Adherence interventions and outcomes of tuberculosis treatment: A systematic review and meta-analysis of trials and observational studies

1Adherence to treatment is challenging, given the complexity, modest tolerability, and long duration of treatment regimens currently available for both drug-susceptible and -resistant tuberculosis (TB). In turn, low adherence increases the risk of poor outcomes, including treatment failure, relapse, and development or amplification of drug resistance. Public health programs have used a variety of strategies to improve adherence at the health system level via financial incentives or enablers to offset the cost of accessing treatment, improving coordination and logistics around TB treatment delivery, and training healthcare providers. Other strategies tackle barriers to completing TB treatment by addressing knowledge gaps, attitudes, and behaviors surrounding adherence to TB treatment. One of the most commonly used adherence interventions is directly observed therapy (DOT), in which a health worker, family member, or community member observes the patient taking TB medications. In recent years, video-observed therapy (VOT) has gained attention as an alternative way of delivering DOT. Other interventions aimed at supporting adherence through DOT include incentives, which are material or financial rewards provided to those adhering treatment, and enablers, which are interventions that allow patients to overcome economic constraints associated with DOT, such as absence from work or the direct and indirect patient costs of accessing TB treatment. Other interventions focus on providing education on TB, its treatment, and prevention to help patients make informed decisions and the healthcare team to deliver patient-centered care. Reminder systems and patient tracers are targeted at assisting patients to keep appointments and to take action when patients miss appointments. These interventions include reminder letters, phone calls, home visits, and, more recently, short message service (SMS) technology as well as electronic pill boxes. Psychological interventions aim to support via psychological or emotional counseling or a social network of peers undergoing TB treatment as a means of improving adherence to TB treatment. Given the significant losses patients and the health system incur as a result of poor TB treatment outcomes, identifying those interventions that are most likely to improve adherence and outcomes, especially in resource-limited settings, is crucial.

2The first pillar of the End TB Strategy of the World Health Organization (WHO)—Integrated, Patient Centered Care and Prevention—calls for “treatment of all people with tuberculosis including drug-resistant tuberculosis; and patient support”. In 2015, WHO commissioned a series of systematic reviews and meta-analyses ahead of a Guideline Development Group meeting tasked with the revision of its TB treatment guidelines in accordance with the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach. Until then, no evidence-based recommendations on TB treatment adherence and delivery existed. In order to inform this process, we conducted a systematic review and meta-analysis of studies of adherence interventions on drug-susceptible and -resistant TB treatment outcomes. Our goal was to identify any adherence interventions associated with improvement in TB treatment outcomes.

3This study has been designed and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA, S1 Text). The full protocol for this study is available in the supplementary material (S2 Text).

4We included all randomized controlled trials (RCTs) as well as prospective and retrospective cohort studies (CSs). The population of interest included all adults or children in any setting undergoing active TB treatment. This included patients with pulmonary, extrapulmonary, smear-positive or -negative, and drug-susceptible and -resistant TB as well as patients with HIV coinfection. To be included, studies must have had an intervention targeted to increase adherence to TB treatment and an internal or historical control group (Table 1). We excluded articles on patients with only latent TB infection. We also excluded studies that compared DOT delivered in a hospital versus clinic setting because of a separate focused systematic review being conducted at the time of this review.

5The following were the primary outcomes of interest as defined by WHO: Cure, a pulmonary TB patient with bacteriologically confirmed TB at the beginning of treatment who was smear or culture negative in the last month of treatment and on at least one previous occasion; Treatment completion, a TB patient who completed treatment without evidence of failure but with no record to show that sputum smear or culture results in the last month of treatment and on at least one previous occasion were negative, either because tests were not done or because results are unavailable; Treatment success, the sum of “cure” and “treatment completion”; Treatment failure, a TB patient whose sputum smear or culture is positive at month 5 or later during treatment; Death, a TB patient who dies for any reason before starting or during the course of treatment; Loss to follow-up, a TB patient who did not start treatment or whose treatment was interrupted for two consecutive months or more; Relapse, a patient with a bacteriologically positive sputum smear or culture after the completion of any of the study TB regimens; Adherence, defined using parent study definitions such as being lost to follow-up (or default), isoniazid (INH) urine test, appointment keeping, etc.; and Development of resistance, identification of new drug resistance in a subsequent isolate otherwise matched with the baseline isolate. In post hoc analyses, given the number of articles that defined adherence as having taken more than 80%–90% of treatment doses, and for ease of comparison between studies, we chose to use this definition for the outcome of adherence when available.

6We searched the literature using Medline with two search strategies, one to include all types of adherence interventions and the second targeted at SMS/VOT through February 3, 2018. The complete search strategy is available in S1 and S2 Tables. We also reviewed references of relevant articles and systematic reviews and contacted experts in the field for unpublished studies. We included all studies in the English language regardless of publication status or date. However, two foreign language articles were included, as data from them were previously abstracted by a different systematic review. Titles and abstracts were reviewed by one investigator (NA) and full manuscript content reviewed by multiple investigators (NA, PN, LJ, CM). Ethics approval was not required for this study, as all information was abstracted from published literature without access to any individual and/or identifiable data.

7Using a standard data abstraction sheet, the following data were recorded from articles that met our inclusion criteria: patient selection, type of TB (pulmonary or extrapulmonary), HIV coinfection, treatment outcomes, method of blinding and randomization, results of drug-susceptibility testing, acid-fast bacilli (AFB) smear and culture results, mode of supervision of therapy, type of adherence intervention, and study results for the outcomes of interest (adjusted or unadjusted risk ratios [RRs], risk differences [RDs], raw data). We chose to use unadjusted RRs preferentially if these data were available. When available, data were gathered from both per protocol and intention-to-treat analyses. The quality of the studies was assessed using the Cochrane Risk of Bias tool for RCTs, and the Newcastle-Ottawa Scale was used for observational studies.

8We grouped the adherence interventions identified across studies into the following categories: (1) DOT referred to the act of observing a patient swallow medications, and self-administered therapy (SAT) was defined as the patient taking each medication dose without supervision; (2) education and counseling interventions were those aimed at providing adequate knowledge and ensuring patient understanding of the disease process and risks and benefits associated with treatment adherence; (3) incentives were interventions to promote treatment adherence through a financial or material reward and enablers were interventions that allowed patients to overcome barriers to treatment adherence (cost, distance, availability); (4) reminders included any intervention made prior to the patient taking medications or attending appointments to serve as a reminder and tracers involved contacts made after a patient had failed adherence in order to improve subsequent adherence to treatment; (5) psychological interventions aimed to provide emotional or psychological support aimed at reducing stigma and increasing treatment adherence; (6) digital health interventions included any of the above categories implemented via mobile electronic devices (SMS, VOT, medication monitors); and lastly, (7) mixed interventions included a combination of the aforementioned interventions to address barriers to adherence based on patient-specific needs and values. Thus, the term “patient-centered DOT” or “enhanced DOT” was used to designate any study using adherence interventions spanning the multiple categories mentioned above in conjunction with DOT. With respect to DOT provider type, a lay provider was defined as an untrained volunteer, including family members or other patient-designated person. A trained health worker included any health worker, community member, or volunteer who had received any form of formal training for DOT. For the categories of patient education and counseling, staff education, incentives and enablers, reminders and tracers, and psychological interventions, studies included in the meta-analysis had to have employed the same supervision modality (either DOT or SAT) for both the intervention and control arms.

9All estimates of effect for dichotomous outcomes were reported as RRs with 95% confidence intervals. When two or more studies were available on a particular outcome, random effects meta-analysis was performed to obtain a pooled estimate of treatment effect and pooled RR between the intervention and control arms. Heterogeneity was assessed visually using forest plots and statistically using the χ2 and I2 tests. If more than 10 studies were available for a particular comparison, we used funnel plots to determine publication bias. All analyses were conducted in RevMan5. All figures were generated using RevMan and GraphPad Prism 7.0. If a study reported no events in the intervention or control group for an outcome of interest, the RD was calculated.

10Title and abstract literature review yielded 7,729 articles, of which 1,092 met the inclusion criteria for full text review (Fig 1). References of 32 systematic reviews found through our online search were also reviewed for relevant articles. A final 129 articles met the inclusion criteria for quantitative analysis. Characteristics of included studies as well as quality assessments are summarized in S3 Table, S1 and S2 Figs. The overall quality of included studies varied significantly amongst CSs. The quality of RCTs was limited predominantly by lack of blinding, given the nature of adherence interventions. Below is a brief summary of findings by each type of adherence intervention (Table 2). When available, data from RCTs and CSs are listed in parentheses. Some of the analyses contained significant heterogeneity, as measured by I2, which has been denoted on their corresponding figures.

11Forty-six studies were included in the meta-analysis, seven of which were RCTs. Six CSs included primarily HIV/TB patients and two CSs involved primarily MDR-TB patients. DOT was offered daily or intermittently at home, clinic, or in the community. DOT providers ranged from family members to trained lay providers and healthcare providers.

12Compared to those undergoing any DOT, participants who received SAT had a lower rate of treatment success (16 CSs: RR 0.81, 95% CI 0.73–0.89; 5 RCTs: RR 0.94, 95% CI 0.89–0.98), cure (18 CSs: RR 0.64, 95% CI 0.54–0.76; 4 RCTs: 0.98, 95% CI 0.83–1.17), and adherence (1 CS: RR 0.84, 95% CI 0.75–0.93; 1 RCT: RR 0.94, 95% CI 0.87–1.02) (Figs 2–5). One RCT found a lower rate of smear conversion at the end of two months amongst SAT patients (RR 0.92, 95% CI 0.87–0.98) (Fig 6). There was no significant difference between DOT and SAT amongst all CSs and RCTs for the outcomes of mortality (23 CSs, 4 RCTs), treatment completion (14 CSs, 5 RCTs), treatment failure (15 CSs, 5 RCTs), loss to follow-up (21 CSs, 4 RCTs), relapse (6 CSs, 1 RCT), or development of drug resistance (3 CSs) (S3–S14 Figs).

13An assessment of the publication bias (S3, S5, S7, S9 and S11 Figs) showed asymmetry for the outcomes of mortality, treatment success, and loss to follow-up. Notably, a few smaller studies found worse outcomes with SAT, as compared with DOT. The plots reflect the significant heterogeneity between the studies and may also be suggestive of publication bias (i.e., reporting bias in CSs).

14In patients with HIV/TB, SAT was associated with lower rates of treatment success (3 CSs: RR 0.41, 95% CI 0.29–0.59), completion (1 CS: RR 0.10, 95% CI 0.01–0.76), and cure (2 CSs: RR 0.40, 95% CI 0.29–0.55) as well as a higher rate of mortality (4 CSs: RR 2.38, 95% CI 1.44–3.92). The difference in the rates of loss to follow-up (3 CSs: RR 1.33, 95% CI 0.72–2.48), treatment failure (2 CSs: RR 1.96, 95% CI 0.61–6.32), and relapse (1 CS: RR 0.90, 95% CI 0.13–6.28) between the two groups (Figs 2, 7 and S16–S20 Figs) was not significant. Although several studies in our meta-analysis included patients with MDR-TB in their cohort, all but two reported aggregate data on treatment outcomes without stratifying by drug-resistant status. Of the 28 MDR-TB patients in one retrospective CS in New York City from 1987–1997, 11 underwent SAT and 17 received DOT. The mortality, treatment completion, and nonadherence rates were not significantly different between the two groups (S21–S24 Figs). A prospective CS in South Africa of rifampicin-resistant patients found no evidence of a difference in rates of mortality, treatment failure, and loss to follow-up between those undergoing DOT for the entire duration of TB treatment and those who only received DOT during the first six months.

15Seven CSs compared treatment outcomes in patients who received DOT by family/lay providers or healthcare workers. Family member DOT was associated with a lower rate of adherence (1 CS: RR 0.86, 95% CI 0.79–0.94) and higher rate of treatment failure (3 CSs: RR 1.79, 95% CI 1.08–2.98) compared with trained health worker DOT (Fig 2). There was no evidence of a difference in rates of treatment success, completion, cure, mortality, and loss to follow-up between the two groups (Figs 2, 8, 9 and S25–S29 Figs). DOT by lay providers showed no significant difference in outcomes when compared with healthcare worker (HCW) DOT.

16Four RCTs and 21 CSs compared DOT offered at different locations, including the home, community, and clinic. In comparison with clinic-based DOT, treatment in the community was associated with higher rates of treatment completion (1 RCT: RR 2.92, 95% CI 1.15–7.41; 3 CSs: RR 0.87, 95% CI 0.87–1.09) and sputum conversion at the end of two months (2 CSs: RR 1.05, 95% CI 1.02–1.08; 1 RCT: RR 1.09, 95% CI 0.99–1.22) as well as lower rates of loss to follow-up (6 CSs: RR 0.63, 95% CI 0.40–0.98; 2 RCTs: RR 1.04, 95% CI 0.34–3.19), treatment failure (6 CSs: RR 0.56, 95% CI 0.33–0.95; 1 RCT: RR 0.68, 95% CI 0.13–3.69), and unfavorable outcome (1 CS: RR 0.63, 95% CI 0.55–0.73) (Figs 10–13 and S30–S35 Figs). No significant differences were noted for the outcomes of mortality or cure. Home-based DOT was associated with a lower rate of treatment adherence when compared with clinic-based DOT in one CS, in which patients were observed by family members or lay providers (1 CS: RR 0.86, 95% CI 0.79–0.94), while another found no significant difference in adherence rates when comparing home and clinic based DOT. Home-based DOT was associated with a higher mortality rate (8 CSs: RR 1.86, 95% CI 1.34–2.59). This effect is nullified with the removal of the study by Mhimbira and colleagues, who defined home-based DOT as DOT delivered at home by a treatment supporter of patients’ choosing. When compared with a particular type of community-based DOT that included offering a free lunch to patients, home-based DOT had a lower success rate in one CS (1 CS: RR 0.94, 95% CI 0.92–0.98; 2 RCTs: RR 1.02, 95% CI 0.94–1.11) but no significant difference in rates of treatment completion, cure, mortality, failure, and loss to follow-up in other studies. Home-based DOT was associated with marginally higher rates of sputum conversion at two months, when compared with clinic-based DOT (3 CSs: RR 1.24, 95% CI 1.01–1.54).

17Four RCTs and one CS evaluated the effect of oral and written educational material as well as counseling on TB treatment outcomes. Education and counseling was associated with a higher rate of treatment completion (1 RCT: RR 1.71, 95% CI 1.32–2.22), cure (1 RCT: RR 2.15, 95% CI 1.58–2.92), and adherence (1 RCT: RR 1.83, 95% CI 1.14–2.92; 1 CS: RR 1.21, 95% CI 1.05–1.40) (Figs 10, 14, 15 and S36–S40 Figs). These interventions had no meaningful impact on rates of mortality, treatment success, failure, or loss to follow-up.

18Incentives and enablers in four RCTs and eleven CSs were associated with lower rates of mortality (3 CSs: RR 0.51, 95% CI 0.37–0.71; 2 RCTs: RR 0.93, 95% CI 0.41–2.09), treatment failure (1 RCT: RR 0.66, 95% CI 0.50–0.87; 2 CSs: RR 0.18, 95% CI 0.02–2.10), and loss to follow-up (1 RCT: RR 0.74, 95% CI 0.60–0.90; 5 CSs: RR 0.48, 95% CI 0.28–0.81) as well as higher rates of treatment success (3 RCTs: RR 1.07, 95% CI 1.03–1.11; 4 CSs: RR 1.25, 95% CI 1.09–1.42), completion (2 RCTs: RR 1.23, 95% CI 1.15–1.31; 4 CSs: RR 1.18, 95% CI 0.97–1.43), cure (4 CSs: 1.13, 95% CI 1.02–1.26; 1 RCT: RR 0.92, 95% CI 0.85–1.01), and sputum conversion at two months (1 RCT: RR 1.21, 95% CI 1.02–1.43). (Figs 16, 17 and S41–S47 Figs).

19Nine RCTs and six CSs found higher rates of treatment success (4 RCTs: RR 1.14, 95% CI 1.02–1.28; 2 CSs: RR 1.12, 95% CI 1.00–1.26), cure (2 CSs: RR 2.16, 95% CI 1.63–2.85; 2 RCTs: RR 1.22, 95% CI 0.78–1.90), adherence (1 RCT: RR 1.65, 95% CI 1.21–2.25), and sputum conversion at two months (3 RCTs: RR 1.29, 95% CI 1.12–1.50) and lower rates of development of drug resistance (1 CS: RR 0.50, 95% CI 0.45–0.55) and loss to follow-up (4 CSs: RR 0.77, 95% CI 0.69–1.13; 4 RCTs: RR 0.69, 95% CI 0.42–1.13) with the use of reminders and tracers (Figs 16, 18–20 and S48–S55 Figs). One study by Bronner and colleagues was removed in sensitivity analyses due to lack of controlling for baseline differences in treatment outcomes prior to intervention in the intervention and control districts (Fig 18).

20Three RCTs and one CS involved interventions such as adherence education for staff, peer training for lay health workers, reminders to initiate adherence discussions, and educational tools and aids for decision-making. These interventions were associated with higher rates of treatment success (1 CS: RR 1.34, 95% CI 1.15–1.55) and lower rates of loss to follow-up (1 CS: RD −0.18, 95% CI −0.26–−0.10) in the one CS. The three RCTs found no significant difference in rates of TB treatment outcomes with the use of such interventions (Figs 21, 22 and S56–S61 Figs).

21One RCT focused on brief counseling interventions for alcohol cessation and another offered self-help groups. One prospective CS evaluated the impact of TB clubs as a support network. Support groups were associated with higher rates of treatment completion (1 CS: RR 1.47, 95% CI 1.08–2.00; 1 RCT: RR 1.20, 95% CI 1.03–1.39) and lower rates of treatment failure (1 RCT: RD −0.12, 95%CI −0.22–−0.01; 1 CS: RD −0.02, 95% CI −0.06–0.03) and loss to follow-up (1 CS: RR 0.31, 95% CI 0.15–0.63; 1 RCT: RR 0.50, 95% CI 0.05–5.31) (Fig 21 and S62–S67 Figs).

22Five RCTs and two CSs looked at daily reminder texts or phone call reminders to take medications for patients undergoing SAT or family DOT. Two CSs evaluated the impact of VOT. In one RCT comparing SMS reminders in patients undergoing SAT versus those undergoing DOT, SAT patients had a higher treatment completion rate (1 RCT: RR 1.11, 95% CI 1.04–1.18). One CS on patients using a combination wireless pill box/SMS reminder system found higher rates of cure (RR 2.32, 95% CI 1.60–3.36) and sputum conversion at two months (RR 1.62, 95% CI 1.09–2.42) compared to those without this system. Other studies found no difference in rates of mortality, treatment success, failure, or loss to follow-up (Figs 23–25 and S68–S74 Figs).

23Electronic medication monitor boxes were associated with lower rates of loss to follow-up (1 CS: RR 0.59, 95% CI 0.43–0.80), poor outcome (1 CS: RR 0.63, 95% CI 0.47–0.83), and poor adherence (1 CS: RR 0.57, 95% CI 0.53–0.61). Phone reminders and medication monitor boxes combined were associated with lower rates of poor adherence (1 CS: RR 0.56, 95% CI 0.52–0.60). Compared with DOT, VOT rates of treatment completion (2 CSs: RR 1.17, 95% CI 0.79–1.72) and mortality (1 CS: RR 1.80, 95% CI 0.19–17) were not significantly different.

24Seven RCTs and eleven CSs combined multiple adherence interventions from the aforementioned categories to DOT or SAT. Patient-centered DOT (enhanced DOT) was associated with lower rates of loss to follow-up (4 RCTs: RR 0.45, 95% CI 0.32–0.65) and higher rates of treatment success (3 CSs: RR 1.27, 95% CI 1.09–1.49; 2 RCTs: RR 1.16, 95% CI 1.11–1.22) and cure (3 RCTs: RR 1.19, 95% CI 1.13–1.25) when compared with DOT alone (Figs 23, 26, 27 and S75–S82 Figs). When compared with SAT, patient-centered DOT was associated with higher rates of treatment success (2 CSs: RR 1.22, 95% CI 1.16–1.27; 1 RCT: RR 1.36, 95% CI 1.06–1.75), treatment completion (2 CSs: RR 1.84, 95% CI 1.52–2.21; 1 RCT: RR 1.41, 95% CI 1.11–1.79), cure (2 CSs: RR 1.42, 95% CI 1.02–1.99; 1 RCT: RR 1.36, 95% CI 1.06–1.75), and sputum conversion at two months (1 RCT: RR 1.65, 95% CI 1.16–2.34). We did not find any studies that compared patient-centered DOT to patient-centered SAT.

25To our knowledge, this is the first comprehensive systematic review of prospective and retrospective CSs as well as RCTs studying any interventions used to increase adherence to treatment of either drug-susceptible or -resistant TB. The studies included represent high- and low-resource settings and therefore are broadly representative across a variety of TB program settings. Our meta-analysis identified several adherence interventions that consistently showed associations with improved adherence and TB treatment outcomes. Of note, DOT was associated with better treatment outcomes compared to SAT, especially for TB patients living with HIV. The addition of other adherence interventions to DOT, such as education (for staff or patients), material or psychological support, or reminder systems (including SMS technology and phone reminders), correlated with reduced rates of mortality and loss to follow-up and higher rates of treatment success and cure. Combining other types of adherence interventions with DOT or SAT was associated with improved outcomes compared to either treatment modality alone. Lastly, VOT had no evidence of a difference in rates of treatment completion and mortality when compared to DOT.

26Our conclusions differ from prior reviews. Karumbi and colleagues focused on data from RCTs and concluded that DOT did not improve TB treatment completion and cure when compared to self-administered therapy. When our meta-analysis was limited to datasets from RCTs, we found similar results on these outcomes. Whereas RCTs are a superior study design because of their strong internal validity, they can lack external validity. For example, RCTs provide rigorous oversight of participants that does not correspond to routine practice in program settings. Furthermore, without allocation concealment, alterations of behavior because of the awareness of being observed can further hamper the ability of RCTs to assess the value of adherence interventions. Another review by Pasipanodya and colleagues included RCTs as well as prospective CSs and concluded that DOT was not significantly better than SAT in preventing microbiologic failure, relapse, or development of drug resistance. Our review found similar results with respect to these particular outcomes of interest. However, because our inclusion criteria were broader than these three outcomes, we included a larger number of CSs and RCTs in our review, resulting in a larger dataset for analysis.

27Important findings in our analysis are that a variety of adherence interventions are effective in improving TB outcomes, certain subpopulations are more likely to benefit from particular types of interventions (i.e., DOT for TB patients with HIV), and certain modalities of DOT are more effective than others. Furthermore, treatment supervision alone in the form of DOT is not likely to guarantee improved TB treatment outcomes in all TB patients across all settings. As such, a patient-centered care approach to TB adherence using a package of interventions tailored to a patient’s needs and values is more likely to improve TB outcomes. It must be noted that most of the literature on adherence interventions included in our meta-analysis did not evaluate the impact of adherence interventions on relapse and development of drug resistance, which are critical outcomes of interest for population-level TB management. Only three CSs compared the impact of DOT on the development of drug resistance compared with SAT, one of which found no significant difference. One CS found reminders correlated with reduced rates of drug resistance, and another CS on incentives found no benefit. With regard to relapse rates, one RCT and four CSs found no significant differences between DOT and SAT, while two CSs saw a decline in relapse rates with DOT. Lastly, relapse rates were similar between patient-centered DOT and SAT in one CS. Given the paucity of data on such critical outcomes, the effectiveness of adherence interventions based on our analyses should be assessed cautiously.

28DOT has several limitations, including the cost imposed on patients and the health system. Furthermore, WHO’s guidance on TB ethics argues that DOT is only ethically justifiable in the context of a patient-centered approach and provides an ethical framework for its implementation. It must be noted that many of the adherence interventions in these studies involve significant resource utilization, which makes it necessary to mobilize the resources needed to facilitate their implementation. Comparing the cost-effectiveness of different adherence interventions would be crucial to strengthen and expand patient-centered approaches for adherence to TB treatment. Data from studies on digital technology and VOT are promising and have the potential to reduce costs to the patient and the health system. The key challenges to address in implementing SMS and VOT adherence interventions are ensuring access to smartphones, coverage of data transmission costs, data encryption, and patient privacy.

29Our meta-analysis has limitations. First, most of the studies on DOT in TB patients living with HIV are from the pre-ART era or were conducted in patients at highest risk of loss to follow-up. Contemporary, more integrated approaches to HIV/TB care were not assessed in this review. Second, the studies in our review were heterogeneous in their methodology. It is difficult to standardize methodology in these types of studies, given the numerous disparate interventions possible, evolving technologies, and the complexities of behavior change. We grouped similar interventions within categories of adherence interventions to draw on their similarities for practice recommendations. As the way in which such interventions are implemented largely determines their success, inter-study variability impacts our meta-analytic findings. Third, our literature review was restricted to English language articles in Medline. The absence of potentially relevant studies published in non-English language journals or via other electronic databases is a limitation. However, we reviewed the references of 32 systematic reviews on this topic with search strategies that spanned European (Embase), Latin American (LILACS), Cochrane Central Register of Controlled Trials (CENTRAL), and metaRegister of Controlled Trials (mRCT) databases. Additionally, we consulted experts in studies of adherence interventions to ascertain that all relevant studies were included. Lastly, the results from meta-analyses of CSs should be interpreted with caution given uncontrolled confounding inherent to such studies, as noted by our quality assessment.

30In this review, we sought to evaluate the impact of adherence interventions in the treatment of drug-resistant TB, and whereas some of the studies included in our review had cohorts of patients with drug resistance, no studies focused only on MDR-TB met our inclusion criteria. A systematic review in 2009 that was limited to case series without an internal or historic control group found that rates of treatment success were higher in patients with MDR-TB who received DOT throughout treatment than those who did not. Another meta-analysis also limited to case series found lower rates of loss to follow-up with DOT delivery at home, use of community health workers for DOT (as opposed to nurses or healthcare providers), a standardized treatment regimen, provision of DOT throughout treatment, and patient education. Given the increased length and duration of treatment necessary for patients with MDR-TB, more studies on adherence interventions tailored to this population are needed.

31We have found that TB treatment outcomes improve with the use of adherence interventions such as patient education and counseling, material support, psychological support interventions, reminders and tracers, and digital health technologies. DOT provided by trained health workers in the community is associated with better treatment outcomes than DOT provided by family members or untrained lay workers. DOT provided in the community is associated with better treatment outcomes than clinic-based DOT. TB patients living with HIV have significantly better outcomes when treated with DOT as opposed to SAT. VOT may be an appropriate alternative to in-person DOT if the resources for its use are available. More importantly, a patient-centered approach to TB treatment using a package of adherence interventions tailored to patients’ needs and values leads to improved TB treatment outcomes. The optimal package of adherence interventions to implement may vary by setting, resources, and the local epidemiology of TB (e.g., prevalence of comorbidities, including HIV coinfection), among other factors. The WHO TB treatment guidelines update, for which this review was conducted, provides additional information on selecting patient-centered approaches for enhancing adherence. Based on our review, studies on adherence interventions in patients with MDR-TB as well as cost-effectiveness analyses will be helpful in identifying the optimal interventions to implement in various settings.