ABSTRACT
OBJECTIVE: In the developmental programming literature, the association of birth weight and blood pressure later in life is modest at best. This article reexamines this issue using Covariate Density Defined mixture of regressions (CDDmr) to determine if a latent variable, based on birth weight, and known to influence infant mortality, provides a stronger indicator of developmental programming. METHODS: CDDmr identifies two latent components in the birth weight distribution, generally interpreted in the infant mortality literature as "compromised" and "normal" fetal development. The data are taken from the 1958 British National Child Development Study. RESULTS: In the 1958 cohort, the "compromised" component consists of about 10% of the birth weight distribution. Compared to "normal" births, the mean systolic blood pressure at age 45 of "compromised" births is 20.3 (females) and 26.4 (males) mm Hg higher than "normal" births. The relative risks of stage-I-II systolic hypertension (>140 mm Hg) of "compromised" births compared to "normal" births are 10.7 (females) and 4.2 (males). The relative risks of stage-II systolic hypertension (>160 mm Hg) are considerably higher, 241.6 (females) and 74.3 (males). In general, the highest blood pressures are observed in "compromised" births with birth weights within the normal (about 3.0-3.5 kg) birth weight range. Results for diastolic blood pressure are similar. CONCLUSIONS: These associations are much stronger than those identified using conventional methods. CDDmr is likely to be useful for studying the dynamics of developmental programming particularly in large cohort studies where birth weight is often the only indicator of programming available. Am. J. Hum. Biol. 28:545-554, 2016. © 2016 Wiley Periodicals, Inc.
