Quantitative data on melatonin in stroke patients are scarce. A gender- and age-matched cross-sectional case-control study in 33 patients with ischaemic stroke was performed and associations between nocturnal melatonin and other factors (e.g. cortisol) were evaluated. Clinical and laboratory (e.g. melatonin and cortisol) measurements (03.00 h and 08.00 h) with statistical techniques [e.g. multifactorial regressions, receiver operating characteristic (ROC) curve and curvilinear estimations] were used. We identified mean value and 95% confidence interval (CI) (69.70 pg/ml; 95% CI = 53.86-85.54) for control levels of nocturnal melatonin in healthy subjects. The patients with stroke had lower melatonin (48.1 +/- 35.9 pg/ml) and higher cortisol (297.3 +/- 157.8 nmol/l) at 03.00 h (P < 0.05) but not at 08.00 h (P > 0.05). Stroke was the strongest factor of disturbed nocturnal cortisol (P < 0.001), whereas decreased melatonin depended on stroke (P = 0.010) and gender (P = 0.018). At the same time, vice versa, only nocturnal measures were associated with an increased probability of the presence of stroke (accuracy > 75%, Pmodel < 0.001). Thus, a hypothesis that a decrease of melatonin with 1.0 pg/ml might be associated with > 2% increase in the probability of the presence of stroke [adjusted odds ratio (OR) = 1.020; 95% CI = 1.002-1.037] was also suggested. The ROC curve (0.67, P = 0.0119) and optimisation techniques indicated that a novel best cut-off < 51.5 pg/ml for decreased nocturnal melatonin in the view of the presence of stroke (OR = 3.12, P = 0.0463) might exist. The classification performance of such a cut-off might be confirmed by existing nocturnal melatonin and cortisol differences between the sub-groups; potential differences in diurnal melatonin were also suggested. In conclusion, a novel melatonin cut-off of 51.5 pg/ml may be associated with the presence of ischaemic stroke. As a single marker (84% sensitivity, 74% specificity), it is hypothesised that modelling performance was independent of age, gender and cortisol. These new results, including the suggested hypothesis, might be further tested in follow-up (cohort), longitudinal studies and be applied to explore melatonin disturbances as targets in high-risk pre-stroke and post-stroke patients.