Fasting Plasma Glucose (FPG) and the Risk of Impaired Glucose Tolerance in Obese Children and Adolescents




A timely diagnosis of impaired glucose tolerance (IGT) is desirable in obesity. The oral glucose tolerance test (OGTT), the gold standard to diagnose this condition, may not be realistically performed in all patients due to discomfort, labor, and cost. The aim of this study was to assess whether one or more biochemical indexes measured in fasting conditions could be used to identify obese children at risk of IGT. A cohort of 563 white obese children and adolescents (M/F: 315/248; aged 4–17 years) was recruited and underwent anthropometric evaluation and OGTT. Anthropometric parameters, fasting plasma glucose (FPG), fasting serum insulin (FSI), and homeostasis model assessment of insulin resistance (HOMAIR) were tested in pursuit of a possible threshold to be used as a predictor of IGT. Thirty-seven children (6.9%) had IGT and one child (0.1%) had type 2 diabetes (T2D). FPG, FSI, and HOMAIR were all significantly higher in children with IGT than in children without IGT. Receiver-operating characteristic (ROC) curve analyses run for gender and puberty-adjusted FPG, FSI, and HOMAIR were all significant: area under the curve (95% confidence interval) equaled 0.68 (0.59–0.76), 0.66 (0.56–0.76), and 0.68 (0.59–0.78), respectively. The three parameters did not show significantly different sensitivity/specificity in the pooled population or in the gender/puberty subgroups. Thresholds varied among gender/puberty subgroups for FSI and HOMAIR, but not for FPG, which showed a fixed threshold of 86 mg/dl. A gender/puberty independent cutoff of FPG may be considered a screening tool to narrow clinical indication to OGTT in obese white children and adolescents.


Impaired glucose tolerance (IGT) is a well-established risk factor for cardiovascular disease, type 2 diabetes (T2D), and cardiovascular mortality (1,2,3,4,5). In obese children, IGT has been associated with >50% risk of short-term IGT persistence (30%) or T2D onset (24%) (6). Moreover, this condition has been associated with cardiovascular risk factors in children (7) and, similarly to what has been observed in adults, it may increase long-term cardiovascular morbidity and mortality, independently of the possible development of T2D (1,2,3,4). Therefore, early diagnosis of IGT could be very helpful for obese children. However, two main questions remain: What is the best test for screening asymptomatic obese children? and What are the criteria for selecting the subsample of obese children to test? At present a consensus is not available, and the American Diabetes Association claims there is a need for large-scale studies on child populations (8).

The oral glucose tolerance test (OGTT) is the gold standard for the diagnosis of IGT, although low reproducibility has been reported as a drawback of this test (9). In fact, studies reporting poor metabolic and/or cardiovascular prognosis significantly associated with IGT were generally based on one, not confirmed baseline OGTT (1,2,3,4,5,6), so that IGT defined by this test should be considered an important risk factor independently of its confirmation. Therefore, at present, the OGTT can still be considered the most suitable test to assess glucose tolerance in a clinical setting.

OGTT showed a relatively low yield (2–14%) of identifying obese children and adolescents with IGT in most studies (7,10,11,12,13,14,15,16,17), even when selectively performed according to traditional clinical and/or familial/ethnical risk factors (10,17), although a few studies reported higher prevalence of IGT (20–28%) in a US Latino population (18,19) or in multiethnic cohorts with particularly high insulin resistance (9,20).

OGTT may not be realistically performed in all obese children and adolescents due to costs, labor, and patient discomfort. A family history of T2D, a well-established risk factor for IGT/T2D, has been reported by >50% of parents of obese children (7), whereas a much lower percentage of obese children are likely to present IGT (7,10,11,12,13,14,15,16,17); then, familiarity for T2D may not be a critical variable for selecting the subsample at high risk of IGT. Ideally, the use of another, more specific parameter could be helpful in distinguishing obese children with higher risk of IGT from those with low risk.

The purpose of this study was to see whether one or more fasting glucose metabolism parameters were able to predict IGT in obese children and whether they could be suitable as screening tools for selecting obese children to be tested with OGTT.

Methods and Procedures


We recruited 563 Italian obese children and adolescents (315 males and 248 females) between the ages of 4 and 17 years. Subjects were recruited from the clinical centers for childhood obesity in Verona and Milan University Hospitals within their first examination for weight excess. Inclusion criteria were white ethnicity, age (4–17 years), and obesity. Exclusion criteria were obesity associated with endocrine disorders, chronic diseases, malformations, and chronic use of drugs. All eligible subjects accepted to undergo OGTT. The characteristics of the sample are shown in Table 1.

Table 1.  Physical characteristics and biochemical parameters of boys and girls
inline image

All participants underwent a physical examination with anthropometric measurements and assessment of pubertal development. Body weight was rounded off to the nearest 0.1 kg on standard physician's beam scales and height was measured to the nearest 0.5 cm on standardized, wall-mounted height boards. BMI was calculated as weight(kg)/(height(m))2. Using International Obesity Task Force BMI cutoffs as reference (21), children with a BMI above the curve that passes through the BMI of 30 at the age of 18 were defined as obese. The standard deviation score of the BMI (BMI z-score) was also calculated according to the Cole's least mean square method, using Italian updated norms (22). Puberty development was assessed on the basis of Tanner stages (23), and accordingly, the children were divided into two groups: prepubertal: boys without pubic hair and gonadal stage I, girls without pubic hair and breast stage I; pubertal: boys with pubic hair or gonadal stage ≥ II and girls with pubic hair or breast stage ≥ II (23).

Blood pressure was measured three times on the left arm with the subject sitting, using a mercury sphygmomanometer and a cuff appropriate for the child's age. The average of the three blood pressure values was used for the analysis. Sex, age, and height cutoff values for systolic and diastolic blood pressure were based on United States references (24) and high blood pressure was defined as systolic and/or diastolic blood pressure higher than the 95th percentile for gender, age, and height.

Children underwent OGTT (1.75-g/kg oral glucose, maximum 75 g), after a 12-h overnight fast; blood samples were obtained at 0, 30, 60, 90, and 120 min for determination of glucose and insulin (25). IFG, IGT, and T2D were defined as plasma glucose between 100 and 125 mg/dl at time 0 min, between 140 and 199 mg/dl at time 120 min and above 199 mg/dl at time 120 min, respectively (10). Homeostasis model assessment of insulin resistance (HOMAIR) was calculated as fasting serum insulin (FSI) (µU/ml) × fasting plasma glucose (FPG) (mmol/l)/22.5 (26).

Parental informed consent and child assent were obtained. The protocol was in accordance with the 1975 Declaration of Helsinki as revised in 1983.

Biochemical parameters

Plasma glucose concentrations were measured using the glucose oxidase method. Insulin was measured by chemiluminescent immunometric assay.

Statistical analysis

Two-way ANOVA was used to compare means of anthropometric and biochemical variables across genders and puberty states. We had previously transformed skewed variables (P < 0.05 at Kolgomorov–Smirnov test) logarithmically to obtain normal distribution and we checked for equivalence of variances for all variables. We saved gender/puberty adjusted residuals for all variables which showed different means according to gender and/or puberty. We also studied dichotomized variables according to gender, puberty, and gender/puberty interaction, by using linear logistic regression.

We searched for possible predictors of IGT by studying variation of parameters according to presence/absence of IGT. We used general linear model for quantitative variables, adjusting for gender, and/or puberty when appropriate; we used linear logistic regression for dichotomized variables, adjusting for gender and/or puberty when appropriate.

We considered all variables with significantly different means or proportion in subjects without IGT compared to subjects with IGT, as predictors of IGT. To evaluate predictive properties of continuous biochemical predictors, we performed receiver-operating characteristic (ROC) analysis on the pooled population for each of them, with IGT as state variable. For variables which showed different means according to gender, puberty, or both, we performed ROC analysis with gender/puberty adjusted residuals instead of raw values, in order to improve ROC curves and to avoid raw threshold values that were identical for gender and puberty state, and then inevitably burdened with heterogeneous predictive properties in gender/puberty subgroups.

By ROC analyses, for all considered variables we obtained threshold values by calculating the Youden index (J), that is the maximum vertical distance between the ROC curve and the reference or chance line (maximum{sensitivity + specificity − 1}) (27). For each threshold value, we calculated sensitivity, specificity, predictive values, and 95% confidence intervals in predicting IGT, by classical Bayesian formulas on cross-table data. We drew cross-tables and predictive properties not only for the pooled population but also for gender and puberty subgroups. In fact, gender and puberty were significantly associated with different prevalence of IGT, thus inevitably conditioning different predictive values for the same sensitivity/specificity. In fact, they could even be associated per se with variation of the sensitivity/specificity of the same predictor. We then tested whether sensitivity and specificity varied significantly for each predictor inter gender/puberty subgroups by comparing sensitivity/specificity 95% confidence interval inter subgroups. Moreover, for each subgroup we compared sensitivity, specificity, and predictive values between the different predictors.

All computed analyses were performed by SPSS 16.0 statistical package (SPSS, Chicago, IL). Significance was fixed at P < 0.05.


All physical characteristics were significantly affected by gender and puberty (Table 1). FPG and FSI, but not HOMAIR, varied significantly according to gender. FSI and HOMAIR, but not FPG, varied significantly according to puberty.

Two-hour plasma glucose (after glucose load) was significantly higher in pubertal than in prepubertal children in both genders. Frequency of IFG was not significantly different between males and females. In females, a gender × puberty interaction was found, the IFG frequency being higher in pubertal than in prepubertal girls.

Children with IGT had significantly higher FPG, higher FSI, and higher HOMAIR than children without IGT (Table 2). IGT frequency was higher in girls than in boys and it was affected by puberty in girls only, being higher in pubertal than in prepubertal girls.

Table 2.  Physical and biochemical characteristics of children with or without IGT
inline image

ROC curve analyses run for gender and puberty-adjusted biochemical parameters (FPG, FSI, HOMAIR), were all significant and were not statistically different from each other, as demonstrated by their widely overlapping 95% confidence interval: area under the curve = 0.68 (0.59–0.76), P = 0.0002; area under the curve = 0.66 (0.56–0.76), P = 0.001; area under the curve = 0.68 (0.59–0.78), P = 0.0001, respectively (Table 3 and Figure 1). In the pooled population, FPG, FSI, and HOMAIR did not show statistically different sensitivity, specificity, or predictive values. This result was also confirmed in each gender/puberty subgroup (Table 3). Sensitivity and specificity of FPG, FSI, and HOMAIR were not significantly different among the four gender/puberty subgroups.

Table 3.  IGT prediction properties of HOMAIR, FPG, and FSI in gender/puberty subgroups and in the pooled population
inline image
Figure 1.

Areas under the ROC curves for gender/puberty-adjusted FPG, FSI, and HOMAIR predicting impaired glucose tolerance. FPG, fasting plasma glucose; FSI, fasting serum insulin; HOMAIR, homeostasis model assessment of insulin resistance.

Threshold values varied among gender/puberty subgroups for FSI and HOMAIR, but not for FPG, due to the minimal variation of this parameter according to gender and to the absence of variation according to puberty.


Our study found a not negligible but relatively low overall prevalence of IGT among obese children and adolescents, similar to that reported in most studies on pediatric populations affected by obesity (7,10,11,12,13,14,15,16,17), and lower than IGT prevalence observed in Latino children (18,19) and in some small mixed ethnicity populations with particularly high insulin resistance (9,20). This result further supports the idea that in several populations performing large-scale OGTT in obese children may have too low a yield to be cost-effective, whereas searching for effective fasting predictors of IGT could be a good strategy to define the subsample of obese children at risk of IGT, for whom OGTT would be justified.

Our study had two main results: firstly, that IFG is not a predictor of IGT in obese children; secondly, that FPG, FSI, and HOMAIR are acceptable negative predictors of IGT, as they succeed to select children with very low to absent risk to present this condition.

The lack of association between FPG in the 100–125 mg/dl (IFG) range and IGT is in agreement with some previous studies done on obese children (11,13,20), and in contrast with others showing a significant higher risk of IGT among children with IFG (9,12,19). However, even when associated with IGT, IFG has shown to be absent in most case (70–90%) of IGT, so that it can not be considered a sensitive predictor of this disorder (9,12,19).

Indeed, IFG and IGT are two different expressions of impaired glucose metabolism. Fasting endogenous glucose production is inappropriately increased in individuals with impaired fasting glucose (28). On the contrary, endogenous glucose production is promptly suppressed in individuals with IGT after glucose ingestion and who show an inappropriately low rate of postprandial glucose disappearance (28).

FPG, FSI, and HOMAIR were all predictors of IGT with similar sensitivity and specificity both in the pooled population and in the gender/puberty subgroups. FPG showed homogeneous sensitivity and specificity according to gender and puberty. The same was found for FSI and HOMAIR. However, FSI and HOMAIR had gender/puberty specific thresholds; for example, they both have four different thresholds—one for each gender/puberty subgroup. On the contrary, FPG had the same threshold for the two genders and in prepubertal and pubertal children. This is the most interesting finding of our study. In fact, although FPG, FSI, and HOMAIR could be indifferently proposed as potential parameters to select obese children at higher risk of IGT, the chance of using the same FPG threshold for both genders despite pubertal stage makes FPG preferable. Other advantages are (i) the widespread availability of plasma glucose measurements in all laboratories; (ii) the lower interlaboratory variability of measuring plasma glucose rather than serum insulin; and (iii) the lower cost of measuring plasma glucose than serum insulin (29,30,31).

The threshold value of 4.8 mmol/l (86 mg/dl) has moderate rather than optimal sensitivity and specificity; nevertheless, thanks to the low prevalence of IGT, it is suitable to individualize children with very high probability of not being affected (negative predictive value): close to 100% in gender/puberty subgroups where IGT has the lowest prevalence. This may have important clinical implications, particularly because it allows a clinician dealing with an obese child with under-threshold FPG to be relatively sure that his patient does not have IGT, which has an overall frequency of one child out of fifteen among obese children. This information enables the clinician to avoid OGTT with acceptable confidence.

Use of the FPG threshold is less effective as regards positive prediction because of suboptimal specificity of the test along with the low prevalence of the disturbance. In fact, children and adolescents with FPG above threshold, though at higher risk of IGT, are far from being at high risk (positive predictive value = 12%), and most of them (88%) have normal glucose tolerance. Therefore, the use of this fasting cutoff improves but does not optimize the yield of OGTT for detecting IGT in obese children. Thus, even if FPG threshold individualizes a subsample at higher risk of IGT, it still implies a large number of obese children to be tested with OGTT (about 40% of the total). Nevertheless, finding FPG above threshold may be of practical interest, independently of the choice of whether to have the child undergo OGTT or not. In fact, “positive” FPG may be a useful tool to increase the index of suspicion for IGT and to further justify family counseling on ways to prevent T2D, as recommended by the International Society of Pediatric and Adolescent Diabetes in children and adolescents “at high risk” (25).

Regarding the possible use of the cutoff when considering proposing OGTT to test-positive children, it is highly possible that clinicians will make their choice according to a number of clinical aspects which must be integrated to the use of FPG. For example, puberty has recently been reported as a significant risk factor of IGT among white obese children and adolescents (12,14), which was confirmed also by the present study, even if only in girls (perhaps because of insufficient proportion of mid-pubertal males in our sample). So puberty may be an important additional clinical factor in favor of OGTT execution. Local economical and organizational conditions may further influence the decision.

However, in an attempt to further distinguish among obese children with a FPG higher than cutoff (86 mg/dl), those with the highest risk of IGT, we looked for another potential predictive parameter among those most frequently used in clinical practice such as BMI z-scores and blood pressure. In our subgroup of obese children with an FPG > 86 mg/dl, mean BMI z-score values were not different between subjects with or without IGT (P = 0.215), whereas blood pressure values were different (P = 0.009 for systolic blood pressure and P = 0.026 for diastolic blood pressure) (data not shown). Interestingly, high blood pressure (systolic and/or diastolic blood pressure higher than the 95th percentile for gender, age, and height) was found in 66.7% of obese children with IGT and in 40.8% of obese children without IGT (χ2 = 4.41, P < 0.04), and it showed a positive predictive value of 15%. Therefore, high blood pressure may slightly increase (from 12 to 15%) the chances of diagnosing IGT in obese children with FPG over-threshold. Interestingly, high blood pressure was a predictor of IGT electively in the FPG over-threshold subsample of obese children, indirectly confirming the reliability of the plasma glucose threshold.

A weak point of this study is that findings cannot be generalized to all obese children. In fact, the above mentioned variation of IGT prevalence according to ethnicity or type of recruitment or other environmental factors implies that any given binary risk factor may have population-restrictive validity and/or predictive properties. However, the fasting glucose threshold of ≤86 mg/dl may be quite stable at least in white populations of obese children. In fact, this threshold has already been associated with the lowest risk of IGT among white obese children and adolescents, with the same negative predictive value observed in the present study: about 3% (13).

In conclusion, a gender/puberty independent cutoff of FPG (86 mg/dl) may be considered a screening tool to narrow clinical indication to OGTT in obese white children and adolescents. Further studies on even larger samples of obese children should be performed in order to confirm these preliminary results and, possibly, to achieve a more precise assessment of predictive properties of the three fasting parameters, along with the definition of a validated cutoff.


The authors declared no conflict of interest.