Objective To estimate the proportion of antiretroviral therapy (ART) eligible adults (15–49 years) with tuberculosis potentially identifiable through tuberculosis services using a CD4 count below 350 cells/mm3 as cut-off value for ART initiation.
Methods Using TB notification rate data, HIV seroprevalence data, and estimates of the size of the adult population (15–49 years) in 18 sub-Saharan African countries with an HIV seroprevalence of > 5%, calculations of the number of ART eligible adults with tuberculosis presenting to tuberculosis services were made. Assumptions were made on the tuberculosis notification rates in the age-group 15–49 years, the HIV-infected population with a CD4 count below 350 cells/mm3 and the relative risk of developing tuberculosis, and average duration from HIV infection to death. The probability of having a CD4+ count below 350 cells/mm3 given a diagnosis of tuberculosis was estimated using Bayes’ theorem, and estimates of the number of patients with a CD4 count below 350 cells/mm3 identifiable through tuberculosis were made. The number needed to screen to identify one ART eligible patient through tuberculosis services was estimated for each country.
Results ART eligible adults with tuberculosis potentially identifiable through tuberculosis services in the 18 countries ranged from 2% to 18% of the total HIV-infected adult population with a CD4+ count below 350 cells/mm3 and would average 10% of all such HIV patients. The number needed to screen to identify ART eligible patients through tuberculosis services ranged from 1.4 to 4.2, against 8.6 to 65.4 if adults aged 15–49 are randomly screened for low CD4 counts.
Conclusion Tuberculosis services are an important entry point for identifying ART eligible patients. Given that dually infected patients identified through tuberculosis services contributed to 10% of the HIV-infected adult population with a CD4 cell count below 350 cells/mm3 in the 18 sub-Saharan African countries, major efforts are required beyond the tuberculosis services in detecting patients that should benefit from ART. However, the low number needed to screen gives opportunity to use tuberculosis services in AIDS control and ART scaling-up programmes.
keywords HIV , tuberculosis , CD4+ count , ART , Bayes’ theorem , screening
Objectif Estimer la proportion d'adultes (15–49 ans) atteints de tuberculose nécessitant un traitement antirétroviral (ART) potentiellement identifiables grâce aux services de lutte contre la tuberculose en utilisant un nombre de CD4 inférieur à 350/mm3 comme valeur seuil pour l'initiation du traitement.
Méthodes Le calcul du nombre d'adultes atteints de tuberculose nécessitant un ART se présentant dans les services de lutte contre la tuberculose a été effectué en utilisant les donnés de déclaration des cas de tuberculose, les données de séroprévalence pour le VIH et l'estimation de la taille de la population adultes (15–49 ans) de 18 pays d'Afrique Sub-Saharienne ayant un taux de séroprévalence pour le VIH de plus de 5%. Les hypothèses de calculs étaient basées sur le taux de déclaration des cas de tuberculose dans le groupe d'age 15–49 ans, sur la population infectée par le VIH avec un nombre de CD4 inférieur à 350/mm3, sur le risque relatif de développer une tuberculose et sur la durée moyenne entre le diagnostic d'infection à VIH et le décès. La probabilité d'avoir un nombre de CD4 inférieur a 350/mm3 donnant un diagnostic de tuberculose était estimé par le théorème de Bayes et des estimations du nombre de patients ayant des CD4 inférieur à 350/mm3 lors d'un diagnostic de tuberculose ont été réalisés. Le nombre de sujets nécessaires àévaluer pour identifier 1 patient éligible pour le ART grâce aux services de lutte contre la tuberculose a été estimé pour chaque pays.
Résultats Les adultes atteints de tuberculose nécessitant un ART potentiellement identifiable grâce aux services de lutte contre la tuberculose dans les 18 pays représentaient 2 à 18% de la population totale d'adultes infectés par le VIH ayant un nombre de CD4 inférieur à 350/mm3, soit une moyenne de 10% de l'ensemble des patients VIH positifs. Le nombre de sujets nécessaires àévaluer pour identifier des patients relevant d'un ART grâce aux services de lutte contre la tuberculose allait de 1,4 à 4,2 patients contre 8,6 à 65,4 si les adultes âgés de 15 à 49 ans étaient dépistés au hasard à la recherche d'un taux de CD4 bas.
Conclusion Les services de lutte contre la tuberculose sont un important lieu d'identification des patients relevants d'un ART. Étant donné que les patients co-infectés décelés grâce aux services de lutte contre la tuberculose représente 10% de la population adulte infectée par le VIH avec un taux de CD4 inférieur à 350/mm3 dans 18 pays d'Afrique Sub-Saharienne, des efforts considérables sont requis au delà des services de lutte contre la tuberculose pour dépister les patients relevant d'un ART. Cependant, le faible nombre de patients àévaluer donne l'opportunité d'utiliser les services de lutte contre la tuberculose pour la lutte contre le SIDA et les programme d'accès au ART.
mots clefs VIH , Tuberculose , nombre de CD4> , ART , théorème de Bayes , dépistage
Objetivo Estimar la proporción de adultos (15–49 años) elegibles para terapia antirretroviral (TARGA) con tuberculosis potencialmente identificable, a través de servicios de tuberculosis, utilizando un conteo de CD4 inferior a 350 cel/mm3 como valor limite para el inicio de la TARGA.
Métodos Utilizando los datos de tasa de notificación de TB, los datos de seroprevalencia de VIH, y estimaciones del tamaño de la población adulta (15–49 años) en 18 países del África Subsahariana con una seroprevalencia de VIH superior al 5%, se realizaron los cálculos del número de adultos elegibles para TARGA con tuberculosis, que se presentaron a los servicios TB. En las tasas de notificación de TB se hicieron suposiciones en el grupo de 15–49 años, sobre la población con infección por VIH con un conteo de CD4 inferior a 350 cel/mm3 y riesgo relativo de desarrollar tuberculosis, y un promedio de duración entre la infección por VIH y la muerte. La probabilidad de tener un conteo de CD4+ por debajo de 350 cel/mm3 habiendo recibido un diagnostico de tuberculosis fue estimado utilizando el Teorema de Bayes, y se realizó una estimación del número de pacientes con conteo de CD4 inferior a 350 cel/mm3 identificables a través de tuberculosis. El número necesario de cribas para identificar a un paciente elegible para TARGA a través de los servicios de tuberculosis fue estimado para cada país.
Resultados El número de adultos elegibles para TARGA con tuberculosis potencialmente identificable a través de los servicios de tuberculosis en los 18 países varió de un 2 a un 18% del total de la población adulta infectado por VIH con un conteo de CD4 inferior a 350 cel/mm3, y constituiría un promedio de un 10% de todos esos pacientes con infección por VIH. El número de cribados necesarios para poder identificar a los pacientes elegibles para TARGA a través de los servicios de tuberculosis varió entre 1.4 a 4.2, contra 8.6 a 65.4 si los adultos comprendidos entre 15–49 años son aleatoriamente cribados buscando bajos conteos de CD4.
Conclusiones Los servicios de tuberculosis son un importante punto de entrada para identificar pacientes elegibles para TARGA. Dado que los pacientes coinfectados identificados a través de los servicios de tuberculosis contribuyen a un 10% de la población adulta con infección por VIH con conteo de CD4 inferior a 350 cel/mm3 en los 18 países del África Subsahariana, se requieren mayores esfuerzos más allá de los servicios de tuberculosis para detectar pacientes que deberían beneficiarse de la TARGA. Sin embargo, el bajo número que es necesario cribar da la oportunidad de utilizar los servicios de tuberculosis en el control del SIDA y de los crecientes programas de TARGA.
The widespread use of antiretroviral therapy (ART) in high-income countries for treatment of people infected with the human immunodeficiency virus (HIV) has resulted in a dramatic decline in morbidity and improvement in survival. Although ART does not eradicate infection, it suppresses HIV replication which leads to partial reconstitution of the immune system and containment of opportunistic infections. ART is not widely available in sub-Sahara Africa, and of the estimated 4.7 million people who currently need ART in this region, only 11% are receiving it (World Health Organization 2005a). However, there is increasing consensus and commitment to redress this inequality and scale up access of antiretroviral drugs to HIV-infected people in resource-constrained environments. Major efforts such as the Global Fund to Fight AIDS, TB, and malaria (GFATM) and the United States President's Emergency Plan for AIDS Relief (PEPFAR) have already committed financial and other resources to amplify global efforts to combat the HIV/AIDS scourge (The President's Emergency Plan for AIDS Relief 2004). In 2003, the World Health Organization (WHO) and its partners launched the three by five initiative, with a global target of treating 3 million people in developing countries with antiretroviral therapy by the end of 2005 (World Health Organization 2003a).
In many developing countries, HIV infection is driving the tuberculosis epidemic and in some settings over 60% of new tuberculosis cases are HIV-infected (Raviglione et al. 1997; Corbett et al. 2003). Progressive immunodeficiency predisposes patients to opportunistic infections, including tuberculosis. Therefore, in settings of a high burden of interacting HIV and TB infection epidemics, existing tuberculosis services could be utilized as entry points for delivering ART (Harries et al. 2001; Williams & Dye 2003; Abdool-Karim et al. 2004). Also the experience and infrastructure of tuberculosis services in providing directly observed therapy (DOT) to tuberculosis patients could be used for delivery of DOT-ART in resource constrained settings, a strategy that has been used with success in Haiti (Farmer et al. 2001). ART is recommended for all patients with tuberculosis who have CD4 counts below 200 cells/mm3, and should be considered for patients with CD4 counts below 350 cells/mm3 (World Health Organization 2003b). In this study, we estimate the proportion of ART eligible adults aged 15–49 years with forms of tuberculosis that might be identified through tuberculosis services, mostly, but not exclusively, pulmonary smear positive tuberculosis. We assume that a CD4 count below 350 cells/mm3 is used either as the cut-off point for ART initiation, or as a cut-off point for regular follow-up until patients become ART eligible (e.g. when CD4 counts drop below 200 cells/mm3). TB services might be specialized TB clinics of a vertical programme or clinics integrated into general health facilities. We have also estimated the number of TB patients that need to be screened to identify one case with a CD4 count below 350 cells/mm3, and thus eligible for ART in this model.
We have used Bayes’ theorem to estimate the number of tuberculosis patients aged 15–49 years with CD4 counts of below 350 cells/mm3 that might be identified by tuberculosis services in 18 sub-Saharan Africa countries, and hence eligible for ART. Countries included in this analysis have an estimated adult HIV seroprevalence > 5% as of the end of 2003. HIV seroprevalence data were retrieved from the Joint United Nations Programme on HIV/AIDS (UNAIDS) report (UNAIDS 2004). Data on the size of the adult population (15–49 years) were obtained from the UNAIDS/WHO Global HIV/AIDS online database (World Health Organization 2004). Data on age-stratified notification rates of all diagnosed tuberculosis cases are not readily available for all countries. The official WHO tuberculosis Figures only present age-stratified data on notifications for smear positive tuberculosis. In a South African study, all age tuberculosis prevalence was estimated at 1358 per 100 000 population with a rate of 2120 and 2202 per 100 000 in the age-group 15–44 and 45–59, respectively, which is just over 1.5 times the all age rate (Vella 2003). We have therefore assumed that notification rates of all tuberculosis cases in the age-group 15–49 for all countries in this analysis, whose demographic structures are roughly similar to that of South Africa, are 1.5 times higher than the overall tuberculosis notification rate, which were taken from the most current World Health Organization global tuberculosis data (World Health Organization 2005b).
Two measures are of interest in assessing the potential of TB services in identifying ART eligible patients, viz. (a) the probability that an individual with TB is ART eligible, the ‘yield’ per TB patients, and (b) the total number of ART eligible patients that can be identified through TB services. The former measure is estimated as follows using Bayes’ theorem.
The probability that a patient's CD4 count is below 350 cells/mm3 (denoted by cd < 350) given a new diagnosis of tuberculosis (tb) is:
where λ(tb|cd < 350) denotes the notification rate (which in good programmes should approximate to the incidence rate) of tuberculosis among individuals with low CD4 counts, and P(cd ≥ 350) is the probability (the prevalence) that an individual in the general population does not have a CD4 count below 350 cells/mm3. Dividing both numerator and denominator of the above equation by λ(tb|cd ≥ 350), which is the TB notification rate among individuals with a CD4 count of at least 350 cells/mm3 gives:
where RR is the relative risk (incidence rate ratio) of developing diagnosed tuberculosis among HIV-infected patients with a CD4+ cell count of below 350 cells/mm3, relative to all individuals with a CD4 count above 350, i.e. either HIV negative, or HIV positive. Note that in Equation (3) the assumption is made that P(cd ≥ 350), which is the probability (the prevalence) that an individual in the general population does not have a CD4 count below 350 cells/mm3 is approximately 1.
P(cd < 350), the probability that an individual has a CD4 count below 350 cells/mm3, can be estimated from adult HIV seroprevalence statistics and from assumptions on the proportion of the HIV infected population with a CD4 cell count below 350 cells/mm3.
The second measure (the total number of ART eligible patients that can be identified through TB services) was calculated as follows: λ(tb|cd<350) is the annual notification rate of tuberculosis among individuals with a CD4 count below 350 cells/mm3 and is by definition
Notification rates are not reported separately for CD4 counts below 350 cells/mm3 or ≥ 350 cells/mm3. However, we can derive these from λ(tb), the notification rate in adults as follows:
Since P(cd<350) = 1−P(cd ≥ 350), Equation (6) can be rearranged algebraically to give λ(tb|cd ≥ 350), which is given by
RR, λ(tb), and P(cd < 350) are parameters that are known. Hence notification rates for both CD4 counts below 350 cells/mm3 or ≥350 cells/mm3 can be calculated. If patients with a CD4 count below 350 cells/mm3 have a survival of t years, and the annual notification rate is λ(tb|cd < 350), then the probability that a patient with a CD4 count below 350 cells/mm3 is diagnosed with tuberculosis at any given time during their lifetime is given by:
Multiplying Equation (8) with the estimated total number of HIV-infected adults with CD4 count below 350 cells/mm3 gives the number of dually infected patients with CD4+ cell count below 350 cells/mm3 and therefore potentially detectable through TB services.
A key parameter in our calculations is the proportion of the HIV infected population with a CD4 count below 350 cells/mm3. Using linear interpolation between an assumed average of 110 cells/mm3 at the time of death and a value of 910 cells/mm3 at the time of infection yields an equilibrium (i.e. when the prevalence is stable) value of 30%. As the decline in CD4 counts may not decline on a linear scale but rather on a square root scale (Brettle et al. 1996), this estimate may be conservative. If we assume that 30% = 100×P(cd < 350)/P(HIV+) of the HIV-infected population have a CD4 count below 350 cells/mm3, and if we assume that the average duration from HIV infection to death is 8 years then t = 2.4 years. Although the HIV specific mortality in Africa is similar to that in Western countries, other causes of death are probably higher so that the overall life expectancy of 8 years would seem realistic (Morgan et al. 2002). We further assumed that RR is 20 for CD4 counts below 350 cells/mm3, corresponding to an average of approximately 7 for all HIV-infected individuals.
Number needed to screen (NNS)
Assuming that TB services are subject to all notified TB cases among the adult population aged 15–49 years for HIV screening, then the number of TB patients needed to screen to identify one patient with CD4 below 350 cells/mm3 , and thus eligible for ART in this model, is given by the reciprocal of P(cd < 350|tb). By contrast, using simple random screening of the population programmes would have to screen the reciprocal of P(cd < 350), which is the product of the HIV prevalence times the fraction of HIV infected who have CD4 counts <350. An excel sheet implementing all the formulas described in this model is available upon request from the corresponding author.
Estimates of the adult population aged 15–49 years with HIV infection and tuberculosis disease eligible for ART and potentially identifiable through tuberculosis services in 18 sub-Saharan countries with an HIV prevalence of >5% are shown in Table 1. The probability that a tuberculosis patient's CD4 count is below 350 cells/mm3 was estimated to range from 0.24 to 0.72, with the probability increasing with rising HIV seroprevalence (Table 1).
Table 1. Estimate of HIV-infected adult population (15–49 years) with CD4 count below 350 cells/mm3 and tuberculosis in 18 sub-Saharan African countries with an HIV prevalence in the general population of >5%
Countries ranked in order of increasing HIV seroprevalence
Est. TB notification rate all cases (15–49 years)/ 100 000 population)
Est. number of HIV-infected adults with CD4 cell count below 350
No. HIV-infected adults with TB and CD4 cell count below 350 detected annually
Probability that an adult TB patient with HIV infection has a CD4 cell count below 350
ART eligible identified through TB services as %age of total patients with CD4 cell count below 350
*Tuberculosis high burden country.
†Figures based on published data; all other variables in the columns are derived from model output.
Central African Republic
1 573 994
Using the methodology as described in this model, we estimated that out of a total of 5 398 811 adults aged 15–49 years with a CD4 cell count below 350 cells/mm3 in the 18 countries, 540 195 tuberculosis patients with HIV infection and CD4 count below 350 cells/mm3 are potentially identifiable by tuberculosis services. Thus TB services would contribute to identifying 10 % of all HIV-infected patients with a CD4+ count below 350 cells/mm3 , and if this CD4 level is used as the threshold for initiation of ART then these patients would qualify for antiretroviral treatment.
Performance of TB services in identifying ART eligible patients would vary considerably between countries (Table 1), with the highest yield, 18%, estimated for Namibia. Six countries (Botswana, Gabon, Lesotho, Namibia, South Africa and Zambia) have a tuberculosis case detection rate of at least 70%. The effect of improving tuberculosis case detection rate to 70% on the yield of ART eligible patients was examined for all countries with a case detection rate below 70%. Increasing case detection rate would result in a greater yield of ART eligible patients identified through tuberculosis services, especially for countries with a low case detection rate (Table 2). In Nigeria, increasing the case detection rate would result in an increase of more than fivefold in ART eligible patients potentially found through tuberculosis services.
Table 2. Estimate of ART eligible patients identified through tuberculosis services in 12 sub-Saharan African country at 70% tuberculosis case detection rate
Countries ranked in order of increasing HIV seroprevalence
Estimate number of HIV-infected adults with TB and CD4 cell count below 350 detected annually
ART eligible identified through TB services as % age of total patients with CD4 cell count below 350
Central African Republic
The number needed to screen for the 18 countries is shown in Table 3. NNS TB patients in order to identify one ART eligible patient varies from 1.4 in Swaziland and Botswana to 4.2 in Rwanda. If the HIV seroprevalence is low, then TB services would have to screen more patients to identify one ART eligible patient in this model (Figure 1). Similarly, the number needed to screen if adults aged 15–49 years are randomly screened for counts below 350/mm3 is higher in HIV low prevalence countries compared with high prevalence countries, ranging from 65.4 in Rwanda to 8.6 in Swaziland (Figure 1).
Table 3. Number needed to screen (NNS) to identify one HIV-infected patient with a CD4 cell count below 350 cells/mm3 in 18 sub-Saharan African countries
Countries ranked in order of increasing HIV seroprevalence
The tuberculosis service model as an entry point for ART is an attractive option for developing countries. In this study, the number of ART eligible adult patients potentially detectable through tuberculosis services in18 sub-Saharan Africa countries with an estimated HIV seroprevalence of >5% is quantified. Overall, tuberculosis services operating at current detection rates have the potential to identify 10% of the HIV-infected tuberculosis patients with a CD4+ count below 350 cells/mm3. Furthermore, improving tuberculosis case detection in countries with a performance below WHO's global target of 70% (World Health Organization 2003c) results in an increase in the number of ART eligible patients identified. For example, increasing tuberculosis case detection in Nigeria from the current 12% to 70% would result in a more than fivefold increase in the number of ART eligible patients detected through tuberculosis services.
Of the 18 countries in this study, six are considered as tuberculosis high burden countries. The prevalence of dually infected patients with HIV and tuberculosis might be expected to be higher in these countries and that the tuberculosis services would potentially reach out to more ART eligible patients. However, in half of the six countries, tuberculosis services would potentially reach <10% of the ART eligible patients. Clearly, if tuberculosis services are to optimize their potential as entry points for identification of ART eligible patients most of the high burden countries will have to improve on tuberculosis case detection, which for the majority of countries is below WHO's global target of achieving case detection of 70%.
In spite of the modest yield in terms of identifying ART eligible patients, tuberculosis services are an important entry point for ART as the number needed to screen (NNS) to identify one ART eligible patient is low compared to the NNS if adults between 15 and 49 years are randomly screened for low CD4+ counts. Therefore, the low NNS gives opportunity to use tuberculosis services in AIDS control and ART scaling up programmes.
Given this modest contribution of tuberculosis services in the identification of ART eligible patients, it might be better if tuberculosis services with a limited ability to identify ART eligible patients gave more priority to strengthening tuberculosis control activities. They could to identify and treat more tuberculosis patients before expanding their activities to include delivering antiretroviral therapy and HIV care themselves, which is a logistically more demanding task (Harries et al. 2002). Concurrent treatment of tuberculosis and HIV is often challenging as the potential of interactions between antiretroviral and antituberculosis drugs in patients undergoing simultaneous HIV and TB treatment may result in ineffectiveness of ARV drugs, infective tuberculosis treatment or an increased risk of drug toxicity (Harries et al. 2004). Nonetheless, both tuberculosis and HIV control services (and other entry points for HIV/AIDS patients to the health care system) should develop mechanisms to ensure that patients identified as eligible for ART do eventually access antiretroviral treatment.
Our model has some limitations, especially regarding the estimates used. Official World Health Organization Figures do not give all case age-stratified tuberculosis notification rates, and these are not easily available for many countries. Our approach of multiplying the official World Health Organization all case notification rates with 1.5 to estimate the rate in the 15–49 year age-group based on the observations in one country might not be representative of the epidemiology in other parts of Africa. Another source of uncertainty is the relative risk estimate of tuberculosis among patients with CD4 counts below 350 cells/mm3 used in this model. The relative risk of the association between HIV and tuberculosis has varied widely among African studies, but most studies suggest a relative risk of about seven (Glynn et al. 1997). Some studies have reported a far higher relative risk of >20 (Braun et al. 1991; Allen et al. 1992). As the risk of tuberculosis is closely correlated with the number of CD4 lymphocytes (Antonucci et al. 1995; Markowitz et al. 1997) we have elected to use a RR of 20, which represents the higher relative risk reported in African studies. If a lower RR of 10 is assumed, the overall yield for all eighteen countries in identifying adult tuberculosis patients eligible for ART is 7%, and 12% if a higher RR of 30 is assumed (results not presented). A further source of uncertainty is the assumption that 30% of the HIV-infected patient population has CD4 counts below 350 cells/mm3. Larger studies have reported higher estimates of >50% (Gupta et al. 2000; Dybul et al. 2002), but these are based on the CD4 cell counts at time of detection, which might be subject to bias.
Our findings have practical implications as countries facing a challenging HIV and tuberculosis epidemic move towards scaling up ART access. Quantifying the magnitude of ART eligible patients that might be seen by tuberculosis services will allow for proper planning and mobilization of resources in order to avoid overwhelming the tuberculosis services. Whilst tuberculosis services are an important entry point for identifying ART eligible patients, the World Health Organization's three by five initiative, and indeed any programme that aims at delivering ART to those in need, would require major efforts beyond TB services.