Type 2 diabetes (T2D) prevalence increases unabated even as interventions focused on individuals at high risk can prevent T2D (1). Best approaches to identify individuals at high risk continue to be refined. Current risk assessment focuses on elevated glycemia, commonly estimated by fasting glucose (FG) and HbA1c levels, despite imperfect sensitivity and specificity (2). We previously showed that risk for T2D increases progressively as HbA1c increases, independently and in addition to increasing FG (3,4). As risk assessment for T2D depends on more than glycemia alone, simultaneous consideration of another physiological axis may improve biomarker-based diagnostic precision. Here, we test the hypothesis that simultaneous consideration of HbA1c and insulin resistance (IR), assessed with fasting insulin as HOMA of IR (HOMA-IR), can substantially improve risk assessment for T2D.
We have previously detailed our statistical approach, IR and T2D diagnostic criteria, and particulars of the Framingham Heart Study (FHS) (3). Using the same FHS data and excluding those with T2D at baseline, we categorized individuals according to HbA1c <5.7% or 5.7–6.49% and into HOMA-IR tertiles and followed them for a mean (SD) of 16.4 (4.5) years for incident T2D.
Age- and sex-adjusted T2D incidence rates and counts of subjects are shown in Fig. 1A for all 2,205 study subjects and 1,583 normoglycemic subjects with FG <100 mg/dL. Incidence was high in those with HbA1c 5.7–6.49% and HOMA-IR in the top two tertiles, or with HbA1c <5.7% and HOMA-IR in the top tertile, relative to those in lower categories. Age- and sex-adjusted models for those with HbA1c 5.7–6.49% compared with those with HbA1c <5.7% and for HOMA-IR tertile 3 versus tertile 2 or 1 are shown in Fig. 1B. Elevated HbA1c and HOMA-IR independently predicted T2D in all and in normoglycemic subjects. The association between HbA1c and risk of T2D did not differ according to HOMA-IR tertile (all first-order interactions P > 0.1). Area under the receiver operating characteristic curves (AUC) and continuous net reclassification indices are shown in Fig. 1C. Addition of HOMA-IR to age- and sex-adjusted HbA1c prediction models in all subjects, and even in normoglycemic subjects, significantly improved correct classification of T2D risk.
![Figure 1. A: Risk for incident T2D increases with increasing HbA1c and HOMA-IR category. The graph shows the predicted probability of incident T2D (y-axis), by tertiles of HOMA-IR stratified by HbA1c <5.7% or 5.7% to <6.5% (x-axis), for all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). The inset numbers indicate the number of T2D events/the total sample size in each category group for all study subjects (black font) and among those with FG <100 mg/dL (blue font). The box plots represent the first quartile (lower hinge), median, and third quartile (upper hinge) of the risk distribution, and the whiskers indicate 1.5 times the interquartile range. B: Elevated HbA1c and HOMA-IR are independent risk factors for incident T2D. The graph shows odds ratios, 95% CIs, and P values for terms for HbA1c <5.7% vs. 5.7% to <6.5%, HOMA-IR tertile 3 vs. tertile 1, and HOMA-IR tertile 2 vs. tertile 1 from a model containing age, sex, HbA1c category, and HOMA-IR category, for models of all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). C: HOMA-IR improves discrimination and reclassification when added to HbA1c in prediction models. The graph shows AUC, with sensitivity on the y-axis and 1 − specificity on the x-axis, for age- and sex-adjusted regression models predicting incident T2D that include categorical HbA1c (solid lines) or HbA1c plus HOMA-IR (dashed lines), for all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). The inset shows the value of AUC for the HbA1c and HbA1c plus HOMA-IR models, the difference [ΔAUC (95% CI)] between those AUCs, the continuous net reclassification indices, and the proportion of T2D events and nonevents correctly reclassified with addition of HOMA-IR to HbA1c prediction models for all study subjects (black font) and among those with FG <100 mg/dL (blue font).](https://ada.silverchair-cdn.com/ada/content_public/journal/care/43/8/10.2337_dc20-0718/5/m_dc200718f1.jpeg?Expires=1730235740&Signature=mP590NMjaY6xY4Ce4jb07f~EVJGNgLVXNrT5t5sbfyUk8pzcnyNybk7sSv0CnAM4CDfAlgFYc6dGn1tBvoNQPVHik5Zualo6RTQ6YQYqiU6AVDndHXsfhlRXwUcp57G35KNt2VfJx0-hyHI~Wjmk44c84Xa58qVAyNWdk1S5uvGPU7oRCYUSIHv4-uDPi3J91tGxCOwBu34Vnv18VjMHIUUc7z0f2VnKGMFn8Yc8URP6oSMfEExzdePihHUmOalLAn3WoHBEU0qVv4BgzLymV7-q8REkFAQdWEckBFhnnmyl0EI-t6vXMxX7026aiuSXEYMEdaJ-LphKQo4kfmRd5A__&Key-Pair-Id=APKAIE5G5CRDK6RD3PGA)
A: Risk for incident T2D increases with increasing HbA1c and HOMA-IR category. The graph shows the predicted probability of incident T2D (y-axis), by tertiles of HOMA-IR stratified by HbA1c <5.7% or 5.7% to <6.5% (x-axis), for all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). The inset numbers indicate the number of T2D events/the total sample size in each category group for all study subjects (black font) and among those with FG <100 mg/dL (blue font). The box plots represent the first quartile (lower hinge), median, and third quartile (upper hinge) of the risk distribution, and the whiskers indicate 1.5 times the interquartile range. B: Elevated HbA1c and HOMA-IR are independent risk factors for incident T2D. The graph shows odds ratios, 95% CIs, and P values for terms for HbA1c <5.7% vs. 5.7% to <6.5%, HOMA-IR tertile 3 vs. tertile 1, and HOMA-IR tertile 2 vs. tertile 1 from a model containing age, sex, HbA1c category, and HOMA-IR category, for models of all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). C: HOMA-IR improves discrimination and reclassification when added to HbA1c in prediction models. The graph shows AUC, with sensitivity on the y-axis and 1 − specificity on the x-axis, for age- and sex-adjusted regression models predicting incident T2D that include categorical HbA1c (solid lines) or HbA1c plus HOMA-IR (dashed lines), for all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). The inset shows the value of AUC for the HbA1c and HbA1c plus HOMA-IR models, the difference [ΔAUC (95% CI)] between those AUCs, the continuous net reclassification indices, and the proportion of T2D events and nonevents correctly reclassified with addition of HOMA-IR to HbA1c prediction models for all study subjects (black font) and among those with FG <100 mg/dL (blue font).
A: Risk for incident T2D increases with increasing HbA1c and HOMA-IR category. The graph shows the predicted probability of incident T2D (y-axis), by tertiles of HOMA-IR stratified by HbA1c <5.7% or 5.7% to <6.5% (x-axis), for all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). The inset numbers indicate the number of T2D events/the total sample size in each category group for all study subjects (black font) and among those with FG <100 mg/dL (blue font). The box plots represent the first quartile (lower hinge), median, and third quartile (upper hinge) of the risk distribution, and the whiskers indicate 1.5 times the interquartile range. B: Elevated HbA1c and HOMA-IR are independent risk factors for incident T2D. The graph shows odds ratios, 95% CIs, and P values for terms for HbA1c <5.7% vs. 5.7% to <6.5%, HOMA-IR tertile 3 vs. tertile 1, and HOMA-IR tertile 2 vs. tertile 1 from a model containing age, sex, HbA1c category, and HOMA-IR category, for models of all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). C: HOMA-IR improves discrimination and reclassification when added to HbA1c in prediction models. The graph shows AUC, with sensitivity on the y-axis and 1 − specificity on the x-axis, for age- and sex-adjusted regression models predicting incident T2D that include categorical HbA1c (solid lines) or HbA1c plus HOMA-IR (dashed lines), for all study subjects (black data series) and among those with FG <100 mg/dL (blue data series). The inset shows the value of AUC for the HbA1c and HbA1c plus HOMA-IR models, the difference [ΔAUC (95% CI)] between those AUCs, the continuous net reclassification indices, and the proportion of T2D events and nonevents correctly reclassified with addition of HOMA-IR to HbA1c prediction models for all study subjects (black font) and among those with FG <100 mg/dL (blue font).
Strengths of this study include testing of a relevant clinical question using a well-validated approach in a well-known cohort. Limitations include study of white individuals only and small numbers of T2D events in some subgroups. Also, a more precise surrogate marker of IR than HOMA-IR would identify a greater proportion of individuals at high risk, but most measures require at least an oral glucose tolerance test. For instance, fasting insulin and C-peptide measured with mass spectrometry afford an alternate simple approach to classify IR (5).
HbA1c and fasting insulin are both commonly available clinical diagnostic tests. Consideration of both tests combined identifies highly increased risk for future T2D in the great majority of white individuals, even those with apparently normal HbA1c.
Article Information
Acknowledgments. The authors thank the staff and participants of FHS for their important contributions.
Funding. The FHS is supported by National Heart, Lung, and Blood Institute contracts N01-HC-25195 and HHSN268201500001I.
Duality of Interest. Quest Diagnostics supported this study. J.J.D., D.S., and M.J.M., from Quest Diagnostics, reviewed and commented on the manuscript. J.B.M. serves as an Academic Associate for Quest Diagnostics. No other potential conflicts of interest relevant to this article were reported.
J.J.D., D.S., and M.J.M. had no role in developing the scientific hypothesis or study design, performing the analysis, or compiling results.
Author Contributions. J.B.M. developed the scientific hypothesis, study design, and analysis plan; compiled results; and drafted the manuscript. B.P. and J.B.M. contributed to the scientific hypothesis, study design, and analysis plan. B.P. and A.L. performed analyses and prepared the figure. J.B.M., B.P., A.L., D.S., J.J.D., and M.J.M. reviewed all aspects of the manuscript and approved of the final version of the manuscript. J.B.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
