Noninvasive methods of detecting fetal anaemia: a systematic review and meta-analysis

Authors

  • SJ Pretlove,

    1. Fetal Medicine Centre, Birmingham Women’s Foundation Trust, Edgbaston
    Search for more papers by this author
  • CE Fox,

    1. School of Clinical and Experimental Medicine (Reproduction, Genes and Development), College of Medical and Dental Sciences, University of Birmingham, Birmingham, UK
    Search for more papers by this author
  • KS Khan,

    1. School of Clinical and Experimental Medicine (Reproduction, Genes and Development), College of Medical and Dental Sciences, University of Birmingham, Birmingham, UK
    Search for more papers by this author
  • MD Kilby

    1. Fetal Medicine Centre, Birmingham Women’s Foundation Trust, Edgbaston
    2. School of Clinical and Experimental Medicine (Reproduction, Genes and Development), College of Medical and Dental Sciences, University of Birmingham, Birmingham, UK
    Search for more papers by this author

Dr SJ Pretlove, Birmingham Women’s Hospital, Metchley Park Road, Edgbaston, Birmingham B15 2TG, UK.
Email sampretlove@doctors.net.uk

Abstract

Objectives  To estimate in a systematic review of the literature the diagnostic value of ultrasound and Doppler blood flow velocity in the evaluation of fetal anaemia.

Study selection and data extraction  Literature from 2000 to 2008 was identified using MEDLINE and EMBASE, the Cochrane Library and relevant specialist register of the Cochrane Collaboration, and by checking reference lists of known primary studies and review articles. Studies were selected if the accuracy of the fetal ultrasound parameters or Doppler studies of blood flow in the fetal vessels was estimated compared with a reference standard. Data from the selected studies were abstracted as 2 × 2 tables comparing the diagnostic test result with the reference standard. Results were pooled where appropriate. Diagnostic accuracy was expressed as likelihood ratios.

Results  Twenty-five primary studies were identified containing suitable data on middle cerebral artery Doppler peak systolic velocity (MCA-PSV). The largest group of studies whose data could be pooled containing nine studies gave a positive likelihood ratio of 4.30 (95% CI: 2.50 to 7.41) and a negative likelihood ratio of 0.30 (95% CI: 0.13 to 0.69) for 675 cases in detecting severe anaemia in the analysis.

Discussion  Although middle cerebral artery peak systolic velocity Doppler has limited diagnostic accuracy, it remains the gold standard for noninvasive screening of fetal anaemia.

Introduction

Classically, the severity of maternal red cell alloimmunisation, causing Rhesus disease has been evaluated in the fetus using invasive methods such as amniocentesis and fetal blood sampling to determine fetal anaemia either directly or indirectly. Invasive testing carries an intrinsic risk of miscarriage, fetal loss or amniorrhexis and associated premature labour. Over the last twenty years, interest has been shown in using ultrasound parameters or Doppler studies of fetal blood to assess fetal anaemia. The ideal test would be able to detect severe anaemia with 100% sensitivity and specificity as this would avoid unnecessary invasive procedures. The ideal test would also not been influenced by previous intrauterine transfusion or aetiology of the anaemia. This publication is designed to update a systematic review of noninvasive methods of determining fetal anaemia initially performed in 2000.1 The studies reviewed contained data on fetal spleen perimeter,2 ductus venosus Doppler,3 thoracic aorta Doppler4–6 and umbilical vein7–9 and artery6 studies. It also contained one study on middle cerebral artery Doppler peak systolic velocity10 and while other tests have been discarded, middle cerebral artery Doppler has taken over and has replaced amniocentesis as a screening test for fetal anaemia in high risk pregnancies. There are now a plethora of individual cohort studies and a small number of randomised controlled trials investigating noninvasive testing evaluating different test thresholds and differing cutoffs for fetal anaemia. Naturally such a diversity of diagnostic investigation has indicated differing estimates of accuracy. Therefore, we performed a systematic review of the literature to evaluate middle cerebral artery Doppler peak systolic velocity for detecting fetal anaemia.

Methods

We developed a protocol using widely recommended methods for systematic reviews of observational studies for this study.

Identification of studies

Search and study selection

The search focused on capturing studies in which the diagnostic technique for determining fetal anaemia was compared with the reference standard.

A search strategy was developed based on existing advice for observational studies.11,12 Medline and EMBASE (January 2000 to October 2008) and the Cochrane library (2008:3) were searched electronically using OVID. Papers from the last eight years were chosen, as a previous systematic review on fetal anaemia was performed in 2000.1 The search of Medline and EMBASE captured citations combining relevant MeSH, key words and word variants for the concept ‘fetal anaemia’ as well as using test descriptions such as ‘middle cerebral artery Doppler’. The search strategy was designed by a specialist librarian and the authors (SJP). In addition, bibliographies of relevant articles were manually searched and experts in the speciality were contacted to identify papers not captured by electronic searches. The search was not restricted by language. Papers not in English were translated by healthcare professional whose first language was that of the paper. Where queries existed regarding the data, the original authors were contacted where possible.

Studies were selected in a two-stage process. First, all abstracts or titles in the electronic searches were scrutinised (SP) and full manuscripts of potentially eligible citations were obtained. Second, studies meeting our predefined criteria were selected. Studies were selected if data on fetal anaemia could be extracted and compared with a reference standard and sufficient data were provided to compute 2 × 2 tables. This was performed by two independent reviewers (SJP and CEF) and a third reviewer (KSK) was available for discussion if a definitive answer could not be reached. Studies without reference standards were excluded as these do not help assess for diagnosis. Where studies reported reference ranges, not reported in other studies, these have been included in the results tables but in the meta-analyses.

Data extraction and study quality assessment

The quality12,13 of all selected papers was assessed for the following attributes: i) Population, ii) intervention and iii) outcome.

Population selection was considered adequate if the sample was prospectively recruited consecutively, rather than on the basis of convenience. Where the article did not clearly state how the population was selected, this has been described as ‘unreported’. The population was considered adequate if it was comprised of women who were at risk of ‘fetal anaemia’ from any aetiology.

The description of the diagnostic intervention was considered adequate if it was reported in sufficient detail or referenced in such a way that reproduction of the technique could be performed in other studies.

Fetal haemoglobin estimates from fetal blood samples or from postpartum cord samples were considered adequate. The cutoffs were analysed and studies were grouped according to the reference standards used.

Data synthesis

Information from each study, stratified by intervention was tabulated by reference standard. Data in individual studies were expressed as positive and negative likelihood ratios with 95% confidence intervals (95% CI). A correction factor of 0.5 was used when the data or a study included a 0 value to allow calculation of the positive and negative likelihood ratio and its confidence interval.14 The likelihood ratio represents the ratio of the probability of a positive (or negative) test results in alloimmunised women with a severely anaemic fetus to the probability of the same test result in those women without a severely anaemic fetus. Likelihood ratio indicates by how much a given test result will raise or lower the probability of having the disease. Heterogeneity in rates was examined graphically and statistically. For graphical assessment, ‘forest’ plots of positive and negative likelihood ratio and their 95% CI’s were used. A random effects model15 was used in the light of heterogeneity.

We combined results for meta-analysis wherever appropriate using MetaDiSc 1.4 (available on Biomed Central: http://www.hrc.es/investigacion/metadiscen.htm).

Results

The electronic search yielded a total of 1762 citations (Figure 1). On examination of titles and abstracts, 79 were found to be since the previous systematic review in 20001 and relevant to the review and full papers were obtained. Neither discussion with experts nor a hand search of the references of the primary articles retrieved or specialist reviews/publications revealed any further studies. No relevant systematic reviews since 2000 were identified. Twenty-five studies10,16–39 with 1639 participants met our inclusion criteria for middle cerebral artery Doppler peak systolic velocity and seven studies2,25–27,35,40–42 contained information on tests other than MCA-PSV, some containing data on more than one test. Study characteristics of studies with data on MCA-PSV are shown in Table 1. Tests other than MCA-PSV have not been included in this review as there was either too little data to warrant meta-analysis, for example, cardio-femoral index, or the tests have been discarded by common consensus because of their lack of diagnostic accuracy, for example, liver length and spleen perimeter.

Figure 1.

 Study selection process.

Table 1.   Study characteristics
StudyGestationDesignPatient selectionRhesus diseaseKellOtherTechnique descriptionReference standard
(a) Papers containing data for MCA-PSV (test threshold 1.5 MoM) for severe anaemia (<0.55 MoM)
Rimon et al.1620–38ProspectiveUnreported080Mari 2000FBS or neonatal sample
Lubusky et al.1717–38ProspectiveUnreported176290 (no disease)Mari 2000FBS or neonatal sample
Collins and Ott18Not statedRetrospectiveConsecutive2117 parvovirus
5 hydrops
2 not stated
Teixeira 2000Fetal or neonatal haematocrit
McLean et al.1918.4–36.5RetrospectiveConsecutive20 (may also have other antibodies)019Mari 2000Neonatal Hb
Alshimmiri et al.2016–40ProspectiveConsecutive16 (D = 14, c = 1, D + E = 1)4 (Kell = 3, Kell + Jka = 1)0Described not referencedHb on fetal blood samples
Pereira et al.2122–34RetrospectiveConsecutive17110Mari NEJM 2000Fetal blood sampling
Zimmerman et al.2216–40ProspectiveConsecutive110 total
63 Anti-D only (3 Rhesus + Kell)
105Mari US O&G 1995Fetal blood sampling
Mari et al.2319–31ProspectiveConsecutive16 total
10 Anti-D alone
1 Rhesus + other
20Mari NEJM 2000Fetal blood sampling
Mari et al.1018–40ProspectiveConsecutive110 total
77 Anti-D alone
141 DuffyDescribedFetal blood sampling
(b) Papers containing data for MCA-PSV (test threshold 1.5 MoM) for severe anaemia (<−5 SDs)
Oepkes et al.2417–38ProspectiveConsecutive16500Mari NEJM 2000FBS or Hb at birth
Van Dongen et al.2521–30ProspectiveConsecutive0216Mari NEJM 2000FBS
Sikkel et al.2618–36ProspectiveConsecutive46140Described, not referencedFetal haematocrit from fetal blood sampling
Dukler et al.2722–40ProspectiveUnreported720Mari NEJM 2000Fetal blood sampling
Nishie et al.28Not statedProspectiveUnreported17 Anti-D only
9 Anti-D and Anti-c
2 Anti-D and others
Other s excludedOthers excludedMari NEJM 2000Fetal blood sampling
Amniotic fluid optical density at 450 nm
StudyGestationDesignPatient selectionRhesus diseaseKellOtherTechnique descriptionMCA thresholds usedReference standard?Reference standard
(c) Papers containing data for MCA-PSV (variable test threshold) for anaemia (variable cutoff)
Ochoa-Torres et al.3318–35Not statedNot stated1200Described not referenced>1.29 MoMCordocentesisSevere anaemia <0.55
Bullock et al.3220.2–38RetrospectiveConsecutive3800Mari NEJM 2000>1.5 MoMFBS<5th centile for gestational age
Hernandez-Andrade et al.3420–32UnreportedUnreported0016 non-immune hydrops
7 parvovirus
1 a-thal
1 primary cardiomyopathy
1 unexplained marrow suppression
7 idiopathic
Vyas45; Mari NEJM 2000>1, 1.5 & 2 SDsCordocentesisNicolaides Lancet4
Scheier et al.2920–40ProspectiveUnreported5800Described not referenced>0.5, 1, 1.5 &2 SDsFBS & cord blood at delivery>2 SDs, >4 SDs, >6 SDs.
Deren and Onderoglu3020–36ProspectiveUnreported52 non hydropic Rhesus D cases00Based on Mari 2000 and Bahado-Singh41,42>1.35 & 1.45 MoMCordocentesisSevere anaemia = <0.60 MoM
 Constructed own median
Haugen et al.3519–30ProspectiveUnreported34 total
16 Anti-D alone
1 Rhesus + Kell
01Oepkes 1993 FBS or neonatal HbHb deficit >2 SD = anaemia
Hb deficit >5 SD = severe anaemia
Teixeira et al.3126 fetuses
gestational age = 15–36 weeks
(1 was hydropic)
ProspectiveConsecutive24 Anti D
2 Anti C
00Unreferenced explanation in the text>1 &2 SDs above meanCordocentesisHaematocrit <1 SD and <2 SD8
StudyGestationDesignPatient SelectionRhesus diseaseKellOtherTechnique descriptionMCA thresholds usedReference standard?Reference standardTransfused
(d) Papers with variable cutoff and previous transfusions
Carbonne et al.38GestationMCAs performednot statedProspectiveConsecutive38 Rhesus D
 6 c
 (some were associated with C and E)
60Mari NEJM 2000>1.5 MoMHb < 0.55 MoMCordocentesisMixed Unable to separate results
Scheier et al.3920–34RetrospectiveNot stated‘Alloimmunisation’Not statedNot statedDescribed not referenced>1.5 MoMHb deficit >6 g/dLFBSAll had 1 previous transfusion
Delle et al.3718–36ProspectiveUnreported30 ‘alloimmunisation’010 parvovirusMari NEJM 2000>1.5 MoMHb < 0.55 MoMFBSMixed unable to separate results
Detti et al.3619–36
All with previous transfusions
ProspectiveUnreported53 Anti-D (5 associated with another antigen)
5 C (4 with D, 1 with e)
2 c
1 E (associated with Kell)
1 e (associated with C)
(Patients included more than once)
(Total 64 patients)
8 Kell (1 associated with E)1 KiddMari NEJM 2000>1.69 MoMHb < 0.55 MoMFBSAll had 1 previous transfusion

Study quality assessment (Table 1) of the twenty-five studies examining middle cerebral artery Doppler peak systolic velocity demonstrated good quality studies. Eight of the studies met all four high quality criteria.20,23–26,31,38,43

Fourteen studies could be pooled in the meta-analysis in two groups. When a test threshold for MCA-PSV of 1.5 multiples of median (MoM) and a cutoff for severe fetal anaemia of haemoglobin ≥0.55 MoM was used, this gave a positive likelihood ratio of 4.30 (95% CI: 2.50 to 7.41) and a negative likelihood ratio of 0.30 (95% CI: 0.13 to 0.69) for 675 cases in detecting severe anaemia. When a test threshold for MCA-PSV of 1.5 MoM was used but the cutoff for severe fetal anaemia was expressed as a haemoglobin deficit of ≥5 SDs, the results were similar with a positive likelihood ratio of 2.45 (95% CI 1.12 to 5.37) and a negative likelihood ratio of 0.35 (95% CI 0.13 to 0.95) for 296 patients in detecting severe fetal anaemia (Figure 2). The results are tabulated in Table 2(a–d).

Figure 2.

 Positive and negative likelihood ratios of MCA-PSV in detection of severe anaemia according to definition of severe anaemia (see text).

Table 2.   Results for middle cerebral artery Doppler peak systolic velocity as a noninvasive screening test for fetal anaemia using a cutoff of 1.5 MoM for the test and a definition of severe anaemia of Haemoglobin (a) <0.55 MoM; (b) >5 SDs. Results for middle cerebral artery Doppler peak systolic velocity as a noninvasive screening test for fetal anaemia (c) with test and anaemia thresholds as stated; (d) in where in-utero transfusion has been performed with test and anaemia thresholds as stated
StudyTPFPFNTNPositive LR (95% CI)Negative LR (95% CI)Sensitivity (95% CIs)Specificity (95% CIs)
(a)
Rimon et al.1670013.75 (0.34 to 41.63)0.08 (0.01 to 1.37)100 (59 to 100)100 (2.5 to 100)
Lubusky et al.1712013288277.88 (16.93 to 4560.90)0.52 (0.36 to 0.75)48 (27.8 to 68.7)100 (98.7 to 100)
Collins and Ott18443254.14 (1.36 to 12.61)0.50 (0.21 to 1.18)57.1 (18.4 to 90.1)86.2 (68.3 to 96.1)
McLean et al.19160354.84 (1.67 to 14.10)0.30 (0.03 to 3.27)100 (2.5 to 100)85.3 (70.3 to 94.4)
Alshimmiri et al.20107124.33 (0.20 to 94.83)0.87 (0.64 to 1.18)12.5 (0.3 to 53)100 (73.5 to 100)
Pereira et al.2131054.08 (0.94 to 17.69)0.16 (0.01 to 2.19)100 (29.2 to 100)83.3 (35.9 to 99.6)
Zimmerman et al.228201894.84 (3.06 to 7.66)0.14 (0.02 to 0.87)88.9 (51.8 to 99.7)81.7 (73.1 to 88.4)
Mari et al.2276051.73 (1.00 to 3.00)0.14 (0.01 to 2.14)100.0 (59.0 to 100.0)45.5 (16.8 to 76.6)
Mari et al.1031160644.83 (3.14 to 7.45)0.02 (0.00 to 0.31)100 (88.8 to 100.0)80.0 (69.6 to 88.1)
Total7453245244.30 (2.50 to 7.41)0.30 (0.13 to 0.69)75.5 (65.8 to 83.6)90.8 (88.2 to 93.0)
(b)
Oepkes et al.2465169755.00 (3.18 to 7.86)0.15 (0.08 to 0.27)87.8 (78.2 to 94.3)82.4 (73 to 89.6)
van Dongen et al.25161288.00 (1.25 to 51.14)0.13 (0.0 3 to 0.47)88.9 (65.3 to 98.6)88. (51.8 to 99.7)
Sikkel et al.262562181.26 (0.65 to 2.45)0.80 (0.46 to 1.39)54.3 (39 to 69.1)57.1 (28.9 to 82.3)
Dukler et al.27601000.75 (0.10 to 5.88)1.25 (0.17 to 9.20)37.5 (15.2 to 64.5)Not applicable (0 to 100)
Nishie et al.28963102.00 (0.98 to 4.08)0.40 (0.14 to 1.14)75 (42.8 to 94.5)62.5 (35.4 to 84.8)
Total cases12129451012.45 (1.12 to 5.37)0.35 (0.13 to 0.95)0.73 (0.66 to 0.80)0.78 (0.69 to 0.84)
StudyHbTest cutoffTPFPFNTNPositive LR (95% CI)Negative LR (95% CI)Sensitivity (95% CIs)Specificity (95% CIs)
(c)
Ochoa-Torres et al.33>0.55 MoM1.29 MoM11011223.00 (1.51 to 350.92)0.12 (0.03 to 0.54)91.7 (61.5 to 99.8)100 (73.5 to 100)
Bullock et al.32<5th centile1.5 MoM1438133.39 (1.16 to 9.87)0.45 (0.25 to 0.82)64.6 (40.7 to 82.8)81.3 (54.4 to 96)
Hernandez-Andrade et al.344 SDs1.5 MoM61086.19 (1.40 to 27.46)0.08 (0.01 to 1.23)100 (54.1 to 100)88.9 (51.8 to 99.7)
Scheier et al.29>6 SDs1.5 MoM2251306.70 (2.96 to 15.14)0.05 (0.01 to 0.35)  
Deren and Onderoglu30>0.60 MoM1.5 MoM1923209.50 (2.51 to 35.99)0.15 (0.05 to 0.43)85 (65.1 to 97.1)90.0 (70.8 to 98.9)
Haugan et al.35<5 SDs95% prediction rate230308.10 (2.66 to 24.66)0.20 (0.02 to 2.34)100 (15.8 to 100)90.9 (75.7 to 98.1)
Teixeira et al.314z scores2 SDs422166.00 (1.45 to 24.91)0.38 (0.12 to 1.18)66.7 (22.4 to 95.7)55.9 (65.3 to 98.6)
Total  56111499    
(d)
Carbonne et al.38<0.551.5 MoM214162675.84 (4.121 to 8.28)0.256 (0.13 to 0.52)77.8 (57.7 to 91.4)86.4 (82.4 to 90.3)
Scheier et al.39Deficit of >6 g/dL1.5 MoM1291213.14 (1.83 to 5.39)0.06 (0.00 to 0.85)100.0 (73.5 to 100.0)70.0 (50.6 to 85.3)
Delle et al.37<0.551.5 MoM841113.33 (1.40 to 7.96)0.15 (0.02 to 0.099)88.9 (51.8 to 99.7)73.3 (44.9 to 92.2)
Detti et al.36<0.551.69 MoM110203324.44 (1.50 to 397.86)0.65 (0.50 to 0.85)100 (71.5 to 100)63 (47.9 to 75.2)
Total  404527311    

Nine studies10,20,22–26,31,38 giving data on MCA-PSV met all four high quality criteria. Seven of these could be pooled in two groups of three and four studies determined by the definition of severe anaemia. Two studies could not be pooled.31,38 When anaemia was defined as Hb < 0.55 MoM, the pooled positive likelihood ratio was 3.83 (95% CI: 2.41 to 6.09) and the negative likelihood ratio was 0.13 (95% CI: 0.08 to 2.80). When anaemia was defined as Hb >5 SDs, the pooled positive likelihood ratio was 3.21 (95% CI 1.04 to 9.90) and the pooled negative likelihood ratio was 0.28 (95% CI: 0.08 to 1.01).

Discussion

Principal findings

This publication is designed to update a systematic review of noninvasive methods of determining fetal anaemia initially performed in 2000.1 Middle cerebral artery peak systolic velocity is most widely assessed with the most data available. Meta-analysis of this screening test for fetal anaemia found the most reliable when sensitivity and specificity are determined against the ‘gold standards’ of fetal blood sampling or neonatal haemoglobins. However, spleen perimeter, liver length and umbilical vein maximum velocity have low sensitivities and specificities when assessed against the gold standard of fetal blood sampling. Cardiofemoral index, splenic artery peak systolic velocity and middle cerebral artery deceleration angle all demonstrate promise, but data informing their use is from relatively small cohort studies (n = 1 for each).

The validity of the findings of this review are dependent on the rigour of its methodology.11 A prospective protocol was used and a concerted effort was made to identify all the available evidence. It met quality criteria laid down in the MOOSE statement.11 Additionally, data was available for 2163 women making this a large cohort. Hence, this systematic review provides the most reliable evidence available on noninvasive methods of assessment of fetal anaemia.

The methodological quality of the included studies was good. Although there was some heterogeneity in cutoff values between the studies, all studies described the technique used for the test adequately and defined an appropriate reference standard. The majority of studies were performed prospectively with a representative population. Where there was heterogeneity in the cutoff values, studies were only grouped for meta-analysis when sufficient data in each group made this viable. When only high-quality studies were analysed, there was no marked difference in the pooled sensitivity and specificity. Studies were performed in a variety of centres worldwide with these tests performed by many different operators.

This systematic review provides compelling evidence that the positive and negative likelihood ratios of MCA-PSV are some way from those of an ideal test. While ultrasonographic methods of determining fetal anaemia appear to hold great promise when compared with invasive methods and the complications associated with them, they need to be interpreted with care in the clinical situation.

Screening typically involves use of a confirmatory test after initial testing, prior to institution of therapy. In this review, the initial screening test is the noninvasive ultrasonographic parameter, but the confirmatory test is the invasive fetal blood sampling ideally performed at the time of intrauterine transfusion. While MCA-PSV is regarded as the best noninvasive screening tool for severe fetal anaemia, it did not generate a positive likelihood ratio greater than ten or a negative likelihood ratio of <0.1 These levels were chosen because it is at these levels that a test is deemed clinically useful (i.e. substantially increases or lowers the pre-test probability and hence ruling in or excluding the disease respectively). We would describe the likelihood ratios, both positive and negative, achieved in this meta-analysis by MCA-PSV as weakly accurate.44 Given the consequences of false positive results (both of morbidity to the fetus from invasive procedures and cost of performing invasive testing), it is important that the positive likelihood ratio of the test is suitably high. Given the consequences of false negative results (both costs and morbidity of cases left untreated), it is important that the negative likelihood ratio of the test is also suitably high. This is because erroneously withholding effective interventions from falsely negative cases leads to excessive morbidity when disease develops later in pregnancy. If available effective interventions are convenient, inexpensive and without particular risk of harm or adverse effects (to both mother and fetus), it is better to have a high positive likelihood ratio.

The sensitivity and specificity of MCA-PSV are markedly similar to that of OD450 from amniocentesis.32 This was previously noted in our comparative cohort study published in 2004,32 noting similar ROC for both of these screening tests. MCA-PSV is chosen as superior as it is noninvasive and therefore carries little fetal risk. Ideally, a large multi-centre randomised trial of MCA-PSV versus amniocentesis as a screening tool for fetal anaemia could be undertaken to ensure that the most appropriate screening tool is being used. However, it would be difficult to justify this ethically with the data currently available. Further work to develop test combinations to improve diagnostic accuracy and to further assess tests post intrauterine transfusion should be undertaken.

Conflict of interest

None known.

Contribution to authorship

Professor Kilby formulated the idea for the systematic review, checked the publications for the systematic review and oversaw the project. Professor Khan advised on the technical details of performing a systematic review and meta-analysis. The systematic review was undertaken by Dr Pretlove and counter checked and corroborated by Dr Fox. All authors were involved in the preparation and writing of the manuscript.

Funding

None.

Acknowledgements

Mary Publicover, Birmingham Women’s Hospital Librarian helped form the search strategy.

Ancillary