Skip to main content

TB-RROC Expert Elicitation

Introduction

Welcome to the TB-RROC Expert Elicitation. Thank-you for agreeing to participate. This short exercise aims to quantify your beliefs about the treatment effects of the TB-RROC intervention, which will be described on the next page.

The results of this elicitation along with the results from other participants will be used in a Bayesian analysis of the trial. Within a Bayesian framework probability is interpreted as subjective uncertainty: how sure we are about the value of a parameter. This prior belief about the value of the parameter can be updated with new data, in this case the data from the trial. You may use external sources of evidence to inform your beliefs.

You will be provided with information about the trial and the intervention. And then the process to elicit you beliefs will be explained using a training exercise. Two round of elicitation will then take place. This should take no more than 20 minutes. When you are ready, please click next.

Background

The problem

Patients on treatment for multidrug-resistant tuberculosis (TB) and those receiving the standardised WHO Category II retreatment regimen require long courses of injectable anti-tuberculous agents. Although there are few data supporting the use of streptomycin in the standardised Category II regimen, and injection-free regimens for MDR-TB are being evaluated, it is likely that the use of injectable agents is still going to be necessary for a proportion of patients for some time to come. There are various options for the delivery of long-term daily injectables as part of TB treatment, but each has major disadvantages:

  • Prolonged admission to hospital is associated with hospital-acquired infections; social and economic consequences for patients and their households; and a disproportionate burden on health systems.
  • A number of community-based approaches to care have been implemented, but once again these often face barriers to success. A system which involves healthcare workers travelling daily to patients at home in order to administer injections incurs additional expense and healthcare worker hours. The alternative is for patients to travel daily to a health facility, but this model simply transfers the time and financial burdens onto already vulnerable health system users.
If current management of MDR and ‘retreatment’ TB is to be successful, sustainable models for the delivery of long-term injectables need to be evaluated.

The intervention

The TB-RROC (TB Retreatment Regimen Outcomes and Care) Trial aimed to assess a novel method of community-based management for adult patients with recurrent TB in Malawi. The intervention involved training lay people (guardians) to administer streptomycin injections to patients receiving retreatment regimen in the community.

The evaluation

The study was a pragmatic, individually randomised controlled clinical trial conducted across two sites in Malawi – Queen Elizabeth Central Hospital in Blantyre and Bwaila Hospital in Lilongwe. Patients starting retreatment regimen at each site were approached and asked to identify a guardian. Consent to participate was obtained from both patient and guardian. Once enrolled, guardians were trained in the technique of intramuscular injection, and evaluated using a structured competency assessment. Once patients were medically fit for discharge, and their guardian had passed the competency assessment, participants were randomised to receive standard care (which in Malawi consists of hospital admission for the 60 day duration of streptomycin treatment) or community-based care in which guardians administered streptomycin to patients in their homes. Participants in both arms of the study were reviewed at 1, 3, 5 and 7 weeks post randomisation. At each visit complications of treatment, adverse events and adherence were assessed. Patients receiving community-based management who became unwell were evaluated by the study team and readmitted to hospital if necessary. The primary outcome of the study was ‘still alive and on treatment at the end of the 60 day period of streptomycin.’

Training Exercise

You will be asked for five values in order to quantify you beliefs about the parameter of interest. This training exercise will illustrate the process. You will be asked about the effect of a treatment on the outcomes of patients that will be studied in a randomised controlled trial. As an illustration of this, we will ask here about your beliefs regarding an intervention with known efficacy: isoniazid prophylaxis for the prevention of TB in children. Please do not research this intervention - use the knowledge you currently have. When you are ready please click next.

Training Exercise

Plausible Range

You are first asked to provide a plausible range for the value of the relative risk. Outside of this range you would consider it very unlikely to observe the relative risk, not necessarily impossible, but you would be very surprised to observe a relative risk outside of this range.

What would be the lower end of this plausible range? (i.e. you would not expect to see a relative risk below this value):

What would be the upper end of this plausible range? (i.e. you would not expect to see a relative risk above this value):

Please click submit.

Median

Your plausible range is indicated by the two red lines above. Now, please provide you belief about the median. There should be an equal probability of the relative risk falling either side of this value.

What is your median value?:

Interquartile Range

Your plausible range is indicated by the two red lines above and your median by the blue line. Now, we want you to provide your interquartile range. There should be an equal probability of the relative risk lying between the median and upper interquartile range and the upper interquartile range and the upper plausible range. The same goes for the lower interquartile range. The interquartile range values may lie closer to the median than the plausible range limits as values near the centre of the distribution are more likely.

What is the upper value for the interquartile range?:

What is the lower value for the interquartile range?:

The answer according to a recent meta-analysis was that the relative risk associated with isoniazid prophylaxis on TB in children was 0.65. This is shown by the orange square above.

Elicitation Round 1

Now you will be asked about your beliefs regarding the TB-RROC trial.

The relative risk you are asked to think about is the relative risk of a patient being alive and on treatment at two months in the home treatment group compared to the hospital treatment group. A relative risk greater than one indicates a patient in the home treatment group is more likely to be alive and on treatment than a patient in the hospital group.

Plausible Range

You are first asked to provide a plausible range for the value of the relative risk. Outside of this range you would consider it very unlikely to observe the relative risk, not necessarily impossible, but you would be very surprised to observe a relative risk outside of this range.

What would be the lower end of this plausible range? (i.e. you would not expect to see a relative risk below this value):

What would be the upper end of this plausible range? (i.e. you would not expect to see a relative risk above this value):

Please click submit.

Median

Your plausible range is indicated by the two red lines above. Now, please provide you belief about the median. There should be an equal probability of the relative risk falling either side of this value.

What is your median value?:

Interquartile Range

Your plausible range is indicated by the two red lines above and your median by the blue line. Now, we want you to provide your interquartile range. There should be an equal probability of the relative risk lying between the median and upper interquartile range and the upper interquartile range and the upper plausible range. The same goes for the lower interquartile range. The interquartile range values may lie closer to the median than the plausible range limits as values near the centre of the distribution are more likely.

What is the upper value for the interquartile range?:

What is the lower value for the interquartile range?:

Your responses suggest the following:
  • There is an equal probability that the relative risk is less than and greater than . Each has a 25% probability.
  • For every 100 hospital based patients, approximately 90 will be alive and on treatment at two months. Your answers imply that you believe there is a 50/50 chance than of 100 patients treated at home, fewer than will be alive and on treatment at two months.
We will conduct the elicitation a second time to allow you to adjust your responses in light of this feedback. When you are ready press next.