Demographic and Clinical Determinants of Tuberculosis and TB Recurrence: A Double-Edged Retrospective Study from Pakistan

Objective TB recurrence is the second episode of TB after initial treatment bringing about an additional 7% load in TB burden intensified by 17.7% of multidrug-resistant recurrent cases. It is necessary to curb recurrence so that attempts to deal with active disease can be made more effective. This study aimed to characterize sociodemographic and clinical factors associated with recurrent TB in a high-burden setting. Methodology. A retrospective case-control study was carried out at two hospitals in Rawalpindi, Pakistan. TB patients and controls were included in the study. Sociodemographic and clinical data were collected by questionnaire from all subjects. Multivariate regression analysis was performed to determine factors associated with TB and TB recurrence respectively. Results In our study cohort, factors significantly associated with TB were low BMI (OR: 0.961 (CI 0.954–0.968), p < 0.001), female gender (OR: 2.065 (CI 1.922–2.219), p < 0.001), being single/unmarried (OR: 1.214 (CI 1.109–1.328), p=0.003), middle-income status (OR: 1.935 (CI 1.616–2.323), p < 0.001), smoking (OR: 1.567 (CI 1.435–1.710), p < 0.001), and diabetes mellitus (OR: 1.142 (CI 1.017–1.278), p=0.023). TB recurrence constituted 11.2% of patients presenting to the hospital. Compared with the first episode of TB, cases with recurrence were more likely to be older (OR: 1.011 (CI 1.004–1.017), p < 0.001), have disease awareness (OR: 1.906 (CI 1.486–2.437), p < 0.001), smear positive (OR: 2.384 (CI 1.650–3.536), p < 0.001), and be drug-resistant (OR: 5.615 (CI 4.265–7.386), p < 0.001). Conclusion In the present study cohort, low BMI, female gender, being single, middle-income status, being unemployed, smoking, and being diabetic came out to be the sociodemographic and clinical risk factors for TB. Further exploring the TB cases increasing age, drug resistance and smear positivity stood out to be the major sociodemographic and clinical factors of TB recurrence despite high disease awareness.


Introduction
Tuberculosis (TB) is an airborne infectious disease, caused by Mycobacterium tuberculosis (Mtb), that is still afecting the world despite global eforts and funding to eradicate it. According to the Global Tuberculosis Report of 2020, over 10.0 million people were afected and 1.4 million people died in 2019 [1,2]. It is representing only a 9% reduction in TB incidence from 2015 to 2019 and is well below the target reduction of 20% needed to be on track with the World Health Organization (WHO) END-TB strategy [1]. Tis slow decline in TB reduction rate underlines the global need for additional efective TB-control measures. Globally, an additional 7% of the TB burden is caused by TB recurrence in previously treated individuals [1] and this number is signifcantly higher in high-incidence countries [3][4][5]. Drug resistance is another tier of disease burden with a prevalence rate of 17.7% in relapse cases as compared to 3.3% in new cases [1].
Pakistan is a TB endemic country with a disease burden of 5.7% [1], accounting for 61% of the burden in WHO's Eastern Mediterranean Region [24]. Tere is no countrylevel report on the recurrence rate from Pakistan; however, recently, a hospital-based study from Rawalpindi, Pakistan, reported 5.83% of prevalent pulmonary TB cases to have the recurrent disease [25]. In a resource-restricted setting like Pakistan where molecular genotyping for strain identifcation and follow-up study needs additional funding sources investigating recurrence rate and social determinants of TB and TB recurrence can facilitate TB case reporting. To investigate this, we performed a retrospective case-control study at two large hospitals in Rawalpindi, a commercial and industrial city in Pakistan, with demographic characteristics representative of the nation.

Methods
A retrospective case-control study involving 3999 subjects was conducted between May, 2018, and September, 2019, to explore sociodemographic and clinical aspects of TB and its recurrence in Pakistan. Te study was approved by the ethics committee of Pir Mehr Ali Shah Arid Agriculture University, Rawalpindi, as well as by the ethics committee of Federal Government Tuberculosis Hospital Asghar Mall (Approval number: E-7(4)/2012-TBC/Training) involving human subjects. Data were collected through a designed questionnaire and informed written consent was taken from all subjects visiting the outpatient departments (OPDs) of Leprosy Hospital, Rawalpindi, and Federal Government TB Hospital, Rawalpindi. Both these hospitals are reference centers for patients of Rawalpindi-Islamabad. Among total subjects, 2000 were TB cases defned by evidence of abnormalities on chest X-ray compatible with TB, bacterially confrmed (sputum positive), and taking anti-TB medication. In some cases, along with chest X-ray presence of Mtb was detected with Xpert Mtb-RIF (Cepheid Inc). Te remaining 1999 subjects were considered controls who visited hospitals either due to any other illness or are household contacts of TB patients who also underwent TB diagnosis and were excluded through chest X-ray.
Individuals of all ages (from <1 year to 92 years) were included in the study and no matching for age and gender was done. Among the sociodemographic factors, marital status, body mass index (BMI), educational status at three levels (uneducated, secondary level, and higher secondary level), socioeconomic status at three levels (low, i.e., PKR < 25000, middle, i.e., PKR 25001-40000, and high, i.e., PKR > 40000), provincial distribution (Punjabi and non-Punjabi), family history of disease, and employment status (employed and unemployed) were recorded. Among the other factors, Bacillus Calmette-Guerin (BCG) vaccination, HIV status, disease awareness, smoking, and co-morbidities like diabetes mellitus (DM) and cardiovascular diseases (CVDs), acid-fast bacillus (AFB) sputum test at three levels (nil, negative and positive), and drug resistance (drugresistant and drug-susceptible) as well as for the presence of Mtb Xpert Mtb-RIF test (tested and not tested) were computed.
Te data were analyzed using R software (R-4.1.1) [26]. Te Shapiro-Wilk test was performed to assess for the distribution of continuous variables (age and BMI). Te data were summarized by the descriptive statistics using frequency (percentage) of the sociodemographic and clinical factors and median and interquartile range of continuous variables. Univariate analysis was done using the Chi-square test (p < 0.05) for the association of sociodemographic and clinical factors with disease status as the dependent variable. Pearson correlation was used for measuring the association of factors with one another and only those variables with a correlation coefcient value of <0.6 were included in the multivariate model (Supplementary File (available here)). Finally, a multivariate logistic regression model was used, and the measure of association was odds ratio (OR) and 95% confdence intervals (95%CI). A p value of <0.05 was considered statistically signifcant.

Sociodemographic and Clinical Characteristics of Patients with TB.
Te sociodemographic and clinical characteristics of both groups are summarized (Table 1). In univariate analyses, compared with controls attending our two hospitals, cases were younger (median age 30), had a lower BMI (median 18.5) comprised a higher proportion of males (60.4%), and a larger proportion was employed (53.4%). 75.1% of the cases belong to Punjab province, 87.2% have monthly income <25000, and 38.4% were uneducated. Of the clinical characteristics compared between the two groups, smoking was more prevalent in cases compared to controls. Te HIV co-infection rate in our cohort was very low (<1% in both groups) and only one-quarter of patients had a history of BCG vaccination. Te univariate analysis resulted in gender, occupation, provincial status, marital status, socioeconomic status, educational status, BCG vaccination, family history, smoking, and diabetes mellitus being associated with cases. Although statistically signifcant, variables like occupation and BCG vaccination were associated with gender and marital status (r ≥ 0.6), respectively. Tese two variables were excluded from multivariate analysis.
We then proceeded to perform a multivariate logistic regression analysis that is shown in Table 1

TB Recurrence.
To identify factors associated with TB recurrence, we compared patients with the second episode of TB with primary TB cases. Tere were 223 patients (11.2%) with recurrent TB in our cohort. Te sociodemographic, clinical, and TB diagnostic characteristics of patients with primary and recurrent TB are given in Table 2. In univariate analyses, patients with recurrent TB were older (median 32), and a lower proportion had a history of BCG vaccination (19.7%). Important diferences were observed in disease characteristics of patients with recurrent TB, with a higher proportion having pulmonary TB (91%), smear positivity in sputum (75.8%), and drug-resistant (39%). We also note that 81.2% of TB patients with recurrent TB have Xpert Mtb-RIF testing; however, this proportion was only 33.6% in patients with primary TB.
In the multivariate regression model, age, BMI, disease awareness, AFB sputum, and drug resistance were included. All of these were found to be signifcantly associated with recurrent TB (

Discussion
Tis large case-control study of patient-level data from two hospitals in Pakistan is representative of the wider community in the region and refects the epidemiology of TB in this high-burden setting. We have investigated the sociodemographic and clinical characteristics of TB patients and identifed that female gender, low BMI, middle-income status, single/unmarried individuals, and smoking as factors associated with TB cases. Further, we found that smear positivity, drug resistance, and increasing age are signifcantly associated with TB recurrence despite disease awareness in this population.
In our study, around 60% of TB patients were males similar to the 65% global prevalence of TB in males reported by the World Health Organization in 2016 [27]. Interestingly, the female gender has two-fold higher odds of developing TB compared to males in our analysis. Our results are consistent with many studies in diferent settings that reported a higher prevalence rate in men but an association with TB is higher in females [28][29][30]. Tis has been explained by sociocultural diferences that place a lower     Journal of Tropical Medicine priority on women's health and are not supposed to seek healthcare for themselves. According to results of the present study, TB risk is 1.2 times higher in unmarried individuals compared to married individuals. In 2018, a systematic review of demographic factors conducted by Mohidem in Malaysia reported that single individuals are more likely to have TB [31]. Similarly, a case-control study in three countries of West Africa reported being single as a risk factor of TB [32]. Lack of emotional support and socioeconomic difculties faced by unmarried individuals might be the potential reason for the higher risk of TB among unmarried persons.
Findings of the current study revealed association of low BMI with TB and are consistent with many studies in other high TB burden countries including South Africa [33], India [34], Sri Lanka [35], Korea [36], Taiwan [37,38], and Colombia [39]. Tere is a reverse logarithmic relationship between BMI and TB. Tis relationship is attributable in part to weight loss before TB infection as evidenced by a recent study in 2020 that showed a 3-fold higher prevalence of underweights in TB than controls [40]. Body mass refects adipose tissues and individuals with low visceral fat are vulnerable to TB because of a decrease in both proinfammatory cytokines like tumor necrosis factor (TNF) and a T-cell population that afects adaptive immune response by T cells in TB [41,42].
We found a medium-income group represented by a monthly income of PKR. 25001-40000 (i.e., USD 123-197) have a 1.9 times higher risk of developing TB compared to those earning more than PKR. 40000. Tis is indicative of a low socioeconomic class as this income is a source of livelihood both for the earner as well as dependents. Socioeconomic life afects lifestyle by changing the priorities such as care seeking, nutritional status, and living conditions. Low socioeconomic status and poverty are important indicators of TB documented in many studies in diferent settings with increasing poverty increasing the risk of TB [43][44][45][46].
We investigated that there is a higher risk of TB in smokers compared to nonsmokers as documented in many previous studies [47][48][49][50][51]. Smoking promotes the adherence of bacteria by changing the mucociliary function of bronchi and exacerbating pulmonary infammation and oxidative stress which contributes to increased lung damage and cavitation with pulmonary TB and in some studies is reported to increase the risk of TB recurrence [52].
Diabetes mellitus (DM) is another risk factor for tuberculosis in our cohort like several previous studies [53][54][55]. Tis is because DM alters the immune response in the host increasing the chance of TB infection [55,56]. DM decreases the IFN-c level [57,58] as well as the antigenpresenting ability of macrophages. IL-8 and IL-22 drop in case of TB-DM co-disease state afecting phagocytosis and TB clearance [59].
Te rate of recurrence diagnosed in our study was 11.2%, which is higher than the previous report (5.83%) from Pakistan published in 2021 [25] and studies from other regions of the globe, i.e., 3.1% in Shanghai [11], 3.1% in Malawi [60], and 8.6% in Vietnam [61] but lower than Uzbekistan where the recurrence rate was 36% [62]. Recurrence rates are higher in high TB burden settings compared to European countries as represented by a recurrence rate of 0.66% in England and Wales [16] and 1.3% in Spain [63]. Te higher recurrence rate we observed in our cohort might refect the patient population attending the hospital although the prevalence of recurrent TB in this setting may be underestimated. Tis is because communitybased TB diagnosis and treatment have not been fully captured in this cohort.
We established that older age is associated with recurrence. As immunity decreases with age and people with suppressed immunity are prone to infectious diseases and in a TB high burden setting chances of recurrence increase. A retrospective case-control study conducted in Singapore reported age ≥60 years to be related to TB recurrence [64] while a study from Pakistan reported a high recurrence rate in the younger age group (15-45 years) [25]. Te diference between these reports likely refects the diference in the etiology of recurrence. Singapore has a moderate TB incidence and TB elimination program has assured treatment completion rate. Terefore, recurrence is more likely to be due to reinfection. In Pakistan, relapse of disease arising from treatment failure is more likely because there is no tracking of treatment completion in Pakistan. Te median age of patients with recurrent TB in our cohort was 32 years, consistent with the reported age range in Pakistan.
Te association of sputum smear positivity with TB recurrence in our cohort refects an increase in mycobacterial load, increasing the chance of recurrence as it is highly infectious. In a study conducted in California for investigating recurrence rate and risk factors associated with late recurrence, it was found that sputum smear-positive disease during primary TB is associated with late recurrence [17]. A patient-level pooled analysis of treatment-shortening regimens involving 3411 participants was conducted in 2018. Te fndings of the study showed that high smear grade resulted in treatment failure elevating the risk of recurrence [65]. In a study conducted in South Africa on 500 smearpositive pulmonary tuberculosis patients, 11% of the patients who successfully completed their treatment had TB recurrence. Recurrence in these patients is associated with high smear grade during the frst episode of TB [66].
Drug resistance is another important factor associated with recurrent disease in the current study. We are unable to determine whether patients with drug-resistant recurrent TB in our study had this phenotype of disease at frst presentation or acquired resistance at recurrence due to a lack of knowledge of drug resistance testing during frst episode. A follow-up study in Henan province of China showed MDR-TB as a predictor for recurrence [22]. Te odds of recurrence for MDR-TB patients reported from China are lower than in the present study and the reason might be a diference in the study design. A retrospective observational study from Uzbekistan also reported MDR-TB to be a risk factor for recurrence [62]. Both the abovementioned studies included participants who have successfully completed treatment. Te higher odds of TB recurrence in the current study might be because of incomplete treatment and Journal of Tropical Medicine it is well-recognized that treatment failure carries an incremental risk of developing a drug-resistant disease.
Besides those sociodemographic and clinical factors included in this study, treatment failure due to inappropriate treatment is reported in 0.4%-45% of patients [67]. A study from India reported a 26% prevalence of inappropriate treatment [68] and another study from Benin reported a ≥10% prevalence in children [69]. Inappropriate treatment leading to treatment failure might be the reason for recurrence because of drugs with low bactericidal efects, inadequate treatment duration, and ignoring pre-existing drug resistance [70]. In Pakistan, a nine-month regimen including four drugs (isoniazid, rifampicin, ethambutol, and pyrazinamide) is recommended for treatment without testing for drug resistance. Still, medication use is not proper according to the physician's prescription. Although a positive response has been recorded for treatment completion by almost 62% of the patients in our cohort; however, this does not represent a true picture in all cases because patients comply in terms of getting medicines from healthcare centers but do not comply with using them. Tough expensive, supervised treatment is a better option to ensure proper medication use and prevent illicit drug deviation [70].
A signifcant proportion of recurrent TB cases in our large case-control study is representative of the national picture. Tese cases are more often highly infectious leading to greater onward transmission, and a higher proportion are drug-resistant requiring longer and more complex treatment regimens to achieve cure. Strategies focused on improved management of patients presenting with TB as a frst episode, including enhanced DOTS (directly observed treatment, short course) program, are likely to be cost-efective if they can succeed in improving treatment failure rates that predispose to recurrent TB.

Limitations of the Study.
Our study is limited by the absence of information relating to previous TB treatment and outcomes in our cohort with a recurrent disease, which requires a longitudinal study. We also do not have data on the infecting strain at the times of initial and recurrent disease. We are therefore unable to distinguish between relapse and reinfection. Data on Xpert Mtb/RIF test for drug resistance are also missing during the frst episode. Overcoming these shortcomings in clinical data integration for each incident TB episode and accessibility of tests for strain identifcation at each episode would empower a better study design in the future. Improved data infrastructure is needed to enhance the utility of surveillance studies like ours.

Conclusion.
In this large case-control study, we sorted individuals with low BMI, having female gender, single/unmarried status, low socioeconomic status, being diabetic, and smokers to be at risk for developing TB. Among the cases, AFB sputum positivity, drug resistance, age, and awareness are identifed as risk factors for TB recurrence. We recommend target screening based on the identifed risk factors and further studies for investigating the time of recurrence and diferentiating between relapse and reinfection.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon request.

Ethical Approval
Tis study was conducted following the principles of the Declaration of Helsinki and approved by the ethics committee of the Federal Government Tuberculosis Centre Asghar Mall (Approval number: E-7(4)/2012-TBC/Training) involving human subjects and by the ethics committee of PMAS-AAUR for conducting research on human subjects.

Consent
Study subjects (or their guardians/ caregivers) willingly participated in this health investigation through informed written consent. Data confdentiality was guaranteed.

Conflicts of Interest
Te authors declare that they have no conficts of interest.

Supplementary Materials
Correlation coefcient values are presented in Table S1 and  Table S2. Table S1: Correlation coefcient values showing the association of factors with one another for inclusion of independent factors in the multivariate model for TB cases.