Methods

Population and data collection

We conducted a retrospective cohort study using data on all infants born in Victoria, Australia, from 2005 to 2015. Data were obtained from the Consultative Council on Obstetric and Paediatric Mortality and Morbidity (CCOPMM), which is the central agency that collects and validates data on obstetric and perinatal outcomes within the state (27, 28).
Prior to data cleaning and analysis, an a priori plan was formulated to determine the inclusion and exclusion criteria, and how implausible data values would be managed. Singleton pregnancies from 24 weeks +0 days to 42 weeks +6 days’ gestation at delivery were included; pregnancies prior to 24 weeks’ were excluded due to highly variable resuscitation preferences and outcomes. Exclusion criteria included multiple pregnancy, congenital anomalies, termination of pregnancy, those with missing or implausible birthweights or missing infant sex, or where gestational age in days was not recorded. Gestation in days was calculated based on the date of birth and the last normal menstrual period, or the date of birth and estimated due date.
Maternal height and weight data were based on that recorded at the obstetric booking visit. Parity was defined as the number of previous births (live or stillborn) over 20 weeks’ gestation. Maternal age was recorded to the nearest year at booking, and birthweight was recorded in grams. Country of birth was also self-reported. Obstetric outcome data were recorded by the attending midwife during pregnancy, delivery and after birth, using the Birthing Outcomes System, as a part of routine clinical care.

Growth charts used

We applied Fetal Medicine Foundation fetal and neonatal (birthweight) population charts (2018) (25). These charts were derived from both newborns, and fetuses still in utero, for a given gestational age, and so were designed to represent the entire obstetric population at any gestational age. They included all well-dated, singleton fetuses from 2011 to 2017, without congenital anomalies (95,579 infants), from two centres within the United Kingdom to derive the reference ranges (25). Using their dataset, reference ranges were derived for both birthweight (birthweight charts) and estimated fetal weight (fetal charts). They assumed both birthweight and estimated fetal weight would have a common median and a bivariate Gaussian distribution, with the main difference being levels of spread (25). A non-parametric quantile regression formula was used to directly estimate both birthweight and estimated fetal weight for each gestational day (29).
These charts were chosen for several reasons. First, the published charts include gestation in days, as well as gestation by completed weeks. This enabled us to compare the difference between the two, without the need for correction according to the methodology of the charts, or the population from which they were derived. Second, these charts provide both an estimated fetal weight and a birthweight centile, which enabled us to assess our findings with charts that are used in two different settings; an antenatal and a postnatal setting. Third, the study population and median birthweights have similarities to an Australian population, allowing us to approximate an appropriate 10th centile, which can be a challenge when applying externally derived growth standards (22).

Outcomes and analysis

Every newborn across all gestations was allocated a “+” number, based on the day of the gestational week that they were born; +0 being born exactly on the first day of a completed week (eg. 36+0 weeks, 252 days), through to +6 being the day before the next completed week of gestation (eg. 36+6, 258 days). This created seven groups available for analysis, with two primary groups of interest (Table S1): i) Group +0 days contained all infants born on the first day of a completed week of gestation (24 weeks +0 days, 25+0, 26+0…. 42+0) and ii) Group +6 days contained all infants born on the final day of a completed week of gestation (24 weeks +6 days, 25+6, 26+6… 42+6).
To every infant available for analysis, we then applied the Fetal Medicine Foundation Charts in the following ways. First, we generated birthweight “week” charts; we applied the neonatal (birthweight) charts, which provide just one series of weight centiles per gestational week. This was based on the median value for that gestational week (equivalent to the exact value at +3 days for each week of the day charts). Second, we generated birthweight “day” charts: We applied the neonatal (birthweight) charts, this time using the day-by-day charts, therefore generating a specific weight centile for each completed day of gestation.
We repeated the above process using the estimated fetal weight (fetal) charts. Both fetal and birthweight charts were assessed, as fetal charts are commonly used antenatally, while birthweight charts are commonly used postnatally. This allowed us to extrapolate the results to both clinical settings.
Every infant therefore had four different birthweight centiles (birthweight week, birthweight day, estimated fetal weight week, and estimated fetal weight day centiles), and every infant was also allocated a “+” value. This allowed us to compare the impact of the four centile classifications between each of the “+” categories. In particular, we were interested in comparing the cohort of infants born at +0 days, and the cohort of infants born at +6 days, because these two cohorts were maximally affected by use of week, or day charts.
First, using week charts, we examined the proportion of infants born on each day (+0 days, +1, +2, +3, +4, +5, +6) that were classified as <10th centile (small for gestational age, SGA) or <3rd centile (an accepted surrogate for true fetal growth restriction). This allowed us to assess differences in classification of small infants across the week. We propose that the exact day of the gestational week that an infant was born was unlikely to have impacted the true probability of fetal growth restriction, and that differences across the week would therefore be an artefact of their classification.
We then applied day charts, and again examined the proportion of infants born on each day that were classified as <10th centile or <3rd centile. For each given day (+0 days to +6 days), we then assessed what proportion of infants werereclassified as small or large after day charts were applied, compared with week charts. This enabled us to assess the magnitude of the impact of using ‘day’ charts.
Finally, we examined stillbirth rates amongst small (<10th and <3rdcentile) infants, focussing our comparisons on those born on +0 days and those born on +6 days. Given at +0 days an infant is at its smallest, and at +6 days its largest, these are the days on which the greatest fetal growth restriction classification error might occur. As we assumed the true probability of fetal growth restriction would occur equally on both days, we also anticipated stillbirth rates to occur equally on both days. Therefore, a different proportion of infants classified as small at +0 or +6 may impact the reported rate of stillbirth amongst small infants.
To assess this, we analysed the data in two ways. First, of those classified as <10th centile by week charts, we compared the relative risk of stillbirth for infants born at +0 days with infants born at +6 days, using +0 days as the reference. We hypothesised that week charts would classify a higher proportion of infants as SGA at +0 days (when the infant is at its smallest), and the least number at +6 days (when the infant has had the full week to grow). We therefore expected the relative risk of stillbirth at +0 days to be smaller (when the SGA cohort was artificially larger and over-represented by healthy infants), and higher at +6 days (when the SGA cohort was smaller, and so represented only the smallest and highest risk infants). Second, of those classified as <10th centile by day charts, we also compared the relative risk of stillbirth for infants born at +0 days to infants born at +6 days, again using +0 days as the reference. We hypothesised that day charts would classify a similar proportion of infants as SGA at both +0 days and +6 days, which would be reflected as a similar relative risk of adverse outcomes within each cohort.
We assessed these outcomes using both the birthweight and fetal standards.

Statistical analysis

Baseline characteristics of the population were summarized by mean (standard deviation), median (25th – 75th percentile) and number (%) according to type and distribution of the data. Relative risks were calculated and reported as point estimates with Wilson 95% confidence intervals. Significance level was two-sided, set at 0.05 and not adjusted for multiple comparisons. Statistical analysis was conducted using Stata Version 16 (StataCorp. 2019. Stata Statistical Software: Release 16.1. College Station, TX, USA).

Ethics

Ethics approval for the project was obtained from the Mercy Health Human Research Ethics Committee (approval project number R16-10) and CCOPMM (approval number RR16-04). As this was a retrospective cohort study using de-identified data, individual patient consent was not required.

Results

Study population

Within the study period, there were 735,590 births in Victoria. After the exclusion criteria were applied, 529,261 infants remained for final analysis (Figure 1). Baseline characteristics of those born <10th centile by birthweight week and day charts, and of those born <10th centile at +0 days or +6 days, are presented in Table 1. As Fetal Medicine Foundation charts are not sex-specific, a smaller proportion of male infants were considered <10th centile.

Proportions classified as small or large by day or week birthweight charts.

Of the total population, 52,987 (10.0%) newborns were classified as <10th centile by birthweight week charts. Of all infants born at +0 days, week charts classified 12.1% as SGA. The proportion of infants classified as SGA dropped steadily across the gestational week cohorts to 7.8% at +6 days (Table 2, Figure 2A). The relative risk of an infant being classified as <10th centile at +6 days compared with +0 days was 0.56 (95% CI 0.54 – 0.58, p<0.0001). This suggests that an infant is 44% less likely, by week charts, to be considered SGA if they are born at the end of the gestational week (+6 days) compared with the beginning (+0 days).
Overall, 52,219 (9.9%) newborns were classified as <10th centile by birthweight day charts. Using day charts, the proportion classified as <10th centile was similar across the gestational week; with 9.5% <10th centile at the beginning (+0 days), and 9.9% at the end (+6 days) (Table 2, Figure 2A).
We then assessed the change in classification if day charts were used compared with week charts. When day charts were applied instead of week charts, 20.9% fewer infants born at +0 days were considered <10th centile (Table 2). When day charts were applied instead of week charts, 27.9% more infants born at +6 days were considered <10th centile.
The same findings were seen in those infants born <3rd centile (Table 2, Figure 2B). Using week charts, 4.9% born at +0 days were considered <3rd centile, dropping to 2.9% at +6 days. The relative risk of being considered <3rdcentile at +6 days compared with +0 days was 0.60 (95% CI 0.57 – 0.63), suggesting that an infant is 40% less likely to be considered <3rd centile on week charts, if born at the end of the gestational week compared with the beginning. When day charts were used, the proportions classified <3rdcentile were equivalent at +0 days and +6 days (3.8% vs 3.8%).

Stillbirth risk

Overall, there was no difference in the rate of stillbirth between all infants born at +0 days (197 stillbirths; 0.26%) and those born at +6 days (196 stillbirths; 0.29%) (p=0.23). We then examined stillbirth rates amongst those classified as <10thcentile by week charts (Table 3). The risk of stillbirth in the SGA cohort (<10th centile) born at the beginning of the week (+0) was used as our reference, and we compared it with the relative risk of stillbirth in the SGA cohort born at the end of the week (+6). Using week charts, there was a higher relative risk of stillbirth amongst SGA infants at +6 compared with +0 (RR 1.47, 95% CI 1.09 – 2.00). When day charts were used, there was no significant difference in the stillbirth risk between the +0 days and +6 days cohorts (Table 3).
When a cutoff of <3rd centile was applied, an even stronger trend was observed (Table 3). Using week charts, the relative risk of stillbirth amongst those born at +6 days was higher than at +0 days (RR 1.73, 95% CI 1.24 – 2.43). If day charts were used, no significant difference in the risk of stillbirth was seen between the two cohorts.
We replicated our analysis, using the estimated fetal weight standard instead of the birthweight standard, to ensure that the same effect size was seen using a fetal growth standard (which has reduced variation around the mean). The same findings were seen, with significantly increased stillbirth risk in small infants born at the end of the week (Table S2).