There were 5621 pregnancies in mothers with diabetes, of which 229 were excluded for reasons including missing gestational age at delivery (n = 7), missing sex of infant (n = 3), delivery before 24 weeks of gestation (n = 5), diabetes diagnosis other than type 1 or type 2 (n = 145) or twin pregnancy (n = 74). Exclusions could have been in more than one category.
The remaining 5392 singleton babies (3778 offspring of mothers with type 1 diabetes, 1614 offspring of mothers with type 2 diabetes) were born to 3847 mothers (2582 mothers with type 1 diabetes, 1265 mothers with type 2 diabetes). Stillbirth rates were 16.1 per 1000 births (95% CI 12.4, 20.8) in type 1 diabetes (n = 61) and 22.9 per 1000 births (95% CI 16.4, 31.8) in type 2 diabetes (n = 37). Stillbirth rates did not change with time in type 1 diabetes. There was a very small reduction with time in type 2 diabetes (p = 0.02).
Type 1 diabetes
Distributions of age, parity, smoking rates and deprivation scores were similar in mothers regardless of whether the pregnancy ended in stillbirth or livebirth (Table 1). Duration of diabetes was lower in the stillbirth group, as was pre-pregnancy BMI (Table 1).
HbA1c was recorded at each stage of pregnancy in 50–60% of cases (Fig. 1). Stillbirth rates were similar in women with an available pre-pregnancy HbA1c measure to those without (1.6% both, p = 0.9), despite higher levels of deprivation and smoking (26.1% vs 17.0%) in those without a pre-pregnancy HbA1c measurement (electronic supplementary material [ESM] Table 1). Overall, one in five women achieved pre-pregnancy glycaemic targets (<53 mmol/mol [7%]), with lower rates in the stillbirth group (Table 1). Mean pre-pregnancy HbA1c was 11 mmol/mol (1%) higher in pregnancies ending in stillbirth (p = 0.0002). Glycaemic control improved from pre-pregnancy across successive trimesters in both stillborn and liveborn groups, but with the stillborn group maintaining a similar relative level of hyperglycaemia compared with the liveborn group at all points (Fig. 1).
Stillborn infants were born 2.8 weeks earlier than liveborn infants (Table 1). Stillbirths occurred between weeks 24 and 38, with over a third (38%) at term (Fig. 2), compared with 63% of livebirths occurring at term. Stillbirth rates (expressed per week of pregnancy) were highest in the 37th and 38th weeks at 5.1 (95% CI 2.8, 9.1) and 7.0 (95% CI 3.7, 12.9) per 1000 ongoing pregnancies, respectively (Fig. 2). There were no stillbirths after 38 weeks, and only 11% of all deliveries occurred beyond this.
, birthweight and stillbirth risk
Stillborn infants had similar corrected birthweights to liveborn infants, with birthweight z scores 1.38 and 1.37, respectively (Table 1). Over half of babies (52% combined live and stillborn) born to mothers with type 1 diabetes were LGA, of which the majority (78%) had birthweights above the 95th centile (ESM Table 2). Higher birthweight was related to higher HbA1c. Mean birthweight z scores for this combined group were 0.9 (0.12) SD higher than the reference population in the subgroup of women in the lowest quartiles for HbA1c pre-pregnancy (<52 mmol/mol [6.9%]) and in the third trimester (<42 mmol/mol [6.0%]), increasing to 1.73 (0.09) SD for those in the highest quartiles (>76 mmol/mol [9.1%] pre-pregnancy and > 56 mmol/mol [7.3%] third trimester) (ESM Table 3). SGA infants had sixfold higher stillbirth rates than birthweight appropriate for gestational age infants (ESM Table 2). There was a significant non-linear relationship of birthweight z score and stillbirth (p = 0.002). When categorised as in ESM Table 2, there was a significant effect of birthweight category (p = 0.005) and SGA vs the reference category in particular.
Type 2 diabetes
Mothers who had a stillbirth had higher mean pre-pregnancy BMI than mothers delivering a live infant (p = 0.01, Table 1) and with a non-significant trend to higher parity (p = 0.06). Stillborn infants were more likely to be male (p = 0.0007). Smoking, deprivation score and duration of diabetes were similar between groups (Table 1). HbA1c was documented at each gestational period in approximately 50% of cases (Fig. 1). There was no difference in stillbirth rates in those with an available pre-pregnancy HbA1c vs those without (2.1% vs 2.5%, respectively; p = 0.6). Almost half of women achieved pre-conceptual targets (<53 mmol/mol [7%]), with lower rates in the stillbirth group. There was no association between gestational blood glucose level and stillbirth (Fig. 1) but preconception HbA1c was 12 mmol/mol (1.1%) higher in the stillbirth group (p = 0.01).
In type 2 diabetes, stillborn infants were delivered over 3 weeks earlier than liveborn infants (p < 0.0001) (Table 1). Stillbirth occurred in low frequency across all gestations from 24 weeks (Fig. 2), with 68% occurring preterm (Fig. 2) compared with 24% of livebirths occurring preterm. Stillbirth rates were highest in the 39th week at 9.3 (95% CI 2.4, 29.2) per 1000 ongoing pregnancies (Fig. 2).
HbA1c, birthweight and stillbirth risk
Mean birthweights were similar between live and stillborn infants (Table 1). Stillbirth odds were highest amongst infants with birthweights <10th centile (threefold) and > 95th centile (2.2-fold) (ESM Table 2). There was a significant non-linear relationship of birthweight z score and stillbirth (p = 0.007) and, when categorised as in ESM Table 2, there was a significant effect of birthweight category (p = 0.04), with suggestion of increased risk for both SGA and birthweight percentile >95th. Similar to type 1, babies (live and stillborn) born to mothers with HbA1c in the lowest pre-pregnancy and third trimester quartiles were 0.39 (0.13) SD heavier than the reference population. This increased to 1.79 (0.26) SD for those with HbA1c in the highest pre-pregnancy and third trimester quartiles (ESM Table 3).
Risk factor effect estimation in stillbirth
In type 1 diabetes, univariate models suggested that shorter diabetes duration (OR 0.96 [95% CI 0.93, 0.99]; p = 0.01) and lower BMI (OR 0.92 [95% CI 0.86, 0.99]; p = 0.04) were weakly associated with stillbirth. There were significant associations between stillbirth and higher HbA1c pre-pregnancy (OR 1.03 [95% CI 1.01, 1.04]; p = 0.0003) and in each trimester (OR 1.04 [95% CI 1.02, 1.05], 1.05 [95% CI 1.03, 1.07] and 1.06 [95% CI 1.04, 1.08] in first, second and third trimesters, respectively; all p < 0.0001). These effects remained in models adjusted for maternal age, SIMD and diabetes duration.
For type 2 diabetes, univariate models showed that higher BMI (OR 1.07 [95% CI 1.01, 1.14]; p = 0.02) and higher pre-pregnancy HbA1c (OR 1.02 [95% CI 1.00, 1.04]; p = 0.016) were associated with stillbirth. In contrast to type 1 diabetes, HbA1c later in pregnancy was not associated. These effects remained when adjusted for maternal age, SIMD and diabetes duration.
For type 1 diabetes, we further explored whether associations between HbA1c and stillbirth varied by timing of HbA1c. Unsurprisingly, HbA1c at various stages of pregnancy was highly correlated (ESM Table 4). In models including maternal age, diabetes duration, deprivation score, and pre-pregnancy and third trimester HbA1c (n = 1382 with all variables), only third trimester HbA1c remained significantly associated with stillbirth (OR 1.05 [95% CI 1.02, 1.08]; p = 0.0008). The effect of HbA1c pre-pregnancy was attenuated (OR 1.02 [95% CI 0.99, 1.04]; p = 0.08). Third trimester HbA1c did not have a significantly greater effect than pre-pregnancy HbA1c (p = 0.12) on stillbirth but analysis was limited by low numbers of women with both HbA1c measures.
For type 2 diabetes, the effect of BMI appeared independent of pre-pregnancy HbA1c, with both remaining associated with stillbirth (OR 1.09 [95% CI 1.02, 1.17]; p = 0.01 and 1.03 [95% CI 1.01, 1.05]; p = 0.006, respectively) in models including both factors, maternal age, diabetes duration and deprivation score (n = 594 with all values).
Variation by region and unit size
There were no significant differences in stillbirth rates by health board area of delivery (p = 0.60) or by hospital unit size (p = 0.39). Regional differences in gestational age at delivery were observed for type 1 (range 33.4 ± 1.6 to 36.9 ± 0.2 weeks, p < 0.0001) and type 2 diabetes (range 36.4 ± 0.3 to 37.8 ± 0.2 weeks, p = 0.006). This appeared partly driven by earlier delivery in very small boards, but remained significant after exclusion of these in type 1 (range 36.1 ± 0.1 to 36.9 ± 0.2 weeks, p < 0.0001) and type 2 diabetes (range 36.5 ± 0.3 to 37.8 ± 0.2 weeks, p = 0.007) (ESM Fig. 1). Babies born at the largest units were delivered 4 days earlier than those delivering 10–19 mothers with diabetes per annum in type 1 (p = 0.002), but with no difference in type 2 diabetes. Mean birthweight z score in type 2 diabetes varied across health boards (p < 0.0001) (ESM Fig. 1), but not in type 1 diabetes (p = 0.05). Birthweight z score did not vary by hospital unit size in type 1 (p = 0.20) or type 2 diabetes (p = 0.15) in models including smoking (lower in smokers), year of delivery (small increase with time), maternal age at delivery (reduced with age), diabetes duration (increased with duration type 2 diabetes) and deprivation score (similar in type 2 diabetes, lower with higher SIMD in type 1 diabetes).