SEARCH

SEARCH BY CITATION

Keywords:

  • α-fetopcoteins;
  • chorionic gonadotrophins;
  • risk;
  • stillbirth

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Objective  To determine whether maternal serum levels of alphafetoprotein (α-FP) and human chorionic gonadotrophin (hCG) at 15–21 weeks provided clinically useful prediction of stillbirth in first pregnancies.

Design  Retrospective study of record linkage of a regional serum screening laboratory to national registries of pregnancy outcome and perinatal death.

Setting  West of Scotland, 1992–2001.

Population  A total of 84 769 eligible primigravid women delivering an infant at or beyond 24 weeks of gestation.

Methods  The risk of stillbirth between 24 and 43 weeks was assessed using the Cox proportional hazards model. Logistic regression models within gestational windows were then used to estimate predicted probability. Screening performance was assessed as area under the receiver operating characteristic (ROC) curve.

Main outcome measure  Antepartum stillbirth unrelated to congenital abnormality.

Results  The odds ratio (95% CI) for stillbirth at 24–28 weeks for women in the top 1% were 11.97 (5.34–26.83) for α-FP and 5.80 (2.19–15.40) for hCG. The corresponding odds ratios for stillbirth at or after 37 weeks were 2.44 (0.74–8.10) and 0.79 (0.11–5.86), respectively. Adding biochemical to maternal data increased the area under the ROC curve from 0.66 to 0.75 for stillbirth between 24 and 28 weeks but only increased it from 0.64 to 0.65 for stillbirth at term and post-term. Women in the top 5% of predicted risk had a positive likelihood ratio of 7.8 at 24–28 weeks, 3.7 at 29–32 weeks, 5.1 at 33–36 weeks and 3.4 at 37–43 weeks, and the corresponding positive predictive values were 0.97, 0.33, 0.47 and 0.63%, respectively.

Conclusions  Maternal serum levels of α-FP and hCG were statistically associated with stillbirth risk. However, the predictive ability was generally poor except for losses at extreme preterm gestations, where prevention may be difficult and interventions have the potential to cause significant harm.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Intrauterine fetal death before the onset of labour (antepartum stillbirth) is one of the most common severe complications of pregnancy. It affects approximately 1 in 200 pregnancies and accounts for almost two-thirds of all perinatal deaths.1 Antepartum stillbirth is potentially preventable by elective delivery.2 The effect of this intervention depends on the gestational age. At extreme preterm gestations, delivery may result in either neonatal death or survival with long-term impairment.3 However, elective delivery at term carries a low risk of infant or maternal morbidity and mortality.4,5 Hence, assessment of stillbirth risk and knowledge of the gestational age where the risk is increased is clinically important.

Nulliparous women are at increased risk of antepartum stillbirth.6 Moreover, the outcome of previous pregnancies is one of the most informative predictors of the risk of complications in future pregnancies,7–9 and clearly, this information is not available among nulliparous women. Hence, developing novel predictors of stillbirth risk is particularly important for the care of nulliparous women. Elevated maternal serum levels of alphafetoprotein (α-FP) and human chorionic gonadotrophin (hCG) are associated with an increased risk of antepartum stillbirth.10,11 However, the gestational age dependence of these associations is currently unclear, as are the properties of these measurements as a screening test for stillbirth risk. Moreover, many maternal characteristics are also associated with stillbirth risk.7 The gestational age dependence and screening performance of these, both in isolation and in combination with α-FP and hCG, are also currently unclear.

We obtained complete demographic and second-trimester biochemical screening data on approximately 85 000 nulliparous women and addressed the following aims: (1) to assess the relationship between maternal characteristics and serum screening data and the risk of stillbirth, (2) to determine whether the associations between maternal and biochemical factors varied according to gestational age and (3) to characterise the screening properties of a combined model at different gestational ages.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Data sources and woman selection

The Scottish Morbidity Record collects information on clinical and demographic characteristics and outcomes for all women discharged from Scottish maternity hospitals. The register is subjected to regular quality assurance checks and has been greater than 99% complete since the late 1970s.12 The Scottish Stillbirth and Infant Death Enquiry is a national register that routinely classifies all perinatal deaths in Scotland.1 All women attending for antenatal care in the West of Scotland are offered biochemical screening, using maternal serum levels of α-FP and hCG to assess their risk of having a fetus affected by Down syndrome and structural fetal abnormality, and approximately 81% of them accept screening.13 The laboratory information management system for the West of Scotland antenatal screening programme in the Institute of Medical Genetics in Glasgow contains a database of the maternal information and biochemical screening results, and electronic storage of these data in their current form was commenced in September 1991. A probability-based matching approach14 was employed using maternal identifiers to link the Scottish Morbidity Record, the Scottish Stillbirth and Infant Death Enquiry and the antenatal screening database in the Institute of Medical Genetics. We excluded multiple births and births outside the range 24–43 weeks of gestation. Ethical approval for the linkage was obtained from the Privacy Advisory Committee of the Information and Statistics Division of the NHS, Scotland.

Definitions

Maternal age, parity, postcode of residence and all outcome data were obtained solely from the Scottish Morbidity Record. Maternal weight was obtained solely from the biochemical database. Maternal height and smoking were obtained from the Scottish Morbidity Record, except in cases where they were missing from the Scottish Morbidity Record, the biochemical database was employed. The smoking status (current, past, never) was determined from information at the time of the first antenatal visit. Maternal age was defined as the age of the mother at the time of delivery. Body mass index (BMI) was calculated from the weight in kilogram recorded at the time of sampling for α-FP assay divided by the height in metres squared. Socio-economic status was estimated based on the postcode of residence, using Carstairs socio-economic deprivation categories15 (based on 1991 Census data on car ownership, unemployment, overcrowding and social class within postcode sectors of residence which contain, on average, around 1600 residents).

The gestational age at birth was defined as the completed weeks of gestation on the basis of the estimated date of delivery in each woman’s clinical record; standard national criteria exist for the estimation of date of delivery using menstrual and ultrasound data.16 The gestational age has been confirmed by ultrasound scan in the first half of pregnancy in more than 95% of pregnancies in the UK from the early 1990s.17 The birthweight was categorised into sex- and gestational-age-specific percentiles, as previously described in detail.18 Small for gestational age (SGA) was defined as birthweight in the smallest 5% for sex and week of gestation of delivery, and those above this threshold were considered appropriate for gestational age (AGA). Maternal serum levels of α-FP and hCG were quantified as multiples of the median (MoM) for week of gestation, corrected for maternal weight.19

The cause of antepartum stillbirth was classified according to a modified version of the Wigglesworth hierarchical system which is described in detail elsewhere.1 Stillbirths were defined as antepartum or intrapartum, and the cause of stillbirth was classified according to direct obstetric causes (in order): congenital abnormality, pre-eclampsia, haemorrhage (antepartum), mechanical, maternal, miscellaneous and unexplained. The hierarchy dictates that a perinatal death where there was severe pre-eclampsia complicated by abruption would be classified as being caused by pre-eclampsia, as pre-eclampsia is above haemorrhage in the hierarchy. Similarly, a stillbirth where the infant was SGA and the mother had pre-eclampsia would be defined as being caused by pre-eclampsia. Deaths caused by congenital anomaly were defined as ‘any structural or genetic defect incompatible with life or potentially treatable but causing death’ and were excluded. Hence, women who had a loss or therapeutic abortion where the fetus was affected by structural or chromosomal abnormalities associated with maternal serum screening were excluded from the analysis. Stillbirths caused by mechanical, maternal and miscellaneous causes were combined into the category of ‘other’. There was no information on whether women had an amniocentesis. However, serum screening was performed no later than 21 weeks, and all procedure-related losses would be expected to have occurred prior to 24 weeks. Classification is performed by a single, medically qualified individual (the Scottish Coordinator) with the results of postmortem investigations, where obtained.

Statistical analysis

Continuous variables were summarised by the median and interquartile range (IQR). Univariate comparisons were performed using the Mann–Whitney U test, the chi-square test and the chi-square test for trend, as appropriate. The P values for all the hypothesis tests were two sided, and statistical significance was assumed at P < 0.05. All the variables were treated as categorical. Biochemical data were categorised into quintiles, with the last quintile split into the top 5, 6–10 and 11–20%. For detailed analysis, the top 5% was further subdivided into the top 1, 2–3 and 4–5%. We preformed univariate and multivariate Cox regression using gestational age as the time scale, antepartum stillbirth as the event, and all other births as censored. The proportional hazards assumption was tested using the method of Grambsch and Therneau.20 The rationale for and advantages of this approach are discussed in detail elsewhere.21 Risk factors for stillbirth within four gestational windows were then explored using logistic regression analysis. Each woman contributed one record for each gestational window covered by her pregnancy. Stillbirth within the gestational window was taken to be the event, and all births at that or a later gestation were taken as the denominator. Factors whose associations with stillbirth had been shown in the proportional hazards model to vary significantly with gestation were included with an interaction term with gestational window, whereas the rest of the predictors were included without interactions. The predicted probability of stillbirth in each gestational window was obtained from the logistic model. The performance of the model in each gestational window was assessed via the area under the receiver operating characteristic (ROC) curve. Further, the positive predictive value, sensitivity, specificity and positive and negative likelihood ratios were estimated using different thresholds of predicted risk as screen positive, specifically, the top 5, 10 and 20%. All statistical analyses were performed using the Stata software package (Stata Corporation, College Station, TX, USA), version 8.2.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

There were 216 563 records of singleton births with a recorded value of maternal serum α-FP and hCG levels which could be linked to Scottish morbidity record 2, and 97 264 (44.9%) of these were first births. Among this group, we excluded 69 (0.1%) deliveries that were outside the range of 24–43 gestational weeks and 133 (0.1%) deaths caused by fetal abnormality or rhesus isoimmunisation. Of the remaining 97 062 records, 12 293 (12.7%) had missing data on one or more variable (2 [<0.1%] maternal age, 1297 [1.3%] height, 6927 [7.1%] BMI, 171 [0.2%] deprivation category, 5724 [5.9%] smoking status and 5 [0.01%] previous spontaneous early pregnancy losses), leaving a study group of 84 769 births (87.3% of eligible births).

The upper limits of percentiles (all expressed as MoM) for the 20th, 40th, 60th, 80th, 90th, 95th, 97th and 99th percentiles were 0.77, 0.93, 1.10, 1.35, 1.59, 1.83, 2.03 and 2.52, respectively, for α-FP and 0.66, 0.88, 1.13, 1.53, 1.92, 2.33, 2.65 and 3.44, respectively, for hCG. Women who experienced antepartum stillbirth in their first pregnancy were shorter, had higher median BMI, were more likely to live in an area of high socio-economic deprivation and were more likely to smoke (Table 1). There was no difference in median maternal age or marital status in relation to stillbirth. Women who experienced stillbirths had higher second-trimester levels of α-FP and hCG. An elevated value of α-FP was associated with an increased risk of stillbirth caused by pre-eclampsia, haemorrhage and of unexplained stillbirth (Table 2). The association with unexplained stillbirth was stronger where the birthweight was SGA. An elevated level of hCG was strongly associated with the risk of stillbirth caused by pre-eclampsia and weakly associated with unexplained stillbirth. The latter relationship was similar for AGA and SGA losses (Table 3).

Table 1.  Comparison of maternal and outcome characteristics by occurrence of antepartum stillbirths
 No antepartum stillbirth (n= 84 363)Antepartum stillbirth (n= 406)P*
  • *

    Mann–Whitney U test, chi-square test or chi-square test for trend, as appropriate.

Maternal characteristics
Age, median (IQR)26 (22–30)27 (22–31)0.1
Height, median (IQR)163 (158–167)162 (157–167)0.007
BMI, median (IQR)23.5 (21.4–26.3)24.0 (21.8–27.9)<0.001
Deprivation category
 1 (least deprived)3121 (3.7)6 (1.5)0.02
 28729 (10.4)33 (8.1) 
 315 481 (18.4)66 (16.3) 
 420 979 (24.9)108 (26.6) 
 514 219 (16.9)64 (15.8) 
 613 143 (15.6)79 (19.5) 
 7 (most deprived)8691 (10.3)50 (12.3) 
Smoking status
 Never52 202 (61.9)185 (45.6)<0.001
 Current24 258 (28.8)185 (45.6) 
 Former7903 (9.4)36 (8.9) 
Marital status
 Married43 375 (51.4)195 (48.0)0.2
 Other40 988 (48.6)211 (52.0) 
Maternal serum α-FP (multiple of the median), median (IQR)1.03 (0.83–1.29)1.16 (0.89 to 1.46)<0.001
Maternal serum hCG (multiple of the median), median (IQR)1.03 (0.74–1.45)1.09 (0.74 to 1.59)0.03
Gestational age at birth (weeks)
24–28328 (0.4)106 (26.1)<0.001
29–32782 (0.9)75 (18.5) 
33–364135 (4.9)76 (18.7) 
37–4379 118 (93.8)149 (36.7) 
Table 2.  Association between second-trimester maternal serum levels of α-FP and different causes of stillbirth
Causes of stillbirthα-FP ≤ 95th percentile (n= 80 408)α-FP > 95th percentile (n= 4361)Odds ratio95% CIP*
Experienced outcomePercentExperienced outcomePercent
  • CI, confidence interval.

  • *

    Chi-square test or Fisher’s exact test as appropriate.

All causes3530.43531.212.792.09–3.73<0.001
Pre-eclampsia260.03110.257.823.91–15.62<0.001
Haemorrhage430.0570.163.001.38–6.550.005
Unexplained (All)2500.31310.712.301.58–3.33<0.001
Unexplained (SGA)620.07120.283.581.94–6.58<0.001
Unexplained (AGA)1880.23190.441.871.17–2.980.009
Other340.0440.092.170.80–5.870.1
Table 3.  Association between second-trimester maternal serum levels of hCG and different causes of stillbirth
Causes of stillbirthhCG ≤ 95th percentile (n= 80 408)hCG > 95th percentile (n= 4361)Odds ratio95% CIP*
Experienced outcomePercentExperienced outcomePercent
  • CI, confidence interval.

  • *

    Chi-square test or Fisher’s exact test as appropriate.

All causes3650.45410.941.931.39–2.66<0.001
Pre-eclampsia260.03110.257.243.62–14.45<0.001
Haemorrhage450.0650.111.900.78–4.640.2
Unexplained (All)2580.32230.531.521.00–2.330.05
Unexplained (SGA)680.0860.141.510.67–3.400.3
Unexplained (AGA)1900.24170.391.530.93–2.500.09
Other360.0420.050.950.00–3.57>0.9

When the risk of all-cause stillbirth was assessed using a Cox proportional hazards model, it was positively associated with maternal age, deprivation category and BMI and negatively associated with height (Table 4). The strength of association with these maternal characteristics did not significantly vary over the range of gestation from 24 to 43 weeks. The risk of stillbirth was also positively associated with second-trimester maternal serum levels of α-FP and hCG. The cumulative probability of stillbirth associated with elevated levels of these analytes is plotted, and the strength of association varied with gestational age (Figure 1). High levels of α-FP and hCG were strongly associated with stillbirth between 24 and 28 weeks of gestation. In both cases, the strength of association became weaker with advancing gestational age, and there was no statistically significant association between levels of these analytes and the risk of stillbirth at or after 37 weeks of gestation (Table 5).

Table 4.  Risk of antepartum stillbirth in relation to maternal second-trimester serum biochemistry and demographic characteristics
VariablesUnivariateMultivariate
Hazard ratio95% CIPHazard ratio95% CIP
  1. CI, confidence interval.

  2. Test of proportional hazards assumption: *P < 0.05,**P < 0.01, all others P≥ 0.05.

  3. Global test of proportional hazards assumption for multivariate model: P= 0.05.

α-FP percentile
1–200.840.59–1.20 0.870.61–1.25 
21–400.960.68–1.36 0.990.70–1.39 
41–60Reference Reference 
61–801.340.98–1.84 1.300.95–1.78 
81–901.441.00–2.08 1.360.95–1.96 
91–951.490.94–2.35 1.330.84–2.10 
96–971.760.95–3.26 1.600.86–2.95 
98–992.861.70–4.81 2.481.47–4.17 
1008.13*5.07–13.06 6.343.92–10.25 
Trend test <0.001 <0.001
Heterogeneity test <0.001 <0.001
hCG percentile
1–201.210.88–1.66 1.080.79–1.48 
21–400.830.59–1.17 0.800.57–1.13 
41–60Reference Reference 
61–801.040.76–1.43 1.100.80–1.51 
81–901.160.80–1.67 1.270.87–1.84 
91–951.29**0.82–2.01 1.44**0.91–2.25 
96–971.530.83–2.81 1.700.92–3.15 
98–992.11*1.25–3.58 2.28*1.34–3.89 
1003.03*1.65–5.58 3.10*1.67–5.77 
Trend test 0.002 <0.001
Heterogeneity test <0.001 <0.001
Age (years)
<201.120.82–1.53 0.890.63–1.25 
20–241.030.78–1.35 0.860.65–1.15 
25–29Reference Reference 
30–341.200.92–1.57 1.280.98–1.67 
35–391.741.19–2.53 1.791.23–2.61 
>392.491.02–6.09 2.290.94–5.62 
Trend test 0.04 <0.001
Heterogeneity test 0.02 0.004
Deprivation category
1 (least deprived)0.380.17–0.86 0.420.19–0.97 
20.730.50–1.08 0.770.52–1.14 
30.830.61–1.13 0.860.63–1.16 
4Reference Reference 
50.880.64–1.20 0.850.63–1.16 
61.180.88–1.58 1.140.85–1.53 
7 (most deprived)1.150.82–1.60 1.050.75–1.48 
Trend test 0.001 0.02
Heterogeneity test 0.02 0.09
Height (cm)
<1501.070.44–2.62 0.880.36–2.16 
150–1541.591.12–2.26 1.471.03–2.10 
155–1591.481.12–1.95 1.441.10–1.90 
160–164Reference Reference 
165–1691.180.90–1.55 1.200.92–1.58 
170–1740.850.59–1.24 0.890.61–1.29 
>1741.080.64–1.83 1.120.66–1.90 
Trend test 0.006 0.045
Heterogeneity test 0.01 0.05
Smoking
NonsmokerReference Reference 
Current smoker2.191.78–2.68 2.481.98–3.11 
Ex-smoker1.280.90–1.83 1.400.98–2.01 
Heterogeneity test <0.001 <0.001
BMI (kg/m2)
<201.431.05–1.97 1.360.99–1.87 
20–24Reference Reference 
25–291.471.16–1.86 1.441.13–1.82 
>302.121.60–2.81 1.961.47–2.60 
Trend test <0.001 <0.001
Heterogeneity test <0.001 <0.001
Marital status
MarriedReference Reference 
Other1.150.95–1.40 1.030.82–1.29 
Heterogeneity test 0.2 0.82
image

Figure 1. Kaplan–Meier plot of cumulative probability of stillbirth from 24 to 43 weeks of gestation comparing women with elevated (top 5%) maternal serum levels (solid line) with all other women (dashed line): (A) α-FP and (B) hCG. Hazard ratio for top 5% (95% CI) is 3.33 (2.32–4.77) for α-FP and 2.06 (1.41–3.02) for hCG. Test of proportional hazards assumption: P < 0.05 for both and remained statistically significant when included in multivariate Cox model with all other maternal characteristics.

Download figure to PowerPoint

Table 5.  Adjusted odds ratios for the risk of antepartum stillbirth in relation to maternal serum levels of α-FP and hCG for four different gestational age categories
VariablesGestational age of 24–28 weeks, n= 84 769Gestational age of 29–32 weeks, n= 84 335Gestational age of 33–36 weeks, n= 83 478Gestational age of 37–43 weeks, n= 79 267
OR95% CIP valueOR95% CIP valueOR95% CIP valueOR95% CIP value
  1. CI, confidence interval, HR, hazard ratio; OR, odds ratio.

  2. The OR for the other variables were virtually identical to the adjusted hazard ratios listed in Table 4. The only OR which differed from the HR at the first decimal place was for maternal age greater than 39 years: the HR was 2.29 and the adjusted OR was 2.14.

Maternal serum α-FP percentile
1–200.780.33–1.820.60.870.40–1.910.71.290.61–2.710.50.750.42, 1.350.3
21–401.180.55–2.510.70.610.25–1.450.30.580.23–1.440.21.290.78, 2.130.3
41–60ReferenceReferenceReferenceReference
61–801.410.70–2.830.31.070.52–2.200.81.530.77–3.070.21.240.75–2.030.4
81–902.121.02–4.420.041.030.43–2.470.91.540.69–3.450.31.000.54–1.88>0.9
91–952.421.06–5.540.041.520.58–3.960.40.550.12–2.440.40.940.41–2.160.9
96–973.041.08–8.550.042.450.80–7.480.11.010.30–3.32>0.9
98–994.061.61–10.230.0032.570.84–7.860.102.120.60–7.470.21.450.51–4.150.5
10011.975.34–26.83<0.0014.791.54–14.860.0075.761.85–17.960.0032.440.74–8.100.1
Maternal serum hCG percentile
1–200.800.37–1.710.61.680.80–3.530.20.890.46–1.730.71.070.65–1.750.8
21–400.960.47–2.000.91.110.49–2.530.80.500.22–1.110.090.760.44–1.310.3
41–60ReferenceReferenceReferenceReferent
61–801.200.60–2.390.61.200.54–2.680.70.800.40–1.610.51.200.74–1.960.5
81–901.990.97–4.090.061.120.41–3.040.81.160.53–2.520.71.010.53–1.90>0.9
91–952.861.31–6.260.0092.580.99–6.670.050.700.21–2.380.60.710.28–1.840.5
96–973.871.49–10.050.0051.860.41–8.450.40.590.08–4.400.61.130.34–3.700.8
98–995.202.25–12.01<0.0011.710.38–7.780.51.130.26–4.900.871.480.52–4.220.5
1005.802.19–15.40<0.0014.901.34–17.940.022.160.49–9.470.30.790.11–5.860.8

There were no statistically significant interactions between α-FP and hCG and maternal characteristics and no significant interactions between the two analytes. Because of the large number of comparisons, the threshold for statistical significance was set at P < 0.01. There was a trend towards an interaction between α-FP levels in the top 5% and hCG levels in the top 5%: the hazard ratio for the interaction term was 2.20 (95% CI 1.06–4.56, P= 0.03). Four hundred and seventy-nine women had levels of both analytes in the top 5%, and 14 (2.9%) of these pregnancies ended in stillbirth. The hazard ratio for stillbirth among this group was 7.3 (95% CI 4.3–12.5, P < 0.001). Again, the association declined significantly with advancing gestational age (P < 0.05), and there was no statistically significant association at term (OR 3.0, 95% CI 0.7–12.1, P= 0.1), although the numbers were too small to exclude a significant association.

The screening characteristics of the model were assessed at the different gestational ages using the predicted probability of stillbirth from the logistic regression models (Table 6). The top 5% of the population for predicted risk had a likelihood ratio for stillbirth of more than seven for 24–28 weeks. The likelihood ratio associated with the top 5% of predicted risk steadily declined with advancing category of gestational age and was lowest for stillbirth at or beyond 37 weeks, being approximately three. The screening characteristics of the predicted probabilities were then assessed using the area under the ROC curve, and models containing only maternal variables were compared with those which also included second-trimester serum screening results (Table 7). Adding biochemical data significantly increased the area under the ROC curve for stillbirth between 24–28 weeks and 29–32 weeks but had a minimal effect for stillbirth prediction at term.

Table 6.  Screening performance of the predicted probability from multivariate logistic regression models containing maternal and biochemical data
Gestational age and predicted riskPositive predicted value (%)Sensitivity (%)Specificity (%)Positive LRNegative LR
  1. LR, likelihood ratio.

  2. The risk (per 1000 subsequent births) of antepartum stillbirth is as follows: 1.25 (95% CI 1.01–1.49) for gestational age group 24–28 weeks, 0.89 (95% CI 0.69–1.09) for gestational age group 29–32 weeks, 0.91 (95% CI 0.71–1.11) for gestational age group 33–36 weeks and 1.88 (95% CI 1.58–2.18) for gestational age 37–43 weeks.

24–28 weeks
Top 5%0.9736.6895.047.800.65
Top 10%0.5241.5190.044.170.65
Top 20%0.3454.7280.042.740.57
29–32 weeks
Top 5%0.3318.6795.013.740.86
Top 10%0.2832.0090.023.210.76
Top 20%0.2044.0080.022.200.70
33–36 weeks
Top 5%0.4725.0095.135.130.79
Top 10%0.3234.2190.233.500.73
Top 20%0.2146.0580.452.360.67
37–43 weeks
Top 5%0.6316.7895.023.380.88
Top 10%0.4222.1590.032.220.86
Top 20%0.3335.5780.031.780.81
Table 7.  Area under the ROC curve from the two multivariate logistic regression models for the four different gestational age groups
Model predictorsGestational age, 24–28 weeksGestational age, 29–32 weeksGestational age, 33–36 weeksGestational age, 37–43 weeks
Area*95% CIArea*95% CIArea*95% CIArea*95% CI
  • CI, confidence interval.

  • *

    Area under the ROC curve.

Maternal only0.660.61–0.710.690.64–0.750.640.58–0.700.640.60–0.68
Maternal and biochemical0.750.70–0.800.740.69–0.790.710.65–0.770.650.61–0.69

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The main finding of this paper is that women with elevated serum levels of α-FP and hCG in the second trimester of pregnancy were at increased risk of stillbirth and that these associations were strongest for stillbirth at extreme preterm gestations. In contrast, the association between stillbirth and maternal characteristics was similar over the period of 24–43 weeks. Consequently, a model combining maternal characteristics and biochemical data performed reasonably well in predicting stillbirth risk at extreme preterm gestations (24–28 weeks): women in the top 5% of predicted risk had a positive likelihood ratio of seven and included 36% of all losses between 24 and 28 weeks of gestation. In contrast, a model combining maternal characteristics and biochemical data performed poorly at predicting stillbirth at term: women in the top 5% of predicted risk had a positive likelihood ratio of three and only included 15% of losses at or after 37 weeks of gestation.

These findings are of clinical relevance. Measurement of α-FP and hCG is in widespread use as a well validated method of screening for Down syndrome.13 It is generally recognised that elevated levels of these proteins are associated with an increased risk of other adverse outcomes, including stillbirth. Some authors suggest that women with elevated levels of α-FP or hCG should have close surveillance of their pregnancies, such as growth scans at intervals of 2–4 weeks.22 However, the current data suggest that if surveillance is thought to be beneficial, it should commence at 24 weeks. Moreover, if serial scans are reassuring up to 36 weeks, it may well be safe to discontinue further assessment.

The current analysis also addresses whether selecting women for further fetal assessment on the basis of second-trimester serum screening data is justifiable. There are no studies that directly indicate that the use of α-FP and hCG measurements to screen for stillbirth improves outcome. A successful screening programme has two major components, namely (1) effective assessment of risk and (2) application of an effective intervention among women who screen positive. The primary intervention to prevent stillbirth is delivery of the infant prior to intrauterine demise. In assessing the effect of delivery of high-risk infants on overall perinatal mortality, the gestational age where the fetus is presumed to be at risk is crucial. The risk of neonatal death is around 30–40% at 24–25 weeks23 and less than 0.1% at term.5 Therefore, at term, elective delivery of an infant at high risk of stillbirth carries a small risk of increasing overall perinatal mortality. Hence, if a risk factor is strongly predictive of stillbirth at term, elective delivery at term is likely to be beneficial and is unlikely to cause serious harm. This practice is widespread in clinical obstetrics. Women with risk factors such as insulin-dependent diabetes mellitus or a history of stillbirth are commonly offered elective delivery at around 38 weeks of gestation. However, the current data do not suggest that routine elective delivery at term is indicated on the basis of elevated levels of α-FP or hCG in the second trimester.

The stronger association with stillbirth at preterm gestations raises the possibility that α-FP and hCG measurements may be used to screen for fetuses at risk of stillbirth at preterm gestations. However, the management of a pregnancy where the fetus is thought to be at increased risk of stillbirth at preterm gestations is more complex. Elective delivery at extreme preterm gestations would clearly not simply be on the basis of a raised α-FP or hCG level. Rather, women with a raised α-FP or hCG would have an increased level of fetal surveillance (such as ultrasound scanning and computerised cardiotocography), and those with severely abnormal findings would be considered for delivery. However, no assessment of fetal wellbeing is completely accurate. Therefore, identifying and delivering fetuses at risk of stillbirth between 24 and 28 weeks could lead to neonatal deaths as a consequence of iatrogenic prematurity among infants who would not have died had they been left in utero. Moreover, even if the infants survived, elective delivery could lead to long-term severe disability as approximately 40% of long-term survivors of extreme preterm birth may be severely disabled.3 It is possible that some parents would prefer a stillbirth to occur than have delivery of an infant which would be severely disabled in the long term. Furthermore, delivery at extreme preterm gestations also carries issues of maternal morbidity. Delivery would usually be by caesarean section, often using the classical method. This carries immediate risks of maternal morbidity and is associated with significant risks in future pregnancies.24 Thus, the beneficial effects of preventing stillbirth in those who would have died may be offset by the harmful effects among those who were false positives. Finally, the use of α-FP and hCG measurements to screen for stillbirth has not been subjected to an economic analysis. From the above, even if it did result in decreased perinatal mortality, the costs of preventing each stillbirth may be substantial, including costs of fetal monitoring, elective preterm delivery, neonatal intensive care and long-term costs, including management of future pregnancy (e.g. after classical caesarean section) and care of disabled infants. We conclude that although statistical associations can be shown between stillbirth risk and both biochemical and maternal data, population-based screening for stillbirth risk on the basis of these associations should not be undertaken without direct evidence that screening and intervention improve outcome and are economically justifiable, particularly when performed at extreme preterm gestations.

The current data are also of biological interest. This is the first large-scale study of α-FP and hCG which had detailed information on the cause of antepartum stillbirth and it considerably extends the understanding of the association between maternal serum markers in the second trimester of pregnancy and the risk of stillbirth. An elevated level of α-FP was most strongly associated with stillbirth caused by pre-eclampsia (seven- to eight-fold risk), was moderately associated with stillbirth associated with haemorrhage (primarily abruption) and SGA unexplained stillbirth (three- to four-fold risk) and weakly associated with unexplained stillbirths which had AGA birthweight (two-fold risk). Elevated levels of hCG were also strongly associated with stillbirth caused by pre-eclampsia (seven- to eight-fold risk) and weakly associated with unexplained stillbirth (1.5-fold risk) The latter seemed to be unrelated to fetal growth. Finally, there was a trend towards a synergistic association between elevated α-FP and hCG and the risk of antepartum stillbirth. These data indicate the complexity of placental determinants of stillbirth caused by different causes. The biochemical data employed in the present study did not perform well as uterine artery Doppler flow velocimetry as a predictor of stillbirth. A recent study of more than 30 000 women has shown that the area under the ROC curve for this modality was 0.84 for losses before 32 weeks, and the women in the top 5% of predicted risk on the basis of Doppler alone had a positive likelihood ratio of 12.1.25 However, similar to α-FP and hCG, adding uterine artery Doppler data to a model including maternal characteristics had a minimal effect (increase of 0.02) on the area under the ROC curve for stillbirths at later gestations.

In conclusion, we show that elevated levels of α-FP and hCG were strongly associated with stillbirth at extreme preterm gestations, and the strength of association with stillbirth declined with advancing gestational age. These findings have implications for the clinical use of second-trimester maternal serum levels of α-FP and hCG in obstetric risk assessment.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The study was supported by a project grant from the Foundation for the Study of Infant Deaths (UK).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • 1
    Information and Statistics Division NHS Scotland. Scottish Perinatal and Infant Mortality Report 2000. Edinburgh, UK: ISD Scotland Publications, 2001.
  • 2
    Fretts RC, Elkin EB, Myers ER, Heffner LJ. Should older women have antepartum testing to prevent unexplained stillbirth? Obstet Gynecol 2004;104:5664.
  • 3
    Marlow N, Wolke D, Bracewell MA, Samara M. Neurologic and developmental disability at six years of age after extremely preterm birth. N Engl J Med 2005;352:919.
  • 4
    Hannah ME, Hannah WJ, Hellmann J, Hewson S, Milner R, Willan A. Induction of labor as compared with serial antenatal monitoring in post-term pregnancy. A randomized controlled trial. The Canadian Multicenter Post-term Pregnancy Trial Group. N Engl J Med 1992;326:158792.
  • 5
    Smith GC. Life-table analysis of the risk of perinatal death at term and post term in singleton pregnancies. Am J Obstet Gynecol 2001;184:48996.
  • 6
    Raymond EG, Cnattingius S, Kiely JL. Effects of maternal age, parity, and smoking on the risk of stillbirth. Br J Obstet Gynaecol 1994;101:3016.
  • 7
    Cnattingius S, Stephansson O. The epidemiology of stillbirth. Semin Perinatol 2002;26:2530.
  • 8
    Surkan PJ, Stephansson O, Dickman PW, Cnattingius S. Previous preterm and small-for-gestational-age births and the subsequent risk of stillbirth. N Engl J Med 2004;350:77785.
  • 9
    Smith GCS, Shah I, White IR, Pell JP, Dobbie R. Previous pre-eclampsia, preterm delivery and delivery of a small for gestational age infant and the risk of unexplained stillbirth in the second pregnancy: a retrospective cohort study, Scotland 1992-2001. Am J Epidemiol 2007;165:194202.
  • 10
    Waller DK, Lustig LS, Smith AH, Hook EB. Alpha-fetoprotein: a biomarker for pregnancy outcome. Epidemiology 1993;4:4716.
  • 11
    Spencer K. Second-trimester prenatal screening for Down syndrome and the relationship of maternal serum biochemical markers to pregnancy complications with adverse outcome. Prenat Diagn 2000;20:6526.
  • 12
    Cole S. Scottish maternity and neonatal records. In: ChalmersI, McllwaineG, editors. Perinatal Audit and Surveillance. Proceedings of the 8th Study Group of the Royal College of Obstetricians and Gynaecologists. London: RCOG; 1980. pp. 3952.
  • 13
    Crossley JA, Aitken DA, Berry E, Connor JM. Impact of a regional screening programme using maternal serum alpha fetoprotein (AFP) and human chorionic gonadotrophin (hCG) on the birth incidence of Down’s syndrome in the west of Scotland. J Med Screen 1994;1:1803.
  • 14
    Kendrick S, Clarke J. The Scottish record linkage system. Health Bull (Edinb) 1993;51:729.
  • 15
    McLoone P, Boddy FA. Deprivation and mortality in Scotland, 1981 and 1991. BMJ 1994;309:146570.
  • 16
    Evans E, Farrant P, Gowland M, McNay MB, Richards B. Clinical Applications of Ultrasonic Fetal Measurements. London: British Medical Ultrasound Society/British Institute of Radiology, 1990.
  • 17
    Campbell S, Soothill P. Detection and management of intrauterine growth retardation: a British approach. In: ChervenakFA, IsaacsonGC, CampbellS, editors. Ultrasound in Obstetrics and Gynecology. Boston, MA: Little Brown and Company;1993 pp. 14315.
  • 18
    Smith GCS. Sex, birth weight and the risk of stillbirth in Scotland, 1980-1996. Am J Epidemiol 2000;151:61419.
  • 19
    Neveux LM, Palomaki GE, Larrivee DA, Knight GJ, Haddow JE. Refinements in managing maternal weight adjustment for interpreting prenatal screening results. Prenat Diagn 1996;16:111519.
  • 20
    Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;81:51526.
  • 21
    Smith GC. Estimating risks of perinatal death. Am J Obstet Gynecol 2005;192:1722.
  • 22
    Coppens M, James DK. Planning prenatal care and identification of risk (screening). In: JamesDK, SteerPJ, WeinerCP, GonikB, editors. High Risk Pregnancy: Management Options. Philadelphia, PA: Elsevier; 2006. pp. 6783.
  • 23
    Alexander GR, Kogan M, Bader D, Carlo W, Allen M, Mor J. US birth weight/gestational age-specific neonatal mortality: 1995-1997 rates for whites, hispanics, and blacks. Pediatrics 2003;111:e616.
  • 24
    Greene RA, Fitzpatrick C, Turner MJ. What are the maternal implications of a classical caesarean section? J Obstet Gynaecol 1998;18:3457.
  • 25
    Smith GCS, Yu CK, Papageorghiou AT, Cacho AM, Nicolaides KH. Maternal uterine artery Doppler flow velocimetry and the risk of stillbirth. Obstet Gynecol 2007;109:14451.