Increased risk of perinatal/neonatal death in infants who were smaller than expected at ultrasound fetometry in early pregnancy

Authors


Abstract

Objectives

To investigate if there is an association between perinatal/neonatal death and a discrepancy between gestational age by ultrasound fetometry in early pregnancy (GAU) and gestational age by last menstrual period (GALMP), and to investigate possible causes for such an association.

Methods

The Swedish Medical Birth Registry was used to identify singleton pregnancies with information available on GALMP and GAU that were delivered in Sweden between 1990 and 2000. A total of 718 011 pregnancies was included and information on the pregnancy and delivery was obtained from the National Board of Health.

Results

Infants with a GAU at least 7 days less than the GALMP were at increased risk for stillbirth (odds ratio (OR), 1.45; 95% CI, 1.32–1.58), neonatal death within 1 month (OR, 1.87; 95% CI, 1.67–2.09), Apgar score < 7 at 5 min (OR, 1.18; 95% CI, 1.11–1.24), birth weight < 2500 g (OR, 1.48; 95% CI, 1.43–1.52), and preterm birth < 37 weeks (OR, 1.45; 95% CI, 1.42–1.49). The association between a postponed expected date of delivery and perinatal/neonatal death increased with gestational length and was especially pronounced among infants who were born after at least 40 completed weeks of pregnancy (GAU).

Conclusions

A discrepancy between GAU and GALMP may indicate early disturbances in fetal/placental development. Furthermore, it can be speculated that, as the risk significantly increased with gestational duration, at least a part of the increased risk for poor pregnancy outcome in adjusted pregnancies was due to consequences of true post-term pregnancies not being recognized as such. Copyright © 2004 ISUOG. Published by John Wiley & Sons, Ltd.

Introduction

Today, almost all Swedish women are offered a free anomaly/dating ultrasound examination during weeks 16–18 of pregnancy. Dating by ultrasound in early pregnancy has frequently been shown to result in a more accurate prediction of the delivery date compared with using information on the last menstrual period (LMP)1–5. However, by converting variations in fetal size to differences in gestational age using ultrasound fetometry, early differences in fetal growth could seriously bias the assessment of gestational age. Several investigators have presented evidence suggesting that the ultrasound estimate of gestational age is biased by gender6–10, maternal smoking habits10, 11, number of fetuses10, and maternal age and parity9, 10 and educational level10. It is not known to what extent these systematic errors may increase the risk for suboptimum obstetric management due to erroneous estimates of gestational age.

It has been reported that infants whose gestational age according to ultrasound fetometry in early pregnancy (GAU) is less than the gestational age according to LMP (GALMP) are at increased risk of low birth weight12–15 and intrauterine growth restriction10, 12, 13, 15. These findings suggest that for some newborn infants who were considered small-for-gestational age at birth (SGA), the growth restriction was already present in early pregnancy. In a study comprising 16 469 pregnancies, Nguyen et al.13 reported that a discrepancy between GALMP and GAU also indicated an increased risk of fetal death (especially in combination with high maternal alpha-fetoprotein). The authors concluded that such a discrepancy might indicate early fetal pathology. However, investigating 15 241 pregnancies, Tunón et al.14 found no increased risk for perinatal death in pregnancies in which the GAU was more than 14 days less than the GALMP. In another Norwegian study15 based on 16 302 pregnancies, a highly significantly increased risk for perinatal death was found in such pregnancies.

In order to confirm whether a discrepancy between GALMP and GAU indicates an increased risk of perinatal/neonatal death, I conducted a study based on 718 011 pregnancies recorded in the Swedish Medical Birth Registry (MBR). If an association was found, the study also aimed to investigate whether this indicates early fetal pathology, or if evidence could be found indicating that the increased risk could also be due to an increased risk of death in infants born after unrecognized post-term pregnancies.

Methods

This study was based on all singleton births in Sweden during 1990–2000 with valid information on expected date of delivery according to ultrasound and LMP (valid information on GAU as well as GALMP was obtainable for 69% of all births). In order to reduce confounding by erroneous LMP, deliveries were excluded if the difference between the GAU and the GALMP exceeded 21 days (4.2% of all deliveries with apparently valid information on GALMP and GAU were excluded for this reason). After exclusions, a total of 718 011 pregnancies was included in the study. Information was obtained from the Swedish MBR, which contains medical information on nearly all deliveries in Sweden (coverage about 99%)16. Standardized record forms are used at all antenatal clinics, delivery units and pediatric examinations of newborn infants, and copies of these forms are sent to the National Board of Health where they are computerized. Nearly all pregnant women receive free antenatal care. At the first visit (usually during weeks 10–12), each woman is interviewed by a midwife and, among other things, the LMP and the smoking habit of the woman (none, < 10 cigarettes per day, or ≥ 10 cigarettes per day) are recorded. Almost all women are offered a free routine ultrasound examination during weeks 16–18 of pregnancy (GALMP) in order to estimate the expected date of parturition, to identify twin pregnancies, and to detect some severe malformations. Most ultrasound units use biparietal diameter (BPD) to assess gestational age, but in some units a combination of BPD and femur length is used. The routine examinations are performed by trained midwives, and the standard policy is to accept the results from the measurements as the best estimate of gestational age.

A Swedish standard for birth weight according to gestational age was used17 to estimate expected birth weight. Infants weighing less than two SD below the expected birth weight for gestational age (GAU) were considered SGA.

All odds ratios (OR) were calculated using Mantel-Haenszel's technique. Data were stratified according to year of birth, maternal age (5-year classes), parity (0, 1, 2, ≥ 3), infant gender, and maternal smoking in early pregnancy (none, < 10, ≥ 10 cigarettes/day). 95% CIs were estimated using Miettinen's method18. When comparing two stratified ORs, two-tailed Z-tests were carried out using the same variance as used to estimate the 95% CI. In order to detect putative linear trends among a series of stratified ORs, weighted linear regression analyses were carried out.

Results

Table 1 shows the numbers of infants according to magnitude of adjustment of expected date of delivery and pregnancy outcome. Infants delivered after pregnancies that were adjusted by at least −7 days (fetuses which were smaller than expected at ultrasound examination) were at increased risk for stillbirth/neonatal death, low Apgar scores, low birth weight and premature birth. The risk estimate for neonatal death within 1 month was significantly higher compared with the corresponding risk estimate for stillbirth (P for homogeneity < 0.001). For fetuses that were larger than expected at ultrasound examination in early pregnancy, there was no increased risk for any of these outcomes. However, these infants were at increased risk of being born in the post-term period (according GAU).

Table 1. Numbers and odds ratios (ORs) for various pregnancy outcomes according to adjustment of pregnancy progress at ultrasound examination in 718 011 singleton pregnancies in Sweden (1990–2000)
Pregnancy outcomeAdjustment
−21 to −7 days+7 to +21 days−6 to +6 days
nOR (95% CI)nOR (95% CI)n
  • Controlled for year of birth, maternal age, parity, smoking, and infant gender.

  • *

    Includes infants with missing 5-min Apgar scores. GA, gestational age.

Total n*184 269 27 891 505 851
Death
 Stillbirth    6851.45 (1.32–1.58)    680.88 (0.69–1.12)   1355
 Neonatal death (0–28 days)    5131.87 (1.67–2.09)    561.16 (0.88–1.53)    785
 Total deaths   11981.60 (1.49–1.72)   1240.99 (0.82–1.18)   2140
 Survival183 071Reference27 767Reference503 711
Apgar score at 5 min
 < 7   17931.18 (1.11–1.24)   2881.11 (0.98–1.25)   4465
 ≥ 7181 128Reference27 425Reference498 431
Birth weight (g)
 < 2500   71941.48 (1.43–1.52)   7680.97 (0.90–1.04) 13 778
 ≥ 2500177 075Reference27 123Reference492 073
GA at delivery (weeks)
 24–32   19711.48 (1.40–1.57)   2071.00 (0.87–1.16)   3771
 33–36 10 1831.45 (1.42–1.49)   9600.92 (0.86–0.99) 19 486
 37–41166 898Reference23 165Reference454 493
 42+   52170.54 (0.52–0.56)  35592.39 (2.31–2.48) 28 101

Figure 1 shows the association between number of days of adjustment of gestational age at ultrasound examination and risk of perinatal/neonatal death for each adjustment class compared with all other classes. For example, the OR for stillbirth/neonatal death was 1.7 among infants whose GAU was 14 days or more less than their GALMP, compared with infants with smaller adjustments of their gestational age. The risk for stillbirth/neonatal death increased linearly with the discrepancy between GALMP and GAU. When infants who were SGA at birth were excluded, there was still an increased risk for perinatal/neonatal death when GAU was at least 5 days less than GALMP, but no linear relationship between negative adjustment and perinatal/neonatal death was demonstrated.

Figure 1.

Odds ratio (OR) for stillbirth/neonatal death within 1 month plotted against adjustment of gestational age at ultrasound examination in 718 011 singleton pregnancies in Sweden (1990–2000). ORs with 95% CIs are for each class of difference between gestational age by ultrasound (GAU) and that by last menstrual period (GALMP) vs. all other classes of adjustment. (e.g. −6 days represents pregnancies that at ultrasound examination were judged to be 5–7 days less progressed than the LMP dates suggested). Controlled for year of birth, infant gender, maternal age, parity, and smoking habits. ————, all infants; - - - - -, excluding small-for-gestational age infants.

Figure 2 shows the OR for perinatal/neonatal death among infants who were smaller than expected at ultrasound examination according to completed weeks of pregnancy (by ultrasound estimate). For example, among infants born after 40 completed weeks of pregnancy (GAU), the OR for perinatal/neonatal death was 1.6 for infants whose GAU was at least 7 days less than the GALMP vs. infants for whom less adjustment of their gestational age was made; the corresponding OR was 1.1 among infants born after 38 completed weeks. The magnitude of the OR increased with gestational length. This trend seemed more pronounced when infants who were SGA at birth were excluded (P for linear trend of the ORs: < 0.01). The association between low Apgar score at 5 min and adjustment of gestational length showed a similar progress with length of gestation (P for linear trend < 0.01, data not shown).

Figure 2.

Odds ratios with 95% CIs for perinatal/neonatal death by gestational age (according to ultrasound) at birth among infants whose gestational age by ultrasound (GAU) was at least 7 days less than the gestational age by last menstrual period (GALMP) vs. infants whose difference between GAU and GALMP was less than 7 days, in 718 011 singleton pregnancies in Sweden (1990–2000). Controlled for year of birth, infant gender, maternal age, parity, and smoking habits. ————, all infants; - - - - -, excluding small-for-gestational age infants.

Further investigations were carried out in order to detect signs of disturbances in early fetal/placental development among pregnancies that were adjusted at the ultrasound examination. There was a small but significant association between a discrepancy between GAU and GALMP of at least −7 days and the presence of a significant congenital malformation at birth (OR, 1.07; 95% CI, 1.02–1.12). This association was of similar magnitude among infants who were classified as SGA at birth and those who were not. Among 184 269 women whose expected date of delivery was postponed by at least 7 days, 5702 (3.1%) were diagnosed with pre-eclampsia during late pregnancy, compared with 13 718/505 851 (2.7%) in women whose length of gestation was not adjusted. The adjusted OR (95% CI) for pre-eclampsia was 1.20 (1.16–1.24).

Discussion

This study confirms the results of Nguyen et al.13 and Nakling and Backe15 of an association between perinatal/neonatal death and a shorter gestational age according to ultrasound than the LMP-based estimate. Tunón et al.14 found no such statistically significant association, but it should be noted that their study was based on 47 perinatal deaths, compared with 188 deaths in the investigation of Nguyen et al.13 and the 3338 deaths in the current study.

The information on LMP used in this study was based on self-reported information obtained at the antenatal units and must be considered somewhat unreliable. However, the misclassification bias is non-differential as this information is obtained in early pregnancy and could not be biased by the pregnancy outcome. Errors of this type would bias the results towards unity. Thus, the association between adverse pregnancy outcome and adjustment of gestational age at ultrasound examination is likely to be underestimated, and would have been even more persuasive if putative erroneous LMP dates could be detected and excluded.

In concordance with the literature, infants who were smaller than expected at ultrasound examination were at increased risk for low birth weight and premature birth. The association between low birth weight and a smaller than expected fetus in the second trimester is by now well-established and is likely to reflect early onset of growth restriction in affected pregnancies. The outcome ‘premature birth’ is more difficult to interpret as possible erroneous adjustments of the gestational durations could seriously bias the results. The increased risk found in this study for post-term birth in pregnancies that were judged more progressed than the LMP date suggested is likely to reflect such a bias. Another example could be fetuses with growth restriction in early pregnancy whose gestational age was adjusted at the ultrasound examination. These infants may erroneously be judged as premature and of appropriate size for gestational age, but would have been SGA at term if their correct gestational age had been considered. This could be the reason why some investigators who found an increased risk for low birth weight in adjusted pregnancies could not detect an increased risk for SGA14. However, despite the fact that the magnitudes of the growth restriction among infants who were smaller than expected at ultrasound examination in the second trimester is likely to be underestimated, a strong association between a ‘smaller than expected’ fetus at ultrasound examination and SGA has been reported by several investigators10, 12, 13, 15.

The existence of an association between early onset of growth restriction and a shorter estimate of gestational length according to ultrasound than the LMP suggests is consistent with the hypothesis that the association between adjustment of gestational age and perinatal/neonatal death is mainly due to early disturbances in fetal/placental development. The association between adjustment of gestational age and the presence of a significant malformation in the newborn or maternal pre-eclampsia later in pregnancy to some extent supports such a hypothesis. An increased risk for chromosomal anomalies in infants who were smaller than expected at ultrasound examination has been reported19.

By scrutinizing seven deaths in which GAU was 14 days less than GALMP, Tunón et al.14 concluded that none of these could have been prevented had the estimated date of delivery according to the LMP been used instead. However, the significance of such a limited investigation must be questioned. In the current study, the association between a postponed date of delivery and perinatal/neonatal death was most pronounced in pregnancies lasting at least 40 completed weeks (especially among infants who died postpartum and were not SGA). A similar trend was shown for the association between adjustment of gestational age and low Apgar score. It is likely that several of these pregnancies entered the post-term period unnoticed, and these results do not exclude the possibility that erroneous adjustments of expected date of delivery may increase the risk for perinatal/neonatal death due to true post-term pregnancies not being recognized as such. The results of this study suggest that special attention should be paid to pregnancies that are adjusted by at least 1 week and last 40 weeks or more.

Acknowledgements

Thanks to K. A. Wallenbergs stiftelse.

Ancillary