Skip to main content
. Author manuscript; available in PMC: 2020 May 19.
Published in final edited form as: Diabetes Res Clin Pract. 2019 Feb 16;150:38–47. doi: 10.1016/j.diabres.2019.01.030

Table 2.

The association of log2(HOMA-IR) with cognition

Cognitive Function Beta±SE (p-value)
Total Learning Recall Post-Interference (Memory) Fluency Digit Symbol Substitution Test
Age −0.13±0.01 (<0.001) −0.06±0.05 (<0.001) 0.005±0.02 (0.81) −0.42±0.02 (<0.001)
log2(HOMA-IR) 0.02±0.11 (0.82) 0.008±0.05 (0.88) 0.18±0.16 (0.25) 0.36±0.22 (0.10)
Effect Modification of Age
Age −0.11±0.02 (<0.001) −0.06±0.01 (<0.001) 0.01±0.03 (0.61) −0.50±0.04 (<0.001)
log2(HOMA-IR) 0.84±0.72 (0.24) 0.38±0.37 (0.30) 0.70±0.97 (0.47) 2.83±1.50 (0.05)
Age*log2(HOMA-IR) −0.01±0.01 (0.24) −0.006±0.006 (0.31) −0.01±0.02 (0.60) 0.05±0.02 (0.02)

Note: log2(HOMA-IR) was log 2 transformed to approximate normality and improve model fit; thus, beta weights reflect the effect of doubling HOMA-IR on cognition. Analyses adjusted for age, sex, education, body mass index, total cholesterol, systolic blood pressure, C-reactive protein, smoking status, and language of test administration. Age is listed in the table given its interaction with log2(HOMA-IR) was central to determining effect modification; however, only significant results with log2(HOMA-IR) and age* log2(HOMA-IR) are denoted with bold text given these were the outcomes of interest.