An intensive ‘household counselling’ intervention (involving TB contact tracing, multiple home visits, TB & HIV counselling and screening with linkages to care) reduced the prevalence of culture positive tuberculosis (TB) by 22% within randomised communities in Zambia and the Western Cape of South Africa, according to the results of ZAMSTAR, a long-awaited, large-scale trial, which were released on October 30th at the 42nd World Conference on Lung Health in Lille, France.
“In the era of HIV, this is the first community-randomised trial of a public health intervention to be shown to have an impact on the epidemiology of TB at community-level,” said Professor Peter Godfrey-Faussett of the London School of Hygiene and Tropical Medicine (LSHTM), one of three ZAMSTAR principal investigators.
However, another study intervention did not have a positive effect on TB prevalence. This intervention was called enhanced case finding (ECF) – but it is very important not to confuse this with intensified case finding (ICF) which involves actively screening people for TB. ECF, as defined by this study, involved a number of activities to make a TB diagnosis easier to access and encourage people with symptoms to seek out diagnosis, but essentially continued to be dependent on a person’s own health-seeking behaviour (more below).
ZAMSTAR was a major study that took several years to run and analyse. It involved 24 communities that contained close to one million residents overall. To underscore the importance of the study, the former Archbishop Desmond Tutu sent a pre-recorded statement that the investigators played before the results were released:
“ZAMSTAR is the biggest study ever undertaken to understand the real situation with TB and HIV, trying to find ways to interrupt the transmission of TB in communities with a high burden of both TB and HIV.”
Consequently, an entire conference symposium was dedicated to discussing the results, some aspects of which were surprising. These data will probably continue to be analysed for years to come. This article will only touch on some of the key findings from the study, and subsequent articles will present further analysis, and expert opinion on the trial’s ramifications.
Background
The communities in which the ZAMSTAR study took place all had an extremely high burden of TB and HIV, much of it undiagnosed.
Prevalence of TB in the ZAMSTAR sites (from the outcome portion of the ZAMSTAR study)
|
Zambian sites |
Prevalence 542/100,000 adults (Stand Deviation (SD) 263) |
Community range 221-10,967/100,000 |
|
South African sites
|
Prevalence 2319/100,000 (SD 487) |
Community range 1489-3054/100,000
|
Prof. Godfrey-Faussett noted that most people with culture positive TB have symptoms (90% overall, and 95% of the HIV-infected) – but not everyone complains of cough. Most did not meet the formal definition of ‘TB suspect’ at the time (cough for two or three weeks), “and might have been missed by routine services even if they had presented,” he said.
“More alarmingly, perhaps, is that even those people who did meet the definition of ‘TB suspect’ when they went to the clinic, were not asked for a sputum specimen, because of the barriers involved in giving sputum and of registering and going through the process of getting a diagnosis made,” said Prof. Godfrey-Faussett. “So there are many opportunities being missed for TB control. The ZAMSTAR philosophy has been that we do need to go beyond the clinic, if we are going to make a difference.”
Consequently, the ZAMSTAR team sought to evaluate, in a rigorous manner, two ways in which to increase TB diagnosis and treatment and interrupt the transmission of TB in communities with a high burden of both TB and HIV (the home counselling intervention and ECF, described in more detail below).
The study was carried out by a consortium of three institutions, Zambia AIDS Related TB (ZAMBART) Project, Desmond Tutu TB Centre (DTTC) at Stellenbosch University (in the Western Cape province of South Africa) and the London School of Hygiene and Tropical Medicine (LSHTM), as one of three large studies within the Consortium to Respond Effectively to the AIDS and TB Epidemics (CREATE), which received major funding from the Bill and Melinda Gates Foundation to do the sort of research which could produce policy-changing results.
The study design
ZAMSTAR randomised communities – not people (though the outcomes were measured in individuals). Thus, though the total population covered was 962,655, the trial had a total sample size of 24 communities: 16 were spread throughout Zambia – and 8 in the Western Cape province of South Africa. The community, for this purpose, was defined as the population attending one TB diagnostic centre. To participate in the trial, the communities had to have a minimum population of 25,000, a TB notification rate of more than 400 cases per 100,000 inhabitants per year and what was considered to be a high HIV prevalence rate.
Randomisation was stratified by country (since they were quite different) and by high and low tuberculin skin test (TST) baseline prevalence in the communities (see below).
It is very important to note that all the communities involved in the study received extra support – including additional dedicated staff – to improve TB and HIV care and service integration over the course of the study. This third non-randomised intervention may be important to understanding some of the study’s more anomalous results.
Communities were randomised into four arms, six communities per arm.
- Group 1 served as the control since the communities in this arm received neither of the two interventions – though as Dr Helen Ayles, also one of the principal investigators, based in Zambia, said (tellingly perhaps) “we don’t like to call it a control arm” because of the additional staff. (Population included: 257,698.)
- Group 2: Enhanced case finding (ECF). (Population included: 148,090.)
- Group 3: The household intervention (HH). Population included: 257,729.)
- Group 4: Both ECF and HH. (Population included: 299,138.)
The primary endpoint was the prevalence of tuberculosis in the communities (compared by intervention) – in other words, ECF vs no ECF; and HH versus no HH. The secondary endpoints, compared in the same way, included the incidence of TB infection at the community level (which was measured by TST conversions in children, indicating TB transmission); and at the household level: difference in TB outcomes, TB incidence, and HIV incidence were measured.
The interventions
The intervention phase of the trial lasted three years (2006 to 2009) carried out by field staff and trained community members – with rigorous monitoring and evaluation and quality assurance.
Enhanced case finding (ECF): ECF consisted of community mobilisation efforts to encourage community members to seek a TB diagnosis, combined with efforts to improve access to a timely TB diagnosis.
Community mobilisation activities could be drama, using megaphones, leafleting, football matches, fashion shows, beauty pageants and “whatever worked for that community,” said Dr Ayles. In addition, there was a school intervention – with TB/HIV education activities for children who, it was hoped, would then take the message of TB and HIV home to their families, and encourage household members to seek diagnosis whenever they had symptoms of TB.
To increase TB diagnostic access, sputum collection points were set up in the community, and at community mobilisation events. Also, open access or fast track points were set up at the community’s TB clinic, so that individuals coming to the clinic with a cough could immediately go and give a sputum sample, and get the result back there without having to sit and wait and go through the normal clinic process.
The guiding principles for ECF were that each person should be able to give a sputum sample within a 30-minute walk of their home; and they should be able to get their sputum smear microscopy results within 48 hours.
Household intervention (HH)
The household intervention (HH) involved contact tracing of a TB patient, “who served as the gateway to a household at risk from TB and HIV,” said Dr Ayles. It consisted of three visits by a trained household counsellor – one at the beginning of a TB patient's treatment, one at two months, and one at the end of TB treatment.
During these visits, the whole household was educated about TB and HIV and each household member was screened for TB using a symptom screen and smear microscopy. HIV counselling and testing was also offered for all members of the household, either to the group as a whole, the couple or individually. Then the household received ongoing counselling and referral for care for HIV and TB as necessary.
Process results
ECF did appear to attract at least some TB suspects, with 20,630 people submitting a sputum sample. Overall, 4.6% of the residents in the communities gave a sputum sample through ECF. Out of the samples submitted, 1699 (8.2%) were found to be smear-positive. This translated into an overall case rate of 277/100,000 (211/100,000 for Zambia, and 561/100,000 in South Africa), accounting for 24% of all the smear positive diagnoses being made in the community. The cost for the intervention was (US) $17,137-24,455 per ECF site, per year, for a total cost of $0.37-0.71 per person, per annum.
Dr Ayles noted that getting the microscopy result back to the patient within 48 hours proved challenging. She said, “We only achieved our target in just under 50% of the time in Zambia,” where diagnosis was decentralised. In South Africa’s centralised system, a result in under 48 hours was obtained in less than 25% of the specimens.
HH reached 9353 households containing 36,751 people, though only 84% (27,074) were reached by all three household visits. This amounted to 5.8% of the population in the HH communities (though, again, some of the individuals at highest risk of TB and HIV). The intervention, if spread across the community cost US $24,126-34,661 per site per year, or $0.48-0.80 per person per annum.
Dr Ayles did not present the number or rate of TB cases identified by contact screening but a large number of HIV cases were identified. 99.3% of the adults (over 18,000 individuals) in each household received HIV education and counselling, of whom 12,000 (66.5%) accepted HIV testing, and 7021 (37.9%) were HIV-positive (the proportion of those who actually tested who were positive was much higher). Around 4000 are accessing ART and cotrimoxazole, but the uptake of isoniazid preventive therapy has been very low in these communities.
Outcomes
After the intervention period, adults in the community were sampled at random to determine the prevalence of TB (the statistical calculations and analytical methods used to measure the impact of the interventions may be discussed in a separate article).
55,450 households were visited – though a number of households (or individuals) refused to offer consent to participate in the survey. 90,601 consenting individuals answered a questionnaire, submitted a respiratory sample (inoculated into two liquid culture tubes), and were offered HIV testing. Blood sugar testing was also offered due to the strong association between diabetes and tuberculosis. One batch involving 16,710 specimens was rejected due to quality assurance problems and 9461 specimens were rejected due to contamination.
Out of the remaining 64,430 evaluable cultures, 884 were found to have TB.
Upon analysis, only the HH intervention appeared to reduce TB prevalence in the community, but ECF had no effect (see table below).
A longitudinal study was performed to evaluate TB transmission in the community, as measured by TST culture conversions among children who were negative at baseline. The baseline survey was conducted in 98 schools in 24 communities, involving 21,393 children in grades 1-3. At baseline, 16.5% were TST positive (with an induration of ≥ 10 mm) in Zambia (95% CI 12-21.1, and 30.5% in South Africa (95% CI: 22.9-38.2). There was little difference seen between children with a BCG scar and those without.
8809 children who were negative at baseline were seen during the second survey roughly four years later – 733 had converted to TST positive. The HH intervention arm appeared to reduce conversions by 55%, though this was not quite statistically significant. However, the ECF intervention arm did not appear to reduce incidence at all.
Effects of the interventions at the community level
|
HH |
ECF |
|
|
TB prevalence in intervention communities |
746/100,000 |
927/100,000 |
|
TB prevalence in non-intervention communities |
883/100,000 |
711/100,000 |
|
Adjusted risk ratio (95% confidence interval) |
0.78 (0.61-1.00) (significant) |
1.11 (0.87-1.42) (NS) |
|
TB incidence in intervention communities |
0.87/100 years |
1.41/100 years |
|
TB incidence in non-intervention communities |
1.71/100 years |
1.05/100 years |
|
Adjusted risk ratio (95% confidence interval) |
0.45 (0.20-1.05) |
1.36 (0.59-3.14) |
The analysis of the secondary outcomes relied on a cohort survey (SOCS), in which TB index cases who were involved in the HH invention were enrolled for follow-up throughout the intervention period, to see whether there were changes in TB outcomes or incidence and HIV incidence in the home. None could really be found partly because TB outcomes were generally good once diagnosed, and the incidence of HIV was already high. The SOCS visits may also have diluted the intervention effects.
“TB is sometimes seen as an unavoidable part of life in poor communities. We need to improve and strengthen our health services, we need better methods to diagnose TB in clinics and we need people affected by TB, usually in poor communities, to demand that things change,” said Dr Godfrey-Faussett.