Background: Covert cerebrovascular disease (CCD) includes white matter disease (WMD) and covert brain infarction (CBI). Incidentally discovered CCD is associated with an increased risk of subsequent symptomatic stroke. However, it is unknown whether the severity of WMD or the location of CBI predicts risk. Objectives: The objective of this study was to examine the association of incidentally discovered WMD severity and CBI location with a risk of subsequent symptomatic stroke. Method: This retrospective cohort study includes patients ≥50 years old from the Kaiser Permanente Southern California health system who received neuroimaging for a nonstroke indication between 2009 and 2019. Incidental CBI and WMD were identified via natural language processing of the neuroimage report, and WMD severity was classified into grades. Results: 261,960 patients received neuroimaging, 78,555 (30.0%) were identified to have incidental WMD, and 12,857 were identified (4.9%) to have incidental CBI. Increasing WMD severity is associated with the increased incidence rate of future stroke. However, the stroke incidence rate in CT-identified WMD is higher at each level of severity compared to rates in MRI-identified WMD. Patients with mild WMD via CT have a stroke incidence rate of 24.9 per 1,000 person-years, similar to that of patients with severe WMD via MRI. Among incidentally discovered CBI patients with a determined CBI location, 97.9% have subcortical infarcts rather than cortical infarcts. CBI confers a similar risk of future stroke, whether cortical or subcortical, or whether MRI detected or CT detected. Conclusions: Increasing severity of incidental WMD is associated with an increased risk of future symptomatic stroke, dependent on the imaging modality. Subcortical and cortical CBI conferred similar risks.

Covert cerebrovascular disease (CCD) is, by definition, a neuroimaging finding suggestive of ischemic brain lesions unaccompanied by overt neurological symptoms and includes both white matter disease (WMD) and covert brain infarction (CBI). As patients are not generally screened for CCD, these findings are typically discovered incidentally on neuroimaging during routine care. Our previous work using natural language processing (NLP) of neuroimaging reports shows incidentally discovered CCD to be associated with the increased risk of subsequent symptomatic stroke [1]. Patients with CBI had a three-fold increased risk, whereas those with WMD had a slightly lower but still significantly increased risk. While there are no current evidence-based guidelines for stroke prevention for these patients, there is increasing interest in considering CCD alongside typical stroke risk factors [2, 3].

Risk stratification of patients with CCD may be enabled by understanding whether different subtypes of WMD and CBI have differential impact on future stroke risk. This study aimed to examine whether the severity of WMD and the location of CBI, as documented on routinely obtained neuroimaging reports of CT scans and MRI, are associated with the risk of future stroke. No previous study thus far has studied these details.

Study Design

This retrospective cohort study involved health plan enrollees of Kaiser Permanente Southern California (KPSC). Structured data were captured from the KPSC Research Data Warehouse, and clinical notes were extracted from Clarity (a data repository of Health Connect). The study protocol was approved by Institutional Review Boards at Tufts Medical Center and KPSC.

Patient Cohort

Inclusion criteria included age ≥50 years old, enrollment in KPSC, and a head CT or MRI for a nonstroke indication between 2009 and 2019. Exclusion criteria included prior ischemic stroke, dementia/Alzheimer’s disease, transient ischemic attack, or a “high probability” stroke symptom (defined in this study as aphasia, hemiparesis [including face, arm, and/or leg weakness], hemisensory loss, hemiataxia, hemineglect, visual disturbance [vision loss, diplopia], dysarthria, and dysphagia). Corresponding ICD diagnosis and ICD/CPT procedure codes are given in online supplementary A1 (for all online suppl. material, see https://doi.org/10.1159/000534581). In the event of multiple imaging studies, the first study was considered the “index scan.” Individuals were only included if they did not receive a new ICD code for cerebral infarction within 60 days of this index scan. Follow-up started 60 days after the index scan and ended with the earliest of the following events: disenrollment from the health plan, end of the study, death, or stroke (outcome).

Identification of WMD and CBI

Patients with incidental WMD or CBI were identified via application of NLP to neuroimaging reports associated with index scans. Our previous studies have discussed this method in detail and its yield as on par with human readers [1, 4‒6]. WMD was classified into three severity grades: mild, moderate, or severe; CBI was classified into cortical (which includes the cerebral cortex), subcortical (which includes subcortical, deep, or infratentorial structures), or unknown (where no information on location was provided). Further details are provided in online supplementary A2.

Statistical Analysis

We examined the crude and adjusted associations of WMD severity with stroke using Cox proportional hazard regression models. For adjusted effects, we included known cardiovascular risk factors for stroke based on prediction models in the literature [7, 8]. Analyses were performed using SAS (version 9.4 for Unix; SAS Institute, Cary, NC, USA) and R version 3.6.0 (R Foundation, Vienna, Austria).

Out of an initial cohort of 1,064,969 patients, 261,960 were included in the study after applying exclusion criteria. 78,555 (30.0%) were identified to have incidental WMD, and 12,857 (4.9%) were identified to have incidental CBI. The online supplementary eTable 1 describes baseline characteristics in the entire cohort, those with WMD (with or without CBI), those with CBI (with or without WMD), and those without CCD. The online supplementary eTable 2 describes baseline characteristics by severity of WMD.

Table 1 demonstrates that WMD presence and severity can be used to stratify patients across a wide range of stroke risk from 5.3 per 1,000 person-years in those free of WMD by MRI to 47.7 per 1,000 person-years for those with severe WMD by CT scan. Increasing WMD severity appears to be associated with a monotonically increased incidence rate of future stroke within each modality. However, the stroke incidence rate in CT-identified WMD is higher at each level of severity when compared to rates in MRI-identified WMD. Patients with mild WMD discovered via CT have a stroke incidence rate of 24.9 (95% CI 23.9–25.9 years), similar to that of patients with severe WMD detected by MRI. Crude and adjusted hazard ratios are significant and increase monotonically with increasing WMD severity for both modalities.

The location of incidentally discovered CBI was identifiable in 84.4% of patients; most patients with an identifiable CBI location have subcortical lesions (97.9%) rather than cortical lesions. CBI confers a similar risk of future stroke, whether cortical or subcortical, or whether detected by MRI or CT (online suppl. eTable 3).

Kaplan-Meier curves demonstrating stroke-free survival probability of patients with WMD stratified by severity are shown in Figure 1. Increasing severity of WMD is associated with worse stroke-free survival. Furthermore, WMD identified via CT has worse stroke-free survival at each severity grade compared to that identified via MRI.

To our knowledge, this is the first study that assesses incidentally discovered WMD severity and CBI location and their association with future symptomatic stroke. In this large observational cohort with over 260,000 patients, an increased severity of WMD is associated with a monotonic increase in the incidence rate of future stroke. This routinely obtained information, ascertainable from reports of 77% of patients with WMD, can be used to stratify risk of future stroke. This resembles observations from prospective cohort studies with protocol-driven neuroimaging of the increasing stroke risk with progression in WMD [9, 10]. Additionally, the incidence of future stroke risk is dependent on the imaging modality used to identify WMD severity, with mild WMD identified via CT having a similar incidence compared to severe WMD identified via MRI. This is a similar pattern to what we observed for the outcome of dementia [11] and likely reflects the much higher sensitivity of MRI for detecting WMD. Indeed, patients with mild disease on MRI are at a similar risk to patients with no WMD on CT.

While patients with CBI had an increased risk of future stroke, no difference in risk emerged between cortical and subcortical locations. This is somewhat surprising as stroke mechanisms that often result in cortical infarcts common in older adults (e.g., large artery atherosclerosis and atrial fibrillation) tend to have higher stroke recurrence rates than small vessel disease mechanisms leading to subcortical infarcts (hypertension). This may underscore the differences in intensity of stroke risk factor management in primary and secondary stroke prevention scenarios and the potential impact of not addressing small vessel disease risk factors (e.g., hypertension) in patients with CBI. It may also reflect issues with selection, since only stroke-free patients with other indications for scanning are included in our study. Additionally, despite the very large size of our cohort, there were relatively few cortical infarcts, which may have limited statistical power.

CCD has been proposed as a stroke equivalent, and these patients may benefit from more intensive stroke risk factor control, potentially resembling secondary stroke prevention. An important next step to assess this potential is a clinical trial involving antiplatelet therapy in patients with incidental CCD and assessing future risk of stroke. Our study reveals an abundance of CCD in the general adult population, particularly with WMD. The new understanding that severity of WMD is positively correlated with an increasing incidence rate of subsequent stroke may be helpful in identifying which patients may benefit most from secondary prevention. However, as patients with severe WMD are often older with limited life expectancy, one must also take into consideration who will be able to benefit from preventive treatments.

This study has several limitations, many of which relate to the use of “real world data.” For example, all variables are ascertained through the electronic health record, which may lead to some misclassifications; in particular, patients who were not assessed for certain characteristics (e.g., carotid artery arteriosclerotic disease) are classified as negative. Identifying CCD using neuroimaging obtained through routine care is subject to selection bias compared to screening-detected CCD. However, this cohort is arguably a clinically relevant target as these are the patients with CCD who, in the absence of screening, come to clinical attention. Furthermore, the use of NLP on routinely ascertained neuroimaging reports may have shortcomings of consistency and reliability as compared to centralized adjudication or advanced imaging analysis-based approaches (e.g., volumetric). We also did not record microbleeds and perivascular spaces, which are inconsistently captured in routine neuroimaging reports. From a pragmatic perspective, however, this may also be seen as strength since these are the types of neuroimaging reports that can be leveraged for “opportunistic screening” and stroke prevention in everyday clinical practice. Finally, our results should not be interpreted as estimating the causal effects of CCD on future stroke risk, since estimating causal effects from observational data typically requires unverifiable assumptions. Our results do show that WMD severity, as ascertained from routine neuroimaging reports, is useful in identifying patients at high risk of future stroke. Adjusted effects remain significant and increase monotonically with increasing severity, lending additional credence to the prognostic value of these findings. Lastly, these data are based on KPSC patients, and there may be issues with external validity to patients outside of this health system and systems with different coding practices. However, KPSC serves 4.8 million individuals at 15 hospitals and 230+ medical offices. The large and diverse study population increases the likelihood of the applicability of our findings to other populations.

In summary, incidentally discovered WMD is common, and increasing WMD severity is associated with a higher risk of future stroke. WMD discovered via CT has a higher stroke incidence risk at all severity levels when compared with those discovered via MRI. In the current study, the CBI location did not demonstrate the differential risk of future stroke.

This article is a reworking of a previous article by us that was retracted following the postpublication identification of a coding error (for more details, see https://doi.org/10.1159/000531108).

This study was approved by the following institutional review boards: the Kaiser Permanente Southern California Institutional Review Board (approval number 12111), Mayo Clinic Institutional Review Board (approval number PR17-006674), and the Tufts Health Sciences Institutional Review Board (Tufts Medical Center, approval number 11953). Informed consent requirements were waived by the Institutional Review Boards.

The authors have no conflicts of interest to declare.

This work was funded by an NIH Grant (R01-NS102233). The funder had no role in the design or conduct of the study: collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Andy Y. Wang wrote the original manuscript draft and contributed to investigation and data curation. Lester Y. Leung, Patrick H. Leutmer, David F. Kallmes, and Sunyang Fu contributed to investigation and methodology. Eric J. Puttock contributed to data curation, formal analysis, and investigation. Jason Nelson contributed to investigation. Chengyi Zheng contributed to data curation, investigation, and methodology. Hongfang Liu contributed to funding acquisition, investigation, methodology, and supervision. Wansu Chen contributed to data curation, formal analysis, investigation, methodology, and supervision. David M. Kent contributed to conceptualization, funding acquisition, investigation, methodology, and supervision and wrote the original manuscript draft. All authors contributed equally to reviewing and editing the final manuscript.

These data that support the findings of this study are not publicly available due to ethical standards. The authors do not have permission to share data. Further inquiries can be directed to the corresponding author.

1.
Kent DM, Leung LY, Zhou Y, Luetmer PH, Kallmes DF, Nelson J, et al. Association of silent cerebrovascular disease identified using natural language processing and future ischemic stroke. Neurology. 2021;97(13):e1313–e1321.
2.
Gupta A, Giambrone AE, Gialdini G, Finn C, Delgado D, Gutierrez J, et al. Silent brain infarction and risk of future stroke: a systematic review and meta-analysis. Stroke. 2016;47(3):719–25.
3.
Debette S, Markus HS. The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: systematic review and meta-analysis. BMJ. 2010;341:c3666.
4.
Fu S, Leung LY, Wang Y, Raulli AO, Kallmes DF, Kinsman KA, et al. Natural Language processing for the identification of silent brain infarcts from neuroimaging reports. JMIR Med Inform. 2019;7(2):e12109.
5.
Leung LY, Fu S, Luetmer PH, Kallmes DF, Madan N, Weinstein G, et al. Agreement between neuroimages and reports for natural language processing-based detection of silent brain infarcts and white matter disease. BMC Neurol. 2021;21(1):189.
6.
Fu S, Leung LY, Raulli AO, Kallmes DF, Kinsman KA, Nelson KB, et al. Assessment of the impact of EHR heterogeneity for clinical research through a case study of silent brain infarction. BMC Med Inform Decis Mak. 2020;20(1):60.
7.
Flueckiger P, Longstreth W, Herrington D, Yeboah J. Revised framingham stroke risk score, nontraditional risk markers, and incident stroke in a multiethnic cohort. Stroke. 2018;49(2):363–9.
8.
Wolf PA, D’Agostino RB, Belanger AJ, Kannel WB. Probability of stroke: a risk profile from the Framingham Study. Stroke. 1991;22(3):312–8.
9.
Kuller LH, Longstreth WTJr, Arnold AM, Bernick C, Bryan RN, Beauchamp NJJr, et al. White matter hyperintensity on cranial magnetic resonance imaging: a predictor of stroke. Stroke. 2004;35(8):1821–5.
10.
Koton S, Schneider ALC, Windham BG, Mosley TH, Gottesman RF, Coresh J. Microvascular brain disease progression and risk of stroke: the ARIC Study. Stroke. 2020;51(11):3264–70.
11.
Kent DM, Leung LY, Zhou Y, Luetmer PH, Kallmes DF, Nelson J, et al. Association of incidentally discovered covert cerebrovascular disease identified using natural language processing and future dementia. J Am Heart Assoc. 2023;12(1):e027672.