Lipoprotein insulin resistance (LPIR) score is a surrogate index of insulin resistance (IR) that predicts incident type 2 diabetes mellitus in epidemiological studies. LPIR is derived from six plasma lipoprotein parameters with weighting scores originally estimated by their relationships with HOMA-IR. HOMA-IR may be limited in its ability to predict IR in African Americans (AA) who also have a metabolically favorable lipoprotein profile. The reliability of LPIR score to predict IR in AA is unknown. In this study, we used a calibration model to evaluate the differences in the relationship of LPIR with insulin sensitivity (Si) in AA and non-Hispanic Whites (NHW).


In a cross-sectional study, non-diabetic AA (n=49; age 36 ±10 years; BMI 31.5±7.2 kg/m2) and NHW (n=74; age 36±12 yr; BMI 31.5±7.7 kg/m2) subjects underwent an intravenous glucose tolerance test (IVGTT) to derive Si using the Minimal Model. Lipoprotein (HDL, LDL, VLDL) particle concentrations and sizes were measured using nuclear magnetic resonance. A calibration model was used to assess the ability of LPIR to predict Si in AA and NHW. The log inverse of LPIR was regressed on the reference logSi to generate a LPIR-predicted Si, which was compared to the observed Si to evaluate its predictive ability.


Age, sex, and % body fat-adjusted Si and LPIR were lower in AA when compared with NHW (Si: 2.86±0.41 vs. 4.44±0.33, p=0.003; LPIR: 35.6±3.0 vs. 45.7±2.5, p=0.01), suggesting a discordance between LPIR and Si. LPIR was inversely related in both ethnic groups (AA: r=-0.36; NHW: r=-0.46, p<0.001). Deming regression suggests a systematic underestimation of LPIR in AA compared to NHW (p=0.03). The linear least-squares fit between LPIR-predicted Si and measured SI showed a modest correlation in both ethnicities, with a lower correlation in AA than in NHW.


LPIR underestimates IR in non-diabetic AAs as compared to NHWs. This suggests that LPIR-scores as a measure of IR should be interpreted with caution in AAs.