Mathematical modelling approach to estimating TB burden in children Pete Dodd (University of Sheffield) & (Imperial College London) Wednesday, 1 April 215 Health Economics & Decision Science School of Health & Related Research University of Sheffield 1
Overview Goal: Circumvent potential shortcomings in paediatric notification data by mathematical modelling starting from adult data. Global estimate (c.f. 22 HBCs in published article.) Two modelling steps:.1 Relate adult prevalence to infection risk.2 Model progression from infection to disease Uncertainty in knowledge of each ingredient included. 2
Bird's eye view TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude Ovals = models; diamonds = data inputs; squares = numbers. 3
TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 4
Number (thousands) 5 4 3 2 1 9 6 3 4 3 2 1 12 8 4 75 5 25 Afghanistan Bangladesh Brazil Cambodia China e+ Democratic Republic of Congo Ethiopia India Indonesia Kenya 125 25 1 6 1 2 75 15 4 5 5 1 2 25 5 Mozambique Myanmar Nigeria Pakistan Philippines 5 2 4 9 2 15 3 2 1 6 1 1 5 3 Russian Federation South Africa Thailand Uganda United Republic of Tanzania 6 8 4 4 6 4 4 2 2 2 2 Viet Nam 4 5 9 1 14 15 19 2 24 25 29 3 34 35 39 4 44 45 49 5 54 55 59 6 64 65 69 7 74 75 79 8 15 1 5 2 15 1 5 Zimbabwe 4 5 9 1 14 15 19 2 24 25 29 3 34 35 39 4 44 45 49 5 54 55 59 6 64 65 69 7 74 75 79 8 15 1 5 Data for these results from 213, for all countries reported. 21 Age 15 1 5 1e+5 5e+4 data: UN ESA 5
TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 6
Zimbabwe 21 Viet Nam United Republic of Tanzania Uganda Thailand South Africa Russian Federation Philippines Pakistan Nigeria country Myanmar Mozambique Kenya Indonesia India Ethiopia Democratic Republic of Congo China Cambodia Brazil Bangladesh Afghanistan 5 1 per 1, Data for these results from 213, for all countries reported. data: WHO 7
TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 8
density.2.15.1.5. 5 1 15 Transmission parameter, β A model of community infection, via an updated Styblo's rule mainly based on a review by Bourdin Trunz et al. 9
TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 1
Age-dependent risks of disease following infection Separated by 5 age groups and type of disease: age quantity median LQ UQ probability of disease.5.298.72 1 probability of disease.215.18.36 2-4 probability of disease.16.2.64 5-9 probability of disease.1..13 1-14 probability of disease.11.43.219 probability disease is EP.255.112.451 1 probability disease is EP.295.17.557 2-4 probability disease is EP.6.17.145 5-9 probability disease is EP.85.29.183 1-14 probability disease is EP...8 distributions based on Marais et al., 24 review of the pre-chemotherapy literature. 11
TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 12
Crude approach UNAIDS under-15 prevalence, not disaggregating by age, ART-status or CD4 count. Single IRR based on a few South African papers: quantity median LQ UQ IRR: TB given infection 2. 1.118 39.261 13
TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 14
Effect Data Potential variation in efficacy by latitude (up to 41% of efficacy) Greater protection against extrapulmonary disease (Rodrigues et al., Colditz et al.) quantity median LQ UQ protection for PTB 54% 38% 69% protection for EPTB 7% 52% 84% Coverage as WHO vaccination coverage estimates from 213. Country latitudes as barycentre of country. 15
Countries included Figure: Countries included with WHO estimates of per capita TB, 213. Matching across all datasets leaves 18 countries, total population > 7 billion. 16
By type & age TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 17
By type & age 4 5 14 PTB type DTB age 18
TB TB prevalence Incident EPTB Numbers at risk Model of exposure Infection probability of progression to disease TB Incident PTB data latitude 19
TB, total 1.2 1 6 density 8 1 7 4 1 7 LAT lat nolat 1 + 5, 1,, 1,5, 2,, paediatric TB (per year) LAT median LQ UQ lat 828,56 549,42 1,243,878 nolat 593,14 379,758 912,563 2
TB, 2 1 6 1.5 1 6 density 1 1 6 5 1 7 1 + 2 1 6 1.5 1 6 1 1 6 4 5 14 LAT lat nolat. age LAT median LQ UQ -4 lat 423,585 27,191 635,439-4 nolat 289,245 177,65 459,425 5-14 lat 352,835 171,139 657,833 5-14 nolat 254,32 118,63 489,681 5 1 7 1 + 25, 5, 75, 1,, paediatric TB (per year) 21
TB, by country country India Nigeria Pakistan China South Africa Indonesia Democratic Republic of the Congo Bangladesh Philippines Ethiopia Mozambique Cameroon Myanmar Kenya Angola United Republic of Tanzania Madagascar Zimbabwe Afghanistan Zambia Somalia Uganda Viet Nam Sudan Canada Brazil Côte d'ivoire Russian Federation Democratic People's Republic of Korea Niger 5, 1, 15, 2, paediatric TB (per year) bcg 8 6 4 data: WHO 22
23 Proportion in children..1.2.3 5 1 TB proportion odds 1 5 1 4 6 4 2 1 1 TB logit(proportion)
M.tb infection. M.tb. infection age group median LQ UQ prevalence -14 62,896,959 48,897,936 8,649,253-14 8,997,41 6,936,881 11,696,688 prevalence -4 7,861,145 6,8,713 1,173,695-4 3,191,81 2,459,671 4,145,84 prevalence 5-14 55,63,579 42,84,117 7,5,851 5-14 5,89,956 4,477,617 7,553,58 24
Comparison with notifications Notifications (log scale) 1, 1 1 US UK 1 1 1, Model prediction (log scale) sqrt(gdp) 3 2 1 Overall CDR 37% 25
Adult TB estimates used as starting point - all limitations inherited. Homogeneous mixing assumption - especially poor where TB is rare. efficacy data difficult to interpret. Crude treatment of. & others 26
systematic review of evidence around effects of /ART drug resistance mortality & morbidity interventions uncertainty 27
. Mechanistic model of TB infection and disease in children Provides results on & prevalence of infection & of disease (by type): quantity measure median LQ UQ infection /yr 9. million 6.9 million 11.7 million infection prevalence 62.9 million 48.9 million 8.6 million disease (lat) /yr 827, 549, 1,245, disease (nolat) /yr 593, 379, 914, Table: Global model estimates for 213. 28