High-throughput non-invasive prenatal testing for fetal rhesus D status in RhD-negative women not known to be sensitised to the RhD antigen: a systematic review and economic evaluation.

Summary of technologies (index tests)

The technology under assessment is high-throughput non-invasive prenatal testing (NIPT) for fetal rhesus D status (International Blood Group Reference Laboratory, NHS Blood and Transplant, Bristol, UK).

High-throughput NIPT of fetal RhD status uses a real-time quantitative polymerase chain reaction (PCR) method for predicting the fetal RhD genotype from fetal deoxyribonucleic acid (DNA) in the plasma of RhD-negative women. The test principle is based on the analysis of cell-free fetal DNA, that is, small fragments of fetal extracellular DNA shed from the placenta and circulating freely in the maternal plasma. The level of cell-free fetal DNA in maternal blood increases throughout the pregnancy. A woman who is RhD negative does not have a copy of the RHD gene; therefore, the presence of a RHD gene in a RhD-negative pregnant woman suggests a RhD-positive fetus.

High-throughput NIPT is performed using samples of maternal anticoagulated blood. DNA extraction is performed using an automated robotic platform, which can rapidly process samples. The robotic platform is used as a liquid handler to dispense samples and reagents. In the UK, primers and probes for specific exons of the RHD gene are used, with a number of controls being tested (such as RhD-positive DNA, RhD-negative DNA, RHD pseudogene positive DNA and no DNA). An algorithm is employed to determine the fetal RhD status. The samples can be tested in batches of between 32 and 88 samples. The time to complete the test from sample receipt to report generation is 5–6 hours.

High-throughput NIPT for fetal RhD status may enable anti-D immunoglobulin to be withheld from RhD-negative women who are carrying a RhD-negative fetus. These women could avoid unnecessary treatment with routine anti-D immunoglobulin, along with the potential risk associated with administration of blood products. In addition, these women may not need the provision of anti-D immunoglobulin following potentially sensitising events and there may no longer be a need for serological cord testing at birth.

Identification of important subgroups

There are potential challenges for the detection of fetal rhesus D status when performing NIPT in pregnant women. Dealing with the presence of RHD pseudogene poses a challenge. The majority of RhD-negative individuals with white European ethnicity have the pseudogene as a result of gene deletion; however, in people with African ethnicity the Rh-negative phenotype is mainly the result of genes that contain RhD sequences but do not produce D antigen (RHD pseudogene). 5 In the presence of the RHD pseudogene, prenatal determination of fetal Rh type from maternal blood would reveal a RhD-positive type, but this would be confirmed as RhD negative by serology because of the abundant maternal D gene sequences that are not expressed but are amplified. This may, therefore, lead to higher rates of false-positive results when performing NIPT in this population.

There is a diverse array of Rh variant genes and it is generally accepted that at least two exons of RHD should be targeted for accurate RhD status prediction. For instance, targeting only exon 7 (or exon 10) would not detect the presence of the RHD pseudogene and other variants and targeting only exon 10 would not detect the presence of the RHD pseudogene or the hybrid RHD-CE-D(^s) gene, which are commonly present in people with African ethnicity.

Evidence suggests that the diagnostic accuracy of NIPT may vary according to different gestational ages at the time of sampling. Two meta-analyses found that the diagnostic accuracy of NIPT was higher in the first trimester than in the second and third trimester. 10,11 However, a recent UK cohort study found that fetal RhD genotyping was more accurate for the prediction of RhD status if it was performed after, rather than before, 11 weeks’ gestation. 12

In this assessment we aim to investigate findings of high-throughput NIPT from a number of subgroups, such as those based on different gestational ages and different ethnicities as well as on the usage of different exons of RHD, if data are available.

Current usage in the NHS

Currently, all high-throughput NIPT for fetal RhD status determination in the UK is performed by the NHS Blood and Transplant International Blood Group Reference Laboratory in Bristol. If all pregnant RhD-negative women in England were to be tested, approximately 100,000 samples would be tested each year. An increased capacity would be required for the International Blood Group Reference Laboratory to be able to cope with this demand by employing additional staff and acquiring more analytical platforms. Beyond this, extending the testing service to other laboratories is an alternative option. Blood samples would need to be transported from local hospital laboratories to the International Blood Group Reference Laboratory in Bristol or other laboratories. The established NHS Blood and Transplant transport system would be used to deliver blood samples across the country. This would need to be achieved in reasonable time, although there is evidence to suggest that cell-free fetal DNA is very stable. 13 There would also need to be reporting systems in place to ensure the accurate transmission of test results back to the women and their physicians and midwives.

Expected costs associated with technology

The potential costs associated with high-throughput NIPT to the NHS comprise two components. First, there is the unit cost of the diagnostic test itself, which varies with the level of throughput and to which a royalty fee may be added. An estimated unit cost for high-throughput NIPT of (confidential information has been removed) and a royalty payment of (confidential information has been removed) were considered. It should be noted that these estimates were provided in confidence by the company with the underlying assumption that the International Blood Group Reference Laboratory in Bristol will be the sole provider of the test nationally. Second, the potential costs of incorporating the test into routine antenatal care must be considered, which may bring additional costs relating to the time for antenatal care appointments to provide information about the test, counselling and delivering test results and also relating to blood drawing and blood sample transportation.

Searches

The literature search aimed to systematically identify studies relating to the clinical effectiveness and cost-effectiveness of high-throughput, non-invasive, prenatal blood testing to determine fetal rhesus D status.

The search strategy was developed in MEDLINE (via Ovid) and then adapted for use in the other resources searched. The strategy included terms for rhesus D status combined, using the Boolean operator AND, with terms for the test. No language, date or geographical limits were applied and study design search filters were not used. EndNote X7 software (Thomson Reuters, CA, USA) was used to manage the references for the project.

Search strategies were developed by an information specialist with input from the project team. The search strategy was checked by a second information specialist.

The following databases were searched for relevant clinical effectiveness or cost-effectiveness studies from inception to November 2015: MEDLINE, MEDLINE In-Process & Other Non-Indexed Citations, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effects (DARE), EMBASE, Health Technology Assessment (HTA) database, Maternity and Infant Care, NHS Economic Evaluations Database (NHS EED), PubMed and the Science Citation Index.

In addition, the following resources were searched for ongoing, unpublished or grey literature: ClinicalTrials.gov, Conference Proceedings Citation Index: Science, EU Clinical Trials Register, PROSPERO and the World Health Organization’s International Clinical Trials Registry Platform portal.

The following websites were searched to identify any relevant guidelines: National Guidelines Clearinghouse, NICE, NHS Evidence, the Royal College of Obstetricians and Gynaecologists, the Turning Research into Practice database and the UK National Screening Committee. Reference lists of relevant reviews and included studies were checked to identify additional potentially relevant reports. The searches were updated in February 2016. A full search strategy can be found in Appendix 1.

Selection criteria

Data extraction

We selected the most recent or most complete report in cases of multiple reports for a given study or when we could not exclude the possibility of overlapping populations.

The data extraction forms were developed and piloted. One reviewer independently extracted details from full-text studies of study design, participants, index, comparator and reference standard tests and outcome data. The data extraction was checked by another reviewer. Any disagreements were resolved by consensus or by recourse to a third reviewer.

For studies reporting diagnostic data, we extracted the number of true positives, true negatives, false positives and false negatives for each index test evaluated in each study to construct 2 × 2 tables. If such data were not provided by the study authors, we attempted to contact them to construct the 2 × 2 table for the study population or the prespecified subgroups. Otherwise, we calculated the number of true positives, true negatives, false positives and false negatives from the summary estimates of sensitivity and specificity of the index test, if available. If reported, we extracted data on the number of undetermined or uninterpretable results. For studies in which only a subgroup of patients was included in the review, we extracted, analysed and presented data for this subgroup only. If some data were unclear or missing, we attempted to contact study authors to obtain additional data.

For studies reporting clinical outcomes, we extracted data as the numbers of women or fetuses experiencing the specified outcome. Mean differences, relative risks (RRs) or odds ratios [with 95% confidence intervals (CIs)] were extracted from comparative studies, when reported as unadjusted data.

For the implementation review, we summarised the findings and conclusions of the included publications using the following broad categories: study results and findings, issues for implementation, practical guidance and recommendations for research.

For the review of anti-D prophylaxis, we extracted summary results from syntheses or meta-analyses of studies on each clinical outcome reported. Mean differences, RRs or odds ratios with 95% CIs were extracted, when reported.

Critical appraisal

One reviewer independently assessed the quality of all included studies in terms of risk of bias. Risk of bias from diagnostic accuracy studies was assessed using a modified version of the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) checklist. 14 The QUADAS-2 tool was adapted to ensure that it is applicable to assessing the quality of studies of non-invasive prenatal tests for detecting rhesus D status. The QUADAS-2 tool consists of four key domains: (1) patient selection, (2) index test, (3) reference standard and (4) flow of patients through the study and timing of the index test(s) and reference standard. Each domain was assessed in terms of the risk of bias. The first three domains were also assessed for concerns regarding their applicability in terms of whether or not the participants and setting; the index test, its conduct or interpretation; and the target condition, as defined by the reference standard, were applicable to nationwide screening in the UK.

A Cochrane Risk Of Bias Assessment Tool: for Non-Randomised Studies of Interventions (ACROBAT-NRSI) was used to assess risk of bias for each outcome of all comparative studies reporting other eligible clinical outcomes. The quality assessment was checked by another reviewer. Any disagreements were resolved by consensus or by recourse to a third party.

The quality of the studies in the implementation review was not assessed, as there is no validated tool for assessing the quality of studies on the implementation of health interventions.

Methods of data synthesis

Using extracted diagnostic accuracy data from the 2 × 2 tables, estimates of sensitivity, specificity, false-positive rates (FPRs) and false-negative rates (FNRs) were calculated and presented on forest plots and in receiver operating characteristic (ROC) space to examine the variability in diagnostic test accuracy within and between studies. In the primary analysis, undetermined or uninterpretable results were counted as being test positive, in accordance with current practice.

The hierarchical bivariate model described by Reitsma et al. 15 was fitted, which calculates summary estimates of sensitivity, specificity, FPRs, FNRs and the associated 95% CIs. The hierarchical summary receiver operating characteristic (HSROC) model16 was fitted to produce summary ROC curves. Results of both models were presented in ROC plots.

Other eligible clinical outcomes were pooled if at least two studies reported on the same outcome and if data were reported consistently enough for analysis to be feasible. Otherwise, results were synthesised narratively. When meta-analyses were performed, data were pooled using standard random-effects DerSimonian and Laird meta-analyses. Analyses were conducted in R version 3 (The R Foundation for Statistical Computing, Vienna, Austria) and/or Stata^® version 14 (StataCorp LP, College Station, TX, USA) software, as appropriate.

Simulation study of clinical effectiveness

During the course of this report we found very little evidence on the probable clinical effectiveness of high-throughput NIPT and its impact on future sensitisation rates and adverse events. In order to investigate these issues, we opted to perform a simulation study to simulate possible outcomes of high-throughput NIPT in the UK, based on results from the diagnostic accuracy review and the results of published systematic reviews of antenatal anti-D prophylaxis and relevant audit data identified through additional literature searches.

The simulation sought to estimate the following in the UK population:

• rates of women with a RhD-positive fetus

• rates of women with positive/negative/inconclusive NIPT results

• rates of women who receive NIPT and/or antenatal anti-D prophylaxis

• number of sensitisations

• number of adverse effects on fetuses in subsequent pregnancies.

Data were extracted from the diagnostic accuracy review, existing systematic reviews of antenatal anti-D prophylaxis and other primary sources, when necessary.

We considered the following clinical scenarios:

• no antenatal anti-D and postpartum anti-D based on cord blood serology only (control)

• antenatal anti-D offered to all RhD-negative women (current practice)

• antenatal anti-D offered based on NIPT and postpartum anti-D based on cord blood test for all RhD-negative women

• antenatal and postpartum anti-D offered based on NIPT only. No cord blood testing.

Scenario 3 is equivalent (in clinical outcomes) to performing cord blood testing on women with negative NIPT but offering postpartum anti-D to all test-positive women without cord blood testing. Scenario 4 is equivalent (in clinical outcomes) to withdrawing cord blood testing and postpartum anti-D for women with negative NIPT but offering cord blood testing and postpartum anti-D (if needed) to all test-positive women.

A Monte Carlo simulation of 10 million women was performed in R. Monte Carlo analysis is a modelling method that uses random number generation to simulate the running of multiple scenarios to define all potential outcomes of an event. We compared the amount of antenatal anti-D prescribed, the level of unnecessary anti-D use and the relative numbers of sensitisations and other adverse outcomes for each scenario.

Quantity and quality of research available

Number of studies included

The literature searches of bibliographic databases identified 3921 references. After initial screening of titles and abstracts, 227 were considered to be potentially relevant and were ordered for full-text paper screening. In total, eight studies12,17–23 were included in the diagnostic review of high-throughput NIPT, seven studies18,20,22,24–27 were included in the clinical effectiveness review and 12 studies13,17,18,20–28 were included in the review of implementation of high-throughput NIPT (with some overlap between studies). Figure 1 shows a flow diagram outlining the screening process with reasons for exclusion of full-text papers.

FIGURE 1.

Flow diagram: study selection process.

All studies except two8,28 were cohort studies. Most cohorts were reported in several papers and abstracts, with considerable overlaps in data and reporting. For each cohort and each review we selected the paper with the most up-to-date and complete data. Consequently, some papers were included in more than one review and some papers (mostly conference abstracts with limited or outdated data) were not included in any analysis. Table 1 presents an overview of these cohort studies, the publications associated with each cohort and in which review the publications were included. Appendix 2 presents a list of all included references.

TABLE 1 - Overview of included cohorts and studies

Cohort (country)	Number of full-text papers	Number of conference abstracts	Papers included in review
			Diagnostic accuracy (full-text paper)	Clinical effectiveness (full-text paper)	Implementation (full-text paper)	Linked conference abstracts
UK (Bristol)	3	6	Chitty et al., 2014;12 Finning et al., 2008;17 and Soothill et al., 201518	Soothill et al., 201518	Finning et al., 2008;17 and Soothill et al., 201518	Chitty et al., 2011;29 Chitty et al., 2012;30 Daniels et al., 2012;31 Finning et al., 2015;32 Finning et al., 2014;33 and Ford and Soothill, 201634
UK (London)	2	0	Akolekar et al., 201119	None	Oxenford et al., 201328	None
Denmark	4	5	Banch Clausen et al., 201420	Banch Clausen et al., 2014;20 Banch Clausen et al., 2012;24 and Damkjaer et al., 201227	Banch Clausen et al., 2014;20 Banch Clausen et al., 2012;24 Clausen et al., 2013;13 and Damkjaer et al., 201227	Banch Clausen 2012;35,36 Dziegiel 2012;37 Banch Clausen et al., 2011;38 and Steffensen et al., 201239
The Netherlands	2	10	Thurik et al., 201521	de Haas et al., 201225	de Haas et al., 2012;25 and Thurik et al., 201521	Veldhuisen et al., 2014;40 Veldhuisen et al., 2013;41 Thurik et al., 2014;42,43 Scheffer et al., 2013;44 van der Schoot et al., 2005;45 de Haas et al., 2012;46 de Haas et al., 2013;47 Grootkerk-Tax et al., 2006;48 and van der Ploeg et al., 201549
Spain	1	0	Grande et al., 201322	Grande et al., 201322	Grande et al., 201322	None
Sweden	2	10	Wikman et al., 201223	Tiblad et al., 201326	Wikman et al., 2012;23 and Tiblad et al., 201326	Wikman et al., 2012;50 Wikman et al., 2011;51 Wikman 2013;52 Wikman et al., 2010;53 Tiblad et al., 2010;54 Tiblad et al., 2012;55 Neovius et al., 2014;56 Tiblad 2012;57 and Neovius et al., 201658
Total			8	7	12	31

Results: assessment of diagnostic accuracy

Meta-analyses of diagnostic accuracy

This section presents the results of the meta-analyses of the diagnostic accuracy studies. One key issue when considering the diagnostic accuracy of NIPT is how women with inconclusive test results are handled. It is expected that, in the UK, such women will be treated as having a positive test with no further testing. Although this was the policy in the three high-quality studies performed in Bristol, data on inconclusive tests were not reported in two studies. 21,22

Given these differences we considered four approaches to the diagnostic analysis:

1. women with inconclusive tests treated as test positive (including Thurik et al. 21 and Grande et al. 22 studies)

2. women with inconclusive tests treated as test positive (excluding Thurik et al. 21 and Grande et al. 22 studies)

3. excluding all women with inconclusive test results

4. studies conducted in Bristol only. 12,17,18

This last analysis is likely to represent the most plausible results for UK practice, assuming that the methods used in Bristol are retained nationwide.

In all analyses, women whose NIPT was conducted at or before 11 weeks’ gestation were excluded when possible because of concerns that the diagnostic accuracy is poorer before 11 weeks and that the test should not be conducted before then (see Subgroup analyses). Some tests were performed between 8 and 11 weeks’ gestation in two studies,17,23 most women were tested between 8 and 12 weeks’ gestation in Wikman et al. 23 and < 8% of tests were performed before 11 weeks in Finning et al. ,12 but it was not possible to remove those women from the analysis.

In diagnostic analyses it is conventional to report results in terms of sensitivity (women who correctly test positive) and specificity (women who correctly test negative). NIPT is highly accurate and the focus should be on women with an incorrect test result, so in these analyses results are presented in terms of the FPRs (women incorrectly testing positive and so offered unnecessary anti-D) and FNRs (women incorrectly testing negative and so at risk of sensitisation, as they do not receive anti-D treatment).

A summary of all the results of the bivariate meta-analyses of FPRs and FNRs is presented in Table 4.

TABLE 4 - Bivariate meta-analyses of FPRs and FNRs

Analysis case	Number of studies	FNR (at risk of sensitisation)		FPR (unnecessary anti-D)
		Estimate (%)	95% CI	Estimate (%)	95% CI
Inconclusive tests treated as test positive (including Thurik et al.21 and Grande et al.22)	8	0.34	0.15 to 0.76	3.86	2.54 to 5.82
Inconclusive tests treated as test positive (excluding Thurik et al.21 and Grande et al.22)	6	0.38	0.15 to 0.94	4.37	2.79 to 6.78
Excluding all women with inconclusive test results	8	0.35	0.15 to 0.82	1.26	0.87 to 1.83
Studies conducted in Bristol only	3	0.21	0.09 to 0.48	5.73	4.58 to 7.16

It can be seen that results are broadly consistent across the four scenarios. NIPT is very accurate among women with a RhD-positive fetus: only 2–4 in 1000 of such women will have a negative test result and so be at risk of sensitisation as a result of not being offered anti-D. NIPT is slightly less accurate among women with a RhD-negative fetus: between 1.3% and 5.7% of such women will test positive (depending on the analysis performed) and so may be offered NIPT unnecessarily. If women with inconclusive test results are excluded from analyses, the FPR was 1.3%, rising to 3.9–4.4% if women with inconclusive test results are treated as having tested positive. This suggests that the main cause of test error is treating women with an inconclusive NIPT result as if they had tested positive.

Assuming that 60% of RhD-negative women have a RhD-positive fetus, about 0.5% of women have a conclusive, but incorrect, positive test result. About 0.1–0.2% of women have a false-negative test result.

We consider the results of each analysis in more detail in the following sections.

Considering inconclusive results as test positive

Figure 2 shows forest plots of FNRs and FPRs when counting an inconclusive test result as being test positive. The results of these figures are slightly different from those in Table 4, because the figure shows separate analyses of FPR and FNR, rather than a full bivariate analysis.

FIGURE 2.

Forest plots of (a) FNR and (b) FPR when counting an inconclusive test result as being test positive.

There was some evidence of inconsistency across studies. The I²-statistic for heterogeneity was 75% for the FNR and 99% for the FPR. It should be noted that these high heterogeneities are, in part, a consequence of the high accuracy of the test and the large size of the studies (and consequent small within-study variance, because I² increases as the average within-study variance declines). They do not necessarily indicate any clinically meaningful differences between studies. The heterogeneity in FPRs is likely to be a consequence of differing reporting and handling of inconclusive tests.

Figure 3 shows the results of each study, the results of the bivariate analysis (black circle) and the summary HSROC curve (black curve) for this analysis. As for other analyses, this is presented in terms of FPR and FNR rather than sensitivity and specificity. This plot shows the consistency of false-negative results, except for two outlying studies. 19,23 The Wikman et al. 23 study performed most NIPT in the first trimester, earlier than other studies. As discussed later (see Subgroup analyses), the timing of NIPT may have an impact on the FNR. The studies are less consistent in FPRs. This is most probably because the studies have different numbers of inconclusive test results and different methods of handling such results. As women with an inconclusive result are treated as RhD positive, women with an inconclusive result but a RhD-negative fetus will have a false-positive result. There may also be some heterogeneity because of differences in the threshold used and how different testing machines operated.

FIGURE 3.

Hierarchical summary receiver operating characteristic and bivariate analysis when counting an inconclusive test result as being RhD positive.

When excluding the two studies that did not report numbers of inconclusive tests,21,22 the results were broadly similar, as seen in Table 4. The forest plots of FPR and FNR for this analysis are given in Appendix 6.

Excluding inconclusive results

We considered the diagnostic accuracy of NIPT, excluding all inconclusive test results, to identify the ‘optimal’ diagnostic accuracy in which a test result is obtained for every woman. This analysis excluded women who were difficult to diagnose, so it may overestimate diagnostic accuracy. Forest plots for FNR and FPR are shown in Figure 4.

FIGURE 4.

Forest plots of (a) FNR and (b) FPR excluding women with inconclusive test results.

Excluding women with inconclusive test results has no meaningful impact on false-negative results (as those women are always assumed to have a positive result). It does, however, considerably reduce the FPR. The FPR, at 1.2%, is low but still considerably higher than the FNR. This suggests that NIPT is more accurate in women with a RhD-positive fetus than in those with a RhD-negative fetus. There was some evidence of heterogeneity across studies. The I²-statistic for heterogeneity was 75% for the FNR and 99% for the FPR. The ROC plot with bivariate and HSROC analyses is given in Appendix 6.

Bristol studies

We performed a subgroup meta-analysis of only the high-quality studies based in Bristol12,17,18 in order to assess the most likely performance of NIPT in the UK. We excluded the study by Akolekar et al. 19 (based in London but with NIPT run in Bristol) from this analysis on the grounds of it having a high risk of bias, as it was not primarily intended to assess NIPT screening and because of the limited applicability of recruited participants. A higher proportion of people with African ethnicity (19.3%) in this study means that it may not be representative of the general population of pregnant women in the UK.

In this analysis, women with an inconclusive test result were treated as having a positive result, in line with the practice in the studies.

As observed in Table 4 and Figure 5, the three Bristol studies have a slightly lower FNR and a higher FPR than other studies. This suggests that the Bristol high-throughput NIPT approach in which the MDx Bio Robot machine is used may be using a different test threshold from other countries, which further minimises false-negative findings, with a consequent increase in the FPR. This may explain some of the heterogeneity observed in previous analyses.

FIGURE 5.

Forest plots of (a) FNR and (b) FPR for the Bristol studies.

If inconclusive tests results were excluded from the Bristol studies, the summary FNR was 0.263% (95% CI 0.13% to 0.56%) and the FPR was 1.474% (95% CI 0.82% to 2.63%). This confirms that most false-positive results arise from treating women with an inconclusive test result as being test positive.

Subgroup analyses

We considered the effect of the timing of NIPT on its diagnostic accuracy. Figure 6 shows the FNRs plotted by gestational age at time of high-throughput NIPT. It suggests that FNRs after 11 weeks’ gestation were consistent, irrespective of timing, but that FNRs were higher before 11 weeks’ gestation. Figure 7 shows the FPRs plotted by gestational age at time of high-throughput NIPT. There was no obvious pattern from this figure. Only one study12 examined test performance at multiple time points. Figure 8 shows the FPRs and FNRs at different times for this study. It indicates that FNRs were higher before 11 weeks’ gestation and were generally stable after 11 weeks’ gestation. We did not perform any formal statistical analyses on the timing data (such as a metaregression) because the relationship appears to be a step change in accuracy, rather than a linear trend over time. These results together suggest that NIPT is insufficiently accurate before around 11 weeks’ gestation (i.e. in first trimester) but is accurate at any time after the end of the first trimester.

FIGURE 6.

False-negative rate by gestational age at time of NIPT.

FIGURE 7.

False-positive rate by gestational age at time of NIPT.

FIGURE 8.

False-negative rate against FPR for Chitty et al. 12 study.

We also considered the impact of the timing of high-throughput NIPT on the number of inconclusive test results (Figure 9). Despite the data from Wikman et al. 23 being heterogeneous, there appears to be a trend that the percentage of inconclusive results for this test reduces as the gestational age increases from 11 weeks. This is most obvious in the Chitty et al. 12 study, which reported numbers of inconclusive tests at different times.

FIGURE 9.

Inconclusive results by test timing.

We were unable to conduct any subgroup analysis based on ethnicity, as the relevant data were not reported in any publication. As all studies were conducted in Europe, numbers of participants of non-white ethnicity were few. Any diagnostic analysis of non-white ethnicities may therefore not give reliable results.

Because each country used a different machine to perform NIPT, a subgroup analysis by type of NIPT method was not feasible, as it would be confounded by study location. We have considered a subgroup analysis including the Bristol-based studies only, as reported in Meta-analyses of diagnostic accuracy.

Results: assessment of clinical effectiveness

Results: assessment of implementation

Diagnostic accuracy

Eight studies12,17–23 were included in the diagnostic review of high-throughput NIPT. There were three studies based in Bristol (UK). 12,17,18 The majority of included studies were judged as having a low risk of bias.

Meta-analyses found that high-throughput NIPT had very good diagnostic accuracy. In the primary analyses, in which women with inconclusive test results were treated as if positive, the summary FNR (women at risk of sensitisation) was 0.34% (95% CI 0.15% to 0.76%) and the FPR (women needlessly receiving anti-D) was 3.86% (95% CI 2.54% to 5.82%).

The three high-quality studies performed at Bristol,12,17,18 which were most representative of UK practice, had a lower FNR of 0.21% (95% CI 0.09% to 0.48%), with a consequently higher FPR of 5.73% (95% CI 4.58% to 7.16%). This difference may be partly because the NIPT used in Bristol had a different test threshold to other countries to further reduce false-negative results.

The FPR found is mostly a consequence of treating women who have an inconclusive test result (approximately 7% of non-invasive prenatal tests in the UK) as if they had a positive test. Excluding these women from analysis gave a lower FPR of 1.26% (95% CI 0.87% to 1.83%). It may therefore be possible to reduce the FPR by further targeted testing of women with an initially inconclusive result.

The diagnostic accuracy performance of high-throughput NIPT varied by gestational age. The data suggest that high-throughput NIPT is insufficiently accurate before around 11 weeks’ gestation (i.e. in first trimester) but is accurate at any time after the end of the first trimester. One study12 also suggested that the number of inconclusive results may decline over time. Hence, NIPT cannot be recommended before the second trimester and may be best performed later in the second trimester.

Clinical effectiveness

Seven studies18,20,22,24–27 were included in the clinical effectiveness review. Only two studies had a control group. All studies were judged as having a high risk of bias. As all except one were conducted in non-UK countries, the generalisability of their findings to the UK setting is limited because of variations in national guidelines and health policies between countries (e.g. prescription of RAADP). One large prospective cohort study26 reported that use of high-throughput NIPT for targeted antenatal anti-D prophylaxis was associated with a significant risk reduction in sensitisation (adjusted odds ratio 0.41, 95% CI 0.22 to 0.87) compared with historical controls (routine management, postpartum anti-D only).

Uptake rates of NIPT were reported in seven studies, ranging from 70% in a pilot study conducted in England to > 95% in an established national programme in Denmark. Uptake rates of RAADP in women who accepted NIPT and received a positive result were moderate to high, ranging from 86% to 96.1% (four studies). Uptake rates of routine postnatal anti-D prophylaxis in women who accepted NIPT and received a positive result were reported in three studies and were generally high, ranging from 92% to 99.7%.

Three non-comparative studies evaluated changes in anti-D use following the implementation of NIPT. All found that the use of NIPT reduced the total use of anti-D immunoglobulin doses, which fell by 29% in one UK study18, because around 35% of RhD-negative women avoided receiving anti-D unnecessarily.

As the quality of the clinical effectiveness evidence was limited, we performed a simulation study, based on the findings of our reviews, to assess the probable clinical consequences of implementing NIPT. Its results were broadly consistent with the review evidence. It suggested that NIPT, when compared with offering anti-D to all RhD-negative women, would substantially reduce the use for anti-D from 99% of women to 65.9%. The number of women receiving anti-D unnecessarily would fall from 38.9% to 5.7%. The number missing out on potentially beneficial anti-D (because of a false-negative test result or non-compliance) depends on the compliance rate but could increase from 0.6% to between 1.2% and 3.1%.

The impact of NIPT on sensitisation rates (compared with universal anti-D use) also depends on compliance. Sensitisation rates may increase by 3–15 sensitisations per 100,000 women if postpartum cord blood testing is continued, or 13–28 per 100,000 women if cord blood testing is withdrawn and postpartum anti-D given on the basis of the NIPT result. Ensuring that women who do not receive NIPT are still offered, and receive, antenatal anti-D will minimise the number of additional sensitisations.

Implementation

Twelve studies were included in the review of implementation. Most of the included studies were large cohort studies reporting implementation data along with diagnostic accuracy data, although one study was a UK-based survey. As most studies were conducted in non-UK countries, the generalisability of their findings to the UK settings is limited because of variations in national guidelines and health policies between countries. All the large cohort studies suggested that high-throughput RhD genotyping of fetuses in all RhD-negative women was feasible and should be recommended. A number of studies reported issues of implementation such as those relating to programme anti-D prophylaxis compliance. Some studies emphasised the importance of short transport times of samples and the need for good management of transporting samples. Some studies also identified the need for greater knowledge of NIPT among physicians, midwives and pregnant women.

Conclusions

High-throughput NIPT for fetal RhD status is an accurate diagnostic test, if performed after 11 weeks’ gestation. It has a FNR (women remain at risk of sensitisation) of around 0.2% and a FPR (women receive unnecessary anti-D) of around 5.7%. The test gives an inconclusive result in around 7% of women in the UK. Owing to limited evidence, the accuracy of NIPT in non-white women and multiple pregnancies is unclear. Treating inconclusive tests as if they were positive is the cause of most false-positive results. Giving antenatal anti-D immunoglobulin on the basis of NIPT, rather than to all RhD-negative women, will reduce the use of anti-D and largely eliminate unnecessary use of anti-D in women who do not need it because they have a RhD-negative fetus. Some women will, however, continue to receive anti-D unnecessarily because of an inconclusive test result.

Although the evidence was limited, it appears that using NIPT will lead, at worst, to only a small increase in the number of sensitisations compared with universal use of anti-D. The simulation suggested that achieving high compliance with both NIPT and antenatal anti-D (particularly in women who do not receive NIPT) is important in order to achieve good clinical effectiveness and to reduce the sensitisation rate. It may be clinically reasonable to withdraw postpartum cord blood testing and base postpartum anti-D administration on the results of NIPT. All large implementation studies suggested that high-throughput NIPT in all RhD-negative women was feasible and should be recommended. Key issues of implementation include ensuring anti-D prophylaxis compliance, effective management of transporting samples and greater knowledge of NIPT among physicians, midwives and pregnant women.

Searches

In addition to the searches conducted for the review of clinical evidence (see Chapter 3), the following databases were searched up to December 2015 for cost-effectiveness evidence: NHS EED, EconLit and IDEAS database via Research Papers in Economics (RePec). The bibliographies of relevant studies were also searched. Citations of identified studies were searched for any relevant publications published after the initial search.

Selection criteria

A broad range of studies was considered in the review, including economic evaluations conducted alongside trials, modelling studies and analyses of administrative databases. Only full economic evaluations that compared two or more options and considered both costs and consequences (i.e. cost-minimisation, cost-effectiveness, cost–utility and cost–benefit analyses) were included in the review.

Study selection

Relevant studies were then selected in two stages. Titles and abstracts identified by the search strategy were examined independently by two researchers (PS and SG) and screened for possible inclusion. Disagreements were resolved by discussion. Full texts of the potentially relevant studies were obtained. Two researchers (PS and SG) examined these independently for inclusion or exclusion, and disagreements were resolved by discussion.

Data extraction

One reviewer (PS) independently extracted details from full-text studies on objectives, setting, population, comparators, analytical approach, data on costs and outcomes (short- and long-term) and main results/conclusions. Another reviewer (SG) checked extracted data and disagreements were resolved by discussion.

Critical appraisal

A quality appraisal was carried out using the checklist of Drummond and Jefferson. 67 This checklist evaluates the extent to which each review result provides detail on different aspects, such as study design, data collected and their use in the economic evaluation and analysis and interpretation of results. One reviewer (PS) independently assessed the quality of all included studies according to all these domains. The quality assessment was checked by another reviewer (SG). Any disagreements were resolved by consensus.

Quantity of research available

Number and type of studies included

The initial search of economic databases identified a total of 31 references. After the initial screening of titles and abstracts, 10 were considered to be potentially relevant and were ordered for full-text paper screening. Of those, seven met the selection criteria and were included in the review. 58,68–73 A flow diagram of the selection process is reported in Figure 10.

FIGURE 10.

Assessment of cost-effectiveness: summary of study selection and exclusion.

Characteristics of included studies

The characteristics of the seven studies are summarised in Table 15. The large majority of studies specified the target population as being unsensitised RhD-negative pregnant women or RhD-negative pregnant women not known to be sensitised to the RhD antigen. Macher et al. 70 and Hawk et al. 72 stated that their analysis considered RhD-negative pregnant women but they were not clear about women’s sensitisation status at study entry. Only two studies68,73 explicitly stated that a high-throughput NIPT method was being used for the comparative assessment, although for the other studies this was considered implicit, as the test diagnostic performance was considered similar to the high-throughput studies. One study71 explicitly focused on providing NIPT to all RhD-negative women, as the test for sensitisation was conducted only if the NIPT result was positive.

Most studies58,68,71–73 evaluated the cost-effectiveness of introducing NIPT in the management pathway of RhD-negative pregnant women compared with alternative strategies. These studies explored a range of alternative strategies to prevent sensitisation. Except for Szczepura et al. 68 and Macher et al. ,70 two strategies were common across the studies: (non-targeted) RAADP at around 28–30 weeks to every (unsensitised) RhD-negative pregnant women; and use of NIPT for fetal RhD typing with prophylaxis guided by test results (targeted RAADP) for RhD-negative pregnant women. Duplantie et al. 71 also explored the immunological determination of the father’s RhD type to target RAADP. Most studies considered the introduction of NIPT at a single time point, usually at first routine antenatal care appointment occurring between 8 and 12 weeks’ gestation. Benachi et al. 69 compared alternative timings of NIPT by considering the cost consequences of performing NIPT during the first and the third gestation trimesters. With the exception of the Duplantie et al. 71 study, for which insufficient information is provided, all cost-effectiveness studies evaluated the consequences of introducing NIPT in terms of avoiding RAADP but also in terms of the impact it had on postpartum treatment.

Three studies58,68,73 aimed to evaluate the short-term costs and consequences of sensitisation in RhD-negative women. Duplantie et al. 71 and Hawk et al. ,72 however, estimated long-term outcomes relating to morbidity and mortality attributable to haemolytic disease of the fetus and/or newborn infant. Furthermore, two studies58,71 explicitly considered in their analysis women’s first and subsequent pregnancies, presenting cost-effectiveness results for each scenario.

Benachi et al. 69 and Macher et al. 70 are cost-minimisation studies, as no health outcomes were considered, restricting their analysis to an evaluation of the impact of the test on the costs of managing the target population. A variety of cost components were considered across these two studies, such as anti-D immunoglobulin, genotyping, antibody testing.

The cost-effectiveness studies evaluated different strategies in different health systems, including England and Wales, Canada, Sweden and the USA. Except for Sweden, where only postpartum administration of anti-D (conditional on having a RhD-positive baby) is recommended, current guidance for the prevention of sensitisation in these countries is routine prophylactic administration of anti-D, with further prophylactic doses for potentially sensitising events and post partum. The two cost-minimisation studies69,70 evaluated the cost implications of introducing NIPT in the French and Spanish (namely the Andalusia region) health-care settings. Current guidance on the prevention of sensitisation in these countries was not clearly stated. Macher et al. 70 focused mainly on addressing questions relating to the accuracy and implementation of different NIPT methodologies into current clinical practice in Spain.

Quality of included studies

A summary of the results of the quality appraisal of the seven included studies is provided in Table 16.

Results of included studies

In terms of conclusions, conflicting results were reported across the existing economic studies. Three studies68,71,72 reported NIPT fetal RhD genotyping not to be cost-effective or of no economic benefit. Hawk et al. 72 and Szczepura et al. 68 reported that the main factor driving these factors was the cost of the test itself (i.e. the clinical and economic benefits were not sufficient to offset the additional costs of the test). Szczepura et al. 68 also stated that the implementation of NIPT in the clinical pathway of the RhD-negative pregnant woman was not expected to produce important clinical benefits. Supporting this was an estimation of the potential rise in the number of sensitised women if NIPT sensitivity fell below 99.9%.

Two studies58,69 reported that NIPT is cost saving compared with no RAADP (i.e. compared with postpartum anti-D only). Only one study73 found NIPT for targeted RAADP to be cost saving compared with non-targeted RAADP, which also estimated no increase in the risk of sensitisation if NIPT were to be used. Duplantie et al. 71 found that targeting of RAADP based on the immunological RhD typing of the father is cost-effective compared with the use of NIPT.

Overall, the quality of the included studies’ findings is uncertain because of a lack of reporting of the validity of the diagnostic accuracy outcomes used. Furthermore, although SA exercises were generally done over some key parameters, the degree of uncertainty in the cost-effectiveness estimates is generally difficult to establish.

Relevance to the NHS and current decision problem

One of the key aspects of this review is to address how relevant study assumptions and findings are to the UK. None of the study approaches and findings reviewed was considered to be generalisable to the decision problem as set out in the NICE scope for the current diagnostic assessment. The scope for this decision problem includes an evaluation of the introduction of NIPT at different gestation points, the impact of the test result on the administration of anti-D immunoglobulin treatment routinely and post partum, and the impact of sensitisation on infant health and/or on subsequent pregnancies. Only one68 of the seven economic studies reviewed directly relates to the UK. This study, however, did not explicitly explore how the introduction of NIPT could impact on costs relating to potentially sensitising events. In addition, it assumed that postpartum testing and treatment would be unaffected by NIPT results. Furthermore, no assessment of the timing of NIPT or any consideration of the impact on subsequent pregnancies was undertaken. Therefore, limited UK-specific information exists that explicitly relates to the decision problem as specified in the scope for this diagnostic appraisal. Although some studies are from Canada and the USA, countries in which similar guidance to that in the UK exists on the prevention of sensitisation, relevance to the UK and generalisability of findings can be questioned, as there are crucial health-care system differences and differences in how anti-D immunoglobulin policies have been implemented over recent decades.

Overall aims and objectives of the independent economic assessment

The cost-effectiveness assessment of the use of high-throughput NIPT to identify fetal rhesus D status had the following overall main objectives:

• To produce a de novo cost-effectiveness model assessing the cost-effectiveness of high-throughput NIPT to identify fetal RhD status in RhD-negative women not known to be sensitised to the RhD antigen.

• To assess the impact of alternative scenarios related to the timing of the test and the impact of the test on the use of antenatal anti-D immunoglobulin prophylaxis for sensitising events and postdelivery testing and postpartum anti-D immunoglobulin administration.

Intervention and comparator pathways

Current NICE clinical guidance on antenatal care7 recommends that women be offered testing for blood group and rhesus D status in early pregnancy. All pregnant women identified as RhD-negative would be tested for the presence of RhD antibodies. Women identified as RhD-negative and found not to have RhD antibodies are not yet sensitised and form the population for this appraisal. In these women, anti-D immunoglobulin is recommended, both as prophylaxis and following potential sensitising events, to prevent sensitisation occurring. 2

Routine antenatal anti-D prophylaxis is recommended to be given as two doses at weeks 28 and 34 of pregnancy, or as a single dose between 28 and 30 weeks. Supplementary doses of anti-D immunoglobulin should also be administered prophylactically after a potentially sensitising event. 2,8 Potentially sensitising events include those that may lead to FMH, such as medical interventions (e.g. chorionic villus sampling, amniocentesis or external cephalic version), terminations, late miscarriages, antepartum haemorrhage and abdominal trauma. Following a potential sensitisation event, the recommended minimum dosage of anti-D immunoglobulin increases with gestational age (i.e. a higher dose for > 20 weeks’ gestation), and FMH testing is used to inform the actual dose after 20 weeks’ gestation.

Following birth, RhD typing should be performed on a cord blood sample to determine the RhD status of the baby. If the baby is confirmed to be RhD positive, it is recommended that previously non-sensitised RhD-negative pregnant women receive anti-D immunoglobulin within 72 hours following delivery, with the actual dose guided by FMH results. This represents the pathway and current clinical practice of the management of RhD-negative pregnant women not known to be sensitised.

The intervention technology of this assessment is high-throughput NIPT for fetal rhesus D status. By analysing cell-free fetal DNA in the plasma of RhD-negative pregnant women, high-throughput NIPT is able to predict fetal RhD genotype. High-throughput NIPT for fetal RhD status may enable prophylactic anti-D immunoglobulin to be withheld from women who are RhD-negative and carrying a RhD-negative fetus. These women could avoid unnecessary treatment with anti-D immunoglobulin, along with the potential risk associated with blood products. The results of NIPT could impact on the care pathway in the following ways:

1. For women in whom the high-throughput NIPT indicates the presence of a RhD-negative fetus:

   • avoidance of RAADP

   • avoidance of prophylactic anti-D immunoglobulin and FMH tests following potentially sensitising events

   • avoidance of cord serology testing, fetal maternal haemorrhage test and administration of anti-D immunoglobulin following delivery.

2. For women in whom the high-throughput NIPT indicates the presence of a RhD-positive fetus:

   • avoidance of cord serology testing in favour of routine FMH testing and postpartum anti-D immunoglobulin following delivery.

Modelling methodology and scope

A decision-analytic model using a decision tree structure simulates the experience of a hypothetical cohort of RhD-negative pregnant women not known to be sensitised to the RhD antigen, with and without the introduction of high-throughput NIPT for fetal RhD status. A pregnant woman enters the model after having been identified as RhD-negative and not yet sensitised based on the results of tests from bloods drawn either at first contact with the doctor or midwife (the date at which pregnancy is reported or established) or at the booking appointment (8–12 weeks’ gestation). All further contacts between the woman and the health service are informed by the recorded test results. At the routine 16-week visit the woman is informed about her RhD status, whether or not she is sensitised and how these results impact on further management. If the woman contacts the health service following any potentially sensitising event she may be offered anti-D immunoglobulin and, if after 20 weeks’ gestation, a FMH test. Women provided with RAADP receive it at either or both of the routine visits at 28 and 34 weeks’ gestation. At delivery, a sample of cord blood may be taken and the baby’s RhD status established to guide the use of FMH tests and the administration of postpartum anti-D immunoglobulin.

All high-throughput NIPT are assumed to be performed early enough to determine the use of RAADP at 28 weeks’ gestation. Figure 11 shows the current schedule of routine antenatal care appointments and the potential placement of NIPT.

FIGURE 11.

Excerpt from NICE schedule of appointments in routine antenatal care. © NICE 2016. All rights reserved. Schedule of Appointments in Routine Care. 79 Available from: http://beta.pathways.nice.org.uk/pathways/antenatal-care/schedule-of-appointments-in-routine-antenatal-care. NICE guidance is prepared for the National Health Service in England, and is subject to regular review and may be updated or withdrawn. NICE has not checked the use of its content in this report to confirm that it accurately reflects the NICE publication from which it is taken.

In addition to the first contact/8–12 weeks’ gestation booking appointment, the points of routine contact at which blood could be drawn for NIPT are the 16-week visit and 18- to 20-week scan (at which outstanding routine screening tests are offered). Other opportunities may include attendance to receive the whooping cough vaccine and the routine 25 weeks’ gestation visit for first pregnancy only. Once the results of any high-throughput NIPT are known, they will be communicated to the woman and recorded with the potential to inform all further contacts and decisions regarding testing and treatment. We assume that RAADP and management for potentially sensitising events would be subsequently offered only to women in whom the test result indicates that their fetus is RhD positive and in whom the test result is inconclusive. For women in whom the high-throughput NIPT result is inconclusive, the existing care pathway will remain unchanged and they would receive the same management as women for whom the results of NIPT indicate a RhD-positive baby. We assume that provision of NIPT can be incorporated into routine antenatal care without requiring additional visits (to undertake the test or to communicate the results of test). Similarly, in the base case we do not model additional resources within existing antenatal care appointments to draw blood.

As previously mentioned, the model may be separated into two main elements: (1) an identification part reflecting the diagnostic performance and costs of the alternative identification strategies and (2) a treatment part evaluating the subsequent costs and outcomes (expressed in QALYs) of alternative care pathways. The main aim of the first model element is to divide the cohort according to fetal RhD status and treatment administered (i.e. routine anti-D immunoglobulin, FMH tests and anti-D immunoglobulin for potentially sensitising events, cord serology, FMH tests and postnatal anti-D immunoglobulin). This determines when receipt of anti-D immunoglobulin is appropriate (true positive in terms of NIPT result and/or postnatal cord serology and inconclusive result but pregnant with RhD-positive fetus), when avoidance of anti-D immunoglobulin is appropriate (true negative in terms of NIPT result), when anti-D immunoglobulin is unnecessary (false positive or inconclusive in terms of NIPT result and carrying a RhD-negative fetus) and when avoidance of anti-D immunoglobulin is potentially harmful (false negative in terms of NIPT result). Aspects such as the diagnostic test performance (including inconclusive results and results at different gestation timings), compliance with high-throughput NIPT and anti-D immunoglobulin treatment and the effectiveness of anti-D immunoglobulin all inform the estimation of the probability of sensitisation for each of these groups. The second model element (i.e. the treatment part) considers the short- and long-term consequences of sensitisations (i.e. fetal or neonatal death, minor and major development problems of the child) for the first, second, third and subsequent pregnancies. Costs and utilities are then evaluated for the different components and for each of the alternative pathways.

Four alternative ways in which the use of high-throughput NIPT may impact on the existing postpartum care pathway were considered.

1. NIPT postpartum scenario 1 (PP1): postpartum cord blood typing and FMH testing would continue to be performed, as per current guidelines, in all women regardless of the fetal RhD status identified through high-throughput NIPT.

2. NIPT postpartum scenario 2 (PP2): postpartum cord blood typing, FMH testing (and by implication anti-D immunoglobulin) would be withheld if high-throughput NIPT of fetal RhD status identifies a RhD-negative fetus but would continue to be performed if high-throughput NIPT was inconclusive or had identified a RhD-positive fetus.

3. NIPT postpartum scenario 3 (PP3): postpartum cord blood typing would be performed if high-throughput NIPT of fetal RhD status identifies a RhD-negative fetus. FMH testing and postdelivery anti-D immunoglobulin would be administered if high-throughput NIPT was inconclusive or identifies a RhD-positive fetus.

4. NIPT postpartum scenario 4 (PP4): postpartum cord blood typing not performed in any women. FMH testing and postdelivery anti-D immunoglobulin administered if high-throughput NIPT was inconclusive or had identified a RhD-positive fetus.

The impact that postdelivery testing has on the cost-effectiveness results is explored using separate scenarios in the model. In reality, these four separate scenarios actually represent separate and distinct testing and management strategies and, hence, could also be considered to represent relevant strategies that should be directly compared in the cost-effectiveness assessment.

The cost-effectiveness of high-throughput NIPT is determined by comparing with current practice (i.e. no use of high-throughput NIPT), which comprises (1) RAADP and supplementary anti-D immunoglobulin (as required based on potentially sensitising events) offered to all RhD-negative pregnant women and (2) further postpartum anti-D immunoglobulin offered to all RhD-negative women whose baby’s RhD status is confirmed to be positive after cord blood typing.

A schematic representation of the model is provided in Figure 12a and b. Note that this figure does not provide a comprehensive representation of all components being considered in each alternative strategy, including the postpartum scenarios. The four postpartum scenarios for how the introduction of NIPT could impact on the use of cord serology, fetal maternal haemorrhage tests and anti-D immunoglobulin use following delivery are detailed in Table 17.

FIGURE 12.

Decision-analytic model schematic representation of RhD-negative pregnant women pathways. (a) No high-throughput NIPT and RAADP (current practice, no test and RAADP); and (b) high-throughput NIPT and targeted RAADP.

The model considers the total number of children who would be born to each RhD-negative woman in order to capture the effect of any sensitisation on all subsequent pregnancies based on national fertility rates. We assume that the consequences of sensitisation do not affect the pregnancy in which it occurs (with respect to treatments and tests administered, management and health outcomes of the resultant RhD-positive baby) but rather only subsequent pregnancies. Under current practice, a woman who is sensitised during pregnancy will be identified at the start of her next pregnancy, when she will be tested for antibodies to the RhD antigen. As a consequence of having been sensitised, the woman will be subject to more intense antenatal care in all subsequent pregnancies (see Cost of management of sensitisation) and any further RhD-positive fetuses are at risk of adverse health consequences (see Cost of high-throughput non-invasive prenatal testing and Cost of management of sensitisation). First and subsequent pregnancies together with long-term consequences of sensitisations, in terms of costs and utilities, are evaluated with a yearly cycle and a lifetime horizon. This lifetime horizon includes the full life expectancy of any fetus lost as a consequence of sensitisation. The decision model follows a NHS perspective, and all costs and effects are discounted at a rate of 3.5% each year. The main outcomes of interest within the model are the total lifetime costs and total lifetime QALYs for each of the alternative pathways. Other outcomes recorded in the model include:

• number of sensitisations and the associated costs

• number of affected fetuses following sensitisation

• number of fetuses lost and associated QALY loss

• cost per life-year gained.

What alternative scenarios have been modelled?

In addition to the five alternative pathways compared in the base-case analysis, we compare the inclusion of the high-throughput NIPT at specific gestational ages. These are determined based on available data that show how the diagnostic accuracy of the test varies with gestational age. The timing of the test is important not only in terms of diagnostic performance but also in terms of the cost of managing potentially sensitising events. Although the majority of these are thought to occur in the third trimester (weeks 29–40), any that occur prior to the use of the high-throughput NIPT will incur the cost of anti-D immunoglobulin for all women regardless of fetal RhD status. We further explore the impact of variation in compliance with anti-D immunoglobulin.

Under current guidance, more recent data on RAADP coverage indicate an uptake of approximately 99.0% in women who are still pregnant at 28 weeks and where the father is not established as RhD negative. 8 In addition, postpartum anti-D immunoglobulin current uptake is believed to be also close to 100%. 8 However, data relating to the uptake of routine and postpartum anti-D immunoglobulin in the presence of fetal RhD status identification are scarce (see Chapter 3). Finally, we consider alternative scenarios for the proportion of women in whom the NIPT result is inconclusive. The rate of inconclusive results may reach > 14% and these are typically managed as RhD-positive results (see Chapter 3, Results: assessment of diagnostic accuracy). However, women in whom the high-throughput NIPT result is inconclusive are likely to differ systematically from those in whom the test result is positive, with ethnicity being the most important factor.

Target population

The number of pregnancies in RhD-negative women in England was estimated to be 99,225 per year. This represents a cross-section of all pregnancies and the proportions of first, second, third and subsequent pregnancies are used to characterise the total fertility rate of a typical RhD-negative woman. This estimate was based on a birth rate of 12.2 per 1000 women per year80 and assumes that 15% of the population is RhD negative. 3

Proportion of RhD-positive babies born to RhD-negative women

The RhD status of babies does not depend solely on the zygosity of the mother but also of the father. The RhD-negative gene is recessive. Following Mendel’s law on inheritance,81 if the father is homozygous (i.e. he has two RhD-positive genes) all of his children will be RhD positive, but if he is heterozygous (i.e. he has one RhD-positive gene and one RhD-negative gene) his children will have a 50% chance of being RhD-negative. Therefore, as in the NICE TA156,62 the model assumes that the proportion of RhD-positive babies born to RhD-negative women is a function of (1) the proportion of RhD-positive men (assumed to be identical to the proportion of RhD-positive women, thus, the complement of the proportion of RhD-negative women), (2) the proportion of heterozygous fathers and (3) the proportion of heterozygous fathers having RhD-positive babies. Although the probability of having a RhD-positive baby in subsequent pregnancies can be estimated conditional on knowledge of the RhD status of the first baby, we do not split the cohort in this way. The use of high-throughput NIPT among RhD-negative women not yet sensitised to the RhD antigen is not expected to be determined on the basis of the RhD status of previous offspring. It is therefore unnecessary to split the cohort according to this characteristic and so we apply the same overall rate of RhD-positive babies across all pregnancies. This equates to approximately 62%, as described in Table 18.

Diagnostic accuracy of non-invasive prenatal testing

Data on the diagnostic accuracy of high-throughput NIPT are based on the meta-analyses summarised in Chapter 3, Results: assessment of diagnostic accuracy. The base case utilises the pooled results for the subgroup of UK (Bristol-based) studies in which inconclusive results are considered as test positive. These were considered to be the most relevant to the English setting. Sensitivity, specificity (with 95% CIs) and the correlation between these two test accuracy dimensions (on the log-odds scale) were used to inform log-normal distributions within the decision model. Note that the correlation estimate for the UK (Bristol) approach was based on only three studies (Table 19). SAs were performed based on pooled results from all studies and when inconclusive results were not considered as test positive. In general, high-throughput NIPT accuracy is consistently high across the different approaches to the diagnostic meta-analysis. The subgroup of only UK studies shows a lower FNR and a slightly higher FPR than other scenarios.

Only one study12 extensively examined the test performance at multiple gestation time points. In scenario analysis these results were used to assess the cost and consequences of introducing high-throughput NIPT at different gestation ages (Table 20). We considered that high-throughput NIPT might be targeted at more specific gestational ages from 11 weeks’ gestation and not after 24 weeks’ gestation, and thus, in the model, we compared the diagnostic accuracy reported for 11–13 weeks, 14–17 weeks and 18–23 weeks (see Sensitivity analyses results).

Non-invasive prenatal testing inconclusive results

In the UK studies that inform the base case for the decision model, the pooled proportion of inconclusive NIPT results was 6.7%. Across all diagnostic studies that report the number of inconclusive results this proportion is lower at 4.0%. The results of the diagnostic accuracy studies suggest that the probability of a RhD-positive baby is higher among women in whom the high-throughput NIPT is inconclusive than in all RhD-negative women (see Chapter 3, Inconclusive test results). In Proportion of RhD-positive babies born to RhD-negative women, it was estimated that the probability of RhD-negative women having RhD-positive babies in the first and subsequent pregnancies was 61.6%. In the presence of high-throughput NIPT inconclusive results it is estimated that this probability is 70.1%, irrespective of the pregnancy. This probability is slightly reduced (70.7%) if only UK studies are considered. These last two probabilities are used to estimate the positive predictive value of NIPT, and in SA around the postpartum management of women with inconclusive NIPT results (SA8).

Effectiveness of anti-D immunoglobulin

The introduction of high-throughput NIPT into the care pathway will be used to determine the level of use of anti-D immunoglobulin. Anti-D immunoglobulin affects the rate of sensitisation in women carrying RhD-positive fetuses and carries a potential risk of adverse effects as it is derived from blood products. The costs and consequences of the introduction of high-throughput NIPT are therefore determined by:

• the efficacy of anti-D immunoglobulin in preventing sensitisation, as this determines the health and cost implications for women from whom this is incorrectly withheld because of a false-negative high-throughput NIPT result

• the costs and adverse effects associated with administration of anti-D immunoglobulin.

The clinical effectiveness and cost-effectiveness of RAADP in RhD-negative women has been previously established in NICE TA4184 and most recently in NICE TA156. 62 No new systematic reviews of RAADP with studies not considered in TA156 were identified. We maintain consistency between the NICE TA process and the diagnostics assessment of high-throughput NIPT for fetal rhesus D status by utilising the RAADP efficacy estimated based on the same set of clinical effectiveness studies that were considered to be most representative of the UK within NICE TA156. The parameter estimates applied in our base-case analyses are based on the synthesis presented within NICE TA156. The impact of using alternative estimates reported in a related publication by Turner et al. ,63 published after NICE TA156 had been completed, is explored within a separate SA. Evidence for the clinical effectiveness of the postpartum use of anti-D immunoglobulin was sourced from a previous Cochrane review. 65 The clinical effectiveness estimates of RAADP and postpartum use of anti-D immunoglobulin reported across these separate sources are reported in Table 21.

Potentially sensitising events

Following potentially sensitising events, the administration and dosage of anti-D immunoglobulin is conditional to the pregnancy stage in which the event occurs. Current guidance7 recommends that only in extraordinary sensitising events (such as ectopic pregnancy, molar pregnancy or therapeutic termination of pregnancy) should anti-D immunoglobulin be administered at < 12 weeks’ gestation. A minimum dose of 250 IU of anti-D immunoglobulin within 72 hours of the event is recommended to be administered if it occurs between 12 and 20 weeks’ gestation. For potentially sensitising events after 20 weeks’ gestation, a minimum anti-D immunoglobulin dose of 500 IU should be administered within 72 hours, with additional doses as guided by a test for FMH.

Evidence on the reported number of potentially sensitising events was found in the recent audit on anti-D immunoglobulin prophylaxis. 8 The probability of women having at least one (reported) potentially sensitising event was estimated to be 15.5%. Of these, 69.3% women were estimated to have had a FMH test and 95.8% women were estimated to have been treated with anti-D immunoglobulin following the event. It was estimated that approximately 80% of these events happened after 20 weeks’ gestation. We assume that these 80% of sensitising events are treated with the minimum required dose of 500 IU of anti-D immunoglobulin. For the remaining 20% of events (pre 20 weeks’ gestation events), we assumed that women received the minimum required dose of 250 IU of anti-D immunoglobulin.

The audit on anti-D immunoglobulin prophylaxis8 also provided information on the type of potentially sensitising event. It was estimated that the probability of women having a miscarriage (including stillbirth and intrauterine death) was 4.7%. We assumed that these fetal deaths were not a consequence of sensitisation and they are incorporated in the model only to adjust the amount of postpartum health resource consumption following delivery.

In contrast to women in whom the high-throughput NIPT result indicates that their fetus is RhD positive, women in whom the test shows that the fetus is RhD negative will not be offered prophylactic anti-D immunoglobulin treatment and will not be subject to FMH testing. This is an issue particularly for the false negatives (RhD-negative women with a RhD-positive fetus but for whom the test result was negative), as these women will, at most, receive only postpartum treatment. For women with false-negative NIPT results who receive only postpartum anti-D immunoglobulin, the model assumes a rate of sensitisation of 0.95%. This is likely to be an underestimate, as it includes receipt of anti-D immunoglobulin for potentially sensitising events. However, the only other estimate for the rate of sensitisation without RAADP is that based on no anti-D immunoglobulin at all, including no postpartum treatment (10.7%), which is likely to be a large overestimate, as the majority of events occur at birth (see Table 20). The true rate of sensitisation is likely to lie between 0.95% and 10.7%, but it appears reasonable that this rate will be closer to 0.95%.

Compliance with routine antenatal anti-D prophylaxis and postpartum anti-D immunoglobulin

The National Comparative Audit of Blood Transfusion 2013 on Anti-D Immunoglobulin Prophylaxis8 reported that, out of all eligible women, 99% received at least one RAADP injection. Full compliance (i.e. correct dose at the correct time) was found to be higher in the single-dose regime (90%) than in the two-dose regime (59%). In addition, the audit shows that a very high proportion of eligible women (98.4%) received postpartum anti-D immunoglobulin prophylaxis. Finally, for documented potentially sensitising events, it showed that approximately 96% of eligible women having these events received anti-D immunoglobulin.

Following the recent audit findings, within the de novo economic model it has been assumed that compliance with RAADP is 99.0%. This value was assumed for the base case and subject to scenario analysis, assuming a rate of 87.5% (i.e. the proportion receiving the correct dose at the correct time). Evidence from the audit points to higher compliance with the single-dose regimen than with the two-dose regimen for a number of reasons (e.g. cost, manufacturer supply, etc.) and there is a move towards the use of the single dose, over the two dose, with its market share reaching approximately 93%. 8 Thus, we did not adjust the compliance rate across RAADP regimen. In the model it has been also assumed that postpartum anti-D immunoglobulin compliance rate is 98.4%, again following evidence from the recent audit. 8 This value was subject to scenario analysis by assuming a rate of 91.6% (i.e. the proportion receiving the correct dose at the correct time).

Compliance with non-invasive prenatal testing given routine antenatal anti-D prophylaxis and postpartum anti-D immunoglobulin

The evidence for compliance with high-throughput NIPT is scarce, particularly in health systems in which the test is introduced after RAADP guidance is in place (see Chapter 3, Results: assessment of clinical effectiveness). In the absence of such evidence, and based on the already high rates of compliance assumed for current practice (99.0% for RAADP and 98.4% of women received at least one dose of anti-D immunoglobulin at RAADP and post-partum, respectively), we subsequently assume that the use of high-throughput NIPT has no additional impact on compliance. Therefore, it has been assumed that RAADP and postpartum anti-D immunoglobulin compliance is 99.0% and 98.4%, respectively, the same as in the no high-throughput NIPT scenarios.

Sensitisation outcomes

As for the independent economic model developed for NICE TA156 on RAADP, the current economic model considered a set of input parameters directly related to the consequences of sensitisation towards the fetus and the newborn infant, namely the implications of haemolytic disease. Three of these model input parameters were key to an appropriate representation of the possible health states, namely (1) the fetal loss rate per RhD-negative women at risk (i.e. carrying a RhD-positive baby), (2) the proportion of babies affected by haemolytic disease that resulted in minor developmental problems (these include, for instance, myopia, squint or delay in language and fine motor skills) and (3) the proportion of babies affected by haemolytic disease that resulted in major developmental problems (these include, for instance, severe permanent neurodevelopmental delay, such as cerebral palsy). Given the long-term consequences of these two last parameters, it was also important to consider the average duration of minor development problems and the life expectancy of an individual with major development problems.

A pragmatic literature search was performed to identify evidence sources for the outcomes associated with haemolytic disease of the fetus and newborn infant, in addition to the ones found in the NICE TA156. The literature review focused particularly on the anti-D immunoglobulin systematic reviews64,65,85 and the high-throughput NIPT diagnostic accuracy studies (see Chapter 3, Results: assessment of clinical effectiveness) as potential sources of data associated with the consequences of sensitisation. Apart from the study published by Finning et al. ,17 no other relevant evidence was found. Evidence from this study relating to the proportion of fetal or neonatal deaths (5%) and to the proportion of babies affected with mild/severe development problems (5%) was used to populate the model. In the absence of more recent data for parameters relating to the proportion of babies affected with minor development problems and the duration of these problems and relating to the life expectancy of people with major developmental problems, we used the same evidence as NICE TA156 with updated costs. It should be noted that owing to the small number of haemolytic disease-related events, the corresponding model estimates are subject to considerable uncertainty.

In the absence of more recent or relevant data, the health-related quality of life evidence used relating to the utilities of minor (0.85) and major (0.42) development problems and the associated uncertainty was assumed to be the same as those used in NICE TA156. 62

Cost of high-throughput non-invasive prenatal testing

For the base-case analysis the cost of high-throughput NIPT per sample was estimated to be (confidential information has been removed). This unit cost takes into account consumables, staffing, equipment, indirect costs and overhead costs. This is the company’s estimated cost of testing at full capacity, that is, dealing with at least 100,000 samples. An estimated royalty payment of (confidential information has been removed) of the test cost is assumed to be added to the unit cost of the test, bringing the base-case estimate of the cost of the test to (confidential information has been removed). The cost of high-throughput NIPT is discounted according to the pregnancy number in which it is being performed, accounting for an expected median time between pregnancies of around 3.2 years. The unit cost per sample may, however, fluctuate, as it is a function of capacity and predicted level of usage of each testing machine annually. The cost applied in the base-case analysis does not include transport costs for the delivery of blood samples for testing. Szczepura et al. 68 included a postage cost of £1.10 per sample in their analysis, although they recognised that the cost would be much reduced if the existing NHS transport service system was to be used.

Cost of routine antenatal anti-D prophylaxis and of anti-D immunoglobulin for potentially sensitising events and post partum

The cost of anti-D immunoglobulin was taken from the British National Formulary (BNF). 86 Currently, two brands [D-Gam^® (Bio Products Laboratory Ltd, Elstree, Hertfordshire, UK) and Rhophylac^®(CSL Behring LLC, Kankakee, IL, USA)] and four doses (250-, 500-, 1500- and 2500-unit vials) are available. At current prices the cost of anti-D immunoglobulin is £23.75 for D-Gam 250 IU, £33.75 for D-Gam 500 IU and £39.52 for Rhophylac. Note that current market prices of anti-D immunoglobulin may vary with supply and demand. Regional and local price negotiations exist that may make the cost anti-D immunoglobulin lower than the values indicated above.

The cost of anti-D immunoglobulin for potentially sensitising events was estimated to be £31.69, representing a weighted average of the cost of anti-D immunoglobulin 250 IU and 500 IU (minimum required) doses and their expected utilisation before and after 20 weeks’ gestation based on evidence from a recent audit. 8 The cost of RAADP was estimated to be £41.58, representing a weighted average of single-dose (1500 IU) and two-dose (2 × 500 IU) regimens and their associated market share, 92.6% versus 7.4%, respectively. 8 Similarly, the cost of anti-D immunoglobulin administered post partum was estimated to be £35.69, which reflects the expected utilisation of ‘standard’ doses: 500 IU (66.3%) and 1500 IU (33.7%). 8 Costs applied in the current economic model were discounted according to the timing of the pregnancy (the pregnancy number) in which the treatments are administered. As in NICE TA156,62 an administration cost of anti-D immunoglobulin was set to £5.

Cost of postpartum health resources used

Following birth, in current practice a cord serology test should be performed to confirm the baby’s RhD type. In addition, maternal blood samples should be tested for FMH. The costs, updated to 2015 prices, for postpartum serology (£4.18) and associated phlebotomy (£3.32) were obtained from Szczepura et al. 68 The cost of FMH testing was provided by personal communication with a NHS Blood and Transplant Manager and estimated to be £128.10 (for test by flow cytometry, NHS Blood and Transport Red Cell Immunohaemotology) (Erika Rutherford, NHS Blood and Transplant, 2016, personal communication). This cost was subject to SA, as Szczepura et al. 68 report a much lower value of £3.17 for a Kleihauer test (when updated to 2015 prices). All costs were discounted according to the timing of the pregnancy in which the resources were consumed.

Cost of management of sensitisation

The list of relevant interventions in the management of maternal and neonatal sensitisation was taken from the previous NICE TA156. 62 The proportion of individuals requiring each intervention, the estimated average number of interventions required per individual and the estimated average number of days were considered to be the same as in NICE TA15662 (Table 22). Utilisation of these resources was validated by our clinical experts, who highlighted that no significant changes in clinical practice have occurred since 2009. Similarly, the estimated annual costs for minor (£111) and major (£574) development problems was assumed to be the same as in NICE TA156 but updated to 2015 prices. Unit costs were sourced from the NHS reference costs 2014–15. 87 The total average cost per sensitisation is estimated to be £3167. Note that, owing to the multiplicity of factors affecting sensitisation and its management, the uncertainty associated with this parameter was taken from NICE TA15662 and assumed to be substantial (standard error £700).

Model parameters and main assumptions

The parameters used within the de novo economic model, and their characteristics, as described in the preceding sections, are outlined in Table 23. Costs refer to 2015 prices.

Within the model the following assumptions are consistent with NICE TA156:62

• Sensitisations do not affect the pregnancy in which they occur.

• Anti-D immunoglobulin used within one pregnancy has no effect on reducing sensitisations during the next pregnancy.

• The proportion of RhD-negative women is based on the white European population given that this group makes up > 90% of the population of England and Wales.

Furthermore, the following assumptions were made:

• The proportion of RhD-positive babies in Rh-negative women is the same irrespective of pregnancy number.

• The probability of having a RhD-positive baby in the general population of Rh-negative women (61.6%) is combined with the diagnostic accuracy results in terms of sensitivity and specificity (in which inconclusive results are treated as test positive) to determine the number of Rh-positive babies in the model.

• The probability of having a RhD-positive baby in women with inconclusive test results is based on the pooled probability in the study populations used to inform the diagnostic accuracy estimates.

• All NIPT is performed early enough to determine the use of RAADP at 28 weeks’ gestation.

• Routine and prophylactic anti-D immunoglobulin is offered only to women whose NIPT result indicates that their fetus is RhD positive or whose results are inconclusive.

• In women with an inconclusive NIPT result the existing care pathway is unchanged and they are treated the same as women who test positive in terms of RAADP, anti-D immunoglobulin and associated tests.

• Women identified to receive RAADP will receive supplementary anti-D immunoglobulin at the minimum dose required for any potentially sensitising events.

• Potentially sensitising events that involve fetal death were independent of previous sensitisation within the same pregnancy.

• Women with false-negative test results but who are provided with cord serology and postpartum anti-D immunoglobulin have a sensitisation rate of 0.95% despite forgoing anti-D immunoglobulin treatment for potentially sensitising events.

• Compliance with RAADP is same with and without NIPT; similarly, compliance for postpartum anti-D immunoglobulin is assumed to be the same with or without NIPT.

Base-case analysis

The set of main assumptions used in the base-case analysis are shown in Table 24.

Sensitivity analyses

A series of scenario analyses and SAs was also conducted. We focused on parameters and assumptions to which we expected that the ICER would be the most sensitive and where the available evidence was limited. The SAs are described in detail but also summarised in Table 24. We focus on the comparison of current practice with the best performing postpartum scenario in all cases unless the results of the SA affect the rank order of postpartum scenarios or suggest that multiple postpartum scenarios could potentially provide the highest NHB.

Model validation

Pedro Saramago developed the model and Susan Griffin checked the model for errors. Comparisons across strategies were done to identify inconsistencies. Comparisons with the previous NICE TA156 were also done to identify the sources of any potential discrepancy.

Base-case results

Table 26 presents the results for each postpartum testing scenario separately against current practice of ‘no test and RAADP’. Total costs, total QALYs, incremental costs and incremental QALYs are presented together with incremental cost per QALY gained (ICER) and population NHBs at £20,000 and £30,000 threshold values. The results of the model suggest that for each additional sensitisation there is a loss of approximately 0.9 QALYs. Any difference in QALYs between strategies is attributable wholly to the difference in the number of sensitisations.

Non-invasive prenatal testing PP1 describes the use of NIPT to guide RAADP only, with all women continuing to receive cord serology with FMH testing and postpartum anti-D immunoglobulin as required, irrespective of NIPT result. This is estimated to reduce costs by £584,000 per 100,000 pregnancies and to result in lower health benefits (0.5 QALYs) than current practice.

Non-invasive prenatal testing PP2 (NIPT PP2) describes the use of NIPT to guide both RAADP and postpartum care to women who test positive or in whom the results are inconclusive, when cord serology is provided only in these women to guide FMH testing and postpartum anti-D immunoglobulin as required. This is estimated to reduce costs compared with current practice by approximately £671,000 but to result in a loss of 19.1 QALYs per 100,000 pregnancies.

Non-invasive prenatal testing PP3 (NIPT PP3) describes the use of NIPT to guide RAADP and postpartum anti-D immunoglobulin to women who test positive or inconclusive and when cord serology is used to guide FMH testing and postpartum anti-D immunoglobulin as required only to women whom NIPT indicates have a RhD-negative fetus. This is estimated to reduce costs compared with current practice by £485,000 but to result in a loss of 0.5 QALYs per 100,000 pregnancies.

Non-invasive prenatal testing PP4 (NIPT PP4) describes the use of NIPT to guide both RAADP and postpartum FMH testing and anti-D immunoglobulin to women who test positive or inconclusive and when cord serology is not provided. This is estimated to reduce costs compared with current practice by approximately £573,000 but results in a loss of 19.1 QALYs per 100,000 pregnancies.

All postpartum scenarios are cost saving but also less effective than no test and RAADP, placing them on the south-west quadrant of the cost-effectiveness plane (Figure 13). The least effective strategies are those that omit cord serology for women who test negative on NIPT. Without cord serology false negatives are not picked up at delivery and are not provided with postpartum anti-D immunoglobulin. In the model, the additional health gains are determined by the management of high-throughput NIPT false-negative test results.

FIGURE 13.

Cost-effectiveness plane of current practice (no test and RAADP) and alternative NIPT scenarios (PP1 to PP4).

Owing to these NIPT strategies being less costly and less effective than no test and RAADP, the ICERs calculated in Table 25 (and Figure 13) show the cost per QALY gained with current practice compared with high-throughput NIPT. Hence, when the ICER is above the cost-effectiveness threshold this would support the use of NIPT (no test and RAADP vs. NIPT PP1, ICER approximately £1,270,000 per QALY gained). The cost-effectiveness threshold can be used to present results in terms of NHBs, in which case the comparison is more straightforward, as the strategy with the highest NHB is preferred. All NIPT strategies have an expected NHB higher than no test and RAADP, both at threshold values of £20,000 and £30,000. Compared with no test and RAADP, NIPT PP1 has greater NHB (incremental NHB at £20,000 of approximately 14; incremental NHB at £30,000 of approximately 16, vs. no test and RAADP).

The base-case analysis assumes no adverse health impacts from use of a blood-based product, such as anti-D immunoglobulin. This is in line with the fact that widespread global use of anti-D immunoglobulin has yet to produce evidence of any adverse consequences. We illustrate how sensitive the ICER is to changes in these assumptions. Using the net benefit framework, it is possible to interpret the results of the SA around the price of anti-D immunoglobulin in terms of health impact. An increase of 20% in the cost of anti-D immunoglobulin represents a cost of £39.50 × 0.2 = £7.90. At a cost-effectiveness threshold of £20,000 per QALY, this is equivalent to assuming a health cost of 7.9/20,000 = 0.0004 QALYs per administration, or a loss of 3.5 hours of full lifetime health from every woman per dose of anti-D immunoglobulin they receive.

The incremental costs of introducing NIPT can be broken down into the cost of NIPT, the cost of managing potentially sensitising events, the cost of RAADP, the cost of postpartum tests and anti-D immunoglobulin and the cost consequences of sensitisations, and these are shown in Table 27. Although the added NIPT cost is similar across strategies (at approximately £1,585,000 per 100,000 pregnancies) it is accumulated over multiple pregnancies and so is affected by the performance of strategy in terms of the number of sensitisations. Strategies with more sensitisations (NIPT PP2 and NIPT PP4) have marginally less test cost, as sensitised women do not receive NIPT to guide RAADP in subsequent pregnancies (however, it is worth noting that it is recommended that NIPT be used in women who are sensitised in order to guide antenatal care). Similarly, all strategies save similar levels of costs from avoiding RAADP (approximately £1,544,000 per 100,000 pregnancies) and the management of potentially sensitising events (approximately £626,000 per 100,000 pregnancies). The NIPT strategies vary more markedly in their impact on postpartum testing and anti-D immunoglobulin costs. Here, NIPT PP1 is essentially the same as current practice, except for the small reduction in costs attributable to increased sensitisations, which makes women ineligible for FMH testing and anti-D immunoglobulin. NIPT PP2 decreases postpartum care costs by avoiding cord serology for women who test negative, but this comes at an increased cost of managing sensitisations, as false negatives are not picked up at delivery and women testing negative falsely are not provided with postpartum FMH tests and anti-D immunoglobulin. NIPT PP3 increases postpartum care costs because, although cord serology is avoided for those who test positive, this results in unnecessary use of FMH tests and anti-D immunoglobulin among women who test false positive (which includes those who test inconclusive but carry a RhD-negative baby). NIPT PP4 decreases postpartum care costs relative to current practice by avoiding cord serology for all women and is a combination of NIPT PP2 and NIPT PP3. As might be expected, the added cost of managing sensitisations and their associated health consequences is largest for the strategies with more sensitisations (NIPT PP2 and NIPT PP4) and is very small for strategies NIPT PP1 and NIPT PP3 (approximately £1700 per 100,000 pregnancies).

The assumption that the results of NIPT can be used to avoid all costs associated with the management of potentially sensitising events is favourable to NIPT and £626,000 represents the maximum cost saving in this regard. If this cost saving is reduced to £52,000, that is, if 92% of potentially sensitising events occur prior to the results of the NIPT being known, the ICER for no test and RAADP compared with NIPT PP1 would fall below £20,000 per QALY. The results of the audit indicate that 80% of potentially sensitising events occur after 20 weeks’ gestation. This suggests that incorporating NIPT into routine antenatal care when it would be provided in week 20 or earlier (see Figure 11 for schedule of appointments) could avoid upwards of 80% of the cost of managing potentially sensitising events.

We calculated the probability that each strategy would be cost-effective compared with no test and RAADP for each pair-wise comparison. NIPT PP1 and NIPT PP3 both have 99% probability of being cost-effective at a threshold of £20,000 per QALY. NIPT PP2 and NIPT PP4 have a lower probability of being cost-effective at £20,000 per QALY, at no higher than 73% compared with no test and RAADP.

Table 28 presents the fully incremental cost-effectiveness probabilistic results for high-throughput NIPT versus other strategies. Fully incremental results do not compare each NIPT strategy with current practice (i.e. no test and RAADP) but compare all NIPT scenarios simultaneously as competing alternative strategies. In this table, strategies are ranked by total costs and total QALYs, with the cheapest strategy coming first (NIPT PP2). Dominated strategies (those that have higher costs than more effective strategies) are in the bottom rows of the table. Incremental costs, incremental QALYs and, consequently, the ICER are incremental to the next cheapest, non-dominated strategy. This means that they represent the difference between the strategy in the current row compared with the strategy in the row above, as the table is ordered from least to most costly. The same applies to the incremental NHBs at £20,000 and £30,000 threshold values.

In NIPT PP2 cord serology is used to identify false-positive results, thereby avoiding unnecessary FMH testing and anti-D immunoglobulin in these women, but this is withheld in women for whom NIPT indicates a RhD-negative fetus. Using the negative results of high-throughput NIPT to rule out postpartum cord serology, FMH and anti-D immunoglobulin (NIPT PP2 and NIPT PP4) has lower QALYs than no test and RAADP, NIPT PP1 and NIPT PP3. Although there are further cost savings from avoiding postpartum cord serology and anti-D immunoglobulin, the majority of sensitisations occur and can be prevented by the administration of anti-D immunoglobulin at delivery. NIPT PP2 is the cheapest strategy and provides the same QALYs as NIPT PP4. Hence, NIPT PP4 is dominated by NIPT PP2.

Providing cord serology to all women, as with NIPT PP1, will identify both the false-positive results (the small number of false positives and the proportion of women with inconclusive results who are carrying RhD-negative babies) and false-negative results. Although NIPT PP1 has higher costs than NIPT PP2 because of the additional cord serology tests, these are offset somewhat by cost savings from avoiding sensitisations in false negatives. Compared with NIPT PP2, NIPT PP1 is estimated to provide approximately 19 additional QALYs per 100,000 pregnancies, at approximately £88,000 in additional costs, corresponding to an ICER of around £5000 per QALY gained.

In NIPT PP3 cord serology is used to identify false-negative results but this is withheld in women with inconclusive results or for whom NIPT indicates a RhD-positive fetus (in favour of FMH testing and anti-D immunoglobulin). Compared with NIPT PP1, the QALY gain is not affected as the model assumes no adverse health benefits from unnecessary use of anti-D immunoglobulin. As NIPT PP3 is more costly than NIPT PP1, in the base case it is dominated by NIPT PP1.

No test and RAADP is more costly than NIPT PP1 and is the most effective strategy. The administration of RAADP and supplementary anti-D immunoglobulin for potentially sensitising events among the false negatives leads to an additional 0.5 QALYs per 100,000 pregnancies compared with NIPT PP1, at an additional cost of £584,000. This means that the ICER for no test and RAADP compared with NIPT PP1 is £1,270,000. Using high-throughput NIPT and performing cord serology irrespective of the result (NIPT PP1) has higher NHB than any other strategy.

The decision uncertainty can be shown graphically with a cost-effectiveness acceptability curve. Figure 14 shows the cost-effectiveness acceptability curves for the different scenarios being compared (i.e. no test and RAADP and alternative high-throughput NIPT scenarios – PP1 to PP4) in which we can depict the probability that each strategy is cost-effective for a range of threshold values. When all strategies are simultaneously compared, for threshold values of £20,000 and £30,000, the highest probability of being cost-effective is obtained by NIPT PP1 with 0.65 and 0.73, respectively. For the same threshold values, the probability of NIPT PP2 being cost-effective is 0.30 and 0.22, respectively. NIPT PP1 is the alternative with the highest probability of being cost-effective and also the expected cost-effective alternative for thresholds above £10,000. An estimate of the maximum value of further research, the expected value of perfect information, is estimated to be approximately £203,000 considering 10 years of cohorts of 100,000 pregnancies and using a cost-effectiveness threshold of £20,000 per QALY. If research to reduce uncertainty in the model values would cost > £203,000 this suggests that it would not represent a good investment.

FIGURE 14.

Cost-effectiveness acceptability curves of current practice (no test and RAADP) and alternative NIPT scenarios (PP1 to PP4).

Sensitivity analyses results

Several SAs were carried out to assess the sensitivity of the base-case cost per QALY findings, as detailed in Table 24. We assessed the impact of using pooled evidence from all relevant NIPT accuracy evidence rather than UK Bristol studies only and, by using recent evidence from a UK study,12 assessed the performance of high-throughput NIPT at different gestation periods. An analysis of the NIPT inconclusive results was also performed by replacing the pooled estimates for the sensitivity and specificity with the individual study results. SA was performed on the effectiveness of RAADP by using a different sensitisation rate pooled from a larger number of studies. An assessment was also carried out for the uptake rates for RAADP and postpartum anti-D immunoglobulin, with and without NIPT, decreasing these to the circumstances when the correct dose at the correct time was administered according to recent evidence. 8 In addition, we analysed the impact of altering the cost of the diagnostic test and the cost of treatment, two key components of this assessment as highlighted in the relevant literature. Finally, we have evaluated the impact of reducing the cost of the FMH test and, under an alternative postpartum scenario, assessed the management of high-throughput inconclusive results separately to the positive test results. The following sections look closely at each of these analyses and provide interpretations of obtained results relative to the base-case findings.

Sensitivity analysis 5: sensitivity analysis on non-invasive prenatal testing inconclusive results

The cost saving achievable by using the high-throughput NIPT to guide anti-D immunoglobulin will depend on the rate of inconclusive test results, as for these women the current care pathway is unchanged. That is, all inconclusive results are managed as if they were test positive and, hence, unnecessary anti-D immunoglobulin continues to be provided in these women carrying a RhD-negative fetus. In order to undertake a SA around the rate of inconclusives, we replaced the pooled estimates for the sensitivity and specificity with the individual study results. Figure 15 shows how the specificity varies with the rate of inconclusives within each study. In general, a higher rate of inconclusive results will lead to a larger number of false positives and, correspondingly, a lower specificity. The cost saving achievable by using high-throughput NIPT to guide anti-D immunoglobulin will depend on the rate of inconclusive test results, as for these women the current care pathway is unchanged, that is, all inconclusive results are managed as if they were test positive, and, hence, unnecessary antenatal anti-D immunoglobulin continues to be provided in those women carrying a RhD-negative fetus.

FIGURE 15.

Specificity by rate of high-throughput NIPT inconclusive results per study.

One study produced no inconclusive results and no false-negative results and so we omitted this from the SA. 22 In general, the NHBs associated with the NIPT strategies fall as the rate of inconclusive results increases, but at no point do the NHBs from NIPT PP1 or NIPT PP3 fall below those offered with no test and RAADP. Figure 16 shows the NHBs for all of the NIPT strategies. When the rate of inconclusive results is low, NIPT PP3 offers the highest NHB. This is because the amount of unnecessary postpartum FMH testing and anti-D immunoglobulin is reduced when the number of false-positive results falls. When the rate of inconclusives is high, NIPT PP1 is preferred. If the rate of inconclusives was very high, no test and RAADP would be preferred. However, the rate would have to be much higher than that observed in the set of studies underlying the evidence synthesis. Akolekar et al. 91 and Wikman et al. 23 diverge from the remaining studies in terms of the number of false-negative results and sensitivity and this impacts on the NHBs of the strategies that do not identify false negatives through cord serology (NIPT PP2 and NIPT PP4). For these two strategies, the NHB falls below that offered by current practice when the results associated with these two studies are used. Figure 17 shows how the NHBs for NIPT PP1 vary only with the rate of inconclusives.

FIGURE 16.

Population NHBs for all NIPT strategies by rate of NIPT inconclusive results per study.

FIGURE 17.

Population NHBs for NIPT PP1 by rate of NIPT inconclusive results per study.

Diagnostic accuracy

Eight studies were included in the diagnostic review of high-throughput NIPT. There were three studies based at Bristol (UK). The majority of included studies were judged as having a low risk of bias.

Meta-analyses showed very high diagnostic accuracy of high-throughput NIPT. In the primary analyses, for which women with inconclusive test results were treated as being testing positive, the summary FNR (i.e. women at risk of sensitisation) was 0.34% (95% CI 0.15% to 0.76%) and the FPR (i.e. women needlessly receiving anti-D) was 3.86% (95% CI 2.54% to 5.82%). SAs did not materially alter the overall result.

A subgroup analysis of three high-quality studies based at Bristol (UK) showed a slightly lower FNR of 0.21% (95% CI 0.09% to 0.48%) and a higher FPR of 5.73% (95% CI 4.58% to 7.16%). This suggests that the Bristol NIPT approach may be using a different threshold for the detection algorithm that further reduces false-negative error rates, consequently increasing the false-positive error rate. The FPR found was mostly as a result of treating the roughly 7% of women (in the UK) who have an inconclusive test result as if they had a positive test. Excluding these women from analysis resulted in a lower FPR of 1.26% (95% CI 0.87% to 1.83%). We were unable to conduct the subgroup analysis based on ethnicity because of lack of relevant data from included studies.

The diagnostic accuracy performance of high-throughput NIPT varied by gestational age. The data suggest that high-throughput NIPT is insufficiently accurate before around 11 weeks’ gestation (i.e. in first trimester) but is consistently accurate at any time after 11 weeks’ gestation. This may be because of a low concentration of cell-free fetal DNA in early pregnancy92 but an increased concentration of cell-free fetal DNA after the end of the first trimester. 93

Clinical effectiveness

Seven studies18,20,22,24–27 were included in the clinical effectiveness review. Only two studies20,26 had a control group, but both studies were judged as having a high risk of bias. One large prospective cohort study26 reported that use of high-throughput NIPT for targeted antenatal anti-D prophylaxis was associated with a significant risk reduction in sensitisation (adjusted odds ratio 0.41, 95% CI 0.22 to 0.87) compared with historical controls (routine management, postpartum anti-D only).

Three non-comparative studies18,20,22 reported outcome measures relating to anti-D doses administrated. All studies found that the use of NIPT reduced the total use of anti-D immunoglobulin doses (decreasing by 29% in one UK study by Soothill et al. 18) because around 35% of RhD-negative women avoided unnecessary anti-D administration.

Four studies20,26,27,49 reported moderate to high compliance with antenatal anti-D immunoglobulin administration. The compliance with antenatal anti-D administration after a positive NIPT result ranged from 86% to 96.1% (four studies20,26,27,49). High-throughput NIPT uptake rates ranged from 70% to > 95% (seven studies. 12,18,20,22,25–27).

The results from the simulation study suggested that the use of NIPT to determine antenatal anti-D use would substantially reduce the number of women receiving anti-D unnecessarily from 38.9% to 5.7%. Results were sensitive to the rate of compliance. NIPT use could increase sensitisation rates by up to 15 sensitisations per 100,000 women if postpartum cord blood testing is continued or up to 28 per 100,000 women if cord blood testing is withdrawn and postpartum anti-D given on the basis of the NIPT result. Sensitisation rates are minimised by ensuring that women who do not receive NIPT are still offered, and receive, antenatal anti-D. The results suggest that NIPT results (if available and conclusive) could potentially be used in place of cord blood testing for administration of postpartum anti-D, if the small increase in sensitisations rates can be considered ethically acceptable.

Implementation

Twelve studies13,17,18,20–28 were included in the review of implementation. Most of the included studies were large cohort studies13,17,20,21,23–27 reporting implementation data along with diagnostic accuracy data, although one study was a survey that was based in the UK (London). 28 All the large cohort studies reported high diagnostic accuracy of high-throughput NIPT and suggested that high-throughput RhD genotyping of fetuses in all RhD-negative women was feasible and should be recommended. A number of studies reported potential issues of implementation such as those relating to programme anti-D prophylaxis compliance. 20,27 Some studies highlighted the importance of short transport times of samples and the need for effective management of transporting samples. 13,17,24 Some studies also identified the need for greater knowledge of NIPT among physicians, midwives and pregnant women. 27,28

Cost-effectiveness

Seven cost-effectiveness studies58,68–73 were included in the review. Conflicting results were identified across the existing economic studies, with three of the studies68,71,72 reporting that NIPT fetal RhD genotyping did not appear to be cost-effective. The unit cost of the test was consistently identified as a key driver of the cost-effectiveness results and the potential for the use of NIPT to result in overall cost savings. Only one of the studies68 was undertaken in a UK context, but this study did not explicitly explore how the introduction of NIPT could impact on costs relating to potentially sensitising events. For the studies undertaken outside the UK, differences in health-care systems and in implementation of anti-D immunoglobulin policies limit their relevance to UK practice. In conclusion, none of the existing studies was considered to be sufficiently generalisable to inform the specific decision problem as set out in the NICE scope for the current assessment.

A de novo independent economic model was developed to assess the cost-effectiveness of high-throughput NIPT to identify fetal rhesus D status in women who are RhD negative and not known to be sensitised to the RhD antigen. The model was made up of two main elements: (1) an identification part reflecting the diagnostic performance and costs of the alternative identification strategies and (2) a treatment part that evaluated the subsequent costs and outcomes (expressed in QALYs) of alternative care pathways. Four alternative ways in which the use of high-throughput NIPT may impact on the existing postpartum care pathway were evaluated (cord serology, FMH testing and postpartum anti-D immunoglobulin). These included scenarios in which the result of NIPT was used to guide RAADP only (with all women continuing to receive cord serology with FMH testing and postpartum anti-D immunoglobulin as required, irrespective of NIPT result) and scenarios for which the NIPT result guided both RAADP and separate aspects of postpartum care. A series of additional sensitivity and scenario analyses was also performed.

Our de novo economic model indicated that the use of high-throughput NIPT to guide the prenatal and postpartum provision of anti-D immunoglobulin prophylaxis is estimated to be cost saving compared with the current practice of providing RAADP to all women who are RhD negative. The magnitude of the cost saving appears to be highly sensitive to the cost of NIPT itself to the NHS, which comprises the base unit cost per test, the level of any royalty fee and any increase in antenatal care costs required to accommodate an additional test. In the base-case analysis, the extent of the cost saving appears sufficient to outweigh the small increase in sensitisations and the associated small QALY loss through using NIPT compared with current practice. However, even a small increase in the cost imposed on the NHS of (confidential information has been removed) or more per test would alter these conclusions.

In the base-case analysis, all four separate postpartum scenarios were estimated to be cost saving but also less effective than current practice. Based on a cross-section of 100,000 pregnancies, the magnitude of cost savings varied between approximately £485,000 and £671,000. The magnitude of the QALY loss varied between 0.5 QALYs and 19.1 QALYs (per 100,000 pregnancies). Although the magnitude of the cost savings was sufficient to outweigh the associated QALY loss when each postpartum scenario was separately compared with current practice, these four separate scenarios potentially represent separate and distinct testing and management strategies that should be directly compared. In the base-case analysis, the strategy in which the NIPT result is used to guide RAADP only (i.e. all women continuing to receive cord serology with FMH testing and postpartum anti-D immunoglobulin) was associated with the highest NHB and had the highest probability of being cost-effective for threshold values of £20,000 and £30,000 per QALY (probability of 0.65 and 0.73, respectively). However, the use of cord serology to capture false-negative results has the potential to undermine the implementation of the test if it impacts on the confidence in the NIPT results. The most efficient postpartum strategy was also shown to vary across several of the main SAs.

A postpartum strategy that distinguishes between inconclusive results and positive results offers the greatest cost savings. In the base-case analysis, women who receive inconclusive results were assumed to be managed as if they test positive, but there may be potential for further cost savings if these are treated as a distinct group in terms of postpartum care. This could allow for a postpartum scenario in which cord serology was applied to women who test negative and who test inconclusive but in which FMH tests and anti-D immunoglobulin are provided without cord serology in women who test positive.

Clinical effectiveness

Extensive literature searches were conducted in an attempt to maximise retrieval of potentially relevant studies. These included electronic searches of a variety of bibliographic databases as well as screening of clinical trial registers and conference proceedings to identify unpublished studies. The search strategy did not restrict by study design. The review process followed recommended methods to minimise the potential for error and/or bias. The quality of the included studies was assessed and accounted for when interpreting the review results. Appropriate synthesis methods were employed by taking into account the heterogeneity of study characteristics.

There was some evidence of inconsistency in the meta-analysis of diagnostic accuracy studies. The observed heterogeneity may be explained by variations in methods used in the high-throughput NIPT approach (including diagnostic accuracy thresholds and number and types of exons targeted), gestational age at the time of testing and different methods of handling inconclusive test results. There were also variations in the reporting of included studies. Particularly, two studies19,21 did not report the number of inconclusive results of the test and some studies12,18,22 did not report detailed reasons for inconclusive results.

There was very limited evidence relating to the clinical effectiveness of high-throughput NIPT. No studies were identified reporting adverse effects of high-throughput NIPT.

Owing to limited evidence, the generalisability of the review findings to non-white women and multiple pregnancies is unclear.

Cost-effectiveness

The de novo economic model was specifically developed to address the limitations of existing studies and concerns regarding the generalisability to current UK practice. The main strength of the decision model is the linkage between the diagnostic accuracy of a given identification strategy, the impact on subsequent treatment decisions and the ultimate effect on health outcomes and costs. A key element of the model is based on the previous economic model underpinning NICE TA15662 on RAADP ensuring consistency between the separate diagnostic and TAs. A broad range of scenario analyses and SAs were undertaken to address key assumptions and uncertainties.

Clinical effectiveness

In this assessment we identified very limited data on the evaluation of clinical effectiveness for using high-throughput NIPT to detect fetal RhD status in RhD-negative women. Therefore, the potential role of high-throughput NIPT in terms of its clinical impact on the care pathway and adverse effects to the mother and fetus remains unclear. In particular, we did not identify any studies reporting comparative data relating to patient-related outcomes, such as quality-of-life measures.

Owing to a lack of sufficient data from included studies, we were unable to conduct subgroup analyses based on ethnicity. Therefore, whether or not the diagnostic performance of high-throughput NIPT differs between different ethnic groups remains unclear.

In terms of implementing high-throughput NIPT in health-care settings, no studies were identified reporting compliance rates to prenatal anti-D treatment in UK settings. Although a few non-UK studies reported compliance rates to prenatal anti-D treatment, the generalisability of their findings to the UK settings remains uncertain because of variations in national guidelines and health policies between different countries.

Cost-effectiveness

There is uncertainty regarding the cost of introducing high-throughput NIPT. The unit cost will vary with throughput and may be subject to an additional royalty fee. Unless NIPT can be incorporated seamlessly into routine antenatal care, it may result in additional costs for blood draw, transportation of samples and antenatal care visits to administer the test and deliver counselling and results. We conducted extensive SAs to address this uncertainty and to identify the threshold cost per NIPT. The cost of high-throughput NIPT has to increase by only (confidential information has been removed) above that modelled in the base case in order for current practice to be the preferred strategy.

Although there remains uncertainty as regards the timing of the test, our analysis suggests that this does not appear to be influential in determining the cost-effectiveness results either in terms of diagnostic accuracy or in terms of the extent of management costs for potentially sensitising events that can be avoided.

Although the evidence to support the diagnostic accuracy of NIPT is of good quality, existing evidence informing the impact of sensitisations in terms of their long-term health and cost consequences are more limited and highly uncertain.

Key

/ = indexing term [medical subject heading (MeSH) heading]

exp = exploded indexing term (MeSH heading)

$ = truncation

# = mandated wildcard – stands for one character

? = optional wildcard – stands for zero or one character

.ti,ab. = terms in either title or abstract fields

.pt. = publication type

adj = terms next to each other (order specified)

adj2 = terms within two words of each other (any order)

Key

MH = indexing term (CINAHL heading)

* = truncation

? = wildcard – stands for one character

# = optional wildcard – stands for zero or one character

TI = words in the title

AB = words in the abstract

“ “ = phrase search

N2 = terms within two words of each other (any order)

PT = publication type

Key

MeSH descriptor = indexing term (MeSH heading)

* = truncation

:ti,ab,kw = terms in either title or abstract or keyword fields

near/2 = terms within two words of each other (any order)

next = terms are next to each other

“ “ = phrase search

Key

MeSH DESCRIPTOR = indexing term (MeSH heading)

* = truncation

NEAR2 = terms within two words of each other (order specified)

“ ” = phrase search

Key

/ = indexing term (Emtree heading)

exp = exploded indexing term (Emtree heading)

$ = truncation

# = mandated wildcard – stands for one character

? = optional wildcard – stands for zero or one character

.ti,ab. = terms in either title or abstract fields

.pt. = publication type

sh. = subject heading field

adj = terms next to each other (order specified)

adj2 = terms within two words of each other (any order)

Key

.de. = subject heading search

$ = truncation

# = mandated wildcard – stands for one character

? = optional wildcard – stands for zero or one character

.ti,ab. = terms in either title or abstract fields

.pt. = publication type

adj = terms next to each other (order specified)

adj2 = terms within two words of each other (any order)

ClinicalTrials.gov (https://clinicaltrials.gov/)

Date searched: 10 November 2015.

Records retrieved: 44.

RhD OR “rhesus D” OR “Rh(D)” OR “Rh-(D)” OR “Rh D” OR “Rh-negative” OR “Rh-positive” OR “Rhesus negative” OR “Rhesus positive”

The search was updated on 26 February 2016, retrieving two new records.

Conference Proceedings Citation Index: Science (via Web of Science, Thomson Reuters, http://thomsonreuters.com/thomson-reuters-web-of-science)

Date range searched: 1990 to 4 November 2015.

Date searched: 6 November 2015.

Records retrieved: 801.

See Science Citation Index for search strategy used. As both databases were searched together, the records retrieved refers to results from both databases.

The searches for Science Citation Index and the Conference Proceedings Citation Index: Science were updated on 26 February 2016, retrieving 811 records.

EU Clinical Trials Register (www.clinicaltrialsregister.eu/ctr-search/search)

Date searched: 10 November 2015.

Records retrieved: 4.

“RhD” OR “rhesus D” OR “Rh(D)” OR “Rh-(D)” OR “Rh D” OR “Rh-negative” OR “Rh-positive” OR “Rh negative” OR “Rh positive” OR “Rhesus negative” OR “Rhesus positive”

The search was updated on 26 February 2016 but no new records were retrieved.

PROSPERO (www.crd.york.ac.uk/PROSPERO)

Date searched: 10 November 2015.

Records retrieved: 4.

RhD or Rh-D or Rh-negative or Rh-positive in all fields.

The search was updated on 26 February 2016, retrieving one new record.

World Health Organization’s International Clinical Trials Registry Platform (www.who.int/ictrp/search/en)

Date searched: 10 November 2015.

Records retrieved: 29.

RhD OR rhesus OR Rh-negative OR Rh-positive.

The search was updated on 26 February 2016 but no new records were retrieved.

National Guidelines Clearinghouse (www.guideline.gov)

Date searched: 17 November 2015.

(rhd or rhesus or ‘rh negative” or “rh positive”)‘ and ‘(pregnan* or maternal or antenatal or ante-natal or prenatal or pre-natal or intrapartum)

A total of 23 results were retrieved and browsed for relevance; 18 relevant guidelines were found.

National Institute for Health and Care Excellence (www.nice.org.uk)

Date searched: 13 November 2015.

1. Browsed for relevant guidance in the fertility, pregnancy and childbirth section: www.nice.org.uk/guidance/conditions-and-diseases/fertility--pregnancy-and-childbirth.

2. Searched NICE website using general search box with keyword RhD.

3. Searched NICE website using general search box with keyword Rhesus.

4. Relevant guidelines found.

NHS Evidence (www.evidence.nhs.uk)

Searched on: 17 November 2015.

(rhd OR rhesus OR “rh negative” OR “rh positive”) AND (pregnan* OR maternal OR antenatal OR ante-natal OR prenatal OR pre-natal OR intrapartum) limited to guidelines.

A total of 81 results were retrieved and browsed for relevance. Seven relevant guidelines were found.

Royal College of Obstetricians and Gynaecologists (www.rcog.org.uk/en)

Date searched: 13 November 2015.

1. Browsed all guidelines.

2. Searched all guidelines by keyword – RhD or rhesus.

Four relevant guidelines were found.

Turning Research into Practice database (www.tripdatabase.com)

Date searched: 17 November 2015.

(rhd OR rhesus OR “rh negative” or “rh positive”) AND title:(pregnan* OR maternal OR antenatal OR ante-natal OR prenatal OR pre-natal OR intrapartum)

A total of 37 results were retrieved and browsed for relevance; 17 relevant guidelines were found.

UK National Screening Committee (www.gov.uk/government/groups/uk-national-screening-committee-uk-nsc)

Date searched: 13 November 2015.

Recommendations list was filtered by antenatal and the resulting list browsed.

One relevant report was found.

MEDLINE and MEDLINE In-Process & Other Non-Indexed Citations (via Ovid, http://ovidsp.ovid.com/)

Date range searched: 1946 to October Week 5 2015.

Date searched: 18 January 2016.

Records retrieved: 45.

The search was updated on 4 March 2016, retrieving 45 records.

1. systematic$ review$.ti,ab. (75,835)

2. meta-analysis as topic/ (14,365)

3. meta-analytic$.ti,ab. (4298)

4. meta-analysis.ti,ab,pt. (89,180)

5. metanalysis.ti,ab. (140)

6. metaanalysis.ti,ab. (1210)

7. meta analysis.ti,ab. (70,616)

8. meta-synthesis.ti,ab. (331)

9. metasynthesis.ti,ab. (166)

10. meta synthesis.ti,ab. (331)

11. meta-regression.ti,ab. (3249)

12. metaregression.ti,ab. (344)

13. meta regression.ti,ab. (3249)

14. (synthes$ adj3 literature).ti,ab. (1689)

15. (synthes$ adj3 evidence).ti,ab. (4926)

16. integrative review.ti,ab. (1177)

17. data synthesis.ti,ab. (7985)

18. (research synthesis or narrative synthesis).ti,ab. (1041)

19. (systematic study or systematic studies).ti,ab. (8551)

20. (systematic comparison$ or systematic overview$).ti,ab. (2200)

21. evidence based review.ti,ab. (1467)

22. comprehensive review.ti,ab. (8251)

23. critical review.ti,ab. (11,964)

24. quantitative review.ti,ab. (517)

25. structured review.ti,ab. (542)

26. realist review.ti,ab. (102)

27. realist synthesis.ti,ab. (73)

28. or/1-27 (187,703)

29. review.pt. (2,049,547)

30. medline.ab. (68,680)

31. pubmed.ab. (46,181)

32. cochrane.ab. (39,786)

33. embase.ab. (40,092)

34. cinahl.ab. (12,936)

35. psyc?lit.ab. (879)

36. psyc?info.ab. (10,559)

37. (literature adj3 search$).ab. (32,390)

38. (database$ adj3 search$).ab. (30,393)

39. (bibliographic adj3 search$).ab. (1461)

40. (electronic adj3 search$).ab. (11,252)

41. (electronic adj3 database$).ab. (13,910)

42. (computeri?ed adj3 search$).ab. (2857)

43. (internet adj3 search$).ab. (2045)

44. included studies.ab. (9670)

45. (inclusion adj3 studies).ab. (8188)

46. inclusion criteria.ab. (44,510)

47. selection criteria.ab. (22,215)

48. predefined criteria.ab. (1258)

49. predetermined criteria.ab. (787)

50. (assess$ adj3 (quality or validity)).ab. (48,127)

51. (select$ adj3 (study or studies)).ab. (43,640)

52. (data adj3 extract$).ab. (34,903)

53. extracted data.ab. (8161)

54. (data adj2 abstracted).ab. (3617)

55. (data adj3 abstraction).ab. (1017)

56. published intervention$.ab. (121)

57. ((study or studies) adj2 evaluat$).ab. (121,595)

58. (intervention$ adj2 evaluat$).ab. (7046)

59. confidence interval$.ab. (258,288)

60. heterogeneity.ab. (106,141)

61. pooled.ab. (53,158)

62. pooling.ab. (8496)

63. odds ratio$.ab. (171,463)

64. (Jadad or coding).ab. (133,119)

65. or/30-64 (923,716)

66. 29 and 65 (141,974)

67. review.ti. (299,976)

68. 67 and 65 (62,549)

69. (review$ adj4 (papers or trials or studies or evidence or intervention$ or evaluation$)).ti,ab. (119,221)

70. 28 or 66 or 68 or 69 (340,645)

71. letter.pt. (897,674)

72. editorial.pt. (391,059)

73. comment.pt. (647,299)

74. 71 or 72 or 73 (1,445,828)

75. 70 not 74 (331856)

76. exp animals/ not humans/ (4,171,020)

77. 75 not 76 (321762)

78. “Rho(D) Immune Globulin”/ (1190)

79. (immune adj2 globulin adj2 rh$).ti,ab. (257)

80. anti-D.ti,ab. (2610)

81. (D-Gam or Partobulin or Rhophylac or WinRho).ti,ab. (47)

82. or/78-81 (3165)

83. 77 and 82 (45)

Cochrane Database of Systematic Reviews (via Wiley Online Library, http://onlinelibrary.wiley.com/)

Issue 1 of 12, January 2016.

Date searched: 18 January 2016.

Records retrieved: 6.

The search was updated on 4 March 2016, retrieving six records from CDSR.

#1 MeSH descriptor: [Rho(D) Immune Globulin] this term only(51)

#2 (immune near/2 globulin near/2 rh*):ti,ab,kw(5)

#3 anti-D:ti,ab,kw(110)

#4 (D-Gam or Partobulin or Rhophylac or WinRho):ti,ab,kw(10)

#5 #1 or #2 or #3 or #4(119)

#6 #1 or #2 or #3 or #4 in Cochrane Reviews (Reviews and Protocols)(6)

Database of Abstracts of Reviews of Effects (via Centre for Reviews and Dissemination, www.crd.york.ac.uk/CRDWeb)

Date range searched: inception to 31 March 2015.

Date searched on: 20 January 2016.

Records retrieved: 8.

1. (anti-D) IN DARE, HTA (15)

2. ((D-Gam or Partobulin or Rhophylac or WinRho)) IN DARE, HTA (1)

3. ((immune NEAR globulin NEAR rh*)) IN DARE, HTA (0)

4. ((immune NEAR rh* NEAR globulin)) IN DARE, HTA (0)

5. ((rh* NEAR immune NEAR globulin)) IN DARE, HTA (5)

6. ((rh* NEAR globulin NEAR immune)) IN DARE, HTA (0)

7. ((globulin NEAR rh* NEAR immune)) IN DARE, HTA (0)

8. ((globulin NEAR immune NEAR rh*)) IN DARE, HTA (0)

9. MeSH DESCRIPTOR Rho(D) Immune Globulin IN DARE,HTA (5)

10. #1 OR #2 OR #5 OR #9 (15)

11. (#1 or #2 or #5 or #9) IN DARE (8)

12. (#1 or #2 or #5 or #9) IN HTA (7)

Health Technology Assessment database (via Centre for Reviews and Dissemination, www.crd.york.ac.uk/CRDWeb)

Date range searched: inception to 31 March 2015.

Date searched: 20 January 2016.

Records retrieved: 7.

See above under Database of Abstracts of Reviews of Effects for search strategy used.

PubMed (www.ncbi.nlm.nih.gov/pubmed)

Date searched on: 20 January 2016.

Records retrieved: 57.

The search was updated on 4 March 2016, retrieving 58 records.

(((“Rho(D) Immune Globulin”[Mesh:noexp] OR “rh* immune globulin”[Title/Abstract]) OR (“RHO(D) antibody”[Supplementary Concept] OR “RHO(D) antibody”[All Fields] OR “anti d”[All Fields])) OR (Partobulin[Title/Abstract] OR Rhophylac[Title/Abstract] OR WinRho[Title/Abstract])) AND systematic[sb]

EconLit (via Ovid, http://ovidsp.ovid.com/)

Date range searched: 1886 to November 2015.

Date searched: 4 December 2015.

Records retrieved: 4.

1. (RhD or “rhesus D” or “Rh(D)” or “Rh-(D)” or Rh D).ti,ab. (3)

2. (Rh-negative or Rh-positive).ti,ab. (0)

3. (Rhesus negative or Rhesus positive).ti,ab. (0)

4. ((rh or rhesus) adj2 (factor or factors or antigen$ or system or group)).ti,ab. (1)

5. ((isoimmuni$ or iso-immuni$ or isoimmune or iso-immune) adj6 (rh or rhesus or maternal or pregnan$)).ti,ab. (0)

6. ((alloimmuni$ or allo-immuni$ or alloimmune or allo-immune) adj6 (rh or rhesus or maternal or pregnan$)).ti,ab. (0)

7. ((unsensiti#ed or un-sensiti#ed or non-sensiti#ed) adj6 (rh or rhesus or maternal or pregnan$)).ti,ab. (0)

8. ((sensiti#ation$ or sensiti#ed) adj6 (rh or rhesus or maternal or pregnan$)).ti,ab. (0)

9. ((fetomaternal or feto-maternal or foetomaternal or foeto-maternal) adj2 immuni#ation).ti,ab. (0)

10. ((rh or rhesus) adj2 (immuni#ation or autoimmuni#ation)).ti,ab. (0)

11. ((hemolytic or haemolytic) adj2 (disease$ or disorder$)).ti,ab. (0)

12. HDFN.ti,ab. (0)

13. ((rhesus or rh) adj2 (disease$ or disorder$)).ti,ab. (0)

14. ((rhesus or rh or RhD) adj2 (incompatib$ or antagonism)).ti,ab. (0)

15. ((erythroblastoses or erythroblastosis) adj2 f?etal$).ti,ab. (0)

16. or/1-15 (4)

NHS Economic Evaluations Database (via Centre for Reviews and Dissemination, www.crd.york.ac.uk/CRDWeb)

Date range searched: inception to 31 March 2015.

Date searched on: 4 December 2015.

Records retrieved: 6.

Please note that the total number of hits at line 54 refers to the total number of results from DARE, HTA database and NHS EED.

Research Papers in Economics (http://repec.org/)

Date searched: 4 December 2015.

Records retrieved: 0.

“RhD” | rhesus | “hemolytic disease” | “haemolytic disease” | HDFN | erythroblastoses | erythroblastosis | “fetomaternal immunisation” | “fetomaternal immunization” | “foetomaternal immunisation” | “foetomaternal immunization”