Dr P. Owen, Princess Royal Maternity Hospital, Alexandra Parade, Glasgow, Scotland, UK.
The study aims to compare the utility of unadjusted with customised weight standards in the identification of intrauterine growth restriction (IUGR) among unexplained stillborn infants undergoing postmortem examination. Unadjusted and customised birthweight centiles were determined for 51 unexplained stillborn infants undergoing perinatal autopsy. Unadjusted centiles were calculated from an ultrasonically derived fetal weight standard. Customised centiles were calculated from an online calculator which adjusts the standard to account for important physiological variables. IUGR was defined as moderate or severe according to brain/liver ratios of >3 and >5, respectively. The relationship between the weight centiles and abnormal brain/liver weight ratios was explored. Neither unadjusted nor customised standards identify stillborn infants with brain/liver ratios >3:1. Both unadjusted and customised weight standards identify stillborn infants with brain/liver ratios >5 equally well with high sensitivity (95%) but low specificities (63% and 66%, respectively). Customising weight standards to account for physiological variables does not identify growth restricted stillborn infants more usefully than an unadjusted fetal weight standard.
The relationship between fetal weight and risk of stillbirth is confounded by the increased risk of fetal demise in intrauterine growth restriction (IUGR) secondary to uteroplacental insufficiency. Birthweight is influenced by several physiological variables as well as pathological influences such that it is inappropriate to diagnose stillbirth due to IUGR on the basis of birthweight centile alone. Customising birthweight standards to account for physiological variables is reported to identify a higher proportion of stillbirths as small for gestational age compared with population-based standards, probably due to improved identification of IUGR.1 This concept, however appealing, has not been explored in a study to verify the association using postmortem confirmed IUGR as the reference standard. We compared the performance of customised and non-customised weight standards in the identification of postmortem confirmed IUGR among otherwise unexplained stillborn infants.
Records of the perinatal autopsies of stillborn infants greater than 24 weeks of gestation delivered at Glasgow Royal Maternity Hospital and The Queen Mother's Hospital, Glasgow were examined for the years 1998 to 2002. All autopsies were performed at the Department of Paediatric Pathology, Royal Hospital for Sick Children, Glasgow. Deaths associated with chromosomal abnormalities, structural abnormalities, maternal disease and twin pregnancies were excluded. Postmortem reports where liver and brain weights were not available were also excluded.
Weight at birth and brain and liver weights were recorded and the ratio of brain/liver weight was calculated. The brain/liver weight ratio is normally less than three in the second half of gestation and a ratio greater than this indicates fetal malnourishment.2 A brain/liver ratio >5 is without the limits of normal variation3 and indicates severe fetal malnourishment and IUGR. We chose brain/liver ratios of both 3 and 5 as diagnostic criteria for IUGR among the stillborn infants.
Pregnancy characteristics were obtained from the maternal case record and included maternal height, weight at booking, parity and ethnic origin. Gestational age at delivery and infant gender were included in order to calculate a customised birthweight centile using computer software from http://www.gestation.net.4 Gestational age was determined from last menstrual period, provided there was agreement within seven days with ultrasound. Where a discrepancy of more than seven days was present, gestational age was determined from ultrasound biometry. Cases where maternal information was incomplete were excluded. Unadjusted birthweight centiles were generated by conversion from standard deviation scores derived from sonographically determined gestational age specific fetal weight standards.5
The ability of both unadjusted and customised weight standards to identify abnormal brain/liver weight ratios was explored by constructing receiver–operator characteristics curves (ROC curves). Where appropriate, a cutoff centile was chosen by inspection of the ROC curve and the sensitivity and specificity values for the prediction of IUGR were determined.
Two hundred and twenty-six perinatal autopsies of stillborn infants from both maternity hospitals were performed during the study period, of which 121 were of ‘unexplained’ stillbirths. Fetal and/or maternal information was incomplete in 70 cases resulting in a study population of 51 cases.
The mean (range) gestational age at delivery for the study population and the excluded cases was 33 weeks (24–41 weeks) and 34 weeks (24–41 weeks), respectively. The mean (range) of unadjusted weight centiles for the study population and the excluded cases was 21 (0–100) and 18 (0–99), respectively. The mean (range) of customised centiles for the study population was 22 (0–100).
Forty-three and 18 of the study cases had brain/liver ratios >3 and >5, respectively. The mean (range) of unadjusted birthweight centiles for cases with brain/liver ratios <3, >3 but <5 and >5 was 47 (0.6–77), 26 (0–78) and 2 (0–33), respectively. The relationship between unadjusted birthweight centiles and brain/liver ratios is presented in Fig. 1a.
The mean (range) of customised birthweight centiles for cases with brain/liver ratios <3, >3 but <5 and >5 was 48 (0–87), 28 (0–100) and 1 (0–9), respectively. The relationship between customised birthweight centiles and brain/liver ratios is presented in Fig. 1b. As anticipated, higher brain/liver ratios are associated with both lower unadjusted and customised birthweight centiles.
With brain/liver ratio >3 as the outcome variable, the area under the ROC curve for unadjusted and customised birthweight centiles was 0.47 and 0.44, respectively, indicating no discriminatory capacity. With a brain/liver ratio >5 as the outcome variable, the area under the ROC curve for unadjusted and customised birthweight centiles was 0.81 in both cases. Inspection of the ROC curve resulted in optimal cutoff values of 13 for unadjusted and 9 for customised centiles. The sensitivity of the unadjusted and customised centile cutoffs in the identification of a brain/liver ratio >5 is 95% in both cases: the respective specificities are 63% and 66%.
The relationship of unadjusted and customised weight standards to objective evidence of IUGR among stillborn infants has not been previously reported. Neither unadjusted nor customised weight categorisation is able to identify infants with a brain/liver ratio of >3 but both methods identify with equal utility the more severely growth restricted infants (brain/liver ratio >5) with high sensitivity. We chose to examine two values of brain/liver ratio to reflect different degrees of fetal malnutrition. The inability of weight centiles to identify infants with brain/liver ratios >3 is perhaps unsurprising as malnutrition in the presence of an apparently normal weight is recognised among both stillborn5 and live born infants,6 reflecting a relative rather than absolute failure of growth potential achievement. Nevertheless, this is disappointing as the loss of opportunity to make a positive diagnosis of IUGR due to the falling incidence of perinatal autopsies among stillborn infants might have been partially offset with the introduction of an accurate non-invasive method of establishing IUGR.
Customising birthweight standards to account for maternal and fetal physiological characteristics did not result in an improvement in the prediction of abnormal brain/liver weight ratios compared with unadjusted standards. This is unexpected because a previous comparison of customised and unadjusted birthweight standards demonstrated a closer relationship between customised birthweight categorisation and risk of stillbirth.1 One explanation is that the risk for stillbirth is multifactorial and while low birthweight is associated with the risk of stillbirth, this risk is manifest via genuine growth restriction in only a proportion of cases. The results of this study do not support the hypothesis that customising birthweight centiles improves the identification of genuine growth restriction among stillborn infants.
A second consideration is our use of an unadjusted fetal weight standard rather than a birthweight standard per se. The importance of employing an appropriate weight standard has previously been highlighted with the recognition that a higher proportion of stillborn infants will be categorised as small for gestational age depending upon whether standards are constructed from gestation specific birthweights or fetal weight standards extrapolated from normal term infant birthweights.7 It is possible that our use of an appropriate fetal weight standard has minimised any potential improvement in the prediction of IUGR that might be anticipated by customising the birthweight standard.
A final consideration is the influence of post demise weight loss in utero upon the relationship of birthweight standards and IUGR. Such weight loss inevitably strengthens any association between birthweight standards and IUGR but can be expected to affect both unadjusted and customised standards equally.
Both unadjusted and customised weight standards are sensitive but non-specific indicators of severe IUGR among stillborn infants. Compared with an appropriate unadjusted fetal weight standard, customising the birthweight standard to take account of the physiological influences upon birthweight does not result in improved prediction of IUGR among stillborn infants.