European Journal of Neurology 2014, 21: 1324–1329
€gren’s syndrome: a Loss of cerebral white matter in primary Sjo controlled volumetric magnetic resonance imaging study M. B. Lauvsnesa, M. K. Beyerb, S. Appenzellerc, O. J. Greved, E. Harboea, L. G. Gøranssona,e, A. B. Tjensvollf and R. Omdala,e a
Clinical Immunology Unit, Department of Internal Medicine, Stavanger University Hospital, Stavanger; bDepartment of Radiology and Nuclear Medicine, Oslo University Hospital, National Hospital, Oslo, Norway; cRheumatology Unit, Department of Medicine, Faculty of Medical Science, University of Campinas, Campinas, SP, Brazil; dDepartment of Radiology, Stavanger University Hospital, Stavanger; e
Department of Clinical Science, University of Bergen, Bergen; and fDepartment of Neurology, Stavanger University Hospital, Stavanger,
imaging, primary Sj€ ogren’s syndrome, voxel-based morphometry, white matter Received 17 February 2014 Accepted 5 May 2014
Background and purpose: Although brain involvement is common in primary Sj€ ogren’s syndrome (pSS), results from cerebral imaging studies are inconsistent. This study aimed to perform both voxel-wise and global brain volume analyses in a nearly population-based pSS cohort to explore whether the patients displayed any focal or diﬀuse volume diﬀerences compared with healthy subjects. Methods: Global grey matter (GM) and white matter (WM) volumes were measured and compared in 60 patients with pSS and 60 age- and gender-matched healthy subjects. Regression models were constructed with potential explanatory variables for GM and WM volumes. In the same groups, voxel-wise morphometric analyses were performed. Results: In analyses of global GM and WM, the patients had lower WM volumes than healthy subjects (540 63 cm3 vs. 564 56 cm3, P = 0.02), but no diﬀerences in GM. Voxel-wise analyses displayed no localized areas of GM or WM volume diﬀerences between pSS patients and healthy subjects. Conclusion: Individuals with pSS have a diﬀuse reduction of cerebral WM but no localized loss of WM or GM. This indicates a general deleterious eﬀect on WM due to pSS itself.
Introduction The central nervous system (CNS) is aﬀected in up to 60% of patients with primary Sj€ ogren’s syndrome (pSS); however, the mechanisms are incompletely understood [1–4]. There are few imaging studies that explore pSS, and the results are inconsistent. An increased load of white matter (WM) hyperintensities (WMH) in T2-weighted magnetic resonance imaging (MRI) is considered to be associated with cerebral manifestations [2,5–9], although this has been disputed [10–12]. Also regarding brain volumes, results are not consistent. In three uncontrolled studies cerebral atrophy was found in 16%–40% of the patients by visual inspection performed by neuroradiologists [3,5,8]. Brain volumes were evaluated with computer based methods in two Correspondence: M. B. Lauvsnes, Clinical Immunology Unit, Department of Internal Medicine, Stavanger University Hospital, Pb. 8100 Forus, 4068 Stavanger, Norway (tel.: +47 51 51 80 00; fax: +47 51519906; e-mail: [email protected]
other studies; Segal and colleagues found no diﬀerences regarding cortical thickness, ventricular volume or brain parenchymal portions between pSS patients and healthy control subjects by applying FreeSurfer; however, disrupted WM microstructure was revealed by diﬀusion tensor imaging . In contrast, a voxelbased morphometry (VBM) study of pSS patients claimed to reveal signiﬁcant volume loss in both grey matter (GM) and WM in pSS patients compared with healthy subjects . These discrepancies, and lack of a uniform cerebral ‘MRI pattern’, may partly be due to the use of diﬀerent classiﬁcation criteria for pSS, different imaging methods and protocols, and selection biases such as tertiary care studies and studies performed in neurological departments. The detection of imaging abnormalities attributable to pSS is an important step in the search of mechanisms for CNS involvement in pSS, and these conﬂicting results encouraged us to investigate and compare our nearly population-based cohort of pSS patients with a group of healthy subjects matched for age and © 2014 The Author(s) European Journal of Neurology © 2014 EAN
White matter in pSS
gender. Further, the potential inﬂuence of selected clinical and laboratory variables on MRI ﬁndings was explored.
Subjects and methods Subjects
Stavanger University Hospital, Norway, serves approximately 330 000 inhabitants in the southern part of Rogaland county. In general, all pSS patients in this region are taken care of by this hospital. The catchment and inclusion procedure for the participants in this study have been described in detail previously . In brief, 72 (73%) of the 99 patients fulﬁlling the American European Consensus Group Criteria (AECC) for pSS  provided informed consent to participate and were thus included for study. Patients without optimal MRI data and patients displaying cortical infarcts were excluded (two were ineligible for MRI examination because of cochlear implant or claustrophobia, eight were excluded after visual inspection of the MRI scans revealed poor image quality, and two patients were excluded because of lack of MRI images of either the patient or the corresponding healthy subject). Thus, 60 patients remained for analysis: 51 (85%) females and nine (15%) males. Selected patient characteristics are provided in Table 1. Sixty age-matched (2 years) and gender-matched healthy subjects were recruited from amongst patients’ unrelated friends and neighbors, hospital staﬀ, and the unrelated friends and neighbors of hospital staﬀ. The mean age SD of the healthy subjects was 55.4 12.6 years. All study participants were Caucasian. For research purposes only, all participants were subjected to a standardized examination by a specialist in internal medicine (EH) and a neurologist (ABT). A detailed neurological history was obtained by structured
Table 1 Selected characteristics of patients with pSS (n = 60) Age in years, mean (SD) Disease duration in years, mean (SD) No. of AECC fulﬁlled, median (range) MSG focus score ≥ 1, no. of patients (%) ANA, no. of patients (%) Anti-SSA antibodies, no. of patients (%) Anti-SSB antibodies, no. of patients (%) aPL, no. of patients (%) Present use of prednisolone, no. of patients (%)
56.0 7.2 5 48 50 45 30 7 15
(12.5) (5.0) (3–6) (80) (83) (75) (50) (12) (25)
pSS, primary Sj€ ogren’s syndrome; AECC, American European Consensus Group Criteria for pSS; MSG, minor salivary glands; ANA, anti-nuclear antibodies; aPL, anti-phospholipid antibodies.
© 2014 The Author(s) European Journal of Neurology © 2014 EAN
interviews. The Beck Depression Inventory was used for evaluation of mood. A score ≥ 13 was regarded as reﬂecting clinical depression . MRI scans were performed within a median of 12 days (range 0–33) for the patients and 20 days (range 0–81) for the healthy subjects. All participants gave informed written consent to participate. This study was approved by the regional research ethics committee and was carried out in compliance with the Helsinki Declaration. Laboratory tests
Screening for anti-SSA and anti-SSB antibodies was performed with the QUANTA Lite ENA 6, and positive results were conﬁrmed using the QUANTA Lite SSA and SSB enzyme-linked immunosorbent assays (ELISAs). Anti-phospholipid antibodies (anti-cardiolipin IgM and IgG antibodies) were screened for with QUANTA LiteTM ACA IgM and IgG ELISAs (all from Inova Diagnostics, San Diego, CA, USA). Cerebral MRI
MRI image acquisition Patients and healthy subjects were examined using a 1.5-T Philips Gyroscan NT Intera Release 10 (Philips Medical Systems, Best, The Netherlands). MRI data from patients and healthy subjects were collected noncontiguously. For the MRI scan protocol axial T2 turbo spin echo was conducted with repetition time (TR) 3240 ms, echo time (TE) 19/80, slices 5 mm, gap l.5 mm. Sagittal T2 ﬂuid-attenuated inversion recovery (FLAIR) was performed with TR 6500 ms, inversion recovery 2200 ms, TE 105 ms, slices 5 mm and gap 1 mm. Axial T1 three-dimensional turbo ﬁeld echo (TFE) had a TR 17 ms, TE 4 ms. Field of view 230 9 230 mm, matrix 256 9 256, nominal resolution 0.9 9 0.9 9 1.4 mm. Axial T1 spin echo (SE) (TR 525 ms, TE 12 ms, slices 5 mm, no gap) was performed before and after intravenous gadolinium contrast. Sagittal T1 SE was performed with TR 525 ms, TE 12 ms, slices 5 mm and no gap. For the image analyses, axial T1 three-dimensional TFE with nominal resolution 0.9 9 0.9 9 1.4 mm was used. WMH evaluation Two radiologists, blinded for clinical data, independently rated WMH according to the visual rating scale of Scheltens, as previously described in detail [12,15]. MRI data preprocessing MRI images were preprocessed using default settings in the VBM8 extension (Gaser, http://dbm.neuro.
M. B. Lauvsnes et al.
uni-jena.de/vbm/download/) of the SPM8 (http:// www.ﬁl.ion.ucl.ac.uk/spm/software/spm8/). First, images were bias-corrected, tissue classiﬁed [GM, WM and cerebrospinal ﬂuid (CSF)] and registered by using linear and non-linear transformations within a uniﬁed model . After the normalization, images were modulated by multiplying the non-linear components from the normalization to preserve GM and WM volumetric changes. The normalization included correction for individual brain size. Finally, the images were smoothed using a 12-mm full-width half-maximum (FWHM) Gaussian kernel to reduce the inter-subject variability and obtain a more normal distribution of the data . Data quality and sample homogeneity were tested by viewing one normalized, unsegmented slice for all subjects and evaluating the covariance amongst all volumes using the VBM8 tool (Gaser, http:// dbm.neuro.uni-jena.de/vbm/download/). The segmentation of GM and WM was visually examined for all participants. Whole brain volume analyses Global GM, WM and CSF volumes for each individual were estimated from segmented, unnormalized images using the VBM8 tool. In addition, total intracranial volume (TIV) was calculated as the sum of GM, WM and CSF volumes. TIV is considered to reﬂect premorbid brain size . Further, the following ratios were calculated: GM volume/TIV, WM volume/TIV and CSF volume/TIV. These ratios and global GM and WM volumes were compared between the groups using the paired samples t test. Models for GM and WM volumes were established by testing potential explanatory variables with univariable linear regression to generate unadjusted eﬀect estimates. These variables included disease characteristics and selected clinical phenomena (Tables 1 and 2). Based on results from these analyses, multivariable regression analyses were performed for brain volumes. Variables with P ≤ 0.25 in univariable analyses and variables thought to have clinical importance independent of P were added in the multivariable models. Variables without signiﬁcant eﬀect in multivariable regression were excluded from the ﬁnal models, with the exceptions of age, gender and disease duration which were regarded as important variables regardless of their signiﬁcance levels. A signiﬁcance threshold of P < 0.05 was used. Voxel-wise analyses Two samples t tests in SPM8 was applied to compare voxel-wise both GM and WM volumes between the groups. All voxels with