Data analysis
The general characteristics of subjects were described according to the
median, ranges, frequencies, and percentages. A Conditional inference
tree, a non-parametric multivariate regression model, using thepartykit library of R 3.5.1 software, was applied to classify
groups according to the presence of any adverse outcomes potentially
related to ZIKV infection (miscarriages, stillbirths, microcephaly with
or without brain abnormities, and brain abnormalities without
microcephaly) from gestational age25. In general, this
method selects, among independent variables, the ones that better
classifies the individuals according to the presence of the outcome. In
the next step, the algorithm considers all the independent variables to
select another (or the same variable previously selected but estimating
different cut-off from the first step) that better classifies the
individuals according to the outcome and the process extends until there
are no more variables selected. The independent variable used by us in
these analyses was the gestational age. The same method was used to
classify groups more likely to present a typical CZS phenotype as a
function of gestational age. Log-binomial regression models were
adjusted to estimate relative risks and their respective 95% confidence
intervals. Data were analyzed using SPSS software version 20.0.