The Percutaneous shunting in Lower Urinary Tract Obstruction (PLUTO) study and randomised controlled trial: evaluation of the effectiveness, cost-effectiveness and acceptability of percutaneous vesicoamniotic shunting for lower urinary tract obstruction

Systematic review of the effectiveness of antenatal interventions in lower urinary tract obstruction

We have undertaken a systematic review of the effectiveness of antenatal intervention (VAS, vesicocentesis, fetal cystoscopy, open procedures), building on a previously published systematic review,13 using methodological advances in search strategies, quality assessment and statistical analysis. 19,20 Authors of papers in the original review were contacted to ensure that the correct and most complete data were obtained. The objective was to systematically review the literature to evaluate the effectiveness of antenatal interventions in improving perinatal survival and postnatal renal function in congenital LUTO. (This systematic review has been published in full21 and we summarise the methods and results here only.)

Extensive electronic searches (from database inception to 2009) were performed using medical subject headings (MeSH) and keywords, without restrictions. Reference lists of included studies were checked and all authors were contacted. Studies were selected according to the following criteria:

• population – fetuses with ultrasonographic evidence of LUTO (enlarged bladder, bilateral hydronephrosis, keyhole sign)

• intervention – bladder drainage by vesicocentesis, VAS, fetoscopic surgery (e.g. cystoscopy and ablation of valves, open fetal bladder surgery)

• outcome – perinatal mortality, measurement of renal function in survivors (e.g. serum creatinine, need for dialysis/transplantation), other morbidity indicators (e.g. need for ventilation)

• study design – RCTs, controlled and uncontrolled observational studies; case reports and case series with less than five cases were excluded.

Data on study design and quality and the results were extracted to construct 2 × 2 tables. The reporting of the study was judged according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement, incorporating assessment of aspects of study quality. 22 We performed meta-analysis to explore the effect of antenatal interventions on outcome according to predefined subgroups. The effect of any intervention compared with no treatment was computed for each outcome: overall survival, survival excluding voluntary termination of pregnancy (TOP), perinatal survival [excluding voluntary TOP and intrauterine death (IUD)] and survival with normal postnatal renal function. Subgroup analyses were performed according to predicted fetal prognosis (according to antenatal ultrasound features or fetal urinalysis), and comparing treatments with one another. Forest plots were constructed for each group. We inspected for heterogeneity visually and statistically, calculating the Cochran’s Q statistic and the I ² value. 19 An I ² value > 50% was felt to demonstrate significant heterogeneity between studies. 23 In the event, low levels of heterogeneity resulted in fixed-effects models being used throughout using Peto’s method.

A total of 20 articles5,6,9,24–40 including 369 fetuses were eligible for inclusion in the review. All fetuses included had ultrasound features suggestive of LUTO (i.e. enlarged fetal bladder with dilated proximal urethra with or without associated hydronephrosis). The reported gestational age at diagnosis ranged from 13 to 38 weeks. Of the 369 fetuses, 261 (71%) received an antenatal intervention intended to relieve the obstruction. In the majority of cases (87%) this was a percutaneous VAS. Nine fetuses underwent an open procedure (by maternal laparotomy and hysterotomy) including open shunt insertion, bladder marsupialisation or cutaneous ureterostomy; 26 underwent fetal cystoscopy and 14 of these had ablation of PUVs using laser fulguration, urethral stent or hydroablation. Vesicocentesis was considered a diagnostic or therapeutic procedure depending on the technique used. Fetuses were classified as having a good or poor predicted prognosis by the study authors based on fetal urinalysis results prior to intervention. The overall study quality was variable ( Figure 3 ). There were no eligible RCTs; included studies represented a combination of prospective and retrospective cohort studies. Over 80% of studies complied with the STROBE statement elements, describing study design, explanation of study size, participant eligibility criteria, patient characteristics and follow-up, and number of outcome events. 22 However, few studies made efforts to address bias and most were poor in their reporting of the technique used for the intervention, the overall results and the precision of the findings.

FIGURE 3.

Bar chart summarising the quality of evidence of papers included in the systematic review of the effectiveness of antenatal intervention in LUTO. Numbers in the bars are the numbers of studies. Reproduced from Morris RK, Malin GL, Khan KS, Kilby MD. Systematic review of the effectiveness of antenatal intervention for the treatment of congenital lower urinary tract obstruction. BJOG 2010;117:382–9021 © 2010 The Authors Journal compilation © RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology, with permission from John Wiley & Sons, Inc.

Meta-analysis of non-randomised observational studies ( Figure 4 ) demonstrated that prenatal bladder drainage for LUTO may improve perinatal survival (OR 3.82, 95% CI 2.14 to 6.84), particularly in those with a poor predicted outcome (identified by fetal urinalysis) (OR 26.19, 95% CI 4.39 to 156.25). However, the observational data suggested that survivors had a high residual risk of poor postnatal renal function ( Figure 5 ). VAS compared with no treatment had an OR of 3.86 (95% CI 2.00 to 7.45) for perinatal survival and an OR of 0.50 (95% CI 0.13 to 1.90) for survival with normal renal function. The conclusions of the systematic review were that the quality of evidence was poor, with heterogeneity in study design and follow-up, and there was the potential for bias in observational studies raising concerns about the validity of the results. 41

FIGURE 4.

Effect of antenatal intervention compared with no treatment on perinatal survival (including voluntary TOP) stratified by predicted prognosis. Reproduced from Morris RK, Malin GL, Khan KS, Kilby MD. Systematic review of the effectiveness of antenatal intervention for the treatment of congenital lower urinary tract obstruction. BJOG 2010;117:382–9021 © 2010 The Authors Journal compilation © RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology, with permission from John Wiley & Sons, Inc.

FIGURE 5.

Effect of antenatal intervention compared with no treatment on postnatal survival with normal renal function (excluding IUDs and voluntary TOP) stratified according to predicted prognosis. Reproduced from Morris RK, Malin GL, Khan KS, Kilby MD. Systematic review of the effectiveness of antenatal intervention for the treatment of congenital lower urinary tract obstruction. BJOG 2010;117:382–9021 © 2010 The Authors Journal compilation © RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology, with permission from John Wiley & Sons, Inc.

The observational evidence thus still justified the need to proceed with a RCT of this rare congenital malformation including long-term follow-up for renal function.

Systematic review of the accuracy of fetal urinalysis as a predictor of postnatal renal function

We systematically reviewed14 the evidence that obtaining a fetal urine sample before therapy by vesicocentesis and urinary analysis to measure sodium, calcium and β2-microglobulin concentrations may be helpful in identifying those babies most at risk of postnatal kidney damage. 42

The systematic review was conducted according to a protocol designed using widely recommended methods. 43–46 We searched MEDLINE (1966–2006), EMBASE (1980–2006), The Cochrane Library (2006, issue 2), the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (from inception to 2006), Medion, System for Information on Grey Literature in Europe (SIGLE), the Index of Scientific and Technical Proceedings, SciSearch^®, the National Research Register and the Medical Conferences Register for relevant citations. In MEDLINE the search consisted of a combination of MeSH (e.g. urethral obstruction, hydronephrosis, fetal diseases) ‘or’ keywords (e.g. enlarged bladder, congenital urinary tract obstruction, posterior urethral valves) for disease. These were combined using ‘and’ with MeSH and keywords for intervention (e.g. shunting) ‘or’ investigation (e.g. ultrasound and fetal urine analysis).

Papers were included if they satisfied the following criteria:

• population – fetuses with ultrasound evidence of congenital urinary tract obstruction

• index test – any test on fetal urine

• reference standard – any reference standard looking at postnatal renal function or renal dysplasia in non-survivors

• study design – test accuracy studies that allowed generation of 2 × 2 tables to compute indices of test accuracy; case series with less than eight cases were excluded.

All articles meeting the selection criteria were also assessed for methodological quality using items from the Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool. 47,48 A study was considered to be of good quality if it utilised a prospective design with consecutive recruitment, included full verification of the test result with a reference standard and had an adequate description of both the index test and the reference standard, thus reducing bias. 43,44,48–50 Sensitivity, specificity and likelihood ratios (LRs) were generated and pooled using meta-analytical methods.

A total of 23 articles5,24,29,39,40,42,51–67 met the selection criteria. All of the studies were observational studies and of poor quality ( Figure 6 ).

FIGURE 6.

Quality of studies included in the review of fetal urinalysis in fetuses with obstructive uropathy to predict poor postnatal renal function. Stack bar chart with numbers inside bars indicating the numbers of studies. Reproduced from Morris RK, Quinlan-Jones E, Kilby MD, Khan KS. Systematic review of accuracy of fetal urine analysis to predict poor postnatal renal function in cases of congenital urinary tract obstruction. Prenatal Diagnosis 2007;27:900–1114 Copyright © 2007 John Wiley & Sons, Ltd, with permission from John Wiley & Sons, Inc.

The two most accurate tests were calcium > 95th centile for gestation (LR+ 6.65, 95% CI 0.23 to 190.96; LR− 0.19, 95% CI 0.05–0.74) and sodium > 95th centile for gestation (LR+ 4.46, 95% CI 1.71 to 11.6; LR− 0.39, 95% CI 0.17 to 0.88). β2-microglobulin was found to be less accurate (LR+ 2.92, 95% CI 1.28 to 6.69; LR− 0.53, 95% CI 0.24 to 1.17). However, there was no individual urinary analyte or threshold that could be shown to be of particular clinical value ( Table 1 ). The conclusion was that the current evidence demonstrated that none of the analytes of fetal urine investigated so far can be shown to yield clinically significant accuracy to predict poor postnatal renal function.

Systematic review of the accuracy of antenatal diagnosis and ultrasound markers in predicting postnatal outcome

Data relating to the natural history of LUTO are difficult to identify. However, recent registry data from Finland68 have presented outcomes in babies with PUVs (n = 46), with 23 diagnosed prenatally (with no intervention) and 23 diagnosed postnatally (with apparently ‘normal’ ultrasound appearances in the majority). This cohort study is unique, in that long-term follow-up over 10 years was performed (mean 12.5 years, range 5.5–20.1 years). Despite postnatal treatment, 13% of this cohort developed chronic renal failure in childhood and adolescence with 17% developing end-stage renal disease. These rates are lower than those reported in the literature for antenatally diagnosed LUTO, possibly reflecting a different pathogenesis. There was no significant difference in the proportion of patients with primary or acquired renal dysplasia between those prenatally diagnosed and those postnatally diagnosed. There was also no difference in long-term renal outcome between those patients with a prenatal diagnosis and those with a postnatal diagnosis, with no difference in the mean age of advancing to end-stage renal disease. The mean age of achieving continence in both prenatal and postnatal diagnostic groups was not significantly different at 5 years of age. Such data indicate that the severity of the ultrasound appearances (in the postnatally diagnosed group at least two prenatal ultrasound examinations had been performed and demonstrated no anomaly) in LUTO have little bearing on the eventual long-term paediatric outcome. A limited amount of literature has reported on the effects of gestational age of < 24 weeks at diagnosis,5 associated prenatal ultrasound features of macrocystic/microcystic renal abnormality (indicative of dysplastic change)6,7 and significant oligohydramnios,69 demonstrating good predictive accuracy for poor outcome (but with considerable heterogeneity of results). No consensus exists at present as to the best prenatal ultrasound sign (or combination of signs) to predict postnatal renal function.

During the study period our group performed and published a systematic review of the literature on antenatal ultrasound to predict postnatal renal function in LUTO. 70 We searched MEDLINE (1966–April 2008), EMBASE (1980–April 2008), The Cochrane Library (2008, issue 4), CINAHL (from inception to 2008), Medion, SIGLE, the Index of Scientific and Technical Proceedings; SciSearch, the National Research Register and the Medical Conferences Register for relevant citations. In MEDLINE the search consisted of a combination of MeSH (e.g. urethral obstruction, hydronephrosis, fetal diseases) ‘or’ keywords (e.g. enlarged bladder, congenital urinary tract obstruction, posterior urethral valves) for disease. These were combined using ‘and’ with MeSH and keywords for intervention (e.g. shunting) ‘or’ investigation (e.g. ultrasound and fetal urine analysis).

Papers were assessed for inclusion using the following criteria:

1. population – fetuses with ultrasound evidence of congenital LUTO

2. diagnostic measure – any prenatal ultrasound parameter

3. outcome measure – any outcome measure looking at postnatal renal function or renal dysplasia in non-survivors

4. study design – RCTs or observational studies that allowed generation of 2 × 2 tables (true positives, false positives, false negatives and true negatives) to compute indices of test accuracy; case series with less than five cases were excluded.

Data were used to construct 2 × 2 tables of test accuracy using the antenatal ultrasound parameter reported in the paper and the authors’ definition of a positive or negative test and comparing these results to the postnatal renal function or outcome for each individual patient. All articles meeting the selection criteria were also assessed for quality using items from validated tools. 47,48,71 A study was considered to be of good quality if it utilised a prospective design with consecutive recruitment, included full verification of the diagnostic measure with the outcome measure and included an adequate description of both of these measures, thus reducing bias. 43,44,72 Sensitivity, specificity and LRs were generated.

Using the 2 × 2 tables we computed sensitivity (true positive rate), specificity (true negative rate) and the LRs73 (the ratio of the probability of the specific test result in people who do have the disease to the probability in people who do not) and 95% CIs for individual studies. Subgroup analysis with pooling of LRs was performed only when there were at least two studies with similar characteristics within that group. All statistical analyses were performed using Meta-DiSc 1.3 software (see www.hrc.es/investigacion/metadisc.html)74 with 0.5 being added to all cells in 2 × 2 tables and using a random-effects model. A p-value of < 0.05 was used throughout for statistical significance and the chi-squared test was used as a statistical test of heterogeneity. When there was still significant statistical heterogeneity, summary receiver operating characteristic curves were drawn and the area under the curve presented as the summary measure of accuracy.

The final data set included 13 articles,11,12,24,26,29,53,54,75–80 with a total of 215 women and 33 2 × 2 tables. All included studies were observational. The populations of included studies all consisted of fetuses with antenatally suspected LUTO that was subsequently confirmed postnatally. The thresholds used for the diagnostic measure varied between the studies as did the outcome measure. The papers all described investigation of a small number of subjects with imprecise results and did not report characteristics of patients or data collection. Meta-analysis was performed using subgroups with the same diagnostic measure and an outcome measure assessing renal function in survivors to minimise clinical heterogeneity ( Table 2 ). The ultrasound parameter that showed the best predictive value for postnatal renal function in survivors was renal cortical appearance, with a sensitivity of 0.57 (95% CI 0.37 to 0.76), a specificity of 0.84 (95% CI 0.71 to 0.94) and an area under the curve of 0.78.

We concluded that measurement of amniotic fluid volume and the appearance of the renal cortex at diagnosis showed promising predictive accuracy for poor postnatal renal function (see Table 2 ).

The need for a large simple trial of vesicoamniotic shunting compared with conservative management for lower urinary tract obstruction

On the basis of published evidence and evidence from specialist advisors, in 2006 NICE produced an interventional procedures guidance for fetal VAS (IPG 2028). The guidance states that the ‘current evidence on safety and efficacy of fetal vesico-amniotic shunting for LUTO does not appear adequate for this procedure to be used without special arrangements for consent and for audit or research’. The guidance thus recommends randomisation to PLUTO: ‘Clinicians are encouraged to enter patients into this trial or the associated registry’ (section 1.1). Thus, from the beginning of the HTA-funded trial, the national recommendation to clinicians was to use VAS only within the PLUTO trial.

Missing data

In 67 cases it could not be determined whether a prenatal diagnosis had been made. However, it was confirmed that no referral was made to the regional fetal medicine centre or local ultrasound centres nor was a prenatal notification sent to the WMCAR. Therefore, within the analysis of prenatal detection rates it was assumed that this group had no prenatal diagnosis. There were four cases (all isolated LUTO cases) in which the fetal sex could not be determined as this information was not included in the prenatal ultrasound report and all were terminated or delivered spontaneously before 20 weeks’ gestation and sex could not be determined macroscopically. These were included in the ‘isolated, non-female subgroup’ as they were likely to be male fetuses (however, we did not have cytogenetic confirmation of this fact).

Prevalence/statistical analyses

Birth prevalence is the preferred measurement when estimating the frequency of congenital anomalies. The calculation of birth incidence requires a precise definition of the denominator,100 including the number of all early spontaneous losses, and these data are not routinely recorded, thus leading to an underestimation of the denominator and consequent overestimation of the prevalence. The prevalence ratio at birth of the congenital anomaly of interest is therefore the number of affected births divided by the total number of live births and stillbirths. In studies in which mid-trimester loss data and TOP data are collected for specific anomalies, the denominator can be adjusted by including TOP data for the population at risk. However, there are no estimates of the total number of mid-trimester losses available routinely for population data. The denominator used is therefore a slight underestimate but it is the best possible denominator available. Numerator and denominator data were linked by maternal postcode at delivery to small-area deprivation scores based on the Index of Multiple Deprivation (IMD) 2004 (see http://data.gov.uk/dataset/imd_2004). Incidence rates were calculated for quintiles of deprivation.

We also present the prevalence of LUTO by maternal ethnic group, as defined at antenatal booking. The lack of population-based data on maternal ethnic group for the study period necessitates the use of estimates for denominator data. The distribution of all births by maternal ethnic group was generated by standardising the total number of births during the study using an ethnic distribution derived from census data. The ethnic distribution applied to the denominator was that of the population aged 0–4 years born in the UK derived from the 2001 census. 101 A sensitivity analysis was undertaken on the denominator to test significant findings.

A chi-squared test was used to examine the association between prevalence and maternal age and to examine secular trends. ORs and confidence limits were calculated using Microsoft Excel 2003 (Microsoft Corporation, Redmond, WA, USA) and trends were calculated within Epi lnfo version 6 (Centers for Disease Control and Prevention, Atlanta, GA, USA).

Aetiology of lower urinary tract obstruction

Table 5 shows the incidence of the different causes of LUTO. The most common underlying aetiology of LUTO was PUVs (63.0%).

Outcomes and survival

The outcomes of cases of LUTO are presented in Table 6 , separately for isolated and complex cases. TOP was undertaken in 24.6% of cases. In the complex and isolated groups, the proportion of pregnancies ending in TOP was 41.3% and 19.9% respectively. However, this will be influenced by prenatal detection rates and parental choice. For cases resulting in a live birth or stillbirth, the perinatal mortality rate was 458/1000 births for complex cases and 120/1000 births for isolated cases.

Table 7 demonstrates perinatal and infant mortality rates according to pathological diagnosis. Of those that survived, 67 had normal renal function (64 PUV, three urethral stenosis), 16 had mild renal impairment (all PUV), eight had chronic renal failure (all PUV) and five had end-stage renal failure (all PUV).

Pathology

The post-mortem rate in babies who died was 73.6% (92/125). For all outcomes (deaths and survivors) a diagnosis was confirmed by pathology (post-mortem or surgical) in 32.4% of cases (92/284).

Chromosomal anomaly

In total, 100 cases had a karyotype available and a chromosomal abnormality was noted in 16 cases ( Table 8 ), all classified in the complex group, which represents a prevalence of chromosomal anomaly of 5.6% for all LUTO cases or 16.0% for cases with a known karyotype. For complex LUTO cases, chromosome analysis was undertaken and was successful in 82.5% of cases and abnormal in 25.4%.

Prenatal detection

Table 9 demonstrates that prenatal diagnosis of LUTO by ultrasound was made in 50.7% of all cases. This rate was higher in complex cases (57.1%) than in isolated cases (48.9%), which would be expected as there was additional pathology. In a further 10.6% of cases coexistent anomalies of the renal tract were suspected and in 4.9% of cases there were abnormal ultrasound features including anomalies of other systems. In total, 66.2% (188/284) of LUTO cases had an abnormal prenatal ultrasound finding and in 76.6% (144/188) of diagnoses this was reported specifically as LUTO. The proportion of cases with any ultrasound anomaly (LUTO, renal, other) was higher in complex cases (87.3%) than in isolated LUTO cases (60.2%).

Over the period of the study there was no significant change in the proportion of cases with a prenatal diagnosis of any ultrasound anomaly (LUTO/renal/other system) (χ² = 2.63, p = 0.105). However, the proportion of cases with a specific prenatal diagnosis of LUTO did increase over time (χ² = 5.68, p = 0.017) ( Figure 8 ).

FIGURE 8.

Proportion of LUTO cases with a prenatal diagnosis 1995–2007. Reproduced from Malin GL, Tonks A, Morris RK, Gardosi J, Kilby MD. Congenital lower urinary tract obstruction: a population based epidemiology study. BJOG 2012;119:1455–64,97 © 2012 The Authors BJOG An International Journal of Obstetrics and Gynaecology © 2012 RCOG, with permission from John Wiley & Sons, Inc.

Table 10 presents pregnancy and infant outcomes according to gestation of diagnosis. Postnatally detected cases were more likely than prenatally detected cases to survive to 1 year of age (91% vs. 50% for isolated cases and 62.5% vs. 15% for complex cases, p < 0.001). There was no significant difference between postnatally detected cases and prenatally detected cases in the proportion of live-born babies who eventually developed end-stage renal failure.

False-positive prenatal diagnoses

There were 53 cases with an antenatal ultrasound suspicion of LUTO that were later excluded from the study cohort ( Table 11 ). This represents 26.9% of the prenatal diagnoses of LUTO (53/197). Of these 53 cases, 38 had bladder signs (25 enlarged bladder, 21 thick-walled bladder, eight keyhole sign), 18 had hydronephrosis (15 bilateral), 10 had oligohydramnios (reduced) or anhydramnios (complete absence of liquor) and 15 had an abnormal appearance of the renal cortex (eight bilateral echogenicity, four unilateral echogenicity, two unilateral macrocysts and one bilateral macrocysts). The most common final postnatal diagnoses in these false-positive cases were ultrasound finding associated with significant renal reflux (24.5%), cloacal dystrophy (18.9%) and hydronephrosis (11.3%). In 10 cases (20.8%) there was no renal/bladder anomaly at delivery and in five of these the appearance of LUTO on ultrasound was seen to resolve during the antenatal period (see Table 11 ).

Cases eligible for antenatal (in utero) vesicoamniotic shunting

When the cohort is restricted to isolated non-female singleton fetuses, there were 211 cases of LUTO (74.3% of all LUTO cases), a total incidence of 2.48 (95% CI 2.14 to 2.81) per 10,000 total births. However, the prenatal diagnosis rate of LUTO within this group was 46.9% (99/211), further reducing the proportion that could be referred for VAS. In this antenatally treated group, two fetuses underwent vesicocentesis, with both pregnancies ending in TOP. A total of eight fetuses were treated with VAS, of whom five had PUV, one had urethral atresia, one had urethral stenosis and one had unspecified LUTO. In the PUV group the outcome was one TOP, one IUD and one neonatal death, with two fetuses surviving, one with mild renal impairment and the other with renal failure. The baby with urethral stenosis survived with normal renal function and the baby with unspecified LUTO was a neonatal death.

In the apparently ‘isolated’ cohort with a specific prenatal diagnosis of LUTO, 33% of parents opted for TOP. Of those that survived to birth, primary resection of valves was performed in 92% with a diagnosis of PUVs at cystoscopy (29% of the total isolated LUTO cohort). For 45% this was within the first week postnatally and for 29% this was within the first month. Of these, 2% (a single case) required repeat cystoscopic resection and 50% were alive at 1 year. Of these cases, 69% had documented ‘normal renal function’ with 10% being classified as having chronic renal failure and 4% end-stage renal failure.

Principal findings

This cohort of pregnancies complicated by LUTO is the largest reported series in the UK. The study design examines a single, large, multiple-source, population-based congenital anomaly register that uses active case finding ensuring that ascertainment is high and incidence rates are reliable and unaffected by information bias.

The incidence of LUTO was reported as 3.34 (95% CI 2.95 to 3.72) per 10,000 total births (1 in 2994). LUTO incidence was found to be significantly associated with maternal ethnic group and deprivation, being highest in the most deprived quintile. This may explain the higher incidence seen in the West Midlands compared with other published series, such as that from NorCAS (a population-based register in the northern region) (2.2 per 10,000 total births 1984–97). 24 As there has been no change in incidence over time it is unlikely that the difference seen is due to the use of a later cohort in our study, and similarly it is unlikely to be due to any differences in the maternal age profile of the two regions. Both WMCAR and NorCAS are part of the BINOCAR, a network of UK regional anomaly registers that use the same methodology for data collection, generating good-quality data with high ascertainment rates,15,16,102,103 and so variations between registry area data are not likely to be the result of ascertainment bias. There is no longer national coverage for anomaly data for England; these data are available only at regional level. Regional register data have been shown to be better than previous national data. 103 It is recognised that as early spontaneous pregnancy losses are not recorded by regional anomaly registers this leads to an underestimate of the denominator and thus an overestimate of the incidence of LUTO. We have produced stratified incidence rates for ethnic group and deprivation that can be applied to other areas to estimate incidence.

As with other population-based series, these data from the WMCAR indicated that, of the 83% of complex prenatal cases for which fetal karyotyping was known, 5.6% were abnormal. In ‘isolated’ LUTO the most prevalent postnatally confirmed diagnosis was PUVs (72%) followed by urethral atresia/stenosis in 12% of cases. In 14% of cases the underlying pathology was not clearly delineated.

These data indicate that overall perinatal and infant survival are significantly worse if the congenital bladder neck obstruction is ‘complex’ (associated with other anomalies) and also if the diagnosis is made antenatally. In this series from the West Midlands (1995–2007) the prenatal detection rate using routine ultrasound screening was 66%. This is not significantly different from detection rates quoted for the northern region for a cohort between 1984 and 1997. 24 This indicates that, despite national recommendations for screening during a detailed ultrasound scan at 20 weeks from the Royal College of Obstetricians and Gynaecologists (see www.rcog.org.uk/womens-health/clinical-guidance/ultrasound-screening) and improved ultrasound equipment and ultrasonographer training, a significant proportion of babies with LUTO were not detected prenatally. This may be because a less severe phenotype is not amenable to early detection. Also, for the first time in a population-based study it was documented that in 26.9% of prenatal diagnoses there was a false-positive diagnosis of LUTO. The majority of these cases were associated with structural anomalies of the urinary tract (i.e. ureterocele or cloacal dystrophy). However, significantly, in 5.9% of cases the ultrasound appearances were secondary to reflux nephropathy and were not distinguished from obstructive renal disease antenatally. Our data demonstrated a significant trend towards an increase in the number of prenatally diagnosed cases with a specific prenatal diagnosis. This could be because of a multitude of factors including improved ultrasound equipment, improved training and improved documentation/reporting. Our cohort study indicated that in the West Midlands it was not common for prenatal assessment of liquor volume and assessment of renal parenchymal appearance to be supplemented by fetal urinalysis. This is in all likelihood because evaluation of this diagnostic test has indicated that it has little value over assessment of renal appearance and liquor volume. 14,70

It was also interesting to note the number of antenatally identified babies with LUTO who were potentially amenable to ‘in utero’ therapy. Of the prenatally suspected, isolated, singleton male LUTO cases, 2% underwent serial vesicocentesis and 8% underwent VAS. The number having an intervention is relatively small indicating that any prenatal intervention aimed at influencing the pathogenesis of this condition will influence relatively low numbers of babies in absolute terms. It is important to realise that this number will have been influenced by the significant proportion of parents who would have opted for TOP rather than intervention and by the potential for not all eligible fetuses to have been referred to the tertiary centre for consideration for intervention. These are local influences and may not be representative of the situation internationally. There were no cases in which fetal cystoscopy was performed. There is very little published evidence on the effectiveness of fetal cystoscopy as a diagnostic and therapeutic intervention for LUTO and the quality of this evidence is poor;21,104 thus, we do not feel that this is unusual or a limitation of our cohort.

Randomised controlled trial

Registry

Women whose fetus was eligible to be randomised into the PLUTO trial but who did not give consent to randomisation, women for whom the fetal medicine specialist had a strong preference for either shunting or conservative management or women with a multiple pregnancy were eligible to take part in the registry (see Figure 9 ). Consent to allow recording of follow-up information was still required in these cases.

Setting of the study

All fetal medicine centres within England, Scotland and the Republic of Ireland agreed to take part within the study and received local research ethics approval and trust R&D approval, either as a centre for randomisation and intervention or as a referring centre. Referral centres identified eligible women and counselled them regarding entry into the trial with referral to Birmingham for consent, randomisation and VAS when appropriate. The full list of centres and the principal investigators can be found in Appendix 2 . This includes 24 randomising centres, three referring centres and three paediatric follow-up centres.

Setting for the randomised controlled trial

The setting for the RCT was six fetal medicine centres in the UK [Birmingham Women’s Hospital NHS Foundation Trust, Liverpool Women’s NHS Foundation Trust, Nottingham University Hospitals NHS Trust, Forth Park Hospital (Fife), Royal Victoria Infirmary (Newcastle), St George’s Healthcare NHS Trust (London)] and one in the Netherlands (Leiden University Medical Centre, receiving referrals from all hospitals within the Netherlands).

Setting for the registry

The setting for the registry was nine fetal medicine centres, eight within the UK [Birmingham Women’s Hospital NHS Foundation Trust, Liverpool Women’s NHS Foundation Trust, St George’s Healthcare NHS Trust (London), St Michael’s Hospital (Bristol), Princess Anne Hospital (Southampton), King’s College Hospital (London), Queen Mother’s Hospital (Glasgow), St Mary’s Hospital (Manchester)] and one in the Netherlands (Leiden University Medical Centre, receiving referrals from all hospitals within the Netherlands).

Screening and consent for the randomised controlled trial and registry

All women were offered a high-resolution scan at about 18–22 weeks of gestation and were counselled about the possibility of fetal abnormalities. 106 The diagnosis of fetal LUTO was made prospectively on the basis of the ultrasound appearances of the fetal bladder and kidneys and the amniotic fluid volume. Some clinicians used fetal vesicocentesis with analysis of urinary analytes (see Chapter 1 ) in an attempt to further aid diagnostic accuracy and inform prognosis. The diagnosis was explained carefully to the parents (often over several days) and the possibility of participation in the PLUTO trial was proposed. A written participant information sheet (see Appendix 3 ) and consent form (see Appendix 4 ) were provided along with supporting literature for those receiving a diagnosis of fetal abnormality. Support and counselling were provided according to local practice and a contact telephone number for a specialist midwife counsellor was provided. In all cases, because of the natural history of the condition, the fetal medicine subspecialist felt that it was appropriate to discuss the option of TOP. The PLUTO trial was not discussed further with those mothers who opted for termination but they were approached at their follow-up counselling appointment about participation in the qualitative research study (see Chapter 9 ). An anonymised register of TOPs (including gestational age, fetal medicine centre) was kept within the site file.

At the next visit, typically up to 7 days following vesicocentesis or at the follow-up scan, the mother was invited to consent to participation. Only fetal medicine subspecialists experienced in fetal bladder shunt insertion performed and consented for the procedure. All participants provided written informed consent.

Randomisation

Eligible consenting participants were allocated to placement of a VAS or observation until delivery (conservative management) by a telephone or internet randomisation service organised by the Birmingham Clinical Trials Unit. Allocation was concealed until a participant’s baseline details had been provided. Minimisation was used to ensure balance of treatment allocation overall and so that the following variables could be used in the prespecified subgroup analyses: gestational age at diagnosis, age of mother at diagnosis, liquor volume by maximum pool depth (MPD). The data collected at randomisation are detailed in Appendix 5 .

Interventions: randomised controlled trial and registry

Registration

Not all patients eligible for the RCT consented to be randomised. Those women who did not agree to be randomised were then asked if they would provide consent for their details to be included in a register of LUTO cases. If a decision was made to enter a patient onto the registry this was ideally done prospectively, but retrospective registration, ideally before delivery, was accepted. All parents entered onto the registry were counselled in the same way as in the RCT, were provided with the patient information leaflet and were consented as in the RCT.

Entry onto the registry was via the internet or by telephone registration when internet access was not available, as for randomisation; all information on the randomisation form was required and recorded.

Study management

Ethical considerations

The study was conducted according to the principles of the 1998 MRC Guidelines for Good Clinical Practice in Clinical Trials 107 and the appropriate NHS research governance frameworks. PLUTO did not fall within the scope of the European Directive on Clinical Trials (2001/20/EU). The University of Birmingham was the sponsor.

The RCT and registry obtained ethical approval from the Nottingham MREC (COREC Ref. 04/Q2404/89), which determined that the trial design respects the rights, safety and well-being of the participants. Every potential UK centre also obtained local research ethics committee (LREC) and trust R&D approval. All UK centres were required to sign an investigator’s agreement with the University of Birmingham, detailing their commitment to accrual, compliance, good clinical practice, confidentiality and publication. The sponsor ensured that all researchers not employed by a NHS organisation held a NHS honorary contract for that organisation if required by current guidelines.

International centres were asked to apply for ethics committee approval according to their own system, providing the PLUTO trial office with written evidence of approval. One local principal investigator in each country was asked to act as a national co-ordinator and monitor the centres in that country for compliance with good clinical practice and any relevant local regulations.

Assessments and outcome measures: randomised controlled trial and registry

All antenatal and perinatal outcome measures were obtained as per standard clinical practice for diagnosis, monitoring and treatment of prenatal bladder outflow obstruction. Long-term follow-up of all children born with this condition is also routine practice but for the trial this was undertaken by a paediatric nephrologist or urologist, who may be at a different hospital from where the shunt was placed.

Statistical considerations

Barriers to recruitment

1. International barriers – There were delays in the international centres being able to recruit. By June 2009 no international centres had approval. Of the two international centres, that in the Netherlands eventually randomised five participants although the centre in Dublin did not randomise anyone. The trial team also noted the barriers that they had encountered in getting international centres involved (namely problems with the sponsor, the University of Birmingham, having to pay an individual insurance premium for the trial of £30,000 per annum for each international centre). This eventually led to the University of Birmingham deciding that no further international centres could be recruited.

2. Delays in approval for UK centres – There were delays in getting UK centres recruiting because of lengthy approval processes partly because some R&D offices were more complex in their paperwork requirements than others. Three centres also negotiated changes to the site agreement with the university to make it better suited to their trust. These negotiations were complex and caused delays.

3. High proportion of women opting for TOP – More women than expected were opting for TOP rather than continuing to delivery after their fetus was diagnosed with LUTO. The reasons for this were explored as part of the patient acceptability study (see Chapter 9 ). By June 2009, 20 women had been randomised, 21 had been entered onto the registry and 41 had opted for TOP.

4. Lower prevalence of LUTO – It appeared that the prevalence of isolated LUTO might be less than that quoted in the literature; this was determined in the epidemiological study (see Chapter 3 ).

5. High proportion of women being entered on the registry – Of those women who did agree to join the study, more than anticipated were being placed on the registry rather than being entered into the RCT. The tendency to use the register was felt to be a result of the fetal medicine specialists having firm preferences for when and how to treat. It was felt that this operational lack of equipoise [despite the assurances of support for the trial in the assessment of clinicians’ prior beliefs (see Chapter 7 )] would ultimately make it impossible to deliver the required number of participants.

A recommendation was thus made that there should be no absolute target for recruitment at the 24-month stage but that a confidential report should be sent on behalf of the TSC (after consultation with the DMEC) to the HTA programme for its views on projected recruitment and whether funding should be continued in September 2010. On receipt of this report the HTA decided to perform a monitoring visit in May 2010.

Amendments to the PLUTO study

The funder and the TSC decided to halt recruitment into the PLUTO trial in December 2010 as it was agreed that there was no realistic possibility of recruiting sufficient participants.

In October 2011 a final closure plan (a revised proposal for the PLUTO study and financial plan) was agreed (see Appendix 7 ). In summary, this was a 12-month plan with the following components:

• completion of follow-up to the primary end point for all women within the randomised and registry arms of the study

• a limited economic analysis and decision-analytic modelling would be performed

• qualitative research would continue in its present form

• a HTA monograph would be prepared.

This was amended in June 2012 after the HTA programme was informed that long-term follow-up to 1 year was now available for all RCT and registry participants and thus 1-year outcomes could be presented.

Thus, the following secondary objectives from the original protocol could not be met and completion of long-term follow-up to 2 and 5 years of age was not funded:

• to determine the long-term effects of VAS with respect to (1) the development of chronic renal failure and need for dialysis or transplantation, (2) the development of incontinence (bladder dysfunction) and (3) disability-free life-years (incorporating assessment of cognitive development, quality of life, micturition and general health)

• to derive a prognostic risk index.

Baseline characteristics of participants

The baseline characteristics of those randomised appeared well balanced ( Table 13 ). The median age of mothers in the trial was 28 years (IQR 23 to 33) and the vast majority of fetuses were of < 24 weeks’ gestation at the time of LUTO diagnosis (27/31, 87%). The majority of the fetuses had a liquor volume less than the fifth centile (19/31, 61%).

Seven fetuses had vesicocentesis and fetal urinalysis before VAS insertion (n = 5 good prognosis; n = 2 poor prognosis). 58 Four fetuses in the VAS arm and five in the conservative arm were karyotyped; all were normal males (46, XY).

Compliance with allocation

In total, 3/16 (19%) of those randomised to VAS did not receive the intervention and 2/15 (13%) randomised to conservative management ended up receiving a VAS. Of the three cases that were randomised to VAS but did not receive it, one patient withdrew consent after randomisation (for treatment but not follow-up), one patient was found to have a contraindication to VAS after randomisation and in the final patient the clinician determined that inserting a VAS was inappropriate because of poor fetal condition. In both of the cases in which the patient was randomised to conservative management but received a VAS this was because of deterioration in the clinical picture (anhydramnios) and the clinician determining that VAS was appropriate. In three cases (one in the VAS group and two in the conservative management group) parents opted for TOP between 18 and 25 weeks because of perceived poor prognosis by health-care professionals and parents. These decisions were unrelated to treatment allocation.

Vesicoamniotic shunt complications

Vesicoamniotic shunting complications are described in Table 17 [as a proportion of the total number who received the VAS intervention (n = 15)]. Total pregnancy losses totalled 4/15 (27%), although one of these was not thought to be treatment related.

Principal findings

The PLUTO trial was stopped early because of poor recruitment over time. Although this is a limitation, results are available for all babies for outcomes to 1 year of age. The results demonstrate that prognosis in both arms (conservative management and VAS) is poor, with only two babies surviving to 1 year of age with no renal impairment. This finding again reinforces the natural pathogenesis of this fetal disease as one of severe and significant mortality and morbidity independent of treatment. In this randomised study all perinatal survivors had pathology confirmed as being secondary to congenital bladder outflow obstruction (PUVs and atresia).

There was a high proportion of pregnancy losses in both arms and pregnancy outcomes (live births) show no significant difference between the two arms. There appears to be a higher survival rate at 28 days (perinatal and neonatal survival) with VAS but this result is not statistically significant. All neonatal deaths were secondary to pulmonary hypoplasia and the possible reduction in perinatal mortality observed with VAS is probably secondary to prevention or amelioration of oligohydramnios at the critical time of lung development (canalicular phase between 16 and 24 weeks’ gestation).

Clinical outcomes at 1 year demonstrate an overall poor prognosis (in both groups) but again with a possible improved survival (but not statistically significant) with VAS. The likelihood of surviving with normal renal function is minimal in both groups, suggesting that the damage to the renal parenchyma has already occurred at the time of diagnosis and is irreversible. There was no difference in baseline characteristics nor other outcomes between the arms of the trial (e.g. gestational age at delivery, birthweight/small for gestational age and admission to NICU).

Another principal finding of the PLUTO trial was the high proportion of women who did not receive the allocated treatment (crossover). This was because of either clinician choice or changing clinical features as the pregnancy progressed. As this proportion was so high it was felt appropriate to perform an analysis according to treatment received. This analysis demonstrated larger differences in outcomes between the treatment arms of the trial. For survival at 28 days and at 1 year, the outcomes favoured VAS and the differences were statistically significant. There was no difference in survival with normal renal function at 28 days or 1 year between the analysis according to treatment received and the ITT analysis. However, it is conceded that there may have been confounding by selection bias.

Finally, the PLUTO trial demonstrated that, in the context of a RCT of an intervention for a rare fetal disease, relatively few women (or their families) were willing to consider randomisation of treatment and opted instead for either entry onto the registry (where they or their health professional could choose treatment) or, more commonly, TOP. This is investigated further in our qualitative research into patient acceptability (see Chapter 9 ).

Strengths and limitations of the study

The strengths of this study lie in its design. This was a RCT that was multicentre and international. The trial had clear objectives with well-defined (designated a priori) primary and secondary end points. The trial was registered and the protocol published. 16 There were no significant changes to this protocol during the lifetime of the study with the exception of those requested by the primary funder in the light of poor recruitment (i.e. removal of follow-up to 5 years of age). These are clearly documented in Appendix 7 .

A major limitation of this trial is the low number of subjects recruited, making the study underpowered. For this reason the trial was sufficiently powered to detect only very large differences in the primary outcome. The reasons for poor recruitment have already been discussed and are explored further in Chapters 3 , 7 and 9 . One of the major barriers to recruitment was the high proportion of women entered onto the registry rather than being randomised. This was often because the clinician was apparently not in ‘equipoise’ as to the benefits of the intervention (despite the findings of a published systematic review of the literature13) and is in contrast to the expert opinions elicited at the start of the study (see Chapter 7 ). This indicated a divergence of opinion among specialists and established clinical equipoise.

A strength in this respect is that all patients completed follow-up and there were no missing data for the primary and secondary end points. All end points were assessed independently by paediatric neonatologists and nephrologists as part of routine practice. The primary analysis reported is an ITT analysis with no exclusions.

Blinding of fetal medicine specialists, parents and end point assessors was not possible because of the nature of the intervention. As the primary end point was survival the primary outcome would not have been affected by this lack of blinding. However, this may have had an impact on the number of women who ‘crossed over’ from the conservative management arm to the VAS arm because of a change in the assessment of the condition as the pregnancy progressed. This selection bias will have been a confounder and the amount or direction of effect this may have had cannot be assessed. Thus, the number of women crossing over from the conservative management arm to the VAS arm is a limitation with regard to the study’s conclusions.

Strengths and limitations in relation to other studies

This is the only RCT of the effectiveness of VAS in LUTO. We therefore cannot make direct comparisons between our results and previously published results of a RCT/prospective study in terms of strengths and limitations. Although the numbers recruited are small (in the context of the prospective power calculation for the RCT), in relation to previously published cohort studies they are comparable. These retrospective cohort studies and their results have been rigorously assessed by our group as part of the systematic reviews discussed in Chapter 1 . The results from these systematic reviews indicated an OR of 3.86 (95% CI 2.00 to 7.45) for perinatal survival with VAS and an OR of 0.50 (95% CI 0.13 to 1.90) for survival with normal renal function. The conclusion of the observational evidence is that VAS improves perinatal survival but the effect on long-term renal function is unclear. The outcome measures used for renal function in these observational studies were also heterogeneous. The results of the PLUTO RCT are consistent with the findings of the systematic review summarising the observational evidence for perinatal survival and suggest that the chance of survival with normal renal function is very low regardless of whether VAS is performed or not.

Recommendations for future practice

As the findings of the PLUTO trial were limited by the small numbers recruited and thus the RCT was underpowered, any recommendations must be interpreted in this light. However, the results of the RCT and the observational evidence are consistent and thus the body of objective evidence suggests that VAS improves overall perinatal survival compared with conservative management but that the long-term prognosis for these babies into infant life is poor (with high rates of mortality and morbidity). Parents should be counselled about the risks of pregnancy loss with or without VAS insertion. The NICE interventional procedures guidance (IPG 2028) should be updated to reflect this new evidence.

Recommendations for future research

As the PLUTO trial was a multicentre international trial and did not recruit the required number of participants to achieve statistical power, it is unlikely that a further RCT would be feasible. It is imperative that the babies recruited into the PLUTO trial are prospectively followed up throughout childhood to determine the effects of VAS on outcomes such as renal function, incontinence, cognitive development and quality of life. However, the number of children surviving to this age (2–5 years) is likely to be small. There should be further research to try and overcome the barriers to recruitment identified within this study, namely in relation to the methodology of RCTs in rare diseases (especially relating to pregnancy). Improvements have already been made in the last few years in the UK relating to achieving ethics and R&D approval for studies recruiting at multiple sites. These changes greatly shortened the time that recruitment sites took to recruit their first patient in the latter period of the PLUTO study. However, the study was limited in its recruitment ability because of the difficulties in obtaining indemnity and thus sponsorship for international centres. This is something that the international academic community, higher education institutions and funders must work hard to resolve so that research questions can be adequately addressed in the study of rare diseases.

Baseline characteristics of registry participants

The baseline characteristics of those registered can be seen in Table 19 . The average age of mothers was 31 years (p = 0.82 between groups). The proportion of fetuses with a liquor volume less than the fifth centile was higher among those opting for VAS than among those opting for conservative management (70% vs. 34%, p = 0.07). The proportion of fetuses for whom age at diagnosis was < 24 weeks also appeared higher in the VAS group (80% vs. 51%), but this was not statistically significant (p = 0.15). There were no other statistically significant differences between the groups but we cannot rule out that this was because of the small sample size.

Compliance with intervention selection

In four cases (one in the VAS group and three in the conservative management group) parents opted for TOP between 19 and 23 weeks. This was because the overall prognosis for the fetus was felt to be poor. These decisions were considered unrelated to intervention choice.

Vesicoamniotic shunt complications

As described in Table 23 , the risk of total pregnancy loss was 3/10 (30%) (one of these was not considered to be a serious adverse event related to the intervention but was considered to be a treatment failure as the fetus had persistent anhydramnios and thus the parents opted to terminate the pregnancy).

Background

We have combined data from the randomised and registry data sets to explore whether the difference in survival to 28 days between treatment groups in the randomised and registry data (the RCT appeared to favour VAS whereas the registry favoured conservative management) is related to the demographics of the randomised and registry participants. We also examined whether the predefined subgroups of gestational age, liquor volume and maternal age were indeed associated with survival to 28 days. Evidence for the importance of these variables in urinary tract obstruction is thus far limited, as discussed in Chapter 1 . The combined data set gives us a larger group to make statistical inferences from (n = 40 survivors from 76 participants in total).

Results

Principal findings

A high proportion of women and clinicians opted to select their management and thus entered the registry. In addition, a much higher than anticipated number of parents underwent TOP as the only option for management. Although not statistically significant, survival at 28 days and at 1 year appeared higher in the conservative arm. Exclusion of non-treatment-related TOPs gave an improved survival to 1 year in the conservative arm that was statistically significant. There was a higher proportion of babies born preterm in the VAS group; thus, there was a shorter interval from registration to delivery and a younger gestational age at delivery in this group. Longer-term outcomes demonstrated a higher proportion of babies within the conservative arm surviving with normal renal function at both 28 days and 1 year, although the outlook was still poor. High proportions of babies required surgery in either the perinatal period or the first year of life in both arms.

Comparing the results of the registry with those of the RCT there appears to be no difference in live birth rates for either VAS or conservative management, highlighting again the poor prognosis with pulmonary hypoplasia. The main difference is seen in the conservative group of the registry, with higher survival to 28 days and 1 year. The chance of surviving with normal renal function is also higher in the group registered for conservative management than among those randomised.

However, there were clinically important differences between fetuses in the VAS arm and fetuses in the conservative arm within the registry. There was a statistically significant higher proportion of babies with a liquor volume less than the fifth centile in the VAS group and they were also more likely to have their diagnosis made earlier in the pregnancy (at < 24 weeks). There was also a difference in the pathological diagnoses made postnatally in these groups, in that all babies with a pathological diagnosis in the VAS group were confirmed to have had LUTO. However, in the conservative group there was a high proportion of babies with a false-positive antenatal diagnosis, which was comparable to that identified in the epidemiological study (see Chapter 3 ). These factors should be seen as confounders, likely to influence the results for both survival and renal impairment and explain the difference in outcomes seen between the RCT and the registry.

The exploratory analysis of the combined PLUTO study data set out to ascertain if apparent differences in effect on 28-day survival between VAS and conservative management in the two data sets (randomised and registry) were statistically important and furthermore to investigate possible reasons for these differences. We also set out to provide some stronger evidence for or against the importance of gestational age, liquor volume and maternal age as influential prognostic factors for survival.

We found that there was statistically important heterogeneity between the survival rate estimates in the randomised and registry data sets (randomised set favoured VAS whereas the registry set favoured conservative management). The composition of the randomised and registry sets clearly differed, with those registered for conservative management comprising a greater proportion of fetuses with a liquor volume greater than the fifth centile and with a gestational age at diagnosis of ≥ 24 weeks. Liquor volume greater than the fifth centile and gestational age at diagnosis of ≥ 24 weeks appear to be associated with increased chance of survival to 28 days.

It is possible that the number of TOPs in the registry was a national underestimate. The denominator (nationally) is unknown with the most complete data arising from the ONS. 115 In 2010 there were 726,879 births in England and Wales and 196,109 TOPs. 116 Of these TOPs, only 9% were performed at > 13 weeks and < 0.1% (n = 147) at > 24 weeks. Of those TOPs performed in 2010 under clause E (termination up to term pregnancy when there is a substantial risk that if the child is born it would suffer from such physical or mental abnormalities as to be seriously handicapped), 100 were performed for urinary tract abnormalities (ICD-10 codes Q60–64). Similar rates were reported for 2011: 196,082 TOPs, 9% at < 13 weeks, < 0.1% at > 24 weeks (n = 146) and, of those carried out under clause E, 107 were performed for urinary tract abnormalities. However, these national data are available for England and Wales only and the ICD-10 codes Q60–64 include all urinary tract abnormalities (e.g. renal agenesis, cystic kidney disease, congenital obstruction of the renal pelvis). BINOCAR reported similar data for 2010 with urinary anomalies having a TOP rate of 5.3 per 10,000 total births (95% CI 4.5 to 6.3). 96

Strengths and limitations of the study

The major limitation of this part of the study is that it was a non-randomised treatment allocation study. However, data collection was prospective and was carried out according to a protocol designed and published a priori as for the RCT. The data collected was identical to that collected in the RCT, as were the outcome measures.

A strength of this study is again the high number of cases with complete follow-up. There were no cases with missing data for the primary end point of survival and only one case in the VAS group with missing data for renal function. As for the RCT, all end points were assessed independently by paediatric neonatologists and nephrologists as part of routine practice.

The logistic regression analysis provides the strongest evidence to date that liquor volume and gestational age at diagnosis are associated with survival to 28 days in fetuses with a confirmed diagnosis of LUTO. The data come from a high-quality RCT and registry with excellent follow-up of all babies. The main limitation is the relatively small data set (even with trial and registry data combined), which means that we are limited in the number of variables that we can examine for possible association with survival to 28 days. The small data set also means that we cannot rule out a possible association of maternal age with survival.

Strengths and limitations in relation to other studies

As discussed in Chapter 1 there have been other observational studies comparing VAS with conservative management. The numbers recruited to this registry are comparable to those in the previously published cohorts. However, these other cohorts were retrospective and did not have the high rate of follow-up and length of follow-up involved in this registry, nor was a diagnosis of LUTO sought in all patients. The outcome measures used for renal function in these observational studies were also heterogeneous. Our registry results are in contrast to the results of our systematic review, which indicated an OR of 3.86 (95% CI 2.00 to 7.45) for perinatal survival with VAS. The reason for the apparent better outcomes in the registry conservative management group may be because of the confounders identified above, which have already been discussed.

Conclusion

Those women entered onto the registry and receiving VAS had similar baseline characteristics and outcomes to those in the VAS arm of the RCT, with a similar but relatively poor prognosis. The conservatively managed group in the registry appeared to be different from the conservatively managed group in the RCT in terms of baseline characteristics (later age of gestation at diagnosis and a higher proportion with a liquor volume greater than the fifth centile) and had a better prognosis for survival. There were a significant number of cases of false-positive antenatal diagnosis within this conservatively managed group, with final postnatal diagnoses being conditions associated with a better prognosis than LUTO. A high proportion of parents opted for TOP.

Normal liquor volume (greater than the fifth centile) and gestational age at diagnosis of ≥ 24 weeks seem to be associated with an increased probability of survival at 28 days in fetuses with a confirmed diagnosis of LUTO. Non-randomised comparisons of survival between groups in the registry are likely to be confounded with the effect of these parameters and are likely to have contributed to the observed statistical heterogeneity between the randomised and registry data sets.

Prior elicitation

We selected as experts consultant members of the British Maternal and Fetal Medicine Society, the British Association for Paediatric Nephrology and the British Association of Paediatric Urologists. These expert groups are henceforth referred to as fetal medicine experts, paediatric nephrologists and paediatric urologists respectively.

We developed a questionnaire that was used to elicit opinions. An electronic copy of the elicitation form and a worked example illustrating the method (see Appendix 8 ) was e-mailed to all experts, with an additional e-mail reminder being sent to those who did not respond. The e-mails were sent over the period February 2007–November 2008. 119

The questionnaire sought to elicit opinions on both perinatal survival under conventional treatment and survival using intrauterine VAS. To this end each expert was asked to specify their beliefs for three components:

1. the most likely percentage mortality under conservative treatment (i.e. control) at 6 weeks

2. the most likely change in mortality under intrauterine VAS

3. the likelihood (or distribution) of change in mortality for each 5-point interval between a 40% relative increase in mortality and a 40% relative decrease in mortality (total of 17 point values).

The primary outcome for the PLUTO study was survival to 28 days. We assume throughout that elicited beliefs for outcomes at 6 weeks are equivalent to those at 28 days.

To elicit these distributions for change (no. 3 above) each expert was asked to use the graph illustrated in Figure 14 to record their beliefs. Experts were told that each vertical line represents a possible true benefit/harm from intrauterine VAS compared with the control treatment. Experts were asked to mark a horizontal line through each vertical line to show how likely it is that this level of benefit/harm is the true level. Experts were provided with illustrative examples.

FIGURE 14.

Graphical figure used to elicit distributions of beliefs from experts used in part 3.

Summarising beliefs elicited for outcomes at 28 days

The elicited beliefs for perinatal mortality under conservative management (no. 1 above) are summarised using medians and IQRs and are graphically displayed in a histogram for all experts combined across specialties and for experts stratified by specialty. Elicited beliefs for change in mortality under VAS (no. 2 above) are similarly summarised.

In part 3 of the elicitation process, for the elicited beliefs for change in perinatal mortality under intrauterine VAS, we first measured the height of each mark made on the graph in mm (see Figure 14 ). The distributions of change for each expert were then computed by dividing the height of each mark (at 5-point intervals between −40 to 40) by the sum of the heights for that individual, henceforth referred to as standardised height. These then formed the expert’s beliefs for the distribution of change in perinatal mortality under intrauterine VAS. The distributions of change are graphically displayed for each expert independently, grouped by specialty.

The individual expert’s distributions for change in perinatal mortality were then pooled across experts, by weighting by standardised height of marks. These pooled distributions are displayed graphically and summarised using medians and IQRs, again by specialty and over all specialties.

Obtaining informative prior distributions from elicited beliefs

The elicited beliefs for perinatal mortality under conservative management and expected change in perinatal mortality using intrauterine VAS were then transformed into ORs for survival (intrauterine VAS compared with conservative treatment) along with corresponding distributions for each OR. It was necessary to parameterise prior distributions on the OR scale as statistical models fitted to the trial data, to obtain posterior distributions, were paramaterised using logistic regression. Outputs from the logistic regression were subsequently transformed into RRs to be consistent with the frequentist analysis (see Obtaining posterior distributions for the outcome of perinatal mortality). This process was implemented according to the following steps:

• Transforming changes in mortality into ORs – The first stage was to convert each expert’s distribution of changes in mortality into a distribution of ORs for the effect of VAS on mortality. From the value elicited to be the expert’s belief for perinatal mortality under conservative treatment, and for each 5-point value (between a 40% increase and a 40% decrease) in the change in mortality, we calculated a corresponding OR (intrauterine VAS vs. conservative management). For example, an expert might believe that the perinatal mortality rate under conservative treatment is 50%. They might then give various weights to between a 40% increase and a 40% decrease in mortality using intrauterine VAS. A scenario including a 40% increase in mortality using intrauterine VAS and 50% mortality under conservative treatment would equate to a mortality rate of 70% [50% × (1 + 0.4)] in the intervention arm, equating to an OR for survival of 0.43 [(0.3/(1 − 0.3))/(0.5/(1 − 0.5))]. Similarly, a scenario including a 40% decrease in mortality using intrauterine VAS and 50% mortality under conservative treatment would equate to a mortality rate of 30% [50% × (1 − 0.4)] under VAS, equating to an OR for survival of 2.33 [(0.7/(1 − 0.7))/(0.5/(1 − 0.5))]. The standardised height for each OR value directly corresponded to the standardised height assigned to the change in mortality value.

• Pooling over experts – The individual expert’s distributions for ORs were combined to produce a pooled distribution for the OR. For each value of OR, the overall standardised height was computed as the mean of the standardised heights for that OR value from the individual OR distributions of the experts. These pooled distributions for ORs then formed the basis for the prior distributions of likely effectiveness of intrauterine VAS, but, to aid assumptions of normality, these ORs were then transformed to log-ORs (we use natural logs throughout).

• Summarising resulting prior distributions – These prior distributions for ORs for survival are graphically displayed using histograms for all experts combined and are summarised using medians and IQRs (weighted by standardised height). Prior distributions for log-ORs are similarly summarised but we also present the means and SDs (again weighted by standardised height) and superimpose normal densities on histograms.

• Evaluating assumptions of normality – We then formally evaluated assumptions of normality for the log-OR through quantile normal plots, by specialty and for all specialties combined. Assuming normality we then formed four informative priors for the log-OR of survival: a fetal medicine prior, a paediatric nephrology prior, a paediatric urology prior and an all experts prior.

Experts who did not specify a value for perinatal mortality under conservative management could not contribute to this estimation of ORs or log-ORs. Furthermore, some experts specified relative changes in perinatal mortality under intrauterine VAS, which, when paired with their specified value for mortality under conservative management, resulted in mortality using intrauterine VAS of either < 0% or > 100%. Any such out-of-range values were excluded.

In summary, we therefore elicited for each expert their belief about the percentage perinatal mortality under conservative management and their belief for the distribution of change in mortality under intrauterine VAS across the range from a 40% increase to a 40% decrease in relative mortality (in 5-point changes). From these elicited values we then formed prior distributions for beliefs for the log-OR for survival.

Default prior distributions

We also constructed three additional default priors: an uninformative prior, a sceptical prior and an enthusiastic prior. To construct default priors we followed methods as outlined by others. 117

For the binary outcome of perinatal survival all priors were for log-ORs and were assumed to be normally distributed. The enthusiastic prior was centred at the alternative hypothesis that PLUTO was designed to test. PLUTO was powered to detect a change in survival at 28 days from 39% under conservative management to 62% using intrauterine VAS (i.e. an absolute increase of 23%), equating to an OR (for survival) of 2.55 (log-OR = 0.94). The SD for this prior was found by finding the value for the SD for which the enthusiastic prior has a small probability (5%) that the treatment effect is negative (i.e. favours conservative management). Therefore, the enthusiastic prior was centred at log-OR = 0.94 and a SD of 0.57 (0.94/1.645).

The sceptical prior was constructed similarly but was centred at the value of no difference (and so harmful effects possible) and with a 5% probability that the treatment effect was greater than a log-OR of 0.94. This prior was therefore N(0, SD 0.57) for the log-OR.

The uninformative prior, sometimes called a uniform or vague prior, was centred at the value of no effect (log-OR = 0) with very large variance (100). 120

For the outcome survival to 1 year, all priors for hazard ratios (HRs) were for the log-scale and were assumed to be normally distributed.

The enthusiastic prior was centred at the alternative hypothesis that PLUTO was designed to test. The results from a meta-analysis13 informed the calculation of the sample size for the PLUTO trial. The original meta-analysis computed the pooled OR; however, the prior for a HR is best approximated by the distribution for a RR. The pooled RR was calculated using the summary data in the paper and by rerunning the meta-analysis, which gives an estimate for the RR of 1.55 (log-RR = 0.44). The SD for this prior was found by finding the value for the SD for which the enthusiastic prior has a small probability (5%) that the treatment effect is negative (i.e. favours conservative management). Therefore, the enthusiastic prior was centred at log-RR = 0.437 and a SD of 0.27 (0.44/1.645). An estimate of the distribution for the RR can be used as a prior for the HR by assuming that the RR is constant and instantaneous for the whole time period.

The sceptical prior was constructed similarly but was centred at the value of no difference (and so harmful effects possible) and with a 5% probability that the treatment effect was greater than a log-RR of 0.44. This prior was therefore N(0, SD 0.27) for the log-RR.

The uninformative prior was centred at the value of no effect (log-RR = 0) with very large variance (100). 3,121

Obtaining posterior distributions for the outcome of perinatal mortality

For the binary outcome of survival to 28 days we used logistic regression models to model the outcome (survival at 28 days) as a function of the treatment (conservative management or VAS) and so, consequently, prior distributions for the treatment effect are on the log-OR scale. However, the treatment effect measure prespecified in the trial protocol was the RR and so this is also the treatment effect measure reported here. We computed the RR using a transformation from the OR, obtained from the logistic regression model, to a RR using the formula risk = odds/(1 + odds).

We used logistic regression to model the outcome (survival at 28 days):

where the probability, p, of the outcome of survival at 28 days is:

From these estimated log-odds (α₀ for the control group and α₀ + α₁ for the treatment group) we computed risks [risk = odds/(1+ odds)] and RRs.

Bayesian analyses are valued for their ability to provide posterior probability estimates of effect sizes being greater (or less) than some clinically meaningful value, sometimes referred to as Bayesian p-values. The PLUTO trial was designed to detect an OR for survival of 2.5, informed by a meta-analysis of non-randomised studies. Repeating this meta-analysis, on the RR scale this OR of 2.5 is equivalent to a RR of 1.55. We therefore additionally calculated the posterior probability that the RR is > 1.55.

Finally, to aid clinical interpretability we also computed the number needed to treat (NNT). Credible intervals (CrIs) and CIs for NNTs that cross the null value include estimates of both harms and benefits of an intervention. We present them as being bounded by a negative NNT (indicating the NNT for one additional person to be harmed) and a positive NNT (indicating the NNT for one additional person to benefit). Note that unlike other CrIs and CIs, more extreme values (i.e. higher than the positive upper limit and lower than the negative lower limit) are included in the interval.

Obtaining posterior distributions for the outcome of survival to 1 year

For the outcome of survival to 1 year we use Cox proportional hazards regression models to model the hazard of death as a function of the treatment (conservative management or VAS).

Mothers entered the study on the date of randomisation and were followed up until death of the fetus or TOP or the end of the study. Patients were censored because of loss to follow-up or if the fetus had not died before the end of the study, which was 1 January 2012. The survival time is measured as the length of time in the study in weeks from the date of conception. The presumed conception date is estimated from the estimated date of delivery minus 267 days (date of the last menstrual period) plus 14 days.

We used a Cox proportional hazards regression model to model the difference in survival over time between the group who received VAS and the group who received conservative management. The Cox proportional hazards regression model specifies the hazard for individual i as:

where λ ₀(t) is the baseline hazard function that is left unspecified and β ₁ X _i1 + . . . + β _k X _ik is a linear function of a set of k fixed covariates. 122

The variable X ₁ denotes the intervention group thus X ₁ = 1 for those subjects who were randomised to receive the control and X ₁ = 0 for those who were randomised to receive the treatment. Therefore, the hazard rates for those in the control and treatment groups are:

Of note, the hazard rate comparison is conservative care compared with treatment. This is so that the effect estimate on the HR scale will be consistent with RR estimates in so far as treatment effects > 1 will favour the intervention.

As for the binary outcome, we computed the posterior probability estimate of the effect size being greater (or less) than some clinically meaningful value. We used the value of 1.55 as a clinically meaningful difference.

Kaplan–Meier plots were produced for categorical variables to show unadjusted differences in the probability of survival over time between the infants randomised to receive VAS and the infants randomised to receive conservative management.

The Cox regression model assumes that the HR for a variable is constant over time. If this assumption does not hold this implies that the HR changes over time and so time-dependent effects need to be incorporated in the model. This assumption was investigated with the visual inspection of log-cumulative hazard plots in which log(−log) survival curves for two or more subgroups are plotted against time (or log-time). If the hazard rates are proportional the resulting lines should be approximately parallel.

The hazard rates for the treatment groups were not found to be proportional as the curves cross in the log-cumulative hazard plots. To deal with this the survival data were split at the time point when the lines cross and a Cox proportional hazards model was fitted with a time-dependent effect, which results in the estimation of two HRs for the treatment effect. The two HRs estimate the difference in 1-year mortality rates between the two groups in the two time periods. The proportional hazards assumption was held within each of the two distinct time periods. For those infants whose survival time was less than the time that the curves cross, their data remains the same. For each infant who has survived beyond the time that the curves cross, two separate records of data for each period were created. The first line of data includes their survival time from conception to the crossing point. Therefore, the survival time for each individual who survived beyond the time that the curves cross was the time that the curves cross, for example if the curves cross at 35 weeks the survival time is calculated as 35 weeks for each infant. The second line of data includes the survival time from the crossing point until the end of each infant’s survival time. The survival time here was calculated from 0 to the end of their survival time. A new binary variable, split_time, was created for the subjects to allow for the interaction between the time period and the intervention variable. For the first time interval split_time = 0 and for the second time interval split_time = 1. This results in the Cox proportional hazards regression model:

Here, exp{β ₁} is the HR estimate of the difference in mortality rates for up to 1 year between the intervention groups when split_time = 0, that is, the mortality ratio from conception to the time that the curves cross. Exp{β ₂} is the change in the HR estimate when split_time = 1 and therefore exp{β ₁ + β ₂} is the HR estimate for the difference in mortality rates between the intervention groups when split_time = 1, that is, the mortality ratio from the time that the curves cross to the end of follow-up.

The proportional hazards assumption was difficult to assess in this study as there are very little data. Therefore, results are calculated under two assumptions. Results are given assuming that the hazard rates are proportional from conception to the end of follow-up and therefore the estimate of a HR is an average of the treatment effect over time. The results are calculated under the assumption of a time-dependent treatment effect as described above.

Implementation

All Bayesian analyses were performed using the WinBUGS software version 14 (MRC Biostatistics Unit, Cambridge, UK) with 200,000 iterations after allowing for a 10,000 iteration burn-in and checking for convergence using several common measures. Summary estimates provided are medians and 95% CrIs, which can be interpreted as one would like to interpret a conventional CI, that is, as there being 95% probability that the true value of the parameter lies between the upper and lower bounds. For parameters other than the treatment effect we used standard uninformative priors: normal distribution centred at 0 and with variance 100. 3,121 The WinBUGS code for the analysis of prenatal mortality is provided in Appendix 9 and that for the survival analysis is provided in Appendix 10 .

Elicitation results

The questionnaire survey was emailed to 248 experts and 59 replies were received, seven of which were too incomplete to be included, leaving 52 partially completed forms. These 248 experts included 132 fetal medicine experts (of whom 37 replied), 56 paediatric nephrologists (of whom 16 replied) and 60 paediatric urologists (of whom six replied). The overall response rate was 24% (59/248).

Two experts did not provide a belief for perinatal mortality under conservative management and so could not contribute to the estimate of the OR for survival (intrauterine VAS compared with conservative management). Furthermore, some values were given that implied mortality outside of the range 0–100 and these were excluded from the analysis. These affected 46 values out of the 884 (52 × 17) values provided by all experts.

Elicited beliefs for perinatal mortality under conservative management

In total, 50 of the 52 experts provided their opinion on the percentage perinatal mortality rate under conservative management. These beliefs were elicited from 35 fetal medicine experts, 10 paediatric nephrologists and five paediatric urologists. The percentage mortality rates under conservative management elicited from the 50 experts are presented in histograms in Figure 15 . There is considerable variability in the elicited percentage mortality rates both within and between specialties.

FIGURE 15.

Histograms of experts’ elicited beliefs for percentage perinatal mortality rate at 28 days when using conservative management.

The fetal medicine experts, on average, showed less optimism for a favourable outcome under conservative management ( Table 28 ), with an elicited median perinatal mortality rate of 60% (IQR 40% to 75%). Paediatric nephrologists and urologists generally believed that perinatal mortality under conservative management would be much lower: the median mortality rate under conservative management elicited from the paediatric nephrologists was 10% (IQR 10% to 20%) and that elicited from the paediatric urologists was 5% (IQR 1% to 20%) (see Table 28 ).

Elicited beliefs for change in perinatal mortality using intrauterine vesicoamniotic shunting

A total of 52 experts provided information on their beliefs for a change in perinatal mortality with intrauterine VAS. These distributions are displayed in Figures 16 and 17 . There is considerable variation between these distributions, again both within and between specialties. Some experts provided distributions for change similar to normal distributions. For example, expert 3 believed that the most likely value for change would be a 20% decrease in perinatal mortality and that this was symmetrically distributed between a 0% change and a 40% decrease. On the other hand, other experts believed that the likelihood of change increased with an increasing reduction in mortality. So, for example, expert 29, although believing that the change might be anywhere between a 40% increase and a 40% decrease, gave steadily increasing weights to more favourable changes. Several experts (e.g. expert 9) placed all of their weight at a single point for change, suggesting no uncertainty in their beliefs, perhaps suggesting a lack of understanding of the question.

FIGURE 16.

Elicited beliefs for percentage change in perinatal mortality when using intrauterine VAS compared with conservative management: fetal medicine specialists (experts 1–35).

FIGURE 17.

Elicited beliefs for percentage change in perinatal mortality when using intrauterine VAS compared with conservative management: paediatric nephrologists specialists (experts 36–52).

The pooled (over experts) distributions for change (in perinatal mortality using intrauterine VAS compared with conservative management) were reasonably similar between specialties and were reasonably normally distributed for fetal medicine and all experts combined ( Figure 18 ). All three specialties, on average, placed more weight on small values for change, giving preference to intrauterine VAS, but all were reasonably uncertain in their beliefs to provide non-zero density across the range of changes given. Summary statistics are provided in Table 28 . The fetal medicine experts believed that the median percentage decrease in mortality would be 10% (IQR 0% to 20%) whereas the paediatric nephrologists believed that the median decrease would be smaller (5%, IQR −5% to 10%), as did the paediatric urologists (5%, IQR 0% to 15%).

FIGURE 18.

Histogram of pooled elicited beliefs for percentage change in perinatal mortality when using intrauterine VAS compared with conservative management.

Pooled elicited beliefs for mortality using intrauterine vesicoamniotic shunting

From both the beliefs elicited for perinatal mortality under conventional treatment and the beliefs elicited for change in mortality with intrauterine VAS we elicited beliefs for mortality under intrauterine VAS. Two experts did not provide a belief of mortality under conservative management and so could not contribute to the estimate of mortality under intrauterine VAS.

Again, there was a wide variation in beliefs between the three specialty groups (see Table 28 ). The fetal medicine experts were the most pessimistic, giving a median perinatal mortality rate with VAS of 51% (IQR 33% to 68%). The paediatric nephrologists and the paediatric urologists were much more optimistic, with a median mortality rate of 10% (IQR 7% to 18%) for the paediatric nephrologists and a median mortality rate of 5% (IQR 1% to 18%) for the paediatric urologists.

Pooled elicited prior distributions for odds ratios

From these elicited beliefs for perinatal mortality under conservative management and change in mortality with VAS along with corresponding distributions, pooled prior distributions were obtained for ORs (VAS vs. conservative management). The paediatric nephrology and paediatric urology experts were more certain in their beliefs and less optimistic for a benefit of intrauterine VAS than the fetal medicine experts: the elicited median OR for survival from experts in fetal medicine was 1.28 (IQR 1.00 to 1.63) compared with 1.07 (IQR 1.00 to 1.19) for paediatric nephrology experts and 1.11 (IQR 1.00 to 1.28) for paediatric urology experts (see Table 28 ).

These elicited ORs for (perinatal) survival were transformed onto a log-scale. Histograms of the distributions for log-ORs pooled over experts (weighted by standardised height) by speciality appear reasonably normally distributed ( Figure 19 ). The distribution of log-ORs for fetal medicine experts is centred slightly above 0, suggesting a slight preference towards the belief that intrauterine VAS will be beneficial, with a mean value of 0.26 and SD of 0.5 (see Table 28 ). The distributions for both paediatric specialties are narrower and centred slightly closer to 0 than that of the fetal medicine experts (paediatric nephrologists: mean 0.07, SD 0.20; paediatric urologists: mean 0.13, SD 0.20).

FIGURE 19.

Elicited pooled log-ORs (overlaid with normal distribution) by specialty and for all specialties combined for odds of perinatal survival with intrauterine VAS compared with conservative management.

Although demonstrating some departure from normality ( Figure 20 ), these data were deemed sufficiently normal and so the log-OR approximated by a normal distribution with mean and SD values taken to be those elicited.

FIGURE 20.

Quantile normal plot for log-OR of perinatal survival by specialty and for all specialties combined.

Summary of elicited priors

From the above elicitation process we therefore arrived at four informative priors. These informative priors were for log-ORs (for perinatal survival) and were assumed to be normally distributed with means and SDs as estimated from the process outlined above (values in Table 28 ). Therefore, the fetal medicine prior for the log-OR was taken to be N(mean 0.26, SD 0.50), the paediatric nephrology prior was taken to be N(mean 0.07, SD 0.20), the paediatric urology prior was taken to be N(mean 0.13, SD 0.20) and the all experts prior was taken to be N(mean 0.21, SD 0.44). These densities are displayed in Figure 21 . Consistent with the results above, the prior for the fetal medicine experts is wider and centred very slightly at a value more favourable to intrauterine VAS than the prior for either the paediatric urologists or the paediatric nephrologists. Because the majority of the experts (35 out of 52) were fetal medicine experts the combined overall expert prior was very similar to that of the fetal medicine experts.

FIGURE 21.

Elicited priors by specialty group for log-OR and OR of perinatal survival with VAS compared with conservative management. (a) Informative priors: log-OR scale; and (b) informative priors: OR scale.

Experts were collectively very sceptical about the ability of VAS to provide a treatment effect as large as that anticipated by the trial (OR 2.55), and from the elicited opinions we estimated that the clinicians believed that the probability that the treatment effect would be this large was only 5% (see Table 28 ).

Default prior distributions

The elicited informative priors were then compared with three default priors, an enthusiastic prior centred at a log-OR of 0.94 and including a small possibility (about 5%) of harm (SD 0.57), a sceptical prior, centred at the value of no difference (log-OR 0) and also with a SD of 0.57, and finally an uninformative prior, once again centred at 0 but with a very large SD (variance 100²), to essentially provide uniform distributions over the range of plausible log-ORs. The sceptical and enthusiastic prior densities for outcomes at 28 days are displayed in Figure 22 .

FIGURE 22.

Sceptical and enthusiastic priors for the outcome of perinatal survival. (a) Default priors: log-OR scale; and (b) default priors: OR scale.

For the outcome survival to one year, the default priors were slightly different: an enthusiastic prior is formed with a centre at a log-RR of 0.44 and including a small possibility (about 5%) of harm (SD 0.27); a sceptical prior is centred at the value of no difference (log-RR 0) and also with a SD of 0.27; and finally a non-informative prior centred at 0 but with a very large SD (variance 100²). The sceptical and enthusiastic prior densities for the outcomes at one year are displayed in Figure 23 .

FIGURE 23.

Sceptical and enthusiastic priors for the outcome of survival to 1 year. (a) Sceptical and enthusiastic prior distributions on the log-HR scale; and (b) sceptical and enthusiastic prior distributions on the HR scale.

PLUTO trial results for the outcome of perinatal survival

As reported in Chapter 5 , 15 participants were randomised to the conservative management arm and 16 to the intrauterine VAS arm of the PLUTO RCT. Of the 15 infants randomised to conservative management, four (27%) were alive at 28 days. Of the 16 randomised to the intrauterine VAS arm, eight (50%) were alive at 28 days ( Table 29 ).

In the event, two patients randomised to conservative management actually received intrauterine VAS and three patients randomised to the intrauterine VAS arm actually had conservative management. Therefore, 16 patients received conservative management and 15 patients received intrauterine VAS; these form the AT analysis (see Table 29 ). Of the 15 patients who received intrauterine VAS nine (60%) were alive at 28 days and, of the 16 patients who received conservative management, three (19%) were alive at 28 days.

Posterior distributions for the outcome of perinatal survival

Table 30 presents both the estimated RR along with the 95% CrI and the associated Bayesian p-value (probability that the RR is > 1.55, which is the effect size that the trial was designed to detect) and the NNT for each of the four informative prior distributions (fetal medicine, paediatric urologists, paediatric nephrologists and all experts) and the three additional prior distributions (uninformative, enthusiastic and sceptical).

Under the ITT analysis and the uninformative prior, the estimated RR for survival is 1.93 (95% CrI 0.77 to 6.25) with a probability of a clinically important benefit (RR > 1.55) of 0.67 and a NNT of 4 (95% CrI −31 to 38). Perhaps surprisingly, the median estimated RR under the enthusiastic prior is a little closer to the null (RR 1.82), although the CrI includes only values that favour the intervention (95% CrI 1.05 to 3.52); the probability of a clinically important effect increases to 0.71; and the 95% CrI for the NNT includes only beneficial effects (NNT 5, 95% CrI 2 to 24). As expected, the sceptical prior pulls the RR even closer towards the null (RR 1.25) and the 95% CrI does not exclude values that favour the conventional treatment (95% CrI 0.73 to 2.27); and the probability of a clinically important effect reduces to 0.23 and the 95% CrI for the NNT is very wide (NNT 7, 95% CrI −103 to 118).

When including the elicited priors the probability that the RR is > 1.55 decreases; the point estimates for the RRs are generally closer to the null and the 95% CrIs include the possibility of either treatment being optimal; and the estimated NNTs, although all positive, have 95% CrIs that are wide and include both beneficial and harmful effects. The posterior estimates using the fetal medicine and all expert priors are similar (RR 1.31, 95% CrI 0.84 to 2.18 for all experts; RR 1.35, 95% CrI 0.82 to 2.34 for fetal medicine experts). Both the paediatric nephrology and paediatric urology posteriors show less optimism for the intervention with the probability that the RR is > 1.55 being just 0.01 for both and the RRs being close to the null and 95% CrI intervals crossing the null but showing increased certainty (RR 1.08, 95% CrI 0.86 to 1.38 for paediatric nephrologists; RR 1.12, 95% CrI 0.89 to 1.44 for paediatric urologists).

Under the AT analysis all estimates of RRs show greater preference for the intervention and the probabilities that the RR is > 1.55 are increased (compared with the ITT analysis), although 95% CrIs are wider, indicating less certainty, and include both positive and harmful effects, as do the 95% CrIs for the NNTs.

A plot of the posterior distributions for the ORs under the seven prior distributions for the ITT analysis is shown in Figure 24 and for the as-treated analysis is shown in Figure 25 . Figure 26 displays the individual prior distributions overlaid with their updated posterior distributions for the ITT analysis only.

FIGURE 24.

Posterior distributions for ORs (perinatal survival) under the ITT analysis.

FIGURE 25.

Posterior distributions for ORs (perinatal survival) under the AT analysis.

FIGURE 26.

Prior and posterior distributions (ITT) overlaid (log-OR scale). Solid line, prior; dashed line, posterior.

Survival analysis

Fifteen participants were randomised to the conservative management arm and 16 to the intrauterine VAS arm ( Table 31 ). Of the 15 infants randomised to conservative management, three (20%) were alive at the end of follow-up; in two cases (13%) the pregnancy had been terminated and 10 (67%) had died. These two TOPs were analysed as censored observations. Of the 16 infants randomised to the intrauterine VAS arm, seven (44%) were alive at the end of the study, in three cases (19%) the pregnancy had been terminated and six (38%) had died. One of the three TOPs was analysed as a censored observation as it was not considered to be as a result of treatment failure. The remaining two TOPs were a result of SROM within 28 days of receiving the shunt and were therefore analysed as failure of the treatment.

In the event, two patients randomised to conservative management actually received intrauterine VAS and three patients randomised to intrauterine VAS arm actually received conservative management. Therefore, 16 patients received conservative management and 15 patients received intrauterine VAS; these form the AT analysis (see Table 31 ). Of the 15 patients who received intrauterine VAS eight (53%) were alive at the end of the study and, of the 16 patients who received conservative management, two (13%) were alive at the end of the study.

Assessing the proportional hazards assumption

The Kaplan–Meier unadjusted survival curves for the treatment arms are shown in Figure 27 . A time point of 0 corresponds to the time of conception. The probability of survival is lower for those subjects who received a VAS than for those who received conservative management for approximately the first 36.5 weeks from conception. After this time the Kaplan–Meier curves cross and the probability of survival is higher for those who received a VAS than for those who received conservative management. The time point that the survival curves cross approximately corresponds to the time of birth of the infants. Therefore, this plot suggests that the infants who received a VAS have a lower probability of survival before birth than the infants who received conservative management. However, the infants who received a VAS have a higher probability of survival than those in the conservative management arm if they survive the antenatal period.

FIGURE 27.

Kaplan–Meier plots for survival estimates by treatment arm.

The log-cumulative hazards plot of the treatment variable is displayed in Figure 28 , which was used to assess the proportional hazards assumption. The lines cross suggesting that the assumption is violated. This is again suggestive of a harmful effect for the VAS arm for the time period of conception to approximately 36.5 weeks and a beneficial effect for the VAS arm in the time period from 36.5 weeks onwards. Therefore, the data were split at the time point of 36.5 weeks and the treatment effect was estimated in the form of two HRs.

FIGURE 28.

Testing the assumption of proportional hazards: ‘log-log’ plot.

Posterior distributions for hazard ratios

Table 32 presents the results under the first assumption that the hazard rates are proportional from conception to the end of follow-up and so the HR is an estimate of the average (over the period from conception to 1 year of age) difference in mortality rates between the two groups. Table 32 presents the estimated HRs with 95% CrIs and the Bayesian posterior probabilities that the HR is > 1.55 for the three default prior distributions (uninformative, enthusiastic and sceptical).

For clarity, it is reiterated that the HR comparison is conservative care compared with VAS. This is so that the effect estimate on the HR scale will be consistent with RR estimates in so far as treatment effects > 1 will favour the intervention.

Under the ITT analysis and uninformative prior, the estimated HR is 1.63 (95% CrI 0.63 to 4.43) with a probability that the HR is > 1.55 estimated to be 0.54 (i.e. 54%). Perhaps surprisingly, the median estimated HR under the enthusiastic prior is a little closer to the null (HR 1.56); however, the CrI (95% CrI 0.99 to 2.47) is narrower than that for the uninformative results. Under the enthusiastic prior the probability that the HR is > 1.55 (favours treatment) is 0.52. As expected, the sceptical prior pulls the HR even closer towards the null and the 95% CrI includes values that favour the conventional treatment (HR 1.12, 95% CrI 0.71 to 1.75). The probability that the HR is > 1.55 reduces to 0.08.

Under the AT analysis all estimates of the HR show a greater preference for the intervention and the probabilities that the HR is > 1.55 are increased (compared with the ITT analysis). However, the 95% CrIs are wider indicating less certainty and the 95% CrI for the sceptical prior still includes both positive and harmful effects.

A plot of the posterior distributions under the three prior assumptions for the ITT analysis is shown in Figure 29 . Figure 30 shows the prior distributions overlaid with the updated posterior distributions.

FIGURE 29.

Posterior distributions for the HRs under the ITT analysis assuming proportional hazard rates.

FIGURE 30.

Prior and posterior distributions for HRs (ITT) overlaid, assuming proportional hazard rates. (a) Posterior distribution; (b) posterior distribution (dashed line) and sceptical prior distribution (solid line); (c) posterior distribution (dashed line) and enthusiastic prior distribution (solid line).

Table 33 presents the results under the assumption of a time-dependent treatment effect. Two HRs estimate the difference in mortality rates between the two intervention groups during two time periods. HR1 is the estimate of the treatment effect between conception and 36.5 weeks and HR2 is the estimate of the treatment effect between 36.5 weeks and the end of follow-up. Table 33 includes the median estimate of both HRs with their 95% CrIs and the probability that the HRs are > 1.55. Once again, a HR > 1 favours treatment.

Under the ITT analysis, and for all prior assumptions, the estimate for HR2 suggests that the intervention has a higher survival rate in proportion to conservative management, compared with the estimate for HR1. The probability that the HR is > 1.55 is greater for the second time period than the first time period.

Under the uninformative prior the estimate of HR1 is 0.90 (95% CrI 0.25 to 3.04) and the probability of 0.19 (19%) that this HR is > 1.55. The estimate of HR1 under the enthusiastic prior is closer to the null (1.42, 95% CrI 0.88 to 2.29) with a probability of 0.37 that this HR is > 1.55. Under the sceptical prior the estimate of HR1 is even closer to the null at 0.99 (95% CrI 0.62 to 1.59) with a probability of 0.03 that HR1 is > 1.55.

Under the uninformative prior the estimate of HR2 is 4.73 (95% CrI 0.96 to 37.13) and the probability that this HR is > 1.55 is 0.91. Under the enthusiastic prior the estimate of HR2 is closer to the null (1.72); however, the CrI is narrower (95% CrI 1.05 to 2.80) and the probability that this HR2 is > 1.55 is 0.66. For the sceptical prior the estimate of HR2 is even closer to the null (1.16, 95% CrI 0.71 to 1.90) with a probability of 0.13 that HR2 is > 1.55.

Under the AT analysis (see Table 33 ) all of the estimates of the HRs show a greater preference for the intervention for both time periods and the probabilities that both HRs are > 1.55 are increased (compared with the ITT analysis). Again, the 95% CrIs are wider indicating less certainty.

A plot of the posterior distributions for HR1 under the three default prior distributions and the ITT analysis is shown in Figure 31 . Figure 32 displays the individual prior distributions overlaid with their updated posterior distributions for the estimate of HR1. Figures 33 and 34 are the corresponding plots for the estimate of HR2.

FIGURE 31.

Posterior distributions for HR1 (period conception to 36.5 weeks) under the ITT analysis.

FIGURE 32.

Prior and posterior distributions for HR1 (period conception to 36.5 weeks) (ITT) overlaid. (a) Posterior distribution; (b) posterior distribution (dashed line), sceptical prior distribution (solid line); and (c) posterior distribution (dashed line), enthusiastic prior distribution (solid line).

FIGURE 33.

Posterior distributions for HR2 (period 36.5 weeks to the end of follow-up) under the ITT analysis.

FIGURE 34.

Prior and posterior distributions for HR2 (period 36.5 weeks to the end of follow up) (ITT) overlaid. (a) Posterior distribution; (b) posterior distribution (dashed line), sceptical prior distribution (solid line); and (c) posterior distribution (dashed line), enthusiastic prior distribution (solid line).

Limitations

A Bayesian analysis is heavily reliant on the choice of prior distributions, especially when the amounts of data available are small, as in the PLUTO trial. Although a large number of opinions were sought, the priors will be valid only if the questionnaires were properly interpreted and we were successful in eliciting opinions. Indeed, with hindsight, one of the questions, although intended to be on a relative scale, may have been interpreted as being on an absolute scale. Furthermore, opinions were elicited on expected outcomes under conservative management and expected change under VAS. Analysis methods therefore required substantial manipulation of these opinions into pooled prior distributions for analysis within the posterior modelling, and so analysis is further reliant on the appropriateness of our pooling techniques and in particular on the assumptions of normality invoked.

Of note, some of the values published here for the pooled expert beliefs do not quite correspond to earlier published values. 119 This is because of a small mistake in the formula used in the earlier analysis. In addition, the questionnaires referred a little ambiguously to the outcome being both perinatal mortality (28 days) and 6 weeks. It is not expected that elicited beliefs were sensitive to this anomaly.

The analysis had to be restricted to an unadjusted perinatal and survival analysis because of the very small sample size. It is likely that there will be other factors related to survival other than the intervention variable; however, these could not be included. The variables that could have been considered for the analysis had missing data, which, if incorporated, would have reduced the number of subjects included in the analysis and further reduced the power. Therefore, any results shown must be interpreted with a high degree of caution.

For perinatal survival, the RR estimated from this Bayesian (ITT) analysis using the uninformative prior is similar to but has a wider CrI than the frequentist counterpart (RR 1.93, 95% CrI 0.77 to 6.25 vs. RR 1.88, 95% CI 0.71 to 4.96). This is somewhat contrary to the Bayesian paradigm, that of increasing the effective sample size by including additional ‘data’, but we are not the first to observe such a phenomenon. 123,124

Finally, the primary outcome stated within the protocol for the PLUTO study was survival at 28 days. In concordance with this protocol all analysis subsequently treated the outcome event as survival (as opposed to death). However, in a survival analysis the natural outcome is death. So that the effect estimate on the HR scale was consistent with RR estimates (in so far as treatment effects > 1 will favour the intervention), the survival analysis had to take the somewhat unusual parameterisation of being a control versus treatment comparison.

Conclusion

In conclusion, the PLUTO trial failed to recruit its target sample size and the results from the analysis of the RCT data were therefore inconclusive. Fortuitously, before the trial had commenced the triallists had sought opinions from consultants in the specialties of fetal medicine, paediatric urology and paediatric nephrology. These opinions were elicited in such a way as to enable the construction of informative Bayesian prior distributions, which, when combined with the PLUTO trial data on the 31 pregnancies, provide Bayesian posterior probabilities that the treatment effect (RR) is greater than the 1.55 value that the study was initially designed to detect. This was complemented with Bayesian analysis based on what are commonly referred to as off-the-shelf priors (an uninformative prior, a sceptical prior and an enthusiastic prior).

Importantly, we found that our conclusions were dependant on the choice of prior. Under the ITT assumption and using an elicited prior, the posterior probability that the RR is > 1.55 (a clinical and prespecified difference) is 25%. Results from the other priors were equally or less optimistic and only under the enthusiastic prior did we find evidence that VAS is beneficial and even then the probability that the RR is > 1.55 is still only 71%. Under all other priors we found that it was not possible to rule out harmful effects of VAS.

Finally, we also observed that the clinicians from whom we elicited opinions were highly sceptical about VAS but also quite variable in their opinions. This has several important implications. First, it is likely that it contributed to the failure of the trial to recruit the number of participants required. Therefore, although not having intentionally set out to explain the low recruitment rate, eliciting Bayesian priors may also have provided an explanation for the failure of the study. Second, because the clinicians were sceptical and the number of patients recruited to the study small, the estimated treatment effect obtained from these Bayesian analyses are pulled towards the null.

Model structure

A model-based economic analysis was carried out based on the data from the PLUTO trial. A decision tree model was felt to be most appropriate for this analysis given the small number of participants and the short time frame of the study. The model was constructed in TreeAge Pro 2012 (TreeAge Software Inc., Williamstown, MA, USA) by the authors. The structure of the model was informed by clinical input with regards to presentation of the condition and treatment pathways as well, as by systematic reviews published in this disease area. 13,21 In this model we examine the antenatal, perinatal and postnatal progress of the affected fetus within the trial following randomisation to VAS ( Figure 35 ) or conservative management ( Figure 36 ). Pathways experienced by the patients within the trial over a period of 1 year are represented. Data collection at 2 years was not complete at the time of writing this report and hence was not included in the analysis. Health states used in the model are defined in Table 34 .

FIGURE 35.

The VAS pathway.

FIGURE 36.

The conservative management pathway.

Data inputs for the model

Outcomes

The survival outcomes explored in this analysis are summarised in Table 14 . In addition to survival to 28 days and 1 year we considered a composite measure, namely remaining disability free at the end of 1 year. This measure incorporated all children who survived up to 1 year of life without any renal or other chronic morbidity.

There is currently no validated method available for reliably estimating quality-adjusted life-years (QALYs) in children aged < 2 years. A cost–utility analysis could not therefore be performed. Although a neurological development assessment using a validated PARCA questionnaire was initially planned in the study protocol, this was not available for all patients at the time of writing this report and therefore was not included.

Assumptions

The following assumptions hold true for all of the analyses performed:

• Data relating to drug use were not collected within the trial and were assumed to be minimal. These costs are therefore not included in the analysis.

• All centres (UK and international) were assumed to have similar expertise and protocols for management of these patients.

Analysis

The results are reported as incremental cost per outcome of interest, that is, as cost per infant who survived to 28 days, who survived to 1 year or who remained disability free at the end of 1 year.

Base-case analysis

The results of the analyses are presented in Tables 44 – 46 . For the base case, in the ITT analysis the average cost incurred within the VAS arm was £20,851 for 1.38 survivors at 28 days. The average cost incurred in the conservative arm was £9868 for 0.67 survivors at 28 days, resulting in an ICER for 28-day survival of £15,506 per survivor in the VAS arm. The ICER indicates that, within the ITT analysis, for every additional child who survived up to 28 days (and 1 year) a cost of £15,506 (and £15,415) was incurred. Similar results were obtained in the method 2 analysis and uniform prior analysis (method 3). The ICER results for being disability free at the end of 1 year were substantially higher for all analyses as very few patients survived to 1 year without any morbidity.

Deterministic sensitivity analysis

1. DSA1 and DSA2 (changing VAS costs). As there was some uncertainty associated with the cost of VAS, we explored the effect on the ICER of changing the VAS cost. Accordingly, VAS costs were doubled (DSA1) or halved (DSA2) and the analysis was repeated. This did not result in significant changes in the ICER values. Thus, the results of the base-case analysis were not sensitive to changes in VAS cost.

2. DSA3 (normalising the high costs of outliers). When the high costs incurred by a few of the patients within the trial (patients 1015, 1020 and 1023, all within the VAS arm) were substituted with the average cost incurred by all other patients in that arm, the overall costs within the VAS arm reduced significantly and the VAS arm dominated over the conservative arm in the ITT and per-protocol analysis. The savings were even higher in the method 2 analysis. Thus, patients receiving VAS incurred lower costs overall whilst accruing more benefits. Although costs in the uniform prior analysis also reduced significantly, dominance was not reached.

3. DSA4 (as per treatment received). Two patients who were randomised to the conservative arm (patients 1027 and 1029) received a VAS and three patients randomised to the VAS arm (patients 1013, 1016 and 1028) did not receive a VAS. When the analysis was repeated as per the treatment received, the incremental costs increased slightly (£14,124 vs. £10,983 in the base-case analysis) and the gains in effectiveness were higher (1.7 vs. 0.71 in the base case), resulting in substantially lower ICERs for survival at 28 days and 1 year. The ICER for those who remained disability free at the end of 1 year did not change significantly however.

Probabilistic sensitivity analysis

Probabilistic analysis using 20,000 simulations as per the Monte Carlo principle did not result in a significant change to the base-case ICERs. The scatter plot representing survival at 28 days (for the ITT analysis) is shown in Figure 37 . The ellipse in this figure represents 95% of all data points and most of these lie in the upper right quadrant of the cost-effectiveness plane. This illustrates that VAS placement is very likely to be more effective whilst being more expensive for this outcome.

FIGURE 37.

Incremental cost-effectiveness scatter plot of VAS vs. conservative for the base-case ITT analysis for survival at 28 days. Note: the dotted lines represent the ‘status quo’ with respect to cost and effectiveness.

Figures 38 – 40 represent the probability of VAS being cost-effective for a given outcome of interest at various willingness-to-pay (WTP) thresholds. Figure 38 represents the cost-effectiveness acceptability curve for survival at 28 days using ITT data, which suggests that, at an arbitrary WTP threshold of £40,000, the probability of VAS being the acceptable treatment modality for achieving survival at 28 days is > 80%. However, if the WTP threshold is low at £10,000, the acceptability of VAS falls to < 30%. Similarly, Figures 39 and 40 represent the cost-effectiveness acceptability curves for survival at 1 year and disability-free survival at 1 year, respectively, using ITT data. Thus, at a WTP of £40,000 per survivor at 1 year, the probability that VAS is cost-effective is > 80%, whereas at a WTP of £40,000 per 1 disability-free year there is a < 50% chance that VAS is cost-effective. It should be noted that the WTP threshold of £40,000 is arbitrary and relates to the cost per infant surviving to 28 days or 1 year or achieving 1 disability-free year and is not comparable to NICE’s preferred cost-effectiveness threshold of £20,000 per QALY.

FIGURE 38.

Cost-effectiveness acceptability curve representing the likelihood of the acceptability of VAS for survival at 28 days at various WTP thresholds (ITT analysis).

FIGURE 39.

Cost-effectiveness acceptability curve representing the likelihood of the acceptability of VAS for survival at 1 year at various WTP thresholds (ITT analysis). Reproduced from Diwakar L, Morris RK, Barton P, Middleton LM, Kilby MD, Roberts TE. Evaluation of the cost effectiveness of vesico-amniotic shunting in the management of congenital lower urinary tract obstruction (based on data from the PLUTO trial). PLOS ONE. 2013; (in press). 125

FIGURE 40.

Cost-effectiveness acceptability curve representing the likelihood of the acceptability of VAS for disability-free survival at 1 year at various WTP thresholds (ITT analysis). Reproduced from Diwakar L, Morris RK, Barton P, Middleton LM, Kilby MD, Roberts TE. Evaluation of the cost effectiveness of vesico-amniotic shunting in the management of congenital lower urinary tract obstruction (based on data from the PLUTO trial). PLOS ONE. 2013; (in press). 125

Figure 41 shows the value (per patient) of resolving the uncertainties in data obtained in this trial to obtain ‘perfect information’. This parameter, known as the expected value of perfect information (EVPI), reflects uncertainties in both costs and outcomes to calculate what additional amount it is worth paying at each WTP threshold to obtain perfect information. For example, at a WTP threshold equivalent to the ICER obtained by PSA (£15,482 for survival at 28 days using the ITT analysis), it is worth just over £3000 per patient to obtain perfect information. The ‘per-patient’ figure must then be multiplied by an estimate of the number of patients who will benefit from the results of a future study to reduce the uncertainty in the results. The result of this multiplication provides an upper limit on the amount that it would be worth spending on such a future study.

FIGURE 41.

A representation of the EVPI for 28-day survival using the ITT analysis.

Principal findings

Percutaneous VAS, although being more expensive, might lead to improved perinatal and 1-year survival in fetuses with LUTO. It does not appear to reduce long-term renal morbidity (at 1 year of age) in these infants as there was very little difference in the probability of being disability free at 1 year between the VAS arm and the conservative management arm.

The results were not sensitive to changes in the cost of the VAS. When the analysis was repeated according to treatment received (and not as per randomisation), the costs in the VAS arm reduced relative to the base-case costs, rendering the VAS less expensive for the same survival outcomes. Given the small number of patients within the trial, it may be argued that the uniform prior analysis is the most appropriate for this study as it considers all possible outcomes for the patients. Nevertheless, as ITT analysis is generally accepted as the least biased method of analysing data from RCTs, the conclusions in this report are based on the results of the ITT analysis. However, as demonstrated in Tables 44 – 46 , the ICER values for the base-case for survival at 28 days and 1 year in the uniform prior analysis do not vary significantly from those of the ITT analysis or indeed from the analysis excluding patients who underwent TOP.

We also found that some patients who received a VAS utilised a disproportionate amount of resources within the first year of life. Normalising these high costs improved the cost-effectiveness of VAS, even conferring dominance in the ITT analysis and analysis excluding TOP patients (method 2). Although for the purposes of this analysis we assumed that the high cost incurred by a few patients in this trial was entirely coincidental, it is unlikely to be the case. It is remarkable that all of the patients with a high overall cost (> £50,000) were within the VAS group. The high costs incurred were almost entirely due to prolonged NICU admission immediately after birth in these infants (data not shown). It is possible that these infants would have succumbed to LUTO in the absence of VAS (the infants with ‘poor prognosis’ on antenatal scans70). These children continued to have higher admission rates in the first year (beyond the neonatal period) and are responsible for the higher costs in the VAS arm relative to the conservative arm. Ironically, these are the patients who are deemed to benefit the most from VAS placement (in terms of survival). 13,21

The National Institute for Health and Care Excellence recommends that cost-effectiveness be expressed in terms of cost per QALY. 133 Cost–utility analysis enables comparison of costs across disease areas and is considered the cornerstone for health policy decisions within the UK. However, as discussed previously, QALYs or other utility measures could not be obtained in this trial. Nevertheless, we could extrapolate the ICER results obtained to surmise that an extra 0.77 QALYs would need to be obtained in the VAS arm compared with the conservative arm to make the intervention cost-effective at 1 year (assuming a cost–utility threshold of £20,000 per QALY). From the data obtained from this trial and other observational evidence, it appears unlikely that VAS would result in such significant QALY gains in these infants.

Strengths and limitations of the study

The main strength of this economic analysis is that it is the first analysis that is based on a RCT in this clinical area. The data are directly derived from the PLUTO study, which is the largest RCT (albeit small in absolute terms) in this clinical area to date. This analysis benefits from collaborative work between clinicians, nurses, statisticians and health economists. Furthermore, resource use data were collected prospectively within the trial and represent actual UK NHS costs incurred by the participants.

One major drawback of the study was the small sample size, which introduces uncertainties in the analysis. However, given that LUTO is a very uncommon condition, it may not be possible to carry out larger studies in this disease area in the future. To improve the robustness of our findings we carried out various different analyses for the base case and also incorporated extensive deterministic and probability sensitivity analyses.

Another drawback is that these findings apply only to those babies who were included in the randomised arms of the PLUTO trial. All those babies for whom the clinicians felt sure of the advantages (or futility) of VAS were included in the registry arm of the trial and were not considered in this analysis. Thus, the overall cost-effectiveness of VAS may have been under- or overstated in this analysis.

Further, it can be argued that the longer-term effects of LUTO (e.g. cognition, continence, infertility and need for dialysis or renal transplantation) will have a bearing on the overall effectiveness of VAS in the management of these patients. Indeed, the actual costs of renal morbidity do not become apparent until the child is aged > 1 year as dialysis is not favoured in infants. Resource use data from the longer-term management of these children will, therefore, be instrumental in judging the cost-effectiveness of the intervention. It was not possible to include such data in the current economic evaluation as no reliable data are currently available and the participants within the PLUTO trial are still very young.

Strengths and limitations in relation to other studies

Despite the difficulties with recruitment, the PLUTO trial is the largest RCT to study the effectiveness of VAS for the management of LUTO and the first to prospectively estimate the cost-effectiveness of the intervention.

Some researchers who have studied these patients prospectively have reported significant renal morbidity (notably end-stage renal failure requiring dialysis or transplantation)68 and others have found acceptable quality-of-life scores in a majority in the long term. 83 Such data, which will be useful in understanding the overall cost-effectiveness of VAS, are not currently available from the PLUTO trial and hence were not included in this analysis.

Meaning of this study

The results of this analysis suggest that VAS is costlier than the conservative option for the management of LUTO and there is no statistically significant increase in survival in these infants after VAS. Most infants who survived to 28 days managed to survive to 1 year in this study.

However, taking into account the uncertainty in the data, the ICER could be much higher than the base-case estimate. It is possible that VAS is both more costly and less effective than conventional treatment.

Using survival alone as the outcome of interest in these patients will be misleading as most of the survivors are left with significant renal morbidity. Therefore, the effectiveness of the intervention can be overstated if only survival is considered in the analysis. It is worth mentioning that a small study of children with LUTO (aged 1–14 years, mean 5.83 years) who received intrauterine VAS reported quality-of-life scores that were comparable to those of the general healthy population. 83 Thus, longer-term follow-up of patients recruited into the PLUTO trial may provide greater clarity regarding VAS effectiveness in this group of patients.

Unanswered questions and future research

To obtain a better understanding of the outcomes of VAS in LUTO, a larger study needs to be carried out. However, this may be difficult given the infrequent occurrence of this condition. The long-term effects of VAS on these patients need to be determined to establish the true clinical effectiveness and cost-effectiveness of the intervention. As reliable longer-term data are not currently available for VAS, we did not attempt to extend our model beyond 1 year in this analysis. However, within the PLUTO trial, longer-term follow-up of patients in the registry as well as in the randomisation arms is planned and these data can be input into the model when available.

Aim of the study

To gain an insight into the experiences and perceptions of pregnant women asked to participate in an interventional fetal medicine trial requiring an invasive procedure.

Objectives

• To critically explore pregnant women’s views and perceptions of the PLUTO trial.

• To critically examine influences on the decision by women to participate or to decline to participate in the PLUTO trial, with particular regard to the impact of the information provided to them.

• To ascertain influences on the decision-making process of women opting for TOP following an antenatal diagnosis of fetal LUTO.

A phenomenological research approach was adopted in this study as this was considered to be consistent with the purpose of the research. By following a phenomenological approach to inquiry it was anticipated that greater insights into how pregnant women in this particular situation make sense of their experiences would be obtained.

Sample

Purposeful sampling is an appropriate technique in phenomenology136 as it permits the selection of interviewees whose experiences enable an understanding of the phenomenon in question. 137 All women who were invited to participate in the PLUTO trial were also approached to take part in this qualitative study, regardless of whether they decided to enter the trial or not. A separate information leaflet was given and a consent form signed for this study.

In qualitative research a priori predictions of sample size can be difficult but observation of the basic qualitative rule of data saturation is the norm, whereby researchers continue to sample until no new or relevant data appear to be emerging in any given category and the relationships between the categories are well established. 138 In this case it was anticipated that to achieve saturation women would be recruited in similar numbers from amongst the randomised and registry patients and from amongst those requesting TOP. As the PLUTO trial was concluded early, recruitment for the qualitative study was prematurely halted and so data saturation was not achieved.

The sample comprised six women who had received an antenatal diagnosis of fetal LUTO between 12 and 29 weeks of gestation. Women interviewed were both nulliparous and multiparous and aged between 25 and 41 years. Five women who took part in this research defined themselves as white British and one described herself as black African of sub-Saharan origin. Of the six women interviewed in this study two were randomised in the PLUTO trial, three were registered and one had opted for TOP following her diagnosis. Interpreters were not required in this study as all of the women spoke English to a sufficient standard, and in only one instance was the woman’s partner present.

Methods

The phenomenological approach to data collection used in this study was consistent with the intention to gather participants’ experiences in a way that was internally meaningful and set within the context of their lives. 139 Data were collected using in-depth conversations between the researcher and the women as coparticipants. 140 An interview guide was used to ensure data collection on relevant topics (see Appendix 11 ) but participants were free to raise any issue of importance or concern to them. 141

Interviews were conducted at a time and in a place that were convenient to the participant and lasted between 35 and 50 minutes. All were tape-recorded with permission and transcribed verbatim. A form of study verification referred to as ‘member checking’ was conducted by providing the participants with a typed copy of their interview transcript to read through; they were asked to indicate whether they considered it to be an accurate reflection of what they had said during their interview. All but one interviewee responded to this request for further verification and all who responded considered the content of their interview transcript to have been accurate and fair.

A reflexive diary was also maintained by the researcher to record contextual details and reflections on the research process,95 including emerging impressions and views. 142 Limited demographic data including current age, parity, gestation at diagnosis and whether the pregnancy was singleton or multiple were collected from participants at the beginning of the interview.

Data analysis

Phenomenological data analysis involves a disciplined process designed to ensure that the details of experiences intimately contribute to an articulation of a level of generality that is helpful to one’s interests. 143 In consideration of this perception the data analysis process described by Burnard144 was adopted in this study. This began with the verbatim transcription of the interview tapes followed by intensive reading and rereading of the transcripts by the researcher to enable a search to be made for regularities, contradictions, patterns and themes within the data. Notes on general themes were made throughout the reading and rereading process to serve as memory joggers and to record ideas that the researcher had whilst working with the transcripts. The aim at this initial stage of the analysis was to become immersed in the data in an attempt to become more fully aware of the ‘life world’ of the interview participants. The transcripts were then reread and summarised again as whole documents. Individually the transcripts were then coded and further analysed to develop themes. The two main themes identified were antenatal diagnosis and participation in the study and emotional impact of diagnosis and sources of support.

A second researcher reviewed the original transcripts and the emerging themes as a form of researcher validation.

Antenatal diagnosis and participation in the PLUTO trial

Three of the women interviewed in this study highlighted the personal significance of visualising the fetal abnormality on ultrasound scan, particularly in terms of the obvious severity of the condition diagnosed. For these women the experience of seeing the abnormal ultrasound image represented a major turning point in their pregnancy, a time when they realised that things might not go well. Despite this visual evidence of an abnormality some women recalled sensing a difficulty in accepting the actual diagnosis and found themselves to be in a state of denial for some time. This they related to having previously achieved a successful and healthy pregnancy as well as the expectation that nothing would go wrong with this pregnancy:

The consultant did a scan and at that point it was clear to see that there was more of an issue because rather than being just a little dot on the scan, the baby’s bladder looked like a little water balloon inside it and even we could see that this was not good. So that was the big time for me when I realised that this was not good.

Mother 1, RCT, conservative management

Literally as she put the scan on me I could see that there was a problem because I couldn’t see anything hardly, I could just see a big black blob.

Mother 4, RCT, intervention

Most women reported the predominant emotions that they experienced following their antenatal diagnosis as being worry, distress and fear. For some the worry continued throughout their pregnancy whereas for others it lessened as time went on although it did not disappear entirely:

It was very upsetting for me to be honest and I worried about how things would be sorted out. I was worried and upset and it was hard for me to deal with especially being my first pregnancy, I didn’t expect anything to go wrong.

Mother 5, registry, conservative management

Emotions-wise I think worry was the main thing. Just worry that things were not going to go well. I think you worry for everyone involved and you worry for the little one inside, what stage he would get to, what would happen if he reached full term, were we making the right decision for him. We didn’t want to give up on him too soon.

Mother 1, RCT, conservative management

Some women described feeling a ‘sense of loss’ or ‘devastation’ after being told of their baby’s condition. Another common recollection for women was the fear that they might lose their baby, which for some was combined with a heartfelt desire to continue their pregnancy:

I guess for me it felt like I had lost something, like a death, that is the best way I can describe it. There was a sense of loss of this baby, I had chosen to have this baby and I wanted this baby, I didn’t want to start all over again.

Mother 4, RCT, Intervention

I was really emotional and I thought I would lose my baby. At the time I just kept thinking ‘is he in pain’ and things like that.

Mother 3, registry, conservative management

I am quite a logical person but in that situation I could not have even pictured a place in my head that I would have felt comfortable saying ‘please end this pregnancy’. There was just an overwhelming urge to continue and to think that everything just might be ok or I would never know.

Mother 1, RCT, conservative management

These extracts illustrate a common theme amongst women who participated in this study – what they described as a ‘rollercoaster of emotions’ following their antenatal diagnosis. Some even suggested that at times the emotions that they experienced seemed to conflict and proved challenging to resolve on a personal level. So, for example, the woman quoted below describes feeling anger at being out of control when first receiving a diagnosis, but a later sense of relief at having decisions taken out of her hands:

That is when the conversations started about the trial and that is when I had feelings of anger. I think this came mainly because of feeling out of control, not having any control of what might happen, and not just us but the doctors not having any control over what was then going to happen. It felt very unfair, that was probably the most difficult time.

When we went for the scan and realised that the baby’s heart was no longer going, we both felt a sense of relief, we didn’t want to make that decision [to terminate the pregnancy] and it got took out of our hands so that was ok because nature had decided and it was not meant to be and we could live with that an awful lot better.

Mother 1, RCT, conservative management

Partners and significant others were also affected by strong emotions on receiving an antenatal diagnosis of LUTO: ‘My partner definitely, it has caused quite a lot of extra ripples for him. He was extremely depressed and still is quite low but getting through it. I think it’s just sadness now really’ (mother 6, TOP, conservative management). However, the feelings of partners were sometimes at variance with the emotions of the woman, in terms of dealing with the diagnosis and getting through the following months:

I think he has just blocked himself off from hearing anything negative and I think that if anything was to happen when the baby is born, I would be more prepared than he would and I think that is when we have a problem.

Mother 4, RCT, intervention

My partner is supporting me but he mentioned that we should perhaps consider a termination and I found that very difficult to accept from him.

Mother 2, registry, conservative management

One woman reported that she worried deeply throughout the pregnancy that her husband had not grasped the potential gravity of the situation and remained somewhat in denial of the fetal diagnosis:

He thinks that everything will be fine and he cannot understand why we are having a baby but we are not preparing for a baby so I think sometimes the person I need to speak to I can’t because he won’t listen to me or he won’t have it any other way.

Mother 4, RCT, intervention

Other family members were also affected by a diagnosis of LUTO and it was felt that this could lead to them experiencing increased levels of anxiety during subsequent pregnancies:

When I lost the baby in the end I think that my family were extremely sad. I don’t know if they would be more wary if I became pregnant again but that would be interesting to see, how people react then.

Mother 6, TOP, conservative management

Three women described feeling a specific need to use the internet to find out further information about LUTO following their antenatal diagnosis, such as its causes, treatments and effects:

The internet is a wonderful thing and you can get so much information, I know we certainly tend to seek out more information to try and understand what has happened or what could happen.

Mother 1, RCT, conservative management

I am a bit of a web hound anyway so I found it very easy to go and find the right information but if you had been pointed to the website and Google and were told go and have a look and you were not really sure what you were doing maybe that would have been difficult.

Mother 6, TOP, conservative management

In contrast, some women reported a lack of reliable information on the internet, which created a sense of confusion and frustration within them as conflict arose between their personal experiences and the experiences of others in managing this particular complication in pregnancy:

To have seen stuff out there and I cannot say whether that is right or wrong, and then to be sitting in a room getting told we cannot say whether it [a treatment] works or not and all it comes down to is whether a computer says yes or no, was very, very hard.

Mother 1, RCT, conservative management

Commonly, women reported that they felt themselves to be in ‘safe hands’ following their diagnosis and appreciated the particular reassurance provided by regular ultrasound scanning. They reported that in general they were satisfied with the standard of care that they received during their pregnancy and felt relieved that the condition was detected antenatally, thus allowing them the opportunity to prepare for the delivery and the potential postnatal course:

Going for the scans put my mind at ease. I hadn’t been looking forward to going for them but I am glad I have had the opportunity to go every 4 weeks and keep an eye on his kidneys and stuff.

Mother 3, registry, conservative management

However, for one mother who reported that each scan seemed to reveal more problems, this was more worrying:

It is actually the scans that frighten me because I think what is going to be coming up next because one minute it is the kidneys and the bladder, then they say the head hasn’t grown much, then it’s the fluid around him and I am thinking every time I go for a scan everything like seems to be going wrong.

Mother 5, registry, conservative management

The obvious value and genuine reliance that some women placed on this form of medical intervention during pregnancy was evident within the interviews, and some women suggested that regular scanning assisted to some degree with bonding with their baby. This contrasts with the ‘turning point’ that women described feeling on undergoing their first scan. Certainly none of the women interviewed appeared to have any safety issues with regard to receiving multiple and sometimes lengthy ultrasound scans during pregnancy. Possibly the perceived benefit that regular ultrasound scanning provided to them both on a personal level and with regard to their fetus outweighed any risks that they may have associated with the actual procedure and its potential cumulative effect in pregnancy.

In general, women interviewed were highly satisfied with the standard of the fetal medicine care that they received and reported a sense of relief that their pregnancy complications had been detected before their baby’s delivery:

We were getting the best advice and that is as much as you can ask for really, to know that the people you are dealing with are the right people and we totally, totally felt that.

Mother 1, RCT, conservative management

I was pleased with the treatment I was given and I was pleased it was picked up so soon and obviously having the extra scans done put my mind at ease.

Mother 3, registry, conservative management

Influences on the decision by women to participate or to decline to participate in the PLUTO trial

Various influences were presented by women to account for their personal reasons for participation in the study. A common perception among the women interviewed involved them deriving some personal benefit from taking part, either for this or subsequent pregnancies. In addition, women frequently reported that they were encouraged to participate for more altruistic reasons. Indeed, altruism is commonly reported by women as a reason for research participation during pregnancy:145,146

It was more that if I did go on to have another child I would know what had possibly caused the problem and I felt that [taking part] would help other mothers, pregnancies and other people’s daughter.

Mother 5, registry, conservative management

Me and my husband wanted to go into the study and I felt happy to do that because at the end of the day it is just to help somebody.

Mother 3, registry, conservative management

It was also reported by some women that participation in the study was considered to be their only perceived real option at the time. In addition, women often viewed the opportunity to participate in the study as being able to offer their baby a ‘chance’, a lifeline as one mother described it:

I really didn’t feel that there was any other option, we were just going along with ‘let’s see how this goes’,

Mother 1, RCT, conservative management

I suppose for both of us really it was just that we wanted the baby and we wanted a chance so the trial for us was still a chance. Although you don’t know, it seems to make sense to us that it [the treatment] is something that could help so why not just do it. When you have been through something so dire, to have a little bit of hope is like being thrown a lifeline.

Mother 4, RCT, Intervention

Further reasons reported to influence participation related to women’s perceptions of the extent of existing ‘damage’ to the baby, particularly in earlier gestation, or of the treatment procedure itself:

It was the results of the scan really when professor described the damage that was already there and knowing that even with the trial intervening the damage was already done.

Mother 6, TOP, conservative management

The actual treatment procedure really scared me and I didn’t want to be randomised in case I was the one chosen to have the shunt put in.

Mother 2, registry, conservative management

Personal faith also represented an influential factor relating to study participation but this was raised by only one woman: ‘I am praying a lot and hoping that the baby will be ok but as a Christian I believe that we shouldn’t interfere with nature’ (Mother 2, registry, conservative management).

In general, the women felt that they had fully understood the concept of the study when it was introduced to them. Most also considered it beneficial to have had the study explained to them by someone closely involved in the project, that is, a midwife linked directly to the research:

I think having someone to explain it to you first would be good and then have information to go away and think about.

Mother 1, RCT, conservative management

I was happy with the way I was told, I had plenty of leaflets about the study and you [research midwife] explained everything to me anyway, so I totally understood.

Mother 5, registry, conservative management

It was suggested by some women that receiving more information might have been helpful when making their decision whether to participate in terms of assisting them to make a more informed choice: ‘I think the way they did it (explained about the study) with us was fine but I think maybe if there is anything where there are case studies or statistics that you can give then that might help’ (mother 6, TOP, conservative management). The concept of informed choice and patient consent is a particularly complex and debated issue and it could certainly be argued that within the context of professional–patient interactions the facilitation of truly informed decision-making and consent is questionable given the subtle power that health professionals naturally exert over their patients.

Some women in the study felt that they did not require any additional time beyond the minimum of 24 hours as encouraged for ethical purposes to make a decision to participate and considered this to have been a hindrance to the recruitment process: ‘I just wanted desperately to be picked and I wanted them to do the shunt that day. I didn’t want to wait, I didn’t want to wait any more than what I had already’ (mother 4, RCT, intervention).

However, some women reported that they would have particularly appreciated being able to communicate with someone who had personal experience of this diagnosis in pregnancy, suggesting that this might have been helpful when making a decision whether or not to participate in the study. The reasons why some women expressed this preference were not altogether clear during the interviews but their need for this type of specific peer support can be surmised. For example, conflicting internet information available to some women in this study may have caused them to doubt what they were being told by professionals involved in their care and encouraged them to want to seek additional reassurance and support from others with direct experience of a similar situation in pregnancy. Similarly, some women may have particularly appreciated the potentially unbiased manner with which a peer with shared experience could have offered them focused support and information, in other words, they could have told them about the reality of the situation:147

We would like to have been told about others that had gone through the same thing as us as this would have been helpful.

Mother 3, registry, conservative management

I just would feel better if I could talk to somebody who has had a baby with it or been through the same.

Mother 4, RCT, intervention

Influences on women opting for termination of pregnancy

Although a number of women decided to terminate their pregnancy following diagnosis of LUTO, only one agreed to be interviewed. It is worth giving some space to her story while recognising that it is individual and cannot be held up as representative of women approached to take part in the trial who subsequently opted for termination.

The woman concerned and her husband received an initial diagnosis of LUTO at a district hospital and were referred to a fetal medicine centre:

You keep like this tiny bit of your brain in saying that we might get to [fetal medicine centre] and it’s a miracle and actually there is nothing really damaged and that something can be done which is a possibility and we know that now but at the same time you are trying to prepare yourself for the worst.

I think that felt like the longest week and the hardest and most horrific week of my life . . . it was a whole process of having to come to terms with the fact that your baby looks healthy and is acting healthy but isn’t and that you are ultimately likely to be facing a terrible decision and I think it was the deep concern, worry and fear about that is a distinct memory for me in that period.

Later in the interview this woman spoke of her feelings during the week between receiving a definitive diagnosis and prognosis and the termination of her pregnancy.

Fear was obviously the first thing and sadness obviously because as soon as something like that happens, maybe not everybody does but we both thought the worst and that this is likely to end badly and it was distinct sadness and grieving even before the decision had been made and I felt that we generally were grieving even until the point where we were offered the trial and told about the damage that was there and made that decision, which may seem premature to do so but we had a very clear idea of what was wrong and that it was almost as if we went into another stage of grieving. We both felt we were in a bit of cloud for days, depressed let’s say, didn’t want to really speak to anybody particularly.

Deciding on a termination was something that was agreed between this woman and her husband before they knew the prognosis, as they both had similar views on severe disability. It was this that influenced their decision-making:

Different people have different views on disability and all those different things and there is a level of disability that you can live with. I think we had both discussed this because of my age and what would do because you have to take these things this seriously because it can happen and I think we both said to ourselves if there is anything seriously wrong it will bring suffering to the child and make their life less than we would want it to be and we would have to make a tough choice and we were both of the same mind. So as soon as this happened I think we both knew what decision we would make unless we can reverse this and it would be fine. You know what I mean? Or if it was a minor ongoing complaint that needed treatment that would have been fine but it is severity of it. It was grief for a healthy baby, grief for what we had planned if I am honest. It is the healthy child. We had worried about all sorts of things, who was going to give up their job, are we going to be able to go out on Friday night and then this happened and you say none of that matters, what mattered was that it was going to be a healthy child so I think it was probably that more than anything.

Mother 6, TOP, conservative management

She later expressed the view that had they been told at the fetal medicine department that the baby’s kidneys were in a better condition than expected and that some treatment could be offered then they would have continued with the pregnancy. A more favourable prognosis would also have encouraged this couple to take part in the trial, but because of the amount of damage already done to the kidneys they felt that it was too late for them.

Summary of findings

The study findings demonstrated that various factors appear to be influential to women when deciding whether to take part in research during pregnancy, which, for the purposes of simplified illustration, were reported in a manner consistent with the objectives of the study.

In summary, the majority of women interviewed in this study described the significant personal impact of visualising the fetal anomaly on ultrasound scan, with some remembering this as a particular ‘turning point’ in their pregnancy when it first became apparent that things might not go as they had expected. Some women also experienced difficulty in initially accepting the diagnosis and found themselves to be in a state of denial for some time. Some women attributed this denial to not expecting anything to go wrong with their pregnancy. The emotional impact for partners and significant others following diagnosis was also highlighted by some women in this study, who suggested that reactions had included sadness and depression and increased anxiety levels during this and potentially even future pregnancies.

The need for more detailed information about the condition and its implications during pregnancy and following delivery was highlighted by a number of women interviewed in this study, with many turning to internet sources for additional knowledge and support. Using the internet in this way was reported by some women to have been hugely beneficial, which is consistent with the findings of a global study on the influence of the internet on decision-making in pregnancy,148 which concluded that its impact on all aspects of pregnancy was highly visible, with information-seeking being the key theme. In this study some women also suggested that the information that they found on the internet led to feelings of confusion and frustration, particularly in terms of the apparent differences in the way in which the condition is managed in the UK and elsewhere.

Most women in this study described feeling in ‘safe hands’ following their diagnosis and found particular reassurance in the regular ultrasound scanning that they subsequently received, although this was not the case for one woman, who observed a worsening problem on subsequent scans. Some suggested that this regular scanning assisted to some degree with bonding with their baby and was especially valuable to them in terms of providing immediate reassurance that their baby was still alive. Most women interviewed reported that in general they were satisfied with the standard of care that they received during their pregnancy and felt relieved that the condition was detected antenatally, thus allowing them the opportunity to prepare for the delivery and the potential postnatal course.

In terms of the influences on participation in this study, personal benefit and altruism were reported as the most common reasons by women interviewed. This is consistent with the literature. Typically, women will agree to be recruited into clinical research when the benefit to the fetus is considered probable,146 and the mother’s/parent’s duty to their (sometimes unborn) child will usually be given a higher priority than consideration for the mothers own health. 145 Women undoubtedly feel an inherent responsibility for the child that they carry to the extent that they often modify their habits and behaviours during pregnancy. 146 As such, pregnancy may impair their ability as women to make ‘free’ choices, making them feel bound to accept interventions that might benefit the unborn child or causing them to decline treatment for themselves because of potential teratogenic fears.

Enrolment into this trial was described by one mother as being ‘her only real option’ in terms of being able to give her baby a chance. Personal faith, fear of the actual procedure and the knowledge that existing ‘damage’ to the baby could not be corrected were other reasons cited as influential in non-participation in this study.

Support from a clinical trial co-ordinator or research nurse with responsibility for trial recruitment has been found to be positively associated with recruitment to research studies. 149 In general, women interviewed for this study felt that they understood the concept of the study when it was introduced to them but reported that it had helped to have the details of the research explained to them by someone closely involved in the study,95 in this case a research midwife. It was acknowledged that the communication style of the researcher responsible for recruitment was particularly important to women considering participation in this study, with women citing that an open, honest and detailed discussion was important to enable them to make an informed choice with regard to participation in the research.

Strengths and limitations of the qualitative study

The strength of the study lies in the rich data obtained on a poorly understood phenomenon. The use of a phenomenological approach resulted in an in-depth understanding of participants’ experiences that were internally meaningful.

The limitation, as with all qualitative research, is that findings are not generalisable outside of the study population. In particular, the one woman who opted for TOP was not representative of either the trial population or the population of women opting for TOP, most of whom did not enter the trial. However, as the aim of the study was to gain insight into experience rather than to extrapolate to a wider population, her story merited inclusion. Moreover, as participation in an interventional fetal medicine trial is an under-researched area, it is not possible to obtain additional verification of findings by embedding them within existing literature. A further limitation is that because of low recruitment numbers it was not possible to achieve data saturation, and therefore it is not known whether there are additional important issues to uncover. Similarly, as the study was terminated early, it is not known how women’s perceptions about inclusion in the RCT would have altered with time. Women’s experiences of the trial were bound up with their experiences of receiving a fetal diagnosis during pregnancy, and the outcome of that pregnancy, in particular a negative outcome, may have influenced their ultimate views on the management of LUTO.

Consideration of the ways in which the participants’ perceptions of the interviewer influenced the interview interactions may have particular significance with regard to this study. Interviews were conducted by the dedicated research midwife involved in the PLUTO trial, who was required to maintain an effective professional relationship with all trial participants. It is possible that some women involved in the trial may have considered that this created some ‘distance’ between the midwife as a ‘professional’ and themselves as ‘patients’. Some participants may have assumed that the midwife would act purely in the best interests of them and their babies and therefore would not discuss involvement in research that might carry risks. 95 However, this patient–professional relationship could also lead to participants responding to questions in a way that they deemed socially or culturally acceptable, or expected.

Implications for practice

As data saturation was not reached within this study, further research with women who are asked to participate in both interventional and non-interventional clinical trials during pregnancy would provide additional data on the expectations of women/couples and their motivations for participating.

However, the findings of the study do raise issues that have implications for practice. First, they have demonstrated some of the influences on decision-making that women report when considering their potential involvement in research during pregnancy. Knowledge of the factors that appear to influence decision-making with regard to participation may be used to tailor future research designs to meet the needs of pregnant women and address the issues of importance to them around this difficult time. In addition, the requirement of women for more appropriate information sharing regarding their particular fetal anomaly seem to be a particularly important finding of this research and one that all clinicians involved in the recruitment of pregnant women to research studies should endeavour to meet.

The importance of ongoing focused discussion with regard to entry into clinical research is also a consideration for clinicians and research staff recruiting to clinical trials. Pregnant women may need to speak to professionals from different disciplines and specialties as well as dedicated research staff as they consider their potential involvement in research and may also require numerous consultations to take place before they feel able to decide whether to agree to participation.

It seems that an individualised approach to recruitment of pregnant women to research might be beneficial not only to ensure that recruitment targets are facilitated but also to enable the individual informational needs of women to be met. An individualised approach to recruitment of pregnant women into research studies will of necessity be mediated by both time restrictions and the availability of various clinical experts, but this could be facilitated by the use of dedicated research staff such as midwives who can focus on support during the entire research process. Indeed, the ability of women to be able to discuss the implications relating to their potential involvement in research in pregnancy was reported to be very helpful by the participants in this study.

Finally, although the experience of receiving a fetal abnormality diagnosis such as LUTO in pregnancy does not appear to deter women from embarking on a subsequent pregnancy, it should be remembered that the emotional impact that receiving this type of diagnosis infers is significant to women and their partners alike. The frequent visualisation of the fetal abnormality on ultrasound scan seems to provoke mixed emotions in families and serves as a reminder of the need for professionals to be highly sensitive during ultrasound scan assessments and during the discussions that follow them, and to be both observant and reactive to the needs of women undergoing regular ultrasound scan assessments in this respect.