I can see absolutely zero use for this information. NOTHING HERE helps survivors recover. Predicting failure to recover helps no one. WILL YOU PLEASE JUST SOLVE STROKE?
Global white matter structural integrity mediates the effect of age on ischemic stroke outcomes
Mark R Ethertonhttps://orcid.org/0000-0002-4739-24491, Markus D Schirmer1,2, Maria Clara Zanon Zotin1,3, Pamela M Rist4, Gregoire Boulouis1, Arne Lauer1, Ona Wu1,5, and Natalia S Rost1
Background
Background
The relationship of global white matter microstructural integrity and ischemic stroke outcomes is not well understood.
Aims
Aims
To investigate the relationship of global white matter microstructural integrity with clinical variables and functional outcomes after acute ischemic stroke.
Methods
Methods
A retrospective analysis of neuroimaging data from 300 acute ischemic stroke patients with magnetic resonance imaging brain obtained within 48 hours of stroke onset and long-term functional outcomes (modified Rankin, mRS) was performed. Peak width of skeletonized mean diffusivity (PSMD), as a measure of global white matter microstructural injury, was calculated in the hemisphere contralateral to the acute infarct. Multivariable linear and logistic regression analyses were performed to identify variables associated with PSMD and excellent functional outcome (mRS < 2) at 90 days, respectively. Mediation analysis was then pursued to characterize how PSMD mediates the effect of age on acute ischemic stroke functional outcomes.
Results
Results
White matter hyperintensity volume, age, pre-stroke disability, and normal-appearing white matter mean diffusivity were independently associated with increased PSMD. In logistic regression analysis, increased infarct volume and PSMD were independent predictors of excellent functional outcome. Additionally, the effect of age on functional outcomes was indirectly mediated by PSMD (P < 0.001).
Conclusions
Conclusions
As a marker of global white matter microstructural injury, increased PSMD mediates the effect of increased age to contribute to poor acute ischemic stroke functional outcomes. PSMD could serve as a putative radiographic marker of brain age for stroke outcomes prognostication.
Keywords
Ischemic stroke, leukoaraiosis, MRI
1JPK Stroke Research Center, Department of Neurology, Massachusetts General Hospital (MGH) and Harvard Medical School, Boston, MA, USA
2Clinic for Neuroradiology, University Hospital Bonn, Bonn, Germany
3Center for Imaging Sciences and Medical Physics, Department of Medical Imaging, Hematology and Clinical Oncology, Ribeirão Preto Medical School, University of São Paulo, Ribeirão Preto, SP, Brazil
4Division of Preventive Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA
5Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, MGH, Boston, MA, USA
Corresponding author(s):
Mark R Etherton, JPK Stroke Research Center, Massachusetts General Hospital, 175 Cambridge Street, Suite 300, Boston, MA 02114, USA. Email: metherton@partners.org
Introduction
In patients with acute ischemic stroke (AIS), cerebral small vessel disease (cSVD) is highly prevalent and a recognized risk factor for poor AIS outcomes.1,2 As one radiographic marker of cSVD, neuroimaging evidence of decreased white matter structural integrity predisposes to poor AIS outcomes by contributing to infarct growth, and, independent of the ischemic tissue outcome, ostensibly affecting stroke recovery.1,3,4 Several radiographic modalities measuring chronic cSVD-mediated white matter injury, including white matter hyperintensities (WMH),5 commonly seen on Fluid-attenuated inversion recovery (FLAIR) MRI sequences, and Diffusion Tensor Imaging (DTI)-based analysis of normal-appearing white matter (NAWM), are associated with poor AIS outcomes.4,6 While WMH volume and NAWM structural integrity are associated with AIS outcomes, they represent subpopulations of cerebral white matter. An open question, therefore, is whether a total assessment of white matter structural injury may better represent cSVD burden and inform on AIS outcomes.
The peak width of skeletonized mean diffusivity (PSMD), a fully-automated DTI-based measure of global white matter microstructural injury,7 represents one appealing cSVD marker for potential application in AIS populations. From the DTI sequences, PSMD captures a tract-based quantification of the heterogeneity of mean diffusivity (MD) in principal white matter tracts to inform on white matter microstructural integrity.7 Importantly, PSMD has been shown to have high inter-scanner reproducibility and important associations with cognitive function in population-based studies and cSVD cohorts, outperforming other neuroimaging markers.7–9 While preliminary data suggests the utility of PSMD as a cSVD marker in healthy aging and cSVD cohorts, it is unknown whether it can be applied to AIS populations and if it is a clinically relevant determinant of AIS outcomes.
Aims
To test these questions, we measured PSMD in an AIS cohort with long-term functional outcomes. Our hypothesis was that increased PSMD would be an independent determinant of poor long-term functional outcomes. The aims of this study were: (1) to identify the clinical and radiographic variables in AIS associated with PSMD; (2) to compare the association of PSMD with long-term functional outcomes to other radiographic markers of white matter structural injury, like WMH volume and NAWM MD; and (3) to determine how PSMD mediates the effect of age on long-term functional outcomes.
Methods
Study design and participants
This study is a retrospective analysis of a single-center prospective study of AIS patients.
The details of study design and participants have been described previously (see Supplementary Materials).4 In brief, all participants 18 years of age or older that presented to the Massachusetts General Hospital Emergency Department between 2003 and 2011 with signs and symptoms consistent with an AIS were eligible for enrollment in the prospective study. Additional inclusion criteria for this analysis included: (1) patients with an acute, unilateral, supratentorial infarct on MRI brain, defined as a DWI-positive lesion compatible with the presenting symptoms; (2) within 48 h of symptom onset; and (3) FLAIR and DTI sequences available for WMH quantification and DTI analysis.
Clinical assessments
Admission stroke severity and pre-stroke disability were assessed by a trained neurologist using the National Institutes of Health Stroke Scale score (NIHSS) and modified Rankin scale (mRS), respectively. Large-vessel occlusion (LVO) was defined as occlusion of the intracranial internal carotid artery, or middle cerebral artery M1 and proximal M2 segments. Functional outcomes were assessed between three and six months post-stroke at a median follow-up time of 154 days from index event (IQR 104–208.5 days).
MRI acquisition and processing
Briefly, a multimodal, whole-brain MRI protocol with axial FLAIR, DWI, and DTI sequences was acquired using a 1.5 T MRI scanner (General Electric Signa scanner) within 48 h of admission (see Supplementary Materials). DWI volume (DWIv) of the acute infarct was determined using a semi-automated approach and normalized for intracranial volume.10
WMH and NAWM segmentation were performed using MRIcro (www.mricro.com) as described previously.4 WMH and chronic infarct masks were constructed on FLAIR images by a blinded expert. To create the probabilistic NAWM mask (90%) in the contralesional hemisphere, all images and masks were subsequently coregistered to the MNI 152 1 mm atlas.11–13 WMHv was normalized to intracranial volume to adjust for differences in head size.14 NAWM MD and FA were extracted as the median voxel values from the contralesional NAWM mask. PSMD was extracted from the contralesional hemisphere using a publicly available script (PSMD Marker Version 1.0; see Supplementary Materials for details, Figure 1).
Statistical analysis
All statistical analyses were performed in R Version 4.0.0. Pre-stroke disability was defined as an mRS ≥2. Excellent functional outcome was dichotomized as mRS < 2. As appropriate, Student’s t-test, Wilcoxon rank sum, and Chi square test were used to evaluate for baseline differences in the group with excellent versus poor outcome. The details of our model building and statistical analysis approaches are described in the supplement. In brief, we performed univariate linear regression to identify clinical and radiographic variables associated with PSMD and then included all statistically significant univariate predictors in a multivariate model. Separate logistic regression models were used to identify clinical or radiographic variables associated with excellent functional outcomes. Variables which were significant in either the clinical or radiographic model were considered for inclusion in a combined model which was constructed using backward stepwise logistic regression (see Supplementary Materials). Receiver operating characteristic (ROC) curves were constructed to compare model performance for prediction of excellent functional outcome. Mediation analysis was performed using the Lavaan package to estimate the direct and indirect causal mediation effects of age and PSMD on mRS score.15
Results
The baseline demographics of the study population have been reported previously.4,10 In brief, of 481 initially enrolled participants, after excluding participants with the aforementioned exclusion criteria or with motion artifact precluding MRI processing and assessment, 300 participants were analyzed in this study (see Supplementary Figure 1), including 74 with chronic contralesional infarcts. Age at enrollment ranged from 18 to 101 years. Median admission NIHSS score was 3 (interquartile range, IQR 1–7) and follow-up mRS was 1 (IQR 0–3). In comparison with the poor outcome group, AIS patients with excellent outcomes were younger, more likely to be male, had lower rates of atrial fibrillation, pre-stroke disability, and LVO, lower NIHSS and DWI volume, and decreased PSMD (Table 1 and Figure 1).
Figure 1. PSMD analysis in the hemisphere contralateral to the acute infarct. Representative (a) FLAIR image; (b) MD image with contralesional skeletonized FA mask shown for PSMD calculation (red); (c) FLAIR image with probabilistic NAWM mask in the contralesional hemisphere. (d) Box plot of PSMD in AIS patients with excellent versus poor outcomes.
PSMD: peak width of skeletonized mean diffusivity; mRS: modified Rankin scale.
In univariate linear regression analysis for determinants of PSMD, increasing age, atrial fibrillation, coronary artery disease, hypertension, and pre-stroke disability were associated with increased PSMD (Table 2). Radiographic markers of increased white matter structural injury, WMHv, and NAWM MD were also associated with increased PSMD. In the multivariable linear regression model, pre-stroke disability and increased age, WMHv, and NAWM MD were independently associated with increased contralesional PSMD (Table 2). Relative importance metrics analysis showed age, and NAWM MD explained the largest proportion of variance of PSMD (Supplementary Tables 1 and 2).
Next, we pursued logistic regression analysis for determinants of excellent functional outcome. In univariable analysis, increasing age, female sex, atrial fibrillation, pre-stroke disability, alteplase treatment, admission NIHSS, DWIv, NAWM MD, and PSMD were associated with decreased likelihood of excellent functional outcome (Table 3). Using the significant variables in the univariable analysis, we constructed clinical and radiographic regression models for predictors of excellent functional outcome. In the clinical model, increasing age, admission NIHSS, and female sex were independent predictors of poor functional outcomes. In the radiographic model, increased DWIv and PSMD were independent predictors of poor functional outcomes (Table 4). Sensitivity analyses further demonstrated that increased PSMD was associated with poor functional outcomes in the LVO stroke and NIHSS >6 subgroups (Supplementary Table 3). Next, we performed backward stepwise logistic regression analysis to define the minimal regression model for excellent functional outcomes using the variables from the clinical and radiographic models. To test the candidacy of PSMD as a putative marker of biologic age and given the strong effect of age on PSMD, separate multivariable regression analyses were performed in which we allowed either age or PSMD to enter the model. In the Age final model, age, female sex, pre-stroke disability, NIHSS, and DWIv were independent determinants of poor outcome (Supplementary Table 4, AUC = 0.78). In contrast, in the PSMD final model, PSMD, NIHSS, DWIv, and female sex were independent predictors of poor outcome (Table 4, AUC = 0.79). ROC curve analysis showed that the PSMD final model had the largest AUC for predicting excellent functional outcome compared to the clinical and radiographic models (Supplementary Figure 2; AUC = 0.79) and that including PSMD in the final model resulted in a small, but statistically significant improvement in model performance (Supplementary Figure 3; final +PSMD: AUC = 0.79 vs. final – PSMD: AUC = 0.76; P value <0.001).
Lastly, we pursued mediation analysis to evaluate the direct and indirect causal mediation effects of age and PSMD on functional outcomes (mRS, Figure 2). The total effect of age on functional outcome (mRS) was significant (β = 2.0; P < 0.001). Additionally, both the direct effect of age on mRS (β = 1.08; P < 0.012) and indirect effects of age on PSMD (β = 0.687; P < 0.001) and PSMD on mRS (β = 1.37; P < 0.001) were significant, suggesting that the effect of increasing age on poor functional outcomes after ischemic stroke is partially mediated by PSMD.
Figure 2. PSMD mediates the effect of age on 90-day modified Rankin Scale. Mediation analysis for the model evaluating the effect of age and PSMD on 90-day modified Rankin Scale. Bold arrows indicate P < 0.05. Standardized path coefficients are used as effect estimate.
PSMD: peak width of skeletonized mean diffusivity; mRS: modified Rankin scale.
Discussion
In this study of PSMD in AIS, we observed several important findings. First, increased age, pre-stroke disability, and markers of white matter structural injury (WMH volume and NAWM DTI measures) were associated with increased PSMD. Second, PSMD is an independent determinant of excellent long-term functional outcomes in regression models including WMH volume and NAWM DTI measures. Lastly, we demonstrate that PSMD partially mediates the effect of increasing age on poor post-stroke functional outcomes.
As a marker of global white matter microstructural integrity, we observed that increasing PSMD values were associated with increased age, pre-stroke disability, increased WMH, and NAWM DTI measures. To our knowledge, this is the first report that PSMD is associated with increased age, pre-stroke disability, and white matter macrostructural injury in patients with AIS. These observations also extend prior observations in AIS patients, showing that increasing age is strongly predictive of greater white matter micro- (NAWM) and macrostructural injury (WMH).16 Our findings are also in agreement with similar observations of a relationship between PSMD and age in patients with sporadic and genetic forms of cSVD,7 stroke-free individuals,17 hemorrhagic variants of cSVD (i.e. cerebral amyloid angiopathy),8,18 and vascular cognitive impairment and dementia.7 Collectively, the association between PSMD and age in multiple clinical cohorts suggest a broader potential clinical relevance for PSMD as a radiographic marker of the ageing process.
In the analysis of determinants of post-stroke functional outcomes, only PSMD and DWI volume were significant predictors of excellent functional outcomes when included in a multivariable model with WMH and NAWM DTI metrics. As a marker of global injury, PSMD probably reflects structural injury both in the form of WMH and microstructural injury, based on the strong association between these variables in our regression analysis. Both WMH volume and NAWM microstructural injury have previously been shown to be predictive of poor post-stroke functional outcomes.1,4,19,20 In contrast to WMH volume and NAWM diffusivity anisotropy metrics, PSMD was the only variable representing white matter structural injury that remained in the final multivariable analysis for radiographic predictors of excellent functional outcomes. This observation supports the hypothesis that PSMD effectively integrates different forms of white matter injury that occur in the context of an adverse vascular risk profile and, as such, co-determines stroke outcomes. PSMD may therefore better reflect total pre-stroke white matter structural injury than WMHv or NAWM diffusivity measures, which is supported by our observation that pre-stroke disability is associated with PSMD.
To inform on the underlying etiology of how PSMD contributes to poor post-stroke functional outcomes and delineate the causal pathway, we pursued mediation analysis. We observed that part of the effect of increased age on poor post-stroke functional outcomes is mediated indirectly by PSMD. The direct effect of increased age on poor post-stroke functional outcomes is likely multifactorial with contributions from both clinical and social factors. Our results, however, suggest that PSMD is an important determinant of AIS functional outcomes, in part, by indirectly mediating the effect of age.
Our study has several important limitations. First, this was a retrospective, cross-sectional analysis of a prospective AIS study of patients with mild stroke severity (admission NIHSS score 3). Supporting the hypothesis that white matter microstructural integrity similarly influences outcomes for more severe stroke syndromes, however, we observed the same associations in the subgroups with NIHSS > 6 and LVOs (see Supplementary Materials). Second, this analysis was restricted to unilateral, supratentorial AIS patients to allow quantification of DTI measures in the contralesional hemisphere as an approach to offset any influence of the acute infarct on the DTI metrics, which could introduce selection bias. While 25% of the patient population had chronic infarcts contralesional to the acute infarct, future studies should include infratentorial strokes to broaden generalizability. Further, whether PSMD values in the ipsilesional hemisphere provide any meaningful information remains unanswered. Differences in the DTI acquisition protocols between scanners could presumably be a source of error. Prior reports, however, have demonstrated that differences in diffusion gradient direction or excitations do not substantially alter the mean FA, MD or PSMD values.8,21 Baykara et al. also reported high inter-scanner reproducibility of PSMD with the highest intraclass correlation coefficient of the white matter structural injury radiographic markers, suggesting that these markers are generally stable across different protocols.7 Overall, we would maintain that this analysis of clinical MRI scans in AIS patients is a strength of this study, as the observed relationship in this setting suggests a more widespread generalizability to other AIS populations.
In summary, in a cohort of AIS patients, we show that PSMD is a clinically feasible measure of global white matter microstructural injury that partially mediates of the effect of age on post-stroke functional outcomes. Future studies investigating the association of PSMD with other stroke outcome domains and comparing its utility to qualitative cSVD scores22 are warranted.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Keywords
Ischemic stroke, leukoaraiosis, MRI
1JPK Stroke Research Center, Department of Neurology, Massachusetts General Hospital (MGH) and Harvard Medical School, Boston, MA, USA
2Clinic for Neuroradiology, University Hospital Bonn, Bonn, Germany
3Center for Imaging Sciences and Medical Physics, Department of Medical Imaging, Hematology and Clinical Oncology, Ribeirão Preto Medical School, University of São Paulo, Ribeirão Preto, SP, Brazil
4Division of Preventive Medicine, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, MA, USA
5Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, MGH, Boston, MA, USA
Corresponding author(s):
Mark R Etherton, JPK Stroke Research Center, Massachusetts General Hospital, 175 Cambridge Street, Suite 300, Boston, MA 02114, USA. Email: metherton@partners.org
Introduction
In patients with acute ischemic stroke (AIS), cerebral small vessel disease (cSVD) is highly prevalent and a recognized risk factor for poor AIS outcomes.1,2 As one radiographic marker of cSVD, neuroimaging evidence of decreased white matter structural integrity predisposes to poor AIS outcomes by contributing to infarct growth, and, independent of the ischemic tissue outcome, ostensibly affecting stroke recovery.1,3,4 Several radiographic modalities measuring chronic cSVD-mediated white matter injury, including white matter hyperintensities (WMH),5 commonly seen on Fluid-attenuated inversion recovery (FLAIR) MRI sequences, and Diffusion Tensor Imaging (DTI)-based analysis of normal-appearing white matter (NAWM), are associated with poor AIS outcomes.4,6 While WMH volume and NAWM structural integrity are associated with AIS outcomes, they represent subpopulations of cerebral white matter. An open question, therefore, is whether a total assessment of white matter structural injury may better represent cSVD burden and inform on AIS outcomes.
The peak width of skeletonized mean diffusivity (PSMD), a fully-automated DTI-based measure of global white matter microstructural injury,7 represents one appealing cSVD marker for potential application in AIS populations. From the DTI sequences, PSMD captures a tract-based quantification of the heterogeneity of mean diffusivity (MD) in principal white matter tracts to inform on white matter microstructural integrity.7 Importantly, PSMD has been shown to have high inter-scanner reproducibility and important associations with cognitive function in population-based studies and cSVD cohorts, outperforming other neuroimaging markers.7–9 While preliminary data suggests the utility of PSMD as a cSVD marker in healthy aging and cSVD cohorts, it is unknown whether it can be applied to AIS populations and if it is a clinically relevant determinant of AIS outcomes.
Aims
To test these questions, we measured PSMD in an AIS cohort with long-term functional outcomes. Our hypothesis was that increased PSMD would be an independent determinant of poor long-term functional outcomes. The aims of this study were: (1) to identify the clinical and radiographic variables in AIS associated with PSMD; (2) to compare the association of PSMD with long-term functional outcomes to other radiographic markers of white matter structural injury, like WMH volume and NAWM MD; and (3) to determine how PSMD mediates the effect of age on long-term functional outcomes.
Methods
Study design and participants
This study is a retrospective analysis of a single-center prospective study of AIS patients.
The details of study design and participants have been described previously (see Supplementary Materials).4 In brief, all participants 18 years of age or older that presented to the Massachusetts General Hospital Emergency Department between 2003 and 2011 with signs and symptoms consistent with an AIS were eligible for enrollment in the prospective study. Additional inclusion criteria for this analysis included: (1) patients with an acute, unilateral, supratentorial infarct on MRI brain, defined as a DWI-positive lesion compatible with the presenting symptoms; (2) within 48 h of symptom onset; and (3) FLAIR and DTI sequences available for WMH quantification and DTI analysis.
Clinical assessments
Admission stroke severity and pre-stroke disability were assessed by a trained neurologist using the National Institutes of Health Stroke Scale score (NIHSS) and modified Rankin scale (mRS), respectively. Large-vessel occlusion (LVO) was defined as occlusion of the intracranial internal carotid artery, or middle cerebral artery M1 and proximal M2 segments. Functional outcomes were assessed between three and six months post-stroke at a median follow-up time of 154 days from index event (IQR 104–208.5 days).
MRI acquisition and processing
Briefly, a multimodal, whole-brain MRI protocol with axial FLAIR, DWI, and DTI sequences was acquired using a 1.5 T MRI scanner (General Electric Signa scanner) within 48 h of admission (see Supplementary Materials). DWI volume (DWIv) of the acute infarct was determined using a semi-automated approach and normalized for intracranial volume.10
WMH and NAWM segmentation were performed using MRIcro (www.mricro.com) as described previously.4 WMH and chronic infarct masks were constructed on FLAIR images by a blinded expert. To create the probabilistic NAWM mask (90%) in the contralesional hemisphere, all images and masks were subsequently coregistered to the MNI 152 1 mm atlas.11–13 WMHv was normalized to intracranial volume to adjust for differences in head size.14 NAWM MD and FA were extracted as the median voxel values from the contralesional NAWM mask. PSMD was extracted from the contralesional hemisphere using a publicly available script (PSMD Marker Version 1.0; see Supplementary Materials for details, Figure 1).
Statistical analysis
All statistical analyses were performed in R Version 4.0.0. Pre-stroke disability was defined as an mRS ≥2. Excellent functional outcome was dichotomized as mRS < 2. As appropriate, Student’s t-test, Wilcoxon rank sum, and Chi square test were used to evaluate for baseline differences in the group with excellent versus poor outcome. The details of our model building and statistical analysis approaches are described in the supplement. In brief, we performed univariate linear regression to identify clinical and radiographic variables associated with PSMD and then included all statistically significant univariate predictors in a multivariate model. Separate logistic regression models were used to identify clinical or radiographic variables associated with excellent functional outcomes. Variables which were significant in either the clinical or radiographic model were considered for inclusion in a combined model which was constructed using backward stepwise logistic regression (see Supplementary Materials). Receiver operating characteristic (ROC) curves were constructed to compare model performance for prediction of excellent functional outcome. Mediation analysis was performed using the Lavaan package to estimate the direct and indirect causal mediation effects of age and PSMD on mRS score.15
Results
The baseline demographics of the study population have been reported previously.4,10 In brief, of 481 initially enrolled participants, after excluding participants with the aforementioned exclusion criteria or with motion artifact precluding MRI processing and assessment, 300 participants were analyzed in this study (see Supplementary Figure 1), including 74 with chronic contralesional infarcts. Age at enrollment ranged from 18 to 101 years. Median admission NIHSS score was 3 (interquartile range, IQR 1–7) and follow-up mRS was 1 (IQR 0–3). In comparison with the poor outcome group, AIS patients with excellent outcomes were younger, more likely to be male, had lower rates of atrial fibrillation, pre-stroke disability, and LVO, lower NIHSS and DWI volume, and decreased PSMD (Table 1 and Figure 1).
Figure 1. PSMD analysis in the hemisphere contralateral to the acute infarct. Representative (a) FLAIR image; (b) MD image with contralesional skeletonized FA mask shown for PSMD calculation (red); (c) FLAIR image with probabilistic NAWM mask in the contralesional hemisphere. (d) Box plot of PSMD in AIS patients with excellent versus poor outcomes.
PSMD: peak width of skeletonized mean diffusivity; mRS: modified Rankin scale.
In univariate linear regression analysis for determinants of PSMD, increasing age, atrial fibrillation, coronary artery disease, hypertension, and pre-stroke disability were associated with increased PSMD (Table 2). Radiographic markers of increased white matter structural injury, WMHv, and NAWM MD were also associated with increased PSMD. In the multivariable linear regression model, pre-stroke disability and increased age, WMHv, and NAWM MD were independently associated with increased contralesional PSMD (Table 2). Relative importance metrics analysis showed age, and NAWM MD explained the largest proportion of variance of PSMD (Supplementary Tables 1 and 2).
Next, we pursued logistic regression analysis for determinants of excellent functional outcome. In univariable analysis, increasing age, female sex, atrial fibrillation, pre-stroke disability, alteplase treatment, admission NIHSS, DWIv, NAWM MD, and PSMD were associated with decreased likelihood of excellent functional outcome (Table 3). Using the significant variables in the univariable analysis, we constructed clinical and radiographic regression models for predictors of excellent functional outcome. In the clinical model, increasing age, admission NIHSS, and female sex were independent predictors of poor functional outcomes. In the radiographic model, increased DWIv and PSMD were independent predictors of poor functional outcomes (Table 4). Sensitivity analyses further demonstrated that increased PSMD was associated with poor functional outcomes in the LVO stroke and NIHSS >6 subgroups (Supplementary Table 3). Next, we performed backward stepwise logistic regression analysis to define the minimal regression model for excellent functional outcomes using the variables from the clinical and radiographic models. To test the candidacy of PSMD as a putative marker of biologic age and given the strong effect of age on PSMD, separate multivariable regression analyses were performed in which we allowed either age or PSMD to enter the model. In the Age final model, age, female sex, pre-stroke disability, NIHSS, and DWIv were independent determinants of poor outcome (Supplementary Table 4, AUC = 0.78). In contrast, in the PSMD final model, PSMD, NIHSS, DWIv, and female sex were independent predictors of poor outcome (Table 4, AUC = 0.79). ROC curve analysis showed that the PSMD final model had the largest AUC for predicting excellent functional outcome compared to the clinical and radiographic models (Supplementary Figure 2; AUC = 0.79) and that including PSMD in the final model resulted in a small, but statistically significant improvement in model performance (Supplementary Figure 3; final +PSMD: AUC = 0.79 vs. final – PSMD: AUC = 0.76; P value <0.001).
Lastly, we pursued mediation analysis to evaluate the direct and indirect causal mediation effects of age and PSMD on functional outcomes (mRS, Figure 2). The total effect of age on functional outcome (mRS) was significant (β = 2.0; P < 0.001). Additionally, both the direct effect of age on mRS (β = 1.08; P < 0.012) and indirect effects of age on PSMD (β = 0.687; P < 0.001) and PSMD on mRS (β = 1.37; P < 0.001) were significant, suggesting that the effect of increasing age on poor functional outcomes after ischemic stroke is partially mediated by PSMD.
Figure 2. PSMD mediates the effect of age on 90-day modified Rankin Scale. Mediation analysis for the model evaluating the effect of age and PSMD on 90-day modified Rankin Scale. Bold arrows indicate P < 0.05. Standardized path coefficients are used as effect estimate.
PSMD: peak width of skeletonized mean diffusivity; mRS: modified Rankin scale.
Discussion
In this study of PSMD in AIS, we observed several important findings. First, increased age, pre-stroke disability, and markers of white matter structural injury (WMH volume and NAWM DTI measures) were associated with increased PSMD. Second, PSMD is an independent determinant of excellent long-term functional outcomes in regression models including WMH volume and NAWM DTI measures. Lastly, we demonstrate that PSMD partially mediates the effect of increasing age on poor post-stroke functional outcomes.
As a marker of global white matter microstructural integrity, we observed that increasing PSMD values were associated with increased age, pre-stroke disability, increased WMH, and NAWM DTI measures. To our knowledge, this is the first report that PSMD is associated with increased age, pre-stroke disability, and white matter macrostructural injury in patients with AIS. These observations also extend prior observations in AIS patients, showing that increasing age is strongly predictive of greater white matter micro- (NAWM) and macrostructural injury (WMH).16 Our findings are also in agreement with similar observations of a relationship between PSMD and age in patients with sporadic and genetic forms of cSVD,7 stroke-free individuals,17 hemorrhagic variants of cSVD (i.e. cerebral amyloid angiopathy),8,18 and vascular cognitive impairment and dementia.7 Collectively, the association between PSMD and age in multiple clinical cohorts suggest a broader potential clinical relevance for PSMD as a radiographic marker of the ageing process.
In the analysis of determinants of post-stroke functional outcomes, only PSMD and DWI volume were significant predictors of excellent functional outcomes when included in a multivariable model with WMH and NAWM DTI metrics. As a marker of global injury, PSMD probably reflects structural injury both in the form of WMH and microstructural injury, based on the strong association between these variables in our regression analysis. Both WMH volume and NAWM microstructural injury have previously been shown to be predictive of poor post-stroke functional outcomes.1,4,19,20 In contrast to WMH volume and NAWM diffusivity anisotropy metrics, PSMD was the only variable representing white matter structural injury that remained in the final multivariable analysis for radiographic predictors of excellent functional outcomes. This observation supports the hypothesis that PSMD effectively integrates different forms of white matter injury that occur in the context of an adverse vascular risk profile and, as such, co-determines stroke outcomes. PSMD may therefore better reflect total pre-stroke white matter structural injury than WMHv or NAWM diffusivity measures, which is supported by our observation that pre-stroke disability is associated with PSMD.
To inform on the underlying etiology of how PSMD contributes to poor post-stroke functional outcomes and delineate the causal pathway, we pursued mediation analysis. We observed that part of the effect of increased age on poor post-stroke functional outcomes is mediated indirectly by PSMD. The direct effect of increased age on poor post-stroke functional outcomes is likely multifactorial with contributions from both clinical and social factors. Our results, however, suggest that PSMD is an important determinant of AIS functional outcomes, in part, by indirectly mediating the effect of age.
Our study has several important limitations. First, this was a retrospective, cross-sectional analysis of a prospective AIS study of patients with mild stroke severity (admission NIHSS score 3). Supporting the hypothesis that white matter microstructural integrity similarly influences outcomes for more severe stroke syndromes, however, we observed the same associations in the subgroups with NIHSS > 6 and LVOs (see Supplementary Materials). Second, this analysis was restricted to unilateral, supratentorial AIS patients to allow quantification of DTI measures in the contralesional hemisphere as an approach to offset any influence of the acute infarct on the DTI metrics, which could introduce selection bias. While 25% of the patient population had chronic infarcts contralesional to the acute infarct, future studies should include infratentorial strokes to broaden generalizability. Further, whether PSMD values in the ipsilesional hemisphere provide any meaningful information remains unanswered. Differences in the DTI acquisition protocols between scanners could presumably be a source of error. Prior reports, however, have demonstrated that differences in diffusion gradient direction or excitations do not substantially alter the mean FA, MD or PSMD values.8,21 Baykara et al. also reported high inter-scanner reproducibility of PSMD with the highest intraclass correlation coefficient of the white matter structural injury radiographic markers, suggesting that these markers are generally stable across different protocols.7 Overall, we would maintain that this analysis of clinical MRI scans in AIS patients is a strength of this study, as the observed relationship in this setting suggests a more widespread generalizability to other AIS populations.
In summary, in a cohort of AIS patients, we show that PSMD is a clinically feasible measure of global white matter microstructural injury that partially mediates of the effect of age on post-stroke functional outcomes. Future studies investigating the association of PSMD with other stroke outcome domains and comparing its utility to qualitative cSVD scores22 are warranted.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
More at link.
No comments:
Post a Comment