Alternative Methods of Insulin Sensitivity Assessment in Obese Children and Adolescents

The study included 191 overweight or obese Swedish participants, ranging from children to young adults. Patients with mental, endocrine, or metabolic disorders (except mild dyslipidemia) were excluded. Anthropometric and pubertal assessments were performed (13,14), BMI SD score calculated (15), and dual-energy X-ray absorptiometry (DEXA) scan performed (16). The study protocol was approved by the local ethics committee, and written informed consent was obtained.

FSIVGTT was performed according to standard protocol (7,16,17) with sensitivity index (S_i) calculated by computer program MINMOD version 3.0 (by R. Bergman 1994). Fasting insulin and glucose were analyzed using standard laboratory techniques, and HOMA-IR and QUICKI were calculated as previously described (8,9).

Results are reported as the mean ± SD. Fasting indexes and S_i were transformed logarithmically. Analyses were sex stratified and the sample divided into groups of prepubertal (Tanner stage I), pubertal (Tanner stages II and III), and postpubertal (Tanner stages IV and V) (13,14). Partial correlation coefficients were calculated with adjustment for BMI and age. To explore the effect of insulin resistance on the relationship between S_i and fasting indexes, a variable denoting high/low S_i was tested in regression models (determined as above/below median). An interaction term between indexes and the high/low variable was included to investigate potential effect modification. When testing for interaction, adjustment was made for main effects. P values <0.05 were considered statistically significant. Because QUICKI and HOMA-IR showed consistently highly similar results, only HOMA-IR results are reported.

The study included 191 overweight or obese participants (82 males and 109 females, aged 13.9 ± 2.9 years [median 14.0 years, range 5.5–22.8], BMI 36.9 ± 6.2 kg/m² [35.8 kg/m², 24.2–60.3], BMI SD score 6.1 ± 1.6 [6.0, 2.7–14.0], DEXA 45.7 ± 5.2% [45.6%, 27.2–58.8]). Tanner stage (I/II/III/IV/V) for males was 37.8/14.6/12.2/14.6/20.7% and for females was 8.3/9.2/2.8/16.5/46.8%, respectively. Males were heavier (3.1 kg, P = 0.001), were taller (2.9 cm, P < 0.001), had higher fasting glucose (0.1 mmol/l, P = 0.022), and were less adipose (2.6%, P = 0.001). Statistically significant differences in Tanner stage were found (P < 0.001). DEXA results showed that 99% were above the 98th percentile in body fat percentage (age and sex adjusted) (18).

The relationships between HOMA-IR, fasting insulin, and S_i are presented in Table 1, stratified by sex and pubertal status. The female prepubertal group was too small to provide any reliable results (n = 6). In both sexes, the pubertal group had the highest correlation coefficients. When including the high/low S_i variable in regression analysis, in females the relationship appeared to vary with degree of insulin sensitivity, while variations in HOMA-IR explained 33.7% of the variation in S_i for subjects with high S_i (P < 0.001); the corresponding number for subjects with low S_i was only 3.2% (P = 0.197). Fasting insulin explained 14.1% of the variation in S_i in female subjects with high S_i (P < 0.004), with a corresponding number for subjects with low S_i of 0.3% (P = 0.715). In males, no such interactions were detected either graphically (parallel lines) or in regression analysis. The Pearson correlation coefficient of HOMA-IR to fasting insulin was 0.81 (males 0.87, females 0.78, P < 0.001 in all instances).

We found sex-dependent differences in the explanatory power of fasting indexes, with consistently better ability to explain the variation in S_i among males. In both sexes the explanatory power varied with pubertal status, surprisingly showing highest correlations in the pubertal groups, where insulin sensitivity is known to be low (19). When stratifying for high/low S_i, the relationship was significantly weaker in the more-insulin-resistant females, in which fasting indexes exhibited virtually no explanatory value at all. HOMA-IR yielded little additional explanatory value compared with fasting insulin alone.

The validity of fasting indexes in children and adolescents has been investigated with correlations of HOMA-IR and clamp or FSIVGTT ranging from 0.4 to 0.9 (10,20–23). Conwell et al. (10) found fasting indexes to correlate strongly with S_i of the FSIVGTT-MMOD, concluding that these are valid tools in obese children and adolescents. The high correlation seen in the Conwell study of 18 individuals in three repeated tests (r = −0.9 for HOMA-IR and S_i, P < 0.001) is not found in our larger study, although the correlation in pubertal males is reasonably high. The results in most of our subgroups are in better agreement with the study by Cutfield et al. (24), in which fasting indexes exhibited only weak correlation with the S_i of the FSIGVTT-MMOD (r = −0.4, P < 0.001 for HOMA-IR and S_i). Brandou et al. (12), who examined fasting indexes in obese and lean children at varying stages of puberty, found no correlation between HOMA-IR or QUICKI and S_i. Also, in the present study, for each given value of the fasting indexes there was a wide spread in S_i, which warrants careful interpretation.

In summary, we have shown the validity of HOMA-IR and fasting insulin in explaining S_i as determined by the FSIVGTT in obese children and adolescents to be sex dependent as well as influenced by pubertal stage and in females by level of insulin sensitivity. In pubertal males, HOMA-IR and fasting insulin correlated reasonably well with FSIVGTT-MMOD, whereas in pre- and postpubertal subgroups the correlation was low. In females, the strength of the association was generally weaker, especially among the most insulin-resistant participants. Thus, in obese pediatric populations, especially those at risk of altered glucose homeostasis, we discourage the use of these measures.

This work was supported by grants from Stiftelsen Samariten, the Frimurare Barnhuset Foundation, the Stockholm City Council, and the Swedish Medical Research Council (9941).