Effectiveness of Directly Observed Therapy (DOT) for Tuberculosis

Introduction Directly observed therapy (DOT) for tuberculosis (TB) is widely recommended by experts (6,20,50,67,71,100) and has been endorsed by local (12,21,42,99), national (17), and international (104–5) health authorities as the “standard of care” for TB treatment, yet it remains controversial in some quarters. Critics have questioned the appropriateness and net effectiveness of “universal” DOT on pragmatic, scientific, and ethical grounds (4,6,25,47,96–7,108). They suggest that DOT may be poorly accepted and hence fail to improve outcome in some settings, that it may consume scarce resources better directed elsewhere, and that concurrent improvements in other aspects of previously ineffective TB programs, in accordance with the “DOTS strategy” of the World Health Organization (WHO) (105), may actually account for much of DOT’s alleged benefit (6,45,47,72,96–7,108). Skeptics may also cite the lack of critically designed trials and question the cost-effectiveness of building an infrastructure for DOT in districts that already achieve good results with unsupervised therapy (however, these often resort to selective DOT) (6,45,72,87,96). Thus an evidence-based review published in June 2001 rated the DOT option as having “unknown effectiveness” (47). Modern chemotherapy of TB is highly successful provided a well-designed drug regimen is taken as planned. Even effective TB programs, however, fail to achieve long-term cure in a small minority of fully treated patients (typically 1%–5%) (40,48), and completion rates can fall below 50% when there are major impediments to “case-holding” (keeping patients engaged in treatment until they complete their designated drug regimen) (9,11,30,82). A poor treatment outcome may occur for behavioral reasons on the part of the patient (missed doses) or the program staff (faulty selection and dosing of drugs; ineffective case-holding due to administrative inconveniences, lax follow-up, nonsupportive interactions with patients, etc.) Other potential reasons are biologic in nature: impaired drug bioavailability related to enteric malabsorption (55) or pill formulation (92), poor penetration of circulating drug or drugs into infected sites (34), deletion of a key drug or drugs due to an adverse effect or acquired drug resistance, and—possibly—inadequate host defences (2,51,78). In addition, reinfection by a new strain of Mycobacterium tuberculosis during or after treatment (83,94), if it manifests clinically, is likely to be misinterpreted as a failure to eradicate the initial strain. To clarify DOT’s actual “track record” we review the published treatment outcomes of contemporary (human immunodeficiency virus [HIV] era) DOT cohorts. We examine how often and why TB treatment was unsuccessful despite DOT, focusing on 2 critical outcome measures, treatment failure and posttreatment relapse (29,40,48). Given that faulty adherence to pill-taking is regarded worldwide as the foremost cause of poor outcome (13,19,40,62,72,74,91), DOT, if successfully implemented, should yield measurable reductions in failure and relapse in a variety of settings. We reasoned that the consistent application of DOT removes nonadherence from consideration: any remaining failures of a first-line regimen could then be attributed to biologic, not behavioral, problems. Methods Cohorts were identified by manual search of relevant studies, reviews, and journals and by a computerized literature search (MEDLINE, National Library of Medicine, Bethesda, MD) combining text word sets: 1) tuberculosis and 2) DOT or supervised therapy crossmatched with 3) cohort, outcome, failure, or relapse. We arbitrarily limited the scope to reports published in English during 1990–2000 (11 yr); this effectively encompasses the contemporary DOT era and includes all reports to our knowledge (in English) on DOT for HIV-infected cohorts. The major omission is the many treatment trials conducted before 1990 (and pre-HIV) under the auspices of the British Medical Research Council that employed DOT throughout; this monumental edifice of clinical research is summarized elsewhere (40,64–5). Studies met our inclusion criteria if 1) they reported treatment outcome in a case cohort or clinical trial involving adult TB patients, including data on treatment failure and/or relapse; 2) treatment was directly observed for the entire course; and 3) standard chemotherapy was used, that is, isoniazid and rifampin for 6 months, pyrazinamide for the first 2 months, with or without other drugs in the intensive phase. We allowed several partial exceptions, as noted in the tables. We considered studies that employed culture for diagnosis (Group 1) separately from those that relied on sputum acid-fast bacillus smear alone (Group 2); for the latter we included data on smear-positive pulmonary TB only. For reports that segregated new cases and retreatment cases, we excluded the latter. As most series included patients with drug resistance, we determined its overall prevalence (to any drug) and noted the impact of multidrug-resistance, defined as resistance to isoniazid and rifampin with or without other drug resistance. In accordance with WHO’s categorization (105), we counted successful treatment (completion, cure), treatment failure, death during therapy, major nonadherence (default, prolonged interruption, or loss to follow-up), and transfer out as 5 mutually exclusive outcomes. Relapse rates were mainly calculated as simple proportions of patients, as time-based rates (for example, events per 100 person-years of observation) were infrequently reported. We contacted some authors directly for additional data or clarification. Results We found 34 eligible studies: 24 culture-based (Group 1;Table 1) and 10 smear-based (Group 2;Table 2), comprising 6,822 and 71,431 analyzable cases, respectively. Most were conducted in resource-poor settings. The main geographic sources were sub-Saharan Africa (12 studies; South Africa in 8), the United States (9 studies), and Asia (9 studies). Caution is required in comparing the studies, as they represent a broad diversity of cultural setting, level of TB program development, clinical and mycobacteriologic methods, and prevalence of drug resistance and HIV coinfection. The institutional settings range from well-established TB programs associated with urban academic medical centers to newly organized programs in underserved and impoverished rural sites. Program staffing, the nature and frequency of clinical monitoring, the methods for providing and supervising DOT, and the use of incentives and enablers vary widely. Data quality is difficult to assess in most reports but probably varies substantially as well; the data range from those acquired prospectively in controlled trials to those recorded by lay workers under field conditions and analyzed retrospectively. There are important differences in outcome analysis, particularly in the availability and adequacy of posttreatment follow-up. The studies are therefore not suitable for formal meta-analysis.TABLE 1: Group 1: TB diagnosed by culture of M. tuberculosis *TABLE 1: Group 1: TB diagnosed by culture of M. tuberculosis— *—ContinuedTABLE 2: Group 2: TB diagnosed by AFB smear onlyTABLE 2: ContinuedDespite the major disparities in their methods, most groups reported good efficacy in patients who remained on therapy. Treatment failure, which most studies assessed after 5–6 months of chemotherapy, occurred in <5% of patients in 26 of 30 evaluable reports. Six studies in Group 1 (comprising 641 evaluable patients) and 1 in Group 2 (929 patients) had no failures, while the highest rate (6.8%) was in a study that used a broad definition of failure (13). The average failure rate (unweighted mean ± SD) was 2.4% ± 2.2% for 21 culture-based studies and 2.5% ± 1.7% for 9 smear-based studies (NS). However, if one includes patients who were nonadherent or lost to follow-up in an intention-to-treat analysis, net program effectiveness was substantially lower in most studies. Thus the combined rate of failure plus default was 11.1% ± 6.7% (n = 20) and 10.0% ± 7.5% (n = 9) in culture-and smear-based studies, respectively (NS). The default rate exceeded the failure rate in 16 of 23 evaluable reports, indicating the greater importance of nonadherence. Additionally, poor adherence may account for some failures in patients who did not default from treatment. The inclusion of TB-related deaths would further increase the aggregate “poor outcome” rate at the end of therapy, but information on cause of death was largely unavailable (most deaths during DOT in the HIV-infected are due to other causes). The crude relapse rate was 3.6% ± 2.4% in the 21 culture-based studies that provided posttreatment follow-up, and 3.2% and 3.3% in 2 smear-based studies, respectively. The quality, planned duration, and actual success of follow-up varied widely and are not documented in several studies. Only 9 of the 23 provide sufficient information to judge that at of or for 2 or in 9 the mean follow-up exceeded 1 Given the of cases with follow-up, or plus the that relapse is excluded from relapse the overall relapse rate was probably in many studies. reinfection may have the relapse rate in some (for example, all with relapse in this was excluded in 1 South cohort of initial and was for patients from 6 studies in on or drug resistance with a new strain was found in 2 patients while In the remaining 2 the but the relapse reinfection There were no differences in failure or relapse when we studies to research study program urban setting, and there to be failures in not However, the of studies are small for studies provide data on overall adherence or the of planned but the fail to how DOT was actually The most nonadherent cases are under their We to poor adherence also to failure and but 10 of the 34 reports provide any relevant In studies, treatment, or were associated with failure or relapse the importance of in treatment, a increase in poor outcomes in patients who met their definition of a at 2 months and cure and were the of failure while in South study a of therapy (to for doses) was associated with cure In nonadherence did not in the remaining studies: the 2 patients in of initial the poor outcomes at and all patients who had taken of reinfection was not of initial drug resistance from to in Group 1 and had In studies resistance in 1) did not to poor outcomes. remained at failure or relapse in of evaluable patients from 6 studies patients from studies who had initial resistance to isoniazid and/or 2 treatment. in these studies, 1 of cases with reported data had initial resistance In 6 other studies, however, drug resistance in did outcomes. 9 of 16 of adherence good or from studies and was the for failure in (to any drug) was in 5 of 6 failures in of 6 in and was a for TB in In the remaining 10 studies initial drug resistance was excluded not related to any poor outcome or not the of drug poor outcomes of TB a major Most cases worldwide in the infrastructure for TB is treatment has often been associated with poor cure rates in and the the of DOT has with major improvements in TB in several and in authorities that DOT has been key in improvements it the most in TB treatment in a However, have and DOT, particularly its application of DOT on treatment completion as an of DOT’s effectiveness a completion rate is to good TB completion not long-term We on failure and default with death and the of and on a standard of regimen efficacy in an to how often and why DOT may be one would assess the effectiveness of DOT with a controlled trial comparing it with concurrent unsupervised therapy in an is not In the of DOT’s unsupervised treatment is on with and/or by clinical and of the 34 studies a of failure and relapse rates with DOT and no to outcomes in their The remaining 9 all noted improvements in completion and/or cure rates with the of In 6 DOT was as part of a to the program infrastructure in an underserved with TB In the other adherence to unsupervised therapy had been poor despite a health these 9 studies, 2 provide evaluable In urban the use of DOT of patients) the rates of and acquired resistance to and of their respectively In DOT completion and cure rates at and but not at centers We found controlled all conducted in settings, that DOT with concurrent has and/or limited The first did not DOT, but its overall results were due largely to nonadherence The of the are for several other many patients were excluded from posttreatment outcome was not and transfer to was counted as the patients had to their adherence to a they were is for unsupervised therapy, which may have their In a related study comparing DOT, DOT supervised by a lay and the authors found no outcome differences the with a DOT by lay health workers study also has outcome and The authors not the behavioral to patient adherence at their DOT In the controlled trial DOT but improvements in case-holding and cure at 6 DOT and patients better a study has of and relapse by that its The controlled trial was not eligible for our after no rifampin in unsupervised therapy, by a at a health or in the and by a The rates did not and were difficult program all were to other aspects of under The and poor of DOT its failure to unsupervised treatment. The Group 1 outcomes 1) those in studies conducted by the British Medical Research Council and in the The studies employed the regimen of with or without or by DOT for 6 months controlled trials treatment failure rates at 6 months that were in some and relapse rates at months posttreatment that were <5% and Most Group 1 outcomes also with those under program conditions by the in and by in the in the the other DOT program results 1) are not in some are in controlled trials that relied on unsupervised pill-taking under program The trials in and the United States relapse rates of and and the failure rate was in the British trial = In patients, relapse rates after DOT varied from to with the observed after unsupervised treatment in and are and the methods and clinical follow-up in most DOT studies. from most reports the poor outcomes under DOT to drug efficacy when or patient adherence and/or program Additionally, may into in that DOT programs with and outcomes are likely to their the the of DOT unsupervised therapy for TB has not yet been in an evidence-based most authorities are that DOT treatment drug resistance and overall TB of DOT poor outcomes acquired drug which are with unsupervised treatment and to DOT can overall for TB despite its initial used to the cost-effectiveness of DOT, on urban in the United that DOT is effective and case unsupervised treatment The questioned the cost-effectiveness of DOT, on outcomes in patients at of default treatment, of and that the quality of TB programs may be DOT A of actual program found that DOT in of net have also DOT studies have the of DOT in settings found that DOT was in rural South Africa but for the were by an initial which would not be for many TB and comparing rates during TB treatment is as deaths are often to other the and or sites of of the TB health conditions in the under study (for example, and the adequacy of medical of diagnosis of and use of first-line HIV is other of overall during TB the cause for most in the 34 was not and some deaths have patients lost to follow-up. was in of evaluable studies, of HIV and was in of The was in a cohort in but the of TB in some of these 26 deaths Given the and the lack of we could not DOT in the a benefit is be likely in with a prevalence of drug resistance, as these cases often poorly with unsupervised treatment. mainly unsupervised therapy in the rifampin national rates from all during TB have been in and rates have been reported from Africa However, rates of have been observed is and in patients the rate at The in most DOT programs 1 and 2) may a but it should be noted that treatment trials and TB programs often have rates case selection and the that deaths from TB often occur before or after the of treatment in which would into many DOT cohorts. to DOT In a of DOT patients this outcome, by the of or by acquired drug resistance, as The effect of partial adherence may on the of the regimen is or and biologic as the and the of many can be without a measurable in long-term cure remains and would be difficult to from that the relevant outcomes by are in and by The 34 studies to clarify the impact of on There are no data adherence to poor and the adverse events it is from the default data alone that major nonadherence remains under methods by which DOT programs may adherence that first-line have a at TB in treatment as as with patients are and to of planned achieve and the loss of isoniazid has effect on outcome in newly treated cases In a trial of and isoniazid in the omission of (to poor in adverse outcomes However, relapse and the of these data to standard is for of planned has been widely in clinical trials and studies and is the standard of a of local and national health on much clinical the a it has not been as the and it was for not To for a of the treatment is recommended clinical is often on the of nonadherence and other Program staff as as patients may be The of DOT success if case-holding is staff are lax or with to actual and pill is not some of pill may The frequency of poor staff is largely or but can be in and out of in 1 program an of were not actually observed resistance from the adverse effect of multidrug-resistance, when the of relapse in Group 1 was not to initial drug resistance. studies that resistance to drugs but not standard treatment. There were a of acquired resistance which during DOT resistance was of importance in new cases overall (Group it probably often to poor outcomes during retreatment prevalence is or in retreatment cases, in is The impact of resistance in smear-based studies is but its prevalence may be in at several of the standard first-line and retreatment can yield results when to with drug resistance, as in and in these treatment failure DOT) a of local drug or resistance HIV effect The in TB patients with and other that the drugs not the most of the of For example, those with HIV circulating often have or yet in most they to chemotherapy and have a Studies in varied settings have found failure rates to be by HIV the of immunodeficiency assessed by was in many patients cure rates are as relapse was not assessed in some patients is often by other and including the of reinfection are difficult to account To studies no effect of HIV on relapse a in 2) but the to the question A found a increase in TB in in HIV-infected a minority of had Other data analyzed on South and from rates in the reinfection is likely to rates from settings, in the an HIV effect Thus the data are but not out a increase in relapse to HIV as a of Studies have found that overall during TB treatment is in patients, but this is largely due to other of TB-related is also in the but it remains the which occur are due to immunodeficiency per or to as diagnosis and treatment, initial drug resistance, poor and concurrent outcomes from TB may with the use of Other biologic Other biologic for poor outcome, as impaired drug bioavailability and the clinical or of the were not for in cohort studies of the The information on these is largely study reported data on drug overall importance remains controversial despite some with treatment failure Only a study with posttreatment follow-up clarify the impact of drug malabsorption on treatment and the would be if drug were to as it would be to their effect on all patients had pulmonary Only 2 reports considered its clinical found that was a for TB (in to and cure in cases were included in but of poor outcomes at sites is to The availability of has that some cases of relapse in represent reinfection to Most of the assessed in Group 1 were from of to TB The reports suggest that but this may not in data are of DOT The net effectiveness of DOT under program conditions from to with the differences largely to poor Most programs good particularly when one their and the conditions that some However, 16 to WHO’s cure rate of and in 9 completion rates were also conditions prevalence of drug resistance, good DOT program long-term cure rates should in when is biologic failures are with the chemotherapy To better poor outcomes biologic behavioral research studies should actual DOT drug in cases, if not host other and the clinical of that fail treatment or and reinfection as a cause of A of unsuccessful cases can programs to for in the and behavioral aspects of TB program to overall success in TB it difficult to the of per improvements in program and methods, drug to patients, may have that are from those of DOT, yet in an the other we it to DOT in the of observed drug while other that are to a DOT The latter vary with the and but at the should staff and patients, the and to adverse drug and concurrent medical and (in most incentives and may to other medical and in the treatment or HIV and to and health In the of our program and in urban these are to DOT case-holding and hence to program particularly with In resources are scarce and DOT be at medical centers and/or supervised by some of DOT be lay or are the program is and probably this in urban with to health In rural or underserved and are often lay is a to unsupervised treatment In settings DOT supervised by local can effective has the potential to probably and DOT However, DOT be fully from reports to outcomes are and data on adherence and relapse are programs have been under they probably represent a TB and they can be results under DOT often nonadherence biologic reasons for treatment To further improve program outcomes DOT be and accepted by patients, which is not the case and it use for local drug resistance data and may the of DOT in well-designed programs, at and To programs should to local cultural as elsewhere are better data on actual adherence patients and program staff of the of partial and further research at case-holding In the published with DOT that it has TB treatment outcomes in a variety of settings, most of the data are and the for DOT’s unsupervised pill-taking is largely DOT has the potential to improve long-term cure rates and be in most treatment settings, its would vary in Studies in which DOT to unsupervised therapy in settings with limited resources and indicating that program quality be for DOT to yield its potential its effectiveness would a controlled a that we would as in our the with DOT is our by published and clinical that the of DOT to TB and The use of directly observed therapy (DOT) for tuberculosis (TB) is widely recommended by TB but some have questioned its effectiveness and We review reports the outcome of DOT in treated in the immunodeficiency virus era The 34 studies represent a variety of and medical settings and program data and clinical Most reported rates of treatment failure and posttreatment relapse ± ± = and ± = effectiveness was substantially lower in many of rates of treatment completion and cure were in 9 and respectively. The methods of outcome and for posttreatment relapse was often (11 or that long-term cure rates in most cohorts. of initial drug resistance was not in the of but substantially at some sites. HIV had impact on outcome deaths during but likely for TB-related deaths and To better why poor outcomes occur despite DOT, we data on actual which was not and on and drug and strain (to DOT is to unsupervised its to improvements in TB be adherence remains the to the to on behavioral and aspects of program