Data analysis

A forest plot of the summary meta-analysis odds ratios and 95% confidence intervals for each measure of being small for gestational age and neonatal mortality is given in Figure 6.2. Birth weight <1.5kg showed the strongest association with neonatal mortality (OR 48.6, 95% CI 28.62 to 82.53), with no between-study heterogeneity in this effect. Raising the birth weight threshold to 2.0kg, 2.5kg or 2.9kg gradually reduced the association and increased the heterogeneity, but the summary effect remained highly significant at each threshold. Population centile charts were also strongly associated with neonatal mortality, but generally showed a weaker

association at all thresholds than absolute birth weight, because the summary ORs were closer to 1, especially compared to an absolute birth weight < 1.5kg or

71 2.0kg.The results for other measures, including ponderal index, birth weight < 2 standard deviations below the population mean, and fetal growth ratio (defined in the primary study as the observed birth weight over the population mean) varied (Figure 6.2).

The association between measures of being small for gestational age and neonatal morbidity are given in Figure 6.3. The analysis was subdivided into reported

neurological morbidity (including seizures, HIE, IVH) and non-neurological morbidity (including hypoglycaemia, respiratory distress syndrome, cardiac failure) according to the definitions given in the primary studies. Birth weight <2.0kg was the most

strongly associated with neurological morbidity, (OR 17.34, 95% CI 5.63 to 53.70) however this was based on a single study of 770 neonates. There was a significant association between weight below the 3rd, 5th and 10th centiles and neurological morbidity. Birth weight <10th centile according to customised growth chart and ponderal index ≤2.25 did not show a significant association with this outcome. For non-neurological morbidity, birth weight <3rd, 5th or 10th centiles on population chart and birth weight > 2 standard deviations (SD) below the population mean showed significant association with this outcome with odds ratios of a similar magnitude.

Infant outcomes

Figure 6.4 gives the association between measures of being small for gestational age and infant outcomes. Birth weight <1.5kg was strongly associated with infant

morbidity on average (OR 21.57, 95% CI 6.31 to 73.70), however significant

heterogeneity was present in the analysis. Birth weight <2.0kg and <2.5kg were also significantly associated with this outcome but the odds ratios were smaller. For

72 neurodevelopmental delay, birth weight <10th centile according to population chart was significantly associated with this outcome (OR 2.03, 95% CI 1.01 to 4.08)

Childhood and adolescent outcomes

A forest plot for the association of definitions of being small for gestational age with childhood and adolescent outcomes is given in Figure 6.5. Meta-analysis was

performed to assess the association of birth weight <2.5kg with a composite group of adverse outcomes reported in primary studies (including obesity, hypertension, type 1 diabetes mellitus, asthma, hypercholesterolaemia, learning difficulties and

strabismus). There was no significant association present (OR 0.98, 95% CI 0.87 to 1.10). A meta-analysis for birth weight <10th centile on population chart showed a small association that was just significant (OR 1.49, 95% CI 1.02 to 2.19) however there was significant heterogeneity present. When the analysis was restricted to learning difficulties or mental handicap, birth weight <3rd centile on population chart and <10th centile both showed a weak but significant association. There was no significant association between any birth weight standard and childhood obesity, hypertension, asthma, visual impairment or psychiatric diagnosis.

Adult outcomes

A forest plot of odds ratios for the association of measures of being small for gestational age and adult outcomes is given in Figure 6.6. A meta-analysis was performed for the association of birth weight <2.5 kg with a composite measure of adult morbidity (including obesity, hypertension, hypercholesterolaemia, type 2 diabetes mellitus, coronary heart disease, and polycystic ovarian syndrome). There was no significant association between birth weight <2.5kg, or birth weight <10th

73 centile on population chart, with this outcome. When individual morbidities were considered, birth weight <10th centile according to population chart was significantly associated with obesity according to a single study (OR 1.86, 95% CI 1.20 to 2.88). Birth weight <2.5kg showed a weak association with hypertension, diabetes mellitus or impaired glucose tolerance and cardiovascular mortality. Ponderal index (kg/m3) <24 was also weakly associated with mortality from cardiovascular disease.

Childhood or adulthood end stage renal disease showed a significant association with birth weight <10th centile on population chart.

Subgroup analyses

The results for subgroup analyses within the meta-analysis groups for each age group and birth weight standard are presented in Table 6.2. Few studies reported ethnicity in enough detail to permit subgroup analysis. For neonatal death, none of the subgroup analyses affected the magnitude or significance of the association between population chart <10th centile and this outcome. Limiting to a singleton population slightly weakened the association between birth weight <1.5kg and neonatal death, but did not affect birth weight <2.5kg for the same outcome. For childhood morbidity, singleton population, Caucasian population or year of birth ≥ 1990 did not significantly influence the results.

Birth weight as a continuous variable

There were seven papers that reported regression analysis using birth weight as a continuous outcome. These studies looked at adult hypertension (age 50 and 60 years) and hypercholesterolaemia, childhood obesity and hypertension, composite childhood metabolic risk index. Only one found a significant association (Anderrson

74 et al 2000): logistic regression analysis to examine the association between birth weight and hypertension (defined as treatment for hypertension and /or systolic BP ≥ 160mmHg and/or diastolic BP >95mmHg) found that at age 60, the OR was 0.96 (95% CI 0.92 to 0.99, p= 0.028 for change in risk of hypertension per 100g birth weight).

Direct comparison of absolute versus population centiles

Only two studies directly compared absolute birth weight and centile on population chart in the same population. For type 1 diabetes in childhood, birth weight <2.5kg had an OR of 0.68 (95% CI 0.22 to 2.12), and population chart <10th centile had an OR of 0.46 (95% CI 0.26 to 0.82) (Algert et al 2009). For neonatal mortality, birth weight <2.9kg had an OR of 2.64 (95% CI 1.45 to 4.82) and population chart <10th centile had an OR of 5.51 (95% CI 2.95 to 10.31) for the same outcome (Balcazar 1990).

Predictive ability of birth weight standards for neonatal death

The outcome that had the strongest prognostic association overall with fetal growth restriction was neonatal death. For those birth weight tests with a large (OR>5) and statistically significant prognostic association for this outcome, their predictive ability for individual babies was summarised by using meta-analysis to calculate summary sensitivity, specificity and likelihood ratios (Table 6.3). These measures reveal the discriminative ability of each test and how test results modify a baby’s odds of having a neonatal death. For each test the specificities and positive likelihood ratios were high, but the sensitivity and negative likelihood ratios were generally poor (Table 6.3). This can be explained by the fact that although a higher proportion of deaths

75 occurred within the low birth weight group, because this group represents a small fraction of the overall population, a large absolute number of deaths still occurred within the normal weight groups, and therefore sensitivity is low and the ‘false negative’ numbers are high, giving a poor negative likelihood ratio (close to 1). For example, the highest positive likelihood ratio was for 1.5kg, indicating that any baby under this weight multiplied their pre-test odds of neonatal death by 49.1 (95% CI: 27.3 to 88.5). However the negative likelihood ratio was only 1.01 (1.00 to 1.01), indicating that the odds of death barely change after a negative test result. Thus although being < 1.5kg substantially increases the odds of a poor outcome, being > 1.5kg does not increase the odds of a good outcome.