Identifying ischemic stroke associated with cancer: a multiple model derived from a case-control analysis

Ischemic stroke in patients with cancer is thought to be associated with a worse prognosis and might be the initial symptom of an unknown malignancy. However, diagnostic algorithms to reliably identify cancer-associated stroke have not been developed. In this retrospective single-centre analysis, 68 patients with ischemic stroke and an active solid malignancy were identified. Neurological assessment and outcome, cardiovascular risk factors, neuroimaging studies as well as laboratory findings were compared to 68 age- and sex-matched control subjects with ischemic stroke without diagnosis of cancer. Lung, pancreatic and renal cancer showed increased prevalences compared to those of the general population in Germany. Diagnosis of cancer was most often made within the 12 months preceding (32.4%) or during the diagnostic work-up for stroke (17.7%). Cancer-associated stroke was characterized by a more severe clinical deficit, frequent clinical deterioration (13.2 vs. 1.5%) or death (25 vs. 4.4%). Ischemic lesions often involved multiple territories (51.6 vs. 12.7%), more often with co-existing subacute and acute infarctions in imaging studies (54.8 vs. 11.1%). Patients with cancer had significantly higher levels of C-reactive protein, relative granulocytosis and serum lactate dehydrogenase activity. Using receiver operating characteristics-based multiple analysis, we developed a model using these parameters which detected cancer-associated stroke with a sensitivity of 75% and specificity of 95%. Our analysis suggests that a multiple algorithm combining the number of territories involved and laboratory signs of inflammation and cell turnover might identify patients with stroke suffering from previously unknown malignancy.