Effect Modification by Baseline Mortality in the MORDOR Azithromycin Trial

Abstract

We examined whether baseline mortality risk, as a function of child age and site, modified the azithromycin mortality-reduction effect in the Macrolide Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) clinical trial. We used the Cox proportional hazards model with an interaction term. Three models were examined representing three sources for the baseline-risk covariate: two using sources external to MORDOR and the third leveraging data within MORDOR. All three models provided moderate evidence for the effect becoming stronger with increasing baseline mortality (P = 0.02, 0.02, and 0.07, respectively) at the rate of approximately 6-12% additional mortality reduction per doubling of baseline mortality. Etiological and programmatic implications of these findings are discussed.

Figure 1.

Externally derived baseline mortality curves used as risk covariates in the model. Left: Demographic and Health Survey (DHS) and Multiple Indicator Cluster Survey (MICS) estimates. Right: Institute of Health Metrics and Evaluation (IHME) local burden estimates. The bands indicate 95% CIs. The dots in both panes indicate crude observed mortality rates in Macrolide Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) placebo arm, calculated by site and 3-month-age bins. This figure appears in color at www.ajtmh.org.

Figure 2.

Observed mortality rates calculated by site and 3-month-age bins: placebo rates (x axis) vs. the difference between the arms (y axis). Negative difference indicates net mortality reduction on the azithromycin arm. Site is indicated by circle color and age group by the circle outlines’ shade of gray. The solid black curve is an empirical locally estimated scatterplot smoothing (LOESS) fit using span parameter 0.75. Compare with ref. 2, Fig. 2. This figure appears in color at www.ajtmh.org.

Figure 3.

Estimates from the effect-modification models with 99% CIs. Left: the main effect at a baseline risk of 15 deaths per 1,000 py. The blue line marks the overall Macrolide Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) estimate of 13.5% mortality reduction. Right: effect modification by baseline risk, represented as additional mortality reduction per doubling of baseline risk. This figure appears in color at www.ajtmh.org.

Figure 4.

Fitted effect-size curves from the three interaction models. The diagonal blue line indicates the average Macrolide Oraux pour Réduire les Décès avec un Oeil sur la Résistance (MORDOR) effect with no interaction. The age-bin observations from Figure 2 are added in the background as visual reference. This figure appears in color at www.ajtmh.org.