We conducted a retrospective cohort study of consecutive patients treated for AIS at a comprehensive stroke center in Bern, Switzerland between January 1, 2015 and December 31, 2020. These patients were prospectively enrolled in the Swiss Stroke Registry, which was accessed for data analysis.

The study was approved by the local ethics committee in accordance with Swiss regulations (project ID: 2022-01560; Kantonale Ethikkommission Bern), and the requirements for written consent was waived. Access to the study data can be requested from the corresponding author and is subject to clearance by the local ethics committee. This analysis adhered to the STROBE checklist guidelines for cohort studies.

The study population consisted of patients with AIS and known active cancer at the time of AIS or new cancer diagnosed during the hospitalization for the index AIS. Active cancer was defined according to the criteria recommended by the International Society on Thrombosis and Haemostasis (11). This comprised a new or recurrent cancer that was diagnosed or treated within six months prior to the index AIS, or known metastatic cancer. Hematological malignancies that were not in complete remission for more than 5 years were also considered active. Patients diagnosed with a new cancer during the index hospitalization were considered to have occult cancer at the time of AIS, and were included in the known active cancer group for analyses (12, 13). We excluded patients who were diagnosed with cancer after hospital discharge for the index AIS, as this sequence of events may have influenced antithrombotic management decisions and clinical outcomes.

We excluded patients with focal non-melanoma skin cancer because of their low risk of dissemination as well as patients with breast cancer in complete remission who were receiving maintenance hormonal therapy (11, 14, 15). Other exclusion criteria were (i) death during the index hospitalization or missing follow-up data regarding vital status, (ii) no antithrombotic therapy prescription at discharge, and (iii) a cardioembolic stroke mechanism at discharge necessitating anticoagulation (e.g., atrial fibrillation, mechanical valve). Patients with a determined stroke mechanism typically treated with antiplatelet therapy, such as large artery atherosclerosis, were not excluded because these patients may still have cancer-mediated hypercoagulability and could preferentially benefit from anticoagulant therapy.

Demographic and clinical characteristic data were collected from the Swiss Stroke Registry and electronic health records. These included age at admission, sex, pre-stroke functional status (independency defined as a pre-stroke modified Rankin Scale [mRS] score ≤2), baseline imaging modality, and history of cardiovascular risk factors (prior stroke, hypertension, diabetes mellitus, hyperlipidemia, smoking history, and atrial fibrillation). The presence of multi-territory brain infarcts (involving at least two different cerebrovascular territories) was determined using data from baseline neuroradiological imaging. Data on the primary cancer type, stage (16), and presence of metastasis at the time of AIS was collected from electronic health records. Laboratory measurements included D-dimer, C-reactive protein (CRP), hemoglobin, platelet count, fibrinogen, and lactate dehydrogenase (LDH). For patients with multiple measurements during the index AIS hospitalization, the baseline value was recorded. Two neurologists determined the AIS mechanism at discharge, classified according to the Trial of Org 10,172 in Acute Stroke Treatment (TOAST) criteria and the embolic stroke of undetermined source (ESUS) classification (17, 18). Recent data have indicated that among patients with AIS, there is an inverse association between the presence of a right-to-left shunt (patent foramen ovale [PFO] or atrial septal defect [ASD]) and the presence of cancer, suggesting that paradoxical embolism is not a major cause of AIS in patients with cancer (19). Therefore, we classified patients with PFO/ASD and no other apparent cause of AIS as “undetermined stroke etiology” and, if meeting criteria, as ESUS. The diagnosis of venous thromboembolism (VTE), including deep vein thrombosis and/or pulmonary embolism, in the year before and after the index AIS was recorded.

The study exposure was the type of antithrombotic therapy (anticoagulant or antiplatelet therapy) prescribed at hospital discharge from the index AIS. The anticoagulant group consisted of patients treated with therapeutic doses of any oral or parenteral anticoagulant. These included vitamin K antagonists, low-molecular-weight heparins (LMWH) such as enoxaparin or tinzaparin, and direct oral anticoagulants (DOACs) such as edoxaban, rivaroxaban, dabigatran, or apixaban. Patients treated with recommended dosing regimens adjusted for age, weight, or renal function were classified as receiving therapeutic anticoagulation. The antiplatelet group included standard-dose aspirin, clopidogrel, or both. Patients treated with both therapeutic-dose anticoagulant and antiplatelet therapy were included in the anticoagulant group. The treatment decision was left to the treating physician. According to our internal stroke guidelines, if D-dimer levels at baseline exceed 3,000 μg/L and cancer is present, paraneoplastic coagulopathy should be considered as AIS etiology. When a paraneoplastic coagulopathy is suspected, anticoagulation with LMWH at therapeutic dosage (twice daily) or oral anticoagulants should be considered. As tolerance to LMWH is known to be low in cancer-related strokes (in opposition to patients with VTE), we reported continuation of LMWH prescribed at discharge versus regimen change for secondary prevention at the 3-month clinical follow-up (20, 21).

The primary study outcome was mortality at 1 year after the index AIS. To determine patients’ vital status, we analyzed data from the Swiss Population Registry, which records the vital status of Swiss residents on a monthly basis. Secondary outcomes were (i) a good functional outcome at 90 days, defined as a mRS score of ≤2, (ii) all-cause long-term mortality, (iii) recurrent AIS at 90 days and during the entire follow-up period (iv) symptomatic intracranial hemorrhage (sICH) at 90 days and during the entire follow-up period, and (v) major bleeding besides intracranial hemorrhage or clinically relevant non-major bleeding during the entire follow-up period. The criteria used to define the composite bleeding outcome is provided in the Supplemental methods. For long-term mortality, follow-up time was defined as the time from the index AIS to the date of death for deceased patients or to the last update of the Swiss Population Registry for surviving patients. For incident cerebrovascular events (recurrent AIS or sICH), follow-up time was defined as the time from index AIS to the date of event or to the last documented follow-up in the electronic health record if no event was reported. For 90 days follow-up, only patients with available clinical or telephone follow-up were assessed. The mRS score and incident cerebrovascular events at 90 days were determined through telephone survey or clinical visit by investigators certified in mRS assessment. Symptomatic ICH was determined based on the ECASS III definition (22).

Baseline characteristics were reported as median and interquartile range (IQR) for continuous variables and frequency (percentage) for categorical variables. Differences between groups were assessed using Fisher’s exact test for categorical variables and the Wilcoxon rank-sum test for continuous variables. Kaplan–Meier curves were used to estimate the cumulative rates of time-to-event endpoints. The log-rank test and multivariable Cox regression were used to compare the outcomes between antithrombotic treatment groups. All multivariable models were adjusted for patient age, sex, initial D-dimer level, documented metastases at the time of the index AIS, and the presence of multi-territory brain infarcts. These covariates were selected because they either significantly differed between study groups or they were previously associated with adverse clinical outcomes in patients with cancer-related stroke (5). Adjusted hazard ratios (aHRs) were reported with their associated 95% confidence interval (CI). To investigate the association between treatment groups and recurrent cerebrovascular events (AIS or sICH), additional analyses using mortality as a competing risk were performed (see Supplementary methods). Patients who received intravenous thrombolysis before D-dimer assessment were excluded from the multivariable analyses because intravenous thrombolysis can influence levels of coagulation parameters such as D-dimer and fibrinogen (23).

Subgroup analyses were performed in (1) patients whose index AIS mechanism was cryptogenic and who met ESUS criteria. (2) patients with ESUS and an elevated D-dimer level (defined as a value above the median for the cohort), and (3) patients with ESUS and multiterritory infarcts.

In sensitivity analyses, we excluded (1) patients with combined antiplatelet and anticoagulant therapy and (2) patients with a history of VTE treated with anticoagulant therapy at discharge for the index AIS. Because antithrombotic prescription patterns likely varied according to physicians’ perceptions of how prothrombotic and high-risk individual patients were, we performed a second set of analyses calculating propensity scores and using the Inverse Probability of Treatment Weighting (IPTW) method with stabilized weights to minimize potential confounding (24). For IPTW, we reported hazard ratios (HRs) with their associated 95% CI.

Continuous variables with skewed distributions were logarithmically transformed. Missing data were not imputed. Statistical significance was defined as a p-value of <0.05. All analyses were performed using Stata 16 (StataCorp LLC) and R (version 3.6.0, R Core Team).

Of 5,012 patients with AIS assessed for eligibility, 306 had active cancer at the time of the index hospitalization (Figure 1 – study flowchart). Among these patients, we excluded 35 who died during the hospitalization, 30 whose cancer was diagnosed after hospital discharge, 33 without available follow-up, 63 with a cardioembolic indication for anticoagulation, and 10 who were not prescribed an antithrombotic medication at discharge. The baseline characteristics of included and excluded AIS patients with active cancer are shown in Supplementary Table 1. The final study population comprised 135 patients, including 58 (43%) in the anticoagulant group and 77 (57%) in the antiplatelet group.

Per TOAST criteria, an undetermined cause (i.e., cryptogenic) was the leading mechanism in both treatment groups: 82% of patients in the anticoagulant group (n = 48/58) and 69% in the antiplatelet group (n = 53/77). In these cryptogenic cases, ESUS criteria were met in 100% of the anticoagulant group (n = 48/48) and 72% of the antiplatelet group (n = 38/53). Large-artery atherosclerosis was identified in 9% of patients treated with anticoagulation (n = 5/58) and 25% of those treated with antiplatelet therapy (n = 19/77). Other determined causes were present in 9% of the anticoagulant group (n = 5/58) and 2% of the antiplatelet group (n = 2/77). Small-vessel disease was only observed in the antiplatelet group (4%, n = 3/77).

Of the 58 patients treated with anticoagulation, 41 (71%) received LMWH, 4 (7%) received vitamin K antagonists, 8 (14%) received DOACs (edoxaban, rivaroxaban, or apixaban), and 5 received (9%) a combination of antiplatelet plus anticoagulant therapy. Among the 41 patients discharged with LMWH, 21 (51%) had died at 3-months, 14 (34%) were lost to follow-up, 4 (10%) were switched to DOAC, and 2 were still receiving LMWH (5%). Median follow-up time for long-term mortality was 495 days (IQR 57–1,029) for the overall cohort, 133 days (IQR 43–506) for the anticoagulant group, and 797 days (IQR 218–1,483) for the antiplatelet group.

Patients treated with anticoagulant therapy, compared to patients treated with antiplatelet therapy, were younger (69 years [IQR 62–75] versus 75 years [IQR 65–82], p = 0.01), had more multi-territory brain infarcts (47% versus 17%, p < 0.001), and more often had an ESUS mechanism (82% versus 50%, p < 0.001; Table 1). Data on the primary cancer site in the overall study population and stratified by the individual treatment groups are provided in Figure 2. The distribution of primary cancer sites differed between groups with higher rates of lung and pancreatic cancers in the anticoagulant group (p < 0.001). Additionally, patients in the anticoagulant group had a higher median cancer stage (4 [IQR 3–4] versus 3 [IQR 2–4], p < 0.001) and more frequent metastases at the time of AIS (72% versus 41%, p < 0.001). There were 23 patients with documented VTEs, of whom 22 were prescribed anticoagulant therapy at hospital discharge.

Four patients in the antiplatelet group received intravenous thrombolysis before D-dimer sampling and were excluded from analyses including laboratory parameters. Compared to patients in the antiplatelet group, patients in the anticoagulant group had higher D-dimer levels in μg/L (median [IQR]: 8536 [2080–13,726] versus 1,010 [495–2,090], p < 0.001), higher CRP levels in mg/L (median [IQR]:18 [5–50] versus 4 [2–24], p = 0.01) and lower hemoglobin levels in g/dL (median [IQR]:11.8 [10.8–13.4] versus 13.0 [10.5–14.2], p = 0.03).

As depicted in Figure 3, the estimated cumulative 1-year mortality rate was higher in the anticoagulant group (66, 95% CI 53–77%) than in the antiplatelet group (33, 95% CI 23–44%) (log-rank test p < 0.001). In multivariable Cox regression analysis, anticoagulant use was associated with a similar 1-year mortality rate as antiplatelet use (aHR 0.76; 95% CI 0.36–1.63; p = 0.47) (Figure 4). Factors independently associated with 1-year mortality after AIS were initial D-dimer levels (aHR 4.59; 95% CI 2.24–9.38; p < 0.001) and multi-territory brain infarction (aHR 2.13; 95% CI 1.19–3.82; p = 0.01). After excluding five patients who received combined antiplatelet and anticoagulant therapy, no difference remained between groups for 1-year mortality (aHR 0.81; 95% CI 0.38–1.74; p = 0.59) (Supplementary Figure 1).

As shown in Table 2, patients in the anticoagulant group had a lower rate of good functional outcomes (mRS ≤ 2) at 90 days compared to those in the antiplatelet group (29% versus 66%, p = 0.003). After adjustment for potential confounders, there was no difference in 90-day good functional outcomes between study groups (aHR 1.36; 95% CI 0.58–3.19, p = 0.49, Supplementary Figure 2).

At 90-days, 0% of patients in the anticoagulant group (n = 0/23) were diagnosed with recurrent AIS compared to 7% in the antiplatelet group (n = 3/43, p = 0.55). There were no sICHs in either group during the first 90 days of follow-up.

Patients treated with anticoagulant therapy had higher long-term mortality compared to patients treated with antiplatelet therapy (88, 95% CI 77–94% versus 52, 95% CI 41–63%, log-rank test p < 0.001). However, after adjustment for potential confounders, there was no difference in long-term mortality between the groups (aHR 1.29; 95% CI 0.67–2.47; p = 0.44) (Supplementary Figure 3).

The median total follow-up time for cerebrovascular events was 165 days (IQR 80–707) for the overall cohort, 107 days (IQR 40–232) for the anticoagulant group, and 178 days (IQR 99–940) for the antiplatelet group. After one year and also during the entire follow-up period, 9% (n = 5/58) of patients treated with anticoagulant therapy had a recurrent AIS compared to 8% (n = 6/77) of patients treated with antiplatelet therapy (aHR 0.49; 95% CI 0.08–2.83; p = 0.83, Supplementary Figure 4). There were no cases of sICH in either group during long-term follow-up. Major bleeding besides intracranial hemorrhage or clinically relevant non-major bleeding were reported in three patients (2% of the entire cohort), all occuring within one year after AIS, with no differences between treatment groups. When accounting for mortality as a competing risk, the recurrence of AIS did not differ between patients receiving anticoagulants and those receiving antiplatelet therapy (Supplementary results I).

Of the 135 index AIS, 86 (64%) were classified as ESUS. In the ESUS subgroup (Supplementary Table 2), patients treated with anticoagulant therapy were on average younger and more often had metastatic disease compared to those treated with antiplatelet therapy. After multivariable adjustment, anticoagulant therapy, compared to antiplatelet therapy, was associated with similar mortality at one year after AIS (aHR 0.51; 95% CI 0.21–1.22; p = 0.11, Supplementary Figure 5). In patients with ESUS and D-dimer levels above the median (2,037 μg/L) for the cohort (n = 35), there was no difference in 1-year mortality rates between antithrombotic treatment groups after multivariable adjustment (aHR 0.26; 95% CI 0.05–1.27; p = 0.10, Supplementary Figure 10). Results were similar in patients with ESUS and multiterritory infarcts (n = 29) (aHR 1.17; 95% CI 0.32–4.19; p = 0.81, Supplementary Figure 11).

After exclusion from the initial study population of 22 patients with documented VTE treated with anticoagulant therapy at the time of index AIS hospital discharge, the risk of 1-year mortality remained similar between patients treated with anticoagulant therapy and those treated with antiplatelet therapy (aHR 0.65; 95% CI 0.28–1.47; p = 0.30, Supplementary Figure 6). Of the remaining 36 patients discharged on anticoagulant therapy, 24 (67%) were treated with LMWH, 3 (8%) with vitamin K antagonists, 7 (19%) with DOACs, and 2 (6%) with combination therapy. After excluding the same 22 patients with documented VTE treated with anticoagulant therapy at index AIS hospital discharge from the ESUS cohort, 1-year mortality remained similar between antithrombotic treatment groups (aHR 0.38; 95% CI 0.14–1.05; p = 0.06, Supplementary Figure 7).

In the second round of analyses employing IPTW, the stabilized weights exhibited a near-normal distribution centered around one, albeit with a few weights surpassing two, which could have a significant impact on study outcomes (Supplementary Figure 8).

The IPTW analysis for 1-year mortality (Figure 5) corroborated the main analyses for the primary AIS cohort (HR 0.82; 95% CI 0.39–1.72, p = 0.61) and the ESUS subgroup (HR 0.51; 95% CI 0.20–1.29; p = 0.15). These results were unchanged after excluding patients with documented VTE from the primary cohort (HR 0.73; 95% CI 0.32–1.66; p = 0.46) and the ESUS subgroup (HR 0.46; 95% CI 0.16–1.34; p = 0.30).