An individual participant data meta-analysis on metabolomics profiles for obesity and insulin resistance in European children

Christian Hellmuth, Franca F. Kirchberg, S. Brandt, Anja Moß, Viola Walter, Dietrich Rothenbacher, Hermann Brenner, Veit Grote, Dariusz Gruszfeld, Piotr Socha, Ricardo Closa‐Monasterolo, Joaquín Escribano, Verónica Luque, Elvira Verduci, Benedetta Mariani, Jean-Paul Langhendries, Pascale Poncelet, Joachim Heinrich, Irina Lehmann, Marie Standl, Olaf Uhl, Berthold Koletzko, Elisabeth Thiering, Martin WabitschScientific 2019 · DOI

Abstract

Childhood obesity prevalence is rising in countries worldwide. A variety of etiologic factors contribute to childhood obesity but little is known about underlying biochemical mechanisms. We performed an individual participant meta-analysis including 1,020 pre-pubertal children from three European studies and investigated the associations of 285 metabolites measured by LC/MS-MS with BMI z-score, height, weight, HOMA, and lipoprotein concentrations. Seventeen metabolites were significantly associated with BMI z-score. Sphingomyelin (SM) 32:2 showed the strongest association with BMI z-score (P = 4.68 × 10<sup>-23</sup>) and was also closely related to weight, and less strongly to height and LDL, but not to HOMA. Mass spectrometric analyses identified SM 32:2 as myristic acid containing SM d18:2/14:0. Thirty-five metabolites were significantly associated to HOMA index. Alanine showed the strongest positive association with HOMA (P = 9.77 × 10<sup>-16</sup>), while acylcarnitines and non-esterified fatty acids were negatively associated with HOMA. SM d18:2/14:0 is a powerful marker for molecular changes in childhood obesity. Tracing back the origin of SM 32:2 to dietary source in combination with genetic predisposition will path the way for early intervention programs. Metabolic profiling might facilitate risk prediction and personalized interventions in overweight children.

Introduction

While the rise in childhood obesity is well-documented, the underlying biochemical "fingerprint" remains fuzzy. Most previous metabolomics studies—the study of small-molecule metabolites—suffered from small sample sizes, often involving fewer than 100 children. This led to low statistical power and inconsistent results across the literature. Furthermore, the lack of standardization across different laboratories makes it difficult to compare data directly.

This paper addresses these gaps by performing a large-scale, standardized meta-analysis to identify robust molecular markers for high BMI and insulin resistance in pre-pubertal children, providing a clearer picture of early metabolic dysfunction.

Methods

The researchers employed an individual participant data (IPD) meta-analysis, pooling raw data from three major European longitudinal cohorts: the European Childhood Obesity Project (CHOP), the Ulm Birth Cohort Study (UBCS), and the GINIplus/LISA birth cohorts. This combined dataset included 1,020 pre-pubertal children.

To ensure consistency, all blood samples were analyzed using the same targeted metabolomics platform (LC-MS/MS and FIA-MS/MS) at a single location (LMU Munich), eliminating the "batch effects" typical of multi-study comparisons. They measured 285 metabolites, including amino acids, acylcarnitines, and various phospholipids. When a significant association was found for SM 32:2, advanced mass spectrometry (lithium-induced fragmentation) was used to identify its exact molecular structure as SM d18:2/14:0.

Key Results and Insights

The "Hidden" Biomarker: SM 32:2

The study identified 17 metabolites significantly associated with BMI z-score (ZBMI). The standout marker was SM 32:2, which showed a massive statistical association ($P = 4.68 \times 10^{-23}$). This metabolite explained between 10% and 23% of the variance in BMI across the different study groups. Interestingly, SM 32:2 is often missing from standard commercial metabolomics kits, which explains why it has not been a primary focus of previous obesity research.

Association of 285 metabolites with BMI z-score. SM C32:2 (top right) shows the strongest positive association and highest explained variance.

The researchers pinpointed the specific structure of this sphingomyelin as containing myristic acid (14:0). Myristic acid is linked to high-energy diets and has been shown in other studies to predict obesity risk in infants as young as six months. The consistency of this association across the different cohorts underscores its robustness as a biomarker.

Forest plot showing the consistent positive association (beta coefficients) of SM 32:2 with BMI z-score across the four study cohorts included in the meta-analysis.

Early Signals of Insulin Resistance

Beyond BMI, 35 metabolites were linked to the HOMA index (a measure of insulin resistance). Alanine showed the strongest positive association ($P = 9.77 \times 10^{-16}$), followed by tyrosine. These amino acids appear to be early markers of metabolic shifts, likely driven by increased insulin secretion even in children who appear healthy by traditional clinical standards. Conversely, non-esterified fatty acids (NEFAs) were negatively associated with insulin resistance, suggesting that high insulin levels in these children are already suppressing the breakdown of fats (lipolysis) in adipose tissue.

Discussion and Future Directions

While these findings provide a powerful molecular map of childhood obesity, several questions remain. Because the analysis was largely cross-sectional, it is unclear if SM 32:2 is a cause or a consequence of obesity. Additionally, the study was limited to European children, and metabolic profiles may vary across different ethnic populations.

Future work should focus on longitudinal studies to determine if these childhood metabolic signatures can predict the development of Type 2 Diabetes or cardiovascular disease in adulthood. Tracing the origin of SM 32:2—whether from dietary sources or internal synthesis—will be crucial for developing early intervention programs and personalized risk prediction for overweight children.

Summary updated

Abstract

Childhood obesity prevalence is rising in countries worldwide. A variety of etiologic factors contribute to childhood obesity but little is known about underlying biochemical mechanisms. We performed an individual participant meta-analysis including 1,020 pre-pubertal children from three European studies and investigated the associations of 285 metabolites measured by LC/MS-MS with BMI z-score, height, weight, HOMA, and lipoprotein concentrations. Seventeen metabolites were significantly associated with BMI z-score. Sphingomyelin (SM) 32:2 showed the strongest association with BMI z-score (P = 4.68 × 10<sup>-23</sup>) and was also closely related to weight, and less strongly to height and LDL, but not to HOMA. Mass spectrometric analyses identified SM 32:2 as myristic acid containing SM d18:2/14:0. Thirty-five metabolites were significantly associated to HOMA index. Alanine showed the strongest positive association with HOMA (P = 9.77 × 10<sup>-16</sup>), while acylcarnitines and non-esterified fatty acids were negatively associated with HOMA. SM d18:2/14:0 is a powerful marker for molecular changes in childhood obesity. Tracing back the origin of SM 32:2 to dietary source in combination with genetic predisposition will path the way for early intervention programs. Metabolic profiling might facilitate risk prediction and personalized interventions in overweight children.
BibTeX
@inproceedings{hellmuth2019individual,
  author = {Christian Hellmuth and Franca F. Kirchberg and S. Brandt and Anja Moß and Viola Walter and Dietrich Rothenbacher and Hermann Brenner and Veit Grote and Dariusz Gruszfeld and Piotr Socha and Ricardo Closa‐Monasterolo and Joaquín Escribano and Verónica Luque and Elvira Verduci and Benedetta Mariani and Jean-Paul Langhendries and Pascale Poncelet and Joachim Heinrich and Irina Lehmann and Marie Standl and Olaf Uhl and Berthold Koletzko and Elisabeth Thiering and Martin Wabitsch},
  title = {{An individual participant data meta-analysis on metabolomics profiles for obesity and insulin resistance in European children}},
  booktitle = {Scientific},
  year = {2019},
  doi = {10.1038/s41598-019-41449-x},
}
Figure