key: cord-0258670-l7s43j0p
authors: Jochems, A. C. C.; Arteaga, C.; Chappell, F. M.; Ritakari, T.; Hooley, M.; Doubal, F. N.; Munoz Maniega, S.; Wardlaw, J. M.
title: Longitudinal changes of white matter hyperintensities in sporadic small vessel disease: a systematic review and meta-analysis
date: 2022-03-23
journal: nan
DOI: 10.1101/2022.03.22.22272696
sha: 4a5e146bd0d258c6c9dd5fa042a3ffaafb872141
doc_id: 258670
cord_uid: l7s43j0p

White matter hyperintensities (WMH) are frequent imaging features of small vessel disease (SVD) and related to poor clinical outcomes. WMH progression over time is well described, but regression was also noted recently, although the frequency and associated factors are unknown. This systematic review and meta-analysis aims to assess longitudinal intra-individual WMH volume changes in sporadic SVD. Following PRISMA guidelines we searched EMBASE and MEDLINE for papers up to 28 January 2022 on WMH volume changes using MRI on [≥]2 time-points in adults with sporadic SVD. We classified populations (healthy/community-dwelling, stroke, cognitive, other vascular risk factors, depression) based on study characteristics. We performed random-effects meta-analyses with Knapp-Hartung adjustment to determine mean WMH volume change (change in mL, % of intracranial volume [%ICV], or mL/year), 95%CI and prediction intervals (PI, limits of increase and decrease) using unadjusted data. Risk of Bias Assessment tool for Non-randomised Studies (RoBANS) was used to assess risk of bias. We included 40 papers, 10,932 participants. Mean WMH volume increased over time by: 1.74 mL (95% CI 1.23, 2.26; PI -1.24, 4.73 mL; 27 papers, N=7411, mean time interval 2.7 years, SD=1.65); 0.25%ICV (95% CI 0.14, 0.36; PI -0.06, 0.56; 6 papers, N=1071, mean time interval 3.5 years, SD =1.54); or 0.61 mL/year (95% CI 0.37, 0.85; PI -0.25, 1.46; 7 papers, N=2450). Additionally, 13 papers specifically mentioned and/or provided data on WMH regression, which occurred in asymptomatic, stroke, and cognitive presentations of SVD. Net mean WMH volume increases over time (e.g. 1[3/4] mL) mask wide-ranging change (e.g.1[1/4]mL decrease to 4[3/4] increase), with regression documented explicitly in up to 1/3 of participants. More knowledge on underlying mechanisms, associated factors and clinical correlates is needed, as WMH regression could be an important intervention target.

White matter hyperintensities (WMH) of presumed vascular origin are the most common neuroimaging feature of small vessel disease (SVD), a disorder of the cerebral microvessels.

WMH are visible on MRI as hyperintense lesions on T2-weighted and hypointense on T1weighted sequences. 1 Around 11-15% of a general middle-aged population have one or more WMH, increasing to over 90% in people aged over 80. WMH are associated with risk factors including age, hypertension, smoking, and diabetes 2 and with symptoms such as apathy, fatigue, delirium, cognitive decline and increased risk of stroke and dementia. 3 WMH are thought to indicate areas of permanent white matter damage due to demyelination and axon loss. 3 In longitudinal studies, WMH progression over time is a common finding, while WMH regression has only been noted in a few recent studies, 4, 5 but otherwise has been disregarded as measurement error or overlooked. If a genuine finding, then WMH regression might suggest that WMH do not only indicate permanently damaged brain tissue.

A better understanding of the frequency and factors associated with WMH regression could help identify potential interventions to delay WMH progression and the devastating clinical consequences.

We performed this systematic review and meta-analysis to assess longitudinal intraindividual WMH volume changes in sporadic SVD over time using volumetric MRI measurements and progression rates.

We searched EMBASE and MEDLINE from January 1985, when MRI became more widely implemented in clinical practice, to January 28, 2022, for studies investigating longitudinal quantification of WMH volume on MRI on at least two different times points in adults over 18 years old (eAppendix 1: Search strategy). We designed the search strategy and complied with the Preferred Reporting in Systematic Review and Meta-Analysis (PRISMA) guidelines.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. We included published full-text articles from peer-reviewed longitudinal studies that used MRI to quantify WMH volumes, defined according to STRIVE criteria 1 , on at least two different time points and that provided numerical analysis of WMH volume change between the time points. Studies included randomised trials, non-randomised trials, cohort studies, and case-control studies. For data from a single cohort published more than once, we included the most relevant paper with the largest sample size and years of follow-up and most useable data to minimise duplication or overlapping samples. We excluded studies of SVD attributable to hereditary causes (e.g., CADASIL, CARASIL) or WMH attributable to other causes (e.g., multiple sclerosis, inflammatory disorders including primary angiitis, secondary vasculitis, post-infectious, and paraneoplastic syndromes).

Title and abstract screening and duplicate paper removal were done independently by one reviewer (CA, ACCJ, TR, MH). A second reviewer screened a random 10% sample of titles and abstracts. Full-text review was assessed independently by one reviewer, and a second reviewer screened a random 20% sample of the full texts (CA, ACCJ) using Covidence software (https://www.covidence.org/). Data extraction was performed by a single reviewer using a pre-specified data collection form (eTable 1), and a second reviewer doubleextracted a random 20% sample. Two reviewers used RoBANS (risk of bias assessment tool for non-randomised studies) 6 to assess all studies including randomised trials, for participant selection, adjustment for confounders, adequacy of WMH descriptors, blinding, incomplete outcome data, and selective outcome reporting (eTable 2). Any disagreements were resolved by discussion between reviewers (CA, ACCJ) with the help of a senior reviewer (JMW).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 

We extracted information on study design, demographic characteristics, vascular risk factors, study population (i.e. healthy and/or community-dwelling population (healthy/CD), patients with depression, patients with stroke, cognitive presentation or other vascular risk factors (VRF) presentations), and follow-up period.

We extracted data related to MRI assessment, including WMH volumes per time point and WMH change between those time points, adjustment of WMH measurements to e.g., intracranial volume, total brain volume; methods of WMH calculation and predictors of change. WMH volume data are often skewed and are log transformed to normalise the data before being used in analyses. We aimed to use unadjusted raw data where possible to reflect real intra-individual changes and include the whole range of least and most growth.

We selected studies that provided unadjusted mean raw WMH volume change, e.g. mL, cm 3 and cc, changes in volume expressed as a percentage of intracranial volume (%ICV), or an annual change rate (e.g. mL per year). We extracted the mean or median with corresponding SD, IQR, range, 95% CI, SEM, and if provided, the baseline WMH volumes. If studies did not provide mean and SD of WMH volume change, we calculated the mean and standard deviation from the sample size, median, range, IQR, SEM or 95% CI. 7 We extracted data from studies that reported volumes per group and for the entire cohort; when the volume for the entire cohort was not available, data for the separate groups were included. For studies that did not report complete volumetric assessments, or where the data were not useable in the meta-analysis, we extracted relevant measures to perform a narrative summary of findings.

We used random-effects meta-analyses (because of expected between-study heterogeneity) to calculate mean WMH change, its confidence interval and prediction interval (PI) using untransformed raw means. More than two studies are needed to calculate a PI. The PI is an estimate of an interval, based on data that have already been observed and are included in . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 23, 2022. ; https://doi.org/10.1101/2022.03.22.22272696 doi: medRxiv preprint the meta-analysis, in which future observations will fall, with 95% confidence. In contrast to the CI, it indicates the range of least and most WMH volume change in a sample, while the CI indicates the range of mean change. We applied the Knapp-Hartung adjustment to control for uncertainty regarding between-study heterogeneity and calculate the confidence interval around the pooled mean. 8 We planned additional explorative subgroup analyses by study population group (e.g. stroke), age strata, and time lapse between MRIs, when possible.

Populations were assigned based on the study characteristics. When papers reported subgroups, e.g. cases, controls, or treatment groups, these were entered as separate groups and mentioned in the main meta-analyses unless data from the total group was used.

However, there were insufficient data to assess trial interventions. We assessed heterogeneity by visual assessment of the forest plots and by calculating the I² and τ 2 statistics to estimate the between-study variance with a restricted maximum-likelihood estimator. We used R version 4.0.2 (https://www.R-project.org/) and the 'meta' package. 9

Data used in this study are available upon reasonable request from the corresponding author.

Our search yielded 1206 publications and our manual search provided 197 further publications ( Figure 1 ). After title and abstract screening, 248 full texts were assessed for eligibility. Most papers were excluded because they only reported WMH volumes at one time point, reported no WMH volumes at all, or were superceded by a publication from the same study reporting more complete relevant data. This left 64 papers that were relevant to the review question, but 24 papers did not provide raw/unadjusted WMH volumes and instead used log-transformed, estimated volumes, or percentages of change compared to baseline volumes. Therefore, 40 papers were included in meta-analysis, comprising 39 different studies and 10,932 participants (summarised in eTables [3] [4] [5] , time between scans median . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 23, 2022. ; https://doi.org/10.1101/2022.03.22.22272696 doi: medRxiv preprint 2.6 years (range 0.25 -8.7 years). For 13/40 papers included in the meta-analysis we had to estimate the mean and SD. The 24 relevant papers that did not provide useable WMH change data are summarised in text, eResults 1-2 and eTable 6.

We identified 27 papers 4, 5, 10-34 (total 7411 participants) that reported raw WMH volume change over time. Overall time between scans was on average 2.7 years (SD=1.65; median=2 years; range 0.25 -8.1 years). We combined all data from all populations in one meta-analysis ( Figure 2 ). Overall WMH increased by mean 1.74 mL over time (95% CI 1.23, 2.26 mL), with PI of -1.24, 4.73 mL).

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Six studies reported WMH and their change over time as %ICV [35] [36] [37] [38] [39] [40] (total 1071 participants), with time between MRI mean 3.5 years (SD=1.54. Median=3; range 1.9 -6.7) were metaanalysed ( Figure 3) . WMH volume as %ICV shows an increase of 0.25 (95% CI 0.14, 0.36; PI -0.06, 0.56) %ICV. 

For explorative purposes, we examined WMH change per population type and, when possible, by age and time lapse between scans.

We performed meta-analyses per population group for all three ways of reporting WMH change (mL, %ICV, mL/year). Where studies reported treatment groups, control or case groups, these subgroups are reported. Not all population groups were available for %ICV and mL/year. 41, 42, 44, 46 In people with memory complaints 22 and dementia 10, 24 WMH increased by mean 1.17 mL (95% CI 0.40, 1.94; PI -0.87, 3.20; eFigure 4). Among mild cognitive impairment and Alzheimer's dementia groups, WMH increased by 0.27mL/year (95% CI 0.02, 0.51; PI -0.07, 0.60; eFigure 5). 42, 44 In people with depression, WMH volume increased by 1.19 mL (95% CI -1.81, 4.18; PI -6.90, 9.27; eFigure 6). 18, 27 Within the depression populations, a group with incident dementia at follow up 27 presented the largest mean WMH volume increase with 4.52mL (95% CI 2.25, 6.79), while the raw mean volume changes for the other depression groups range from 0.08-1.10 mL.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 1 

Most studies had a follow up time between scans of around two years. The 27 papers reporting raw volume in mL had a follow-up time of 2.7±1.65 years (median=2, range 0.25 -8.7). Longer follow up times between scans appeared associated with larger WMH increase in mL ( Figure 5 ). In a similar bubble plot of mean WMH change as %ICV (3.5 ± 1.54 years, median=3; range 1.9 -6.7), there is no clear relation between longer time between scans and larger WMH change ( Figure 6 ) but there were far fewer studies.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Patterns in the bubble plots of WMH change versus mean age at baseline suggest that WMH volume change increases at older ages across WMH in mL (Figure 7) , %ICV (eFigure 12) and mL/year (eFigure 13), with a younger mean age at baseline generally corresponding to smaller WMH change over time.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Thirteen of the 39 papers (eTables 3-5) included in the meta-analysis mentioned WMH volume regression, of which WMH volume regression was discussed by 8/13 papers, 4, 5, 11, 16, 24, 26, 45, 46 but only 7/8 papers 5, 10, 16, 22, 23, 25, 28 provided data. One paper 46 did not mention regression but showed WMH regression in a figure with individual trajectories.

WMH regression was found in healthy/community-dwelling participants (in ~34% 11 and 17% 26 ) and in stroke participants (4% 45 and 37% 5 ). Areas of WMH shrinkage were found in AD participants with high SVD load, with less WMH regression in cognitively normal controls who had less WMH at baseline than the patients with AD and high SVD load. 24 Over 9-years follow-up 4 (n=276), one participant (0.4%) showed net WMH regression, while 9% showed regression in the first follow-up period (2006) (2007) (2008) (2009) (2010) (2011) and 2% in the second period (2011) (2012) (2013) (2014) (2015) . Factors associated with WMH regression were not found. 48 Observations of WMH . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 23, 2022. ; https://doi.org/10.1101/2022.03.22.22272696 doi: medRxiv preprint 1 5 regression from papers not included in meta-analysis (eTable 6) can be found in eResult1, and associations with WMH change in eResult2.

Our meta-analyses suggest that although WMH volumes increase on average, WMH volumes also regress, with WMH volume regression occurring explicitly in up to 1/3 of participants. The prediction intervals of the main analyses of WMH change capture the extent of WMH volume regression (-1.24 mL; -0.06 %ICV; -0.25 mL/year) and increase (4.73 mL; 0.56 %ICV; 1.46 mL/year). We show that WMH regression can occur in all typical populations affected by SVD, greater WMH volume change might occur at older age, and WMH regression might occur over a wide range of follow-up times.

As WMH progression is the main focus of most papers and WMH regression is regarded as an accidental finding or even an 'error', the underlying mechanisms of WMH regression are unknown, or whether regression represents improvement in tissue health and translates to a positive effect on clinical outcomes. 48 However, some evidence suggests that regression is linked to less cognitive decline, recurrent stroke or dependency. 49 Furthermore, if patients with less WMH progression have less cognitive impairment than those with more WMH progression, then there is reason to think that WMH regression might translate to even better outcomes.

The papers included had several limitations. First, methodologies to measure and report WMH volume change, i.e. mL, %ICV, or mL/year, varied and complicated the comparison of volume change, and means that the findings should be interpreted with caution. Although there are many methods to assess WMH volume, including artificial intelligence approaches, there is little cross-validation or standardisation. Furthermore, methods specifically addressing WMH volume change that account for registration steps are only beginning to emerge. 50 Second, papers that report WMH volume change as an annualised rate might represent a bias by assuming that change is linear. We did not find a clear relation between longer follow-up times and larger WMH volume change but this may reflect the heterogeneity . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. of studies, populations and follow-up times. A study with a 9-year follow-up period 4 shows that WMH volumes can both increase and decrease within the 9 years and that more people had WMH regression in the first five years than in the second part. Thus, studies reporting annualized rates should also report the absolute change over time at the final time point. Our analyses were limited by not being able to examine factors related to regression since too few papers examined these. For the same reason, we were unable to assess the effect of interventions on WMH volume regression. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 23, 2022. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted March 23, 2022. ; https://doi.org/10.1101/2022.03.22.22272696 doi: medRxiv preprint

Neuroimaging standards for research into small vessel disease and its contribution to ageing and neurodegeneration

Rate of, and risk factors for, white matter hyperintensity growth: a systematic review and meta-analysis with implications for clinical trial design

The clinical importance of white matter hyperintensities on brain magnetic resonance imaging: systematic review and meta-analysis

Nonlinear temporal dynamics of cerebral small vessel disease: The RUN DMC study

White matter hyperintensity reduction and outcomes after minor stroke

Testing a tool for assessing the risk of bias for nonrandomized studies showed moderate reliability and promising validity

Optimally estimating the sample standard deviation from the five-number summary

Thrombogenic microvesicles and white matter hyperintensities in postmenopausal women

Progression of white matter hyperintensities in elderly individuals over 3 years

Longitudinal Magnetic Resonance Imaging Vascular Changes

Development of Dementia in the Neurocognitive Outcomes of Depression in the Elderly Study. The American Journal of Geriatric Psychiatry

Longitudinal decrease in blood oxygenation level dependent response in cerebral amyloid angiopathy

Serial MR imaging of volumes of hyperintense white matter lesions in elderly patients: Correlation with vascular risk factors

Progression of brain atrophy and cognitive decline in diabetes mellitus: A 3-year follow-up

Cerebral small vessel disease and association with higher incidence of depressive symptoms in a general elderly population

The AGES-reykjavik study

Is antiplatelet treatment effective at attenuating the progression of white matter hyperintensities? PLoS ONE

A prospective study of cerebral white matter abnormalities in older people with gait dysfunction

Progression of White Matter Hyperintensities, and Cognitive Decline in Stroke Patients with Confluent White Matter Hyperintensities

Progression of Cerebral Atrophy and White Matter Hyperintensities in Patients With Type 2 Diabetes

Progression of White Matter Disease and Cortical Thinning Are Not Related in Older Community-Dwelling Subjects

Mobility decline in the elderly relates to lesion accrual in the splenium of the corpus callosum

Effects of intensive versus standard ambulatory blood pressure control on cerebrovascular outcomes in older people (INFINITY)

Combining Neurovascular and Neurodegenerative Magnetic Resonance Imaging Measures in Stroke

White matter hyperintensities associated with progression of cerebral small vessel disease: a 7-year Chinese urban community study

Cerebral microbleeds are associated with the progression of ischemic vascular lesions

Longitudinal Changes in White Matter Disease and Cognition in the First Year of the Alzheimer Disease Neuroimaging Initiative

Progression of white matter lesions and hemorrhages in cerebral amyloid angiopathy

FLAIR and Diffusion MRI Signals Are Independent Predictors of White Matter Hyperintensities

Development of White Matter Hyperintensity Is Preceded by Reduced Cerebrovascular Reactivity

Longitudinal Progression of White Matter Hyperintensity

Effect of Long-Term Vascular Care on Progression of Cerebrovascular Lesions

The Effects of Longitudinal White Matter Hyperintensity Change on Cognitive Decline and Cortical Thinning over Three Years

White Matter Hyperintensity Regression: Comparison of Brain Atrophy and Cognitive Profiles with Progression and Stable Groups

CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity

Preboske GM, Schwarz CG, Senjem ML, Gunter JL, et al. Cardiometabolic Health and . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)