Mediation of the effect of malaria in pregnancy on stillbirth and neonatal death in an area of low transmission: observational data analysis

Table 2 Mediation of the association between malaria in pregnancy and antepartum stillbirth or neonatal death

Mediating variable/s	Natural indirect effect, risk ratio (95% CI)	Natural direct effect, risk ratio (95% CI)	Proportion indirect
Falciparum malaria and antepartum stillbirth
SGA only	1.09 (1.05, 1.14)	1.61 (0.84, 2.67)	18%
Maternal anaemia and SGA	1.23 (1.15, 1.34)	1.48 (0.77, 2.46)	42%
Symptomatic vivax malaria and antepartum stillbirth
SGA only	1.10 (1.04, 1.17)	1.91 (0.65, 3.49)	16%
Maternal anaemia and SGA	1.13 (1.06, 1.23)	1.73 (0.59, 3.20)	24%
Falciparum malaria and neonatal death
Preterm birth only	1.25 (1.10, 1.43)	2.27 (1.29, 3.63)	26%
SGA and preterm birth	1.30 (1.15, 1.49)	1.82 (1.07, 2.89)	40%
Vivax malaria and neonatal death
Preterm birth only	1.20 (1.05, 1.37)	1.58 (0.94, 2.44)	31%
SGA and preterm birth	1.25 (1.09, 1.44)	1.32 (0.81, 1.98)	50%

Numbers are risk ratios (95% confidence interval). Natural direct effect (NDE, the effect of malaria on birth outcome, not mediated through the specified mediator/s); natural indirect effect (NIE, the effect of malaria on birth outcome, mediated through the specified mediator/s); SGA small-for-gestational-age. Note that total effects (NDE + NIE) from mediation analysis are not equivalent to the magnitude of associations obtained from Cox regression, which accounts for lost to follow-up and time-varying exposure and risk, because methods that incorporate time-to-event data in mediation analyses are limited. Maternal anaemia was defined as a final haematocrit during pregnancy of <30% (measured fortnightly throughout pregnancy). SGA was defined as a birthweight for gestational age below the 10th centile of INTERGROWTH-21st centiles. Maternal anaemia and SGA were included as binary variables because the association between the continuous variables, haematocrit and birthweight-for-gestational-age Z-score, and log_e(odds) of stillbirth and/or neonatal death was non-linear. Maternal anaemia was not considered as a mediator of the association between malaria in pregnancy and neonatal death because <1% of women in this sub-set were anaemic. For further details on calculation and formal definitions of NDEs and NIEs, see Additional file 4

ISSN: 1741-7015