Estimation of annual risk of Mycobacterium tuberculosis infection accounting for tuberculin skin test reversion: A mathematical modelling approach

Abstract

Antecedentes: Se estima que una cuarta parte de la población mundial está infectada con Mycobacterium tuberculosis. El riesgo anual de infección (ARI) se calcula con la prevalencia de M.tuberculosis utilizando la prueba inmunorreactiva cutánea de la tuberculina (TST). Sin embargo, el ARI generalmente se estima asumiendo que la inmunorreactividad es persistente. Métodos: estudio de modelado matemático basado en un ejercicio teórico por Sutherland, que explora el efecto de los niveles variables de reversión de TST en la estimación del ARI. El modelo se amplió mediante el uso de tasas empíricas de reversión por edad. Luego, se aplicó el modelo para volver a estimar las ARI de encuestas publicadas, ajustando la positividad de TST observada. Resultados: Las tasas constantes de reversión de TST de más del 1% tuvieron un efecto significativo en la prevalencia de inmunorreactividad, disminuyendo la prevalencia en un 9%; tasas de reversión del 10% redujeron la prevalencia en un 55% a los 19 años. Cuando se ajustó con la reversión, el modelo mostró que las ARI reestimadas eran entre 50-450% más altas que las derivadas de las encuestas de TST. Conclusiones: La estimación de ARI sin tener en cuenta la reversión subestimará sistemáticamente el riesgo de infección verdadero.

Background: Over one quarter of the global population is estimated to be infected with Mycobacterium tuberculosis. One of the key metrics is the Annual Risk of Infection (ARI), derived from M. tuberculosis prevalence data from surveys using the immunoreactive Tuberculin Skin Test (TST). However, the ARI is generally estimated using the dual assumptions of lifelong viable infection as well as persistent immunoreactivity, both of which have been challenged. The study will explore the implications of TST-reversion on ARI estimates. Methods: This mathematical modelling study is building upon an existing theoretical exercise by Ian Sutherland, exploring the effect of the varying levels of TSTreversion on the estimate of the ARI. The model was expanded by using empirically estimated age-specific TST-reversion rates, daily timesteps and cubic spline interpolation. Uncertainty was estimated by determining confidence intervals for the TST-reversion proportion. The model was then applied to manually re-estimate ARIs from published TST surveys in Vietnam (Hoa et al.) and South Africa (Wood et al.), by manually fitting observed TST-positivity. Sensitivity analyses included multiple TST-reversion rates and the use of linear interpolation for the ascertainment of TST-positive estimates. Results: Constant TST-reversion rates of over 1% had a significant effect on TST-positive prevalence decreasing prevalence by 9%; TST-reversion rates of 10% decreased prevalence by 55% by age 19. When fitted with reversion, the model showed that re-estimated ARIs were 50% to 450% higher than those derived in the TST surveys. Sensitivity analyses also produced similar results. Conclusions: Estimation of ARI from TST data without accounting for reversion will consistently underestimate the risk of infection. If we are to understand transmission, we will need to incorporate reversion into our estimates of ARI to facilitate insights into the population at-risk and cost-benefit applications.