replications. In brief, the population used to derive models
forced vital capacity (FVC) and the FEV1/FVC ratio. was resampled with replacement 2000 times, each sample
FEV1 and FVC values retained were the best out of the being of the same size and sex composition as the original
three tests which fulfilled the acceptability criteria (maximal population. From each bootstrap sample, new sex-specific
difference below 5% or 150 ml). All these measurements models were developed using predictors in the final model.
were expressed in absolute values and as a percentage of the The regression coefficients from each bootstrap sample were
theoretical or predicted value. Predicted or theoretical values then applied both to the relevant bootstrap sample and the
were estimated using Global Lung Initiative (GLI) 2012 original population to predict the spirometric indices of
173 Eric Walter Pefura-Yone et al.: Prediction of Post-Tuberculosis Lung Function Impairment
interest. The difference in the performance measures of the performance on the derivation sample that should be
model on the bootstrap sample and the original population expected when the model is tested on a new sample.
was then recorded and averaged across the 2000 replicates to On the basis of these prediction equations, it will be
characterize the optimism. A prediction model would always possible to know if a given patient has a high probability of
tend to perform better when it is tested on the same sample presenting a PTB-LFI, by comparing the value of the
used to develop the model (self-fulfilling prophecy), as spirometric parameter found with the prediction equations
opposed to when the same model is tested on a new sample. with the theoretical threshold (LLN or fixed threshold).
Optimism reflects the drop in the observed models’
3. Results
3.1. Study Population
A total of 513 patients were invited to participate in the men than in women (35% vs 4.3%, p<0.001). Regular
study. Seventy six patients were excluded for unavailability alcohol consumption was also more prevalent among men
of lung function testing and 37 patients for unusable than women (29.1% vs 4.3%, p<0.001). The previous history
spirometry curves. In overall, 400 participants were of TB was found in 25 (11.7%) men and in 11 (5.9%) women
definitively included in the study. The flow chart of (p=0.044). The median (25th -75th percentiles) duration of
participant’s inclusion is shown on the figure 1. symptoms before TB treatment was 8 (4-16) weeks in total
group. The prevalence of HIV infection was 35.5% for
3.2. Baseline Characteristics of Study Participants women and 31.6% for men (p = 0.081). In overall sample,
The baseline characteristics of participants are shown in 70.5% of the participants had a least one type of radiographic
Table 1. In this study, 214 (53.5%) men and 186 (46.5%) sequelae of TB. Persistent cavity lesions and parenchymal
women were included. The median age (25th -75th lung destruction were respectively found in 21.3% and 3.5%
percentiles) of men was 40 (31-50) years and that of women of participants. There was no significant difference between
was 36 (27.8-46) years (p=0.002). The majority of men and women concerning the frequency and type of
participants had reached the secondary school level of radiographic sequelae of TB. The median (25th -75th
education (274 participants, 68.5%). In this study, 100 (25%) percentiles) proportion of extension of pulmonary TB
participants were active smokers and 70 (17.6%) consumed sequelae was 5% (0.8-15.8%) in total sample with no
alcohol regularly. The prevalence of smoking was higher in significant difference between men and women.
American Journal of Internal Medicine 2018; 6(6): 170-181 174
Characteristics Overall N=400 (%) Men N=214 (%) Women N=186 (%) p-value
Age, years 38(29-49) 40(31-50) 36 (27.8-46)
Residency
Urban 314 (78.5) 167 (78) 147 (79) 0.967
Semi-urban 49 (12.3) 20(9.3) 17 (9.1)
Rural 37 (9.3) 27 (12.6) 22 (11.8)
Level of education
No formal education 19 (4.8) 7 (3.3) 12 (6.5) 0.143
Primary 107 (26.8) 51 (23.8) 56 (30.1)
Secondary 198 (49.5) 110 (51.4) 88 (47.3)
University 76 (19) 46 (21.5) 30 (16.1)
Marital Status
Non-unionized 226/398 (56.8) 112 (52.6) 114 (61.6) 0.069
Unionized 172/398 (43.2) 101 (47.4) 71 (38.4)
Smoking
Current 83 (20.8) 75 (35) 8 (4.3) <0.001
Ex-smokers 14 (3.5) 12 (5.6) 2 (1.1)
Non smokers 303 (75.8) 127 (59.3) 176 (94.6)
Biomass exposure 148/397 (37.3) 82/212 (38.7) 66/185 (35.7) 0.537
Alcohol consumption
Regular 70/398 (17.6) 62 (29.1) 8 (4.3) <0.001
Occasional 124/398 (31.2) 71 (33.3) 53 (28.6)
Ex-drinker 87/398 (21.9) 45 (21.1) 42 (22.7)
Non drinker 117/398 (29.4) 35 (16.4) 82 (44.3)
Previous history of TB 36(9) 25(11.7) 11(5.9) 0.044
Past history of pneumonia 5(1.3) 0(0) 5(2.7) 0.021
Duration of respiratory symptoms, weeks, m (25th -75th percentiles) 8(4-16) 8(4-12) 8(4-16) 0.017
Persistent respiratory symptoms 111(27.8) 56(26.2) 55(29.6) 0.449
Weight, kg, m (25th -75th percentiles) 62(56-69.8) 65(59.6-72) 58(54-64) <0.001
Height, cm, m (25th -75th percentiles) 168(163-173.8) 168(164-178) 160(156-167) <0.001
BMI, Kg/m2 , m (25th -75th percentiles) 22 (20.2-24.1) 22.1(20.4-24.1) 220(20.2-24.2) 0.836
HIV infection 141/397 (35.5) 67(31.6) 141(35.5) 0.081
Radiographic sequelae
No sequelae 118 (29.5) 58 (27.1) 60 (32.3) 0.259
Significant pleural thickness 11 (2.8) 8 (3.7) 3 (1.6) 0.195
Fibro-nodular sequelae 212 (53) 118 (55.1) 94 (50.5) 0.358
Persistent cavity lesions 85 (21.3) 46 (21.5) 39 (21) 0.898
Parenchymal destruction 14 (3.5) 8 (3.7) 6 (3.2) 0.781
Extension of pulmonary sequelae, %, m (25th -75th percentiles) 5 (0.8-15.8) 5.81 (1-16.2) 5 (0.7-15.3) 0.306
TB, tuberculosis; m, median; BMI, body mass index; HIV, human immunodeficiency virus; PA, pack-years; BMI, body mass index; HIV, human
immunodeficiency virus.
FEV1, forced expiratory volume in 1st second; FVC, forced vital capacity; m, median, M, mean, SD, standard deviation.
175 Eric Walter Pefura-Yone et al.: Prediction of Post-Tuberculosis Lung Function Impairment
FEV1, forced expiratory volume in 1 s FVC, forced vital capacity; TB, tuberculosis, RSD, residual standard deviation; R2, determination coefficient,
Parameters estimation = Intercept + β1*P1 + β2*P2 +…βn*Pn, βn, regression coefficient for corresponding predictor (Pn)
FEV1, forced expiratory volume in 1 s FVC, forced vital capacity; TB, tuberculosis, RSD, residual standard deviation; R2, determination coefficient,
Parameters estimation = Intercept + β1*P1 + β2*P2 +…βn*Pn,
βn, regression coefficient for corresponding predictor (Pn)
Table 5. Optimism in bootstrap internal validation of prediction models in men and women.
from 44 to 100 kg for weight and 140 to 183 cm for height. The impairment: restrictive pattern, airflow limitation or
estimation of post-tuberculosis spirometric indices should take obstructive lung disease, and the mixed pattern associating
into account the limits of validity of the anthropometric restrictive pattern and obstructive lung disease [32]. The
parameters. In fact, the normal spirometric indices depend on pathogenic and pathophysiology mechanisms explaining these
these parameters [20–24]. These equations should not be applied alterations are not completely understood. Anatomically, the
to people who do not have the characteristics of the population restrictive pattern is due to pulmonary fibrotic lesions, loss of
studied, in the absence of validation of these equations in these parenchymal elasticity, broncho-vascular distortion and pleural
groups. Application outside populations with similar sequelae. Obstructive post-tuberculous lung disease may be
characteristics to those used for development may be misleading. related to bronchial stenosis, diffuse bronchial inflammation,
The absolute values of FEV1 and FVC found in this study para-cicatricial emphysema, or bronchiectasis [32].
are higher in men than in women. These findings corroborate Progress has been made in understanding the inflammatory
the variation of these indices when they are obtained in mechanisms that are likely responsible for tissue damage and
healthy subjects. Indeed, in black, Caucasian, Asian or mixed impairment of lung function. Inflammatory mediators involved
populations, the normative absolute values of FEV1 and FVC in cavity formation and airflow obstruction are tumor necrosis
are higher in men than in women even if the other parameters factor-α (TNF-α), interleukins 1β, 6, 8, 12, hypoxia inducible
(age, height) are otherwise identical [25, 26]. In contrast, the factor (HIF), nuclear factor (NF) , matrix metalloproteinases 1,
mean FEV1/FVC ratio was 0.80 in men and 0.83 in women 3, 8, 9, 12 and interferon γ. The immune factors associated with
with a significant difference. This difference is explained by fibrosis and restriction include transforming growth factor β
a larger decline in FVC compared to FEV1 in men. (TGF-β), TNF-α, and interleukins-1 β, 6, 8, 12 [32].
The other main predictors of each spirometric indice are The equations developed for FEV1 and FVC can be used
variable and include: smoking, clinical and radiographic data. in two ways. The first way is to determine the residual
Tobacco use was associated with a lower FEV1 and absolute value of FEV1 and FVC and then compare these
FEV1/FVC for men, as well as a lower FVC as a percentage values to normative values using specific prediction
of predicted value (% predicted FVC). In fact, smoking has a equations. This method can be used regardless of the set of
known deleterious effect on the decline of lung function and reference equations chosen. The second way is to use
is the main risk factor for the occurrence of chronic equations based on the predicted values (% predicted) to
obstructive pulmonary disease, COPD [27]. In a study directly obtain FEV1 and FVC values based on the
conducted in a low-smoking Indian women population theoretical values for blacks according to the 2012 GLI
(smoking prevalence at 5% and 73.6% of participants reference equations [20]. For the FEV1/FVC ratio, the
cumulating less than 2.5 pack-years), smoking was absolute value found from the prediction equations will be
associated with lower FEV1 [28]. compared to the lower limit of the expected normal value to
A longer duration of respiratory symptoms before TB define post-tuberculous airflow obstruction.
treatment was associated with a lower spirometric parameters For example, considering a 30-year-old smoking man with
in men but not in women. Indeed, a longer duration of the following characteristics after treatment of PTB: duration
symptoms of PTB is associated with larger radiographic of symptoms= 9 weeks, no persistence of symptoms after
lesions, sources of pulmonary sequelae [29]. The persistence treatment, height = 170 cm, weight = 60 kg, persistence of a
of respiratory symptoms at the end of TB treatment was also pulmonary cavity, extension of pulmonary sequelae = 10%.
associated with a decrease in FEV 1 and FVC. This persistence His residual post-tuberculosis FEV1 will be 2.42 L. Using
of respiratory symptoms should prompt an investigation of the 2012 GLI equations; the normal theoretical FEV1 value
impaired respiratory function. Persistence of respiratory signs will be 3.47 L. He has a predicted post-tuberculosis residual
in patients treated for PTB is associated with distal obstruction, FEV1 of 70% (2.42 L/3.47 L). Similarly, the other values are:
indicating distal lung function impairment in subjects with no post-tuberculous FVC = 3.22 L, % predicted FVC post-
proximal airflow limitation [30]. tuberculosis = 78% (3.22 L/4.13 L), FEV1/FVC = 0.75
In this study, the radiographic extension of sequelae of TB (compared to the predicted value of 0.84). This patient should
summarized in terms of the proportion of radiographic perform lung function test to investigate possible ventilatory
sequelae is negatively correlated with all spirometric abnormalities, in particular the airflow obstruction given a
parameters in men and women with correlation coefficients low value of post-tuberculosis FEV1/FVC ratio.
ranging from -0.173 to - 0.608. Few studies have investigated For a non-smoking 40 years old female patient, with
the association between the extent of initial radiographic symptoms of pulmonary tuberculosis lasting for 16 weeks,
lesions or radiographic sequelae and impaired lung function. persistence of respiratory signs at the end of tuberculosis
Radovic et al. on a small sample size (n = 40 patients) treatment, height = 1.60 m, weight = 60 kg, absence of residual
showed that initial radiographic extension is associated with cavity, extension of pulmonary sequelae = 30%: post-
low FEV1 and post-tuberculous airflow obstruction [31]. In tuberculosis FEV1 = 1.53L, post-tuberculosis FVC = 2.05 L,
another study, including 591 patients in 21 hospitals in South FEV1/FVC = 0.81. Using the 2012 GLI standards, the following
Korea, the number of pulmonary lobes affected by TB values are obtained: % predicted post-tuberculosis FEV1 =
sequelae was negatively correlated with FEV1 and FVC [31]. 61%, % predicted post-tuberculosis FVC = 67%. The predicted
There are three types of post-tuberculosis lung function values of FEV 1 and FVC found in this patient would encourage
American Journal of Internal Medicine 2018; 6(6): 170-181 178
clinicians to perform spirometry to investigate possible function have been established in this study. Determinants of
restrictive pattern. In this case, FEV1/FVC ratio approaches the the post-tuberculosis spirometric indices vary according to
expected value (0.83) and airflow obstruction is unlikely the each indice and include age, weight, height, body mass
main type of lung function impairment. index, smoking, duration of symptoms before TB treatment,
Overall, the coefficient of determination (R2) is higher for persistent of respiratory symptoms after TB treatment,
FEV1 and FVC than for FEV1/FVC ratio. Thus, the equations persistent of cavity lesions and extension of lung sequelae.
developed in this work are more robust for FEV 1 and FVC. The equations developed and validated in this study could
Obtaining an abnormal FEV1 or FVC value regardless of the help the selection of patients in whom spirometry should be a
value of the FEV1/ FVC ratio should prompt clinicians to priority after TB treatment. The models had good
request functional lung investigations for their patients. performance in the derivation sample and during internal
The boostrap method was used to validate the prediction validation using bootstrap resampling methods.
equations. The boostrap validation method is validated and These prediction equations can be used to select patients in
allows the use of the entire sample available for whom PFT should be a priority. Successful application of the
development. The validation is done by a multiple replication models in routine TB care settings can assist the
of random choice of subsamples (n = 2000) to check the identification of a group of patients whose risk of
stability of the developed model. A weak and non-significant development of lung function impairment is high. Therefore
optimism makes it possible to conclude on the stability of the corrective measures will be taken to treat the observed
model in spite of the multiple replications [33]. deteriorations and improve the quality of life of these
patients. Like any newly developed model, results from this
5. Conclusions study are just preliminary findings. Models will require
independent external validation to establish the performance
The prediction equations for post-tuberculous residual lung both in the study setting and in other settings including.
Appendix
Supplementary Tables (Prediction of Post-Tuberculosis Lung Function Impairment)
Table A1. Correlation coefficients (r) and p-values between spirometric parameters, and age/anthropometric predictors.
r, correlation coefficient; FEV1, forced expiratory volume in 1 s FVC, forced vital capacity.
Table A2. Correlation coefficient (r) and p-values between spirometric parameters and continuous clinical and radiological predictors.
r, correlation coefficient; FEV1, forced expiratory volume in 1 s FVC, forced vital capacity; TB, tuberculosis.
179 Eric Walter Pefura-Yone et al.: Prediction of Post-Tuberculosis Lung Function Impairment
Table A3. Association between spirometric parameters and categorical predictors in men.
FEV1, forced expiratory volume in 1 s FVC, forced vital capacity; TB, tuberculosis.
Table A4. Association between spirometric parameters and categorical predictors in women.
FEV1, forced expiratory volume in 1 s FVC, forced vital capacity; TB, tuberculosis.
[16] Austin PC, Steyerberg EW. The number of subjects per variable [26] Hankinson JL, Odencrantz JR, Fedan KB. Spirometric
required in linear regression analyses. J. Clin. Epidemiol. Reference Values from a Sample of the General U. S.
2015;68:627–36. Population. Am. J. Respir. Crit. Care Med. 1999;159:179–87.
[17] Mahler DA, Wells CK. Evaluation of clinical methods for rating [27] Vijayan VK. Chronic obstructive pulmonary disease. Indian J.
dyspnea. Chest. 1988;93:580–6. Med. Res. 2013;137:251–69.
[18] Ralph AP, Ardian M, Wiguna A, Maguire GP, Becker NG, [28] Arora S, Rasania S, Bachani D, Gandhi A, Chhabra S. Air
Drogumuller G, et al. A simple, valid, numerical score for grading pollution and environmental risk factors for altered lung
chest x-ray severity in adult smear-positive pulmonary function among adult women of an urban slum area of Delhi:
tuberculosis. Thorax. 2010;65:863–9. A prevalence study. Lung India. 2018;35:193.
[19] Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, [29] Pefura-Yone EW, Kuaban C, Assamba-Mpom SA, Moifo B,
Coates A, et al. Standardisation of spirometry. Eur. Respir. J. Kengne AP. Derivation, validation and comparative
2005;26:319–38. performance of a simplified chest X-ray score for assessing
the severity and outcome of pulmonary tuberculosis. Clin.
[20] Quanjer PH, Stanojevic S, Cole TJ, Baur X, Hall GL, Culver BH, Respir. J. 2015;9:157–64.
et al. Multi-ethnic reference values for spirometry for the 3-95-yr
age range: the global lung function 2012 equations. Eur. Respir. J. [30] Pefura-Yone EW, Kengne AP, Tagne-Kamdem PE, Afane-Ze
2012;40:1324–43. E. Clinical significance of low forced expiratory flow between
25% and 75% of vital capacity following treated pulmonary
[21] Pellegrino R, Viegi G, Brusasco V, Crapo RO, Burgos F, tuberculosis: a cross-sectional study. BMJ Open.
Casaburi R, et al. Interpretative strategies for lung function tests. 2014;4:e005361.
Eur. Respir. J. 2005;26:948–68.
[31] Radovic M, Ristic L, Ciric Z, Dinic-Radovic V, Stankovic I,
[22] Stanojevic S, Wade A, Stocks J. Reference values for lung Pejcic T, et al. Changes in respiratory function impairment
function: past, present and future. Eur. Respir. J. 2010;36:12–9. following the treatment of severe pulmonary tuberculosis -
limitations for the underlying COPD detection. Int. J. Chron.
[23] Ratomaharo J, Linares Perdomo O, Collingridge DS, Obstruct. Pulmon. Dis. 2016;11:1307–16.
Andriamihaja R, Hegewald M, Jensen RL, et al. Spirometric
reference values for Malagasy adults aged 18-73 years. Eur. [32] Ravimohan S, Kornfeld H, Weissman D, Bisson GP.
Respir. J. 2015;45:1046–54. Tuberculosis and lung damage: from epidemiology to
pathophysiology. Eur. Respir. Rev. 2018;27.
[24] Pefura-Yone EW, Kanko-Nguekam NF, Kengne AP, Balkissou
AD, Kuaban C. Équations spirométriques de référence des [33] Harrell FE. Regression Modeling Strategies. 1995. Available
Bantous camerounais. Rev. Mal. Respir. from: http://biostat.mc.vanderbilt.edu/tmp/course.pdf.
Accessed September 5, 2018.
[25] Bougrida M, Ben Saad H, Kheireddinne Bourahli M,
Bougmiza I, Mehdioui H. [Spirometric reference equations for
Algerians aged 19 to 73 years]. Rev. Mal. Respir.
2008;25:577–90.