Study design
As shown in Figure 1, a case-control study was conducted to estimate the association between cigarette smoking and risk of TB. Among current smokers, smoking cessation behavior after TB diagnosis was observed and its correlation to anti-tuberculosis treatment adherence was explored.
Study sites
This study was conducted in Yangzhong and Wujin County, both of which were relatively rich rural areas, located in the southeast part of Jiangsu Province, China. The local TB dispensary was affiliated to the county Centre for Disease Prevention and Control (CDC) which was formerly called Anti-Epidemic Station. All newly notified TB patients were registered and managed under the DOTS (directly observed therapy, short-course) program. The annual notification rate of TB in 2007 was 79/100,000 in Yangzhong and 78/100,000 in Wujin, respectively.
Study subjects
TB patients diagnosed in Yangzhong (905 cases from 2004 to 2007) and Wujin (1476 cases from 2006 to 2007) were target population of this study. By using simple random sampling method, 700 cases in each site were selected and those with eligible phone numbers recorded in the registry book were contacted by telephone. A total of 613 patients, including 381 patients in Yangzhong and 232 patients in Wujin County, were successfully recruited in this study. The reasons for non-response included migration, death, and stopping or changing the phone numbers, etc. To further explore the potential selection bias in this study, we compared the distribution of key characteristics between source population and these 613 cases. No significant difference was found on the distribution of gender and sputum smear test results (P > 0.05). However, the average age of patients included in this study was older than that of registered patients (56 ± 16 years vs. 50 ± 19 years). It might be attributed to the higher migration rate among young patients. A total of 1226 controls were randomly selected from a pool of more than 30,000 individuals (overall response rate: 86%) who participated in a community-based health examination program conducted in Jiangsu Province during the same time period as the cases were recruited [9]. These control subjects had no self-reported history of TB, diabetes and malignancy and were frequency matched to the cases on gender, age (± 5 years) and residential areas (urban and rural). The cases and control subjects were all genetically-unrelated Han Chinese.
Data collection
Based on the local TB management system, all patients were followed up by TB dispensary for the whole treatment episode. To further explore the role of cigarette smoking in the risk of TB and to examine longitudinal patterns of smoking cessation among TB patients, we followed these patients again in March, 2008, by inviting them to CDC for personal interview. The interview was undertaken by trained investigators using a structured questionnaire. To ensure that all items were correctly filled in, the research supervisor double checked the questionnaires every day. For cigarette smoking status, a person who smoked at least once a day and lasted for more than 6 months in his or her lifetime was regarded as a smoker. Ex-smoker (former smoker) was defined as a smoker who had stopped smoking for at least 3 months. Smoking relapse among TB patients was defined as those who quitted smoking (≥ 3 months) after being diagnosed with TB but smoked again in the follow-up period. For alcohol drinking, it was defined as those who reported drinking alcohol more than 3 times per week. Non-adherence to anti-tuberculosis treatment was defined as those who had interrupted treatment for more than 2 weeks, including those who refused to take drugs or stopped treatment for two weeks before the end of the prescribed course.
Data analysis
Data were managed with EpiData 3.1 (Denmark) and analyzed by STATA 10.0 (College station, TX, USA). Differences in the distribution of demographic characteristics (gender, age, education and marital status) and selected variables (alcohol drinking history) between cases and controls were tested by using the Student t-test (for continuous variables) or χ2-test (for categorical variables). The associations between cigarette smoking and risk of TB were estimated by computing odds ratios (ORs) and their 95% confidence intervals (CIs) using logistic regression model. Crude odds ratios and corresponding adjusted odds ratios by controlling age, gender and alcohol drinking history for cigarette smoking history, age of starting smoking, smoking years and cumulative pack-years were calculated respectively. To evaluate the effects of cigarette smoking on TB risk according to selected variables (gender, age, education and alcohol drinking), stratified analysis was also performed. Multivariate Cox proportional hazards model was applied to calculate hazard ratios (HRs) and 95% confidence intervals (95% CIs) in analyzing the risk factors for smoking relapse. The Kaplan-Meier estimate was computed to plot the ability of smoking-free after cessation among different groups, with the Log-rank test being used to compare the difference. During the follow-up period, if the quitter smoked again, it was defined as one event in the model; otherwise it was regarded as the censor. The criterion for significance was set at P < 0.05 based on a two-sided test. The continuous variables of age, age of starting smoking, smoking years, cigarettes per day and cumulative pack-year were transferred to categories by using median among controls as cutoff point.
Ethical consideration
This project has been approved by Institutional Review Board of Nanjing Medical University. Written informed consent was obtained from all participants. Ethics has been respected throughout the whole study period.