A population-based study of the prenatal diagnosis of congenital malformation over 16 years

Authors


S. Richmond, Regional Maternity Survey Office, 25 Claremont Place, Newcastle upon Tyne NE2 4AA, UK.

Abstract

Objective  A population-based study of the trends in accuracy over time of prenatal diagnosis of congenital malformations.

Design  A prospective study of all congenital malformations.

Setting  The counties of Cleveland, Durham, North Cumbria, Northumberland and Tyne and Wear.

Population  All 573,471 babies born to residents at 24 weeks of gestation or more, and all terminations for fetal abnormality, occurring between 1985 and 2000.

Methods  An analysis of all congenital anomalies whether antenatally detected or diagnosed before one year of age to examine the accuracy of prenatal diagnosis and its trends over four consecutive four-year periods.

Main outcome measures  The proportion of cases where a diagnosis before birth was correctly made, wrongly made or not made in a range of important structural malformations.

Results  There was a substantial improvement in the accuracy of prenatal diagnosis of all conditions reported here. However, the extent to which different conditions were diagnosed varied widely. It now exceeds 90% for anencephaly and for abdominal wall defects but is still less than 70% for diaphragmatic hernia, bladder outlet obstruction and many major skeletal defects. Termination of pregnancy for fetal malformation rose from 23 to 47 per 10,000 registrable births.

Conclusion  This large prospective study provides data on the accuracy of prenatal diagnosis of a number of malformations in a geographically defined population under service conditions. Diagnostic accuracy has improved over time but may well be lower than many clinicians assume for some important conditions and may not match public expectations.

INTRODUCTION

There has been rapid development in prenatal diagnosis over the last 20 years. Much has been due to the increased availability and sophistication of ultrasound used both as a diagnostic tool and as part of a screening process. However, there have also been many other developments in this period both in screening and diagnostic tests. A survey sponsored by the Royal College of Obstetricians and Gynaecologists and the Royal College of Radiologists showed that by 1995 over 80% of UK maternity units offered an anomaly scan to all expectant mothers.1

This paper evaluates the final accuracy of prenatal diagnosis for residents of the former Northern Regional Health Authority (excluding South Cumbria) and compares this over each of four consecutive four-year periods ending in 2000. It is a study of the final effectiveness of prenatal screening and diagnosis, not a measurement of its individual parts.

It also demonstrates notable changes in prevalence for some important conditions.

METHODS

The Northern Congenital Abnormality Survey (NorCAS), previously known as the Fetal Abnormality Survey, is a voluntary collaborative survey collecting data prospectively from a population of about three million living in the former Northern health region.2 This comprises the counties of Cleveland, Durham, North Cumbria, Northumberland and Tyne and Wear. This geographical area is served by 15 consultant units and four general practitioner units and is bounded by the sea to east and west. There is a sparsely populated national border to the north, and an area of low population to the south, both of which have little cross-boundary flow for maternity care. All paediatric surgery and paediatric cardiology services for the region are provided in Newcastle and all genetics services are centralised to three centres within the region.

South Cumbria is in the region but, for geographical reasons, mothers resident there are frequently transferred outside the region when complications arise. Births to South Cumbria residents have therefore been excluded. Similarly some mothers resident outside the region deliver in hospitals in the region but these pregnancies have also been excluded.

Data are collected on malformations in any baby born to a resident, wherever birth occurred and regardless of whether the abnormality was diagnosed in utero, at birth or during the first year of life.2,3 Notification to the survey during pregnancy is encouraged as soon as detection of a possible abnormality leads to further investigation. All units have offered an anomaly scan to all pregnant mothers since the mid-1980s and more than 99% of women book early enough to avail themselves of this opportunity. Antenatally notified cases of possible malformation remain on the register even if the baby is later found to be normal after birth. All terminations of pregnancy for fetal malformation are included. Minor abnormalities such as skin tags, birthmarks, failed testicular descent, glandular hypospadias, isolated talipes and minor syndactyly are excluded. All conditions requiring surgery in the first year of life (except inguinal hernia, patent ductus and developmental dysplasia of the hip) are included. Children with congenital hypothyroidism and familial conditions such as phenylketonuria, haemophilia, Duchenne dystrophy and cystic fibrosis are excluded unless the pregnancy is terminated on these grounds. The reasoning behind this apparent contradiction is that we wished to take note of all terminations for fetal reasons but recognised that ascertainment of liveborn babies with some of these conditions is likely to be patchy.

Before the survey was set up a letter was sent to all consultant obstetricians, paediatricians, paediatric pathologists, paediatric surgeons, paediatric cardiologists and clinical geneticists followed by a personal visit by one of the original survey steering group. Support for this survey, and for its sister survey of fetal and infant mortality from 20 weeks of pregnancy to one year of age, remains enthusiastic. The survey uses multiple sources of ascertainment including obstetric ultrasonographers, obstetricians, neonatologists, clinical geneticists, cytogeneticists, perinatal pathologists, paediatric cardiologists, paediatric surgeons and general paediatricians. Details of any antenatal or postnatal investigations are reported to the survey office. Postdelivery diagnoses are cross checked with the appropriate subspecialty. Reminders are sent from the survey office to those reporting antenatal abnormalities if no postnatal report is received three months after the expected date of delivery.

The original establishment of the survey was approved by the ethics committee of all districts involved. Data are held by NorCAS in compliance with the Data Protection Act 1998 and the survey has appropriate exemption under section 60 of the Health and Social Care Act 2001. Access to the database is controlled by the steering group. A recent comparison of data from four local surveys, including NorCAS, with national data found that ascertainment by local surveys was considerably more complete and that NorCAS had the most complete ascertainment of postnatally diagnosed anomalies currently available.4

The survey has been validated from the start by cross checking the information available from District Health Authority SD 56 returns to the Office of Population Censuses and Surveys [OPCS—now the Office of National Statistics (ONS)]. This showed an almost complete match in respect of the number of registered births with a neural tube defect, isolated congenital hydrocephalus, exomphalos, anal atresia and cleft lip and palate. For all other conditions notifications to the survey consistently outnumbered those submitted to OPCS.

Cross checks with the region's three cytogenetic laboratories confirmed that every pregnancy terminated for a confirmed chromosomal anomaly had been reported to the survey by the clinicians involved. Cross checks against the mandatory termination of pregnancy returns to OPCS seemed to show a shortfall in terminations for ‘suspected chromosome abnormality’, but a confidential analysis of these returns for 1985–1987 showed the discrepancy was accounted for by terminations before 12 weeks of pregnancy. These were usually found to be terminations for an increased age-related risk, undertaken without further diagnostic procedures. Since then the survey has continued to validate completeness of reporting in detail, and ascertainment has continued to exceed the national system. The main numbers and groupings are set out in Table 1a.

Table 1a.  A comparison of numbers in the time periods reported.
Years1985–19881989–19921993–19961997–2000
  • Spontaneous losses before 24 completed weeks are excluded. All terminations for fetal abnormality are included regardless of gestation.

  • Figures in parentheses () are prevalence per 10,000 births at 24 weeks of gestation or over, those in [ ] brackets per 100 antenatal reports, and those in { } brackets per 100 babies with a postnatally confirmed abnormality.

  • *

    DRP = dilated renal pelvis/hydronephrosis.

  • VSD = ventricular septal defect.

  • ¥

    Before October 1992, babies born dead after 28 weeks of gestation were not registered in the UK. After that date all babies born dead from 24 weeks of gestation were registered.

A. Total registered births¥154,459153,339140,180125,386
B. Registrable births and all births of 24 weeks of gestation or more154,502153,403140,180125,386
C. Antenatally reported cases1047 (68)1889 (123)1874 (134)1734 (138)
D. Antenatally reported cases normal at birth183 [17%]631 [33%]498 [27%]201 [12%]
E. Antenatally reported cases of dilated renal pelvis (DRP*)200 [19%]735 [39%]580 [31%]287 [17%]
F. Antenatally reported cases of DRP* normal at birth [% of DRP*]76 [38%]434 [59%]320 [55%]91 [31%]
G. Antenatally reported cases excluding all DRP*847 (54)1154 (75)1294 (92)1447 (112)
H. Antenatally reported cases normal at birth (DRP excluded)107 [13%]197 [17%]178 [14%]110 [8%]
I. Babies with malformations found or confirmed after birth or after termination for fetal abnormality2768 (179)3329 (217)3343 (238)3208 (256)
J. Babies with malformations where pregnancy was not screened32 {1.2}22 {0.66}16 {0.47}16 {0.49}
K. Babies with isolated VSDs220 (14)330 (22)457 (33)477 (38)
L. Babies with malformations thought to be normal before birth1876 {68%}2052 {62%}1952 {58%}1676 {52%}
M. Babies with malformations found or confirmed after birth or after termination for fetal abnormality—isolated VSDs excluded2548 (165)2999 (195)2889 (206)2731 (218)
N. Babies with malformations thought to be normal before birth—isolated VSDs excluded1657 {60%}1722 {52%}1498 {45%}1190 {37%}
O. Terminations for confirmed abnormality (all gestations)349 (22.6)490 (31.9)554 (39.5)595 (47.4)

Every case in the database was examined. Minor differences between what had been suspected antenatally and what was found after delivery were discounted. Thus, for example, confusion between anencephaly and iniencephaly was not counted as an error. However, where confusion could well affect management, as when an exomphalos was thought to be a gastroschisis, or what was thought to be a diaphragmatic hernia turned out to be a cystadenomatoid malformation of the lung, the antenatal diagnosis was classed as erroneous. Babies were considered normal if they reached their first birthday without a congenital abnormality being reported.

In cases of possible false-negative or false-positive diagnosis the authors discussed each case and came to an agreement based on these principles and the following rules:

  • Terminations for malformation:

  • Rule 1. If the indication for termination was confirmed, then failure to report or search for another anomaly after termination had been agreed was not considered an error on the basis that, once the condition defining the indication for termination had been identified, further more detailed examination for further anomalies could be seen as an academic exercise unnecessary for management and prolonging the distress of the patient. For example, in trisomy 21 if the trisomy was confirmed after termination then the additional postnatal finding of an atrio-ventricular septal defect (AVSD) was not counted as a failure to diagnose the cardiac defect prenatally.

  • Rule 2. If an additional prenatally diagnosed anomaly was not mentioned postnatally this was also not necessarily considered erroneous, unless it failed to be reported following a full postmortem examination. For example, in prenatal diagnosis of trisomy 18 with exomphalos, if the trisomy was confirmed postnatally then the fact that, in the absence of a formal postmortem examination, an exomphalos was not mentioned postnatally was not used to suggest that the prenatal diagnosis of exomphalos was incorrect.

  • In live births or in utero deaths at or after 24 weeks of gestation:

  • Rule 3. If a second significant anomaly was not diagnosed prenatally, then this was considered erroneous on the grounds that parents might have requested a termination had they known the full extent of the anomaly. For example, if a prenatal diagnosis of trisomy 21 was made and the baby, when liveborn, was found also to have an AVSD this was considered a failure to diagnose the cardiac defect prenatally.

  • Rule 4. If the baby was already dead in utero at the time of scanning, then noting ‘multiple anomalies’ without further differentiation was allowed, provided significant anomalies were confirmed postnatally. The reasoning behind this is that it would seem unnecessary and improper to prolong the examination of an abnormal fetus in utero once it was clear that the baby was dead.

When an outcome was unclear, the original notification and any subsequent correspondence in the survey files was examined. In a few cases a final postnatal diagnosis was not available despite requests to the unit concerned. Our system of multiple sources of ascertainment can be relied upon to ensure that any cardiac or chromosomal anomaly, any anomaly requiring surgery or resulting in death would have been reported. In the absence of notification from any of these sources, it was assumed that the baby was normal. Overall there were few such cases.

Where a baby had more than one condition (for example, trisomy 18 with spina bifida) it was considered separately under each of the diagnoses reported. However, each baby appears only once in rows P–W of Table 1b. The UK government definition of a stillbirth changed in 1992 to include all babies born dead from 24 rather than 28 weeks of gestation. For consistency we have classified every baby born dead after a pregnancy lasting 24 weeks or more as a stillbirth even if the child was born before October 1992.

Table 1b.  A comparison of the number of abnormalities in the time periods reported. Grouped conditions.
Years1985–19881989–19921993–19961997–2000
  1. Number of fetuses with confirmed abnormalities reaching termination or 24 weeks of gestation. Hierarchical classification (P through W) ensures that each fetus is only counted once.

  2. Figures in parentheses ( ) are birth prevalence per 10,000 births at 24 weeks of gestation or more.

P. Chromosomal abnormalities402 (26.0)498 (32.5)563 (40.2)581 (46.3)
Q. Neural tube defects and other CNS malformations419 (27.1)397 (25.8)328 (23.4)314 (25.0)
R. Abdominal wall defects including cloacal extrophy and fetal disruption80 (5.2)84 (5.7)89 (6.3)113 (9.0)
S. Renal tract anomalies233 (15.1)401 (26.1)414 (29.5)351 (27.9)
T. Gastrointestinal and intrathoracic defects181 (11.7)216 (14.1)189 (13.5)185 (14.7)
U. Cardiac malformations871 (56.4)959 (62.5)1057 (75.4)918 (73.2)
V. Cleft lip, palate or both118 (7.6)158 (10.3)138 (9.8)148 (11.8)
W. Other445 (28.8)585 (38.1)526 (37.5)531 (42.3)

RESULTS

Total registrable births during the period were 573,471 (Table 1a, row B). The birth rate has fallen by 21% from an average of 39,000 per year in the first four years to 31,000 per year in the last four years. Most of this fall has occurred since 1992. Despite this reporting rates rose throughout.

The prevalence of confirmed malformation appears to have increased more than a third over the 16-year study period. We suggest two explanations for this. The increase between the first and second four years probably represents better ascertainment as the survey became established. However, much of the subsequent increase is probably due to the greater use made of cardiac ultrasound in the investigation of children with asymptomatic cardiac murmurs. This has resulted in a significant increase in identification of minor cardiac malformations, the main single contribution being children whose only malformation was a ventricular septal defect (VSD).5 Isolated VSDs (Table 1a, row K) rose from 14 to 38 per 10,000 over the study period. If these cases are removed, the prevalence of malformations varies little between the last three cohorts (Table 1a, row M).

In calculating specificity we have had to assume that all pregnancies resulting in normal babies born after 24 weeks of gestation were screened. However, we do know how many abnormal babies were not screened and what proportion of the abnormal babies these represent (Table 1a, row J). In the absence of any data, we feel it is reasonable to assume that the same proportion of normal babies were not screened. Though we have not included unscreened cases in our calculations of sensitivity or specificity, we have not reduced the denominator for specificity to allow for the presumed 0.5% or so of normal pregnancies which were probably not screened.

We considered as a false-negative report any baby born with a malformation whose anomaly scan or other antenatal investigations had been interpreted as normal. Analysis of the raw figures would suggest that, despite a continuing improvement, almost half the babies with malformation were not suspected to be abnormal before birth during the final four-year period (Table 1a, row L). However, removing cases of isolated VSD reveals a marked improvement (Table 1a, row N). The overall figure includes some malformations that are relatively minor, such as partial or complete digital reduction deformities as well as conditions not easily diagnosable before birth such as cleft palate with intact lip. The prenatal diagnosis rates are shown in Table 2a.

Table 2a.  Prenatal diagnosis rates.
Registrable birthsNo. (birth prevalence per 10,000) and proportion prenatally diagnosed, %
1985–19881989–19921993–19961997–2000
154,502153,403140,180125,386
  1. All cases where pregnancy was terminated for fetal abnormality or which reached 24 weeks are included.

  2. Spontaneous losses before 24 weeks of gestation are excluded.

  3. Figures in parentheses ( ) are prevalence per 10,000 births at 24 weeks of gestation or over.

Malformation ICD-9 code
Anencephaly 7400, 74002 (7401)123 (7.9), 93%124 (8.1), 99%91 (6.5), 100%80 (6.4), 99%
Spina bifida ± hydrocephalus 7410, 7419 (7402)186 (12.0), 56%125 (8.1), 66%108 (7.7), 85%80 (6.4), 84%
Isolated hydrocephalus or aqueduct stenosis 74230, 7423945 (2.9), 75%44 (2.9), 79%54 (3.8), 77%63 (5.1), 89%
Encephalocele 742022 (1.4), 58%17 (1.1), 78%22 (1.6), 77%27 (2.1), 90%
Holoprosencephaly 74226, 7598013 (0.84), 31%19 (1.2), 42%19 (1.4), 75%24 (1.9), 75%
Dandy–Walker 742316 (0.39), 17%7 (0.45), 43%12 (0.86), 58%12 (0.96), 77%
Hydranencephaly 742321 (0.06), 0%7 (0.39), 13%2 (0.14), 50%2 (0.16), 50%
Exomphalos 7567048 (3.1), 66%35 (2.3), 68%41 (2.9), 87%43 (3.4), 96%
Gastroschisis 7567145 (2.9), 42%39 (2.5), 59%56 (3.9), 84%59 (4.7), 94%
Bilateral renal agenesis 7530026 (1.7), 52%14 (0.9), 87%17 (1.2), 88%22 (1.8), 93%
Diaphragmatic hernia (and eventration) 75661, 7566256 (3.6), 13%60 (3.9), 49%54 (3.8), 59%63 (5.0), 63%
Bladder outlet obstruction 75360, 75361, 7536226 (1.9), 42%20 (1.3), 42%30 (2.1), 56%30 (2.4), 69%
Hypoplastic left heart 746726 (1.7), 12%30 (2.3), 30%30 (2.1), 47%33 (2.6), 88%
AVSD 745640 (2.6), 12%59 (3.8), 15%63 (4.5), 35%42 (3.3), 38%
Osteogenesis imperfecta 7656013 (0.8), 23%14 (0.9), 14%13 (0.9), 8%10 (0.8), 60%
Trisomy 13 758115 (0.8), 50%22 (1.4), 45%24 (1.7), 79%16 (1.3), 83%
Trisomy 18 758227 (1.7), 44%46 (2.9), 40%45 (3.2), 80%57 (4.5), 69%
Trisomy 21 7580206 (13.3), 16%220 (14.3), 31%246 (17.5), 47%233 (18.5), 45%
Trisomy 21—mothers 35 years and over53%66%74%65%
Trisomy 21—mothers under 35 years4%17%33%31%

We considered as a false-positive report any baby notified antenatally as possibly malformed who was later born without any malformation. During the second and third four-year periods a number of units enthusiastically reported cases of dilated renal pelvis using varying criteria for deciding that dilatation was present. A standard measurement for reporting was agreed in 1994. Within the data reported for individual conditions we have also noted as false-positives those cases where the baby did indeed have a malformation but a malformation that was significantly different from that predicted (Table 2b).

Table 2b.  False-positive rates.
Registrable birthsOdds of having no significant abnormality given a positive antenatal reportOdds of having a different anomaly given a positive antenatal report
1985–19881989–19921993–19961997–20001985–19881989–19921993–19961997–2000
154,502153,403140,180125,386154,502153,403140,180125,386
  1. All cases where pregnancy was terminated for fetal abnormality or which reached 24 weeks are included. Spontaneous losses before 24 weeks of gestation are excluded.

Malformation ICD-9 code
Anencephaly 7400, 74002, (7401)1:1151:∞1:∞1:∞1:1151:421:991:∞
Spina bifida ± hydrocephalus 7410, 7419 (7402)1:∞1:301:491:761:1021:161:521: 20
Isolated hydrocephalus or aqueduct stenosis 74230, 742391:181:301:31:31:31:31:61:6
Encephalocele 74201:191:171:∞1:∞1:31:221:91:9
Holoprosencephaly 74226, 759801:∞1:∞1:∞1:∞1:∞1:91:∞1:∞
Dandy–Walker 742311:∞1:41:∞1:121:∞1:41:71:7
Hydranencephaly 742321:∞1:∞1:∞1:∞1:∞1:11:∞1:∞
Exomphalos 756701:181:81:∞1:∞1:61:41:201:20
Gastroschisis 756711:∞1:∞1:471:∞1:81:71:∞1:∞
Bilateral renal agenesis 753001:41:41:191:131:71:∞1:211:∞
Diaphragmatic hernia (and eventration) 75661, 756621:61:161:∞1:∞1:31:61:111:19
Bladder outlet obstruction 75360, 75361, 753621:31:41:∞1:111:51:31:71:8
Hypoplastic left heart 74671:∞1:∞1:∞1:341:41:31:41:7
AVSD 74561:∞1:∞1:∞1:91:∞1:41:201:7
Osteogenesis imperfecta 765601:∞1:∞1:∞1:∞1:∞1:21:21:∞
Trisomy 13 75811:∞1:71:∞1:∞1:∞1:∞1:∞1:∞
Trisomy 18 75821:∞1:∞1:∞1:∞1:∞1:∞1:161:∞
Trisomy 21 75801:∞1:331:1051:∞1:∞1:651:∞1:∞

In almost every case of termination of pregnancy for fetal malformation sufficient information to clarify the postnatal diagnosis is available. There were 86 first trimester terminations. In 82 of these the antenatal diagnosis was of cystic hygroma and in 40 it was possible to confirm a diagnosis of Turner's syndrome. Of the 1902 later terminations, 1166 had a postmortem examination and 517 were chromosome diagnoses. In 187 the diagnosis depended on clinical description with or without photographs and in a further 13 cases the diagnosis was made on X-ray. There were only 19 cases without a clear indication of how the diagnosis was confirmed.The rate of termination for fetal malformation more than doubled over the study period (Table 1a, row O).

Table 2a compares the extent to which certain specific anomalies were diagnosed prenatally.Table 2b shows the odds of having a normal baby despite a particular antenatal diagnosis and the odds of having a baby with an abnormality but not the exact abnormality predicted.

These particular conditions were chosen for more detailed analysis because, with the exception of AVSDs, they are obvious at birth or within hours of birth. Thus, ascertainment is likely to be complete. AVSD was included because it is frequently an issue in the diagnosis of trisomy 21. The figures quoted include all terminations for fetal malformation at any gestation as well as all babies of 24 weeks of gestation or more with malformations. Malformations reported in spontaneous pregnancy losses before 24 weeks of gestation are not included. This is because the reporting of such losses is sporadic and the proportion of such losses professionally examined after birth is variable and would reflect the interest of individual units rather than being a true reflection of prevalence.

The following might help correct interpretation of the tables:

  • Anencephaly and spina bifida. The region relies mainly on ultrasound rather than serum screening to detect neural tube defects.6 The figures for anencephaly include craniorrhachischisis and those for spina bifida include iniencephaly but exclude anencephaly. The prevalence of spina bifida has fallen to a similar level to that of anencephaly. The prenatal diagnosis rate for anencephaly is now virtually 100%, and that for spina bifida has improved. In the most recent c cohort the possibility of a fetus being normal after an antenatal diagnosis of spina bifida is low at 1 in 76, but the possibility that this diagnosis may not be absolutely correct is somewhat higher at 1 in 20. The rise in false-positive diagnoses of spina bifida from a very low level to 1:30 in the second cohort which then falls steadily over the following two cohorts is striking. We suggest that this may be due to an increasing appreciation of the importance of cerebral abnormalities in the diagnosis of spina bifida, which improved the sensitivity of ultrasound diagnosis in this condition. This was much publicised from about 1987 to 1990.7 Care is needed in interpretation of this sign as slight malrotation of the transducer may include a view of the superior aspects of the orbits giving a false lemon sign.8 Errors of this type may have led to a significant number of referrals at that time, falling off as experience was gained.

  • Isolated hydrocephaly and aqueduct stenosis. These were considered together. Overall, the detection rate was similar to that for spina bifida, but the final false-positive rates were surprisingly high. Antenatally diagnosed isolated ventriculomegaly will have been included in this group. A region-wide study of ventriculomegaly was set up during the second cohort. This study is investigating the long term outcome of all babies with atrial measurement of 10 mm or greater.

  • Encephalocele. Though the numbers are small the detection rate has nearly doubled over the study period, reaching 90% in the latest four-year period.

  • Gastroschisis and exomphalos. Body stalk anomalies, cloacal extrophies and amniotic disruption sequence are excluded. Confusion between exomphalos and gastroschisis was considered a diagnostic error because the higher rate of associated malformations in exomphalos, particularly aneuploidies, requires different management.

    The prevalence of gastroschisis rose from 2.9 per 10,000 births in the first cohort to 4.7 in the fourth. When plotted annually the rise appears to start in 1993. Other studies have noted this increase and have also commented on an increasing south to north prevalence in these conditions but particularly in exomphalos.9

  • Bilateral renal agenesis. The prevalence in the four time periods was fairly constant, the proportion prenatally diagnosed almost doubling. False-positive and false-negative rates improved but were still quite high. These relatively poor rates are probably explained by the difficulty of achieving an ultrasound diagnosis in the face of severe oligohydramnios, which may result from premature rupture of membranes or alternative severe renal pathology rather than from renal agenesis.

  • Diaphragmatic hernia and eventration. The most common error was to confuse diaphragmatic hernia and pulmonary cystadenomatoid malformation. Despite the recognised difficulty in antenatal differentiation this was considered a significant error as the prognosis and management are different.

  • Bladder outlet obstruction. The odds of a significant but different malformation were 1:5 in the first period and 1:8 in the last.

  • Hypoplastic left heart and AVSD. Detection of some congenital cardiac malformations showed a marked improvement. For hypoplastic left heart, the improvement was more than sevenfold with a low false-positive rate. For AV septal defect the results were not so good, although there was a more than threefold improvement in detection.

  • Trisomy 21. Screening strategies across the region varied over this period and knowledge of this variation is important in interpreting data on the accuracy of antenatal diagnosis of trisomy 21. Considerable caution should be applied when comparing the results in this region to reports from populations with a more uniform screening policy. There was no significant use of first trimester screening for trisomy 21 in the study area during this period. Serum screening was first introduced into the region in 1991 and units responsible for about 50% of deliveries offered it to all pregnant women, regardless of age, from 1992. The remaining units offered karyotyping on an age-related basis. However, during the period 1997–2000, two of these units responsible for about 8% of deliveries began to offer serum screening to all mothers. Thus, by 2000, serum screening was available to nearly 60% of mothers in this region. In the study area it would appear that, where serum screening for trisomy 21 is offered regardless of age, it is taken up by about 50% of women. The prevalence of trisomy 21 has risen between the first and fourth cohorts almost certainly due to an increase in the average age of mothers. The proportion of mothers delivering aged 35 or over increased from 6% in the first cohort to 10.6% in the last, with a peak of just over 12.9% in 2000. In England and Wales, as a whole the proportion of mothers delivering aged 35 years or over in 2000 was 17.2%. Out of the 905 cases of Down's Syndrome 17 were allowed under Rule 1 (above). Fourteen of these cases were terminated for a cystic hygroma, one for prune belly, one for hydrocephalus and one for non-immune hydrops.

    In a few cases the diagnosis of trisomy 21 was suggested by ‘soft markers’ but further antenatal investigation was not carried out, probably because permission for invasive investigation was refused. This mechanism was responsible for all false-positive diagnoses.

  • Limb reduction defects (Table 3). These were originally coded merely as upper or lower limb defects. The majority of cases were digital anomalies. Those where the defects were limited to digits are not reported here. Accessing the original notifications enabled the authors to classify major anomalies into three groups.

    Amputations—where the limb ‘ends too soon’ but was relatively normal up to that point. Hemimelias—where the length of the limb was largely preserved but there was a missing long bone in the forearm or lower leg (such as absent radius or fibula) or a ray defect. Phocomelias—where part or all of a hand or foot was present but a substantial portion of a proximal long bone was missing. Proximal focal femoral deficiency was included in this group.

    This reclassification could not be applied reliably to the antenatal scan reports so we limited our analysis to prevalence and the proportion prenatally diagnosed. In all three categories, prevalence has changed little and there has been some improvement in prenatal diagnosis. However, most amputations and half the hemimelias are still not diagnosed before birth.

  • Inter-unit comparisons. The 15 consultant obstetric units contributing to the survey vary greatly in size. The smallest unit had only 650 births in 2000 whereas the largest had 4730. To compare the overall efficacy of the entire antenatal screening process, we looked at the success with which nine important conditions were identified before birth in the most recent four-year period in each unit. The conditions were chosen because they are important, they cover a range of different body systems and they are all (with the exception of AVSD) easily identifiable in the neonatal period. The conditions chosen were spina bifida aperta, holoprosencephaly, diaphragmatic hernia or eventration, gastroschisis, renal agenesis, bladder outlet obstruction, AVSD in the absence of trisomy 21, hypoplastic left heart and lethal short limbed dwarfism.

Table 3.  Birth prevalence of limb defects and proportion antenatally diagnosed.
 No. (birth prevalence per 10,000) and proportion prenatally diagnosed, %
  1. All cases where pregnancy was terminated for fetal abnormality or where the pregnancy reached at least 24 weeks are included.

  2. Cases resulting in spontaneous loss before 24 weeks of gestation are excluded.

  3. Cases with missing digits only are not included.

  4. Figures in parentheses ( ) are prevalence per 10,000 births at 24 weeks of gestation or over.

Years1985–19881989–19921993–19961997–2000
Registrable births154,502153,403140,180125,386
Amputation18 (1.2), 5%18 (1.2), 17%15 (1.1), 33%12 (0.96), 25%
Hemimelia21 (1.4), 5%17 (1.1), 29%14 (0.99), 29%10 (0.8), 50%
Phocomelia2 (0.13), 0%8 (0.5), 25%5 (0.36), 0%4 (0.32), 75%

The three most successful units identified 46/50 (92% [95% CI 81–98%]) of these defects, and the three least successful units identified 29/43 (67% [95% CI 52–79%]). The three largest (tertiary referral) units were neither among the top three units nor among the bottom three units when centres were ranked in this way. Given the rarity of the conditions used for this assessment, ranking is only going to be possible when the results are averaged over a number of years, and will still be subject to much chance variation even then. Such an approach does, however, help to establish that most units in the region are currently performing in a broadly comparable way.

DISCUSSION

Despite recent centralisation of the legal management of negligence cases against the National Health Service, data are not collected in such a way as to allow measurement of the frequency with which errors in the antenatal detection of abnormality form the basis of legal action. However, few would quarrel with the assertion that public expectation of the accuracy of this process is often unrealistically high, and this undue optimism frequently extends to their medical and midwifery advisers.

Strictly speaking, prenatal diagnosis is the result of two distinct processes. The first of these is a screening test such as the routine ultrasound examination, which might pick up an indication of abnormality (lemon sign, abnormal four-chamber view of heart or any of a selection of ‘soft-markers’), which then leads to the second which is a more focussed test which attempts to diagnose the problem. This diagnostic test may be, for example, a more detailed ultrasound examination of the heart or perhaps an amniocentesis with a view to obtaining a karyotype. The importance of the distinction is that one can only really expect high degrees of accuracy from the diagnostic tests. This paper is reporting the result of the entire episode without distinguishing between these two individual processes.

Many studies have attempted to assess the accuracy of prenatal diagnosis but most have been based on the experience of single units, either alone or in groups, rather than on a defined population. Such centres usually have a special interest in antenatal diagnosis and thus the service they provide is not likely to be typical of that available to the average pregnant mother. A recent review emphasised problems of small numbers and heterogeneity between studies.10 A Health Technology Assessment report on ultrasound screening in pregnancy published in 2000 recommended that ‘general detection rates should be assessed by linkage with high-ascertainment fetal abnormality registers at a regional level’.11 This report attempts to address this recommendation.

Two population-based studies have been published recently. A Scottish study of 264,481 pregnancies between 1989 and 1994 looked at trisomies 13, 18 and 21; and 12 major structural abnormalities discovered in the first year of life.12 A smaller German study reported on 20,248 pregnancies between 1990 and 1994.13 Neither reported false-positive results.

The problems associated with measuring false-positives and their potential hazards have been pointed out, as have the difficulties in diagnosing malformations antenatally when screening a low risk population.12,14 Parents may be seriously distressed by information they did not seek or do not wish to know before birth. They may also be distressed unnecessarily by an incorrect diagnosis of malformation in what is in fact an entirely normal pregnancy.

We wish to inform these debates by presenting data showing the accuracy of this process for some significant anomalies, over a large and finite population, demonstrating how accuracy has changed over time. We also wish to present a realistic population-based assessment of the accuracy or otherwise of prenatal diagnosis for each time cohort over the study period, so that units defending errors of prenatal diagnosis can have a yardstick of overall accuracy for that condition at that time.

In the case of central nervous system abnormalities, results have improved both in terms of prenatal diagnosis and also in reduction in false-positive diagnoses, apart from isolated hydrocephaly and hydranencephaly. There has also been improvement in the prenatal diagnosis of all trisomies. The improvements in these areas probably reflect better training of all staff providing maternity services, improved ultrasound equipment and the proper allocation of time to scan. However, the proportion prenatally diagnosed remains low for diaphragmatic hernia, for bladder outlet obstruction and for osteogenesis imperfecta.

Prenatal diagnosis of hypoplastic left heart improved substantially, perhaps due to training of ultrasonographers in recognition of the four chamber view and the main connections15; but there is room for improvement. In the most recent cohort, 60% of AVSDs diagnosed in the first year of life were unsuspected before birth, despite a threefold improvement in the proportion prenatally diagnosed.

The proportion of limb defects diagnosed prenatally remained low, particularly for amputation, where in three out of four affected babies it was unsuspected even in the most recent cohort. Our results are broadly similar to those from a recent report on prenatal diagnosis of limb reduction between 1990 and 1994 from a number of single centres in Europe.16 It can be very difficult for parents to understand how something as apparently obvious as a missing limb can be missed by a technique that readily detects brain and heart malformations in utero.

In summary a 16-year prospective study of prenatal diagnosis of fetal malformation in a defined population covering over half a million registrable births has shown substantial improvement in accuracy with time. Our rates of successful prenatal diagnosis during the appropriate time periods are broadly in line with those found in a number of smaller12–14 or less extended studies.16,17 Some conditions are clearly easier to diagnose prenatally than others. It is also possible that some units may be performing somewhat above or below the regional norm. We believe publication of this data will allow the development of a more balanced view of what could have been expected of prenatal diagnosis over the past 16 years.

Acknowledgments

The authors would like to thank the Survey Steering Group for permission to use the register and for their encouragement, Mrs Marjorie Renwick and her staff at the survey office for their tireless help and support in accessing the register and tracing incomplete returns. This register can only be maintained by the continuing active support of obstetricians, paediatricians, radiologists, ultrasonographers, geneticists, paediatric surgeons, paediatric cardiologists, paediatric pathologists and above all the secretarial staff working with all of the above in the health districts covered by the survey to whom we are indebted. We would also like to thank Edmund Hey for his continuing encouragement and support.

Contributors

SR and JA jointly conceived the study, analysed the data and wrote the manuscript. SR is the guarantor and accepts full responsibility for the study, had access to the data and controlled the decision to publish.

Funding

This analysis of the data was undertaken without funding.

Competing interest

None declared.

Accepted 21 February 2005

Ancillary