Table 8. The relationship between antibiotic use and successful treatment of tuberculosis (versus death or failure—excluding relapse and loss to follow-up) applying different analytic methods to account for confounding.
Method of adjustment | Earlier generation FQN a | Later generation FQNs a | Macrolide antibiotics b | |||
---|---|---|---|---|---|---|
OR | 95% CI | OR | 95% CI | OR | 95% CI | |
Conventional approaches | ||||||
Unadjusted estimate | 2.3 | (1.5, 3.3) | 1.5 | (0.9, 2.7) | 0.3 | (0.2, 0.5) |
Multivariable analysis | 2.1 | (1.4, 3.2) | 2.0 | (1.1, 3.6) | 0.3 | (0.2, 0.5) |
Unmatched PS based methods | ||||||
Regression with PS quintile as covariate | 2.1 | (1.4, 3.0) | 1.8 | (1.0, 3.1) | 0.3 | (0.2, 0.5) |
Regression with continuous PS as covariate | 1.9 | (1.2, 2.8) | 1.5 | (0.9, 2.7) | 0.4 | (0.3, 0.5) |
Inverse probability of treatment weighting | 1.8 | (0.7–4.3) | 1.1 | (0.2, 5.4) | 0.4 | (0.3, 0.6) |
PS matching within studies | ||||||
1:1 matching,¶ no replacement | 1.5 | (0.5, 4.3) | 0.7 | (0.1, 4.8) | 0.5 | (0.3, 0.9) |
1:1 matching with replacement | 0.5 | (0.1, 2.2) | 0.2 | (0.0, 2.1) | 0.5 | (0.3, 0.8) |
1:4 matchingΔ | na | na | 0.4 | (0.3, 0.7) | ||
PS matching across studies | ||||||
1:1 matching+ no replacement | 2.4 | (0.8, 7.6) | 2.0 | (0.3, 12.3) | 0.4 | (0.2, 0.7) |
1:1 matching¶ with replacement | 0.5 | (0.1, 2.2) | 0.4 | (0.1, 1.9) | 0.4 | (0.2, 0.8) |
1:4 matchingΔ | na | na | 0.4 | (0.2, 0.6) |
a Comparison between those taking fluoroquinolones and those not taking fluoroquinolones, in studies where fluoroquinolones were used.
b Comparison between those taking macrolides and those not taking macrolides, in studies where macrolides were used.
¶ Propensity scores calculated within individual studies. Ten studies were excluded owing to insufficient numbers of patients to perform analysis.
+ Matching by propensity score calculated across all studies. PS = Propensity score. OR = odds ratios. 95% CI = 95% confidence intervals. Bolded text indicates p <0.05. Macrolides includes azithromycin, clarithromycin and roxithromycin. na = not applicable, as insufficient unexposed subjects available to perform four to one matching. FQN = fluoroquinolone. These analyses excluded individuals with relapse or loss-to follow-up.