The term ‘glucose effectiveness’ has been used to indicate and quantify the processes that occur to normalize the glucose concentration during hyperglycemia, independently from any change in insulin concentration [1]. In fact, glucose per se has the ability to stimulate its own uptake and suppress its own production even at basal insulin concentrations [1], contributing to about 60% of all glucose disposal in normal human subjects and animals [1,2,3]. The role of glucose effectiveness in predicting possible future derangements of glucose tolerance has gained increasing interest in the last years and the available evidence has been recently reviewed by Alford et al. [4], concluding that a lower glucose effectiveness over time is a critical determinant of future glucose intolerance and type 2 diabetes mellitus (T2DM) in subjects at increased risk. Recent evidence has also shown that overweight and obese subjects, who are at recognized increased risk of developing T2DM, exhibit lower glucose effectiveness [5]. On this basis, specific therapies aimed at enhancing glucose effectiveness in at-risk or intolerant subjects may play an important role in preventing T2DM [4]. Such treatments may include sodium glucose cotransporter 2 (SGLT2) inhibitors, recently proposed as antidiabetic pharmacologic agents, which exert their function by acting directly on non-insulin-mediated processes [6]. Thus, quantification of glucose effectiveness should be performed to have a complete description of the glucose tolerance status.

The history of gestational diabetes mellitus (GDM) is one of the most representative conditions leading to the increased risk for developing glucose intolerance and eventually T2DM [7], thus suggesting regular follow-ups for early detection of possible deterioration of glucose tolerance. Women with previous GDM (pGDM), among several metabolic abnormalities, may also exhibit lower glucose effectiveness [8]. However, to the best of our knowledge, no study has focused on the analysis of glucose effectiveness in this population; specifically, possible alterations in glucose effectiveness after a follow-up period have never been explored in pGDM.

Originally, two tests were employed to assess glucose effectiveness: the hyperglycemic glucose clamp and the intravenous glucose tolerance test (IVGTT) with minimal model interpretation [3,9]. The estimation of glucose effectiveness provided by the minimal model, known as the S_G index, has been proven to be reliable and equivalent to the one provided by the clamp technique [10]. Although less experimentally demanding compared to the clamp technique, the minimal model S_G assessment requires a 3-hour IVGTT, specific skills, and dedicated computer programs for model parameter estimation. To overcome these limitations, a simple index, calculated S_G (CS_G), has been recently proposed, allowing the reliable assessment of S_G without using complex mathematical models and resorting to a shorter—only 1-hour—IVGTT [11].

It has been shown that increasing the dynamics of the IVGTT may yield more accurate evaluation of the metabolic parameters. To this aim, an infusion of insulin can be added to the traditional IVGTT, a test termed the insulin-modified IVGTT (IM-IVGTT). This test is particularly recommended in subjects at increased risk of developing T2DM, such as women with pGDM, where the IM-IVGTT has been used to accurately assess possible defects in insulin sensitivity [12]. A simple index has already been proposed to assess insulin sensitivity [13] from the IM-IVGTT, but this has not been conducted for glucose effectiveness. Thus, the aims of this study were: (1) to derive—from a short IM-IVGTT—a simple surrogate marker (i.e., a predictor) of the S_G; (2) to test the reliability of the derived surrogate marker in a population of women with pGDM; and (3) to analyze possible alterations in glucose effectiveness in this population, at baseline and after a follow-up period.

The stepwise model selection showed that, among all the tested explanatory variables, K_G/G_peak was the only variable that had a significant role in predicting the minimal model S_G. Thus, the CS_G formulation was:

The related Bland-Altman plots (in the training dataset and in the jackknife-obtained validation dataset) (Fig. 1) showed a substantial agreement between CS_G and S_G, with less than 5% of the data falling outside the limits of agreement and no fixed bias affecting CS_G estimation (mean of S_G-CS_G not significantly different from zero). Furthermore, according to the TOST procedure, CS_G and S_G were shown to be equivalent both in the training and validation dataset, with small equivalence margins: epsilon was in fact equal to 0.1 standard deviation (SD) of S_G in both cases.

With regard to the pGDM group, the mean follow-up duration in the NONPROG subgroup was 5.2±0.1 years. In the PROG subgroup, the follow-up was interrupted at the onset of T2DM: the mean follow-up duration was 4.6±0.4 years. Notably, in our dataset, almost 80% (11 out of 14) of those women progressing to T2DM developed the disease within 5 years of follow-up.

In the pGDM group and in the related NONPROG and PROG subgroups, Bland-Altman plots at baseline and at follow-up (Figs. 2 and 3) showed substantial agreement between CS_G and S_G, again with less than 5% of the data falling outside the limits of agreement, with no fixed bias affecting CS_G estimation. Also, in the pGDM group, CS_G and S_G were shown to be equivalent according to the TOST procedure, both at baseline and at follow-up, with the equivalence margin epsilon equal to 0.2 SD and 0.3 SD, respectively. In the NONPROG and PROG subgroups, at baseline, the epsilon was equal to 0.3 and 0.5 SD, respectively; at follow-up, the epsilon was equal to 0.5 and 1 SD, respectively.

Glucose effectiveness values assessed by S_G and CS_G at baseline and at follow-up in the PROG and NONPROG subgroups are reported in Fig. 4. At baseline, the PROG subgroup showed significantly lower S_G and CS_G values compared to the NONPROG subgroup (P=0.01 and P<0.0001, respectively). Similar results were found at follow-up (P=0.03 and P<0.0001, respectively).

The present study proposed and validated a simple index for the assessment of CS_G that predicts the minimal model S_G, when a simple 1-hour IM-IVGTT is performed. IM-IVGTT may be preferred to the regular IVGTT in subjects with high insulin resistance, to obtain a more reliable estimation of this metabolic parameter [14,24,25,26]. The introduction of CS_G, together with the previously proposed calculated insulin sensitivity (CS_I) [13], overcomes the limitations of minimal model methodology, allowing a simple assessment of the two main processes regulating glucose disposal from a short, 1-hour, IM-IVGTT. The CS_G index proposed in this study exhibited excellent performances on the general population ranging from normal glucose tolerance to T2DM, as shown by the Bland-Altman plot (Fig. 1) and by the results of the test for equivalence (TOST), showing an equivalence margin parameter much lower than the SD of S_G.

The CS_G index was then tested in a population of women with a previous history of GDM, commonly recognized for an increased risk of developing T2DM. This population was analyzed early postpartum and then re-examined up to 7 years later. To the best of our knowledge, this is the first study to analyze, in detail, the glucose effectiveness in women with pGDM, coupled with a specific focus on the possible differences of this metabolic parameter in women progressing to T2DM during a follow-up period, compared to women that did not progress to T2DM. As detected by values at baseline and at follow-up, we found that glucose effectiveness was significantly and chronically reduced in those women progressing to T2DM compared to women remaining nondiabetic (Fig. 4). These findings suggest that in this population of women who progress to T2DM, alterations in glucose effectiveness, together with the already well-known defects in insulin resistance and secretion, could contribute to the development of T2DM. In fact, results obtained in previous studies showed that glucose effectiveness independently predicts conversion to T2DM across ethnic groups, family history of T2DM, and obesity [27]. Notably, a substantial equivalence between S_G and CS_G was proven again by the Bland-Altman plots (Figs. 2 and 3) and by the TOST analysis (showing equivalence margin always not higher than one SD), both at baseline and at follow-up.

In a previous study, CS_G was assessed from a regular 1-hour IVGTT [11]. Comparing the “regular” and the “insulin-modified” indices, one can notice that the formulation is essentially the same, i.e., the ratio between the rate of glucose disappearance, K_G, and the maximum of the glucose curve during the test, G_peak. This confirms the central role assumed by K_G/G_peak in the prediction of S_G. Conversely, other regressors (sex, age, BMI, basal glucose, and insulin concentrations), which were proven to be relevant variables in other studies [28], were not able to improve the S_G prediction.

The proposed CS_G index was designed and validated on a general population, including subjects with glucose tolerance abnormalities, ranging from defective glucose regulation to overt T2DM, but also healthy subjects. Insulin administration during IM-IVGTT can induce transient hypoglycemia and a consequent counterregulatory response in healthy insulin-sensitive subjects, thus determining the inaccuracy of minimal model-based parameter estimation [29]. However, it was shown that this inaccuracy pertains only to SI and not to S_G estimation [29], making inclusion of insulin-sensitive healthy subjects in CS_G design an advantage rather than a disadvantage.

Since this study represents a post hoc analysis of datasets collected from previous studies, IM-IVGTT data were obtained according to different insulin administration protocols (i.e., 0.03 or 0.05 U/kg at 20 minutes). The choice to include data derived from different protocols has been justified by the observation that using different insulin doses and protocols during the IM-IVGTT does not affect the minimal model estimation of S_G [30]. Also, K_G estimation is not affected by the insulin dose [30]. Moreover, the sampling protocol was not exactly the same in all studies. Nonetheless, the obtained results have proven the robustness of the proposed CS_G, which can be computed on the basis of somewhat different IVGTT protocols.

In this study, K_G was computed as the mean value between K_G1 and K_G2, i.e., the rates of glucose disappearance before (10 to 25 minutes) and after (25 to 50 minutes) the insulin infusion, as previously carried out for CS_I [13]. In the pGDM group, we analyzed the effect of replacing K_G with K_G1 in the equation (as above) for CS_G assessment. Since samples drawn before the 25-minute are not influenced by insulin infusion, using K_G1 rather than K_G would mean resorting to a regular, very short, IVGTT. In that condition, CS_G estimation was not equivalent to S_G (results not shown), corroborating the hypothesis that IM-IVGTT may be superior to regular IVGTT; thus, it should be preferred when possible.

Important factors differentiating women at highest risk of progression to T2DM among all the pGDM women are family history of diabetes, certain ethnicities, age, early onset of GDM, hypertension during pregnancy, high BMI, insulin resistance, impaired β-cell function [31]. However, other possible markers for later risk of T2DM have been hypothesized recently. For instance, an increased circulating fatty acid-binding protein 4 (FABP4) has been shown to differentiate between women progressing to and those not progressing to T2DM after GDM [32]. In fact, circulating FABP4 has pleiotropic roles that include the stimulation of hepatic glucose production [33]. On the other hand, glucose effectiveness is defined as the ability of glucose per se to stimulate its own uptake, but also to suppress its own production [1]. Thus, the lower glucose effectiveness characterizing the pGDM women progressing to T2DM may reflect, among other factors, increased levels of FABP4, which have shown to be a condition of risk for T2DM progression [32]. However, further studies are required to corroborate this hypothesis.

It is also worth noting that all the subjects included in this study were white Caucasians, and hence, the application of CS_G to assess glucose effectiveness in other ethnicities (e.g., Asian subjects) has to be tested.

In conclusion, this study developed a simple index for the assessment of glucose effectiveness when an IM-IVGTT is performed. Testing the new index, CS_G, in a population of women with a previous history of GDM confirmed the index reliability in a category of subjects at a particularly high risk for developing T2DM. Since glucose effectiveness appears to be impaired in this population of subjects progressing to T2DM development, its simple and reliable quantification by CS_G may facilitate a timely assessment of this aspect of glucose metabolism, thus possibly prompting corrective actions that have now become available with pharmacological agents, such as SGLT2 inhibitors, which act independently on insulin.