Tuberculosis (TB) has reappeared as a serious public health problem. Non-compliance to antituber-culous drug treatment is cited as one of the major obstacles to the containment of the epidemic. Compliance may be optimized by Directly Observed Treatment (DOT) and short-course treatment regimens. Since 1986, Tanzanian TB patients have received daily DOT at health facilities for the first 2 months of the treatment course. However, adherence and cure rates have been falling as the number of TB cases continues to increase and the burden on already stretched health facilities threatens to become unmanageable. We used an open cluster randomized controlled trial to compare community-based DOT (CBDOT) using a short-course drug regimen with institutional-based DOT (IBDOT). A total of 522 (301 IBDOT and 221 CBDOT) patients with sputum-positive TB were recruited. Overall, there was no significant difference in conversion and cure rates between the two strategies [M-H pooled odds ratio (OR) 0.62; 95% confidence interval (CI) 0.23, 1.71 and OR = 1.58; 95% CI 0.32, 7.88, respectively] suggesting that CBDOT may be a viable alternative to IBDOT. CBDOT may be particularly useful in parts of the country where people live far from health facilities.
Mycobacterium tuberculosis is probably the single largest infectious cause of death in the world, with almost 3 million deaths annually (Dolin et al. 1994; Raviglione et al. 1995). The increasing global burden of tuberculosis (TB) on public health is exacerbated by the HIV/AIDS pandemic (Harries et al. 1997). In developing countries, the major burden of TB disease and death is concentrated on the economically most productive age group (Arata 1991), intensifying the economic and developmental implications of this curable disease.
Recent trends in TB case notification in Tanzania are alarming: in 1980 there were 5103 notifications across the country, in 1986 this had risen to 15 452 and in 1997 a total of 46 433 were notified (WHO 2000). In Kilombero District of southern Tanzania, the number of TB cases doubled between 1995 and 1996 [District TB and Leprosy Co-ordinator (DTLC) 1995, 1996]. At the same time, 58% of the registered patients were lost to follow-up and less than 40% of patients were cured (Ifakara Health Research and Development Centre (IHRDC), unpublished data 1996]. In order to counter these alarming developments, the Kilombero District Health Management Team (KDHMT) launched the Kilombero Tuberculosis Research Programme (KITUPA) in collaboration with a district-based research centre (IHRDC) and the Tanzania National Tuberculosis and Leprosy Programme (NTLP).
The current epidemic is marked by high rates of non-compliance with anti-TB regimens (Raviglione et al. 2001). In response to this, Directly Observed Treatment (DOT) with short-course drug regimens has been recommended as the main means of ensuring compliance and as the most important step in the prevention of drug-resistant disease (Kent 1993; WHO 1997). An evaluation of the country TB programmes conducted by the International Union Against Tuberculosis and Lung Disease (IUATLD), found that Malawi, Mozambique, Nicaragua, Tanzania, South Africa and Botswana achieved cure rates of more than 80%, mainly by means of institutional-based direct observed therapy (IBDOT) (Broekmans 1994). The DOT has been hailed as one of the best ways of ensuring that patients take all of their medication (Westaway et al. 1991; Wilkinson & Moore 1996).
However, over-stretched health facilities are struggling to manage IBDOT effectively and if current trends in case notification continue, IBDOT is likely to become unsustainable. IBDOT is not always convenient and accessible to TB patients (Akkslip et al. 1999), and there is no single formula for implementation of DOT (Karim 1997). We decided to evaluate the efficiency, cost-effectiveness and acceptability of community-based directly observed treatment (CBDOT) with a short-course treatment regimen, in comparison with the existing IBDOT strategy. The aim of the study was to assess whether in new patients with smear-positive pulmonary tuberculosis (PTB) the smear conversion rate at 2 months and treatment outcome (specifically cure rates) were similar using CBDOT and IBDOT. This pragmatic cluster-randomized controlled trial was conducted under programme conditions.
Materials and methods
The study was based in Kilombero District in the Morogoro region of southern Tanzania (Figure 1). The district covers an area of 14 918 km2 and has an estimated population of more than 250 000 (1988 national census). Most of the people of this district are subsistence farmers, with a very low overall family income (Alonso et al. 1994). Health services are offered in 40 health units, including a 380-bed designated district hospital, one parastatal hospital with 120-beds, four health centres, each with a maximum of 20 beds, and 34 dispensaries. Curative TB services are offered in 18 health facilities among which seven have diagnostic facilities based on microscopy for acid-fast bacilli (AFB). Anti-TB drugs (rifampicin, isoniazid, pyraminazide, ethambutol and streptomycin) are available in the treatment centres and there is a specific health worker who is responsible for TB case finding, treatment and tracing of defaulters as part of his/her daily activities. The DTLC is the representative of the NTLP at district level and is responsible for the management of all TB and leprosy activities. All trial activities were conducted in conjunction with the DTLC. Patients from health facilities without TB and leprosy treatment facilities [private health facilities, non-governmental organizations (NGOs)] are referred to health facilities with diagnostic and curative facilities.
This study was an unmasked cluster-randomized controlled trial. The 18 TB treatment units were matched in pairs on the basis of available data on type of health facility and number of TB patients registered in the past year. Each pair was given a unique number (from 1 to 9), and one treatment unit within each pair was randomly assigned to IBDOT or CBDOT.
Diagnosis was based on the standard NTLP practice. All patients presenting to treatment units between January 1999 and June 2000 with complaints of prolonged cough, wasting or night fevers were suspected TB cases. Each patient was sent to a diagnostic centre where three sputum specimens were collected for direct microscopy. The first sputum was collected on the first day of visit under the supervision of a health worker. The patient was given a container to bring early morning sputum the next day and during the second visit the third specimen was collected under the supervision of a health worker. All sputum specimens were assessed for the presence of AFB by Ziehl-Nielsen staining and light microscopy. In our study, smear positive PTB was defined as a suspected TB case with at least one positive smear examination. Confirmed cases were referred back to their nearest treatment unit for therapy.
In Tanzania, the treatment for TB is offered free of charge. The treatment recommended by the Tanzanian National TB and Leprosy Program was used in both the CBDOT and IBDOT arms and consisted of an intensive phase (2 months) under direct supervision. The drugs given during the intensive phase are rifampicin (10 mg/kg) and isoniazid (5 mg/kg) in a combined formulation tablet (RH); pyrazinamide (Z) at a dose of 15–30 mg/kg and ethambutol (E) at a dose of 15–25 mg/kg. The intensive phase is followed by ethambutol (10 mg/kg) and isoniazid (5 mg/kg) in a combined formulation tablet (EH) daily for 6 months. Patients with other forms of TB were treated and followed up according to the national TB guidelines (Tuberculosis/Leprosy Control Unit 1995).
Details of the intervention
For CBDOT, a community member (CBDOT observer), living in the same village as that of the patient, observed the patient daily during the first 2 months (intensive treatment phase). The CBDOT observers who volunteered to work were interviewed and selected by the village leaders and trained by the DTLC to supervise the patient's intake of the TB medications during the intensive treatment phase. During this phase, the CBDOT observers were visited fortnightly by the TB health worker in charge of the nearby health facility and once per month by the DTLC. During house visits, the health workers monitored adherence to treatment by checking the treatment card and counting pills remaining from the patient's monthly drug supply. In the intensive treatment phase (first 2 months), of the standard IBDOT, the patient had to visit the health facility daily to be observed while swallowing the medications.
During the whole course of treatment, patients under CBDOT had to visit the health facility monthly for drug collection and those under IBDOT had to do so everyday for the first 2 months of treatment, and every month in the continuation phase of 6 months. In patients under CBDOT, drugs were collected from a nearby health facility by a patient or her/his CBDOT observer, depending on the condition of the patient.
In the continuation treatment phase of 6 months, patients from both approaches had to swallow the medications daily under self-supervision. Moreover, on completion of 2, 5 and 7 months of treatment, patients from both arms were asked to visit the diagnostic centres for a sputum check. As well as the community participating in the choice of CBDOT observers, verbal consent was obtained from all patients who participated in the study.
Endpoints, sample size and statistical methods
We aimed to compare the effect of the two strategies during the intensive treatment phase, which was under the supervision of either the community observer or health facility worker. Our analysis included only patients who had AFB-positive PTB treated with ambulatory short course chemotherapy. Our primary outcome was the conversion rate, defined as the proportion with AFB-negative sputum after 2 months of treatment. Our secondary outcome was the cure rate at 7 months, defined by AFB-negative sputum at 2 months, remaining negative at month 5 and/or at month 7 after the start of treatment.
We estimated that a total of 250 patients in each arm would provide more than 80% power to detect a 20% difference in the conversion rates and cure rates during follow-up, at a 5% significance level (Hayes 1991). Continuous variables and categorical variables were compared using Student t- and chi-squared tests, respectively.
Since communities (treatment units) rather than individuals were used as the unit of randomization, endpoints were calculated for each of the eight paired clusters and the method described by Mantel and Haenszel (1959) used to obtain a pooled estimate for the completeness of follow-up, conversion rate and cure rate. The probabilities of converting or being cured were also compared between the groups using regression methods, the standard error of the coefficients for clustering on treatment units being corrected using the Huber-White-Sandwich method (Huber 1967). Analysis was performed by intention to treat and sensitivity analyses performed to evaluate the potential effect of missing responses.
The IHRDC and Scientific Ethical Review Committee and the Tanzania Medical Research Co-ordinating Committee approved the study.
Figure 1 shows the name, location and intervention group of each health unit participating in the study. Table 1 shows that at baseline, the groups were similar in terms of age and sex of patients. The trial profile (Figure 2) shows that, of 1619 patients registered during the course of the study, 522 with AFB-positive PTB were prescribed ambulatory short-course chemotherapy and were included in the analysis presented here. More patients were randomized to IBDOT (301) than CBDOT (221).
Table 1. Baseline characteristics of patients by intervention group
Completeness of follow-up at 2 months was 87% (455 of 522) and 68% (311 of 473) at 7 months. Overall losses to follow-up 2 months after recruitment were similar in the IBDOT (33 of 301) and the CBDOT (24 of 221) groups [odds ratio (OR) 1.14; 95% confidence interval (CI) 0.64, 2.01]. However, the clusters were not homogenous in this respect (χ2 test 9.8, d.f. = 5, P = 0.08) with significantly more losses to follow-up in CBDOT patients in Mlimba–Kidodi [zero of 53 vs. nine of 64, OR 18.0 (95% CI 1.0, 322)] and Idete–Zignali (zero of two vs. one of four, OR 2.14; 95% CI 0.06, 77.5). At 7 months, the overall loss to follow-up was higher in CBDOT compared with IBDOT (88 of 214 vs. 74 of 259, OR 1.92; 95% CI 1.29, 2.8) and significant heterogeneity between the two groups persisted (χ2 test = 30.4, d.f. = 7, P < 0.001). Fewer patients randomized to CBDOT died than under IBDOT (OR at 2 months 0.13; 95% CI 0.03, 0.65 P = 0.01 and OR at 7 months 0.12; 95% CI 0.02, 0.89, P = 0.04). Excluding St Francis Designated District Hospital (SFDDH) and Kilombero pairs did not significantly change these results (OR 0.13; 95% CI 0.17, 0.99, P = 0.05 at 7 months).
The proportion of patients transferred out from CBDOT was also lower than in IBDOT (OR 0.29; CI 0.07, 1.15, P = 0.08 at 2 months and OR 0.30; 95% CI 0.13, 0.67, P = 0.04, at 7 months).
Table 2 shows that the sputum conversion rates were similar in the groups at 2 months (OR 1.17, P = 0.5). This result was unaltered when the analysis was adjusted for sex (data not shown) and sensitivity analyses did not suggest that the result would be dramatically altered by more complete follow-up. After 7 months of treatment, the proportion of patients who were cured under CBDOT was similar to that in IBDOT (Table 3). Again, adjusting the analysis for sex and the sensitivity analyses did not materially affect this result.
Table 2. Smear conversion at 2 months
OR (95% CI)
0.29 (0.08, 1.14)
Lost before 2 months
0.99 (0.44, 2.21)
Died before 2 months
0.13 (0.03, 0.65)
Known to be alive at 2 months
2.03 (0.79, 5.20)
0.62 (0.23, 1.71)
Table 3. Treatment outcome at 7 months
OR (95% CI)
0.29 (0.07, 1.28)
Lost to follow-up
1.52 (0.73, 3.14)
0.12 (0.04, 0.35)
0.63 (0.13, 3.15)
1.58 (0.32, 7.88)
We have reported a cluster-randomized controlled trial of IBDOT vs. CBDOT in the treatment of AFB-positive PTB. The primary endpoint, sputum conversion rate at 2 months, suggested that CBDOT fares no worse than IBDOT. Similarly, the cure rate 7 months after starting treatment was not significantly different in the two groups. This suggests that in terms of conversion and cure rates, it is not necessary that a health worker supervises the patient: a trained community observer can perform this task at least as effectively. Furthermore, by delivering health care closer to the patients' homes it can be expected that compliance will improve, an idea supported by the observation of a reduced proportion of transfers out in CBDOT compared with IBDOT patients. The CBDOT was practical and is sustainable in a setting where IBDOT is rapidly becoming unmanageable as a result of a huge increase in TB case notification. Finally, CBDOT can free up some health workers' time for other duties.
Our results, however, need to be interpreted with caution as some selection bias may have resulted in an overestimate of the benefits of CBDOT. SFDDH was the treatment facility with the single largest number of TB cases, with more than twice the number of the next largest treatment centre. SFDDH was also the designated district hospital and is likely to have attracted the more sick patients, with the highest risk of adverse outcome. Because SFDDH was randomized to deliver care through IBDOT, this may well have biased the results in favour of CBDOT. For this reason, we are reluctant to place too great an emphasis on the differential mortality rates, although it was encouraging that exclusion of the SFDDH-containing cluster suggested that CBDOT may still afford a benefit in terms of reduced mortality. It is also important to note that there was some heterogeneity of completeness of follow-up and outcome suggesting that some community observers may be better than others.
This was a pragmatic trial, with the intervention delivered by the DTLC and the DHMT with support from IHRDC. Although this was largely the reason for a higher-than-desirable number of missing outcomes, it is of considerable value to note that CBDOT operated satisfactory and continues to operate 2 years after the end of the trial. The issue of voluntarism of CBDOT observers should not be overlooked: motivating them is vital for the sustainability of the strategy. The modalities may vary from place to place; in our study, a feeling of prestige related to helping a fellow villager, a good relationship with the nearby health facility workers, refresher training and respect from the community were cited as some of the motivating factors of CBDOT observers. We recommend that, prior to the introduction of this strategy, communities should play the central role in identifying CBDOT observers. It is through this approach that the communities will develop mechanisms of sustaining their own selected CBDOT observers. Furthermore, the greater awareness of TB and the additional health education that results from CBDOT may help to reduce the high drop-out rates.
We have presented data suggesting that CBDOT is practicable and not inferior to IBDOT in the management of PTB. We believe that CBDOT should be used as an addendum to the existing IBDOT strategy to assure DOT in rural Tanzania, especially in areas where many patients live 10 km or more away from the nearest TB treatment unit. We do not believe that the CBDOT strategy should replace IBDOT as giving patients a variety of supervision options, and focusing on their convenience, are likely to be important determinants of the success of TB control programmes.
We thank the patients who volunteered to participate in the study. We also thank the CBDOT observers for their excellent job on a voluntary basis. The good job done by health workers in charge of TB services in each treatment unit cannot be overemphasized. We are grateful to Rashid Khatibu, Elisante Mchomvu, Faustin Gabriel and Godfrey Ndauka for supervising the fieldwork of the study. We would like to commend our coinvestigators at NIMR (Muhimbili TB Reference Laboratory) for the good job done. Very special thanks go to Dr Msamanga who was the monitor of the study for his invaluable contributions. The contributions by Rehema Kilonzo and Kesheni Senkoro cannot be overemphasized. The Tanzanian Health Users Research Fund, the Spanish Agency for International Co-operation (AECI), Fons Catalan and La Caixa, both from Spain, kindly donated the financial support for the study. Technical support, of which we are very much grateful, was received from WHO Global Tuberculosis programme through the ‘Community TB Care in Africa’ project.
Authors Fred Lwilla, District Health Office, Ifakara, Tanzania. Fred Lwilla, David Schellenberg, Camilo Acosta, Claudia Galindo, John Aponte and Pedro Alonso, Centre for International Health, Institut d'Investigacions Biomedicas August Pi i Sunyer, Barcelona, Spain. David Schellenberg, Honorath Masanja, Camilo Acosta, Claudia Galindo and Carlos Ascaso, Ifakara Health Research and Development Centre, Ifakara, Tanzania. Marcel Tanner, Swiss Tropical Institute, Basle, Switzerland. Said Egwaga and Blasdus Njako, Ministry of Health, Dar es Salaam, Tanzania.