Hypothalamic Inflammation Improves Through Bariatric Surgery, and Hypothalamic Volume Predicts Short-Term Weight Loss Response in Adults With or Without Type 2 Diabetes Free

Preclinical studies have suggested the participation of hypothalamic inflammation (HI) in obesity and type 2 diabetes pathophysiology (1). In mice, a high-fat diet leads to an inflammatory response primarily mediated by glial cells (astrocytes and microglia) and mainly located in the arcuate nucleus (Arc) (1–3). This inflammation has been related to diminished responsiveness to peripheral anorexigenic signals and has been causally associated with obesity (1–3). While the direct participation of HI in type 2 diabetes is less well established, a significant direct contribution has also been proposed (1,4,5).

In vivo, brain inflammatory processes might be detectable through different MRI modalities. T1-weighted MRI images offer detailed macrostructural and morphological information, while T2-weighted MRI images and diffusion tensor imaging metrics from diffusion-weighted imaging (DWI), like mean diffusivity (MD), reveal magnetic properties and microstructural characteristics of brain tissue. Previous cross-sectional neuroimaging studies have consistently identified micro- and macrostructural hypothalamic radiological abnormalities in individuals with obesity or type 2 diabetes (6–12). These abnormalities included an increased signal on T2-weighted MRI, increased MD on DWI, and hypothalamic enlargement in T1-weighted MRI. Although all of them might indicate the presence of HI, none of these measures is specific for inflammation. Of note, both MD, which reflects water diffusivity in the brain tissue, and T2 signal, which reflects water content, increase not only in the presence of inflammation but also in other conditions such as atrophy (6). Similarly, brain volume is influenced by various components such as cell number or size, synaptic and vascular density, and extracellular water content. Thus, although increased gray matter volume is often viewed as a sign of neuronal integrity, it may also indicate the presence of inflammation and the existence of cytotoxic or vasogenic edema (7).

On the other hand, the physiopathological relevance of the above-mentioned radiological alterations still needs to be fully defined. Very few prospective studies have assessed the relationship among radiological biomarkers of HI, future adiposity gain, and type 2 diabetes development, and whether these structural biomarkers are related to weight loss (WL) trajectories after lifestyle interventions or bariatric surgery (BS) has not been explored (13,14,15). Additionally, although preclinical models have shown that HI and its metabolic consequences can be reversed by caloric restriction or BS, whether HI might also be reversible in humans remains unclear (16,17). Of note, the few studies investigating the impact of BS on hypothalamic microstructure showed mixed results (6,11,18). The limited sample in these previous studies (n = 10–28) and the possibility that MRI might not be sensitive enough to detect gliosis reversion might explain negative findings (6,11,17,18).

In this scenario, longitudinal multimodal studies including a larger number of participants and combining both micro- and macrostructural radiological evaluations with the analysis of novel biochemical biomarkers indicative of glial activation might improve our understanding of the involvement of HI in obesity and type 2 diabetes pathophysiology and might provide insight into hypothalamic cell dynamics in response to BS. Among these biochemical biomarkers, extracellular vesicles (EVs) might be especially interesting. EVs are small blebs (30–500 nm) surrounded by a phospholipid bilayer and released by most cell types in response to different stimuli (19). Previous studies have demonstrated that microglia shed EVs upon activation, and an increased concentration of EVs enriched for markers of glial cell origin isolated from the cerebrospinal fluid (CSF) has been described in patients with diseases associated with glial activation (20–22).

Our main aims were to compare micro- and macrohypothalamic structure between individuals with obesity (with and without type 2 diabetes) and healthy control individuals to assess the reversibility of radiological hypothalamic alterations 1 year after BS and to explore the association between these pre- and post-BS MRI-derived metrics and the extent of WL at short-term assessment after BS. As a secondary aim, we analyzed the utility of CSF-EVs enriched for markers of glial cell origin in studying HI in vivo.

We conducted a prospective longitudinal study across two centers between 2017 and 2021 that involved 72 individuals with obesity eligible for BS at Hospital Clínic de Barcelona (52 without type 2 diabetes [obesity group] and 20 with type 2 diabetes [diabetes group]) frequency matched by age and sex in a 3:1 ratio with control individuals (n = 24) sourced from the Sant Pau Initiative on Neurodegeneration (SPIN) cohort (23). Inclusion criteria were age between 18 and 70 years and BMI ≥40 or ≥35 kg/m² in the presence of obesity-related comorbidities. Exclusion criteria were a personal history of cardiovascular or chronic inflammatory diseases, neurodegenerative or unstable psychiatric disorders, common MRI and lumbar puncture contraindications, and a body weight change ≥5.0% in the 3 months before the baseline assessment. Type 2 diabetes and normal glucose tolerance were defined according to American Diabetes Association guidelines (24). Participants with type 2 diabetes using more than once daily basal insulin and presenting microvascular complications or HbA_1c >8.5% were also excluded.

The study protocol was approved by the institutional ethics committee of both study centers (reg. HCB/2018/0619; reg. IIBSP-DOW-2014-30). Written informed consent was obtained from all the study participants.

The obesity and type 2 diabetes groups were evaluated before and 1 year after BS. Forty-seven of 71 (66.2%) study participants included in the obesity and type 2 diabetes groups underwent Roux-en-Y gastric bypass (RYGB) and 24 (33.8%) sleeve gastrectomy (SG). The proportion of individuals with type 2 diabetes undergoing RYGB and SG was similar (55.0% vs. 45.0%, respectively, P = 0.212). Body weight after BS was recorded for 2 years. The control group was assessed on a single occasion. Study procedures included body composition analysis by DXA, fasting blood analysis, mixed-meal tolerance test, structural T1-weighted and DWI scans, and lumbar puncture. The flow diagram for study participants is presented in Supplementary Fig. 1. The MRI acquisition site and protocol were modified in the 2nd year of the study. Thus, DWI sequences were only available for a subgroup of participants (69.0%). Participants with or without DWI sequences were comparable in age, sex, and BMI (Supplementary Table 1). However, in the DWI subgroup, participants with obesity were slightly younger than control participants and participants with type 2 diabetes (P = 0.064).

Mixed-meal tolerance test and DXA, including estimated visceral adipose tissue (eVAT), were conducted as previously described (25). HOMA of insulin resistance (HOMA-IR) and Matsuda indexes were calculated (26,27).

Neurologists with expertise in the procedure performed the lumbar puncture. CSF was collected by free-flow/dripping following international consensus recommendations (28).

Large EVs (lEVs) were isolated from CSF as previously described, with slight modifications (25). A detailed description of these analyses is provided in the Supplementary Methods.

CSF analyses were conducted in control group participants with available CSF samples and those from the obesity and diabetes groups with available CSF samples at the baseline and post-BS assessments (Supplementary Fig. 1). Individuals included in the CSF analyses did not differ in age or BMI from those not included. Sex proportion was also maintained except for a larger representation of female participants in the control group (Supplementary Table 1).

MRI data were acquired at two different sites in Barcelona, Spain: Hospital Clínic used the Siemens 3T MAGNETOM Prisma scanner, and Hospital del Mar used the Philips 3T Achieva scanner. Baseline and longitudinal MRI scans were obtained at the identical site for each individual participant. DWI was only available at Hospital Clínic. High-resolution three-dimensional T1-weighted structural images were acquired in the sagittal plane (Hospital Clínic: repetition time [TR] = 2,300 ms, echo time [TE] = 2.98 ms, voxel size = 1.0 × 1.0 × 1.0 mm; Hospital del Mar: TR = 6,672 ms, TE = 60 ms, voxel size = 1.64 × 1.64 × 1.64 mm) for anatomical segmentation. A single-shot two-dimensional (2D) echo planar imaging (EPI) sequence was used (TR = 7,700 ms, TE = 89 ms, voxel size = 2.05 × 2.05 × 2.0 mm), applying diffusion gradients along 30 directions with a B value of 1,000 s/mm² and a baseline image without diffusion weighting. All the MRI scans were inspected by an expert neuroradiologist in each center before the analyses to check for incidental lesions and quality control.

From the structural T1-weighted images, putamen volume and intracranial volume (ICV) were estimated using automated FreeSurfer version 7.0 software (https://surfer.nmr.mgh.harvard.edu). The automated protocol introduced by Billot et al. (29) was used to segment the hypothalamus and its subunits as follows: anterior-superior (a-sHyp), anterior-inferior (a-iHyp), superior tuberal (supTub), inferior tuberal (infTub), and posterior (posHyp). The hypothalamic segmentations were visually inspected by trained neuroimaging processing technicians.

The regional volumes were normalized by ICV to assess intergroup differences. Raw hypothalamic volumes were used for paired data before and after BS. Further details concerning the correction for ICV are available at https://surfer.nmr.mgh.harvard.edu/fswiki/eTIV.

The hypothalamic location and composition are presented in Fig. 1. Hypothalamic subunit images were generated with ITK-SNAP (open-source software for segmenting anatomical structures; https://www.itksnap.org).

Processing of DWI included eddy current–induced distortion correction, denoising with MRtrix3 (an open-source, cross-platform software for medical image processing; https://www.mrtrix.org), and EPI distortion correction. Diffusion tensor imaging data were estimated using diffusion imaging in Python (Dipy) (30) to subsequently derive MD maps. Finally, the mean MD for the entire hypothalamus and its main subunits was calculated. To overcome challenges in accurately quantifying MD in small hypothalamic subunits, larger regions comprising both right and left sides (supTub, infTub, posHyp) were defined. Notably, the analysis excluded regions a-sHyp and a-iHyp because of the limited number of voxels corresponding to these areas.

Data are presented as median (25th–75th percentile), mean ± SD, or number (percentage). Among primary study variables, MD and raw hypothalamic volumes displayed a normal distribution, and head-standardized hypothalamic volumes and CSF-lEV concentration exhibited a nonnormal distribution even after log transformation. Intergroup difference assessments and correlation analyses were conducted using parametric (ANOVA and ANCOVA with Bonferroni-corrected posttests and pairwise Pearson test) or nonparametric tests (Kruskal-Wallis rank test, followed by Wilcoxon-Mann-Whitney test and Spearman rank test).

Pre- and post-BS changes in raw hypothalamic and putamen volumes and MD were assessed in the whole cohort with repeated-measures ANCOVA test. An interaction term was introduced in these models to test the effect of type 2 diabetes status and the type of BS on the study outcomes. Stratified analyses were conducted when a significant interaction was present. Pre- and post-BS CSF-lEV profiles were compared using the Wilcoxon signed rank test.

The hypothalamic volumetric changes were calculated as the percentage of initial volume minus the final volume (1-year post-BS) divided by the initial volume. The association between baseline and post-BS MRI-derived metrics and the extent of WL achieved 1 and 2 years after BS were analyzed using linear regression. Backward stepwise linear regression models were applied to evaluate the contribution of those variables associated with the percentage of total WL (TWL). A P value threshold of 0.1 was used to limit the total number of variables at each step. The association between baseline CSF-lEV concentrations and WL was explored by applying Spearman rank test.

The statistical analysis was performed with Stata/IC 15.0 (StataCorp, College Station, TX) for Windows. All P values are two-sided, and the significance level was defined as P < 0.05.

Demographic, clinical, anthropometric, biochemical, and neuroimaging data of the study groups at baseline is displayed in Table 1.

In univariate ANOVA analysis, hypothalamic MD tended to be higher in the obesity and diabetes groups than in the control group. When examining various hypothalamic subregions, the obese and diabetes groups showed higher MD in the infTub subunit than the control group, with no differences in posHyp and supTub regions or putamen (selected as reference region) (Table 1).

In correlation analyses, higher hypothalamic MD was associated with higher hs-CRP, greater BMI, higher fasting plasma glucose, and lower peripheral insulin sensitivity measured by the Matsuda index (Supplementary Fig. 2). However, in a multivariable linear regression analysis incorporating age, sex, BMI, and these metabolic variables, only age (β = 0.350, P = 0.027), female sex (β = −0.397, P = 0.012), and BMI (β = 0.550, P = 0.048) remained significantly associated with hypothalamic MD.

In ANCOVAs, after adjusting for age and sex, differences in whole hypothalamic MD among the three study groups were reinforced (P_group = 0.005). In post hoc comparisons, both the obesity and diabetes groups (P = 0.002 and P = 0.035, respectively) showed higher age- and sex-adjusted whole hypothalamic MD compared with the control group, without differences between them (P = 0.298). These differences were present across all the explored hypothalamic subunits, including posHyp (P_group = 0.014), infTub (P_group = 0.006), and supTub (P_group = 0.006). In contrast, the lack of association among study groups and putamen MD was unmodified with the inclusion of age and sex as covariates (P_group = 0.311).

As shown in Table 1, the obesity and diabetes groups showed a larger head-standardized right infTub subunit (the region comprising the Arc) than the control group, without differences among them. A similar tendency was found for the left infTub subunit, albeit statistical significance was not achieved (P = 0.074). No significant volumetric differences were observed among the three study groups for the whole left or right hypothalamus, other hypothalamic subunits, or the putamen.

In correlation analyses, infTub volumes were positively and bilaterally associated with hs-CRP, eVAT, and HbA_1c (Supplementary Fig. 3). No significant associations between left or right infTub volumes and age, BMI, or indexes of insulin sensitivity were observed.

The concentration of CSF-lEVs enriched for markers of microglial or astrocyte origins was comparable among the study groups (Table 1). There were no associations between microglia/astrocyte-derived CSF-lEVs and age, anthropometric, metabolic, inflammatory, or neuroimaging variables (Supplementary Table 2).

Post-BS changes in anthropometric, metabolic, and biochemical variables in the diabetes and obesity groups are detailed in Supplementary Table 3.

MD hypothalamic values decreased 1 year after BS (15.5 ± 0.2 × 10⁻⁴ vs. 15.2 ± 0.2 × 10⁻⁴, P = 0.035). Hypothalamic MD changes were comparable in individuals with or without type 2 diabetes (P_{time × type 2 diabetes} = 0.820) and between those who underwent RYGB or SG (P_{time × type of BS} = 0.750). The post-BS reduction in MD was mainly mediated by MD decreases in the infTub and supTub subunits (Supplementary Fig. 4). There were no changes in putaminal MD values (24.4 ± 0.1 × 10⁻³ vs. 24.5 ± 0.2 × 10⁻³, P = 0.264). Baseline between-group differences in age- and sex-adjusted hypothalamic MD were no longer observed after surgery (P_group = 0.083).

At the 1-year evaluation, there was a reduction in the left but not the right whole-hypothalamic volume (P < 0.001 and P = 0.681, respectively). When analyzing the different hypothalamic regions, significant volume size reductions were observed in the left a-iHyp (P = 0.037), bilateral a-sHyp (left: P = 0.013; right: P = 0.010), left posHyp (P = 0.004), left infTub (P = 0.015), and left supTub (P = 0.023) subunits (Fig. 2A and Supplementary Table 3). No variation in the putamen volume was observed bilaterally (left: P = 0.587; right: P = 0.184).

The extent of volumetric changes in the whole left hypothalamus did not vary according to the type of BS (P_{time × type of BS} = 0.277) but differed by pre-BS type 2 diabetes status (P_{time × type 2 diabetes} = 0.014). In stratified analyses, volumetric changes in the diabetes group were nonsignificant and numerically minor compared with the obesity group (Fig. 2B and C and Supplementary Table 3). At the post-BS assessment, participants with type 2 diabetes antedating surgery, but not those without, still presented larger head-standardized right infTub volume than control participants (P = 0.004 and P = 0.236 in post hoc analyses, respectively).

At the 1-year follow-up, the concentration of lEV in CSF was not significantly modified (median 10,145.4 [25th–75th percentile 6,491.8–12,420.6] per μL CSF pre-BS vs. 10,454.7 [7,529.1–13,269.2] per μL CSF post-BS, P = 0.079). The concentration of CSF-lEVs enriched for markers of microglial and astrocyte origins significantly increased post-BS (Supplementary Fig. 5).

Changes in CSF-lEV profile did not vary according to the pre-BS type 2 diabetes status or by type of BS (P > 0.05 for all). The post-BS CSF-lEV profile in the diabetes and obesity groups did not significantly differ from that observed in the control group (Supplementary Fig. 5).

The mean TWL at the 1 and 2 years post-BS evaluation was 31.5 ± 8.5% and 31.8 ± 9.1%, respectively. TWL was comparable between participants who had undergone RYGB and SG (1 year: P = 0.145; 2 years: P = 0.242) and the obesity and diabetes groups (1 year: P = 0.464; 2 years: P = 0.375).

Among the different baseline MRI-derived metrics, only infTub subunit size was associated with TWL in univariate linear regression analyses (Table 2). Larger infTub subunit volumes at baseline were bilaterally associated with lesser TWL 1 and 2 years after BS (model 1). This association was independent of age, sex, baseline BMI, type 2 diabetes, and type of BS (model 2 and model 3). In backward stepwise linear regression analyses including infTub volume, age, sex, type of BS, baseline BMI, and pre-BS type 2 diabetes status, only the infTub subunit size at baseline (β = −0.098, P = 0.014) remained as a significant predictor of 1-year TWL, explaining 8.2% of the TWL variability at this time point. At the 2-year follow-up, only larger infTub volume at baseline (β = −0.124, P = 0.007), undergoing SG instead of RYGB (β = 4.331, P = 0.088), and older age (β = −0.240, P = 0.065) remained significantly associated with a lesser TWL. These variables explained 18.6% of weight variability at this time point. No association was observed between the 1-year MRI-derived hypothalamic metrics and the TWL at the 2-year follow-up (data not shown).

No associations were observed between baseline or 1-year concentrations of CSF-lEVs enriched for markers of microglia or astrocyte origin and the TWL at 1 or 2 years post-BS (data not shown).

Our study reinforces the existence of radiological hypothalamic alterations that might indicate HI in individuals with obesity (with or without type 2 diabetes) and expands previous literature by providing insight into their improvement through BS. Furthermore, it supports the physiological importance of such neuroimaging findings by demonstrating an independent association with worse weight outcomes after BS.

First, in our study, as in previous ones, obesity was found to be associated with the presence of altered hypothalamic microstructure that might indicate inflammation and hypothalamic enlargement within the subregion where the Arc is located (10,13,31,32). Furthermore, our multimodal neuroimaging assessment strengthens the indication that inflammation is the potential underlying mechanism for the observed volumetric differences. Of note, we found a colocalized increase in hypothalamic volume and MD and a joint decrease in these measures at the post-BS reassessment. As previously mentioned, increased brain volume may arise from heightened cell number or size, increased synaptic or vascular density, or vasogenic edema. However, only this latter condition is linked to MD increases, as heightened cellularity or vascularization restricts water diffusion, causing decreased MD (33). Our neuroimaging findings were also consistent with earlier preclinical studies conducted in mice, showing increased vascular permeability and augmentation in hypothalamic volume attributable to vasogenic edema in response to prolonged exposure to a high-fat diet (34,35).

On the other hand, we did not observe any differences in hypothalamic volumetrics or MD between participants with and without type 2 diabetes. This is in contrast to a recent prospective study showing a progressive rise in gliosis (measured as T2 relaxation time) in the mediobasal hypothalamus across normal glucose tolerance, prediabetes, and type 2 diabetes states independent of adiposity (13). The limited number of participants with diabetes and MD data in our study and the possibility that microstructural indexes might be more sensitive than volume to detect HI might explain this discrepancy. Nonetheless, we observed a positive association between HbA_1c and infTub volume that might support an association between HI and altered glucose homeostasis.

Second, 1 year after BS, we detected a significant decrease in hypothalamic MD and a size reduction of the left hypothalamus, suggesting a partial reversal of HI. Volumetric reduction affecting other than the infTub subunit might indicate that obesity-associated HI is not restricted to Arc. In this regard, in the study by Brown et al. (31), with a larger sample size than ours (N = 1,111 adults from the Human Connectome Project), the positive association between BMI and hypothalamic volumes was more widespread. Thus, it might be possible that our study was underpowered to fully detect baseline volumetric differences but powered enough to capture its changes after BS. Similarly, although the significance of lateralized volumetric changes in our study is unclear, it deserves further investigation. Preclinical studies suggest a lateralized hypothalamic control for food intake and energy metabolism with dominance for the right side on the anorexigenic response (36).

Our data might also suggest impaired hypothalamic recovery in people with type 2 diabetes. Although the extent of MD changes was comparable, hypothalamic volumetric changes were smaller in participants with diabetes than in those without. In addition, at the post-BS assessment, individuals with type 2 diabetes still showed increased hypothalamic volume in the Arc area compared with control participants.

Only three previous studies have evaluated changes in hypothalamic microstructure after BS. In contrast to our findings, Kreutzer et al. (6) and Rebelos et al. (18) reported no significant changes at 6 and 10 months, respectively. Discrepancies with our study may arise from limited sample sizes (N = 10 in Kreutzer et al. and N = 24 in Rebelos et al.), varying neuroimaging protocols, and differences in the time elapsed between BS and the radiological reassessment. Conversely, Van de Sande-Lee et al. (13) found a significant reduction in hypothalamic relaxation time 9 months post-RYGB in a cohort of 11 participants with and 17 without diabetes. Contrary to our findings, they observed larger radiological improvement in individuals with diabetes. However, in this research, hypothalamic microstructure in the non–type 2 diabetes group was similar to that observed in the control group. Additionally, at the post-BS assessment, T2 relaxation time was still longer in participants with type 2 diabetes, which would also suggest more permanent hypothalamic damage in this group. Further studies, with larger sample sizes and longer follow-ups would be necessary to define better whether type 2 diabetes modulates the effects of BS on HI and to establish the clinical relevance of post-BS hypothalamic changes on sustaining BS benefits on body adiposity and glucose homeostasis (37).

Third, as a main addition to the literature, we demonstrate an association between structural biomarkers of HI, specifically at the Arc location, and post-BS WL trajectories. In our study, a larger infTub subunit at baseline was independently and bilaterally associated with lesser WL in the short term after BS. This finding might reinforce the pathological significance of the baseline volumetric and microstructural alterations observed in our study. It also complements data from previous studies showing an association between a greater degree of microstructural mediobasal hypothalamus alterations and increased susceptibility to future adiposity gain and the development of metabolic complications (13,14,15).

Finally, we did not observe between-group differences in the glial cell–derived CSF-lEVs or an association between its concentrations and anthropometric, metabolic, or neuroimaging variables. Furthermore, contrary to our hypothesis, we detected a significant increase in the CSF-lEV concentration post-BS. As we only evaluated lEV concentration but not cargo, our analyses cannot distinguish between glial phenotypes or localize the regions primarily involved in glial and astrocyte-derived lEV production. Thus, the significance of post-BS increases in the analyzed CSF-lEVs remains to be elucidated. In any case, our data did not support its utility in studying HI.

Our study has limitations. The lower-than-anticipated number of individuals with DWI acquisitions might have limited our statistical power to detect differences between groups in radiological measures at the baseline and post-BS evaluations. Women were overrepresented, and thus, our results might not be generalizable to men. The applied hypothalamic segmentation protocol cannot delineate the Arc in an isolated manner but locates it within the infTub subunit, which also comprises other nuclei. Also, both MD and volume lack specificity for inflammation, preventing the establishment of histopathological changes underlying the neuroimaging findings. Our study cannot disentangle whether hypothalamic radiological improvement was primarily mediated by WL, BS, or dietary changes. In this regard, a high-fat diet has been identified as the primary mechanism involved in the development of HI in mice. CSF-lEV cargo was not examined, so whether BS impacts lEV functional properties deserves further research. Finally, our study was observational, and causal relationships cannot be established.

In conclusion, people with obesity present micro- and macrostructural hypothalamic alterations that might indicate inflammation. These radiological alterations are partially reversible through BS and independently associated with a poorer WL response. These data suggest HI as a potential target for treating obesity and its comorbidities.

Acknowledgments. The authors express sincere gratitude to Sara Caelles (independent graphic designer) for assistance in the preparation of figures and the graphical abstract, Jaume Llopis (Genetics, Microbiology, and Statistics Department, Universitat de Barcelona) for guidance in the statistical analysis, and Emma Muñoz-Moreno (Neuroimaging Core, Institut d’Investigacions Biomèdiques August Pi Sunyer) for assistance with the interpretation of the MRI acquisitions. The authors thank Antoni Pané Ripoll (veterinary nutritionist dedicated to translational research, Cooperativa d’Ivars) for heartening them anytime throughout the research process. The images presented in the graphical abstract were acquired through Shutterstock (https://www.shutterstock.com) and Freepik (https://www.freepik.com) repository graphs. Parts of the graphical abstract were drawn by using pictures from Servier Medical Art (https://creativecommons.org/licenses/by/3.0). Finally, the authors are indebted to the Flow Cytometry and Cell Sorting Core Facility of the Institut d’Investigacions Biomèdiques August Pi i Sunyer.

Funding. This study was funded by Instituto de Salud Carlos III (ISCIII) and cofounded by the European Union through projects PI20/00424 and PI17/00279 to A.J. and PI19/01512 to V.C. This research was also supported by Hospital Clínic de Barcelona grant SLT008/18/00127 “Pla Estratègic de Recerca i Innovació en Salut” (PERIS) to A.J. and by the “Ajut a la Recerca Josep Font” (2018) to A.P. CIBEROBN is an initiative of ISCIII, Spain. M.R.-A. acknowledges financial support from Alzheimer's Association Research Fellowship to Promote Diversity (AARF-D) Program (AARFD-21-852492).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. A.P., L.V., J.Via., A.I., and V.M. acquired and processed all the demographic, clinical, and biochemical data. A.P., L.V., and À.C. wrote the first draft of the manuscript. A.P., J.Vid., E.O., G.C.-B., and A.J. facilitated the literature research and review. A.P., G.C.-B., J.F., and A.J. participated in the data analysis and interpretation. L.V., À.C., M.R.-A., J.P., and L.V.-A. were responsible for the neuroimaging acquisition and processing. J.Vid., E.O., G.C.-B., J.F., and A.J. revised the manuscript. I.B., V.C., and G.C.-B. were responsible for the lumbar puncture procedure and CSF-derived analysis. J.F. and A.J. designed the study. All authors helped to interpret the data, read the final version of the manuscript, and approve its submission. A.P. and A.J. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented as an oral communication at the 64th Meeting of the Spanish Society of Endocrinology and Nutrition, Barcelona, Spain, 18–20 October 2023.

Handling Editors. The journal editors responsible for overseeing the review of the manuscript were Elizabeth Selvin and Steven E. Kahn.