Abstract Background Atrial structural abnormalities linked to the risk of atrial fibrillation (AF), as indicated by P-wave indices on surface ECG, are associated with the fibrotic transformation of the atrial myocardium. To what extent incident AF can be predicted by ECG characteristics in patients with advanced atrial structural remodeling, such as patients with heart failure (HF), is less well studied. Purpose To evaluate the association between P-wave indices and biomarkers implicated in cardiac fibrosis, and incident AF in patients with HF. Methods Patients with new-onset or worsening of HF were prospectively recruited in a observational study (n=501) and followed-up for 5 years. Blood samples were collected for analysis using proximity extension assay that included a subset of biomarkers that were associated with fibrosis development (TIMP-2, MMP-2, MMP-3, MMP-9, ST-2, GDF-15, Gal-3). All ECGs ever recorded in the hospital catchment area were exported in digital format and processed using Glasgow algorithm for calculation of the P-wave indices (P-wave duration [PWD], P-wave terminal force in V1 [PTFV1] and P-wave axis [Paxis]) and rhythm identification. AF diagnosis was ascertained by manual ECG review. Only patients who did not have AF reported or ECG documented at enrolment were included in the analysis. Of those subjects, 121 patients also had biomarkers of fibrosis analyzed. Cox-regression analyses were used to assess biomarkers’ and p-wave indices’ associations with incident AF. Results After exclusion of patients with prevalent AF (n=317), 184 patients comprised the study group (mean age of 71 years, 33.2% women, 121 patients with biomarker data available). During follow-up, incident AF was observed in 39 patients. At enrolment, PWD>120 ms was observed in 38.7%, abnormal PTFV1>40 mm*ms in 24.5% and Paxis<0° in 5.4%. Among the seven biomarkers assessed, five demonstrated significant associations with incident AF in adjusted analyses: TIMP4 (HR 1.78; 95%CI 1.03-3.05; p=0.037); MMP2 (HR 1.97; 95%CI 1.03-3.80; p=0.042), MMP3 (HR 1.39; 95%CI 1.03-1.88; p=0.031); ST2 (HR 1.91; 95%CI 1.29-2.83; p=0.001) and GDF-15 (HR 2.56; 95%CI 1.50-4.35; p=0.001) (Figure 1). However, none of the tested P-wave variables (P-wave duration, P-wave terminal force, and P-wave axis) were associated to incident AF. Conclusions In this prospective long-term follow-up study of patients admitted for acute HF, biomarkers with proven associations to fibrosis were strongly associated with incident AF. P-wave indices used for prediction of AF in epidemiological cohorts appear to have limited value in patients with HF who have high prevalence of ECG atrial abnormalities at baseline.