|Home | About | Journals | Submit | Contact Us | Français|
Crowded townships of Cape Town, South Africa, where HIV prevalence and tuberculosis (TB) notification rates are among the highest in the world.
To determine age-specific prevalence rates of latent tuberculosis infection (LTBI) among HIV-negative individuals and determine the annual risk and force of infection during childhood and adolescence.
A cross-sectional survey using standardised tuberculin skin testing (TST) in HIV-negative individuals aged between 5 and 40 years. A TST diameter of ≥10 mm was defined as indicative of LTBI.
Among 1061 individuals, the proportions of patients with low grade TST responses of 1-5mm or 5-9mm were low (0.8% and 3.9%, respectively). However, the proportions of patients with a TST diameter of ≥10 mm increased from 28.0% in the 5-10 year age stratum to 88.0% in 31-35 year age stratum. Mean annual risk of infection up to age of 5 years was 3.9%. However, the estimated force of infection increased throughout childhood to a maximum of 7.9% per year at 15 years of age.
An extremely high rate of acquisition of TB infection in childhood and adolescence results in very high LTBI prevalence rates in young adults who are most at risk of acquiring HIV infection.
South African tuberculosis (TB) notifications have increased 6-fold over the last two decades, largely as a result of increasing HIV prevalence.1 A total of 461,000 new cases of TB in 2007 reflected the one of the highest national TB notification rates in the world (948/100,000 population). An estimated 73% of these cases were co-infected with HIV, and South Africa alone accounted for approximately 25% of the global burden of HIV-associated TB.1
Total TB notifications in Cape Town, a city of 3.2 million people, reached 27,000 in 2006.2 The distribution of TB cases within this population, however, is very unequal, with unprecedentedly high burdens in the crowded and socially deprived African townships. Here TB annual notification rates were reported to exceed 1500/100,000 in 2006.3-5 Fuelled by high HIV prevalence, TB notifications are now most frequent between the ages of 20 years and 40 years.6
This marked deterioration of TB control in South Africa over the past two decades of increasing HIV prevalence has occurred despite reported progress towards National TB control programme case management targets.1,7 Coverage using the World Health Organization (WHO) directly observed treatment, short–course (DOTS) strategy is 100% and case detection rates have remained above target since 2003.1 In the decade 1996 to 2005, treatment default and treatment failure during treatment improved from 18.1% to 10.4% and from 3.5% to 1.7%, respectively.8
In 1990, Karel Styblo, a leading protagonist of effective case management for TB control, postulated that “the impact of HIV infection on the epidemiological situation of TB is so large that under some conditions, the tools available at present for TB control will fail to restrain the incidence of TB caused by HIV-infection”.9 He further proposed that the impact of HIV infection would depend not only on the prevalence of HIV infection but also on the prevalence of latent TB infection (LTBI) and the annual risk of TB infection in the general population. However, although the HIV epidemic in South Africa has been carefully monitored by annual sero-prevalence surveys among antenatal women and by household surveys,7,10 few data describe the prevalence of LTBI or the annual risk of TB infection in the general population.
We therefore undertook a study to measure the prevalence of LTBI in the township populations of Cape Town. Intra-dermal tuberculin skin testing (TST) has been the most widely used of all the immunological tests for estimation of prevalence, incidence and trend of Mycobacterium tuberculosis infection in populations, despite concerns over its sensitivity and specificity.11 We assessed TST responses among healthy HIV-negative residents of high density townships aged between 5 years and 40 years and used these to derive estimates of the age-specific prevalence of LTBI. Further analyses estimated the annual risk of TB infection in the study population and the force of TB infection (the rate of acquisition of TB infection among the residual pool of uninfected individuals). These data provide important insights into the explosive impact of the high rates of HIV acquisition in young adults on TB control.
TST responses were assessed in 1061 healthy individuals, including children and adolescents aged 5-17 years (n=832) and HIV-negative adults aged 18-40 years (n=229). Participants were residents of high density predominantly black townships of Cape Town where annual TB notification rates in 2006 exceeded 1500/100,000.3-5 Children were all school attendees and healthy HIV-negative adults were recruited from a prospective cohort recruited for a phase III HIV vaccine study (n=60) or from HIV voluntary counselling and testing centres (n=169). A history of previous TB, isoniazid preventive therapy, steroid therapy and pregnancy were exclusion criteria. These studies were approved by the University of Cape Town Research Ethics Committee.
TST responses were assessed using the WHO-recommended standard methodology.12 Two tuberculin units of purified protein derivative (PPD) RT23 (Statens Seruminstitut, Copenhagen, Denmark) were administered intradermally to the volar surface of the left forearm. Induration was assessed between 48 hours and 72 hours after inoculation and the diameter was measured. Age and gender were also recorded for each participant as was Bacille Calmette-Guerin (BCG) scar status in children.
TST induration of ≥10mm at the inoculation site was considered indicative of LTBI.
The relationship between TST diameter and age was explored by regression analysis. A generalized non-parametric logistic model 13 was proposed to quantify the relationship between LTBI prevalence rate (as defined by TST diameter ≥10mm) and age. We fitted the model using the penalized regression spline approach.14 To evaluate gender differences, prevalence rates for female and male subjects were obtained separately by using a generalized semi-parametric logistic model which was then fitted to the data.15 The proportional rate of change in TB prevalence per year increase in age was estimated using the derivative of the prevalence rate with respect to the age of subjects. The annual risk of infection was calculated as: 1-(1-prevalence)1/mean age+0.5. The force of infection was also calculated at specific ages for the pool of individuals who remained uninfected [annual change in prevalence/(1-prevalence)]
TST responses were available from 1061 individuals (524 male and 537 female). Diameters of induration ranged between 0 and 35 mm with a mode within the 16-20 mm stratum (Figure 1). No TST response was recorded in 535 individuals (50.4%), 1-5mm in 8 individuals (0.8%), 5-9mm in 41 individuals (3.9%) and ≥10mm in 477 (45%) individuals (Figure 1).
To establish whether the magnitude of TST responses was associated with age, TST diameters recorded as ≥1 mm were plotted against age (Figure 2). The linear regression line for these data approximated to a horizontal line, indicating no overall significant relationship. Furthermore, among those aged 5-17 years, there was no significant difference in the mean TST diameter comparing those who had observable BCG scars (n=213) and those who did not (P=0.28).
Using a TST diameter of ≥10mm to define LTBI, the age-stratified prevalence of infection for the total population and for male and female sub-populations were calculated (Table 1). LTBI prevalence increased steadily from 28.0% in the 5-10 years age stratum to a peak of 88.2% in the 31-35 years age stratum.
To better demonstrate the quantitative relationship between LTBI prevalence and age, a non-parametric curve fitting approach was used (Figure 3). The smoothed curve shows best fit of the prevalence with increasing age together with 95% confidence bands. This analysis was also used to derive estimates of the LTBI prevalence at specific ages (Table 2). This shows that LTBI was estimated to be present in approximately one third of 10 year old children, one half of adolescents aged 15 years, two-thirds of 20 year olds and three quarters of 25 year-olds (Table 2).
Estimates of LTBI prevalence were separately derived for female and male subjects (Figure 4). The difference in prevalence rates between the male and female subjects was greatest among the older study participants. The maximum estimated LTBI prevalence among males was at 33 years (85%) compared to 29 years among females (78%). The decreasing trend in prevalence at ages greater than 29 years was mainly due to changes in prevalence among females. However, the main gender effect (P=0.218) and the age-gender interaction effect (P=0.294) were not statistically significant.
Estimates of the change in LTBI prevalence rate per year increment in age were derived for males and females (Figure 5). The patterns of the curves were similar for both groups. The rate of change in prevalence rate increased from 5 years of age and peaked at the age of 13 years. Maximum rates were 4.4% per year in males and 3.7% per year in females (Figure 5). After age 13 years, the annual rate of change in LTBI prevalence decreased.
Further analyses estimated the annual risk of TB infection among children aged 5 years, 10 years and 15 years, showing a range of 3.9% per year to 4.8% per year (Table 2). The force of infection was also calculated at specific ages for the pool of individuals who remained uninfected [annual change in prevalence/(1-prevalence)] (Table 2). This parameter reached a peak of 7.9% (95% CI 2.5,13.2) per year among individuals aged 15 years and was negative above the age of 30 years. This analysis of force of infection did not adjust for active TB disease (an exclusion criterion from the cohort) or TB-associated mortality.
We have shown rapidly increasing prevalence of TST responses in healthy HIV-negative township residents aged between 5 years and 40 years of age. Using a cut-off of ≥10 mm diameter of induration as evidence of LTBI, we found that by the age of school entry almost a fifth of children were already infected. By the average age of sexual debut at 15 years,16 50% of adolescents in these communities were infected. By the age of 25 years when HIV prevalence peaks in South Africa,7 approximately 75% of individuals had evidence of LTBI. The rate of increase in the prevalence of LTBI was maximal at 13 years of age in both males and females. Between the ages of 5 years and 15 years the mean annual risk of TB infection (ARTI) remained exceptionally high (range 3.9% to 4.8%) while the force of infection (the risk of infection in the residual pool of uninfected individuals) was maximal at 7.8% at the age of 15 years.
The complexities and limitations of the TST have been extensively discussed elsewhere.11,12,17 Controversies have included the choice of the antigen utilised, the threshold for defining positivity and the performance characteristics of the test in different settings. The tuberculin reagent employed in this study was the standard WHO recommended PPD RT23, which allows comparison with other population surveys.18 Non-specific sensitivity resulting from exposure to environmental mycobacteria impairs test performance in some populations,11 manifesting with a high proportion of low-positive results. In our study population there was a relatively clear separation between negative and positive results; the mode of positive results was in the 16-20 mm diameter stratum and the proportion of low positive results was very small. Furthermore, diameter of induration was not directly associated with known confounders, such as presence of BCG scar or the age of participants. We therefore used the conventional but conservative cut-off for test positivity of ≥10 mm.
The relationship between prevalence of LTBI, age and sex is determined by individuals’ current and lifetime exposure to infection. While prevalence in children reflects more recent transmission, prevalence in adults reflects overall historical trends in transmission risk. It has been observed that at low prevalence rates of LTBI, instability of TST reactions may lead to false interpretations of secular trends.19 However, the predictive value of TST responses is much greater when prevalence of infection is high.11 In our study population, TST performance characteristics appeared good and LTBI prevalence rates were exceptionally high. Furthermore, trends in sputum smear-positive TB notifications are well documented and have increased relentlessly over the last two decades in South Africa and particularly in the study communities.1,6,20
To explore the relationship between LTBI prevalence, age and sex, we fitted a generalised non-parametric (ie. no assumptions about the data distribution) logistic model. The resultant “S”-shaped curve had strong similarities with age and gender prevalence curves from TST surveys performed over 50 years ago in urban and rural sites of Bechuanaland (Botswana), Ghana, Liberia, Nigeria, Sierra Leone and Swaziland.21
Between 5 and 15 years of age, the curve is concave upwards (Figure 3). This could be indicative of decreasing transmission in recent years 22 but, in the known context of rapidly increasing TB notifications over the past two decades, this is much more likely to be indicative of increasing risk of infection with age.23 The finding of maximal risk of acquisition in the mid-teenage years may reflect social mixing patterns and associated TB exposure in this age-group.23,24
From 15 years to 30 years the curve had an exponential form (Figure 3), compatible with either a steady or decreasing infection risk with age.23 From 30 years onwards declining prevalence with age was more marked in females. Declining of immunological memory with age did not appear to be a significant contributor as reaction size was not shown to diminish significantly with age. Alternative explanations could include immigration from lower prevalence settings, survival from a period of lower TB transmission or preferential progression of individuals with TST positivity to TB disease, as a history of TB treatment was an exclusion criterion for the study.
Larger TST studies, such as those performed between 1955 and 1960, included heterogeneous populations with very variable TB exposure rates.23 This study which was modest in size (n=1061), particularly focussed on a population within a single city which has an extraordinary high burden of HIV and TB disease. The study population was not representative of the total community, as HIV-infected individuals and those who had already developed TB disease were excluded. We believe that the risk of TB infection and performance characteristics of TST in HIV-infected and patients with TB disease may be very different. The results are therefore only descriptive of those remaining free of both TB and HIV. The mean ARI, a measure of sequential annual risks in children under 5 years was approximately 4%, although the more dynamic measure of force of infection at a specific age was similar at 5 years, it reached a peak of 7.9% at age 15 years. Such rates of TB infection, as reported elsewhere in Cape Town,25 are unprecedented in the modern post chemotherapy era. Furthermore, in the context of deteriorating TB control, use of the change in prevalence per year may tend to underestimate the force of infection.
In summary we have shown an extremely high rate of acquisition of LTBI in childhood and adolescence in poor African townships of Cape Town. The combination of high prevalence and force of infection in adolescence before the acquisition of HIV infection may be a key factor underlying the explosive HIV-associated TB epidemic in South Africa. The long term aim of TB control is to lower infection rates in successive generations. Present facility-based TB control is failing to decrease TB infection rates in children and adolescents in these communities. Systematic evaluation and reduction of infection rates in these high burden communities of Southern Africa should be incorporated as a target of TB control.
This research was a collaborative project of the Developmental Centre for AIDS Research (D-CFAR) University of Rochester, NY, USA. R.W. is funded in part by the National Institutes of Health (NIH/NAID) grants 1U19AI53217-01 and RO1 A1058736-01A1. H.L is funded by NIH/NAID grants AI59773 and NSF grant DMS 0806097. H.W is funded by NIH/NAID grants AI50020 and AI055290. S.D.L is funded by the Wellcome Trust, London, UK with grant number 074641. MXR is funded by the Wellcome Trust, London, UK with grant number 072070 and by the European and Developing Countries Clinical Trials Partnership grant 060613.