Directly observed therapy for treating tuberculosis.

BACKGROUND: Tuberculosis (TB) requires at least six months of treatment. If treatment is incomplete, patients may not be cured and drug resistance may develop. Directly Observed Therapy (DOT) is a specific strategy, endorsed by the World Health Organization, to improve adherence by requiring health workers, community volunteers or family members to observe and record patients taking each dose.
OBJECTIVES: To evaluate DOT compared to self-administered therapy in people on treatment for active TB or on prophylaxis to prevent active disease. We also compared the effects of different forms of DOT. SEARCH
METHODS: We searched the following databases up to 13 January 2015: the Cochrane Infectious Diseases Group Specialized Register; the Cochrane Central Register of Controlled Trials (CENTRAL), published in the Cochrane Library; MEDLINE; EMBASE; LILACS and mRCT. We also checked article reference lists and contacted relevant researchers and organizations. SELECTION CRITERIA: Randomized controlled trials (RCTs) and quasi-RCTs comparing DOT with routine self-administration of treatment or prophylaxis at home. DATA COLLECTION AND ANALYSIS: Two review authors independently assessed risk of bias of each included trial and extracted data. We compared interventions using risk ratios (RR) with 95% confidence intervals (CI). We used a random-effects model if meta-analysis was appropriate but heterogeneity present (I(2) statistic > 50%). We assessed the quality of the evidence using the GRADE approach. MAIN
RESULTS: Eleven trials including 5662 participants met the inclusion criteria. DOT was performed by a range of people (nurses, community health workers, family members or former TB patients) in a variety of settings (clinic, the patient's home or the home of a community volunteer). DOT versus self-administered Six trials from South Africa, Thailand, Taiwan, Pakistan and Australia compared DOT with self-administered therapy for treatment. Trials included DOT at home by family members, community health workers (who were usually supervised); DOT at home by health staff; and DOT at health facilities. TB cure was low with self-administration across all studies (range 41% to 67%), and direct observation did not substantially improve this (RR 1.08, 95% CI 0.91 to 1.27; five trials, 1645 participants, moderate quality evidence). In a subgroup analysis stratified by the frequency of contact between health services in the self-treatment arm, daily DOT may improve TB cure when compared to self-administered treatment where patients in the self-administered group only visited the clinic every month (RR 1.15, 95% CI 1.06 to 1.25; two trials, 900 participants); but with contact in the control becoming more frequent, this small effect was not apparent (every two weeks: RR 0.96, 95% CI 0.83 to 1.12; one trial, 497 participants; every week: RR 0.90, 95% CI 0.68 to 1.21; two trials, 248 participants).Treatment completion showed a similar pattern, ranging from 59% to 78% in the self-treatment groups, and direct observation did not improve this (RR 1.07, 95% CI 0.96 to 1.19; six trials, 1839 participants, moderate quality evidence). DOT at home versus DOT at health facility In four trials that compared DOT at home by family members, or community health workers, with DOT by health workers at a health facility there was little or no difference in cure or treatment completion (cure: RR 1.02, 95% CI 0.88 to 1.18, four trials, 1556 participants, moderate quality evidence; treatment completion: RR 1.04, 95% CI 0.91 to 1.17, three trials, 1029 participants, moderate quality evidence). DOT by family member versus DOT by community health workerTwo trials compared DOT at home by family members with DOT at home by community health workers. There was also little or no difference in cure or treatment completion (cure: RR 1.02, 95% CI 0.86 to 1.21; two trials, 1493 participants, moderate quality evidence; completion: RR 1.05, 95% CI 0.90 to 1.22; two trials, 1493 participants, low quality evidence). Specific patient categoriesA trial of 300 intravenous drug users in the USA evaluated direct observation with no observation in TB prophylaxis to prevent active disease and showed little difference in treatment completion (RR 1.00, 95% CI 0.88 to 1.13; one trial, 300 participants, low quality evidence). AUTHORS'
CONCLUSIONS: From the existing trials, DOT did not provide a solution to poor adherence in TB treatment. Given the large resource and cost implications of DOT, policy makers might want to reconsider strategies that depend on direct observation. Other options might take into account financial and logistical barriers to care; approaches that motivate patients and staff; and defaulter follow-up.

Background

Description of the condition

Tuberculosis (TB) remains a leading cause of death in low‐ and middle‐income countries despite the availability of effective treatments. In 2012, the World Health Organization (WHO) estimated that there were 8.6 million people infected with TB, of whom 1.3 million died (WHO 2013). Most people infected with Mycobacterium tuberculosis develop 'latent TB', where the bacteria are contained by the person's immune system and they do not develop symptoms. The risk of progression to active TB and development of symptoms is about 10% over the course of a lifetime (Frieden 2003), but co‐infection with human immune deficiency virus (HIV) increases this risk to about 10% per year (Sepkowitz 1995).

The WHO currently recommends at least six months of treatment for active disease, and 12 months for latent TB (Smieja 2010; WHO 2010). These long durations of treatment can be difficult for patients to adhere to, especially once they are well and need to return to work. Poor adherence can lead to relapse and even death in individuals, and also has important public health consequences, such as increased transmission and the development of drug resistance (Hirpa 2013; Moonan 2011).

Munro 2007 synthesized evidence from qualitative studies among patients and health workers and identified eight factors that influence adherence:

Organization of treatment and care for TB patients.

Interpretation of illness and wellness by the patient.

Financial cost.

Patient knowledge, attitudes and beliefs about treatment.

Law and immigration status.

Gender and substance abuse.

Drug side effects.

Influence of the family, community and peers.

We adapted an existing conceptual framework by van den Boogaard 2012 to develop a model for understanding approaches to improving adherence (Figure 1).There are health system level barriers (staff, inconvenient location, expensive and a poorly organized healthcare system) and personal level barriers (stigmatisation, poverty, competing demands and health beliefs) to adherence and the patient has to work through these barriers in order to be adherent. Healthcare workers have devised several strategies targeted at some of the key barriers to improve adherence, some of which are addressed by Cochrane Reviews:

1

Factors influencing adherence and possible intervention points.

Reminder systems and late patient tracers in the diagnosis and management of TB (Liu 2008).

Patient education and counselling for promoting adherence to treatment for TB (M'Imunya 2012).

Material incentives and enablers in the management of TB (Lutge 2012).

Contracts: written or verbal agreements to return for an appointment or course of treatment (Bosch‐Capblanch 2007).

Description of the intervention

'Directly observed therapy' (DOT) is one component of a wider WHO strategy called 'Directly Observed Therapy Short course' (DOTS). This strategy incorporates wide ranging health system improvements, political commitment to improving TB programmes, improved TB laboratory services, free TB drugs for all TB patients, and accurate documentation and monitoring of TB diagnosis and treatment outcomes (WHO 2002). The DOT component is an attempt to improve adherence by active monitoring and recording of the consumption of each and every drug dose by an 'observer' acceptable to the patient and the health system (Hopewell 2006). This approach was first adopted in studies in Madras, India and Hong Kong as early as the 1960s (Bayer 1995), and is now considered a core component of TB programmes by the WHO to ensure cure and prevent the emergence of drug resistance (Chien 2013; Hirpa 2013). Proponents of DOT argue that the close monitoring has a social effect and acts as a peer pressure which leads to behavior change towards improved adherence (Macq 2003) and it has strong proponents (Chaulk 1998; Frieden 2007). However, to opponents it has been seen as a coercive model which leaves the patient as a passive recipient of therapy thereby eroding the gains made in involving patients in management of their own health (Zwarenstein 1998 ZAF).

The initial Cochrane Review (Volmink 1997) and subsequent updates (Volmink 2000a; Volmink 2001; Volmink 2003; Volmink 2006; Volmink 2007) challenged the dogma that DOT improved cure and thus helped prevent drug resistance developing. The debate has continued, with some even advocating for a shift of resources away from DOT programmes (Barbara 2013; Gross 2009; Moonan 2011; Pasipanodya 2013). There are also debates as to the best delivery of DOT, for example, should it be through healthcare workers or family members (Anuwatnonthakate 2008; Dick 2005).

Why it is important to do this review

Full implementation of DOT requires considerable resources. For example, in Pakistan it has been shown that direct observation at a health facility costs two times more than self‐supervision (USD310 versus USD164). Therefore, it is important to evaluate the effects in order to inform decisions about whether the benefits are worth investing in (Khan 2003). This Cochrane Review is an update of Volmink 2007.

Objectives

To evaluate DOT compared to self‐administered therapy in people on treatment for active TB or on prophylaxis to prevent active disease. We also compare the effects of different forms of DOT.

Methods

Criteria for considering studies for this review

Types of studies

Individually randomized controlled trials (RCTs) or cluster‐RCTs. We also included quasi‐RCTs.

Types of participants

People on treatment for active TB or receiving prophylaxis to prevent the development of active TB disease.

Types of interventions

Intervention

DOT where a health worker, community volunteer or family member, routinely observes participants taking their antituberculous drugs.

Control

Self‐administered therapy or an alternative form of DOT.

Types of outcome measures

Primary

Cure (having a negative sputum smear test in the last month of treatment having been smear‐positive initially).

Treatment completion.

Development of clinical TB (in trials of drug prophylaxis).

Secondary

Proportion of outpatient appointments attended.

Search methods for identification of studies

We attempted to identify all relevant trials regardless of language or publication status (published, unpublished, in press and in progress).

Databases

We searched the following databases using the search terms and strategy described in Appendix 1: Cochrane Infectious Diseases Group (CIDG) Specialized Register (13 January 2015); the Cochrane Central Register of Controlled Trials (CENTRAL), published in the Cochrane Library; MEDLINE (1966 to 13 January 2015); EMBASE (1974 to 13 January 2015); and LILACS (1982 to 13 January 2015). We also searched the metaRegister of Controlled Trials (mRCT) using 'tuberculosis AND DOT*' (13 January 2015).

Researchers and organizations

For unpublished and ongoing trials, we contacted individual researchers working in the field and the following organizations: WHO, the International Union Against Tuberculosis and Lung Disease, and the Centers for Disease Control and Prevention (CDC).

Reference lists

We also checked the reference lists of all studies identified by the above methods.

Data collection and analysis

Selection of studies

We independently applied the inclusion criteria to all identified trials. We used the titles and abstracts of the identified citations to exclude trials that clearly did not meet the inclusion criteria. If either review author judged that the trial might be eligible for inclusion, we obtained the full text article. We independently screened the full text articles of selected trials to confirm eligibility and resolved any disagreements by discussion.

Data extraction and management

We independently extracted the data and checked whether trial authors had conducted an intention‐to‐treat analysis. We contacted trial authors to obtain missing information and to clarify issues. We resolved discrepancies through discussion. For the outcomes, we extracted the number of participants experiencing the event.

Assessment of risk of bias in included studies

We independently evaluated the methodological quality of each trial, classifying the generation of allocation sequence and concealment of allocation as either adequate, inadequate or unclear, according to Jüni 2001. We classified blinding as adequate if the trial authors took steps to ensure the people recording the main outcome of the trial were blinded to the assigned interventions, and inadequate if this was not the case or if there was no mention of attempts to blind the observers. We assessed completeness of follow‐up as adequate if 90% or more of the enrolled participants had outcome data reported, inadequate if less than 90% of the participants had outcome data reported, or unclear if not mentioned in the trial.

Measures of treatment effect

For dichotomous outcomes, we calculated the results using the risk ratio (RR). We presented the effect estimates with 95% confidence intervals (CIs).

Assessment of heterogeneity

We looked for statistical heterogeneity by inspecting the forest plots for overlapping CIs, applying the Chi2 test (P value < 0.10 considered statistically significant) and the I2 statistic (I2 value of 50% used to denote moderate levels of heterogeneity). We assessed whether a difference in the intensity of supervision between the intervention and control group could explain heterogeneity.

Data synthesis

We used Review Manager 5 to analyse the data, using risk ratio (RR) with 95% CIs to assess estimates of effect. We used the fixed‐effect model when there was no statistically significant heterogeneity (Chi2 test, P > 0.1) and a random‐effects model when heterogeneity was present (I2 statistic > 50). We assessed the quality of the evidence using the GRADE approach.

Results

Description of studies

Eleven trials, enrolling 5662 participants, met the inclusion criteria (see 'Characteristics of included studies'), and we excluded 13 studies for the reasons listed in the 'Characteristics of excluded studies' table.

Nine included trials were individually RCTs, and two were cluster‐RCTs (Lwilla 2003 TZA; Newell 2006 NPL). One trial used a quasi‐random method of allocation (MacIntyre 2003 AUS).

Three trials were conducted in low‐income countries (Tanzania: Lwilla 2003 TZA; Wandwalo 2004 TZA; Nepal: Newell 2006 NPL); six in middle‐income countries (Taiwan: Hsieh 2008 TWN; Pakistan: Walley 2001 PAK; Thailand: Kamolratanakul 1999 THA; South Africa: Zwarenstein 1998 ZAF; Zwarenstein 2000 ZAF; and Swaziland: Wright 2004 SWZ); and two were from high‐income countries (Australia: MacIntyre 2003 AUS; USA: Chaisson 2001 USA).

Populations targeted

Ten trials evaluated DOT in people on treatment for active TB, and one evaluated directly observed prophylaxis in intravenous drug users (see Table 9; Table 10; Table 11).

1

Summary of interventions in trials of DOT versus self‐administered

Trial ID	DOT						Self administered therapy
	Who observed?	Where?	How often?		Adherence recorded at each contact	Cure	Frequency of contact with health service	Adherence recorded at each contact	Cure
			Intensive phase	Consolidation phase
Zwarenstein 1998 ZAF	Nurses	Clinic	5 times per week	3 times per week	Yes	38%(42/111)	Weekly	Yes	51%(31/61
Zwarenstein 2000 ZAF	Nurse	Clinic	5 times per week	3 times per week	Yes	57%(31/54)	Weekly	Yes	41%(9/22)
	Lay health worker	Lay health workers home
Kamolratanakul 1999 THA1	Healthcare worker	Clinic	Daily	Daily	Yes	76% (315/414)	Monthly	Unclear	67%(283/422)
	Community health worker	Home	Daily	Daily
	Family member	Home	Daily	Daily
Walley 2001 PAK	Healthcare worker	Clinic	6 times per week	2 times per month	Yes	59%(199/335)	Every two weeks	Unclear	62%(100/162)
	Community health worker	Home
	Family member	Home	Daily	Daily
MacIntyre 2003 AUS2	Family member	Home	Daily	Daily	Yes	Not reported	Monthly	Yes	Not reported
Hsieh 2008 TWN3	Case manager orHospital care	Hospital	Daily	Once per week	Yes	94%(30/32)	Monthly unscheduled visit	Yes	69%(22/32)

1In Kamolratanakul 1999 THA patients could choose which observer they preferred and there a more intense supervision of observers in the intensive phase.
 2In MacIntyre 2003 AUS nurses made weekly calls to the patients who were observed by a family member.
 3In Hsieh 2008 TWN the case manager directly supervised medicine intake for first two months (Intensive phase), then self‐administration with weekly unscheduled visit.

2

Interventions comparing home versus clinic direct observation

Trial ID	DOT at patient's home					DOT at clinic
	Who observed?	How often?		Supervision of observer	Cure	Who observed?	How often?		Cure
		Intensive phase	Consolidation phase				Intensive phase	Consolidation phase
Walley 2001 PAK1	Family member	Daily	Not described	Observers collected drugs from the clinic every 2 weeks	55%(91/165)	Health worker	6 times per week	Self‐supervised	64%(108/170)
Wandwalo 2004 TZA1	Family member or former TB patient	Daily	Self‐supervised	Observers collected drugs from clinic weekly and spot checks were conducted by health worker	43%(111/260)	Health worker	Daily	Self‐supervised	43%(141/327)
Zwarenstein 2000 ZAF	Lay health worker2	'Several times a week'	Not described	Observer collected drugs monthly	57%(31/54)	Health worker	5 times a week	3 times a week	41%(24/58)
Lwilla 2003 TZA1	Community volunteer	Daily	Self‐supervised	Observer was visited every two weeks by the health worker and every month by the district co‐ordinator3	53%(117/221)	Health worker	Daily	Self‐supervised	49%(148/301)

1In Lwilla 2003 TZA, Walley 2001 PAK and Wandwalo 2004 TZA observation was during the intensive phase, while in the clinic observation arm of Zwarenstein 2000 ZAF it continued in the consolidated phase.
 2In Zwarenstein 2000 ZAF the observation took place in the lay health worker's home, not the patient's home.
 3In Lwilla 2003 TZA there was additional supervision by the district coordinator.

3

Interventions comparing family‐administered DOT versus community health worker DOT

Trial ID	Who observed?	Where?	How often?		Additional intervention	Who observed?	Where?	How often?
			Intensive phase	Consolidation phase				Intensive phase	Consolidation phase
Newell 2006 NPL	Family member	Patient's home	Daily	Daily	Drugs supplied to supervisor every week	Community health worker	Patient's home1	Daily	Daily
Wright 2004 SWZ	Family member	Patient's home	Daily	Daily	Patient reviewed at the diagnostic centre once per monthRecorded in a patient adherence card	Community health worker	Community health worker's home	Daily	Daily

1In Newell 2006 NPL the community health worker mainly visited the patients at their homes but occasionally the patients came to the health worker's home.

Six trials compared DOT with self‐administered therapy:

Zwarenstein 1998 ZAF included two arms in two locations (Elsies River and Khayelitsha); Zwarenstein 2000 ZAF was the same trial containing data from one of these two locations, and had an additional arms (lay health worker administered DOT). The control arms (self‐administered treatment) were therefore the same for Elsies River in both trials, and so in the meta‐analysis we adjusted the data to ensure they were not counted twice.

Kamolratanakul 1999 THA allowed participants to choose between DOT by a health worker, community leader or family member; 85% chose the latter.

Walley 2001 PAK compared DOT by a health worker or community health worker with DOT by a family member and with self‐administration of treatment.

MacIntyre 2003 AUS evaluated DOT by a family member compared to self‐administration.

Hsieh 2008 TWN had DOT by case managers; there were three arms; weekly observation, monthly observation and the control group was patients admitted to hospital (Inpatient care).

Four trials compared different forms of DOT:

Newell 2006 NPL compared community health worker observation to family member observation.

Wandwalo 2004 TZA trials compared DOT by a family member with either DOT by a health worker at a health facility or DOT by a community health worker.

Wright 2004 SWZ compared community health worker observation to family member observation coupled with a once per week visit by a community health worker.

Lwilla 2003 TZA compared a community health worker DOT at home to DOT at a health facility.

One trial evaluated DOT in injecting (intravenous) drug users in the USA:

Chaisson 2001 USA involved intravenous drug users, and studied DOT by an outreach nurse with self‐administration either with monthly peer support or monthly clinic visits.

Intensity of supervision

Intensity of supervision varied in the included trials. For the six RCTs of DOT compared to self‐administered treatment, three trials appeared to be a direct comparison of healthcare worker administered DOT versus self‐administered. Another three trials appeared to have more intense supervision in the DOT arm only, with health workers visiting patients at home every two weeks. In Kamolratanakul 1999 THA, community health workers and family members received additional supervision by health centre staff once every two weeks. In MacIntyre 2003 AUS, nurses had weekly calls to the patients who were observed by family members. In Hsieh 2008 TWN the case manager visited the patients in the intervention arm and was supervised by weekly unscheduled supervision. In the control group of these trials no such intensive supervision was described.

Adjustment for clustering

Both cluster‐RCTs adjusted for clustering appropriately: standard error of the coefficients for clustering on units corrected using the Huber‐White‐Sandwich method (Lwilla 2003 TZA); and, in Newell 2006 NPL, using the coefficient of variation between clusters.

Risk of bias in included studies

We have summarized the 'Risk of bias' assessments Figure 2 and Figure 3, and have listed the reasons in the Characteristics of included studies section.

2

Risk of bias summary: review authors' judgements about each risk of bias item for each included trial.

3

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included trials.

Allocation

Seven trials used adequate methods to generate a random sequence: computer‐generated random sequences (Chaisson 2001 USA; Walley 2001 PAK; Zwarenstein 1998 ZAF; Zwarenstein 2000 ZAF), a random‐number table (Kamolratanakul 1999 THA), coin tossing (Wandwalo 2004 TZA) or drawing of lots from a basket (Newell 2006 NPL). One trial used alternate allocation, which was an inadequate randomization method (MacIntyre 2003 AUS). The remaining trials reports did not provide information (Hsieh 2008 TWN; Lwilla 2003 TZA; Wright 2004 SWZ).

Four trials employed adequate methods for concealing allocation (Newell 2006 NPL; Walley 2001 PAK; Zwarenstein 1998 ZAF; Zwarenstein 2000 ZAF). Five trials had unclear allocation concealment (Chaisson 2001 USA; Hsieh 2008 TWN; Kamolratanakul 1999 THA; Lwilla 2003 TZA; Wright 2004 SWZ) and the remaining two trials did not use allocation concealment (MacIntyre 2003 AUS; Wandwalo 2004 TZA).

Blinding

Only four trials blinded outcome assessment (MacIntyre 2003 AUS; Newell 2006 NPL; Walley 2001 PAK; Wright 2004 SWZ). It was not used in three trials (Chaisson 2001 USA; Kamolratanakul 1999 THA; Lwilla 2003 TZA) and unclear in the remaining trials (Hsieh 2008 TWN; Wandwalo 2004 TZA; Zwarenstein 1998 ZAF; Zwarenstein 2000 ZAF).

Incomplete outcome data

Two trials excluded more than 10% of participants from the analyses (Lwilla 2003 TZA; Zwarenstein 1998 ZAF). A further three trials did not provide sufficient information to assess this aspect of trial quality (MacIntyre 2003 AUS; Newell 2006 NPL; Zwarenstein 2000 ZAF). The remaining trials had adequate follow‐up.

Selective reporting

We found no evidence of selective reporting.

Other potential sources of bias

Hsieh 2008 TWN had a control group which was inpatient based. We have not included this group in our analyses. Two trials had the same control groups (Zwarenstein 1998 ZAF; Zwarenstein 2000 ZAF; see Table 9).

Lwilla 2003 TZA had one cluster in the community observed arm lost to follow‐up and we therefore did not include it in the final analysis. Two cluster‐RCTs had cluster adjustment.

Effects of interventions

See: Table 1; Table 2; Table 3; Table 4

Summary of findings for the main comparison

Directly observed therapy (DOT) versus self‐administered TB treatment

Directly observed therapy (DOT) versus self‐administered TB treatment
Patient or population: Patients on TB treatment Settings: Low‐, middle‐ or high‐income countries Intervention: DOT Comparison: Self‐administered therapy
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (trials)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	Self‐administered therapy	DOT
Cure Follow‐up: up to 6 months	617 per 1000	666 per 1000 (561 to 784)	RR 1.08 (0.91 to 1.27)	1645 (5 trials)	⊕⊕⊕⊝ moderate1,2,3,4
Treatment completionFollow‐up: 2 to 8 months5	709 per 1000	751 per 1000 (680 to 829)	RR 1.07 (0.96 to 1.19)	1839 (6 trials)	⊕⊕⊕⊝ moderate1,2,3,4
The basis for the assumed risk is the median control group risk across studies. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; DOT: directly observed therapy; TB: tuberculosis.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

1No serious risk of bias: three trials adequately described allocation concealment. Exclusion of trials at unclear or high risk of bias did not substantially change the result.
 2Downgraded by 1 for inconsistency: trials include qualitative differences in effect size and direction. The benefit reached standard levels of statistical significance in the two trials where those receiving self‐administered therapy had less frequent contact with health services compared to the directly observed group, so any effect probably due to confounding.
 3No serious indirectness: The trials were conducted in low‐, middle‐ and high‐income countries between 1995 and 2008.
 4No serious imprecision: The analysis is adequately powered to detect clinically important differences between treatment arms.
 5Some trials checked for completion of intensive phase treatment and others the completion of the whole therapy, hence the 2 to 8 months.

Summary of findings 2

Home DOT versus clinic DOT

Home DOT versus clinic DOT
Patient or population: Patients with TB treatment Settings: Low‐, middle‐ or high‐income countries Intervention: Home observation Comparison: Clinic observation
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (trials)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	Clinic observation	Home observation
Cure Follow‐up: up to 6 months	492 per 1000	502 per 1000 (433 to 580)	RR 1.02 (0.88 to 1.18)	1556 (4 trials)	⊕⊕⊕⊝ moderate1,2,3
Treatment completion4 Follow‐up: 2 to 6 months	751 per 1000	781 per 1000 (684 to 879)	RR 1.04 (0.91 to 1.17)	1029 (3 trials)	⊕⊕⊕⊝ moderate1,2,3
The basis for the assumed risk is the median control group risk across studies. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; DOT: directly observed therapy; TB: tuberculosis.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

1Downgraded by 1 for risk of bias: selection bias is probable in one trial, Wandwalo 2004 TZA, as there was no blinding and no allocation concealment. In Lwilla 2003 TZA, sequence generation and allocation concealment were unclear and there was no blinding. This could bias the measurement of treatment completion.
 2No serious indirectness: The trials were conducted in low‐, middle‐ and high‐income countries between 1995 and 2008.
 3No serious imprecision: The analysis is adequately powered to detect clinically important differences between treatment arms.
 4Some trials checked for completion of intensive phase treatment and others the completion of the whole therapy, hence the 2 to 6 months.

Summary of findings 3

Summary of findings table 3

Community DOT versus family DOT
Patient or population: Patients on TB treatment Settings: Low‐, middle‐ or high‐income countries Intervention: Community DOT Comparison: Family DOT
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (trials)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	Family DOT	Community DOT
CureFollow‐up: up to 6 months	766 per 1000	781 per 1000 (659 to 927)	RR 1.02 (0.86 to 1.21)	1493(2 trials)	⊕⊕⊕⊝ moderate1
Treatment completionFollow‐up: 2 to 6 months	827 per 1000	869 per 1000 (744 to 1000)	RR 1.05 (0.90 to 1.22)	1493(2 trials)	⊕⊕⊝⊝ low1,2
*The basis for the assumed risk (eg the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; DOT: directly observed therapy; TB: tuberculosis.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

1Downgraded by 1 for risk of bias. Both trials had unclear random sequence generation and recruitment bias could not be ruled out for Newell 2006 NPL.
 2Downgraded by 1 for risk of bias for the outcome of treatment completion as there was no allocation concealment and selective reporting could not be ruled out in Wright 2004 SWZ.

Summary of findings 4

DOT versus self‐administered therapy for intravenous drug users

DOT versus self‐administered therapy for intravenous drug users
Patient or population: Patients on TB treatment Settings: Low‐, middle‐ or high‐income countries Intervention: DOT Comparison: Self‐administered treatment
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of participants (trials)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	Self‐administered therapy	DOT
Treatment completionFollow‐up for 6 months	79 per 100	79 per 1000 (70 to 89)	RR 1.00 (0.88 to 1.13)	300 (1 trial)	⊕⊕⊝⊝ low1,2,3
*The basis for the assumed risk (eg the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio; DOT: directly observed therapy; TB: tuberculosis.
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

1Downgraded by 1 for risk of bias. There was no blinding of outcome assessment and allocation concealment was unclear and treatment completion can be a bit subjective hence the results might be biased. The level of completeness to follow‐up was 88%.
 2Downgraded by 1 for indirectness. The self‐administered group had a 10 dollar stipend which is may have enhanced adherence in this group.
 3There may have been some imprecision. The study was had a small sample size and may have been underpowered to detect clinically important differences.

1. DOT versus self‐administered therapy

The details of the interventions are described in Table 9; and see Table 1.

Six trials compared DOT and self‐administered therapy. The observers were described as either nurses (Zwarenstein 1998 ZAF; Zwarenstein 2000 ZAF), healthcare workers (Kamolratanakul 1999 THA; Walley 2001 PAK), community health workers (Kamolratanakul 1999 THA; Walley 2001 PAK), lay health workers (Zwarenstein 2000 ZAF), case managers (Hsieh 2008 TWN) or family members (Kamolratanakul 1999 THA; MacIntyre 2003 AUS; Walley 2001 PAK). In one trial participants were allowed to choose either a healthcare worker, a community health worker or a family member (Kamolratanakul 1999 THA).

Overall TB cure was low with self‐administered therapy, ranging from 41% to 69% across trials, but on average this did not substantially improve with DOT (RR 1.08, 95% CI 0.91 to 1.27; five trials, 1645 participants, moderate quality evidence; Analysis 1.1). However, there was moderate statistical heterogeneity between trials (I2 statistic = 68%; P = 0.01), with two trials finding benefits which reached standard levels of statistical significance (Kamolratanakul 1999 THA; Hsieh 2008 TWN). These differential effects may be explained by differences in the intensity of follow‐up between the intervention and control arms (Analysis 1.2). However it is important to note that these two trials were also at unclear or high risk of selection bias and detection bias.

1.1

Analysis

Comparison 1 Directly observed versus self‐administered, Outcome 1 Cure (negative sputum smear in last month of Rx in patients +ve initially).

1.2

Analysis

Comparison 1 Directly observed versus self‐administered, Outcome 2 Cure (by intensity of monitoring in control group).

Kamolratanakul 1999 THA is the largest trial of DOT to date and found a higher TB cure with the intervention (76% versus 67%; RR 1.13, 95% CI 1.04 to 1.24; one trial, 836 participants). This trial had the least supervised control group (patients picked up their medication monthly), and one of the most intensely supervised intervention groups (doses were directly observed daily by a choice of health worker, community health worker or family member, and a health worker visited the patient at home every two weeks to check on adherence). This difference in intensity was similar in the second trial showing a difference (94% versus 69%; RR 1.36, 95% CI 1.06 to 1.75; one trial, 64 participants), but this trial was small and underpowered to have full confidence in this effect (Hsieh 2008 TWN).

Similarly, TB treatment completion ranged from 59% to 78% in those allocated to self‐administration, and on average did not substantially improve with DOT (RR 1.07, 95% CI 0.96 to 1.19; six trials, 1839 participants, moderate quality evidence; Analysis 1.3). There was again moderate heterogeneity between trials (I2 = 57%; P = 0.04), with the same two trials finding statistically significant benefits (Hsieh 2008 TWN; Kamolratanakul 1999 THA).

1.3

Analysis

Comparison 1 Directly observed versus self‐administered, Outcome 3 Treatment completion (both with smear sputum test at end and those without).

Different levels of monitoring in the self‐administered groups did not yield substantially different levels of completion. (Self‐treatment group: monthly monitoring RR 1.12, 95% 0.95 to 1.31; three trials, 1073 participants; every two weeks: RR 0.99, 95% 0.87 to 1.14; one trial, 497 participants); weekly monitoring RR 1.04, 95% 0.74 to 1.46; two trials, 269 participants; Analysis 1.4).

1.4

Analysis

Comparison 1 Directly observed versus self‐administered, Outcome 4 Treatment completion (grouped by frequency of monitoring in the self‐administered therapy group).

2. Home observation versus clinic observation

The trials are described in Table 10; and see Table 2.

Four trials compared home with clinic observation. Two tested family member direct observation against direct observation at clinic (Walley 2001 PAK; Wandwalo 2004 TZA) while the other two tested community health worker home visits to direct observation at clinic (Lwilla 2003 TZA; Zwarenstein 2000 ZAF).

TB cure was generally low for both the home observation groups (ranging from 43% to 57%) and for clinic observation (ranging from 41% to 64%). On average there was little or no difference between the two strategies (RR 1.02, 95% CI 0.88 to 1.18; four trials, 1556 participants, moderate quality evidence; Analysis 2.1).

2.1

Analysis

Comparison 2 Home observed versus clinic observed, Outcome 1 Cure (having a negative sputum smear test in the last month of treatment having been smear‐positive initially).

Treatment completion ranged from 62% to 85% in those being observed at home and between 57% and 83% in clinic observation. On average there was little or no difference between the two locations (RR 1.04, 95% CI 0.91 to 1.17; three trials, 1034 participants, moderate quality evidence; Analysis 2.2).

2.2

Analysis

Comparison 2 Home observed versus clinic observed, Outcome 2 Treatment completion (both with smear sputum test at end and those without).

One trial, Lwilla 2003 TZA, had more intense supervision of the observer than the other three trials. This intense supervision however did not improve cure rates (53% for home observation and 49% for clinic observation; RR 1.01, 95% CI 0.91 to 1.11; four trials, 1556 participants, Analysis 2.3). This trial however did not report on completion of treatment.

2.3

Analysis

Comparison 2 Home observed versus clinic observed, Outcome 3 Cure (stratified by intensity of observation).

Wandwalo 2004 TZA had high completion rates (85% in home and 83% in clinic observation arm) but the cure rates were quite low in either arm (43% in both arms).

3. Community observed versus family observed

Two trials compared community health worker based observation with family based observation (Newell 2006 NPL; Wright 2004 SWZ). The trials are described in Table 11.

The Nepal trial, Newell 2006 NPL, had higher cure rates across the two arms (85% for community and 89% for family observed) compared with Wright 2004 SWZ (68% for community and 61% for family observed). There was little or no difference between community and family observation (RR 1.02, 95% CI 0.86 to 1.21; two trials, 1493 participants, moderate quality evidence; Analysis 3.1). However there was high statistical heterogeneity (I2 = 86%; P = 0.009).

3.1

Analysis

Comparison 3 Community observed vs family observed, Outcome 1 Cure (having a negative sputum smear test in the last month of treatment having been smear‐positive initially).

Similarly, for the completion of treatment outcome Newell 2006 NPL had higher rates across the two arms (96% in both arms) compared with Wright 2004 SWZ (74% for community and 67% for family observed). There was little or no difference between community and family observation (RR 1.05, 95% CI 0.90 to 1.22; two trials, 1493 participants, low quality evidence; Analysis 3.2). Again there was high statistical heterogeneity (I2 = 87%; P = 0.005).

3.2

Analysis

Comparison 3 Community observed vs family observed, Outcome 2 Treatment completion (both with smear sputum test at end and those without).

4. DOT versus self‐administered therapy for intravenous drug users

One trial, Chaisson 2001 USA, had three arms, supervision at a clinic, peer group supervision and self‐administered treatment. The level of treatment completion was similar in those self‐administering (79%) and those under peer or clinic supervision (79%) (RR 1.00, 95% CI 0.88 to 1.13; one trial, 300 participants, low quality evidence; Analysis 4.1).

4.1

Analysis

Comparison 4 Injecting drug users, Outcome 1 Treatment completion.

Discussion

Summary of main results

TB cure and treatment completion were low with self‐administered therapy in these trials, and direct observation did not substantially improve this. Positive effects with direct observation were seen in two trials where patients in the control group were only seen in clinic once a month, but not in the three trials where the controls were seen more frequently (every one or two weeks).

Trials comparing home observation (community observer or family observer) to clinic or healthcare worker led observation did not show any difference in TB cure or treatment completion. Within home‐based direct observation, there were no differences between direct observation by a family member and direct observation by a community health worker.

Overall completeness and applicability of evidence

This Cochrane Review includes trials from both high‐ and low‐burden countries, conducted between 1994 and 2008. Direct observation was implemented in line with current recommendations, and the findings remain applicable to TB treatment programmes today.

Cure and treatment completion with self‐administered treatment were low in these trials, consistent with the outcomes seen in TB programmes at the time, and consequently it is remarkable that direct observation failed to substantially improve these.

One interpretation, offered by Frieden 2007, is that these trials failed to implement direct observation effectively. This interpretation seems unreasonable to us, as the authors of the included trials did what they could to implement direct observation, and it may even be harder to implement, and less successful, outside of a clinical trial . Alternative interpretations are that the health systems were struggling to deliver TB treatment, and direct observation on its own did not resolve these underlying issues, or that TB patients experience financial or logistical barriers to compliance with direct observation. For example, it may cost patients money if they have to visit a health facility as was the case in Walley 2001 PAK.

Direct observation as a strategy is still debated in TB and other chronic diseases. These debates and the findings of this review and others are important given the often huge resource implications of implementing a direct observation therapy programme. In two studies conducted in Brazil to evaluate the cost effectiveness of direct observation strategy; one reported a doubling of indirect costs to the patient compared to self‐administered therapy (Mohan 2007) while the other reported an incremental cost‐effectiveness ratio (ICER) of USD6616 per completed direct observed treatment compared to self‐administered therapy (Steffen 2010).

Agreements and disagreements with other studies or reviews

The findings of this Cochrane Review are similar to the findings of a meta‐analysis by Pasipanodya 2013, who reported no difference between DOT and self‐administered treatment in terms of reduction in microbiological failure, adverse drug reactions acquired resistance and relapses. The meta‐analysis included ten studies, five RCTs and five observational studies. However, it is worth noting that the included trials were quite heterogenous for a meta‐analysis, the quality scales used were quite unclear and their findings were probably influenced by one observational study.

In their clinical review, Chan 2002 reported that direct observation is essential and effective for treatment and, by extension, TB elimination. The review was not systematic and mainly looked at areas where DOT has been done in conjunction with other interventions. It is probable that the other interventions or inputs, rather than specifically observing a patient as they take their medication, were beneficial for benefit, as highlighted by Volmink 2000b. A review by Tian 2014 reported that direct observation at a clinic was not more effective than self‐treatment; but that community direct observation may be, with no difference detected between family and non‐family direct observation. The review did not assess the inputs and associated supervision to the extent that we did in this Cochrane Review.

Ford 2009, a review of direct observation in HIV therapy, also reported no effect on virological suppression. Though it might be argued that the therapy of these two diseases is different given that TB is for a finite duration whereas HIV is lifelong and thus adherence issues are different.

Authors' conclusions

The available evidence indicates that direct observation, even when supervised by health staff, does not resolve poor adherence in TB treatment. Given the huge cost implications of direct observation, policy makers therefore might want to rethink their strategies for improving adherence. It is probably worthwhile in considering financial and logistical barriers to care, motivating patients and staff, and enhancing defaulter tracing mechanisms.

The lack of effects of direct observation in improving cure and completion rates is surprising but reflects the complexity of adherence. Further research in well functioning health systems is needed to assess alternative and complementary strategies to direct observation. Qualitative work focusing on defaulters, where defaulter mechanisms exists and how clinicians interact with patients would especially be important. Evaluation of the cost of DOT for patients and providers would also help in properly assessing these strategies.

What's new

History

Protocol first published: Issue 4, 2001
 Review first published: Issue 4, 2001

Study	Reason for exclusion
Batki 2002	Compared direct observation plus with methadone treatment for injecting drug users with routine TB treatment without methadone.
Carroll 2004	Before‐and‐after study; no control group.
Hwang 2004	Not randomized.
Jasmer 2004	Different criteria for allocation to self‐administration or direct observation.
Lewin 2004	An educational intervention was evaluated.
Malotte 2001	Evaluates incentives for IV drug users within the context of a direct observation programme.
Matthew 2002	Cohort study.
Moulding 2002	Trial evaluating devices that monitor treatment using uranium along a strip of photographic film.
Pungrassami 2002a	Not randomly allocated; A publication reporting same data as Pungrassami 2002b.
Pungrassami 2002b	Not randomly allocated; A publication reporting same data as Pungrassami 2002a.
Sorete‐Abore 2002	Cohort study.
Tandon 2002	Described as a RCT, but the randomization led to very different numbers in the 2 groups; subsequently over 50 participants (out of a total of 379) crossed over from self‐treatment to direct observation and were excluded from the analysis; little detail for the rest of the study provided.
Thiam 2007	Multifaceted intervention including DOT.
Toyota 2003	Patients in hospital.

Comparison 1

Directly observed versus self‐administered

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Cure (negative sputum smear in last month of Rx in patients +ve initially)	5	1645	Risk Ratio (M‐H, Random, 95% CI)	1.08 [0.91, 1.27]
2 Cure (by intensity of monitoring in control group)	5	1645	Risk Ratio (M‐H, Fixed, 95% CI)	1.07 [1.00, 1.15]
2.1 Monthly monitoring of patients in self administered group	2	900	Risk Ratio (M‐H, Fixed, 95% CI)	1.15 [1.06, 1.25]
2.2 Once every two weeks monitoring of patients in self‐administered group	1	497	Risk Ratio (M‐H, Fixed, 95% CI)	0.96 [0.83, 1.12]
2.3 Weekly monitoring of patients in self‐administered group	2	248	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.68, 1.21]
3 Treatment completion (both with smear sputum test at end and those without)	6	1839	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.96, 1.19]
4 Treatment completion (grouped by frequency of monitoring in the self‐administered therapy group)	6	1839	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.96, 1.19]
4.1 Monthly monitoring of self‐administered treatment	3	1073	Risk Ratio (M‐H, Random, 95% CI)	1.12 [0.95, 1.31]
4.2 Once every two weeks monitoring of self‐administered treatment	1	497	Risk Ratio (M‐H, Random, 95% CI)	0.99 [0.87, 1.14]
4.3 Weekly monitoring of self‐administered treatment	2	269	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.74, 1.46]

Comparison 2

Home observed versus clinic observed

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Cure (having a negative sputum smear test in the last month of treatment having been smear‐positive initially)	4	1556	Risk Ratio (M‐H, Random, 95% CI)	1.02 [0.88, 1.18]
2 Treatment completion (both with smear sputum test at end and those without)	3	1034	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.91, 1.17]
3 Cure (stratified by intensity of observation)	4	1556	Risk Ratio (M‐H, Fixed, 95% CI)	1.01 [0.91, 1.11]
3.1 DOT (Intense supervision of observer)	1	522	Risk Ratio (M‐H, Fixed, 95% CI)	1.08 [0.91, 1.28]
3.2 Routine supervision of DOT	3	1034	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.86, 1.10]

Comparison 3

Community observed vs family observed

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Cure (having a negative sputum smear test in the last month of treatment having been smear‐positive initially)	2	1493	Risk Ratio (M‐H, Random, 95% CI)	1.02 [0.86, 1.21]
2 Treatment completion (both with smear sputum test at end and those without)	2	1493	Risk Ratio (M‐H, Random, 95% CI)	1.05 [0.90, 1.22]

Comparison 4

Injecting drug users

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Treatment completion	1	300	Risk Ratio (M‐H, Fixed, 95% CI)	1.0 [0.88, 1.13]