Time Trends in Sputum Mycobacterial Load and Two-Day Bactericidal Activity of Isoniazid-Containing Antituberculosis Therapies

Abstract

Recent early bactericidal activity (EBA) studies of isoniazid-based antituberculosis therapies have shown a lower EBA over the first two treatment days than in earlier years. To quantify this trend and evaluate factors contributing to it, we extracted individual data from 18 studies with a total of 182 participants using isoniazid-containing therapies between 1992 and 2015 at a single site and laboratory in Cape Town, South Africa. We recalculated EBA as the daily fall in CFU per milliliter sputum up to day 2 of therapy (EBA_0-2) for individual patients and treatment groups and used mixed-effects linear models to investigate the correlation between pretreatment CFU, EBA_0-2, and year of study. We found that mean pretreatment CFU and year of study accounted for 46% and 47%, respectively, of the variation in mean EBA_0-2 Mean pretreatment CFU differed between the periods 1992 to 2001 and 2007 to 2015 by 0.92 log₁₀ CFU (95% confidence interval [CI], 0.57 to 1.28; P < 0.0001). On average, pretreatment CFU dropped by 0.053 log₁₀ CFU (95% CI, 0.029 to 0.076; P = 0.0004) and EBA_0-2 by 0.012 log₁₀ CFU (95% CI, 0.006 to 0.018; P = 0.001) per year. The EBA_0-2 of isoniazid-based antituberculosis therapy is strongly correlated with baseline mycobacterial load and shows a declining trend over the past 2 decades.

Keywords: early bactericidal activity; isoniazid; tuberculosis.

FIG 1

Recalculated individual pretreatment CFU counts, EBA_0–2, and year of study. (A) EBA_0–2 increases with increasing pretreatment CFU counts. Pretreatment CFU (B) and EBA_0–2 (C) decrease with the year of study completion. The illustrations show lines estimated from linear mixed-effects models of individual participant outcome as function of predictor, including study as a random effect. The slopes and P values, respectively, are 0.120 and <0.0001 (A), −0.053 and 0.0004 (B), and −0.012 and 0.0031 (C). Values with the same year are jittered to avoid them falling on top of one another. EBA_0–2, early bactericidal activity expressed as daily fall of log₁₀ CFU/ml sputum over the first 2 days of treatment.

FIG 2

Study mean pretreatment log₁₀ CFU count and year of study completion. Means per study of pretreatment CFU, calculated with general linear regression models, are grouped in time periods from 1992 to 2001 and 2007 to 2015, within which there is no significant trend over time. Between these time periods, the means differ by 0.92 log₁₀ CFU (95% CI, 0.57 to 1.28; P < 0.0001). The circle size is proportional to the number of participants in each study. Bars are 95% confidence intervals.

FIG 3

Recalculated study mean pretreatment log₁₀ CFU count and mean EBA_0–2. There is a clear association of lower pretreatment CFU with lower EBA_0–2. The circle size is proportional to the number of participants in each study. The 95% confidence intervals of the study means were calculated with general linear models. EBA_0–2, early bactericidal activity expressed as daily fall of log₁₀ CFU/ml sputum over the first 2 days of treatment.