Answers are grounded in this paper and cite their sources.
To the Editor. Are small for gestational age (SGA) preterm infants at increased or reduced risk of neonatal death? Currently, it seems either answer will do! In two recent papers, Horbar et al1,2concluded after an analysis using logistic regression that SGA infants have a significantly reduced risk of dying (odds ratio [OR] 0.63, confidence interval [CI] 0.55–0.86). In both cases, a birth weight (BW)-defined cohort was studied (501–1500 g). As a consequence of the 1500-g cutoff, well-grown babies 29 to 31 weeks old were excluded, while numerous SGA infants with gestations >31 weeks were included (about 15%). These, as well as other studies using BW-defined cohorts,3 continue to add support to the widely held belief that SGA infants mature more rapidly than appropriate for gestational age (AGA) infants.However, other authors,4–7 in studies using case-control design matching for gestational age (GA) or using GA-defined cohorts, have arrived at the opposite conclusion that SGA is associated with an increased risk of both mortality and morbidity in very preterm infants.We undertook a study of two cohorts drawn from the 32 926 babies cared for at King George V Hospital during the years 1992–1996 inclusive. Infants with major congenital abnormalities were excluded. The first cohort contained all infants born with GA between 24 and 31 weeks inclusive (n = 700; mean BW 1208 g; mean GA 28.5 weeks; 28-day mortality 9.4%), and the second cohort contained all infants born with BW of 501 to 1500 g (n = 636; mean BW 1112 g; mean GA 28.6 weeks; 28-day mortality 9.9%). Logistic regression was used to estimate ORs for SGA (<10th BW percentile8) for the outcome of 28-day mortality. Factors adjusted for in all analyses were gender, 1-minute Apgar score, multiple birth, antenatal steroids, and type of delivery. BW or GA were separately entered in addition to the above factors. The results are presented in Table 1. In the GA-defined cohort adjusting for GA, SGA was associated with a significantly increased risk of death. In stark contrast, in the BW-defined cohort adjusting for BW, SGA was associated with a significantly reduced risk of death. In both the GA-defined cohort adjusting for BW and the BW-defined cohort adjusting for GA, SGA was not significantly associated with death.This simple exercise was undertaken to demonstrate that the result one obtains using logistic regression will vary significantly depending on whether a BW- or GA-defined cohort is used, and also depending on how one handles GA and BW. So which answer is correct? Perhaps one could say that for a given GA, SGA is associated with an increased risk of death, while for a given BW, SGA is associated with a reduced risk of death. However, this distinction is probably lost on the casual reader, with the result that the myth of SGA being associated with reduced mortality or respiratory distress syndrome is perpetuated. Horbar et al1 do concede that “SGA infants have a lower risk of mortality than AGA infants at any given birth weight because they are more mature.” But this is not just an academic argument, as there are important implications for student and resident teaching, and counseling of parents.GA remains the best measure of maturity and predictor of outcome, especially before birth when only GA is known. On clinical grounds GA should be included in the logistic model, rather than allowing the computer to exclude it via stepwise selection. The argument that BW is more reliable data than GA doesn't make sense since GA is required to calculate SGA.To get back to the original question, the one and only correct answer must be that SGA is associated with an increased risk of death. The above exercise also highlights the need for caution in interpreting the results of logistic regression analyses.
No takes yet. Share an insight, caveat, or question.
Philip J. Beeby (1998) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: