Three provinces, in South Africa, with the highest TB caseload were selected for inclusion in the study. One district in each province (N = 3) with the highest TB caseloads were consequently included. These districts were Siyanda in the Northern Cape Province, Nelson Mandela Metro in the Eastern Cape Province, and eThekwini in KwaZulu-Natal Province. Within each of these three study districts 14 public primary health care facilities were selected based on the highest TB caseloads per clinic (N = 42). The type of health facility was a primary health care clinic or community health centre. The study catchment areas within the study health districts and randomization procedures enabled a broad coverage of major population groups.
In order to assess the effectiveness of the Screening and Brief Interventions (SBI) among tuberculosis patients found to be misusing alcohol, a cluster randomized controlled trial design was implemented. All new TB and retreatment patients were screened using the AUDIT. TB patients who met the cut-off for misusing alcohol both in the intervention and control arms were reassessed after baseline assessment at time 2 (3 months following intervention) and time 3 (6 months following intervention). The intervention comprised the following: personalized feedback on AUDIT results, a health education leaflet, simple advice and brief counselling about reducing excessive drinking, during one −20 minute- session. The trial incorporated cluster randomization of public primary health care facilities to avoid the risk of contamination.
Principles for recruitment
Public primary health care clinics
Clinics with a high TB caseload (based on statistics provided by the Department of Health) in each of the three study districts with a high burden of TB were included in the study.
Patients with active TB
New tuberculosis and retreatment patients (i.e. those patients who have been initiated or have been on anti-TB treatment for less than one month), males and females, 18 years and above who visited the primary health care facility and who scored 8 or more for men and 7 or more for women on the AUDIT questionnaire after the screening were included in this study. A few studies have reported that brief interventions are effective even among excessive drinkers [22–24]. Therefore patients scoring high on the AUDIT were also included in the study. Under 18 years of age were excluded since in South Africa the legal minimum age for alcohol use is 18 years and the study aimed at providing an intervention for hazardous, or harmful drinking or alcohol dependence.
Randomisation was conducted using a secure remote randomization service. Within each district in each of the three provinces the 14 public primary health clinics with the highest TB caseloads were randomly assigned to the treatment and control arms. A secure remote randomisation service carried out randomisation. Twenty one allocations were initially generated for each of the possible factorial combinations of clinic type, TB case load, and intervention. Randomisation was stratified by clinic type (clinic and community health centre) and TB case load. The standby and reallocated clinics were subsequently randomly allocated in a similar manner. TB patients misusing alcohol in intervention clinics were receiving treatment and those in control clinics received a health education leaflet. At the clinic level all consecutive new and retreatment TB patients were systematically recruited over a period of six months from mid- April to mid- October 2011.
Participants (clinic staff members and TB patients) were not blinded to their intervention or control status. However, to protect against information biases in the reporting of alcohol use and TB treatment adherence behaviour, the data collection team who assessed the outcomes were blinded to the clinic’s status as intervention or control group.
Universal screening of all new tuberculosis and retreatment patients was used where all consecutive patients visiting the public primary health care facility were screened for alcohol misuse and offered a brief counselling intervention if they met the criteria for misuse of alcohol. A health care provider who identified a new or retreatment TB patient (within one month of treatment initiation) informed the patient about the study and referred the patient for participation if interested. A research assistant asked for consent from patients attending the public primary care facility to participate in stage 1 of the study, i.e. screening or baseline assessment using the AUDIT questionnaire. To increase the reliability of the AUDIT, researchers have suggested locating alcohol consumption within the context of other health-related behaviours . The interviewer administered questionnaire, therefore, included questions on mental and physical health status, tobacco use and chronic diseases. The research assistant was not involved in delivering the brief counselling intervention. All participants underwent the initial assessment and the research assistant scored the results of the alcohol test section of the questionnaire. Tuberculosis patients who scored 8 or more for men and 7 or more for women on the AUDIT questionnaire after the screening (risky drinkers) were then approached by the research assistant for a second informed consent for enrolment in stage 2, the intervention study. For patients included in the study, the research assistant referred the patient to a clinic lay counsellor who carried out the intervention (experimental = brief counselling intervention or control = provision of health education leaflet) for all the participants during the same visit. Thereafter patients were scheduled for follow-up sessions at 3 and 6 months at the health facility, and the follow-up assessments were conducted by the research assistant. The experimental brief intervention included a brief counselling intervention on alcohol risk reduction consisting of two sessions, the first immediately after alcohol screening and the second within a month thereafter. In the control condition the clinic lay counsellor provided an alcohol education leaflet. Sampling occurred throughout all hours of clinic operation over a 6 months period from mid March 2011 to mid October 2011. Research assistants conducted follow-up interviews at 3 and 6 months following baseline assessment at scheduled clinic visits. Several attempts were made to follow-up participants. Firstly, HSRC co-ordinators visited the clinics to access participants fieldwork records with permission of the professional staff in an attempt to get contact details of participants. Secondly, fieldworkers did home visits where no telephone numbers were available for participants. Finally, a special team of research assistants were employed to make direct calls to both fieldworkers and participants to help find participants to schedule appointments. Participants received Rand 60 (8.5 US$) transport reimbursement at 6-month follow-up. Ethical approval was obtained from the Human Sciences Research Council Research Ethics Committee (Protocol REC No.1/16/02/11). The Department of Health in South Africa has also provided approval for this study. Consent to participate was obtained in a 2-stage process. Research assistants initially asked for informed consent to conduct a health screen and collected basic information and checked eligibility to take part. No identifiable information was collected at this stage. Those patients who had a positive score on the AUDIT (alcohol risk score), had the study explained to them verbally and in writing (using the patient information sheet). Informed consent was obtained at this second stage which included permission to give the contact details to the research staff, participate in the experimental or control condition and follow up after 3 and 6 months by the research assistant and provide permission to medical file information.
Screening and baseline assessment
To identify drinking and other health behaviours, interviews with potential study volunteers were conducted by trained research assistants using a standardized questionnaire. The 10-item AUDIT  assessed alcohol consumption level (3 items), symptoms of alcohol dependence (3 items), and problems associated with alcohol use (4 items). Responses to items in the AUDIT were rated on a 4-point Likert scale from 0 to 4, for a maximum score of 40 points. Higher AUDIT scores indicate more severe levels of risk: a score of 8 or more indicates a tendency for problematic drinking, and a score of 20 or above probable alcohol dependence . Heavy episodic drinking is defined as the consumption of six standard drinks (10 g alcohol) or more on a single occasion . In South Africa a standard drink is 12 g alcohol. Because AUDIT is reported to be less sensitive at identifying risk drinking in women , the cut-off points of binge drinking for women (4 units) were reduced by one unit as compared with men (5 units), as recommended by Freeborn and others . The AUDIT has been validated in HIV patients in South Africa showing excellent sensi-tivity and specificity in detecting MINI-defined dependence/abuse (sensitivity, 100% and correctly classified 79% of individuals who did not have alcohol abuse or dependence)  and among TB and HIV patients in primary care in Zambia demonstrating good discriminatory ability in detecting MINI-defined current AUDs (AUDIT = 0.98 for women and 0.75 for men) . To comply with the timeline of this study, all subjects were asked for their alcohol consumption in the previous 3 months rather than in the past 12 months.
A researcher-designed questionnaire was used to record information on participants’ age, gender, educational level, marital status, income, employment status, dwelling characteristics and residential status. Poverty was assessed with 5 items on the availability or non-availability of shelter, fuel or electricity, clean water, food and cash income in the past week. Response options ranged from 1 = “Not one day” to 4 = “Every day of the week”. Poverty was defined as higher scores on non-availability of essential items. The total score ranged from 5 to 20, 5 = being low, 6-12 = medium and 13-20 = high poverty. Cronbach alpha for this poverty index was 0.89 in this sample.
Two questions were asked about the use of tobacco products. 1) Do you currently use one or more of the following tobacco products (cigarettes, snuff, chewing tobacco, cigars, etc.)? Response options were “yes” or “no”. 2) In the past month, how often have you used one or more of the following tobacco products (cigarettes, snuff, chewing tobacco, cigars, etc.)? Response options were once or twice, weekly, almost daily and daily.
TB treatment status, HIV status and antiretroviral treatment were assessed by self-report and from medical information. Patients were also asked about a list of chronic and other illness conditions they had been diagnosed with such as diabetes.
Health education leaflet
All the patients randomized and allocated to the control group completed the baseline measures and received a health education leaflet on responsible drinking.
The patients randomized and allocated to the intervention arm completed baseline measures and received brief counselling for alcohol reduction intervention. The goals for brief counselling were as follows: 1) To identify any alcohol- related problems mentioned in the interview, 2) To introduce the sensible drinking leaflet, emphasis the idea of sensible drinking limits, and make sure that patients realize that they are in the risk drinking category, 3) To provide feedback on the relationship between alcohol and TB treatment [21, 29], 4) To work through the first 3 sections of the problem solving manual while mentioning the value of reviewing the other sections, 5) To describe drinking diary cards, 6) To identify a helper, and 7) To plan a follow-up counselling session. The Information-Motivation-Behavioural Skills (IMB) Model was used in the study to guide the alcohol reduction intervention. The IMB model [30–32] proposes that information about alcohol misuse and methods of reducing and preventing harmful and/or hazardous drinking is a necessary precursor to risk reduction. Motivation to change, however, also directly affects whether one acts on information about risk and risk reduction. Finally, the IMB model holds that behavioural skills related to preventive actions represent a final common pathway for information and motivation to result in alcohol risk behaviour change. The IMB model posits that information and motivation activate behavioural skills to ultimately enact risk reduction behaviours. The IMB model also shows that information or motivation alone can have direct effects on some preventive behaviour, such as when information about risky alcohol drinking prompts drinking at moderate levels or to stop drinking more detail, . The intervention was composed of two scheduled sessions: on Day 1 and within one month after the baseline evaluation. Each session comprised 15–20 minutes of counselling.
Counsellor training and intervention quality assurance
The intervention counsellors consisted of lay HIV counsellors from the study clinics who spoke the predominant languages, namely English, Afrikaans, i-Zulu, i-Xhosa and Tswana, in the respective areas. These counsellors delivered the interventions to the participants of the study in addition to the “treatment as usual”. The treatment “as usual” consisted of the provision of standardised short-course chemotherapy under direct observation of treatment (DOT) for at least the initial phase of treatment to all identified smear-positive TB cases in all study clinics . All lay counsellors and up to four nurses per study clinic who were suitable to deliver the brief counselling intervention received formal training (lay counsellors 3 days and nurses 2 days) and supervision prior to the start of the study. It was agreed that the HIV lay counsellors would implement the intervention and the trained nurses will assist when necessary. The training took a practical, systems approach, aiming to facilitate the implementation of SBI in clinic operations rather than merely educating staff. The training curriculum contained modules addressing practical issues deemed essential to implementing the programme. For early identification of alcohol problems in public primary care the AUDIT  and for the brief intervention the WHO brief intervention package for hazardous and harmful drinking  was used. Both were adapted to the South African context, e.g. in terms of standard unit of alcoholic drink and drinking limits. The AUDIT was translated and back translated according to scientific standard procedures  into four of the major languages (Tsonga, Northern Sotho, Venda, and Afrikaans) [16, 18] and also translated into additional languages (Xhosa, Zulu and Tswana). The self-help booklet for patients and a hand-out on “cutting back” showing the drinking limits and health effects of risky alcohol consumption were also made available in the languages relevant for the study. The AUDIT manual explains the purpose of screening for alcohol problems in primary care, the context of alcohol screening, the development and validation of the AUDIT, administration guidelines, scoring and interpretation. The Brief Intervention manual defines concepts and terms, roles and responsibilities of Public Primary Health Care, SBI: a risk management and case finding approach, alcohol education for low-risk drinkers, abstainers and others, and simple advice a brief counselling for risk zone drinkers, self-help booklet and training sources. Critical administrative activities included administration and scoring of the screening instruments, assuring availability of patient brochures, sequencing of interventions with treatment of presenting health problems, the essential elements of an intervention, and the management of SBI records.
The training comprised of four elements: orientation to the relevant practice, standardised power point presentation, tape recorded simulated consultations with trained actors and on-going clinical supervision by experienced HSRC staff. The simulated consultations were recorded and rated by two independent clinical assessors. The brief intervention counsellor was assessed for adherence to the brief counselling protocol in addition to their behaviour and skills using a Behaviour Change Counselling Index .
Assessors submitted the ratings, comments and supervision points for each consultation. This information supported clinical supervision and training until the brief intervention counsellor reached a required standard of practice agreed by the assessors .
To help protect against counsellor drift (deviating from the existing treatment intervention protocol), the brief intervention was completely manualized and was used to guide the counsellor through the content of the session. Lay counsellors received bi-weekly support visits by the project trainers during the implementation of the project. Monthly visits were paid by research staff to the clinics in support for the project to assist them with any technical aspects of the counselling intervention. In terms of control around the quality and consistency of the implementation of the intervention, intervention counsellors filled in a patient monitoring form for each counselling session conducted. It included sections on the client’s AUDIT score, stage of change, action and intervention plan, handing of action plan to client and comments. A readiness ladder was used to assess a patients’ readiness to change their drinking, i.e. their stage of change. Patients were asked to rate on a ladder from ‘not at all’, ‘a little’, ‘somewhat’, ‘very’ to ‘extremely’, “How important is it for you to change your drinking?” Patients who score in the lower end (‘not at all’) of the scale were classified as pre-contemplators, those who score in the middle range (‘a little’ – ‘somewhat’) as contemplators, and those scoring in the higher range were considered as ready to take action. Intervention counsellors were compensated with R20 (US$2.6) for each counselling session conducted and monitoring form completed. In addition, study fieldworkers were able to report to their coordinators regarding any problems they may be having in implementing the brief interventions. Regular meetings between the researchers and the project managers allowed for any problems to be resolved timeously.
A health survey questionnaire was administered in a face-to-face interview in scheduled appointments at the clinic was used to collect outcome information at 3 and 6 months following the baseline assessment. In addition, medical file information was collected for HIV and TB treatment status and outcome. Non-attenders were followed up by telephone and home visits arranged as necessary. The primary outcome was the (1) change in the mean score on the AUDIT in the last 3 months and the number of AUDIT negative drinkers in the last month of the study period compared with baseline, as measured by the AUDIT. The secondary outcome was the successful TB response, classified by WHO as cured or treatment completed (versus treatment failure, defaulted, died or transferred out to another health facility .
Sample size calculation
Using the Practihc Trial Protocol tool sample size is calculated with cluster size as a binary outcome based on an average cluster size of 350, an intra-cluster correlation coefficient of 0.2, with 90% power and 5% significance level. A previous review  found a clinically important difference in negative status on AUDIT between brief intervention and controls of 13% (5% reduction in controls and 18% in brief intervention recipients). With an expected 20% AUDIT negative in the intervention and 5% in the control, a minimum of 200 sample per group (total 400), with 20 clusters per group, would be needed. Assuming a loss to follow-up of 40%, the sample size was inflated to 280 per group (total 560). As initially only new TB patients were considered in the trial but later new retreatment patients were also included in the trial the sample size was doubled to 1120.
Data management and analyses
Data were captured by dedicated data capturers. Data cleaning and data verification was supervised by key project staff, including the principal investigator and methodologist. The verification process included double data entry of all questionnaires and its fields, doing programmed range checks by computer to identify outlying values, checking for missing values, and checking for inconsistencies in the data. Cluster-specific methods of data analysis (Generalized Estimating Equations) were used because we randomized clinics rather than patients. Data were analysed using STATA Version 11.
Intention to treat
The principle of the statistical analysis was intention to treat. It was applied to the two hierarchical levels of the trial: the clinics randomized at the cluster level and the patients recruited within each of these units.
In all 36 units randomized were included in the analysis. Units were analysed according to the intervention group allocated at randomization. The stratification used for the randomization of clinics (TB caseload and type of facility) were taken into account in the analysis.
In all 853 patients recruited to the study were used in the analysis of the primary and secondary outcomes. Patients were analysed as a member of the clinic in which he/she was recruited.
The primary outcome was measured at three time points: baseline, three and at six months. If a patient dropped out, and is not present on the day of the interview or refuses to answer questions the primary outcome at the end point of the trial was missing. Therefore, except for the baseline measurement, no post-randomization information was available for these participants. The extent of the missing component was 29% at six months. The method used to take account of the stratified cluster trial design, the repeated binary and linear nature of the primary and secondary outcome (Risky drinking, TB treatment outcome) and the missing data at follow-up is a generalized estimation equations (GEE) approach . The total AUDIT score had to be log-transformed [x’ = log(x + 1)] beforehand, with the aim of making the distribution less skewed. The logarithmically transformed AUDIT score was compared between groups by Student t test. Because the randomization had been on clinic level, and to correct for baseline differences between the two groups, multilevel logistic regression was performed for binominal, and multilevel linear regression for continuous variables. Estimated treatment effects are reported with 95% confidence intervals. Descriptive statistics were calculated for baseline and follow-up.