Four Groups of Thirty: NMR Metabolomics Across the MS Subtypes
Most metabolomics work in multiple sclerosis (MS) is built on relapsing patients, because they are the easiest group to recruit and make up roughly 85% of diagnoses. Alwahsh and colleagues built their cohort the other way round: 30 patients with relapsing-remitting disease, 30 with primary progressive, 30 with secondary progressive, and 30 healthy controls. Progressive patients account for two thirds of the sample, which is what makes a three-way subtype comparison possible at all. Recruitment ran through neurology clinics in Amman, Jordan, with diagnosis by the revised McDonald criteria and severity scored on the Expanded Disability Status Scale. Groups were matched on age, body mass index, and sex, both against controls and across the three subtypes. Serum went through cryogenic proton nuclear magnetic resonance at 600 MHz with ERETIC referencing against a sucrose standard, which returns absolute concentrations rather than relative peak intensities and needs no calibration curve for each compound. Forty metabolites were quantified across 120 samples.
Twenty-Five Metabolites Separate Patients From Controls
The largest effects are losses. Lysine fell to 0.48 of control levels (p = 3.11E-19), myo-inositol to 0.36 (p = 6.13E-18), glutamate to 0.57 (p = 1.56E-16), threonine to 0.38 (p = 7.64E-16), and glycine to 0.45 (p = 9.51E-15). Tyrosine, choline, O-phosphocholine, serine, phenylalanine, cysteine, and creatine followed the same direction. Nine metabolites rose: formate at 2.47-fold, ATP at 1.98, NAD+ at 1.90, histidine at 1.90, succinate at 1.58, inosine at 1.61, glutathione at 1.47, pantothenate at 1.37, and tryptophan at 3.34, the largest fold change anywhere in the study. Both multivariate models cleared the usual threshold for a valid separation, with predictive ability of 0.526 for the orthogonal partial least squares model and 0.635 for the partial least squares model. Lysine, myo-inositol, and glutamate produced areas under the curve of 0.93 (95% CI 0.869–0.981), 0.92 (0.859–0.969), and 0.91 (0.843–0.968).
Those Curves Come With a Condition the Paper Does Not State
The three best-discriminating metabolites were picked from the same 120 samples used to compute their areas under the curve. With no held-out test set and no independent cohort, values above 0.9 are optimistic by construction. The honest reading is that these are the molecules that separate this dataset best, which is not the same thing as validated diagnostic performance, and the authors point that way themselves when they write that the metabolites need assessing in a larger cohort. A second gap sits in the statistics. The methods set significance at p < 0.05 from t-tests and one-way ANOVA, and no correction for testing 40 metabolites is described anywhere in the paper. Most headline results are untouched by this, since p-values around E-15 survive any correction. But a Bonferroni threshold of 0.00125 would drop five of the 25: sn-glycero-3-phosphocholine, histidine, glutathione, pantothenate, and fumarate. It would cut deeper into the subtype analysis, where many p-values fall between 0.01 and 0.05.
The Subtype Comparison Is the Part Worth Having
Twenty-one metabolites differed across the three subtypes by ANOVA, and this is where the cohort design pays off. Two moved consistently across all three groups: tryptophan and succinate. Relapsing patients carried higher tryptophan than both progressive groups, and within the progressive subtypes, tryptophan and pantothenate were higher in secondary than in primary progressive disease. Comparing relapsing with primary progressive gave eight metabolites past the fold-change cutoff, with tryptophan and ascorbate higher in relapsing patients and histidine, proline, phenylalanine, cysteine, formate, and fumarate lower. Comparing relapsing with secondary progressive gave only two, inosine and NAD+, both higher in relapsing disease. Glutamate and lactate were higher in primary than in secondary progressive patients. The asymmetry is itself informative: relapsing disease looks metabolically much closer to secondary progressive than to primary progressive, which is what you would expect if the two progressive phenotypes have different origins.
What the Title Promises and What the Design Can Deliver
The paper is titled as identifying biomarkers of disease progression, and the design cannot supply that. Every sample was drawn once. No patient was followed over time, and no patient was observed moving from one subtype into another. What the study compares is three groups of different people who currently hold different diagnoses. Differences between them may reflect progression, or the biology that made them different subtypes from the outset, or the treatment histories and disease durations that accumulated differently in each group, and a single time point separates none of these. The conclusion is more careful than the title, describing inosine and O-phosphoethanolamine as showing a trend with disease progression that will need validating in a larger cohort using targeted mass spectrometry. Reading the study as a map of how the three subtypes differ right now, rather than as a set of progression markers, keeps the claim inside what the data support.
The Pathways, and One Number the Discussion Leaves Alone
Pathway analysis put inositol phosphate metabolism and the phosphatidylinositol signalling system at the top by significance, with phenylalanine, tyrosine, and tryptophan biosynthesis carrying the highest pathway impact. Glycine, serine, and threonine metabolism, histidine metabolism, glutamine and glutamate metabolism, and glutathione metabolism followed. The biological readings offered are reasonable and sourced. Myo-inositol is a constituent of myelin, so lower serum levels may track disruption of inositol phosphate metabolism caused by demyelination. Glycine is the main inhibitory neurotransmitter in the brain and has anti-inflammatory effects on immune cells, so its reduction may mark an imbalance between pro- and anti-inflammatory mediators. Succinate is a tricarboxylic acid cycle intermediate that rises in inflammatory conditions and may reflect the higher energy demand of demyelinated axons. Pantothenate feeds coenzyme A biosynthesis and mitochondrial reactions. Tryptophan gets less attention than its numbers deserve: at 3.34-fold it is the biggest single change in the paper and it also separates all three subtypes, yet the discussion mentions tryptophan metabolism as an altered pathway without addressing the direction or size of the shift.
Reporting Slips and Where This Leaves Things
Two internal inconsistencies are worth knowing before citing the numbers. The text states that inosine increased in relapsing patients compared with both progressive stages, while Table 4 marks it significant only against secondary progressive, matching the paper's own earlier statement that inosine and NAD+ were the only two metabolites separating relapsing from secondary progressive disease. Separately, the discussion reports reduced choline as agreeing with cerebrospinal fluid studies while calling it inconsistent with blood studies that also report reduced choline, which reads as a slip rather than a finding. One feature of the cohort also deserves attention: the control group's age standard deviation is 2.5 years against 8.9 in patients, and body mass index 2.2 against 6.5. The groups match on means, as the reported p-values of 0.77 and 0.24 confirm, but the controls are drawn from a far narrower band, which matters when serum metabolite levels move with age and body mass. What this study earns is the balanced subtype design and measurement in real concentrations, both of which are uncommon in this literature. What it has not yet earned is the diagnostic claim. The next step is the one the authors name: an independent cohort, the same three subtypes, and the areas under the curve recomputed in samples that played no part in choosing the metabolites.
Disclaimer: This blog post is based on the cited research article and is intended for informational purposes only. It is not intended to provide medical advice. Please consult with a healthcare professional for any health concerns.
Reference:
Alwahsh, M., Nimer, R. M., Dahabiyeh, L. A., Hamadneh, L., Hasan, A., Alejel, R., & Hergenröder, R. (2024). NMR-based metabolomics identification of potential serum biomarkers of disease progression in patients with multiple sclerosis. Scientific Reports, 14, 14806. https://doi.org/10.1038/s41598-024-64490-x
