The aim of this study was to evaluate clinical, adrenergic and endocrine factors that could predict sinus rhythm maintenance after direct current cardioversion in chronic atrial fibrillation. Nineteen patients with chronic non-rheumatic atrial fibrillation (mean duration 6 +/- 5 months) were studied. They were exercised 24 h before cardioversion at maximum effort with the Naughton protocol. Heart rate and blood pressure at rest and exercise were recorded and blood samples were taken for the assessment of adrenergic activity, by measuring cyclic adenosine monophosphate, heart endocrine function, atrial natriuretic peptide and its second messenger, cyclic guanosine monophosphate. Fifteen of the 19 patients were initially converted to sinus rhythm (eight patients with external and seven patients with internal DC shocks). After 3 months eight patients remained in sinus rhythm and 11 had relapsed, most of them within the first month. On exercise the chronotropic response was lower in the group who remained in sinus rhythm than in the group in atrial fibrillation (peak heart rate 147 +/- 11 beats.min-1 vs 165 +/- 24 beats.min-1 P = 0.02). During exercise, the systolic blood pressure in the sinus group reached higher values than in the group who relapsed (192 +/- 17 mmHg vs 176 +/- 18 mmHg, P = 0.03). Cyclic adenosine monophosphate increased significantly from rest to peak exercise in the sinus rhythm group (from 23 +/- 9 pmol.ml-1 to 31 +/- 15 mol.ml-1, P = 0.02) while it remained unchanged in the atrial fibrillation group (25 +/- 10 pmol.ml-1 to 24 +/- 8 pmol.ml-1, P = 0.02). For all 19 patients the difference in cyclic adenosine monophosphate between rest and exercise was negatively correlated with maximum heart rate (r = 0.58, P = 0.009). Atrial natriuretic peptide increased from rest to peak exercise in the sinus rhythm group (from 129 +/- 58 fmol.ml-1 to 140 +/- 66 fmol.ml-1) while it remained unchanged in the group in which atrial fibrillation persisted or recurred (from 112 +/- 58 fmol.ml-1 to 111 +/- 53 fmol.ml-1, P = 0.002). A significant correlation between atrial natriuretic peptide and cyclic guanosine monophosphate levels at exercise before cardioversion was found for the sinus rhythm group only (r = 0.76, P = 0.02). In patients with non-rheumatic chronic atrial fibrillation evaluation of clinical parameters such as heart rate and blood pressure changes during maximal exercise can be useful in the choice of suitable therapy. An inadequate increase in plasma cyclic-adenosine monophosphate and atrial natriuretic peptide on exercise could predict patients with more severe underlying disease, where cardioversion should not be recommended.