Individual and Polyparasitic Helminth Infections as Determinants of Anemia in School-Age Children with Formal Assessment of Departures from Additive Risk Profile.

Analyses are based on multivariable logistic regression models clustered by household of residence and adjusted for the following confounders: age, sex, socioeconomic status, village of residence, nutritional status and tanner stage.†The estimates below are from logistic regression models that adjusted for the intensity of the 3 other parasitic helminth species (each as an ordinal covariate).‡These analyses exclude concomitant low intensity infections for any two helminth species under investigation. Sample size varied from 378 to 451 in multivariable analyses as a result of the exclusion of concomitant low intensity infections. For any pair of infections being evaluated, the intensity of the other 2 helminth species are controlled for in regression models. For example, for regression models examining the effect of hookworm and S. japonicum co-infection, ascaris and trichuris intensity are included in regression models as ordinal covariates.§AP = Attributable Proportion.**SI = synergy index. 95% Confidence Interval estimated as described by Kenneth J. Rothman.γ The interaction variables significantly improved the capacity of the regression model to explain the variability in anemia on the basis of likelihood ratio tests. All χ23df ranged from 25.6 to 9.6558; associated p-values ranged from ††The reference population for any pair of interactions is defined to include the following: a) children uninfected by either species and b) children infected by only 1 of the species at low intensity.