Study population

All pregnant women who had a second trimester FAS in six of seven Welsh Health Boards between July 2008 and March 2011 were eligible for inclusion. At recruitment, women were asked to give written consent that the data from their ultrasound scan could be linked with routinely collected data on their child.

Inclusion and exclusion criteria

The population for this analysis was singleton children who were live-born between January 1, 2009, and December 31, 2011, to mothers in Wales who consented to take part in the study and for whom validated FAS data were available (Fig 1). Pregnancies with an unknown outcome (for example, because the birth happened outside of Wales) were excluded. Children whose information could not be assigned an anonymised linking field (ALF; for example, because they did not receive their healthcare in Wales or did not have a valid NHS number or other identification variables) were also excluded, as linkage with the healthcare datasets was not possible. Children were followed from birth until the occurrence of death, migration out of Wales, 3rd birthday, or December 31, 2014 (end of follow-up). Person-time was censored in cases of death and migration.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Cohort flow diagram.

ALF, anonymised linking field; QA, quality assurance; RPD, renal pelvis dilatation.

https://doi.org/10.1371/journal.pmed.1002859.g001

Definition of exposure

RPD was defined as fluid-filled dilatation of the renal pelvis measured on the axial section, with an AP diameter of 5.0 to 9.9 mm (with the callipers to be placed on the inner AP margins of the renal pelvic wall) at the FAS [25]. An additional reporting screen was added to the information system for radiological data storage and reporting in Wales (Radiology Information Service 2 [RadIS2]) to capture the scan data. Data were also collected on whether the dilatation was unilateral or bilateral. An expert quality assurance (QA) panel reviewed FAS images to confirm the presence of dilatation in accordance with the study definition; the number of cases of RPD reduced from 221 to 144 after this process (for detail, see [26]).

We also sought to obtain information from radiological investigations conducted later in pregnancy from RadIS2 and the Congenital Anomaly Register for Wales (CARIS). Data on the presence of dilatation identified later in pregnancy and/or up to 12 months postpartum, plus any measurements recorded, were extracted from radiological reports. In accordance with international guidance [6,12], measurements of 7.1 mm or greater later in pregnancy or postpartum were considered evidence of later dilatation. The exposure groups in the analysis were then stratified into (i) no RPD at FAS or later, (ii) no RPD at FAS but RPD present later, (iii) RPD at FAS but not later, and (iv) RPD at FAS and later (Fig 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Definition of exposure groups for the analysis, and characteristics of hospital admissions.

IQR, interquartile range; RPD, renal pelvis dilatation; UTI, urinary tract infection.

https://doi.org/10.1371/journal.pmed.1002859.g002

Definition of outcomes

The outcome was a hospital admission for urinary tract causes in the first three years of life. An admission was defined as a continuous stay using a hospital bed provided by the NHS in Wales under one or more consultants, and included transfers between hospitals. A list of codes for possible causes of admissions in children for which RPD could be considered a possible precursor (including renal and lower urinary tract problems) was agreed by the study steering group, which included a consultant paediatric nephrologist, consultant radiologists, and an academic general practitioner (see S1 Table). The list was based on the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) [27] and the procedure code list used in the NHS (Operating Procedure Codes Classification of Interventions and Procedures version 4 or OPCS-4) [28]. Hospital admissions with any of these codes in any coding position were identified from the Patient Episode Database for Wales (PEDW). Admissions as a day case for postnatal investigations alone are not a part of this dataset, and these admissions would not therefore have been included.

Data linkage

Data from the Welsh Study of Mothers and Babies were exported to the Secure Anonymised Information Linkage (SAIL) Databank [29] to enable the radiological data to be linked with data on hospital admissions (from PEDW), congenital anomalies (from CARIS), deaths (from the Office for National Statistics Annual District Death Extract), and migration (from the Welsh Demographic Service data). For each of these datasets, individuals were assigned a unique identifier (the ALF) provided by the NHS Wales Informatics Service. The linkage system uses a combination of deterministic (based on NHS numbers) and probabilistic record linkage (based on first name, surname, date of birth, gender, and phonex and soundex versions of names); this linkage is more than 99.85% accurate [30]. Second-stage encryption is used by the data bank before storing data, and third-stage encryption is used to create project-specific linked datasets. Approval from the Information Governance Review Panel of the SAIL Databank was obtained for the analysis.

Sample size

Preliminary analyses of PEDW data showed that 2 per 100 children were admitted to hospital with a urinary tract cause before the age of five years. We therefore estimated that we would need approximately 21,000 children without RPD and 140 children with RPD to detect, at a 5% type 1 error rate, a 3-fold increase in urinary tract hospital admissions with RPD with 80% power. There were enough children in our cohort for this analysis.

Statistical methods

The statistical analysis was planned by the investigators of the study, in collaboration with the study steering group, in advance of the conduct of the analysis (for detail, see [24] and the protocol for this analysis, S1 Renal Study Protocol). This included the specification of the ICD-10 codes to be included in the definition of the outcome. An initial analysis of the association between RPD and hospitalisations was presented to the study steering group in 2016, and they recommended that the additional radiological data (from later pregnancy and postpartum) were sought and added to the analyses. We used Cox regression to model time to the first urinary tract hospital admission to three years of age. Our primary outcome was time to first hospital admission because most children with an admission were only admitted once (81.3%). We estimated hazard ratios (HRs) with 95% confidence intervals (CIs) to examine the risk of hospital admissions associated with the presence of RPD at the FAS, and then according to whether the child had RPD and/or later dilatation. The proportional hazards assumption was assessed graphically using log-minus-log plots and was tested based on the Schoenfeld residuals. We originally planned to include follow-up time until the child’s fifth birthday, but there was evidence that the proportional hazards assumption was violated in this analysis because all of the admissions in the RPD group occurred before the age of three years, whereas children without RPD continued to be admitted for the first time after three years of age. Based on this and on peer review comments, we chose a cutoff of three years of age for the follow-up period (instead of the preplanned five years), and formal testing confirmed there was no strong evidence that the proportional hazards assumption was violated in these models (for example, p = 0.11 for the binary RPD variable). Conclusions from models including five years of follow-up were similar to the models including three years of follow-up.

We examined associations in unadjusted models and conditional on other predictors of hospital admissions (sex, maternal age in three categories [<25, 25–34, 35+ years], deprivation quintile based on the UK Townsend Deprivation Score [31], and prematurity). We repeated the analyses to estimate HRs stratified according to whether the RPD was unilateral or bilateral. We also conducted sensitivity analyses: (i) adding information on dilatation identified during hospital admissions into the definition of exposure subgroups (that is, also using codes Q62.0, N13.0, N13.1, N13.2, or N13.3 in any position for a PEDW admission as evidence of later dilatation) and (ii) using an Anderson-Gill model to account for multiple urinary tract admissions during the follow-up period.

Data from radiological investigations to assess for dilatation later in pregnancy or postpartum were missing for 50 of the 138 children with RPD. There was also a low percentage of children with missing data on covariates (0.7% for Townsend score, 0.3% for prematurity). Multiple imputation with chained equations [32] was used to impute values for the missing data (10 imputations) under the missing at random assumption. The imputation model included all covariates, the outcome variable (urinary tract admissions), and the cumulative baseline hazard [33]. Conclusions from a complete case analysis and from the multiple imputation were similar. In response to peer-review comments, we present the results from the multiply imputed datasets.

All analyses were conducted within the SAIL Gateway using Stata version 15.1.

The study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (see S1 STROBE Checklist).

Strengths

This was a large population-based study in a cohort that was representative of all pregnant women in Wales. The stringent QA process was a strength of the study, with scan images reviewed by an expert panel to confirm that all cases of RPD conformed to the study definition. Linkage to routinely collected healthcare records was possible for 97% of women and children, ensuring that few were lost to follow-up. Unlike previous studies, we could compare outcomes in children with and without RPD. We used hospital admissions for urinary tract causes (including operations but not investigations) as a proxy for significant morbidity.

Limitations

The prevalence of RPD in this study was similar to estimates in previous studies. However, the absolute number of cases was small (n = 138), and these numbers reduced further when the data were stratified by the presence or absence of later dilatation. The estimates from the study must therefore be interpreted with caution. Although the small numbers mean that the CIs around the point estimates are wide, the effect sizes for the associations examined are large, suggesting that the sample size for examining these associations is adequate.

The SAIL Databank only includes healthcare information from Wales, which means that we were unable to access information on pregnancy outcomes that occurred in facilities outside of Wales or on mothers and children who ordinarily received their healthcare outside of Wales (for example, because they live along the border with England). The distribution of pregnancy outcomes, congenital anomalies, and premature deliveries in the cohort included in our analysis was comparable with those in published data for Wales [26], suggesting that they were representative of the general obstetric population in Wales.

We used routinely available healthcare records on hospital admissions to capture data on outcomes in this study, as these are available for the whole population in Wales. They are also an indicator of clinically significant morbidity. However, we acknowledge that this means that we have not included less severe outcomes in our analysis, such as UTIs managed in the community or in outpatients, underestimating the total burden of these outcomes in the general population. It is also possible that provider practice may explain some of the increased risk in admissions seen with persistent dilatation, as they may be more likely to admit a child with a known kidney abnormality for treatment and observation. Without access to community or outpatient data, we cannot assess the effect of this potential bias on the estimates obtained.

Obtaining records of radiological investigations later in pregnancy and postpartum was challenging. Although guidelines state that all cases of RPD should receive follow-up investigations, we found later tests for only 88 of the 138 children with RPD (64%). This does not mean that they did not have later investigations, but that we were unable to find a record of these. Previous studies have restricted their samples to children with complete data on later pregnancy or postpartum scans (for example, see [34]), but this leads to an incomplete picture of the natural history of RPD, especially as the characteristics of the excluded sample are usually unreported. We were able to identify additional cases of ‘hydronephrosis’ from hospital admission codes and use this information in a sensitivity analysis to reclassify some children who had no data on later radiological investigations. This strengthened our conclusions, as admission rates for children without persistent dilatation reduced in this analysis.

When information from later radiological investigations was available, it was clear that the number, timing, and type of test was not consistent between children and that different terminology and measurement thresholds were used to identify dilatation in the radiological reports. Our findings suggest that regular training, clear reporting protocols, and frequent audits to monitor follow-up rates and maintain reporting standards are needed if the management of RPD is to be standardised.

Implications for further research

Our results are consistent with findings from previous case series that have used repeated prenatal measurement of dilatation to predict which children require postnatal follow-up [35]. However, there were some additional questions that we were unable to answer in this study. For example, RPD was defined as an AP diameter of between 5.0 and 9.9 mm at the FAS, in accordance with the definition used by the antenatal screening programme in Wales when the data were collected [25]. Other classification systems [12] have further defined dilatation at the FAS as ‘mild’ (<7.0 mm) and ‘moderate’ (7.0 to ≤10 mm), and some screening programmes recommend that follow-up is only conducted when an RPD of ≥7.0 mm is identified at the FAS [8]. We could not assess whether there was a difference in hospital admissions between children with ‘mild’ or ‘moderate’ dilatation at the FAS, because the scan data collected for the study only contained information on whether RPD was present or not (but no measurements). Further studies are needed to document the natural history of dilatation and outcomes in children according to the actual measurement of the AP diameter at the FAS. We were also unable to draw firm conclusions about other predictors of poorer outcomes, such as unilateral compared with bilateral dilatation or the gender of the infant, because of sample size constraints. A larger study is needed to compare these subgroups. However, there was some evidence that children with bilateral RPD had a higher risk of hospital admissions even when there was no later evidence of dilatation, which is consistent with how bilateral RPD is currently managed in practice in Wales.

Two retrospective case series of patients known to specialist services have suggested that including other parameters (such as calyceal dilatation, renal parenchymal thickness, renal parenchymal appearance, bladder abnormalities, ureteral abnormalities, and oligohydramnios) in a formalised classification system is helpful in predicting which cases of dilatation will resolve spontaneously and which children require later surgery [36,37]. Further research is needed to fully understand the best combination of factors for predicting later pathology [9], and therefore allow for the development of care pathways with fewer follow-up visits or investigations for children identified as low risk [34]. Small studies have also examined different methods of visualising fetal urinary tract pathology (for example, conventional versus three-dimensional ultrasound) [38] and the use of serum and urinary biomarkers in the postpartum period to predict the risk of obstruction or impairment of renal function [39], but this research is currently inconclusive.

We also found that there were children who did not have RPD at the FAS (either 5.0 to 9.9 mm or ≥10 mm) who were found to have dilatation at later investigations. This is consistent with previous findings that not all cases of congenital dilatation are identified at the FAS [12]. In our study, this was a small group (44 of 21,101, 0.2%), but they had the highest rates of hospital admissions overall. As we did not have access to the full report from the FAS or later medical notes, we do not know why further investigations were conducted for these children. However, their hospital admission codes suggested that they had multiple congenital renal abnormalities and required more renal surgical procedures than children in the other groups. It is therefore likely that these children had other anomalies identified at the FAS, which led to the follow-up investigations and the repeated hospital admissions from an early age.

Implications for clinical practice

Clear protocols for reporting and further investigation and management of RPD are being developed (for example, [10]), and regular audits are needed to ensure that these are followed. The effectiveness of continuous prophylactic antibiotics in the prevention of UTI in children with prenatally detected RPD is unclear, with evidence based on observational studies. The first trials of prophylactic antibiotics for antenatal hydronephrosis with postnatally diagnosed vesicoureteral reflux are under way (ClinicalTrials.gov Identifier: NCT01140516). Postnatal infections still occur despite prophylaxis [40], and antibiotic resistance is higher with increased use [41]. Studies have found an increased risk of UTI with severe prenatal dilatation in females, uncircumcised males, and with specific postnatal diagnoses (ureteral dilatation, vesicoureteral reflux, or vesico-ureteric junction obstruction) [42,43]. We did not have access to labour ward, outpatient, or general practitioner records in this study and, as such, did not have information on which children were given antibiotics after birth. Although we are therefore unable to comment on the role of antibiotics in preventing UTI, the main cause of admission in our study was for UTIs and our results support the need for close postnatal follow-up of children with persistent dilatation.

Conclusion

RPD at the FAS is an important finding because it identifies fetuses who require later investigations. When these investigations are normal, we do not yet have sufficient evidence to tell whether there is an increase in hospital admissions for urinary tract problems in childhood. When there is persistent dilatation, hospital admission rates in childhood are higher. These results can be used to improve counselling of parents. Although this was a large study in a representative population of pregnant women in Wales, obtaining records of radiological investigations after the FAS was challenging, and scans were not conducted or reported consistently. Clear protocols for reporting and further investigation of RPD are being developed [10], and regular audits are needed to ensure that these are followed. Further studies should examine whether other characteristics at the FAS could improve the detection of renal pathology during antenatal screening.