Optimal glycaemic control is of the utmost importance to achieve the best possible outcome of a pregnancy complicated by diabetes. This holds for pregnancies in women with preconceptional type 1 or type 2 diabetes as well as for pregnancies complicated by gestational diabetes. Glycaemic control is conventionally expressed in the HbA1c value but the HbA1c value does not completely capture the complexity of glycaemic control. The daily glucose profile measured by the patients themselves through measurements performed in capillary blood obtained by finger stick provides valuable information needed to adjust insulin therapy. Hypoglycaemia is the major threat to the pregnant woman or the woman with tight glycaemic control in the run-up to pregnancy. Repetitive hypoglycaemia can lead to hypoglycaemia unawareness, which is reversible with prevention of hypoglycaemia. A delicate balance should be struck between preventing hyperglycaemia and hypoglycaemia. Insulin requirements are not uniform across the day: it is low during the night with a more or less pronounced rise at dawn, followed by a gradual decrease during the remainder of the day. A basal amount of insulin is needed to regulate the endogenous glucose production, short-acting insulin shots are needed to handle exogenous glucose loads. Insulin therapy means two choices: the type of insulin used and the method of insulin administration. Regarding the type of insulin, the choice is between human and analogue insulins. The analogue short-acting insulin aspart has been shown to be safe during pregnancy in a randomised trial and has received registration for this indication; the short-acting analogue insulin lispro has been shown to be safe in observational studies. No such information is available on the long-acting insulin analogues detemir and glargine and both are prescribed off-label with human long-acting insulin as obvious alternatives. Randomised trials have not been able to show superiority of continuous subcutaneous insulin administration (CSII (insulin pump)) over intensive insulin injection therapy (multiple-dose insulin (MDI)) on any maternal or foeto-neonatal end point. However, group sizes were far too small to allow assessment of superiority and issues such as manageability of the disease and quality of life were never assessed. These two issues are of major importance to patients. The first trimester is often the period of most hypoglycaemic events, and insulin therapy should be especially closely monitored and adjusted in this period. After midterm, insulin requirements increase. Continuous glucose monitoring can offer better insights into the glycaemic profile than self-monitoring of blood glucose levels by the patients but the place of these new monitoring techniques has yet to be established more clearly. Insulin therapy during labour means short-acting insulin adjusted to achieve glucose levels between 4 and 8 mmol l −1 to prevent neonatal hypoglycaemia as much as possible. After delivery, glycaemic control must be relaxed to prevent hypoglycaemia, especially in women who breastfeed.
Insulin therapy is the mainstay to achieve optimal glycaemic control during pregnancy complicated by any form of diabetes. In general, the diabetic pregnancy is associated with an excess of adverse foetal/neonatal and maternal pregnancy outcomes. Foetal/neonatal risks include congenital malformations, foetal growth acceleration and macrosomia, premature birth, birth trauma and neonatal hypoglycaemia and hyperbilirubinaemia; maternal complications are pre-eclampsia and haemolysis, elevated liver enzymes, low platelets (HELLP) syndrome and primary or secondary caesarean section. These complications are directly or indirectly linked to the degree of glycaemic control.
Insulin therapy is a complex therapy and success of treatment depends on many factors and choices. Choices include the type of insulin used (human insulin or analogue insulin), the method of subcutaneous insulin administration (multiple-dose insulin injection therapy (MDI)) or externally worn insulin pump with subcutaneous insulin delivery (continuous subcutaneous insulin infusion (CSII)), the possibilities of self-monitoring of blood glucose levels by patients (self-measurement of blood glucose (SMBG) levels), the possibility of continuous glucose monitoring (continuous glucose monitoring (CGM)), the risk of (severe) hypoglycaemia and the targets used for blood glucose levels. All these issues will be dealt with in this article.
The diabetic pregnancy is a very complex adventure for the woman involved, her family and the health-care team, and still constitutes a high-risk pregnancy fraught with difficulties. A reliable estimation of the risk of adverse outcomes can only be established in large series. Apart from optimal insulin therapy, other issues are of great importance in this area to achieve the best result. Preconceptional planning and preparation are essential. The issue of fertility, pregnancy and adequate birth control should be discussed at appropriate moments with all women of fertile age. At that moment, choices about optimal insulin therapy should be discussed, considered and made. Potentially teratogenic medications must be discussed and at an appropriate time, when there is a wish to have a pregnancy, conducting a full clinical survey and laboratory assessment is mandatory.
General considerations on glycaemic control
Expressing glycaemic control using HbA1c values
Adequate glycaemic control is a major goal in treating diabetes in pregnancy. The goal defined in general terms is to achieve an HbA1c-value within or as close to the normal range as possible. HbA1c has gained the stature of ‘gold standard’ and provides an easy digital parameter but it has some drawbacks. First, during the first part of pregnancy, there is a normal physiological drop in HbA1c with a mean of approximately 0.5%. This can be attributed to increased erythropoiesis with younger erythrocytes being exposed to glucose levels for a shorter period of time. Second, HbA1c does not reflect the complexities of glycaemic control and large glucose variations (including recurrent hypoglycaemia and also severe hypoglycaemia) can exist in patients with an apparently adequate HbA1c. Glucose variability currently receives much attention but the pathophysiological meaning of glucose variability during pregnancy is still unclear and remains to be proved in terms of association with adverse pregnancy outcomes. The mathematical expression of glucose variability is another challenge and many different parameters have been used. Recently, Rodbarth et al. have shown in a more general population of patients with diabetes that standard deviation is closely related to all other parameters of variability but is most easily calculated. To reliably assess glucose variability, a continuous glucose measurement of at least 24 h is required; however, ideally, it should be done at least on two consecutive occasions.
Summary
Expressing glycaemic control remains a difficult problem and has an important bearing for both clinical management and advancing research. In the clinical research setting, HbA1c is the obvious choice as primary end point, despite the difficulties of using HbA1c as has been outlined. Perhaps the best parameter would be the combination of an HbA1c as close to the normal range as possible without excessive incidence of hypoglycaemia.
Hypoglycaemia
The major threat to the pregnant woman with diabetes or the woman trying to achieve the best control in the run-up to conception is hypoglycaemia. Hypoglycaemic symptoms can be divided in adrenergic and neuroglycopenic ( Fig. 1 ). With decreasing blood glucose levels, adrenergic symptoms (palpitations, transpiration, tremor and hunger) alert the individual to hypoglycaemia and lead to actions to resolve the hypoglycaemia. The blood glucose threshold for neuroglycopenic symptoms is usually lower than that for adrenergic ones. Neuroglycopenic symptoms are altered behaviour, mood swings and, finally, lowering of the level of consciousness and convulsions, posing a direct threat to the individual. Severe hypoglycaemia is defined as hypoglycaemia requiring assistance of another person and/or associated with loss of consciousness, convulsion, coma and death. Severe hypoglycaemia is linked to neuroglycopenia and is a potential fatal complication of glucose-lowering treatment. In a national survey in the Netherlands in 2000, on the preparation for pregnancy and pregnancy outcome in type 1 diabetes, Evers et al. found that there were two fatalities in 324 women, one directly attributed to hypoglycaemia. The incidence of maternal mortality was 0.6%, roughly 60 times the normal incidence. Hypoglycaemia unawareness is closely related to severe hypoglycaemia and implies that signals of hypoglycaemia are detected less well and/or at a lower level of blood glucose concentration, that is, appropriate counteractions are not taken or until at a late stage. Hypoglycaemia unawareness can develop in the setting of recurrent hypoglycaemia and is a temporary phenomenon – disappearing when glycaemic control is relaxed to some extent and hypoglycaemia is prevented for a while. In some patients with type 1 diabetes, hypoglycaemia unawareness is a feature of their diabetic condition and, in such a situation, it is wise to consider defining the target of glycaemic control with less strict criteria. In case of severe hypoglycaemia, oral administration of glucose can lead to aspiration and aspiration pneumonia. In this situation, glucagon should be administered intramuscularly. Patients should have glucagon ready-to-use devices at home and possibly carry it with them or have it at their workplace.
Summary
Hypoglycaemia is frequent in the diabetic pregnancy and repetitive hypoglycaemia can lead to temporary increase in hypoglycaemia unawareness. Hypoglycaemia is a hazardous phenomenon that can contribute to maternal morbidity and mortality.
Physiological profile
The physiological insulin requirement has two distinct parts: the prandial requirement and the non-prandial (basal) requirement. The basal requirement of insulin denotes the insulin level needed to adequately regulate the endogenous glucose production by the liver. In the fasting or post-absorptive phase, glucose levels are maintained by endogenous glucose production through glycogenolysis and gluconeogenesis, mainly in the liver. This is most evident during the night but is also present during the day and prandial periods. Insulin (produced by the pancreatic β-cells) and glucagon (produced by the pancreatic α-cells) regulate endogenous glucose production with insulin inhibiting glucose production and glucagon stimulating production. Insulin levels required during the day and night to achieve adequate glucose production may differ and levels differ between individuals. In normal physiology, endogenous glucose production is necessary to prevent hypoglycaemia in a period of fasting. In diabetes. it means that any insulin regimen used should aim at achieving two things: adequately provide insulin to handle exogenous glucose loads (our meals) and adequately manage endogenous glucose production, especially during the night. To complicate matters, insulin resistance also determines insulin requirement and there is a transient physiological rise in insulin resistance at the end of the night between 4 and 6 am. This rise is attributed to increasing levels of cortisol and growth hormone. These considerations are important when designing the individual insulin therapy and help to choose between different insulin administration methods. To complicate matters even further, insulin resistance (and thus insulin requirement) increases after the 20th week of gestation and, in some individuals, insulin requirement is lower during the first weeks of pregnancy, leading to hypoglycaemia when insulin dose is not adjusted.
Summary
There is a distinct pattern of insulin requirement; hence, insulin therapy has to meet the demands as much as possible. Insulin therapy has to cover the basal insulin need to limit endogenous glucose production and needs to cover insulin requirements after a meal, that is, deal with the exogenous glucose loads.
Method of insulin delivery
Types of insulins used
Do analogue insulins offer an advantage over human insulins? The situation with short-acting analogue insulin has become clearer during recent years. Insulin aspart has been studied in large randomised clinical trial in patients with type 1 diabetes. The aim of the study was to compare the outcome in women with type 1 diabetes who were treated with either an insulin analogue aspart and long-acting human insulin combination or an insulin actrapid and long-acting human insulin combination. Primary end point was the incidence of major hypoglycaemia, which dictated a group size of at least 190 patients in each group (accommodating drop-outs including miscarriages). Women either were early in pregnancy (gestational age ≤10 weeks) or were planning to become pregnant. From the latter group, only women who became pregnant within 1 year could take part in the study. The HbA1c level had to be 8.0% or less early in pregnancy. Major exclusion criteria were having had three or more spontaneous abortions or stillbirths, previous birth of a child with a major congenital malformation, severe renal impairment, active proliferative retinopathy, other endocrine disorders and current fertility treatment. The target values for glucose control were strict: fasting or preprandial glucose between 4.1 and 6.1 mmol l −1 , glucose levels 1 h after a meal less than 8.6 mmol l −1 or less than 7.5 mmol l −1 2 h after a meal.
Recruitment took place between 2002 and 2005. A total of 322 patients took part in the actual study and 264 completed the whole trial. Reasons for withdrawal were adverse events and other reasons, evenly distributed between the two groups. Perinatal mortality (stillbirth from 22 weeks gestational age to delivery + neonatal death ≤6 days) was comparable between the two groups (aspart + human long-acting insulin vs. actrapid + human long-acting insulin: 1.4% vs. 2.2%). No significant difference in birth weight or percentage appropriate for gestational age neonates was observed; pre-term delivery with aspart was less frequent than with actrapid (20.3 vs. 30.6%, p = 0.053) just failing to reach the threshold of statistical significance. Incidence of major congenital malformation was 4.4% in the aspart group and 6.6% in the actrapid group (no significant difference). Neonatal hypoglycaemia (less than 2.6 mmol l −1 ) was comparable in both groups (aspart 33.6% and actrapid 39.7%). In summary, regarding foetal and neonatal outcomes, no significant difference between the groups was seen, except perhaps for a tendency for a lower risk for pre-term delivery with the aspart combination. However, the study was not powered for this outcome and maybe a slightly larger study group might have provided a statistically significant outcome. Mathiesen et al. described the maternal outcomes, including the primary end-point incidence of major (maternal) hypoglycaemia of this trial. There was no statistically significant difference between the rate of major hypoglycaemia, major nocturnal hypoglycaemia or major daytime hypoglycaemia between the two groups. The risk for major hypoglycaemia with the aspart combination was 28% lower but this difference was not sufficient for statistical significance. The risk for nocturnal hypoglycaemia was 52% lower with aspart but this large difference was not sufficient for statistical significance because of the limited number of events in absolute terms. No difference in HbA1c levels was observed. However, aspart led to significantly lower postprandial glucose levels during the first and the third trimester compared with actrapid.
McCance et al. have demonstrated in a spin-off study that there is minimal and no increase in insulin antibodies during pregnancy for either insulin aspart or human long-acting insulin, and that there is no appreciable transplacental transfer of either insulins. Insulin antibodies are not associated with foetal birth weight. A similar low immunogenicity of insulin aspart as well as the comparator drug short-acting human insulin was observed in a small randomised trial in gestational diabetes by Pettitt et al. No randomised trials have been performed with insulin lispro or the newest one, insulin glulysine.
A large observational study on insulin lispro has shown that there is no indication that the use of insulin lispro is related to a higher incidence of congenital malformations. This was an unblinded, retrospective multi-national multi-continental study of 496 women with type 1 or type 2 diabetes mellitus from 1996 to 2001, who had used insulin lispro at least 1 month before conception and throughout the first trimester. Data capture was on 88.9% of eligible women. Because of legal restrictions, it was not possible to obtain essential data from the women who declined information; hence, no comparison could be made between women who did and did not participate. The 496 women provided information on 533 pregnancies, resulting in 500 live births. The majority of women had type 1 diabetes (481 (97%)). Although not the purpose of the study, it showed that only 28.4% had had preconceptional counselling. Macrosomia (birth weight ≥4000 g occurred in 23.4% and 71.6% of the women delivered by caesarean section). Mean gestational age at delivery was 36.7 weeks, which means that the incidence of large for gestational age babies was much higher than the macrosomia rate, since pregnancy duration did not allow the foetus to grow to ≥4000 g. The incidence of major malformations (defined as abnormal structure leading to abnormal function or requiring surgery) was 5.4% with an incidence of all malformations (major and minor) of 5.8%. These figures are not significantly different from data available from other groups not treated with insulin lispro. Periconceptional HbA1c was the single predictor of major malformations. Mean HBA1c was 8.9%, considerably higher than, for example, in the randomised trial with aspart described earlier.
Boskvic et al. and Holchberg et al. have described that there is no transplacental transport of insulin lispro in the term placenta. Similarly, Jovanovich et al. also observed no insulin lispro or regular insulin in the foetal circulation after delivery in women with gestational diabetes. This randomised study showed, just as insulin aspart, that analogue short-acting insulin has a significantly favourable effect on the postprandial metabolic excursions compared with regular short-acting insulin.
The long-acting analogue insulins glargine and detemir are not registered for use in pregnancy and are prescribed off-label. They both demonstrate less intra-individual variability in absorption from the subcutaneous tissue and have a duration of action that is longer than that of human long-acting insulin (22–24 h vs. 18 h). This leads to the paradoxical situation that currently non-pregnant women use analogue short-acting and long-acting insulin to achieve the best glycaemic control, preventing as much as possible acute and chronic complications, whereas, at the same time, when they want to become pregnant they may have to revert to the ‘old’ human insulin with probably less satisfactory control. It is the choice between the devil one knows (‘old’ insulins which are safe but with less glycaemic control) and the devil one does not know (i.e., analogue long-acting insulin with better control but theoretically potential teratogenic effects). A recent ex vivo perfusion study using term placenta has shown that there is no appreciable transplacental transfer of insulin glargine in the normal range of insulin doses. No such data exist for detemir yet.
Clinical experience with both insulins is limited; some articles have appeared describing in general successful pregnancies with these insulins. The numbers are too small to draw conclusions. A large randomised trial comparing insulin detemir with long-acting human insulin is currently underway.
Summary
Putting all data together, short-acting insulin analogues are the drugs of choice in pregnancy over human short-acting ones, with insulin aspart being the only one studied in a randomised trial. The situation is less-clear regarding the new long-acting insulin analogues glargine or detemir.
Method of insulin administration
Basically, the choice is between MDI injection therapy and subcutaneous insulin infusion with an externally worn pump (CSII). With MDI, long-acting insulin once or twice daily is combined with short-acting insulin with each meal or major snack. The long-acting insulin serves to suppress endogenous glucose production. The dose of long-acting insulin is determined by the fasting glucose level, which should ideally be generally between 4 and 6 mmol l −1 . Choosing the right dose is not as simple as it may seem. The long-acting insulin dose should have a smooth pharmacodynamic profile, that is, a stable overnight and day profile with every injection. The first caveat is what to do when the fasting glucose level is too high. First, it can be as a consequence of a dose of insulin that is too low. Second, it can be the consequence of nocturnal hypoglycaemia and this hypoglycaemia may go unnoticed. With hypoglycaemia, the acute adrenergic response in combination with a surge in glucagon, cortisol and growth hormone causes a rapid rise in blood glucose values, leading to hyperglycaemia (the Somogyi phenomenon). In the first instance, one should increase the dose and with the second possibility one should lower it. Third, fasting hyperglycaemia may be caused by an exaggerated dawn phenomenon in pregnancy. The dawn phenomenon describes the autonomous rise in plasma glucose caused by a rise in cortisol and growth hormone levels between 4 and 6 am described earlier. The three nocturnal glucose profiles are illustrated in Fig. 2 . When someone has fasting hyperglycaemia caused by an exaggerated dawn phenomenon, mistakenly identifying it as caused by insufficient long-acting insulin, increasing the dose of insulin creates an even higher fasting hyperglycaemia because of the Somogyi effect. To distinguish between these possibilities, glucose levels during the night should be obtained which is notoriously difficult. One can advice to measure glucose levels at 6.00 am on day one, at 5.00 am on day two, at 4.00 am at day three and so on. Alternatively, continuous glucose monitoring can measure the glucose profiles with the patient asleep and identify the individual pattern. This is described in more detail later.
