Obstetric ultrasound has universal use in developed countries but at different levels of expertise between the various centers. Because the correct practice of ultrasound is strongly operator-dependent, education and training are of paramount importance. Ultrasound malformation screening is the best example of this need for education and training. Screening for fetal malformations is generally regarded as a worthwhile endeavor and is taken for granted in the field of obstetric ultrasound. Obviously, the best results in detecting malformations will be obtained by the most skilled operators. However, it is unrealistic to consider organizing a screening program that is dependent on the participation of expert operators only. Such expert operators are naturally requested for targeted examinations, however they are not sufficiently numerous to be able to handle a mass screening program involving all the pregnant patients attending their institution of which only some 2% will have a malformed fetus.
Discussion as to whether antenatal screening for fetal malformation is justified will not be included here. It is generally accepted that ultrasound screening is a useful procedure for a number of important reasons. Ultrasound screening permits the diagnosis of the majority of structural abnormalities and a more rational approach to the diagnosis of chromosomal abnormalities; it allows termination of pregnancy at the parents' request for severe abnormalities, while for operable lesions it permits the organization of optimal care or surgery for the newborn and psychological support for the parents. My approach to the purpose, organization, methods and benefits of screening have been extensively discussed elsewhere1, 2.
In this Editorial I will address the efficiency of mass population screening. For this reason it is important to consider sensitivity studies performed on large population samples. Only a few papers in the literature have analyzed routine scanning on samples from more than 100 000 pregnant women. Two papers published prior to 19963, 4 reported a screening sensitivity for major malformations of 26.7% and 28.5%, respectively, with a comparable prevalence of malformed fetuses (1.9% and 2.0%, respectively). Another two studies appeared in journals in 1999 and 2001, reporting results from the Eurofetus study5 and the Euroscan study6.
The Eurofetus project7 was designed to test the efficiency of ultrasound in detecting fetal malformation in a mass screening program. Eurofetus was conducted prospectively between 1990 and 1993 in institutions in which good conditions for fetal examination were present, and the program was focused on screening for congenital anomalies in non-selected pregnant women. The ultrasound units involved included personnel (nurses, midwives, technicians, physicians) with daily experience of routine obstetric ultrasound and who practiced malformation screening on a regular basis as a full- or part-time occupation. Additional training personnel could be part of the examiners' team under the supervision of the experienced professionals. Although most of the collaborating units included at least one individual with average to high expertise in fetal ultrasound, the staff members in general were not expert in fetal malformation assessment at tertiary level experience.
All the patients with an indication for ultrasound examination were systematically removed from the Eurofetus study in order to comply with the non-selection principle and to retain the average frequency of anomalies found in a normal population, i.e. 2%8. We have commented previously on the relationship between the frequency of the important group of congenital heart defects and the sensitivity of fetal malformation screening2, 9.
In 2001 the Editorial in the October issue of this Journal was dedicated to the Euroscan study6. The Eurofetus and Euroscan studies had similar aims and the European Union sponsored both studies. Fortunately these studies were not duplicative but complementary, therefore making appropriate use of the European Union funding. In addition to sharing certain similarities, the two projects also exhibited fundamental differences.
Let us examine the similarities first. The aims of the two projects were similar, i.e. evaluation of the prenatal detection by ultrasound of all congenital malformations in unselected populations. Both studies adopted the same endpoints, such as excluding identical minor malformations together with anatomical ‘peculiarities’ regarded as ultrasound markers of a potential genetic malformation.
However, the material selection for the two studies differed. Indeed, Eurofetus was a prospective institution-based study while Euroscan was a retrospective population-based study (the data were collected from malformation registries covering several defined geographical areas).
Thus the comparability of the studies might be altered by subtle bias as follows:
The expertise of examiners from a geographical area-based study can be very heterogeneous, reflecting the totality of ultrasound practice from basic office to tertiary ultrasound units. Basic office practice—characterized by relatively few patients and a rare occurrence of malformation—is usually linked with low expertise. Specialized units usually have a high throughput of malformations and are usually run by expert personnel.
Low-risk and documented high-risk pregnant women are mixed in a population-based study although an excessive number of high-risk patients might be included in samples from areas characterized by a high concentration of obstetric tertiary centers.
Although a standard region should have a full range of ultrasound skills and an average number of fetal malformations, the weight of each of these very different practices in making up the totality is not known. Conversely, Eurofetus had a more homogeneous practice, thus making it easier to evaluate the overall and the individual unit performance.
We have tracked down samples exceeding 30 000 pregnant women and observed that the sensitivity observed in population-based studies (24% to 38%)4, 10–12 was lower than in the Eurofetus institution-based (61%) study5.
Some methodological differences are observed also:
The proportion of false-positives is not available for the Euroscan study.
Major malformations not detectable by ultrasound were excluded from the Euroscan study.
Unlike the Euroscan project, all the Eurofetus centers have a routine screening policy: at least one scan around the 20th week for nearly all the pregnant women attending the institutions.
The frequency of malformed fetuses or babies in the Euroscan study is 1.15%. This is considerably lower than the mean frequency disclosed in the Eurocat Registry of malformations (2.27%)8, which includes most of the registries covered by the Euroscan study, and also much lower than the frequency disclosed by Eurofetus (2.0%). (NB. Extrapolation from available data concerning 76% or 2600 out of the 3400 of the Eurofetus malformed fetuses).
The Euroscan detection rate was not given as a whole but only for specific classes of anomalies.
I will now illustrate some rarely considered facts and misconceptions by considering the Eurofetus study. The severity of the anomalies and the outcome for malformed fetuses are rarely considered in screening studies. I will illustrate this with reference to congenital heart defects (CHD) to show the influence of severity on the outcome of screened fetuses. CHD are chosen for two reasons:
CHD can be equally distributed into serious and relatively benign lesions.
Antenatal diagnosis of CHD is difficult, unlike other severe lesions such as those affecting the central nervous system (CNS), and the number of detected cases of CHD does not greatly exceed the non-detected cases. Hence, comparison between missed and detected anomalies becomes statistically assessable13.
Eurofetus has shown that the global sensitivity of CHD detection (34%) is significantly lower when compared to CNS (88%) and to urogenital malformations (89%). This observation would appear to support some of the reservations expressed with regard to CHD screening. However, the study also demonstrated that there are large differences in sensitivities between isolated and associated CHD sensitivity, 23% vs. 67% respectively, and between severe and benign isolated lesions, i.e. 56% vs. 5%, respectively. It is therefore obvious that a shift in the proportion between the subgroups can alter significantly the sensitivity figures.
Conversely, let us reflect on what we can expect from antenatal detection. Although discussed on several occasions, it remains paradoxical that the fetuses that are diagnosed antenatally have a poorer outcome. This can be explained by the fact that the majority of the detected abnormalities are severe malformations compared with those not diagnosed (85% vs. 36%, respectively).
The detection of fetal malformation reduces significantly the number of affected newborns. Among the pregnancies with the most severe CHD (68 cases), 52% of the detected cases were terminated, while an additional 12% died in utero and 24% died before day 6. Conversely, all of the missed abnormal fetuses were born alive but 51% died before day 6. It is possible to rule out the effect of pregnancy termination by examining the less severely affected fetuses (297 cases) since none of these was terminated. The total spontaneous loss was 19% in the detected group vs. 3% of the missed cases. Obviously, the set of detected anomalies included more severe lesions than the set of non-detected anomalies.
The antenatal detection of cardiac malformation does not influence the rate of Cesarean section. The complexity of the defect favors per se the occurrence of premature birth, as observed in both the antenatal and postnatal diagnosis groups.
We encountered a cardinal problem in attempting to complete one aspect of the Eurofetus study. We failed to achieve a satisfactory analysis of the cost–benefit aspect of the screening project. Despite the expertise of the specialized team in charge of the project, it appeared that the costs of the scanning process and the consequences of false-positive and false-negative cases were impossible to calculate. Our health economist experts were unable to obtain financial data of sufficient quality from the institutions involved, and in addition the available health statistical data were often incomplete. The shortage of records required to calculate the costs with even minimal accuracy to achieve average scientific standards meant that the cost–benefit calculation had to be discarded. However, a few studies on the costs associated with specific malformation identification can be found in the literature together with additional references 1, 14–18.
The high number of malformed fetuses in the Eurofetus database provides robust information on the frequency of particular malformations in the population together with meaningful statistics on the detection rates of these malformations.
At the beginning of this article the clear dependence of ultrasound screening efficiency on education and expertise was stressed. However, at a similar level of expertise, screening efficiency might be altered by other often quite simple factors such as an excess of routine work or the inclusion of too many scanning staff members. Too many routine anatomical scans done each day, combined with the expected low prevalence of anomalies, can result in lowered operator attention. Too many part-time scanning staff members could reduce the average expertise level. Indeed, looking at the overall detection rate in the Eurofetus study (sensitivity 61%) we noticed that the highest sensitivity (71%) was reached by 45 centers scanning a population of less than 1500 pregnant patients per year, while the three centers scanning in excess of 4000 patients per year achieved a much lower average sensitivity (47%), thus suggesting that a high throughput lowers screening efficiency.