Counterpoint: The Oral Glucose Tolerance Test Is Superfluous

The oral glucose tolerance test (OGTT) has been used in clinical medicine for nearly 90 years (1). As heretical as this may sound, it is time for it to be retired. It no longer provides unique and important clinical information that cannot be obtained by other means. The main reason for performing it is to diagnose impaired glucose tolerance (IGT) or diabetes by virtue of the 2-h value (2,3). Both of these are risk factors for cardiovascular disease (CVD) (4,5), and IGT predicts the development of diabetes (6). However, abnormal carbohydrate metabolism is only one of a number of risk factors for CVD, the combination of which is termed the insulin resistance syndrome. Clinically, the insulin resistance syndrome is diagnosed if three of the following five risk factors for CVD are present: central obesity (waist circumference >88 cm in women and >102 cm in men), elevated triglyceride concentrations (>1.7 mmol/l or 150 mg/dl), decreased HDL cholesterol levels (<1.0 mmol/l or 40 mg/dl in men, <1.3 mmol/l or 50 mg/dl in women), hypertension, and hyperglycemia (impaired fasting glucose, IGT, or diabetes) (7).

Two articles in this issue of Diabetes Care demonstrate that the 2-h value on the OGTT adds nothing (8) or very little (9) for identifying CVD risk if the other risk factors are taken into account. Stern et al. (8) followed 2,662 Mexican Americans and 1,595 non-Hispanic whites who were between 25 and 64 years of age and did not have diabetes or CVD at baseline for 7–8 years. Addition of 2-h glucose values from the OGTT to models incorporating readily available CVD risk factors did not improve their power to predict CVD. More specifically, further unpublished analysis of these data revealed that if 1,000 people were screened for CVD risk factors and those who fell into the top 20% of the distribution were considered to be at increased risk, 24 of 37 cases of future CVD events would be identified with the multivariate model that excluded the 2-h value on the OGTT. Using the multivariate model that included the 2-h value identifies 25 of 37 cases of future CVD events. In other words, one would have to perform 1,000 OGTTs to identify one additional case of future CVD events (personal communication, M.P. Stern).

Meigs et al. (9) evaluated 3,370 subjects from the Framingham Offspring Study who had no clinical evidence of CVD, i.e., coronary artery disease, stroke, or intermittent claudication, during 1991–1995 and followed them for 4 years for incident CVD. They showed that if standard CVD risk factors were taken into account, “the marginal predictive capacity of postchallenge hyperglycemia was small” and concluded that “eliminating the OGTT for the screening and diagnosis of diabetes would have minimal impact in terms of identifying CVD risk.” Further unpublished analysis of this Framingham data set revealed that the addition of the 2-h value on the OGTT to the multivariate proportional hazards regression model lead to no more than two additional future CVD cases being detected per 1,000 individuals screened (personal communication, J.B. Meigs).

The results of the DECODE study (10) are the major argument offered for the importance of performing an OGTT to diagnose IGT that places these individuals at an increased risk for CVD. OGTTs were performed in 18,048 men and 7,316 women over the age of 30 years from 13 prospective European cohort studies; the subjects were then followed for a median of 7.3 years with mortality as the outcome. Men and women with IGT at baseline had hazard ratios of 1.51 and 1.60, respectively, for death. Men and women with newly diagnosed diabetes by virtue of a 2-h value on the OGTT ≥11.1 mmol/l at baseline had hazard ratios of 2.02 and 2.77, respectively. Although causes of death were not specified, it seems reasonable to assume that the majority was from CVD in these groups of subjects. However, it cannot be known whether the individuals with IGT had progressed to diabetes in the interim before death.

This may be an important question given the follow-up data of the Mexican American subjects enrolled in the San Antonio Heart Study who underwent a repeat OGTT 7.7 years after the baseline OGTT and who were followed for an additional 7.7 years after the second OGTT (11). Compared with individuals who had normal glucose tolerance (NGT) at baseline and did not progress to diabetes during the follow-up period, increased all-cause and CVD mortality occurred only in those who did progress to diabetes from either IGT or NGT at baseline. Those with IGT at baseline who remained free of diabetes at follow-up did not have an increase in either all-cause or CVD mortality. Thus, in this cohort, IGT that did not progress to diabetes was not a risk factor for CVD over a 15-year period.

There are a number of drawbacks to performing an OGTT. From a clinical perspective, the most serious is its lack of reproducibility (12–15). This is why both the American Diabetes Association (ADA) (2) and the World Health Organization (3) require a second OGTT to confirm the diagnosis of diabetes. Two studies (14,15) have compared the diagnosis in the same individual on the first and second OGTTs performed within 2–6 weeks of each other. In those with IGT on the first OGTT, this diagnosis of IGT was sustained less than one-half of the time on the second OGTT. In one study of 198 people having IGT on the first OGTT, the second diagnosis was IGT in 48%, NGT in 39%, and diabetes in 13% of the subjects (14). The results of the other study (15), in which 93 individuals had IGT on the first OGTT, agreed remarkably well, with the second diagnosis being IGT in 44%, NGT in 46%, and diabetes in 10%. These data would suggest that we should repeat the OGTT to confirm the diagnosis of IGT as is recommended for the diagnosis of diabetes. Other drawbacks are that it is inconvenient, sometimes causes symptoms, and, at least in the U.S., few physicians are using it (16,17).

How about the utility of the OGTT for diagnosing diabetes? In 1979 the National Diabetes Data Group (NDDG) promulgated OGTT criteria for the diagnosis of diabetes using the principle that the level of glycemia chosen should be associated with the specific microvascular complication of diabetic retinopathy (18). The 2-h criterion of ≥11.1 mmol/l (200 mg/dl) was chosen based on the development of this diabetic complication in 77 of 1,213 subjects followed for 3–8 years after a baseline OGTT (19). Thus, this sacrosanct value is based on <100 individuals who had one 2-h value on an OGTT at one point in time. Since then, several long-term intervention studies have conclusively shown that microvascular complications are causally related to HbA_1c levels (20–22). However, these values are normal in two-thirds of people with 2-h glucose concentrations on an OGTT of 11.1–13.3 mmol/l (200–239 mg/dl) and in 20–40% of those with 2-h levels >13.3 mmol/l (240 mg/dl) (23). If the principle of setting a glycemic level for the diagnosis of diabetes that is associated with the microvascular complications of the disease is valid (and I believe that it is), how can we diagnose diabetes in people with normal HbA_1c levels? This is not a benign misdiagnosis. In this country, there are certain negative insurance, and possibly employment, consequences of carrying the diagnosis of diabetes. For instance, people with the diagnosis of diabetes are eight times more likely to be unable to obtain medical insurance because of poor health or illness than people without diabetes (24).

How about the utility of the OGTT for helping to guide therapy? It is of no help in this arena as well. Subjects with IGT have normal glucose levels in their daily lives (25). The results of an OGTT are also not helpful for deciding therapy in patients diagnosed with diabetes. Although CVD is associated with glucose levels (4), the risk is a continuous one occurring throughout the normal range of glucose (5). For example, men between the ages of 45 and 74 years whose HbA_1c levels were between 5.0 and 5.4% had a 2.7-fold increase in myocardial infarctions over 4 years compared with those with HbA_1c values <5.0% (26). Even if a goal HbA_1c level of <5.0% was realistic (which it’s not), there is no evidence to date that lowering glycemia in diabetic patients benefits patients with macrovascular disease (22,27).

Regarding microvascular disease, lowering glycemia (as judged by HbA_1c levels) is helpful (20–22). However, >90% of people who would be diagnosed with diabetes by an OGTT already meet the ADA’s goal HbA_1c level of <7.0% (assuming an upper limit of normal of 6.0%) (28). For example, in the subjects in the data set from the Meta-Analysis Research Group (MRG) on the Diagnosis of Diabetes Using Glycated Hemoglobin (29) who were diagnosed with diabetes with 2-h glucose values ≥11.1 mmol/l (200 mg/dl), but had fasting plasma glucose concentrations (FPG) of <7.0 mmol/l (126 mg/dl), 94% of those with a normal FPG concentration of <6.1 mmol/l (110 mg/dl) and 90% of those with impaired fasting glucose (IFG), i.e., FPG concentrations of 6.1–7.0 mmol/l (110–125 mg/dl), had HbA_1c levels <7% (personal communication, David Schriger). The MRG data were obtained from 10 published studies in which subjects were nonrandomly self-referred from the general population or referred from high-risk populations and thus more likely to have hyperglycemia. People from a randomly selected population who were diagnosed with diabetes by 2-h values on the OGTT, but did not meet the FPG concentration criterion, are even less likely to have HbA_1c levels above the ADA goal of 7%. In the weighted data from the Third National Health and Nutrition Examination Survey (NHANES III), no one with a normal FPG concentration had an HbA_1c >7%, and 95% of those with IFG had values <7% (personal communication, Brett Lorber).

The therapeutic approach to all those diagnosed with IGT and >90% of those diagnosed with diabetes by the OGTT is lifestyle modification of diet and exercise patterns. Indeed, even those with HbA_1c values >7% at diagnosis should initially receive lifestyle modification unless they have symptoms of uncontrolled diabetes (30). Treatment of hypertension and hyperlipidemia is also important, but the OGTT is not involved in identifying these risk factors.

In summary, the OGTT is not helpful clinically. It is not necessary for identifying individuals at risk for CVD. Far too many people whose diagnosis of diabetes depends solely on the 2-h value have normal HbA_1c levels and therefore (in my view) should not be given the diagnosis of diabetes. Furthermore, the OGTT is useless in guiding therapy, either in those with IGT or with diabetes, regardless of the criteria used to diagnose the latter. An argument perhaps can be made to use it for epidemiological studies. Indeed, the ADA Expert Committee on the Diagnosis and Classification of Diabetes Mellitus decided to retain the 2-h value of ≥11.1 mmol/l (200 mg/dl) because a large number of epidemiological studies in the literature used this value to define diabetes and changing it “would be very disruptive” (2). However, based on the above data, the OGTT has clearly outlived its clinical usefulness and should be retired.