[1] 11w Preservation of brain metabolism in recently diagnosed Parkinson's impulse control disorders

[1] 179w Background Impulse control disorders (ICD) are a common and disrupting complication of Parkinson's disease (PD) treatment.Although their relationship with dopaminergic activity is well studied, their brain metabolic correlates are mostly unknown. Methods In this work we studied brain metabolism using brain 18 F-FDG-PET. We performed a case-control study nested within a cohort of PD patients free of ICD at baseline to compare ICD patients right after ICD diagnosis and prior to any treatment modification with matched ICD-free patients. We also compared both PD groups with healthy controls. Results When compared with ICD-free PD patients, PD patients with recently diagnosed ICD showed higher glucose metabolism in widespread areas comprising prefrontal cortices, both amygdalae and default mode network hubs (p < 0.05, corrected). When compared to healthy controls, they did not show hypermetabolism, and the only hypometabolic region was the right caudate. In turn, ICD-free patients showed diffuse hypometabolism when compared to healthy controls. Conclusion Our results suggest brain metabolism is more preserved in PD patients with ICD than patients without ICD. This metabolic preservation could be related to ICD development.

[1] 91w Impulse control disorders (ICD) are a common and frequently disabling complication of Parkinson's disease (PD). They represent a group of disorders characterized by an exaggerated increase in pleasurable activities which, even though rarely being noxious per se, can become harmful due to the prominence they acquire in the patient's life. ICD in PD are frequently related to sex, eating, shopping, gambling, and a great variety of hobbies [1]. The amplified pursue of these activities leads to personal, social, financial, and sometimes even legal consequences to the patients and their relatives [2,3].

[2] 126w ICD are more frequent in PD patients receiving treatment than in the general population, because they are related to dopaminergic replacement, particularly to a frequently used family of drugs, dopamine agonists (DA) [4]. Nevertheless, a great number of PD patients under DA treatment do not develop ICD [5,6]. This suggests that factors other than DA may explain why some patients develop this disorder and others do not. Even though longitudinal studies targeting causality of PD-ICD are scarce [7], several sociodemographic and clinical risk factors identified in cross-sectional studies are usually accepted. These include younger age [5], impulsivity [8][9][10], history of depression [11,12], personal history of alcohol or tobacco use, and family history of pathological gambling [5, This article is part of the Topical Collection on Neurology.

[1] 212w One hundred and twelve patients were included in the cohort study. During follow-up, 18 ICD cases were detected. Seven patients did not consent PET scans. Two scans could not be acquired within 2 weeks for technical reasons, and thus they were not acquired to avoid any delay in treatment modification. Finally, we obtained nine PD-ICD PET scans. We selected 18 PD-nonICD controls for PET imaging. However, one withdrew the consent, and two acquisitions were excluded because they were performed using a different scan (i.e., they were performed in a Philips Vereos). The resulting PD-nonICD group had therefore 15 patients. Nine HC with available PET scan were included. Table 1 describes the sample's clinical and sociodemographic information. No relevant differences were found comparing PD-ICD and PD-nonICD. PD-ICD patients who did not participate in the PET study had similar age (67 vs 70 years, p = 0.41), sex (female 4/9 vs 4/9, p = 1), PD evolution (5.2 vs 4.5 years, p = 0.6) LEDD (761 vs 634 mg p = 0.42), and DA-LEDD (231 vs 175 mg, p = 0.3) than those who did. Among PD-ICD there were two cases of isolated hypersexuality, three cases of isolated binge eating, three cases of isolated hobbism, and one case of hobbism combined with binge eating.

[2] 117w As seen in Fig. 1A, PD-ICD cases showed higher glucose metabolism than PD-nonICD controls. Given the spatial extent of the resulting voxelwise statistical map, only voxels showing p < 0.005 voxels are displayed and discussed. The p < 0.05 map is shown in Supplementary Fig. S1. This metabolic pattern included the posterior cingulate cortex (PCC), bilateral supramarginal gyrus, right precuneus, bilateral fusiform gyrus, bilateral lingual, parahippocampal gyrus, left anterior insula, bilateral amygdala, bilateral uncus, bilateral inferior orbitofrontal cortex (OFC), right BA10, left BA46, and left BA6. The exploratory correlation analysis only showed a significant correlation: a positive correlation between right amygdala SUVr and QUIP-RS scores within the PD-ICD group (ρ = 0.66, p = 0.05, Fig. 1b).

[3] 32w The result from intracortical PVC-SUVr analysis (Fig. 2) was similar to that of voxelwise uptake analysis. There were not significant differences between PD-ICD and PD-nonICD either in Cth or in subcortical volume.

[4] 89w As vertexwise measures are restricted to cortical structures, we also addressed differences in subcortical PVC-SUVr and volumetric information controlling for the same variables. The following subcortical structures were considered: caudate, pupallidum, nucleus accumbens, thalamus, amygdala, and hippocampus. The following regions showed higher SUVr values in the PD-ICD group with respect to PD-nonICD: left caudate (p = 0.022), right putamen (p = 0.043), left hippocampus (p = 0.025), and right hippocampus (p = 0.025). However none of these survived significance correction by permutations. No significant subcortical structural differences were found.

[5] 99w To ensure that the PET-FDG uptake differences found between PD groups were not caused by the (non-statistically significant) 30% greater agonist LEDD of PD-ICD patients (175.3 vs 135.6 mg), we explored a new set of analyses including agonist LEDD as a nuisance covariate (along with the covariates used in the main analysis). The results are shown in Supplementary Fig. S2. Due to the decreased power caused by lowering degrees of freedom, the areas surviving 0.05 FWE decrease, but the relative hypermetabolic pattern of PD-ICD patients with respect to PD-nonICD was still significant when further controlling for agonist LEDD dosage.

[6] 70w Finally, in order to investigate whether this uptake difference was driven by an abnormal increase in FDG uptake in PD-ICD patients or rather by a hypometabolism of the PD-nonICD, PD subgroups were compared to HC (Fig. 3). Given the spatial extent of the HC vs PD-nonICD statistical map, only voxels showing p < 0.005 voxels are displayed and discussed. The p < 0.05 map is shown in Supplementary Fig. S3.

[7] 80w None of the PD groups showed hypermetabolism with respect to the HC group (Fig. 3). Whereas PD-ICD hypometabolic pattern only included the right caudate, PD-nonICD hypometabolic pattern was widespread and similar to the one obtained by the comparison of PD-ICD and PD-nonICD patients. It included bilateral angular gyrus, bilateral showing p < 0.005 are displayed. (b) Correlation between FDG-SUVr values in the right amygdala cluster and demeaned QUIP-RS scores within the PD-ICD group, controlling for the same set of covariates

[1] 46w We report a significant pattern of metabolic preservation in recently diagnosed PD-ICD patients prior to any medication changes. Importantly, this is one of the few studies addressing both structural and functional correlates of recently diagnosed ICD in a matched sample of PD patients and healthy controls.

[2] 106w Our results show a diffuse pattern of relative brain hypermetabolism in PD-ICD patients compared to PD-nonICD which included critical regions of the reward system (OFC, amygdala, insula) but also key nodes of neurocognitive networks (PCC, parahippocampus) and important heteromodal hubs (supramarginal gyri). This relative hypermetabolism is better understood as metabolic preservation when PD-ICD metabolism is compared with that of HC. Although antiparkinsonian medication influences brain metabolism [27,28,30], the differences between PD-ICD and PD-nonICD remained after controlling for the main ICD causal factor, dopamine agonists [7]. Our results go in line with previous evidence of lower regional FDG-PET uptake in PD-nonICD patients with respect to PD-ICD [23].

[3] 79w As similar widespreadand sometimes topographically inconsistent-multimodal brain differences in this population have been reported [18][19][20]. This suggests that the presence of PD-ICD is associated with a global brain staterather than with well-defined localized abnormalities. Nevertheless, the exploratory correlational analysis showed an association between glucose metabolism in the amygdala and ICD severity. This finding may be related to the greater preservation of the uncinated fasciculus previously found in PD-ICD [20] and suggests a specific role of this structure in ICD.

[4] 161w Prior studies have shown lower activation (mainly showing lower increase in BOLD signal) during some tasks, specially risk taking tasks such as gambling. This has been found in PD-ICD [53] and in other addictive disorders [54]. These results apparently contradict our findings but they do not. Activation in functional imaging is determined showing the changes in BOLD signal (fMRI) or 15 O-H 2 O comparing different times of the same voxel between rest and the specific task. FDG-PET studies are almost exclusively performed during resting (as diagnostic studies) [46], and the signal is measured against Fig. 3 Voxelwise FDG-PET uptake differences between HC and PD subgroups, using age and sex as covariates of no interest (p < 0.05 FWE). No significant hypometabolic regions were found in HC. In the HC > PD-nonICD contrast, for illustration purposes, only voxels showing p < 0.005 are displayed the reference region. That is, PD-ICD shows a metabolism during resting but lower activation during gambling tasks.

[5] 132w A recent 18 study also found interesting results that might seem contradictory and merit a specific comment [24]. They found a positive correlation between dopamine transporter (DaT) availability and 18 FDG uptake in several regions among PD-ICD. Although the results did not survive multiple testing, the authors used a conservative α. They also found lower DaT availability in PD-ICD than in PD-nonICD as other studies previously reported. However 18 FDG was not available for PD-nonICD. Therefore, it is perfectly possible that PD-ICD patients have lower DaT availability and greater FDG uptake than PD-nonICD and still to exist a positive DaT-FDG correlation in both groups or that the correlation does not hold within PD-nonICD patients, as brain metabolism is not determined by DaT but also influenced by benign [26] and pathological factors [55].

[6] 73w Overall, these findings reinforce the view of PD-ICD as a multidimensional disorder where an abnormal overdrive of the meso-cortico-limbic pathway in response to DA therapy [15] would require an underlying preservation of metabolism to take action. In other words, patients with lower metabolic preservation would exhibit a lower risk of ICD under dopamine replacement therapy. This reasoning is supported by the fact that DA alters risk and reward processing in healthy controls [56,57].

[7] 101w The metabolic preservation we found might be related to previously reported PD-ICD association such as younger age [5], drug-abuse [58], or cognitive preservation [59]. Although some neuropsychological studies have found worse performance in some cognitive domains, mainly set-shifting [60] longitudinal evidence shows lower cognitive decline in PD-ICD patients [59], probably due to the higher metabolic preservation this condition associates. Noteworthy, PD-nonICD patients showed widespread hypometabolism with respect to healthy controls of similar age and sex profiles. Of particular interest is hypometabolism observed in key nodes of the default mode network (DMN) such as the PCC, angular gyri, and medial prefrontal cortex.

[8] 107w With the current knowledge of the role of resting-state networks such as the default mode network on cognitive decline in PD [61] and current evidence of neurodegeneration in early PD extending beyond frontostriatal regions [62,63], our findings suggest that the cortex of PD-ICD patients might be less affected by the disease's neuropathological mechanisms. Correspondingly, the preservation of cortical metabolism observed in PD-ICD patients might explain their lower cognitive decline [59]. Further research is needed for confirmation of this hypothesis. Nonetheless, the metabolic preservation found in PD-ICD could be a possible source of discrepancy among the so far reported metabolic signatures of PD with respect to HC [64][65][66].

[9] 113w Opposite findings have been reported for PD apathy. Apathetic patients show a more severe structural [67] compromise, an increased risk of cognitive impairment [68], and a more conservative response in gambling tasks [69]. This backs the idea that apathy and ICD are related but opposite phenomena [70]. However, although apathy is related to hypometabolism in Alzheimer's disease [71,72] and Huntington's disease [73], in the context of PD apathy has been related to both hypermetabolism and hypometabolism. We consider that the interpretation of the findings of this study in relation to the metabolism of apathetic PD patients would be too speculative until the metabolic correlates of both PD apathy and PD ICD are settled.

[10] 90w Besides the expected cognitive preservation, the main clinical implication of our study is not current but future. Based on our results, we expect brain metabolic preservation to be a risk factor for ICD in PD patients. If confirmed by longitudinal FDG-PET studies, FDG-PET could be used to tailor dopaminergic treatment to PD patients. Furthermore, this preservation might not be restricted to FDG-PET but also measurable with wet biomarkers such as CSF TAU or neurofilaments. In this case they could also be used as biomarkers of ICD risk to personalize treatment.

[11] 92w The main strengths of this study are the use of a longitudinal cohort to attain the adequate timing (before any therapeutic change) of PET scans and to avoid Neyman bias, the matching that allowed a robust clinical-imaging avoiding potential confounders, and the absence of significant structural brain differences between PD-ICD and PD-nonICD groups, which could also act as a potential confounding effect. Notwithstanding, this work has some limitations: the sample size is relatively low although the sample represents a larger longitudinal cohort, and its cross-sectional nature limits the interpretation of the results.

[12] 42w To conclude, we found that preservation of brain glucose metabolism characterize recently diagnosed ICDs in PD. We hypothesize that this metabolic preservation fosters the appearance of ICD in patients receiving dopaminergic medication. This hypothesis will require a longitudinal design to be confirmed.

[1] 30w The online version of this article (https://doi.org/10.1007/s00259-019-04664-2) contains supplementary material, which is available to authorized users. 13]. Nonetheless, current predictive models leave a considerable part of this risk unexplained [14].

[2] 205w Studies targeting neural mechanisms of ICD may also help to explain why some patients develop ICD and may lead to strategies to prevent or treat them. In this vein, neuroimage is frequently used to investigate neuropsychiatric symptoms of PD. Neuroimaging studies aiming the study of ICD in PD can be classified according to their image modality. Studies targeting the dopaminergic system and the reward circuit have found significant differences such as lower presynaptic dopamine transporter and higher dopamine release related to reward [15,16] in PD-ICD patients. Studies targeting brain structure in this context showed conflicting results [17][18][19][20]. Studies targeting restingstate connectivity have found differences between ICD and nonICD PD patients, but its results are inconsistent [20][21][22]. Finally, two studies targeted brain glucose metabolism using PET imaging. One found a relative increase in the right middle and inferior temporal gyri of PD-ICD patients compared to that of patients those without ICD [23]. The other one found a correlation between striatal presynaptic dopamine transporter and 18 F-Fluorodeoxyglucose uptake in limbic and associative areas [24]. This last study compared dopamine transporter availability in patients with and without ICD-obtaining results similar to the aforementioned, but 18 F-Fluorodeoxyglucose ( 18 FDG) uptake was not evaluated in patients free of ICD.

[3] 118w An important limitation of most neuroimaging studies is that they only considered prevalent ICD cases without specifying either the time elapsed from ICD inception or treatment changes performed after ICD diagnosis. Nonetheless, two works studied preclinical ICD using a longitudinal neuroimaging setting. These studies found interesting differences in both DAT ligands uptake [25] and resting-state connectivity [21] in patients who are about to develop ICD. Importantly, although the patterns of DAT ligands uptake are similar in patients with ICD and patients who will develop ICD, connectivity patterns found in preclinical ICD differed from those found in patients with established ICD [20][21][22]. This divergence could be explained by the fact that functional correlations in PD-ICD evolve with time [20].

[4] 125w Overall, previous neuroimaging findings suggest that timing is a critical issue in the study of the functional brain correlates of PD-ICD. 18 FDG uptake varies with different factors, the most relevant for PD-ICD being medication [26], especially dopaminergic drugs [27][28][29][30]. As dopaminergic drugs are usually modified after ICD diagnosis, timing of PET acquisition is also paramount for 18 FDG-PET studies. Furthermore, evaluation just after the phenomenon identification is crucial to avoid Neyman (prevalence-incidence) bias. This highlights the necessity of studies targeting a population of recently diagnosed ICD cases, allowing a comprehensive characterization of the imaging correlates of PD-ICD in the continuum ranging from preclinical, recently diagnosed and well-established phenomena. Such studies could also provide useful imaging biomarkers to predict ICD development or monitor therapeutic strategies.

[5] 105w In this work, we describe the brain metabolism in recently diagnosed PD-ICD patients prior to any treatment modification. We use the relatively widely available 18 F-Fluorodeoxyglucose Positron Emission Tomography ( 18 FDG-PET) imaging modality. This technique has shown several advantages with respect to functional MRI in terms of interpretability, sensitivity, magnetic-artifacts, signal-to-noise ratio, and out-of-sample replication at single-level [31]. We profit from structural MRI to study FDG uptake of intracortical and subcortical gray matter ensuring the results are not related to structural differences. We used a prospective cohort of PD patients who were free of ICD when recruited and underwent PET closely after ICD inception.

[6] 137w We conducted a nested case-control study within a cohort study of PD patients targeting risk markers for the development of ICDs. Inclusion criteria for the cohort study were PD diagnosis according to the Movement Disorders Society clinical diagnostic criteria [32], informed consent to perform the proposed evaluations, and willingness to take part in biannual follow-up visits. Exclusion criteria were the use of neuroleptics, other suspected cause of parkinsonism, dementia [33] (mild cognitive impairment was not an exclusion criterion), inability to perform the proposed evaluations, and current or previous ICD diagnosis based on the Questionnaire for Impulsive-Compulsive Disorders (QUIP current and QUIP anytime, respectively [34]). This study also comprised a healthy control (HC) cohort with the same criteria (except that excludes PD and any neurological or psychiatric disorder). Participants in the HC cohort also perform the same evaluations.

[7] 104w For this nested case-control study, cases were participants who were diagnosed with PD-ICD in follow-up visits during the first 2 years of the cohort study. Right after ICD diagnosis, we proposed them to participate in the PET study. The PET scan was performed before any treatment modification and always within a fortnight. For each included PD-ICD case, two matched PD controls (PD-nonICD) were selected. They were PD patients free of ICD, the same sex, similar age, and similar disease duration. Finally, for each PD-ICD patient, an age and sex matched HC who agreed to perform a PET scan was selected from the corresponding cohort.

[8] 18w The study was designed according to the declaration of Helsinki and was approved by the local ethics committee.

[9] 20w As part of the cohort study, we performed a comprehensive interview targeting ICD to all PD participants every 6 months.

[10] 133w This interview was based on the core criteria for behavioral addictions proposed by Brown and modified by Griffiths [9,35,36]. The interview also included the short version of the Questionnaire for Impulsive-Compulsive Disorders in PD (QUIPs) [34]. When an ICD case was detected, we also performed the rating scale version (QUIP-RS) to evaluate its severity [37]. We assessed motor status using part III of the Movement Disorders Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS) [38], levodopa equivalent daily dose (LEDD) according to previous literature [39], depressive symptoms using the depression subscale of the Hospital Anxiety and Depression Scale (HADS) [40,41], apathy using the Starkstein Apathy Scale (SAS) [42], the Barratt Impulsiveness Scale (BIS-11) [43] and cognition using the Parkinson's Disease Cognitive Rating Scale (PD-CRS) [44,45]. All the evaluations were performed in the on state.

[11] 29w We compared clinical and sociodemographic data across groups using t-test for continuous variables and X 2 for categorical variables. We considered significant differences with a p value < 0.05.

[12] 97w We assessed voxelwise and vertexwise neuroimaging differences between groups using a generalized linear model (GLM). We included age and sex as nuisance variables. When comparing PD-ICD and PD-nonICD, we also included disease duration and PD-CRS values as nuisance variables. We considered significant the clusters surviving p < 0.05 and family-wise error (FWE) correction for multiple comparison by permutation testing using 5000 permutations [52]. In the subsample of participants with available T1-MRI imaging, we also compared parceled subcortical volumetric and PVC-SUVr information across groups using the same GLM. For this analysis we used the same significance and correction.

[13] 36w Finally, within the PD-ICD group, we studied the Spearman's correlations of ICD severity (QUIP-RS) with the clusters identified in the group analysis. This analysis was considered exploratory, and a p value < 0.05 was considered significant.

[1] 58w F-FDG-PET scans were acquired for all participants on a Philips Gemini TF station 60 min after the intravenous injection of 277 MBq/ml of radiotracer and following the European Association of Nuclear Medicine procedural guidelines for PET brain imaging [46]. PET scans of PD cases were performed within 15 days of ICD diagnosis and prior to any therapeutic change.

[2] 183w We compared voxelwise brain FDG uptake across groups using the SPM12 software package (http://www.fil.ion.ucl.ac. uk/spm/) and the SnPM toolbox (https://warwick.ac.uk/ snpm). We applied a previously reported PET preprocessing pipeline [47], which included intensity normalization by a pons-vermis reference region, spatial normalization to MNI space, and smoothing with a Gaussian kernel of 12 mm fullwidth at half maximum (FWHM). All the neuroimaging acquisitions were performed in the on state to represent the most frequent state of the patients and the state in which clinical evaluations were performed. We avoided PET acquisitions in the off state because they would lead to a great variability (i.e., clinical off arises within hours of levodopa discontinuation, while the effect of some dopaminergic drugs lasts for days or weeks [48]) and because the on state better represents the most frequent state of the sample. Full clearance of dopaminergic medication could theoretically avoid this variability, but probably it would not be considered ethical and would require at least 2 weeks making the PET acquisition within a fortnight timeframe almost impossible and, more importantly, would probably modify the studied phenomenon, ICD.

[3] 181w PD patients had an available MRI scan from the cohort study acquired in a 3-Tesla Philips Achieva station within 6 months (median 2.5 months) of the PET scan. None of the patients had abnormalities in T2/FLAIR sequences. We used the T1 sequence to obtain structural information. This sequence was acquired using the following parameters: MPRAGE, Repetition time/Echo time (TR/TE) 500/50 milliseconds, flip-angle = 8°, field of view (FOV) =23 cm, in-plane resolution of 256 × 256, and 1-mm slice thickness. Structural MRI information allowed us to perform a cortical thickness (Cth) and subcortical volumetric group comparison (PD-ICD vs PD-nonICD) using the FreeSurfer 6.0 software package (https://surfer.nmr.mgh.harvard.edu/). The specific methods employed for cortical and subcortical reconstruction of structural T1-MRI images have been fully described elsewhere [49]. Briefly, optimized surface deformation models following intensity gradients are able to accurately identify white matter and gray matter boundaries in the cerebral cortex, from which Cth values are computed at each vertex. On visual inspection, no reconstruction errors were observed. Cth values in fsaverage space were subsequently smoothed by a vertexwise Gaussian kernel of 10-mm FWHM.

[4] 142w Using the information provided by the T1-MRI scans, we performed an additional surface-based intracortical FDG-PET uptake analysis. This approach provides substantial improvements in reliability and detectability of metabolic effects within the cerebral cortex [50]. Additionally, the application of partial volume correction (PVC) techniques allows the correction of PET signal spillover and also its adjustment by a possible underlying structural atrophy. For this, we used the PetSurfer pipeline (https://surfer.nmr.mgh.harvard.edu/ fswiki/PetSurfer) [50,51]. In short, PET images were registered to its associated T1-MRI scans, intensity scaled with respect to the pons region to obtain relative standardized uptake values (SUVr), partial-volume corrected using the Muller-Gartner method, sampled halfway between the white and pial surfaces, and vertexwise smoothed using a Gaussian kernel of 10-mm FWHM. The cohort study did not include MRI for HC, and therefore, analyses based on T1 could not be deployed for this group.