Diagnosis of Incident Cancer After Cryptogenic Stroke

Abstract

Objectives

The objective of this study was to estimate the incidence, timing, and type of new cancer diagnosis among patients with cryptogenic stroke.

Methods

We used data from the ARCADIA trial, which enrolled patients with cryptogenic stroke and atrial cardiopathy. Participants were prospectively followed, and serious adverse events were assessed every 3 months or sooner if investigators were alerted between visits to an event. Kaplan-Meier statistics were used to estimate the cumulative incidence of a cancer diagnosis within the first year after randomization.

Results

Among 878 participants without baseline history of cancer, 13 (1.5%) were diagnosed with incident cancer in the year after randomization, comprising 12 solid cancers (3 prostate, 2 breast, 2 gastrointestinal, and 5 other primary sites) and 1 hematologic cancer (non-Hodgkin lymphoma). The cumulative incidences of a cancer diagnosis were 0% at 3 months, 0.6% (95% CI 0.2%–1.5%) at 6 months, and 2.0% (95 CI 1.1%–3.4%) at 1 year. The median time from index stroke to cancer diagnosis was 261 days (interquartile range 183–358).

Discussion

In a multicenter cryptogenic stroke cohort with prospective follow-up, the 1-year cumulative incidence of a cancer diagnosis was 2%. This rate may be an underestimation because of the clinical trial population and exclusion of cancers diagnosed immediately after stroke.

Trial Registration Information

ClinicalTrials.gov Identifier: NCT03192215. Registered June 20, 2017. First patient enrolled February 1, 2018.

Introduction

Cancer is associated with a heightened risk of stroke.^1,2 This increased risk is highest near the time of cancer diagnosis.³ In the first 6 months after cancer diagnosis, ischemic stroke risk is approximately doubled.⁴ Similarly, in the year before cancer diagnosis, ischemic stroke risk is increased by approximately 60% and first becomes significantly elevated 5 months before the cancer diagnosis date.⁵ Because cancer generally takes months to years to develop, occult cancer may be present at the time of index stroke and serve as the underlying trigger or cause. Prior studies have reported that 1%–10% of patients with ischemic stroke will be diagnosed with incident cancer in the following year.^6-9 However, these studies were nearly all retrospective, and most were small, single-center, and racially homogeneous. In addition, few studies have examined patients with cryptogenic stroke, a group lacking an identified source of stroke.^7,9-12 Therefore, we used data from the ARCADIA trial to evaluate the incidence of a new cancer diagnosis in the year after cryptogenic stroke.

Methods

Design

This was an exploratory analysis of ARCADIA, a prospective randomized trial, which compared apixaban with aspirin in patients aged 45 years and older with cryptogenic stroke and biomarker evidence of atrial cardiopathy.¹³ Cryptogenic stroke was defined per the consensus criteria for embolic stroke of a undetermined source.¹⁴ For our primary analysis, we excluded patients with history of known cancer at the time of randomization. Details of the ARCADIA trial's design were previously published.¹³ The study, supported by National Institutes of Neurological Disorders and Stroke grant U01NS095869, was conducted from 2018 to 2023 at 185 North American sites. The first participant was enrolled on February 1, 2018.

Measurements

The primary outcome for this analysis was a new cancer diagnosis within 1 year of randomization. All malignant, solid, hematologic, or CNS cancers were included. Localized skin carcinomas were excluded. Data on new cancer diagnoses were obtained through adverse event reporting. Cancer was considered a mandatory reporting event. Data were available on cancer type but not stage. Scheduled visits were performed at approximately 30, 90, 180, 270, and 360 days after randomization. Data on baseline demographics, comorbidities, stroke severity, functional status, and atrial cardiopathy biomarkers were collected.

Analysis

We used Kaplan-Meier statistics to estimate the cumulative incidence of new cancer diagnoses after cryptogenic stroke. Censoring occurred at the time of death, withdrawal, loss to follow-up, or trial completion. We used Fisher exact test and Wilcoxon rank-sum test to compare baseline characteristics between participants diagnosed with cancer in the year after randomization vs those who were not. Univariate p values were provided for hypothesis generation. We did not perform multivariable analyses because of the low number of events. An α error of <0.05 was considered significant. In sensitivity analysis, we accounted for the competing risk of death. In secondary analysis, we included participants with known history of cancer upon randomization, among whom a new different cancer diagnosis was considered an outcome, whereas recurrence of the previous cancer was not.

Standard Protocol Approvals, Registrations, and Patient Consents

The ARCADIA trial was approved by site ethical review boards and first registered at ClinicalTrials.gov (ID NCT03192215) on June 20, 2017. All patients provided written informed consent to participate.

Data Availability

Data were accessible through ARCADIA's Publication Committee.

Results

Among 1,015 participants, 137 (13.5%) had baseline history of cancer at randomization. After excluding participants with known cancer, 878 participants were included in our main analysis. Participants diagnosed with incident cancer in the year after randomization did not differ significantly on baseline characteristics from those not diagnosed with cancer (Table). The median time from index stroke to study randomization was 48 days (interquartile range [IQR] 21–97). During the first year of follow-up, 13 participants (1.5%) were diagnosed with new cancer. This included 12 solid cancers (3 prostate, 2 breast, 2 gastrointestinal, and 5 other primary sites) and 1 hematologic cancer (lymphoma). The cumulative incidences of a cancer diagnosis were 0% at 3 months, 0.6% (95% CI 0.2%–1.5%) at 6 months, and 2.0% (95% CI 1.1%–3.4%) at 1 year (Figure). Accounting for the competing risk of death, cumulative incidence rates were 0% at 3 months, 0.6% (95% CI 0.2%–1.3%) at 6 months, and 2.0% (95% CI 1.1%–3.2%) at 1 year. The median time from randomization to cancer diagnosis was 217 days (IQR 142–273). The median time from index stroke to cancer diagnosis was 261 days (IQR 183–358).

Table.

Participant Characteristics at Enrollment Stratified by the Subsequent Diagnosis of Incident Cancer During 1-Year Follow-Up

Characteristica,b	New cancer diagnosis (n = 13)	No new cancer diagnosis (n = 865)	p Value
Demographics
Age, y	69 (64–74)	67 (59–75)	0.392
Female sex	6 (46)	470 (54)	0.586
Race			0.533
Black or African American	2 (15)	197 (23)
White	10 (77)	631 (73)
Other	1 (8)	37 (4)
Ethnicity			1.000
Hispanic or Latino	1 (8)	73 (8)
Not Hispanic or Latino	12 (92)	788 (91)
Unknown	0 (0)	4 (1)
Weight, kg	77 (71–88)	83 (71–98)	0.356
Medical comorbidities
Coronary artery disease	1 (8)	86 (10)	1.000
Congestive heart failure	1 (8)	61 (7)	1.000
Diabetes mellitus	3 (23)	279 (32)	0.566
Hypertension	9 (69)	672 (78)	0.503
Prior stroke	2 (15)	172 (20)	1.000
Tobacco use	7 (54)	365 (42)	0.412
Stroke severity/functional status
Baseline NIH Stroke Scale	2 (0–3)	1 (0–3)	0.677
Baseline mRS score	1 (1–2)	1 (0–2)	0.860
Atrial cardiopathy biomarkers
NT-proBNP, pg/mL	277 (42–553)	302 (107–548)	0.443
PTFV1, μV × ms	5,500 (4,500–6,600)	5,200 (2,750–6,000)	0.329
LA diameter index, cm/m2	2.0 (1.9–2.2)	1.9 (1.6–2.2)	0.137
Randomized treatment arm
Apixaban	5 (38)	441 (51)	0.414
Aspirin	8 (62)	424 (49)

Abbreviations: LA = left atrial; NT-proBNP = N-terminal prohormone of brain natriuretic peptide; mRS = modified Rankin Scale; PTFV₁ = P-wave terminal force in lead V₁.

^aData reported as n (%) for categorical variables and median (interquartile range) for continuous and ordinal variables.

^bFisher exact test and Wilcoxon rank-sum test were to compare baseline characteristics between participants diagnosed with cancer in the year after randomization vs those who were not.

Figure. Cumulative Incidence of New Cancer Diagnoses Among Patients With Cryptogenic Stroke.

Kaplan-Meier curve estimating the cumulative incidence of cancer diagnoses among patients with cryptogenic stroke and atrial cardiopathy. Participants were followed for a diagnosis of incident cancer from the day of study randomization to day 365. The shaded areas represent the 95% CIs for the estimated rates. Hash marks represent participant censoring for death, withdrawal, loss to follow-up, or trial completion. The numbers at risk over time are listed at the bottom.

Three of 137 participants (2.2%) with baseline history of cancer were diagnosed with new cancer by 1 year of follow-up. These included 2 leukemias and 1 lymphoma. When including these participants in Kaplan-Meier analysis, cumulative incidences of a new cancer diagnosis were 0% at 3 months, 0.6% (95% CI 0.2%–1.4%) at 6 months, and 2.1% (95% CI 1.3%–3.4%) at 1 year.

Discussion

In a multicenter, prospective, racially and ethnically heterogeneous cohort of 878 patients with cryptogenic stroke and atrial cardiopathy, the 1-year cumulative incidence of a new cancer diagnosis was 2%. The median time from index stroke to cancer diagnosis was 261 days. The most frequent cancer types were prostate, breast, and gastrointestinal.

Our estimated 1-year incidence of new cancer diagnoses after cryptogenic stroke falls within the lower range of prior reports.^6-9,12 A meta-analysis from 2021 evaluated 2 studies comprising 397 patients with cryptogenic stroke and reported a 1-year incidence rate of new cancer diagnoses of 6.2%.⁶ More recent estimated rates of incident cancer after cryptogenic stroke have included 7.1% at 1 year from Bern, Switzerland (n = 521), and 5.9% after 10-month follow-up from Atlanta, GA (n = 236).^7,9 These studies were all single-center and 3 of 4 were retrospective. They also differed from ours in that they included cancers diagnosed during the index stroke hospitalization and soon thereafter, whereas our cohort was from a secondary stroke prevention trial in which participants were typically randomized 1–2 months after cryptogenic stroke. Therefore, the lower incidence rate identified in our study may be an underestimation. Furthermore, some patients diagnosed with cancer after their index stroke may have been excluded from the ARCADIA trial because investigators focused on instituting time-sensitive cancer treatments, or they no longer considered the patient's stroke cryptogenic. Only one other multicenter prospective study has reported the rate of new cancer diagnoses after cryptogenic stroke, and that was from another randomized trial (NAVIGATE ESUS).⁸ Among 7,213 participants with cryptogenic stroke, 124 (1.7%) were diagnosed with a new cancer during 11months of average follow-up, a rate similar to ours.

Our study was limited by its clinical trial population and the inherent risk of selection bias, its reliance on adverse event reporting to capture diagnoses of incident cancer, a median 48-day delay from index stroke to the start of follow-up, and lack of available data on how cancers were diagnosed and their histopathology and stage. If systematic cancer screening were performed, incidence rates may have been higher. We also lacked data on D-dimer and other biomarkers linked to cancer-associated stroke.¹⁵ In addition, the ARCADIA trial required biomarker evidence for atrial cardiopathy, and because atrial cardiopathy is a potential mechanism for cryptogenic stroke, our study population could have been less likely to harbor occult cancer.

In this prospective cryptogenic stroke cohort, 2% were diagnosed with new cancer in the subsequent year. Most of the cancers were diagnosed between 6 and 12 months after patients' cryptogenic stroke, a fairly long interval between events, suggesting that many of the cancers identified during follow-up may not have contributed to index stroke development. Because cancers diagnosed immediately after stroke were excluded from this analysis, our results may not generalize to the average hospitalized patient with cryptogenic stroke. Future studies are needed to determine which characteristics predict occult cancer in patients with cryptogenic stroke and whether systematic screening can expedite cancer discovery and improve outcomes.

Appendix. Authors

Name	Location	Contribution
Babak B. Navi, MD, MS	Clinical and Translational Neuroscience Unit, Feil Family Brain and Mind Research Institute and Department of Neurology, Weill Cornell Medicine; Department of Neurology, Memorial Sloan Kettering Cancer Center, New York, NY	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Cenai Zhang, MS	Clinical and Translational Neuroscience Unit, Feil Family Brain and Mind Research Institute and Department of Neurology, Weill Cornell Medicine, New York, NY	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Benjamin R. Miller, MD	Department of Neurology, University of Minnesota, Minneapolis	Drafting/revision of the manuscript for content, including medical writing for content
Anokhi Pawar, BS	Clinical and Translational Neuroscience Unit, Feil Family Brain and Mind Research Institute and Department of Neurology, Weill Cornell Medicine, New York, NY	Drafting/revision of the manuscript for content, including medical writing for content
Mary Cushman, MD, MSc	Division of Hematology and Oncology, Department of Medicine, University of Vermont Larner College of Medicine, Burlington	Drafting/revision of the manuscript for content, including medical writing for content
Scott E. Kasner, MS, MSCE	Department of Neurology, University of Pennsylvania School of Medicine, Philadelphia	Drafting/revision of the manuscript for content, including medical writing for content
David Tirschwell, MD, MSc	Department of Neurology, University of Washington, Seattle, WA	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design
W. T. Longstreth, Jr., MD, MPH	Department of Neurology, and Department of Epidemiology, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design
Richard Kronmal, PhD	Department of Biostatistics, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; analysis or interpretation of data
Jordan Elm, PhD	Department of Biostatistics, Medical University of South Carolina, Charleston	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Richard M. Zweifler, MD	Ochsner Neuroscience Institute, Ochsner Health, New Orleans, LA	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design
Joseph Tarsia, MD	Ochsner Neuroscience Institute, Ochsner Health, New Orleans, LA	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design
Joseph P. Broderick, MD	Department of Neurology and Rehabilitation Medicine, University of Cincinnati College of Medicine, OH	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design
David J. Gladstone, MD, FRCPC	Sunnybrook Research Institute, Hurvitz Brain Sciences Program, Sunnybrook Health Sciences Centre, and Division of Neurology, Department of Medicine, University of Toronto, Ontario, Canada	Drafting/revision of the manuscript for content, including medical writing for content
Morin Beyeler, MD	Clinical and Translational Neuroscience Unit, Feil Family Brain and Mind Research Institute and Department of Neurology, Weill Cornell Medicine, New York, NY; Department of Neurology, Inselspital, Bern University Hospital and University of Bern, Switzerland	Drafting/revision of the manuscript for content, including medical writing for content
Hooman Kamel MD, MS	Clinical and Translational Neuroscience Unit, Feil Family Brain and Mind Research Institute and Department of Neurology, Weill Cornell Medicine, New York, NY	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Mitchell S.V. Elkind, MD, MS	Department of Neurology, Vagelos College of Physicians and Surgeons, and Department of Epidemiology, Mailman School of Public Health, Columbia University, New York, NY	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Christopher Streib, MD, MS	Department of Neurology, University of Minnesota, Minneapolis	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design

Study Funding

The study was supported by National Institutes of Neurological Disorders and Stroke grant U01NS095869. The BMS-Pfizer Alliance for Eliquis provided study drug in kind for the trial and Roche provided ancillary funding to help support laboratory assays.

Disclosure

B.B. Navi discloses receiving consulting fees for serving on an adjudication committee for MindRhythm Inc. C. Zhang, B. Miller, A. Pawar, and M. Cushman report no disclosures relevant to the manuscript. S.E. Kasner discloses receiving grant funding from Diamedica, Bayer, Bristol Myers Squibb, Genentech, and WL Gore, consulting fees for serving on a DSMB for AstraZeneca and an adjudication committee for NovoNordisk, and royalties from UpToDate. D.L. Tirschwell, W.T. Longstreth Jr, R.A. Kronmal, J. Elm, R.M. Zweifler, and J. Tarsia report no disclosures relevant to the manuscript. J.P. Broderick reports receiving multiple research awards from the National Institutes of Neurological Disorders and Stroke (NINDS), receiving study medication and monetary support from Novo Nordisk for temperature monitoring and after-hour enrollment for the ongoing NINDS-funded FASTEST trial, and receiving monies for educational and research funds for the University of Cincinnati Department of Neurology and Rehabilitation Medicine from Genentech (for his role on the Executive Committee of the TIMELESS trial), Roche, BrainsGate Ltd., Basking Biosciences, and the Pharmacy and Therapeutics Committee of Kroger Prescriptions Plans Inc., as a voting member of the committee. D.J. Gladstone reports no disclosures relevant to the manuscript. M. Beyeler reports grant support from the University of Bern, Switzerland. H. Kamel reports no disclosures relevant to the manuscript. M.S.V. Elkind discloses receiving salary as an employee of the American Heart Association and royalties for chapters on cryptogenic stroke in UpToDate. C. Streib reports no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.