Abstract
Objective : Comparison of birthweight references for diagnosing SGA. To provide denominator data for suspicion and diagnosis of SGA.
Design : A retrospective cohort study of 10,616 babies.
Setting : A regional obstetric centre.
Population : 10,616 consecutive newborns, born in Derriford Hospital, University Hospitals Plymouth NHS Trust (UPHT), whilst using the GROW package,1 compared with using Intergrowth 21st (IG21),2 and British 1990 (UK90) references.3
Methods : Statistical analysis of centile data from GROW, IG21 and UK90 references.
Main outcomes : Induction rates, detection of suspected and/or diagnosed SGA. Assessment of goodness of fit to the Plymouth population.
Results : GROW and IG21 showed bias. GROW had a systematic bias towards smaller centiles (skewness 0.169). IG21 had a systematic bias towards larger centiles (skewness -0.452). UK90 was best fit to the Plymouth dataset with insignificant bias across centiles (skewness -0.047).
Conclusions : GROW and IG21 are not appropriate gold standards for our population for allocation of birthweight centile. The size of the population suggests the conclusions may be extrapolatable to other centres. UK90 does not have everyday accessible tools compared with GROW and IG21. A continual local audit of birthweight would be ideal, enabling accurate local centile allocation. If a national SGA screening programme monitoring units’ ability to detect SGA was introduced, it could not start without validated, unit specific birthweight data.
Keywords : Birthweight, centile, SGA, small for gestational age, LGA, large for gestational age, GROW, Intergrowth 21, IG21, British 1990 reference values, UK90
Introduction :
In 2013 the Royal College of Obstetricians and Gynaecologists (RCOG) released its revised guideline for The Investigation and Management of the SGA Fetus (SGA GTG).4 In 2016 NHS England released “Saving Babies’ Lives” (SBL), a care bundle for reducing stillbirth.5 This required units to provide the SGA birth rate, detection rate, false positive and false negative rate.
The availability of packages like GROW, provided the required data. UHPT implemented SGA screening using GROW software in April 2015. The unit noticed a near doubling of the induction rate using the combination of GROW and the SGA GTG recommendations. It was perceived that GROW allocated more babies to the SGA group. The unit also noticed a massive increase in ultrasound examinations requested (after the 18-20 week anomaly scan), linked contemporaneously to use of GROW and actions required by the SGA screening programme. Confidence in the GROW package decreased. Monitoring of our Stillbirth rate had shown an almost halving of the rate prior to the introduction of GROW.
The inability to access patient data in the GROW programme locally, only summary data being available, prevented our local assessment of interventions. Alternatives to GROW were sought and a decision was made to exit GROW and move to IG21.
GROW has a significant cost compared with IG21, which including all the analysis tools, is open access, so there was a cost saving with the change.
The core principle of GROW providing ‘customised’ resources based on maternal demographics versus the uncustomised approach of IG21 were discussed. Neither the RCOG SGA guideline nor the SBL care package made a clear recommendation about customisation. The SGA GTG says, “No trials were identified that compared customised with non–customised symphysis fundal height (SFH) charts and thus evidence for their effectiveness on outcomes such as perinatal morbidity/mortality is lacking”. It also says, “Use of a customised fetal weight reference may improve prediction of a SGA neonate (good practice point)”. In SBL the only reference is a footnote which says “Customised or other established growth chart”
This lack of evidence combined with a potentially adverse impact from GROW allowed a change. IG21 was implemented at booking in 13th November 2018 replacing GROW in our local SGA guideline and practice.
Methods:
At delivery the birthweight and gestation were recorded on the GROW software. The software also required the reporting midwife to tick a field to say if the pregnancy had been ‘suspected’ of SGA or growth restriction or if either had been ‘detected’. A departmental standard policy was used to answer both questions (online Box A). Routine data collection included whether the delivery was induced or not.
We requested, in June 2018, from the GROW programme, all the individual patient data that had been entered during our use. After the last birthweights were entered, we received the whole dataset in June 2019. The dataset identified patients only by their GROW identity number (GROW ID), which was generated as the patient was entered on the programme at their 12 week scan and contained no patient identifiers. All analyses were done using the GROW ID as the only identifier. The GROW dataset provided the gestation (days), birthweight (grammes), the calculated customised GROW centile and the status at birth (live/stillbirth). It also provided the response to the delivery questions about ‘suspicion’ or ‘detection’ of SGA (online Box A). GROW does not produce information about high birthweights other than centile.
The IG21 Neonatal Size Calculator (newborn infants between 24+0 and 42+6 weeks gestation) app6 was used on the same dataset to derive the IG21 birthweight centile and Z score. The same approach was used to derive the UK90 birthweight centile and Z score using a free tool (LMSgrowth programme, v2.77 authors Huiqi Pan and Tim Cole, copyright MRC 2002-12).7 The UK90 references were reanalysed in 20098 when new UK-WHO growth charts were introduced. Review by the joint committee of the Royal College of Paediatrics and Child Health and the Scientific Advisory Committee on Nutrition, concluded the UK90 reference should be retained for assessment at birth, as the WHO mean birth weight for term infants was considerably lower than for UK infants, and the WHO values omit preterm births.9
This provided us with a single large dataset, with each patient identified by GROW ID, but with centiles from GROW, IG21 and UK90 references. We derived Z scores from IG21 and UK90 we were not provided with Z scores for the GROW data calculations.
We examined the distribution of the centiles (by plotting histograms and calculating corresponding skewness values) and the number of births deemed SGA (birthweight < 10th centile) according to each of the reference standards. Analysis was carried out in R.10
Results:
For all analyses we excluded the 40 stillborn infants (overall cohort singleton stillbirth rate = 3.77/1000) as both the cause of the stillbirth and the time from diagnosis to delivery may affect the weight at delivery and thus the allocated centile. This resulted in a cohort of 10576 liveborn singletons.
We plotted frequency histograms of birthweight centiles for each reference standard. In the absence of bias, the centiles would be expected to be uniformly distributed (showing approximately equal numbers of observations in each bar of the histogram and a skewness value near zero).
The histogram of the GROW centiles (Figure 1) confirmed a systematic bias, with a positive (right) skew of 0.169, towards allocating birthweights to a lower (customised) centile when used on our local population .
We found the opposite bias with the IG21 centile centiles (Figure 1), with a negative (left) skew of -0.452.
The UK90 reference range gave the most uniform distribution of centiles (Figure 1) with a minor negative left skew of -0.047, for our population.
The Saving Babies lives care bundle requires units to report to NHS England their SGA, true positive rate and false negative rates. These values were calculated for the population by each package (Table 1). The rate of SGA derived from each package varied from 4.9% to 11% with newborn SGA numbers varying from 518 for IG21 to 1167 for GROW. IG21 gave the lowest true positive rate (32.8%) and highest false negative rate (67.2%). The UK90 analysis produced the highest true positive rate identifying 427 (59.4%) SGA births with the lowest false negative rate (40.8%).
A previous unpublished internal review of 200 consecutive stillbirths at UPHT, over a decade, showed our singleton stillbirth rate to have fallen from 5.84/1000 in 2011 to 1.98/1000 in 2013 (figure 2). Ultrasound examination numbers (post routine anomaly scan) were stable at 15-16,000 between 2010 and 2014 but increased between 2015 and 2017 to over 21100 (figure 3). Labour induction rate was calculated as part of our continuing quality dashboard. It rose from 21% in Dec 2014 to a peak of 36.4% in July 2018 (online figure A).
The birthweight (online figure B) and gestation data (online figure C) produced the expected left skewed distributions. The median birthweight was 3425 grammes (interquartile range: 3100 – 3750g). Median gestation was 277 days (interquartile range: 271 – 284 days). The prematurity rate (<37 weeks) was 5.4%.