Choosing options for ultrasound screening in pregnancy and comparing cost effectiveness: a decision analysis approach

Authors


Correspondence: Ms T. Roberts, Health Economics Facility, University of Birmingham, Park House, 40 Edgbaston Park Road, Birmingham B15 2RT, UK.

Abstract

Objective To compare the cost effectiveness of different programmes of routine antenatal ultrasound screening to detect four key fetal anomalies: serious cardiac anomalies, spina bifida, Down's syndrome and lethal anomalies, using existing evidence.

Design Decision analysis was used based on the best data currently available, including expert opinion from the Royal College of Obstetricians and Gynaecologists, Working Party and secondary data from the literature, to predict the likely outcomes in terms of malformations detected by each screening programme.

Setting Results applicable in clinics, hospitals or GP practices delivering antenatal screening.

Main outcome measure The number of cases with a ‘target’ malformation correctly detected antenatally.

Results There was substantial overlap between the cost ranges of each screening programme demonstrating considerable uncertainty about the relative economic efficiency of alternative programmes for ultrasound screening. The cheapest, but not the most effective, screening programme consisted of one second trimester ultrasound scan. The cost per target anomaly detected (cost effectiveness) for this programme was in the range £5,000-£109,000, but in any 1000 women it will also fail to detect between 3.6 and 4.7 target anomalies.

Conclusions The range of uncertainty in the costs did not allow selection of any one programme as a clear choice for NHS purchasers. The results suggested that the overall allocation of resources for routine ultrasound screening in the UK is not currently economically efficient, but that certain scenarios for ultrasound screening are potentially within the range of cost effectiveness reached by other, possibly competing, screening programmes. The model highlighted the weakness of available evidence and demonstrated the need for more information both about current practice and costs.

INTRODUCTION

Offering women ultrasound screening for fetal anomalies during their pregnancy is often included as part of a package of routine antenatal care in the UK and the rest of Europe. It is well documented that there has been very little formal evaluation of this practice1–3. Like any other health intervention resources used for ultrasound screening might be used in other ways. To maximise benefits to society it is important to assess the relative advantages of the alternatives.

The Royal College of Obstetricians and Gynaecologists (RCOG) Working Party on Ultrasound Screening for Fetal Abnormalities4 was set up in 1995, in response to concerns raised by purchasers, professionals in the speciality, and occasionally patients, about the justification for routine ultrasound screening for fetal anomalies and the number of scans. Little real evidence exists, but anecdotal evidence suggests that the package of antenatal screening varies widely depending on the area of the country, health authority or antenatal clinic providing the care. For most women, the antenatal package will consist of at least one ultrasound scan5. But there is no consistent policy in hospitals or amongst clinicians regarding either the optimal number of scans, or when during pregnancy the scans should be carried out. There have been randomised controlled trials of various aspects of ultrasound use in pregnancy, but a randomised controlled trial addressing the optimal timing of ultrasound examinations for detecting fetal anomalies has not been carried out to date2,3.

A complicating factor is that the scans undertaken at different times during pregnancy potentially serve different purposes. The value of an early scan is mainly for dating purposes or checking viability. Accurate dating is considered important because it adds to the effectiveness of tests for Down's syndrome and other anomalies. A more accurately dated pregnancy can also lead to fewer post-term inductions6. A scan undertaken in the second or third trimester is useful for detecting certain anomalies but is of little value in dating23.

The authors of this paper, two of whom were members of the Working Party (MM, JP), undertook to review the economic evidence for routine ultrasound for anomaly screening, based on the best available data, and we report the methods and results in this paper.

METHODS

We compzred the cost effectiveness of alternative permutations of ultrasound screening. Since the alternatives vary in both costs and effects, cost effectiveness ratios, in terms of cost per anomaly detected, were calculated for each option. For the purpose of our evaluation we adopt the viewpoint of the health authority purchasing the screening. Therefore only direct costs to the health service are considered. We assume that their objective is the identification of as many true cases of malformations as possible, while minimising the identification of false positives and false negatives in time for women to make a decision about whether or not to proceed with their pregnancy.

The RCOG Working Party established that there were few strong data on patterns, timing and effectiveness of scans4. Despite this dearth, economic evaluation can have an important role. Economic evaluation in health care technology assessment should ideally be an iterative process. Undertaking analysis early in the diffusion of a new technology helps to identify the gaps in the evidence required for a complete economic evaluation but can also indicate options not likely to be viable. Furthermore, the identification of areas of high expenditure, widespread utilisation, but little clinical evidence, can help to justify future research. The characteristics of economic evaluations at different stages have been described elsewhere7. We carried out a stage 2 economic evaluation: that is, where there is some existing information on costs and outcomes, but it is not based on evidence from adequately powered randomised controlled trials (which are the gold standard8) or high quality costing exercises. The analysis lends itself well to the use of a decision analytic framework7. This consists of a logical presentation of the alternative patterns of care, and it guides us to the identification of the relevant cost data and clinical evidence required for the economic analysis9. It also has the advantage of allowing explicit synthesis of expert clinical opinion with the results of published and unpublished data from various sources. Obviously, the results of this type of analysis are subject to uncertainty especially since sufficiently robust evidence is not available. So rather than provide a definite, and possibly wrong, statement about the point estimates for cost effectiveness of ultrasound screening, we have considered it less misleading to identify ranges into which the cost effectiveness ratios may be expected to fall.

The Model

Routine antenatal ultrasound screening is often used alongside other screening procedures, such as serum screening, but our analysis focused only on the use of ultrasound screening. We considered this to be a necessary first step to develop a general model for comparing technologies. The model could be developed further to include options that combine ultrasound screening with other screening methods, but this was not the objective here. In the model serum screening is included only as part of a package of further tests used for confirmation of diagnosis and not for screening purposes.

Data

There are a range of abnormalities which can be detected through use of Ultrasound, but insufficient evidence and consensus about the value and effectiveness of routine screening for many of the possible anomalies reduced the range to four key anomalies. These were agreed by the Working Party as serious cardiac abnormality; spina bifida; Down's syndrome; and other lethal abnormalities4.

The patterns of screening options, timing and the effectiveness of the scans was based on the work of the RCOG Working Party and is documented in their report4. The Working Party defined 12 distinct options for use of ultrasound in screening for anomalies in pregnancy, based on combinations of one or more of the following types of scan:

  • 1First trimester scan (or early dating scan usually < 18 weeks)*.
  • 2First trimester anomaly scan (including nuchal translucency examination, usually 10–14 weeks).
  • 3Second trimester scan (usually carried out at 18–20 weeks gestation).
  • 4Third trimester scan (2840 weeks).

The two ‘first trimester’ scans are genuine alternatives. Women would not routinely receive both types of scan. One reason for including the alternative first trimester scans as separate options is that the examination of nuchal translucency is not widely adopted and is the subject of a multi-centre study at present10. However, for the four primary anomalies described above there seem to be genuine differences in diagnostic efficiency between the two first trimester scanning alternatives.

A list of the potential options for consideration is shown in Table 1. Option 4, for example, would consist of a third trimester scan only, and Option 5 would involve a first trimester dating scan and a second trimester scan. Options 2a, 5a, 6a and 8a are identical to Options 2, 5, 6 and 8 except for the fact that the first trimester dating scan is replaced by the first trimester anomaly scan. Option 1 (no scanning at all) is considered to be the baseline option, against which the other options are assessed.

Table 1.  Options for routine ultrasound scanning policies: source: RCOG4.
OptionFirst trimester datingFirst trimester anomalySecond trimesterThird trimester
1
2*   
3  * 
4   *
5* * 
6*  *
7  **
8* **
2a *  
5a ** 
6a * *
8a ***

The clinical effectiveness data required for the analysis was the likelihood of each screening option detecting any of the key abnormalities.

Each option consists of combinations of ultrasound tests. The outcome of each option therefore depends on the combined performance of the ultrasound tests defining the option. The final outcome is defined as the number of true cases of fetal anomaly detected through antenatal ultrasound screening, before women are faced with the decision of whether to terminate their pregnancy.

The performance of each ultrasound test is based on three criteria. These are the sensitivity of the screening test, which for this purpose means the proportion of true cases of abnormality identified correctly; the false positive rate (false positive rate = 1 - specificity) which is the proportion of true negatives identified as positive; and the population prevalence of the anomaly. These are the data required to assess the clinical effectiveness of each option. Where possible the evidence for sensitivity and specificity of each test and population prevalence of each target anomaly was derived from an extensive literature search carried out on behalf of the RCOG Working Party. Table 2 and Table 3 summarise these data which were subsequently graded by the Working Party committee according to their level of confidence in any particular estimate. These data were supplemented as necessary by professional judgements from individual members of the Working Party. The final estimates took the form of ranges of opinion within the committee and where there was extreme uncertainty with no research evidence it has been indicated in the tables. It was impossible to ascribe specific references to each final estimate because of our methods of deriving consensus.

Table 2.  Estimatedassumed sensitivity of the test for each anomaly.
Scan typeCardiac %Spina Bifida %Down's syndrome %Other lethal abnormalities %
  1. Source: RCOG literature review. The data in the above table vary in the strength of evidence. A query (?) indicates extreme uncertainty where no research evidence came to light (clinical hunch4).

First trimester dating00?30> 10
Anomaly
  First trimester0050–8510–30
  Second trimester10–2580–9010–2080–95
  Third trimester(? 50)(? 70–80)0–1080–95
Table 3.  Estimated combined diagnostic ability for each option.
OptionCardiacSpina bifidaDown's syndromeOther lethal abnormalities
  1. Source: Authors own calculations based on RCOG Working Party Report4. See Appendix 1.

10.000.000.000.00
20.000.000.300.10
30.10–0.250.80–0.900.10–0.200.80–0.95
40.050.70–0.800.00–0.100.80–0.95
50.10–0.250.80–0.900.37–0.440.82–0.96
60.500.70–0.800.30–0.330.82–0.96
70.55–0.630.94–0.980.10–0.280.96–0.99
80.55–0.630.94–0.980.37–0.500.96–0.99
2a0.000.000.50–0.850.10–0.30
5a0.10–0.250.80–0.900.55–0.880.82–0.97
6a0.50.7–0.800.50–0.870.82–0.97
8a0.55–0.630.94–0.980.55–0.890.96–0.99

Sensitivity

Many gaps were identified by the literature search. The data reflecting sensitivity of the test for each anomaly is shown in Table 2. To model the effectiveness of options involving more than one scan, we combined the diagnostic efficiency of each of the relevant building blocks. We assumed that the probabilities of an abnormality being detected for each scan were independent. In Appendix 1 we illustrate the method used to combine the diagnostic efficiency of each scan. We cannot over-emphasise that the quality of the evidence behind the estimates was variable. We estimated the effectiveness for all the target anomalies in the same manner (Table 3).

False positives

The false positive rate (ie, diagnosis of an anomaly when the fetus is not in fact affected) is important for two reasons. First, even if a positive test result is found to be false before parents make a decision about whether to terminate the pregnancy, such a result inflicts considerable and unnecessary psycho- logical distress to the parents. Secondly, costs to the NHS and risks to women may be incurred as a result of any tests and procedures which are carried out to confirm the diagnosis (ie, procedures which would not have been undertaken after a negative result). A false positives diagnosis can lead to two main outcomes: 1. there are false positives which are not corrected, and many women, believing their fetus to be malformed may choose termination of a healthy fetus without further diagnosis; and 2. there are false positives which are further investigated and found to be true negatives before women are faced with the decision about whether or not to terminate the pregnancy.

In our model, we assumed that all cases detected as positive after the combination of scans defined by the option path undergo a package of further nonroutine tests for confirmation of diagnosis. In reality, the additional non-routine tests may include one or more of a range of tests such as another scan, serum screening, amniocentesis or chorionic villus sampling combined with counselling at different stages. We have assumed that the package of additional tests included a scan, serum screening and amniocentesis. In order to ensure we could calculate an endpoint for each woman in terms of cost per case of fetal anomaly detected, we further assumed that the package conclusively detected the true anomalies and ruled out any false positive diagnosis made after the routine combination of scans defined by each option. The false positive rate in Table 4 represents an average for any group of scans defining an option.

Table 4.  Assumed prevalence per 1000 population and false positive rate for each anomaly. Source: RCOG Working Party Report4.
 Prevalence in 1000Average false positive rate (%)
Congenital cardiac disease8–10 (overall)(1)
 2–4 (serious)(1–2)
Open spina bifida1–4(1)
Down's syndrome1.5(5–12)
Lethal abnormalities1(<1)

Population prevalence

We used population based estimates for the prevalence of each ‘target’ malformation. These data were supplied by the RCOG Working Party and are shown in Table 4. For certain malformations, such as Down's syndrome, there are data available in the literature for the prevalence of the malformations for specific groups, such as women over 35 years of age. But to ensure consistency in the comparison with the other target malformations under analysis, we have used estimates of the overall population prevalence in all pregnant women. The available data were recognised to be crude and did not allow for changes in the prevalence of specific malformations over the pregnancy period. This may be an importantomission, as natural fetal loss is known to reduce birth prevalence at term in some cases11,12. Routine statistics reporting malformations are also known to be incomplete13,14.

Cost data

Most of the cost data were based on sources found in a very inclusive search of international health evaluation literature. We set a low threshold for our criteria of quality of cost evidence in our review of the studies identified. The cost data presented relate to the time period 1995–1996 and data from earlier sources have been adjusted accordingly using the NHS Executive Hospital and Community Health Services pay and price inflation index. The basis for the costs described in the literature varied widely (Table 5). In general little explanation was provided on the methods or types of cost reported. It was not clear if the various cost estimates included the associated costs of a scan, which includes for example, staff, equipment, overheads and counselling. Unfortunately, such detailed cost information is not currently available in the literature.

Table 5.  Cost at each node of decision model. Values shown are in £ sterling at 1996 prices. Source: Authors’ review of literature.
ProcedureBest scenario lower estimateWorst scenario upper estimateSource
No testsNil  
Scan30149References 20 to 23
 2065Northern & Yorkshire
   Regional Health Authority
   Extra Contractual Referral 1996)
Serum screening1525References 18 & 24
Amniocentesis95250References 18 & 19
CounsellingNot included No evidence found
Serum, scan & amniocentesis130424 

For the cost of ultrasound scans, we have used two cost ranges. The first is derived from the literature search and as can be seen, includes a wide range of costs. This could represent difference in the factors which are included or excluded in the estimates, and different input costs in different settings, and also might simply reflect cost calculation methodology in different studies. The second range of costs is derived from published prices charged by NHS providers in the North of England. These prices are extra-contractual referral prices, but are appropriate since these prices are supposed generally to be charged at full average cost. The range of cost estimates using this approach is much narrower.

Analysis

1. Clinical effectiveness

We constructed a decision tree to reflect the alternative options and to identify the clinical outcomes and link the options to these outcomes (Figs 1 and 2). The chance of following any particular path is determined by a node preceding the choice of paths. Decision nodes (square) reflect where choices must be made by women and their carers. Chance nodes (round) represent the uncertainty of an outcome after a clinical intervention, determined by the population incidence of anomalies and the assumed effectiveness of the ultrasound scan or scans. Terminal nodes (triangles) represent the outcome for a patient whose uncertainty is resolved in the manner indicated by the probabilities leading to the node.

The main clinical outcome measure used was the estimated number of women with a ‘target’ malformation detected antenatally. In this model our analysis looks only at outcomes in terms of diagnostic suc- cess in the short term. We do not extend the model to consider longer term survival and quality of life issues.

Because of the uncertainty of the quality of the data, it was decided to consider ‘best’ and ‘worst’ scenarios, from an economic viewpoint, for clinical effectiveness. The best scenario is that with highest prevalence for the target anomalies, highest likely sensitivity of the test for each anomaly and highest specificity (lowest false positive rate) for each anomaly. The worst scenario is with lowest prevalence for the target anomalies, lowest sensitivity of the tests and lowest specificity (highest false positive rate) for the detection of anomalies.

For each of the four target anomalies, we estimated outcomes of each of the screening options for both scenarios using the upper and lower limits of the data as appropriate. For each of the target anomalies the prevalence was used to estimate how many true fetal anomalies there would be in a population of 1000 women.

The combined sensitivities, presented in Table 3, were used to estimate the number of positive diagnoses for each of the four anomalies individually. We assumed that all women who are positively diagnosed will receive a package of further tests. Obviously for each option there will be a proportion of women who will have been given a false positive diagnosis because the specificity of the test for any option is not 100%.

The women who have been given a negative diagnosis will continue with their pregnancy. The prevalence minus the true positives is used to estimate the number of women who have been given a false negative diagnosis (ie, missed malformations) and who will proceed with their pregnancy unaware that they have a malformed fetus. For the best scenario, the proportion of women, in each scenario, receiving either diagnosis who follow the ‘proceed with pregnancy’ path or ‘further tests’ path are illustrated in Figs 1 and 2.

The true positives and the false positives, respectively, were first calculated for each individual anomaly within each scanning option and then combined to calculate the total number of true positives and false positives for each option respectively. For example, the total number of true positive Down's syndrome cases for Option 5 is estimated first, using the prevalence of Down's and the combined sensitivity of the scans for Option 5. This result was then added to the total number of true cardiac anomalies detected in Option 5, which is defined by its prevalence and the sensitivity of the option. This is repeated to calculate the total true positives for each target anomaly. The sum of the four target anomalies provides the number of true cases detected by each option. The total false positive cases detected by each option is calculated in the same way using the (1-specifity) instead of the sensitivity for each option. In our model there are three possible final outcomes: no malformations; malformations detected; and missed malformations. Although we acknowledge that miscarriage could occur at any stage, it is not explicitly considered. We do not depict branches for outcomes with (assumed) zero probability at each terminal node; therefore only three outcomes are shown as the final point of each option in Figs 1 and 2.

2. Cost effectiveness

To estimate cost effectiveness for the best and worst scenarios we attach the lower limit of costs to the best scenario and the upper limit of costs to the worst scenario. The cost of the initial package of scans which define each option path is multiplied by 1000 women to give the initial cost of that programme. The cost of the additional non-routine tests encountered by the women who have fetal anomalies detected, including the false positives is added to the initial cost of the programme to give the total cost of that programme. For example, the best scenario for Option 5 involves two scans: a first trimester dating scan and a second trimester anomaly scan. The best possible cost for both is the lowest estimate of £20 each, thus the total scanning cost for this option is £40 per woman. For 1000 women the initial cost of this option is £40,000. The number of cases detected by Option 5 is 79.78 false positive cases and 6.22 true positive cases. This implies that 86 women will undergo a package of additional tests at £130 per woman, which adds £11,180 to the cost. The total cost of Option 5 is £40,000 +£1 1,180 =£51,180, and the total number of true anomaly cases detected is 6.22. The cost per case detected for Option 5 is £8,228.

RESULTS

Table 6 shows the results of the analysis, defined in terms of best and worse scenarios for cost, target defects detected, missed target defects, and cost per case detected. These final results are based on the main outcome measure-number of women cor- rectly diagnosed with a ‘target’ malformation combined with the total cost of the screening option. However, there is a clear trade off between the outcome ‘true target cases detected’ and ‘target cases not detected’ or missed (ie, false negatives). The best scenario has a higher number of target malformations undetected than the worst scenario.

Table 6.  Preliminary estimates of costs and effects of ultrasound screening options per 1000 women screened. T1 (d) = first trimester dating scan; T1 = first trimester anomaly scan; T2 = second trimester scan; T3 = third trimester scan.
 Best scenarioWorst scenario
Options*Cost 1996 £Target defects detectedCost per case detected£Target defects missed*Cost 1996 £Target defects detectedCost per case detected£Target defects missed
  1. *Costs rounded to nearest £1.

2(T1(d))30,4500.5357,4539.95216,9710.55394,4934.95
3(T2)31,1395.855,3234.65217,5651.95111,5723.55
4(T3)31,1986.304,9524.20217,7982.5087,1193.00
5(T1(d)& T2)51,1806.228,2284.28366,7452.38154,0953.13
6(T1(d)& T3)51,2446.667,9643.85366,9982.67123,5682.53
7(T2 & T3)51,3997.856,5482.65367,0743.15116,5312.53
8(T1(d), T2 & T3)71,4428.188,7342.32516,2463.56145,0131.95
2a(T1)30,5831.5819,3568.96217,0990.85255,4104.65
5a(T1 & T2)51,2746.897,4423.61366,8602.65138,4382.89
6a(T1 & T3)51,3507.486,8653.03367,1253.27112,2712.23
8a(T1, T2 & T3)71,5188.778,1551.74516,3603.83134,8201.68

The most widely adopted current practice across the UK is difficult to assess because there are so few national data about this aspect of antenatal care. The consensus of the RCOG Working Party was that most women receive at least one ultrasound scan in pregnancy, and this is often in the first trimester for the purpose of dating which is defined by Option 2 (T1 (d) scan). The cost of this option per case detected, using the limits of our worst and best sce- narios is the range £50,750 to £361,618 per target anomaly detected; between 5 and 10 fetal anomalies would remain undetected in 1000 women who might follow such a practice. Alternatively, if the single scan performed is the first trimester anomaly scan, defined by Option 2a, then the cost per case detected is in the range £19,114 to £277,888 with between 4.7 to 8.5 remaining undetected target fetal anomalies in 1000 women. The majority of routine anomaly scanning thought to be done in the second trimester with the performance of a single anomaly scan (T2 scan)1, which is defined by Option 3. The cost per case detected for this programme is in the range £5,278 to £108,782 with between 3.6 and 4.7 undetected fetal anomalies.

There is clearly enormous variation in the costs and cost effectiveness of each of the options discussed. These variations are due to two compounding factors: uncertainty surrounding the clinical evidence regarding the detection rates, sensitivity and specificity of each option; and large differences in cost estimates according to the alternative sources used. Future work must address these issues with the aim of improving the quality of data both in terms of effectiveness and costs. Only then will it prove possible to make defensible decisions about which should be the preferred option (or options) for routine practice. Equally, this exercise considers routine screening for all women, with no mention of risk factor status. If routine screening was restricted to certain categories of women perceived to be at higher risk, as happens for Down's syndrome screening, then the cost effectiveness ratios will look more favourable.

DISCUSSION

Our preliminary economic evaluation has illustrated the wide range of uncertainty surrounding the comparative cost effectiveness of using the ultrasound screening options for the detection of fetal anomalies. The ranges presented are potentially within the range of cost effectiveness reached by other competing screening programmes. Direct comparisons between studies which have used different methods for cost and effectiveness estimation should be made with caution. However, the cost effectiveness approach has been adopted in comparisons of screening programmes for cystic fibrosis15–17 and Down's syndrome18. Estimates of costs per detected fetus affected by cystic fibrosis varied from £40,000 to £43,000. There is a higher cost per unwanted cystic fibrosis birth averted, because not all families would choose termination. This could also be the case in programmes for ultrasound screening.

We acknowledge that there are many respects in which our model is severely limited. We identified several aspects of the model where we had some information, but it was not obvious how it could be incorporated into the model. In particular, in estimating effectiveness in detecting one of a combined package of target anomalies, derived from known chances of detecting each individual type of anomaly, we assumed the risks of each anomaly to be independent, which is not the case. Heart disease and Down's syndrome are known to be related for example. We were also unable, in this model, to allow for the varying strength of evidence for different components of the model, arising from the diverse quality of sources of data and the professional consensus required to reach the final estimates used. A wide band of costs was used to reflect the real uncertainty about the true cost of each anomaly being targeted. The ideal cost data would take into account all the input costs for a single scan and such cost inputs are likely to vary with the size of the centre or unit per- forming the scan and the number of scans performed within a specified session. It is also unsatisfactory to use, as we did, a single generic cost for scans which are conducted for different purposes.

Furthermore, we did not account for the fact that each test can alter the risk level in the population of moving on to the next routine test. We have acknowledged the fact ultrasound screening is often used alongside other methods of antenatal screening such as serum screening and the model could be developed further to include these. However, our focus is on the use ultrasound alone in order to improve the quality of the information and data for its use. This is a necessary first step before development of the model to include other technologies is considered. In our analysis serum screening was included to confirm positive diagnosis as part of a package of further tests and not for screening purposes.

A further limitation is that the use of the cost effectiveness approach provides a unidimensional approach to outcome measurement which is potentially limiting because of the conflicting outcomes such as the trade off between target anomalies detected and target anomalies undetected. There are three main outcomes in our model which are related by the nature of test characteristics: 1. the number of true anomalies detected by each programme; 2. the number of women who had an unwanted target fetal anomaly that was missed; and 3. the number of women who are falsely identified as carrying an affected fetus. These are potentially devastating events for which a woman may be unprepared. Where the woman receives a false positive diagnosis, considerable anxiety can be associated with such a diagnosis until, and if, it is ruled out by further tests. Apart from the extra expense associated with the further non-routine tests required to rule out a false positive, some of the further tests, for example amniocentesis, have an associated raised risk of miscarriage. Rarely, false positive cases result in a terminated, wanted healthy fetus. A cost utility approach would capture the morbidity and mortality effects, and the trade- off between outcomes, in a single metric. This will only be worth addressing when there is both better cost data and better clinical evidence. However, the negative impact of a false diagnosis must be weighed up by decision makers, in addition to the evidence of the costs of achieving the desired outcomes.

Given the uncertainty illustrated by the wide over- lapping ranges, and the unknown factors which we had to exclude, it was not possible to select any single programme as a clear choice for NHS purchasers. The lack of evidence and the paucity of robust data are the fundamental reasons for this uncertainty. Yet we have analysed the information that is currently available for decision makers, and upon which current decisions are presumably based. We have shown therefore, the difficulty for decision makers in taking any necessary action in connection with policies on ultrasound screening to maximise the total health gain from the fixed resources at their disposal.

We are led to conclude that the overall allocation of resources for routine ultrasound screening in the UK is currently not economically efficient, in that there may be better ways of using the resources taken up in the screening process. However, to discontinue the process of routine ultrasound screening may be equally inappropriate, since the best cost-consequence ratios that we have estimated suggest that the intervention is comparable with other technologies for anomaly screening. Our economic evaluation has demonstrated one stage of an iterative process which has served an important purpose by identifying an area of high expenditure on an accepted clinical practice for which there is little evidence. Given the large amount of resources used in routine ultrasound screening, it is essential that robust evidence is obtained which would allow the areas of uncertainty outlined in our model to be resolved. There is clearly therefore a need for further research. We are now engaged in further work, funded by the NHS Research and Development programme to refine the model as part of a systematic review to assess the cost effectiveness of antenatal ultrasound screening, alongside the Cochrane Pregnancy and Childbirth Review Group.

Acknowledgements

We would like to thank the RCOG Working Party on ultrasound screening for fetal abnormalities for their permission to use their data and their ‘expert’ opinion in the absence of available data, and for their comments on earlier drafts of this paper. We would like to thank the members of the UK Health Economics Study group for their helpful comments on an earlier draft. We would also like to thank our colleagues at the National Perinatal Epidemiology Unit who commented on drafts. Ms T. Roberts and Dr M. Mugford were based at the National Perinatal Epidemiology Unit and funded by the Department of Health; Mr. Piercy was funded by the Trent Regional Health Authority.

Footnotes

  • *

    The first trimester is defined as < 13 weeks. The early dating scan is often carried out in the first trimester but can be carried before 18 weeks. For consistency with the RCOG Report we include both.

Appendix

Appendix 1.

To illustrate the method used to combine diagnostic efficiency of each scan in the determination of the detection rate for Down's syndrome under option 5 (first trimester dating plus second trimester scan).

  • 1Probability of detecting Down's syndrome using a dating scan=0.3
  • 2Probability of detecting Down's syndrome using a second trimester scan = 0.1–0.2
  • 3The probability of detecting Down's syndrome at either scan is between 0.37 and 0.44:
    image

Ancillary