Development of a model to assess the cost-effectiveness of gestational diabetes mellitus screening and lifestyle change for the prevention of type 2 diabetes mellitus.
G estational diabetes mellitus (GDM), defined as carbohydrate intolerance first diagnosed during pregnancy (1), is a common medical complication of pregnancy, affecting 1.1–14.3% of pregnant women depending on the ethnic and clinical characteristics of the population and the diagnostic test employed (2). Ever since O’Sullivan and Mahan (3) published their criteria for diagnosis of GDM using a 100-g oral glucose tolerance test (GTT), clinicians worldwide have been struggling to determine whether screening for GDM should be offered routinely in pregnancy and, if so, the optimal method of screening. There have been no adequately designed randomized controlled trials to answer the question of whether screening for GDM is both beneficial and cost effective, leading to a wide variance in screening practices worldwide. The Australian Carbohydrate Intolerance Study in Pregnant Women (ACHOIS), a multicenter randomized controlled trial of GDM treatment versus routine prenatal care, found a reduction in a composite of severe perinatal outcomes in the treatment group compared with a control group receiving routine prenatal care (4). Although it was not a trial of screening, its results have convinced many practitioners to adopt some type of screening for GDM because logically, in order to treat GDM (an asymptomatic entity), one must first screen for it. Recently, the U.S. Preventive Services Task Force (USPSTF), in an update to its policy statement on screening for GDM, recognized that treatment of GDM after 24 weeks of gestation improves some maternal and neonatal outcomes but, conversely, also stated that there is still insufficient evidence to support screening of all pregnant women either before or after 24 weeks of gestation (5). Despite this, most clinicians use some method of screening for GDM. Ideally, the chosen screening protocol should identify subjects at maximal risk of adverse pregnancy outcomes who would most benefit from intensified management and surveillance, while freeing the rest from the burden of excessive interventions. Unfortunately, the policy of universal or near-universal screening that is recommended by numerous professional medical organizations (1,6–8) will lead to the blanket labeling of a large group of women as having GDM, without differentiating between those at high and those at low risk of pregnancy complications. It has very clearly been shown that glucose intolerance in pregnancy is not a threshold phenomenon but, rather, is linked to several adverse pregnancy outcomes along a continuum of measured glucose values in both fasting and postprandial (or glucose challenge) states (9– 11). Not all of these outcomes are of equal clinical importance; to justify widespread screening, one would prefer to show a reduction in serious outcomes such as perinatal death and permanent birth injury. These are rare outcomes and, thus, are difficult to study. The best prospective data on the effect of increasing levels of glucose intolerance in pregnancy come from the Toronto Tri-Hospital Gestational Diabetes Project (9) and have been confirmed by the recently published results of the Hyperglycemia and Adverse Pregnancy Outcome (HAPO) study. The data from the Toronto Tri-Hospital Gestational Diabetes Project showed a graded increase in the risk of macrosomia and preeclampsia, the need for phototherapy, and maternal and neonatal length of stay. The HAPO study largely confirmed these data and Pederson’s hypothesis (12), showing a positive association between plasma glucose values after a 75-g oral GTT, birth weight above the 90th percentile, and cord blood C-peptide levels (11). Neither of these studies was able to identify clear outcome-based thresholds that could lead to new clinically relevant diagnostic criteria for GDM. This brings us to the current status in 2009 where we have evidence that treatment of GDM reduces some perinatal outcomes without significantly increasing the cesarean section rate but still lack a gold standard diagnostic or screening test. In North America, the most common method of screening is a two-stage test comprised of a 50-g oral glucose challenge test (GCT) at 24–28 weeks of gestation followed by either a 75-g or a 100-g oral GTT for women who screen positive on the GCT (13). Using a cutoff of 7.8 mmol/l (140 mg/dl) for the GCT, 14– 18% will test positive and need to proceed to the diagnostic test. The reported sensitivity and specificity of this test strategy are 80 and 90%, respectively, whereas the positive and negative predictive values vary according to the prevalence of GDM in the population tested (14). Thus, with this strategy some 20% of women with GDM will remain undiagnosed even with universal screening. In many parts of Europe, a risk factor–based approach to GDM screening is still the most common approach (15–17). Common risk factors can be found in Table 1. Overall, the performance of risk factor–based screening has been poor, with reported sensitivity and specificity of 63 and 56%, respectively (18). This poor performance is in part due to the inability to apply historic obstetric risk factors to women during their first pregnancy; thus, primiparas who develop GDM will likely remain undiagnosed unless they develop macrosomia, glycosuria, or polyhydramnios during the index pregnancy. The pragmatic utility of applying risk factor–based screening will largely depend on the frequency of these risk factors in the screened population. For example, if the screened population is mostly slim young Caucasians, then many women will be spared biochemical screening. On the other hand, in heavier, older, multiethnic obstetric populations, applying risk factor–based screening will likely lead to the majority of women undergoing biochemical screening. Studies from North America have shown that using the American Diabetes Association criteria for selective screening based on risk factors will lead to some 90% of the obstetric population still undergoing some form of biochemical screening (19). This leaves us with the question of whether it is possible to apply a form of risk factor–based screening that will spare the lower-risk population unnecessary tests while maintaining the performance of a universal screening protocol. The answer is available from a secondary analysis of data from the Toronto Tri-Hospital Gestational Diabetes Project (20). The subject population inE d i t o r i a l s