Cost effectiveness analysis of strategies for tuberculosis control in developing countriesBMJ 2005; 331 doi: http://dx.doi.org/10.1136/bmj.38645.660093.68 (Published 08 December 2005) Cite this as: BMJ 2005;331:1364
- Rob Baltussen, health economist ()1,
- Katherine Floyd, health economist2,
- Christopher Dye, coordinator3
- 1Institute for Medical Technology Assessment (iMTA), Erasmus Medical Centre, PO Box 1738, 3000 DR Rotterdam, Netherlands
- 2Stop TB Department, World Health Organization, Geneva, Switzerland
- 3Stop TB Department, World Health Organization, Geneva, Switzerland
- Correspondence to: R Baltussen
- Accepted 12 October 2005
Objective To assess the costs and health effects of tuberculosis control interventions in Africa and South East Asia in the context of the millennium development goals.
Design Cost effectiveness analysis based on an epidemiological model.
Setting Analyses undertaken for two regions classified by WHO according to their epidemiological grouping—Afr-E, countries in sub-Saharan Africa with very high adult and high child mortality, and Sear-D, countries in South East Asia with high adult and high child mortality.
Data sources Published studies, costing databases, expert opinion.
Main outcome measures Costs per disability adjusted life year (DALY) averted in 2000 international dollars ($Int).
Results Treatment of new cases of smear-positive tuberculosis in DOTS programmes cost $Int6-8 per DALY averted in Afr-E and $Int7 per DALY averted in Sear-D at coverage levels of 50-95%. In Afr-E, adding treatment of smear-negative and extra-pulmonary cases at a coverage level of 95% cost $Int95 per DALY averted; the addition of DOTS-Plus treatment for multidrug resistant cases cost $Int123. In Sear-D, these costs were $Int52 and $Int226, respectively. The full combination of interventions could reduce prevalence and mortality by over 50% in Sear-D between 1990 and 2010, and by almost 50% between 2000 and 2010 in Afr-E.
Conclusions DOTS treatment of new smear-positive cases is the first priority in tuberculosis control, including in countries with high HIV prevalence. DOTS treatment of smear-negative and extra-pulmonary cases and DOTS-Plus treatment of multidrug resistant cases are also highly cost effective. To achieve the millennium development goal for tuberculosis control, substantial extra investment is needed to increase case finding and implement interventions on a wider scale.
This article is part of a series examining the cost effectiveness of strategies to achieve the millennium development goals for health
Every year almost nine million people contract tuberculosis, and almost two million die from the disease.1 In many parts of the world it is reappearing in almost epidemic proportions, mainly because of coinfection with HIV/AIDS and increasing multidrug resistance.1 2 In developing countries, tuberculosis is second only to HIV/AIDS as the most common cause of adult death and is one of the top public health problems almost everywhere. For this reason, the United Nations millennium development goals include targets and indicators related to tuberculosis control, which have been adopted and extended by the international Stop TB Partnership. The targets include reversing tuberculosis incidence by 2015, halving tuberculosis prevalence and mortality by 2015 (compared with 1990), and diagnosing 70% of new smear-positive cases and curing 85% of these cases by 2015 (see box 1).3
Box 1: Goals, targets, and indicators for tuberculosis control
Millennium development goal 6: Combat HIV/AIDS, malaria, and other diseases
Target 8: Have halted by 2015 and begun to reverse the incidence of malaria and other major diseases
Indicator 23: Prevalence and death rates associated with tuberculosis
Indicator 24: Proportion of tuberculosis cases detected and cured under DOTS (the internationally recommended tuberculosis control strategy)
Stop TB Partnership targets
By 2005: At least 70% of people with infectious tuberculosis will be diagnosed (that is, under the DOTS strategy), and at least 85% cured
By 2015: The global burden of tuberculosis (prevalence and death rates) will be reduced by half compared with 1990 levels. This means reducing prevalence to ≤ 150/100 000 and deaths to ≤ 15/100 000/year by 2015 (including cases coinfected with HIV). The number of people dying from tuberculosis in 2015 should be < 1 million, including those coinfected with HIV
By 2050: The global incidence of tuberculosis disease will be < 1 case/million population/year (the criterion for tuberculosis “elimination” adopted in the United States)
For many countries, the targets will not be achieved at current rates of progress.4 This is despite the existence of effective interventions to diagnose and cure tuberculosis, and thus to decrease transmission. A key question, therefore, is whether the correct mix of interventions is currently being used, and what strategies should be scaled up if current international efforts to raise extra funds for health care are successful. Cost and cost effectiveness analyses can provide valuable inputs to these decisions by identifying the most efficient ways of delivering diagnosis and treatment services at different levels of resource availability.
Box 2:Definitions of types of tuberculosis and recommended control strategies
Types of tuberculosis
Pulmonary tuberculosis—Commonest form of tuberculosis (about 70-90% of all cases), which affects the lungs
Smear-positive pulmonary tuberculosis—The most infectious cases can be diagnosed bacteriologically by means of sputum smear microscopy (about 60% of all pulmonary cases)
Smear-negative pulmonary tuberculosis—Diagnosed on the basis of clinical signs and symptoms, a chest x ray, and failure to respond to a standard course of antibiotics
Extra-pulmonary tuberculosis—Tuberculosis that occurs outside the lungs
Drug susceptible tuberculosis—Tuberculosis bacteria susceptible to standard antituberculosis drugs
Multidrug resistant tuberculosis—Resistance to at least rifampicin and isoniazid, the two most effective first line antituberculosis drugs
Recommended tuberculosis control strategies
DOTS—Internationally recommended tuberculosis control strategy, developed in the mid-1990s and has been implemented in 182 countries. It has five essential components: political commitment, diagnosis by sputum smear microscopy, short course treatment with standard first line drug regimens, a reliable drug supply, and a recording and reporting system that allows assessment of individual patient outcomes and overall programme performance
DOTS-Plus—Strategy for management of cases with multidrug resistant tuberculosis, developed by the World Health Organization and partner agencies from 1999. It is based on the same principles as the DOTS strategy but includes use of sputum cultures and drug susceptibility tests for diagnosis, and use of second line as well as first line drugs
Stop TB strategy—Developed by WHO during 2005, designed to guide tuberculosis control efforts during 2006 to 2015. It builds on the DOTS and DOTS-Plus strategies and has six major components: pursuing expansion and enhancement of DOTS; addressing tuberculosis and HIV coinfection, multidrug resistant tuberculosis, and other special challenges; helping to strengthen healthcare systems; engaging all healthcare providers; empowering patients and communities; and promoting research. The strategy underpins the second “Global Plan to Stop TB,” which also covers the period 2006-15
The main interventions recommended to control tuberculosis are short course treatment with first line drugs for drug-susceptible tuberculosis (smear-positive pulmonary, smear-negative pulmonary, and extra-pulmonary) within the framework of the DOTS strategy, and treatment of cases with multidrug resistant tuberculosis with longer and more complex drug regimens that include second line as well as first line drugs within the framework of the DOTS-Plus strategy (see box 2 for definitions).
To date, most economic studies of tuberculosis interventions in developing countries have evaluated short course treatment for drug susceptible, smear-positive pulmonary tuberculosis,5–7 since these cases are the most infectious and therefore of greatest concern from a public health perspective. Most of these studies are from Africa,8 although Asia has the highest burden of tuberculosis. Two studies in Africa have also reported the cost effectiveness of treating smear-negative cases.9 10 There is one published study, from Peru, of treatment for multidrug resistant tuberculosis with first line and second line drugs.11
Most of these studies did not assess the impact of interventions on transmission, and most used indicators of effectiveness that are specific to tuberculosis control. This prevents the cost effectiveness of tuberculosis control being compared with that of interventions for other diseases. Moreover, interventions have generally been considered individually and not in combination with complementary control strategies—for example, the cost effectiveness of providing simultaneous treatment for new smear-positive and new smear-negative and extra-pulmonary cases has not been evaluated even though in practice they are usually undertaken at the same time.
Five years after the adoption of the millennium declaration, an up to date assessment of the cost effectiveness of tuberculosis control strategies is needed. In this paper we address the question of what are the costs and effects of treatment of new smear-positive cases and of new smear-negative and extra-pulmonary cases in DOTS programmes, and of DOTS-Plus treatment for multidrug resistant cases that have not responded to first line treatments, both singly and in combination. Our analysis includes assessment of the impact of interventions on transmission, a generic measure of effectiveness, and covers Asia as well as Africa.
In common with the other papers in this series,12–17 we evaluated interventions for two particular regions classified by the World Health Organization according to their epidemiological characteristics: Afr-E, which includes countries in sub-Saharan Africa with very high adult and high child mortality, and Sear-D, which includes countries in South East Asia with high adult and high child mortality.18 Table 1 shows the existing levels of tuberculosis control globally and for these two regions.
Interventions run for the 10 years 2000-9, and we included all benefits accruing during the period 2000-100. We evaluated the three standard levels of geographical coverage—50%, 80%, and 95%—which in this case mean the percentage of eligible cases living in areas where treatment is available. We assessed costs from a societal perspective, and used a population model to translate disease-specific results into a generic measure of health effects. Details of the standardised analytical approach are available in Evans et al.18
Because the technologies available to tackle tuberculosis are well known, we restricted our analysis to four interventions:
Minimal DOTS—Treatment in DOTS programmes for new smear-positive cases only. We assume that the percentage of cases diagnosed and treated in areas covered by DOTS increases linearly from year 2000 levels to the WHO target of 70% in 2009 and that the cure rate is at the WHO target level of 85% from 2000 to 2009. In areas not covered by DOTS, we assume that no cases are treated. In all areas, no cases are treated from 2010 onwards.
Full DOTS—As for minimal DOTS plus treatment of smear-negative and extra-pulmonary cases in DOTS programmes. We assume that the percentage of cases diagnosed and cured is the same as for smear-positive cases. We did not consider the treatment of smear-negative and extra-pulmonary cases separately because in practice it would not be introduced in the absence of treatment for the more infectious smear-positive cases.
Minimal DOTS plus resistant cases—As for minimal DOTS plus treatment of multidrug resistant cases in DOTS-Plus programmes with an 18 month regimen that includes first and second line drugs. We assume that patients are tested for multidrug resistance after failing treatment with the short course of first line drugs. Treatment of multidrug resistant tuberculosis must be combined with the basic strategy because multidrug resistance does not exist without initial treatment. We assume the cure rate to vary from 48% (baseline analysis) to 70% (sensitivity analysis).11
Full combination—As for full DOTS plus DOTS-Plus treatment for multidrug resistant tuberculosis as defined above.
The maximum scale at which we considered each intervention is much greater than the level of tuberculosis control efforts in 2003 (table 1).
Estimating health effects
We estimated health effects in three steps. Firstly, we calibrated a published tuberculosis-HIV model19 20 to produce tuberculosis incidence, prevalence, and mortality for each region that matched those observed between 1950 and 2000. We applied parameters similar to those that were specified in the original paper.19 Our regional population estimates, including background mortality, were based on WHO estimates.21 Regional estimates of HIV/AIDS incidence, prevalence, and mortality for the period were based on internal projections by UNAIDS (the Joint UN Programme on HIV/AIDS). Full details of the model and parameters are available in the appendix on bmj.com.
Secondly, we used the calibrated tuberculosis-HIV model to project incidence, prevalence, and mortality for the period 2000-100 for the base case of no interventions, and then for each of the intervention scenarios.
Thirdly, we used the population model PopMod22 to combine the projected incidence, prevalence, and mortality data with the standard health state valuations23 to estimate the population impact of the different interventions in terms of healthy years lived.18 We ran the model for the length of time necessary for all people affected by the interventions to have died. The difference between the healthy years lived in each intervention scenario and the no-intervention scenario is the health gain of the intervention, or the number of disability adjusted life years (DALYs) averted.
We based our estimates of the resources required—diagnostic tests, drug use, health centre visits for supervision and monitoring, and hospitalisation—for each intervention on WHO treatment protocols and expert opinion of actual practice. We based drug costs on the latest WHO negotiated prices, with a mark-up for international and local transportation costs.24 25 Unit costs of health centre visits and hospital inpatient days were taken from Adam et al,26 while those for laboratory tests and x rays were based on the best available international cost information included in WHO's costing database. We combined unit costs with patterns of resource use to estimate the cost per patient treated. We then calculated total patient costs as the cost per patient treated multiplied by the number of patients treated (calculated as the annual incidence of disease from the model multiplied by the relevant coverage level and then by the percentage of cases diagnosed and treated in the areas covered).
We estimated the costs of running the programmes (that is, costs above the individual patient level, such as managerial staff) using a standardised approach.18 All costs are reported in international dollars ($Int) for the year 2000, and the conversion from $Int to US$ is explained elsewhere.18 Details of all cost calculations are found in the appendix on bmj.com.
The tuberculosis model replicated the strong increase in the incidence of infectious disease in Afr-E from around 1990, with an annual growth rate of about 10% between 1990 and 2000. In Sear-D, the tuberculosis model estimates an annual decline in incidence of 1% in the same period.
Tables 2 and 3 show the health effects, costs, and cost effectiveness of the different interventions in Afr-E and Sear-D. When only smear-positive cases are treated in DOTS programmes and the geographical coverage level is 95%, an average of 0.62 million people are treated in Afr-E and 1.38 million in Sear-D each year. The annual cost averages $Int366m in Afr-E and $Int536m in Sear-D. The total number of DALYs averted per year averages 44.8 million in Afr-E and 76.6 million in Sear-D. Adding treatment of smear-negative and extra-pulmonary cases or of multidrug resistance cases increases costs considerably but increases the DALYs averted only slightly. Increasing the coverage level from 50% to 95% roughly doubles both costs and effects for each of the four interventions considered.
In both regions, treating only smear-positive cases is the most cost effective intervention, with an average cost per DALY averted of ≤ $Int8 at all coverage levels. The next most cost effective intervention in both regions is treatment for both smear-positive and smear-negative and extra-pulmonary cases at a coverage level of 95%, at a cost per DALY averted of $Int95 in Afr-E and $Int52 in Sear-D. This is followed by implementing the full combination of interventions, including treatment for multidrug resistant tuberculosis, at a cost per DALY averted of $Int123 in Afr-E and $Int226 in Sear-D.
The figure shows the order in which interventions should be introduced according to their cost effectiveness for Afr-E (that is, the expansion path). Treating only smear-positive cases at a coverage level of 50% would be introduced first. With more resources, coverage would be expanded to 80% and then to 95%. With yet more resources, treatment of smear-negative and extra-pulmonary cases would be introduced, followed by the addition of treatment for multidrug resistant cases. The expansion path is similar in Sear-D.
In Sear-D, our model suggests that implementing the full combination of interventions could reduce tuberculosis prevalence and mortality by 71% and 64% respectively between 1990 and 2010. In Afr-E prevalence and mortality increase substantially between 1990 and 2000, because of the HIV epidemic, but could fall by 50% and 40% respectively between 2000 and 2010.
Sensitivity and uncertainty analyses
We undertook various sensitivity analyses, and table 4 shows the results for Afr-E. Changes to the parameters that were most uncertain, such as cure rate of standardised second line treatment of multidrug resistant cases, had little impact on our cost per DALY averted results. Similar results applied for Sear-D (data not shown).
Since the early 1990s, short course drug treatment for new smear-positive cases of tuberculosis has been promoted as one of the most cost effective healthcare interventions available, based on a study in three low income African countries in the late 1980s that reported a cost per DALY averted of US$1-3.5 Our updated analysis, covering countries with some of the highest rates of tuberculosis infection in sub-Saharan Africa and South East Asia, supports this result, with the cost per DALY averted at around $Int8 (< US$2) in both regions. The addition of the other interventions that we considered—treatment of smear-negative and extra-pulmonary cases in DOTS programmes and treatment of multidrug resistant cases in DOTS-Plus programmes—is also highly cost effective compared with commonly used benchmarks.17
Limitations of study
Our study has several limitations. Some of these are related to the general methodological approach to cost effectiveness analysis, and are discussed in more detail elsewhere in this series.18 Others are more specific to tuberculosis control.
In the absence of better data, we assumed that key model parameters such as tuberculosis transmission rates are the same across regions. Studies of the transmissibility of multidrug resistant tuberculosis have produced variable results, and our assumption that multidrug resistant tuberculosis and drug susceptible tuberculosis are equally transmissible contrasts with the more conservative range of assumptions considered in an earlier study.11
Evidence about the costs of increasing the percentage of tuberculosis cases that are treated in DOTS programmes remains limited, and, despite building in extra costs to allow for this, we may have underestimated them. The only published cost data for DOTS-Plus programmes are from Peru.
Our study results may not be directly generalisable to other settings because of differences in regional epidemiological and economic profiles. However, the results of studies for other regions that used similar methods show similar results.27
The strengths of our study include the use of a tuberculosis model that has been published and widely applied,20 consideration of combinations of interventions, inclusion of transmission in the analysis, use of a generic measure of effectiveness, and testing of important assumptions through sensitivity analyses.
Implications of results
Our results have three major policy implications. Firstly, they reinforce the principle that treatment of smear-positive cases in DOTS programmes must be the basis of any tuberculosis control strategy, as has become standard practice in almost all control programmes.
Secondly, they show that there is a strong economic case for treating smear-negative and extra-pulmonary cases in DOTS programmes and for treating multidrug resistant cases in DOTS-Plus programmes, as set out in WHO's new “Stop TB” strategy and the second global plan for tuberculosis control (see box 2).
Finally, our study shows that substantial scaling up of all three interventions is needed in the next 10 years if the millennium development goal and related targets for tuberculosis control are to be reached. In particular, the case detection rate must be improved so that many more tuberculosis cases are diagnosed and successfully treated, in line with existing targets. Improving the case detection rate will mean ensuring that people who currently have access to treatment facilities are covered and that coverage is expanded to people who do not currently have access. Such scaling up would bring the millennium development goal and related Stop TB Partnership targets within reach in South East Asia and achieve major progress towards these targets in Africa.
What is already known on this topic
Studies have shown that DOTS treatment of new cases of smear-positive tuberculosis to be a cost effective intervention in Africa, but data for other regions of the world or for treating smear-negative and extra-pulmonary cases and multidrug resistant tuberculosis are scarce
Most studies have not considered the impact of interventions on transmission or interactions among interventions and have used measures of effectiveness that do not allow comparisons with other health interventions
What this study adds
This comprehensive and standardised analysis of different interventions in Africa and South East Asia accounts for both transmission and interactions among interventions
Treatment of smear-positive, smear-negative, and extra-pulmonary cases in DOTS programmes and treatment of multidrug resistant cases in DOTS-Plus programmes are cost effective in both regions
These results provide a strong case for substantial investment to improve case finding and to implement these interventions on a much wider scale
Further details of the methods used appear on bmj.com
Contributors All authors contributed to the conception and design of the study, interpretation of data, and drafting of the manuscript. RB performed the technical analysis. All authors approved the submitted version of the manuscript. RB is guarantor for the manuscript.
Funding RB received funding from WHO and conducted the research in collaboration with WHO staff. For WHO staff, no external funding was received.
Competing interests None declared.