Abstract
Background: A common approach to reducing sample sizes for late-stage tuberculosis vaccine trials is to restrict enrolment to interferon-gamma release assay (IGRA)-positive participants to maximize tuberculosis case accrual. The efficiency gain, if any, from this screening strategy is unknown.
Methods: We estimated the age specific IGRA positivity prevalence for transmission levels generally considered in tuberculosis vaccine trials (annual risk of tuberculosis infection [ARTI] 2-6 %) and calculated the expected tuberculosis incidence at each age by IGRA status, using a difference equation model. We modelled scenarios that assumed constant or increasing ARTI during adolescence and differing levels of partial protection afforded by previous Mycobacterium tuberculosis infection. We then estimated sample size requirements for tuberculosis vaccine trials enrolling only IGRA-positive participants or participants without prior IGRA testing ('Mixed' trial). We assumed participants were 15-44 years at enrolment and followed-up for 3 years.
Results: Estimated tuberculosis incidence was 4.7 times higher in IGRA-positive compared to IGRA-negative participants at age 15 years, but 0.9 times lower at age 44 years (assuming ARTI 4 %). This age-cohort effect was exacerbated when assuming partial protection and attenuated when assuming increasing ARTI during adolescence. In a model that included both these assumptions, the sample size required for a Mixed trial compared to that for an IGRA-positive participants-only trial was 124 % larger at 2 % ARTI, 36 % larger at 4 % ARTI but only 8 % larger at 6 % ARTI. Prioritizing enrolment of participants aged 15-29 years improved sample size efficiency for an IGRA-positive participants-only trial. These results were largely unaffected by our model assumptions.
Conclusion: In late-stage tuberculosis vaccine trials among adults and adolescents, pre-enrolment screening by IGRA testing provides a large sample size efficiency when M. tuberculosis transmission levels are relatively low, but modest or no sample size benefits at high transmission levels.