Quantitative determination of insulin resistance is technically demanding and expensive. We have recently shown (1) that insulin resistance determined using the homeostasis model assessment of insulin resistance (HOMA-IR) has a significantly greater biological variability in individuals with type 2 diabetes than in healthy ones. A surrogate marker of insulin resistance that was reproducible, stable, and easily measured would be invaluable for both research and clinical practice, particularly for following insulin-sensitizing therapy, such as metformin and the thiazolidinediones. A low sex hormone-binding globulin (SHBG) concentration reflects hyperinsulinemic insulin resistance and has been proposed as such a surrogate measure (2–4). This study aimed to compare the biological variation of SHBG and insulin resistance in type 2 diabetes to determine the potential for SHBG as a surrogate marker of insulin resistance in type 2 diabetes.
Subjects were initially recruited for a study to assess the biological variation of insulin resistance in individuals with type 2 diabetes (1). Postmenopausal Caucasian subjects (n = 12) with type 2 diabetes (median age 62 years, range 50–73, and median BMI 31.6 kg/m2, range 25.1–35.7) and 11 age- and weight-matched, healthy, postmenopausal Caucasian control subjects (median age 56 years, range 48–70, and median BMI 32.0 kg/m2, range 26.6–44.4) participated. Fasting blood samples were collected at 4-day intervals on 10 consecutive occasions. Plasma glucose was analyzed in singleton within 4 h of collection. Duplicate samples (i.e., two per visit) of stored serum were randomized and then analyzed (in a single continuous batch using a single batch of reagents) for SHBG and insulin (on a DPC Immulite 2000 analyzer; Euro/DPC, Llanberis, U.K.). The coefficient of variation for serum insulin and SHBG was 10.6% and 8.5%, respectively. The analytical sensitivity of the insulin assay was 2 μU/ml, and there was no stated cross-reactivity with proinsulin. All subjects were asked to have an unrestricted diet and instructed not to modify their eating patterns during the sampling period. The subjects were also advised to refrain from excessive physical exercise and alcohol before each fasting blood test.
The insulin resistance was calculated using the HOMA-IR method (HOMA-IR = [insulin × glucose]/22.5). Biovariability data were analyzed by calculating analytical, within-subject, and between-subject variances (5,6). The critical difference (i.e., the smallest percentage change unlikely to be due to biological variability) between two consecutive SHBG samples in an individual subject with type 2 diabetes was calculated using the formula 2.77(CVI) (5), where CVI is the within-subject biological coefficient of variation.
Figure 1 shows the mean and range of HOMA-IR and SHBG for the individuals in the two groups. In the type 2 diabetic group, SHBG concentrations were lower than those in the control subjects (mean ± SD; 38.8 ± 18.2 vs. 42.2 ± 17.1 nmol/l, P = 0.001), the insulin levels were higher (13.1 ± 5.4 vs. 9.42 ± 3.4 μIU/ml, P = 0.0001), and the HOMA-IR greater (4.33 ± 2.3 vs. 2.11 ± 0.79 units, P = 0.0001).
An inverse relationship was demonstrated between SHBG concentration and HOMA-IR in the group with type 2 diabetes (r = −0.32, P = 0.001) and in control subjects (r = −0.28, P = 0.003). The intraindividual variance of SHBG rose linearly with increasing SHBG concentrations (r = 0.82, P = 0.0001), and after accounting for analytical variation, the intraindividual variation of SHBG for the group with type 2 diabetes was similar to that seen in the control group (mean 2.35 vs. 2.44 nmol/l, P = 0.93). In contrast, the mean intraindividual variation of serum insulin (mean 2.38 vs. 1.45 μU/ml, P = 0.016) and HOMA-IR (mean 1.05 vs. 0.15 units, P = 0.001) was significantly greater in the group with type 2 diabetes than in the control subjects.
The critical difference between two consecutive SHBG samples in an individual patient with type 2 diabetes was 14.5% at any initial level of SHBG, indicating that a subsequent sample must rise or fall by >14.5% to be considered significantly different from the first.
The subjects with type 2 diabetes were hyperinsulinemic, insulin resistant, and demonstrated lower SHBG levels than control subjects. However, the more variable fasting insulin/insulin resistance in the subjects with type 2 diabetes was not reflected by similarly more variable SHBG readings compared with those of the control subjects. This suggests that a low SHBG concentration is a stable integrated marker of insulin resistance and therefore has the characteristics to be potentially used as a surrogate measure of insulin resistance, perhaps in monitoring the response of an individual to insulin sensitizers. However, although SHBG levels differed significantly between those with and without diabetes, the absolute mean difference was small, indicating that measurement of SHBG cannot be used as a simple test for insulin resistance in diabetes. A much larger study is required to investigate whether diagnostic cutoff values for low SHBG concentrations and insulin resistance in type 2 diabetes can be established. Without these parameters, the utility of a low SHBG concentration as a reflection of insulin resistance in type 2 diabetes will be for the serial monitoring of insulin resistance in individuals on treatment after the presence of insulin resistance has been established by conventional means. The low variation of SHBG compared with insulin resistance is likely due to the inherent temporal volatility of insulin and glucose levels as compared with SHBG. In conclusion, in the evaluation of serial measurements of SHBG concentration for an insulin-resistant individual with type 2 diabetes, such as before and after therapeutic intervention, the critical difference value of 14.5% reported here will identify whether any change is beyond that of natural biological variation and therefore a true response.