Prevalence and Natural History of Superficial Siderosis
A Population-Based Study
Background and Purpose—Superficial siderosis (SS) is characterized by hemosiderin deposition in the superficial layers of the central nervous system and can be seen during postmortem examination or with iron-sensitive magnetic resonance imaging techniques. The distribution of SS may predict the probable underlying cause. This study aimed to report the prevalence and natural history of SS in a population-based study.
Methods—Brain magnetic resonance imaging scans from the MCSA (Mayo Clinic Study of Aging), a population-based study of residents 50 to 89 years of age in Olmsted County, Minnesota, were reviewed. Participants with imaging consistent with SS were identified from 2011 through 2016. An inverse probability weighting approach was used to convert our observed frequencies to population prevalence of SS. Additional data abstracted included amyloid positron emission tomography, Apolipoprotein E genotype, coexisting cerebral microbleeds, and extent of SS.
Results—A total of 1412 participants had eligible magnetic resonance imaging scans. Two participants had infratentorial SS, restricted to the posterior fossa. Thirteen participants had cortical SS involving the cerebral convexities (7 focal and 6 disseminated). Only 3 of the participants with cortical SS (23%) also had cerebral microbleeds. The population prevalence of SS was 0.21% (95% confidence interval, 0–0.45) in those 50 to 69 years old and 1.43% (confidence interval, 0.53–2.34) in those over 69 years old. Apolipoprotein E ε2 allele was more common in those with SS (57.1% versus 15.0%; P<0.001). Compared with participants without SS, those with SS were also more likely to have a positive amyloid positron emission tomographic scan (76.9% versus 29.8%; P<0.001).
Conclusions—SS may be encountered in the general elderly population. The association with increased amyloid burden and Apolipoprotein E ε2 genotype supports cerebral amyloid angiopathy as the most common mechanism. Longitudinal follow-up is needed to evaluate the risk of subsequent hemorrhage in cases of incidentally discovered SS.
Superficial siderosis (SS) is characterized by hemosiderin deposition in the superficial layers of the central nervous system and can be seen during postmortem examination or with iron-sensitive magnetic resonance imaging (MRI) techniques.1,2
The distribution of SS may predict the probable underlying cause. In classical or infratentorial SS, hemosiderin deposition occurs in the posterior fossa, brain stem, and spinal cord with chronic intermittent bleeding into the subarachnoid space identified as a potential cause.3 Bleeding is often caused by dural or nerve root pathologies, prior surgeries, or tumors. These patients typically present with slowly progressive cerebellar ataxia and sensorineural hearing loss.4
In contrast, cortical SS (cSS) refers to hemosiderin deposition restricted to the cerebral convexities and is associated with cerebral amyloid angiopathy (CAA), an age-related disease of cerebral vasculature caused by β-amyloid deposition within the walls of small cortical and leptomeningeal arteries.2,5 cSS has been suggested as a useful radiological finding to establish a diagnosis of CAA.6,7
The prevalence and clinical significance of SS in the general elderly population remains unclear. The aim of this study was to report the prevalence of SS in a population-based study and to determine the clinical outcome of SS.
Materials and Methods
The MCSA (Mayo Clinic Study of Aging) is a population-based study in Olmsted County, Minnesota. Details of the MCSA have been published previously.8 In brief, Olmsted County residents 50 to 89 years of age were enumerated using Rochester Epidemiology Project resources.9 Eligible subjects were sampled from the population and invited to participate. Recruitment started in 2004, with a subset undergoing a research MRI and Pittsburgh compound B (PiB) positron emission tomographic (PET) scans. MRI was performed on one of three 3T scanners from the same vendor (General Electric, Waukesha, WI). MRI sequences sensitive to hemosiderin, a T2* gradient echo (repetition time/echo time=200/20 ms; flip angle=12°; field of view=20 cm; in-plane matrix=256×224; phase field of view=1.00; slice thickness=3.3 mm), were introduced in October 2011. The Figure demonstrates patient ascertainment.
The study protocols were approved by the Mayo Clinic and Olmsted Medical Center Institutional Review Boards. All subjects provided signed informed consent to participate in the study and in the imaging protocols.
Brain MRI scans were reviewed and documented by a board-certified neuroradiologist for abnormalities. The radiological report was searched for the terms SS or hemosiderin deposition. All MRI scans were also systemically reviewed by a trained image analyst, and areas of hemosiderin deposition including SS were identified. Findings from the radiology report and analyst review were confirmed by a vascular neurologist. Participants with SS were identified with the following inclusion criteria: (1) initial head MRI from October 2011 through June 2016 showing linear pattern of hypointensity on gradient echo imaging consistent with SS. Participants with an alternative explanation for MRI findings, such as aneurysmal subarachnoid hemorrhage, intracranial surgery, or significant head trauma, were not counted as SS. Additional data abstracted included age at the time of MRI, cognitive status, medications, PiB-PET imaging, Apolipoprotein E (APOE) genotype, number and location of coexisting cerebral microbleeds (CMBs), and extent of cSS. Cognitive status was categorized as normal, mild cognitive impairment, or dementia based on consensus diagnosis from clinician, neuropsychologist, and study coordinator. The details of the cognitive evaluation have previously been published.8 PiB-PET findings were classified as PiB positive if the standardized uptake value ratio was >1.42.10 CMBs were defined according to consensus criteria11 as homogenous hypointense lesions in the gray matter or white matter distinct from iron or calcium deposits and vessel flow voids on T2* gradient echo. Location of CMBs was classified as lobar, deep, or combination. Similar to prior publications,6 the extent of cSS was defined as focal if restricted to 3 or fewer sulci or disseminated if affecting >3 sulci. Cases of SS were further categorized as either infratentorial SS or cSS based on distribution of hemosiderin deposition. Cases involving only the posterior fossa (cerebellum and brain stem) were considered infratentorial SS, whereas those with hemosiderin deposition restricted to supratentorial locations (superficial layers of cerebral cortex) were defined as cSS.2 In participants with a diagnosis of SS, serial MRI scans were reviewed when available to evaluate for progression of SS, CMBs, or hemorrhage. However, only the initial MRI scan obtained during the study period was used when calculating prevalence. The electronic medical record was reviewed to evaluate for interval events such as symptomatic hemorrhage or stroke.
Descriptive statistics including the median and interquartile range for continuous variables and the frequency and percent for categorical variables were used to summarize participant characteristics by the presence of SS. When comparing those with SS to those without SS, Kruskal–Wallis tests were performed for continuous measures, whereas χ2 tests (or Fisher exact test where appropriate) were performed for categorical measures.
Logistic regression models were used to determine whether age at imaging, mild cognitive impairment or dementia status at imaging, presence of an APOE ε2 allele, or abnormal PiB-PET scan was associated with SS. Each covariate was considered individually with age to explore confounding as well as in multivariable models. Our final multivariate model included adjustment for age, abnormal PiB-PET, and presence of an APOE ε2 allele.
Frequencies of SS were calculated following the original sampling scheme by dividing the number of observed SS by the number of imaged participants per age/sex strata. A 2-stage inverse probability weighting approach was used to adjust observed frequencies for 2 possible causes for bias: study nonparticipation and imaging nonparticipation.12,13 We derived weights from a logistic regression model adjusting for age, sex, and education for whether an individual recruited ultimately had an in-person MCSA study visit during the time period studied. Further covariates abstracted from the medical record were also considered for inclusion into the logistic models with very little effect so the weights derived using age, sex, and education were used. Similarly, weights were derived among those who had an in-person MCSA study visit on whether or not they had a usable MRI scan for SS ascertainment adjusting for age, sex, education, and mild cognitive impairment/dementia prevalence at recruitment for imaging. These 2 sets of weights were then multiplied to give each imaged participant a single weight to adjust their observation. The frequency of SS was then standardized to the Olmsted County population (2010 Census) directly by age and sex according to our population-based sampling to give population prevalence estimates.14,15 All statistical testing presented was performed at the conventional 2-tailed α level of 0.05. All analyses were performed using SAS, version 9.4 (SAS Institute, Cary, NC).
Among a total of 1412 participants who had eligible MRI scans, 1381 (98%) had APOE data, and 1239 (88%) had PiB/PET. Eleven participants were identified through the radiology report with an additional 4 identified through the trained analysts. Two participants had infratentorial SS (0.14%), and 13 had cSS (0.92%). When stratified by age, the observed frequency of SS was 0.39% in those 50 to 69 years old, with estimated population prevalence 0.21% (95% confidence interval [CI], 0–0.45). The observed frequency was 1.89% in those over 69 years old, with estimated prevalence 1.43% (CI, 0.53–2.34). The overall estimated population prevalence in those 50 to 89 years old was 0.56% (CI, 0.25–0.86).
Demographics of participants with and without SS are shown in Table 1. The presence of an APOE ε4 allele did not differ between those with and without SS (28.6% versus 28.5%; P=0.99). Presence of an APOE ε2 allele was more common in those with SS (57.1% versus 15.0%; P<0.001). Participants with SS were also more likely to be PiB positive (76.9% versus 29.8%; P<0.001). In logistic regression models (Table 2), SS was associated with an elevated (positive) PiB standardized uptake value ratio (odds ratio, 5.68; 95% CI, 1.33–24.22; P=0.019) and the APOE ε2 allele (odds ratio, 10.67; 95% CI, 3.23–35.28; P<0.001) after adjusting for age at imaging.
APOE and PiB-PET results were reviewed further based on location of SS. Seven of 12 participants with cSS and APOE data available (58.3%) were APOE ε2 positive; neither of the participants with infratentorial SS had an APOE ε2 allele. Of those with PiB-PET results available, 9 participants with cSS (75%) and 1 participant (50%) with infratentorial SS were PiB positive. Table 3 describes the clinical, genetic, and imaging features for participants with SS. Of the 13 participants with cSS, 9 (69%) were on aspirin at the time of diagnosis and none were on anticoagulation. Seven participants had focal cSS, and 6 had disseminated involvement. None of the participants with focal cSS had associated CMBs, but 3 of the 6 (50%) with disseminated cSS had associated CMBs (range 1–90).
Three participants with cSS (1 focal, 2 disseminated) had follow-up MRI scans a mean of 31.3 months after the initial scan. There was no further hemorrhage in the participant with focal cSS. However, both of the participants with disseminated cSS and follow-up MRI experienced additional hemorrhage: one with multiple new asymptomatic CMBs and one with multiple new CMBs and symptomatic lobar intracerebral hemorrhage distant to the site of cSS.
Our findings indicate that cSS can be seen in ≈1% of people over age 70 in the general population. The subset of individuals with disseminated cSS and associated microbleeds may be at risk of developing additional hemorrhage, including lobar hematomas, but future studies will be necessary to confirm this increased risk. The association of cSS with APOE ε2 allele, which is a risk factor lobar hemorrhage, further supports an increased risk of hemorrhage among these individuals.
The overall prevalence of SS in our study was similar to another epidemiological study, the Rotterdam study, which focused only on those with cSS. That study found cSS in 0.7% of nondemented elderly individuals, all with coexisting lobar CMBs suggestive of CAA.16 Unlike the Rotterdam study, the majority (10 of 13) of our cSS cases occurred in the absence of CMBs. Our study extends the findings of the Rotterdam study by including (1) PiB-PET and APOE profiles and (2) participants with infratentorial SS. The association of cSS with increased brain β-amyloid burden and the APOE ε2 genotype in our study provides supportive evidence of CAA as a possible cause of cSS in older patients, but none of the participants with cSS had pathological confirmation.17 APOE ε2 has previously been associated with both focal and disseminated cSS in patients with pathologically proven CAA.18,19
The finding that cSS occurs in the absence of CMBs is not surprising. Among patients with CAA, the presence of cSS has been associated with a lower CMB burden.19 Increased CMB count has been associated with APOE ε4.19 The variations in clinical and imaging characteristics among patients with CAA suggest that although APOE ε2 and ε4 are both risk factors for CAA, they present with different phenotypes and may contribute to CAA-related hemorrhage through distinct mechanisms.19,20 The APOE ε2 allele is associated with a reduced risk of Alzheimer dementia but a higher risk of clinical intracerebral hemorrhage, whereas the APOE ε4 allele is a risk factor for both Alzheimer dementia and CAA.20
cSS is a risk factor for CAA-related hemorrhage, independent of CMB burden.21,22 The ability to determine which patients are at increased risk of hemorrhage may influence decisions on antithrombotic medications. CAA is an important cause of warfarin-associated lobar intracerebral hemorrhage in the elderly, and APOE ε2 status is overrepresented among these patients.23 The 2 participants with interval hemorrhage during follow-up had disseminated cSS and CMBs at the time of the initial MRI. This finding supports prior studies demonstrating extent of cSS and number of CMBs as important risk factors for future hemorrhage.21,24
The prevalence of SS is low, and associations of SS with APOE status and PiB-PET results must be interpreted cautiously in the setting of such small numbers which limits the generalizability. In addition, some participants were missing information such as APOE status, PiB-PET results, and cognitive profile. Lack of follow-up data in a majority of eligible participants with SS is another limitation and limits our ability to estimate the risk of subsequent hemorrhage. Although most participants with siderosis had multiple MRI scans, a minority had subsequent scans after diagnosis of SS because gradient echo imaging was introduced in October 2011. Although a strength of this study is that we were able to account for 2 potentially large sources of bias, study nonparticipation and imaging nonparticipation bias, through inverse probability weighting, unmeasured bias not accounted for by this technique remains a possibility.
CAA exists on a continuum with additional factors such as APOE status influencing disease progression. Asymptomatic cases of focal cSS may represent an early manifestation of CAA, and more extensive longitudinal follow-up is needed to determine whether these participants develop CMBs, disseminated cSS, or intracerebral hemorrhage. Detecting imaging changes while still asymptomatic is an important way to track the disease course and to identify risk factors for progression.
Sources of Funding
Research reported in this publication was supported by the National Institute on Aging of the National Institutes of Health (NIH) under Award Number K76AG057015 (PI: Dr Graff-Radford) and NIH grants R01 NS097495 (PI: Dr Vemuri), U01 AG06786 (PI: Dr Petersen), P50 AG16574 (PI: Dr Petersen), R01 AG034676 (PI: Dr Rocca), R01 AG11378 (PI: Dr Jack), R01 AG041851 (PIs: Drs Jack and Knopman); the Gerald and Henrietta Rauenhorst Foundation grant, the Alexander Family Alzheimer’s Disease Research Professorship of the Mayo Foundation, and the Elsie and Marvin Dekelboum Family Foundation. The funding sources were not involved in the article review or approval. MCSA (Mayo Clinic Study of Aging) data were as follows: U01 AG006786: (Dr Petersen) Mayo Clinic Study of Aging (MCSA); P50 AG016574: (Dr Petersen) Alzheimer’s Disease Research Center (ADRC); R01 AG034676: (Dr Rocca) Rochester Epidemiology Project (REP); R01 AG011378: (Dr Jack) Evaluating and Extending Our Hypothetical Model of Alzheimer’s Biomarkers; R01 AG041851: (Drs Jack/Knopman) Validating the New Criteria for Preclinical Alzheimer’s disease; and R01 NS097495: (Dr Vemuri) Development, Validation, and Application of an Imaging-based CVD Scale.
Dr Vemuri receives research support from the National Institues of Health (NIH)/National Institute of Aging (NIA). Dr Kremers receives research funding from NIH, Department of Defense, Agency for Healthcare Research and Quality, AstraZeneca, and Roche. Dr Mielke is a consultant for Eli Lilly and Lysosomal Therapeutics, Inc. She receives research grants from the NIH/NIA, Department of Defense, Biogen, Roche, and Lundbeck. Dr Knopman serves on a Data Safety Monitoring Board for Lundbeck Pharmaceuticals and for the DIAN study (Dominantly Inherited Alzheimer Network); is an investigator in clinical trials sponsored by Biogen, TauRX Pharmaceuticals, Lilly Pharmaceuticals, and the Alzheimer’s Disease Cooperative Study; and receives research support from the NIH. Dr Jack receives research support from the NIH/NIA, and the Alexander Family Alzheimer’s Disease Research Professorship of the Mayo Foundation. Dr Petersen serves on data monitoring committees for Pfizer, Inc, Janssen Alzheimer Immunotherapy; is a consultant for Biogen, Roche, Inc, Merck, Inc, and Genentech, Inc; receives publishing royalties from Mild Cognitive Impairment (Oxford University Press, 2003); and receives research support from the National Institute of Health. Dr Graff-Radford is supported by NIH/NIA, the Mayo Clinic Myron and Jane Hanley Career Development Award in Stroke Research. The other authors report no conflicts.
- Received August 2, 2017.
- Revision received August 30, 2017.
- Accepted September 21, 2017.
- © 2017 American Heart Association, Inc.
- Wilson D,
- Chatterjee F,
- Farmer SF,
- Rudge P,
- McCarron MO,
- Cowley P,
- et al
- Linn J,
- Herms J,
- Dichgans M,
- Brückmann H,
- Fesl G,
- Freilinger T,
- et al
- Feldman HH,
- Maia LF,
- Mackenzie IR,
- Forster BB,
- Martzke J,
- Woolfenden A
- Jack CR Jr,
- Wiste HJ,
- Weigand SD,
- Therneau TM,
- Lowe VJ,
- Knopman DS,
- et al
- Ahmad OB,
- Boschi-Pinto C,
- Lopez AD,
- Murray CJ,
- Lozano R,
- Inoue M
- Kahn HA,
- Sempos CT,
- Kahn HA
- Lummel N,
- Wollenweber FA,
- Demaerel P,
- Bochmann K,
- Malik R,
- Opherk C,
- et al