Dysregulated Cholesterol Metabolism, Chronic Inflammation, and Disability in Severe Progressive Multiple Sclerosis Authors: Taylor R Wicks, Mujeeb Shittu, Anna Wolska, Diala Ghazal, Bianca Weinstock‐Guttman, Alexander Burnham, Alan T Remaley, Robert Zivadinov, Murali Ramanathan Published: 2026 ABSTRACT ABSTRACTBackgroundDysregulated lipid metabolism, which can affect remyelination and neural repair, has not been extensively investigated in progressive multiple sclerosis (PMS). The goal was to examine whether dysregulated cholesterol metabolism was associated with disability in PMS.MethodsThe study sample comprised 53 Severe PMS patients (Expanded Disability Status Scale, EDSS score: 7.0–9.5) requiring skilled nursing and 53 age, sex‐ and disease duration‐matched community‐dwelling Control PMS patients (EDSS score: 3.0–6.5). Patients underwent neurological examinations and provided blood samples. Cholesterol and apolipoprotein (Apo) profiles, nuclear magnetic resonance spectroscopy metabolomic profiles, inflammatory cytokines, neuroaxonal, and glial injury biomarkers were obtained. Disability was measured on the EDSS and the Combinatorial Weight‐Adjusted Disability (CombiWISE) scales.ResultsThe Severe PMS group had lower high‐density lipoprotein cholesterol (HDL‐C), ApoA‐I, and ApoA‐II levels (all p < 0.001) and higher ApoB and ApoE levels. The nutritional status biomarker, albumin, was lower, and the chronic inflammation marker, GlycA, was higher in Severe PMS versus Control PMS (all p < 0.001). Total cholesterol, ApoA‐II, and ApoB were associated with disability on the EDSS and CombiWISE scale. Difference‐correlogram analyses revealed substantial remodeling of biomarker interdependencies in Severe PMS. ApoE showed stronger associations with lipid and nutritional biomarkers in Severe PMS, but with neuroaxonal markers (serum neurofilament light chain, glial fibrillary acidic protein) in Control PMS. Inflammatory cytokines (tumor necrosis factor, interleukin‐6) showed stronger associations with apolipoproteins in Severe PMS, whereas several lipid‐inflammation and lipid nutritional correlations were attenuated.ConclusionsDisrupted cholesterol metabolism and chronic inflammation are strongly linked to greater disability in Severe PMS. Our results also indicate marked alterations in metabolic, inflammatory, nutritional, and neuroglial biomarker networks in Severe PMS.Keywords: biomarkers, cholesterol, metabolic dysfunction, multiple sclerosis, progressive Introduction Multiple sclerosis (MS) is a disease of the central nervous system (CNS) defined by loss of blood–brain barrier (BBB) integrity, demyelination, neuroinflammation, and neurodegeneration that causes physical, cognitive, and neurological disability. Approximately 10%–15% of people with MS have a progressive subtype (PMS) of the disease, comprising secondary‐progressive (SPMS) and primary‐progressive (PPMS) forms, where disability worsens without relapses or new MRI activity. PMS can be further categorized by disability trajectory as early‐aggressive or late‐severe subtypes. Early‐aggressive PMS is marked by rapid disability progression from onset, with patients requiring walking assistance (Expanded Disability Status Scale, EDSS ≥ 6.0) within 10 years of onset. Late‐severe PMS is defined as disability accrual > 10 years from onset [1]. Early‐aggressive and late‐severe MS can be differentiated based on spinal cord involvement, while sharing global gray matter atrophy [1]. Research on patients with PMS is limited compared to relapsing forms of MS, with fewer effective treatment options for slowing disability progression. Therefore, it is important to consider other contributing factors, such as dysmetabolism. Cholesterol is a central structural component of myelin that is crucial to the functioning and maintenance of neurons and astrocytes [2, 3]. Defects in cholesterol signaling and metabolism have been linked to impaired neuronal activity, leading to neurodegeneration of dendritic spines and synapses [4, 5]. Loss of BBB integrity during MS neuroinflammation can allow lipoproteins to enter the CNS, and the entry of low‐density lipoprotein‐cholesterol (LDL‐C) and its oxidation in the inflammatory milieu promotes the production of proinflammatory cytokines such as interleukin‐6 (IL‐6), IL‐8, and tumor necrosis factor (TNF) [6]. High‐density lipoprotein‐cholesterol (HDL‐C) has been shown to exert anti‐inflammatory and antioxidant effects by reducing BBB permeability through its promotion of cholesterol efflux. Apolipoprotein A‐I (ApoA‐I) exerts an inhibitory effect on cytokine‐induced endothelial cell activation [7], while the role of ApoA‐II remains unknown. Dysregulation of lipid and cholesterol metabolism is associated with worsened outcomes in PMS. Low levels of HDL‐C and ApoA‐I were associated with greater gray matter and cortical volume loss, and conversion from RRMS to SPMS [8]. Higher serum very low‐density lipoprotein (VLDL) levels were associated with increased EDSS scores [9]. Increases in serum LDL‐C were positively correlated with contrast‐enhancing lesions, new or enlarging T2 lesions, and EDSS progression [9]. The objective of this study was to investigate the roles of lipids and cholesterol, their associations with disability, and to identify differences in crosstalk between metabolic, inflammatory, nutritional, and neuroglial biomarker networks in a cohort of early‐aggressive and late‐severe PMS. Materials and Methods Study Design Study Setting The patient groups and data in this study were collected from a prospective, cross‐sectional multi‐center study: the Comprehensive Assessment of Severely Affected Multiple Sclerosis (CASA‐MS). Recruitment for Control PMS occurred at the Jacobs Center for Multiple Sclerosis Treatment and Research at the University at Buffalo, Buffalo, NY, while patients with late‐severe PMS who were enrolled at The Boston Home, a skilled nursing facility in Dorchester, MA. We refer to the late‐severe PMS group as “Severe PMS” and the control group as “Control PMS”. Patients were excluded if they could not provide informed consent, and/or if they had a relapse or received corticosteroids within 30 days of study enrollment. The inclusion criteria were a diagnosis of MS based upon the 2017 McDonald criteria [10], 30–80 years of age, and an EDSS score of 3.0–6.5 for the Control PMS group, and 7.0–9.5 for the Severe PMS group to differentiate Severe MS. Both PMS groups were age‐, sex‐, and disease duration‐matched within ±2 years. The University at Buffalo Human Subjects Institutional Review Board approved the study, and data were collected in accordance with the Declaration of Helsinki. Patients provided written informed consent; an authorized legal healthcare proxy provided consent on the patient's behalf if the patient was unable to provide consent. Study Participants A Neurostatus‐certified researcher conducted the patient examination under the supervision of a board‐certified MS neurologist. Demographic, neurological, and clinical assessment information was collected. MS was classified according to the 2013 Lublin criteria [11]. The EDSS was utilized as a measure of disability. The Combinatorial Weight‐Adjusted Disability Score (CombiWISE), a combination of the EDSS, timed 25‐ft walk, Scripps Neurological Rating Scale (SNRS), and 9‐hole peg test, was also obtained [12]. Serum Sample Analysis Patient blood samples were collected, separated into serum and plasma within 24 h of collection, and frozen at −80°C. Samples were de‐identified for analysts. Cholesterol and Lipids Total cholesterol (TC), triglycerides, HDL‐C, LDL‐C, apolipoproteins (Apo) ApoA‐I, ApoA‐II, ApoB, and ApoE levels were measured as previously published [13] on a Pentra 400 chemical analyzer utilizing immunoturbidometric diagnostic reagent kits, quality control materials, and calibrators (Kamiya Biomedical, Thousand Oaks, CA). All concentrations are reported in mg/dL. Albumin and Prealbumin Serum albumin and prealbumin levels were measured using immunoturbidimetry (Kamiya Biomedical, Thousand Oaks, CA). Albumin concentrations were expressed in g/dL, and prealbumin concentrations in mg/dL. Neurofilament Light Chain (NfL), Glial Fibrillary Acidic Protein (GFAP), and Growth Differentiation Factor‐15 (GDF15) Serum NfL (sNfL) and GFAP concentrations were obtained by utilizing a single‐molecule array (SIMOA) by collaborators at the University of Basel. GDF15 levels were measured using an immunoassay (R & D Systems Inc., Minneapolis, MN, USA). All concentrations are expressed in pg/mL. Integrative Metabolomics Measures GlycA (a systemic inflammation marker corresponding to the number of N‐acetyl methyl groups on circulating plasma glycoproteins), lipoprotein particle size distributions, branched chain amino acids (valine, leucine, and isoleucine), and ketone body concentrations were measured via nuclear magnetic resonance (NMR) spectroscopy on a Vantera analyzer using the LP4 algorithm (LabCorp, Madison, WI). Small HDL particles (SHDLP) were defined as the sum of H3P (8.7 nm) + H2P (7.8 nm) + H1P (7.4 nm). Metabolite concentrations were expressed in μmol/L. The Inflammatory Vulnerability Index (IVX), Metabolic‐Malnutrition Index (MMX), Diabetic Risk Index (DRI), and Metabolic Vulnerability Index (MVX), which are clinically validated integrative markers for interactions among lipid, carbohydrate, and branched‐chain amino acid metabolic pathways [14, 15, 16], were computed using the equations in Methods S1. Cytokines The proinflammatory cytokines TNF IL‐6, and the anabolic peptide hormone, insulin‐like growth factor‐1 (IGF‐1), were measured using immunoassays (R & D Systems, Minneapolis, MN, USA). Data Analysis The R statistical computing platform (4.1.2) was used for statistical analyses and visualization. Severe PMS status was an indicator variable taking the value of 1 for individuals in the Severe PMS group and 0 for the Control PMS group. A binary statin‐use variable was created based on patients' use of statin medications. Cholesterol, apolipoproteins, albumin, prealbumin, sNfL, GFAP, and GDF15 levels were logarithmically transformed (base 10). All regression analyses included age, sex, body mass index (BMI), and metabolic and inflammatory biomarkers as predictors. The Severe PMS status variable was analyzed as the dependent variable in logistic regression. EDSS and CombiWISE disability measures were analyzed as dependent variables in a linear regression model that included an interaction term between the Severe PMS status variable and the biomarker of interest. Regression predictor coefficients, standard errors, and p‐values were extracted. The ggplot2 package was used for plotting. The Spearman correlation matrices for the interdependencies among metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarker variables were computed for the Control and Severe PMS groups. Differences in Spearman correlation values between groups were calculated. The difference correlogram was plotted using ggcorplot. The variables were organized using hierarchical clustering (with the default complete‐linkage agglomerative method) to facilitate the identification of patterns. Bayesian network (BN) modeling was employed to assess directed associations and conditional dependencies among the biomarkers. Sex, age, BMI, Control versus Severe PMS status, and metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarkers were included as predictors. EDSS was the target node. The derived variables, DRI, IVX, MMX, and MVX, were excluded. The BN structure was learned using the hill‐climbing (HC) algorithm in the bnlearn R package, with the Bayesian Information Criterion (BIC) used to penalize model complexity [17]. A blacklist was applied to preclude edges into age and sex predictor variables. Arc strength was defined by the frequency of appearance, and the final consensus network structure was constructed with edges that exceeded an optimal inclusion threshold of 0.5. Since this was an exploratory study with a small sample size, p‐values ≤ 0.05 were considered statistically significant. Results Clinical and Demographic Characteristics Clinical and demographic information for the Control and Severe PMS groups is displayed in Table 1. TABLE 1Clinical and demographic characteristics of control and severe progressive multiple sclerosis groups.Control PMSLate severe PMS p Sample size n5353Gender, Female (%)33 (62.2%)33 (62.2%)1.00Age, years61.8 (9.46)61.7 (10.2)0.88Body mass index, kg/m2 28.5 (6.27)26.0 (5.47) 0.048 Race0.44Caucasian48 (90.6%)51 (96.2%)African American5 (9.4%)2 (3.8%)Hispanic/Latino00Asian00Other00Missing00Disease duration, years27.8 ± 1129.8 ± 9.80.25EDSS (median, IQR)6.0 (3.5–6.0)8.0 (8.0–8.5) < 0.001 SNRS71.9 ± 1124.4 ± 13 < 0.001 CombiWISE39.0 ± 1482.3 ± 6.9 < 0.001 Disease‐modifying treatments 0.030 No treatment20 (37.7%)32 (60.4%)Interferon6 (11.3%)2 (3.77)Glatiramer acetate5 (9.43%)9 (17.0%)Anti‐CD20 therapies11 (20.8%)4 (7.55%)Other11 (20.8%)6 (11.3%)Statins1.00None36 (67.9%)35 (66.0%)Atorvastatin9 (17.0%)8 (15.1%)Pravastatin2 (3.77%)3 (5.66%)Simvastatin2 (3.77%)6 (11.3%)Rosuvastatin4 (7.55%)1 (1.89%) Note: The p‐values were calculated using the Mann–Whitney test for continuous variables and the Fisher Exact test for sex, race, and disease‐modifying treatment variables. The bold font indicates p ≤ 0.05.Abbreviations: EDSS, Expanded disability status scale; IQR, Interquartile range; PMS, Progressive multiple sclerosis; SNRS, Scripps neurological rating scale. Clinical and demographic characteristics of control and severe progressive multiple sclerosis groups. Note: The p‐values were calculated using the Mann–Whitney test for continuous variables and the Fisher Exact test for sex, race, and disease‐modifying treatment variables. The bold font indicates p ≤ 0.05. Abbreviations: EDSS, Expanded disability status scale; IQR, Interquartile range; PMS, Progressive multiple sclerosis; SNRS, Scripps neurological rating scale. The mean age, disease duration, and sex distributions were similar in the Control PMS and Severe PMS groups. The Severe PMS group had a lower average BMI (26.0 vs. 28.5, p = 0.048, Mann–Whitney test), as well as higher median EDSS (8.0 vs. 6.0), SNRS, and CombiWISE scores compared to the matched Control PMS group (all p < 0.001, Mann–Whitney test). Metabolic, Inflammatory, Nutritional, Neuroaxonal, and Glial Biomarkers Table S1 compares the means and standard deviations of the metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarkers in the Control PMS and Severe PMS groups. Figure 1 compares the distributions of the cholesterol and apolipoprotein biomarkers: TC, HDL‐C, LDL‐C, ApoA‐I, ApoA‐II, ApoB, and ApoE in the Severe PMS vs. Control PMS groups. The Severe PMS group had lower HDL‐C, ApoA‐I, and ApoA‐II levels (all p < 0.001) and higher ApoB and ApoE levels. Boxplots of the levels of total cholesterol (A), high‐density lipoprotein cholesterol (HDL, B), low‐density lipoprotein cholesterol (LDL, C), and the apolipoproteins (Apo), ApoA‐I (D), ApoA‐II (E), ApoB (F), and ApoE (G), in control progressive multiple sclerosis (Control PMS, salmon fill) and severe progressive multiple sclerosis (Severe PMS, teal fill). All concentrations are in mg/dl. The lines on the boxes correspond to the 25th, median, and 75th percentiles; the error bars correspond to the median ±1.5 interquartile range. The filled circles are outliers. Figure 2 summarizes the distributions of the nutritional status biomarkers, albumin and prealbumin, as well as GlycA and SHDLP. Albumin and SHDLP concentrations were lower, and GlycA concentrations were higher, in the Severe PMS vs. Control PMS (all p < 0.001). Boxplots of the levels of albumin (A), prealbumin (B), albumin‐prealbumin ratio (C), GlycA (D, a measure of glycosylated acute‐phase proteins), and small HDL particles (SHDLP) in Control progressive multiple sclerosis (Control PMS, salmon fill) and severe progressive multiple sclerosis (Severe PMS, teal fill). Albumin concentrations are in g/dl, prealbumin in mg/dl, GlycA, and SHDLP are in μM. The lines on the boxes correspond to the 25th, median, and 75th percentile; the error bars correspond to the median ±1.5 interquartile range. The filled circles are outliers. The integrative NMR metabolomic indices, IVX, DRI, and MVX, were higher in the Severe PMS group compared to Control PMS (all p < 0.001, Figure 3). Boxplots of the levels of the inflammatory vulnerability index (A), metabolic malnutrition vulnerability (MMX, B), diabetes risk index (DRI, C), and metabolic vulnerability index (D), in control progressive multiple sclerosis (Control PMS, salmon fill) and severe progressive multiple sclerosis (Severe PMS, teal fill). All concentrations are unitless. The lines on the box correspond to the 25th, median, and 75th percentile; the error bars correspond to the median ±1.5 interquartile range. The filled circles are outliers. TNF levels were only slightly higher in the Severe PMS group (p = 0.016); we found no evidence of differences in IL‐6 or IGF‐1 (Figure S1). sNfL and GFAP levels were similar in both PMS groups, whereas GDF15 levels were higher in Severe PMS compared to Control PMS (p = 0.016, Figure S2). Metabolic, Inflammatory, Nutritional, Neuroaxonal, and Glial Biomarkers and Disability in Severe PMS Table 2 and Table S2 summarize the associations of the metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarkers with the EDSS and CombiWISE disability measures, respectively. The regression analyses adjusted for age, sex, and BMI and included main effects and interaction terms between Severe PMS status and the biomarker of interest. TABLE 2Associations of EDSS with biomarkers.PredictorMain effectsInteractionSevere PMSBiomarkerSevere PMS:biomarkerCholesterol and apolipoproteinsTC, mg/dL−7.66 ± 4.51 (0.092) −2.91 ± 1.41 (0.041) 4.70 ± 1.96 (0.018) HDL‐C, mg/dL1.82 ± 3.43 (0.60)−0.579 ± 1.54 (0.71)0.726 ± 1.92 (0.71)LDL‐C, mg/dL−2.48 ± 2.67 (0.36)−1.60 ± 0.881 (0.072) 2.78 ± 1.32 (0.038) ApoA‐I, mg/dL−4.53 ± 5.87 (0.44)−3.44 ± 2.20 (0.12)3.57 ± 2.77 (0.20)ApoA‐II, mg/dL −9.79 ± 4.11 (0.019) −8.03 ± 2.09 (< 0.001) 8.18 ± 2.63 (0.0024) ApoB, mg/dL−5.40 ± 3.69 (0.15) −2.65 ± 1.12 (0.020) 4.40 ± 1.88 (0.021) ApoE, mg/dL 2.51 ± 0.901 (0.0064)−0.523 ± 0.992 (0.60)1.11 ± 1.52 (0.47)Nutritional status biomarkersAlbumin, g/dL3.53 ± 3.43 (0.31)0.833 ± 3.62 (0.82)−0.551 ± 5.34 (0.92)Prealbumin, mg/dL4.7 ± 3.09 (0.13)1.28 ± 1.71 (0.46)−1.10 ± 2.19 (0.62)NMR metabolomics‐derived indicesGlycA, μM1.91 ± 6.60 (0.77)0.657 ± 1.68 (0.70)0.42 ± 2.45 (0.86)SHDLP, μM−0.861 ± 1.6 (0.59) −3.42 ± 1.11 (0.0027) 3.81 ± 1.58 (0.017) IVX 11 ± 4.72 (0.021) 4.35 ± 1.62 (0.0085) −4.60 ± 2.65 (0.087)MMX6.85 ± 6.62 (0.30)2.38 ± 2.54 (0.35)−2.22 ± 3.94 (0.57)DRI2.72 ± 1.65 (0.10)−0.541 ± 0.805 (0.50)0.28 ± 0.989 (0.78)MVX 16.5 ± 6.07 (0.0078) 7.36 ± 2.17 (< 0.001) −7.83 ± 3.47 (0.026) Proinflammatory cytokinesTNF, pg/mL 4.72 ± 0.737 (< 0.001) 0.603 ± 0.201 (0.0034) −0.813 ± 0.343 (0.02) IL‐6, pg/mL 6.05 ± 0.939 (< 0.001) 1.05 ± 0.266 (< 0.001) −1.25 ± 0.391 (0.0018) IGF‐1, ng/mL 5.24 ± 1.37 (< 0.001) 0.41 ± 0.206 (0.050) −0.559 ± 0.36 (0.12)Neuroaxonal and Glial injury biomarkerssNfL, pg/mL 4.60 ± 1.14 (< 0.001)1.3 ± 0.741 (0.083)−1.16 ± 0.866 (0.18)GFAP, pg/mL3.40 ± 2.13 (0.11)0.185 ± 0.738 (0.80)−0.128 ± 1 (0.90)GDF‐15, pg/mL 6.71 ± 2.35 (0.005) 1.04 ± 0.698 (0.14)−1.46 ± 0.962 (0.13) Note: The coefficients ± standard errors (p‐values) from linear regression, adjusted for age, sex, and body mass index, are shown. All biomarkers were logarithm (base 10) transformed. Terms with p‐values ≤ 0.05 are in bold font.Abbreviations: Apo, Apolipoprotein; DRI, Diabetes risk index; GDF15, Growth differentiation factor‐15; GFAP, Glial fibrillary acidic protein; HDL‐C, High‐density lipoprotein cholesterol; IGF‐1, Insulin‐like growth factor‐1; IVX, Inflammatory vulnerability index; LDL‐C, Low‐density lipoprotein cholesterol; MMX, Metabolic malnutrition index; MVX, Metabolic vulnerability index; SHDLP, Small HDL particles; sNfL, Serum neurofilament light chain; TC, Total cholesterol; TNF, Tumor necrosis factor. Associations of EDSS with biomarkers. Note: The coefficients ± standard errors (p‐values) from linear regression, adjusted for age, sex, and body mass index, are shown. All biomarkers were logarithm (base 10) transformed. Terms with p‐values ≤ 0.05 are in bold font. Abbreviations: Apo, Apolipoprotein; DRI, Diabetes risk index; GDF15, Growth differentiation factor‐15; GFAP, Glial fibrillary acidic protein; HDL‐C, High‐density lipoprotein cholesterol; IGF‐1, Insulin‐like growth factor‐1; IVX, Inflammatory vulnerability index; LDL‐C, Low‐density lipoprotein cholesterol; MMX, Metabolic malnutrition index; MVX, Metabolic vulnerability index; SHDLP, Small HDL particles; sNfL, Serum neurofilament light chain; TC, Total cholesterol; TNF, Tumor necrosis factor. The main effect and interaction terms for TC, ApoA‐II, and ApoB were associated with both EDSS and CombiWISE scores. The LDL interaction term was associated with EDSS, whereas the main effect and interaction terms for LDL were associated with CombiWISE. The main effect and interaction terms for TC, ApoA‐II, and ApoB were associated with EDSS. The SHDLP interaction term and the IVX main effect term were associated with CombiWISE. Both the main effect and the interaction terms of MVX were associated with CombiWISE. Interestingly, EDSS and CombiWISE were associated with the main effect and interaction terms for TNF and IL‐6. Distinctive Changes in the Interdependencies of Metabolic, Inflammatory, Nutritional, Neuroaxonal, and Glial Biomarkers in Severe PMS Figure S3A,B show the correlations among the biomarkers in the PMS groups. We visualized the key differences in biomarker interdependencies between the Control PMS and Severe PMS groups using a difference correlogram (Figure 4). Variables were ordered by hierarchical clustering to facilitate pattern recognition. Correlogram of the differences in Spearman correlations among the metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarker variables between Control PMS and Severe PMS. The color scale and Spearman correlation difference (R Control PMS − R Severe PMS) values are shown. The variables are arranged in hierarchical clustering order to highlight the findings. The darker salmon color indicates that the correlation value for the variable pair is greater in the Control PMS group, whereas a darker blue indicates that the correlation value is greater in the Severe PMS group; a white color indicates that the difference in correlation values is zero. The figure represents the Bayesian network for EDSS. The rectangles represent the variables at the network's nodes, and the edges denote the conditional dependencies between variables. The thicknesses of the edges are proportional to the arc strength values, and the vertical positions of the nodes are in topological order. Alb, Albumin; Apo, Apolipoprotein; GDF15, Growth differentiation factor‐15; GFAP, Glial fibrillary acidic protein; HDL‐C, High‐density lipoprotein cholesterol; IGF‐1, Insulin‐like growth factor‐1; IVX, Inflammatory vulnerability index; LDL‐C, Low‐density lipoprotein cholesterol; MMX, Metabolic malnutrition index; MVX, Metabolic vulnerability index; Prealb, Prealbumin; SHDLP, Small HDL particles; sNfL, Serum neurofilament light chain; TC, Total cholesterol; TNF, Tumor necrosis factor. The associations of ApoE with the neuroaxonal markers sNfL and GFAP were greater (difference in Spearman correlation (∆R = 0.73 and 0.45, respectively)) in the Control group. In contrast, the associations of ApoE with HDL cholesterol biomarkers, including TC (∆R = −0.51), HDL (∆R = −0.40), and ApoAII (∆R = −0.37), as well as albumin and prealbumin (∆R = −0.40 and −0.51, respectively), were greater in the Severe PMS group. The associations of ApoE with inflammatory biomarkers, including TNF, IL‐6, and GlycA, were not substantially changed. The associations of GFAP with IVX and MVX were greater in the Control PMS compared to the Severe PMS. The associations of MVX with GlycA were also greater (∆R = −0.38) in the Severe PMS group. The association of TNF with ApoA‐II was greater in the Severe PMS group (∆R = −0.33). IL‐6 associations with HDL (∆R = −0.32), ApoA‐I (∆R = −0.32), and ApoA‐II (∆R = −0.41) were greater in Severe PMS. The ApoA‐II‐GlycA association was greater (∆R = 0.42) in the Control PMS group, but the ApoA‐II associations with IL‐6, TNF, and ApoE were greater in Severe PMS. The associations of ApoA‐I with GlycA (∆R = 0.37) and SHDLP (∆R = 0.34), and of SHDLP with GlycA (∆R = 0.44), were greater in Control PMS. These results indicate that there are consequential alterations in the interdependencies among key metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarkers. Among the metabolic biomarkers, the associations of ApoE and ApoA‐II with the inflammatory biomarkers, TNF, IL‐6, and GlycA were notable. In the difference correlogram comparison between Control PMS and Severe MS, GFAP showed more altered associations with metabolic and inflammatory biomarkers than sNfL. We used BN modeling to identify the patterns and candidate interactions among the biomarker associations with Severe MS status. Severe PMS status was the major intermediary hub in the BN (Figure 4B) and the only parent node for EDSS, indicating that EDSS variance is explained by differences between the Severe and Control PMS groups. Age was the primary driver of sNfL and GFAP. The cascade downstream of age included GDF15, which was linked to Severe PMS via albumin. This supports a key role for GDF15 in mediating the nexus between inflammation, nutritional, and metabolic status. The lipid/apolipoprotein networks contain separate LDL‐biomarker and HDL‐biomarker modules, with ApoB in the LDL module influencing ApoE. In the inflammatory cytokine module, TNF influenced both IL‐6 and IGF‐1. The covariation of TNF with IL‐6 is expected in systemic inflammation, and the effects on IGF‐1 could reflect modulation of anabolic metabolism. The Severe PMS status node was the parent node connected to chronic inflammation (GlycA), HDL, and LDL pathway biomarkers, and the inflammatory cytokine pathway via IGF‐1. Effects of Statins on Biomarkers Statin use was reported in 17 (32%) of Control PMS and 18 (34%) of Severe PMS. The patients were on one of the following medications: Atorvastatin, Pravastatin, Simvastatin, or Rosuvastatin. Figure 5 shows the associations of the metabolic, inflammatory, and neuroglial biomarkers with the statin use variable. Figures A‐L show the effects of statin use in the Control PMS (blue region) and Progressive PMS (gray region) groups on total cholesterol (mg/dl), apolipoprotein B (ApoB, mg/dl), apolipoprotein E (ApoE, mg/dl), high‐density lipoprotein cholesterol (HDL‐C, mg/dl), apolipoprotein A‐I (ApoAI, mg/dl), apolipoprotein A‐II (ApoA‐II, mg/dl), albumin (g/dl), tumor necrosis factor (TNF, ng/ml), metabolic vulnerability index (MVX, no units), serum neurofilament light chain (sNfL, pg/ml), glial fibrillary acidic protein (GFAP, pg/ml), and Expanded Disability Status Scale (EDSS). The green bars correspond to those not on statins, and the magenta bars to the group on statins. As expected, the statin‐use group had lower levels of LDL‐pathway biomarkers, including TC, LDL‐C, and ApoB, in the Severe PMS and Control PMS groups. We did not obtain evidence for statin effects on HDL‐C, ApoA‐I, and ApoA‐II (Figure 5D–F) or ApoE in either group. We did not obtain evidence for inter‐group or intra‐group differences between statin use and albumin, TNF, MVX, sNfL, and GFAP (Figure 5G–K). Statin use was not a significant modifier of EDSS in Control (p = 0.93) or Severe (p = 0.45) PMS (Figure 5L). We did not find any differences between statin use and CombiWISE in either group (results not shown). Discussion This research examined the role of dysregulated cholesterol metabolism in disability and the crosstalk among metabolic, inflammatory, nutritional, and neuroglial biomarker networks in PMS. The Severe PMS group had lower HDL‐C, ApoA‐I, ApoA‐II, and SHDLP levels, and higher ApoB and ApoE levels. Nutritional and inflammatory biomarkers differed, with lower albumin and higher GlycA and TNF in the Severe PMS cohort (p < 0.001). Integrative metabolomic indices (IVX, DRI, MVX) were greater in Severe PMS (p < 0.001). TNF levels were modestly higher (p = 0.016), whereas IL‐6 and IGF‐1 were unchanged. However, sNfL and GFAP were similar between groups. Intriguingly, GDF‐15, an emerging biomarker at the intersections of metabolism, inflammation, stress, and aging pathways, was higher in Severe PMS. Regression analyses adjusted for age, sex, and BMI showed that cholesterol and apolipoprotein biomarkers (TC, ApoAII, ApoB, and LDL), as well as for SHDLP, IVX, MVX, TNF, and IL‐6, were associated with disability (EDSS and/or CombiWISE). Difference‐correlogram analyses revealed substantial remodeling of biomarker interdependencies in Severe PMS. Notably, ApoE showed stronger associations with lipid and nutritional biomarkers in Severe PMS but stronger associations with neuroaxonal markers (sNfL, GFAP) in Control PMS. Inflammatory cytokines (TNF, IL‐6) exhibited stronger associations with apolipoproteins in Severe PMS. Overall, these findings indicate marked alterations in metabolic, inflammatory, nutritional, and neuroglial biomarker networks in Severe PMS. It is informative to consider any metabolomic findings in PMS in the context of the Phase 2 MS‐STAT (NCT00647348) and Phase 3 MS‐STAT2 (NCT03387670) trials [18, 19], which provided the initial clinical trial evidence that pharmacological targeting of cholesterol biosynthetic pathways could modify neurodegeneration in SPMS [20]. Both trials were randomized, placebo‐controlled studies of 80 mg (high dose) simvastatin, a hydroxymethylglutaryl coenzyme A (HMG‐CoA) reductase inhibitor that blocks cholesterol synthesis. Simvastatin has high lipophilicity and is hydrophobic, allowing it to cross the BBB and exhibit more direct CNS effects than other statins [21]. MS‐STAT (n = 140, 70 per arm) reported a 43% reduction in brain atrophy in the simvastatin‐treated group vs. placebo, a potent atrophy‐slowing effect for PMS at the time [18]. The atrophy effects suggested that the lipid‐lowering, vascular, and other effects of simvastatin may be neuroprotective. MS‐STAT2 (n = 964), designed to confirm if simvastatin had an impact on disability progression, did not find differences in disability progression in the simvastatin‐treated group [19]. The clinico‐radiological disconnect between marked atrophy effects and disability progression raises fundamental issues regarding the causality of atrophy and the sensitivity of disability endpoints in SPMS. In addition to the clinico‐radiological disconnect, the MS‐STAT results indicated a possible biomarker‐atrophy disconnect, as NfL in the simvastatin group was similar to placebo; higher sNfL was linked to greater atrophy progression [22]. This disconnect may be underpinned by the limitations of relying on a single, specific, and selective biomarker (sNfL) in PMS, where clinical outcomes, such as disability progression and cognition, are influenced by multiple pathways. Our multimodal approach of interrogating validated sentinel biomarkers, spanning diverse physiological pathways, in the context of Severe PMS represents a promising strategy for identifying new treatment approaches. In conclusion, our results demonstrate marked alterations in metabolic, inflammatory, nutritional, and neuroglial biomarker networks in Severe PMS, and that disrupted cholesterol metabolism and chronic inflammation are associated with greater disability in Severe PMS. Author Contributions Taylor R. Wicks, Anna Wolska: data analysis, manuscript preparation. Mujeeb Shittu: data analysis. Anna Wolska: data analysis; Diala Ghazal: data analysis. Bianca Weinstock‐Guttman, Alexander Burnham, Robert Zivadinov, and Alan T. Remaley: manuscript preparation. Murali Ramanathan: study concept and design, data analysis, manuscript preparation. Funding The authors have nothing to report. Disclosure The underlying clinical studies were funded by the Annette Funicello Research Fund for Neurological Diseases, Jacquemin Family Foundation, and private donations to the Buffalo Neuroimaging Analysis Center, Department of Neurology, Jacobs School of Medicine and Biomedical Sciences, University at Buffalo, State University of New York, Buffalo, NY. The funder had no role in the study's design or data analysis. Anna Wolska and Alan T. Remaley's research was supported by the Research Program of the National Heart, Lung, and Blood Institute (NHLBI) at the National Institutes of Health. This publication is solely the work of the authors and should not be interpreted to represent the views, perspectives, or policies of the U.S. Department of Health and Human Services, the National Institutes of Health, or the National Heart, Lung, and Blood Institute. Taylor Wicks, Mujeeb Shittu, Anna Wolska, Diala Ghazal, Alexander Burnham have nothing to disclose. Bianca Weinstock‐Guttman has participated in speaker's bureaus and/or served as a consultant for EMD Serono, Novartis, Genentech, SANA Biotechnologies, and Immunic Inc. Dr. Weinstock‐Guttman has also received grant/research support from the agencies listed. She serves on the editorial boards for Children, CNS Drugs, MS International Journal of Neurology, and Frontiers in Epidemiology. Alan T. Remaley has a CRADA research agreement with Nissui Inc. Dr. Robert Zivadinov has received speaker honoraria and consultant fees from Bristol Myers Squibb and EMD Serono. He has received research support from Protembis, Bristol Myers Squibb, Genentech, and Neurogenesis. Dr. Murali Ramanathan has received research funding from the Department of Defense and Gates Foundation. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Method S1 Methods and biomarker index calculations. This file provides detailed descriptions of the computational methods used to derive the Inflammatory Vulnerability Index (IVX), Metabolic‐Malnutrition Index (MMX), Diabetes Risk Index (DRI), and Metabolic Vulnerability Index (MVX). Full equations, variable definitions, and scaling procedures are included to support reproducibility of all metabolomics‐derived indices. Table S1: This table summarizes cholesterol‐related biomarkers, nutritional status markers, NMR‐derived metabolomic indices, proinflammatory cytokines, and neuroaxonal/glial injury biomarkers in control progressive MS and late severe progressive MS groups. Group means, standard deviations, and Mann‐Whitney p‐values are provided. Table S2: This table presents linear regression coefficients (± standard errors) evaluating associations between CombiWISE scores and individual biomarkers, adjusted for age, sex, and BMI. Main effects and interaction terms for severe PMS status are included. Figure S1: Boxplots depicting natural‐log–transformed concentrations of TNF, IL‐6, and IGF‐1 in control PMS and severe PMS groups. Figures display median, interquartile ranges, and outliers. Figure S2: Boxplots showing natural‐log–transformed levels of serum neurofilament light chain (sNfL), glial fibrillary acidic protein (GFAP), and growth differentiation factor‐15 (GDF‐15) across PMS groups, with standard boxplot elements indicated. Figure S3: Spearman correlation matrices illustrating relationships among metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarkers in control PMS and severe PMS groups. Variables are hierarchically clustered to highlight correlation patterns. Acknowledgments This project was a collaboration between the Boston Home and the University at Buffalo. We express our profound gratitude to the patients and caregivers whose invaluable contributions have been vital to our research on severe progressive MS. Use of Artificial Intelligence: Artificial intelligence‐enabled tools, such as Grammarly, ChatGPT, and Microsoft Word, were used to enhance readability and language; Google and EndNote were used to collect and curate references, add citations, and generate the bibliography. Data Availability Statement The data that supports the findings of this study are available upon reasonable request from the principal investigator of the clinical study (Dr. Robert Zivadinov). The data are not publicly available due to containing information that could compromise the privacy of research participants. References Associated Data Supplementary Materials Method S1 Methods and biomarker index calculations. This file provides detailed descriptions of the computational methods used to derive the Inflammatory Vulnerability Index (IVX), Metabolic‐Malnutrition Index (MMX), Diabetes Risk Index (DRI), and Metabolic Vulnerability Index (MVX). Full equations, variable definitions, and scaling procedures are included to support reproducibility of all metabolomics‐derived indices. Table S1: This table summarizes cholesterol‐related biomarkers, nutritional status markers, NMR‐derived metabolomic indices, proinflammatory cytokines, and neuroaxonal/glial injury biomarkers in control progressive MS and late severe progressive MS groups. Group means, standard deviations, and Mann‐Whitney p‐values are provided. Table S2: This table presents linear regression coefficients (± standard errors) evaluating associations between CombiWISE scores and individual biomarkers, adjusted for age, sex, and BMI. Main effects and interaction terms for severe PMS status are included. Figure S1: Boxplots depicting natural‐log–transformed concentrations of TNF, IL‐6, and IGF‐1 in control PMS and severe PMS groups. Figures display median, interquartile ranges, and outliers. Figure S2: Boxplots showing natural‐log–transformed levels of serum neurofilament light chain (sNfL), glial fibrillary acidic protein (GFAP), and growth differentiation factor‐15 (GDF‐15) across PMS groups, with standard boxplot elements indicated. Figure S3: Spearman correlation matrices illustrating relationships among metabolic, inflammatory, nutritional, neuroaxonal, and glial biomarkers in control PMS and severe PMS groups. Variables are hierarchically clustered to highlight correlation patterns. Data Availability Statement The data that supports the findings of this study are available upon reasonable request from the principal investigator of the clinical study (Dr. Robert Zivadinov). The data are not publicly available due to containing information that could compromise the privacy of research participants.