Risk and Recurrence of Serious Adverse Outcomes in the First and Second Pregnancies of Women With Preexisting Diabetes

Serious adverse pregnancy outcomes, such as miscarriages, stillbirths, and congenital anomalies, are associated with significant psychological distress, and parents who experience such events are often very anxious about their chances of recurrence (1,2). In the general population, the risks of miscarriage, stillbirth, and congenital anomaly in the second pregnancy are approximately two times greater in women who experienced the same event in their first pregnancy (3–5), although the absolute risks remain low in the absence of clear genetic or physiological factors.

Despite significant improvements in preconception and antenatal care, women with preexisting (type 1 or type 2) diabetes still experience substantially increased risks of serious adverse pregnancy outcomes, including miscarriages (6), congenital anomalies (7), stillbirths (8), and infant deaths (8). Little is known, however, about the absolute risks of these outcomes in first and second pregnancy, specifically, and whether women with diabetes experience the same patterns of recurrence as the general population. Suboptimal glycemic control at the start of pregnancy explains a large proportion of the excess risk of congenital anomalies and fetal and infant death (7,8); however, the extent that interpregnancy changes in glycemic control can modify the risk in subsequent pregnancies has not been demonstrated.

This study used unique data from the U.K.’s longest-running survey of women with preexisting diabetes to estimate: 1) preparation for and change in preparatory behavior between the first and second pregnancies of 220 women with preexisting diabetes, including the effect of adverse outcome in the first pregnancy, and 2) risk of, change in risk of, and predictors of serious adverse outcome in each pregnancy, including the effect of adverse outcome in the first pregnancy.

The North of England is a distinct region of the U.K. with a population of 3 million and ∼32,000 births per year (Supplementary Fig. 1). The sample comprises 440 pregnancies occurring in 220 women with preexisting diabetes who completed two successive singleton pregnancies at any gestational age—regardless of outcome—in the North of England during 1996–2008.

Miscarriage is the spontaneous loss of a fetus at ≤23 weeks’ gestation. Stillbirth is the delivery of a fetus showing no signs of life at ≥24 weeks’ gestation. Spontaneous fetal death comprises miscarriages and stillbirths. Infant death is the death of a live-born infant aged ≤1 year. Congenital anomalies are any major chromosomal, genetic, or structural abnormality defined by the European Surveillance of Congenital Anomalies (EUROCAT) criteria (9). Termination of pregnancy is the induced loss of a fetus for therapeutic or elective reasons. Serious adverse outcomes comprise congenital anomalies, spontaneous fetal deaths, and infant deaths.

The Northern Diabetes in Pregnancy Survey (NorDIP) records details of all pregnancies occurring in women resident in the region and diagnosed with diabetes at least 6 months before conception. Clinicians within the region’s nine maternity units collect and supply information on a range of clinical and sociodemographic variables (10).

Pregnancies affected by congenital anomaly (regardless of outcome) were identified from the Northern Congenital Abnormality Survey (NorCAS) (11). Stillbirths and infant deaths were identified from the Northern Perinatal Mortality Survey (PMS) (12).

Available variables with a hypothesized influence on serious adverse pregnancy outcome were obtained for analyses. Maternal ethnicity; diabetes type; prepregnancy history of clinically diagnosed nephropathy, neuropathy, and retinopathy; attendance at preconception care; preconception folic acid supplementation (self-reported); smoking during pregnancy (self-reported); and attendance at the first antenatal appointment before 10 weeks’ gestation were all analyzed as dichotomous variables. Socioeconomic circumstances at birth were estimated from the Index of Multiple Deprivation, an area-based measure of disadvantage (13), and analyzed in tertiles of ranks. Year of delivery, maternal age at delivery, duration of diabetes, maternal BMI (derived from height and weight at the first antenatal visit), duration of diabetes, and mean periconception glycated hemoglobin (A1C) concentration were analyzed as continuous variables. Periconception A1C was defined as the closest measurement within 3 months before the last menstrual period (available for 52.9% of pregnancies) or mean first trimester measurement (<14 weeks’ gestation) (available for 82.0% of pregnancies) for women with no preconception measurement. Periconception A1C was considered a reasonable proxy for preconception A1C because the first trimester A1C was highly correlated with the preconception A1C (Spearman correlation coefficient 0.73) (14). Gestational age at delivery and the first antenatal appointment were determined during the first ultrasound examination or (rarely) from the date of the last menstrual period. Small for gestational age (SGA; birth weight <10th centile) and large for gestational age (LGA; birth weight >90th centile) were determined from birth weight, standardized for sex, parity, and gestational age against a Scottish reference population (15).

Four variables were selected as markers of pregnancy preparation due to their established associations with outcome (16,17) and integration within care guidelines for women with preexisting diabetes:

1. Periconception A1C <53 mmol/mol (7.0%)—recommended by the American Diabetes Association (based in the U.S.) (18).

2. Self-reported preconception folic acid—the National Institute for Health and Care Excellence (NICE; based in England) recommends that women with diabetes take 5 mg/day folic acid before conception (19).

3. Attendance at the first antenatal visit before 10 weeks’ gestation—recommended by NICE (19).

4. Record of attending specialist preconception care services—recommended by NICE to be offered to all women with preexisting diabetes (19). Regional guidelines advise those responsible for routine care to inquire about pregnancy intention, discuss the benefits of preparation, and refer those with plans to specialist services.

The proportion of women achieving each marker of preparation was calculated per 100 pregnancies. Changes in prevalence and prevalence ratios for repeat behavior were estimated by Poisson regression. The association between an adverse outcome in the first pregnancy and preparation in the second was examined by logistic regression, with adjustment for baseline.

The prevalence of miscarriage, stillbirth, spontaneous fetal death, and congenital anomaly was calculated per 100 pregnancies. The prevalence of infant death, delivery by Caesarean section, SGA, and LGA was calculated per 100 births. Changes in prevalence and relative risks (RRs) of recurrence were estimated by Poisson regression.

The total probabilities of spontaneous fetal death from 6 weeks, 12 weeks, and 24 weeks were estimated using Kaplan-Meier. Pregnancies were “at risk” between the gestation at the first antenatal appointment and the gestation at delivery. Miscarriages and stillbirths were events, elective terminations of pregnancy were censored, and live births were modeled as surviving throughout.

Predictors of serious adverse outcome in each pregnancy were examined by competing-risks regression (20). Pregnancies were “at risk” between the first antenatal appointment and delivery. The primary event was any serious adverse outcome, and the competing events were live births or terminations without evidence of congenital anomaly. Unadjusted subdistribution hazard ratios (SHRs) were calculated for each variable in relation to serious adverse outcome in each pregnancy separately. Adjusted SHRs (aSHRs) were estimated within multivariable models constructed using a backward stepwise approach. Variables with an unadjusted P < 0.5 were entered and removed iteratively (by descending P value) until only those with P < 0.1 remained. The shape of association between periconception A1C and a serious adverse pregnancy outcome was explored by locally weighted scatterplot smoothing (21). Because a J shape was observed, periconception A1C was modeled by piecewise linear regression with a knot at the lowest modeled value (47 mmol/mol [6.5%]). Differences in effect by type of diabetes were not explored due to small numbers with type 2 diabetes. The absolute risks of serious adverse outcome in the second pregnancy, stratified by outcome in the first and values of periconception A1C, were estimated by taking marginal values of a simplified logistic regression model (conditioning for first-pregnancy outcome and periconception A1C), with 95% CIs estimated using the delta method (22).

Missing data were more likely in women who experienced adverse pregnancy outcomes. Calculations were hence evaluated across 100 multiply imputed data sets. Missing values were estimated by multivariate imputation by chained equations using the variables described plus second and third trimester A1C. Conditional prevalence proportions were estimated by taking marginal values from Poisson regression models, with 95% CIs predicted using the delta method (22). For complete data, 95% CIs for proportions were estimated by the Clopper-Pearson (exact) method (23). Missing values were not predicted for gestational age when required for competing-risk regression. Analyses were performed using Stata 11.1 software (StataCorp LP, College Station, TX). P < 0.05 was considered statistically significant.

The Newcastle Research Ethics Committee originally granted approval for the NorDIP in 1993. Data are now obtained and held with informed consent.

Of the 220 participating women with preexisting diabetes, 89% had type 1 diabetes and 95% were white. The median interpregnancy interval (the time between the end of the first pregnancy and the start of the second) was 1.8 years (interquartile range [IQR] 0.9–3.0), although this was shorter in women whose first pregnancy ended in serious adverse outcome (1.0 years [IQR 0.4–2.1] vs. 2.0 years [1.2–3.2], P < 0.0001). Maternal characteristics during each pregnancy are summarized in Supplementary Tables 1 and 2.

A quarter of women achieved a periconception A1C <53 mmol/mol (7.0%) before their first and second pregnancies (22.6% and 28.9%, respectively), one-half attended preconception care (54.1% and 55.5%, respectively), and two-thirds made their first antenatal visit before 10 weeks (61.6% and 66.2%, respectively) (Table 1). The proportion of women who consumed folic acid supplements before pregnancy increased from 27.1% before the first pregnancy to 43.0% before the second (P = 0.01) (Table 1), although this was not significant after adjusting for year of birth (P = 0.07). Less than half of the women attended both first antenatal visits before 10 weeks (43.2% [95% CI 36.5–49.8]), a third attended preconception care before both pregnancies (35.0% [95% CI 28.6–41.4]), and less than a fifth achieved a periconception A1C <53 mmol/mol or consumed folic acid supplements before both pregnancies (14.4% [95% CI 9.3–19.4] and 15.9% [10.4–21.5], respectively).

Preparation for pregnancy was correlated between pregnancies. Women who in their first pregnancy achieved a periconception A1C <53 mmol/mol, consumed folic acid supplements, and attended preconception care were, respectively, 3.33 (P < 0.0001), 1.57 (P = 0.04), and 1.45 (P = 0.047) times more likely to do so again in the second (Table 1).

A serious adverse outcome in the first pregnancy was not associated with improved preparation in the second. Achieving a periconception A1C <53 mmol/mol, attending the first antenatal visit before 10 weeks, and attendance of preconception care were, if anything, less likely in the second pregnancy among those who had experienced a previous adverse outcome, although none of the associations were statistically significant (A1C: OR adjusted [aOR] for behavior in the first pregnancy 0.65 [95% CI 0.29–1.42], P = 0.28; first appointment before 10 weeks: aOR 0.74 [0.40–1.37], P = 0.34; attendance of preconception care: aOR 0.80 [0.44–1.44], P = 0.45). There was no association between outcome in the first pregnancy and folic acid consumption in the second (aOR 1.01 [0.54–1.88], P = 0.98).

A serious adverse outcome occurred in 39.1% of women (95% CI 32.6–45.9) in at least one pregnancy, and 8.2% (4.9–12.6) experienced serious adverse outcomes in both pregnancies.

A serious adverse outcome affected 30.5% of first pregnancies. There was no difference in prevalence by diabetes type (type 1: 30.8% [95% CI 24.4–37.8] vs. type 2: 28.0% [12.1–49.4], P = 0.78). A total of 17.3% ended in miscarriage, 5.5% in stillbirth, and 1.4% in infant death, and 6.4% were affected by congenital anomaly (Table 1). Of the 14 first pregnancies affected by congenital anomaly, <5 (<35.7%—the count is censored to conform to U.K. disclosure regulations) ended in termination of pregnancy. The total probability of spontaneous fetal death from 6 weeks’ gestation was 33.9% (95% CI 24.7–45.3); from 12 weeks’ gestation was 16.1% (11.4–22.4); and from 24 weeks’ gestation was 6.3% (3.5–11.0).

A total of 178 first pregnancies (80.9% [95% CI 75.1–85.9]) resulted in a registered birth. Of these, 54.5% were delivered by Caesarean section, 4.5% of offspring were SGA, and 42.7% were LGA (Table 1).

Nonwhite ethnicity (P = 0.02), prepregnancy neuropathy (P < 0.0001), increasing maternal age (P = 0.03), smoking during pregnancy (P = 0.01), and increasing periconception A1C ≥47 mmol/mol (P = 0.003) were all independently associated with an increased risk of a serious adverse outcome in the first pregnancy (Table 2).

A serious adverse outcome affected 16.8% of second pregnancies, 0.55 times (P = 0.004) the rate among first pregnancies (Table 1). There was no difference in prevalence by diabetes type (type 1: 16.9% [95% CI 11.9–22.9] vs. type 2: 16.0% [4.5–36.1], P = 0.91). The proportions of second pregnancies ending in miscarriage (5.5%) and stillbirth (1.4%), respectively, were 0.32 times (P = 0.0005) and 0.25 times (P = 0.03) the rate among first pregnancies (Table 1). The proportions of second pregnancies that ended in infant death (0.5%) or were affected by congenital anomaly (9.5%) were not significantly different from the rates among first pregnancies (P = 0.28 and P = 0.24, respectively) (Table 1). Of the 21 second pregnancies affected by congenital anomaly, <5 (<23.8%) ended in termination of pregnancy. The total probability of spontaneous fetal death from 6 weeks’ gestation was 11.9% (95% CI 6.1–22.6), from 12 weeks’ gestation was 2.7% (1.1–6.4), and from 24 weeks’ gestation was 1.5% (0.5–4.6).

A total of 201 second pregnancies (91.4% [95% CI 87.6–95.1]) resulted in a registered birth. Of these, 60.7% were delivered by Caesarean section. The proportion of births delivered by Caesarean section in the second pregnancy was 2.81 times greater (P < 0.0001) in women whose previous birth was delivered by Caesarean section (88.6%); 11.4% (95% CI 5.6–19.9) delivered by vaginal birth after a Caesarean delivery (Table 1). Of births in the second pregnancy, 3.5% were SGA and 58.2% were LGA (Table 1). The proportion of LGA births in the second pregnancy was 1.55 times greater (P = 0.045) in women whose first birth was LGA (69.5%) (Table 1).

Women whose first pregnancy resulted in a serious adverse outcome experienced more than twice the prevalence of a serious adverse outcome in the second (26.9% vs. 12.4%; SHR 2.59 [95% CI 1.35–4.96], P = 0.004) (Table 1 and Table 3). Nearly a third of this was explained by other factors. Nonwhite ethnicity (P = 0.02), prepregnancy nephropathy (P = 0.02), increasing periconception A1C ≥47 mmol/mol (P = 0.0008), and earlier year of delivery (P = 0.002) were all independently associated with increased risk of serious adverse outcome in the second pregnancy (Table 3). After adjusting for these and other variables with P < 0.1, the association between previous adverse outcome and risk in the second pregnancy was not statistically significant (adjusted SHR 1.83 [95% CI 0.96–3.47], P = 0.07) (Table 3).

To establish the relative importance of contemporaneous compared with historical A1C, additional analyses included periconception A1C in the previous pregnancy. There was no crude association between the first pregnancy A1C and the risk of a serious adverse outcome in the second (A1C <47 mmol/mol: SHR 1.13 [95% CI 0.92–1.40], P = 0.25); A1C ≥47mmo/mol: SHR 1.00 [0.99–1.02], P = 0.46). After adjusting for other model variables; however, there was some suggestion of a lower conditional risk for increasing values of A1C ≥47 mmol/mol, although the effect was outside the nominal significance level (A1C <47 mmol/mol: aSHR 1.15 [95% CI 0.9–1.44], P = 0.24; A1C ≥47 mmol/mol: aSHR 0.98 [0.95–1.00], P = 0.054).

The absolute risks of a serious adverse outcome in the second pregnancy, stratified by periconception A1C and first pregnancy outcome, are reported in Table 4.

This study describes the preparation for and outcome of the first and second pregnancy in women with preexisting diabetes. The overall risk of a serious adverse outcome fell from 30% in the first pregnancy to 17% in the second, predominately due to a fall from 34 to 12% in the probability of spontaneous fetal death.

Women who experienced a serious adverse outcome in their first pregnancy were more than twice as likely to experience another serious adverse outcome in their second. A third of this was explained by persistent risk factors such as ethnicity and periconception A1C.

A greater proportion of women achieved a favorable periconception A1C and consumed folic acid supplements before their second pregnancy than before their first, although both remained minority behaviors. There were no differences in the proportion of women who attended preconception care or their first antenatal visit before 10 weeks. Achieving a periconception A1C <53 mmol/mol, prepregnancy folic acid supplement use, and attendance of preconception care were all more likely in the second pregnancy if they had occurred in the first, but there was no evidence that an adverse outcome in the first pregnancy was associated with preparation for the second.

This study benefitted from the North of England’s unique range of population-based registers. The NorDIP is England’s longest-running uninterrupted audit of pregnancies in women with preexisting diabetes and one of few registers that supports the study of repeated pregnancies in the same mother. Detailed information is gathered before and during pregnancy on a range of clinical and sociodemographic variables. Cases of congenital anomaly were identified by the U.K.’s longest-running regional register of congenital anomalies, which receives information, regardless of pregnancy outcome, from multiple sources up to 12 years after birth. The PMS has been collecting information on all stillbirths and infant deaths within the region since 1981 and cross-references with mortality records from the U.K. Office for National Statistics. The results of this study are likely to be generalizable to any predominately white population with similar standards of peripregnancy care.

Several limitations result from low statistical power. Owing to small numbers with each outcome specifically, multivariable analyses was used to examine the composite variable, serious adverse pregnancy outcome, despite possible heterogeneity. Only associations that apply to all constituent outcomes are likely to have been detected. The small numbers with type 2 diabetes (n = 25) prohibited examination of effect modification by diabetes type, although previous studies have found negligible evidence of such differences (7,8,16). Several important exposures also had low numbers. Despite significant associations, the sample was too small to stratify the second pregnancy absolute risks by ethnicity or prepregnancy nephropathy. Lack of statistical significance should not be considered evidence of no effect, as demonstrated by the biologically implausible disagreement in the influence of smoking between the first and second pregnancies (Tables 2 and 3). Similarly, the interpregnancy differences in the contributions of neuropathy and nephropathy are consistent with sampling variation. Data were more likely to be missing in women who experienced serious adverse outcomes. Multivariate imputation by chained equations was used to reduce any consequent bias but cannot account for unknown predictors of missingness. Individuals with mild microvascular complications may not have been ascertained because only “clinically diagnosed” cases (regardless of method) were recorded. Other potentially relevant exposures, most notably medication use, were not collected.

It is unlikely that all pregnancies ending in miscarriage were ascertained. Losses before 6 weeks are typically undetected (24), whereas later losses may be recognized but not reported. The earliest miscarriage in a registered pregnancy occurred at 6 weeks, by which time a quarter of the women had attended their first antenatal appointment. Kaplan-Meier scales the denominator to account for different entry and exit times (25); thus, this study should provide accurate estimates of the risks of spontaneous intrauterine death from 6 weeks onwards. The total risk of miscarriage from conception, however, may be underestimated.

Approximately half of women in the North of England with preexisting diabetes do not seek preconception care (16). For these women, we used first trimester A1C values to approximate preconception A1C. Although highly correlated, this will have increased variation and biased our estimates toward the null. A1C provides an incomplete profile of overall glycemic control because it provides no information on potentially salient glycemic excursions or hypoglycemic episodes (26). Unfortunately, continuous glucose monitoring is not yet routinely available in the U.K.

This study is the first to explore the risk of recurrence of adverse pregnancy outcomes in women with preexisting diabetes and to describe the absolute risks in first and second pregnancies specifically. Nevertheless, there are analogous observations in the general population. The RRs of recurrence for both congenital anomalies (at 1.55 [95% CI 0.40–5.99]) and fetal or infant death (at 2.45 [0.96–6.26]), for example, were highly consistent with the doubling of risk in the general population (3–5). Across both pregnancies, the prevalence of congenital anomaly (8.0% [95% CI 5.4–10.5]), stillbirth (3.4% [1.7–5.1]), and infant death (1.2% (<0.1–2.4]) was consistent with previous observations in larger samples from the same population (7.7% [95% CI 6.5–9.1], 2.7% [1.9–3.6], and 0.7% [0.3–1.2], respectively) (7,8). The proportion of pregnancies ending in miscarriages (11.4% [8.4–14.3]) was consistent with the 5–20% that is typically reported in women with diabetes (27–29).

Comparisons of the change in risk between pregnancies are more problematic, due to large differences in the profiles of primiparous and multiparous women (30). This likely explains the discrepancy between the current study and a previous cross-sectional analysis, where no association was found between parity and risk of stillbirth (8). Even in longitudinal studies, the attributable risk of parity may be masked by changes in other risk factors such as maternal age and BMI (31). Nevertheless, it is broadly recognized that the prevalence of a serious adverse pregnancy is greater among first pregnancies. In the general population, the Flenady et al. (32) meta-analysis estimated the risk of stillbirth as 1.40 (95% CI 1.33–1.42) times higher among primiparous women than multiparous. Though smaller than we observed (RR for primiparity vs. multiparity 4.02 [95% CI 1.15–14.04]), the difference is consistent with sampling variation (P = 0.10). This was similar for miscarriage, with the crude RR for the current study (3.17 [95% CI 1.70–5.90]) being higher, but not significantly (P = 0.07), than in a U.K. sample of women of reproductive age (1.75 [1.42–2.14], comparing first and second pregnancies). We did not find a relationship between pregnancy order and the risk of congenital anomaly, despite it being observed in the general population (33). This may reflect our modest sample size or the aforementioned problems comparing longitudinal and cross-sectional data.

Women with preexisting diabetes continue to experience very high risks of serious adverse pregnancy outcomes. In the first pregnancy, 30.5% (95% CI 24.4–37.0) were affected. In the second, as in the general population, outcomes were more favorable, especially among those who had not experienced previous adverse outcome (12.4% [7.6–18.7]). This was not explained by changes in the known risk factors and may instead reflect constitutionally lower risks of, for example, preterm delivery, preeclampsia, and intrauterine growth restriction (34–36).

Among those whose first pregnancy was affected by serious adverse outcome, the risk in the second remained very high (26.9% [95% CI 16.8–39.1]). A third of this was explained by persistent and known exposures. Adverse outcomes were more common in both pregnancies among women from minority ethnic groups, consistent with previous observations (37). This may reflect genetic factors or enduring environmental or behavioral influences. Preparation for pregnancy is particularly poor in nonwhite women in the North of England (16), indicating they may require alternative methods of support such as community-based approaches (38).

We observed a familiar J-shaped association between periconception A1C and adverse outcome (39), with the risk increasing by 2–3% per mmol/mol ≥47 mmol/mol. This reiterates the benefits of good, though not overly strict, prepregnancy glycemic control (8). Notably, although periconception A1C levels were correlated across both pregnancies, only current values were associated with outcome, suggesting a causal and reversible association. However, after adjusting for current values, there was suggestion of a protective effect of A1C in the previous pregnancy, indicating the highest risk may occur in women whose glycemic control deteriorates substantially between pregnancies.

Preparation for pregnancy among our sample was poor. Only a quarter managed the preconception A1C target or took folic acid supplements before their first pregnancy, and only just over half attended preconception care or attended their first antenatal appointment before 10 weeks. Although favorable preparation in the first pregnancy was broadly predictive of repeat behavior in the second, this exposes a disheartening converse. Women whose first pregnancy ended in a serious adverse pregnancy outcome did not prepare any differently for their subsequent pregnancy. With an average interpregnancy interval of only 1 year, there is a narrow window for intervention. Because many of the circumstances that inhibited preparation for the first pregnancy likely remain, this motivates a change in approach such as providing intensive postnatal support covering various aspects of care, including control, contraception, and well-being (38). Such interventions, however, would have to be carefully balanced against the distressing consequences of discussing future pregnancies during a period of grief (40). Regardless, because preconception care was equally poor across both pregnancies, changes or greater choice may be needed in style and setting (38). The barriers to improved pregnancy preparation are multifaceted and complex (41), but further progress is urgently needed to reduce the risk of recurrent tragedy.

Acknowledgments. The authors are grateful to all of the district conveners and coordinators in the North of England for their continued collaboration and support of the NorDIP, PMS, and NorCAS. The authors also thank the staff at the Regional Maternity Survey Office (Public Health England, Newcastle upon Tyne, U.K.) for their help in data tracing and checking.

Funding. This study was partly funded by South Tees Hospitals National Health Service Foundation Trust. The NorDIP, PMS, and NorCAS are funded by Public Health England.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. The views expressed in this article are entirely those of the authors and do not necessarily reflect those of the funders.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. P.W.G.T. performed the primary analysis and drafted the manuscript. P.W.G.T., R.W.B., and R.B. designed the study. R.W.B. acquired the data. R.W.B. and R.B. conceived the project. S.P. prepared the data and performed preliminary analyses. All authors contributed to interpretation of the data, critically reviewed the manuscript, and approved the final version of the manuscript before submission. P.W.G.T. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were published in abstract form in the following publications: 1) Tennant PWG, Prathapan S, Bilous RW, Bell R. Recurrence of adverse pregnancy outcomes in women with pre-gestational diabetes. J Epidemiol Community Health 2011;65(S1):A214–A215; 2) Bell R, Tennant PWG, Prathapan S, Bilous RW. Risk of recurrent adverse pregnancy outcome in women with diabetes. Arch Dis Child (Fetal Neonatal Ed) 2011;96(S1):Fa129; 3) Tennant PWG, Prathapan S, Bilous RW, Bell R. Outcomes in successive pregnancies in women with pre-gestational diabetes. Diabet Med 2011;28(S1):171; and 4) Tennant PWG, Bilous RW, Prathapan S, Bell R. Risk of serious adverse outcomes in the first and second pregnancies of women with pre-existing diabetes: a cohort study from the North of England. J Epidemiol Community Health 2014;68(S1):A19.