To investigate patterns of prognostic associations over time of insulin- and obesity-related variables measured at diagnosis of early breast cancer (BC), focusing on whether the prognostic associations with distant recurrence and death changed over time. Five hundred thirty-five nondiabetic women with T1-3, N0-1, M0 invasive BC diagnosed from 1989 to 1996 were included in the study. Insulin-related variables included fasting insulin, Homeostasis Model Assessment, C-peptide, and glucose. Obesity-related variables included weight, body mass index (BMI), waist and hip circumference, and leptin. Correlations were examined using the Pearson correlation coefficient and prognostic associations using the Cox model. There was evidence that associations of baseline insulin-related variables with distant recurrence and death were not constant over time; univariable adverse prognostic associations were significant only during the first 5 years (eg, insulin quartile 4 v 1: hazard ratio HR, 2.32; 95% CI, 1.39 to 3.86; P < .001 for distant disease-free survival DDFS; and HR, 2.85; 95% CI, 1.48 to 5.50; P = .002 for overall survival OS, with little attenuation of this pattern in multivariable analyses). In contrast, obesity-related variables (BMI, weight, leptin) exerted significant adverse univariable associations that were constant over time (eg, BMI quartile 4 v 2: HR, 1.40; 95% CI, 1.07 to 1.82 for DDFS; P = .014; and HR, 1.50; 95% CI, 1.16 to 1.93; P < .001 for OS); prognostic associations of leptin remained significant in multivariable analyses. Baseline insulin- and obesity-related variables exert different patterns of prognostic associations over time in early BC.