Patient-centred tuberculosis treatment delivery under programmatic conditions in Tanzania: a cohort study

Tanzania is one of the 22 high-burden countries with respect to the number of incident tuberculosis (TB) cases[1]. Up to the mid-eighties of the previous century, the annual number of cases identified and treated in the country was manageable, but due to the developing epidemic with the human immunodeficiency virus (HIV), the caseload in Tanzania increased rapidly. Within 5 years, the number of TB cases notified tripled from 5000 to 15,000 between 1980 and 1985. In 2006, Tanzania reported over 62,000 TB cases (157 cases per 100,000 population), while the World Health Organization (WHO) estimated the incidence to be 312 per 100,000 population [1]. Thirty-five percent of smear-positive TB patients were HIV-positive in a recent study [2].

The currently recommended strategy for TB control is the directly observed treatment, short course (DOTS) [3, 4]. This strategy includes, amongst others, the daily observation of patients' drug intake at the treating facility during the initial 2 months of therapy. The reason behind this observation is that it will ensure adequate treatment adherence during the period, in which the treatment regimen contains rifampicin which will prevent the selection of Mycobacterium tuberculosis strains resistant to this powerful drug [5–8].

The National Tuberculosis and Leprosy Programme (NTLP) of Tanzania changed the standard first-line treatment regimen in 2006 from an 8-month regimen to a 6-month regimen. In the latter, both the intensive phase treatment and the continuation phase treatment contains rifampicin, while in the former rifampicin is only given for 2 months of the intensive phase. The rifampicin-containing regimen throughout the course of treatment has a higher efficacy in situations with a high prevalence of HIV-infection in the general population [9–11]. The WHO recommends the observation of treatment throughout the 6 months of this regimen at least three times a week [3]. This is clearly not feasible at the health facility level for either the patient or the health staff. Initial studies in urban and rural areas of Tanzania showed that decentralized treatment observation by guardians or former TB patients was feasible without a detrimental effect on treatment outcome [12, 13]. The main drawback encountered in a possible scale-up of this approach was the need for incentives expressed by the supporters which made the strategy non-sustainable. The experiences from these studies formed the basis of a new approach to the implementation of DOTS in Tanzania in which the observation of treatment intake could be transferred from the health facility to the patient's home and the observer could be changed from a health-care worker to a supporter of choice.

These changes maintain the core principle of daily observation of treatment (DOT) for the full duration of treatment, prevent health facilities from being overburdened and enable the patients to choose a supervision model within the constraints of their daily life. This should have a positive effect on treatment adherence and, subsequently, the efficacy of treatment. The new strategy was labelled 'patient centred treatment' (PCT).

From previous studies it is known that DOT outside the health facility, by either health personnel or laymen, is feasible and can result in treatment outcomes that are similar to those under conventional DOT at the health facility [14–18]. In Nepal, both DOT at the health facility and DOT by family members obtained success rates that met the international target of 85% [17]. However, most of these studies have been designed as randomized clinical trials and/or put restrictions on the type of supporter who can observe treatment intake. Clinical trials have the inherent limitation of coming with increased supervision and monitoring which can have a positive effect of its own on treatment delivery and adherence. Restrictions on supporters complicate the interpretation and replication of results within the context of national control programmes with a wider choice of potential treatment supporters. Thus, there is an urgent need to test the concept of delegating treatment observation outside the health facility under routine programmatic conditions. This need also holds true for fields outside the context of TB where strategies other than health facility-based treatment delivery are being examined, such as home-base care programmes for the delivery of antiretroviral therapy to HIV-infected patients.

The concept of PCT was tested and well perceived by patients and health care professionals when assessed in a qualitative survey. Furthermore, the proposed strategy was seen to be a positive contribution to adherence to therapy [2]. The strategy was formally tested in three districts under programmatic conditions with no restriction on the choice of treatment supporter. The objective of this study was to assess whether favourable treatment outcomes (cure and success) in new patients (that is, those without previous TB treatment) treated under the new strategy were no lower than those in new patients who were treated under the conventional strategy. If so, this would lend strong support to the practice of observing treatment intake outside the health facility by laymen, under programmatic conditions.

This study showed that treatment outcomes in a cohort of new TB patients receiving treatment under the PCT strategy were comparable to treatment outcomes in patients who received treatment with the conventional health facility-based DOT and administration of loose drugs. The higher success rate in the PCT cohort is driven by a larger proportion of patients completing their treatment. This finding is not easy to interpret. It might be just an effect caused by the fact that the treatment was for 2 months less. For the verifiable outcome cure, there is no difference between the historical cohort and the PCT cohort. In addition, the study did not identify any supporter characteristic that influenced the achievement of cure or treatment success in patients who had their treatment observed at home.

These findings are in line with the results from other studies performed in a variety of settings. A cluster-randomized trial in Nepal showed that 89% of the patients who had their treatment intake observed by a family member had a treatment success. This was comparable with the 85% success in patients who were supervised by a community health worker and higher than the treatment targets from the WHO (85%) [17]. A retrospective study in Thailand assessed treatment outcomes in patients who accepted direct observation of treatment by staff of the health facility, a village health worker or a family member. Ninety percent of the patients opted for observation by a family member and 10% for the health staff of the facility. Only one of the 216 patients chose observation by a village health worker. The success rate was higher in patients observed by a family member (87%) compared to patients observed by health staff (76%) [15]. A randomized trial in Swaziland reported that treatment observation by family members obtained similar success rates in TB patients as did treatment observation by community health workers (66% and 68%, respectively) [18]. Similarly, a randomized trial in Pakistan did not find any differences in treatment success rates in patients observed by family members (62%), observation at the health facility (67%) or self-administration (62%) [19].

Despite these findings, the ultimate proof of principle for the efficacy of home-based treatment observation is not given by the short-term cure rates or success rate as assessed in this and other studies. Only when, in settings with home-based DOT, the relapse rates do not increase (and preferentially decrease due to improved adherence) can home-based DOT be regarded as effective.

The concept of treatment observation by family members is still controversial within the TB community. The WHO treatment guidelines state that family members should, in general, not be considered as treatment supporters [3]. However, in the updated guidelines on community involvement in TB care, there is a shift towards the idea that there is a need to provide 'an increased range of treatment support options', although this statement was mainly geared at TB patients co-infected with HIV [20].

Despite the above mentioned results of the PCT approach, the success rate in Tanzania remains below the 85% target set by WHO. This relatively low success rate is mainly driven by the death rate. It is more than likely that this rate is largely a result of co-infection with HIV. It is known that HIV-associated death in TB patients is generally seen at the start of TB-treatment. Even the PCT approach will not have a major impact on these events, which is reflected in a death rate of 12%.

There are several arguments formulated by those who are against the use of family members as treatment supporters. First, family members are not always able to be firm on treatment intake, especially if the patient suffers side effects or if the social context makes it difficult or inappropriate to do so [21]. Second, the design of most studies showing similar, or better, treatment outcomes with observation by family members than by health workers actually test multifaceted interventions including increased monitoring, supervision and training. In the Nepal study, there were frequent visits to the households by research staff in order to monitor the implementation and to collect data. In Thailand, clinic staff were strongly involved in the identification of a treatment supporter and special boxes were prepared in order to facilitate treatment delivery at home [15]. In Senegal, the intervention was clearly multifaceted, including improved communication between health staff and patients and increased training and supervision, making it impossible to tease out the effect of decentralized treatment observation [22]. Third, treatment observation outside the health facility is, in practise, more frequently self administration of treatment. This was shown in a study by a study in Thailand where 11% of the health personnel did not perform any treatment observation, compared to 23% of community members and 35% of family members [23].

The major limitation of the study was the change in design which meant that the present study was not able to adequately test the independent effect of the type of treatment delivery. Instead, it tested a multifaceted intervention. Each part of this intervention might have contributed to the positive findings. In the PCT cohort, all drugs were taken as FDC during the full treatment period. In the control cohort, FDC was only available for isoniazid and rifampicin during the intensive phase. All other drugs were administered as loose drugs. Drug delivery by FDCs might be associated with an improved adherence to therapy [24]. At the same time, the treatment regimen was changed to include rifampicin throughout the duration. An additional 4 months of this potent drug might positively influence the cure rate in the PCT cohort. However, the additional efficacy of this treatment regimen over the conventional regimen used in the control cohort is based largely on the relapse rates and not on the initial cure rates. If the new regimen had an effect on the outcome it was most likely due to its shorter duration and, therefore, the possible higher completion rates, rather than an increased drug efficacy. However, the total intervention will be rolled-out in the rest of the country making the individual contributions of the multifaceted intervention towards the overall positive effect less important. The important finding, that the type of treatment supporter does not influence treatment outcome, makes this nationwide scale-up of the intervention more feasible.

Another limitation was the absence of demographic information for a considerable number of treatment supporters. As mentioned, this did not introduce a bias because of the concurrent start of the study in the districts and the retaining of the freedom of choice for the type of treatment delivery strategy. There is also a marked absence of HIV data form the enrolled patients. At the time of the study (2005 for the historical cohort and 2006 of the PCT cohort), routine HIV testing for newly registered patients had not been fully implemented. Collection of this information for inclusion in statistical models would lead to a bias, given the large proportion of patients who would not have this information. A comparison of this information between the two cohorts would also be not possible due to the increase in routine testing between 2005 and 2006.

There is some indication that ascertainment bias might have influenced the results in Arusha. In this district case notification increased considerably during the study and there was, also, a significantly higher probability of cure in the PCT cohort. This may indicate improved diagnosis and/or microscopy. It could also be a result of the quality control measures that were imposed in the PCT cohort but not the historical cohort. However, adjusting the multivariable models did not give any change in the point estimates or the confidence intervals, making this potential bias of little value.

Acceptable and feasible treatment observation strategies are highly context specific [25]. In Tanzania, home-based treatment observation by a supporter of choice is widely appreciated and does not lead to an increase in unfavourable treatment outcomes compared to conventional treatment delivery strategies. It allows patients to rest and recover, as daily health facility DOT places considerable physical strain on patients. Being at home also made it possible for the patient to engage in productive activities and reduce the cost of travel associated with the daily health facility visits. At the same time, it could ease the burden on health facilities [2]. Studies from Malawi and Australia have shown that, in a decentralized setting, adherence to treatment can be as high as seen in health-facility based treatment observation [16, 26]. However, it is important to assess adherence to treatment in the setting of Tanzania.

The programmatic condition in which this study was carried out lends strong support to the scaling up of the PCT strategy in Tanzania. This must coincide with rigorous supervision of the health facilities implementing the strategy. Furthermore, follow-up of this first cohort of patients treated under the PCT strategy will give much needed data on the risk of relapse and the resistance patterns in patient who need retreatment: two factors that determine the long-term efficacy of the intervention.

The positive effect of home-based treatment in a programmatic setting found in this study can be of value for areas outside the context of TB. HIV-programmes in especially resource-poor areas are experimenting with such an approach [27, 28]. The findings that home-based treatment can be achieved under programmatic conditions might give the much needed impetus to move from smaller projects to a more general implementation within national treatment programmes.

The PCT approach showed similar cure rates and better treatment success rates compared to daily health-facility DOT. The results indicate that there are no specific prerequisites for the supporter chosen by the patient. The programmatic setting of the study lends strong support for scaling-up of TB treatment observation outside the health facility.