Background. In August 2005, the World Health Organization declared the tuberculosis (TB) epidemic in Africa to be a regional emergency. Current TB-control measures are failing, largely as a result of the human immunodeficiency virus (HIV) epidemic. Evaluation of additional control interventions requires detailed understanding of the epidemiological relationship between these diseases at the community level.
Methods. We examined age- and sex-specific trends in TB notifications and their association with the prevalence of HIV infection in a peri-urban township in South Africa during 1996–2004. Denominators for TB notifications were derived from population census data. The local TB-control program used the World Health Organization directly observed treatment, short-course (DOTS) strategy.
Results. TB notification rates increased 2.5-fold during the period, reaching a rate of 1468 cases per 100,000 persons in 2004 (P = .007, by test for trend); the estimated population prevalence of HIV infection increased from 6% to 22% during the same period. After stabilization of prevalence of HIV infection, the TB notification rate continued to increase steeply, indicating ongoing amplification of the TB epidemic. In 2004, at least 50% of children aged 0–9 years who developed TB were HIV infected. Annual TB notification rates among adolescents increased from 0 cases in 1996–1997 to 436 cases per 100,000 persons in 2003–2004, and these increases were predominantly among female. However, 20–39-year-old persons were affected most, with TB notification rates increasing from 706 to 2600 cases per 100,000 persons among subjects in their 30s. In contrast, TB rates among persons aged >50 years did not change.
Conclusions. HIV infection is driving the TB epidemic in this population, and use of the DOTS strategy alone is insufficient. TB notifications have reached unprecedented levels, and additional targeted, age-specific interventions for control of TB and HIV infection in such populations are needed.
Annual tuberculosis (TB) incidence rates have increased by 2- to 3-fold in many countries in sub-Saharan Africa since 1990 . The African continent, which contains just 11% of world's population, now accounts for 27% of the global burden of TB and 30% of TB-related deaths; an estimated 2.4 million new TB cases and 540,000 TB-related deaths occur in Africa annually . The World Health Organization (WHO) TB-control strategy, which is based on the directly observed treatment, short course (DOTS) strategy, has failed to contain the African TB epidemic, primarily because of the effects of the HIV epidemic in the region. In August 2005, the WHO Committee for Africa declared the TB epidemic to be an African regional emergency . Rapid escalation of the disease rate in the region is undermining progress towards the Millennium Development Goals for TB control and was deemed as mandating “urgent and extraordinary actions” .
An interaction between the HIV and TB epidemics has been well established by demonstration of increased TB case load in populations affected by HIV infection; an increased prevalence of HIV infection among persons with TB, compared with the general population; and progressively increasing risk of TB with advancing immunodeficiency [3–5]. The impact of HIV infection is accentuated in populations with preexisting high TB case rates . Moreover, the impact on TB notification rates may increase during the course of a rapidly maturing HIV epidemic as the proportion of patients with advanced immunodeficiency increases . Thus, the relationship between HIV and TB epidemics in a community is likely to evolve over time.
Existing data on the population-level interaction between the TB and HIV epidemics in sub-Saharan Africa are derived largely from national or regional data . Such data are often incomplete and do not define the relationship within individual communities bearing the brunt of disease. Insights into the dual epidemic have also been derived from studies of rural districts or unusual populations, such as gold miners [6–10]. However, neither population is likely to reflect the situation in urban and peri-urban communities.
We investigated epidemiological changes in TB notifications and the prevalence of HIV infection during 1996–2004 in a well-defined, peri-urban community in the Western Cape, South Africa. The study objective was to document the epidemiological association of these 2 epidemics in a community during a period in which the prevalence of HIV infection rapidly increased. Such information is needed to provide a scientific basis for any modification of present TB-control strategies.
Study population. We studied a peri-urban township near Cape Town, South Africa, which was established in 1992 and has grown to its current population size of ∼13,000 people. The township is home to an almost exclusively African population, and the majority of persons have low socioeconomic status. Unemployment rates exceed 50%, and housing predominately consists of closely aggregated, informal structures with high levels of overcrowding. The township is geographically clearly demarcated and constitutes a well-defined population for research studies and community health interventions. This research was approved by the Research Ethics Committee of the University of Cape Town.
Health services. The study community is served by a single government primary health care clinic with a dedicated TB service. All patients with TB in the community are treated at this facility. The program adhered to the South African National TB Control Programme guidelines throughout the study period and included the WHO DOTS strategy. DOTS coverage in this community was complete, and cure rates for smear-positive disease exceeded 80% in the district . All sputum testing was performed at the National Health Laboratory Services facilities in Cape Town.
Case definitions. Pulmonary sputum-positive TB was diagnosed on the basis at least 1 positive sputum culture of Mycobacterium tuberculosis or 2 sputum smears containing acid-fast bacilli in the context of a compatible clinical illness. Pulmonary sputum-negative TB was diagnosed on the basis of negative smears and cultures for M. tuberculosis in the context of clinically and radiologically compatible illness of at least 3 weeks' duration that did not respond to administration of simple antibiotic treatment but that responded to subsequent antituberculosis treatment. Diagnosis of extrapulmonary TB was based on a combination of clinical, radiological, and histopathological findings, as well as a response to antituberculosis treatment.
Data sources. Numbers of TB notifications, demographic characteristics, history of previous TB, sputum microbiologic test findings, and TB classification data were obtained from the community TB clinic register. Demographic data for the community were derived from the 1996 South African national census and from a household census performed in 2004 as part of ongoing health research. Prenatal HIV prevalence data were collected within the local Prevention of Mother to Child Transmission program. The prevalence of HIV infection among patients who had TB diagnosed was obtained from the results of voluntary counseling and testing routinely performed at the TB clinic in 2004 and from re-treatment cases tested in 2003–2004.
Estimates of the prevalence of HIV infection. Age-specific estimates of the prevalence of HIV infection were made using the Actuarial Society of South Africa 2002 demographic model for the African population. This model estimates the prevalence of HIV infection in South Africa using prenatal surveys and demographic data. The assumptions and results of the model have been described elsewhere [12, 13], and this model was validated for use in this community using a cross-sectional survey of 6% of the population in 2005.
Data analysis. Denominators for rate calculations were based on extrapolations of 1996 and 2004 population census data, assuming a linear trend in age- and sex-specific distributions over the study period. TB notification rates, TB re-treatment rates, and age- and sex-specific rates of TB were calculated using the data from the TB register and population census data. Ninety-five percent CIs for rates are based on Poisson-distributed standard errors. Trend analyses were conducted using Cuzick's nonparametric test for trend . Changes over time in the median age of subjects with TB diagnoses were compared using the Wilcoxon rank sum test. All tests were 2-sided at α = 0.05.
TB notification rates and prevalence of HIV infection. During 1996–2004, a total of 968 cases of TB were diagnosed. The annual TB notification rate increased 2.5-fold over the study period, reaching 1468 cases per 100,000 persons in 2004 (table 1). Rates of TB re-treatment also increased to >20% of notifications in 2003–2004.
The estimated prevalence of HIV infection in the community increased from 6% to 22% (table 1). In 2004, among persons aged >15 years, 98 (59%) of 166 persons who provided TB notifications were HIV seropositive, whereas 30% were HIV seronegative, and 11% had not been tested. On the basis of these data, the TB notification rate among HIV-infected individuals was calculated to be 4381 cases per 100,000 persons (95% CI, 3570–5313 cases per 100,000 persons), and the rate among HIV-uninfected individuals was 656 cases per 100,000 persons (95% CI, 486–866 cases per 100,000 persons). Even if persons who were not tested in 2004 were assumed to be HIV seronegative, a statistically significant increase in TB notification rates among HIV-uninfected individuals in this community could not be demonstrated (data not shown). Among patients who underwent re-treatment, cure of the previous episode of TB was microbiologically confirmed among 32 patients in 2003–2004. Of these 31 tested patients, 27 (87%) were found to be HIV seropositive.
Spectrum of TB. Of the 968 TB cases diagnosed, 450 (46.5%) were pulmonary sputum positive, 247 (25.5%) were pulmonary sputum negative, 140 (14.5%) involved extrapulmonary disease (predominantly lymph node), and 108 (11.2%) were primary disease among children. Sputum test results were missing for 23 cases of pulmonary disease, representing 2.3% of all TB notifications.
The spectrum of TB diagnoses changed over the course of the study period (figure 1). The notification rate of sputum-positive TB cases increased 1.9-fold, from 326 cases per 100,000 persons (95% CI, 181–471 cases per 100,000 persons) in 1996 to 617 cases per 100,000 persons (95% CI, 481–752 cases per 100,000 persons) in 2004 (P < .01, by test for trend). However, increases in the rates of extrapulmonary and sputum-negative pulmonary disease were proportionately greater, increasing 4.3-fold from 167 cases per 100,000 persons (95% CI, 63.8–271.0 cases per 100,000 persons) in 1996 to 719 cases per 100,000 persons (95% CI, 572–865 cases per 100,000 persons) in 2004 (P < .01, by test for trend).
TB among adults. Predominantly young people lived in this community, with the modal age lying in the 20–29-year-old age range; 88% of the population were <39 years of age. Comparison of census data from 1996 with data obtained in 2004 showed that the age-sex composition of this population had not changed significantly. Subsequent analysis of the age-specific distribution of TB notifications during 1996–2004 revealed a marked increase in the number of notifications among 20–39-year-old persons (figure 2A). This pattern was similar to increases in the estimated age-specific prevalence of HIV infection in this population during the same period (figure 2B). Thus, increases in the prevalence of HIV infection and the escalating burden of TB were both concentrated among young adults.
During 1996–2004, the median age of patients with TB (excluding those aged <10 years) decreased from 43 years to 32 years (P < .001). Additional comparison of age- and sex-specific notifications showed that the modal age group for patients with TB was 20–29 years for women and 30–39 years for men (figure 2C).
Although the distribution of age-specific TB notification rates observed during 1996–1997 conformed to that observed historically in this region , the distribution changed substantially over the study period. The highest TB notification rates in 2003–2004 occurred among adults aged <50 years (figure 3), and the greatest increase in notification rates was among persons aged 30–39 years, reaching 2600 cases per 100,000 persons in 2004. However, no significant increase in rate occurred among those aged ⩾50 years.
TB among adolescents and children. TB was notified for 44 adolescents aged 10–19 years (31 of whom were female, and 13 of whom were male). Although no cases occurred in 1996–1997, the TB notification rate among adolescents steadily increased from 1998 onwards, particularly among female subjects (figure 2A and 2C), to reach an average annual rate of 436 cases per 100,000 persons in 2003–2004 (figure 3). Of 11 TB cases notified in 2004, 5 involved persons aged <16 years who had not undergone HIV testing. Among the remaining 6 adolescents tested, 3 were HIV seropositive.
The increase in TB notification rates among children aged 0–9 years was not statistically significant. However, 18 cases were diagnosed in 2004, and of 13 children tested, 9 (50% of total cases) were found to be HIV infected.
Temporal association between HIV and TB epidemics. We next analyzed the rates of these 2 diseases among persons aged 20–39 years in whom both epidemics were concentrated (figure 4). Major increases in the prevalence of HIV infection occurred before 2000, and yet a high rate of increase in the TB notification rate continued thereafter. With data from 1996–1997 used as baseline values, we calculated the increases in the TB notification rate per unit increase in the prevalence of HIV infection. For each 1.0% increase in the prevalence of HIV infection, TB notification rates increased by 54.7 cases per 100,000 persons in 1998–1999 and by 80.6 cases per 100,000 persons in 2004. Thus, increases in the prevalence of HIV infection were associated with ongoing amplification of the TB epidemic several years later.
To our knowledge, this is the first community-based study from sub-Saharan Africa to characterize in detail the epidemiological associations over time between rapidly evolving epidemics of TB and HIV infection. The 2.5-fold increase in overall TB notification rates culminated in annual rates exceeding 1400 cases per 100,000 persons, ∼9-fold higher than the TB notification rate for sub-Saharan Africa and 280-fold greater than that for the United States . A striking temporal association between the evolving age-specific burdens of TB and HIV infection was demonstrated. Despite the existence of a well-run TB-control program, the DOTS strategy alone is failing to control TB in this population, and additional measures are urgently needed. These data provide an epidemiological basis with which to target interventions.
The strength of this study lies in fact that this clearly defined community was well characterized by total-population census surveys at the beginning and the end of the study period. This permitted calculation of disease rates, taking into account the changing demographic profile of the community. Moreover, the community is served by a single health center, where all patients with TB in this community are treated. Thus, data completeness is likely to be high. Trends in TB incidence in this community before the HIV epidemic are not known, and we cannot exclude other factors in addition to HIV infection that may have contributed to the increasing burden of TB. However, TB notification rates in the Western Cape were relatively stable in the 1980s, before the advent of the HIV epidemic in South Africa. Whether these data can be generalized is unknown, and the findings may apply only to similar communities with a high burden of TB and HIV infection.
Annual TB notification rates exceeding 1400 cases per 100,000 persons are almost unparalleled in the era of modern antituberculosis treatment, except for rates reported among the Inuit population of Northern Canada in the early 1960s . Three key factors probably underlie the explosive TB epidemic in this and similar communities: the high pre-existing TB notification rate, the rapid increase in the prevalence of HIV infection, and the population composition, in which 20–40-year-old persons who bear the brunt of these 2 diseases are disproportionately represented. The strongest factor pointing towards HIV infection as the principal cause underlying the increasing burden of TB was the temporal association between the evolving age-specific burdens of the 2 epidemics (figure 2A and 2B). Other factors corroborating this association were the disproportionate increases in the rates of extrapulmonary and pulmonary sputum-negative TB and the high prevalence of HIV infection among patients with TB.
The epidemiology of TB in this community has changed profoundly over a short period of time. Historically, TB notification rates in this community have been highest among persons >60 years of age . However, in 2003–2004, the rates were highest among persons aged 30–49 years (figure 3), indicating a major shift in the burden of disease to individuals in the economically productive age groups who also have the most dependents. A particularly concerning finding was the burden of TB among adolescents, especially girls, with TB rates among 10–19-year-old adolescents reaching 436 cases per 100,000 persons in 2004. We have since documented that the prevalence of HIV infection among adolescents aged 10–19 years in this community was 7% among male subjects and 11% among female subjects in 2004 (unpublished data), indicating that increases in the burden of TB in this age group were very likely to be, at least in part, attributable to HIV infection.
The risk of TB increases with advancing immunodeficiency [3, 4], so as the HIV epidemic in a community matures, the burden of HIV-associated TB may be expected to increase, even after the prevalence of HIV infection has stabilized. This is likely to explain why the major increases in the prevalence of HIV infection in this community occurred before 2000, yet TB notification rates continued to increase steeply thereafter (figure 4). These data suggest that ongoing amplification of the TB epidemic continues for several years as a result of earlier increases in the prevalence of HIV infection.
Although increases in the number of notifications were more pronounced for less infectious types of TB (i.e., pulmonary sputum-negative and extrapulmonary TB), the number of notifications of infectious, sputum-positive disease also increased substantially (figure 1). The secondary impact of HIV-associated burden of sputum-positive TB in the general community remains incompletely defined [7, 9, 17], but we were unable to demonstrate a statistically significant increase in rates among HIV-uninfected individuals. Moreover, TB notification rates among persons aged ⩾50 years, among whom the prevalence of HIV infection is low, did not increase (figure 3). HIV-associated sputum-positive disease clearly does not cause a proportionate increase in transmission within the general community. This is almost certainly related to the fact that HIV-associated TB becomes symptomatic more quickly and is diagnosed earlier than disease in HIV-uninfected individuals , diminishing the period for potential transmission.
Although the DOTS strategy is central to the WHO policy for TB control, the explosive increase in TB notifications in this community and others like it show that the DOTS strategy alone is not sufficient, and additional measures are required. We suggest that these strategies may be targeted in various ways on an age-specific basis among children, adolescents, those aged 20–49 years, and those aged ⩾50 years. A substantial proportion of children who developed TB in 2004 were HIV infected. Thus, prevention of mother-to-child transmission services need to be strengthened, and voluntary counseling and testing for HIV in prenatal clinics and access to antiretroviral treatment must be provided for HIV-infected pregnant women. Secondly, administration of isoniazid prophylaxis to household contacts of adults with TB needs to be more rigorously implemented.
The escalating burden of TB and HIV infection among adolescents, among whom TB was previously rare, is of great concern. Schools in this and similar communities across Africa are extremely crowded and may provide a key site of TB transmission. Moreover, in South Africa, the prevalence of HIV infection is high among teachers , who may, therefore, represent an important source of TB transmission. Therefore, we suggest that TB and HIV diagnostic and prevention services for adolescents and teachers should be introduced into schools.
Annual TB notification rates among persons aged ⩾50 years did not increase but remained ∼1800 cases per 100,000 persons (figure 3). The major burden of disease in this age group was among men who were unlikely to be HIV infected (figure 2B and 2C). Because HIV-uninfected individuals with TB are more infectious , this age group may represent a key source of TB transmission and may be targeted by intensified case-finding.
Most of the increasing burden of TB occurred among 20–49-year-old persons. TB and HIV interventions need to be rigorously implemented in health care facilities, and provision of such services in workplaces may also facilitate access for this age group. In this community, HIV infection is frequently diagnosed once individuals have presented with TB; strategies to diagnose HIV infection before the development of TB are urgently needed. This mandates intensification of voluntary counseling and testing in prenatal, sexual health, and family-planning clinics and whenever individuals interface with the health care system.
In addition to the DOTS strategy, antiretroviral therapy may play an important adjunctive role in addressing this TB epidemic. Antiretroviral therapy reduces TB risk by 70%–90% in treated cohorts over 1–2 years of follow-up [4, 20]. Voluntary counseling and testing among patients with TB on an opt-out basis may be accompanied by use of antiretroviral therapy among all persons who are found to be HIV seropositive. The majority of such patients here and elsewhere fall within the eligibility criteria for antiretroviral therapy [21, 22]. This intervention may reduce both the rate of new cases of TB, as well as the high TB recurrence rate. However, there is concern that, although antiretroviral therapy is very effective in reducing an individual's risk of developing TB per unit time, the lifetime risk of disease may not decrease [20, 23]. Thus, the overall burden of TB in the community may not substantially diminish, despite good antiretroviral coverage. Other adjunctive interventions that may also be important include use of primary and secondary isoniazid prophylaxis [24, 25].
In conclusion, we have carefully documented the differential impact of HIV infection on TB notifications in age- and sex-specific groups in this community. These data show that additional interventions may need to be targeted in an age-specific manner to address this regional emergency.
We gratefully acknowledge Heather Jaspan and Leigh Johnson, respectively, for input on the data concerning adolescents and the Actuarial Society of South Africa model for estimating the prevalence of HIV infection.
Financial support. Wellcome Trust (074641/Z/04/Z, to S.D.L.) and National Institutes of Health (Comprehensive Integrated Programme of Research on AIDS grant 1U19AI53217-01, to L.G.B., K.M., L.M., and R.W.).
Potential conflicts of interest. All authors: no conflicts.