Table 5. Longitudinal biomarker measurements: Hazard ratios for incident HF estimated by joint models.

	HR per IQR (95% CI)	p-value
GDF-15	2.12 (1.54, 2.92)	<0.001
CRP	2.27 (1.64, 3.15)	<0.001
Cystatin C	1.14 (0.91,1.45)	0.25

The interquartile range (IQR) of the round 1 biomarker measurement is used. Biomarkers are used after log-transformation. The hazard ratio is derived from the proportional hazards model part of the joint model and it measures the association of the time-course of the corresponding biomarker with time to HF. The proportional hazards model part of the joint model was adjusted for age, sex, overweight (BMI > 25 kg/m²), systolic blood pressure, diabetes, daily smoker and renal insufficiency (eGFR > 60 ml/min or 1.73m³).