Background: Stroke mortality comprises different specific diagnoses as cerebral infarction, different haemorrhagic conditions and unspecified stroke. This study seeks to explore the prediction of oral health indicators versus known cardiovascular disease risk factors for stroke mortality. Methods: Altogether, 12,764 men aged 58 to 77 years were invited to the health screening Oslo II in the year 2000. It included general medical measurements and questionnaire information. Mortality data were supplied by Statistics Norway for the 6530 attending men. Cox proportional hazards regression analyses were used to establish prediction models for mortality. Results: Oral health by number of tooth extractions >10 was found to be an independent predictor for cerebral infarction hazard ratio = 2.92, 95% confidence interval (1.24–6.89). This was independent of HDL-Cholesterol (inversely) hazard ratio = 0.21, 95% confidence interval (0.06–0.76), frequent alcohol consumption (drinking 4–7 times per week) hazard ratio = 3.58, 95% confidence interval (1.40–9.13) and diabetes hazard ratio = 4.28, 95% confidence interval (1.68–10.89). Predictors for cerebral haemorrhage were age, hs-C-reactive protein and body mass index (inversely). Age and total cholesterol (inversely) were predictors for unspecified stroke. Conclusions: Oral health measured by number of tooth extractions >10 was an independent predictor for cerebral infarction in addition to age, HDL-C, hs-C-reactive protein and diabetes. The pattern of risk factors varied between the specific stroke diagnoses.