Characteristics and Determinants of Treatment Default Among Smokers With Tuberculosis in an Industrial State of Malaysia: A Registry-based Study of the Years 2013-2017


 Background: The increased risk of treatment default among smokers raises concern over the secondary spread within the community. This study aimed to determine the prevalence and factors associated with treatment default among TB patients who smoke.Methods: A retrospective cohort of all registered TB patients who smoke in the state of Selangor between 2013 and 2017 via the Malaysian National MyTB database was included for analysis. TB patients who smoke were considered those with an active smoking status during the notification, while treatment default was defined as a TB patient who had interrupted treatment for 2 months or longer. There were 4 main variable domains included for analysis: sociodemographic profiles, disease profiles, treatment profiles, and comorbidities. Logistic regression analysis was used to identify determinants of treatment default among TB patients who smoke.Results: A total of 27.6% (N=6278) of the TB patients registered in Selangor were active smokers, and 15.1% (N=813) of the TB patients who smoke experienced defaulted TB treatment. The determinants of treatment default among TB patients who smoke were patients staying in an urban area (AOR 1.47; 95% CI 1.11,1.96), median income level less than RM2160 (AOR 2.0; 95% CI 1.34,2.99), no formal education (AOR 2.12; 95% CI 1.31,3.44), previously treated cases (AOR 2.78; 95% CI 1.99,3.88), active TB case detection methods (AOR 2.05; 95% CI 1.21,3.47), treatment duration of less than 6 months (AOR 7.56; 95% CI 5.74,9.92), and patients not on DOTS during the continuation phase (AOR 27.96; 95% CI 21.1,37.1). All the significant factors gave rise to the final model of determinants, with a predictability of 92.9% (95% CI 92.0,93.7).Conclusions: Our findings highlighted the high prevalence of treatment default among TB patients who smoke compared to the general TB population. Early risk detection that examines the two main domains of risk factors (socioeconomic factors and treatment profiles) should be provided for those who smoke in the TB population. Interventions should aim to reduce the prevalence of smoking among TB patients, together with close supervision during DOTS.


Background TB current situation
An estimated 10 million cases of tuberculosis (TB) occurred in 2018, leading to 1.3 million deaths worldwide (1). In Malaysia, TB is the leading cause of death (mortality rates uctuate from 4.8-6.2 cases per 100,000 population) from a single infectious disease; TB ranked above HIV/AIDS, dengue fever and malaria in the 5-year period from 2012-2016. With the current trend of the TB noti cation rate, it is projected that the incidence of TB will continue to increase through 2030 (2). There are 11 states in the Peninsular of Malaysia and Selangor recorded the highest number of TB cases. Selangor is an industrialised state with > 90% urban population, creating more than 802,000 employment opportunities as reported in 2019 (3).
TB is a mandatory national noti able infectious disease under the Prevention and Control of Infectious Disease Act 1988. All suspected and con rmed cases of TB should be reported and noti ed to the nearest district health o ce by submission of the noti cation form. Data from the noti cation form are entered into the national TB registry by the health inspectors of the respective district health o cer. The TBIS is a web-based application administered by the Ministry of Health Malaysia to record activities related to the noti cation, registration, investigation, and treatment of TB diseases in all states in Malaysia.
Smoking and TB remain major public health challenges globally. Tobacco smoking is responsible for 20% of the global burden of tuberculosis and will be responsible for a total of 18 million new cases and 40 million deaths in the 2010-2050 period (4). It is projected that TB incidence will increase up to 7% if we incorporate the effect of smoking compared to the effect without smoking. Numerous studies have identi ed smoking as a risk factor for the development of TB (5) and that it has a signi cant association with undesirable treatment adherence and default(6). Smoking has also been associated with more extensive lung disease and delayed sputum conversion even after 2 months of treatment in both current smokers and ex-smokers, which makes their treatment outcome worse (7). A higher prevalence of active smokers has been noted among TB patients compared to the general population in countries such as Indonesia, Africa, and India(8-10). In Malaysia, the prevalence of active smokers among TB patients was 34.0% in 2012 (11) compared to only 22.7% among the general population (NHMS, 2015).

Treatment default
In addition to smoking, defaulted TB treatment is also another threat faced by the TB control team in Malaysia. Patients with incomplete treatment for TB will become a signi cant economic burden on the government, with an average of RM901.63 per patient four times higher compared to the cost of completed treatment. The extra cost is normally attributed to hospital stays and patient care for complicated TB cases (12). The prevalence of treatment default varies among different countries and ranges from 2.5-44.9% (13). A very high proportion of 44.9% has been observed in rural northern Mozambique, where treatment default rates are a very serious problem (14). The prevalence of TB treatment default in Malaysia ranged from 4.0-4.8% in the years 2010-2015 among the general TB population (Ministry of Health, 2016); this number has increased to 5.6% according to the latest study (15). Studies from many other countries have shown a signi cant association between smoking and the tendency to default TB treatment; however, there are limited data on the prevalence of treatment default among those who smoke in the TB population. A local study in Penang found that smoking among TB patients is signi cantly associated with poorer treatment outcomes and increased treatment default by OR 7.17 compared to non-smokers(16). In Hong Kong, studies have concluded that smoking is the key contributor to defaulting, with a doubled risk compared to that of TB patients who never smoked (17); furthermore, a smoking habit is a good indicator for evaluating the risk of defaulting from TB treatment under DOTS(6). Similar ndings have also been found in studies from Morocco and Tehran, where smokers have a double and triple increased likelihood of defaulting on their TB treatment, respectively(18, 19).

Determinants of treatment defaults
Previous studies have identi ed several reasons and risk factors for treatment default among TB patients, including smoking, alcohol use, comorbidities (HIV and diabetes mellitus), accessibility to a healthcare centre, socioeconomic factors (age, sex, education level, and income), and poor family support (20)(21)(22). Treatment default is also common among those who previously defaulted on TB treatment and among relapse cases; defaulting is the most common during the intensive phase of treatment (23). The evidence for a connection between smoking and treatment default, however, is inconclusive. Some studies have hypothesized that smokers with TB disease are less likely to comply with their TB treatment (24), while other studies have found that smokers have low levels of concern for their health. The behaviour of delaying seeking medical care at a more severe phase of illness and noncompliance among smokers may result in their worse prognosis (25); however, there is no solid conclusion for this hypothesis.
Despite knowing the impact of smoking and incomplete treatment among TB patients, there have been limited studies that examine these two issues. The current practice in our healthcare system is to refer all TB patients who smoke to the 'stop smoking clinic' for smoking cessation programs. However, there are no speci c interventions to identify smoking populations who are at higher risk of defaulted TB treatment. This present study aimed to determine the factors associated with treatment default among TB patients who smoke in a 5-year cohort (2013-2017) of patients registered in the Selangor My-TB database.

Methods
Study setting, inclusion, and exclusion criteria. This was a cross-sectional study that utilized data from the National MyTB database version 2.1 (TBIS) in 2013 to 2017 from the Disease Control Division, Ministry of Health. TB data from all states in Malaysia are consolidated at the national level for the surveillance database; however, in this study, we included only TB patients who smoke who were registered in the Selangor MyTB database from 2013 until 2017. The state of Selangor was selected based on its high TB incidence rate, i.e., 5,071 cases in 2018 (Malaysian MOH, 2019). Both Malaysians and non-Malaysians were included in this study. The exclusion criteria in this study included cases initially registered as TB but ultimately diagnosed as something other than TB disease, cases with missing data on treatment outcomes and cases with the outcomes not evaluated ("transferred out" to another treatment unit and whose treatment outcome is unknown) (WHO, 2013). We also excluded cases with multidrug-resistant TB (MDR-TB), as the treatment outcome de nition for MDR-TB cases is different from the non-MDR-TB classi cation. TB treatment outcome and disease classi cation were de ned according to the World Health Organization de nition and the National TB database protocol.

Operational de nition
According to the Clinical Practice Guideline for Management of Tuberculosis by Ministry of Health Malaysia and the de nition and reporting framework of the WHO(26), treatment outcomes of TB can be broadly categorized into successful treatment outcomes (cured and completed treatment) and unsuccessful treatment outcomes (treatment default, treatment failure, death). In this study, the outcomes were divided into "treatment default" and "nontreatment default". A defaulter (case) was de ned as TB patients who interrupted treatment for ≥ 2 months or longer before the end of the treatment period, according to the WHO de nition. Nontreatment default was de ned as an outcome other than treatment default, which included being cured, completed treatment, failed treatment, and death. Being cured was de ned as a negative sputum culture in the last month of treatment and on at least one previous occasion. Treatment completion was de ned as completing treatment without meeting the criteria for cured or treatment failure. Treatment failure was considered as TB patients whose sputum smear or culture was positive in the fth month or later during treatment. A patient who died before starting a treatment or during treatment was classi ed as a death, irrespective of the cause. The outcomes that were not included in this study included transferred-out cases or cases lost to follow-up (transfer to another country with unknown treatment outcome) and outcomes that were not evaluated (patients who were undergoing treatment and whose outcomes were not known).

Sample size calculation
The sample size was determined based on the 34.0% prevalence of smoking in the TB patient population registered in National MyTB version 2.1 in 2012 (11). The sample size was calculated using Epi-Info software based on an alpha of 0.05, a power of 80%, and a design effect of 1. By adjusting to 30% of the attrition rate, in which we consider the probability of missing data, the minimum sample required in this study was 420. However, all registered TB patients who smoke in the Selangor MyTB 2.1 database from 2013-2017 were considered for the analysis, making up the total sample size of 5396.

Variable
Twenty variables related to the objective of the study were examined. The four domains of the independent variables were sociodemographic characteristics (age, sex, nationality, ethnicity, locality, education level, personal income, and occupation), disease pro les (TB category, method of TB case detection, BCG scarring, type of TB, chest X-ray status and sputum status), comorbidities (diabetes mellitus and HIV) and treatment pro les (DOTS during the intensive phase, DOTS during the continuation phase, duration of TB treatment and place of TB treatment initiated). Smoking status was speci cally obtained by con rming whether the patient was an active smoker. There was no further information in terms of the duration or severity of smoking available from the database.

Data management
The data extraction ow is summarized in a ow diagram (Fig. 1). Data cleaning and processing were performed using R programming in view of the large amount of data. A total of 882 cases with exclusion criteria were excluded from the overall TB patients who smoke and were registered in the Selangor database. Any redundant data were reviewed, and missing data were treated with data imputation. Data were kept in two backup storage in both hard and soft copies.

Con dentiality
Patient identi cation and information in the database remained anonymous and were kept con dential.
Data will be stored for ve years in a password-protected hard disk and will be destroyed after that.

Statistical analysis
Data were analysed using the SPSS statistical software package, version 23.0. The descriptive analysis of the four main independent variable domains is presented in the form of a frequency table. For continuous variables, the description is expressed as the mean ± SD, and for categorical variables, descriptions are presented as the frequency (n) and percentage (%). Simple logistic regression and multiple logistic regressions (MlogR) were performed to estimate the risk of treatment default among TB patients who smoke. Signi cant results with p < 0.05 from the univariable analysis were considered in MlogR. Multivariable analysis was performed using backward LR. A value of p < 0.05 from the nal logistic model was considered statistically signi cant. Adjusted odds ratios (AORs) were used to present the results. The assumption of "linearity of logit" was checked for continuous independent variables (age). The presence of multicollinearity and interaction between the independent variables was checked.

Result
The prevalence A total of 22785 TB patients were registered in the Selangor MyTB database from 2013-2017. Out of these 22785 TB patients, 27.6% (N = 6278) were smokers. After removing 14.0% (N = 882) of the cases from the database that met the exclusion criteria, i.e., patients with missing data on treatment outcome, 16.2% (N = 143); cases that had a changed diagnosis other than TB, 11.9% (105); duplicate cases, 29.5% (N = 260); and cases with an outcome not evaluated, 40.8% (N = 360), a total of 5396 patients were included for analysis (Fig. 1). The prevalence of treatment default among the overall TB patients in Selangor was 10.3% (95% CI 9.9,10.7) versus a default rate of 15.1% (95% CI 14.1,15.9) for TB patients who smoke in Selangor.

Characteristics of TB patients who smoke
The mean age of the TB patients who smoked was 42.36 ± 14.61 years and ranged from 8 to 96 years old. The majority were male 95.8% (N = 5,167), Malaysian citizens 89.3% (N = 4,821), lived in the urban area 82.7% (N = 4,462), and of Malay ethnicity 57.3% (N = 3,091). Most of the patients had an education level equal to at least to secondary school, worked and reported an income level of less than RM 2160 (median personal income, DOSM 2019). In terms of the disease pro le, 91.8% (N = 4,955) were new TB cases, and 88.4% (N = 4.770) were pulmonary TB cases. A quarter of the TB patients who smoked had DM comorbidity 21% (N = 11,320), while 9.1% (N = 452) were reported as having HIV. TB treatment was mostly initiated in 3,727 government hospitals (69.1%), with a treatment duration of more than 6 months at 3,523 (65.3%). DOTS was utilized by 87.7% (N = 4,733) of the patients during the intensive phase, and the percentage was reduced to 72.6% (N = 3,917) during the continuation phase (Table 1).   among TB patients by itself has a poor prognosis for TB treatment outcomes. TB patients who smoke and who default on TB treatment will experience worse outcomes. Therefore, it is important to examine the factors associated with treatment default among TB patients who smoke to optimize their treatment adherence and assist them in quitting smoking.
Factors associated with treatment default.
Adherence to TB treatment strongly in uences the outcome of patients and has an effect on the development of multidrug resistance (MDR-TB)(28). In the current study, the majority of patients who default on TB treatment had sputum samples that were smear positive when they returned for retreatment care, which indicates a high risk of transmission to others (29). Despite full supervision in the form of the directly observed therapy that is currently being delivered in our setting, the outcome was not consistently improved. Sociodemographic factors associated with treatment default among TB patients who smoke were residing in an urban area, having a low education level and a low-income level (median individual income < RM2160), all of which had signi cant contribution to TB treatment default. This outcome could be possible because smoking is a marker for other social and behavioural factors that make defaulting on treatment more likely (29). A similar nding was also found in a qualitative study performed in urban Morocco, where a low income and a low level of education were barrier resources among TB patients which led to default. The reasons were due to lack of money for transportation, the need to work despite illness, and no one aiding in obtaining medication (30). This shows that socioeconomic support plays important roles in ensuring the continuation of TB treatment. Despite the full subsidization of anti-TB treatment to all TB patients, some out-of-pocket expenditures still exist, especially related transportation costs. An average of RM439.42 out-of-pocket money per patient has been estimated in order to complete a 6-month TB treatment (31). Any intervention that could reduce the cost of TB treatment will help to improve patient compliance with TB treatment.
Other studies from Hong Kong and Morocco have found that male sex and being a nonreligious person have signi cant associations with smoking habits and defaulting on TB treatment(6, 24). This could account for the large observed differences in the proportion of males and females in this study; however, sex was not included in the nal model in this study. The in uence of religious belief on the effect of patient adherence to TB treatment could not be quanti ed, as religious status was not available in the database.
Under the disease pro les domain, patients with a history of previously being treated for TB had an almost threefold risk of default treatment compared to new TB cases. It has been reported that previous experience with TB is a risk factor for defaulting only when there was a previous treatment default(6).
This could be related to smoking habits, as studies conducted among TB patients who smoke in Hong Kong have revealed a signi cant association between retreatment cases among current smokers and TB patients who never smoke (17). The complex psychosocial factors of smoking may explain its association with defaults and non-adherence; however, in this study, we did not address the underlying mechanism. Additional studies from other countries, such as Sudan, Morocco, and Brazil, also found a signi cant association between retreatment cases and TB treatment default, with ORs ranging from 3.2 to 6.5; this indicates that patients who had been defaulting their treatment will be at higher risk of defaulting on their TB treatment again (30,32,33).
This study also found that TB patients who smoke and who were detected through active screening methods had a double-risk AOR of 2.047 (95% CI 1.206-3.473) for defaulted treatment compared to those who were detected through passive detection methods. This nding could be the result of the implementation of the national guideline for systematic screening for TB high-risk groups, such as TB/HIV comorbidities, inmate prisoners, diabetes patients, elderly individuals and patients in methadone replacement therapy since 2015, where the majority of these high-risk groups were signi cantly associated with unfavourable TB treatment outcomes (NSPTB 2016-2020). This action was intended to improve TB detection rates and to provide early TB treatment to them. Most detected TB comorbidities, for example, TBHIV and TBDM, are known to be predictors of poor TB treatment outcomes in many studies. Studies from Brazil, Kenya and Peru have found that HIV-infected patients are at higher risk of defaulted treatment than are HIV-negative patients (23,34,35). The outcomes are similar for diabetes mellitus (DM); clinical evidence has found DM to be a signi cant risk factor for poor TB treatment outcomes, including treatment default. The literature has suggested that DM is signi cantly associated with the development of adverse drug reactions and delayed sputum conversion at the end of 2 months of treatment(36), which explains the high default rates among TB/DM patients in certain countries, including Kuwait and Brazil (37). In this study, TB patients who smoked and had HIV were signi cant in the univariable analysis but were not found to have an independent relationship to treatment default in the multiple logistic regression analysis. While TB/DM patients had no association with treatment default. The lower rates of treatment default among TB patients with comorbidities may be due to their better treatment compliance, as these patients are frequently followed up with for anti-retroviral treatment in HIV patients and chronic diabetes mellitus management (11 (39). Other common signi cant factors associated with non-adherence to DOTS are poor knowledge towards TB and its treatment, the cost of transportation for DOTS at every visit and the distance of the DOTS centre from individual's home (40).
Addressing smoking issues among TB patients requires a strategic plan on its own. A study performed among TB patients who smoke in Penang showed a poor score of tobacco use knowledge and its health consequences in general among newly diagnosed patients (27). Most patients report that they are not informed about the impact of continued smoking on TB outcomes and have only received general health information and not TB-speci c information (41). Evidence on the effects of smoking cessation on TB treatment outcomes, especially on treatment default, is limited. It is also not well known whether quitting smoking during TB treatment would have an immediate impact and produce similar outcomes as those of individuals who have never smoked. However, a study performed by Wang and Shen in Hong Kong found that TB patients without smoking cessation are twice as likely to default on TB treatment than are those who achieve cessation (OR 2.03; 95% CI 0.99-4.18). This outcome is similar to a local study performed in Malaysia, which found that TB patients who receive tobacco cessation intervention during DOTS have a lower rate of treatment default than TB patients in the usual care group (42). Additionally, the literature has shown that smoking-related immunological abnormalities in TB are reversible within six weeks of smoking cessation (43). Therefore, initiating tobacco cessation intervention during DOTS will bene t TB patients in terms of their treatment outcome and improve their adherence to TB treatment. A prospective cohort in Hong Kong also found that 49.6% of smokers who quit will remain nonsmoking 5 years after the cessation intervention while receiving TB treatment (44). This nding supports the longterm effect of the TB smoking cessation intervention delivered by TB chest clinics in reducing smoking prevalence among TB patients who smoke.

Strength And Limitations
The ndings from this study provide new knowledge about the characteristics and determinants of treatment default among TB patients who smoke. It also highlights the high prevalence of treatment default among TB patients who smoke for the attention of stakeholders to implement integrated intervention programmes for this speci c high-risk group by addressing both TB and smoking issues.
Other items of importance are the ndings on the in uence of social determinants among smokers on TB treatment defaults, such as education and economic status, and those on TB burden, especially in urban areas. In addition to patient characteristics, disease and treatment pro les were also identi ed as having signi cant associations with defaults. We did not nd a signi cant association between defaults and TB comorbidities, which is similar to other studies; this outcome could be due to different clinical or preventive approaches practised by healthcare settings. The large sample size taken from the registered TB patients in the Selangor MyTB database from 2013-2017 allowed the generalizability of the study ndings to the general TB population in Malaysia. However, this study has several limitations. Due to the massive data, advanced data software such as R software or STATA is required for data management and processing. The completeness of the database due to a high proportion of missing data affects the analysis for certain variables (HIV status, sputum status and X-ray status with missing values ranging from 1.3%-14.1%). There were also important missing variables from the database that could not be analysed, which limited our factor association analysis. Due to the limited information in the database in terms of smoking characteristics, we were unable to see the temporality of the associations between different smoking statuses among the patients (active smoker, ex-smoker, quit smoker) throughout TB treatment. Some TB patients might stop smoking after being diagnosed with TB; therefore, their smoking status should no longer be active. Our study also lacks details on the timely duration of patient treatment to quantify at which phases of treatment should we intensify our supervision to prevent defaults.

Recommendations
The reduction of the smoking prevalence among TB patients could be achieved through strengthening the inter-sectoral collaboration between units in the TB healthcare management system and reinforcing the communication or educational counselling quality between patients and health care providers.
Integrated TB-tobacco cessation invention programs must be initiated together in TB chest clinics to cater to both issues holistically at the same time. This integrated intervention should be included in the National Strategic Plan of TB for its better implementation.
Further observational studies with primary data are highly recommended. Another area for further research is to develop prognostic scoring tools for the earlier detection of the high-risk group to default TB treatment among those who smoke in the TB population based on the determinants that have been identi ed from this study. Extra supervision is required for TB patients who smoke and have strong predictors of defaulting.

Conclusion
This study identi ed determinants for treatment defaults among TB patients who smoke and focused on patients' socioeconomic pro les, disease and treatment pro les and comorbidities. The logistic models showed that the risk of defaults could be predicted for those who reside in an urban area, have no or low levels of formal education, have a low income, are a retreatment case, are a TB case detected through active methods, are TB patients receiving treatment for less than 6 months and are not on DOTS during the continuation phase of treatment. Early risk detection for combined smoking cessation intervention and close supervision during DOTS should be provided to reduce smoking prevalence among TB patients and ultimately improve TB treatment outcomes, speci cally regarding treatment default. Abbreviations TB: tuberculosis, DOTS: direct observation therapy short course, DM: diabetes mellitus, HIV: human immune de ciency virus, AOR: adjusted odds ratio, COR: crude odds ratio, SD: standard deviation. Figure 1 Flow diagram of data management ow