Statistics from Altmetric.com
Babies with lower birthweights have higher risks of dying in infancy. Populations with lower mean birthweights usually have higher infant mortality rates. So is low birthweight, of itself, an adequate explanation of increased infant mortality? It has been argued that it is not (
If you plot neonatal mortality (y-axis, logarithmic) against birthweight (x-axis) you get a reversed J-curve with neonatal mortality falling from a very high level at very low birthweights to a minimum at about 3.5 kg (US data) and then increasing again at higher birthweights. (Optimal birthweight tends to be somewhat higher than mean birthweight.) Changing circumstances tend to change the level but not the shape of the curve. Thus, in the USA neonatal mortality fell for all birthweights between 1950 and 1988 so the 1998 curve lies below but parallel to the 1950 curve. (There is, incidentally, no change in the curve at 2.5 kg so the distinction between low birthweight and normal birthweight is arbitrary). Factors, such as maternal smoking or high altitude residence, which reduce birthweight in populations simply shift the reversed-J to the left. This produces the “low birthweight paradox” because low birthweight babies in the reduced-birthweight group then have lower mortality rates than babies of the same birthweight in the standard group. Maternal smoking then appears to be “beneficial” for lower birthweight babies. Wilcox solves the paradox by plotting neonatal mortality against birthweight z-scores for each group. It is then found that the neonatal mortality of babies of smoking mothers exceeds that of babies of non-smoking mothers at all points of the curve. Therefore, maternal smoking reduces birthweight at all levels but the effect on neonatal mortality is independent of birthweight. Wilcox argues that attention should be focussed on preterm births either by recording of gestational age or by estimation of the proportion of small preterm births from the “residual distribution” of the birthweight frequency distribution. (The “residual” distribution is the lower tail lying outside the normal, bell-shaped, curve and is almost entirely due to small preterm births.)
Two commentators (Ibid: 1241–3 and 1243–4) accept that the low birthweight/normal birthweight dichotomy is outdated but challenge Wilcox's conclusions, one because he believes that Wilcox takes too little heed of the social context and the other because she still believes that birthweight can be informative about population health.