Smoking Cessation in Tuberculosis Patients and the Risk of Tuberculosis Infection in Child Household Contacts

Abstract

Background

While previous studies have shown that cigarette smoking increases the infectiousness of tuberculosis patients, the impact of smoking cessation on tuberculosis transmissibility has not been evaluated.

Methods

Between 2009 and 2012, we enrolled 4500 tuberculosis patients and followed 14 044 household contacts in Lima, Peru. Tuberculosis patients were classified into 4 categories: never smoked, quit in the distant past (stopped smoking >2 months prior to time of diagnosis), recently quit (stopped smoking ≤2 months prior to time of diagnosis), and active smokers. We used a modified Poisson generalized estimating equation to assess the risk of tuberculosis infection of child contacts at enrollment and by 6 months of follow-up.

Results

In total, 1371 (76.8%) child contacts were exposed to patients who had never smoked, 211 (11.8%) were exposed to distant quitters, 155 (8.7%) were exposed to recent quitters, and 49 (2.7%) were exposed to active smokers. Compared with child contacts of index patients who had never smoked, child contacts of recent quitters had a similar risk of tuberculosis infection at enrollment (adjusted risk ratio, 95% confidence intervals [0.81, 0.50–1.32]) and by six months of follow-up (0.76, 0.51–1.13); and by 6 months of follow-up (aRR, 0.76; 95% CI, .51–1.13); child contacts of recent quitters had a significantly reduced risk of tuberculosis infection compared with contacts of active smokers (enrollment 0.45, 0.24–0.87; 6-month follow-up 0.48, 0.29–0.79).

Conclusions

Our results show that the adverse effects of smoking on the transmissibility of tuberculosis are significantly reduced shortly after quitting smoking, reinforcing the importance of smoking cessation interventions in tuberculosis control.

We report that child household contacts exposed to tuberculosis patients who had quit smoking shortly before time of diagnosis had a reduced risk of tuberculosis infection compared with contacts of tuberculosis patients who continued to smoke in Lima, Peru.

Tobacco consumption, particularly cigarette smoking, threatens ongoing and future global tuberculosis control efforts. Recent estimates indicate that 18% of tuberculosis cases and 15% of tuberculosis mortality can be attributed to tobacco smoking in countries with a high burden of tuberculosis [1]. Despite declining global estimates of smoking prevalence over the past 2 decades, the World Health Organization estimated the global tobacco smoking prevalence among adults to be 19% in 2018 [2]. Previous studies have shown that active cigarette smoking in tuberculosis patients is associated with an increased risk of transmitting tuberculosis to close contacts [3–5] and of poor treatment outcomes [6, 7]. These findings have led researchers to call for the incorporation and scaling up of smoking cessation interventions in national tuberculosis programs [8–10]. However, little is known about the effect of smoking cessation in tuberculosis patients on the transmissibility of tuberculosis. Here, we assessed the association between different smoking histories in pulmonary tuberculosis patients and the risk of tuberculosis infection in child household contacts in Lima, Peru.

METHODS

Study Setting and Design

We conducted this study in a metropolitan area of Lima, Peru, consisting of 20 districts and approximately 3.3 million residents living in urban areas and peri-urban, informal shantytown settlements. The study design, methods, and setting have been previously described in detail [15]. In brief, between September 2009 and August 2012, we identified and invited patients aged ≥16 years with newly diagnosed pulmonary tuberculosis as index patients. We confirmed the microbiological status of their tuberculosis disease using either sputum smear or mycobacterial culture. At the time of enrollment, we collected index patients’ characteristics, including sociodemographic information, number of rooms in the household, history of previous tuberculosis disease, tobacco and alcohol history, duration of tuberculosis symptoms, presence of cavitary lesions, and comorbidities including diabetes mellitus and human immunodeficiency virus (HIV) status. We enrolled all consenting household contacts within 2 weeks following the enrollment of an index patient. We assessed baseline tuberculosis infection status using the tuberculin skin test (TST) in household contacts without a prior positive TST result or history of tuberculosis disease. Household contacts identified with signs or symptoms of tuberculosis disease were referred to their local health center where they underwent further clinical evaluation, including undergoing a chest radiograph and sputum smear test. We reassessed the tuberculosis infection status of household contacts using the TST at 6- and 12-month follow-ups among those who had a negative TST result at the previous test. In addition to tuberculosis infection status, we also collected information on household contact age, sex, sociodemographic data, height, weight, alcohol and tobacco use, comorbidities such as diabetes mellitus and HIV status, bacille Calmette-Guérin (BCG) vaccination status, use of isoniazid preventive therapy, history of tuberculosis disease, number of social contacts per day, household density, secondhand smoke exposure, and residential housing features.

Classification of Index Patient Smoking History and Status

At the time of index patient enrollment, we obtained detailed information on smoking history. We classified index patients as had never smoked, quit in the distant past, recently quit, or actively smoked (Supplementary Figure 1). Those who had stopped smoking longer than or within 2 months prior to their time of tuberculosis diagnosis were classified as distant quitters or recent quitters, respectively. We used 2 months as the primary cutoff to define recent smoking cessation based on previous studies suggesting that the adverse biological effects of cigarette smoking begin to improve as early as 1 month after initiation of smoking cessation [11, 12].

Study Outcomes

We considered tuberculosis infection at baseline as the first outcome of our analysis. Child contacts were considered to have been infected with tuberculosis at baseline if they had previously received a positive TST result within 6 months before enrollment or had a TST induration size >10 mm (>5 mm for HIV-positive contacts). We included TST-positivity status by 6 months of follow-up as the second outcome of our analysis. We classified contacts as infected with tuberculosis by 6 months of follow-up if they were considered infected at baseline (same criteria as above) or had converted their TST status from negative to positive during 6 months of follow-up [13].

Statistical Analyses

We focused our analyses specifically on child household contacts (aged 15 years) because we considered child contacts in our cohort to be more likely to be infected by index tuberculosis patients. Because tuberculosis drug resistance patterns may modify the effect of smoking on the transmissibility of tuberculosis, we also further restricted our analyses to child contacts of index patients with drug-susceptible tuberculosis [14]. We used a modified Poisson generalized estimating equation to measure the association between index patient smoking status and the risk of tuberculosis infection among child contacts at baseline and by 6 months of follow-up. We ascertained empirical standard error estimates to compute Wald-type 95% confidence intervals (CIs) for robust inference. We first performed bivariate analyses for covariates that may modify the risk of tuberculosis infection among child contacts. The covariates were chosen based on a priori background knowledge (age, sex, alcohol intake, and number of rooms in a household for index patients; age, sex, BCG vaccination status, number of social contacts per day, socioeconomic status, and secondhand smoke exposure for household contacts). All of these covariates were entered into a backward stepwise algorithm to construct the multivariate models. We retained the covariates with P < .1 in our final multivariate models and used complete data for our multivariate analyses.

We conducted 3 sensitivity analyses. First, to determine whether our findings were sensitive to the choice of time frame cutoff in defining index patient smoking status, we repeated the analyses using 4 months as the cutoff to classify recent and distant quitters. Second, we considered the possibility that indicators of index patient tuberculosis disease severity may mediate the association between index patient smoking status and tuberculosis infection risk in child contacts. We therefore further adjusted for index patient baseline smear status, treatment delay, and presence of cavitary disease in our multivariate models. Third, as older children, particularly adolescents, are also likely to have been exposed to tuberculosis in the past or in the community, we repeated the main analysis among child household contacts aged 5 years, a study population that was most likely to acquire recent tuberculosis infection from household exposure to the index patient. All statistical analyses were performed using R Version 3.6.2 (R Foundation for Statistical Computing, Vienna, Austria).

Ethics Statement

The Harvard School of Public Health Institutional Review Board and the National Institute of Health of Peru Research Ethics Committee approved the study. Written, informed consent was provided by all study participants or their guardians, and assent was provided by children aged < 18 years.

RESULTS

Index Patient and Household Contact Characteristics and Bivariate Analyses

We enrolled 1811 child contacts of 905 drug-susceptible index pulmonary tuberculosis patients (Figure 1). Of these, 696 (78.0%) had never smoked, 104 (11.7%) had quit in the distant past, 70 (7.9%) had recently quit, and 22 (2.5%) were active smokers (Table 1). Table 2 details the baseline characteristics of index patients and child household contacts by child household contact baseline TST-positivity status. In total, 442 (24.4%) of 1811 child contacts were found to be TST-positive at baseline, and 583 (37.3%) of 1565 child contacts were TST-positive by 6 months of follow-up. In the bivariate analyses, we found that baseline TST-positivity status among child household contacts was associated with index patient age, sex, alcohol intake, cavitary disease status, and treatment delay and with the household contact’s age, frequency of social contacts per day, and socioeconomic status (Table 2).

Figure 1.

Study flow chart for index patients and child household contacts. Abbreviations: TB, tuberculosis; TST, tuberculin skin test.

Table 1.

Baseline Characteristics of Drug-Sensitive Index Tuberculosis Patients With Child Household Contacts

Characteristic	No. (%), N = 905a
Age, y
16–30	541 (59.8)
31–45	217 (24.0)
>45	147 (16.2)
Female sex	417 (46.1)
Human immunodeficiency virus seropositive	17 (1.9)
Self-reported diabetes	50 (5.6)
Smoking status
Never smoked	696 (78.0)
Quit in the distant past (>2 m)	104 (11.7)
Recently quit (2 m)	70 (7.9)
Currently smoke	22 (2.5)
Sputum smear statusb
Negative	227 (25.1)
+	244 (27.0)
++	179 (19.8)
+++	254 (28.1)
Cavitary disease	246 (27.6)
Treatment delay, d
0	90 (10.6)
1–14	103 (11.5)
15–28	292 (32.7)
>28	409 (45.7)

^aTotal number may vary due to missing data.

^b+, 1–99 acid-fast bacilli (AFB) per field in 100 observed fields; ++, 1–10 AFB per field in 50 observed fields; +++; >10 AFB per field in 20 observed fields.

Table 2.

Child Household Contact Baseline Tuberculin Skin Test-Positivity Status by Index Patient and Household Contact Characteristics

		Baseline Tuberculin Skin Test–Positivity Status, No. (%)
Characteristic	Total No. (%), N = 1811a	Negative	Positive	Pairwise P Valueb	Aggregate P Valuec
Index patient characteristics
Age, y					<.001
16–30	1056 (58.3)	824 (78.0)	232 (22.0)	…
31–45	445 (24.6)	292 (65.6)	153 (34.4)	.001
>45	310 (17.1)	253 (81.6)	57 (18.4)	.484
Female sex	835 (46.1)	595 (71.3)	240 (28.7)	...	.002
Alcohol intake					.021
Nondrinker	987 (56.8)	758 (76.8)	229 (23.2)	...
Light	572 (32.9)	440 (76.9)	132 (23.1)	.690
Heavy	179 (10.3)	115 (64.2)	64 (35.8)	.010
Smoking status					.110
Never smoked	1371 (76.8)	1034 (75.4)	337 (24.6)	...
Quit in the distant past (>2 m)	211 (11.8)	162 (76.8)	49 (23.2)	.491
Recently quit (2 m)	155 (8.7)	125 (80.6)	30 (19.4)	.367
Currently smoke	49 (2.7)	30 (61.2)	19 (38.8)	.038
Baseline sputum smear statusd					.400
Negative	463 (25.6)	367 (79.3)	96 (20.7)	...
+	485 (26.8)	360 (74.2)	125 (20.7)	.140
++	383 (21.2)	282 (73.6)	101 (26.4)	.130
+++	478 (26.4)	358 (74.9)	120 (25.1)	.210
Cavitary disease	488 (27.4)	352 (72.1)	136 (27.9)	...	.047
Treatment delay, d					.001
0	151 (8.4)	130 (86.1)	21 (13.9)	...
1–14	202 (11.3)	156 (77.2)	46 (22.8)	.071
15–28	613 (34.2)	484 (79.0)	129 (21.0)	.134
>28	828 (46.2)	586 (70.8)	242 (29.2)	.002
Number of rooms in household					.210
1	303 (16.8)	217 (71.6)	86 (28.4)	...
2–3	716 (39.7)	544 (76.0)	172 (24.0)	.243
>3	784 (43.5)	602 (76.8)	182 (23.2)	.079
Child household contact characteristics
Age, y					<.001
0–5	772 (42.6)	627 (81.2)	145 (18.8)	...
6–10	518 (28.6)	385 (74.3)	133 (25.7)	<.001
11–15	521 (28.8)	357 (68.5)	164 (31.5)	<.001
Female sex	901 (49.8)	683 (75.8)	218 (24.2)	...	.890
Bacille Calmette-Guérin scar	1456 (80.4)	1103 (75.8)	353 (24.2)	...	.940
Frequency of social contacts per day					.004
None	145 (8.2)	127 (87.6)	18 (12.4)	...
Light	1145 (64.7)	847 (74.0)	298 (26.0)	.001
Heavy	480 (27.1)	368 (76.7)	112 (23.3)	.007
Socioeconomic status					.038
Low	700 (39.4)	501 (71.6)	199 (28.4)	...
Middle	784 (44.2)	622 (79.3)	162 (20.7)	.013
High	291 (16.4)	224 (77.0)	67 (23.0)	.142
Secondhand smoke exposure	179 (9.9)	134 (74.9)	45 (25.1)	...	.430

^aTotal number may vary due to missing data.

^bPairwise P values were obtained using the Wald test in the modified Poisson generalized estimating equation.

^cAggregate P values were obtained using analysis of variance.

d+, 1 – 99 acid-fast bacilli (AFB) per field in 100 observed fields; ++, 1 – 10 AFB per field in 50 observed fields; +++; >10 AFB per field in 20 observed fields.

Multivariate Analyses

After adjusting for potential confounders, child contacts exposed to recent quitters were less likely to be infected at baseline and by 6 months of follow-up when compared with those exposed to index patients who were active smokers (baseline: adjusted risk ratio [aRR], 0.45; 95% CI, .24–.87; Figure 2A; by 6 months of follow-up: aRR, 0.48; 95% CI, .29–.79; Figure 3A). The risk of tuberculosis infection at baseline and by 6 months of follow-up among child contacts exposed to recent quitters was similar to that of child contacts exposed to index patients who had never smoked (baseline: aRR, 0.81; 95% CI, .50–1.32; Figure 2B; by 6 months of follow-up: aRR, 0.76; 95% CI, .53–1.09; Figure 3B; see Supplementary Tables 1 and 2 for full univariate and multivariate results). When we repeated this analysis using a cutoff for smoking cessation of 4 months prior to the time of diagnosis, the effect sizes were altered by 10% or less (baseline: Supplementary Table 3; by 6 months of follow-up: Supplementary Table 4). When we further adjusted for indicators of index patient tuberculosis disease severity in the multivariate model, the results remained unchanged (baseline: Supplementary Table 1; by 6 months of follow-up: Supplementary Table 2). When we restricted our analysis to child contacts aged 5 years, the results for baseline TST positivity remained unchanged, and the effect sizes for recent smoking cessation and active smoking on TST positivity by 6 months became slightly more pronounced (baseline: Supplementary Table 5; by 6 months of follow-up: Supplementary Table 6).

Figure 2.

Risk of baseline TB infection among child household contacts by index patient smoking status. The forest plots show the risk ratios (95% CIs) for TST positivity at baseline for child contacts of index patients after adjusting for index patient (age, sex, alcohol intake) and household contact (age, number of social contacts per day, socioeconomic status, and number of rooms in the household) variable for (A) child contacts of active smokers as the reference category and (B) child contacts of index patients who had never smoked. Recent smoking cessation among index patients was defined as quitting smoking within 2 months prior to time of TB diagnosis. Abbreviations: CI, confidence interval; TB, tuberculosis; TST, tuberculin skin test.

Figure 3.

Risk of TB infection by 6 months of follow-up among child household contacts by index patient smoking status. The forest plots show the risk ratios (95% CIs) for TST positivity by 6 months of follow-up for child contacts of index patients after adjusting for index patient (age, sex, alcohol intake) and household contact (age, number of social contacts per day, socioeconomic status, and number of rooms in the household) covariates in the multivariate model for (A) child contacts of active smokers as the reference category and (B) child contacts of index patients who had never smoked. Recent smoking cessation among index patients was defined as quitting smoking within 2 months prior to time of TB diagnosis. Abbreviations: CI, confidence interval; TB, tuberculosis; TST, tuberculin skin test.

DISCUSSION

Here, we show that child contacts exposed to pulmonary tuberculosis patients who had quit cigarette smoking within 2 months prior to time of diagnosis had a similar risk of tuberculosis infection as those exposed to index patients who had never smoked and were half as likely to be infected as children exposed to tuberculosis patients who continued smoking. These observations remained largely unaltered when we took indicators of index patient tuberculosis disease severity into account, when we assessed the impact of smoking cessation periods longer than 2 months in duration, and when we only considered young children aged 5 years.

Several observational studies have previously reported that active smoking increases the infectiousness of tuberculosis patients. In Spain, Godoy et al reported that contacts of active smokers were 1.5 times more likely to be TST-positive compared with contacts of nonsmokers [4]. In India, Singh et al found that child household contacts who were exposed to tobacco smoke were 2.68 times more likely to be TST-positive than those who were not regularly exposed [3]. In Peru, we previously reported that child contacts exposed to tuberculosis patients who actively smoked had twice the risk of tuberculosis infection compared with those exposed to nonsmoking tuberculosis patients [5]. Although some studies showed no statistically significant association between index patient smoking and risk of tuberculosis infection among close contacts, the reported effect estimates were consistently greater than 1.3 [15, 16]. To our knowledge, no study has directly investigated the effect of having quit smoking in tuberculosis patients on the transmissibility of tuberculosis.

Some previous studies have assessed the effect of smoking cessation in in vitro and in vivo models of tuberculosis and have produced mixed results. In alveolar macrophages taken from bronchoalveolar lavage fluid of nonsmokers, ex-smokers, and smokers, O’Leary et al found that macrophages from both ex-smokers and smokers were unable to mount a proinflammatory cytokine response as effectively as macrophages from nonsmokers following Mycobacterium tuberculosis H37Ra infection. The mean smoking cessation period for ex-smokers was 10 years [17]. However, in a separate study that also assessed alveolar macrophages from nonsmokers, ex-smokers, and smokers, Berg et al showed that macrophages from both nonsmokers and ex-smokers were able to effectively migrate toward M. tuberculosis H37Ra within 2 hours, while migration of vacuolated macrophages from smokers was impaired [18]. In animal models, Shaler et al reported that mice that stopped receiving cigarette smoke after 4 weeks of initial exposure had less evidence of histopathological lung tissue damage, reduced mycobacterial burden in both the lung and spleen, and improved indicators of Th1 protective immunity (eg, improved innate cell recruitment and lung T-lymphocyte activation) following challenge with Mycobacterium bovis BCG compared with mice that continued to receive cigarette smoke following initial exposure. Notably, the authors also found that the effects of smoking cessation began to take place as early as 4 to 6 weeks after M. bovis BCG challenge [12]. These findings suggest that pulmonary inflammatory and immune responses are correlated with the infectiousness of an individual infected with tuberculosis.

We evaluated several possible explanations for the effect of recent smoking cessation on the transmissibility of tuberculosis in children. First, index patients who recently quit may be more likely to cough less and, thereby, transmit less to their household contacts. Indeed, longitudinal studies have shown that coughing frequency among individuals who quit reduce within 1 to 2 months following smoking cessation [19–21]. Second, previous observational studies have shown that tuberculosis patients who smoke have a higher risk of poor treatment outcomes compared with nonsmokers [7, 22]. Tuberculosis patients who recently quit smoking may achieve culture conversion faster than those who continue to smoke during treatment and may therefore be less likely to transmit tuberculosis following their diagnosis [23–25]. Third, we also considered the possible role that severity of tuberculosis disease in index patients may play in relation to quitting smoking. On the one hand, it is possible that index patients with more severe tuberculosis disease were more likely to quit smoking prior to diagnosis. On the other hand, it is also possible that the effects of smoking cessation could improve disease severity. While our data do not allow us to determine whether disease severity acts upstream or downstream to the decision to quit smoking, our results did not change appreciably when we adjusted for indicators of index patient disease severity (baseline sputum smear status, cavitary disease, and treatment delay) in our sensitivity analyses.

Our study has several limitations. First, index patient smoking status may be misclassified due to recall bias. However, the impact of such misclassification may be unsubstantial, as our results were altered by less than 10% when we used 4 months as the cutoff to classify recent and distant quitters. Second, although we restricted our analyses to child contacts, children could have still been infected outside the household or in the past. Nondifferential misclassification of the exposure status for contacts would have resulted in an underestimation of the association. Indeed, we observed slightly more pronounced effect estimates for recent smoking cessation and active smoking on TST positivity by 6 months when we analyzed child household contacts aged 5 years who were less likely to be infected outside the household or in the past. Third, as TST-positivity status at baseline is unable to differentiate between tuberculosis infection acquired in the recent or distant past, some child contacts may have been infected with tuberculosis prior to index patients quitting smoking. In our current analysis, we included tuberculosis infection by 6 months of follow-up to include contacts who had converted their TST status after information regarding index patient smoking history had been recorded. Last, we were unable to determine if index patient smoking behavior had changed following study enrollment. This limitation would have resulted in nondifferential misclassification of the exposure and bias the results of tuberculosis infection by 6 months of follow-up toward the null. Given what we report regarding the effect of smoking cessation shortly before time of diagnosis, we would also expect a decrease in the risk of tuberculosis infection among child household contacts of index patients who quit smoking following their diagnosis.

Our findings have potential public health implications for both tobacco and tuberculosis control. Based on our results, we estimated that we would prevent 1 tuberculosis infection among child household contacts for every 2 smokers who decided to quit before diagnosis. In Peru, previous studies have shown that smoking cessation interventions such as the use of text messages among young adults [26] or increases in cigarette excise taxes can be effective in reducing cigarette consumption [27]. In settings where the prevalence of smoking is high, smoking cessation interventions or policies have a greater potential to prevent a higher absolute number of tuberculosis infections in the population [28].

In conclusion, we found that children exposed to index patients who had quit smoking shortly before time of diagnosis had a reduced risk of tuberculosis infection compared with children exposed to active smokers. Our findings suggest that the effect of smoking cessation is both rapid and effective in reducing the transmissibility of tuberculosis. As the onset of tuberculosis symptoms ranges between weeks to several months prior to time of diagnosis, promoting general smoking cessation education and support resources for symptomatic individuals may help reduce transmission of tuberculosis in high-burden communities.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Acknowledgments. The authors thank the patients and their families who gave their time and energy to contribute to this study, the National Strategy for Tuberculosis Control at the Peruvian Ministry of Health, and the healthcare personnel at the 106 participating health centers in Lima, Peru.

Financial support. This work was supported by the National Institutes of Health, the National Institute of Allergy and Infectious Diseases (U01AI057786, U19AI076217), Centers of Excellence for Translational Research (U19AI109755), and Tuberculosis Research Units Network (U19AI111224). (M. B. M. reports these are grants on which she was the principal investigator but which were contracts with Harvard Medical School. C. C. H. reports being a co-investigator of all these grants.)

Potential conflicts of interest. M. B. M. reports direct payment for expert testimony on coronavirus disease 2019 and presidential elections from Fair Elections, American Civil Liberties Union, Arnold and Porter, and Wilmer Hale. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.