In this study we have evaluated for the first time with FDG-PET the brain dysfunction in j-NPSLE. Using a whole pediatric brain voxel-based methodology, FDG-PET group analysis identified a robust and widespread pattern of bilateral cortico- subcortical hypermetabolism, predominantly subcortical and mostly thalamic, but no hypometabolic area. Hypermetabolism was essentially subcortical in case of major NP symptoms, whereas it also extended to cerebral cortex in cases of minor NP symptoms. At the individual level, hypermetabolic abnormalities were detected in 95% of PET exams provided visual analysis was complemented by SPM analysis; otherwise, they were missed in 40% of cases. This particular metabolic profile could be of some diagnostic help, especially in patients with psychiatric symptoms but normal neurological examination and negative MRI. Of note, the patients reported in this cohort presented homogeneous j-NPSLE NP features, mainly psychiatric; none displayed neurological defect or seizures (nor PNS impairment), as it may sometimes more rarely be reported in j-NPSLE. Therefore, we do not know if such clinical signs would be associated to similar FDG patterns, and these observations would need to be evaluated on a larger cohort.

The metabolic abnormalities identified in the present study are particularly robust. A strict methodology was used to minimize the biases of most clinical PET studies in this field. Firstly, this is an exhaustive series, with all eligible patients included for analysis; it is a monocentric series, with a confirmed and relatively recent diagnosis, which reduces clinical and treatment heterogeneity; performing PET without sedation prevents the impact of drugs on metabolism; the absence of epileptic seizures and abnormal movements eliminates confounding factors for hypermetabolism, and the absence of neurological deficits and extensive/multiple clastic lesions confounders for hypometabolism. Secondly, the classic clinical visual analysis of the images, which is highly investigator-dependent and has limited sensitivity, was supplemented by an SPM analysis, which enables the detection of all the clusters significantly different from controls, without any a priori on pre-determined regions of interest. An age-matched pediatric database was used to avoid any bias associated with adult controls [19]. This database has proved useful in previous studies of childhood epilepsy [21,22,23]. In addition to the already relatively high uncorrected threshold p < 0.001 used in individual analysis, a more stringent corrected one was added for group analysis. Finally, the patients were explored on a PET-MR scanner, a notable benefit to optimize PET/MRI registration and thus the anatomical localization of PET anomalies.

To date, there are very few cases of j-NPSLE having experienced FDG-PET: in 3 patients, only hypermetabolism was detected, in basal ganglia using a priori ROI-localization [10], while in 3 others, analyzed visually, there was hypometabolism associated in cerebral cortex [17, 18]. Note that the latter had a purely psychiatric form of j-SLE, while the former exhibit solely neurological signs (seizures, ataxia). By contrast, many FDG-PET studies were carried out in adult NPSLE patients 10 to 20 years ago, but none using SPM: they all reported cortical hypometabolisms, frontal, parietal, temporal, and/or occipital in location, with a predominance on parieto-occipital, including in patients without any lesion on MRI [13,14,15,16]. It is likely that hypermetabolisms were missed at this time, partly because the methodology used for image analysis was not favorable to identify any. Hypermetabolisms had only been searched for in patients with choreiform movements (thus with a clinical hypothesis) and visually identified in basal ganglia [24].

Interest in hypermetabolism has really developed in the last few years, principally in the autoimmune encephalitis (AE). A meta-analysis on more than 700 adult patients shows an excellent sensitivity of PET (90% detection of abnormalities against 60% for MRI) with a strong diagnostic value [11]. Various metabolic patterns are reported according to the type of AE hypometabolism only (mesial temporal lobes, cerebellum, etc.), hypermetabolism only (basal ganglia, mesial temporal lobes, etc.), or both. Similar results are reported in pediatrics: in 34 children with AE, PET was retrospectively abnormal in 100% of cases, associating large cortical hypometabolisms to hypermetabolisms in 82%, the latter in basal ganglia in 59% of cases [10]; in 104 other children, 70% of whom with NP disorders, PET prospectively shows large cortical hypometabolisms with (57%) or without (36%) hypermetabolisms (especially in the basal ganglia 82%) [12]. In this last series PET sensitivity, specificity, positive predictive value and negative predictive value for AE reach 93%, 84%, 89% and 91% respectively. Based on these data, PET is now considered a diagnostic marker in AEs and hypermetabolism of basal ganglia suggestive of an autoimmune process. However, note that patients with associated clastic lesions have largely been included in these AE studies, increasing the incidence of hypometabolism, and none of them have used SPM, making the definite detection and localization of hypo- and hypermetabolisms open to discussion. Moreover, the potential differences in metabolic pattern according to the type of AE have not been formally studied. Interestingly, the 3 SLE children included in the aforementioned AE series disclosed exclusively hypermetabolisms (in basal ganglia), while hypometabolisms were missing, as in the present series [10].

Only one series has used SPM so far in SLE, but in adults and without NP features [25, 26]. They directly compared the whole group of SLE patients to healthy subjects. The pattern reported is close to ours at comparable thresholds: a largely predominant hypermetabolism, bilaterally in putamen/pallidum/thalamus, hippocampus, occipital and frontal cortex at uncorrected p < 0.001, surviving in hippocampi and unilaterally in putamen/pallidum/thalamus and frontal cortex at corrected p < 0.05. However, in our pediatric series of NPSLE, subcortical hypermetabolism is more intense (Z-score max at 6.6 vs. 4.7), it remains bilateral at corrected threshold, and it predominates in thalamus (instead of striatum in Mackay’s). It also significantly involves the subthalamic nuclei and cerebellum, which are not affected in Mackay’s study. These differences could reflect the gap in age between both studies (15 vs. 40 years in mean), particularly for the thalamus, whose metabolism is known to increase with brain maturation until around 25 years and then decrease [27]. However, age-matched control groups are similarly affected by this physiological phenomenon, thus eliminating any interference of aging process. Clinical symptomatology could therefore be responsible for the slight differences rather than age, NP symptoms having been excluded by Mackay, whereas our patients were more severely impaired, with NPSLE diagnosis confirmed. One might hypothesize a physiopathology continuum of NP features in j-SLE, as adult patients reported mood and cognitive disorders to a lesser extent than our patients.

The correspondence between hypermetabolism and NP features is a difficult issue. As opposed to hypometabolism, specific cortical patterns of which have been significantly linked to many conditions - dementia [28,29,30], hallucinations [31, 32], or anxiety/depression, including in children [33], hypermetabolism is much rarer. It has been found, apart from AEs, in narcolepsy [34] and Tourette’s [35]. The classical SLEDAI score in SLE was not presently optimal for correlations as it includes many non-NP components. We therefore used a NP focused scoring. A significant association between hippocampal hypermetabolism and poor spatial memory performance was identified in adult non-NPSLE patients [26]. In the present j-NPSLE study, no significant correlation was found between NP features, their type, number or severity and hypermetabolism, its localization, extension or severity. These results were expected given the limited number of patients and the heterogeneity of NP symptoms, despite a homogeneous cohort, as no patient displayed vascular or focal j-NPSLE feature. However, our group analyses suggest a metabolic spectrum, which seems to follow the intensification of NP symptoms: the cortical hypermetabolism tends to vanish behind the subcortical one and concentrate particularly in thalamus, to the detriment of the visual cortex and hippocampus. This is consistent with the fact that the thalamus plays a major role of connective hub in brain organization, even stronger than cortex does [36]. Thalamic hubs are thought to be involved in many cognitive and behavioral domains, including those involved in j-NPSLE.

Despite stringent methodology, our study retains some limitations. Small samples and symptoms heterogeneity, inherent to such a rare disease, preclude drawing definite conclusions on the course of hypermetabolism and its potential relationships with clinical features. Similarly, exclusively psychiatric patients should be compared with exclusively neurological patients, in order to gain a better understanding of their respective metabolic signatures. Exploring a control group of non-j-NPSLE patients would be the only way to determine what metabolic abnormalities are related to j-SLE pathology per se, but such a study is ethically not feasible in children.

Finally, having identified such a hypermetabolic group pattern of j-NPSLE does not mean that it can be identified in clinical practice on a given patient’s PET. Visual analysis is prone to significant interpretation bias, especially in the case of bilateral abnormalities, which prevent the perception of any asymmetry. In addition, the interpretation of an anomaly as hyper- or hypometabolism depends on the contrast scale setting. Moreover, this pattern may be variably pronounced in different individuals and associated with other anomalies. Single subject against controls SPM analysis is therefore highly recommended for the clinical interpretation of FDG-PET images in this context: in our experience, it has been beneficial in 40% of cases.