Smoking and tuberculosis among silicotic patients

The aim of the present study was to investigate the relationship between smoking and tuberculosis (TB) among high-risk silicotic patients in Hong Kong. A cohort of 435 silicotic patients tuberculin tested from 1995–2002 was prospectively followed-up until the end of 2005. Baseline characteristics were analysed with respect to positive tuberculin reaction (≥10 mm) at baseline and subsequent development of TB. Smoking, alcohol use and body mass index were independent predictors of positive tuberculin reaction at baseline in multiple logistic regression analysis. Total cigarette pack-yrs did not demonstrate any significant effect. The annual incidences of TB were 1,841, 2,294 and 4,181 per 100,000 for never-, ex- and current smokers, respectively. On Cox proportional hazard analysis, current smokers have a significantly higher risk of TB than other silicotic patients (adjusted hazard ratio (95% confidence interval (CI)): 1.96 (1.14–3.35)) after controlling for age, alcohol use, tuberculin status, treatment for latent TB infection and other relevant background/disease factors. A significant dose-response relationship was also observed with the daily number of cigarettes currently smoked. Smoking cessation may reduce 32.4% (95% CI: 6.5–54.0) of the risk. Smoking increases the risk of both tuberculosis infection and subsequent development of the disease among silicotic patients.

I n recent years, there has been increasing evidence on the relationship between smoking and tuberculosis (TB). Smoking has been associated with latent TB infection (LTBI) [1][2][3], active TB disease [4][5][6][7][8] and TB-related death [9][10][11][12]. However, the underlying mechanisms of such associations remain largely elusive. As both conditions are strongly related to a multitude of socioeconomic factors, increased risk of exposure could explain, fully or in part, all the associations observed [13,14]. The long latent period and relatively low incidence of active disease among immunocompetent subjects renders it rather difficult to dissect the complex relationships.
In Hong Kong, both smoking and TB are common conditions [15]. Silicosis is also the second most common statutorily notifiable occupational disease. Silicotic patients are regularly referred to the Pneumoconiosis Medical Board in the Tuberculosis and Chest Service of the Department of Health (Hong Kong, China) for compensation assessment. Confirmed silicotic patients are reassessed every 2 yrs for delineation of additional disability and also at death for contribution of the disease. During compensation assessment, detailed occupational and smoking histories are available. A very high risk of TB has been reported among these silicotic patients [5,16]. Since 1995, tuberculin tests are also regularly offered to newly confirmed cases without previous history of TB. The availability of detailed assessment and tuberculin testing records in the present high-risk cohort allowed examination of the effect of smoking on TB prospectively and in greater detail.

Study subjects
All silicotic patients without previous history of TB, tuberculin skin testing or treatment for LTBI were offered a Mantoux test with purified protein derivative RT23 (1 unit from August 1, 1995 to June 30, 2000; 2 units from July 1, 2000 to December 31, 2002) as part of the regular service. Due to their existing small numbers, female subjects were excluded from the present study. The study was approved by the ethics committee of the Department of Health of Hong Kong.

Methods
The following baseline information was collected at the time of tuberculin testing: date of tuberculin testing; name; age; identity card number; ethnicity; smoking history; alcohol use; bacilli Calmette-Guérin (BCG) scar; coexisting medical conditions; occupation; duration of dust exposure; and disease indices according to the International Labour Organisation classification (profusion/size/shape of lung nodules, progressive massive fibrosis). A never-smoker was defined as an individual who had never smoked as many as 1 cigarette?day -1 or equivalent for the duration of 1 yr. An ever-smoker was defined as an individual who had smoked o1 cigarette?day -1 for o1 yr. An ex-smoker was defined as an ever-smoker who had stopped smoking for o1 yr prior to study commencement. A current smoker was defined as an ever-smoker who was still smoking within the previous 1 yr. A regular drinker was defined as an individual who drank alcohol regularly (o4 days?week -1 ) at the time of tuberculin testing. The cohort was followed-up prospectively in the Pneumoconiosis Clinic (Hong Kong) or other chest clinics of the Tuberculosis and Chest Service until December 31, 2005. The cohort was also cross-matched with the statutory TB notification registry and death registry of Hong Kong for further case finding and ascertainment of vital status. The period of follow-up was defined as the period from enrolment to the date of notification of TB, death or December 31, 2005, whichever was earlier. Date of TB notification, bacteriological status, form of TB and previous TB history were retrieved from the notification registry. The diagnosis and clinical information of all identified TB cases were verified with the medical records retrieved from the Pneumoconiosis Clinic and other relevant sources. An active case of TB was defined as disease proven by isolation of Mycobacterium tuberculosis or, in the absence of bacteriological confirmation, disease diagnosed on clinical, radiological and/ or histological grounds together with an appropriate response to treatment.

Analysis
The effect of background and disease characteristics on LTBI (defined as a tuberculin reaction o10 mm) on the baseline was first analysed in a cross-sectional analysis, followed by a prospective analysis with respect to the development of active TB disease and culture-proven TB. For univariate analysis, Chi-squared or Fisher exact tests were used for categorical variables and ANOVA for numerical variables as appropriate. Multiple logistic regression analysis and Cox proportional hazards analysis were then employed in the cross-sectional and prospective analysis, respectively, to control for all the confounding background/disease variables. A two-tailed pvalue ,0.05 was taken as statistically significant. The risk of TB attributable to smoking was derived by applying the adjusted hazard ratio to the Levin formula [17]: where p is the proportion of the population exposed and r is the relative risk of disease for the exposed group.

DISCUSSION
With the potential ethical concerns in conducting a randomised controlled experiment on human subjects, a prospective cohort analysis is probably the best option to address the relationship between smoking and TB. Even in high incidence areas, the annual rate of TB among immunocompetent individuals is usually just in the order of one in 1,000. Therefore, it is necessary to follow-up a very large cohort to get sufficient cases for analysis, and with this, critical details about potential confounders are often incomplete or missing. The present cohort involved a relatively homogeneous group of male silicotic patients of Chinese ethnicity who were at a  In the present study, there was a high prevalence of LTBI as measured by a tuberculin reaction o10 mm (table 1), similar to what was previously reported among the silicotic patients in Hong Kong [5,16,18]. Both current and ex-smokers were associated with LTBI (table 1) and such association was independent of alcohol intake, BMI and other background variables (table 2). Never-smokers, ex-smokers and current smokers were at different risks of TB (table 3). Current smoking was consistently associated with higher rates of both active TB and culture-confirmed TB, even after exclusion of regular alcohol use or stratification by tuberculin status/ treatment of LTBI (  fig. 1). Past cigarette pack-yrs showed only a suggestive but insignificant effect on the development of TB (table 5). The independent and dose-response relationship between current smoking and the subsequent development of TB disease is consistent with the authors' previous observations in another prospective cohort study [8]. In contrast with that study, none of the subjects had past history of TB and much more information was available regarding cigarette pack-yrs, time of giving up, tuberculin status and treatment history for LTBI.
With the closer examination allowed in the present study, there is further support of a causal link between smoking and TB. There is an increased vulnerability of developing disease, in addition to an increased risk of acquiring infection. For many years smoking has been associated with changes in pulmonary macrophages [19] and lymphocytes [20]. Pulmonary and upper zone predominance of TB lesions in smokers has been reported in a previous study [7]. More recent studies suggest that nicotine [21], similar to vagus efferent discharge [22], inhibits the release of tumour necrosis factor, which plays a key role in the cellular defence against TB [23]. The overall evidences are therefore in line with a direct airborne insult promoting the risk of disease development.
Tuberculin test was used to measure the LTBI because the specific interferon-release assays were not available locally at the time of the study [24]. As BCG vaccination was only introduced into Hong Kong in the 1950s [25] and the vast majority of the present cohort was born well before that time, BCG vaccination is unlikely to affect the interpretation of the test excessively. Indeed, only 3.2% of the present cohort showed a BCG vaccination scar and there was no significant association between a BCG scar and a positive tuberculin reaction.
The present study focused on a homogenous group of ethnically Chinese male patients with silicosis. It would be  ideal if it were indeed possible to examine a cohort involving all age groups in the general population. However, tuberculin test data are seldom available among other groups and their relatively low disease incidences pose a genuine difficulty in showing a difference between different smoking categories using active disease as an end-point. Therefore, the current cohort provides a possible window to examine the intricate relationship between smoking, LTBI and active TB. While there might be genuine concern that the findings might not necessarily be transferable to a general population, the higher risk of active TB in current smokers, as observed, is consistent with the findings in a similar study among a larger cohort of elderly subjects within the general community [8].
Incomplete case ascertainment is a major concern for prospective cohort analysis. As the cohort was under regular follow-up for compensation purposes and attempts have also been made to screen for missed cases through the territory-wide TB notification registry, the number of such cases should be low. In any case, incomplete case ascertainment, unless highly selective, is unlikely to significantly affect the internal comparison among different smoking categories. The vital status of patients missing follow-up was also ascertained through cross-matching with the territory-wide death registry. A relatively long period of cessation for 1 yr was employed to distinguish between current and ex-smokers and this would have minimised the short-term fluctuations associated with abortive smoking cessation attempts. The diagnosis of TB could have been complicated by the progression of the underlying silicotic process. As current smokers were more likely to experience symptoms, this might increase the possibility of over-diagnosis of bacteriologically negative TB. However, the association persisted even if only culture-proven cases were considered.
Although alcohol use is a well-known confounder [26,27], its prevalence was relatively low in the present cohort. Smoking showed independent effect on LTBI and development of TB, even after controlling for the effect of alcohol (tables 2, 4 and 5).
In the absence of detailed local data on dust exposure levels for various jobs, only job type and duration of exposure were used as proxy measures of dust exposure and disease indices were also taken into account in the analysis. The prevalence of LTBI was not associated with specific job types (table 1) and the increased risk of current smokers persisted after adjustment of all relevant background variables and disease indices (table 5).
In the present study, one-third of the risk of tuberculosis disease was attributable to current smoking, partly reflecting the high smoking prevalence in the current cohort. Tuberculosis is an important cause of morbidity and mortality among silicotic patients [5] and both smoking and tuberculosis are independently associated with lung function impairment [28]. Smoking cessation is, therefore, expected to have major impact on the quality of life for patients suffering from this largely irreversible disease. Ageing of the population is also increasingly observed in developing countries, many of which still have a very high tuberculosis burden. In Hong Kong, ,40% of tuberculosis patients are o65 yr of age and males of .80 yrs of age have an eight-fold higher tuberculosis risk than the population average [15]. The interaction of smoking with the huge latently infected pool in an ageing population is therefore a serious concern. As endogenous reactivation accounted for .70% of the tuberculosis risk in Hong Kong [29], the use of directly observed treatment, short course alone, is unlikely to be able to bring down the disease incidence rapidly. With the high male and rising female smoking prevalence in many parts of Asia, it would appear rational to include smoking cessation as one of the armaments for tuberculosis control.