Table 1

Sensitivity analysis: the association between baseline ML ratio and TB outcomes holds across TB outcome grouping
Endpoint Number (cases/at risk) Multivariable regression adjusted hazard ratio (95% CI)a Pvalue
Primary endpoint: TB diseaseb or death in HIV + infants; and TB diseaseb, death or latent MTB infection in HEU infants 187/1336 1.17 (1.01 to 1.34) 0.03
TB diseaseb or death 166c/1336 1.18 (1.02 to 1.37) 0.02
TB diseaseb or latent MTB infection 152/1336 1.21 (1.04 to 1.41) 0.02
TB diseaseb 126/1336 1.23 (1.04 to 1.45) 0.02
Probable and definite TB only 49/1336 1.50 (1.19 to 1.89) 0.006
Death 50/1336 1.25 (0.99 to 1.57) 0.06
Latent MTB infectiond 33/725 1.00 (0.66 to 1.5) 0.99

aEstimated from Cox Proportional Hazard regression models stratified by infant HIV status, modeling ML as linear variable following fractional polynomial exploration, adjusted for weight-for-age z-score, cART receipt at baseline, maternal history of TB, ever breastfed and randomized INH prophylaxis/placebo group. Hazard ratios are per unit increase in the ML ratio. bPossible, probable or definite TB. cParticipants who had TB disease (possible, probable or definite TB) and subsequently died were only counted once, hence this number is not the sum of the TB disease and death numerators below. dLatent MTB infection was only assessed in HEU infants. Text in bold highlights statistically significant findings. cART, combination antiretroviral therapy; CI, confidence interval; HEU, HIV-exposed uninfected; INH, isoniazid; MTB, Mycobacterium tuberculosis; TB, tuberculosis.

Naranbhai et al.

Naranbhai et al. BMC Medicine 2014 12:120   doi:10.1186/s12916-014-0120-7

Open Data