In 1956 Doll and Hill¹ wrote that “the relationship between smoking and mortality from pulmonary tuberculosis is distinct, but with a disease so influenced by social factors more precise data are needed to justify a direct cause and effect hypothesis”.

The essential risk factors for human tuberculosis are (1) the tubercle bacillus, (2) a susceptible host, and (3) an environment which allows the tubercle bacilli to survive transit from one host to the next. All other risk factors are subsumed under these headings. If smoking is a risk factor for tuberculosis, then it must act by increasing the susceptibility of the human host or the probability of transmission by encouraging infectious individuals to cough (this requires smoking to be a social as much as an individual pursuit). If the association between smoking and tuberculosis is more apparent than real, then smoking may be a pointer to other risk factors. These include social class—itself a marker for overcrowding, poor ventilation and rooms with no natural light as well as poor nutrition—general ill health and, increasingly, HIV infection with prostitution and intravenous drug use.

TUBERCULIN SENSITIVITY AND SMOKING

Tuberculous infection and tuberculosis as a disease are entirely different states. The former is commonly characterised by tuberculin reactivity (despite the known problem of exposure to non-tuberculous mycobacteria and the cross reactivity of the mixture of antigens²). Approximately one third of the world’s population may be infected with the tubercle bacillus, but only eight million (0.4%) develop tuberculosis each year.

Studies conflict as to whether smoking affects delayed hypersensitivity to tuberculin. Kuemmerer and Comstock³ noted that tuberculin reactions were greater in children where both parents smoked, but they also observed that education, urban residence, immigration, and overcrowding were significant associations. Although the Heaf test grade was found to be directly related to pack-years of smoking in residential homes for the elderly in the UK, social class was not determined as a possible confounding factor.⁴ A similar study in Hong Kong showed no association with smoking,⁵ while a survey of Vietnamese immigrants in Australia suggested a positive association between tuberculin responses and smoking history.⁶ In Norway, linear regression analysis associated smoking and male sex with greater tuberculin reactivity.⁷ In a study in Kuwait tuberculin reactivity was greater in smokers, and univariate analysis of variance showed a dose-response to pack-years in healthy controls but not in patients with tuberculosis.⁸ None of these studies accounted for socioeconomic status and its possible confounding effect on smoking and tuberculosis.

In this issue of Thorax a study by Den Boon et al⁹ in a socially homogeneous group again shows that tuberculin reactivity can be related to smoking, especially if this constituted more than 15 pack-years. More interestingly, a prospective study in the United States examined whether smokers were more likely to have developed a positive tuberculin reaction while in prison.¹⁰ Only those who had smoked for more than 15 years were more likely to show tuberculin conversion (relative risk 2.12, 95% confidence interval 1.03 to 4.36). The authors concluded that the cumulative effect of prolonged smoking was more significant than the number of cigarettes smoked in increasing the likelihood of infection by Mycobacterium tuberculosis in prison. The broad confidence intervals, with the lower figure giving an attributable risk of <3% to smoking for tuberculin conversion, recommend a further study with greater power and with an assessment of the duration of imprisonment. Exposure to M avium-intracellulare is common in the southern United States and may affect tuberculin conversion. Concurrent HIV infection is more common in those aged 20–35 years and might also confound tuberculin sensitivity. A future prospective study would need to examine the social behaviour of chronic smokers and the likelihood of the index case being within those who have smoked for >15 years.

The evidence to support a direct effect of smoking on tuberculin reactivity is therefore poor.

SMOKING AND TUBERCULOSIS

As already mentioned, tuberculosis as a disease is very different from tuberculous infection, defined by a positive tuberculin response. Several studies, beginning in 1956, have linked smoking with tuberculosis (reviewed by Maurya et al¹¹). As in a more recent study,¹² the possible confounding of socioeconomic factors with both smoking and tuberculosis has only occasionally been examined.¹³ Yu et al¹⁴ used binomial regression to propose that heavy smoking was associated with pulmonary tuberculosis, although both were associated with male sex and increasing age. A case-control study matching street based postcode, sex, date of birth, and ethnic origin in Liverpool suggested that smoking for >30 years was associated with the development of all forms of tuberculosis,¹⁵ but this was not significant when corrected for the number of factors examined. Prospective evaluation of 42 655 individuals registered with the Elderly Health Service in Hong Kong noted that pulmonary tuberculosis was more common in current smokers than in ex-smokers, and both were more common than in never smokers. Cox proportional hazards analysis accounted for 18 potential confounding factors including alcohol and several relating to socioeconomic status. There was also a dose-response relationship in current smokers for the development of pulmonary tuberculosis. A small study of contacts of sputum smear positive tuberculosis compared 46 adult patients with culture positive pulmonary tuberculosis and 46 tuberculin positive subjects without active tuberculosis.¹⁶ Adjusting the odds ratios for age, sex, and socioeconomic status demonstrated a dose-response relationship between the number of cigarettes smoked and the risk of active pulmonary tuberculosis. A similar study from the same group compared children in contact with tuberculosis who later developed the disease with those who remained well during the period of study despite a positive tuberculin skin test.¹⁷ Passive smoking, as assessed from the smoking history of the adult case and from urinary cotinine levels, was associated with a risk of developing pulmonary tuberculosis (∼80% primary complex disease). One would have to postulate that the infectious load, related to coughing behaviour of the smoking adult and using cotinine levels as a marker of proximity of the adult, was significant if passive smoking itself was not the significant factor in the development of active tuberculosis. The data are consistent with the hypotheses that both a long duration of smoking and current smoking might be related to the development of active pulmonary tuberculosis.

EFFECTS OF SMOKING AND THE IMMUNE RESPONSE IN THE LUNG

The alveolar macrophage is probably the first cell to ingest a tubercle bacillus following infection. These cells suppress the local immune response in order to preserve lung architecture.¹⁸ Silicosis is associated with an increased incidence of tuberculosis, suggesting that proper function of these cells is protective. Smoking impairs the phagocytic function of alveolar macrophages.¹⁹ Both smoking and tuberculosis induce apoptosis of these cells.^20,21 However, smoking quadruples the number of macrophages that can be harvested by bronchial lavage.²² Cigarette smoke activates alveolar macrophages to produce a local inflammatory response,²³ but nicotine suppresses the antigen presentation function to develop a specific immune response.²⁴ Chronic exposure to cigarette smoke reduces expression of surface proteins related to antigen presentation by pulmonary macrophages.^25,26 Others have suggested that human alveolar type II pneumocytes may be the first port of call for the tubercle bacillus,²⁷ but again this site of infection may promote innate over specific immunity.²⁸ Smoking is not yet known to affect the expression of significant surface proteins in these cells.

Interleukin-18 (previously known as interferon-γ inducing factor) is reduced in induced sputum from smokers.²⁹ Nicotine impairs antigen receptor mediated signal transduction³⁰ and induces T cell anergy.³¹ Natural killer cell activity is also less and significantly suppressed by alveolar macrophages from bronchoalveolar lavage fluid of smokers compared with non-smokers.³²

CONCLUSION

The association of smoking with pulmonary tuberculosis might be explained by a reduced specific immunity and possibly enhanced non-specific inflammatory response. Such a model would predict that smoking during exposure to tubercle bacilli is less likely to induce delayed hypersensitivity but more likely to produce disease. If correct, any increase in tuberculin sensitivity associated with smoking would have to be explained by social behaviour rather than the host response.