Cancer is an important risk factor for ischemic stroke. It can cause both arterial and venous thrombosis through various mechanisms, including cancer-related hypercoagulability, systemic inflammation and cardiac dysfunction. ^{Reference Lun, Siegal and Ramsay1} Accordingly, the underlying causes of stroke in patients with cancer may differ from those in the general population.

The combination of cancer and right-to-left shunt through a patent foramen ovale (PFO) or intrapulmonary fistula can predispose individuals to paradoxical venous embolization. ^{Reference Dearborn, Urrutia and Zeiler2} While right-to-left shunting is a compelling mechanism for stroke in patients with cancer, there is surprisingly little evidence regarding its frequency. Therefore, we sought to further evaluate this association by comparing the prevalence of right-to-left shunt in patients with ischemic stroke, with and without a new cancer diagnosis.

This is a retrospective cohort study with a population of patients aged 18 to 60 presenting to The Ottawa Hospital with ischemic stroke between January 01, 2020, and December 31, 2022. The Ottawa Hospital is a tertiary care comprehensive stroke center servicing the national capital region of Canada in the province of Ontario, with a catchment population of 1.6 million. For the purposes of our study, we defined Ischemic stroke using International Classification of Disease Tenth Edition (ICD-10) diagnostic code [I63].

At The Ottawa Hospital, an agitated saline (bubble) study with echocardiography is performed as a protocolled standard of care to investigate right-to-left shunts in patients ≤60 years of age presenting with ischemic stroke. Accordingly, our population of interest was restricted to patients ≤60 years old with stroke who had undergone echocardiography with documented saline study. We purposefully excluded those older than 60 to avoid potential selection bias. The primary exposure of interest in our study is the diagnosis of cancer within 1 year of the ischemic stroke (before or after), which was selected based on recent literature confirming patients have the highest risk of ischemic stroke within 1 year of a new cancer diagnosis, ^{Reference Lun, Cerasuolo and Carrier3–Reference Rioux, L. and Nehme5} and because thromboembolism can be the presenting symptom of an undiagnosed occult malignancy. We defined cancer diagnosis using the ICD-10 diagnostic codes [C00 to D48] encompassing all cancer subtypes except primary central nervous system tumors (which may be misclassified as stroke based on imaging) and non-melanoma skin cancers such as squamous or basal cell carcinoma. Non-melanoma skin cancers have favorable prognoses and do not require systemic therapies leading to inaccuracies in administrative diagnostic coding. ^{Reference Lun, Roy and Hao4} We considered occult cancers as active cancer at the time of stroke diagnosis, consistent with previous studies. ^{Reference Lun, Roy and Hao4} Our primary outcome of interest is the presence of right-to-left shunt.

Data extracted from The Ottawa Hospital Data Warehouse ^{Reference Lun, Siegal and Ramsay6} included sex, age at stroke diagnosis, relevant radiology exams such as transthoracic echocardiography and transesophageal echocardiography, prior history of cancer, diabetes, hypertension, hyperlipidemia, coronary artery disease, venous thromboembolism, deep vein thrombosis, pulmonary embolism and relevant medications at the time of stroke such as antiplatelet or anticoagulant therapy. We performed exploratory univariate analyses to identify potential associations between baseline patient characteristics and the presence of shunt. We subsequently built a multivariate model to assess the association between cancer diagnosis and shunt; baseline characteristics from univariable analyses with p ≤ 0.10 were included in the model as potential covariates. Non-significant variables (p > 0.05) were eliminated in a backwards stepwise fashion to create a minimal model. The use of data in this project was authorized by the Ottawa Health Science Network Research Ethics Board (Protocol ID#: 20240331-01H). The study data can be made available upon request to the corresponding author and following clearance by local ethics committee. The Strengthening the Reporting of Observational Studies in Epidemiology guidelines were followed in this study.

From the 650 eligible patients presenting to the Ottawa Hospital with ischemic stroke between January 1, 2020, and December 31^st, 2022, 491 patients (36.9 % female, median age 53 years) were included in this study (Figure 1). Demographic data comparing excluded and included patients suggest excluded patients were more likely to have coronary artery disease (23.9% versus 7.5%, p < 0.001) (Supplementary Table 1). We identified a cancer diagnosis within one year of stroke in 43 (8.8%) patients, 12 of whom (27.9%; 95% CI 15–44) were found to have a right-to-left shunt. In contrast, 448 stroke patients (91.2%) did not have a cancer diagnosis, 133 of whom (29.7%; 95% CI 25–34) were found to have a right-to-left shunt (Table 1). The distribution of cancer types identified in our population is detailed in Figure S1. When adjusting for covariates, there was a non-significant trend towards lower prevalence of right-to-left shunting in patients presenting with ischemic stroke and an active cancer diagnosis (adjusted odds ratio (aOR) 0.83; 95% CI 0.37–1.84; Figure 2). Multivariate analysis revealed the presence of shunt was associated with a history of venous thromboembolism (aOR, 3.60; 95% CI 1.49–8.67).

Figure 1. Flowchart indicating included versus excluded patients.

Table 1. Baseline characteristics of included patients stratified by absence or presence of active cancer

Note: DOAC = Direct oral anticoagulants; VTE = Venous thromboembolism; DVT = Deep vein thromboembolism; PE= Pulmonary embolism.

Figure 2. Multivariable logistic regression using adjusted odds ratios (aOR) and 95% CI for assessing the association between the presence of shunt and covariates. aOR = Adjusted Odds Ratio; DVT = Deep Vein Thromboembolism; PE = Pulmonary Embolism; VTE = Venous Thromboembolism; 95% CI = 95% confidence interval.

In this retrospective cohort study, the prevalence of right-to-left shunting did not significantly correlate with the presence of active cancer in our cohort of 650 patients aged ≤60 years old presenting with ischemic stroke. This result is consistent with a recent study suggesting the dual diagnosis of stroke and active cancer is associated with an absence of shunt (aOR 2.29; 95% CI 1.14–4.58). ^{Reference Steinauer, Bücke and Buffle7} These studies challenge the concept of paradoxical venous embolism as a common mechanism of ischemic stroke in patients with active cancer. However, our study did show an association between shunt and a history of venous thromboembolism (VTE), supporting the mechanism of paradoxical embolism in patients with history of venous hypercoagulability but not with cancer.

There are several mechanisms independent of right-to-left shunting that may explain the increased risk of arterial thromboembolism in patients with active cancer diagnosis. These include arterial hypercoagulability related to the cancer-itself, which likely account for the majority of the cases as well as systemic and cardiovascular complications of therapy. ^{Reference Sener and Keser8,Reference Abdelsalam, Abu-Hegazy and El-Hadaad9} Other rare mechanisms of stroke may include infective endocarditis related to immunosuppression or central line placement, ^{Reference Sener and Keser8} nonbacterial endocarditis with sterile vegetations of aortic valves, or from antithrombotic cessation due to thrombocytopenia or the need for invasive procedrues. ^{Reference Abdelsalam, Abu-Hegazy and El-Hadaad9}

Despite its retrospective nature, our study was strengthened through its use of a well-established comprehensive data repository, The Ottawa Hospital Data Warehouse. Furthermore, given our study was conducted in tertiary care stroke center servicing a diverse patient population with varied ethnic and socioeconomic backgrounds in the national capital region of Canada, our study findings may be generalizable to other multicultural urban healthcare settings.

However, this study has important limitations. First, patients over 60 years old were excluded from the study as they did not routinely undergo an agitated saline study in accordance with our institution’s guidelines. Similarly, patients who were unable to cooperate with Valsalva were excluded from the study due to unreliable echocardiography results, which may have biased the sample against patients with more severe presentations. Nevertheless, our study provides important information to further our understanding of stroke etiology in a relatively younger cohort of patients with cancer.

Overall, our finding does not support the hypothesis that cancer-associated stroke is related to PFO or right-to-left shunting from other sources such as pulmonary fistula. Future studies should focus on alternative mechanisms for ischemic stroke occurrence among patients with active cancer.