Protective effect of leptin on recurrent VTE and death persists after adjustment by potential confounding. Shown in this figure are the HRs with 95% CIs for a standard deviation increase in leptin concentrations from 4 Cox regression models, all predicting the combined recurrent VTE or death end point. All models are adjusted for age, sex, cardiovascular disease, history of VTE, diabetes, arterial hypertension, medication intake at baseline (cardiovascular medication, antithrombotic therapy, corticosteroids, contraceptives, NSAIDs), and the BMI-associated proteomic signature. All models except the first (denoted by BMI-unadjusted) are additionally adjusted for continuous BMI. The third and fourth models contain interaction terms. Low/high BMI and low/high MMP-2 were defined based on median splits.