Table 3.

Association of baseline biomarker levels with baseline microbiological characteristics.

[A]
Smear grade > 1 at baseline
	Unadjusted model			Adjusted (demographics)
Assay	Est (95%CI)	P value	AUC	Est (95%CI)	P value	AUC
SAA1	1.68 (1.45, 1.95)	0.0004	0.62	1.52 (1.32, 1.74)	0.0027	0.60
IL-1β	1.57 (1.38, 1.78)	0.0004	0.62	1.39 (1.24, 1.56)	0.0033	0.60
IL-6	1.52 (1.34, 1.72)	0.0008	0.62	1.37 (1.22, 1.54)	0.0060	0.59
PTX-3	1.49 (1.36, 1.63)	0.0000	0.64	1.37 (1.26, 1.48)	0.0002	0.63
TNF-RI	1.19 (1.14, 1.24)	0.0001	0.62	1.15 (1.11, 1.20)	0.0006	0.61

[B]
MGIT time-to-detection
	Unadjusted model			Adjusted (demographics)
Assay	Est (95%CI)	P value	AUC	Est (95%CI)	P value	AUC
IFN-γ	1.82 (1.56, 2.12)	0.0001	0.67	1.48 (1.28, 1.70)	0.0062	0.63
IL-15	1.27 (1.19, 1.34)	0.0001	0.67	1.19 (1.12, 1.27)	0.0038	0.63

The estimated association (Est) is 10^coeff, where coeff is the coefficient of a linear model for the effect of the indicated baseline microbiological characteristics on log₁₀ transformed baseline biomarker levels. The Est value can be interpreted as the expected factor change in baseline biomarker level for patients with [A] baseline smear grade > 1 relative to patients with smear grade < 1; [B] baseline MGIT time-to-detection ≤ 5 days relative to patients with MGIT time-to-detection > 5 days. The tables show values for an unadjusted model as well as a model that is adjusted for demographic covariates (gender, age, BMI, HIV status, region (Africa vs. Not-Africa) and study arm). The AUC value for ROC curves is a non-parametric indicator of effect size, comparing the distributions of biomarker ratios for the two outcome classes (either unadjusted or adjusted for the covariates). Only biomarkers with statistically significant associations (p < 0.05/62 assays = 0.0008) in at least one of the models are shown.