Effect of treatment adherence on the association between sex and unfavourable treatment outcomes among tuberculosis patients in Puducherry, India: a mediation analysis

Abstract

Background

A better understanding of the complex interplay between risk factors of tuberculosis (TB) is essential. This study was part of the Regional Prospective Observational Research for Tuberculosis (RePORT) India consortium and includes newly diagnosed TB patients in Puducherry between 2014 and 2018. We employed mediation analysis to identify the effect of treatment adherence on association between sex and unfavourable TB treatment outcomes.

Methods

Required demographic and treatment-related variables were extracted from the RePORT India consortium database and causal mediation analysis using parametric regression models was done.

Results

Of the 712 TB patients, ~87 (12.2%) had unfavourable TB treatment outcomes. Total effect of male sex was significantly associated with the unfavourable TB treatment outcomes [adjusted odds ratio (aOR) = 2.48; 95% confidence interval (CI): 1.11–5.55]. However, the overall association between male sex and TB treatment outcomes was dominated by the indirect pathway, as the direct pathway does not show significant association (aOR = 1.67; 95% CI: 0.75–3.75), while the indirect pathway shows significantly higher odds of TB treatment outcomes (aOR = 1.48; 95% CI:1.27–1.73), indicating complete mediation by the treatment adherence.

Conclusions

The study has shown a complete mediation of sexes through TB treatment adherence for unfavourable treatment outcomes. Developing of treatment strategies require better understanding between the biological and social factors related to TB.

Introduction

Tuberculosis (TB) is one of the leading causes of infectious disease-related morbidity and mortality around the world.¹ Although significant progress has been achieved in disease control activities, TB continues to be a global public health threat.¹ According to the Global TB report 2019, around 7 million people were effectively covered with better quality TB care in 2018, a rise from 6.4 million in 2017. There had been a substantial drop in TB-related deaths in 2018, compared to 2017.²

India remains as one of the prominent countries in the TB map of the world, by contributing nearly one-fourth of the global TB burden.³ Ever since the establishment of the Revised National Tuberculosis Control Program, India has ensured better planning, implementation and evaluation of TB prevention, diagnosis and treatment services. However, the program still faces several challenges such as lack of funding, poor access to health resources (infrastructure, testing facilities, drug availability) and lack of adherence. Early identification of disease, proper treatment and follow-up can aid us in progressing our steps towards achieving End TB targets in 2035.^4–6 Although the national TB control programme has expanded TB treatment coverage over years, very little progress is made in improving successful TB treatment outcomes due to unsuccessful efforts.⁷

A better understanding of the complex interplay between risk factors of TB treatment outcomes has critical implications for the global health. One such factor is the sex differences in the TB treatment outcomes. Though several studies have reported significant sex differences in the TB treatment outcomes with significantly worse outcomes among males, the biological reasoning for such differences is poorly understood and it requires further exploration.^8–10 Previous research on sex differences in TB outcomes has not focused on intermediate outcomes such as treatment adherence, as it might act as an intermediate pathway leading to unfavourable treatment outcomes.¹⁰ Mediation analysis answers such questions as it tells how much effect is directly caused by a factor, by calculating the extent to which the factors exert their effects on the outcomes via intermediate mediating variables.¹¹ Hence, we employed mediation analysis in a longitudinal data of newly diagnosed TB patients, to identify the effect of treatment adherence on the association between sex and unfavourable treatment outcomes among TB patients in Puducherry, South India.

Methods

Study setting and study population

We conducted a longitudinal analysis of data from an ongoing large-scale cohort study under the Regional Prospective Observational Research for Tuberculosis (RePORT) India consortium.¹² Jawaharlal Institute of Postgraduate Medical Education and Research (JIPMER) (in collaboration with the New Jersey Medical University—Rutgers University and Boston Medical Center) has developed two observational prospective cohorts. The first cohort has participants with active pulmonary TB and the second one has the household contacts of the active pulmonary TB patients in the first cohort. For our analysis, we used the cohort having participants with active TB.

This study included TB patients from three districts in the Southern part of India, Pondicherry from the Union territory (UT) of Puducherry, and two adjacent districts from the neighbouring state, Tamil Nadu—Villupuram and Cuddalore. Study enrolment began in May 2014 in Pondicherry, August 2014 in the Cuddalore district and November 2015 in the Villupuram district. For recruitment of the study participants, one Tuberculosis unit (TU) in Pondicherry and two TUs each in Cuddalore and Villupuram were selected. Designated microscopy centres (DMCs) and the Peripheral Health Institutions (PHIs) are responsible for service delivery with TU as nodal point. For our study, we enrolled only the newly diagnosed smear-positive TB patients (with at least 1+ acid-fast bacilli) aged 6 years or more and diagnosed at the DMCs and PHIs from the selected districts. We excluded patients with previous history of TB condition or treatment, patients who had received TB treatment for ≥1 week and patients with multidrug-resistant TB from the study. The detailed study protocol has been described previously.^13–16

Study procedure

We obtained ethical approval for the study from the JIPMER Scientific Advisory Committee and Institutional Ethics Committee of JIPMER, and institute review boards at New Jersey Medical University—Rutgers University and Boston Medical Center. We began data collection after obtaining informed written consent from adult participants (≥18 years) and assent form from participants less than 18 years in addition to parents’ consent. We used a pretested semi-structured questionnaire to gather the sociodemographic of the participants such as age, sex, education, occupation and marital status and measured height and weight; body mass index (BMI) was calculated. We also obtained details related to diabetes mellitus (DM), which was defined as random blood sugar >200 mg/dl or self-reported known case of DM. After the baseline assessment, we ascertained adherence to medications, sputum smear and culture results from NTEP records. Study personnel also performed repeat sputum smear and culture confirmation at the time of enrolment. We used NTEP data for adherence, sputum smear and culture results after 2 months of treatment (intensive phase) and end of treatment (continuation phase). All enrolled participants were encouraged to come for mycobateriological testing at the end of the treatment. Sputum AFB test was done for those who were able to produce sputum and declared cured if the smear was negative. However, many participants were not able to produce sputum at the end of treatment. Those who had completed the treatment successfully were declared as ‘treatment completed’ as per the guidelines of the national TB program.

There was a change in the algorithm for diagnosis of TB under the national TB program. However, this did not affect the type of patients being enrolled into the study as the inclusion criteria from the beginning of the study was culture positivity. We reported final treatment outcomes at the end of the treatment as bacteriological cure, clinical relapse, bacteriological/clinical failure, loss to follow-up, emerging resistance and death.

Study definitions

BMI category

We classified BMI into underweight (BMI < 18.50 kg/m²), normal (18.50–22.99 kg/m²), overweight (23.00–24.99 kg/m²) and obesity (≥25.00 kg/m²) based on the Asia-Pacific guidelines.¹⁷

Adherence to TB treatment

We retrieved these data from the treatment cards of patients at the end of intensive phase and the end of treatment. We considered patients who never skipped medications during both the phases to have good adherence to TB treatment.

Treatment adherence was assessed through the TB treatment cards being maintained by the national TB programme personnel. Directly Observed Treatment Short Course (DOTS) was in implementation during the period of the study. DOTS providers in the study area were either the medical social workers, auxiliary nurse midwifes, TB health visitors, staff nurses or pharmacists of the Primary Health Centres from where the patients were receiving the treatment. DOTS providers ticked on the boxes assigned for the days when a patient was supposed to take the drugs during intensive phase and weekly once on showing the empty blister packs during the continuation phase. The treatment cards for the participants enrolled in the study were accessed and collected as source documents at the end of the study.

TB treatment outcomes

We defined treatment outcomes based on the initial RePORT consortium protocol and not the recently published WHO definitions for TB treatment outcomes.¹⁸

Bacteriological cure

Participants completing the recommended drug regimen and has a documented negative culture result at the end of the treatment.

Bacteriological failure

Participants having positive sputum culture results even after the complete period of treatment regimen for Tuberculosis and the result was not found to be false positive.

Clinical failure

Participants who had completed at least 4 months of treatment regimen, (6 months) but still have persistent signs and symptoms, progression of Tdisease or recurrence of symptoms and/or signs of TB, found to be because of TB only and not because of any other underlying causes.

Emerging resistance

Participant who had a change in baseline drug sensitivity before the drug sensitive bacteriological failure can be determined (i.e. after baseline visit, but before 5 months of treatment).

Death

Participant has died of any cause after enrolment into the study and before the study completion.

Unfavourable treatment outcome

Participants satisfying the criteria of bacteriological or clinical failure/emerging resistance/clinical relapse/death were classified as having unfavourable treatment outcome.

Statistical analysis

We scanned and transmitted completed questionnaires to the Boston Medical Center using Verity tele-form information capture system version 10.8 (Sunnyvale, CA, USA) so that they could be read into the Microsoft Access database (Seattle, WA, USA). We reviewed the data entry process for errors and the onsite team in India made necessary corrections. For our analysis, we extracted data from the RePORT India project database for the JIPMER site and performed analysis using the Stata version 14.2 software. We generated descriptive analysis by summarizing the continuous variables as mean and standard deviation (SD) or median (IQR) and categorical variables as proportions.

We performed causal mediation analysis using parametric regression models to estimate the natural direct effect (NDE), controlled direct effect (CDE) and natural indirect effect (NIE) of sex and TB treatment outcomes keeping treatment adherence as mediator variable. The NDE reports the relationship between exposure and outcome without mediator variable; and NIE reports the part of exposure–outcome relationship incorporating the mediator.

The following equations fit a simple mediation model with multiple regression with separate model fitted for each equation.¹⁹ Since the mediator variable is dichotomous, we fitted all the following equations using logistic regression.

(1)

(2)

(3)

In Equation (1), Y represents the outcome variable (TB treatment outcome), and cX represents the slope of exposure variable (sex). In Equation (2), M represents the mediator variable (treatment adherence), and aX represents the slope of exposure variable (sex). In Equation (3), Y represents the outcome variable (TB treatment outcome), c′X represents the slope of exposure variable (sex) and bM represents the slope of mediator variable (treatment variable). In all the above equations, i represents the intercept.

The odds ratio for NIE and NDE is calculated using the following formula¹⁹:

(4)

(5)

Where i₂ represents the intercept from Equation (2), b represents the b coefficient from Equation (3), a represents Coefficient in Equation 2 and c′ represents the c’ coefficient from Equation (3).

We calculated the relative magnitude of mediated effect using proportion mediated (PM) and ratio of mediated to the non-mediated effect. We used MacKinnon’s formula to calculate the PM²⁰:

(6)

(7)

We depict the assumed association between the variables in Fig. 1. The direction of the arrow represents the direction of the effect. We hypothesized that the sex has direct and indirect effects (via treatment adherence) on unfavourable TB treatment outcomes.

Fig. 1.

Causal diagram on the hypothesized relationships between sex, treatment adherence and tuberculosis treatment outcome.

We adopted the Baron and Kenny approach²¹ to assess the association between independent (sex) and mediator variable (treatment adherence), mediator variable (treatment adherence) and dependent variable (TB treatment outcomes) and independent (sex) and dependent variable (TB treatment outcomes) with logistic regression models. The association between independent variable (sex) and outcome variable (TB treatment outcomes) as well as mediator variable (TB treatment adherence) was statistically significant with P-value <0.05. Since the basic assumptions were satisfied, we performed mediation analysis using the ‘paramed’ package.²² We have adjusted for potential confounders such as age, education, occupation, marital status, household income, smoking, BMI category and presence of DM in our causal pathway using backward stepwise logistic regression and reported the final effect estimate as adjusted odds ratio (aOR) with 95% confidence interval (CI). Using the variables which became significant (P value <0.2) in the univariate analysis, multivariable analysis was performed. We have also applied bootstrapping technique (500 iterations) to examine the significance of indirect and direct effects with the bootstrapped bias-corrected CIs.

Results

In total, 712 eligible TB patients had all the required information and included in the analysis. The characteristics of the study participants are described in Table 1. The median Inter-quartile range (IQR) age of the study participants was 45 (33–55) years. Majority (537; 75.4%) participants were males and 517 (72.6%) were employed; about 118 (16.6%) participants had no formal education; majority (623 participants; 87.8%) were Hindus; 508 participants (71.3%) were currently married. Eighty-seven (12.2%) of the study participants had unfavourable TB treatment outcomes. Males had higher proportion of unfavourable TB treatment outcomes (14.3%) when compared to females (5.7%) and this difference was statistically significant in univariable analysis (P = 0.002).

Table 1.

Socio-demographic characteristics and treatment outcomes of the study participants (N = 712)

Sl no.	Characteristics	Frequency n (%)
1	Age categories, in years
	≤19	50 (7.0)
	20–29	94 (13.2)
	30–39	116 (16.3)
	40–49	185 (26.0)
	50–59	149 (20.9)
	≥60	118 (16.6)
2	Sex
	Female	175 (24.6)
	Male	537 (75.4)
3	BMI categorya
	Underweight (<18.50)	419 (58.8)
	Normal (18.50–22.99)	219 (30.8)
	Overweight (23.00–24.99)	45 (6.3)
	Obesity (≥25.00)	29 (4.1)
4	Educational status
	No formal education	118 (16.6)
	Primary school	156 (21.9)
	Secondary school	295 (41.4)
	Higher secondary	143 (20.1)
5	Marital status
	Currently married	508 (71.3)
	Never married	128 (18.0)
	Widowed/divorced/separated	76 (10.7)
6	Employment status
	Unemployed	71 (10.0)
	Employed	517 (72.6)
	Othersb	124 (17.4)
7	Diabetes mellitus status
	Present	252 (35.4)
	Absent	460 (64.6)
8	Treatment outcome
	Favourable	625 (87.8)
	Unfavourable	87 (12.2)

^aAsia Pacific guidelines for obesity.

^bHousewife, students.

Table 2 reports the causal NDE, NIE and total effects mediated by the treatment adherence in our casual pathway. In the mediation analysis, we found that the male sex was associated with significantly higher odds of non-adherence to treatment (aOR = 2.51; 95%CI: 1.51–4.18), while the non-adherence was associated with higher odds of unfavourable TB treatment outcomes (aOR = 11.51; 95%CI: 6.68–19.84). The total effect of male sex was significantly associated with the unfavourable TB treatment outcomes (aOR = 2.48; 95%CI: 1.11–5.55). However, the overall association between male sex and TB treatment outcomes was dominated by the indirect pathway. We also found that the direct pathway (NDE) does not show significant association between sex and TB treatment outcomes (aOR = 1.67; 95%CI: 0.75–3.75), while the indirect pathway (NIE) shows significantly higher odds of TB treatment outcomes (aOR = 1.48; 95%CI: 1.27–1.73), indicating complete mediation by the treatment adherence. We also estimated the CI for the NDE, NIE and total effects using bootstrapping. However, there were no meaningful differences in the CI even with the large number of replications. The relative magnitude of mediated effect was calculated using PM and ratio of mediated to the non-mediated effect. PM was estimated to be 59.7% and ratio of mediated to non-mediated effect was 0.89.

Table 2.

Results of causal mediation analysis with non-adherence to TB treatment as mediator between sex and unfavourable TB treatment outcomes in the cohort of newly diagnosed TB patients at Puducherry, South India

Effect	Exposure	Adjusted OR (95% CI)a	P-value	Bias-corrected CIs
Total effect of sex on unfavourable TB treatment outcomes	Males	2.48 (1.11–5.55)	0.03	1.18–6.53
Females	Ref	—	—
Direct effect of sex on unfavourable TB treatment outcomes	Males	1.67 (0.75–3.75)	0.21	0.76–4.21
Females	Ref	—	—
Indirect effect of sex on unfavourable TB treatment outcomes	Males	1.48 (1.27–1.73)	<0.001	1.27–1.74
Females	Ref	—	—
Effect of sex on non-adherence to TB treatment	Males	2.51 (1.51–4.18)	<0.001	[Not Applicable]
Females	Ref	—
Effect of treatment adherence on unfavourable TB treatment outcomes	Non-compliant	11.51 (6.68–19.84)
Compliant	Ref	—

Ref, reference category.

^aThe mediation analysis was done by constructing two logistic regression models: (i) the first model regressed the risk of non-adherence to TB treatment based on sex, adjusting for age (categorical), marital status, education, employment status, BMI category and DM; (ii) the second model regressed the risk of adverse TB treatment outcomes based on sex and non-adherence to TB treatment, adjusting for age (categorical), marital status, education, employment status, BMI category and DM.

Discussion

Main findings

A successful TB treatment outcome in any country is a factor of patient characteristics, effectiveness and coverage of treatment programs. Unfavourable TB treatment outcomes are influenced by several sociodemographic, behavioural and clinical characteristics. The effect of sociodemographic factors on the treatment outcomes is a complex phenomenon.²³ These factors, by interacting with each other, finally affect the outcome directly or indirectly.^24,25 Examining the role of sex in the TB treatment outcomes is necessary for overcoming the barriers to effective management and treatment coverage of TB. Adherence to TB treatment is another vital dimension and plays a key role in achieving successful treatment outcomes. Though several lines of evidence have explored these factors individually, the possibility of these factors in a causal pathway has yet not been explored.^8–10 Hence, we have conducted this study to identify the role of treatment adherence in the association between sex and successful TB treatment outcomes via causal mediation analysis.

Our study had an unfavourable outcome rate in line with the neighbours such as China.⁹ However, the figures were contrastingly lower when compared to studies done in Northern India, postulating a better TB control activity in the Southern part of the country.²⁶ Our study showed that adherence to TB treatment has a strong indirect effect in the association between sex and unfavourable TB treatment outcomes. We also found that there was no direct effect between sex and treatment outcomes, indicating complete mediation via treatment adherence in the causal mediation pathway.

What is already known on this topic

Previous reports have assessed the association between sex and treatment adherence and have tried to reason out why the males have poorer adherence compared to females.^27–29 One such reason is economic aspect of the family, as the males are major contributors for family income, which makes it difficult for them to take time out for a visit to hospital or clinic for follow-up medications. In addition to family and financial pressure, work-related pressure conditions were also cited as common reasons by males for not complying to the TB treatment.^27–29 Among females, feeling of responsibility to care of their family and children was reported to be a major motivating factor for adherence to treatment.³⁰ The same feeling of care has also been seen in males.³¹ However, females especially mothers taking care of multiple children also has difficulty to visiting the clinic for medication. Hence, sex-specific support systems that differentially target towards men and women should be implemented within the current standard treatment guidelines in National TB Elimination Programme in the country and it requires some operational consideration.

What this study adds

Research on sex differences in TB outcomes has thus far focused on co-existing comorbidities and various sociodemographic factors like age, income, etc.³⁰ Only limited research has focused on the adherence part as a factor responsible for sex differences in the treatment outcomes. It is axiomatic that the program needs to improve the patient engagement and adherence to approved treatment regions. However, our study suggests that a greater focus on males is required to improve their TB treatment outcomes. This is also supported by the recent meta-analysis findings that reported males do not access TB services and suggested that males were a high-risk group who require an improved access to TB services.³² Other sociodemographic, clinical or genetic factors responsible for unfavourable outcomes should be explored to further assist in directing the public health responses.

The major strength of the study is the use of longitudinal data to develop and model a causal mediation pathway. Another strength of this paper is that we have provided NDE, NIE and CDE, in the presence of exposure–mediator interaction. To the best of our knowledge, this is the first study to assess the causal pathway between sex and TB treatment outcomes via adherence to TB treatment.

Limitations

Despite these strengths, the study estimates should be carefully interpreted owing to certain limitations. We have utilized the TB treatment cards verified by DOTS providers to assess the level of adherence and treatment outcomes. It can be less accurate when compared to urine drug testing or other better adherence measures. But, compared to self reporting of adherence by patients, TB treatment cards verified by DOTS providers limits the possibility of recall or reporting bias. The dose–response relationship has not been assessed in this study. We could not adjust for socioeconomic status, delay in diagnosis, tobacco use, alcohol use, type of residence, sputum smear grading at baseline, cavitary lung disease, due to missing information for large number of patients, limiting the sample size to conduct mediation analysis. The estimation of NDE, NIE and causal interpretation requires that there is no unmeasured exposure–mediator confounding, and no mediator–outcome confounding be affected by the exposure.³³ It is unrealistic to satisfy both these assumptions, given the limited set of variables adjusted in the model. We could not perform sensitivity analysis that estimates the PM using different method, i.e. structural equation modelling (SEM), owing to the dichotomous nature of our variables. SEM requires continuous variables for application.³⁴

Despite these limitations, our study has important clinical and public health implications for understanding the association between sex, treatment adherence and outcomes. On the basis of longitudinal association identified in our study, adherence to treatment appears to be intertwined with the association between sex and adverse treatment outcomes. From a public health or policy standpoint, this possibility has important implications for developing a sex-specific treatment strategy for TB patients. However, this study finding does not diminish the importance of sex, which has significant role in the clinical endpoints of multitude of diseases including TB. Strategies for engaging males in care may be different than those that are effective for females. Further research on the influence of sex and adherence to treatment in the causal pathway of treatment outcomes could shed light on the possible reasons, mechanisms and establish the causal link. Better understanding of the complex interplay between these biological and social factors is necessary to develop appropriate treatment strategies and help India fight the burden of TB epidemic, so that we will be in the path to achieve the aspiring End TB goal for 2035.

Conflict of interest

The authors have declared that no competing interests exist.

Funding

This work was supported by in whole or in part with extramural grant of Federal funds from the Government of India’s (GOI) Department of Biotechnology (DBT), the United States National Institutes of Health (NIH), National Institute of Allergy and Infectious Diseases (NIAID), Office of AIDS Research (OAR) and distributed in part by CRDF Global. The contents of this publication are solely the responsibility of the authors and do not represent the official views of the DBT, the NIH or CRDF Global. Any mention of trade names, commercial projects or organizations does not imply endorsement by any of the sponsoring organizations. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. There was no additional external funding received for this study.

Arivarasan Barathi, Senior Resident

Yuvaraj Krishnamoorthy, Senior Resident

Pranay Sinha, Post-doctoral Fellow

Charles Horsburgh, Professor

Natasha Hochberg, Associate Professor

Evan Johnson, Associate Professor

Padmini Salgame, Professor

Soundappan Govindarajan, State TB Officer

Senbagavalli Prakash Babu, Chief Manager

Subitha Lakshinarayanan, Additional Professor

Gautam Roy, Professor (Senior Scale)

Sonali Sarkar, Professor