In this sample of adolescents who received an OGTT as part of NHANES, we were surprised to find a relatively poor association between two important markers of risk for T2DM: MetS and IGT. While these processes are clearly linked—seen in a 3-fold higher prevalence of MetS among adolescents with IGT—we found that the majority of cases of IGT were not associated with MetS, either using traditional ATP-III criteria or using a sex- and race/ethnicity-specific linear MetS Z-score. This low prevalence of MetS in IGT was true both in the overall population and in a sub-set of adolescents who were overweight/obese and thus more likely to experience the processes underlying MetS, such as adipocyte dysfunction and oxidative stress. Altogether, this lower-than-anticipated association between MetS and IGT may implicate a predominance of non-MetS causes of IGT in adolescents.
The low association of MetS with IGT in adolescents is likely related to the complex physiology of glucose control. Elevations in BG following a glucose challenge are influenced by insulin release and tissue responsiveness to insulin . Insulin release following glucose ingestion occurs in two phases: an initial spike (the first phase, blunting of which is an early occurrence in the pathophysiology of T2DM) and a second more gradual rise that can be heightened in T2DM but remains inadequate to lower BG . Insulin secretion is influenced by underlying genes affecting beta cell function (which constitute the largest group of genes implicated in T2DM pathogenesis in large-scale evaluations ) but can be suppressed further by elevations in lipids [36, 37] and glucose itself , as well as dysfunction of other hormones including the incretin GLP-1 . Insulin resistance can contribute to BG elevations by necessitating that beta cells secrete higher amounts of insulin to mediate sufficient glucose disposal. Once the minimum threshold of insulin level for maintenance of BG levels is exceeded, post-prandial BG begins to rise. The combination of inadequate insulin secretion and insulin resistance is the primary cause of IGT and subsequent diabetes in adults [39–41].
Among children and adolescents, the major cause of IGT is unknown, while the major cause of diabetes remains isolated defects in insulin production as seen following auto-immune beta cell destruction in T1DM, with a prevalence of 0.23% among adolescents compared to a prevalence of 0.042% for T2DM in this age range . Limitations in insulin release are also seen in monogenic forms of diabetes such as the group of genes comprising MODY, also with a low overall prevalence at approximately 1% of pediatric diabetes cases [42, 43]. Whereas insulin resistance and MetS are more prevalent in overweight/obese children and adolescents  (contributing to an improved sensitivity of MetS for IGT identification among overweight/obese adolescents, Table 2), primary defects in insulin release such as T1DM and MODY would be expected to be present across the weight spectrum in childhood. T1DM itself, if poorly controlled, is clearly associated with abnormalities in MetS components, particularly hypertriglyceridemia and low HDL , further complicating the potential relationships between MetS and IGT. We excluded participants with known diabetes from our analysis, but it is possible that a small percentage of the adolescents in our sample had early, undiagnosed T1DM or MODY . Nevertheless, the proportion of all participants in our sample with non-MetS IGT (3.6%) far exceeded the expected number of cases of undiagnosed T1DM and MODY (which together have an incidence of 0.022% per year in childhood [21, 46]), potentially suggesting a high prevalence of other limitations of insulin secretion, including polygenic defects in beta-cell function that have been implicated in T2DM in adults [18, 47].
We hypothesized that one limitation in the ability of MetS to identify individuals with IGT was due to variation in the diagnostic accuracy of MetS criteria by racial/ethnic group [22–26]. Because of this, we performed our analysis using both a common adolescent adaptation of ATP-III MetS criteria [20, 29], as well as a sex- and race/ethnicity-specific MetS severity score . This severity score has the potential to have cut-off levels adjusted based on outcomes-based data (which we currently lack) or based on a desire to identify higher numbers of adolescents at risk. In using this score we first tested a cut-off level of 1 standard deviation above the mean, which provides prevalence of MetS similar to that determined by ATP-III criteria—and which produced sensitivity values for IGT prediction similar to ATP-III criteria. We then tested a more liberal cut-off of 0.75 standard deviations above the mean, exhibiting increased sensitivity but worsened specificity compared to ATP-III-based criteria. This type of approach could be used to improve identification of adolescents at increased risk for long-term diseases associated with MetS—which could be important since improved tools for risk detection are badly needed to target interventions to help avert disease progression .
The MetS Z-score exhibited a differential response in IGT prediction by racial/ethnic group, with the score overall exhibiting worsened sensitivity among non-Hispanic white adolescents (using a MetS Z-score cut-off 1.0, sensitivity was 11.2% vs. 17.4% for ATP-III MetS) but an improved sensitivity among non-Hispanic blacks (55.9% vs. 31.8%). This improvement in sensitivity among non-Hispanic black adolescents may not be surprising, since traditional MetS criteria utilize population-based cut-off values for the individual components and do not take into account that non-Hispanic-black adolescents have lower baseline levels of triglycerides and are less likely to exhibit abnormalities in triglycerides or HDL despite having more insulin resistance and higher rates of diabetes [24–26]. These inter-ethnicity differences are what stimulated our formulation of the sex- and race/ethnicity-specific MetS Z-score in the first place. Overall, the MetS Z-score appeared to work best in the identification of non-Hispanic black adolescents with IGT, though we were limited in many of our comparisons between racial/ethnic groups by the small sample size of adolescents in NHANES who underwent OGTT’s and by the overall lower prevalence of IGT.
However, while we noted an increased sensitivity using a lower cur-off of the race/ethnicity-specific MetS Z-score, the presence of MetS overall was a poor screening test to identify adolescents with IGT. Acceptable degrees of sensitivity and specificity in a screening test depend on the importance of the outcome being screened for and the downside to missing detection of that outcome. In this case, the specificity of these tests was overall reasonable clinically, 78-97% among all adolescents. However, the low sensitivity and thus number of non-MetS cases of IGT reflects the potential for false reassurance regarding the risk of IGT in an adolescent based on the absence of MetS.
Interestingly, there appeared to be a sizable number of cases of elevated fasting insulin levels (as an estimate of insulin resistance) in this sample that were not identified by ATP-III MetS or our MetS Z-score. Among adolescents with IGT there was a higher prevalence of elevated fasting insulin—present in 52% of adolescents with IGT—than of ATP-III MetS (24%) or any of the individual components of MetS (with prevalences of 19-45%). There is a clear difficulty in using measures of fasting insulin as an estimate of insulin resistance in settings of glucose excursions, since the mere elevation of BG implies a limitation in secretion of adequate amounts of insulin to normalize BG. Additionally, while fasting insulin correlates highly with other surrogate markers of insulin resistance (HOMA-IR, QUICKI), it clearly lacks the precision of more robust measures of insulin resistance, such as an insulin clamp [44, 48]. Thus, our measure of those with elevated insulin levels (fasting insulin above 16.0 IU/mL, approximately the 95th percentile among lean adolescents in NHANES  and used elsewhere [32–34]) may not reflect the full number of participants with insulin resistance. Overall, however, there were 48% of children with IGT who did not exhibit elevations in fasting insulin, again suggesting a high prevalence of IGT unrelated to insulin resistance.
This study had multiple limitations, including the cross-sectional design of NHANES, which limits any conclusions regarding causality. In addition, while NHANES often represents a powerful study, there was only a small subset of adolescents who underwent the OGTT. Finally, we lacked important additional information such as antibodies associated with T1DM as well as genetic information on participants.
In conclusion, we found that the presence of MetS and elevated fasting insulin in adolescents had a poor correlation with IGT, an important precursor T2DM as well as a potential finding during the short pre-symptomatic phase of T1DM. This lack of overlap between MetS and IGT may indicate that assessment of MetS is not likely to be a good indicator of which adolescents to screen using an OGTT. These data further underscore the need for further research to assess for other potential contributors to IGT, including T1DM, MODY and polymorphisms associated with poorer beta-cell function. Practitioners should keep in mind other potential causes of IGT, even when evaluating obese adolescents with IGT.