High-throughput non-invasive prenatal testing for fetal RHD genotype

4.1 Eight studies reported the diagnostic accuracy of high-throughput NIPT for fetal RHD genotype, all of which were prospective studies carried out in European countries. Four studies were done in England, 3 of which were based in Bristol. Cord blood typing was the reference standard in all studies. Six studies were considered to be at low risk of bias and 2 studies (Akolekar et al. 2011; Thurik et al. 2015) were judged to be at high risk of bias. Except for 2 studies (Akolekar et al. 2001; Wikman et al. 2012), the results of the studies were considered broadly applicable to using high-throughput NIPT for fetal RHD genotype for nationwide testing in the UK.

4.2 It is expected that, in the UK, women with inconclusive NIPT results will be treated as having a positive test with no further testing. Data on inconclusive results were not reported in 2 studies (Thurik et al. 2015; Grande et al. 2013). So, 4 approaches to the diagnostic accuracy analysis were considered:

4.3 Results of the hierarchical bivariate meta-analyses are shown in table 1. In all analyses, women in whom NIPT was carried out at or before 11 weeks' gestation were excluded because the test is known to be less accurate before 11 weeks. NIPT for fetal RHD genotype is very accurate among women with a rhesus‑D (D) positive fetus; only 2 to 4 in 1,000 such women will have a negative test result and so be at risk of sensitisation because they would not be offered antenatal anti‑D immunoglobulin. NIPT for fetal RHD genotype is slightly less accurate among women with a D‑negative fetus; between 13 and 57 in 1,000 such women will have a positive test result and so be offered antenatal anti‑D immunoglobulin unnecessarily.

Abbreviation: CI, confidence interval.

4.4 The analysis of the 3 Bristol studies gave a slightly lower false-negative rate (0.21%; 95% confidence interval [CI] 0.09 to 0.48) and a higher false-positive rate (5.73%; 95% CI 4.58 to 7.16) than analyses including other studies. This suggests that the Bristol high-throughput NIPT approach may use a different test threshold compared with the testing done in other studies; minimising false-negative findings, with a consequent increase in the false-positive rate.

4.5 There was considerable variation in rates of inconclusive tests across studies, ranging from 0.4% to 14.3%. The most likely causes for this variability are differences in how the NIPT was done (such as different numbers and types of exons considered) and differences in characteristics of study populations (for example, different proportions of women of black African family origin). Based on a meta-analysis, the average rate of inconclusive test results was estimated to be 4.0% (95% CI 1.5 to 10.3) if all studies reporting inconclusive results were included, and 6.7% (95% CI 3.7 to 11.7) if only the Bristol studies were included.

4.6 An analysis of the effect of the timing of high-throughput NIPT for fetal RHD genotype on diagnostic accuracy suggested that false-negative rates were higher before 11 weeks' gestation, and thereafter false-negative rates were consistent, irrespective of timing. The effect of the timing of high-throughput NIPT on the number of inconclusive test results suggested that the percentage of inconclusive results drops as the gestational age increases from 11 weeks.

4.7 Seven studies reported the clinical effectiveness of NIPT for fetal RHD genotype, all of which were observational and carried out in European countries. The sample size of the studies ranged from 284 to 15,126 and most participants were of white European family origin. Only 2 studies compared women having NIPT for fetal RHD genotype with controls (Tiblad et al. 2013; Banch Clausen et al. 2014). Tiblad et al. (2013) was considered to be at serious risk of bias, and Banch Clausen et al. (2014) was considered to be at critical risk of bias. The generalisability of these 2 studies to NHS clinical practice was limited because participants in the control group did not have routine antenatal anti‑D prophylaxis (RAADP). The other 5 studies only reported non-comparative-effectiveness data for women having NIPT for fetal RHD genotype. Data from these studies were supplemented with data from a UK audit on anti‑D immunoglobulin use (National comparative audit of blood transfusion: 2013 audit of anti-D immunoglobulin prophylaxis) for a comparison with current practice.

4.8 Tiblad et al. (2013) compared targeted RAADP in the first trimester with routine care (postpartum anti‑D prophylaxis only) in Sweden. They reported the incidence of D sensitisation in the cohort that had high-throughput NIPT for fetal RHD genotype as 0.26% (95% CI 0.15 to 0.36%; n=8347) compared with 0.46% (95% CI 0.37 to 0.56%; n=18,546) in the historical control cohort. High-throughput NIPT for fetal RHD genotype was associated with a significant risk reduction in sensitisation (unadjusted risk ratio [RR] 0.55; 95% CI 0.35 to 0.87) compared with historical controls. An updated analysis reported in a linked conference abstract (Neovius et al. 2015) found an adjusted odds ratio of 0.41 (95% CI 0.22 to 0.87).

4.9 Seven studies reported uptake rates of NIPT for fetal RHD genotype. Uptake rates ranged from 70% to more than 95% across the studies. In a pilot study done by Soothill et al. (2015) in 3 maternity services in the south-west of England, only 70% of eligible women joined the study in the first 6 months. The larger English study done by Chitty et al. (2014) reported that 88% of the 3,069 participants consented to have NIPT for fetal RHD genotype. The only country that reported nationwide uptake data was the Netherlands, where more than 95% of eligible women had NIPT for fetal RHD genotype. The studies generally noted that uptake is likely to increase over time if a nationwide screening programme is implemented.

4.10 The uptake of RAADP in women who accepted NIPT and had a positive result was reported in 4 studies and ranged from 86.0% to 96.1%. Of the larger studies, Van der Ploeg et al. (2015) reported nationwide data on women having NIPT for fetal RHD genotype in the Netherlands, where 96.1% of about 18,383 women with a positive test result had RAADP. Tiblad et al. (2013) reported a slightly lower rate, with 90% of 5,104 women with a positive NIPT result having RAADP. Data on the uptake of RAADP in women who had a negative test result, those who had an inconclusive test result, and those who refused NIPT for fetal RHD genotype, were limited. None of the studies reported whether all the women who had RAADP had the intended dosage at the intended time, or what proportion of women had additional anti‑D immunoglobulin because of a potentially sensitising event.

4.11 The uptake of postpartum anti‑D prophylaxis in women who accepted NIPT for fetal RHD genotype and had a positive test result was reported in 3 studies. Van der Ploeg et al. (2015) reported nationwide data on women having NIPT for fetal RHD genotype in the Netherlands, where 92% of about 18,383 women had postpartum anti‑D prophylaxis. A subgroup analysis by Banch Clausen et al. (2014) found slightly higher uptake of postpartum anti‑D prophylaxis among women who had NIPT (99.7%, 353/354) compared with those who did not have NIPT (95.7%, 66/69). Damkjaer et al. (2012) reported a similar rate among women who had NIPT (99.3%, 151/152). None of the included studies reported whether all women who had postpartum anti‑D prophylaxis had the intended dosage at the intended time.

4.12 Outcome measures relating to anti‑D immunoglobulin administration were reported in 3 non‑comparative studies. Soothill et al. (2015) reported a significant 6% reduction per month in anti‑D immunoglobulin administration (95% CI 4 to 8) over a 6‑month period in 3 maternity services in the south west of England. The total use of anti‑D immunoglobulin fell by about 29%, corresponding to 35% of D‑negative women not having anti‑D immunoglobulin in their pregnancy unnecessarily. Similar results were also seen by Banch Clausen et al. (2014), who reported that 37.1% of women avoided unnecessary anti‑D immunoglobulin within 2 years of the introduction of a programme of NIPT for fetal RHD genotype. Grande et al. (2013) reported that, of 95 women carrying a D‑negative fetus, 5 women requested anti‑D immunoglobulin; so, unnecessary anti‑D immunoglobulin was avoided in 95% of women carrying a D‑negative fetus.

4.13 To better understand the likely consequences of implementing NIPT for fetal RHD genotype and basing antenatal anti‑D immunoglobulin administration on its results, the external assessment group did a simulation study. The following assumptions were made:

4.14 The results of the simulation study, summarised in table 2, showed that using NIPT for fetal RHD genotype leads to a substantial reduction in RAADP use, from 99% of D‑negative women to 65.9%. This decline is similar in size to that seen by Soothill et al. (2015). The decrease is because of the drop (from 39% to 5.7%) in women with D‑negative fetuses needlessly having anti‑D immunoglobulin. Using NIPT for fetal RHD genotype means that about 1.2% of women miss having possibly beneficial RAADP, compared with 0.6% when using a universal RAADP approach with no testing.

Abbreviation: NIPT, non invasive prenatal testing.

4.15 Assuming all women still have postpartum cord blood typing and postpartum anti‑D prophylaxis if needed, the simulation study showed that NIPT would result in about 3 extra sensitisations per 100,000 women. If cord blood typing is not done, there would be about 13 extra sensitisations per 100,000 women. These increases are small compared with the total number of sensitisations because of anti‑D immunoglobulin failure and non‑adherence to anti‑D immunoglobulin treatment (around 281 per 100,000 women), and compared with not using RAADP at all (around 641 per 100,000).

4.16 Results of the simulation study also showed that using NIPT for fetal RHD genotype is unlikely to have any meaningful effect on mortality in later pregnancies; if women with a negative NIPT result never have postpartum anti‑D prophylaxis, there would be about 5 extra fetal or neonatal deaths per 1 million D‑negative women. In current practice, there are an estimated 86 fetal or neonatal deaths per 1 million D‑negative women.

4.17 Twelve studies were identified in a review of implementation of NIPT for fetal RHD genotype. Most studies reported that NIPT for fetal RHD genotype was feasible. Several studies reported potential issues relating to implementation, such as adherence to the anti‑D prophylaxis programme. Some studies highlighted the importance of short transport times for samples and effective management of transporting samples. The need for greater awareness of NIPT among physicians and midwives was also identified in some studies.

4.18 A UK-based survey (Oxenford et al. 2013) showed that, although most of the women surveyed supported the implementation of NIPT, their current knowledge of Rh blood groups and anti‑D treatment was limited, which could be a barrier to implementation.

4.19 Seven studies were identified in a review of existing studies on the cost effectiveness of high-throughput NIPT to determine fetal RHD genotype in pregnant women who are D negative and are not sensitised to the D antigen. The quality of the included studies' findings was uncertain because they did not report the validity of the diagnostic accuracy outcomes used. The degree of uncertainty in the cost-effectiveness estimates was also difficult to establish.

4.20 Results across the existing economic studies were conflicting. Only 1 study found NIPT for targeting RAADP to be cost saving compared with non‑targeted RAADP. Two studies reported that NIPT for fetal RHD genotype was cost saving compared with no RAADP, that is, compared with postpartum anti‑D prophylaxis only. Three studies reported that NIPT for fetal RHD genotype was not cost effective or was of no economic benefit. Only 1 study directly related to the UK (Szczepura et al. 2011).

4.21 The external assessment group developed a de novo economic model designed to assess the cost effectiveness of high-throughput NIPT to determine fetal RHD genotype in pregnant women who are D negative and are not sensitised to D antigen.

4.36 Key assumptions made in the model were:

4.37 Results show that all NIPT strategies cost less, but are less effective than the comparator, current clinical practice (table 5). Strategies PP1 and PP3 are associated with smaller quality-adjusted life year (QALY) losses than PP2 and PP4. This is because in both PP1 and PP3, cord blood typing is used to identify false-negative results, which would allow women who had been incorrectly identified as having a D‑negative baby, and so had not been offered RAADP, to have postpartum anti‑D prophylaxis. This would reduce the number of sensitisations, therefore reducing QALY losses.

Abbreviations: ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted life year; RAADP, routine antenatal anti D prophylaxis.

4.38 The variations in costs between the 4 strategies were mainly driven by different postpartum testing costs and postpartum anti‑D prophylaxis costs. The added cost of managing sensitisations and their associated health consequences in later pregnancies was largest for the strategies with more sensitisations (PP2 and PP4), and was very small for strategies PP1 and PP3.

4.39 In the fully incremental analysis of NIPT for fetal RHD genotype for the different postpartum testing strategies, PP3 and PP4 were dominated. Strategy PP4 was dominated by strategy PP2 because it had the same number of QALYs but was more expensive than PP2. Strategy PP3 was dominated by strategy PP1 because it had the same number of QALYs but was more expensive than PP1.

4.40 The cost-effectiveness acceptability curve showed that PP1 had the highest probability of being cost effective, with 0.65 and 0.73 for maximum acceptable incremental cost-effectiveness ratio (ICER) values of £20,000 and £30,000 per QALY gained respectively. For the same maximum acceptable ICER values, the probability of PP2 being cost effective was 0.30 and 0.22 respectively.

4.41 Sensitivity analyses showed that the results of the economic model are robust to small changes in the clinical effectiveness of RAADP, the timing of testing (between 11 and 23 weeks) and adherence to anti‑D immunoglobulin treatment.

4.42 A sensitivity analysis was done on the diagnostic accuracy of NIPT. When the diagnostic accuracy of NIPT was based on the meta-analysis of all studies rather than the Bristol studies alone, specificity increased by 2%, sensitivity decreased by 0.2%, the total cost across all NIPT strategies reduced, but total QALYs were only marginally affected. PP1 and PP3 remained the most cost-effective strategies.

4.43 In a sensitivity analysis on the rates of inconclusive results, NIPT became less cost effective as the rate of inconclusive results increased, but strategies PP1 and PP3 always remained more cost effective than current practice. When the rate of inconclusive results was low, PP3 became the most cost-effective strategy. When the rate of inconclusive results was high, PP1 became the most cost-effective strategy.

4.44 A 2‑way sensitivity analysis was done on test and treatment costs. The unit cost of NIPT is subject to uncertainty because it depends on throughput (the annual total number of samples analysed) and the level of the royalty fee. Similarly, the cost of anti‑D immunoglobulin may differ from the list price depending on negotiated discounts. The results of a 2‑way analysis on these unit costs showed that the base case is very sensitive to both the price of NIPT and the price of anti‑D immunoglobulin. A small increase in price of high-throughput NIPT or a small fall in the price of anti‑D immunoglobulin would result in current practice becoming more cost effective than NIPT strategies.

4.45 The cost of high-throughput NIPT for fetal RHD genotype was uncertain when this diagnostics guidance was written because it is highly dependent on the number of tests processed. The external assessment group did a threshold analysis to identify the point at which the test would move from being considered cost effective to being considered not cost effective, using a maximum acceptable ICER of £20,000 per QALY gained. Results show that raising the cost for high-throughput NIPT to £24.64 or more would result in current practice becoming more cost effective than NIPT strategies.

4.46 A sensitivity analysis was done on the cost of fetomaternal haemorrhage testing. Reducing the cost of a fetomaternal haemorrhage test to £3.17 (Szczepura et al. 2011; updated to 2015 prices) halved the estimated total costs of all strategies when compared with the total costs of the base-case scenarios, with total QALYs remaining similar to base-case results. When the cost of fetomaternal haemorrhage test was reduced, PP2 and PP4 became less cost effective than current practice, whereas PP1 and PP3 remained more cost effective compared with current practice.

4.47 An alternative postpartum-testing strategy to those included in the scope was assessed. The strategy separated women in whom NIPT identified a D‑positive fetus from women in whom NIPT gave an inconclusive result (and were therefore treated as if the fetus was D‑positive). Cord blood typing was done for women identified as having either a D‑negative fetus by NIPT or who had an inconclusive NIPT result, but not done for women in whom NIPT indicated a D‑positive fetus, and resulted in total costs of £15,230,372 and £2,433,756 QALYs per 100,000 pregnancies. This postpartum approach dominated all other NIPT strategies, and the ICER for this strategy compared with current practice was £1,638,356 saved per QALY lost.