Diagnosis and Management of Diabetes Mellitus in Pregnancy

Few pregnancy problems have been more dramatically impacted by the advances in technology and in our understanding of physiology during the past century than has pregnancy in the mother with diabetes. Prior to the availability of insulin in the early 1920s maternal death was common, and perinatal death was expected in women with diabetes who became pregnant. The advent of insulin in 1921 was the single most important event in an unbroken chain of advances which led first to the virtual elimination of increased maternal mortality, and then much more slowly to a reduction in the perinatal loss rate for diabetic pregnancies such that there is no longer any valid medical reason to dissuade most young women with diabetes from reproducing. A large body of information has accumulated thanks to the efforts of researchers and clinicians in multiple disciplines, including obstetrics and maternal–fetal medicine, pediatrics, and neonatology, internal medicine with its many subspecialties, pathology and laboratory medicine, psychology and psychiatry, nursing, social work, nutrition, and anesthesiology. Scientific organizations have been formed to foster the interdisciplinary interchange of information; these include the International Association of Diabetes in Pregnancy Study Groups (IADPSG), the Diabetes in Pregnancy Council of the American Diabetes Association (ADA), and similar organizations throughout the world. It is likely that no single disorder of pregnancy has sparked the interest of a more diverse group of investigators, and no single entity is a better example of the interplay of advances in all of the above-mentioned fields.

It has been estimated that 0.3% of pregnancies in the United States occur in women with preexisting diabetes mellitus,¹ and that 4–8% more are complicated by gestational diabetes.^{2, 3} This problem is likely to be encountered by every clinician caring for pregnant women. The above-mentioned advances in our understanding and management of diabetic pregnancy may have inspired increased confidence that perinatal outcomes are likely to be positive, but we must never lose sight of the fact that perinatal mortality and morbidity are markedly increased in diabetic pregnancies that are not managed by modern approaches, and that continued vigilance is necessary in order to optimize results. Ideally all pregnancies in women with preexisting diabetes should be cared for in a modern perinatal center with a team approach; those with gestational diabetes should be managed in a systematic fashion with appropriate resources available to address the obstetric, metabolic, pediatric, dietary, nursing, and social needs of the mother and fetus/neonate.

 

Gestational diabetes

Gestational diabetes is defined as "diabetes diagnosed in the second or third trimester of pregnancy that is not clearly either type 1 or type 2 diabetes."⁴ The existence of this disorder was first suspected on the basis of increased perinatal mortality rates among pregnancies in women later developing overt diabetes,⁵ and was first diagnosed by the same criteria as were used for the diagnosis of diabetes in the nonpregnant state.⁶ Because pregnancy exerts significant effects on maternal metabolism, with a fall in fasting and rise in postprandial circulating glucose levels and a marked increase in insulin resistance,⁷ the development of pregnancy-specific glucose tolerance test criteria for gestational diabetes by O'Sullivan and Mahan⁸ in 1964 marked another important milestone in clinical obstetrics. The criteria were based on tests administered to 752 consecutive prenatal patients, primarily in the second and third trimesters, and were validated based upon the likelihood of developing overt diabetes during 7 year follow-up of a separate group of subjects. Long term follow-up by the same investigator has revealed an approximately 60% prevalence of overt diabetes, utilizing current criteria for nonpregnant individuals, at an average of 20 years after the index pregnancies.⁹ It is important to note that the O'Sullivan criteria were validated on the basis of their predictive value for subsequent maternal diabetes, and not for pregnancy outcome.

GLUCOSE TOLERANCE TESTING

The glucose tolerance test, which is currently standard for pregnant women in the United States and is recommended for use in pregnancy by the American College of Obstetricians and Gynecologists,¹⁰ is the 100-g, 3-hour oral test (OGTT). The 100-g glucose challenge is administered after an overnight fast of 8–14 hours, and after 2–3 days of an unrestricted diet containing at least 150 g of carbohydrate. The dietary preparation is important, since individuals who are carbohydrate depleted will not mount as effective an insulin response to a glucose challenge, and may manifest higher glucose levels than they would have if properly prepared. Blood samples are drawn in the fasting state, and at 1, 2 and 3 hours after ingestion of the glucose challenge. Glucose should be analyzed with a standard laboratory technology, and not with test strips and reflectance meters which are designed for self glucose monitoring. Although quite useful in management of the individual with already diagnosed diabetes, the latter method is too imprecise for use in diagnostic testing.¹¹

When O'Sullivan and Mahan originally derived their criteria for the diagnosis of gestational diabetes,⁸ they opted for the presence of at least two out of four values exceeding the given thresholds in order to improve specificity, and to avoid reliance on a single laboratory value to make an important diagnosis. Venous whole blood samples were analyzed using the Somogyi-Nelson method of glucose analysis, and the thresholds originally established are depicted in Table 1. Currently most laboratories measure glucose in plasma or serum specimens, which yield results approximately 14% higher than does whole blood. The National Diabetes Data Group (NDDG)¹² published in 1979 an adaptation of O'Sullivan and Mahan's OGTT thresholds, applying them to plasma and serum specimens. These conversions are shown in Table 2, and in the past have been recommended by the ADA and the American College of Obstetricians and Gynecologists (ACOG).

 

Table 1. Pregnancy oral glucose tolerance test thresholds of O'Sullivan and Mahan*

*Whole blood samples, Somogyi-Nelson method of analysis. Gestational diabetes is diagnosed if two or more threshold values are met or exceeded.

(Adapted from O'Sullivan JB, Mahan CM: Criteria for the oral glucose tolerance test in pregnancy. Diabetes 13:278, 1964)

 

Table 2.  National Diabetes Data Group (NDDG) pregnancy oral glucose tolerance test thresholds*

*Plasma or serum samples. Gestational diabetes is diagnosed if two or more threshold values are met or exceeded.

(Adapted from American Diabetes Association: Position statement of gestational diabetes mellitus. Diabetes Care 18(suppl 1):24, 1995; and National Diabetes Data Group: Classification and diagnosis of diabetes mellitus and other categories of glucose intolerance. Diabetes 28:1039, 1979)

 

The original method of glucose analysis, which also measured some nonglucose reducing substances, has been replaced by more specific enzymatic methods such as glucose oxidase and hexokinase. Consequently, our center¹³ derived O’Sullivan OGTT criteria which are somewhat lower than those derived by the NDDG, having been based on the unrounded original cutoffs, then corrected for the more specific enzymatic methodology by subtracting 5 mg/dL from each value, then increased by 14% to compensate for the change from whole blood samples to plasma or serum. These thresholds are shown in Table 3.

 

Table 3. Pregnancy oral glucose tolerance test thresholds established by Carpenter and Coustan,¹³ based on those of O'Sullivan and Mahan⁸*

*Plasma or serum samples, enzymatic (e.g., hexokinase, glucose oxidase) method. Gestational diabetes is diagnosed if two or more threshold values are met or exceeded.

(Carpenter MW, Coustan DR: Criteria for screening tests for gestational diabetes. Am J Obstet Gynecol 144:768-773, 1982)

 

When performing the OGTT in pregnancy, it is important that the laboratory understand that a 100-g, 3-hour test is required rather than the 75-g, 2-hour test that is standard in nonpregnant individuals. Furthermore, diagnostic criteria for pregnancy should be applied to the results. If the laboratory is unaware of the fact that the patient is pregnant, or is unaware of the need for different criteria in pregnancy, it is possible that gestational diabetes could be over- or underdiagnosed. Another problem that may arise is the patient who is unable to undergo oral glucose tolerance testing because of vomiting. It may help to try a different flavor of glucose challenge, and to administer the oral load on crushed ice. Another option is the use of a relatively tasteless glucose polymer as the challenge.¹⁴ If oral testing is still not possible, an intravenous glucose tolerance test¹⁵ may be substituted, although there is less evidence available to support the use of specific thresholds during pregnancy.

Although the thresholds shown in Table 2 were once widely used in the United States, it should be pointed out that both sets (Tables 2 and 3) are theoretical conversions from those shown in Table 1. When Sacks et al.¹⁶ recreated O'Sullivan and Mahan's⁸ original methodology and ran parallel samples with current methodology, the NDDG thresholds were above 95% confidence limits at all times except fasting, while those in Table 3 were within those confidence limits at all intervals. Such a finding may help to explain reports of increased perinatal morbidity in pregnancies where the glucose tolerance test criteria were nearly, but not quite, exceeded.^{17, 18} Furthermore, a large observational study demonstrated that the criteria listed in Table 3 identify individuals with significantly increased perinatal morbidity compared to the general population, at rates similar to those in the population identified by the criteria in Table 2.¹⁹  In a sub-analysis of patients with GDM enrolled in a multi-institutional randomized clinical trial,²⁰ adverse outcomes (pregnancy-induced hypertension, shoulder dystocia, macrosomia, neonatal fat mass) were reduced with identification and treatment of GDM to a similar extent compared to untreated controls in subjects who met the NDDG criteria in Table 2, and those meeting only the Carpenter & Coustan criteria in Table 3. A likely explanation of these findings is the determination that maternal hyperglycemia is associated with perinatal risk in a continuum, and that any set of thresholds for gestational diabetes must be rather arbitrary with borderline cases on either side. A blinded multi-institutional multinational observational study has demonstrated a continuous relationship between 75-g glucose tolerance test results and various outcomes, including macrosomia, cesarean section rate, neonatal hypoglycemia, and cord blood C-peptide (a surrogate for insulin) among individuals without gestational diabetes.²¹

 

INTERNATIONAL ASSOCIATION OF THE DIABETES AND PREGNANCY STUDY GROUPS CRITERIA

In addition to the various derivations of cutoffs for the 100 g, 3-hour OGTT based on the O’Sullivan and Mahan criteria⁸ described above, a number of other diagnostic schemes for GDM have been in use around the world. These include, but are not limited to, the 1998 World Health Organization criteria²²  which were simply the same criteria as in use for non-pregnant individuals to diagnose impaired glucose tolerance, and utilized a 75 g, 2-hour OGTT. None were based on pregnancy outcomes.  Because the 75 g, 2-hour OGTT was universally accepted for diagnosing diabetes and prediabetes in nonpregnant individuals, and because there was a need for diagnostic criteria which were based on their predictive value for potentially preventable adverse pregnancy outcomes, the Hyperglycemia and Adverse Pregnancy Outcome (HAPO) Study²¹ was undertaken. This was a large (>23,000 completed subjects), multinational observational study designed to determine the relationship between each of the three plasma glucose values during a 75 g, 2-hour OGTT and a number of adverse pregnancy outcomes including fetal macrosomia, primary cesarean section, preeclampsia, shoulder dystocia and neonatal hypoglycaemia. If the relationship between OGTT glucose values and adverse outcomes demonstrated an inflection point, above which problems were markedly increased and below which adverse outcomes were minimal, the choice of cutoffs for diagnosing gestational diabetes would have been relatedly straightforward. As noted above, the relationships were all continuous, with no obvious inflection points. This meant that the cutoffs for diagnosing GDM would be somewhat arbitrary. In order to achieve the hoped for worldwide acceptance of outcomes-based GDM criteria, the International Association of Diabetes in Pregnancy Study Groups (IADPSG) convened a large panel of experts from around the world, and carried out an iterative process to achieve consensus. The final recommendation²³ was to use diagnostic criteria based upon odds ratios for adverse outcomes (including macrosomia, primary cesarean section and cord blood C-peptide above the 90th centile, which is reflective of fetal hyperinsulinemia) of 1.75, compared to mean glucose values for the population at large. These criteria are shown in Table 4 below:

Table 4. IADPSG criteria for gestational diabetes mellitus²³

Based upon a 75 g, 2-hour OGTT; if one or more of the cutoffs is met or exceeded, gestational diabetes is diagnosed.

 

These criteria would have identified over 16% of subjects in the HAPO study as having gestational diabetes. A great deal of controversy has arisen over such a high prevalence. Since the IADPSG cutoffs for GDM are not dissimilar to the generally accepted cutoffs for prediabetes in nonpregnant individuals (fasting plasma glucose 100–125 mg/dL [5.55–6.9 mmol/L), it is worth noting that the prevalence of diabetes among individuals aged 20-44 years in the US was approximately 5% in 2011–12, and prediabetes was estimated at approximately 30% in that age group, totalling over one third of reproductive aged Americans having glucose dysfunction.²⁴ From this perspective, a GDM rate of >16% does not seem excessive. The global epidemic of obesity and type 2 diabetes is a challenge for all health care systems, but the solution is not to revise diagnostic thresholds upwards in order to lower the prevalence, but rather to develop innovative approaches to providing care.

The IADPSG diagnostic criteria for gestational diabetes have now been adopted by a number of leading organizations around the world, including the World Health Organization²⁵ and the International Federation of Gynecology and Obstetrics²⁶ and The Endocrine Society in the US.²⁷ In the United States a consensus conference held in 2013²⁸ recommended that the 100 g, 3-hour glucose tolerance test described above continue to be utilized because of a number of concerns with the IADPSG recommendations, including the increased number of patients diagnosed with GDM leading to an increased burden on the health care system, along with a perceived need for more evidence regarding the potential benefits of identifying and treating so many gravidas. Subsequently the American College of Obstetricians and Gynecologists (ACOG) echoed those recommendations.¹⁰ The American Diabetes Association recommends either the IADPSG 75 g, 2-hour OGTT or the 100 g, 3-hour test, while noting that the IADPSG criteria are based on pregnancy outcomes. A number of critiques of these IADPSG criteria have been published.²⁹  A response to these concerns is too long for inclusion in this chapter, but is very much worth reading.³⁰  The data on which the response is based are largely contained in another recent publication.³¹

SCREENING

Population based screening for gestational diabetes was previously called into question because of the lack of well controlled outcomes research to determine the overall benefit to society of such programs.³² More recently, in 2014 the US Preventive Services Task Force recommended screening for GDM in asymptomatic pregnant women after 24 weeks of gestation.³³

The risks of gestational diabetes to the fetus and newborn, and its implications for the mother’s future likelihood of diabetes, are well documented and are discussed later in this chapter. Furthermore, in every series of gestational diabetic patients there are some who have previously undiagnosed preexisting diabetes. Some sort of testing for this disorder is deemed appropriate by most caregivers in obstetrics and gynecology. The simplest approach to screening for gestational diabetes is the taking of a history and the performance of an OGTT whenever risk factors for diabetes are present. Traditional risk factors include the family history of diabetes in a first (or second) degree relative, the history of a previous perinatal loss or other adverse pregnancy outcome, the history of the previous birth of a large infant (variously defined as 8.5 pounds, 9 pounds, 4000 g, 4500 g), the presence of glycosuria during the present pregnancy, and obesity. Unfortunately, when risk factors such as these have been sought in universally screened populations, only approximately 50% of individuals with gestational diabetes have such factors.^{2, 34, 35, 36} It is worth noting that in a primarily Caucasian population of 18,000 pregnant women, use of the “low risk” category described above would have eliminated only 10% of the population from glucose testing.³⁷ 

The ADA³⁸ recommends the use of a different set of risk factors for diabetes testing at the first prenatal visit in order to identify those at highest risk. These include overweight (BMI ≥25 kg/m²) plus additional risk factors including physical inactivity, first-degree relatives with diabetes, high risk race/ethnicity, previous history of gestational diabetes or birth of a baby ≥9 pounds, hypertension, polycystic ovary syndrome or other clinical conditions associated with diabetes The remainder of patients are considered of average risk, and (along with high risk patients who had normal glucose testing earlier in pregnancy) should undergo glucose testing between 24 and 28 weeks. The American College of Obstetricians and Gynecologists¹⁰ also recommends early testing of high risk individuals. 

The use of a glucose screening test followed by a full OGTT for individuals with elevated screening test values is often referred to as the “two-step” approach. The most common glucose screening test is a 50-g, 1-hour glucose challenge.¹⁰ If the venous plasma or serum glucose exceeds a predetermined threshold, a full 100-g, 3-hour OGTT is performed. A systematic review for the US Preventive Services Task Force³⁹ concluded that the sensitivity for detecting GDM using a threshold of 140 mg/dL (7.8 mmol/L) was 70–88%, whereas it was 88–99% using a threshold of 130 mg/dL (7.2 mmol/L). Glycated hemoglobin and fasting plasma glucose did not function as well. ADA³⁸ and ACOG¹⁰ note that either a 130 (in the case of ADA), 135 or 140 mg/dL threshold may be utilized. If the higher threshold is used, then approximately 14% of gravidas will need the full OGTT, whereas if the lower threshold is used, 23% will need further testing.¹³

Other options have been described. As mentioned above, the IADPSG recommendations for a one-step process using the new diagnostic criteria and the 2-hour, 75 g OGTT have been adopted in many parts of the world.  Even when the 3-hour, 100 g OGTT is used, there are some populations at high enough risk that a one step process (immediate OGTT without a screening test) is reasonable. For example, some Native American populations have such a high prevalence of gestational diabetes that membership in such an ethnic group carries a higher a priori risk than does a positive screening test. Thus it may be appropriate to perform a full OGTT on all such individuals, without prior glucose challenge screening. Individual practices may choose to modify screening protocols depending upon the characteristics of the population served and the resources available.

As noted above, early testing of gravidas at high risk for previously undiagnosed type 2 diabetes is recommended by ADA³⁸ and a number of other professional organizations. Any of the diagnostic criteria for diabetes in the nonpregnant state may be used, including HbA1c ≥6.5%, fasting plasma glucose ≥126 mg/dL (7 mmol/L) or 2-hour plasma glucose (75 g, 2-h OGTT) ≥200 mg/dL (11.1 mmol/L).  While the IADPSG recommended that if the fasting plasma glucose in early pregnancy is ≥92 mg/dL (5.1 mmol/L) but <126 mg/dL (7 mmol/L) gestational diabetes could be diagnosed, it should be remembered that the diagnostic cut-off of 92 mg/dL (5.1 mmol/L) was based on data from the late second and early third trimesters, so the use of this threshold in the first or early second trimesters is unsupported.  

Preexisting diabetes

A number of different classification schemes have been proposed for pregnancies in women with preexisting diabetes. Dr Priscilla White devised a classification of diabetes⁴⁰ which was based on the likelihood of the presence of vascular disease (Table 5); this classification was intended to offer help in prognosticating the outcomes of pregnancies. It is worth noting that the White classification did not include a category for gestational diabetes. The class A individual, as described by White, had an abnormal glucose tolerance test prior to pregnancy but was treated only by diet, and not insulin. Although the White classification scheme was once learned by every resident in obstetrics and gynecology, its current value in clinical management is dubious. All individuals with preexisting diabetes mellitus who become pregnant should be considered to carry high risk pregnancies. While the presence of vascular disease clearly increases certain maternal and fetal risks, the apparent absence of vascular disease should not be considered reassuring. Another approach to classification is that of Pedersen,⁴¹ who based his prognostically bad signs in pregnancy (PBSP) system on easily recognizable signs associated with perinatal mortality (Table 6). In our own anecdotal experience the patient who is a "neglector" has the highest risk. While such classification schemes have not been altogether helpful in the care of the individual patient, they are a very useful means of comparing data from different centers.

 

Table 5. Priscilla White classification of diabetes mellitus in pregnancy

(Modified from White P: Pregnancy and diabetes. In Marbel A, White P, Bradley RF, Krall LP (eds): Joslin's Diabetes Mellitus, 11th, p588. Philadelphia, Lea and Febiger,1971)

 

Table 6.  Pedersen's prognostically bad signs in pregnancy (PBSP)

(Adapted from Pedersen J: The Pregnant Diabetic and Her Newborn. Baltimore, Williams and Wilkins, 1967)

 

Currently, preexisting diabetes is classified into type 1 (formerly called “juvenile onset” or “insulin dependent”), and type 2, (formerly called “maturity onset” or “non-insulin dependent”) diabetes. The former terms were often a source of confusion. The term "insulin dependent" is not synonymous with "insulin treated." A patient who is "insulin dependent" will lapse into diabetic ketoacidosis (DKA) if insulin is withheld for a substantial length of time; she is "dependent" on insulin to avoid DKA, and for survival. This term described the individual with type 1 diabetes, whose primary defect is pancreatic islet cell dysfunction or obliteration, widely believed to occur on an autoimmune basis. Type 1 diabetes usually, but not always, has its onset during the first one to three decades of life. Individuals with type 2 diabetes have a defect in the action of insulin at the insulin receptor or postreceptor level. Early in the course of their disorder their metabolic derangement may be characterized by insulin resistance. However, many become insulinopenic later in the progression of their disease. Individuals with type 2 diabetes may be "insulin treated," but if they require insulin to avoid DKA they should then be considered to have type 1 diabetes. In general terms, individuals with type 2 diabetes are more likely than not to be obese, older, and to have residual pancreatic islet cell function as measured by C-peptide immunoreactivity. Individuals with type 1 diabetes of childbearing age are more likely to have diabetic vascular complications, but such problems may also be found in people with longstanding type 2 diabetes and appear to be more related to the duration and severity of hyperglycemia than to the etiology of their diabetes. Patients with preexisting diabetes face increased maternal and fetal risks during pregnancy. Specific organ system involvement may entail specific types of risk, as described below. It is best to classify preexisting diabetes as type 1 or type 2, and then describe whatever accompanying diabetic sequelae are present.

We are currently in the midst of a growing epidemic of type 2 diabetes²⁴ primarily related to the increasing prevalence of obesity in the population.⁴² Furthermore, type 2 diabetes is developing at younger ages, and is now common among adolescents. Among individuals younger than 44 years, the prevalence of diabetes doubled between 1988 and 2012, from 2.7% to 4.5%.²⁴ One of the results of this epidemic appears to be a shift in the population of pregnant women with preexisting diabetes. While in the past we primarily saw women with type 1 diabetes presenting for prenatal care, in our clinic pregnant women with type 2 diabetes now far outnumber those with type 1 disease. Furthermore, we seem to be encountering more women who have had type 2 diabetes long enough to have vascular disease, a relatively new phenomenon in the reproductive age group.

 

Mortality

Prior to the availability of insulin, maternal mortality in diabetic women approached 50%.⁴³ This rate fell immediately after insulin was discovered, and in the early 21st century occurred in 0.04% or less of diabetic pregnancies.⁴⁴ This maternal death rate is still approximately four times higher than in the general obstetric population, and may be thought of as the sum of the general maternal mortality rate plus the mortality rate attributable to having diabetes for any 9 month period of time. Individuals with particularly high risk for maternal death are those with previous myocardial infarction,⁴⁵ and it is important to counsel such women about the grave risks associated with pregnancy. However, diabetic women without coronary artery disease need not be advised to avoid pregnancy as may have been the case in previous years.

Glucose control

HYPERGLYCEMIA AND DIABETIC KETOACIDOSIS

Perinatal mortality and morbidity appear to be directly related to the degree to which ambient maternal glucose levels during pregnancy exceed nondiabetic limits. The achievement of near euglycemia is the usual goal in managing diabetic pregnancy. The diabetogenic effects of pregnancy upon maternal metabolism complicate attempts at glucoregulation, but an understanding of these effects can aid the clinician in formulating a strategy. Fasting plasma glucose levels tend to fall by the end of the first trimester, while the response to a glucose challenge increases in amplitude and duration as pregnancy progresses.⁴⁶ The latter change is generally ascribed to increased insulin resistance⁷ and is most pronounced after a pure glucose challenge rather than a mixed nutrient meal.

Among individuals with diabetes, pregnancy has been associated with increasing insulin requirements⁴⁷ particularly in the second half of gestation and, in those with type 1 diabetes, a tendency toward diabetic ketoacidosis at lower ambient glucose levels, presumably because of the insulin resistance mentioned above. Because of the extremely high perinatal mortality rate associated with any episode of ketoacidosis, prompt and intensive treatment of this disorder is essential. It should be remembered that ketoacidosis is not a disorder of "too much glucose," but rather is related to "too little insulin" causing increased hepatic glucose production and the breakdown of stored fats to produce ketone bodies, acetoacetate, and beta-hydroxybutyric acid. Because one of the major problems in DKA is osmotic diuresis with resultant dehydration, large volumes of fluid should be infused initially, along with a constant intravenous insulin infusion. Further details of the management of this problem are available elsewhere.⁴⁸

HYPOGLYCEMIA 

Whenever individuals with type 1 diabetes attempt near-normalization of circulating glucose levels the likelihood of hypoglycemia is  increased. Symptomatic hypoglycemia may occur with particular frequency during attempts at glucoregulation in the first half of pregnancy,^{49, 50, 51} particularly among patients with a previous history of symptomatic hypoglycemia.⁴⁹ There is conflicting evidence suggesting that the glucose counter-regulatory response may⁵² or may not⁵³ be impaired during pregnancy in individuals with type 1 diabetes. Adverse fetal effects of maternal hypoglycemia have not been documented in human pregnancy, and fetal heart rate^{52, 53} as well as fetal movements⁵⁴ and placental perfusion^{53, 54} appear to be unchanged during conditions of maternal hypoglycemia in the range of 45–55 mg/dL. It is clear, however, that maternal compromise may result, i.e., from coma, convulsions, trauma, etc. Thus maternal hypoglycemia should be avoided if possible. However, the relaxing of metabolic control to levels of hyperglycemia that are potentially dangerous to the fetus is not usually justified. Studies of glucose tolerance tests among nondiabetic pregnant women suggest that mild degrees of hyperglycemia and hypoglycemia⁵⁵ may carry similar fetal risk, and that the safest levels of glucose are those that are within normal ranges for pregnancy. The work of Lev-Ran et al.⁵⁶ suggests that "brittleness," or the tendency toward wide swings in glucose levels even under the best of circumstances, tends to ameliorate by mid-pregnancy, and improved metabolic control may be more easily accomplished at this time than in the nonpregnant state.

Medical problems of diabetes

NEPHROPATHY

Overt diabetic nephropathy, generally detected by the presence of albuminuria ≥300 mg/day, complicates both type 1 and type 2 diabetes. However, because type 2 diabetes tends to have a later age at onset than type 1 diabetes, and because the prevalence of nephropathy is associated with the duration of diabetes, pregnant women with type 1 diabetes are considerably more likely to have nephropathy than are those with type 2 diabetes. In one series from a tertiary care institution 23% of pregnant women with preexisting diabetes manifested nephropathy, defined as a urinary protein excretion ≥300 mg/day prior to the third trimester.⁵⁷ Nephropathy develops in stages,⁵⁸ with the first stage manifested by renal hypertrophy and hyperfunction, and then renal lesions without clinical signs. It is only in the next stage, occurring within 7–15 years in 25–40% of individuals with type 1 diabetes, that albuminuria is manifest, and at that point it is microalbuminuria, only 30–299 mg/day. Currently spot urine albumin to creatinine ratio (UACR) is the most convenient approach, with increased albumin excretion defined as ≥30 mg/g creatinine.⁵⁹ Once microalbuminuria is present, it is highly likely that a given individual will progress to clinically evident diabetic nephropathy and ultimately end stage renal disease, a progression which usually occurs at a rate of 10 mL/min decrease in glomerular filtration rate (GFR) per year. Glomerular filtration rate (GFR) is estimated (eGFR) based upon serum creatinine levels, and is reported by most laboratories. Normal eGFR is defined as ≥90 mL/min/1.73 m².Once eGFR is below that threshold, stage 2 diabetic nephropathy is present. Hypertension and retinopathy are often present when nephropathy is detected. Risk factors predisposing toward the development of diabetic nephropathy include genetic factors, hyperglycemia in the first stages, hypertension in the later stages, and possibly the protein content of the diet which is believed to be associated with increased intracapillary pressure in the glomeruli. Restriction of dietary protein may slow the progression of renal damage in diabetic individuals with clinically manifest nephropathy.⁶⁰ The ADA recommends limiting protein intake to 0.8 g/kg body weight (Standards of Medical Care in Diabetes, Chapter 9 referenced above). Such protein restriction is generally not advisable during pregnancy. The use of angiotensin converting enzyme (ACE) inhibition has been shown to significantly impede progression of nephropathy in a randomized trial involving nonpregnant, nonhypertensive individuals with type 1 diabetes.⁶¹ Unfortunately, ACE inhibitors and angiotensin receptor blockers (ARBs) are contraindicated during pregnancy because of potential adverse fetal effects.^{62, 63} The Diabetes Control and Complications Trial (DCCT), published in 1993,⁶⁴ demonstrated, in a randomized study of 1441 patients with type 1 diabetes, that intensive control of ambient glucose levels to an average of approximately 150 mg/dL, over a 6–7 year interval, significantly reduced the appearance and progression of nephropathy. The ADA and the National Kidney Foundation subsequently issued a consensus statement⁶⁵ recommending that optimization of metabolic control is the cornerstone of prevention of progression of diabetic nephropathy. Protein restriction may be of value, and control of hypertension may also be advantageous, particularly with ACE inhibitors or ARBs in the absence of pregnancy.

During normal pregnancy renal plasma flow and GFR rate are markedly increased. In women with diabetic nephropathy the quantity of albuminuria tends to increase dramatically, such that many individuals with microalbuminuria in early pregnancy may excrete more than 5 g/day by the third trimester. This not only raises concerns about a possible adverse effect of pregnancy on nephropathy, but can also make the diagnosis of hypertensive disorders of pregnancy quite difficult. Most of the evidence available to date suggests that pregnancy has no long-lasting effect on diabetic nephropathy, and that most women who develop marked proteinuria during pregnancy revert to their pre-pregnancy renal status after delivery.^{58, 66} Approximately one third of women with nephropathy experience a fall in creatinine clearance during the course of pregnancy.⁵⁸

Diabetic nephropathy clearly is associated with greater risk of adverse outcomes in pregnancy.^{57, 58} Perinatal morality is higher, and preterm delivery is necessary in over half the patients, one quarter prior to 34 weeks. Approximately 18% of infants are growth restricted, and nonreassuring fetal status is common. Almost half of such individuals develop preeclampsia. Severe anemia often accompanies pregnancy in women with nephropathy, because of the renal disease as well as the usual hemodilution of pregnancy. A systematic review in 2013⁶⁷ concluded that the published studies were so heterogeneous with respect to definitions and outcomes that it was not possible aggregate them. The authors interpreted the data as showing that the various studies were in general agreement regarding the negative association between nephropathy in type 1 diabetes and adverse outcomes such as preeclampsia, prematurity, intrauterine growth restriction (IUGR), and stillbirth, with risk increasing as nephropathy advances. In 2015 a retrospective review of 108 consecutive type 1 diabetes patients delivering between 1988 and 2011 in the only hospital caring for such patients in Helsinki, Finland⁶⁸ reported that preeclampsia rates remained high over that time period (52–42%) as did preterm birth (14–21% before 32 weeks and 71–77% before 37 weeks). Perinatal mortality occurred in 4.6% of pregnancies, and major malformations were identified in 9.5%. More than 50% of patients with nephropathy also had proliferative retinopathy. Almost all patients with diabetic nephropathy were delivered by elective or emergency cesarean section. 

The management of nephropathy during pregnancy may be difficult since risks and benefits must be constantly balanced. For example it is usual to prescribe a relatively high protein diet to pregnant women in order to provide amino acids for fetal growth and development, and also to help maintain relative euglycemia in individuals with diabetes. However, high protein intake may have an adverse effect on the progression of nephropathy. A middle ground is probably most appropriate, and it may be reasonable to decrease the protein content of the diet but to no less than 60 g/day.⁵⁸ Renal function should be evaluated early in the course of pregnancy in order to detect the presence of incipient nephropathy, lest its discovery later in pregnancy be confused with preeclampsia. This evaluation should include 24-hour urinary protein and creatinine measurements, although a protein/creatinine ratio may be satisfactory if it is normal, as well as serum creatinine. At least two studies have demonstrated an increased risk of preeclampsia in diabetic women who manifest mild to moderate microalbuminuria during the first half of pregnancy.^{69, 70} We have found it helpful to obtain a baseline serum uric acid measurement in order to compare values obtained later in pregnancy if preeclampsia is suspected. Follow-up renal function studies are obtained at least once per trimester if proteinuria was present early in pregnancy. Hypertension should be treated in order to maintain diastolic blood pressure below 100 torr. ACE inhibitors and ARBs should be avoided during pregnancy because of potential adverse fetal effects. Methyldopa, beta blockers, and hydralazine all have been used in this situation. One group has reported the successful use of recombinant human erythropoietin to treat severe anemia during pregnancy in three patients with chronic renal disease, one of whom had diabetic nephropathy.⁷¹ Because of the propensity of patients with diabetic nephropathy for IUGR it is appropriate to evaluate fetal growth on a regular basis, using ultrasound. Antepartum fetal testing is instituted earlier in pregnancy in women with vascular disease than in those without.

RETINOPATHY

Diabetic retinopathy is a form of microvascular disease, resulting from damage to retinal capillaries and arterioles; the exact mechanism is unclear at present. An association between elevated glycosylated hemoglobin and retinopathy has been demonstrated.^{72, 73} Intensive control of diabetes over the long term has been shown to slow the progression of retinopathy in both type 1 and type 2⁷⁴ diabetes. Background retinopathy, the earliest manifestation, includes microaneurysms, small vessel obstruction, cotton wool spots, intraretinal microvascular abnormalities (venous abnormalities and small retinal hemorrhages), and hard exudates.⁷⁵ Vision is generally not threatened by background retinopathy unless macular edema or ischemia supervene. Background retinopathy is almost universally present in diabetic individuals 20 years after the diagnosis is made.⁷⁵ Proliferative diabetic retinopathy is characterized by new vessel formation, or neovascularization. This growth may be a response to underlying retinal ischemia. The new vessels are poorly supported, and may leak and adhere to the vitreous. Resultant shrinkage and contraction of the vitreous may put pressure on the new vessels, causing retinal hemorrhages. These hemorrhages may lead to scarring and/or retinal detachment, and are the primary cause of visual loss in affected patients. One of the most important advances in management of diabetic retinopathy has been the discovery that laser photocoagulation therapy may prevent or forestall retinal hemorrhage and visual loss if applied at the appropriate time. Proliferative retinopathy is found in approximately 17% of individuals with type 1 diabetes 5 years after the diagnosis of diabetes, and afflicts most diabetic individuals after 15 years.⁷⁵ The DCCT results⁷⁶ demonstrated that improved metabolic control in individuals with type 1 diabetes is associated with a decreased rate of development of retinopathy, as well as nephropathy. Among individuals with preexisting nonproliferative retinopathy, intensive management was associated with a higher likelihood of progression of retinopathy during the first year. However, by 36 months of therapy retinopathy was less likely to advance in the intensive management group, and this benefit continued throughout the study period. Overall, subjects in the intensive management group manifested a 50% lower likelihood of significant progression compared to those in the conventional treatment group. This and other studies confirm that the institution of tight metabolic control may be associated with a transient worsening of diabetic retinopathy, but that long term good control has a salutary effect.

It is currently uncertain whether pregnancy exerts any independent effect on accelerating the progression of retinopathy, although the most appropriately controlled prospective study suggests such an effect independent of glycemic control.⁷⁷ Factors such as hyperglycemia and hypertension⁷⁸ appear to accelerate the course of retinopathy in pregnancy. The sudden institution of intensive metabolic control that is generally carried out in early pregnancy may be expected to have a transient adverse effect on retinopathy. This phenomenon has been demonstrated in individuals with glycohemoglobin as low as 6 standard deviations above the control mean.⁷⁹ In an observational study of 185 women with preexisting diabetes (72% type 1, 28% type 2) who underwent at least two retinal examinations in different trimesters 26% experienced progression of their retinopathy.⁸⁰ While 28% of patient with no retinopathy at baseline experienced progression, none developed sight-threatening disease or required laser therapy during pregnancy. Progression was more likely in those with type 1 than type 2 diabetes (31% vs 12%). In the DCCT women who were in the intensive control group when they became pregnant were less likely to experience progression of diabetic retinopathy than those in the conventional control group.⁸¹ Among gravidas with type 2 diabetes, duration of known diabetes is associated with increasing likelihood of progression of retinopathy; in one study the average duration of type 2 diabetes in gravidas whose retinopathy progressed was 6.7 years, compared to 3.3 years in those whose retinopathy did not progress.⁸² An excellent review of retinopathy in pregnancy was published in 2016.⁸³

The effect of diabetic retinopathy on pregnancy outcome has not been widely investigated. One small retrospective series of 20 subjects with advanced retinopathy, none of whom underwent pregnancy termination, reported a perinatal survival rate of 94%.⁸⁴

NEUROPATHY

Little information is available about the frequency with which diabetic neuropathy complicates pregnancy. However, since various forms of neuropathy are common among individuals with diabetes, and since pregnancy is not known to be protective against these problems, it is likely that they will be encountered among pregnant women.⁴⁵ Autonomic neuropathy may cause symptoms referable to the bladder or gastrointestinal tract. Gastroparesis, with gastric atony and delayed emptying, may cause early satiety, fullness, nausea, vomiting, and pain.⁴⁵ It has been described in pregnant women with diabetes, and in one case series of four affected individuals the gastric symptoms were associated with cardiovascular problems, particularly postural blood pressure changes.⁸⁵ The vomiting can be intractable and interfere with nutrition, and thus fetal growth. Symptoms may be confused with hyperemesis, but their persistence and severity in a woman with diabetes should alert the clinician. Various treatments, including metoclopromide and intravenous erythromycin,⁸⁶ have been utilized in nonpregnant individuals. Based on available case reports pregnancy outcome is often suboptimal in cases of gastroparesis.⁸⁷ Autonomic dysfunction may also cause defective glucose counterregulation, and its presence may be suggested by the absence of normal respiratory variation in heart rate as well as the presence of postural hypotension. In one controlled study autonomic dysfunction was associated with a higher frequency of pregnancy complications, but not with any specific complication.⁸⁸

Peripheral neuropathies are the most commonly encountered nerve disorder in individuals with diabetes.⁴⁵ These neuropathies are usually sensory in nature with paresthesias and later anesthesia. They most often occur in the lower extremities, and may lead to skin ulceration, deep seated infection, and in extreme cases amputation. Gabapentin and pregabalin are frequently used to treat painful peripheral neuropathy outside of pregnancy. These medications are FDA category C, since both have been associated with birth defects in animal models. A prospective cohort study of 223 gabapentin-exposed and 223 non-exposed pregnancies reported similar major malformation rates (4.1% vs 2,5%, p = 0.55) and a higher rate of premature delivery (10.5% vs 3.9%, p = 0.019) in the gabapentin exposed group.⁸⁹  When neuropathy pain is severe these drugs may be considered but it is important to counsel patients about the lack of solid information regarding fetal risk.

HYPERTENSION

Hypertension is commonly present in diabetic gravidas with nephropathy. However, even in the absence of nephropathy women with diabetes are more likely than nondiabetic individuals to have hypertensive complications during pregnancy. Hinton and Sabai⁹⁰ found that 50% of diabetic gravidas with nephropathy develop preeclampsia, and that 10–20% of gravidas with type 1 diabetes but without nephropathy do so. Cousins, in an earlier thorough literature review,⁹¹ found that chronic hypertension was present in 8% of pregnant women with preexisting diabetes, without known vascular disease, and in 2.5% of patients with gestational diabetes. Preeclampsia or hypertensive disorders of pregnancy were reported in 14% of women with preexisting diabetes who did not have vascular disease and those with gestational diabetes, but in nearly 30% of those with known vascular disease. In a 2016 systematic review and meta-analysis that evaluated the risk of pre-eclampsia associated with risk factors that could be known by 16 weeks of gestation, 11% of gravidas with pre-existing diabetes experienced pre-eclampsia which was 3.7 times the risk of those without diabetes. Women with chronic hypertension manifested a relative risk for pre-eclampsia which was 5.1 times that in those without hypertension.⁹²

URINARY TRACT INFECTIONS

Urinary tract infections appear to be generally more common among individuals with diabetes than nondiabetic patients. In Cousins' literature review⁹¹ pyelonephritis complicated approximately 3% of pregnancies in women with preexisting diabetes. Urinary tract infections are of particular significance in women with diabetes because metabolic control may be adversely impacted by such problems. Diabetic ketoacidosis as well as more subtle derangements of control are commonly brought about by infections.

HYDRAMNIOS

Hydramnios, or excessive amniotic fluid, was found to complicate approximately 18% of overt diabetic pregnancies in Cousins' literature review,⁹¹ but only 2% of gestational diabetic pregnancies. However, varying definitions of this complication, along with our inability to accurately measure amniotic fluid volume noninvasively, make its diagnosis problematic. Although it is logical that maternal hyperglycemia with fetal osmotic diuresis is responsible for hydramnios, supportive data have not been convincing, and the cause of this complication in diabetic pregnancies must be considered unknown at the present time. One study reported significantly higher mean blood glucose values among 13 gestational diabetic subjects when hydramnios was present compared to the same patients at times when amniotic fluid volumes were more normal.⁹³ Another descriptive study demonstrated a correlation between amniotic fluid volume and amniotic fluid glucose levels in diabetic pregnancies.⁹⁴ In the general population, the presence of hydramnios is associated with an increased risk for congenital malformations. While this is no doubt also true in diabetic pregnancies, in which the a priori risk for malformations is increased, most diabetic pregnancies with hydramnios are productive of structurally normal offspring. Hydramnios is also considered to be a potential inducer of preterm labor, and can cause maternal respiratory embarrassment. Nevertheless, it is distinctly unusual for diabetes‑associated hydramnios to require therapeutic amniocentesis in the absence of congenital malformations.

Perinatal mortality

Although the perinatal mortality rates associated with diabetes in pregnancy have declined considerably during the past eight decades and are now near those in the general population, it is only through continuing vigilance that such advances can be maintained. Both fetal and neonatal deaths occurred with increased frequency in diabetic pregnancies before the advent of modern management methods, and fetal deaths nationwide continue to be significantly higher among diabetic than non-diabetic pregnancies.⁹⁵ The most recent available data indicate that the relative risk of stillbirth in pregnancies complicated by type 1 diabetes (compared to the general population) is 2.9–4.3 fold, and for type 2 diabetes 2.5–4.5 fold.⁹⁶ While some studies suggest an increased perinatal mortality rate with undiagnosed and untreated gestational diabetes, the evidence is equivocal for patients whose GDM was treated during the current era. The cause of fetal death remains incompletely understood. Maternal ketoacidosis, which was at one time associated with a 90% fetal mortality rate,⁴⁸ is currently rare among appropriately treated diabetic women but still carries a fetal mortality rate of 10–30%.⁹⁵ Even in the absence of DKA there appears to be a clear association between suboptimal metabolic control and perinatal death,⁹⁷ but appropriately controlled studies have not been carried out. Animal studies⁹⁸ suggest that fetal hyperinsulinemia which may be brought about by maternal hyperglycemia may cause fetal hypoxemia and lactic acidosis, with fetal death in extreme cases. Similarly, the infusion of large amounts of glucose containing solutions to pregnant women prior to delivery has been associated with fetal acidosis.^{99, 100} It thus seems likely that maternal hyperglycemia is at least partially the cause of the increased fetal death rate among diabetic pregnancies. These observations are consistent with the "Pedersen hypothesis,"¹⁰¹ which states that maternal hyperglycemia is translated into fetal hyperglycemia, which in turn causes fetal hyperinsulinemia, which is the cause of most of the abnormal outcomes in the fetus of the diabetic mother. Perinatal mortality rates are higher in poorly controlled diabetic pregnancies, as evidenced by the study of Karlsson and Kjellmer,¹⁰² in which perinatal deaths ranged from 4% when mean third trimester blood glucose was <100 mg/dL, to 16% when glucose was maintained between 100 and 150 mg/dL, and 24% for diabetic patients whose mean glucose levels exceeded 150 mg/dL. In the aforementioned study modern technology for intervention in jeopardized pregnancies (i.e., tests of lung maturity, fetal monitoring, and antepartum testing) was not utilized. The differences in perinatal mortality rate could not be attributed to timely delivery since it was not generally possible to identify fetuses in jeopardy, and the goal of treatment was to bring the pregnancies to term. A subsequent series from our center¹⁰³ included 73 mothers with preexisting diabetes whose mean plasma glucose levels (14% higher than corresponding whole blood glucose levels) in the third trimester averaged 108 mg/dL, with 77% having mean levels under 120 mg/dL. The perinatal mortality rate was 4%, and only two of the 77 babies were delivered because of deterioration of fetoplacental function test results. Countless other series have demonstrated perinatal mortality rates of 4% or less with attempts to achieve maternal near-euglycemia. Strict metabolic control thus may be considered prevention, as opposed to intervention attributable to technologic advances allowing identification of the fetus at high risk for death in utero.

Perinatal mortality consists of both fetal and neonatal deaths. While fetal death is probably directly related to metabolic derangement in diabetic pregnancies as described above, neonatal deaths appear to be caused more indirectly. The threat of fetal death has, in the past, prompted attempts at prevention by early delivery. Thus, prematurity and its sequelae increased the neonatal death rate. In addition, infants of mothers whose diabetes was poorly controlled are more likely to develop respiratory distress syndrome at a given gestational age than are infants of normal control mothers.¹⁰⁴ Studies in which maternal metabolism was maintained at near normal levels suggest that respiratory distress does not disproportionately occur in infants of diabetic mothers.¹⁰⁵ Congenital anomalies, as described below, are another important cause of increased neonatal mortality. Approaches to maintaining perinatal mortality rates in diabetic pregnancies at levels similar to the general population are discussed later in this chapter in the section on management.

Congenital malformations

At a time when perinatal mortality ranged from 10 to 33% the fact that birth defects occurred in 7–10% of the offspring of diabetic mothers was seemingly of minor significance, and attention was directed toward reversing the unacceptably high stillbirth rate. Now that, with maternal glycemia well controlled, perinatal mortality risks are similar to those seen in the general population, it is apparent that at least half of the perinatal mortality observed, and a good deal of the morbidity, is related to congenital anomalies. Infants of diabetic mothers are two to three times more likely than infants in general to manifest all types of birth defects.^{106, 107} Cardiac, neural tube, and skeletal defects are most common, but a particular set of anomalies affecting the lower half of the body, the "caudal regression syndrome," is highly specific for diabetic pregnancy. Therefore, much effort has been focused on improving our understanding of the genesis of these malformations, and in their prevention. Remarkable success has been achieved in both of these endeavors.

In order to prevent birth defects in the offspring of diabetic women, it was first necessary to improve our understanding of their genesis. In reviewing the embryology literature Mills and colleagues¹⁰⁸ determined that the structural birth defects seen in infants of diabetic mothers have all occurred by the end of the 6th week after conception, i.e., the 8th week after the last menstrual period. This finding presented a major problem in understanding the etiology of such birth defects. Since most women are not under a physician's care during the first 6 weeks after conception, it is difficult to observe possible contributors to teratogenesis. Studies in animal models suggested that maternal hyperglycemia during the period of organogenesis is capable of inducing anomalies in the offspring.¹⁰⁹ Currently there is not total agreement as to whether hyperglycemia itself, or other factors such as ketone bodies¹¹⁰ or arachidonic acid deficiency,¹¹¹ may be of greater importance. Some investigators postulate that free oxygen radicals, resulting from enhanced mitochondrial substrate oxidation in the developing embryo, may be causative in teratogenesis.^{112, 113} One group has proposed that the embryonic yolk sac is the site most critical for teratogenesis.¹¹¹ It can only be stated that the exact mechanisms causing birth defects in diabetic pregnancies remain unknown.¹¹⁴

When the blood test for glycosylated hemoglobin became available, enabling researchers to observe retrospectively over the previous 2–3 months a marker that correlated with maternal glycemic control, an association between elevated glycosylated hemoglobin in the first trimester and congenital anomalies in the offspring was quickly recognized.^{115, 116} Such an association raised the possibility that improving maternal glucose control around the time of conception and organogenesis might lower the birth defect rate, and soon thereafter Fuhrmann et al.¹¹⁷ reported that attendance at a prepregnancy and early pregnancy program of tight metabolic control of diabetes was associated with a congenital malformation rate similar to the 2–3% in the general population, compared with a rate of 7% among diabetic pregnancies whose care commenced after the 8th week of gestation. In 1988 a multicenter observational study of metabolic control in early pregnancy¹¹⁸ found that diabetic women recruited before conception, or within 21 days after conception, had approximately half the likelihood of birth defects in their offspring compared to women enrolled into the study later. However, it was disappointing that their malformation rate (4.9%) was still approximately twice that seen in nondiabetic control pregnancies, and that there was no apparent relationship between indices of glycemic control in early pregnancy and the likelihood of malformations in the offspring. It is possible that this lack of relationship between glycemic control and malformations was due to the characteristics of the population studied: only 5% of mothers had near-euglycemia, with mean glucose values under 140 mg/dL during organogenesis, while a similar 5% had extreme hyperglycemia, with mean glucose values above 235 mg/dL. Thus there may not have been enough subjects at the extremes of glucose control to demonstrate a relationship, if one exists. A subsequent study from the Sweet Success Program in the San Francisco Bay area demonstrated a birth defect rate of 10.9% in diabetic patients beginning care after conception compared to 1.2% in those enrolling in a program of preconception intensified care.¹¹⁹ A number of other studies have demonstrated a correlation between first trimester glycohemoglobin levels and birth defects.^{120, 121, 122} However, what is not clear at this time is exactly how close to normal maternal glucose levels must be in order to achieve the lowest possible risk of anomalies. One study of a series of 215 pregnancies in women with type 1 diabetes suggested that those with hyperglycemia in the first trimester exceeding an apparent threshold (glycohemoglobin >12% or median glucose value >120 mg/dL) were at increased risk for spontaneous abortion or congenital malformations.¹²³ A 2007 review of seven studies published during the previous 11 years suggested a continuum of risk over periconceptional A1c levels with malformation rates increasing from approximately 3% at the nondiabetic mean of 5.5% through 20% or more at an A1c level of 14% in the first trimester.¹²⁴ Although this question has not yet been settled, there is general agreement that the onset of care prior to conception or soon thereafter, along with efforts to attain improved metabolic control, can lower the malformation rate among infants of diabetic mothers. Interestingly, similar effects of glycemic control on the rate of spontaneous abortion have been reported from a number of centers,^{121, 122, 125} suggesting that there may be a continuum of reproductive damage and loss associated with maternal hyperglycemia or other metabolic derangements.

Macrosomia

As described above, efforts to reduce the perinatal mortality rate in diabetic pregnancies have been relatively successful, and the ability now exists to lower the prevalence of congenital anomalies. However, other forms of morbidity continue to complicate these pregnancies. Macrosomia, or the "large for dates" baby, is usually defined in terms of a specific birth weight (i.e., 4000 or 4500 g), or as a relative weight for gestational age, gender, birth order, etc. (i.e., the 90th, 95th or 97.5th centile). The latter designation, usually designated “large-for-gestational-age", is more biologically based, since a premature baby may be macrosomic for its age but not exceed any of the usual absolute weight thresholds. Not all studies report birth weight in such a standardized manner, so that data are not all comparable.¹²⁶ Nevertheless, it is clear that however it is defined, macrosomia is considerably more prevalent in offspring of diabetic, as compared to nondiabetic, pregnancies.¹²⁷ This is true for gestational diabetes as well as preexisting diabetes. Typically macrosomic infants of diabetic mothers have increased body fat and increased length, but not increased head or brain size. Fat mass has been estimated to account for almost 50% of the variance in birth weight, although fat comprises only about 14% of birth weight.¹²⁸ Because the shoulders of infants of diabetic mothers may be abnormally broad in comparison to their head size,¹²⁹ shoulder dystocia may be a problem when they are delivered.¹³⁰ Infants of diabetic mothers have been shown to be at greater risk for shoulder dystocia than infants of nondiabetic mothers, even when corrected for birth weight.¹³¹

The cause of macrosomia in infants of diabetic mothers appears to be similar to that of other problems encountered in these pregnancies: fetal hyperinsulinemia.^{101, 127} A number of studies have demonstrated increased pancreatic insulin content in infants of diabetic mothers, as well as increased C-peptide and proinsulin in their cord blood.¹³² The landmark studies of Susa et al.¹³³ demonstrated that fetal hyperinsulinemia, even in the absence of hyperglycemia, can cause macrosomia in the rhesus monkey. Nevertheless, a number of unanswered questions remain. Glucose is not the only important secretogogue for fetal pancreatic insulin production and release, and Freinkel¹³⁴ proposed the "modified Pedersen hypothesis," which invokes other substrates such as amino acids as being capable of stimulating the fetal pancreas to produce and release excessive insulin. It is also not certain that insulin itself is the cause of macrosomia, or whether some other insulin-like factor might act at the same or similar cellular receptors to stimulate fetal growth in utero in such patients.

The prevention of macrosomia has received much attention, since this form of morbidity is easily measurable and has definite adverse consequences including a greater likelihood of operative delivery and birth trauma. Given the "Pedersen hypothesis,"¹⁰¹ it would seem that improving maternal metabolic control should help to avoid fetal macrosomia. While this has generally been true, there is still a significant proportion of macrosomic newborns when maternal glucose metabolism is maintained within generally accepted goals, even among gestational diabetic pregnancies.¹³⁵ A number of possible explanations exist, and evidence exists to support each of them.¹³⁶ For example, our current understanding of "good control" may not be "good" enough. We attempt to maintain circulating glucose levels below approximately 100 mg/dL fasting, and 120 mg/dL 2 hours postprandial, based on studies of normal pregnant women.¹³⁷ The goals utilized represent two standard deviations above the mean, but most of us do not at exist at two standard deviations above the mean; such a value defines the upper limit of normal, not the average. Thus, our therapeutic goals may be too high. Another possibility is that we are not sampling glucose levels often enough. While we may measure glycemia 4–6 times per day, the pancreas samples ambient glucose levels constantly, and is constantly readjusting insulin release to maintain a steady glycemic level.¹³⁸ When the results of continuous glucose monitoring were compared to the more traditional 6–8 finger stick blood sugars per day in 57 women with gestational diabetes, the intermittent testing missed both hypoglycemic and hyperglycemic episodes throughout the day and night.¹³⁹ Finally, it is possible that glucose does not tell the whole story. Glucose is easy to measure, but other substrates may be important as well. For example, Persson et al.¹⁴⁰ have found a correlation between maternal branched chain amino acid levels and fetal hyperinsulinemia, and Freinkel's group¹⁴¹ demonstrated a relationship between maternal levels of alanine, serine, and leucine and birthweight. The HAPO study²¹ demonstrated a direct and significant relationship between maternal glucose tolerance test results at 24–28 weeks' gestation and neonatal macrosomia and body fat, even in the normal range of glucose values. This supports a basic biologic relationship between maternal glucose and fetal growth. Identification and treatment of gestational diabetes appears to decrease the risk of shoulder dystocia.^{142, 143} 

Neonatal hypoglycemia

Given the presence of fetal hyperinsulinemia in the offspring of diabetic mothers it is easy to understand how neonatal hypoglycemia can occur once the maternal glucose contribution to the fetal circulation has been interrupted by the process of delivery, when the umbilical cord circulation is discontinued. This complication is usually defined as a plasma glucose concentration <35 mg/dL in a term infant, or <25 mg/dL in a preterm infant¹⁴⁴ and its likelihood in an infant of a diabetic mother has been positively associated with maternal hyperglycemia at delivery¹⁴⁵ and the rapid infusion of glucose-containing intravenous fluids during labor.¹⁴⁶ Hypoglycemia is most likely to occur during the first 60–90 minutes of life, and is often asymptomatic. Infants who manifest symptoms usually are irritable, jittery, may have apneic spells, tachypnea, hypotonia, and at the extreme, convulsions. If hypoglycemia does not occur early in the neonatal period, it is unlikely to show up later. Early institution of oral feeding may be helpful in preventing hypoglycemia. Hypoglycemia which is promptly treated generally is not associated with adverse sequelae.¹⁴⁴

Neonatal respiratory problems

Respiratory distress syndrome (RDS), as well as other forms of neonatal respiratory difficulty, occurs with increased frequency in infants of diabetic mothers.^{104, 144} Because maturation of the surfactant system may be delayed in some diabetic pregnancies, it used to be common to use a test of stable lung maturity, such as the presence of phosphatidylglycerol (PG) in amniotic fluid, prior to elective delivery of a diabetic pregnancy. A number of investigators have concluded that with the maintenance of good metabolic control during pregnancy, RDS may be no more common in infants of diabetic mothers than in the general population.^{147, 148, 149, 150}

Other neonatal problems

A number of other problems are reported to occur with increased frequency in infants of diabetic mothers.¹⁴⁴ These include polycythemia and hyperviscosity, hyperbilirubinemia, and hypocalcemia, and are all likely explainable on the basis of hyperinsulinemia. Studies of fetal blood obtained by cordocentesis at 36–40 weeks' gestation demonstrated significantly higher hematocrits and lower platelet counts, compared to normative values, in fetuses of diabetic mothers.¹⁵¹ Umbilical venous pH was also significantly lower than in controls, although still within the normal range. Transient cardiac dysfunction, presumably due to increased thickness of the intraventricular septum, has been reported in neonates of diabetic mothers even when metabolic control has been reportedly good during pregnancy.^{152, 153, 154, 155, 156}

Growth and development

The intrauterine environment may have lasting effects on subsequent development in the offspring. Childhood and adolescent obesity have been demonstrated in offspring of diabetic Pima Indian mothers¹⁵⁷ as well as in macrosomic offspring of diabetic mothers followed in the Collaborative Perinatal Project.¹⁵⁸ Similarly the group at Northwestern University reported that obesity was more common among offspring of diabetic mothers by the age of 6, and that a relationship existed between the degree of maternal hyperglycemia during pregnancy and subsequent childhood obesity.¹⁵⁹ Follow-up studies performed at 4–7 years of age demonstrated that large for gestational age (LGA) infants of gestational diabetic mothers were heavier, with increased skinfold thickness and higher body mass index (BMI) than average for gestational age (AGA) infants of gestational diabetic mothers or normal controls.¹⁶⁰ These findings are consistent with studies showing increased adipocyte numbers in the offspring of diabetic mothers.¹⁶¹ It appears likely that the current increasing prevalence of obesity and other features of the metabolic syndrome in children and young adults may be at least partially related to disorders of glucose metabolism during pregnancy.^{162, 163} Investigators at Northwestern University have also demonstrated a relationship between maternal hyperglycemia, during the second and third trimesters, and worsening performance of the offspring on the neonatal Brazelton behavioral assessment scale.¹⁶⁴ An inverse correlation also was found between maternal ketone body levels in the second half of pregnancy and intellectual development of the offspring up to 5 years of age,^{165, 166} a relationship which could not be explained by obstetric and neonatal complications or socioeconomic status.¹⁶⁶ In animal studies, exposure to a hyperglycemic intrauterine environment has been demonstrated to cause gestational diabetes in the female offspring for at least two generations.^{167, 168} Studies of the Pima Indian population have demonstrated a correlation between maternal hyperglycemia during pregnancy and subsequent diabetes, and gestational diabetes, in the offspring, even when corrected for obesity, genetic predisposition to diabetes, and age of the offspring.¹⁶⁹ A similar association was found between maternal hyperglycemia during pregnancy and impaired glucose tolerance in teenage offspring in the Northwestern University study.¹⁷⁰ Thus, the concept of "fuel-mediated teratogenesis"¹⁷¹ may be applied not only to structural birth defects caused during the period of organogenesis, but may also describe the way in which maternal metabolic aberrations may leave a legacy which continues to be played out into adulthood, and possibly into the next generation.¹⁷² Indeed, an increasing prevalence of gestational diabetes has been reported in a large health maintenance organization population between 1994 and 2002.¹⁷³

 

Preexisting diabetes

ORAL ANTIDIABETIC AGENTS

Oral antidiabetic agents are not likely to maintain euglycemia in pregnancies complicated by preexisting diabetes, and a description of their role will be deferred to the section on gestational diabetes. However, one aspect of their use merits the following discussion. During the current epidemic of obesity and type 2 diabetes it is not uncommon to encounter women who have conceived while taking metformin, a biguanide drug which is an insulin sensitizer. Metformin is commonly used to treat insulin resistance syndrome and polycystic ovary syndrome (PCOS) and is capable of facilitating ovulation, and thus pregnancy, in affected individuals. Some case series, using historical controls, have suggested that continuing metformin after conception is associated with a lower spontaneous abortion rate.¹⁷⁴ Patients coming for their initial obstetrical visit may have been advised by their primary caregivers that metformin treatment should continue throughout the pregnancy. Because metformin crosses the placenta¹⁷⁵ and fetal levels may significantly exceed maternal levels,¹⁷⁶ this practice is of some concern. Two randomized placebo controlled double blind trials,^{177, 178} in which patients conceiving on metformin discontinued the drug immediately when a sensitive pregnancy test was positive, demonstrated that the lower spontaneous abortion rate was evident even with this early discontinuation. A larger randomized controlled trial did not report a decreased spontaneous abortion rate with metformin compared to clomiphene.¹⁷⁹ There is no evidence to support continuing metformin once pregnancy has been established.

MATERNAL GLUCOSE CONTROL

As outlined above, maternal metabolic control to near euglycemic levels, at least in the latter half of pregnancy, can lower the perinatal mortality rate and may prevent various manifestations of perinatal morbidity, by preventing or lessening fetal overproduction of insulin. Prior to pregnancy and during early pregnancy improved metabolic control has the potential to prevent congenital malformations. Indeed, a program of preconception counseling and care has been demonstrated to be effective in reducing the likelihood of birth defects in the offspring^{120, 180} and also has been shown to be cost effective.^{181, 182} Unfortunately, only about one third of individuals with preexisting diabetes receive preconception care according to a multicenter study conducted in 1995.¹⁸³ It is critical that obstetricians and other primary care providers, such as internists, family physicians, and pediatricians, provide effective counseling to all diabetic women in or approaching the reproductive age. The ADA recommends that women with diabetes be encouraged to achieve HbA1c levels <6.5% prior to conception in order to lower the likelihood of congenital malformations in their offspring.¹⁸⁴ The importance of family planning, the evaluation of co-morbid conditions, and the importance of good metabolic control prior to conception are all part of preconception care.¹⁸⁵

The question is still unsettled as to precisely how close to normal maternal circulating glucose values must be in order to maximize the likelihood of a favorable outcome. Various opinions exist, and the suggestions outlined below must be considered tentative, as they are based on the “opinions of experts”¹⁸⁵ as well as the practices at our institution and are neither proven nor universally agreed upon. Nevertheless, virtually all investigators in the field agree that the best outcomes are associated with "good" control.

The ideal approach to management of diabetic pregnancy is the "team approach," in which the high risk obstetrician and the diabetologist collaborate as part of a multidisciplinary team which also includes specialized nurses, dieticians, social workers, and other health care providers. However, each component of the team is not available in every center. In addition, even if the diabetes is being managed by a diabetologist, it is important for the obstetrician to understand what is being done, and why. Therefore, the foregoing discussion of medical management is included in this obstetrics chapter.

Modern management of diabetes in pregnancy relies on the use of self glucose monitoring. Test strips, impregnated with glucose oxidase and an indicator, are covered with a drop of blood and the test strip is inserted into a reflectance meter which "reads" the blood glucose. These machines, originally quite complicated to use and expensive to purchase, have come down in purchase price and are all relatively simple to work with. They are battery powered and portable. Currently, many machines read the glucose level automatically from the strip which is inserted into the meter before the drop of blood is applied. Various schemes exist as to the timing and frequency of self blood glucose testing. Nonpregnant patients are usually asked to measure glucose just prior to meals and snacks, three to four times per day.  During pregnancy glucose levels are checked fasting and either 1 or 2 hours after each meal, or else the time after meals of glucose measurement is determined by initial every 15 minute testing, then choosing the interval of peak glucose levels. Relatively few comparative data are available. Diabetologists tend to rely more heavily on fasting and preprandial than postprandial glucose values in managing nonpregnant adults with diabetes, but existing studies suggest that in pregnancy postprandial glucose values are better predictors of fetal macrosomia than are fasting values.^{186, 187, 188} When 1 hour postprandial values exceeded 130 mg/dL¹⁸⁸ or average glucose values exceeded 130 mg/dL,¹⁸⁹ macrosomia was significantly more likely. One randomized trial¹⁹⁰ demonstrated that using postprandial glucose measurements to inform insulin dosage was more effective than preprandial testing at preventing adverse outcomes in insulin-requiring gestational diabetes. In our own center we ask the patient to measure glucose at a minimum of four times per day. This includes fasting and 1 or 2 hours after each meal. Additional times are chosen if they seem clinically necessary (i.e., 2 AM measurement if there is a question of morning rebound hyperglycemia [i.e., the "Somogyi effect"] or early morning insulin depletion [i.e., the "dawn phenomenon"]). An outline of our goals is depicted in Table 7.

 

Table 7. Daily self-monitoring of glucose in diabetic pregnancy

 

There are also a number of different approaches to insulin administration in diabetic pregnancy.⁴⁷ It is rare for satisfactory metabolic control to be attained in a pregnant woman with preexisting diabetes by a once daily insulin regimen. We have found that a minimum of two injections of mixtures of short- and intermediate-acting insulins are generally necessary, and that many individuals require three or four injections daily. Weight based insulin regimens are commonly employed in pregnancy.⁴⁷ In our unit, in order to begin insulin therapy in a previously uncontrolled patient, we use a formula that was based on the normal insulin release pattern of nondiabetic pregnant women.¹⁹¹ Insulin is injected before breakfast and before dinner. The morning total dose is twice the evening total dose. The ratio of intermediate- to short-acting insulin in the morning is 2:1; in the evening it is 1:1. For example, a patient may receive 40 units of NPH insulin and 20 units of regular insulin (or a rapid-acting analog), mixed in the same syringe, prior to breakfast each day. The total morning dose is thus 60 units. Prior to dinner she would receive half that dose, or 30 units, and it would be composed of a 1:1 ratio, or 15 units each of NPH and regular insulins. This formula should be considered a "starting point." Subsequent adjustments are based upon self glucose monitoring results at particular times of day. For example, if the 2-hour post-breakfast glucose exceeds our goals, we would increase the dose of short-acting insulin given the following morning. If the post-lunch glucose is high, an adjustment in the morning dose of intermediate-acting insulin would be made. The fasting glucose reflects the pre-dinner intermediate dose, etc. Few patients remain on the formula for very long; individual dose adjustments are the rule. A few observations may be useful to the reader:

1. “Regular” insulin was the main short-acting insulin used during pregnancy until insulin analogs such as insulin lispro and insulin aspart became available. Each of these insulins has a single amino acid substitution compared to natural human insulin, which hastens absorption after a subcutaneous injection. These rapid-acting insulin analogs can thus be given immediately before a meal, enhancing flexibility for the patient. Insulin lispro has been shown to demonstrate minimal passage across the placenta at typical therapeutic concentrations,¹⁹² and insulin aspart, while less widely evaluated, appears to have similar properties.¹⁹³
2. There are at least two long-acting insulin analogs on the market in the US, insulin glargine and insulin detemir. Both have 24 hour duration of action, and they appear to demonstrate no absorption peak. These insulins are becoming increasingly popular in the management of non-pregnant individuals with diabetes, since they can imitate the basal insulin secretion typical of the pancreas as well as of continuous subcutaneous insulin infusion pumps. Both insulin glargine^{194, 195} and insulin detemir¹⁹⁶ appear not to cross the placenta and therefore should be safe for use in pregnancy. Insulin detemir has been demonstrated in a randomized controlled trial to be similar in efficacy to NPH insulin when used for pregnant women with type 1 diabetes, and may be more effective in normalizing fasting glucose values.¹⁹⁷  Because these long-acting insulin analogs are considerably more expensive than NPH insulin, we use them in situations when NPH insulin is not optimizing glucose control, or in situations where the patient is already taking one of these analogs prior to pregnancy and is in good metabolic control. They are not usually necessary for managing gestational diabetes.  
   
3. If a patient manifests consistently high fasting glucose levels despite increases in her pre-dinner intermediate insulin dose, and particularly if her 2-hour post-dinner glucose value is normal or low, her intermediate insulin may be peaking too early in the morning, and wearing off by the time she arises (the "dawn phenomenon"). Have her measure her blood glucose at 2 AM. If it is normal or low, move the pre-dinner intermediate acting insulin dose to bedtime in order to get a longer effect. If the 2 AM glucose is quite high, it is reasonable to simply increase the pre-dinner intermediate acting insulin.
4. Some patients will manifest very high glucose levels after breakfast, then become relatively hypoglycemic prior to lunch. This is often due to morning short-acting (regular) insulin absorption lagging behind the absorption of the food eaten at breakfast. A useful solution is to increase the time interval between the morning insulin injection and breakfast. We have occasionally had to give the morning insulin as long as an hour before breakfast is consumed. This problem is much less likely to be encountered when rapid-acting insulin analogs such as insulin lispro or insulin aspart are utilized.

One approach to insulin administration is the use of the continuous, subcutaneous insulin infusion pump (the "insulin pump") in order to more closely approximate physiologic insulin release. With the pump, a continuous basal insulin infusion is provided, and bolus doses are administered prior to each meal. An advantage of the pump is greater flexibility with regard to meal timing. Furthermore multiple basal insulin infusion rates can be tailored to the particular patient’s needs. Disadvantages include the expense, and the fact that only rapid-acting insulin is infused, and interruption of infusion can more quickly lead to ketoacidosis. A prospective randomized trial of pump versus intensive conventional insulin therapy in pregnant women with diabetes failed to reveal any difference in metabolic control or perinatal outcomes,¹⁹⁸ and so the pump may be viewed as "elective" for most pregnant patients. The highly motivated and intelligent subject with diabetes which is very difficult to control conventionally may indeed benefit from a trial of pump therapy. Patients already using a pump should continue to do so throughout pregnancy. A number of different pumps with various features are on the market.

Another reasonable approach for the patient with preexisting diabetes, particularly type 1 diabetes, may be a combination of long-acting insulin given at bedtime (”basal insulin”) and rapid-acting insulin before each meal. This concept is somewhat similar to the use of the insulin pump, although it does not allow for changes in the basal insulin infusion rate at various times of day.  It is clearly much less expensive. It has been our experience that the duration of action of long-acting insulins may be shortened during pregnancy in some patients, necessitating twice daily dosing.

While relatively strict goals for metabolic control in the second half of pregnancy are outlined in Table 7, it is likely that somewhat more liberal goals can be used in the periconceptional period. As pointed out earlier, there is not universal agreement as to exactly how close to normal maternal glucose levels need to be during the period of organogenesis in order to minimize the risk of birth defects. A Standard of Care from the ADA¹⁸⁴ states: “Provide preconception counselling that addresses the importance of glycemic control as close to normal as is safely possible, ideally A1c <6.5% (48 mmol/mol) to reduce the risk of congenital anomalies.” Once pregnancy is established, the ADA recommends an optimal target for A1C at <6% (42 mmol/mol) if this can be achieved without significant hypoglycemia, while higher targets (<6.5% or even <7%) may be necessary to prevent hypoglycemia in some patients. In our center, if we are fortunate enough to make initial contact with a diabetic woman prior to conception we attempt to maintain glucose levels under 150 mg/dL, with A1c levels in the near-normal range (depending on the method). Once conception has been established we aim for the glucose levels outlined in Table 7. When a patient is first seen during the first 8 weeks of pregnancy, and is poorly controlled, we will often hospitalize her in order to improve control as rapidly as possible. This is not the ideal situation, as rapid normalization of metabolic control may be associated with deterioration of retinopathy, at least transiently. Therefore, ophthalmologic consultation is appropriate for these patients, as for all pregnant women with preexisting diabetes.

Insulin requirements can be expected to increase markedly as pregnancy progresses¹⁹⁹ and this should not be cause for alarm. "Brittleness," the marked instability of diabetic control seen in some individuals who are apparently following their regimen assiduously, appears to decrease during the second half of pregnancy,⁵⁶ a phenomenon which may be quite helpful in managing these difficult cases. Although there is a persistent "clinical pearl" which states that falling insulin requirement is an ominous sign for perinatal well-being in diabetic pregnancy, proof of this assertion is lacking. When a significant fall in insulin requirement occurs in the third trimester, it is reasonable to step up the pace of fetal surveillance, but there is little or no evidence that early delivery should be based on this finding in the absence of signs of fetal compromise.

In addition to daily self glucose monitoring, frequent contact with a member of the health care team is important in diabetes management. In our center a Diabetes Nurse Educator is in telephone contact with each patient initially on a daily basis, with longer intervals instituted later as deemed appropriate. Such a practice is invaluable not only in helping maintain good metabolic control, but also in reassuring the patient that help and counsel are always readily available, and preventing such problems as mild upper respiratory infections from developing into diabetic ketoacidosis. Indeed, one of the most common misconceptions among our patients is that insulin dosage must be reduced if food intake is decreased due to illness, so as to avoid hypoglycemia. Of course, the opposite is true. Illness often causes insulin resistance, and hepatic glucose production increases in the face of illness despite decreased caloric intake. Therefore insulin dosage should be increased or should remain the same, depending on circulating glucose levels and the presence or absence of ketonuria. 

PRENATAL SCREENING AND DIAGNOSIS

Because congenital anomalies occur with increased frequency among diabetic pregnancies, it is important that available methods of prenatal diagnosis be applied, where appropriate, in accordance with the patient’s wishes. Serum alpha-fetoprotein (AFP) testing, in combination with measurement of human chorionic gonadotropin (hCG), inhibin, and estriol, is commonly offered to pregnant women throughout the United States in order to detect open neural tube defects, ventral wall defects, and a number of other potential problems including chromosomal trisomies. Currently a variety of combinations and permutations of the above serum screening tests plus nuchal translucency measurement and serum testing of pregnancy associated plasma protein A (PAPP-A) and free beta subunit of hCG in the first trimester are offered as an integrated test. Women with diabetes should likewise be offered these tests, although diabetes does not predispose to chromosomal trisomies. It is important to note that maternal serum AFP values have been demonstrated to be lower in diabetic pregnancies than in other pregnancies,^{200, 201, 202} while diabetic pregnancies have been demonstrated to be at increased risk for neural tube defects.²⁰³ In one study,²⁰¹ maternal serum unconjugated estriol levels were also significantly lower than controls. Evidence exists that the lowering of AFP values may be related to poor metabolic control,^{204, 205} suggesting that the patients whose glucose levels place their babies at the greatest risk for congenital anomalies are also the ones most likely to have lower AFP levels, which might thus be misleading. First trimester testing of nuchal translucency and serum markers does not appear to be impacted by maternal diabetes.²⁰⁶ There are a number of serum tests available to measure cell-free fetal DNA in the mother’s circulation, in order to provide a risk assessment for trisomy 21, 13 and 18 along with a number of other chromosome abnormalities. Since maternal diabetes does not raise the risk of fetal trisomy these tests are not specifically inidicated for diabetic pregnancy, although they may be offered on the basis of maternal age or other risk factors (cell-free DNA screening for fetal aneuploidy).²⁰⁷

Structural anomalies, known to occur with increased frequency in infants of diabetic mothers, often may be diagnosed prenatally with specialized ultrasound examinations.²⁰⁸ Virtually all types of anomalies occur with increased frequency in these pregnancies. While it might not be necessary to perform specialized ultrasound examinations in diabetic pregnancies which were in good control at the time of organogenesis, a specific cutoff level cannot be supplied. In one series of diabetic pregnancies, the “critical threshold” of glycohemoglobin for cardiac defects was the upper limit of normal (6.1% in the laboratory reporting these results).²⁰⁹ The authors of this report recommended fetal echocardiography for every diabetic pregnancy unless glycohemoglobin levels were within the normal range. Another report found that congenital heart defects were increased even in diabetic pregnancies where the A1c was only slightly above normal.²¹⁰ Ultrasound is also used to monitor fetal growth and amniotic fluid volume at intervals throughout the third trimester. Umbilical artery blood flow velocity measured by Doppler velocimetry has been found to be similar in diabetic and nondiabetic pregnancies.²¹¹ Such measurements are not a standard method of fetal evaluation at the present time, and their value is uncertain except in the case of intrauterine growth restriction.

Other screening tests that are helpful include renal function testing in early pregnancy, including serum creatinine, uric acid, 24-hour urinary protein, and 24-hour urinary creatinine clearance determinations. Since American College of Obstetricians and Gynecologists (ACOG)’s recommendation that a spot urine protein/creatinine (P/C) ratio of 0.3 or greater may be used to diagnose pre-eclampsia, a P/C ratio of 0.0 or 0.1 in the first half of pregnancy may be sufficient to rule out significant preexisting proteinuria (for later comparison in cases where preeclampsia is suspected). These help to evaluate early nephropathy, and serve as a baseline for later comparison, since diabetic individuals are more likely to develop hypertensive disorders of pregnancy. Ophthalmologic examination should be carried out by an ophthalmologist, and the pupils should be dilated for this exam. Urine cultures are obtained in order to diagnose asymptomatic bacteriuria before pyelonephritis can develop. Electrocardiography should be obtained in diabetic women with vascular disease or longstanding diabetes. The ADA has developed a set of evidence based recommendations for evaluating women with preexisting diabetes prior to pregnancy.²¹²

TIMING OF DELIVERY

Although at one time maternal diabetes meant automatic early delivery because of the high risk of fetal death near term, the advances already described have made it possible for most diabetic pregnant women managed with contemporary approaches to deliver at or near term. The achievement of near-euglycemia in diabetic pregnancy is the major preventive measure that has allowed this dramatic change in management.

 The recommendations of the ACOG²¹³ are as follows:

 Type 1 or type 2 diabetes:

            Well-controlled – late preterm or early term delivery not indicated

            Poorly controlled – 34–37 weeks 6 days (individualize, based on severity and other factors)

            With vascular complications – 34–37 weeks 6 days (individualize)

Gestational diabetes

            Well-controlled on diet, insulin or oral agents – late preterm or early term delivery not indicated

            Poorly controlled – 34–37 weeks 6 days (individualize, based on severity and other factors)

FETAL–PLACENTAL FUNCTION TESTING

A number of different techniques for assessing fetal well-being in diabetic pregnancy are available. These include fetal movement determinations ("kick counting"), nonstress testing, contraction stress testing, and use of the biophysical profile. Most current approaches utilize the modified biophysical profile (nonstress test plus amniotic fluid index) as the primary testing modality. The scheduling of twice weekly testing in diabetic pregnancies is widely utilized.²¹⁴ It should be pointed out that no antepartum test is predictive of fetal compromise that may result from an acute event, such as maternal diabetic ketoacidosis or umbilical cord accident. Poor metabolic control, even in the absence of ketoacidosis, has been associated with pathologic antepartum and intrapartum fetal heart rate tracings.²¹⁵ The potential effects of maternal hypoglycemia on fetal heart rate tracings are controversial. Increased fetal activity has been noted when blood glucose levels below 60 mg/dL were induced in one study,²¹⁶ and in another report no decrease in reactivity was present when gravidas with diabetes underwent insulin clamp studies to lower their glucose levels to the range of 40 mg/dL.²¹⁷ In our center the gestational age when antepartum testing is to be initiated varies according to the individual patient's circumstances. There is no point in initiating antepartum testing at a point in pregnancy when delivery is not a viable alternative, i.e., before 23–24 weeks. Even after this gestational age, patients with a low likelihood of fetal compromise, if tested positive, are likely to have a false positive rather than a true positive result, and delivery may be more likely to harm than benefit the pregnancy. Therefore patients with very high likelihood of fetal compromise, such as those with evident IUGR and vascular disease, may be tested as early as 23–24 weeks.^{218, 219} The majority of diabetic pregnant women begin testing at 28–34 weeks, depending upon the adequacy of metabolic control and the presence or absence of vascular complications.

MODE OF DELIVERY

Maternal diabetes is not an indication for cesarean section, but its complications may be. Fetal macrosomia occurs with increased frequency in diabetic pregnancies, as described earlier in this chapter. Presumably because the fetal head size is not increased, but the trunk is large,¹²⁹ shoulder dystocia may be a significant problem in such cases. If shoulder dystocia could be anticipated, it would be best to deliver the fetus by cesarean section. However, current approaches to estimating fetal weight are imprecise at best, and no prospectively tested approach to the prenatal diagnosis of macrosomia, or potential shoulder dystocia, is clearly effective. Therefore a compromise must be reached between an inappropriately aggressive approach toward cesarean section and an increased likelihood of shoulder dystocia with attempts at vaginal delivery. In our center diabetic mothers with a fetus estimated to weigh >4500 g are offered delivery by cesarean section without a trial of labor. When the estimated fetal weight is between 4000 and 4499 g, other factors are taken into consideration, including the mother's past obstetric history, clinical assessment of pelvic architecture, and the progress of labor. Forceps or vacuum deliveries, other than true outlet applications, are generally inadvisable when fetal macrosomia is believed to be present and the second stage of labor is prolonged.²²⁰ Conway and Langer²²¹ reported that a policy of elective cesarean section when the estimated fetal weight in a diabetic  pregnancy is 4250 g or more was associated with a reduction of shoulder dystocia of approximately 50% compared to data from an epoch when this policy was not in place. The cesarean section rate increased by 16% with the above policy. Rouse et al.²²² performed a decision analysis and predicted that if all diabetic gravidas carrying a fetus with an estimated fetal weight of 4500 g or more underwent cesarean section without labor, it would take 443 cesarean sections to prevent one permanent brachial plexus palsy, at a cost of $930,000 (1998 dollars). Using an estimated fetal weight of 4000 g would yield 489 cesareans to prevent one permanent brachial plexus palsy. A similar analysis for nondiabetic pregnancies predicted the need for approximately 10-fold more cesarean sections for each brachial plexus palsy prevented. A previous history of shoulder dystocia is also a relative contraindication to vaginal delivery unless the circumstances are clearly favorable, i.e., a much smaller fetus. In the absence of macrosomia the decision for cesarean section should be based on the usual obstetric indications. While cesarean section is more likely to be necessary when labor is induced in the presence of an unfavorable cervix, there is no compelling reason to proceed with operative delivery without an adequate trial of labor in a diabetic patient, just as in the nondiabetic pregnancy. Similarly, diabetes is not a contraindication to attempted vaginal birth after previous cesarean section.

METABOLIC MANAGEMENT FOR LABOR AND DELIVERY

During labor, attention should be paid to the maintenance of maternal euglycemia in order to lessen the likelihood of neonatal hypoglycemia, and in extreme cases to avoid ketoacidosis. Studies utilizing animal models suggest that maternal hyperglycemia may predispose to fetal lactic acidemia and hypoxemia.^{223, 224} The combination of hypoxia and hyperglycemia in primates has been associated with central nervous system damage.²²⁵ Studies of pregnant women have linked the infusion of large amounts of glucose-containing solutions to fetal acidosis.^{99, 100} The need for meticulous control of glucose metabolism during labor has led to the application of constant intravenous glucose and insulin infusions in this setting. Glucose is needed in order to avoid starvation ketosis; studies utilizing an artificial pancreas^{226, 227} during labor have demonstrated that many women with type 1 diabetes require no insulin during the first stage, despite glucose infusion rates of 6–10 g/hour. Monitoring of circulating glucose levels throughout labor is thus critical, so that insulin can be supplemented if hyperglycemia is present.

When labor is to be induced, it is reasonable to begin early in the morning, after the usual insulin and snack the night before. If a patient's diabetes has been well-controlled, it can be anticipated that her fasting glucose levels will be within the target range. An intravenous line is established, and we infuse 5% dextrose in half normal saline with an infusion pump at a constant rate of 100–125 mL/h in order to provide for basic caloric requirements. Blood glucose is measured every 1–2 hours. The target for glucose is 70–120 mg/dL. If this level is exceeded, an insulin infusion is begun, usually at a rate of 1–1.25 units per hour. Adjustments in the insulin infusion rate can be made if glucose falls below the target range, or exceeds the target range. Patients entering the hospital in spontaneous labor may require adjustments in this approach, depending upon when their last insulin injection occurred. If induction is carried out for more than 1 day, and the patient is to be allowed to rest for the night, care should be taken to avoid disturbed glucose metabolism. One useful approach is to discontinue induction just prior to dinner time, and give the patient her usual subcutaneous pre-dinner (and bedtime) insulin and her usual evening meal and snack. The procedure is begun the next morning as before. If the evening insulin dose is omitted, it is possible for ketoacidosis to ensue during the night, so close attention is warranted.

Patients admitted for elective cesarean section are best scheduled in the early morning as for induction. The usual meal, snack, and insulin are administered the night before so that the morning blood glucose will be within the target range. An intravenous line is initiated and normal saline or some other non-glucose containing solution is infused. Once delivery has been accomplished, glucose can be infused in order to maintain the circulating glucose value within a reasonable range.

During the immediate postpartum period insulin requirements often fall precipitously to levels below the prepregnancy dose. This is particularly true in women undergoing cesarean section who are not allowed to eat during the first day or two. Fortunately, the kind of meticulous metabolic control which is the cornerstone of management during pregnancy is no longer necessary after delivery, and glucose levels may be allowed to range up to 150 or so without major short term risk. The subcutaneous insulin dose can then be slowly increased as the patient's dietary restrictions are lifted, and "fine tuning" of control can be accomplished at home after discharge. 

Gestational diabetes

There has been considerable controversy regarding the significance, or lack thereof, of gestational diabetes. In 2005 the Australian Carbohydrate Intolerance Study (ACHOIS)²²⁸ was published. This was a randomized trial comparing two approaches to gestational diabetes. The intervention group underwent universal screening for GDM and those who were diagnosed were managed with diet and insulin as needed. The routine care group were not screened for GDM but caregivers were allowed to test if they felt this was indicated. The intervention group had significantly fewer serious perinatal complications (death, shoulder dystocia, fracture, nerve palsy) and significantly fewer macrosomic babies. In 2009 a similar randomized trial of identification and treatment of mild gestational diabetes, carried out in the US by the Eunice Kennedy Shriver Maternal-Fetal Medicine Units Network of the NICHD, reported similar results.²²⁹ These findings support the benefit to society of screening for and treating gestational diabetes. 

The principles guiding management of the pregnancy complicated by gestational diabetes are the same as for preexisting diabetes: maintenance of relative euglycemia in order to prevent perinatal mortality and morbidity. Once diagnosed with gestational diabetes, patients are counseled by a dietician for a diabetic pregnancy diet.²³⁰ Daily self monitoring of blood glucose is generally prescribed, although patients whose circulating glucose levels are within the target range while on diet alone may monitor less often, such as every other day.²³¹ An approach we have found to be useful is to ask patients with newly diagnosed GDM to perform daily self glucose monitoring, fasting and after each meal for the first week while they are following the prescribed diet. If all or most values are within targets the first week we offer the option of every other day testing. If one third or more of values at any one time of testing are above targets, our dietician assesses whether further dietary refinements are likely to normalize glucose values, or whether insulin should be initiated. In our center fasting, 2-hours post-breakfast, 2-hours post-lunch, and 2-hours post-dinner glucose values are monitored. If the fasting value exceeds 95 mg/dL or any of the other two values exceed 120 mg/dL a third of the time while the patient is following the prescribed diet we recommend additional therapy. Some other centers recommend testing 1 hour after meals rather than 2 hours, with targets of <140 mg/dL or <130 mg/dL. Either approach is reasonable, since the only randomized trial comparing 1-hour and 2-hour post meal testing failed to show any difference in pregnancy outcomes.²³² Insulin therapy initiated in response to elevated maternal glucose levels, as described above, is aimed at reducing the perinatal mortality and morbidity risk; in our experience approximately 28% of women with gestational diabetes require such treatment.²³³ While it is generally true that the more abnormal the glucose tolerance test results, the more likely is the patient to require insulin, 26% of individuals whose glucose tolerance test results were in the lower third of GDMs manifested hyperglycemia to the extent that insulin was needed.²³³ Therefore glucose tolerance test results should not be the sole determinant of the mode of therapy to be used.

Buchanan and colleagues²³⁴ studied gestational diabetic pregnant women whose fasting serum glucose was below 105 mg/dL and who underwent ultrasound evaluation at 29–33 weeks' gestation. The 98 individuals whose fetal abdominal circumference measurement met or exceeded the 75th centile for gestational age were randomized to diet only (N = 29) or diet plus twice daily insulin (N = 30). The prevalence of large for gestational age babies was reduced from 45% in the diet group to 13% in the insulin treated group. The authors concluded that ultrasound in the early third trimester can identify candidates who would benefit from insulin treatment. This same group performed a second randomized trial²³⁵ in patients with fasting glucose values of 105–120 mg/dL. One group received insulin and the other received insulin only if the abdominal circumference (measured monthly) was above the 79th centile or if any fasting glucose value exceeded 120 mg/dL while on diet. Outcomes in the two groups were similar and 38% of the patients in the second group avoided the use of insulin.

While insulin has been the standard treatment for gestational diabetes when diet is unsuccessful at normalizing circulating glucose levels, and oral antidiabetic agents have the disadvantage of the potential for transplacental passage and possible adverse fetal effects, there has recently been a resurgence of interest in this form of therapy.²³⁶ After placental perfusion studies demonstrated that little or no glyburide (a second generation sulfonylurea) crossed the placenta,²³⁷ Langer et al.’s randomized trial²³⁸ of glyburide versus insulin demonstrated similar outcomes in gestational diabetic pregnancies in which diet was unsuccessful. Additionally, glyburide levels in cord blood were undetectable despite significant levels in maternal blood. While in this original trial only 5% of glyburide treated patients required insulin after the maximum dosage of 20 mg/day was reached without adequate metabolic control, a subsequent series from the same institution²³⁹ reported that 20% of patients could not be managed successfully with glyburide alone. Subsequent studies, using more sophisticated measurement techniques, demonstrated cord blood glyburide levels to be approximately 50–70% of maternal levels.^{240, 241} Because glyburide is a pancreatic secretogogue, it is not clear whether potential effects on the fetal pancreas are beneficial or harmful. No high quality long term follow-up studies of the offspring have been carried out, nor have animal models been evaluated. In a systematic review and meta-analysis,²⁴² when compared with the use of insulin in GDM, glyburide was associated with higher birth weight and a greater risk of macrosomia (RR 2.62) and neonatal hypoglycemia (RR 2.04). When patients requiring treatment in addition to diet are unable to take insulin we occasionally prescribe glyburide, but only after we counsel the patient about the lack of long term data, and we document that discussion in the medical record.

As was mentioned in an earlier section of this chapter, metformin is increasingly popular as a treatment for PCOS and insulin resistance syndrome. It carries distinct advantages over insulin secretagogues such as sulfonylureas, particularly the fact that it does not tend to cause hypoglycemia nor is weight gain a common side effect. On the other hand, as noted above, metformin crosses the placenta and is preferentially concentrated in the fetal compartment¹⁷⁴ and its fetal effects have not been well characterized. If metformin, an insulin sensitizer, rendered the fetus more sensitive to its own insulin, this potentially could worsen diabetic fetopathy. In the metformin in gestational diabetes trial, women with gestational diabetes who required pharmacologic intervention were randomized to treatment with insulin versus treatment with metformin.²⁴³ Maternal and neonatal outcomes were similar in the two groups, and women randomized to metformin were more likely to express willingness to use the same treatment in future pregnancy. However, almost half the women randomized to metformin required the addition of supplemental insulin because the predetermined maximum dose of metformin (2500 mg/day) did not adequately normalize their circulating glucose levels. The only published follow-up studies of offspring randomly exposed to metformin vs insulin in utero found more subcutaneous fat (but similar overall body fat) at 2 years of age in those exposed to metformin, the meaning of which is unclear²⁴⁴ and no difference in neurodevelopmental outcomes at 2 years of age.²⁴⁵ Blood pressure readings at 2 years of age were not significantly different between metformin-exposed and insulin-exposed offspring.²⁴⁶ An 8-year follow-up of children exposed to metformin in utero as part of a randomized trial found no differences in anthropometrics but a higher average fasting glucose level in metformin-exposed offspring.²⁴⁷ The need for more long term studies on the effect of metformin on the exposed fetus, as well as the high rate of supplementation with insulin, call into some question the current applicability of this form of treatment for gestational diabetes.

Antepartum testing of fetal and placental well being is often recommended in gestational diabetes, but there is no universal agreement as to the optimum time for initiation of testing, the most appropriate test to be used or the most appropriate patients to be tested. Landon and Gabbe²⁴⁸ suggest daily maternal assessment of fetal movements beginning at 28 weeks' gestation, with weekly nonstress testing beginning at 40 weeks' gestation, in otherwise uncomplicated gestational diabetic pregnancies without identifiable risk factors. Earlier initiation of nonstress testing is recommended if risk factors such as any type of hypertensive disorder, previous stillbirth or poor or undocumented metabolic control are present. In our center we initiate weekly nonstress testing and amniotic fluid index measurement at approximately 36 weeks' gestation in otherwise uncomplicated gestational diabetic pregnancies, in order to avoid inadvertently omitting testing for patients with prospectively or retrospectively identified risk factors. Testing may be initiated earlier, and performed more often than once weekly, in patients who have additional risk factors. ACOG²⁴⁹ notes that there is a lack of conclusive data, and recommends,  “…for women with GDM with poor glycemic control, fetal surveillance may be beneficial. There is no consensus regarding antepartum testing in women with well-controlled GDM. The specific antepartum test and frequency of testing may be chosen according to local practice.”

Decisions concerning the timing and mode of delivery in gestational diabetic pregnancy should be made using the same considerations described for those with preexisting diabetes. A randomized trial of induction of labor at 38 weeks vs. expectant management included 200 women who required insulin during pregnancy, 187 of whom had gestational diabetes and 13 of whom had pre-existing diabetes.²⁵⁰ Cesarean section rates were not significantly different in the induction group (25%) and the expectant management group (31%). While 23% of newborns in the expectant management group were >90th centile, only 10% of those in the induction group were LGA.

In a population-based retrospective cohort study of over 193,000 GDMs delivering in California between 1997 and 2006, the combined fetal and infant mortality rate risk of delivery at 36 weeks exceeded the risk of expectant management, while the risk of expectant management exceeded the risk of delivery at 39 weeks.²⁵¹  A decision analysis concluded that delivery of diet treated GDMs at either 38 or 39 weeks was associated with the best overall quality outcomes (including fetal and neonatal death, cerebral palsy and other morbidities) depending upon the assumed infant death rate.²⁵² A secondary analysis of a randomized controlled trial of identification and treatment of mild GDM²⁵³ found that induction of labor in women with mild gestational diabetes was not associated with an increased caesarean section rate at 37, 38, 39 or 40 weeks gestational age. ACOG²⁴⁹ recommends that in well controlled gestational diabetic pregnancies there is no good evidence to support routine delivery before 39 weeks. As noted above, epidemiologic studies have found that the lowest perinatal mortality rate for GDM was associated with delivery at 38 or 39 weeks and that delivery at 37, 38 or 39 weeks was not associated with a higher caesarean rate than was delivery at 40 weeks. As noted above, ACOG/SMFM²¹³ recommendations are:

Gestational diabetes

            Well-controlled on diet, insulin or oral agents – late preterm or early term delivery not indicated

            Poorly controlled – 34 to 37 weeks 6 days (individualize, based on severity and other factors)

Insulin is usually not required during labor, but circulating glucose levels should be monitored in order to identify those individuals who would benefit from this form of therapy, i.e., those with glucose levels >120 mg/dL.

FOLLOW-UP AFTER GESTATIONAL DIABETES

Because gestational diabetes is a relatively potent predictor of the later development of type 2 diabetes, with 40% or more of such individuals developing this disorder within 20 years of their index pregnancies,⁹ it is appropriate to test women with previous gestational diabetes every 1–3 years¹⁸⁴ for disordered glucose tolerance. The first such test is usually recommended for the time of postpartum visit, since the patient is being seen for other reasons and is more likely to find the testing convenient at that time. Diagnostic criteria for the 75-g, 2-hour oral glucose tolerance test, recommended for nonpregnant patients, are outlined in Table 8.

Table 8. Diagnostic criteria for the diabetes and prediabetes in the nonpregnant state³⁸

 

While fasting plasma glucose and A1c are accepted diagnostic tests for prediabetes, we believe that the 75 g, 2-hour OGTT is most appropriate for post-pregnancy testing, because of its greater sensitivity.²⁵⁴ By definition, patients with a recent history of GDM are at significant “risk” for another pregnancy during subsequent years. It is important to assess the risk of developing type 2 diabetes in such individuals; they can be counselled about prevention and if they should develop type 2 diabetes prior to their next pregnancy they can be helped to attain good metabolic control prior to conception.

Diagnosis of gestational diabetes early in pregnancy has been found to be significantly predictive of persistent glucose abnormality at the time of postpartum testing.^{255, 256} However, it should be remembered that early testing is usually performed because the patient has some indication or risk factor, and in neither of these studies were all gravidas tested early. Nevertheless, it makes physiologic sense that individuals whose glucose abnormality is present early in pregnancy are at higher likelihood of having had abnormal glucose tolerance prior to pregnancy. Apparently independent risk factors for subsequent diabetes include elevated fasting glucose in the OGTT performed during pregnancy,^{255, 256, 257, 258, 259} obesity,^{255, 258, 260, 259} and elapsed time since the index pregnancy.²⁵⁸ Insulin resistance syndrome, a marker for subsequent type 2 diabetes, was seen in 27% of former GDMs and 11% of normal controls within 11 years of delivery.²⁶¹ This finding suggests a link between pregnancy related insulin resistance associated with GDM, and intrinsic insulin resistance leading to the later development of type 2 diabetes. The degree of abnormality of the postpartum glucose tolerance test also predicts future diabetes, as might be expected.²⁵⁷ The likelihood of persistent diabetes after pregnancy, and of subsequent development of diabetes, may vary with the particular population being studied. For example, in the Danish population studied by Damm et al.²⁵⁷ almost one third of the subsequent diabetes reported was type 1 diabetes. This is a possible explanation for the lack of correlation between maternal obesity and subsequent diabetes in this study, since obesity is not a feature of type 1 diabetes. In fact, the data from this study suggest that BMI >25 kg/m² was present in 30% of those with subsequent normal glucose tolerance, 46% of those with subsequent impaired glucose tolerance, and 58% of those with type 2 diabetes, but only 11% of those with subsequent type 1 diabetes. In contrast, the study of Kjos et al.²⁵⁰ encompassed a Mexican-American population with a high prevalence of type 2 diabetes (9% had diabetes during the first 2 months postpartum and an additional 10% had impaired glucose tolerance). Because the average BMI in these patients was approximately 30 kg/m², it was not surprising that obesity was not discernible as an independent risk factor. None of these studies of risk factors for future diabetes allow the exclusion of particular subsets of former gestational diabetic individuals from further testing. Indeed, in the Kjos study even the lowest risk group, who had maintained all fasting glucose levels during pregnancy below 105 mg/dL, manifested a 10% prevalence of impaired glucose tolerance or diabetes within 2 months postpartum.²⁵⁰

Contraception

Pregnancy in any woman with diabetes ought to be planned, and not unintended. Unless the individual is in excellent metabolic control throughout her reproductive years, unplanned pregnancies are likely to be associated with a greater risk of major congenital malformations than are those in which prepregnancy counseling and planning have taken place. Therefore, family planning, including a discussion of contraception, is appropriate, in fact mandatory, for any adolescent or adult woman with diabetes who has reached menarche. Virtually the full range of contraceptive options is open to the individual with diabetes mellitus and previous gestational diabetes;^{262, 263} however, specific risks may be different for such women than for the general population. Because the health risks associated with pregnancy are greater for a woman with diabetes than for women in general, the benefits of contraception may also be considered to be greater.²⁶⁴

If a woman with diabetes and her husband have completed their family, permanent surgical sterilization for one or the other may be the method of choice. The method risks are all taken at the time of surgery, and no further attention need be paid by the couple to contraception, other than an awareness that no method is 100% effective. Since the woman with diabetes assumes all of the risks of pregnancy, and because not all marriages are permanent, it could be argued that the mother with diabetes is the one who should undergo a sterilization procedure. The counter-argument is that vasectomy is less risky than tubal sterilization, and its efficacy can be determined by a sperm count 6 months afterwards, while the only way to know if a tubal sterilization has failed is when pregnancy occurs!

Oral contraceptive preparations are second in effectiveness to permanent sterilization. Potential risks associated with oral contraceptive use include cardiovascular, thromboembolic, and lipoprotein disorders. Since many of these risks are also increased in individuals with diabetes there has been longstanding concern about this method of contraception in diabetic women. However, data suggesting a multiplicative effect are lacking, and current very low dose formulations appear to have minimal demonstrable risk,²⁶⁵ although severe, active microvascular disease and cardiovascular disease may contraindicate their use.²⁶⁶ Another concern with oral contraceptives containing progestational agents is the potential worsening of glucose metabolism in individuals who are unable to compensate fully with an augmented insulin response. Theoretically individuals already requiring insulin can simply increase their insulin dose. However, women with type 2 diabetes who are not taking insulin might need to do so if their metabolism worsened. For this reason it is necessary to closely monitor glucose metabolism in women with diabetes, and women with former disturbances in carbohydrate metabolism such as gestational diabetes, when they take birth control pills. Use of formulations with the lowest dose of estrogens and progestins make sense when prescribing oral contraceptives to such women. At least two studies have failed to demonstrate an adverse effect of modern low dose oral contraceptives on carbohydrate metabolism in individuals with previous gestational diabetes.^{267, 268}

Progestins are used in a number of non-oral forms for contraception, including implantable, transdermal, and injectable preparations. There are no available long term studies specific to diabetes, but mild alterations in carbohydrate metabolism have been reported with their use in nondiabetic women.^{269, 270} A few studies have raised the issue of an increased likelihood of the development of type 2 diabetes with the use of depot-medroxyprogesterone acetate in former GDMs, but it is possible that this finding was confounded by the presence of other risk factors such as obesity and weight gain.²⁵⁷ In the absence of further data, these forms of contraception should not be considered as the first option in women with diabetes or previous gestational diabetes, but there may be circumstances where the risk–benefit ratio favors their use, such as in patients who are incapable or unlikely to faithfully use daily oral contraceptives or barrier methods.

The intrauterine device, slightly less effective than combined oral contraceptives, has the advantage of requiring no active effort by the user once it has been inserted. The known risk of infection and subsequent infertility is a disadvantage for this method, although there is currently no evidence to suggest that this risk is higher in diabetic individuals. While some studies have reported a high failure rate of this method in individuals with diabetes,²⁷¹ others^{272, 273, 274} have found no such problem.

Barrier methods of contraception appear to have virtually no risk of medical complications, but have the disadvantage that they require faithful adherence by the user and partner and have a relatively higher failure rate compared to the above approaches. The condom also offers an advantage in protection against sexually transmitted diseases.