This study proposes reference values for serum fasting insulin concentrations and HOMA-IR index in healthy Brazilian subjects, selected from an extensive database of an important laboratory serving the entire Rio de Janeiro state.
It is noteworthy that, even before applying exclusion criteria, 69% of our subjects were female, which may be explained, most likely, by the fact that women in our country seek more medical care and, consequently, undergo more laboratory tests than men. After applying the exclusion criteria, the proportion of women in our cohort increased even more (86%), showing that proportionally more men had pathologies that were part of our exclusion criteria, such as type 2 diabetes, MS, hypertension, obesity, and dyslipidemia.
The RIs for fasting insulin levels in our study were, respectively, 2.54–13.30, 2.43–11.89, and 2.52–13.14 µU/mL (15.3-80.12, 14.6–71.7, and 15.2–79.2 pmol/L) in women, men, and the total population. In line with these results, Tohidi et al. studied 309 “healthy” Iranian subjects with similar characteristics to our population (124 men and 185 women, 24–83 years, mean age 40 years) and using the same manufacturer (Roche Diagnostics kits and the Roche/Hitachi Cobas e-411 analyzer), reported RIs for serum insulin of 2.34–11.98, 1.61–11.37 and 2.11–12.49 µU/mL (14.0-71.9, 9.7–68.2 and 12.7–74.9 pmol/L) in non-menopausal women, men, and total population [30]. Similarly, Francois et al., studying a non-overweight French population, have reported RIs for serum insulin concentration of 2.34–11.25 and 2.08–11.82 µU/mL (14.0-67.5 and 12.5–70.9 and pmol/L) in women (aged 33–55 years, n = 157) and men (aged 37–58 years, n = 162), respectively [31]. On the contrary, Li et al. recently reported RIs for serum insulin concentrations in 1434 “healthy” non-diabetic Chinese men (age range 20–69 years) to be 1.57–16.32 µU/mL (9.42–97.9 pmol/L) [32]; their higher upper limit compared to our value (16.32 µU/mL or 77.3 pmol/L) may be due to our more stringent criteria for selecting healthy subject and/or related to ethnic and racial differences.
Mean insulin levels in our study were slightly higher in women than in men, a finding similar to other studies like Tohidi et al., who also found mean serum insulin levels to be higher in women than men [30]. The trend of higher insulin levels in women than men may be related to body composition since men tend to have a proportionally higher lean mass/fat mass ratio than women.
In addition to an initial increase in insulin levels at the beginning of adulthood, probably related to the increase in insulin levels observed in late adolescence [33], we highlight a trend of increasing insulin, glucose, and Tgc levels as well as increasing BMI with aging. This finding is corroborated by epidemiological studies and is associated with decreased peripheral insulin sensitivity, probably due to the reduction of lean body mass, increased BMI, and physical inactivity with aging [30]. In the very elderly (from the age of 80 onwards) however, these markers show a downward trend, probably due to nutritional factors such as malnutrition, common in this age group (data not shown) [34].
Despite the statistically significant difference between mean insulin levels in men and women, it does not justify using sex-specific reference values since the ratio of standard deviations was < 1.5 (0.05) and the calculated z value of our cohort was less than the criterion dictating partitioning of reference values [11].
The RIs for the HOMA-IR index in our study were, respectively, 0.39–2.86, 0.38–2.81, and 0.39–2.86 in women, men, and the total population. These cut-off points are relatively higher compared to other studies [23, 30, 35]. The differences reflect differences in selection criteria for the reference cohort and the diversity of populations studied in terms of ethnicity. On the other hand, our upper limit of the HOMA-IR index (2.86) was similar to values reported by Geloneze et al. (2.71), who, however, considered the 90th percentile (we considered 95th percentile), also studying a normal glucose-tolerant non-obese Brazilian population [36]. Applying the same percentile used by Geloneze et al. to our population, we would obtain a HOMA-IR cut-off point of 2,63.
As seen with fasting insulin levels, it does not justify using sex-specific reference values for the HOMA-IR index since the ratio of standard deviations was < 1.5 (0.03), and the calculated z value was less than the criterion dictating partitioning of reference values [11].
Since MS, by definition, is associated with IR, our reference population consisted necessarily of subjects without MS. According to the National Cholesterol Education Program Adult Treatment Panel III (NCEP-ATP III), MS represents the combination of at least three of the five following variables: diabetes, hypertension, abdominal obesity, hypertriglyceridemia, and low HDL-c [37]. As we did have no access to information about abdominal circumference and pressure levels, our reference population had to have the absence of the other three criteria to ensure the exclusion of MS. We did have no access to abdominal circumference nor blood pressure levels, but we had access to self-reported weight and height and to medications in use. Therefore, we were able to exclude overweight and obese patients, and through access to the anti-hypertensive drugs in use, we could also exclude most hypertensive subjects, minimizing the presence of conditions associated with IR.
A concern with indirect methods has been that the RIs may be wider than they should be or skewed because of the inclusion of eventually unhealthy subjects. This was different in the present study. On the contrary, we observed a narrower RI than most manufacturers, probably because of the large number of subjects included (27). As discussed above, our RIs coincide with other studies but are quite different from RIs recommended by manufacturers of commercially available test kits, which even vary widely among themselves [38, 39, 40].
Finally, a potential bias to be considered is that there may have been some inconsistency between self-reported weight and height compared to measured values. However, studies have shown that although self-reported data may differ from measured ones, this difference usually has a negligible impact on BMI. Carvalho AM et al, studying 299 individuals living in São Paulo (112 men and 187 women), observed that differences between reported and measured weight and height were not significant for either gender or age groups [41]. Another potential bias is the self-reported use of medications that might be incomplete or incorrect in some cases, increasing the risk of including potentially unhealthy patients in the study and thus resulting in a falsely higher result for our insulin and HOMA-IR levels. However, considering the large number of individuals included in the present study, we believe this potential bias was also minimized.