Johnstone, F D. MD1; Lindsay, R S. MB, PhD2; Steel, J MD3
The management of type 1 diabetes in pregnancy has changed beyond recognition over the past few decades. The principles underlying management are now better understood. Prepregnancy preparation has increased. Blood glucose monitoring has become outpatient based and more accurate. Glycated hemoglobin assays have become available. Insulin delivery systems have greatly improved. There has been increased attention to team working and the systematic organization of care.1 There has been a huge and widespread reduction in perinatal mortality rates.2
The intuitive expectation is that these changes should be accompanied by a progressive reduction in birth weight. After all, the classic feature of the pregnancy complicated by maternal diabetes is accelerated fetal growth, with a birth weight distribution that is unimodal, approximately normal, but shifted markedly to the right.3,4 The underlying treatment paradigm is that the degree of glycemic control relates directly to the abnormality of fetal growth, and there is evidence supportive of this.5 However, despite the enormous improvements in care of diabetes, the anticipated reduction in fetal growth rate in such a population has not been clinically apparent, and indeed there is current concern that birth weights remain high.3,4,6
This article examines birth weight for infants born to women with probable type 1 diabetes, in a single center, over the last 40 years. This time period encompasses all the major improvements in care described above.
MATERIALS AND METHODS
All pregnancies were included if they met the following criteria: the mother had a diagnosis of diabetes and was taking insulin before pregnancy; she had antenatal care and delivered in the Simpson Memorial Maternity Pavilion, Edinburgh; the gestational age at delivery was 28 weeks or greater; the pregnancy was singleton. Given the age, ethnicity, and requirement for insulin therapy before pregnancy, it is likely that the cohort would overwhelmingly be classified as having type 1 diabetes in modern terms. Mothers with clinical diagnoses of diabetes of other types (maturity onset diabetes of the young, type 2 or secondary forms) were excluded.
The following information was recorded from the case notes: the mother's weight at first visit (only if earlier than 14 weeks pregnant), age at delivery, parity, smoking history, years since diagnosis of diabetes, and White class of diabetes. For the infant, the gestation at delivery was recorded in completed weeks, along with birth weight, sex, the existence of any congenital abnormality causing death or known to affect growth, stillbirth, and early neonatal death.
Information was gathered for the 40 years 1960–1999 inclusive. From 1975, all cases had pregnancy dating by ultrasonography, and clinical records were more comprehensive. Where menstrual dates were certain and ultrasound dating agreed within 6 days, gestation was based on last menstrual period. Where ultrasound estimate differed by 7 or more days in the first 14 weeks of pregnancy, ultrasound dating was used.
Normal ranges for birth weight for the total hospital population were established for births between 1987 and 1991.4 These pregnancies all had ultrasound assessment of gestation in early pregnancy. After exclusion of multiple pregnancies, there were 23,234 births in total for the period between 1987 and 1991. Singleton pregnancies were classified by sex, parity (first born or later born), and gestation. The Shapiro-Wilk statistic, a test of the presence of a normal distribution,7 was carried out at each stratification, and the distribution of birth weights at each gestation was found to be normal. Standardized birth weight (Z score) for the cases was calculated as birth weight minus mean birth weight in that total population cell (of the same gestation, fetal sex, and parity), divided by the standard deviation of birth weight in that cell. Standardized birth weight could not be calculated for gestations less 32 weeks.
The univariable association between standardized birth weight (Z score) and year of delivery was investigated by using Spearman's rank correlation and linear regression. A few exclusions were forced by missing data and numbers used for each variable are indicated in the text. Importantly certain measures (maternal smoking, height, weight at booking) were only consistently available after 1975. The effect of birth year on occurrence of perinatal mortality was assessed by logistic regression. In multivariable analysis, a variety of iterative procedures (stepwise, forward, and backward selection) were used in the context of a multiple linear regression model to produce parsimonious models between standardized birth weight as the dependent variable, and a number of independent variables, including year of delivery. All analyses were performed using the statistical analysis system SAS v8.0 (SAS Institute Inc, Cary, NC). The study was considered by the chair of the local research ethics committee and found not to need formal review.
A total of 643 pregnancies satisfied the criteria set out in the methods section. In this selected population, there were 29 stillbirths, 20 neonatal deaths, and 15 severe congenital abnormalities. Birth weight was available for 640 (99.5%). The perinatal mortality rate per thousand total births, after 28 weeks of gestation, was 225 in the 1960s, 102 in the 1970s, 21 in the 1980s, and 10 in the 1990s (Fig. 1). In univariable analysis, there was a highly significant effect of year of birth on perinatal mortality (odds ratio 0.88 [95% confidence interval 0.85–0.92], P < .001, for effect of birth year in 49 cases of perinatal mortality in 643 births). Other covariates (parity, maternal age, White class, gestational age at delivery, duration of diabetes) were not available for all pregnancies, but the effect of year of birth on perinatal mortality remained highly significant after addition of these factors (odds ratio 0.86 [0.82–0.91], P < .001, for effect of birth year in 33 cases of perinatal mortality in 602 births). Excluding major congenital abnormalities, the perinatal mortality rate per 1,000 total births was 197, 71, 10, and 10 per 1,000 total births for successive decades. After exclusion of major congenital abnormalities, birth year remained a highly significant predictor of perinatal mortality (P < .001, for effect of birth year in 38 cases of perinatal mortality in 628 births).
There were considerable demographic changes in the population with time (Table 1). Statistically significant associations were seen between year of delivery (as a continuous variable) and increasing age (n = 641), more advanced White class (n = 611), longer duration of diabetes (n = 607), increased maternal weight at booking (n = 429), and more advanced gestation at delivery (n = 643). There was no significant change in maternal parity, height, or smoking habit. There was a small, but significant, increase in birth weight with time (n = 640) but no difference in standardized birth weight (n = 625, Fig. 2), which was on average 1.41 standard deviations above the population norm (P < .001).
After exclusion of stillbirths and offspring with severe congenital abnormalities, there was no significant relationship of year of birth and unadjusted birth weight (r = 0.07, P = .07, n = 589) and a marginal negative relationship of year of birth and standardized birth weight (r = −0.09, P = .04, Fig. 2). In contrast in multivariable analysis, year of delivery was not a predictor of standardized birth weight (in models including parity, gestational age at delivery, White class, maternal weight, and smoking). Despite standardization by use of the z score, there was a negative relationship of gestation at delivery and standardized birth weight in simple correlation (Spearman r = −0.21, P < .001), and this was also apparent in multivariable analysis (β −0.18, P < .001 adjusted for parity, year of delivery, White class, maternal weight, and smoking). Babies delivered earlier tended to be larger even after adjustment for gestational age. Analysis of this relationship by decade of delivery showed that there was a strong negative relationship between gestational age at delivery and standardized birth weight in the decade 1990–99 (r = −0.30, P < .001), but this was not apparent or significant in earlier decades (data not shown).
The conclusion is straightforward, although the interpretation is not. Over the last 40 years, mothers with type 1 diabetes have experienced no major decrease in standardized birth weight. This is despite the enormous reduction in perinatal mortality, attributable to advances in obstetric and neonatal management, and to the huge improvements in control of diabetes. A recent survey of birth weight in type 1 diabetes over a shorter 16-year period (1979–1995) found no significant change in birth weight with year of birth.8 We are now able to extend these findings to 40 years encompassing a period with a more dramatic change in blood glucose control during pregnancy. Similarly, birth weight has not changed greatly over the past 40 years in offspring of Pima women with type 2 diabetes, despite a similar improvement in perinatal mortality.9
The strength of these data lies in the very long period of data collection. There are a number of weaknesses. Methods used in the clinic—notably ultrasound dating of pregnancy, recording of pregnancy complications, and assessment of glycemia by HbA1c—were either not available or standardized over the course of the last 40 years. There is likely to have been more variable dating of pregnancy early in the cohort leading to potential error in calculation of standardized birth weights particularly in the first decade. Lack of standardized measures of glycemia means that we cannot address precisely the nature or extent of improvements in glycemic control over the course of study. Nevertheless, those changes were substantial. Standard treatment changed from two injections a day in the 1960s to four injections a day in a basal-bolus pattern in the 1990s. No home monitoring of blood glucose was available in the 1960s, semi-quantitative measures only in the 1970s, while by the 1990s meter home monitoring was standard. Given these changes, it is striking that there is so little change in birth weight over such a long course of data collection.
Over the four decades there were important demographic changes that could have an impact on birth weight. In the general population, an upward trend in birth weight over time is recognized, but it is relatively small in absolute terms and very little of this trend remains unexplained after accounting for factors such as maternal height, age, and smoking.10,11 In our cohort of offspring of mothers with diabetes, the increase in maternal weight (about 4 kg per decade) is likely to be associated with an increase in babies' weight. Against this, more women with longstanding diabetes, hypertension, and renal disease are becoming pregnant, and these characteristics are all associated with reduction in birth weight. However, multivariable analysis would suggest that there is little change in birth weight over time even after accounting for maternal weight or the presence of maternal diabetes complications.
It is unlikely that the findings only reflect local practice. No comparison data are available for previous decades from other centers. However, other centers in Scotland have recently reported a similar contemporary mean standardized birth weight 1.57 standard deviations above that expected for nondiabetic pregnancies.12 Perinatal mortality in this series for the 1990s is lower than, but comparable to, national series of births in 1998–1999 (27.8 per 1,000 births: 95% confidence interval 10.2–59.4).12 A current audit shows that variations in practice between different centers in Scotland are not large (unpublished data). In Scotland the rate of congenital abnormality in offspring of mothers with type 1 diabetes in 1998–1999 was 60 per 1,000 births (95% confidence interval 32–101).12 The rate in this series is lower and not comparable because we have recorded only those anomalies thought severe enough to cause death or known to affect growth, stillbirth, and neonatal death. It is likely that some, mainly minor, anomalies have been missed, but the impact on average birth weight should be small.
Overall then the paradox remains. Despite the marked improvement in perinatal mortality over the study period, birth weight remains little changed, and this cannot be explained by known changes in demographics or maternal behavior. Perhaps this finding should make us rethink the paradigms on which management of diabetes in pregnancy is based. Fetal growth is a balance between stimulus and constraint, and maternal diabetes may have complex effects on growth.
A number of effects might be involved. First, hyperglycemia itself might act via changes in placental function to restrain growth. Sustained high levels of glucose inhibit the proliferation of first-trimester trophoblast.13 High glucose levels in early pregnancy may therefore be a critical determinant of uteroplacental function in later pregnancy.
Second, while overall glycemic control may have improved, maternal blood glucose during pregnancy in women with type 1 diabetes remains far from normal, and this may be enough to significantly affect growth. A number of studies have highlighted the relatively weak relationship of maternal HbA1c and birth weight.14–17 In part this may be methodological, studies using multiple HbA1c assays generally showing a weaker relationship,14,15 while those at a single center or using a single assay show a stronger relationship.16,17 Nevertheless in such studies maternal HbA1c accounts for, at most, only 10–20% of the variance of birth weight. In animal models, short pulsatile surges of hyperglycemia later in pregnancy may be more important than sustained hyperglycemia in inducing fetal hyperinsulinemia and subsequent fetal overgrowth.18 This is in keeping with human data where postprandial glucose has been shown to more closely correlate with birth weight in pregnancies complicated by type 1 diabetes in most19,20 but not all studies.21 Use of continuous glucose monitoring systems in women with type 1 diabetes confirms the presence of large fluctuations, including hypoglycemic and hyperglycemic events, which were not recognized by standard self glucose monitoring.22,23 At the same time recent data in women without diabetes find that mean capillary blood glucose is generally not greater than 5.8 mmol/L (104 mg/dL) even postprandially.24 The same authors suggest that birth weight remains normal in offspring of women with type 1 diabetes only when mean maternal glucose remains less than 5.3 mmol/L (95 mg/dL).25 In keeping with this, many babies in our population are still hyperinsulinemic at birth despite modern management.6
If these mechanisms are acting, the overall effect could be as follows. In the 1960s, when control of diabetes was very poor by contemporary standards, very high glucose levels in early pregnancy caused limited proliferation of first-trimester trophoblast and hence reduced uteroplacental function in late pregnancy. The babies became markedly hyperinsulinemic because of pulsatile surges in maternal glucose in the second trimester, and there was a high rate of intrauterine death because growth overran supply. Although they were big, many of the surviving babies may have been relatively growth restricted. More recently, overall glucose control has improved greatly. Mean glucose levels in early pregnancy are not so high as in previous years, and uteroplacental function is now generally satisfactory. However, we still cannot prevent maternal glucose fluctuations in the second trimester. Because they now likely enjoy a greater uteroplacental capacity, they are able to achieve an accelerated growth without running the same risks of sudden decompensation and intrauterine death.
Unfortunately it is impossible to test these hypotheses retrospectively. Consistent measures are not available across the study to assess glycemia. Glycated hemoglobin only became available in 1980, and the type and standardization of the assay have changed several times since. Alternative outcome measures, such as cord insulin, are not available across the study. It is also remains likely that effects on uteroplacental function and fetal insulinemia are not the whole story. There is no reason why these should balance each other out. They do not explain clinical observations that the biggest babies are sometimes born to women who would be expected to have smaller swings in glucose levels such as those with short duration of type 1 diabetes, type 2 diabetes, or gestational diabetes.
Gestation at delivery may be of more significance. The mean gestation at delivery increased by approximately 1.3 weeks over the 40 years of the study period. This largely reflects changes in the timing of induction of labor or elective caesarean delivery. Longer exposure to maternal diabetes would be expected to result in greater eventual deviation in birth weight from the norm.
Overall, our study attests to the dramatic improvement in perinatal mortality over the past 40 years but suggests that further innovation will be needed before significant changes in birth weight are likely to occur.
1. Steel JM, Johnstone FD, Hepburn DA, Smith AF. Can prepregnancy care of diabetic women reduce the risk of abnormal babies? BMJ 1990;301:1070–4.
2. Coustan DR. Perinatal mortality and morbidity. In: Reece EA, Coustan DR, Gabbe SG, editors. Diabetes in Women. Philadelphia (PA): Lippincott Williams & Wilkins; 2004. p. 205–10.
3. Casson IF, Clarke CA, Howard CV, et al. Outcomes of pregnancy in insulin dependent diabetic women: results of a five year population cohort study. BMJ 1997;315:275–8.
4. Johnstone FD, Mao JH, Steel JM, Prescott RJ, Hume R. Factors affecting fetal weight distribution in women with type I diabetes. BJOG 2000;107:1001–6.
5. Freinkel N. Banting Lecture 1980: of pregnancy and progeny. Diabetes 1980;29:1023–35.
6. Lindsay RS, Walker JD, Halsall I, Hales CN, Calder AA, Hamilton BA, et al. Insulin and insulin propeptides at birth in offspring of diabetic mothers. J Clin Endocrinol Metab 2003;88:1664–1671.
7. Shapiro S, Wilk MB. An analysis of variance test for normality (complete samples). Biometrika 1965;52:591–611.
8. Silva IS, Higgins C, Swerdlow AJ, Laing SP, Slater SD, Pearson DW, et al. Birthweight and other pregnancy outcomes in a cohort of women with pre-gestational insulin-treated diabetes mellitus, Scotland, 1979–95. Diabet Med 2005;22:440–7.
9. Lindsay RS, Hanson RL, Bennett PH, Knowler WC. Secular trends in birth weight, BMI, and diabetes in the offspring of diabetic mothers. Diabetes Care 2000;23:1249–54.
10. Bonellie SR, Raab GM. Why are babies getting heavier? Comparison of Scottish births from 1980 to 1992. BMJ 1997;315:1205.
11. Surkan PJ, Hsieh CC, Johansson AL, Dickman PW, Cnattingius S. Reasons for increasing trends in large for gestational age births. Obstet Gynecol 2004;104:720–6.
12. Penney GC, Mair G, Pearson DW. Outcomes of pregnancies in women with type 1 diabetes in Scotland: a national population-based study. BJOG 2003;110:315–8.
13. Weiss U, Cervar M, Puerstner P, et al. Hyperglycaemia in vitro alters the proliferation and mitochondrial activity of the choriocarcinoma cell lines BeWo, JAR and JEG-3 as models for human first-trimester trophoblast. Diabetologia 2001;44:209–19.
14. Penney GC, Mair G, Pearson DW. The relationship between birth weight and maternal glycated haemoglobin (HbA1c) concentration in pregnancies complicated by Type 1 diabetes. Diabet Med 2003;20:162–6.
15. Evers IM, de Valk HW, Mol BW, ter Braak EW, Visser GH. Macrosomia despite good glycaemic control in Type I diabetic pregnancy; results of a nationwide study in The Netherlands. Diabetologia 2002;45:1484–9.
16. Small M, Cameron A, Lunan CB, MacCuish AC. Macrosomia in pregnancy complicated by insulin-dependent diabetes mellitus. Diabetes Care 1987;10:594–9.
17. Lindsay RS, Hamilton BA, Calder AA, Johnstone FD, Walker JD. The relation of insulin, leptin and IGF-1 to birthweight in offspring of women with type 1 diabetes. Clin Endocrinol (Oxf) 2004;61:353–9.
18. Carver TD, Anderson SM, Aldoretta PW, Hay WW. Effect of low-level basal plus marked “pulsatile” hyperglycemia on insulin secretion in fetal sheep. Am J Physiol. 1996;271:E865–71.
19. Combs CA, Gunderson E, Kitzmiller JL, Gavin LA, Main EK. Relationship of fetal macrosomia to maternal postprandial glucose control during pregnancy. Diabetes Care 1992;15:1251–7.
20. Jovanovic-Peterson L, Peterson CM, Reed GF, Metzger BE, Mills JL, Knobb RH, et al. Maternal postprandial glucose levels and infant birth weight: the Diabetes in Early Pregnancy Study. The National Institute of Child Health and Human Development–Diabetes in Early Pregnancy Study. Am J Obstet Gynecol 1991;164:103–11.
21. Persson B, Hanson U. Fetal size at birth in relation to quality of blood glucose control in pregnancies complicated by pregestational diabetes mellitus. Br J Obstet Gynaecol 1996;103:427–33.
22. Kerssen A, de Valk HW, Visser GH. Day-to-day glucose variability during pregnancy in women with Type 1 diabetes mellitus: glucose profiles measured with the Continuous Glucose Monitoring System. BJOG 2004;111:919–24.
23. Yogev Y, Chen R, Ben Haroush A, Phillip M, Jovanovic L, Hod M. Continuous glucose monitoring for the evaluation of gravid women with type 1 diabetes mellitus. Obstet Gynecol 2003;101:633–8.
24. Parretti E, Mecacci F, Papini M, et al. Third-trimester maternal glucose levels from diurnal profiles in nondiabetic pregnancies: correlation with sonographic parameters of fetal growth. Diabetes Care 2001;24:1319–23.
25. Mello G, Parretti E, Mecacci F, La Torre P, Cioni R, Cianciulli D, et al. What degree of maternal metabolic control in women with type 1 diabetes is associated with normal body size and proportions in full-term infants? Diabetes Care 2000;23:1494–8.
© 2006 by The American College of Obstetricians and Gynecologists.