Table 3.

Estimates of hazard ratios from marginal structural models of the main analyses of the effect on all-cause mortality of current treatment for pulmonary tuberculosis (PTB) at time of HAART initiation, among 7512 individuals initiating HAART in Johannesburg, South Africa between 1 April 2004 and 31 March 2007.

	Main exposure	Crude (unweighted) estimates		Adjusted for confounding and censoring
		HR	95% CI	HR	95% CI
Model 1	AllPTB	1.71	1.31-2.23	1.06	0.75-1.49
Model 2	> 120 days	1.89	1.17-3.06	1.25	0.65-2.43
	61-120 days	1.34	0.80-2.23	0.89	0.48-1.63
	31-60 days	1.62	0.97-2.72	1.08	0.59-1.98
	≤ 30 days	2.02	1.31-3.11	1.28	0.78-2.10

Model 1estimates the effect of prevalent, treated PTB on outcome of mortality. Model 2 estimates the effect of the same exposure by length delay between initiation of PTB treatment and initiation of HAART. Both models adjust for both confounding by baseline factors and LTFU using inverse probability of exposure and censoring weights; components and specification of weight models are described in the Appendix.