Diagnostic reliability of magnetic resonance imaging for central nervous system syndromes in systemic lupus erythematosus: a prospective cohort study

Central nervous system (CNS) lupus is a serious and potentially life-threatening manifestation of systemic lupus erythematosus (SLE), occurring in 37-95% of cases, and is associated with an increased risk of death [1]. Despite its frequency and severity, the lack of a diagnostic gold standard makes it challenging to differentiate primary CNS lupus from secondary neuropsychiatric (NP) manifestations unrelated to SLE at their onsets [1–3]. The American College of Rheumatology (ACR) has developed a standardized nomenclature system that provides case definitions for 19 NP syndromes associated with SLE, including reporting standards and recommendations for laboratory and imaging tests [2]. Although this standardized nomenclature has helped to clarify a complicated situation, its usefulness as a clinical diagnostic criterion remains to be determined.

While neurological or psychological examinations are still the cornerstones for the diagnosis of CNS lupus, neuroimaging, electroencephalography and cerebrospinal fluid tests are also used [1, 2, 4–8]. Among conventional and more recent neuroimaging tools, conventional magnetic resonance imaging (MRI) still remains the modality of choice because of its availability and accessibility. Newer tools, such as magnetic resonance spectroscopy (MRS), magnetization transfer imaging (MTI), diffusion weighted imaging (DWI), perfusion weighted imaging (PWI), single photon emission computed tomography (SPECT), and positron emission tomography (PET) are still investigational (SPECT has probably been most established among these), although they have provided data to greatly improve our understanding of CNS lupus [4–7]. However, even the diagnostic validity and usefulness of conventional MRI is not yet fully established, because most published studies are small in size (n < 100, mostly < 50), retrospective, or based on non-standardized case definitions and some include only NPSLE, not SLE patients without CNS syndromes [2, 4–7, 9]. In order to overcome these problems of the previous studies we prospectively enrolled the largest number of SLE patients with or without active CNS syndrome which was categorized based on the ACR standardized case definitions and evaluated the accuracy and usefulness of conventional MRI in the diagnosis of CNS syndrome. We also assessed the suitability of conventional MRI as a surrogate marker for CNS lupus.

First, from the largest prospective cohort of SLE patients with or without active CNS syndrome, we found that abnormal signals from conventional MRI alone did not have sufficient predictive value to be practical for diagnosis of active CNS lupus, although they were associated with active CNS lupus. Second, after assortment by size, large signals (ø ≥ 10 mm) were found to occur only in the CNS group while small signals (ø < 10 mm) occurred in both the CNS and non-CNS groups with no significant difference in frequency. Finally, we did find that large signals were always reduced or resolved completely when CNS syndromes were ameliorated by treatment but did not change when CNS syndromes did not improve. In contrast, chronological changes in small signals were not related to clinical outcome. We believe this is the first report analyzing predictive values of MRI for diagnosis of active CNS lupus using a prospective study of a large cohort of patients with and without NP syndromes.

This study demonstrated that active CNS lupus was not associated with elevated "SLEDAI scores without CNS syndrome," elevated anti-DNA antibody levels, or decreased serum CH50 levels. The latter two are widely recognized and clinically used as reliable laboratory tests for estimating disease activity in patients with SLE [20, 21]. Our data support the common view that CNS syndrome in SLE may occur independently from and in the absence of serological activity or other organ involvement [22].

Although we observed abnormal signal intensities in the brain MRIs of the CNS group more frequently than in those of the non-CNS group, the PPV (42%) and the NPV (76%) of MRI for detecting active CNS lupus indicate that conventional MRI alone is insufficiently specific to serve as a diagnostic tool for active CNS lupus. Active CNS lupus was diagnosed in less than half of the patients with abnormal signal intensities in brain MRI, even though normal MRI findings tended to be associated with non-CNS lupus. We also categorized patients by CNS manifestations indicating neurologic or psychiatric disorders but did not find any better results. In fact, MRI has been shown to be neither very sensitive nor specific for the diagnosis of NPSLE, estimates of sensitivity and specificity being in the 30 and 40%, respectively [12]. We would like to point out that these figures had been only estimates without solid evidence until this study. Thus, the diagnosis of CNS syndrome in SLE remains a difficult task requiring careful clinical and laboratory assessment together with such evaluation techniques as neuroimaging, electroencephalography and cerebrospinal fluid tests [7, 12].

While exploring the usefulness of conventional MRI, we found that a large MRI signal is useful as a diagnostic and surrogate marker for CNS lupus. Large abnormal MRI signals (ø ≥ 10 mm) were seen only in the CNS group whereas small abnormal MRI signals (ø < 10 mm) were seen in both groups with no significant difference. In addition, chronological changes in large signals always corresponded with clinical outcome while those of small signals did not. This kind of reversible large abnormal MRI signals has been reported [4, 12, 15, 23, 24]. These large abnormal MRI signals may include new infarcts, discrete gray matter lesions, diffuse gray matter hyperintensities, and cerebral edema. These lesions, except for infarcts, are often transitory and resolve completely with time and treatment. The large abnormal MRI signals in the present study included several types of features, such as infarcts, diffuse gray matter hyperintensities, and widespread grey and white matter T2 high signal changes resembling those in acute disseminated encephalomyelitis, accompanied by various CNS syndromes. Although these large abnormal MRI signals were attributed to different pathomechanisms, they exhibited equal usefulness as a diagnostic and surrogate marker. Therefore, our simple criterion (i.e., MRI signals with ø ≥ 10 mm) is applicable to daily clinical practice without specialized knowledge of neuroimaging and regardless of the type of CNS syndromes encountered. We believe this test will prove very beneficial for clinical assessment of the condition.

In agreement with our present findings, small punctate focal lesions in the white matter have been reported to be the most common brain MRI finding, tending to be more frequently observed in the CNS than in the non-CNS SLE patients in most published reports [6, 7, 12, 25, 26]. These lesions appear anywhere in the brain, including the brain stem, and are accompanied by a variety of clinical presentation. These small punctuate hyperintense T2-weighted focal lesions are frequently nonspecific and interpreted as consistent with focal ischemia, demyelination, vasculitis, microinfarcts, gliosis, or other conditions. These small abnormal MRI signals frequently correlate poorly with clinical manifestations, and occur in patients without marked CNS signs and symptoms. In addition, conventional MRI is not capable of differentiating between older chronic and recent acute lesions of this type.

Neuroimaging must be evaluated in the context of the brain pathology of CNS lupus [4]. At this time, pathologic studies have basically been limited to post mortem studies of end stage SLE patients, which has seriously limited understanding of the sequence of events in the pathogenesis and progression of the disease [5]. Moreover, there have been few paired imaging-autopsy studies [4]. Despite the dramatic neurologic manifestations, minimal histopathologic findings in CNS lupus are typical, with apparently normal findings or nonspecific changes in the brain predominating. This general lack of specific histopathology has complicated interpretation of results found with nearly any imaging technique. For example, the autopsy of one of the patients in the present study who died from alveolar hemorrhage 5 years after showing acute confusional state and several foci of small-sized abnormal MRI signals in the white-matter showed no remarkable findings in the brain.

One of the limitations of the present study is that its subject number was insufficient for definitive conclusions about significance of conventional MRI for diagnosis of CNS lupus, although it is appreciably larger than previous studies dealing with NPSLE and MRI [15, 24, 27]. Another potential weakness arises from the fact that the MRI scans were not evaluated by multiple radiologists blinded to the clinical condition. Fortunately, it has been shown that evaluation of these scans is generally consistent among observers, so cranial MRI may be considered to be a valid technique to assess disease progression or reversal [15]. Because we did not include SLE patients who had CNS syndromes caused by non-SLE pathologies, the value of MRI for diagnosis of these conditions was not determined by this study. Thus, it is not clear from this study whether CNS syndromes caused by non-SLE pathologies can be differentiated by MRI, including the large MRI abnormal signals.

MRI alone does not have sufficient predictive value for the diagnosis of active CNS lupus, although abnormal MRI signals showed a significant association with CNS syndrome. Importantly, a large MRI signal is a reliable and practical diagnostic and surrogate marker for CNS syndrome in SLE. The relatively small number of subjects having large MRI signals in the study indicates the need for further studies in larger numbers of patients to validate the findings.