Differences in late fetal death rates in association with determinants of small for gestational age fetuses: population based cohort studyBMJ 1998; 316 doi: http://dx.doi.org/10.1136/bmj.316.7143.1483 (Published 16 May 1998) Cite this as: BMJ 1998;316:1483
- Sven Cnattingius, associate professor ()a,
- Bengt Haglund, associate professorb,
- Michael S Kramer, professorc
- a Department of Medical Epidemiology, Karolinska Institute, S-171 77 Stockholm, Sweden
- b Centre for Epidemiology, National Board of Health and Welfare, Stockholm, Sweden,
- c Department of Pediatrics, McGill University Faculty of Medicine, Montreal, Quebec, Canada
- Correspondence to: Dr Cnattingius
- Accepted 4 February 1998
Objective: To examine differences in late fetal death rates in association with determinants of small for gestational age fetuses.
Design: Population based cohort study.
Subjects: 1 026 249 pregnancies without congenital malformations.
Setting: Sweden 1983-92.
Main outcome measure: Late fetal death rate.
Results: Depending on underlying determinants late fetal death rates were greatly increased in extremely small for gestational age fetuses (range 16 to 45 per 1000) compared with non-small for gestational age fetuses (1.4 to 4.6). In extremely small for gestational age fetuses late fetal death rates were increased from 31 per 1000 in mothers aged less than 35 years to 45 per 1000 in older mothers, and from 22 per 1000 in women <155 cm in height to 33 per 1000 in women ≥175 cm tall. Late fetal death rates were also higher in extremely small for gestational age fetuses in singleton compared with twin pregnancies and in non-hypertensive pregnancies compared with pregnancies complicated by severe pre-eclampsia or other hypertensive disorders. Slightly higher late fetal death rates were observed in nulliparous compared with parous women and in non-smokers compared with smokers.
Conclusions: Although the risk of late fetal death is greatly increased in fetuses that are extremely small for gestational age the risk is strongly modified by underlying determinants—for example, there is a lower risk of late fetal death in a small for gestational age fetus if the mother is of short stature, has a twin pregnancy, or has hypertension.
Small for gestational age fetuses are at increased risk of late fetal death regardless of the underlying determinants
The effect of birthweight ratio on risk of late fetal death is modified by underlying determinants, except maternal age
Regardless of birthweight ratio the rates of late fetal death are higher among women aged 35 years or older compared with younger women
In pregnancies of extremely small for gestational age fetuses lower rates of late fetal death are associated with a maternal age of less than 35 years, short maternal stature, multiple births, and hypertensive disorders
In pregnancies with non-malformed fetuses late fetal death rates are increased in smokers, in multiple births, and in women with severe pre-eclampsia.
A wide range of risk factors are associated with fetuses being small for gestational age,1 and the prognosis for such fetuses varies according to the presence of chromosomal or other congenital malformations.2–4 In pregnancies of small for gestational age fetuses without congenital malformations it has been assumed that constitutionally small fetuses—for example, twins, or infants born to short mothers—are at lower risk of adverse outcomes than fetuses affected by other situations such as pre-eclampsia or cigarette consumption. 5 6 In fact, such an assumption underlies recent pleas for customised or individualised definitions of a fetus being small for gestational age. 7 8 Unfortunately, this assumption has rarely been critically tested. 9 10
In Sweden the population based birth register includes information on risk factors for fetuses being small for gestational age including maternal age, height, parity, smoking habits, blood pressure status, and type of pregnancy (single or multiple). We used this information to study the differences in late fetal death rates in association with fetal size and underlying determinants of a fetus being small for gestational age.
Subjects and methods
Swedish birth register
In Sweden data from all hospital births, including demographics, reproductive history, and complications during pregnancy and delivery, are collected prospectively and recorded in a birth register.11 From 1983 to 1992 1 083 367 births were recorded. Our study was restricted to singleton or twin pregnancies without congenital malformations, according to ICD-8 (international classification of diseases, 8th revision) and ICD-9 (9th revision) codes 740-759, in women aged 15 to 44 years (n=1 026 249).
Maternal height and smoking habits are recorded at the time of registration for antenatal care. Parity is defined as the number of previous births, including stillbirths. Maternal age is defined as completed years at delivery. Any maternal disorders are noted by an obstetrician at the time of the woman's discharge. Hypertension is defined as12: essential (ICD-8 code 401 and ICD-9 codes 642A-C); gestational (non-proteinuric) (ICD-8 code 637.01 and ICD-9 codes 640D and 642X); mild (proteinuric) pre-eclampsia (ICD-8 code 637.03 and ICD-9 code 642E); severe (proteinuric) pre-eclampsia (ICD-8 code 637.04 and ICD-9 code 642F); and eclampsia (ICD-8 code 637.1 and ICD-9 code 642G). The 250 women with eclampsia were grouped with 5145 women with severe pre-eclampsia.
Birthweight ratio was defined as the ratio of observed to expected birth weight on the basis of the fetuses' gestational age and sex. Explanatory variables were the fetuses' sex, a third degree polynomial of gestational age in days, and interaction between the fetuses' sex and gestational age. A normal birthweight ratio was defined as ≥0.90 and that of mildly and extremely small for gestational age fetuses as >0.75 but <0.90 or ≤0.75 respectively. Late fetal death was defined as a stillbirth delivered at 28 completed weeks of gestation or later. When available, ultrasonography was used to estimate gestational age during the second trimester otherwise gestational age was estimated from the last menstrual period. At the start of the study 50% of the obstetric departments performed routine ultrasonography. From 1990 onwards all pregnant women in Sweden were offered ultrasonography before 18 weeks' gestation. 13 14 We used multiple logistic regression analyses to estimate the effect of independent variables on late fetal death.
Late fetal death rates were increased in women who were 35 years or older, were nulliparous, smoked, or were <155 cm in height (table 1). Essential hypertension, severe pre-eclampsia, and twin pregnancies were associated with greatly increased late fetal death rates. In pregnancies of normal birthweight ratio the late fetal death rate was 2.1 per 1000 compared with 4.7 and 32.8 per 1000 in pregnancies of mildly and extremely small for gestational age fetuses respectively.
Table 1 shows the relation between maternal and fetal characteristics and birthweight ratio. Maternal smoking habits and height influenced the mean birthweight ratio in a dose dependent manner but the largest effects were observed in women with severe pre-eclampsia or twin pregnancies.
Late fetal death rates and mean birthweight ratios were similar for women aged 20-24, 25-29, and 30-34 years. Compared with severe pre-eclampsia other hypertensive disorders were either uncommon or had a comparatively small influence on late fetal death or birthweight ratio. In the logistic regression analyses maternal age was grouped as 15-19, 20-34, and 35-44 years, and hypertension as none, severe pre-eclampsia, and other hypertensive disorders. Multiple logistic regression analyses showed that there was an increased risk of late fetal death in women who were 35 years or older, were nulliparous, smoked, had twin pregnancies, or had severe pre-eclampsia (table 2).
In a second logistic model we estimated the crude effect of being small for gestational age on the risk of late fetal death. Compared with fetuses of a normal birthweight ratio the risk was doubled in fetuses mildly small for gestational age (odds ratio 2.3, 95% confidence interval 2.1 to 2.5) and greatly increased in fetuses extremely small for gestational age (16.5, 15.2 to 17.9).
Risk factors and birthweight ratio
To determine whether the increased risk of late fetal death related to a fetus being small for gestational age was modified by underlying determinants, we introduced interaction terms in the logistic regression models. Important interaction terms were found between birthweight ratio and all determinants studied except maternal age, and predicted rates of late fetal death were calculated as a function of birthweight ratio and its determinants (table 3). In extremely small for gestational age fetuses late fetal death rates ranged from 16 to 45 per 1000, which varied according to underlying determinants. In mildly small for gestational age fetuses late fetal death rates ranged from 2.3 to 8.7 per 1000, and from 1.3 to 4.6 per 1000 in fetuses of a normal birthweight ratio.
In extremely small for gestational age fetuses the predicted late fetal death rate was 33.1 per 1000 in women aged less than 35 years v 44.9 per 1000 in older women. Late fetal death rates also increased with maternal height, and were higher in singletons compared with twins and in non-hypertensive pregnancies compared with pregnancies complicated by hypertensive disorders. Slightly higher late fetal death rates were also observed in nulliparous women and in non-smokers. In mildly small for gestational age fetuses late fetal death rates were increased in women who were 35 years or older, were ≥155 cm in height, had twin pregnancies, or had severe pre-eclampsia. In fetuses of a normal birthweight ratio late fetal death rates were increased among women who were 35 years or older, were <155 cm in height, were nulliparous, smoked, had twin pregnancies, or had severe pre-eclampsia. Of the late fetal deaths, 51% were in the preterm period and 49% were at or after term. When the analyses in table 3 were restricted to full term pregnancies late fetal death rates were reduced: 0.9 per 1000 in non-small for gestational age fetuses and 2.3 and 9.8 in mildly and extremely small for gestational age fetuses respectively. The comparative increase in late fetal death rates by risk factors and birthweight ratio, however, remained unchanged (data not shown).
The impact of maternal height on late fetal death rate is shown when using birthweight ratio as a continuous variable (figure). Late fetal death rates consistently increased with decreasing birthweight ratio regardless of maternal height, but the increase was most pronounced among women ≥175 cm in height.
This study shows that late fetal death rates associated with a fetus being extremely small for gestational age are low in women who are less than 35 years, ≤155 cm in height, have twin pregnancies, or have severe pre-eclampsia or other hypertensive disorders. Hypertensive disorders are well documented determinants of a fetus being small for gestational age, and close antenatal supervision may contribute to the favourable prognosis in these pregnancies. The reduced risks associated with short stature and multiple births, however, support the assumption that a small for gestational age fetus may be the result of constitutional rather than pathological factors.5–8 Even in short mothers, however, late fetal death rates were more than 10-fold higher in extremely small for gestational age fetuses compared with fetuses of a normal birthweight ratio clearly indicating that even in short mothers the consequences of a fetus being small for gestational age may be serious.
In pregnancies of a normal birthweight ratio the risks of late fetal death were influenced by maternal age, smoking habits, multiple births, and severe pre-eclampsia. These factors increase the risks of severe placental complications,15–18 which may cause fetal death without affecting the birthweight ratio.
Late fetal death rates were higher among women older than 35 years regardless of the birthweight ratio. The risk of late fetal death has been reported to increase progressively with gestational age and this increase is pronounced among women older than 35 years. 19 20 Risks of placental complications increase with maternal age, and vascular degenerative changes have been observed in the uterine and myometrial arteries of women of older childbearing age 21 22; this suggests that late fetal death may be due to an age related effect as a consequence of uteroplacental underperfusion.
Chance is an unlikely explanation for our findings because of the large size of the study and the correspondingly narrow confidence intervals of our observed rates and odds ratios. The prospective nature of data collection precludes recall bias. As some risk factors of a fetus being small for gestational age are associated with increased risks of congenital anomalies the study was restricted to pregnancies without congenital malformations.
The conclusions from this investigation are, however, limited by the risk factors included in the birth register. We lacked information on the womens' socioeconomic position, prepregnancy body mass index, and weight gain during pregnancy, which are important determinants of birthweight ratio.1 Socioeconomic position and prepregnancy body mass index are also associated with the risk of late fetal death. 23 24 The association between weight gain during pregnancy and late fetal death, however, is less certain. 24 25 The risks of a fetus being small for gestational age and late fetal death related to maternal age and smoking habits, however, seem to be largely independent of socioeconomic position and prepregnancy body mass index. 1 23 24 26 Maternal height was not an important factor in the overall risk of late fetal death, which agrees with a previous investigation.24 The effect of maternal height on birthweight ratio is reported to be the same both before and after adjusting for confounders. The effect of maternal height on the risk of late fetal death associated with birthweight ratio is therefore unlikely to be due to residual confounding. 27 28 An accurate estimation of the birthweight ratio in cases of late fetal death is limited as estimates of gestational age and fetal weight are based on time of delivery rather than time of death. This not only leads to an overestimation of gestational age but the dead fetus may also have lost weight before delivery,29 and the birthweight ratio may therefore be underestimated in these pregnancies. The extent of this bias is, however, probably limited. Firstly, almost all pregnant women in Sweden follow the routine schedule of visiting their antenatal clinic every second week from 24 weeks' gestation and weekly from 36 weeks, and fetal heart activity is registered at each visit. Secondly, the women are routinely told to contact their antenatal clinic or obstetric department immediately if there is a decrease in fetal movements. Thirdly, the time from diagnosis of a late fetal death to delivery is generally reported as less than 24 hours.30 These factors should ensure a comparatively short delay between the time of fetal death and the time of delivery. In a study of fetal histology and stillbirth in the United States, it was estimated that 80% of all stillbirths were delivered within one week of death.31
Highlighting the risk factors of late fetal death in small for gestational age fetuses is especially important in countries with low infant mortality rates. In the present study, in which all infants with congenital malformations were excluded, late fetal deaths accounted for more than 50% of all late fetal and infant deaths in Sweden from 1983 to 1992. This study shows that the risk of late fetal death in a small for gestational age fetus may be modified by the underlying determinants of birthweight ratio. The very strong relation between late fetal death and a small for gestational age fetus should, however, be re-emphasised; a fetus that is extremely small for gestational age is associated with a high risk of late fetal death, regardless of cause, and must therefore be monitored.
Dr Kramer is a distinguished scientist of the Medical Research Council of Canada.
Contributors: SC participated in the discussion of the study hypothesis and study design, contributed to the analyses, and was mainly responsible for writing the paper; he will act as guarantor of the paper. BH participated in the discussion of the study hypothesis and study design, performed the analyses, and contributed to writing the paper. MSK initiated the study, participated in the discussions of the study hypothesis and study design, contributed to the analyses, and helped write the paper.