Study design
A household cluster randomised trial.
Study setting
The study is being conducted in two South African Districts, selected based primarily on their differing TB and HIV burden and existing community research links (Botshabelo Subdistrict, Mangaung Metropolitan Municipality in the Free State, population, 2011 population: 181,712 [30]; and the Capricorn Health District in Limpopo Province, 2011 population: 1,293,000 [31]) – map: Additional file 1: Figure S1. The antenatal HIV seroprevalence and the annual TB incidence for 2012 in Mangaung Municipality and in the Capricorn Health District were 32% and 686/100,000 and 15% and 371/100,000 respectively [32, 33]. These two settings provide valuable information on the intervention when implemented in a TB “hot spot” compared to one with a lower annual TB burden.
Identification of index TB cases
Index TB cases eligible to participate in the study will be any patient (living or recently deceased within 6 weeks of confirmed TB diagnosis) who is a permanent resident of one of the study districts with no plans to relocate over the planned duration of the study, and who has been diagnosed with TB. In index cases older than 7 years, only those with laboratory confirmed pulmonary TB were eligible; if 7 years or younger, we will include patients with TB diagnosed by a doctor without bacteriological confirmation, or bacteriologically-confirmed TB of any organ. Index cases will be diagnosed at any health facility in the study districts in the 6 weeks prior to recruitment. Severely ill and deceased patients will be included as index cases if we obtain written informed consent from a close family member. Severely ill patients with TB will be requested to confirm their participation in the study once they are capable of providing informed consent. We will exclude individuals who are incarcerated or are in long term in-patient care for reasons apart from TB.
Research teams will identify potentially eligible index TB cases by visiting hospital wards and clinics, from routine reporting of TB cases in public-sector health facilities in the two study districts (TB registers and notifications), and from lists of specimens positive for TB generated by public sector laboratories.
For each index TB case who provides informed consent, a baseline questionnaire will be completed; in paediatric index cases or where the index case is deceased or too ill to provide informed consent, this information will be obtained from the caregiver or participant’s representative. Additionally, we will ask index cases – or their representative – to complete a line-list of all household members.
Design of interventions
The households of index TB cases will be randomly allocated in a 1:1 ratio to one of two interventions: intensified household contact tracing and treatment support (intervention); or enhanced standard of care (control).
Randomisation, allocation and blinding
The unit of randomization is households of recruited index TB cases. The dwelling within which the household members live is defined as all rooms under a contiguous roofed area linked by doorways or windows through which air can pass. For this study, household members are defined as all individuals who shared dwelling airspace by either having slept overnight at least once, or shared at least two meals in the same household as the index case in the 14 days prior to the index case’s diagnosis of TB.
Index cases and their household members will be randomly allocated using a computer-generated allocation sequence stratified by district and with a block size of 10 generated prior to the first allocation by the study statistician. For each index case at both sites, allocation is accessed through a central study telephone line, staffed by a data manager in Johannesburg, unlinked to other aspects of the study. Randomisation only occurs after consent procedures and baseline questionnaires and after all eligibility criteria have been satisfied.
This is an open study as we cannot blind index cases, their households, or study field research teams to allocation. However, study research assistants are blinded to the allocation of households when they conduct outcome assessments, and investigator blinding will be maintained until final analysis; no interim analyses which reveal results of the study or provide an indication by group laboratory results, or other outcomes, will be provided to site staff or investigators.
Intensified household contact tracing and treatment support (intervention group)
In households allocated to the intervention group, within 14 days of recruiting the index TB case, a research team comprising of an enrolled nurse and counsellors will conduct an initial household visit, record characteristics of the dwelling, seek consent to participate from all household members, and administer a short questionnaire to each recruited household member that will include: demographic details including relationship and exposure to the recruited index case; risk factors for TB; presence of TB symptoms; and history of previous and current HIV and TB treatment.
All household members will be offered confidential HIV testing and counselling in accordance with South African guidelines - or anonymised testing for study purposes if consent for this is provided - followed by blood sampling for CD4 cell count measurement in household members diagnosed HIV-infected. We will collect one spot sputum sample from each household member capable of producing a sputum sample (irrespective of the presence of TB symptoms), which will be transported to the public sector laboratory at each site, and tested using the GeneXpert MTB/Rif platform until approximately the end of 2017, when the Xpert Ultra was rolled out nationally in South Africa. Sputa will also be cultured using the liquid Mycobacteria growth indicator tube (MGIT) platform. Tuberculin skin testing (TST) will be offered to all household members without confirmed TB.
At a follow-up household visit 1 week later, results of sputum Xpert testing will be communicated in confidence to household members. Participants with microbiologically-confirmed tuberculosis or HIV requiring care and treatment will be supported to attend their nearest local primary care clinic to register for care and treatment. Household members will be offered isoniazid preventive therapy (IPT) according to South African guidelines - which have changed over the course of the trial (in brief those who test HIV-positive and whose sputum tests are negative for TB or those who are under 5 years of age). Additionally, we will offer IPT to HIV-seronegative individuals whose tuberculin skin test is positive (≥10 mm), and who agree to take isoniazid preventive treatment. Patients who accept isoniazid preventive treatment are dispensed the first month of isoniazid preventive therapy by study nurses at their home, and are referred for ongoing preventive treatment at their local clinics.
Because previous studies have demonstrated suboptimal rates of linkage into TB and HIV care [34, 35], a re-visit will be conducted 3 months after recruitment to support access to health facilities where required. If there has been no response to accessing care by individuals in a household, the local clinic will be approached to visit the household and take over their care.
Enhanced standard of care (control group)
Each recruited index case – or their representative – allocated to the enhanced standard of care group will be provided with a referral letter and information materials for every household member they identify at baseline interview, and will be requested to give a pre-printed referral letter to each person in their household. The referral letter contains information about TB and HIV, recommends screening for TB and HIV, and provides details of local health facilities where screening and further care (if required) may be accessed. Additionally, the referral letter contains directions for health providers, including recommendations that TB screening, HIV testing and further care and prevention services (including ART and TB preventive therapy as required) should be offered in accordance with South African National guidelines because the individual was exposed to a likely infectious case of TB.
A week after recruitment, study research assistants will attempt to make a follow up telephone call to the index case, or their caregiver, to ensure the referral letters were given to household members, and to confirm that any ill household members have accessed their local health facility.
Outcome measures
The primary outcome of the trial is a composite one: TB-free survival in the household starting 1 month after randomisation to the final ascertainment visit (an expected total follow-up time of ~ 14 months); the one-month lead-in time is to allow evaluation of the preventive effect of the study interventions, by excluding from analysis incident TB diagnoses that occur between months 0 and 1 that would likely detected as part of the intensified investigations in the intervention households. The follow-up time for those reaching the endpoint will be the time from a month after randomisation, or date of becoming a permanent household member (if entered the household at least 1 month after randomisation) to date of loss to follow-up or the date of the 15-month final outcome ascertainment visit; the follow-up time for those not reaching the endpoint will be the time from 1 month after randomisation or date of becoming a permanent household member (if entered the household at least 1 month after randomisation) to date of tuberculosis diagnosis or to date of death (or if both occur during the follow-up, the earliest of these). All analyses will be done using two different definitions of TB: (1) including all cases where TB was diagnosed or TB treatment started, irrespective of the diagnostic method (“all TB”); (2) including only those where hard copy of a laboratory confirmation of TB was seen by the study team or for whom there is written medical record confirmation of a positive laboratory test for TB (“microbiologically-confirmed TB”).
Secondary outcomes, together with their definitions, are:
-
1)
Prevalence of TB infection at month 15 visit among household children who are aged under 14 years of age at the month 15 visit: Measured by tuberculin skin test (TST) reactivity – defined in three ways:
-
a)
TST diameter ≥ 10 mm (irrespective of HIV serostatus or other comorbid condition) at the month 15 visit. This is our main definition of interest for this key secondary outcome.
-
b)
TST diameter ≥ 15 mm in all participants irrespective of HIV serostatus, or other comorbid conditions at the month 15 visit. This is likely to be more specific.
-
c)
Differential TST diameter: ≥5 mm in HIV infected children and in children who are underweight (defined as either height for age or weight for age Z-score of <− 2); TST diameter ≥ 10 mm in seronegative and non- malnourished children.
We will also report the distribution of TST readings in children < 14 years, and the estimated annual risk of TB infection (ARI).
-
2)
Time between first onset of symptoms and initiation of anti-tuberculosis treatment, among household members diagnosed with TB between baseline and month 15 visit
-
3)
HIV prevalence (reported as a percentage of those HIV-tested at the month 15 visit) of previously undiagnosed, or previously diagnosed, but untreated HIV infection
-
4)
Cumulative incidence of all-cause mortality at the month 15 visit
-
5)
HIV viral load, defined as the prevalence of detectable viraemia (> 400 copies per/ml) within each trial arm (should we have sufficient data from HIV-infected individuals to do this)
-
6)
Estimation of the cost-effectiveness of the intensified household contact tracing and treatment support intervention compared to the enhanced standard of care from a societal perspective. Analyses for this outcome will be described in a separate economic statistical analysis plan.
Subgroup analyses
-
1.
Subgroup analysis by study site (Capricorn District and Botshabelo District) will be conducted for all outcomes, in order to assess whether the effect of the intervention differs between a high (Botshabelo) and lower (Capricorn) TB burden setting.
-
2.
For the secondary outcome of TB infection as assessed by TST reactivity among children aged < 14 years, we will conduct a subgroup analysis comparing children < 5 years of age with those aged ≥5 years at the month 15 visit.
Outcome assessment
We will conduct an outcome household visit to all households, irrespective of their study allocation, at 15 months after the recruitment of the index case. Two months prior to the visit, we will attempt to locate the standard of care households (that have not been visited prior to this time) using a What3Words a mapping application – this to reduce the potential bias of travelling directly to the household versus attempting to find households for the first time at the outcome visit. Moreover, to reduce potential bias, study teams that have not visited that intervention household previously will be used to conduct the outcome visit.
At the outcome visit, Research Assistants blinded to the original allocation of the household will 1) trace all household participants, ascertain vital status and attempt to assess whether those who died, died with TB and/or HIV; 2) record episodes of TB and first HIV diagnosis, treatment and other hospitalizations from verbal report and inspection of patient-held records; 3) investigate participants with symptoms of TB (any of: cough, fever, weight loss, night sweats) by collecting sputum for smear microscopy, sputum culture for M. tuberculosis and Xpert; 4) offer repeat HIV testing to participants negative or not previously tested at baseline; and 5) conduct a prevalence survey for latent TB infection by testing all children under 14 years old with the tuberculin skin test.
Health economics and cost-effectiveness
All costing and economic evaluations will be primarily from the perspective of the health care system to enhance comparability with other studies. However, estimates of household costs of illness and deaths will also be included in certain cost effectiveness measures to reflect a societal perspective. The analysis will also explore the magnitude of household costs, potential for catastrophic costs of HIV and TB, and implications for universal access.
We will use a direct costing approach to assess incremental costs of the intervention in achieving the primary objective. This will consider costs of diagnostic investigations, staff and travel and any other substantial items that may be identified. Intervention costs will be differentiated carefully from various services and also from research activities.
Interviews with household members will record direct and indirect costs (e.g. lost income) of illness and accessing healthcare, and about effect on various household asset holdings to explore dissaving and potential for catastrophic costs.
The principal cost effectiveness indicators will be incremental costs per person diagnosed with HIV and/or TB, incremental cost per person linked to HIV and/or TB care through the intervention in each setting, and incremental cost per incident TB case and death avoided. Further indicators such as estimated incremental cost per life-year saved or per TB treatment completion will be generated if the frequency and profile of reported clinical outcomes support this. In addition, we will measure health-related quality of life of household members at baseline and follow-up using the EuroQol EQ-5D [36].
Records of health service utilization (in-patient days and outpatient visits) for each group will be combined with unit costs for each type of care derived from step-down costings in relevant facilities to estimate possible cost savings and net costs.
Incremental cost effectiveness ratios will be compared to benchmarks for South African HIV, TB and other health interventions to facilitate interpretation for policymakers.
Sensitivity analyses will assess the strength and robustness of cost effectiveness findings, and aggregated and disaggregated unit cost measures will be reported. To clarify comparisons with other interventions and relevance for other settings, reports will clearly describe the case finding initiative and context, and explore system strengthening issues such as capacity needs, effects of scale and overall budget requirements.
Statistical analysis plan
A sample size of 1200 index cases in each district (2400 in total) is proposed, based on previous studies in the districts, and assuming mean household size of 5.5 members, and household TB incidence that is the mean of that reported for each district. This will provide 80% power to detect a 30% difference (two-sided p < 0.05) in TB-free survival between the trial arms with an intra-cluster correlation coefficient of k = 0.3. Routine data from both sites suggest that they diagnose and treat at least double this number of TB cases per annum. No interim analysis will be undertaken.
Data will be captured using paper case record forms, and doubled entered into a secure RedCap Database hosted on a server at the University of Witwatersrand, Johannesburg, South Africa. To maintain confidentiality of participants, completed case record forms will be stored securely at study site research offices, and participant identifying information will be removed during data capture. Logical rules and check ranges for values will be implemented, and data quality checks will be run regularly by the study Data Manager.
All analysis will be done on an intention to treat basis, with denominators comprising households and household members randomly allocated to trial groups. We will compare baseline household-, and index case-, and household members-level socio-demographic and clinical characteristics between allocated trial groups. This will use absolute numbers and proportions for categorical variables. Means and standard deviations, or medians and interquartile ranges, will be used for continuous variables.
Outcomes will be assessed in two analysis populations: A) all household members who were listed at baseline for whom information is available at final study visit (“baseline cohort”, with analysis including and excluding TB index case conducted); B) all household members for whom information is available at final study visit irrespective of presence at baseline (“final visit cohort”). Cohort A will be the primary analysis population. Results from Cohort B will be used to investigate whether the effect of the intervention extends beyond those exposed to the intervention.
All analyses will be done on an intention-to-treat basis, i.e. data from household members included in populations A and B described above, will be included in the analysis, regardless of whether they underwent all intervention procedures in their trial arm.
Additionally, for secondary outcome 1, the analysis population will be restricted to those individuals aged less than 14 years at the time of outcome assessment at 15 months and for secondary outcome 2, the analysis population will be restricted to those individuals with a bacteriologically-confirmed TB diagnosis during the follow-up period.
For the primary trial outcome (TB free survival), Poisson regression with random effects to account for clustering at the household level, will be used to estimate a rate ratio, 95% confidence interval and p-value for the effect of the intervention. To allow for the stratified randomisation by district, a fixed effect term for district will be included in the regression model.
Analysis of secondary and exploratory outcomes will follow a similar approach, except where binomial outcomes are assessed (e.g. proportion of child household members with MTB infection; proportion of household members with undiagnosed or untreated HIV), in which case we will use logistic regression with random effects to account for household clustering. For secondary outcome 5 (community HIV load), we will calculate this as the fraction of HIV-positive individuals with a measured viral load who have detectable viraemia multiplied by the number of individuals testing HIV-positive and divided by the total population with a measured HIV status.
Subgroup analysis will be done by examining stratum-specific (district-specific) rate ratios (or odds ratios for binary outcomes) and by fitting terms for the interaction between trial arm and district in random effects regression models.
This trial offers a number of important health benefits, including early diagnosis and supported access to TB and HIV care, which are known to be problematic and expensive for individuals under current health systems in South Africa. Moreover, the trial will provide important information to assist in the development of improved HIV and TB services and pathways to care in local communities and in South Africa.
All index TB cases (or their representative if deceased) and household members will provide informed consent to participate in the study. Parental consent to participate in research will be obtained for children < 18 years and assent will be obtained from children 7-14 years. Individuals who are illiterate, or unable to write, will be invited to provide a thumbprint confirmation of consent to participant, accompanied by a witness signature.
Household tracing has the potential to inadvertently disclose confidential clinical information such as TB or HIV status, and to perpetrate stigma and discrimination, although previous cluster randomised trials in sub-Saharan Africa have demonstrated that these potential risks are minimal, with trial withdrawal due to stigma or discrimination extremely uncommon [37, 38]. Partner disclosure of HIV status and TB status have been clearly shown to be beneficial to household members and partners. Nevertheless, the trial team will take extensive steps to ensure that recruited index cases are fully aware of the possible implications of household tracing and follow-up, and they will be supported to disclose results to members of their household by trained research assistants and study nurses. HIV and TB testing will use South African National guideline procedures, which have been well-validated in terms of diagnostic accuracy.