Clues to occult cancer in patients with ischemic stroke

Abstract

Background: We hypothesized that hidden malignancy could be detected in patients with cryptogenic stroke without active cancer when they showed the distinctive characteristics of cancer-related stroke.

Methods and findings: Among 2,562 consecutive patients with acute ischemic stroke, patients with cryptogenic stroke were analyzed and categorized into two groups according to the presence of active cancer: cryptogenic stroke with active cancer (cancer-related stroke, CA-stroke) group and without active cancer (CR-stroke) group. Patients with active lung cancer without stroke were also recruited for comparison purposes (CA-control). Clinical factors, lesion patterns on diffusion-weighted MRI (DWI), and laboratory findings were analyzed among groups. A total of 348 patients with cryptogenic stroke were enrolled in this study. Among them, 71 (20.4%) patients had active cancer at the time of stroke. The D-dimer levels were significantly higher in patients with CA-stroke than those with CR-stroke or CA-control (both p<0.001). Regarding lesion patterns, patients with CA-stroke mostly had multiple lesions in multiple vascular territories, while more than 80% of patients with CR-stroke had single/multiple lesions in a single vascular territory (P<0.001). D-dimer levels (OR 1.11 per 1 µg/mL increase; 95% CI 1.06-1.15; P<0.001) and DWI lesion patterns (OR 7.13; 95% CI 3.42-14.87; P<0.001) were independently associated with CA-stroke. Workup for hidden malignancy was performed during hospitalization in 10 patients who showed elevated D-dimer levels and multiple infarcts involving multiple vascular territories but had no known cancer, and it revealed hidden malignancies in all the patients.

Conclusion: Patients with CA-stroke have distinctive D-dimer levels and lesion patterns. These characteristics can serve as clues to occult cancer in patients with cryptogenic stroke.

Figure 1. A typical example of cryptogenic stroke with active cancer (cancer-related stroke).

The patient was a 55-year-old female, who was diagnosed with metastatic adenocarcinoma of lung one year ago. Neurologic examination revealed motor aphasia, right-central facial paralysis, and right hemiparesis. Initial D-dimer levels were 20.56 µg/mL. Diffusion-weighted MRI (A) shows multiple lesions involving multiple vascular territories. MR angiography (B) reveals no significant stenosis or occlusion of the craniocervical arterial vasculature. Comprehensive workup including transesophageal echocardiogram and 24-hour Holter monitoring demonstrated no sources of cardioembolism.

Figure 2. Distribution of plasma D-dimer levels in patients with cancer-related stroke (N = 71), cryptogenic stroke without active cancer (N = 277), and active lung cancer without a history of stroke (N = 33).

D-dimer levels were significantly higher in the cancer-related stroke group than in the cryptogenic stroke without active cancer or lung cancer-control groups (both P<0.001).

Figure 3. Receiver operating characteristic (ROC) curve for predicting cancer-related stroke from plasma D-dimer levels.

The area under the ROC curve (AUC) ± standard error was 0.945±0.182 (P<0.001). The dotted lines indicate 95% confidence interval (0.916 to 0.967).