Table 6. Summary of association of use of individual drugs with treatment success in children treated for confirmed MDR-TB (n = 641)^a,b,c,d.

Drug used	N (%)	aORe	95% CI	p-value
Pyrazinamide	599 (93)	1.63g	(0.41–6.56)	0.484
Second-line injectable agentsf, j	584 (91)	2.94g	(1.05–8.28)	0.041
Ethionamide/ prothionamide	590 (92)	2.19	(0.42–11.54)	0.332
Cycloserine/ terizidone	356 (56)	1.66h	(0.91–3.05)	0.104
Clofazimine	23 (4)	0.55	(0.02–19.20)	0.714
High-dose isoniazidj	133 (21)	5.86g	(1.68–20.51)	0.007
Para-aminosalicylic acid	147 (23)	0.70i	(0.25–1.96)	0.483
Clarithromycin	32 (5)	0.29i	(0.05–1.53)	0.132

Treatment success was compared to Failure/Death by drug use.

^aAdjusted estimates for the clinically diagnosed children were not possible due to very low rates of failure.

^bLost to follow-up was excluded from analysis.

^cAll random effects (random intercept and random slope) models used maximum likelihood estimation with quadrature approximation and were specified with an unstructured variance–covariance matrix parameterized through its Cholesky root unless otherwise stated.

^dToo few children were treated with late-generation fluoroquinolones, carbapenems, and linezolid to be analyzed. No children in these cohorts were treated with bedaquiline or delamanid.

^eaOR, for use of drug, with nonuse as reference category. Adjusted for age, sex, HIV infection, malnutrition, severity of disease on chest radiograph, and severity of extrapulmonary disease.

^fSecond-line injectable agents are amikacin, kanamycin, and capreomycin.

^gRandom-slope only model without random intercept, specified with standard variance components.

^hRandom-intercept only model without random slope, specified with standard variance components.

ⁱModel specified with standard variance components (not unstructured).

^jBolded results met the prespecified criteria for statistical significance.

Abbreviations: aOR, adjusted odds ratio; MDR-TB, multidrug-resistant tuberculosis.