Insulin Resistance and Systemic Inflammation Below the HbA1c Prediabetes Threshold: Age-Dependent Associations in U.S. Adults, NHANES 2015–March 2020

Background: Insulin resistance may be present despite glycemic measures remaining below conventional thresholds for prediabetes, but whether its relationship with systemic inflammation varies across adulthood is unclear. This study examined the association between homeostatic model assessment of insulin resistance (HOMA-IR) and high-sensitivity C-reactive protein (hs-CRP) among U.S. adults without diagnosed diabetes and with HbA1c <5.7%, with particular attention to modification of this association by age. Methods and results: We analyzed 3,079 adults aged ≥20 years from the pre-pandemic NHANES 2015–March 2020 cycles using survey-weighted regression with log-transformed hs-CRP as the outcome. The adjusted HOMA-IR–hs-CRP association differed across age groups (20–39, 40–54, and ≥55 years; interaction F(2,40)=5.69, p=0.00671). Positive adjusted associations were observed in younger and middle-aged adults, whereas no positive association was observed among adults aged ≥55 years. Weighted median HOMA-IR values were similar across age groups, suggesting that the age pattern reflected differences in the relationship between insulin resistance and inflammation rather than large differences in HOMA-IR levels themselves. Sensitivity and replication analyses broadly supported the qualitative pattern, although the strength of formal interaction evidence varied across model specifications. Archive contents: This record contains the study manuscript, supplementary material, and the archived v1.0.0 computational release. The computational package includes the reproducible Python analysis pipeline, aggregate analysis outputs, and an independent R implementation used for cross-software validation. Raw NHANES source files and derived participant-level datasets are not redistributed; the required source data are publicly available from the National Center for Health Statistics. The manuscript and supplementary material are made available under CC BY 4.0. Computational code is distributed under the MIT License as specified in the archived repository.