The asymptomatic TB evidence gap: Evidence required to evaluate the use of bacteriologically-confirmed asymptomatic TB as a primary endpoint in prevention of disease TB vaccine licensure trials

Abstract

Current licensure trials of new vaccines to prevent Tuberculosis (TB) disease use bacteriologically-confirmed symptomatic TB disease (sTB) as the primary endpoint. Globally, the incidence of sTB is relatively low, making licensure trials large, long and expensive. New data suggest bacteriologically-confirmed asymptomatic TB (aTB) may occur more frequently than sTB. Therefore, if vaccines have efficacy against aTB, TB vaccine licensure trials could be smaller/shorter by including aTB in the primary endpoint.

We describe potential benefits and risks of including aTB in the primary endpoint of TB vaccine licensure trials. We simulate licensure trial endpoint accrual. We summarise feedback from anonymous regulators and policymakers on what they need to know about aTB to consider this proposal, and research studies to fill these evidence gaps.

Research studies are needed to evaluate if bacteriologically-confirmed aTB could be included in primary endpoint of TB licensure trials. If so, it could lead to cheaper/more rapid TB vaccine development.

Introduction

Tuberculosis (TB) kills more people globally than any other infectious disease due to a single pathogen (1). New TB vaccines could contribute towards addressing this pandemic. We do not have a correlate of protection to use as a surrogate endpoint for TB prevention of disease vaccine licensure trials. Therefore, vaccine licensure trials use bacteriologically-confirmed symptomatic TB, which is associated with risk of TB-related morbidity, mortality and post-TB lung and cardiovascular sequelae, as the primary efficacy endpoint. However, sTB occurs only in 5–10% of those infected with Mycobacterium tuberculosis (Mtb), and after months or years (2), making licensure trials large, long, and expensive (3).

It is increasingly recognised that Mtb invokes a range of conditions, from infection to symptomatic disease (4–9). These manifestations include infectious and non-infectious, and asymptomatic and symptomatic TB. Following recent WHO recommendations (10), we define bacteriologically-confirmed asymptomatic TB (aTB) as “A person with bacteriologically-confirmed TB who did not report symptoms suggestive of TB during screening”, bacteriologically-confirmed symptomatic TB as “A person with bacteriologically-confirmed TB who did report symptoms suggestive of TB during screening” (sTB), and bacteriologically-unconfirmed asymptomatic TB as “A person with bacteriologically-unconfirmed TB who did not report symptoms suggestive of TB during screening”. Further, in line with the recent ICE-TB definitions, we assume individuals with bacteriological-confirmed TB are more likely to be infectious, and individuals with bacteriological-unconfirmed TB are less likely to be infectious (4).

A recent analysis of historical and contemporary data suggests aTB occurs more frequently than sTB (6), raising the possibility that, if a vaccine candidate also has similar efficacy against aTB, vaccine licensure trials could be smaller or shorter, relative to trials against sTB, by including aTB (detected by sputum testing for Mtb of all participants), in the primary endpoint. This is important given the extremely large cost of these trials, e.g., the current M72:AS01_E-4 Ph3 trial is estimated to cost $550m and is enrolling 20,000 participants (3).

This paper focuses on bacteriologically-confirmed pulmonary (and therefore likely infectious) aTB as a new potential endpoint in vaccine development. The evidence required to support regulatory acceptance of bacteriologically-unconfirmed pulmonary aTB, which lacks evidence of the causative pathogen, is likely to be much more difficult to obtain.

First, we describe some key potential benefits and risks of including aTB in the primary endpoint of a licensure trial. Then, we summarise the anonymous feedback from a small selection of regulators and policymakers (global and country) who were consulted and willing to comment on what they would need to know about aTB for it to be used as an endpoint for licensure and policymaking. Then, we describe data on the characteristics of endpoints similar to aTB. Finally, we discuss the research studies that could be completed to collect the key data needed on aTB.

Potential benefits and risks of including bacteriologically-confirmed asymptomatic TB (aTB) in the primary endpoint for TB vaccine licensure trials

We illustrate two benefits and two risks of including aTB in the primary endpoint for TB vaccine licensure trials. The potential benefits include increasing the number of trial endpoints in a given trial population over a defined period of time (if aTB is more frequent than sTB (6)), enabling a reduction in the size or duration of licensure trials, and providing an opportunity to treat participants before they develop symptoms. The potential risks include not licensing a useful vaccine against sTB because trial underpowered for sTB endpoint (if the vaccine does not have efficacy against aTB), and potential overtreatment of aTB, which may self-cure (6).

Potential benefits - more endpoints enabling smaller/shorter licensure trials, and providing treatment before symptoms develop

aTB may occur more frequently than sTB, and therefore, may reduce the size or duration of TB vaccine licensure trials if the vaccine also has efficacy on aTB (6). Figure 1 shows two scenarios of modelling estimates of the number of sTB, aTB, and total trial endpoints, for the control arm of a three-year prevention of disease vaccine trial in a high disease burden setting. Full methods are in the supporting material, but briefly, we used a compartmental deterministic transmission model of Mtb infection and non-infectious TB, aTB, and sTB, employing the limited available data on transition rates between TB states from a systematic review (11, 12), and calibrated to a steady state trial population with sTB incidence of 300/100,000/year. This model assumed individuals enter bacteriologically-confirmed, and infectious, states via aTB (Figure S1) and estimated that 48% (95% uncertainty interval: 41–54%) of individuals with aTB progressed to sTB. Here we assume sTB is symptom-screen positive, bacteriologically confirmed, infectious, pulmonary TB, diagnosed with sputum obtained before initiation of treatment for TB and confirmed by at least two bacteriologic tests. Asymptomatic TB is defined as sTB, but symptom-screen negative and less infectious (see supporting material).

Figure 1 -. Estimated number of trial endpoints in the control arm of a prevention of disease vaccine trial in a high disease burden setting, for bacteriologically-confirmed symptomatic TB (sTB) endpoints only (first scenario, top) and combined bacteriologically-confirmed asymptomatic TB (aTB) and sTB endpoints (second scenario, bottom).

Scenario assumes: sTB incidence before trial screening = 300/100k/year and following 10,000 individuals over three years. Prevalent sTB and aTB are screened out in month zero. Active screening every six months for sTB (top) and aTB and sTB (bottom); Participants self-presenting with symptoms suggestive of TB would also be investigated for TB.

The first scenario (Figure 1, top graph) shows a conventional TB vaccine licensure trial (eg the current M72/AS01_E-4 phase 3 trial (13) in which sTB and aTB are screened out in month zero pre-randomisation. During follow up, participants are actively screened for symptoms suggestive of TB (here every six months) and investigated for TB if they have one or more symptoms, to detect and treat participants with sTB. Participants self-presenting with symptoms suggestive of TB would also be investigated for TB.

The second scenario (Figure 1, bottom graph) shows an alternative design in which sTB and aTB are also screened out in month zero, but the follow-up aims to detect and treat both sTB and aTB by collecting sputum from all participants, regardless of symptoms, at regular intervals, here every six months (14). Participants self-presenting with symptoms suggestive of TB would also be investigated for TB.

In the first scenario with a sTB endpoint, we estimated that a cohort of 10,000 individuals in the trial control arm would accrue around 68 sTB endpoints over three years. In the second scenario, using a combined aTB + sTB endpoint, we estimate the same control arm trial population would accrue 151 endpoints over three years, 33 sTB and 118 aTB endpoints (in total, 2.2 times as many as in first scenario). The number of sTB endpoints in the second scenario is lower because some cases of sTB were prevented by detecting and treating individuals with aTB, thereby preventing progression to sTB.

If true, and the vaccine also has 50% efficacy against aTB and sTB, including both sTB and aTB endpoints might allow for trials that are ~45% shorter or that enrol about half as many participants (Figure 2). For example, based on the simulation shown in Figure 2, there would be 93% power to detect a vaccine with an efficacy 95% CI lower bound greater than 0% after 18 months of follow-up for the combined sTB + aTB endpoint; for comparison, a trial with only sTB endpoints would require 33 months to achieve the same power. If the vaccine had lower or higher efficacy against aTB, then the power of the study would be lower or higher, respectively.

Figure 2. Precision (a) and power (b) for estimating vaccine efficacy (VE) from a trial with a bacteriologically-confirmed symptomatic only (sTB) endpoint, versus a composite bacteriologically-confirmed sTB + bacteriologically-confirmed asymptomatic (aTB) endpoint.

Based on simulations accruing 68 TB endpoints in a trial with surveillance for sTB only, versus 151 endpoints for a trial with a composite sTB + aTB endpoint. Precision (a) and power (b) for VE were estimated assuming a vaccine with 50% VE against sTB and aTB. (a) Shaded regions show the 95% confidence interval around VE = 50% for the sTB (pink shading) and sTB + aTB (purple shading) endpoint definitions, as a function of follow-up time. The number of sTB only, and sTB + aTB endpoints (upper x-axes) accrue with follow-up time at different rates. (b) Solid lines indicate statistical power to detect VE > 0% as a function of follow-up time and the number of accrued endpoints in the trial with sTB only (pink line) versus sTB + aTB composite (purple line) endpoints (Fisher exact test, two-sided alternative, alpha=0.05).

We also modelled a third scenario (see supporting results) of an intermediate design identical to the first scenario (sTB only), except that aTB is detected only in the final round by collecting sputum from all participants regardless of symptoms (14). In this scenario, we estimated accrual of 97 endpoints, 68 sTB and 29 aTB (1.4 times as many as in first scenario).

A second potential benefit of a trial powered to include aTB endpoints could be a more ethical approach, by providing treatment to participants before they develop symptoms.

Potential risks – not licensing a useful vaccine and overtreatment

Including aTB in the primary endpoint for TB vaccine licensure trials also has risks, most importantly failing to detect efficacy for a vaccine that has higher efficacy against sTB but lower efficacy against aTB. Table 1 summarises different theoretical scenarios of vaccines with high or low efficacy against aTB and sTB assuming sTB prevention is the primary goal, the potential outcome of a trial that used a composite or sTB endpoint, and whether the licensing decision would likely be correct or not.

Table 1. Different theoretical scenarios of vaccines with high (>60%) or low (<30%) efficacy against bacteriologically-confirmed asymptomatic TB (aTB) and bacteriologically-confirmed symptomatic TB (sTB) assuming sTB prevention is primary goal, the potential outcome of a trial that used a composite or bacteriologically-confirmed sTB endpoint, and whether the licensing decision would likely be correct or not.

Trial is assumed to be powered to detect at least 50% efficacy on selected endpoint.

Scenario	Assumed true vaccine efficacy		Trial outcome using composite or sTB-only endpoint		Likely consequence for licensure of using composite endpoint
	aTB	sTB	Composite aTB+sTB	sTB
A	<30%	<30%	True negative	True negative	Correct decision - candidate not licensed.
B	>60%	<30%	False positive	True negative	May or may not be correct decision - candidate licensed.
C	<30%	>60%	False negative	True positive	Not correct decision - candidate not licensed.
D	>60%	>60%	True positive	True positive	Correct decision - candidate licensed.

Scenarios A (joint low efficacy) and D (joint high efficacy) would likely lead to the correct decision being taken, with a vaccine with low efficacy against sTB not being licensed, and a vaccine with high efficacy against sTB being licensed.

However, scenario B and C may lead to the incorrect licensure decision being made. In scenario B (high aTB, low sTB efficacy), using a composite endpoint could result in a vaccine being licensed that has high efficacy against aTB but low efficacy on sTB. This may turn out to be a correct decision, if aTB is later found to cause substantial direct morbidity or transmission, and would have been found more efficiently. However, it would have been an incorrect decision if aTB does not cause substantial direct morbidity or transmission and does not prevent sTB. This outcome may also discourage funding and research efforts to develop a vaccine that would have greater efficacy against sTB because it would be thought that we have found a good vaccine, and it would take many years to know that this was not the case.

In scenario C (low aTB, high sTB efficacy) using a composite endpoint may result in a vaccine not being licensed that has low efficacy against aTB but high efficacy against sTB, because the trial would be underpowered for efficacy on sTB alone. If we had run a trial powered against sTB only, we would have found the vaccine we need (but that trial may be much larger and expensive). One approach to reduce the risk of this scenario in a phase III licensure trial, would be to use a study design in the preceding phase IIb trial that would allow efficacy of the vaccine in preventing aTB and sTB to be assessed independently of one another. In the phase IIb trial, in addition to actively screening for symptoms suggestive of TB during and at the end of follow up, a sputum specimen would be collected from all participants at each study visit during and at the end of follow up. The sputum specimens would be stored and tested at the end of follow up to identify participants with aTB. This design has previously been described and has precedent (14). Although this approach has challenges, it is likely to be feasible and acceptable given the experience from other trials (14). In participants that are not able to spontaneously produce sputum, consideration may be given to using sputum induction with saline nebulisations, teaching of huffing and puffing techniques, and other alternative methods of collecting respiratory specimens once approved, such as use of tongue swabs, lung flute and breath condensate. The use of non-sputum-based diagnostic tests, such as urine tests for lipoarabinomannan and tongue swabs, are not currently recommended for screening in general populations. Once highly specific non-sputum-based TB diagnostics are approved they can potentially be included in the endpoint definition of TB disease.

Collecting and storing sputum specimens for processing at the end of the trial will add to the cost of the trial, which may be justifiable if the approach derisks future phase III trials. If the phase IIb trial suggests similar efficacy in preventing sTB and aTB, then a composite endpoint could be considered in the phase III trial. If the phase IIb trial shows greater efficacy in preventing sTB versus aTB, then a composite endpoint may not be used, and the phase III trial could be powered to only show efficacy in preventing sTB.

Overall, all scenarios benefit from running a faster, more resource-efficient trial (albeit offset by the need to collect more sputa) using a composite endpoint. Scenarios A and D illustrate why a composite endpoint may get us to the correct licensure decision more efficiently, whilst scenarios B and C illustrate how using a composite endpoint may result in an incorrect licensure decision being made.

A second potential risk is of overtreatment of individuals with aTB found and treated during active screening, which may be unethical. Around 50% of aTB episodes may resolve spontaneously (6), and as the morbidity of aTB is unknown, the individual benefits of treating aTB may be outweighed by the individual treatment side effects, costs, and inconvenience.

Further, is it worth noting that symptom thresholds are somewhat subjective, and symptomatic TB detected by repeated active symptom-based screening in TB vaccine trials is likely to have less extensive disease than symptomatic TB detected through self-presentation with symptoms to health services. However, in clinical vaccine trials, the symptom threshold is more rigorously standardised than in routine health services, and there is evidence to suggest that asymptomatic TB remains present after screening for symptomatic TB during trials. In the ACT3 trial that did active, community-based TB screening using symptoms and Xpert in Vietnam, TB symptoms were not readily reported by many people who were later shown to have bacteriologically confirmed TB (personal communication, Dr G Fox), and in the M72 trial, much of the TB was more extensive than would be anticipated among asymptomatic TB cases detected by symptom-agnostic screening (15).

Also, as with all mathematical modelling, these results are uncertain. In particular, we may be over or underestimating the number of aTB endpoints because the probability of progression of aTB has not been empirically measured, and whether the benefit of overtreatment outweigh the costs, toxicity and inconvenience is unknown because the morbidity of aTB has not been empirically measured. See supporting material for more discussion. This uncertainty supports the main message of our manuscript – we urgently we need better data on aTB.

Regulator and policymaker perspectives on including bacteriologically-confirmed asymptomatic TB (aTB) in licensure trial endpoint

To start with the end in mind, we need to understand what key decision-makers would need to know to base decisions on the results of a prevention of disease TB vaccine licensure trial that included aTB in the primary endpoint. We explored two perspectives: the off-the-record views of regulators, and the views of global and country vaccine policymakers. We informally discussed with senior staff from three different international and high-TB-burden country regulators, two vaccine policymakers from different high-burden countries, and reviewed the published vaccine investment strategy criteria from GAVI (16). The number of these discussions was small, and they are not representative, but they illustrate a range of views held by current senior vaccine decision makers

Figure 3 summarises their feedback. Regulators were relatively consistent in their feedback. With their focus on individual benefit/risk, they said aTB needs to be reliably measurable using existing tools. This need is already met, as aTB is similar to sTB in terms of diagnostic methods, just symptom-screen negative. Regulators also said that aTB would also need to cause either significant morbidity in and of itself (the rationale for using sTB for existing vaccine licensure), or there would need to be a significant probability that aTB predicts the development of sTB. Finally, regulators also said that before considering including aTB in the endpoint for TB vaccine licensure trials, they would need to know if treatment of aTB prevented sTB. These criteria are all consistent with the FDA’s guidelines for accelerated approval of new drugs, requiring that a surrogate endpoint (such as aTB) be “reasonable likely to predict clinical benefit” (17, 18). Note, there is precedence for these rationale for vaccine licensure with the initial licensing of HPV vaccines, which relied on evidence of efficacy against moderate or high-grade cervical intraepithelial neoplasias (CIN2/3) as opposed to HPV-associated cervical cancer (19). Importantly, there was sufficient evidence that CIN2/3 was predictive of cervical cancer, and that removal of CIN2/3 lesions could prevent further progression (20, 21).

Figure 3 – Visual summary of the evidence needs from small anonymous convenience sample of senior regulators and global and country vaccine policymakers.

Tx = treatment. Explored two perspectives: the off-the-record views of regulators and the views of global and country vaccine policymakers. We informally discussed with senior staff from three different international and high-TB-burden country regulators, two vaccine policymakers from different high-burden countries, and reviewed the published vaccine investment strategy criteria from GAVI (16). The number of these discussions was small, and they are not representative, but they illustrate a range of views held by current senior vaccine decision makers

Given the regulatory focus on individual benefit/risk, our three interviewees said that the potential for reducing Mtb transmission via vaccine-induced prevention of aTB, a population benefit, was not thought to be helpful for a TB vaccine licensing decision. Indeed, historically, transmission impact has never been a primary consideration for any initial vaccine licensure.

Consistent with their broader perspective of overall population benefit, global and country vaccine policymakers said that they needed to know all that regulators needed to know (above), plus they also needed additional evidence. This included impact on transmission (all global and country vaccine policymakers consulted), and also a larger set of health and economic impacts. Two different high-burden country National Immunization Technical Advisory Group members said they would want to know if the vaccine reduced transmission, and one said they would want evidence that it decreased the incidence of paediatric TB. Published guidance shows that GAVI would also need evidence on the population health impact, value for money, equity and social protection impact, and the potential economic impact of the new vaccine (16)

Unfortunately, there are no reliable direct empirical data on aTB morbidity, aTB predicting sTB, aTB treatment preventing sTB, or significant Mtb transmission from aTB.

Evidence on similar endpoints

Although there is a paucity of direct empirical data on these characteristics of aTB, some evidence exists for these characteristics on other types of asymptomatic TB (eg bacteriologically unconfirmed asymptomatic TB) which may support allocating resources to collecting direct empirical data on aTB:

Asymptomatic TB morbidity

South African gold miners are screened at least annually using symptoms, mass miniature radiographs and lung function testing. In a study of gold miners that evaluated the effect of TB on lung function comparing pre and post TB lung function tests, miners whose TB was detected by mass miniature radiographic screening, and were likely to have bacteriologically confirmed or unconfirmed aTB, and those with characteristics associated with aTB, (i.e., less radiological extent of disease, or were smear and or culture negative), had less lung function loss post TB treatment compared to those with sTB (22).

Progression from asymptomatic TB to symptomatic TB

In a recent systematic review and meta-analysis of persons with radiological evidence of TB, negative microbiology, untreated, and with, without or unknown symptoms suggestive of TB, from nine studies, progression from bacteriologically negative to sTB occurred at a rate of 10% per year (11). Among the three studies that included people with bacteriologically unconfirmed asymptomatic TB, the rates of progression to bacteriologically positive TB were similar (range 4–12% per year).

Of note, the initial approval of an HPV vaccine based on prevention of an alternative, pre-cancerous, endpoint provides a precedence and a framework for considering the kinds of evidence that may be needed to seek licensure of a TB vaccine for prevention of endpoints other than symptomatic disease. The probability of progression from HPV CIN3 to cervical cancer was around 30%, representing another point of comparison for a condition that was targeted by vaccination for prevention of its associated disease progression (23). There are few empirical data, but modelling suggests around 50% of aTB progresses to sTB, which could be an over or under estimate (6).

Asymptomatic TB treatment effects on progression to symptomatic TB and mortality

The optimal management of patients with radiological evidence of TB but negative bacteriology, many of whom are asymptomatic, is unknown. A recent systematic review and meta-analysis of TB treatment trials that enrolled patients with radiological evidence of TB and negative sputum cultures, including patients detected by active case finding and likely to have bacteriologically unconfirmed asymptomatic TB, showed that treatment, particularly with multi-drug regimens, significantly reduced the risk of progressing to bacteriologically-confirmed TB disease (24). In one treatment study of people with drug-susceptible aTB (n=75) and sTB (n=308), treatment of patients with aTB, did not result in any individuals with recurrent TB(25).

In a recent scoping review, treatment success in patients with bacteriologically confirmed and unconfirmed aTB versus sTB was similar, and the death rate and risk of recurrence was lower in patients with aTB (25, 26).

Transmission from asymptomatic TB

Mtb is transmitted through respiratory aerosols. Respiratory droplets can be generated in the absence of symptoms, particularly cough, through talking, singing and tidal breathing (27–30). aTB may therefore contribute to Mtb transmission at a population level due to its high prevalence, long duration (31), high sputum-smear positivity (a known strong indicator of infectiousness) (32), and cross-sectional studies suggest it may (33).

Overall, although there is evidence that endpoints similar to aTB do have morbidity, progress to sTB, cause Mtb transmission, and treatment of these similar endpoints prevents progression to sTB, this evidence is not on aTB specifically, and therefore is unlikely to be sufficient for regulators to include aTB in the endpoint of licensure trials. As such the evidence needs of regulators and policymakers leave substantial research gaps, which may need addressing before aTB can be included in a licensure trial endpoint.

Research that could be carried out to collect the key evidence

Table 2 summarises the studies that could address these key evidence gaps on aTB.

Table 2 -. Research that could be carried out to collect the key evidence on bacteriologically-confirmed asymptomatic TB (aTB), ordered approximately from lower to higher anticipated resource requirements.

sTB - bacteriologically-confirmed symptomatic TB. RCT - Randomized controlled trial. HHCs - household contacts (of person with sTB). ACF – active case finding.

Design	Morbidity	aTB predicts progression to sTB	Treatment of aTB prevents sTB	Transmission potential of aTB
Systematic reviews and meta-analyses	X	X	X	X
Secondary analysis of existing data	X	X	X
Cross-sectional studies of symptom-agnostic community-based screening or prevalence surveys	X
Prospective observational cohort studies	X	X	X
Prospective follow-up of HHCs without symptoms suggestive of TB randomised to continued follow-up or TB investigations 3–6 monthly	X	X	X
Randomised trial of treatment or no-treatment of aTB	X	X	X
Cluster randomised trial of symptom-based vs. symptom-agnostic ACF	X			X

Morbidity of aTB

TB morbidity is defined as the direct effect on current health of TB disease, and the future health impact of post TB lung disease. The severe morbidity and mortality of sTB is well established and is a key reason sTB is accepted as a suitable primary endpoint for TB vaccine licensure trials. The key question from regulators for aTB here, is whether aTB has morbidity independent of progression to sTB.

Measures of morbidity include health related quality of life, lung function decline, evidence of inflammation, and the risk of progression to sTB or to developing other respiratory infections and post TB lung and cardiovascular disease.

Determining morbidity related to aTB can be done relatively cost-effectively, as any study that diagnoses aTB real time can include measures of morbidity and compare to individuals with sTB. Morbidity can be evaluated in cross-sectional studies that compare quality-of-life (eg EQ-5D) in persons with aTB to those with sTB or no TB. Examples of such studies include: symptom-agnostic community screening studies, national prevalence surveys, demographic and health surveys, and in settings where community based active case finding using chest radiographic screening is being implemented (34, 35). Persons with aTB, sTB and a sample of those without TB identified in cross-sectional studies could be followed prospectively to evaluate the risk of post TB lung disease. Morbidity (both direct and post TB lung disease) could also be assessed in prospective cohort studies, either nested in randomised controlled trials or natural history studies. With additional resources, more complex measurements of morbidity, e.g., measures of lung function, inflammation, and heart disease, could be made.

aTB progression to sTB

To empirically quantify how aTB predicts sTB we need to measure the proportion of aTB that progresses to sTB. For this, longitudinal data are required. Sputum (or other highly specific approved diagnostic) samples from individuals with bacteriologically confirmed TB but without symptoms suggestive of TB would need to be followed prospectively without treatment and sputum collected, stored and tested at the end of follow up. This approach will allow the natural history of untreated aTB to be better described. The ethical considerations of such studies may be challenging (14). Such studies would only be considered because the morbidity from aTB is currently unknown, and as such the individual benefits of treatment are not obviously greater than the individual side effects of treatment, costs and inconvenience of the treatment. Such studies would need to happen in a highly monitored environment, where detection of sTB is quick and comprehensive through regular symptom screens, thus minimising the risk to participants. It is worth noting that if empirical data show that aTB results in direct harm to the affected individuals, then such studies would no longer be ethical. Therefore, the time window in which these data could be collected may be limited.

Potential study designs include nesting a study of aTB into TB vaccine, TPT, or TB screening RCT's with sTB as the primary endpoint. Sputum would need to be collected on all individuals at scheduled study visits, but not tested for Mtb unless symptoms suggestive of TB were present. At the end of the study, stored samples would be tested, and probabilities of progression and regression could be calculated. The ethical and logistical considerations of such an approach have previously been discussed (14). An alternative design would be an observational study in which asymptomatic patients that may have aTB are closely followed up without treatment. An innovative study design would be to nest an observational study into a routine contact investigation programme. Household contacts without symptoms suggestive of TB at initial screening would be followed up every three to six months over a period of two to three years. At each scheduled visit, household contacts that remain without symptoms would be randomised to investigation for TB, and treatment if positive, or continued follow up without being investigated for TB. Household contacts with symptoms suggestive of TB either at follow up or self-presenting with symptoms between study visits would be investigated for TB and treated if positive. Ethically this would be acceptable as people with aTB would be identified if randomised to investigation or if they develop symptoms compatible with sTB. The two study designs described above would provide data on progression/resolution rates and are likely ethically acceptable.

Asymptomatic TB treatment on progression to sTB

To know if treatment of aTB prevents sTB, individuals with aTB would need to be identified and randomised to treatment or no-treatment, and followed prospectively to compare the risk of developing sTB between trial arms. A cluster randomised trial may also be used. In the control arm, bacteriological testing would be restricted to those reporting symptoms only, leading to only individuals with sTB receiving treatment. In the intervention arm all participants would receive bacteriological testing, regardless of symptoms, and all patients with aTB and sTB would receive treatment. The incidence rates of sTB in the two arms would then be compared, and the difference would be a measure of the impact of treating aTB on sTB incidence. A limitation of the cluster randomised design is that it will be challenging to determine the proportion of sTB that arose due to progression versus transmission. Such studies could only be implemented in settings where symptom-agnostic active case finding is not being implemented.

Transmission from aTB

Finally, global and country vaccine introduction decision makers need to know how much TB Mtb transmission is due to aTB, compared to sTB.(4)

A potential study design would be to identify household and close contacts of adults with aTB, sTB, and control households without TB, to compare the proportion of household contacts with a positive test of infection and the incidence of sTB (36–38). A limitation of this design is that it does not distinguish between infectiousness and the duration of infectiousness. Natural history studies and mathematical modelling could provide insight into the duration of infectiousness of aTB. Other techniques to assess infectiousness could be considered, such as using cough aerosol sampling or cough chambers and using very sensitive assays to identify viable Mtb; however, these have not been validated. The Active Case Finding for Tuberculosis 3 (ACT3) trial compared the effectiveness of symptom agnostic annual community-wide active case-finding using sputum collection and testing with Xpert MTB/RIF to passive case detection alone, in reducing the prevalence of TB disease and infection in Vietnam (39). Only a small proportion of participants reported symptoms suggestive of TB at each screening visit. The intervention identified and treated persons with sTB and aTB. After three years, the intervention, compared to the control arm, had reduced the prevalence of bacteriologically confirmed TB disease in adolescents and adults and Mtb infection in children. The ACT3 results suggest that repeated rounds of active case finding to detect both sTB and aTB reduced Mtb transmission at a population level. A definitive study design to evaluate the transmission potential of aTB, would be to compare symptom-agnostic ACF with symptom-based ACF and compare TB incidence rather than TB prevalence as the study outcome (as prevalence reflects both incidence and duration of disease and the intervention may affect duration rather than incidence).

The consistent challenge across these studies is identifying enough persons with aTB, which means screening or monitoring a large number of individuals. However, those costs would likely be dwarfed by the potential savings for even a single multi-country vaccine phase 3 trial. As nine candidates are currently in phase 1–2 studies (40), research that could reduce the costs of licensure trials may yield a good return on investment.

Other issues

Regardless of which endpoint is used as the primary endpoint in TB vaccine licensure trials, it will also be important to measure the vaccine impact on all forms of TB either in secondary endpoints in ph3 trials or ph4 studies, and while we wait for trial data to explore uncertainty using math modelling.

Summary

There are benefits and risks to including aTB in the primary endpoint for TB vaccine licensure trials, but their implications remain uncertain and require careful consideration. There is a need for collecting data to better characterise aTB. Such studies would generate the evidence needed to evaluate if aTB could be included in the primary endpoint of a TB licensure trial and may lead to cheaper and more rapid TB vaccine development. Even if aTB is not judged to be a licensable indication, collecting these data would greatly improve our understanding of aTB that may support important changes in global TB care and prevention policy and practice.

Key Messages.

• TB vaccine licensure trials, using a symptomatic TB as the primary endpoint, are large, long and expensive

• Including asymptomatic TB in the primary endpoint might double the number of endpoints leading to smaller or shorter TB vaccine trials, but it also might not double the number of endpoints, and it may also lead us to miss a useful vaccine

• Crucially, research studies are needed to generate the evidence needed to evaluate if asymptomatic TB could be included in the primary endpoint of TB vaccine licensure trials

• If asymptomatic TB can be included, it could lead to cheaper and more rapid TB vaccine development