The causes of preterm birth

W.H. James

The Galton Laboratory, University College London, Wolfson House, 4 Stephenson Way, London NW1 2HE, UK

Dear Sir,

Astolfi and Zonta (1999) analysed the data on 2.8 million births by duration of gestation from last menstrual period (LMP), birth order, maternal age, maternal education and sex. Univariate analysis showed that the risk of preterm birth varied significantly with each of these latter four variables: higher risks were associated with first births, elderly mothers, poorly educated mothers, and male infants. Logistic regression analysis suggested that though all four of these risk factors remained significant, maternal age was the strongest, and infant sex the weakest of them.

I want here to consider the effect of two sorts of error to which such data are susceptible: firstly, the error of misreporting LMP, and secondly, the error involved in using LMP as a surrogate for the initiation of pregnancy (i.e. conception).

Error A

Milner and Richards (1974) analysed the birthweights of infants by reported gestational duration and sex. They noted that the distribution of birthweights at a given gestation is normal at >36 weeks, but skewed or bimodal in preterm infants. These authors noted that the non-normal distributions of birthweights of preterm infants could be described as mixtures of two normal distributions with different means but the same standard deviation. They found that the sub-sample of preterm infants with the higher mean birthweight comprised ~33.3% of the observations each week at 28–34 weeks, but was only 0.79% of all births. The mean birthweight of the high birthweight subsample was about that of infants of 38.4 weeks gestation. The sex ratio of this subsample was similar to that of term infants. The sex ratio of the low birthweight subsample rose with increasing prematurity and was 128 (males per 100 females) at 28–34 weeks. These data suggest the important inference drawn by the authors, i.e. that the high birthweight subsample is composed of pregnancies in which the length of gestation has been underestimated. This is a small proportion of all births, but an important fraction of `preterm' births (about one third).

The role of misreporting LMP here is emphasised by the data of Cooperstock and Campbell (1996). They noted that the reported excess of males in preterm births is greater in those births which may be presumed to have been the subject of more accurate LMP reporting viz. of white versus black births; maternal age 20+ versus <=19 years; maternal education 12+ versus <=11 years; and married versus unmarried. It will be seen that if this inference of error is correct, then the true statistical association between infant sex and duration of gestation is substantially stronger than as calculated by Astolfi and Zonta (1999).

Error B

Pregnancy starts at conception. For the purpose of predicting confinement, pregnancy may be conveniently dated from LMP, but this involves an error in the sense that conception may occur across some period of time (a fertile interval of several days) during a woman's cycle. There are grounds for supposing that the probability of a boy has a U-shaped regression across this fertile interval (James, 2000Go). Given the substantial variation in the conception-confinement interval, I have noted that this postulated U-shaped regression at conception is followed by a much-damped \/ -shaped regression 9 months later (James, 1994Go). In this paper I showed numerically that the latter regression (of sex ratio on duration of gestation from reported LMP) could be explained by the former. In other words, we have an explanation of why more boys are `premature'. They are born about a day earlier because ex hypothesi they were conceived about a day earlier. This argument is strengthened by the observation that the postulated greater variance of cycle day of conception of boys is followed (as expected on this hypothesis) by a greater variance of birthweight of boys for a given reported duration of gestation (Milner and Richards 1974Go). This is so because for a given reported duration of gestation, boys ex hypothesi have a greater variance of true duration of gestation.

If this argument were substantially correct, then (contrary to the hypothesis of Cooperstock and Campbell, 1996, and implicitly followed by Astolfi and Zonta, 1999), infant sex is not a causal agent (nor associated with a causal agent) in initiating labour. My suggestion is supported by other authors (McGregor et al., 1992Go), who noted that the increase in preterm births in males was not accompanied by an increased number of males with low birthweight. Moreover, they found that the difference in preterm births by sex could not be explained by increased occurrences of premature rupture of membranes, chorioamnionitis, endometriosis or other infections. They concluded that the shorter gestation in males may be related to their relatively increased size and birthweight. However, this argument seems vitiated by the fact that the effect occurs most powerfully earlier in pregnancy when both sexes are small, and not at all in late pregnancy when males are substantially bigger. Lastly this explanation fails to account for the shape of the regression of sex ratio on reported duration of gestation, i.e. \/ -shaped rather than monotonically declining (James, 1994Go). Indeed I have seen no other attempt to explain the shape of this regression.

Reverting to the paper of Astolfi and Zonta (1999), one may suggest that unlike first births, poor maternal education, or high maternal age, (which all are associated with sub-optimal obstetric conditions), male infant sex is associated with underestimates of the duration of true gestation (i.e. conception to confinement). If I am right, male fetal sex is only a statistical risk factor (not a causal risk factor) for prematurity. So it may be misleading for Astolfi and Zonta (1999) to liken this to the biological and genetic weaknesses of male fetuses, e.g. their high susceptibility to stillbirth.

References

Astolfi, P. and Zonta, L.A. (1999) Risks of preterm delivery and association with maternal age, birth order and fetal gender. Hum. Reprod., 14, 2891–2894[Abstract/Free Full Text]

Cooperstock, M. and Campbell, J. (1996) Excess males in preterm birth: interactions with gestational age, race and multiple birth. Obstet. Gynecol., 88, 189–193[Abstract/Free Full Text]

James, W.H. (1994) Cycle day of insemination, sex ratio of offspring and duration of gestation. Ann. Hum. Biol., 21, 263–266[ISI][Medline]

James, W.H. (2000) Analysing data on the sex ratio of human births by cycle day of conception. Hum. Reprod., 15, in press.

McGregor, J.A., Leff, M., Orleans, M. and Baron, A. (1992) Fetal gender differences in preterm births: findings in a North American cohort. Am. J. Perinatol., 9, 43–48.[ISI][Medline]

Milner, R.D.G., and Richards, B. (1974) An analysis of birth weight by gestational age of infants born in England & Wales 1967–71. J. Obstet. Gynaecol. Br. Commonw., 81, 956–967.[ISI][Medline]


 
Paola Astolfi and Laura A. Zonta

Dept. of Genetics and Microbiology, Via Abbiategrasso 207, 27100 Pavia, Italy

Dear Sir,

Dr James draws attention on two types of errors connected with pregnancy duration when estimated on the basis of last menstrual period (LMP), and the repercussion they may have on the analysis of the sex ratio by gestational age.

As quoted by James, the error of misreported LMP is clearly detectable by inspecting the distribution of birth weight by week of gestational age (Milner and Richards, 1974Go). The analysis of birth weight for preterm babies in our sample of legitimate single liveborns of first and second birth order confirmed the skewness or, in a few cases, even a clear bimodality of the distribution. However, an approximate estimate of the overall proportion of high weight preterm babies (25–36 weeks) was ~3%. The high weight subsamples never reached the 30% observed at 28–34 weeks by the cited authors. In fact, the highest proportion of heavy babies, found at 30–31 weeks of reported LMP pregnancy duration, was ~20%, and all the other values were <10%.

On the assumption that the high weight subsamples might indeed correspond to misclassified babies of longer pregnancy duration, we removed them from the preterm samples and re-analysed our data. The proportion of male births by week of gestational age, obtained after the correction, varied from 0.521 to 0.551 and was still higher than the value of ~0.51 observed in the full term newborns. To check whether the presumed misclassification might bias the results of our previous logistic analysis (Astolfi and Zonta, 1999Go), we re-evaluated the contribution of the sex of the baby, the mother's age (age groups 30–34, 35–39, 40+ years) and educational level (<13 and 13+ years of schooling) to the risk of preterm delivery. The odds ratio turned out only slightly different from the published values (Table IGo). Although we realize the need of a more accurate determination of pregnancy duration and agree on the role that the mother's low education can play in misreporting the LMP, we feel quite confident about the robustness of our results, namely that the association of low gestational age is in fact stronger with advanced mother's age than with the male sex of the conceptus. Our results strongly suggest that the advanced mother's age can in fact represent an important risk factor for preterm delivery: even in the most favourable circumstances, that is female secondborns of highly educated mothers, the observed frequency of preterm deliveries in mothers aged 40+ years is almost twice that of mothers aged 30–34 years (6.21 versus 3.21% respectively).


View this table:
[in this window]
[in a new window]
 
Table I. The probability of a preterm baby associated with the baby gender (SX), birth order (BO), mother age (MA) and educational level (ED) was analysed by logistic regression. Odds ratios and 95% confidence intervals (in parentheses) are shown for each factor
 
As for the interpretation of the excess of males among preterm babies, we are well aware of the long debate and contradictory results on the association between sex ratio and timing of insemination in several mammalian species besides humans (Berstein, 1998aGray et al., 1998Go,bGo; ). Several studies support the finding that, due to the cycle day of conception, the probability of a son assumes a U shaped distribution over the woman fertile interval (James, 1994Go) with a consequent greater variance of cycle day of conception in boys. The larger variance in male birth weight observed each gestation week is, therefore, in agreement with the U shaped distribution theory (Milner and Richards, 1974Go). Even our samples of both first and second newborns show the tendency of a greater variance in the weight of males than females at most gestational ages. Therefore, it is reasonable to assume that male sex can be a statistical factor for prematurity as suggested by James. On the other hand, since preterm deliveries are adverse pregnancy outcomes, the hypothesis of biological, and possibly genetic, `weakness' of the male sex as one of the factors contributing to the male excess in preterm deliveries cannot be overruled. In fact not only the higher proportion of males than of females among stillborns in unfavourable environments and in adverse pregnancy outcomes (Ulizzi and Zonta, 1993Go; Bracero et al., 1996Go; Zonta et al., 1997Go), but also the decline in the offspring sex ratio of selected samples of parents exposed to specific chemicals and to environmental hazards (James, 1996Go; Mocarelli et al., 1996Go) seem a fairly good support to the hypothesis of such weakness.

References

Astolfi, P. and Zonta, L.A. (1999) Risks of preterm delivery and association with maternal age, birth order and fetal gender. Hum. Reprod., 14, 2891–2894.[Abstract/Free Full Text]

Bernstein, M.E. (1998a) Gestation length and sex of child. Hum. Reprod., 13, 2975.[Free Full Text]

Bernstein, M.E. (1998b) Time of insemination within the cycle and offspring sex ratio. Hum. Reprod., 13, 3280.[ISI][Medline]

Bracero, L.A.,Cassidy, S. and Byrne, D.W. (1996) Effect of gender on perinatal outcome in pregnancy complicated by diabetes. Gynecol. Obstet. Invest., 41, 10–14.[ISI][Medline]

Gray, R.H., Simpson, J.L., Bitto, A.C. et al. (1998) Sex ratio associated with timing of insemination and length of the follicular phase in planned and unplanned pregnancies during use of natural family planning. Hum. Reprod., 13, 1397–1400.[Abstract]

James, W.H. (1994) Cycle day of insemination, sex ratio of offspring and duration of gestation. Ann. Hum. Biol., 21, 263–266.[ISI][Medline]

James, W.H. (1996) Male reproductive hazard and occupation. Lancet, 347, 773.

Milner, R.D.G. and Richards, B. (1974) An analysis of birth weight by gestational age of infants born in England & Wales 1967–71. J. Obstet. Gynaecol. Br. Commonw., 81, 956–967.[ISI][Medline]

Mocarelli, P., Brambilla, P., Gertoux, P.M. et al. (1996) Change in sex ratio with exposure to dioxin. Lancet, 348, 409.[ISI][Medline]

Ulizzi, L. and Zonta, L.A. (1993) Sex ratio and natural selection in humans: a comparative analysis of two Caucasian populations. Ann. Hum. Genet., 57, 211–219.[ISI][Medline]

Zonta, L.A., Astolfi, P. and Ulizzi, L. (1997) Heterogeneous effects of natural selection on the Italian newborns. Ann. Hum. Genet., 61, 137–142.[ISI][Medline]