Screening for small-for-gestational-age fetuses

Abstract

Introduction

It is well established that correct antenatal identification of small-for-gestational-age (SGA) fetuses reduces their risk of adverse perinatal outcome with long-term consequences. Ultrasound estimates of fetal weight (EFW_us) are the ultimate tool for this identification. It can be conducted as a “universal screening”, that is, all pregnant women at a specific gestational age. However, in Denmark it is conducted as “selective screening”, that is, only on clinical indication. The aim of this study was to assess the performance of the Danish national SGA screening program and the consequences of false-positive and false-negative SGA cases.

Material and methods

In this retrospective cohort study, we included 2928 women with singleton pregnancies with due dates in 2015. We defined “risk of SGA” by an EFW_us ≤ −15% of expected for the gestational age and “SGA” as birthweight ≤−22% of expected for gestational age.

Results

At birth, the prevalence of SGA was 3.3%. The overall sensitivity of the Danish screening program was 62% at a false-positive rate of 5.6%. Within the entire cohort, 63% had an EFW_us compared with 79% of the SGA cases. The sensitivity was 79% for those born before 37 weeks of gestation but only 40% for those born after 40 weeks of gestation. The sensitivity was also associated with birthweight deviation; 73% among extreme SGA cases (birthweight deviation ≤−33%) and 55% among mild SGA (birthweight deviation between −22% and −27%). False diagnosis of SGA was associated with an increased rate of induction of labor (OR_adj = 2.51, 95% CI 1.70-3.71) and cesarean section (OR_adj = 1.44, 95% CI 0.96-2.18).

Conclusions

The performance of the Danish national screening program for SGA based on selective EFW_us on clinical indication has improved considerably over the last 20 years. Limitations of the program are the large proportion of women referred to ultrasound scan and the low performance post-term.

Abbreviations

AGA

appropriate-for-gestational-age

BW

birthweight

EFW

estimated fetal weight

EFW_us

ultrasound-based estimation of fetal weight

FPR

false-positive rate

GA

gestational age

OR

odds ratio

OR_adj

adjusted odds ratio

SGA

small-for-gestational-age

Key message

We assessed the performance of the Danish screening program for small-for-gestational-age fetuses for the first time in 20 years. Performance has improved considerably. In spite of a large proportion of women referred for additional ultrasound, the detection of small-for-gestational-age babies post-term remains low.

1 INTRODUCTION

It is well established that small-for-gestational-age (SGA) fetuses are at increased risk of adverse perinatal outcome1 and long-term consequences such as metabolic syndrome2 and cardiovascular disease.3 Antenatal detection of SGA improves the perinatal outcome by enabling timely delivery.4 Unfortunately, false suspicion of SGA may increase the risk of unnecessary obstetric interventions and thereby increase the risk of adverse neonatal outcome.5

When using ultrasound in the screening for SGA, there are two approaches; “universal ultrasound screening” with routine ultrasound estimation of fetal weight (EFW_us) conducted in all pregnant women at a specific gestational age (GA) and “selective ultrasound screening” with EFW_us conducted only on clinical indication.

Universal EFW_us screening has a sensitivity of 68%-77% (birthweight [BW] ≤3rd centile) at a false-positive rate (FPR) of 5%-13%6, 7 compared with selective EFW_us screening with a sensitivity of 29%-32% at an FPR of 3%.7, 8 The performance of selective ultrasound screening is highly depended on the clinical indications used for referral of patients to EFW_us, which dictates the proportion of women referred for ultrasound.

In Denmark, SGA is defined by a BW below −22% of the expected for GA9 and the prevalence is approximately 3% among singleton pregnancies.8 The routine antenatal fetal growth assessment includes clinical examination and symphysis-fundal height measurements performed by midwives and general practitioners every 3-4 weeks in pregnancy from 14 weeks of gestation until delivery. Only high-risk pregnancies based on the 1^st-trimester risk stratification (previous obstetric or medical history) and complications in current pregnancy are referred to obstetric control including EFW_us. If EFW_us is ≤−15% of the expected weight for the GA, the fetus will be considered at risk of SGA.10

The most recent publication on selective EFW_us screening in Denmark was based on data from 1997-1998, where only 3.7% had an EFW, giving a sensitivity of 29% at an FPR of 0.26%.8 However, based on clinical experience, the proportion of women referred for EFW_us has increased considerably over the last decades. Therefore, the actual performance of the screening for SGA in Denmark is currently unknown.

The aim of this study was to investigate the performance of the Danish national screening program for SGA including selective EFW_us. In addition, the obstetric consequences of false-positive and false-negative SGA cases are investigated.

2 MATERIAL AND METHODS

We included all 3113 women with singleton pregnancies from Aalborg University Hospital, who according to their nuchal translucency scan had a due date between 1 January 2015 and 31 December 2015. The staff members were all certified by the Fetal Medicine Foundation.11 A total of 185 women were excluded because of either abortion/miscarriage before 22 weeks of gestation or delivery outside the North Denmark Region. Consequently, a total of 2928 women remained in the study. EFW_us (g) was calculated using the formula by Hadlock et al (based on head circumference, abdominal circumference and femur length)12 and EFW_us deviation (%) was calculated using the reference curve by Maršál et al.9

Information regarding maternal characteristics, pregnancy and delivery were obtained from electronic patient records (Clinical Suite™ version 18.0.4.0; DXC Technology, Tysons, VA, USA) and the local Fetal Medicine database (Astraia software gmbh version 1.24.10).

2.1 Statistical analyses

The performance of the national screening program for SGA was described by calculation of the sensitivity and the FPR using the binary cut-off value for expected SGA during pregnancy as EFW_us ≤−15% and binary outcome for SGA at birth as BW ≤−22%. Logistic regression was performed to compare the odds ratio (OR) of obstetric and neonatal outcomes between expected and unexpected groups of SGA and appropriate-for-gestational-age (AGA) neonates. ORs were adjusted for GA at birth, BW deviation, maternal body mass index and parity (OR_adj).

The statistical software package Stata MP version 15.0 (StataCorp LP, College Station, TX, USA) was used for data analysis. P values <0.05 were considered statistically significant.

2.2 Ethical approval

The study was approved on 18 August 2016 and additional protocol was approved on 3 August 2018 by the Danish Patient Safety Authority, journal number 3-3013-1673/1. Data storage was approved by a regional notification to the Danish Data Protection Agency, journal number 2008-58-0028 with local reference-ID 2016-61 (31 March 2016) and 2018-104 (12 June 2018).

3 RESULTS

Within this cohort of 2928 unselected singleton pregnancies, 3.3% had SGA when defined as BW ≤ −22% (Figure 1) and 63% had a selective EFW_us. Concerning the entire cohort, the sensitivity was 62% given an FPR of 5.6% for SGA defined as BW ≤ −22%. Performance for SGA defined as BW ≤ −15% is added for comparison (Table 1). For those with mild SGA (BW −22% to −27%), the sensitivity of the screening program was 55%; for those with extreme SGA (BW ≤ 33%) it was 73% (see Supplementary material, Table S1); whereas it was only 40% (14/35) for newborns delivered after 40⁺⁰ weeks (Table 2).

Figure 1

Flowchart of the study population. AGA, appropriate-for-gestational-age; BW, birthweight; EFW_us, estimated fetal weight by ultrasound scan; SGA, small-for-gestational-age. SGA is defined by BW ≤ −22% (in accordance with Danish national guidelines) and for comparison by BW ≤ −15%

Table 1. Performance of the screening program for small-for-gestational-age in Denmark

Total population, n	2928
Women never referred to ultrasound	37% (1079/2928)
Women referred to ultrasound	63% (1849/2928)
Time between last ultrasound and birth (days), median (interquartile range)	11 (2, 28)
SGA cut-off	BW ≤ −22%	BW ≤ −15%
SGA at birth	3.3% (98/2928)	10.3% (303/2928)
Last EFWus ≤ −15%	7.5% (219/2928)	7.5% (219/2928)
Sensitivity (last EFWus ≤ −15% and SGA at birth)	62% (61/98)	41.6% (126/303)
False-positive rate	5.6% (158/2830)	3.5% (93/2625)

Note

• SGA is defined by BW ≤ –22% (in accordance with Danish national guidelines) and for comparison by BW ≤ –15%.
• Abbreviations: BW, birthweight; EFW_us, estimated fetal weight by ultrasound scan; SGA, small-for-gestational-age.

Table 2. Performance of the screening program for small-for-gestational-age in Denmark in relation to gestational age at birth

	Overall	Gestational age at birth
		<34 weeks	34^0-36^6 weeks	37^0-39^6 weeks	40^0-40^6 weeks	≥41 weeks
Total population, n	2928	46	130	1,146	845	761
SGA at birth (BW ≤ −22%)	3.3% (98/2928)	24% (11/46)	10% (13/130)	3.4% (39/1146)	2.2% (19/845)	2.1% (16/761)
Sensitivity of screening program	62% (61/98)	73% (8/11)	85% (11/13)	72% (28/39)	42% (8/19)	38% (6/16)
SGA referred to ultrasound	79% (77/98)	82% (9/11)	92% (12/13)	90% (35/39)	68% (13/19)	50% (8/16)
Last EFWus ≤ −15%	7.5% (219/2928)	26% (12/46)	15% (20/130)	8.6% (99/1146)	4.7% (40/845)	6.3% (48/761)
False-positive rate	5.6% (158/2830)	11% (4/35)	7.7% (9/117)	6.4% (71/1107)	3.9% (32/826)	5.6% (42/745)

Note

• Sensitivity for SGA defined by BW ≤ −22% using the following cut-off; EFW_us ≤ −15% at last ultrasound scan.
• Abbreviations: BW, birthweight; EFW_us, estimated fetal weight by ultrasound scan; SGA, small-for-gestational-age.

For the calculation of these sensitivities, we defined “screen positive” by the last EFW_us ≤ −15%.10 Using an EFW_us ≤ −12% would give a sensitivity of 86% at an FPR of 17%; using EFW_us ≤ −22%, would give a sensitivity 57% at a FPR of 1.6% (see Supplementary material, Table S2).

The maternal and neonatal characteristics for the SGA and AGA pregnancies are presented in the Supplementary material, Table S3 (SGA) and Table S4 (AGA).

Among the SGA fetuses, we could not demonstrate different perinatal outcomes among those identified by EFW_us and those not identified by EFW_us (Table 3) even though the identified SGA were more likely to have induction of labor (OR_adj = 0.13, 95% CI 0.04-0.41) and elective cesarean section (27% vs 0%, P < 0.01). Among the AGA fetuses, 5.6% were falsely expected to be SGA and these were more likely to have induction of labor (OR_adj = 2.51, 95% CI 1.70-3.71) and cesarean delivery (OR_adj = 1.44, 95% CI 0.96-2.18) (Table 4).

Table 3. Outcome for small-for-gestational-age pregnancies

Outcome	SGA			OR (95% CI), P-value	Adjusteda OR (95% CI), P-value
	Total n = 98	Expected SGA (Last EFWus ≤ −15%) n = 61	Expected AGA (Last EFWus > −15% or no EFWus) n = 37
Cesarean delivery	36% (35/98)	47% (26/61)	24% (9/37)	0.43 (0.17-1.07), P = 0.07	0.71 (0.24-2.13), P = 0.54
Elective cesarean section among all cesarean sections	20% (7/35)	27% (7/26)	0	**	**
Intended vaginal delivery	77% (75/98)	72% (44/61)	84% (31/37)	2.31 (0.93-5.72), P = 0.07	1.41 (0.47-4.22), P = 0.54
Induction among intended vaginal deliveries	57% (44/75)	83% (34/44)	35% (10/31)	0.14 (0.05-0.39), P = 0.00*	0.13 (0.04-0.41), P = 0.00*
Vacuum among vaginal deliveries	13% (8/63)	17% (6/35)	7.1% (2/28)	0.37 (0.07-2.01), P = 0.25	0.41 (0.07-2.30), P = 0.31
Umbilical artery pH <7.1	7.0% (6/86)	7.1% (4/56)	6.7% (2/30)	0.93 (0.16-5.39), P = 0.93	0.54 (0.08-3.58), P = 0.52
Apgar score <7 after 5 min	4.2% (4/95)	5.0% (3/60)	2.9% (1/35)	0.56 (0.06-5.59), P = 0.62	0.50 (0.04-5.78), P = 0.58
Stillborn	2.0% (2/98)	1.6% (1/61)	2.7% (1/37)	1.67 (0.10-27.47), P = 0.72	0.66 (0.02-27.39), P = 0.83
Neonatal death	1.0% (1/98)	1.6% (1/61)	0	**	**
Adverse outcomeb	11% (11/98)	13% (8/61)	8.1% (3/37)	0.58 (0.14-2.36), P = 0.45	0.53 (0.12-2.37), P = 0.41

Note

• SGA = BW ≤ −22%. Expected SGA = EFW_us ≤ −15% at last ultrasound scan. Expected AGA = normal symphysis-fundal height measurements and/or EFW_us > −15% at last ultrasound scan. Logistic regressions are used to compare the groups of SGA (expected SGA and expected AGA) using SGA-expected SGA as a reference.
• Abbreviations: AGA, appropriate-for-gestational-age; CI, confidence interval; EFW_us, estimated fetal weight by ultrasound scan; OR, odds ratio; SGA, small-for-gestational-age.
• ^a Adjusted for gestational age at birth (weeks in total), birthweight deviation (%), maternal body mass index and parity.
• ^b Umbilical artery pH <7.1, Apgar score <7 after 5 min, stillborn and neonatal death in one variable.
• * P<0.05.
• ** Logistic regression not possible, because no cases within expected AGA group.

Table 4. Outcome for appropriate-for-gestational-age pregnancies

Outcome	AGA			OR (95% CI), P-value	Adjusteda OR (95% CI), P-value
	Total n = 2830	Expected AGA (Last EFWus > −15% or no EFWus) n = 2672	Expected SGA (Last EFWus ≤ −15%) n = 158
Cesarean delivery	20% (572/2830)	20% (535/2672)	23% (37/158)	1.22 (0.84-1.79), P = 0.30	1.44 (0.96-2.18), P = 0.08
Elective cesarean section among all cesarean sections	38% (215/572)	38% (203/535)	32% (12/37)	0.79 (0.39-1.60), P = 0.50	1.49 (0.68-3.26), P = 0.32
Intended vaginal delivery	83% (2349/2830)	83% (2221/2672)	81% (128/158)	0.82 (0.56-1.20), P = 0.30	0.69 (0.46-1.05), P = 0.08
Induction among intended vaginal deliveries	29% (688/2349)	28% (631/2221)	45% (57/128)	2.02 (1.41-2.90), P = 0.00*	2.51 (1.70-3.71), P = 0.00*
Vacuum among vaginal deliveries	8.3% (187/2258)	8.4% (179/2137)	6.6% (8/121)	0.77 (0.37-1.61), P = 0.49	0.66 (0.31-1.44), P = 0.30
Umbilical artery pH <7.1	4.6% (118/2590)	4.6% (113/2443)	3.4% (5/147)	0.73 (0.29-1.81), P = 0.49	0.73 (0.28-1.87), P = 0.51
Apgar score <7 after 5 min	0.8% (22/2812)	0.8% (21/2654)	0.6% (1/158)	0.80 (0.11-5.98), P = 0.83	0.65 (0.08-5.22), P = 0.68
Stillborn	0.3% (9/2830)	0.3% (9/2672)	0	**	**
Neonatal death	0.1% (2/2830)	0.8% (2/2672)	0	**	**
Adverse outcomeb	5.1% (144/2830)	5.2% (138/2672)	3.8% (6/158)	0.72 (0.31-1.67), P = 0.45	0.63 (0.27-1.50), P = 0.30

Note

• AGA = BW > −22%. Expected SGA = EFW_us ≤ −15% at last ultrasound scan. Expected AGA = normal symphysis-fundal height measurements and/or EFW_us > −15% at last ultrasound scan. Logistic regressions are used to compare the groups of AGA (expected AGA and expected SGA) using AGA-expected AGA as a reference
• Abbreviations: AGA, appropriate-for-gestational-age; CI, confidence interval; EFW_us, estimated fetal weight by ultrasound scan; OR, odds ratio; SGA, small-for-gestational-age.
• ^a Adjusted for gestational age at birth (weeks in total), birthweight deviation (%), maternal body mass index and parity.
• ^b Umbilical artery pH <7.1, Apgar score <7 after 5 min, stillborn and neonatal death in one variable.
• * P<0.05.
• ** Logistic regression not possible, because no cases within expected SGA group.

4 DISCUSSION

In this study, we investigated the performance of the screening program for SGA defined as BW ≤ −22% in the North Denmark Region based on selective EFW_us. No less than 63% had an EFW_us giving a sensitivity of 62%; however this was much higher for fetuses with extreme SGA. We could not demonstrate improved perinatal outcome among SGA fetuses identified by EFW_us when compared with those not identified by EFW_us. The FPR was 5.6% and false-positive SGA cases were at an increased risk of obstetric interventions.

It is a strength of this study that the cohort can be classified as unselected, because we included >95% of the pregnant population in a well-defined geographic area13 with a lost to follow-up rate of only 5.9%. Furthermore, the validity of the data was very high because it is based on the unique Danish personal identification number. It is a limitation that the study is not powered to assess rare neonatal outcomes. In addition, referral for EFW_us followed the national guidelines. Unfortunately, the specific indication for referral is not consistently available in the patient record, and therefore the association between SGA and specific indications cannot be evaluated in this study.

In this study, the sensitivity of SGA screening using selective EFW_us on clinical indication was 62%, which is markedly higher than previous studies on selective ultrasound screening reporting a sensitivity of 29%-46%.7, 8, 14 This could be explained by a larger proportion of women referred to EFW_us in our study (63%) compared with previous studies (3.7%-42%).7, 8, 14 The large proportion of women referred to EFW_us may partly be explained by the inclusion of multiparous women in our study (54% of the total cohort). Among multiparous women, indications for EFW_us include previous obstetric complications such as SGA or preeclampsia, which lead to a higher number of referrals.10, 15 Moreover, the referral pattern in Aalborg and Denmark might be different from that in other countries.10 Even with such a large proportion of all women referred for ultrasound, 21% of SGA pregnancies were not referred for EFW_us. Moreover, a large proportion of AGA pregnancies (63%) underwent EFW_us.

We defined a screen-positive case in accordance with the national guidelines on SGA screening as last EFW_us ≤ −15%, that is, by a relatively slight estimated weight deviation. Therefore, it is disappointing that the sensitivity was only 73% (8/11) for extreme SGA (BW ≤ −33%); 2 cases did not have an EFW_us due to a false-negative “clinical screening” based on risk factors and symphysis-fundal height measurement; one case had an EFW_us > −15%, which was performed 29 days before birth (see Supplementary material, Table S1). Previous publications have not addressed this extreme SGA sensitivity even though these cases are most in need of prenatal detection and must be the primary target when we consider potential improvements to our screening program, as discussed below. It is more acceptable that the sensitivity for mild SGA (BW between −22% and −27%) was only 55% (29/53) even though they might also benefit from prenatal detection, especially when born post term.16 In fact, the EFW_us standard deviation of 8% when using The Hadlock Formula12 implies that a significant fraction of mild SGA fetuses will remain undetected (EFW_us > −15%) even when identified by the clinical screening (eg symphysis-fundal height) with correct referral for EFW_us.

The sensitivity decreased markedly with increasing GA; from 72% (GA 37⁺⁰-39⁺⁶) to 38% (GA ≥ 41⁺⁰) leaving 22% (21/98) undiagnosed at birth after term (Table 2). Among these, 67% (14/21) did not have an EFW_us, whereas 33% (7/21) had an EFW_us > −15%. This is highly problematical, because it is generally accepted that SGA babies need to be delivered at least at term.16

We confirmed the results from a Swedish4 study showing increased risk of interventions among SGA cases identified correctly before birth (Table 3). However, our study did not have statistical power to address their finding of improved neonatal outcome. The FPR is also of interest, that is, AGA cases falsely expected to be SGA (Table 4), showed an increased rate of labor induction (OR_adj 2.51, 95% CI 1.70-3.71) and an increased cesarean section rate (OR_adj 1.44, 95% CI 0.96-2.18) confirming results from one previous study.5

In order to improve the screening for SGA in Denmark, several issues could be considered: selection of pregnancies for EFW_us, accuracy of EFW_us, and 3rd trimester routine EFW_us (“universal ultrasound screening”). Improved selection of pregnancies for EFW_us might be achieved by the use of 1st trimester maternal serum markers17, 18 and uterine artery Doppler flow,19, 20 and by improved symphysis-fundal height measurements using a single person throughout the pregnancy.21, 22 An obvious possibility would be to change the “risk of SGA” definition to EFW_us ≤ −12% on the expense of a doubled FPR. Furthermore 3-dimensional ultrasound23 and magnetic resonance imaging24, 25 for better estimates of EFW could be considered. Introduction of routine EFW_us (“universal ultrasound screening”) has been shown to increase the sensitivity from 29%-33% to 42%-80%, but at the expense of increased FPR from 0.26%-3% to 5%-13% in previous studies.7, 14, 26, 27 Routine EFW_us performs best when applied close to delivery26, 27; that is, a sensitivity of 89% if delivery is within 2 weeks from routine EFW_us in GA 35-37 weeks, at an FPR of 5%.27 As suggested by our data, the main limitation of the Danish SGA screening program was in the antenatal detection of post-term SGA infants. Therefore, introducing a late routine EFW_us either at term or post-term (GA 41⁺⁰) would likely increase the sensitivity but also the FPR.

This manuscript focuses on screening for SGA, as small fetal size is regarded as a proxy of placental dysfunction. However, fetal size is not a perfect marker of placental function, and even a perfect screening for SGA may not identify all fetuses at risk because of placental dysfunction.28 New markers of placental dysfunction based on maternal serum18 or placental MRI29 may be able to identify placental dysfunction directly, and the clinical potential of these methods is currently being investigated.

5 CONCLUSION

The performance of the Danish national screening program for SGA based on selective EFW_us on clinical indication has improved considerably over the last 20 years with an increased sensitivity from 29% (1998) to 62% (2015) and FPR from 0.26% (1998) to 5.6% (2015).8 However, the selection of pregnancies for ultrasound is a limitation of the program as a large proportion of AGA pregnancies are referred to ultrasound and a large proportion of SGA pregnancies are not. In addition, the detection of SGA babies post-term remains rather low when compared with earlier gestation. This paper gives a detailed insight into the current screening program, and provides ideas for further improvement of SGA screening.

ACKNOWLEDGMENTS

We wish to thank Hans-Victor Nielsen, academic secretary, for fruitful help with data collection.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.