Statistical modelling of the seasonality of preterm delivery and intrauterine growth restriction in rural Gambia

We have developed a methodology for comparing the seasonal influences on two outcomes, when those influences may act cumulatively or instantaneously. We have used this to compare the seasonal pattern of intrauterine growth retardation (IUGR, as reflected by weight-for-gestational-age) and preterm delivery (as assessed by Dubowitz scoring) among 1718 infants born in rural Gambia. Both outcomes were analysed as binary variables: small-for-gestational-age (SGA, < 10th centile of reference standard) and preterm (< 37 weeks) respectively. Percentages of preterm and SGA babies show divergent seasonal patterns that might indicate separate aetiologies. However, seasonal effects influencing intrauterine growth are likely to be cumulative over the last few months of pregnancy. By modelling seasonality with truncated Fourier series we were able to deconvolve the underlying seasonal influences on fetal growth from the pattern for SGA. This enabled us to use seemingly unrelated biprobit regression to compare the underlying seasonal pattern of intrauterine growth with that governing the incidence of preterm delivery. We conclude that, if the seasonal factors affecting intrauterine growth operate over more than the last 2 months of pregnancy, then the seasonal patterns of the factors causing IUGR and preterm delivery are indistinguishable if the factors are assumed to trigger preterm delivery immediately, but differ if preterm delivery is assumed to be programmed by factors acting at conception.